470

The Oxford Handbook of Construction Grammar

Embed Size (px)

DESCRIPTION

The Oxford Handbook of Construction Grammar

Citation preview

Page 1: The Oxford Handbook of Construction Grammar
Page 2: The Oxford Handbook of Construction Grammar

[UNTITLED]

Page 1 of 2

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

[UNTITLED]TheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

OxfordUniversityPressisadepartmentoftheUniversityofOxford.ItfurtherstheUniversity'sobjectiveofexcellenceinresearch,scholarship,andeducationbypublishingworldwide.OxfordNewYorkAuckland CapeTown DaresSalaam HongKong KarachiKualaLumpur Madrid Melbourne MexicoCity NairobiNewDelhi Shanghai Taipei TorontoWithofficesinArgentina Austria Brazil Chile CzechRepublic France GreeceGuatemala Hungary Italy Japan Poland Portugal SingaporeSouthKorea Switzerland Thailand Turkey Ukraine VietnamOxfordisaregisteredtrademarkofOxfordUniversityPressintheUKandincertainothercountriesPublishedintheUnitedStatesofAmericabyOxfordUniversityPress198MadisonAvenue,NewYork,NY10016©OxfordUniversityPress2013Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,ortransmitted,inanyformorbyanymeans,withoutthepriorpermissioninwritingofOxfordUniversityPress,orasexpresslypermittedbylaw,bylicense,orundertermsagreedwiththeappropriatereproductionrightsorganization.InquiriesconcerningreproductionoutsidethescopeoftheaboveshouldbesenttotheRightsDepartment,OxfordUniversityPress,attheaddressaboveYoumustnotcirculatethisworkinanyotherformandyoumustimposethissameconditiononanyacquirerLibraryofCongressCataloging-in-PublicationData

Page 3: The Oxford Handbook of Construction Grammar

[UNTITLED]

TheOxfordhandbookofconstructiongrammar/editedbyThomasHoffmannandGraemeTrousdalep.cm.Includesbibliographicalreferencesandindex.ISBN978–0–19–539668–3(alk.paper)1.ConstructiongrammarI.Hoffmann,Thomas,1976–II.Trousdale,Graeme,1971–P163.5.O942012415′.018—dc232012016017ISBN978–0–19–539668–31 3 5 7 9 8 6 4 2

Page 4: The Oxford Handbook of Construction Grammar

Acknowledgments

Page 1 of 2

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

AcknowledgmentsTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

Acknowledgments

ConstructionGrammarhasbeenaroundformorethantwentyyearsandduringthattimehasattractedtheinterestofformalistsandfunctionalists,psycho-andneurolinguists,languageacquisitionresearchers,andlinguistsworkingonlanguagevariationandchange.Andyet,untilnow,nosinglevolumehasbeenavailabletostudentsandresearchersthatgivesafulloverviewofwhatthevariousConstructionGrammarapproacheshaveincommon,whattheirmajordifferencesare,andhowConstructionGrammarmaybeappliedtootherdomainsoflinguisticenquiry,suchassecondlanguageacquisitionandsociolinguisticvariation.WewerethereforeverygratefulwhenOxfordUniversityPressaskeduswhetherwewantedtoedittheOxfordHandbookofConstructionGrammar.

Aprojectthissizecaneasilyturnintoanightmarefortheeditors,butwewerefortunatetohaveagreatteamofauthors,aswellascontinuoussupportfromOUP.WeweregratefulthattheleadingexpertsinConstructionGrammarwerewillingandabletocontributetothevolume.

OurthanksgotoPeterOhlin,BrianHurley,andtheentireteamatOUP.ItwasPeterwhofirstsuggestedthatweeditaConstructionGrammarvolumefortheprestigiousOUPhandbookseries,andwehopeheisaspleasedwiththefinalproductasweare.Brian,ontheotherhand,providedpracticalsupportandpublishingfeedbackthroughouttheeditorialprocess.Ontopofthat,wewouldliketothankThomas'sstudentassistants,AnnJeanineBraband,JasminPalaszandRenatavonGottberg,fortheirmeticulousworkwiththereferencesection.

ThomasHoffmannandGraemeTrousdale

OsnabrückandEdinburgh

December2012

Page 5: The Oxford Handbook of Construction Grammar

Contributors

Page 1 of 10

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

ContributorsTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

Contributors

JóhannaBarðdalisaResearchAssociateProfessorattheUniversityofBergen.Shehasworkedoncasemarking,obliquesubjects,grammaticalrelations,constructionalsemantics,andsyntacticproductivityinasynchronicanddiachronicperspective.HerlastbookProductivity:EvidencefromCaseandArgumentStructureinIcelandicwaspublishedbyBenjaminsin2008.ShehaspublishedarticlesinNordicJournalofLinguistics,JournalofLinguistics,Language,Morphology,Linguistics,Lingua,andDiachronica.SheisafoundingcoeditoroftheJournalofHistoricalLinguistics.SheiscurrentlyrunningalargeresearchprojectonnoncanonicalsubjectmarkingintheearlyandarchaicIndo-Europeanlanguages,fundedbytheUniversityofBergen,BergenResearchFoundation,andtheNorwegianResearchCouncil.

GiuliaM.L.BenciniisAssistantProfessorintheCommunicationSciencesProgramattheHunterCollegeBrookdaleHealthSciencescampus.SheobtainedherPh.D.inlinguisticsfromtheUniversityofIllinoisin2002andafterthatworkedasaNationalInstituteofHealthfundedpostdoctoralresearchfellowattheLanguageAcquisitionResearchCenter,atHunterCollege-CUNY.Sheisanexpertinpsycholinguisticsandcombinesinsightsandmethodsfromlinguistictheory,psycholinguistics,andlanguagepathologyinherwork.

Page 6: The Oxford Handbook of Construction Grammar

Contributors

Page 2 of 10

BenjaminBergen(Ph.D.2001,UCBerkeley)isAssociateProfessorintheCognitiveScienceDepartmentattheUniversityofCalifornia,SanDiego,wherehedirectstheLanguageandCognitionLab.Hisresearchfocusesonlanguagecomprehension,production,andlearning,andhasappearedinjournalssuchasLanguage,CognitiveLinguistics,CognitiveScience,BrainandLanguage,MemoryandCognition,andtheJournalofChildLanguage.Heisalsotheauthorofaforthcomingbookonmentalsimulationandmeaning,tobepublishedbyBasicBooks.

HansC.BoasisanAssociateProfessorintheDepartmentofLinguisticsandtheDepartmentofGermanicStudiesattheUniversityofTexasatAustin.BeforecomingtoAustin,hewasapostdoctoralresearcherwiththeFrameNetproject(http://framenet.icsi.berkeley.edu)attheInternationalComputerScienceInstituteinBerkeley,California.Hispublicationsincludetwomonographs:AConstructionalApproachtoResultatives(CSLIPublications,2003)andTheLifeandDeathofTexasGerman(DukeUniversityPress,2009),aswellasfour(co-)editedvolumesandvariousarticlesonConstructionGrammarandFrameSemantics.HeisthedirectoroftheTexasGermanDialectProject(http://www.tgdp.org)andtherecipientofthe2007HugoMoserPrizeforGermanicLinguistics(InstitutefortheGermanLanguage,Mannheim),aswellasthe2011LeonardBloomfieldBookAward(LinguisticSocietyofAmerica).

GeertBooijisProfessorofLinguisticsattheUniversityofLeiden.HeobtainedhisPh.D.fromtheUniversityofAmsterdam,wherehelaterworkedasProfessorofGeneralLinguistics.Hismainresearchinterestismorphologyanditsinterfaceswithphonology,syntax,semantics,andthelexicon.HismajorpublicationsincludeThePhonologyofDutch(1995),Morphology:AnInternationalHandbookofInflectionandWordFormation,2vols.(withCh.LehmannandJ.Mugdan,eds.,2000–2),TheMorphologyofDutch(2002),TheGrammarofWords:AnIntroductiontoLinguisticMorphology(2007)andConstructionMorphology(2010).Moreover,heisoneofthefoundersandeditorsoftheYearbookofMorphology(1988–2005)anditssuccessor,thejournalMorphology.

CristianoBrocciasisanAssociateProfessorofEnglishLanguageattheUniversityofGenoa(Italy).Hisresearchfocusesoncognitivetheoriesofgrammar,Englishsyntaxandphonology,bothsynchronicanddiachronic.HispublicationsincludeamonographonEnglishchangeconstructions(TheEnglishChangeNetwork:ForcingChangesintoSchemas,2003)andvariouspapersonresultativeconstructions,simultaneityconstructions,-lyadverbs,andcognitiveapproachestogrammar.

Page 7: The Oxford Handbook of Construction Grammar

Contributors

Page 3 of 10

JoanL.Bybee(B.A.UniversityofTexas,1966;M.A.SanDiegoStateUniversity,1970;Ph.D.,Linguistics,UCLA,1973)wasonthefacultyattheStateUniversityofNewYorkatBuffalofrom1973to1989andisnowDistinguishedProfessorEmeritaoftheDepartmentofLinguisticsattheUniversityofNewMexico,whereshehasservedasdepartmentchair,associatedean,anddirectorofthe1995LSALinguisticInstitute.Bybee'sresearchinterestsincludetheoreticalissuesinphonologyandmorphology,languageuniversals,andlinguisticchange.Herworkutilizinglargecrosslinguisticdatabases,e.g.Morphology:AStudyoftheRelationbetweenMeaningandForm(1985),TheEvolutionofGrammar:Tense,AspectandModalityintheLanguagesoftheWorld(withReverePerkinsandWilliamPagliuca,1994),providesdiachronicexplanationsfortypologicalphenomena.Inaddition,herbookspresentingausage-basedperspectiveonsynchronyanddiachronyincludePhonologyandLanguageUse(2001),FrequencyofUseandtheOrganizationofLanguage(2007),andLanguage,UsageandCognition(2010).

BertCappelleisaLecturerofEnglishLinguisticsattheUniversityofLilleNorddeFrance.Hehaspublishedarangeofjournalarticlesandbookchaptersonverb-particleconstructionsinEnglish.Inaddition,hehascollaboratedonresearchprojectsinthecoregrammarareasoftense,aspect,andmodality.Hislonger-standingresearchinterestsincludethelinguisticrepresentationofmotionandchangeofstate,andthetensionbetweenconventionandinnovationinlanguageuse.

NancyChangisaPostdoctoralResearcherattheUniversitéSorbonneNouvelle(Paris3),combiningcomputationalanddevelopmentalperspectivesontheemergenceofgrammar.ShehaspreviouslybeenaresearcherattheUniversityofCaliforniaatBerkeley,wheresheearnedherdoctoratein2008,andSonyComputerScienceLaboratoryinParis.Incurrentworksheisextendingpreviousmodelsofcomputationalconstructiongrammar,simulationsemantics,andusage-basedgrammarlearningtoencompassmorecomplexargumentstructureandaspectualconstructions.

Page 8: The Oxford Handbook of Construction Grammar

Contributors

Page 4 of 10

WilliamCroftisProfessorofLinguisticsattheUniversityofNewMexico.HereceivedhisPh.D.fromStanfordUniversityin1986and—besidesnumerousvisitingpositionsacrosstheworld—workedattheUniversityofMichigan,AnnArbor,aswellastheUniversityofManchester.Hismainresearchinterestsaretypology,semantics,cognitivelinguistics,constructiongrammar,andlanguagechange.Hehaspublishedwidelyonthesetopicsininternationally-renownedpeer-reviewedjournalsandhasauthoredseminalbookssuchasExplainingLanguageChange:AnEvolutionaryApproach(2000),RadicalConstructionGrammar:SyntacticTheoryinTypologicalPerspective(2001),TypologyandUniversals(2003),CognitiveLinguistics(withD.A.Cruse2004).

HolgerDiesselisProfessorforEnglishLinguisticsattheUniversityofJena.AfterhisPh.D.fromtheStateUniversityofNewYorkatBuffaloin1998,heworkedasajuniorresearcherattheMax-Planck-InstituteforEvolutionaryAnthropologyinLeipziginthedepartmentofcomparativeanddevelopmentalpsychologyuntilhewasofferedtheprofessorshipattheUniversityofJena.Heisanexpertonlanguageacquisitionandhismainresearchinterestisintherelationshipbetweentheuseandstructureofcomplexsentencesandthepsychologicalandlinguisticfoundationsforausage-basedtheoryofgrammar.HeisontheeditorialboardofCognitiveLinguisticsandCorpusLinguisticsandLinguisticTheoryandaconsultingeditorofStudiesinLanguage.

NickEllisisProfessorofPsychology,ProfessorofLinguistics,ResearchScientistattheEnglishLanguageInstitute,andAssociateattheCenterfortheStudyofComplexSystems,UniversityofMichigan.Hisinterestsincludelanguageacquisition,cognition,emergentism,corpuslinguistics,cognitivelinguistics,andpsycholinguistics.Hislinguisticresearchconcerns(1)explicitandimplicitlanguagelearningandtheirinterface,(2)usage-basedacquisitionandstatisticallearning,(3)vocabularyandphraseology,and(4)learnedattentionandlanguagetransfer.Hisemergentistresearchconcernslanguageasacomplexadaptivesystem,networksanalysisoflanguage,scale-freelinguisticdistributionsandrobustlearning,andcomputationalmodeling.HeservesasGeneralEditorofLanguageLearning.

Page 9: The Oxford Handbook of Construction Grammar

Contributors

Page 5 of 10

CharlesJ.FillmoreisProfessorEmeritusofLinguisticsattheUniversityofCaliforniaatBerkeley,anddirectoroftheFrameNetprojectattheInternationalComputerScienceInstitute,alsoinBerkeley.Hiswritingshaveincludedcontributionstogenerativesyntax(1963),theintroductionofagrammaticalframeworkinwhichsemanticroles(‘deepcases’)werecentral(1968),therecognitionoftheimportanceofdeicticelementsinlinguisticdescriptions(1971/1997),suggestionsforfoundinglexicalmeaningson‘semanticframes’(1982,1985),andassociatedcontributionstopracticallexicography(1992,withB.T.S.Atkins),andcontributionstoagrammaticaltheorybasedonconstructions(1988andpresentvolume),withseveralcolleagues.HewasnamedFellowoftheAmericanAcademyofArtsandSciencesin1984,andheservedasPresidentoftheLinguisticSocietyofAmericain1990.

MirjamFriedisProfessorofLinguisticsattheCharlesUniversityPrague.SheobtainedherPh.D.inGeneralLinguisticsfromtheUniversityofCaliforniaatBerkeleyandhastaughtatseveralAmericanuniversities(UCBerkeley,UniversityofOregon,PrincetonUniversity).Herresearchfocusesonthecognitiveandfunctionalgroundingoflinguisticstructure,particularlyinmorphosyntax(e.g.casemarkingalternations,subordination,gradientcategoriality,interactionbetweenmorphologicalstructureandsyntaxinlanguagechange).

AdeleE.GoldbergiscurrentlyaProfessorofLinguisticsatPrincetonUniversity.Herworkinvestigatesourknowledgeoflanguageandhowthatknowledgeislearnedandconstrainedbydomain-generalprocessessuchascategorization,rationalinferences,andsocialcognition.Sheisparticularlyinterestedinconstructions,learnedpairingsofformandfunctionatthewordandphrasallevel.SheistheauthorofConstructions:AConstructionGrammarApproachtoArgumentStructure(1995)andConstructionsatWork:TheNatureofGeneralizationinLanguage(2006),aswellasnumerousexperimentalandtheoreticalstudies.

StefanTh.GriesisProfessorofLinguisticsattheUniversityofCalifornia,SantaBarbara,aswellasHonoraryLiebigProfessorattheJustus-Liebig-UniversitätGiessen.Theoretically,heisacognitivelyandConstructionGrammarorientedlinguistinthewidersenseofseekingexplanationsintermsofcognitiveprocesses;methodologically,heisaquantitativecorpuslinguistattheintersectionofcorpus,cognitive,computational,andquantitativelinguistics.Heisfoundingeditor-in-chiefoftheinternationalpeer-reviewedjournalCorpusLinguisticsandLinguisticTheory,associateeditorofCognitiveLinguistics,andontheeditorialboardsofCogniTextes,ConstructionsandFrames,Corpora,LanguageandCognition,andothers.

Page 10: The Oxford Handbook of Construction Grammar

Contributors

Page 6 of 10

martinhilpertisaJuniorResearchFellowattheFreiburgInstituteforAdvancedStudies(FRIAS).Hisresearchinterestsincludegrammaticalization,ConstructionGrammar,andcognitivelinguistics.Martinapproachesthesetopicswiththetoolsandmethodsofcorpuslinguistics.HismonographonGermanicFutureConstructions(JohnBenjamins,2008)documentschangeinthedomainoffuturetimereferenceacrossseverallanguagesandconstructions.Heiscurrentlyworkingonaprojectthatworksoutthenotionofconstructionalchangeinmoredetail.

ThomasHoffmannisAssistantProfessorattheUniversityofOsnabrück.Hismainresearchinterestsareusage-basedConstructionGrammar,synchronicsyntacticvariation,andWorldEnglishes.HehaspublishedarticlesininternationaljournalssuchasCognitiveLinguistics,EnglishWorld-Wide,andCorpusLinguisticsandLinguisticTheory,andhismonographPrepositionPlacementinEnglish(2011)waspublishedbyCambridgeUniversityPress.Ontopofthat,heisalsoeditoroftheopenacesson-linejournalConstructions.

WillemB.HollmannhasbeenaLecturerinLinguisticsatLancasterUniversitysince2003.HispublicationsincludearticlesinCognitiveLinguisticsonsummaryvs.sequentialscanning(2007)andcognitivesociolinguisticsandLancashiredialect(2011).HehaspublishedonLancashiredialectgrammarelsewhereaswell,e.g.inEnglishLanguageandLinguistics(2007).Hehasedited(withAnnaSiewierska)aspecialissueofFunctionsofLanguageonditransitiveconstructions,towhichhecontributedapaperaswell(2007).Oneofthethingsheisworkingoncurrentlyistheroleofphonologicalanddistributionalcuesinlexicalcategorization.OnepaperonthistopicwillappearinaspecialissueofStudiesinLanguage,whichheiseditingwithNikolasGisborne.

RayJackendoffisSethMerrinProfessorofPhilosophyandCo-DirectoroftheCenterforCognitiveStudiesatTuftsUniversity.Hewasthe2003recipientoftheJeanNicodPrizeinCognitivePhilosophyandhasbeenPresidentofboththeLinguisticSocietyofAmericaandtheSocietyforPhilosophyandPsychology.HismostrecentbooksareFoundationsofLanguage(Oxford,2002),SimplerSyntax(withPeterCulicover,Oxford,2005),Language,Consciousness,Culture(MITPress,2007),MeaningandtheLexicon(Oxford,2010),andAUser'sGuidetoThoughtandMeaning(Oxford,2011).

Page 11: The Oxford Handbook of Construction Grammar

Contributors

Page 7 of 10

PaulKayisProfessorofLinguisticsEmeritusattheUniversityofCalifornia,Berkeley,ConsultingProfessorofLinguisticsatStanfordUniversity,andSeniorResearchScientistattheInternationalComputerScienceInstitute,Berkeley.Hehasworkedontherelationofcolornamingtothepsychologicalrepresentationofcolor,onconstructionalapproachestogrammar,onthecontextsensitivityofgrammar,andonlanguagevariationandothertopicsinlinguisticsandanthropologicallinguistics.HeiscurrentlyworkingwithIvanSag,CharlesFillmore,andLauraMichaelis-CummingsonthedevelopmentofSign-BasedConstructionGrammar,andwithTerryRegier,Li-HaiTan,MichaelA.Webster,andothersonseekingexplanationforuniversalsandcross-languagevariationincolornamingandontheeffectofcolornamesoncolorprocessing.

JaakkoLeinoisProfessorofFinnishattheUniversityofHelsinki.HehasworkedascoordinatoroftheFinnishgraduateschoolforlanguagestudies,Langnet,asProfessorofFinnishattheUniversityofJyväskyläandattheÅboAkademiUniversity,asLecturerofGeneralLinguisticsattheUniversityofHelsinki,andasresearcherattheResearchInstitutefortheLanguagesofFinland.HehaspublishedonFinnishhistoricalsyntax,cognitivedimensionsoftheConstructionGrammarframework,diachronyandvariationinConstructionGrammar,nonfiniteconstructions,wordorder,analyticalcausatives,andthesyntaxofspokenlanguage,Finnishdialectsyntaxinparticular.

LauraA.MichaelisisanAssociateProfessorintheDepartmentofLinguisticsandafacultyfellowintheInstituteofCognitiveScienceattheUniversityofColoradoatBoulder.SheisalsoafoundingmemberoftheeditorialstaffofthenewMoutonjournalLanguageandCognition.SheistheauthorofAspectualGrammarandPast-TimeReference(1998)andBeyondAlternations:AConstructionalModeloftheGermanApplicativePattern(2001),withJosefRuppenhofer.Sheistheco-editor,withElaineFrancis,ofMismatch:Form-FunctionIncongruityandtheArchitectureofGrammar(2003).SheiscurrentlyatworkonabookthatexploresnonstandardgrammaticalpatternsinEnglishconversationalspeech.Thiswork,tentativelyentitled,ConstructionGrammarandSyntacticInnovation,willbepublishedbyOxfordUniversityPress.Alongwithnumeroushandbooksandvolumes,herworkhasappearedinthejournalsLanguage,JournalofSemantics,JournalofPragmatics,JournalofLinguistics,Lingua,LinguisticsandPhilosophy,LinguisticsandStudiesinLanguage.

Page 12: The Oxford Handbook of Construction Grammar

Contributors

Page 8 of 10

Jan-olaÖstmanisProfessorofScandinavianLanguagesintheDepartmentofFinnish,Finno-UgrianandScandinavianStudiesattheUniversityofHelsinki.HehasanM.A.inlinguisticsfromReadingUniversity(1976),andaPh.D.inlinguisticsfromUniversityofCalifornia,Berkeley(1986).Since1988hehasheldvariouspositionsattheUniversityofHelsinki(includingAssociateProfessorofEnglishLinguistics,ActingProfessorofGeneralLinguistics,andProfessorofEnglishPhilology).Since2006hehasalsoworkedasProfessorIIofScandinavianlanguagesattheUniversityofTromsø,Norway.HismainresearchinterestsaretheroleofdiscourseandgeneralpragmaticissuesinConstructionGrammar,aswellasvariabilityandcontact-inducedchange.

FriedemannPulvermüllerworkedasProgrammeLeaderattheMRCCognitionandBrainSciencesUnit,Cambridge;recentlyheacceptedthepositionofProfessorofNeuroscienceofLanguageandPragmaticsattheFreieUniversitätBerlin.Hediscoveredthatthebraindiscriminatesearlybetweenmeaningfulwordsandsenselesspseudowords,andbetweengrammaticalandsemanticwordkinds;healsoreportedearlybrainactivationpatternsindicatingthemeaningofwordsandsentences.Hehaspublishedfourbooksandover200articles,puttingforwardthefirstneurobiologicalmodeloflanguageintegratingneurobiologicalprincipleswithlinguistictheoryandofferingmechanisticnervecellcircuitsunderpinninglanguageinthehumanbrain.Neuroscienceinsightswerebroughttofruitbyhisworkinthedevelopmentofnewtreatmentproceduresforpatientswithpost-strokeaphasia.

YuryShtyrovisProfessor,SeniorScientist,andtheHeadofMEGLaboratoryattheMedicalResearchCouncil'sCognitionandBrainSciencesUnitinCambridgeandcurrentlyalsoaco-directoroftheCognitiveBrainResearchUnitattheUniversityofHelsinki.Hisresearchinthefieldofcognitiveneuroscienceiscenteredoninvestigatingneurophysiologicaldynamicsoflanguageprocessinginthehumanbrain.Hisparticularcontributiontothisareaisindescribingearlyandautomaticstagesofneurallanguageprocessingandinestablishingfunctionalparallelismfortheneuralprocessingofdifferentlinguisticinformationtypes.Hehasauthoredandco-authoreddozensofarticlesinleadingneuroscientificjournalsandbookchapters.

Page 13: The Oxford Handbook of Construction Grammar

Contributors

Page 9 of 10

LucSteelsisICREAResearchProfessorattheInstituteforEvolutionaryBiology(CSIC,UPF).HestudiedlinguisticsattheUniversityofAntwerp(Belgium)andcomputerscienceattheMassachusettsInstituteofTechnology(USA).In1983hebecameaprofessorofcomputerscienceattheUniversityofBrussels(VUB)andco-foundedtheVUBComputerScienceDepartment(FacultyofSciences).HealsofoundedtheSonyComputerScienceLaboratoryinParisin1996andservedasitsfirstdirector.HismainresearchfieldisArtificialIntelligence,andhecurrentlyfocusesontheoriesoftheoriginsandevolutionoflanguageusingcomputersimulationsandroboticexperimentstodiscoverandtestthem.

AnatolStefanowitsch(Ph.D.2001)isProfessorofEnglishLinguisticsattheUniversityofHamburg.HereceivedhisPh.D.fromRiceUniversity,Houston,Texas,andwasProfessorofEnglishandGeneralLinguisticsattheUniversityofBremenbeforetakinghiscurrentposition.Hismajorresearchinterestsincludeconstructiongrammar,metaphor,andquantitativecorpuslinguistics,withspecialemphasisonassociationsbetweenlexicalitemsandotherelementsofsyntacticandsemanticstructure.Hehaspublishedanumberofpapersoncollostructionalanalysis(manyofthemco-authoredwithStefanTh.Gries)andwasafoundingeditorofthejournalCorpusLinguisticsandLinguisticTheory.

GraemeTrousdaleisaSeniorLecturerattheUniversityofEdinburgh.WithNikolasGisborne,heeditedConstructionalApproachestoEnglishGrammar(deGruyter,2008).HehaspublishedanumberofjournalarticlesandbookchaptersonconstructionalapproachestovariationandchangeinEnglish,andiscurrentlyworkingonabookonconstructionalizationwithElizabethClossTraugott.HeisalsotheauthorofAnIntroductiontoEnglishSociolinguistics(EUP,2010).

StefanieWulffisAssistantProfessorattheUniversityofFlorida.AfterherPh.D.(UniversityofBremen),shefirstworkedasavisitingscholarandlaterasalecturerattheUniversityofCaliforniaatSantaBarbara,aswellasaPostdoctoralResearchFellowattheUniversityofMichiganandanAssistantProfessorattheUniversityofNorthTexas.Hermainresearchinterestsareusage-basedcognitivelinguisticvariation(includingissuessuchaslexico-syntacticvs.constructional,aswellasdialectalvariation)andsecondlanguageacquisition.ShehaspublishedwidelyininternationaljournalssuchasCorpusLinguisticsandLinguisticTheory,CorporaandInternationalJournalofCorpusLinguistics,co-editedthevolumeCorpusLinguisticApplications:CurrentStudies,NewDirections(withStefanGriesandMarkDavis,2009)andistheauthorofRethinkingIdiomaticity:AUsage-basedApproach(2008).

Page 14: The Oxford Handbook of Construction Grammar

Abbreviations

Page 1 of 4

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

AbbreviationsTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

Abbreviations

3SG 3rdpersonsingular

A(DJ) adjective

ADV adverb

ADVP adverbphrase

ACC accusativecase

AGR agreement

ALL allativecase

ASP aspectualmarker

AVM attribute-value-matrix

BCG BerkeleyConstructionGrammar

BNC BritishNationalCorpus

DAR definitearticlereduction

CAT category

Page 15: The Oxford Handbook of Construction Grammar

Abbreviations

Page 2 of 4

CCG CognitiveConstructionGrammar

CG CognitiveGrammar

CEEC CorpusofEarlyEnglishCorrespondence

CLMET CorpusofLateModernEnglishTexts

COCA CorpusofContemporaryAmericanEnglish

DAT dativecase

DET determiner

ECG EmbodiedConstructionGrammar

FCG FluidConstructionGrammar

GEN genitivecase

HPSG Head-DrivenPhrase-StructureGrammar

ICE InternationalCorpusofEnglish

L1 firstlanguage

L1A firstlanguageacquisition

L2 secondlanguage

L2A secondlanguageacquisition

LFG LexicalFunctionalGrammar

LOB Lancaster-Oslo-BergenCorpus

LVC LanguageVariationandChange

MGG MainstreamGenerativeGrammar

MP MinimalistProgram

Page 16: The Oxford Handbook of Construction Grammar

Abbreviations

Page 3 of 4

N noun

NLP naturallanguageprocessing

NOM nominativecase

NP nounphrase

Obj object

Obl oblique

OE OldEnglish

OED OxfordEnglishDictionary

OVS object–verb–subject

P preposition

PDE PresentDayEnglish

PLUR plural

PP prepositionalphrase

PPCEME ThePennParsedCorpusofEarlyModernEnglish

Prt particle

PST pasttense

RCG RadicalConstructionGrammar

RRG RoleandReferenceGrammar

SAI subject-auxiliaryinversion

SBCG Sign-BasedConstructionGrammar

SED SurveyofEnglishDialects

Page 17: The Oxford Handbook of Construction Grammar

Abbreviations

SEM semanticmeaning

Subj subject

SVO subject–verb–object

SUP superlative

TAM tense-aspect-mood

THAT-CL that-clause

UG UniversalGrammar

V verb

VP verbphrase

V transitiveverb

WXDY What'sXdoingY?construction

YCOE TheYork-Toronto-HelsinkiParsedCorpusofOldEnglishProse

YORK CorpusofYorkEnglish

YPC TheYork-HelsinkiParsedCorpusofOldEnglishPoetry

TR

Page 18: The Oxford Handbook of Construction Grammar

The Oxford Handbook of construction Grammar

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

TheOxfordHandbookofconstructionGrammarTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

TheOxfordHandbookofconstructionGrammar

Page 19: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 1 of 9

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0001

ConstructionGrammar:IntroductionThomasHoffmannandGraemeTrousdaleTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterdiscussesthehistoryandprinciplesofConstructionGrammar.ItexplainsthatConstructionGrammarhasitsrootsintheSaussureannotionofthelinguisticsign,andthechapteroutlinessomeoftheunifyingprinciplesofconstructionalapproachestothearchitectureoflanguage.Thechapteralsoprovidesasummaryoftheothercontributionstothevolume.

Keywords:ConstructionGrammar,linguisticsign

OneofthecentralconceptsoflinguisticsistheSaussureannotionofthelinguisticsignasanarbitraryandconventionalpairingofform(orsoundpattern/signifiant)andmeaning(ormentalconcept/signifé;cf.,e.g.,deSaussure[1916]2006:65–70).Underthisview,theGermansignApfelanditsHungarianequivalentalmahavethesameunderlyingmeaning‘apple,’butdifferentassociatedconventionalforms([ˡɑpfl]and[ˡɑlmɑ]).OverseventyyearsafterSaussure'sdeath,severallinguiststhenexplicitlystartedtoexploretheideathatarbitraryform-meaningpairingsmightnotonlybeausefulconceptfordescribingwordsormorphemesbutthatperhapsalllevelsofgrammaticaldescriptioninvolvesuchconventionalizedform-meaningpairings.ThisextendednotionoftheSaussureansignhasbecomeknownasa‘construction’(whichincludesmorphemes,words,idioms,andabstractphrasalpatterns)andthevariouslinguisticapproachesexploringthisideawerelabeled‘ConstructionGrammar.’

Insteadofassumingaclear-cutdivisionoflexiconandsyntax,ConstructionGrammariansthusconsiderallconstructionstobepartofalexicon-syntaxcontinuum(a‘constructicon,’Fillmore1988;seealsoJurafsky1992andGoldberg2003:223).Examplesfromthiscontinuumaregivenin(1)–(4)(employingafairlyinformaldescriptionoftheformandmeaningparts;forvariousdifferentapproachestotherepresentationofconstructions,cf.thechaptersinPartII):

(1)wordconstruction:apple[æpl]—‘apple’(2)idiomconstruction:e.g.,XtakeYforgranted[XTAKEYfәɡɹɑːntιd]—‘Xdoesn’tvalueY’(3)comparativeconstruction:e.g.,Johnistallerthanyou[XBEAdj ðәnY]—‘XismoreAdjthanY’(4)resultativeconstruction:e.g.,Sherocksthebabytosleep[XVYZ]—‘XcausesYtobecomeZbyV-ing’

Thewordin(1)isaclassicpairingofformandmeaningandthusqualifiesasaconstruction.Themeaningoftheidiomin(2)isnotcompletelycompositionalandmustthereforebestoredinaspeaker'smentallexicon.Moreover,while(2)isanidiomthatispartlysubstantive(i.e.,whosephonologicalformisfixedinseveralplaces,i.e.,[fә]and[ɡɹɑːntιd]),italsocontains‘slots’initssubjectandobjectpositionthatcanbefilledbyvariouselements(cf.Manypeopletaketheirfriendsforgranted./Shetakeshimforgranted./JohnandJilltakeeachotherforgranted./…).

comparative1

Page 20: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 2 of 9

Constructionsthatcontainsuchslotsaresaidtobeschematicandthecomparativeconstructionin(3)canthereforebesaidtobemoreschematicthantheidiomin(2),sincetheformeronlyhasonesubstantiveelement([ðæn])andseveralschematicslots(forthesubjectX,theformofBE,thecomparativeadjectiveandY).Finally,theresultativeconstructionin(4)iscompletelyschematic,sinceitonlycontainsslotsforthecauseX,theverbV,theaffectedcomplementY,andtheresultingstateZ(andthuslicensessuchdiversestructuresasShekisseditbetter./Theywipedthetableclean./Hecoloredhisbeardred./…).

Whentracingthehistoryoftheterm‘construction’inlinguistics,itisinterestingtonotethattheanalysisofsyntacticstructuresasform-meaningpairingswascommonplaceintraditionalgrammars(andtheuseofconstructionsinthissensecanbetracedbackatleastasfarasthetwelfth-centuryMedievalLinguists;cf.GoldbergandCasenhiser2006:343).Infact,eventheearlyChomskyantransformationalmodels(1957,1965)employedthenotion‘constructions’andincludedconstruction-specificrules.Inhisdiscussionofphenomenasuchas(2)–(4)inhisseminalAspectsoftheTheoryofSyntax,Chomsky,forexample,speaksofthe‘Verb-with-Complement’(1965:190)constructiontakeforgranted,thecomparativeconstructionsunderlyingJohnismorecleverthanBill(1965:178),oracausativeconstructionakinto(4)(1965:189).Yet,whenthefirstmajorConstructionGrammarpublicationsappearedinthe1980s(inparticularFillmore1985a,1988;Fillmore,Kay,andO’Connor1988;Lakoff1987orWierzbicka1988),thesewereallconsideredtoberevolutionaryandinstarkoppositiontothedominatingsyntactictheoryatthetime,theChomskyanPrinciples-and-Parametersapproaches(asexemplifiedby,e.g.,Chomsky1981).

ThereasonforthisliesinthefactthatthePrinciples-and-Parametersmodels,whichsupersededtheearlierAspectsapproach, wereallcharacterizedbytheclaimthatconstructionsareonlyepiphenomena,thatis,merelyacollectionofstructuresthataretheresultoftheinteractionofuniversalprinciplesandparametersettings(cf.Chomsky1995:129).Assumingtheexistenceofaninnatelanguagefaculty(calledUniversalGrammar),thefocusoftheseChomskyanMainstreamGenerativeGrammarapproachesshiftedfromlanguage-specificcomputationalmodelsofnativespeakercompetencetoaninvestigationofthementalprinciplesandparametersunderlyingallhumanlanguages.Yet,whileproponentsofthePrinciples-and-Parameterstheoryclaimtohaveachievedahighlevelofdescriptiveandexplanatoryadequacy(inthesenseofChomsky1965:24–26),CulicoverandJackendoff(2005:25–37)drawattentiontothefactthatthisallegedsuccessisonlyachievedbyignoringagreatnumberoflinguisticphenomena.ThestandardapproachofthePrinciples-and-Parametersparadigmis“tofocusonthecoresystem,puttingasidephenomenathatresultfromhistoricalaccident,dialectmixture,personalidiosyncrasies,andthelike”(Chomsky1995:20).Asitturnsout,however,thisresultsinagreatmanyphenomena—allofwhichmustbepartofaspeaker'smentalrepresentationoflanguage—whichcannotbeaccountedforbythePrinciples-and-Parameterstheory.Besides,asBorsley(2006,2007)pointsout,inordertoachieveanalysesthatcancapturethedifferencesbetweenclausetypes(aswellasotherabstractconstructions)Principles-and-Parametersapproachesheavilyrelyoninvisibleelementsthatheadfunctionalprojections(suchasvP,T,orC;cf.Chomsky1995,2000b)andwhichhavespecificpropertiesandinterpretations—ananalysis“thatislittledifferentfromaconstruction-basedaccountthatassociatesthesamesetofpropertiesandinterpretationsdirectly”(Sag2010:488).

Asthepresentvolumewillshow,incontrasttothis,ConstructionistGrammarapproachesprovideauniformanalysisofmoreidiosyncratic‘peripheral,’aswellas‘core,’linguisticfeatures.Theyachievethiswithoutrecoursetotransformations/derivationsortheemploymentofemptyelements.Instead,thementalgrammarofspeakersisclaimedtoconsistofanetworkofschematicandsubstantiveconstructions(‘constructicon’)anditistheparallelactivationofconstructionsthatunderliesasetofparticularutterances(‘constructs’).Thisviewofgrammarasamentalnetworkofconstructionshasrecentlyreceivedgreatempiricalsupportbyindependentresearchonfirstandsecondlanguageacquisition(seeDiessel,thisvolume,andEllis,thisvolume,respectively),psycholinguistics(seeBencini,thisvolume),aswellasneurolinguistics(seePulvermüller,Cappelle,Shtyrov,thisvolume).Moreover,constructionalapproacheshavealsoprovidedimportantnewinsightintothediachronicevolutionoflanguages(seeFried,thisvolume;Barðdal,thisvolume;Hilpert,thisvolume),aswellassociolinguistic(seeHollmann,thisvolume)anddialectalanddiscoursevariation(seeÖstmanandTrousdale,thisvolume).

SinceConstructionistGrammaroffersapsychologicallyplausible,generativetheoryofhumanlanguage,ithasspawnedalargebodyofconstructionistresearch.Thepresenthandbooknowprovidesthefirstauthoritativeoverviewofthetheoryanditsapplications,aswellasthevariousconstructionistapproaches.Eachchapteriswrittenbyoneoftheleadingresearchersinthefield,andthebookthereforeoffersthereaderthemostexhaustiveoverviewofConstructionGrammarpresentlyavailable.Next,wewillbrieflyoutlinethehandbook'soverallstructure

2

Page 21: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 3 of 9

andprovideashortoverviewofthevariouschapters.

Followingthisintroduction,PartI.PrinciplesandMethodsstartsoffwithAdeleGoldberg'scontribution(‘ConstructionistApproaches’),whichhighlightsthefundamentalassumptionssharedbyallconstructionistapproaches.AsGoldbergpointsout,therearefourmaintenetsthatdistinguishConstructionGrammarapproachesfromChomskyanMainstreamGenerativeGrammar:

•analysisofphrasalandclausalphenomenaasgrammaticalconstructions

•thefocusonsurfaceformandtheresultingeschewmentoftransformationsorderivations

•theclaimthatconstructionsformanetwork

•andtheacknowledgmentandcarefulanalysisofcrosslinguisticvariabilityandtheroleofdomain-generalcognitiveprocessesinformingcrosslinguisticgeneralizations.

Thenextthreechaptersthenillustratethatdespitethesecommonassumptions,thequestionastowhatcountsasaconstructionisfarfromclear.WhileallConstructionGrammarianswouldnormallyagreethatanyidiosyncraticpairingofmorphological,syntactic,lexical,semantic,pragmatic,ordiscourse-functionalproperties(suchastheidiomXtakeYforgranted)mustbeconsideredaconstruction(cf.CroftandCruse2004:263),otherissuessuchastheroleoffrequencyorthestatusofsemiregularphenomenaremaindisputed:in‘TheLimitsof(Construction)Grammar,’PaulKay,forexample,advocatesthemostconservativeview,whichonlyconsidersthoselinguisticphenomenaasconstructionsthataspeakerneedstoknowto“produceandunderstandallpossibleutterancesofalanguageandnomore.”Consequently,hemakesadistinctionbetweenconstructionsasfullyproductiveprocessesthatarepartofaspeaker'sgrammar,ontheonehand(suchastheAll-cleftconstruction),andsemiregularprocessessuchasthe‘AasNP’-pattern(asexemplifiedbydumbasanox,deadasadoornail,orgreenasgrass),ontheother,whichheseesasmerepatternsofcoining(andwhicharepartofmetagrammar,i.e.“acompendiumofusefulstatementsaboutthegrammar”).

Incontrasttothis,JoanL.Bybee(‘Usage-basedTheoryandExemplarRepresentationsofConstructions’)outlinesaviewinwhichthementalgrammarofspeakersisshapedbytherepeatedexposuretospecificutterancesandinwhichdomain-generalcognitiveprocessessuchascategorizationandcross-modalassociationplayacrucialroleintheentrenchmentofconstructions.Underthisview,therewouldbenodistinctionbetweengrammarandmetagrammar,butalllinguisticknowledgeisinsteadviewedasemergent(Hopper1987;KemmerandBarlow2000;EllisandLarsenFreeman2006a)andconstantlychanging.Moreover,specificphonologicallinguisticusage-eventsaresaidtobestoredtogetherwithrichsemanticandpragmaticinformation(including,e.g.,positive,aswellasnegative,contextsyieldingso-calledsemanticprosodyeffects;cf.Sinclair1996b).Intheseapproaches,(typeandtoken)frequencyplaysamajorroleandanythingthathasbeenencounteredoftenenoughtobeaccessedasawholeisconsideredaconstruction,evenifitexhibitsnoidiosyncrasyofformandmeaning(andthereforecouldalsobeassembledon-lineasaconstructthatisfullylicensedbythecombinationofotherconstructions).

FrequencyalsoplaysanimportantroleinRayJackendoff'sParallelArchitectureapproach(‘ConstructionsintheParallelArchitecture’).IntheParallelArchitecturemodelphonology,syntaxandsemanticsaredescribedasindependentgenerativecomponentswithinaspeaker'smentalgrammar.Moreover,thesmall-scaleinterfacerulesbetweenthethreegenerativecomponents,whichJackendoffcalls‘lexicalitems’(i.e.,pairingsofphonological,(morpho)syntactic,andsemanticstructures),aresynonymouswithwhatmostotherconstructionistapproacheswouldcallconstructions.Ontopofthat,similartousage-basedapproaches,Jackendoffcontendsthatitisfrequencythatdecideswhichlexicalitems/constructionsbecomelistedinaspeaker'slong-termmemory.Yet,incontrasttomostConstructionGrammarapproaches,lexicalitemsintheParallelArchitecturedonotalwayscompriseSaussureansigns(i.e.,pairingsofformandmeaning)butalsoinclude‘defective’lexicalitems,thatis,abstractsyntacticprincipleswithoutmeaning(suchastheVP-construction[ VNP])orsemanticprincipleswithoutsyntacticform,suchasreferencetransferinIhavereadShakespeare(Shakespeare=‘books/playsbytheauthorShakespeare’).

Thefinalchapterinthispart,byStefanTh.Gries(‘DatainConstructionGrammar’),thenmovesawayfromthesetheoreticalissuesbutneverthelessaddressesafundamentalissueofConstructionGrammar:WhatkindsofdataareusedinconstructionistapproachesandwhatinferencesfortheirtheoriesdoConstructionGrammarians

VP

Page 22: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 4 of 9

makefromthesedatatypes?AsGriesshows,ConstructionGrammarresearchersuseawidevarietyofdifferentdatatypesfromintrospectionthroughcorpusevidencetoexperiments.However,hischapterisnotmerelyavaluablesurveyofpreviousresearch.Griesalsometiculouslyoutlinestheparametersalongwhichdatatypescanbeclassified(withrespecttothenaturalnessofthedata,thetypeofstimuliused,andtypeofresponsecollected)andhowthisimpactsthekindofinferencethatcanbemadeforConstructionGrammarmodelsofspeakers’mentalgrammars.

PartII.ConstructionistApproachesthenmovesontoparticularframeworksinwhichthenotionofconstructionsplaysacentralrole.Historically,thefirstapproachthatexplicitlylabeleditselfasa‘ConstructionGrammar’wasBerkeleyConstructionGrammar(BCG).Infact,whenthefirstpapersappearedinthisframeworkinthelate1980stoearlytomid-1990s(Fillmore1985a,1988;Fillmore,Kay,andO’Connor1988;Michaelis1994;MichaelisandLambrecht1996),theysimplycalledtheirapproachConstructionGrammar(withtheattribute‘Berkeley’onlyaddedrecentlytodistinguishitfromotherconstructionistapproachesthatemergedlateron).Inthepresentchapter,oneofthefoundingfathersofBCG,CharlesJ.Fillmore,outlineswhatmotivatedhimandPaulKaytopursueananalysisofgrammarinwhichperipheral/‘noncore,’aswellascoregrammaticalpatternsweredescribedandexplainedbythesamemechanisms.Thechapterintroducesthe‘boxeswithinboxes’representationandtheattribute-value-matrix(AVM)notationofBCGandalsodiscussesissuessuchasunification,constituency,unaryconstructions(‘pumping’),valence,complementation,co-instantiation,andnullinstantiation.Moreover,FillmorepresentsBCGanalysesofvariousconstructionssuchastheSubject-Predicateconstruction,Head-Complementconstruction,Inversionconstruction,orLeft-Isolationconstruction.

WhileworkwithinBCGisstillbeingcarriedoutandimportantconstructionistworkispublishedin(revisedversions)oftheframework(cf.theworkofMirjamFried,Jan-OlaÖstman,ToshioOhori,andothers;e.g.,FriedandÖstman2004a;Ohori2005;Fried2008,2009b),neitherFillmorenorKaycontinuetopursuethisapproach.ThesedaysFillmore'sworkfocusesonissuesofframesemantics,thatis,theconceptualstructureoflexicalmeaning(Petruck1996),andthetechnologicalimplementationofframesemanticandsyntacticinformationintheFrameNetdatabase<http://framenet.icsi.berkeley.edu/〉; (cf.,e.g.,Fillmore,Johnson,andPetruck2003;Boas2006,2009b).YeteventhoughhisworkonframesemanticsisnotexplicitlycouchedinanyparticularConstructionGrammarframework,itisneverthelesshighlyrelevantforconstructionistanalyses(cf.,e.g.,Iwata2008;Boas2009a;Hasegawaetal.2010;Fillmore,Lee-Goldman,andRhodes2012).

PaulKay,ontheotherhand,hasstartedtocollaboratewithIvanSagandLauraA.MichaelisonanapproachcalledSign-BasedConstructionGrammar(SBCG).AsMichaelispointsout,SBCGevolvedoutofideasfromBCGandconstruction-basedHead-DrivenPhrase-StructureGrammar(HPSG)(GinzburgandSag2000).LikeBCGandHPSG,SBCGemploysAVMsorfeaturestructurestomodellinguisticphenomena.However,unlikeBCGbutsimilartoHPSG,thefeaturestructuresofSBCGaretyped,thatis,arrangedinahierarchicalinheritanceclassification(Carpenter1992).SBCGisahighlyformalizedtheoryinwhichconstructionsaredefinedastypeconstraintsonlocaltrees(so-called‘constructs,’notehowthistechnicaldefinitionofconstructandconstructiondiffersfromtheinformalonegivenabove;cf.Michaelis2009,thisvolume;andBoasandSag2012fordetails).Inhercontribution,MichaelistracksthehistoricalconnectionbetweenSBCGandBCGandalsoshowsinwhichaspectstheformerdiffersfromthelatter.OtherissuescoveredinherchapterincludethestrictlocalityconstraintofSBCG,theavoidanceofovergeneralization,andinheritance,aswellasthetreatmentofinflectionalandderivationprocesses.

TwootherframeworksthatarealsohighlyformalizedandbasedontheunificationofAVMsareFluidConstructionGrammar(FCG)andEmbodiedConstructionGrammar(ECG).Besidestheirformalnature,thesetwoapproachesalsosharethefactthattheyarespecificallyaimedatcomputationalimplementation.Inhiscontribution,LucSteelsdrawsattentiontothefactthatFCGisaformalismthatallowsConstructionGrammarresearcherstoformulatetheirfindingsinaprecisemannerandtotesttheimplicationsoftheirtheoriesforlanguageparsing,production,andlearning.Moreover,FCGhasbeenfullyimplementedandisfreelyavailablefromtheproject'shomepage<http://www.fcg-net.org/>.TheFCGchapterinthisvolumeprovidesanoverviewoftheformalismconcerningitsfundamentalprimitivedatastructuresandoperations,aswellasitscomputationalbackgroundtoparsingandproduction.Ontopofthat,Steelsalsodiscussesissuesofdesignpatterns(i.e.,themethodsandtechniquesadoptedfortheimplementationofcomplexlinguisticphenomena).

ECGisanothercomputationalimplementationofConstructionGrammar,butasBenjaminBergenandNancy

3

Page 23: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 5 of 9

Changpointout,thedrivingquestionofthisframeworkishowlanguageisusedinactualphysicalandsocialcontexts.Thus,ECGisanattempttocomputationallymodelthecognitiveandneuralmechanismsthatunderliehumanlinguisticbehavior.OneofthemaintenetsofECGisthatmentalsimulationplaysacrucialroleinprocessinglanguage.ChangandBergensurveythisroleofmentalsimulationinprocessingandoutlinehowlanguagecanbeseenasaninterfacetosimulation.Inaddition,theyillustratetheroleofconstructionalknowledgeformentalsimulationandshowhowconstructionsarerepresentedinECG.Finally,basedonpsycholinguisticevidence,theyalsooutlineanECG-basedmodeloflanguagecomprehension.

Incontrasttotheseformalapproaches,theremainingapproachesinthispartexplicitlyeschewformalizationandinsteadadoptastrongusage-basedperspective.Thefirstofthese,Langacker'sCognitiveGrammar(CG),developedindependentlyofConstructionGrammar,butasCristianoBrocciasnotesalsosharesmanyassumptionswithotherconstructionistapproaches.ThebasicpremiseofCGisthatlanguageisgroundedinembodiedhumanexperienceandlanguage-independentcognitiveprocessessuchasassociation,automatization,schematization,andcategorization.Moreover,sinceCGassumesgrammartobeinherentlymeaningful,constructionsasform-meaningpairingshaveacentralplaceintheframework.InCG,constructionsaredefinedasbipolarstructuresthatpairasemanticpolewithaphonologicalone.Consequently,unlikeotherapproaches,CGdoesnotincludesyntacticinformationintheformpole.Inthechapterforthishandbook,BrocciasoutlinestheroleofconstructionsinCG,butalsoaddressestopicssuchassemanticandgrammaticalrolesinCG,thestatusandanalysisofgrammaticalclasses,aswellastheviewthatmeaningarisesthroughthedynamicprocessofconceptualization.

RadicalConstructionGrammar(RCG),ontheotherhand,emergedasaresultofWilliamCroft'sextensivetypologicalresearch.Arguingthatallgrammaticalcategoriesarelanguage-specificandconstruction-specific,RCGassumesnoformalsyntacticstructureotherthanthepart-wholestructureofconstructionsandthegrammaticalrolesthatoccurinconstructions.RCGisthusanonreductionisttheorythattakesconstructionstobeholisticgestaltphenomena(inwhichpart-partrelationshipsplaynorole).Moreover,RCGadvocatesanexemplarsemanticsmodelofthesyntax-semanticsmappinginwhichspecificsituationtypesareorganizedinamultidimensionalconceptualspace.Formalconstructiontypesarethensaidtohaveafrequencydistributionoverthatconceptualspace.OthercentraltopicsinCroft'schapterincludethecentralroleofdistributionalanalysisinRCG,theinternalformalstructureofconstructions,aswellastheanalysisoftypologicalaswellaslanguage-internalvariation,andthestatusofgrammaticalgeneralizations.

Finally,CognitiveConstructionGrammar(CCG)aimsatprovidingapsychologicallyplausibleaccountoflanguagebyinvestigatingthegeneralcognitiveprinciplesthatservetostructurethenetworkoflanguage-specificconstructions.AsHansC.Boasnotes,oneresultofthisapproachis,forexample,theattempttoexplainhowtheformofconstructionsismotivatedbytheirmeaning.Inhischapter,BoastracesthefoundationsofCCGbyoutliningAdeleGoldberg'sseminalworkonArgumentStructureconstructions.Ontopofthis,hediscussesconstructionsthathaveimportantdiscourse-relevantproperties(suchastheNominalExtrapositionconstructionortheDeprofiledObjectconstruction),andtheroleofphonologicalfactorsforlicensingconstructions(cf.,e.g.,theroleofprosodyintheNominalExtrapositionconstructionorrhetoricalquestions).Besidesthis,Boasalsolooksattheinteractionofmultipleconstructions,theroleofnetworksandinheritancehierarchies,aswellasfrequencyandproductivity,fromaCCGperspective.

PartIII.Constructicon:FromMorphemestoClausesandBeyondthenillustrateshowconstructionistapproachescanbeusedfortheanalysisofalltypesof(morpho)syntacticphenomenafromthelexicon-syntaxcline.Importantly,thechaptersinthispartarenotcommittedtoaparticularConstructionGrammarframeworkbutareinsteadcouchedinageneralconstructionistterminologyinordertomaketheirfindingsandresultsmaximallyaccessibletoallreaders.PartIIIstartsoffatthebottomendofthelexicon-syntaxclinewithadiscussionofthesmallesttypesofconstructions.InMorphologyinConstructionGrammar,GeertBooijpresentsawholerangeofargumentsforaConstructionGrammarapproachtomorphology.First,heshowshowspecificlexicalitemssuchasplayableordebatableinteractwiththeircorrespondingabstractconstruction.AsBooijemphasizes,aconstructiconwithboththesetypesofconstructionsarrangedinadefaultinheritancenetworknotonlylicensesthespecificlexicalconventionsofalanguagebutalsoaccountsforthegenerativepowerofgrammar.Thenhediscussesthefactthatpsychologicalevidencesupportstheviewthatmorphologicalconstructionshaveholisticproperties.Inadditiontoword-formation,Booijalsocoversinflectionalmorphologyanddiscussestherelationshipbetweenmorphologicalandsyntacticconstructions.

Page 24: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 6 of 9

Onestepfurtheruponthelexis–syntaxcline,StefanieWulffthentouchesuponimportantissuesintheconstructionistanalysisofWordsandIdioms.WulffsurveysthemainargumentsfromFillmore,Kay,andO’Connor's(1988)classicstudyonidiomsandillustrateswhatleadFillmoreetal.totheconclusionthatthesephenomenaarebestanalyzedasconstructions.BesidespresentingFillmoreetal.'sclassificationofidiomsalongthecategoriesof‘familiarvs.unfamiliarpieces’and‘familiarlyvs.unfamiliarlyarranged,’shealsodiscussestheinterpretationofidiomsfromausage-basedConstructionGrammarpointofview.Inparticular,sheaddressestheproblemofhowthedegreeofsemanticandsyntacticirregularityofdifferentconstructionscanbemeasuredquantitatively.AsWulffpointsout,thenotionofcollocatesmightprovideanimportantcluetotheanswerofthisquestion:shearguesthatthecompositionalityofanidiomconstruction(e.g.,begDETquestion)cruciallydependsonhowsemanticallysimilaritscollocatesarecomparedtothoseofthesimplewordconstructionsthattheidiomisformallyassociatedwith(herebegandquestion).Inparticular,drawingoncorpusdata,sheproposesaquantitativemeasure(weightedR)thatnumericallyexpressesthisdegreeofsemanticsimilarityandthatcanbeconsideredanumericalmeasureofcompositionality.

Inasimilarvein,AnatolStefanowitschpresentsCollostructionalAnalysis,afamilyofquantitativecorpus-linguisticmethodsthatallowresearcherstoexpressthestrengthoftherelationshipbetweenwordconstructionsandthegrammaticalstructurestheyoccurin.Afterfamiliarizingthereaderwiththethreemajorvariantsofcollostructionalanalysis,namely‘simplecollexemeanalysis,’‘distinctivecollexemeanalysis,’and‘co-varyingcollexemeanalysis,’Stefanowitschgivesseveralcasestudiesandshowshowtheresultsfromthesevaryingcollostructionalmeasurescanenlightenconstructionistanalysesoflexicalandgrammaticalconstructions.Finally,headdressesimportantmethodologicalissuesofcollostructionalanalysissuchastheuseofinferentialstatistics,thecognitivemechanismsassumed,aswellasthechoiceofstatisticaltests.

Next,movingtowardthemoresyntacticendofthelexis-syntaxcline,ThomasHoffmannlooksatAbstractPhrasalandClausalConstructions.Asthetitlesuggests,thischapteroutlineshowconstructionistapproachescandescribeandmodeleventhemostabstractofsyntacticstructures.Ontopofthis,however,Hoffmannemphasizesthatsuchanalysesalsofurtherourunderstandingofthephenomenainquestion.First,heturnstoabstractclausalconstructionsanddiscussesthearrangementofdeclarative,interrogative,imperative,andrelativeclausesindefaultinheritancenetworksandpointsoutdifferencesbetweenthoseConstructionGrammarframeworksthattakeausage-basedapproachandthosethatdonot.Thenhemovesontofiller-gapconstructionsandillustrateshowthesecanbeanalyzedwithinaConstructionGrammarapproachandhowtheseconstructionsinteractinEnglishwithprepositionplacement(pied-pipingTowhomshouldItalk?vs.strandingWhoshouldItalkto?).Finally,hepresentsananalysisoftheEnglishcomparativecorrelativeconstruction(e.g.,themoreyouread,thesmarteryou’llget)andprovidesempiricalevidenceforaspecificintonationalsignatureofthisconstruction.

Whilethepreviouschaptersinthispartfocusedmoreontherelationshipbetweenlexicalandsyntacticconstructions,JaakkoLeinoinsteadconcentratesontheroleofInformationStructureinconstructionistapproaches.Sinceinformationstructureinfluencesbywhichconstructionaparticularmeaningisexpressedandwhyspeakersthereforechooseoneconstructionoveralternativeonesinspecificsituations,itisanelementofsentencegrammarthathasattractedconsiderableinterestwithintheConstructionGrammarcommunity.Inhiscontribution,Leinosurveysthecentralnotionsofinformationstructure(presuppositionandassertion,identifiabilityandactivation,topicandfocus)andhowthesefactorscanbeincorporatedintoaConstructionGrammarviewofmentalgrammar.

PartIV.AcquisitionandCognitionshiftsthefocustothepsycholinguisticandneurolinguisticunderpinningsofConstructionGrammar,withaparticularfocusonhowchildrenandadultsacquireconstructionalknowledgeatdifferentstagesofdevelopment.Thefirstchapter,byHolgerDiessel,isconcernedwithfirstlanguage(L1)acquisition.DiesselcontraststhegenerativeapproachtoL1acquisitionwiththeusage-basedapproach,thelatterofwhichhasbeenstronglyinfluencedby—andhasinturninfluenced—themoregeneraltheoryofconstructionalknowledge.DrawingonworkbyMichaelTomasello,Diesselconsidersthelexicalspecificityofearlypivotschemasthatchildrenuse,andtheirrelationshiptotheverb-argumentconstructionsthattypifyadultlanguage.Thisinvolvesaconsiderationofhowthemorespecificknowledgedisplayedbychildreninpivotschemasgeneralizestothemoreschematicconstructionalknowledgeassociatedwithadultgrammars.Theevidencesuggeststhatchildren“learn…anetworkofrelatedconstructionsinwhichthesameeventisconstruedfromdifferentperspectivessothatspeakerscanchoosetheconstructionthatismostappropriatetorealizetheircommunicativeintentioninaparticularsituation.”Itappearsthatchildren'sknowledgeofcomplexsentencesinvolvingfiniteand

Page 25: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 7 of 9

nonfinitecomplementsdevelopsinawaythatissimilartothatoftheirknowledgeofverb-argumentconstructions.

Theusage-basedapproachfeaturesheavilyinNickEllis'scontributiononsecondlanguage(L2)acquisition.ObservingthatL2acquisitioninvolves“processesofconstructionandreconstruction,”Ellisshowshowthefrequency,form,andfunctionofconstructionsinteractaslearnersacquireanotherlanguage.LearnersofEnglish,forinstance,typically“firstacquirethemostfrequent,prototypical,andgenericexemplaroftheverbisland”(suchasgiveintheDoubleObjectconstruction);thisprocessappearstoholdtrueforacquisitionofotherconstructionaltypes.Crosslinguistictransfereffectsarealsoconsidered,andEllisdrawsattentiontopotentialbiaseffectsofknowledgeofconstructionsinL1ondevelopingacquisitionofL2constructions.ThecontributionconcludeswithasummaryofpossiblefuturedirectionsforresearchintoconstructionalapproachestoL2acquisition.

Afterthesetwoacquisitionchapters,GiuliaBenciniturnstopsycholinguisticissuesoflanguageproduction.Shefirstoutlinesthepsycholinguistic“consensusmodel”forproduction,whichsetsoutthenecessaryexplanandaforasuccessfultheoryoflanguageproduction,aswellastheoverallarchitectureofthelinguisticsystem.Thenshesurveystheempiricalevidenceforandagainstthetwo-stagemodeloflanguageproduction,whichassumesseparatelevelsforfunctional(semanticandsyntactic)processing,aswellasforpositionalprocessing(linearizationandphonologicalrealization).Inaddition,shegivesanoverviewofexperimentsthatprovidesupportfor“theexistenceoflexicallyindependentstructurebuildingoperationsinlanguageproductioninadditiontolexicalrepresentations.”Assheargues,suchlexicallyindependentstructuralprocessesreceiveastraightforwardinterpretationasabstractconstructionsinaConstructionGrammarframework.Finally,shealsodiscussesthehypothesisthatthesameprocessesunderlieunderstandingandproducinglanguageandadducescross-modalprimingevidenceinfavorofthisclaim.

Thefinalcontributiontothispart,byFriedemannPulvermüller,BertCappelle,andYuryShtyrov,movesthediscussionintotherealmsofneuroscience,andthe“neurophysiologicalplausibility”ofsomeoftheclaimsofConstructionGrammarwithregardtosyntacticstructures.Astheauthorsshow,evidencefromneurosciencehashighlyimportantrepercussionsforlinguistictheorybuildingingeneral.Theyalsoarguethattheconstructionistenterprisereceivesconsiderableempiricalsupportfromneurolinguisticstudies.Specifically,theauthorsattempttoillustratesomeofthewaysinwhichtherelationshipbetweensyntacticprocessingandconstructionalknowledgemaybeelucidatedbydrawingon“braintheoriesofgrammar.”Theresearchreportedoninthischaptersuggeststhat“[r]ecurrentwordsequencesandmoreabstractconstructionsgeneralizingover…specificsequencesarealsostoredinthebrain,”thatthebrainrespondsdifferentlytogrammaticalandungrammaticalsequences,andthathumanlanguagemaylegitimatelybeconsidered“asanextensionofother(andolder)neurocognitivefeatures,”whichsupportsmanyoftheclaimsmadebyproponentsofConstructionGrammar.Incontrasttosomeconstructionistapproaches,theyalsocontendthatneurologicalevidenceindicatesthatitmakessensetopostulateflexibleconstructionaltemplates(i.e.,syntacticlinkageoperationsbetweenmeaningfulunits),asdistinctfromlexicalconstructionstorage.

ThefinalpartofthehandbookisconcernedwithV.LanguageVariationandChange.MirjamFried'scontributionconsiderstheapplicationofprinciplesofConstructionGrammartolanguagechange.InadditiontothepresentationofaparticularchangeinanOldCzechmorphologicalconstruction,Friedaddressesmoregeneralissuesforaconstructionalapproachtogrammaticalizationphenomena,andconsiderssomeofthewaysinwhichconstructionsmaychange‘internally,’aswellaschangesthatrelatetoaconstruction'sdistributionwithinlargerconstructions.Theresearchreportedinthischapterdrawsattentiontothegradualnatureofconstructionalchange,themicro-stepsinvolvedatdifferentconstructionallevels,andtheimportanceofcontext.

JóhannaBarðdal'scontributiontakesaconstructionalapproachtothedifficultissueofsyntacticreconstruction.Barðdalpointsoutthathistorical-comparativereconstructionisusuallyconcernedwithlexical,morphological,orphonologicalreconstruction,anddoubthasbeencastonthevalidityofsyntacticreconstruction.Shearguesthataconstructionalapproachtolanguage,whichtakesform-meaningpairingsasgrammaticalprimitives,isan“optimaltheoreticalframeworkforreconstructingsyntax.”DrawingonevidencefromArgumentStructureconstructionsinarangeoflanguages,Barðdalillustratesthataconstructionalapproachovercomessomeoftheallegeddifficultieswithsyntacticreconstruction(e.g.,thatsyntacticcognatesarelacking,orthatsyntacticchangelacksregularity).

AsthechaptersinotherpartsbyStefanowitschandGriesshow,thereisacloserelationshipbetweenconstructionalapproachestolanguageandcorpuslinguistics.MartinHilpert'schaptershowshowcorpus

Page 26: The Oxford Handbook of Construction Grammar

Construction Grammar

Page 8 of 9

linguisticshassignificantlyaffectedourunderstandingofconstructionalchange.Hilpertfocusesonthreedifferentkindsofconstructionalchange:onchangesinaconstruction'sfrequency,initsform,andinitsfunction.Frequencychangesinvolvechangeintextfrequencyandproductivity;formchangesincludemorphophonemicchange,aswellaschangeinargumentstructure;functionalchangesmayinvolvemetaphoricandmetonymicchange,aswellaschangesinvolvinganalogicalextension.Inallcases,Hilpertshowsthatdataretrievedfromcorporacanrevealfine-graineddetailregardingthegradualnatureofconstructionalchange,andfacilitatethequantitativeanalysisthattypifiesmuchworkonlinguisticvariationandchange.

Ongoingchangeinlanguagemaybemanifestinsynchronicvariation,andsynchronicvariationindifferentdialectsanddiscoursesisthesubjectofthecontributionbyJan-OlaÖstmanandGraemeTrousdale.Inthischapter,themodelingofinherentvariabilityincognitivelinguisticsingeneralandinConstructionGrammarinparticularisillustratedbywayofthreecasestudiesthatpresentdatathatdemonstratesstructuralvariation,contextualvariation,andvariationinfrequencyofuseassociatedwithdifferentregionalvarietiesofthesamelanguage.Furthermore,theauthorsshowhowtheusage-basednatureofmuchresearchinConstructionGrammarmaybeappliedtoemergentvariationindiscoursestructures,particularlyindialogiccontexts.

InadiscussionoftherelationshipbetweenConstructionGrammarandsociolinguistics,WillemHollmannillustratessomeofthewaysinwhichbothsociolinguistictheoryandtheoriesofconstructionsinlanguagemaybesynthesized.HollmanndrawsonCroft'snotionof‘social-cognitivelinguistics,’thatis,ontheinteractionalnatureoflanguageinuse,onGries'sworkoncorrelationsbetweenparticleplacementinEnglishverb-particleconstructionsandspokenvs.writtenlanguage,onGrondelaers'sworkoncorrelationsbetweenvariationinDutcher-presentatives,regionalandregisterdifferences,andfinallyonhisownworkwithAnnaSiewierskaongrammaticalandmorphophonologicalvariationinLancashireEnglishconstructions.Throughoutthearticle,HollmannstressestheneedforgreaterdialoguebetweenConstructionGrammariansandsociolinguisticsinordertoachieveapsychologicallyandsociallyplausibleaccountoflinguisticvariation.

Insum,thishandbookpresentsacomprehensiveaccountofcurrentworkonConstructionGrammar(itstheoreticalfoundations,anditsapplicationstoandrelationshipwithotherkindsoflinguisticenquiry).Wearegratefultothecontributorsfortheirdetailed,informed,andauthoritativeaccountsofresearchintoconstructionalapproachestohumanlanguage.

Notes:

(1.)http://lyrics.doheth.co.uk/songs/aaron-watson/shutupanddance/stuck-between-a-rock-and-a-heartache.php.

(2.)ForanoverviewoftheevolutionofChomskyanMainstreamGenerativeGrammar,cf.CulicoverandJackendoff2005;Jackendoff2002a.

(3.)AFrameNetanalysisofJapanesecanbefoundat<http://jfn.st.hc.keio.ac.jp/people.html>(cf.Ohara2009,2011).

ThomasHoffmannThomasHoffmannisAssistantProfessorattheUniversityofOsnabrück.Hismainresearchinterestsareusage-basedConstructionGrammar,synchronicsyntacticvariationandWorldEnglishes.HehaspublishedarticlesininternationaljournalssuchasCognitiveLinguistics,EnglishWorld-WideandCorpusLinguisticsandLinguisticTheory.Ontopofthat,hismonographPrepositionPlacementinEnglish(2011)waspublishedbyCambridgeUniversityPress.Currently,heiswritingthetextbookConstructionGrammar:TheStructureofEnglishfortheCambridgeTextbooksinLinguisticsseries.

GraemeTrousdaleGraemeTrousdaleisaSeniorLecturerattheUniversityofEdinburgh.WithNikolasGisborne,heeditedConstructionalApproachestoEnglishGrammar(deGruyter,2008).HehaspublishedanumberofjournalarticlesandbookchaptersonconstructionalapproachestovariationandchangeinEnglish.HeistheauthorofAnIntroductiontoEnglishSociolinguistics(EUP,2010),andisco-editorofGradience,GradualnessandGrammaticalization(withE.C.Traugott,Benjamins2010),andTheOxfordHandbookofConstructionGrammar(withT.Hoffmann,OUP,2013).HislatestbookisConstructionalizationandConstructionalChanges(withE.C.Traugott,OUP,inpress)[email protected]

Page 27: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 1 of 12

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0002

ConstructionistApproachesAdeleE.GoldbergTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterhighlightsthefundamentalassumptionssharedbyallconstructionistapproaches,distinguishingthemfrommainstreamgenerativegrammar.Inparticular,phrasalconstructions,liketraditionallexicalitems,arelearnedpairingsofformandfunction.Grammardoesnotinvolveanytransformationalorderivationalcomponent.Phrasalconstructions,words,andpartiallyfilledwords(akamorphemes)arerelatedinanetworkinwhichnodesarerelatedbyinheritancelinks.Languagesareacknowledgedtovaryinwide-rangingways;thecross-linguisticgeneralizationsthatdoexistareexplainedbydomain-generalcognitiveprocessesorbythefunctionsoftheconstructionsinvolved.

Keywords:constructionistapproaches,generativegrammar,phrasalconstructions,surfacestructure,cross-linguisticvariability,cognitivelinguistics,linguistictheories

2.1.Introduction

Mostoftheapproachesrepresentedinthisvolumeshareimportantunderlyingassumptionsthatpositiontheentiretyoftheseapproachesatafarremovefrommainstreamgenerativegrammar.Itisnotclearthatthedifferencesinapproachandemphasisbydifferentresearchersrisetothelevelofmandatingdistinctframeworks,sothischapterwillemphasizewhattheyhaveincommon(cf.alsoHoffmannandTrousdale,thisvolume).Themajortenetsthataresharedbyconstructionistapproaches,distinguishingthemfrommainstreamgenerativegrammar,includethefollowing:

1.Grammaticalconstructions:Phrasalconstructions,liketraditionallexicalitems,arelearnedpairingsofformandfunction(Lakoff1987;Fillmoreetal.1988;Wierzbicka1988;Goldberg1995,2006a;BirnerandWard1998).2.Surfacestructure:Grammardoesnotinvolveanytransformationalorderivationalcomponent.Semanticsisassociateddirectlywithsurfaceform(Goldberg2002;CulicoverandJackendoff2005).3.Anetworkofconstructions:Phrasalconstructions,words,andpartiallyfilledwords(akamorphemes)arerelatedinanetworkinwhichnodesarerelatedbyinheritancelinks(Lakoff1987;Langacker1987;Fillmoreetal.1988;Wierzbicka1988;Hudson1990,2007b;Goldberg1995;Booij2010).4.Crosslinguisticvariabilityandgeneralization:Languagesareacknowledgedtovaryinwide-rangingways.Thecrosslinguisticgeneralizationsthatdoexistareexplainedbydomain-generalcognitiveprocessesorbythefunctionsoftheconstructionsinvolved(Croft2001;Haspelmath2008;EvansandLevinson2009;Boas2010a).

Afinaltenetisincluded,eventhoughitisnotsharedequallybyallconstructionistapproaches,becauseitallowsforconstructionistapproachestointerfacenaturallywiththeoriesofacquisition,languageprocessing,andlanguagechange:

Page 28: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 2 of 12

5.Usage-based:Knowledgeoflanguageincludesbothitemsandgeneralizations,atvaryinglevelsofspecificity(Langacker1988b;BarlowandKemmer2000;Lievenetal.2003;Tomasello2003;BybeeandEddington2006;Goldberg2006a).

Itisnotalogicalnecessitythatallofthesetenetsbeadopted,andnotallresearchersemphasizeeachoneequally(Gonzálvez-GarcíaandButler2006).However,thetenetsdocohereinimportantways.Eachofthesefiveprinciplesisdiscussedbelowinsections2.2–2.6.

MostconstructionistapproacheshavebeenheavilyinfluencedbyseveraladditionaltenetstraditionallyassociatedwithCognitiveLinguistics.Ileavetheseasideforpresentpurposes(butcf.Goldberg1996,2011bandreferencescitedthereinfordiscussion).

a.Semanticsisbasedonspeakers’‘construals’ofsituations,notonobjectivetruthconditions(Talmy1976;Jackendoff1983;Fauconnier1985;Lakoff1987;Langacker1987,1988a;Goldberg1995;Michaelis2004).b.Semantics,informationstructure,andpragmaticsareinterrelated;allplayaroleinlinguisticfunction.Suchfunctionsarepartofouroverallconceptualsystemandnotaseparatemodularcomponent(Haiman1980;Fillmore1985b;Lakoff1987;Langacker1987;Talmy1988;Lambrecht1994;Goldberg1995;Leino,thisvolume)c.Categorizationdoesnotinvolvenecessaryandsufficientconditions,butrathergeneralizationsoverexemplars,typicallyincludingprototypicalexemplarsandconventionalizedextensions(Rosch1973;Fillmore1977b;RoschandMervis1975a;Haiman1985b;Lakoff1987;Brugman1988;Heine1992;Taylor1995;RossandMakin1999;Talmy2000;Abbot-SmithandTomasello2006;Ibbotsonetal.2012).d.Theprimaryfunctionoflanguageistoconveyinformation.Thus,formaldistinctionsareusefultotheextentthattheyconveysemanticorpragmatic(includinginformationtheoretic)distinctions(Bolinger1977;Haiman1985b;Lakoff1987;Langacker1987;Wierzbicka1988;Croft1991;Goldberg1995;AckermanandGoldberg1996).e.Theroleofsocialcognitionandbodilyexperienceareviewedasessentialtoaccountsoflearningandmeaning(Fillmore1977b;LakoffandJohnson1980;Matlock2004;Bergenetal.2007;Tomasello2008;Croft2009b).

2.2.GrammaticalConstructions

Constructionsaredefinedtobeconventional,learnedform-functionpairingsatvaryinglevelsofcomplexityandabstraction(Goldberg1995,2006a).Thisdefinitionismeanttohighlightthecommonalitybetweenwordsandlargerphrasalunits.VarioustypesofconstructionsareprovidedinTable2.1.

Noteveryresearcherwhoadoptsaconstructionistperspectiveexplicitlyusestheterm.Nordoeseveryoneapplythetermconstructiontothewordormorphemiclevel,althoughsincedeSaussure(1916),linguistshavebeencomfortableacknowledgingthatroots(orlemmas)andaffixesareconventionalpairingsofformandfunction.Byacknowledgingthatmorphemesareemergentgeneralizationsoverexistingwordsintheformofpartiallyfilledtemplates,theperspectiveisbroughtinlinewiththegrowingconsensusamongmorphologists(Aronoff1983;Blevins2001;AckermanandNikolaeva2004;Booij2010).

Descriptiveadequacyensuresthatsubtlefactsaboutsemanticsanduseofparticularconstructionsneedtobeaccountedfor.Moreover,constructionistsaimtoaccountforallaspectsofgrammar,includingnotonly‘core’aspectsofgrammarbutalsolow-frequencyorunusualconstructionsthatothertheoriesmightrelegatetothe‘periphery’or‘residue.’Inthisspirit,weaimtofullyaccountforsubregularities,aswellasthemostgeneralpatterns(Lakoff1970;Fillmoreetal.1988;Culicover1999).

Asanexampleofasubregularity,considerthePNconstructionthatinvolvesthecombinationofaprepositionandabarecountnounasintheexamplesbelow:

(1)Shewasinprison.(2)Shecamefromschool.(3)Sheleftforwork.(4)Theyareonvacation.

Page 29: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 3 of 12

(5)Hewasinhospital.(BritishEnglish)

Table2.1.Constructionsatvaryinglevelsofcomplexityandabstraction

Construction Examples

Word Iran,another,banana

Word(partiallyfilled) pre-N,V-ing

Idiom(filled) Goinggreatguns,givetheDevilhisdue

Idiom(partiallyfilled) Jog<someone's>memory,<someone's>fortheasking

Idiom(minimallyfilled)TheXertheYer Themoreyouthinkaboutit,thelessyouunderstand

Ditransitiveconstruction:SubjVObj1Obj2(unfilled)

Hegaveherafishtaco;Hebakedheramuffin

Passive:SubjauxVPpp(PPby)(unfilled) Thearmadillowashitbyacar

Normallycommoncountnounswithinprepositionalphrasesrequiredeterminers,sotheformofthisconstructionisclearlyunusual.Moreover,nomodificationofthebarenounisallowed:

(6)*Shewenttobigbed.(Cf.Shewenttothebigbed.)

Theconstructionalsorequiresaspecialinterpretation.Onedoesnotgotoprisonasawarden,butonlyasaprisoner.Peoplearelikewisenotinschoolascustodialstaffastheymightbeiftheywereintheschool.Theconstructionconveysthatastereotypicalactivityorroleassociatedwiththenominalreferentisinvolved(cf.alsoCarlson2006).Theconstructionismotivatedbythefactthatbarenominalsareunmarkedfordefinitenessandspecificityandthenounsinvolvedarenonspecific.Noticethattheyarenotnaturallyreferredtobyanaphora(e.g.,Shewenttobed.??Itwasveryfluffy.).Moreover,theinseparabilitythatidentifiesthePNconstructionisahallmarkofphrasesthatarelexicalizedtosomeextent,andlexicalizationonlygenerallyoccurswhenawordis‘name-worthy’(e.g.,Carlson2006;cf.alsoGoldberg2010).

ThePNpatternisclearlynotfullyproductive.Forexample,AmericanEnglishspeakersdonotgotohospitaloronholidayalthoughBritishEnglishspeakersdo.Andneithergroupcango*tobath,*tocomputer,*tooffice,*tokitchen,or*toairporteventhoughthereexiststereotypicalrolesoractivitiesassociatedwiththeseplaces.

Constructionsthatshowlimitedgeneralizationarenothardtocomeby.Therearethousandsofcollocations,prefabricatedutterances(‘prefabs’),idioms,andminorconstructionsthatbuckthetrendsofalanguageinunexpectedways(Lakoff1970;PawleyandSnyder1983;Culicover1999;Jackendoff2002a;Wray2002b;Goldberg2006a).Thesecanoftenberelatedtomoregeneralconstructions,justasthePNconstructionisaspecialinstanceofthemoregeneralPrepositionalPhrase(PNP)construction.Byrelatingconstructionstooneanotherinadefaultinheritancehierarchy,wecancaptureboththeregularandtheirregularaspectsofeachconstruction(seesection2.4;seealsoKay,thisvolume,foradifferentperspective).

Muchofourknowledgeoflanguageisquitegeneral,andthisknowledgeiscapturedwithinthenetworkofconstructionsaswell.Forexample,ArgumentStructureconstructions(Goldberg1995)aretheform-functionpairingsthatconstitutethebasic‘linking’patternsofsimpleclauses(e.g.,Jackendoff1983;BresnanandKanerva1989;Pinker1989;Grimshaw1990;Dowty1991;Gleitman1994;LevinandHovav1995);or‘lexicaltemplates’overlainonspecificverbs(HovavandLevin1998).Constructionalterminologyisadoptedhere,becausethephrasalform-meaningcorrespondenceshavebeenarguedtoexistindependentlyofparticularverbs(Goldberg

Page 30: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 4 of 12

1989,1995;Jackendoff2002a).

AsanexampleofanArgumentStructureconstruction,considertheDitransitiveconstruction,whichinactiveformhastwopostverbalNPsasin(7):

(7)Shegavehimabook.(Ditransitiveconstruction)

Thisconstructionisreliablyassociatedwiththemeaningof‘transfer’(Green1974;Erteschik-Shir1979;Pinker1989;Goldberg1992;Goldbergetal.2005;HovavandLevin2005).Evidenceoftheassociationbetweentheformandmeaningcomesfromtheinterpretationofnonsenseverbs.Whenpeopleareaskedwhatthenonsenseverbmoopmeansin(8),afull60%ofthemrespondwith‘give,’andtherestoffermeaningsthatpreservethemeaningofliteralormetaphoricaltransfer(e.g.,‘tell’)(Ahrens1995):

(8)Shemoopedhimsomething.

KaschakandGlenberg(2000)havedemonstratedexperimentallythatsubjectsrelyonconstructionalmeaningwhentheyencounternounsusedasverbsinnovelways(e.g.,tocrutch).Inparticular,theyshowedthatdifferentconstructionsdifferentiallyinfluencetheinterpretationsofthenovelverbs.Forexample,Shecrutchedhimtheball(ditransitive)isinterpretedtomeanthatsheusedthecrutchtotransfertheballtohim,perhapsusingitasonewouldahockeystick.Ontheotherhand,Shecrutchedhim(transitive)ismorelikelytomeanthatshehithimovertheheadwithacrutch.KaschakandGlenbergsuggestthattheconstructionalpatternspecifiesageneralsceneandthatthe‘affordances’ofparticularobjectsareusedtospecifythesceneindetail(cf.Pustejovsky's1995:qualiastructures).Itcannotbethesemanticsoftheverbthatdeterminestheinterpretationinisolation,becausethewordform(crutch)isnotrepresentedinthelexiconasaverbbutasanoun(cf.alsoGoldwaterandMarkman2009;Allenetal.2012;JohnsonandGoldberg2012).

Jackendoff(thisvolume)takesissuewiththeideathatconstructionsgenerallyrelateformtofunction,arguinginsteadthattherearemanypurelysyntactic(andpurelysemantic)generalizations.Buttheapparentdisagreementmaybemoreterminologicalthancontentful,inthatJackendoffseemstoassumethatfunctionsarenecessarilyquiteconcrete.Infact,thefunctionofabstractformaltemplatesisoftenquiteabstract.Forexample,theVerbPhraseconstructionthatJackendoffsuggestsismeaninglessdoeshaveafunction,namely,predication.TheTransitiveconstructionislikewiseassociatedwithtwo-argumentpropositionalsemantics(Fisheretal.1994;Gleitman1994).Sinceitisfunctionthatisimportant,notpropositionalsemantics,eventhe‘meaningless’phrasefalaladoeshaveafunctionasanemotivepartofaparticularChristmascarol.

Itispossibleforaconstructiontobeambiguous:thesameformbeingpairedwithunrelatedfunctions,justlikeambiguouswords.However,asistrueofwords,ambiguityisrelativelyrarecomparedwithpolysemy.Moreover,ifasinglephrasalpatternweretrulyassociatedwithunrelatedfunctions,suchasthefunctionsofconjunctionandpassivization,orreflexivizationandmodification,thentheirdistributionalbehaviorisnotlikelytobeidentical.Whenbehaviordiverges,wegenerallydecidethatthesyntaxinvolvedisnotthesame.Thisisperhapswhypurelysyntacticgeneralizationsaresohardtofind.Whatistypicalisneitherambiguitynorfunctionalidentity,butratherpolysemy:thesameformoftenhasdifferentbutrelatedfunctions(Lakoff1987;Goldberg1992;Taylor1995).

2.3.SurfaceGeneralizations

Sincetheearliestdaysofgenerativegrammar,therehasexistedastrongtendencytofocusonalternations,consideringoneArgumentStructureconstructioninrelationtoanotherspecificroughparaphrase(Lakoff1970;Levin1993).Initiallythiswasaresultoftheemphasisontransformationsthatderivedonepatternfromanother.Whiletodaythereexistmanynonderivationaltheoriesforwhichthismotivationnolongerexists,thetraditionaloutlookhasnotcompletelylostitsgrip,ascanbeseenfromcontinuingfocusonpartialorincompletegeneralizationssuchasthe‘dative’alternationorthe‘locative’alternation.YetitisprofitabletolookbeyondalternationsandtoconsidereachArgumentStructureconstructiononitsownterms.Forexample,althoughonly(9a)canbeparaphrasedbyaditransitiveexpression(giveninparentheses),itpatternstogetherwith(9b,c,d)bothsyntacticallyandsemantically;theexamplesin(9)allinvolvesubjects,objects,andobliquepathphrases,andtheyallexpresscausedmotion.AsBakernotes,“itseemsartificialtosaythatthePPin[exampleslike9a]isnotalocationalpathaswell”(Baker1997:96;cf.alsoMarantz1997).TheCausedMotionconstructioncaptures

Page 31: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 5 of 12

thegeneralizationacrossalloftheexamplesin9(a–d)(Pinker1989;Goldberg1995):

(9)a.MinasentabooktoMel.(≈MinasentMelabook.)b.MinasentabooktoChicago.c.Minasentabooktowardthefrontoftheroom.d.Minasentabookthroughthemetaldetector.

RecognizingtheCausedMotionconstructionallowsustoaccountforthefactthataverbsuchascough,thatdoesnotlexicallyentailcausedmotion,canconveycausedmotionwhenitappearsinthisconstruction:

(10)“Iactuallyhadamothgoupmynoseonce.I…coughedhimoutofmymouth”(bikeforums.net/archive/index.php/t-292132)

Thus,wecanidentifyamoregeneralcorrespondencebetweenformandmeaningthatisnotnaturallycapturedattheleveloftheindividualverb:theCausedMotionconstruction.

Whenfacedwithpairsofsentencesthatsharewordsandinvolvedifferentargumentstructurepatterns,similaritiesaremostnaturallyattributedtothewords,anddifferencestotheargumentstructurepatterns.Paraphraserelationshipscanbecapturedbyexplicitreferencetothewords,particularlyrelationalwordssuchasverbs.

TheDitransitiveconstruction(asin(7)–(8))alsoprovidesaverybroadgeneralizationifweattendtosurfacestructureinsteadoftopossiblealternations.Althoughmanylinguistscontinuetotreat(regular)ditransitivesand“benefactive”ditransitives(suchasMinabakedMelacake)asdistinctconstructionsbecauseoftheirdifferentparaphrases(MinasentabooktoMel/MinabakedacakeforMel),bothtypesofditransitiveexamplespatternalikebothsemanticallyandsyntactically(Goldberg1989;2006a:26–33).

2.4.ANetworkofConstructions

Importantly,theconstructionsofagivenlanguagedonotformanunstructuredset.Relationshipsbetweenandamongconstructionsarecapturedviaadefaultinheritancenetwork.ForexamplethePNconstruction(e.g.,tobed)inheritsitswordorderfromthemoregeneral,abstractPrepositionalPhraseconstructionasinFigure2.1.ItdiffersfromthegeneralPPconstructioninspecifyinganNinsteadofanNPdaughter.ThePPconstructioncapturesthefactthatEnglishhasprepositionsinsteadofpostpositions;sincethewordorderoftheconstructionisfixed,itsphrasestructureisspecified.Degreeof‘entrenchment’(roughly,frequency)iscapturedviadegreeofboldface(followingthenotationofLangacker1987);clearlythegeneralPPconstructionismoreentrenchedthanthePNconstruction.

Notallconstructionsspecifywordorder.Forexample,ArgumentStructureconstructionsdonot,andthustheydonotdirectlydeterminephrasestructuretrees(paceMüller2006).Instead,wordorderisdeterminedbycombiningArgumentStructureconstructionswithconstructionssuchastheVPconstruction,Subject-Predicateconstruction,and/oraLong-distanceDependencyconstruction.

Figure2.1. Toyexampleofdefault(usage-based)inheritancehierarchythatrelatesinstancesofthePNconstructiontothemoregeneralprepositionalphrase(PP)construction

LetusconsidertheResultativeconstruction,asin(11)and(12),andthesortoffactsaboutitthatahierarchyis

Page 32: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 6 of 12

naturallyabletocapture.TheResultativeconstructioninvolvesasecondarypredicationanddesignatesachangeofstate(Simpson1983;Wechsler1997;Boas2003;HovavandLevin2001;GoldbergandJackendoff2004;Iwata2006).TheintransitiveandtransitiveversionsoftheResultativeconstructionarerelatedbya‘subpart’inheritancelinksincetheformerisapropersubpartofthelatter(Goldberg1995):

(11)Thenutbrokeapart.IntransitiveResultativeconstruction(12)Shebrokethenutapart.TransitiveResultativeconstruction

RelatingtheEnglish(Transitive)Resultativeconstructionexemplifiedin(13)totheCausedMotionconstruction(e.g.,14)byametaphoricalrelationshiplinkallowsustocapturecertaindistributionalpropertiesoftheformer.

(13)Shedrovehimtodistraction.Resultativeconstruction(14)ShedrovehimtoDetroit.CausedMotionconstruction

Inparticular,theresultativephrasebehaveslikealiteralpathphraseinsofarasit(a)cannotoccurwithditransitives(15,16)and(b)cannotbecombinedtocreateadiscontinuous‘path’(17)(seethe“UniquePathConstraint”ofGoldberg1992,1995;Matsumoto2007):

(15)*Shekickedhimtheballblackandblue.(16)*Shekickedhimtheballintothewindow.(17)*Shekickedhimblackandbluedownthestairs.

Thedefaultinheritancehierarchythatincludesbothwordandphrasalpatternsiswellsuitedforcapturingtrickycasesthathavebedeviledtraditionaltheories,forexample,complexpredicatesthatareinsomewayswordlikeandinotherwaysphrasal.Goldberg(2002)offersanaccountofthePersianComplexPredicate(CP)constructionthattreatstheCPasawordbydefault(cf.Family2006;Mueller2010).Thatis,aCPvoconstructionhastheexternalsyntaxofaV0,whichaccountsforawiderangeofpropertiesstronglyassociatedwithzerolevelstatus,includingtheformationofnominalizations,theresistancetoseparation,certainstressfactsandfrequentnoncompositionalverblikesemantics.

Figure2.2. AsummaryoftheconstructionscriticalforaccountingforPersianCPsandtheirzero-levelandphrasalproperties(seeGoldberg2002)

Atthesametimethegrammaralsohasmorespecificconstructionsthatservetooverridethespecificationsofthegeneralconstruction.Inparticular,thefutureauxiliary,fullyinflected,necessarilyintervenesbetweenhostandlightverb;thedirectobjectcliticalsomayintervenebetweenhostandlightverb.Thesefactsareindependentlymotivated,bygeneralwordorderandstressfactsofPersianandthediachronichistoryoftheCPasacomplement+verb.Adiagramoftherelevantnecessaryconstructions,whetherCP-specificornot,andtheirinteractionsisgiveninFigure2.2.“<”indicateslinearprecedence.Stressisindicatedbyan“/”accent.

Threeconstructionsarerequiredtoaccountforthefutureauxiliarygenerally(A),complexpredicatesgenerally(B),andcliticsgenerally(C).Thedaughtertwoconstructions(DandE)onlyservetospecifyconflictsbetweentheirmotherconstructions.Sincedefaultinheritanceisused,conflictsareallowed.Butthedaughternodemustexplicitlyresolvethem.

2.5.CrosslinguisticVariabilityandGeneralization

Constructionstypicallydonotexistsuigeneris;thatis,theyaregenerallynotfullyarbitrary.Relationshipsbetweenformandmeaningarenormallymotivated,andthuswefindrecurrentpatternscrosslinguistically.Butbecauseconstructionistapproachesdonotrelyoninnateuniversalprinciples,constructionsareexpectedtovaryintheirspecificscrosslinguistically,andthisdoesseemtobethecase(Croft2001;Haspelmath2008;EvansandLevinson

Page 33: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 7 of 12

2009;Boas2010a).Whatmaybecalledapassiveinonelanguagemaydifferfromapassiveconstructioninanotherlanguageinanumberofsubtlewaysincludingthepresenceorchoiceofauxiliary,thepresenceorchoiceofadpositionorcasethatmarkstheagentargument,possiblesemanticordiscourserestrictions,andoverallfrequencyinthelanguage.Findingtwoconstructionsintwodifferentlanguagesthatareabsolutelyidenticalinform,function,anddistributionisarareoccurrenceoutsideofcasesofshareddiachronichistoryorlanguagecontact.

Croft(2001),forexample,notesthatwordsthattranslateintoEnglishasnouns,adjectives,andadverbs,aswellasverbs,areinflectedforperson,aspect,andmoodinMakah,anAmericanIndianlanguage,andthatnowordsareinflectedforthesecategoriesinVietnamese.Hepointsoutthatthereforetense-aspect-moodinflectioncannotbetakenascriticalfordeterminingthecategoryofverbcrosslinguistically(unless,ofcourse,oneiswillingtosaythatallwordsareverbsinMakah,andnowordsareverbsinVietnamese).Croftgoesontopointoutthatnosyntactictestwillpickoutallandonlyentitiesthatonemightwishtocallverbs,nouns,adjectives,subjects,orobjectsacrossalllanguages.Moreover,heobserves,thatevenwithinasinglelanguage,agivencriterionoftenonlyapplieswithincertainlargerconstructions.Forexample,

IfonetakespassivizabilityasthecriterionforDirectObjectinEnglish,thenone'sconclusionswilltellussomethingaboutthePassive,notaboutsomeallegedlyglobalcategoryDirectObject….(C)onstructions,notcategoriesorrelations,arethebasic,primitiveunitsofsyntacticrepresentation.(Croft2001:46)

Thisisnottosaythattherearenostronguniversaltendenciesorimplicationaluniversalstobefoundacrosslanguages.Constructionistsarguethatsuchcrosslinguisticgeneralizationsarebetterexplainedviagrammar-externalexplanationssuchasuniversalfunctionalpressures,iconicprinciples,andprocessingandlearningconstraints.Letusconsiderafewexamplesinordertoillustratethepoint.

Thereisatendencyforlanguagestohavesomesortofconstructionwemightrefertoasa‘passive.’ThesePassiveconstructionsareidentifiedbytheirrelatedfunctions:theyareconstructionsinwhichthetopicand/oragentiveargumentisessentially‘demoted,’appearingoptionallyornotatall.Thefactthattopicsandagentsaretypicallyexpressedinsyntacticallyprominentslotsismotivatedbytheirfunctions.ThefactthatspecialPassiveconstructionsexistthatallowthemtobeexpressedinnonprominentslotsallowsspeakersadegreeofflexibilityandisthereforealsomotivated.Atthesametime,certainlanguages,suchasChoctaw,donotcontainsubject-demotingpassives(Heath1977;VanValin1980).

Anothercrosslinguistictendencyistheso-called‘head-direction’parameter.Languagesinwhichverbsappearattheendofsentenceshavebeenshowntogenerallyhavepostpositionsandpostnominalmodifierswhilelanguageswithverbsappearingbeforetheirnonsubjectcomplementstendtohaveprepositionsandpronominalmodifiers.Thisisshownin(18)below:

(18)HeadInitialLanguages: [V…] [P…]

HeadFinalLanguages: […V] […P]

This‘head-directionparameter’haslongbeenusedasanexampleofapurelysyntacticgeneralizationthatrequiresaninnateuniversalstipulation.Childrenwouldthenonlyhavetodeterminewhereinthesentencetheverbsintheirlanguageappearandcouldthendeducefromthatwheretoexpectallothertypesofheads.

However,asiswellknown,thisgeneralizationisnotwithoutexceptions.Persian,forexample,isaverb-finallanguage,buthasprepositionsinsteadofpostpositions.Inaddition,sincechildrenmuststilllearntheformsandmeaningsofthewordsintheirlanguage,includingverbs,adpositions,andnouns,theorderingofelementsinthesentenceisapparentduringthislearning,whichcallsintoquestiontheconceptualnecessityofanyinnategeneralization.

Still,thecrosslinguistictendencyislikelynotaccidental.Diachronicprocessesmaywellprovideabetteraccountoftherelationshipbetweenverbsandadpositions,sinceadpositionsoftendevelopfromverbs(Nichols1986).Ithasalsobeenhypothesizedthatthetendencyforheadstosystematicallyeitherprecedeorfollowtheircomplementslendsaprocessingadvantage(Hawkins1994,2004).RealiandChristiansen(2005)havesupportedthisideabydemonstratingthatsimple-recurrentnetworksfoundconsistenthead-complementorderingseasiertolearnthanmixedsystems;thatis,giventhechancetoadapttotheir‘learners,’thelanguagesintheirsimulation,

VP PP

VP PP

Page 34: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 8 of 12

overtime,cametohaveaconsistenthead-complementordersimilartothatfoundinreal-worldlanguages.

Othergeneralizationsthathavebeenarguedtorequirerecoursetoinnateprincipleshavealsofoundbetterexplanationselsewhere.Lidzetal.(2003)proposedthatchildrencomehard-wiredwithspecificknowledgethatthenumberofovertlyexpressedcomplementsshouldmatchthenumberofsemanticparticipants.Wecanagreethatlearnerspayattentiontothenumberofreferringexpressionsinaclauseasanindicatorofthepropositionalmeaning.However,thetendencyforthenumberofnounstoreflectthenumberofsemanticparticipantsfindsanaturalexplanationfromthedomainofGriceanpragmatics.Anyreferringexpressionshouldbeassumedtoberelevanttothetopicathand,andanyargumentthatisrelevantandnonrecoverablefromdiscourseneedstobeindicatedinsomeway(Goldberg2004b).Beyondthispragmaticgeneralization,nosyntacticstipulationisneeded.Thepragmaticgeneralizationssaynothingaboutargumentsthatareirrelevantorrecoverable;thisisanadvantage,sincelanguagesandconstructionswithinagivenlanguagetreatrecoverableandirrelevantargumentsdifferently.Forexample,manyoftheworld'slanguages(e.g.,ChineseandKorean)readilyallowrecoverableargumentstobeomitted;andevenEnglishallowspatientargumentsthatareparticularlydeemphasizedinthediscoursevis-à-vistheverbtobeomitted(Goldberg2000).

Theoftencitedlinkinggeneralizationsof,forexample,Dowty(1991)canbesummedupasfollows:insimpleactiveclauses,ifthereisasubjectandanobject,andifthereisanagentandapatient,thentheagentrolewillbeexpressedbythesubjectandtheundergoerroleasdirectobject.Thisisamodestproposalthathasbeentakenbysometoexpressaninnatelinguisticuniversal.Inreality,thefactsareevenmoremodest:therearesyntacticallyergativelanguagesinwhichagentsarenotgenerallyexpressedassubjects,therearemanylanguagesthatdonothavecanonicalsubjects,andtherearemanyconstructionswithinagivenlanguagethatviolatethegeneralizations(e.g.,passiveswhichexpresstheagentargumentasanoblique).Butagain,thereissomethingtothegeneralization.Thefactscanberestatedasfollows:semanticactorsandundergoerstendtobeexpressedinformallyprominentslots.Prominentsyntacticpositionscanbedefinedasthosepositionsthatlicenseagreementand/orlackovertcaseand/ormaybeobligatory.Oncestatedthisway,thegeneralizationismuchlessmysterious:actorandundergoerargumentsaregenerallyexpressedinprominentslotscrosslinguisticallybecausehumanbeings’attentionisnaturallydrawntotheactorsandundergoersinevents(Goldberg2006a).

Othergeneralizationsabouthowformandmeaningtendtobelinkedacrosslanguagescanbeexplainedbyappealtoiconicandanalogicalprocesses(Haiman1985a;Givón1991)andtopragmaticgeneralizations(Levinson1983).ConstraintsonLong-DistanceDependencyconstructions(traditional‘islandconstraints’)appeartoyieldtoprocessingexplanationsthattakeintoaccountthefunctionoftheconstructionsinvolved(Deane1991;Erteschik-Shir1998;Kluender1998;VanValin1998;AmbridgeandGoldberg2008).

Thisshiftofperspectivefromseekingexplanationsintermsofsyntactic,innatestipulationstotryingtoaccountforgeneralizationsbyappealingtoindependentlymotivatedgeneralcognitivemechanismshasbeenechoedtosomeextentwithinmainstreamgenerativegrammaraswell.Forexample,thefactthatalllanguagesappeartohavenounandverbcategoriesmaybeexplainedbytheexistenceofcorrespondingbasicsemanticcategories(Baker2004).Inarecentpaper,Chomskygoessofarastosuggestthattheonlylanguage-specificinnateabilitythatisabsolutelyrequiredisrecursion,andthepointisraisedthatevenrecursionmightturnoutnottobespecifictolanguage(Hauseretal.2002).Infact,theclaimthatrecursionisdomain-specificisnotwellsupported(JackendoffandPinker2005).

Languagearisesfromthehumanconceptualsystem.Constraintsonlanguagesarisefromboundaryconditionsimposedbyconstraintsonmemory,attentionalbiases,andpragmaticrequirementsaboutthecontentthathumanscareaboutconveying.

2.6.Usage-BasedModel

Speakersareatonceimpressivelycreativeandimpressivelyrepetitive.Thecreativeaspectoflanguageallowsspeakerstoexpressnewideasinnewcontexts;thisaspectoflanguagehasbeenemphasizedwithinmuchoftheoreticallinguisticsforoverfiftyyears.

Newerresearchhasemphasizedtheextenttowhichspeechconsistsofformulaicorprefabricatedphrases(‘prefabs’)(PawleyandSnyder1983;Wray2002b).Theformulaicaspectpresumablysimplifiesthespeakers’task,

Page 35: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 9 of 12

inthatmanyutterancesorpartsofutterancescanbepulleddirectlyoffthementalshelfwithouthavingtobuildeveryutteranceanew,fromscratch.Inonestudyofa2-year-oldchild'sspontaneousspeech,63%ofthechild'sutteranceswerefoundtobeverbatimrepetitionsofthechild'sownearlierutterances(Lievenetal.2003).Itturnsoutthatchildrenarealsofasterandmoreaccurateatrepeatingutterancesthatoccurwithhighfrequencyinchildren'scorporawhenlexicalfrequencyandlengtharecontrolledfor(BannardandMatthews2008);thesameistrueforadults(Bod1998).Thiswouldseemtorequirethatsomethingabouthighfrequencycollocationsisretainedinmemory.

Oftentimesutterancesareatoncenovelandformulaicinthattheyinvolveconstructionsthatdictatemuchofthelexicalcontent,whilefillingopenslotsoftheconstructionwithnewlexicalcontent.Forexample,eveninyoungchildren'sspeech,nounsarefreelysubstitutedforoneanotherfromveryearlyon(Tomaselloetal.1997).Openslotsofvarioustypesexistinallphrasalconstructionsthatarenotcompletelyfixedidioms.Possessivephrasesareoftennotfixedinadvanced:onecanjoghismemory/hermemory/theirmemory/thedog'smemory.Verbtenseandaspectisalsooftenflexibleaswell:onecancommitaphonenumbertomemoryorhavecommittedittomemory.Onecantakeastrolloratripdownmemorylane(butcannot?visitmemorylane).Theseidiomscanberepresentedasin(19)–(21),wheretheverballexemeisrepresentedbycapitalletterstoindicatethattenseandaspectarenotspecified.

(19)JOG<someone's>memory(20)COMMIT<something>tomemory(21)TAKE<atrip>downmemorylane.

Clearlybothaspectsoflanguage,itscreativeanditsformulaiccharacter,areimportanttounderstandingourgreatfacilitywithlanguage.Largelyduetoarecognitionofthis,therehasbeenagrowingconvergenceon‘usage-based’modelsoflanguage.Thesemodelsarebasedontheideathatknowledgeoflanguageconsistsofanetworkofform-functioncorrespondencesatvaryinglevelsofspecificity.Particularlanguagesarelearnedbygeneralizingoverutterancesthatalearnerhasheardused,whilelanguageproductionandcomprehensioninvolvecombiningordecomposinganutteranceintoitsmorebasicform-functioncorrespondences(Bybee1985,2006,2010,thisvolume;Langacker1988b;BarlowandKemmer2000;Verhagen2002;Lievenetal.2003;Tomasello2003;Goldberg2006a;Abbot-Smithetal.2007;AlishahiandStevenson2008).

Justhowspecificisspeakers’knowledgeoflanguage?Memoryforanyexperienceisnecessarilypartiallyabstractinsofarastheexperienceisnotrecordedcompletely.Wemightrememberseeingakumquatbutwehaveabstractedawayfromthecolorofthekitchentableuponwhichitsat;wealsomayhavenotnoticedthetinyscratchinitssurfaceortheexactlengthofitsstem.Soourmentalrepresentationofanexperience,nomatterhowvivid,ispartiallyabstractedfromtheactualexperience.Thesamemustbetrueforanymemoryoflanguage.Moreover,memorydecaysovertime,leadingtoadditionalabstraction.

Instancesofphrasalidiomsthatareclearlyretainedinlong-termmemoryincludethoseinTable2.2.

Thereissomerecentevidencethatweretainnontrivialamountsofverbatimlanguageevenafterasingleexposure.Gurevich,Johnson,andGoldberg(2010)demonstratedthatparticipantsreliablyrecognizeandrecallfullsentencesthattheyareexposedtoonlyonceatabovechancerates(Experiments1and3).Thisfindingwasofinterestbecausethetextswerefairlylong(300words),thecontextwasnoninteractive,andnoadvancedwarningofamemory

Page 36: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 10 of 12

Table2.2.Examplesofphrasesthatarelikelystoredaspartofourknowledgeoflanguage,whetherornottheirinterpretationsarecompositional

You’vegottobekidding! Doublewhammy

wearout<one's>welcomeWhat'sup?Whatfor?

Eat,drinkandbemerryExcuse<poss>FrenchFacethemusic

ShootthebreezeAreyouallright?Tellmewhathappened.I’msorrytohearthat.Itjustgoestoshow

soonerorlaterWhatdidyousay?CanIcomein?Needanyhelp?Iseewhatyoumean.Blitheringidiot

testwasgiven.Verbatimmemorywasdemonstratedevenwhenlexicalcontentandmemoryforgistwerecontrolledfor(Experiments2and4).Themoststrikingfindingwasoneofincidentalrecall:evenafterasix-daydelay,participantsreliablyrecalledsentencestheyhadheardbeforewhenaskedtodescribescenes,eventhoughtheywerenotaskedtoreproducewhattheyhadheard(Experiment5).Thusthereisagreatdealofevidencedemonstratingthatweretainanimpressiveamountofitem-specificknowledgeincludingrelativefrequenciesofusage.

Wealsocategorizeorgeneralizetheinputwehearintopatternsbasedonformandfunction(e.g.,Langacker1988b;BarlowandKemmer2000;Tomasello2003;Goldberg2006a;BoydandGoldberg2011).Thisisdiscussedinthefollowingsection.

2.7.Creativity

Creativitystemsfromgeneralizinginstancestoformmoreabstractconstructionswithopenslots.Theopenslotsofaconstructionallowconstructionstocombine.WhenaconstructionoftypeYcontainsaslotofthesametypeY,theconstructionisrecursive.

Severalfactorsthatareknowntoberelatedtoinductionmoregenerallydeterminehowproductiveaconstructionis.Theseincludetypefrequency,variabilityofattestedinstances,similarity,andstatisticalpreemption(Barðdal2008;BoydandGoldberg2011;SuttleandGoldberg2011).Thus,productivitydependsonhowthewitnessedinstancesaredistributedandhowrelatedthepotentialcoinageistothoseinstances(andwhetherthereexistsaprepackagedalternativethatstatisticallypreemptsit).

Anactualexpressiontypicallyincludesthecombinationofatleasthalfadozendifferentconstructions.Forexample,theexpressionin(22)involvesthelistofconstructionsgivenin(23a–g).

(22)WhatdidMinabuyMel?(23)

a.Ditransitiveconstructionb.NonsubjectQuestionconstructionc.Subject-AuxiliaryInversionconstructiond.VPconstructione.NPconstructionf.IndefiniteDeterminerconstructiong.Mina,buy,Mel,what,doLexicalconstructions

Inthisway,thesameDitransitiveconstructionisinvolvedintheactivedeclarativeform,aswellasintopicalized,clefted,orquestionedforms.Thatis,therecipientargumentisanObjectwhetherornotitappearsdirectlyaftertheverborwhetheritappearsasadistantlyinstantiatedquestionword.Itis,forexample,the(Non-echo)Question

1

Page 37: The Oxford Handbook of Construction Grammar

Constructionist Approaches

Page 11 of 12

constructionthatdeterminesthefactthatthewh-wordappearssentence-initiallyinEnglish.

AmbridgeandGoldberg(2008)provideevidencethatsuggeststhatspeakersdeterminewhichconstructionscanbecombined,atleastinpart,onthebasisofthecompatibilityoftheinformationstructurepropertiesoftheconstructionsinvolved.Weinvestigatedtherelative‘island’statusofthefollowingSentenceComplementconstructions:‘bridge’verbcomplements,manner-of-speakingverbcomplements,andfactiveverbcomplements.Questionnairedatademonstratedastrongcorrelationbetweenacceptabilityjudgmentsandanegationtestusedtooperationalizethenotionof‘backgroundedness.’Thefindingsstronglysupporttheideathattheseconstructionsactasislandstowh-extractiontothedegreethattheyarebackgroundedindiscourse.

2.8.RepresentationofConstructions

Therearevariousformalismsforconstructions.Arguably,themostcompleteandwell-workedoutisthefeature-valuematricesusedbyseveralresearchersintheconstructionistframework(cf.MichaelisandLambrecht1996;Sag1997;KayandFillmore1999;Fillmore,thisvolume;Michaelis,thisvolume).Agreatdealofworkwithincomputationallinguisticshasbeenbuiltuponthatformalism;nomainstreamgenerativegrammarformalismcomesclosetoitintermsofdetail,rigor,orcoverage.Jackendoffhasdevelopedanotherexcellentformalismthatfocusesmoreonsemanticrepresentation(Jackendoff1990;CulicoverandJackendoff2005).WorkbyJeffElman,AmyPerfors,MortenChristensen,HaroldBayaan,BenBergenandNancyChang,andRensBodhasdevelopedotherformalisms.

IhaveavoidedusingallbutthemostminimalformalizationinmyownworkbecauseIbelievethenecessaryuseoffeaturesthatformalismrequiresmisleadsresearchersintobelievingthattheremightbeafinitelistoffeaturesorthatmanyormostofthefeaturesarevalidincrosslinguisticwork.Thefactsbeliethisimplication.Themeaningsorfunctionsofwordsandconstructionsdonotlendthemselvestosemanticdecomposition(cf.Fillmore1977b;Lakoff1987;Langacker1987;Goldberg1995;paceJackendoff1983;Wierzbicka1988),andoften-suggestedsyntacticprimitivessuchasnoun,subject,agreement,oragentactuallyvarycrosslinguisticallyaswell(cf.,e.g.,Croft2001).Again,however,decisionsaboutwhetherandhowtoformalizeneednotnecessarilydivideconstructionists.

2.9.ComparisonwithMainstreamGenerativeandTraditionalFunctionalFrameworks

TheemphasisonthepairingoffunctionwithformiswhatsetsConstructionGrammarsapartfrombothothergenerativeapproaches(whichtendtodownplayfunction)andotherfunctionalapproaches(whichtendtodownplayform).Atthesametime,constructionistsbringthesetwoapproachestogetherinsomeways.Constructionistsrecognizetheimportanceoftwomajorquestionsthathavebeenbroughttotheforebygenerativegrammarians:(1)Howcanallofthecomplexitiesoflanguagebelearnedsuchthatweareabletoproduceanopen-endedsetofutterances?And(2)howarecrosslinguisticgeneralizations(andlanguageinternalgeneralizations)accountedfor?

Certainmainstreamgenerativegrammarframeworkssharethebasicideathatsometypeofmeaningisdirectlyassociatedwithsometypeofform,independentlyofparticularlexicalitems(cf.alsoBorer1994;HaleandKeyser1997;Marantz1997).Becausesyntaxplaysaroleincontentfulmeaning,theseotherapproacheshavebeencalled“neo-constructionist.”However,theapproachesarefundamentallydifferentfromthetypeofconstructionistapproachesjustoutlined.Forexample,thesemainstreamgenerativeaccountsdonotadoptanonderivational(monostratal)approachtosyntax,butappealinsteadtounderlyinglevelsofrepresentationinwhichconstituents(orentitiesthatareneverrealized)movearoundabstracttrees.Moreover,theseaccountsemphasizeroughparaphrasesinsteadofspeakers’detailedconstrualsofsituations.Thus,neitherthesyntaxnorthesemanticsisakintothatadoptedbyconstructionistsinthepresentvolume.Empiricalchallengesthat‘neo-constructionist’generativeaccountsfacearediscussedinsomedetailinGoldberg(2006a).

Constructionistshavealsoborrowedmuchfromtraditional‘functionalist’approaches,andwerelyheavilyontraditionalfunctionalistcorpusmethodologies(cf.Gries,thisvolume).Bothapproachesalsoemphasizetheusage-basednatureoflanguageandtheimportanceofgeneralcognitiveprocesses.

1

Page 38: The Oxford Handbook of Construction Grammar

Constructionist Approaches

2.10.Conclusion

Itisfairtosaythattheconstructionistapproachisthefastestgrowinglinguisticandinterdisciplinaryapproachtolanguage.Agrowingnumberofresearchershavemadeconstructionsthecornerstoneoftheirlinguistictheories(e.g.,Fillmoreetal.1988;Goldberg1995,2006a;Croft2001;Sagetal.2003),aswellastheoriesoflanguageacquisition(e.g.,Tomasello2003;seealsoAmbridgeandLieven2011),historicalchange(Traugott2008a,b,c;Bybee2010);morphology(Blevins2001;Spencer2001;Booij2002,2005),processing(Gries2003a),andmachinelearning(Bod1998;Perforsetal.2007).

Iprefertousetheterm,constructionistapproaches,ratherthanXorYGrammar.Thereisahugeamountofworktodobothintermsoflinguisticresearchandintermsofbuildingbridgesbetweenlinguisticsandotherdisciplines.Toomanyresearchersinrelatedfields(psychology,communication,anthropology,computerscience)gaveuponlinguisticswhenitbecameclearthatgenerativegrammarwaseitherunfathomable,wrong-headed,orboth.Thesameresearcherstendtofindconstructionistapproachesmuchmoreinteresting,accessible,anduseful,andwecouldstandtogainagreatdealbytalkingandworkingwiththem.Thereis,ofcourse,alsoagreatdealofworktobedoneinvolvingstandardlinguisticanalyseswiththegoalofgainingadeeperunderstandingoftherichnessandcomplexityoflanguage.Whileitis,ofcourse,importanttoopenlydiscusssubstantivepointsofdisagreement,thefieldmightbewellservedbymovingawayfromcreatingdivisionsamongcloselyrelatedframeworksbylabelingthemX,Y,orZGrammars.

Notes:

(1.)GiventhesyntacticspecificationsoftheDitransitiveconstructionashavingtwononsubjectNParguments,aseparatebutrelatedconstructionisrequiredtoaccountforpassivesofditransitive.SupportingtheideathatthereexistsaPassive-Ditransitiveconstructionisthefactthattheactualformofthepassive-ditransitiveisnotstrictlypredictable.AtonetimeinthehistoryofEnglish,nopassivewasallowedatall(Allen2001).Insomelanguages,boththerecipientandpatientargumentscanpassivize,whereasinEnglishonlytherecipientargumentcanpassivize(AlsinaandMchombo1990;Polinsky1998).

AdeleE.GoldbergAdeleE.GoldbergiscurrentlyaProfessorofLinguisticsatPrincetonUniversity.Herworkinvestigatesourknowledgeoflanguageandhowthatknowledgeislearnedandconstrainedbydomain-generalprocessessuchascategorization,rationalinferences,andsocialcognition.Sheisparticularlyinterestedinconstructions,learnedpairingsofformandfunctionatthewordandphrasallevel.SheistheauthorofConstructions:AConstructionGrammarApproachtoArgumentStructure(1995)andConstructionsatWork:TheNatureofGeneralizationinLanguage(2006),aswellasnumerousexperimentalandtheoreticalstudies.

Page 39: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 1 of 13

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0003

TheLimitsof(Construction)GrammarPaulKayTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapteranalyzesthelimitsofConstructionGrammar.Itadvocatestheconservativeviewwhichonlyconsidersthoselinguisticphenomenaasconstructionsthataspeakerneedstoknowto"produceandunderstandallpossibleutterancesofalanguageandnomore."Thechapterarguesthattherearemanypatternswhichappearinlanguagedatathatdonotqualifyaspartsofagrammar,andthatthesepatternsareneithernecessarynorsufficienttoproduceorinterpretanysetofexpressionsofthelanguage.Thechapterhighlightstheneedtodistinguishcoiningfromthetrueconstructionsbecausethefailuretoobservethedistinctionbetweengrammaticalconstructionsandpatternsofcoiningcanhaveundesirableconsequencesbeyondgrammaticaltheoryperse,forexampleincomparativelexicalsemantics.

Keywords:ConstructionGrammar,linguisticphenomena,languagedata,coining,trueconstructions,grammaticaltheory

Allofthemanycompetingaccountsoftheworkingoflanguagedrawadistinctioninonewayoranotherbetweenwhatitisthatspeakersknowoutrightabouttheirlanguageandwhatitisthattheyhavetobeabletofigureout.Forexample,speakersofEnglishhavetoknowwhatredmeansandthatitisanadjective,andtheyhavetoknowwhatballmeansandthatitisanoun.Theyhavetoknowthatadjectivescanco-occurwithnounsinamodificationstructure(asinaphraselikeredball),andtheyhavetoknowtheproperstrategiesforgivingasemanticinterpretationtosuchadjective-nouncombinations.Buttheydonothavetoknowseparately,ortobetold,whatthephraseredballmeans.Thatissomethingwhichwhattheyalreadyknowenablesthemtofigureout.(Fillmore,Kay,andO’Connor1988:502)

3.1.Introduction

Onthewidelyacceptedviewillustratedintheepigraph,agrammarshouldcontainthestrictlylinguisticinformationrequiredtoproduceandunderstandallpossibleutterancesofalanguageandnomore.Inthischapter,Iwillarguethattherearemanypatternsthatappearinlanguagedatathatdonotqualifyaspartsofagrammar(i.e.,asgrammaticalconstructions)because,unliketheconstructionthatlicensesredball, thesepatternsareneithernecessarynorsufficienttoproduceorinterpretanysetofexpressionsofthelanguage:eachexpressionthatexemplifiesoneofthesepatternshastobelearnedandrememberedonitsown.Withregardtosynchrony,suchpatternsarenonproductivegeneralizationsoveralexicallyfixedsetofphrases:diachronicallytheycanmotivatetheentryofnewexpressionsintothelanguagebutcannotbereliedontodosounderanycircumstancepredictableinadvance.Ontheviewofgrammarinwhichthegrammarcontainsallandonlythestuffaspeakerhastoknowinordertospeakandunderstandthelanguageofthemoment,thesepatternshavenoplace,eventhoughtheyareimplicitinthelanguagedataandsporadicallyproductivediachronically.

ThefirsttimeIheardthewordunderwhelm,IwasimpressedbywhatItooktobethespeaker'sclevernessin

1

Page 40: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 2 of 13

creatingthisinstantlyunderstandableneologismonthefly.AstimewentbyandIheardmoretokensofunderwhelm,itoccurredtomethatpossiblythepersonIhadfirstheardthisexpressionfromwasnotreallyitscreator.(Thelatterconjecturewasprobablycorrect.Merriam-WebsterOnlinedatesunderwhelmsixteenyearsbeforeIfirstnoticedit.)Nevertheless,someonehadtobethefirstpersontosayunderwhelm.SupposethispersonwasamannamedPercivalandconsiderPercival'sactinproducingthefirsttokenofunderwhelm.Twocompetinganalysesofthisactsuggestthemselves.OneisthatPercivalsimplyusedhisknowledgeofEnglishgrammartoproduceanovelutterance,justashewouldhavedoneif,say,hehadneverheardthenouninvidiousnessbutconstructeditontheflyfromhisknowledgeoftheadjectiveinvidiousandoftheness-suffixingderivationalconstruction.AsecondviewisthatPercivalwasnotusinghisgrammar,butwasaddingtoit.Onthisanalysis,PercivalcoinedanewlexemebyanalogywiththingsaboutEnglishhealreadyknew,involvingthewordsoverwhelm,overandunder,theanalogicalproportionbeing:over:overwhelm::under:underwhelm.

Accordingtothefirstview,Percivalwasjustputtingtousehisgrammaticalresourcestocreateanovelutterance.Accordingtothesecondview,Percivalhadtoaddtohisgrammaticalresourcesbeforeusingtheresultinggrammar,augmentedwithabrandnewlexicalitem,toconstructhisutterance.IhopeyouwillfindvalidthedistinctionIhavedrawnbetweenthefirstandsecondanalysesofPercival'shypotheticalactandagreewithmethatthesecondanalysisispreferable.

Fillmore(1997)introducedthedistinctionbetweenconstructionsproperandpatternsofcoining:“Wecandistinguishtwokindsof‘creativity’inlanguage.Inonecasethereistheabilityofspeakers,usingexistingresources[viz.,constructions],toproduceandunderstandnovelexpressions.Intheothercase,theoneforwhichweusethetermcoining,aspeakerusesexistingpatternsinthelanguageforcreatingnewresources.”Thischapterarguesthatpatternsofcoining,althoughfrequentlyandproperlystudiedbylinguists,needtobedistinguishedfromthetrueconstructions.Thesevereviewofgrammaradoptedhereexcludespartiallyproductiveprocesses andconsignsthemtothemeta-grammar:acompendiumofusefulstatementsaboutthegrammar.Amongthemanyvictimsofthispurgewillbealargenumberofimperfectlyproductivederivationalprocesses.

Failuretoobservethedistinctionbetweengrammaticalconstructionsandpatternsofcoiningcanhaveundesirableconsequencesbeyondgrammaticaltheoryperse.InanotherwiseexemplarystudyofthecolortermsofYélîDnye,anunaffiliatedlanguageofRosselIsland(PapuaNewGuinea),Levinson(1997)notesthatthethreemostprominentcolortermsareallrecognizableasreduplicationsofrootswhosedenotata,intwocasesatleast,salientlydisplaythecolorinquestion.kpêdekpêde‘black’isderivedfromkpêde,thenameofaspeciesoftreethatisperhapsnotsalientlyblack,buttheothertwomaincolortermskpaapîkpaapî‘white’andmtyemtye(dialectalternate:taataa)representreduplicationsofthenamesofasalientlywhitecockatooandasalientlyredparrot,respectively.Levinsonreportsthatthereisa‘regular’(thatis,widespread)derivationalpatterninYélîDnyeaccordingtowhichanadjectivedenotingapropertymaybeformedbyreduplicationofanominalrootthatdenotessomethingwhichsalientlydisplaysthatproperty.Forexample,theadjectivemty:aamty:aa‘sweet’isderivedfromthenominalrootmty:aa‘honey.’

LevinsonnotesfurtherthatBerlinandKay'sfirstcriterionforabasiccolortermwasthat“itsmeaningisnotpredictablefromthemeaningofitsparts”(BerlinandKay1969:6),andhepointsoutthatsomeonefamiliarwiththebirdsinquestionandtheirnamesmightwellbeabletoguessthemeaningsofwhiteandredforkpaapîkpaapîandmtyemtye.FromtheseobservationsLevinsonconcludesthatthewhiteandredtermsofYélîDnyearearguablynot‘basiccolorterms’asdefinedbyBerlinandKay.AndfromthefurtherobservationthatmanyAustralianandOceaniclanguagesdisplaysimilarpatternsofreduplication,heopinesthatperhapsseveraloftheselanguageshaveno‘basiccolorterms’inthedefinedsense.

ButitisclearfromthefactsthatLevinsonreportsthatthemeaningsof‘white’forkpaapîkpaapîand‘red’formtyemtyearenotpredictablefromthemeaningsoftheirparts,becausethepartiallyproductivereduplicationprocessofYélîDnyeisnotapredictiveconstructionbutamerepatternofcoining.Thewordsforwhiteandredmighthavebeenbasedonthewordsforsandandblood,respectively,ornotbasedonanynominalroot.NotallYélîDnyeadjectivesareformedbyreduplicationandfortheonesthatarethereisnosurewaytoknowwhichofthenounrootswhosedenotataprominentlydisplaythepropertyinquestionnamestheactualeponym.Forexample,supposewekneweverythingtherewastoknowaboutthegrammarofYélîDnyeexceptthatkpaapîkpaapîisthewordfor‘white,’includingthefactsthatkpaapîisthenameofthewhitecockatooandthataYélîDnyeadjectivemaybeformedbyreduplicatinganounrootwhosedenotatasalientlydisplaytheproperty

1

2 3

Page 41: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 3 of 13

expressedbytheadjective.WiththisknowledgewecouldnotdeducethattheYélîDnyewordfor‘white’iskpaapîkpaapîbecausethewordfor‘white’mightbeanopaqueroot,mightbederivedbysomeotherprocess,ormightbederivedbyreduplicationfromanothernounroot.Nor,ifonesimplyoverheardthewordkpaapîkpaapîcouldonededucethatitsmeaningis‘white’,since—asLevinsonpointsout—themeaningofkpaapîkpaapîmightbebasedon,say,thecockatoo'sdistinctivescreech.

3.2.Constructionsvs.PatternsofCoining

IwouldlikenowtoconsidertwopatternsofEnglish,oneofwhichIwillarguequalifiesasaconstruction,theotheronlyapatternofcoining.Theconstruction,whichIwillcalltheAll-cleftconstruction,isillustratedin(1).

(1)a.Allthatwehadtosaytothemwasthatweintendedtotaxthemmoreseverely.[BritishNationalCorpus(BNC )]b.Allthatonehastodoistostarttrainingearlier.[BNC]c.AllIwantistogetitoutoftheflat…[BNC]d.Allwecanreasonablyconcludeisthattheyhappenedatthesametime.[BNC]e.Allasmemother'sgottodothatdayisthedinners.[BNC]f.All'sIseeisacrazywomanthrowingawayoursupplies.[BNC]g.…soall'swereallyneediscigarettes…[BNC]

Withregardtosyntax,all-cleftscanbedescribedasidenticaltowh-cleftsexceptthattheextractedconstituentofthesubjectclausecontains,insteadofawh-word,eitherallthat(asin1a,b),all(asin1c,d),allas(asin1e),orall's(asin1f,g). Thesyntacticidentityofall-cleftsandwh-cleftscanbeappreciatedbycomparingthewh-cleftsentencesin(2)tothecorrespondingexamplesin(1).

(2)a.Whatwehadtosaytothemwasthatweintendedtotaxthemmoreseverely.b.Whatonehastodoistostarttrainingearlier.c.WhatIwantistogetitoutoftheflat…d.Whatwecanreasonablyconcludeisthattheyhappenedatthesametime.e.Whatmemother'sgottodothatdayisthedinners.f.WhatIseeisacrazywomanthrowingawayoursupplies.g.…sowhatwereallyneediscigarettes…

Thesyntacticidentityofall-cleftsandwh-cleftsincludesthepropertiesofconnectivityandreversibility.Theconnectivitypropertyisillustratedin(3)andthereversibilitypropertyin(4).

(3)a.Allthepresidentwantedwastosucceedhimself.b.Whatthepresidentwantedwastosucceedhimself.c.*Thatthepresidentwouldbere-electedpleasedhimself.d.*Thepresident'scertainre-electiondelightedhimself.

(4)a.Tosucceedhimselfwasallthepresidentwanted.b.Tosucceedhimselfwaswhatthepresidentwanted.

Sinceall-cleftsareidenticaltowh-cleftssyntactically,wepostulatethatall-cleftsrepresentaconstructionthatinheritsitssyntaxfromanabstractconstruction,whichisalsoinherited,bytheWh-cleftconstruction.WecandispensewithfurtherdiscussionofthesyntaxofAll-clefts.

Thesemanticallyinterestingpropertyofall-cleftsentencesisthattheydonotmeanwhattheymightbethoughttomeancontaining,astheydo,theuniversalquantifierall.Asentencelike(5a)doesnotmean(5b).Rather(5a)isglossedreasonablywellby(5c).

(5)

4

5

Page 42: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 4 of 13

a.AllIcaneatishalfapizza.b.EverythingIcaneatishalfapizza.c.ThemostIcaneatishalfapizza.

Moregenerally,utteranceofanall-cleftsentencemayexpressaproposition(e.g.,Icaneathalfapizza)thatistakentorepresentalowerpointinapresupposedscalarmodel(Fauconnier1975;Fillmore,KayandO’Connor1988;Kay1990)thansomecontextuallygivenalternative(e.g.,Icaneatawholepizza).Forexample,in(6),B1isanappropriateanswertoAandB2isnotanappropriateanswer,despitethefactthatB3is.

(6)a.A:Ijumpedsixfeet.b.B :That'sgood.Allyouneededtojumpwasfivefeet.c.B :*That'sbad.Allyouneededtojumpwassevenfeet.d.B :That'sbad.Whatyouneededtojumpwassevenfeet.

Letuscallthisparticularscalarreadingofall-cleftsentencesthe‘belowexpectation’reading.Notallsentenceswithall-cleftsyntaxhavethisreading.Somedoexpressuniversalquantificationofthesubjectclause,asexemplifiedin(7).Comparetheexamplesin(7)withtheparallelexamplesin(8)

(7)a.AllthatIcommandisyoursnow.[BNC]b.Allthatwecansee,feel,touch,tasteandhearisofone,all-pervadingforce—thegodforce.[BNC]c.Allthatweuseinourmodernworldisacommentuponthedelicatebalanceofhumanhand,eyeandbrain.[BNC]

(8)a.EverythingthatIcommandisyoursnow.b.Everythingthatwecansee,feel,touch,tasteandhearisofone,all-pervadingforce—thegodforce.c.Everythingthatweuseinourmodernworldisacommentuponthedelicatebalanceofhumanhand,eyeandbrain.

Althoughall-cleftsentenceswithuniversallyquantifiedreadings,suchasthosein(7),appearincorpora,theyarenotablylessfrequentthanall-cleftsentenceswithbelow-expectationreadings.Noneoftheexamplesofuniversallyquantifiedall-cleftsthatIhavefoundareoftheallasorall'svarieties.Moreover,althoughcorporacanneverpresentdirectevidenceofungrammaticality,Ibelieveallasandall'sversionsofall-cleftformforcethebelow-expectationreading.

(9)a.*All's/*AllasIcommandisyoursnow.b.*All's/*Allaswecansee,feel,touch,tasteandhearisofone,all-pervadingforce—thegodforce.c.*All's/*Allasweuseinourmodernworldisacommentuponthedelicatebalanceofhumanhand,eyeandbrain.

Itcouldconceivablybearguedthatinthecaseofbelow-expectationall-cleftsofbareallorallthatform,theliteralmeaningisoneofuniversalquantificationandthebelow-expectationreadingisderivedbyconversationalimplicature.Theanalystwhotakesthislinewould,however,berequiredtoexplainwhythistypeofconversationalimplicaturedoesnotapplytootherwiseidenticalsentenceswitheverything,everyoneorotherexpressionsofuniversalquantificationintheextractedposition.Wecannot,forexamplesay(10b)toexpresswhatisexpressedby(1b)(repeatedhereas(10a)).

(10)a.Allthatonehastodoistostarttrainingearlier.[BNC]b.#Everythingonehastodoistostarttrainingearlier.

Moreover,ifitiscorrectthatonlythebelow-expectationreadingisavailablefortheallasandall'sversions,thenaspecialconstructionstipulatingall-cleftsyntax(i.e.,wh-cleftsyntax)andbelow-expectationinterpretationwillberequiredanyway.WeconcludethatthegrammarofEnglishcontainsanall-cleftconstructionwithwh-cleftsyntaxandbelow-expectationinterpretationandthatthisconstructionisfullyproductive,beinglexicallyconstrainedonlywithrespecttothefillerconstituentofthesubjectphrase.

1

2

3

Page 43: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 5 of 13

Ourexampleofanonproductive,nonconstructionalpatternofcoiningappearsin(11).

(11)a.dumbasanoxb.greenasgrassc.deadasadoornaild.happyasalarke.strongasanoxf.flatasapancakeg.bigasahouseh.stubbornasamulei.darkasnightj.plainasthenoseonyourfacek.quickasawinkl.hardasarockm.freeasabirdn.dryasaboneo.lightasafeatherp.thinasarailq.hotasblazesr.clearasabells.blackascoalt.blackasnightu.blackaspitchv.coldashellw.hotashellx.easyasducksoupy.easyaspie

Thepatternexemplifiedin(11)ischaracterizedbytheformulain(12).

(12)AasNP[interpretation:‘veryA’]

DespitetheexistenceofmanymoreformulaicexpressionsfittingtheAasNPformulathanareshownin(11),thatformuladoesnotconstituteaconstructionbecauseitisnotproductive.First,knowledgeofformula(12)plusknowledgeoftheconstituentwordsisnotsufficienttolicenseanyoftheexpressionsin(11).Ifayoung,foreign,orshelteredspeakerofEnglishknewwhateasymeant,andknewwhatpie,duck,andsoupmeant,andknewalltheexpressionsin(12)plusmanymorebuiltonthesamepattern,theywouldstillnotknowthateasyaspieandeasyasducksouparewaysofsayingveryeasy.Secondly,onecannotfreelyusethepatterntocoinnewexpressions.Compare(13a)withafixedcollocationand(13b),whichalthoughunderstandableisnotsimple,prosaicEnglish.

(13)a.Whatahealthybaby,strongasahorse!b.Whatahealthybaby,heavyasatruck!

TherearemanymembersoftheAasNPpattern,anditislikelythatnewonescomeintoexistenceeverynowandthenasanalogicalcreations,likeunderwhelm,but,unlikeunderwhelm,dieaborning.AnunsystematicsearchoftheBNCturnsupseveralnovel,singletoncasesoftheAasNPpattern,whosecontentsandcontextssuggestself-conscious,literaryusage.

(14)a.Becauseyou’realwaysutterlyalonethen.Theremaybealover,afriend,asleepbesideyou,butwhoiswide-eyedasamarigoldinthetracklessdark?Justyouandyou.b.SheselectedGoyescas.Themusicwassureasaswinginhighsummer,toandfro,lightasracingoverasunnylawntotheblessedshadeunderthetrees.Upthegardenpathandafrissonofunease

6

Page 44: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 6 of 13

c.Theyflirtedwitheverybodyespeciallyeachother.Onlylastsummer,yearsafteritwasalloverandAstridhadfoundherstonecottagewithrosesroundthedoor,golden/greyinalatesummerorchard,shemetJay,andhereyesweregreenastheflamesonashwood.Firelight,eveninglight.d.Shewouldchangetheirlives.ThehospitalinLeninakanisbareasagarage.Theredrobetheygavehertherewasthefirstbrightcolourshehadworn.

ManyoftheexpressionsintheAasNPpatternaremotivatedbythemeaningoftheNP,butquiteafewarenot.Feathersarethoughtofaslightandleadasheavy,butthereisnothingparticularlyeasyaboutpieorducksoup.FurtherevidencethatindividualexpressionsoftheAasNPvarietymustbelearnedindividuallyisthatevenamongthosethatappeartobehighlymotivated,thereareseveralkindsofidiosyncrasytobeobserved.ThefirsthastodowiththeassociationofsomeAasNPcollocationswithliteralmeaningsoftheadjective,otherswithmetaphoricalmeaningsoftheadjectivewhileyetothersoccurwithbothliteralandfigurativemeanings.Considersomefixedexpressionswithhotandcold.Theseadjectivescanbeusedmetaphoricallyinsportscontextstoindicateskillfulandunskillfulplay,respectively.

(15)Ourbestshooterwashot/coldtonight.

Theexpressionshotasafirecrackerandhotasatwo-dollarpistolcanbeusedinthismetaphoricalsense,butnotintheliteralsenseofelevatedtemperature.Adifferentmetaphoropposeswarmandcoldpersonalities.Wecansaythatapersoniscoldasice,butnotthattheweatheriscoldasice.However,Ithinkwecansaythatourbestshooterwascoldasicetonight.AlthoughwecansayboththattheweatherishotasHellandcoldasHell,wecannotsaythatoutbestshooterishotasHell,althoughitsoundsalmostnaturaltometosayOurbestshooterwascoldasHelltonight.(Yourmileagemaydiffer.)

AsecondkindofidiosyncrasyofAasNPexpressionsisthatsomeofthemcanoccurwithathan-phrase,whileotherscannot.Thisistrueoftheexpressionsin(11),someofwhicharegivenbelowincomparativeform.Someexpressionsofthistypedonotoccurincomparativeform,asillustratedin(16e–h).

(16)a.deaderthanadoornailb.hotterthanHellc.biggerthanahoused.flatterthanapancakee.*happierthanalarkf.*quickerthanawinkg.*easierthanpieh.*drierthanabone

WeconcludethatalthoughthereisaverylargenumberofcollocationsbuildontheAasNPpattern,thispatternisnotproductive.ExpressionsofthisformcannotbefreelyproducedasnovelexpressionsusingtheexistingresourcesofEnglishgrammar.TheAasNPpatternwiththemeaning‘veryA’hasseeminglyprovidedafecundsourceofanalogyforcoiningnewEnglishcollocations,butitisnotaconstructionofEnglishgrammar.

Wehaveseenthatthereexistfamiliesoflexicallyrestrictedexpressions,originallyidentifiedbyFillmoreaspatternsofcoining,whichalthoughsporadicallyproductivediachronicallyarenotsystematicallyproductivesynchronically.Fillmore(1997)writes:

Thereisaviewofgrammaraccordingtowhichthegrammarproperwillidentifyonlytheproductiveprocesses.Sincetheabilitytocreatenewwords,usingnon-productiveprocesses,isclearlyalinguisticability,itismyopinionthatagrammarofalanguageneedstoidentifyconstructionsthatexistfor“coining”purposesaswell.Technically,thecoiningconstructionswillsimplybethoughtofasboundconstructions,constructionsthatare“bound”to—inheritedby—particularcomplexwords.Theywillservetomotivateandrepresentthesubstructureofmorphologicallycomplexwordsandsomeidiomaticphrases.Buttheyarealsoavailableforthecoiningofnewwords.

Thenarrowerview,expressedinthefirstsentenceofthequotedpassage,ismoreinkeepingwiththenotionofgrammarexpressedintheepigraphandadoptedhere:thatagrammarrepresentstheminimalamountofwhata

Page 45: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 7 of 13

speaker-interpreterneedstoknowaboutthelanguageinordertobeabletofigureouttherest.Patternsofcoiningarenotpartofwhataspeakermustknowbecause,aswehaveseeninexaminingtheAasNPpattern,thespeakerofEnglishhastoknoweachofthemembersofthesetofexpressionsexemplifyingthispatternindividually,anyway.SpeakersofEnglishwhoalreadyknowthewordsredandballandtheconstructionlicensingmodifiednominalsdonotalsohavetoknowtheformandmeaningoftheexpressionredball.Thatissomethingtheycanfigureoutfromwhattheyalreadyknow.ButspeakerswhoknowthewordslightandfeatherandtheAasNPpatternofcoining,cannotdeducefromthisknowledgethatlightasafeatherisawaytosayextremelylight.Thatissomethingtheymustpossessasaseparatepieceofknowledge,ornotatall.

HavingarguedthattheAll-cleftpatternandtheAasNPpatternprovideclearillustrationsofaproductiveconstructionandapatternofcoining,respectively,itmaybeofinteresttoconsideracasethatislikelytoelicitmoredisagreementandtherebyhighlightamethodologicalissue.Ihavearguedelsewhere(Kay2005)thattheCausedMotionpattern,illustratedin(17)isproperlyconsideredapatternofcoiningandnotaproductiveconstruction.

(17)a.Kimpushedtheshoeunderthesofa.b.TheylaughedhisHamletoffthestage.c.Tracysneezedthetissueoffthetable.

First,followingsomeideasofGawron(1986),whichwereconsideredbutnottomymindsuccessfullyrefutedbyGoldberg(1995),IreviewedsomereasonsfornotpositingaCausedMotionconstructionaspartofthegrammar.ThegrammarofEnglishwillneedatransitivizingconstruction,whichaddsanagentargumenttoanintransitiveverb,inordertoexplainexampleslike(18a,b),alongwithmanyothersinvolvingnonmotionverbssuchasboil,melt,grow,evaporate,freeze,andsoon.

(18)a.Thetopspun.b.Kimspunthetop.c.Thetopspunoffthetable.d.Kimspunthetopoffthetable.

ThegrammarofEnglishwillalsoneedaconstructionthataddsapathargumenttointransitiveverbs,asillustratedby(18a,c). IfwenowpositaCausedMotionconstruction,licensingtheverbalvalenceexemplifiedin(18d),thatsentencewillbeaccordedaspuriousambiguitybyourgrammar,itsvalencebeinglicensedbothbytheCausedMotionconstructionaloneandbythecombinationoftransitivizingandpath-addingconstructions.Weneedagent-addingandpath-addingconstructionswhetherornotwepositaCausedMotionconstruction,andoncewehaveagent-addingandpath-addingconstructions,thereisnothingleftforaCausedMotionconstructiontodo.Itdoesnotworkintheotherdirection.IfweposittheCausedMotionconstruction,withagent,theme,andpatharguments,wewillstillneedagent-addingfor(18b)andpath-addingfor(18c).

WenotedthattheAasNPpatternofcoiningisnotproductive.Onediagnosticofapatternthatisnotproductiveisthattreatingitasaproductiveconstructionleadstoovergeneration.OvergenerationwiththeCausedMotionpatternisillustratedin(19).

(19)a.*Hebraggedheroutoftheroom.b.*Shescreamedhimunderthebed.c.*Shesmiledhimtoher.d.*Hescreechedherhandsoverherears.

AlthoughtheseobservationsprovidereasonsnottopositaCausedMotionconstruction,thereremainthreetypesofexamplesofacceptablecaused-motionsentencesthatdonotyieldtoananalysisintermsofagent-addingandpath-addingconstructions.Inthefirst,thethemeargumentcannotoccurwithoutthepathargument,asin(17b,c),(20),and(21).

(20)

7

8

Page 46: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 8 of 13

a.Theylaughedhimoffthestage.b.*Theylaughedhim.

(21)a.Weletthewateroutofthepool.b.*Weletthewater.

Thesecond,andrelated,groupofexamplesinvolvesverbsthatdonothavemotionmeaningsunlessthepathisadded.

(22)a.Sheshowedhimtothedoor.b.Hesqueezedtheshimunderthepedestal.

Iwouldsuggestthattheseexamplesareindividuallyconventionalizedinthesensethat,forexample,thereisaconventionalizedusageorsenseoftheverbletthatissubcategorizedforboththemeandpatharguments.Thesamevalenceconfigurationisnotpossiblewiththesemanticallysimilarverbsallow,or(colloquial)leave,orwiththesomewhatlesscloselyrelatedverbs,permitandenable.IftherewereaCausedMotionconstruction,wewouldexpectalltheexamplesin(23)tobeacceptable.

(23)a.*Weallowedthewateroutofthepool.b.*Wepermittedthewateroutofthepool.c.*Leavethewateroutofthepool/thedogstayinthehoused.*Weenabledthewateroutofthepool.

Observing,forexample,thatpermitisinherentlyvolitionalwhileletisnot(Aleakyrooflet/allowedwater(get)intothebarn),onemightbeinclinedtoarguethatthereissomesemanticdistinctionwhichallowslet,butnosimilarverbs,tounifywiththeCausedMotionconstruction(althoughthisparticulardistinctionwillnotruleoutallowin(23a)).Takingthisline,however,willcommitthedefenderofCausedMotionasaconstructiontospecifyingjustwhatthatdistinctionis.TherearealargenumberofnearsynonymsonlyoneofwhichpermitstheCausedMotionpattern,asillustratedin(24).

(24)a.Sheshowedhimtothedoor.b.*Shedisplayed/demonstrated/illustrated/revealed/exhibited…himtothedoor.

WhatthedefenderofCausedMotionasaconstructionwillhavetoproduceisasemanticpropertywhichcharacterizesjusttheverbsthatwork,e.g.,let,laugh,andshow,andrulesoutthemanysemanticallysimilarverbsthatdonot.Otherwise,wemustconcludethattheacceptableexamplesin(17–24)areindividuallyconventionalizedaccordingtoaCausedMotionpatternofcoining.

Thethirdandfinaltypeofexampleconsistsofadmittedlyunconventionalizedexpressionslike(25).Possiblythistypeofexampleisrestrictedtotheverbsneeze,orafewothers.Goldberg(personalcommunication)pointsoutthat(25)occursinachildren'sbookbyRobertMunsch.

(25)Hesneezedhistoothrightacrosstown.

Exampleslike(25)arebestconsiderednonceformulations,formedonanalogywiththemanyconventionalizedexamplesoftheCausedMotionpatternofcoining,astheexamplesin(11)weredoubtlesscoinedonanalogywithotherconventionalizedexamplesoftheAasNPpattern.AcceptingthisanalysisleadsusconcludethattheCausedMotionpatternisnotaconstructionofEnglish.

3.3.Conclusion:PartialProductivityandDiachrony

Thischapterhasadvancedanuncompromisingviewofthelinebetweentheinformationtobeincludedinagrammarandtheinformationtobeexcluded.Itwasclaimedattheoutsetthatthisviewiswidespreadinthefield.Thatclaimistrue,butwidespreadneednotbeunanimous.Infactthereareleasttwofamiliesofreasonable

9

Page 47: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 9 of 13

objectionstothesevereviewadvancedhere.Bothinvolvethedistinctionspartialvs.fullproductivityandsynchronyvs.diachrony.

TheobjectionpresentedbyJackendoff(thisvolumeandtheliteraturecitedthere)hastwoforks.Thefirstinvolvespatternsformakingupnewwordsandmultiwordexpressions.Forexample,mostderivationalprocessesinEnglisharenotoriouslysemiproductive.Jackendoffcitestheexampleofdenominalverbs:Recitehasthetwonominalsrecitalandrecitation,withdifferentmeanings.Thephonologicallyparallelincitehasincitementratherthan*incitalor*incitation,andexcitehasexcitationandexcitementbutnot*excital.Onemayreasonablyaskwhethersuchpartiallyproductiveprocessesshouldnotbemadepartofthegrammar.Theansweristhatinaconstructionallybasedgrammarsuchpartialproductivityinthelexiconcanbeperspicuouslyrepresentedbyamultipleinheritancehierarchyoftypedfeaturestructures,coupledwithanarrayofderivationalconstructionsthatspecifythemother'sformasafunctionofthedaughter'sform.Imperfectlyproductiveprocesses,like-ate->-ation verb-to-nounderivation,canbeexpressedbymother-daughterconstructionswhichincorporatemorphologicalfunctionsrelatingthedaughter'sformtothemother'sformbutwhichalsolistintheirdomaineachstemnotsubjecttothegeneralrulewiththecorrespondingformintherangeandwhichprovidean‘elsewhere’provisionforthemostpopularpattern(seeSag2012:section2.6.2).Thus,inbroadstrokes,themorphologicalfunctionFspecifiedinthederivationalconstructionderivingnounsin-ationfromverbsin-atemightlooksomethinglikeTable3.1.Theconstructionitselfwouldspecifythatthemother's(morphologicalormorphophonological)FORMvalueistheimageofthatoftheuniquedaughterunderasimplefunctionoverthedefinedverbaldomainwiththenameF :

Inperfectlyproductivecases,thefunctionreducestothe“otherwise”specification.Incasesinwhichthereismorethanonepopularpattern,suchasthederivationofadjectivesfromverbsinEnglish,themorphologicalfunctionforeachpattern,say-alaffixation,willspecifyallexceptionalstemssuchasinciteasundefined.Completelyirregularcases,includingthoseinwhichthereisarbitraryassociationofformwithmeaning,arefullyspecifiedinthelexicalentries. Inshort,acarefullyconstructedconstruction-basedgrammarthattakesthesevereview,forexampleSign-BasedConstructionGrammar(e.g.,Sag,Wasow,andBender2003;KayandSag2009;Sag2012;Michaelis,thisvolume),candealperspicuouslywith

Table3.1.TheVerbin<…-ate>toNounin<…-ation>functionF

verbin<…-ate>:x noun:F(x)

Abate undefined

Berate undefined

… undefined

otherwise:

<x-ate> <x-ation>

partialproductivityinderivational

morphology,extractingthepatternsthatarethereandassociatingthelessthanfullyproductivepatternsindividuallywiththedaughterformsonwhichtheyoperate.

Jackendoffalsodiscussesexamplesofpartialproductivityatthesyntacticlevel.Henotes,forexample,thatwhileSluicing(Somebodylovesme,Iwonderwho?)isperfectlygeneral,that‘Sluicestranding’,asheexemplifiesin(26),ishighlyrestrictedinproductivity.

(26)…butIdon’tknowa.whowith/to/from/for/*nextto/*about/*besideb.whatwith/for/from/of/on/in/about/at/*before/*into/*near/*besidec.howmuchfor/*by/*with(notealso*howmanyfor)d.whereto/from/*near

10

verb->noun<-ation>

11

Page 48: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 10 of 13

e.*which(book)with/to/from/nextto/about/beside

Heobservesthat“theacceptablecasesofSluice-strandingmustbelearnedindividually,”andarguesthatsinceSluice-strandingpartakesoftheSluicingpattern,thattheSluicingpatternitselfmustbeseenasapartiallyproductiveconstruction.ThesevereviewwouldratherformulatetheSluicingconstructionsoastobarprepositionstranding,andwouldconsidertheSluice-strandingpatternasanextra-grammaticalpatternofcoinage,perhapswiththepotentialtogrowdiachronically(see,e.g.,Barðdal,thisvolume).

Jackendoff(thisvolume)alsoproposesexamplesinwhichthereareregular,andalsomerelypopular,patternsforcoiningnewpropernames.Forexample,ifwearenaminganewlakewehaveavailableboththepatterns[Lake__](LakeSuperior)and[__Lake](ConvictLake).However,ifwearetonameamountainwiththewordMount,wehaveonlythepattern[Mount__].Jackendoffcountsthelatterpatternasfullyproductive,whichitis,butonlydiachronically.Suchpatternstellushowtocreatenewlinguistictypes,whichaugmenttheexistinggrammar,ratherthanhowtocreate,inconformitywiththeexistinggrammar,newutterancetokens.Ifwerestrictournotionofthegrammartowhatissynchronicallyproductive,totheproductionofnewutterancetokens,wecanthinkofthiskindofpattern,whichaugmentstheexistinggrammar,asadiachronicallyproductivepatternofcoinage.Thiswouldcontrastwithmyparadeexampleofapatternofcoinage,[AasNP],whichisnotdiachronicallyproductive.Whatthe[Mount__]andthe[AasNP]patternshaveincommon,however,isthatneitherissynchronicallyproductive.Fullsynchronicproductivityoveradomaindefinedotherthanbyalistiswhatqualifiesapatternforconstructional,thatisgrammatical,status,ontheviewproposedhere.

Jackendoffrecognizestheimportanceofdistinguishingproductivityfromsemiproductivity,buthealsoimpliesthattheinterpenetrationofproductiveandnon-orsemiproductivepatternsrendersitunwisetorelegatetheformertothegrammarandthelattertosomethingelse.HisperhapsmostcarefullyworkedoutexamplehereistheNoun-Preposition-Noun(NPN)pattern(daybyday,dollarfordollar,bumpertobumper,etc.).Hepointsoutthat,ontheonehand,thispatterncoversacollectionofidioms(handoverfist,limbfromlimb,tongueincheek,etc.),while,ontheotherhand,therearefourcompletelyproductivesubpatterns,giventherelevantsemanticconstraintsineachcase,basedontheprepositionsby,for,after,and(up)on.Theproblem,asheseesit,turnsonthepartialproductivityofNPNexpressionsinto.IfoneinsistsoncallingtheNPNpatternaconstruction,thenthisconstructionhasidiosyncraticidioms,productivesubconstructions,andonesemiproductivesubconstructionasspecialcases.Butthisisnotaproblemsolongasonerealizesthataconstructioncanconsistexclusivelyofasyntacticpattern(asarguedbyFillmore(1999)fortheEnglishinvertedclausepattern,‘SAI’,paceGoldbergandDelGiudice(2005)).Underthesevereview,forexample,Sign-basedConstructionGrammar(SBCG),onepositsasingle,abstract,syntax-onlytype(NPN)mentionedintheindividualby-,for-,after-,and(up)on-NPNconstructions,eachwithitsownsemanticconstraints.Eachofthelistemes(lexicalentries)fortheidiomswouldindividuallyinheritNPN.Totheextentthatthepartialproductiveexpressionsbasedontoexhibitsemanticallycoherentsubpatterns,anNPNtypeandmatchingconstructionforeachsuchsubpattern(inheritingtheabstractconstruction)wouldbeposited,andtheremainingtocasestreatedasidioms. Inshort,Jackendoffraisesanumberofinterestingcases,eachrequiringcarefulstudy.Butnone,Ithink,presentsapersuasiveargumentagainstthesevereview.

Jackendoff'scritiqueofthesevereviewiscouchedwithinthebroadgenerativetradition,bywhichIintendtoincludeallformallyorientedapproachestoasynchronicgrammar,primarilyofsentences,basedonsomeversionofChomsky's(1965)idealizationoflinguisticcompetence(uniformdialect,singleflawlessspeaker,homogeneousspeechcommunity,etc.).Manylinguistsarecurrentlydoingimportantresearchpartiallyorentirelyoutsideofthisframework.Ofparticularrelevancetothischapteraretheexemplar-basedand/orusage-basedapproaches(Bybee,thisvolumeandreferencestherein).Thestudiesinthisburgeoningtraditiondonotemphasizethesynchrony/diachronydistinctionanddonotaccepttheChomskyan‘competence’idealization,butratherviewalanguageasastatisticalaggregateofforms-cum-meaningswhosefrequenciesmay,andusuallydo,changeovertime.Moreover,theseapproachesexplicitlyeschewthataspectofthesevereviewwhichholdsthatagrammarmustnotcontaineliminableredundancies—thatitcontainnomorethantheminimalinformationneededtoproduceandinterpretallutterancesofthelanguage.Ontheusage-basedview,alanguageis“amassive,highlyredundantinventory”comprisingpatternsthat“runthegamutfromfullgeneralitytocompleteidiosyncrasy”(Langacker1988b:131).Fromthisperspectiveadiscretesynchronicgrammarmaybelikenedtoavisuallydegradedversionofanotparticularlyinformativeframeofarichandcomplexmovie,whoseredundantpictorialsplendorandtemporalstructureareessentialtoitsnature.Theusage-basedapproachprovidesavaluableadditiontothearmamentariumoflinguisticresearch,butitproceedsfromavieworthogonaltothebroadgenerativeframeworkin

12

Page 49: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 11 of 13

whichthischapterissituated.

Therearenot,itwouldseem,manyempiricalpointsofcontactbetweentheusage-basedandgenerativeapproaches,butonethatdoescometomindinthepresentcontextpresentstheissueofwhether,andifsounderwhatconditions,patternsofcoiningmaygrowintoconstructions.Thatis,dosyntactico-semanticpatternsthataremanifestinseeminglyarbitrarycollectionsoflexicalmaterialeverbecome‘rules’or‘constructions’—patternsthatcanbegivenabstractdefinition,independentoftheidentitiesoftheparticularlexicalitemsthatparticipateinthem?AnexemplarystudythattakesaconstructionofmodernEnglishandseekstouncoveritshistoricalrootsisIsrael's(1996)studyofthehistoryoftheWayconstruction(s)(Shehackedherwayoutofthethicket/whistledherwaytotown.)Didthisconstructionstartoutasapatternofcoining,witharbitrarylexicalmembershiplikethecurrent[AasNP]pattern,andgraduallygetexpanded(andalsocontracted)intothegleamingandsmooth-edgedconstruction(s)oftoday?Theshortansweris:probably,butwecannotbeabsolutelysure.InthecaseoftheWayconstructionweknowthatinpresent-dayEnglishwehaveaconstruction,orperhapsafamilyofconstructions.Theempiricalquestioniswhetherthelexicalmaterialofattestedcasesofthispatternattheearlieststagesconstitutedsufficientlysemanticallyhomogenousclassestodefineoneormoreconstructionsorwhetherthepresent-dayconstruction(s)grewoutofasemanticallyheterogeneouslexicalhodgepodge.Israeldistinguishesthemannerthread(e.g.,ploddingone'sway)fromthemeansthread(e.g.,cuttingone'sway).Ofthemannerthread,hewrites:

asnewusagesmodestlybuildontherangeofestablishedofpredicates,theconstructiongraduallyincreasesinproductivity.Longstringsofanalogicalextensionsleadtodiscreteclustersofusage,whichthenlicensetheextractionofmoreabstractschemasfortheconstruction.Thesebasicobservationsturnouttoholdequallyforthemeansthread.(Israel1996:23)

InanswertothequestionwhetherwecansafelyconsidertheantecedentsofthecurrentWayconstruction(s)tohavebeenpattern(s)ofcoininginthesensethattermhasbeendefinedhere,Israelwrites:

Finally,sincethenubofyourquestionseemstobewhetherornottheverbsattestedintheconstruction(orwhateveritis)atanygivenpointconstituteasemanticallydefinedclass,itisworthnotingthatIhadtodefinemydataintoseveralsemanticallyverydistinct“threads”,soifanyofthesedoreflectatruesemanticclass,thatmaybe(inpartatleast)anartifactofmyanalysis,whichsoughttogroupverbsalongsemanticlines.

So,Idon’treallythinkIcouldproveit,butmygeneralsenseisthatthemodernconstructionheredidemergeastheresultofanalogicalextensionsfromasmallgroup[of](twoorthree)patternsofcoining.(personalcommunication)

Tosumup,withinthebroadgenerativetradition,itseemsthatthedistinctionbetweenapatternofcoining,asourceofpotentialbutnotguaranteeddiachronicanalogy,andatruegrammaticalconstructionisarelevantone.Withintheusage-basedapproach,whichseesgrammarasessentially,heterogeneous,redundant,statistical,andinastateofflux—perhapsbetteranalogizedtoagameofdicethanagameofchess—theutilityofthedistinctionislessclear.Bothapproachesmayfindinterest,however,intheempiricalquestionaccordingtowhichpatternsgainandlosefullproductivity,accordingtothegenerativeapproachenteringorexitingthesynchronicgrammar,accordingtotheusage-basedapproachsimplywaxingandwaninginrelativestrength.

Acknowledgment

IamgratefultoCharlesFillmore,ThomasHoffmann,GraemeTrousdale,RayJackendoff,JóhannaBarðdal,andMichaelIsraelforcommentsonpreviousdrafts.

See,forexample,KayandSag(2009).

Notes:

(1.)WhetherunderwhelmhasfullymadeitintoEnglishisapparentlystillunsettled.ThespellcheckerofWord2001votesnay.TheauthorsoftheMerriam-WebsterDictionaryofEnglishUsagearemoresanguine:

Page 50: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

Page 12 of 13

Underwhelmiscertainlyaninnocuousword.Itservesasamildlyhumorouswayofdescribingsomethingunimpressive,anditscommonusehassofarbeenlargelyuncontroversial.TheonlycriticismthatweknowofisbytheHarperusagepanelists,whofinditunacceptablebyalargemajority,essentiallybecausetheyseeitasajokethatisnolongerfunny.Severalofthepanelistsregarditspopularityasafad,butover40yearsofincreasingusestronglysuggestthatunderwhelmisheretostay.

MorerecentversionsofWordarealsomoreaccepting.

(2.)“Agrammaticalprocessorpatternorrule(or‘construction’)canbesaidtobeproductiveiftheconditionsofitsapplicabilitydonotrequirethelistingofexceptions.Actually,productivityisanotionofdegree.Allgrammaticalconstructionshavesomeconstraintsontheirapplicability,buttheextenttowhichthoseconstraintscanthemselvesbeformulatedingeneralwaysistheextenttowhichwecansaythattheconstructionisproductive.Someconstructionsonlyworkwithmonosyllabicwords;someonlywithcertaingrammaticalcategories.Buttheyaregeneraltotheextentthatsuchnon-lexicalconstraintsinvolvegeneral(Boolean)conditionsinvolvingpropertiessharedbyclassesoflexicalitems,ratherthanlistsofspecificwords”(Fillmore1997).

(3.)Ameta-grammarcontainsusefulinformationaboutalanguageandisthereforeofinteresttothelinguist.Thedegreetowhichspeakersofalanguageshareacommonmeta-grammarofthatlanguageis,however,particularlyhardtoevaluate,sincemeta-grammaticalstatementsdonotnormallyyieldconcretepredictions,aswillbedemonstratedbelow.

(4.)BritishNationalCorpus,http://www.natcorp.ox.ac.uk/.

(5.)Theall'sformseemstobeacontractionoftheallasform.Examples(2e),(2f),and(2g)weretheonlyexamplesoftheallasandall'svarieties,respectively,thatIcouldfindintheBNC.Isuspectthismaybedueinparttothecolloquialstatusoftheallasandall'sversions.

(6.)AneditorsuggeststhatthesemaynotberealinstancesoftheAasNPpatterninthattheymayjustbereadasunadornedsimiles,notindicating‘very’anything.InthatcasethescantevidenceforproductivityoftheAasNPpatternisfurtherweakened.

(7.)RayJackendoff(personalcommunication)challengestheexistenceofsuchafullyproductiveargumentstructureconstruction—thatis,avalencepatternthatcanbedescribedwithoutlistingtheindividuallexicalheadsitappliesto.Myargumentthusdependsonthepossibilityofdelimiting,presumablyonthebasisofsharedsemantics,asubsetofintransitiveverbstowhichthispatternapplies,withoutresortingtoalist.

(8.)MostoftheexamplesinthissectionareGoldberg'sexamplesorminorvariationsonGoldberg'sexamples.

(9.)Boas(2005)makesasimilarpointanddrawssimilarconclusionswithregardtotheresultativepattern,whichIsuspectisalsobestthoughtofasapatternofcoining:“Atissueisthequestionofwhattypesofrulesorschemasonehastoproposeinordertoaccuratelyaccountforthedistributionofresultatives…anadequateaccountofresultativesshouldinprincipleincludemechanismsthatcapturetheknowledgeneededtoUNDERSTANDandGENERATEsentencessuchasJerrydancedhimselftoexhaustionorNancytalkedherselfhoarsewhileatthesametimerulingoutsentencessuchas*Jerrydancedhimselfexhaustedor*Nancytalkedherselftohoarseness.”Ifyouhaveawaytorulein(withoutlistingthem)thegoodcasesandruleoutthebadcases,youhaveaproductivepattern,orconstruction.Otherwisenot.

(10.)Jackendoffaversthat“verbsendingin-ateinvariablyhaveanominalin-ation”andthough“invariably”appearstobetoostrong(TheOEDhasnoentriesfor*abation,*beration,*sation,*wation,or*hation),thatfactjustreinforceshislargerpointregardingtheirregularityofEnglishderivationalmorphology.

(11.)Formoredetaileddiscussion,seeSag(2012:section2.6.2).

(12.)SBCGmakesaformaldistinctionbetweenconstructionsandtypes.Sincemostotherconstructionalapproachesdonotmakethisdistinctionandmostlyterm‘construction’theconstraintpatternsthatSBCGdesignates‘type,’Ihavenotmaintainedthetype/constructiondistinctionhereinsketchingSBCGanalyses.

PaulKay

Page 51: The Oxford Handbook of Construction Grammar

The Limits of (Construction) Grammar

PaulKayisProfessorofLinguisticsEmeritusattheUniversityofCalifornia,Berkeley,ConsultingProfessorofLinguisticsatStanfordUniversity,andSeniorResearchScientistattheInternationalComputerScienceInstitute,Berkeley.Hehasworkedontherelationofcolornamingtothepsychologicalrepresentationofcolor,onconstructionalapproachestogrammar,onthecontextsensitivityofgrammar,andonlanguagevariationandothertopicsinlinguisticsandanthropologicallinguistics.HeiscurrentlyworkingwithIvanSag,CharlesFillmore,andLauraMichaelis-CummingsonthedevelopmentofSign-BasedConstructionGrammar,andwithTerryRegier,Li-HaiTan,MichaelA.Webster,andothersonseekingexplanationforuniversalsandcross-languagevariationincolornamingandontheeffectofcolornamesoncolorprocessing.

Page 52: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0004

Usage-basedTheoryandExemplarRepresentationsofConstructionsJoanL.BybeeTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapteroutlinesaviewofConstructionGrammarinwhichthementalgrammarofspeakersisshapedbytherepeatedexposuretospecificutterances,andinwhichdomain-generalcognitiveprocessessuchascategorizationandcross-modalassociationplayacrucialroleintheentrenchmentofconstructions.Underthisview,alllinguisticknowledgeisviewedasemergentandconstantlychanging.Thechapteremphasizesthattheprocessofchunkingalongwithcategorizationleadstothecreationofconstructions.Italsoprovidessemantic/pragmaticandphoneticargumentsforexemplarrepresentationandadiscussionoftheroleoftypeandtokenfrequencyindeterminingthestructureoftheschematicslotsinconstructions,aswellastheproductivityofconstructions.

Keywords:categorization,cross-modalassociation,typeandtokenfrequency,exemplars,chunking,usage-basedtheory,cognitiverepresentations,productivity

4.1.Introduction:Usage-basedTheory

ThebasicpremiseofUsage-basedTheoryisthatexperiencewithlanguagecreatesandimpactsthecognitiverepresentationsforlanguage(Langacker1987,2000b;KemmerandBarlow2000).Cognitiverepresentationsarebuiltupaslanguageusersencodeutterancesandcategorizethemonthebasisofphoneticform,meaning,andcontext.Asincomingutterancesaresortedandmatchedbysimilaritytoexistingrepresentations,unitssuchassyllable,word,andconstructionemerge.Thus,grammarcanbeviewedasthecognitiveorganizationofone'sexperiencewithlanguage(Bybee2006).

Itisacceptedwithinthisapproachthatitiswisertobeginthesearchforthecognitiveprocessesinvolvedinlanguagebyfirstconsideringdomain-generalcognitiveprocesses—thatis,thoseknowntofunctionindomainsotherthanlanguage,suchasvisionorneuromotorprocessing—ratherthantoassumeapriorithatlanguagerequiresspecialadaptationsofcognitivefunctions(ElmanandBates1997;Tomasello2003;BybeeandBeckner2009).Someofthesedomain-generalprocessesarecategorization,cross-modalassociation,andneuromotorautomation.Considerhowtheseprocessesapplytolanguage.Categorizationappliesateverylevelofformandmeaning:phonesandtheircombinationsarecategorizedbasedonexistingrepresentations,asarefeaturesofcontextandmeaning(Langacker2000b;Bybee2010).Cross-modalassociationallowshumanstomatchupthephonetic(ormanual)formexperiencedwithpropertiesofthecontextandmeaning.Automationallowstheunitsoflanguagetobecombinedinconventionalwaysthatleadtofluencyinbothproductionandperception(Bybee2002).

Animportantcharacteristicofhumanlanguageisthattheindividualunitsandsequencesofunitsaresubjecttohighlevelsofrepetition.Itisrepetitionthatleadstoconventionalizationofcategoriesandassociations,aswellastotheautomationofsequences.Becausesomeunitsandsequencesarerepeatedmorethanothers,ithasbeen

Page 53: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 2 of 14

possibletoidentifythepropertiesofcognitiverepresentationsthatdependupontheextenttowhichtheyhavebeenaccessedforproductionorperception.Thus,withinUsage-BasedTheorythestudyoffrequencyeffectsofvarioussortshascontributedtotheunderstandingofthenatureofgrammaticalorganization(Bybee2007).

Itisrecognizedthatlanguagesareconstantlychanging,andthischangeisgradualandtakesplaceaslanguageisused.Changeisattributedtothewayparticularcognitiveprocessesapplyinlanguageuse,thuschangeprovidesanimportantwindowintotheunderstandingofthecognitiveprocessesunderlyinglanguage.Aschangeisgradual,thecategoriesandunitsoflanguagearevariableandtheyformgradientratherthanstrictlyboundedcategories.Thus,linguisticstructureisviewedasemergent—governedbycertainregularprocesses,butalwayschangingasitisre-createdintheindividualandinspecificusagesituations(Hopper1987;KemmerandBarlow2000;EllisandLarsen-Freeman2006b).Thus,ratherthanafixed,staticsetofrepresentations,languageisviewedasbeingaffectedbyexperienceinanongoingwayeveninadults.Italsofollowsthatweshouldnotexpectlinguisticconstructssuchassegment,syllable,morpheme,word,orconstructiontohavestrictdefinitions,nordoweexpectallthemanifestationsoftheseconstructsinlanguagestoexhibitexactlythesamebehavior(Bybee2010:chapter1).

Asthetermimplies,theobjectofstudyinUsage-basedTheoryisnotonlythenativeuser'scompetence,whichresidesinthecognitiverepresentation,butalltheperceptionandproductionprocessesthatarebroughttothetaskofusinglanguage(KemmerandBarlow2000).Thus,thedataconsideredapplicabletoformulatingandtestinghypothesescanbeverybroadlydrawnfromexperiments,childlanguageacquisition,languagechange,andlargecorporarepresentingnaturalusage.Thelattersourceofdata(nowmuchmoreavailablethaninthepast)partiallysupplantstheuseofnativespeakerintuitions.Whilesuchintuitionsareinterestingandimportant,researchwithnaturalusageindicatesthatlanguageusersareoftenunawareofthenatureandfrequencyofcertainstructuresthattheyuse.Forthisreason,itisimportanttogroundlinguisticresearchinrecordedinstancesoflanguageuseinnaturalsituations.

Theterm‘Usage-based’wascoinedbyLangacker(1987),buttherootsofthisviewoflanguagecanbefoundstartinginthe1960swiththefunctional-typologicalapproachtolanguagetakenbyJosephGreenbergandcolleagues,andthenfloweringinthe1970sundertheleadershipofGivón,Li,Thompson,Hopper,andmanyothers(see,e.g.,Givón1979;Li1976;HopperandThompson1980,1984).Theselinguistsproposedthatgrammariscreatedbytheconventionalizationofcommonlyuseddiscoursepatterns.Thisproposal,whilenotexplicitlycognitive,encapsulatestheusage-basedpremisethatlinguisticstructureisformedbytherepetitionofcertainlinguisticpatternsinlanguageuse.Theproposalisenrichedbytherecognitionofthecognitiveprocessesthatfeedintolinguisticstructure,suchascategorization,asmentionedabove,butalsoentrenchmentandschematization(Langacker1987,2000b),metaphor(Lakoff1987),inference,andothertypesofconstrual(Traugott1989).

4.2.ConstructionsandUsage-basedTheory

Constructions,withtheirdirectpairingofformtomeaningwithoutintermediatestructures,areparticularlyappropriateforusage-basedmodels.AsNoonan(1998)observes,functionalistusagemodelsarematerialist,‘what-you-see-is-what-you-get’models,inwhichthelanguageuser'sexperiencewithlanguageisrepresentedratherdirectlyincognition.Itwouldappearnottobeentirelyaccidentalthatresearchonconstructionsandresearchonusageandcognitiveeffectshaveconvergedonthisviewinrecentyears.

Fromagrammarian'spointofview,constructionsareidentifiablebecausetheyaregroupingsofwordsthathaveidiosyncraticbehavioratsomelevel:theymightbeformallyspecial,butmoreoftentheytakeonanunpredictablemeaningorpragmaticeffect(Fillmore,Kay,andO’Connor1988;Goldberg1995).Fromthebroaderperspectiveofusage-basedtheory,however,constructionscanbeviewedasprocessingunitsorchunks—sequencesofwords(ormorphemes)thathavebeenusedoftenenoughtobeaccessedtogether.Thiswouldmeanthatwordsequencesthatareoftenusedareconstructionseveniftheydonothaveidiosyncrasiesofmeaningorform(Bybee2001:173,2006;Goldberg2006a).Itisinteresting,ofcourse,thatsuchchunksorconventionalizedsequenceshaveatendencyovertimetodevelopspecialpragmaticimplicationsthatcanleadtospecialmeaning.Theycanalsodevelopidiosyncrasiesofforminavarietyofways.

Page 54: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 3 of 14

4.3.ExemplarRepresentationforConstructions

Despitethemoresurface-orientedviewofConstructionGrammarincomparisontogenerativegrammarinitsvariousmanifestationsoverthedecades,abstractionawayfromparticulartokensdoestakeplace,asevidencedbyinnovativeusesofconstructions.Oneimportantquestionaddressedinrecentliteratureisthenatureofthisabstractionandhowtheexperiencedtokensofaconstructioncontributetotheformationofacognitiverepresentationofthatconstruction(Goldberg2006a;Bybee2010).Thischapterdiscussesthesequestionstakinganexemplarviewofconstructions.

Thenatureofexemplarrepresentationswillbeexplainedinmoredetailinthenextsection,butforpresentpurposeswecansaythatexemplarmodelsproposethatmemoryforlinguisticexperienceislikememoryforothertypesofexperience:eachtokenofexperiencedlinguisticbehaviorhasanimpactoncognitiverepresentation;whenstoredrepresentationsareaccessedineitherencodingordecoding,therepresentationsthemselveschange.Inaddition,memorystorageforlinguisticexperienceincludesdetailedinformationaboutthetokensthathavebeenprocessed,includingtheirformandthecontextsinwhichtheywereused.Exemplarrepresentationscontrastwiththemoreabstractrepresentationsofstructuralorgenerativetheories(atalllevels—phonetic,morphosyntactic,andsemantic/pragmatic),inthatvariationandfeaturespredictablefromgeneralprincipleshavenotbeenremoved.Insuchamodel,thegeneralcategoriesandunitsofgrammarcanemergefromtheexperiencethatisrecordedinmemorybecauseexemplarsarecategorizedbysimilaritytooneanotherandbecausecontiguousexperiences—suchasmeaningandacousticshape—arerecordedaslinkedtooneanother.

Thefollowingsectionswilldescribethegeneralpropertiesofexemplarmodelsanddiscusshowtheycanbeappliedtoconstructions.Thefollowingargumentsforchoosingexemplarrepresentationsforconstructionswillbepresentedinthischapter:

1.AsFillmoreetal.(1988)argue,manyconstructionshaveidiosyncraticfeaturesofmorphosyntax,semantics,pragmatics,andphonology,andmuchofwhataspeaker/hearerknowsabouthis/herlanguageisnotpredictablefromtheverygeneralrulesthathaveoccupiedtheattentionofmostsyntacticiansinthepast,butratherconsistsofspecificinformationthatmustbeassociatedwithspecificconstructions.Specificinformationfindsanaturalexpressioninanexemplarmodel,wherethestorageandcategorizationofalldetailbothpredictableandidiosyncraticisconsideredtobeabasicresponsetolinguisticinputandappliestoallconstructions,specificorgeneral(seesections4.4and4.5).2.Thestoredrepresentationsintheformofexemplarsrespondtousagebyallowingtherepresentationofbothtokenandtypefrequency;thesefrequencypatternsareimportantforunderstandingthecategoriesthatareformedfortheschematicslotsinconstructions(seesection4.6).3.Specificinstancesofconstructionsdevelopintonewconstructions,thusspecificexemplarsofconstructionsneedtohavecognitiverepresentation(seesection4.7).4.Exemplarmodelsallowforspecificmeaningfromthecontextofusetoimpactcognitiverepresentation,whichthenaccountsforthewaythatwordsandconstructionsareaffectedbythemeaningthatoccursinthecontext.Theimpactofcontextisseenbothinthedevelopmentofspecialmeaningsandimplicationsofconstructionsandalsoindevelopmentssuchasthenegativeprosodyidentifiedincorporaforconstructionswithcause(seesection4.8).5.Exemplarmodelswerenotdevelopedespeciallyforlanguage;rathertheyapplyequallytolinguisticandnonlinguisticcategories.Suchmodelsthentaketheusage-basedviewthatlanguageisapartofgeneralcognitionandallowustoaccessexplanationsforlinguisticphenomenaoutsideoflanguage.

4.4.ExemplarsofConstruction:ConstructionsasChunks

Exemplarsarecategoriesformedfromtokensofexperiencethatarejudgedtobethesame(Pierrehumbert2001).Linguisticexemplarscomeinavarietyofsizes,rangingfromasinglesegment,suchasavowel,towholeparagraphs,suchasthePledgeofAllegiance.Theexemplarsthemselvesaregroupedtogetherbysimilarity.Thus,thevowelsofhit,swim,andsipmaybegroupedtogether,thedifferentphoneticrealizationsofaword,suchasprettywillbegroupedtogether,aswellasexemplarsforlongersequences,suchasallofasudden.Theseexemplarclouds,astheyarecalled,constitutecategories.Exemplarcategoriesarestructuredbysimilarityandfrequency(Nosofsky1988)andoftenexhibitprototypeeffects.Becausetheyaregroupedtogetherbasedonsimilarity,intheoryanytypeofcategorycouldberepresentedinexemplars.However,itiscommontofind

Page 55: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 4 of 14

prototypeeffectsemergingfromcategoriesbecauseofdifferencesindegreesofsimilarity.

Inaddition,exemplarsmaydifferinstrengthdependinguponthenumberoftokensthatcomprisethem.Thatis,exemplarsbuiltupfromalargenumberoftokenswillberepresentedmorestronglythanthosebuiltupfromasmallernumberoftokens.Thestrongerexemplarorsetofexemplarsoftenformsthecenterofacategoryandotherexemplarsaremoreorlesssimilartothestrongerexemplarorsetofexemplars(Pierrehumbert2001).

Asexemplarsarebasedonperceptualstimuli,andexemplarsaregroupedtogetherbasedonsimilarity,wedistinguishexemplarcategoriesformedbydifferentcriteria:thusexemplarcloudscanbeformedonphonetic,semantic,pragmatic,orcontextualcriteria.Foranyword,phrase,orconstruction,exemplarcategoriesfromthesedifferentdomainsarelinked.Thus,anexemplarmodeloflanguagemustallowforlinksacrossdomains,basedonco-occurrenceinexperience.Suchlinkscreatetheform-meaningcorrespondencesthatconstituteconstructions.Thus,exemplars,likeconstructions,providefordirectpairingsofformwithmeaningwithoutintermediaterepresentations(suchasphrasestructureorphonemicrepresentations).

Inadditiontocross-domainlinking,mapping,orlinkingalsooccursamongunitsindifferentcontexts.Justasacategorymaybeformedoverthevowelsinhit,swim,anddip,acategorymaybeformedatwordleveldespitethefactthatawordmayoccurinmanydifferentcontexts.Infact,theexemplarcloudofawordwouldincludeallthemeaningsandcontextsinwhichthewordhasbeenexperienced.

Alinguist'sfirstreactiontosuchamodelisthatitincludeswaytoomuchinformationandwouldrequirewaytoomuchstoragecapacity.Therearetworesponsestothisobjection:first,neuralcapacityismuchgreaterthanpreviouslybelieved,andsecond,informationisstoredinahighlystructuredandefficientway.Highlystructuredstorageofinformationresultswhencategoriesareformedandsimilaritemsarestoredinproximitytooneanother.Also,asweknowfromexperience,memoriesforalltypesofsensoryinputcanbelost(throughforgetting)whenthatinformationisnotreinforcedbyrepetitionorrecency.Ofgreatestinteresttolinguists,however,isthequestionofhowmuchgeneralizationandabstractionoccursintheorganizationoflinguisticexperience.Thisisanempiricalquestionthatthemodelallowsustoposeandinvestigateinarealisticway,asweseeinthefollowingsections.

Memorystorageofcomplexunitssuchasidiomaticphrasesorconstructionsrequireslinksofvarioussorts.First,thereisthesequentiallinkingthatcomesaboutthroughrepetitionofsequencesofunits.Inthedomain-generalprocessofchunking,repeatedsequencesofelements,betheylinguisticornot,arerepresentedtogetherasunitsthatcanbeaccesseddirectlyratherthanformedcompositionally(Newell1990;Ellis1996;Bybee2002).Bythesemeansrepeatedsequencesbecomemorefluent.Withinachunk,sequentiallinksaregradedinstrengthbasedonthefrequencyofthechunkorperhapsthetransitionsbetweentheelementsofachunk.Aconstructionisachunkeventhoughitmaycontainschematicslots,thatis,theelementsofachunkcanbeinterrupted.

Second,therearelinksfromtheindividualelementsofachunktothoseelementsinothercontexts.Eventhoughaphrasesuchasallofasudden,oraconstructionsuchasdrivesomeonecrazy,mad,upthewall…formsachunk,thatdoesnotmeanthattheitemsthatcomposeitarenotstillanalyzableaswordsthatoccurelsewhereincognitiverepresentation.Analyzability(Langacker1987)canberepresentedaslinksfromtheexemplarclusterofawordwithinaconstructiontothegeneralexemplarclusteroftheword.Aninterestingphenomenonisthelossofanalyzabilityofchunks,aprocessthatoccursgradually.Aschunksaremorefrequentlyused,andaccesseddirectly,theybecomemoreautonomousandtheircomponentwordscanlosetheirassociationwithexemplarsoftheetymologicallysameword(Bybee2003;BecknerandBybee2009;Bybee2010).Ingrammaticalizationthisprocessisreferredtoasdecategorialization(Hopper1991).Forinstance,inthephraseinspiteof,theerstwhilenoun,spite,haslostitsabilitytofunctionasanoun,asinthisphraseitisnotmodifiable.Ithasalsolostitsearliermeaningof‘indefiance’andaddsnothingtothegeneralmeaningofconcessivethatthephrasenowhas.Thislosshastakenplacegradually,whichmeansthattherearedegreesoflinkingbetweenwordsindifferentcontexts(BecknerandBybee2009).Notethatcomplexprepositionssuchasinspiteofarenotextremelyfrequent.AsHoffmann(2004,2005)haspointedout,lossofanalyzabilitycancomeaboutwithouthighlevelsoffrequency,encouragedbychangesindistribution,meaning,andfunction.

Anotheraspectofchunkingthatisimportantfortheunderstandingofhowconstructionsemergeisthatbothnonlinguisticandlinguisticchunkstendtohavemeaningassignedtothematthehighestlevelpossible(Ellis1996;Bybee2010).Forinstance,onemightwitnesstherepetitionofasequenceofeventssuchasahumanthrowingaball,adogrunningafterit,catchingit,andreturningtothehuman,settingtheballatthehuman'sfeet.Thehuman

Page 56: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 5 of 14

picksuptheballandthrowsitagain.Itisahumanpropensitytotrytounderstandsuchsequencesaschunksofbehaviorandtoassignlabels,suchas‘playingfetch’tosuchchunks.Similarly,sequencesoflinguisticunitsthatoccurtogetherrepeatedlytendtobeassignedmeaningsasawholeratherthansimplyasasumoftheparts,ascanbeseeninmanyexpressionssuchasinspiteof,hereandthere,orallofasudden,whichnolongerarefullycompositional.Itisthischunkingandlabelingphenomenonthatprovideslanguagewithconstructions.

4.5.DetailsinExemplarRepresentation

4.5.1DetailsofSemanticsandPragmatics

Theoriginalmotivationforproposingconstructionsaspartofgrammar,asexpressedinFillmoreetal.(1988),wasthatconstructionshaveidiosyncraticstructureandmeaning.Asmentionedabove,thisattributeistakenbysometobecriterialindefiningconstructions,thoughintheusage-basedviewconstructioncanbedefinedsimplyasfrequentlyusedandthusconventionalwordsequences.Asmentionedabove,detailsofformandusageareautomaticallyregisteredinexemplarrepresentationandthroughentrenchmentcanbecomeaninherentpartoftheconstruction.Inthissection,basedonBybee(2006),Idemonstratethatanexemplarmodelisessentialtoexplaininghowconstructionsacquireidiosyncraticsemantic/pragmaticandphoneticproperties.Ialsoarguethatredundantdetailmayturnouttobeimportantinformingthebasesofnewusesofconstructions.

IndiscussingtheWhat'sXdoingY?(WXDY)construction,KayandFillmore(1999)notethatonereasonforregardingthisasaconstructionisitsspecialpragmaticimplications,whichinsomecasesoccurwithoutanyaccompanyingliteralmeaning,asin(1).

(1)What'sthatboxdoingupthere?

Asfortheoriginsofthissensefortheconstruction,theysay:

WhiletheWXDYconstructionmayhavehaditsorigininconversationalimplicatures—throughsituationsinwhichanindividualAisclearlyuptonogoodandBaskswhatAisdoing—thesemanticsofincongruityisnowconventionallyassociatedwiththespecialmorphosyntaxofWXDYconstructs.(KayandFillmore1995:5;emphasisoriginal)

Theconventionalizationofimplicature(orfromthehearer'spointofview,inference)isalsowell-knownfromgrammaticalizationresearch(Bybee1988b;Traugott1989;Bybeeetal.1994;TraugottandDasher2002).Itisthoughtthatthefrequentco-occurrenceofaninferencewithaparticularconstructioncanleadtothatinferencebeingtakenaspartofthemeaningoftheconstruction.Theoriginallyinferentialmeaningcanevenreplacetheearliermeaning.

Ifweconsiderhowthisconventionalizationoccurs,weseethatweneedanexemplarmodeltoaccountforit.Inamodelinwhichsemanticrepresentationscontainonlyacoreorabstractmeaningandinferencesarecalculatedontheflyineachcontext,thereisnowayforanimplication/inferencetosticktoaconstruction.However,anexemplarmodelwouldrecordtheinferencesmadeineachinstanceofuse,andifthesameinferenceismadeonmultipleoccasions,thestrengthofthatinferenceincreases.Withsufficientstrengthamongthesemanticexemplarsforaconstruction,theinferencecanbecomeconventionalizedaspartofthemeaningoftheconstruction.Thiscanhappenoverdiachronictimeanditcanalsohappenoverthecourseoflanguageacquisition,asthelearnerestablisheswhattheimplicationsofacquiredconstructionsare.Notethatthespeaker/hearercannotwaituntils/hehasexperiencedtheinferencemorethanoncetostoreitinmemory,becausethereisnowaytoknowiftheinferencehasbeenexperiencedbeforeifitisnotregisteredinmemoryfromtheveryfirstexperience(Bybee2006).

4.5.2.PhoneticDetail

Similarly,phoneticchangecanaccruetothewordsofaconstructionthroughinstancesofuseinreducingcontexts.Grammaticalizingconstructionsundergoreductionasfrequencyofuseincreases.Wordsingrammaticalizingconstructionsoftenhaveawiderangeofvariation,aswiththephoneticvariantsofEnglishgoingto,whichrangefromthefullformtovariantsspelledgonnatoextremereductionasinI’mgonna[рimə̃nə]The

Page 57: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 6 of 14

entrenchmentofthereductionintheconstructionhelpstodistinguishoneconstructionfromanother,asinthecaseofusedtoasfoundinthefollowingtwoexamples,wherethefirst,withpasthabitualmeaning,hasthephoneticshape[juztə]whilethesecondhastheform[juzdtʊ].(ExamplesfromtheTimeMagazineCorpus,2000. )

(2)itisbecomingdifficulttodistinguishbetweenwhatusedtobeconsideredelitecultureandmassculture(3)Allthatmoneywouldbeusedtobattlenotonlythedrugtraffickersbutalsotheguerrillaswhoarealignedwiththem

BybeeandScheibman(1999)arguethatphoneticreductionandassimilationtakeplaceinproductionandhigherfrequencywordsandphraseshavemoreopportunitiestoundergotheseprocesses.Thereductionorassimilationthatoccursinproductionisminimal,butinanexemplarmodelitseffectsareregisteredinmemorythroughtheestablishmentofanewexemplar.Ifthatexemplarischosenforproductionlateron,andalteredmoreintheprocess,yetanotherexemplarwillbeestablished(Bybee2000;Pierrehumbert2001).Forveryhighfrequencywordsthatoccurinlowstresspositions,thisgradualprocesscanleadtoextremedegreesofreduction,suchasthosewhichoccurwithI’mgoingtoandIdon’tknow.Asinthecaseofpragmatic/semanticchange,anexemplarmodelisessentialtoexplaininghowthesechangescomeabout.

Somesuchchangesarequitesubtleandcanonlybedetectedbyinstrumentalstudyovercorpustokens.Forexample,inalargecorpusstudy,Berkenfield(2001)foundphoneticdifferencesinthevowelofthatwhichincludedbothdurationanddegreeofcentralization,accordingtowhetheritwasusedasademonstrativepronoun,ademonstrativeadjective,acomplementizer,orarelativeclausemarker.Inotherwords,thatdifferedaccordingtotheconstructionitwasusedin.Similarly,HayandBresnan(2006)findthattheraisingof/æ/inNewZealandEnglishismoreadvancedinthewordhandwhenitreferstoalimbthanwhenitisusedinphrasessuchasgiveahand.Theyalsofindthatthecentralizationof/ι/ingiveismoreadvancedinphraseswithanabstracttheme,suchasgiveachancethaninitsmoreconcreteuseoftransferringanobject,suchasgiveapen.

4.5.3.SchematicSlotsasExemplarCategories

Anotherroleforexemplarsinaconstruction'srepresentationisseeninthewaytheschematicslotsinconstructionsexpand.Aschematicslotinaconstructionmightconsistofalistofalltheitemsthathaveoccurredinthatslot(aspredictedbyanexemplarmodel),oritmightbeconsideredasetofabstractsemanticfeaturesthatconstrainstheslot,asusuallyproposed.Itcould,ofcourse,beboth.However,theimportanceofthespecificexemplarsthathaveoccurredintheconstructioncanbeseenwhenfeaturesthatareredundantforparticularitemsarereferredtoinanextensionoftheconstruction,aswewillseebeloworincaseswhereasingleabstractfeaturedoesnotcharacterizeaclassorexplainitsextension.

Forinstance,itisoftenfoundthatthesetoflexicalitemsthatcanoccupyaslotinaconstructionmaybeconstitutedoftwoormoreclustersofcloselyrelateditems(Goldberg1995;Israel1996).Anexampleisthewell-knownwayconstructionwhichisusedwithverbsthatindicateeithermannerofmotion(4)ormeans(5)(seeJackendoff1990).

(4)RomanaKryzanowskaweavesherwaythroughDrago'sGyminmidtownManhattanlikeamotherhen…(TimeMagazineCorpus,2004)(5)Hisweakness,whichHolmesunflinchinglydescribes,wasaninabilitytoresistthefinancialandsexualrewardsthatcamealongasheclawedhiswaytothetop.(TimeMagazineCorpus,2000)

However,onealsofindswithinthesegroupsmorespecificclustersofverbssuchasthoseindicatingwindingmotion(pick,thread,wind,wend,worm,weave,etc.)orlaboriousmotion(plod,crawl,grind,slog,stumble,etc.)andsoon(Goldberg1995;Israel1996).Suchclustersofhighlyrelateditemsaretheresultsofitem-basedanalogy(Skousen1989;Eddington2000;Bybee2010:chapter4).Inotherwords,ratherthanmakingreferencetoageneralsemanticfeaturewhenusingaconstruction,thespeakermayverywellreferenceaparticularlexicalitemthathasalreadybeenusedintheconstructionandstoredinmemory.Exemplar-basedrepresentationsofconstructionswillincludealistofallthewordsexperiencedinacertainslotinaconstruction.Thislistofwords(organizedinclustersbysimilarityofmeaning)isthebasisfornewextensionsoftheconstruction.Barðdal(2011a)showsthatthemajorityofverbsborrowedintoIcelandictakethecaseconstructionofsynonymousverbs.Boas(2003)hasarguedforresultativeconstructionsinEnglishthatmanymicro-constructionsareavailablefor

1

Page 58: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 7 of 14

referencewhenextendingexistingconstructionstonewlexicalitems.

Item-basedextensionleadstoafamilyresemblancestructureamongthelexicalitemsthatconstituteaschematiccategoryinaconstruction.Forinstance,thesetofadjectivals(includingprepositionalphrases)thatoccurintheSpanish‘become’constructionusingtheverbquedarsecanbeanalyzedintoseveralclustersbasedonsemanticsimilarity(BybeeandEddington2006).Someoftheseclustersarelargeenoughthattheyincludeadjectivalsthathavenosemanticrelationtooneanother,butrathersharepropertieswithmorecentralmembersinfamilyresemblancefashion.Thus,ourcorpusstudyturnedupquedarsewiththeadjectivequieto,whichmeans‘still.’Itis,ofcourse,relatedtoinmóvil‘immobile’butalsotoanadjectivewithverydifferentimplications,tranquilo‘tranquil,peaceful.’Thiscouldinturnberelatedcloselytoadjectivesexpressingthemorepositivesideof‘tranquility,satisfaction’withconforme‘satisfied’andthisinturntoagusto‘pleased’asshownin(6).

(6)inmóvilquietotranquiloconformeagusto

‘motionless’‘still’‘tranquil’‘satisfied’‘pleased’

Aswithotherfamilyresemblancechains,theendsofthechainmayhaveverylittletodowithoneanother;inmóvil‘motionless’andagusto‘pleased’sharefewfeatures.Buttheybothsharesomefeatureswithtranquilo‘tranquil.’Whiletranquilomighthaveoriginallybeenusedinthisconstructionbecauseofitssimilaritytoquieto,itbroughtalongwithitthe‘peacefulness’sense,whichthencouldbereferencedforfurtherextensions.Thistypeofstructure,then,pointstolocalextensionsbyitem-basedanalogy,ratherthanglobalsummationsofabstractfeatures.Inorderforitem-basedanalogyorlocalsimilaritytoapply,thecognitiverepresentationofthisconstructionmustcontainthespecificadjectivalsthathavebeenexperiencedinit(seealsoBybee2010:chapter5).

Localextensionsmayalsobebasedonform,showingthatlexicalslotsinconstructionsarenotpurelysemantic,butcontaininformationregisteredinmemoryfortheformofitems.ThehistoryoftheSpanish‘become’constructionwithquedarse,asreportedinWilson(2009),showsthatitfirstappearedwiththeadjectivesolo‘alone’inthethirteenthcentury,aswellaswithafewotheradjectivals.The‘alone’sensegaveriseinlatercenturiestouseswiththeadjectiveshuérfano‘orphaned’andviudo‘widowed,’aswellassomephrasessuchassinheredero‘withoutanheir’andsinpadre‘withoutafather.’Bythefifteenthcentury,therearemanyexampleswithsin‘without’buttheyextendwellbeyondtheoriginalsemanticsrelatedtosolo‘alone.’Theyincludephrasessuchassinarmas‘withoutweapons,’sinpluma‘withoutapen,’sindeuda‘withoutdebt,’sinpena‘withoutgrief,’andsinquexa‘withoutcomplaint.’Thesemanticsofthesephrasesdonotfitwiththe‘alone’semantics,nordothesephrasesshareabstractsemanticfeatures.Alltheyhaveincommonistheuseoftheprepositionsin‘without.’Thispatternsuggeststhatfeaturesofformmayalsoinfluencethechoiceofitemsoccurringinalexicalslotinaconstruction.Anexemplarrepresentationwouldincludethesepropertiesoftheform,asalltokensandalltheirpropertiespotentiallyhaveanimpactontherepresentation.Incontrast,ifonlysemanticfeaturesarerecordedintheconstruction'srepresentation,therewouldbenowaytoaccountforthisextensionpattern.

4.6.TokenandTypeFrequency

4.6.1.TokenFrequency

Anotherimportantargumentforanexemplarmodeloflinguisticrepresentationsisthatexemplars,bytheirverynature,providearecordofthefrequencyofoccurrenceoftokensinlinguisticexperience.Asmentionedabove,eachtokenofexperiencehasanimpactonthememoryforlinguisticitems(Naderetal.2000).InBybee(1985),Iproposedthatrepresentationsarestrengthenedbyrepetition,makingthemeasiertoaccess.Thisproposalappliesprimarilytotokenfrequency—thenumberoftimesaparticularstringoccursinatextorcorpus.Thus,everytimeaconstructionisused,theconstantpartsofitarestrengthened.Butconstructionsalsohaveschematicslotswhichwillberepresentedbydifferentexemplarsandwhichformcategoriesasdescribedintheprecedingsection.Thesecategoriescanvaryintheirtypefrequency,thatis,inthenumberofitemsthatoccurintheslot.Exemplarmodelsalsoallowustokeeptrackoftypefrequency.Thus,thefrequencyprofileofaconstructioncanbequitecomplex.Inthissectionweexaminewhatisknownaboutthetokenandtypefrequenciesofconstructions.

LetustakeasourexampletheResultativeconstructionwithdriveconsistingofananimateobjectwithanadjective

Page 59: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 8 of 14

orprepositionalphrasesynonymouswithcrazy.Thisconstructionhasrecentlybeguntoappearinhyperbolicusesinwhichtheobjectisnotliterallyinsane,butonlyextremelyirritatedordistraught(seeBybee2010:chapter5).Herearesomeexamplesofdifferenttokensoftheconstruction:

(7)Forthesingersitwas“upanddown,upanddown,fromhighCtolowF,”saidTenorLawrenceWhite.“It'senoughtodriveyoucrazy.”(TimeMagazineCorpus,1965)(8)“Theywon’tletmeoutoftheirsight.It'sdrivingmecrazy.”(TimeMagazineCorpus,1982)(9)ThoughLindsayfinallylosthimhisgoodjobandnearlydrovehimwildwithanxietyabouthermorals,hisfondnessforhergrew.(TimeMagazineCorpus,1933)

Aconstructionisbuiltupfromsuchtokenswiththesetsofitemsthatoccurineachpositionformingcategories.Aswithmanyconstructionsinvolvingverbsandtheircomplements,thereisanopenpositionthatcorrespondstosubjectpositioninotherconstructions.Inmostcasesinthisconstruction,itisinanimate,butitneednotbe.Thiswouldseemtobeatrulyopenpositionandinthatsensenotdefinedbytheconstruction.Theobjectposition(followingdrive)isoccupiedbyanimateobjects,whichcanalsoberealizedsyntacticallyaspronouns.

Theitemsthatarespecifictothisconstructionare:

1.Theverbdriveoccursindifferentinflectedandperiphrasticforms,noneofwhicharespecifictothisconstruction,yettheywouldalloccurinexemplarrepresentationoftheconstruction,linkedtothesetofformsthatdrivehaselsewhere.Thisconstructionoccursoccasionallywithsendandmakeastheverb(Boas2003).2.Thelastpositionisoccupiedbyasetofitemssynonymouswithcrazy.Theseincludemad,nuts,batty,wild,upthewall,andovertheedge.

AlanguageuserwhohasadultlikeexperiencewithEnglishwillhaveencounteredanumberoftokensoftheconstruction,suchasthosein(7)–(9).Allthetokensofthisconstructionencounteredinacorpuswouldconstituteitstokenfrequency.Forinstance,Boas(2003)finds253tokensofthisconstructionwithadjectivesintheBritishNationalCorpus(BNC).Onecanalsocountthetokenfrequencyofaparticularadjectiveintheconstruction,suchaswild,whichoccurstwenty-twotimesintheBNC.Orwecanbeevenmorespecificandcountthenumberoftimestheexactformoccursinacorpus.Forinstance,Itdrivesmecrazyoccursthirty-threetimesinCorpusofContemporaryAmericanEnglish(COCA).

Whatdothesepatternsofoccurrencemeanforcognitiverepresentation?Ifeachtokenofexperiencehasaneffectonrepresentation,theneachoccurrenceoftheconstructionmapsontotheexemplarcloudfortheconstruction.Thefigurein(10)schematizesthisexemplarcloud.Notethattheparticularitemsintheschema,suchasherormadarerepresentedbycloudsofexemplarsthemselves.

In(10)largerfontisusedtorepresentitemsthathaveahighertokenfrequencyintheconstruction,thatis,DRIVE,me,andcrazy.WehaveshowninBybeeandEddington(2006)thatitemswithhighertokenfrequencywithintheconstructionserveasthecentralmembersofthecategoriesthatformforschematicslotswithintheconstruction.Thatis,driveisthecentralverbfortheconstructionandcrazy(forAmericanEnglish)isthecentraladjective.Extensionsoftheconstructionwillbebasedonthesecentralmembers.Inaddition,theseitemsofhigherfrequencywillalsobeeasiertoaccess,whichwillincreasetheirfrequencyevenmore.

Asmentionedinsection4.4,constructionsareformedbythedomain-generalprocessofchunking,bywhich

2

Page 60: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 9 of 14

repeatedsequencesofexperiencecometoberememberedandprocessedasunits.Theseprocessesoperategraduallyinlanguageuse,sothatthecreationofaconstructionasaunitdoesnotmeanthatanalyzabilityislostimmediatelywhenconstructionsareformed.However,extensiverepetitioncanleadtoautonomy,aswewillseeinthenextsection.Forthemoment,notethatafrequentinstanceoftheconstruction,itdrivesmecrazy,whileitmapsontotheschemain(10),alsohasitsownsequentialrelationswhichmaybestrongerthanotherinstancesoftheconstruction,suchasLindsaynearlydrovehimwildwithanxiety.

4.6.2.TypeFrequency

Typefrequencyismeasuredinthecontextofaconstructionandcountshowmanydifferentitemsoccurintheschematicslotsofconstructions.Thus,wemightcountthenumberofadjectivesthatoccurinthedriveXcrazyconstructioninacorpus,asBoasdidfortheBNC,andfindthattherewereeight(providedthatnutsandcrackersinthisconstructionareconsideredadjectives).Countingtheprepositionalphraseswouldincreasethetypefrequencyofthisslot.Constructionalslotsallhavetheirowntypefrequency:theverbslotin(10)hasaverylowtypefrequencywhilethesubjectslotisvirtuallyunlimitedwithinthedomainofnounphrasesandthushasaveryhightypefrequency.

Theimportanceofrepresentingtypefrequencyintherepresentationofaconstructionisthattypefrequencyrelatesdirectlytoproductivity.Ingeneral,thehigherthetypefrequencyofaconstruction,themorelikelyitistooccurwithanovelitem(Bybee1985;Baayen1993;Barðdal2008,2011a).Theeffectoftypefrequencyonproductivityisconstrainedbytwofactors.First,itemswithveryhightokenfrequencymayhaveformedamoreautonomouschunkanddonotactivatetheconstruction'sexemplarcluster;thussuchitemswillnotcontributetoproductivity(Bybee1985).Second,thesemantic(andinsomecasesformal)propertiesoftheschematicslotmayrestrictproductivity.Thus,thecomplementtodrivein(10)mayincludeadjectivessynonymouswithcrazyorphrasessuchasovertheedge,todistraction,tosuicide,butwedonotfindusessuchasdriveXtopneumoniaordriveXtofear.Intermsofformalproperties,thereisaconstructionhavethenounto,wheremanyofthenounsendinthesuffix-ityor-ability,suchashavethecapabilityto.Thisperhapsaccountsfortheextensionhavetheavailabilityto(whichseemsonlymarginallyacceptabletome,butwaswrittenonasheetprovidedbyanoptometrist).

Thesemanticorformalpropertiesofaslotdetermineitsschematicity.Ahighlyschematicslot,suchastheobjectpositionin(10)hasabroadsetofproperties,suchas‘animatenoun.’Alessschematicslotmighthaveanarrowerdefinition,suchas‘synonymouswithcrazy.’Schematicityandproductivityareindependentofoneanother:the‘crazy’slotin(10)isfairlyproductivewithinthissemanticdomain.Thus,drovemebananas,bonkers,berserkareallpossibleextensionsofthisslot.Barðdal(2008)notestheinversecorrelationbetweentypefrequencyandsemanticcoherence,inthatbothpropertiesencourageproductivity,eventhoughaveryhightypefrequencyoftencorrelateswithhighschematicityandlowschematicity(greatercoherence)alsogivesproductivityinthedomaindelimited.Despitethetrade-off,themostproductivecategorieshavebothhighschematicityandhightypefrequency(e.g.,theEnglishregularverbformationwiththesuffix-ed).Bothproductivityandschematicitycanbedetermineddirectlyfromanexemplarrepresentationwhereallthetypesthathaveoccurredintheconstructionarerepresented.

Atleasttwofactorscontributetotherelationbetweenhightypefrequencyandproductivity.First,thegreaterthenumberoftypesinaconstruction,themorebasestherearefortheitem-basedanalogythatcreatesnovelinstancesoftheconstruction.Second,asBaayen(1993)pointsout,theprocessingoflowfrequencyinstancesofconstructionsrequiresparsingoractivationoftherepresentationoftheconstruction,whichstrengthenstheconstruction.Processinghighfrequencyinstancesofconstructions,aswehavealreadyseen,canoccurwithoutactivatingtheconstructionandthereforedoesnotstrengthentheconstructions.

AnotherapproachtofrequencyinconstructionsisCollostructionalAnalysis(StefanowitschandGries2003;Stefanowitsch,thisvolume).Thisapproachrollstokenandtypefrequencyintoonemeasurealongwithcontrolsforoverallfrequencyofthelexicalitemandacountofconstructionsinthelanguage.Inthisapproachitisnotpossibletodistinguishbetweentheeffectsoftokenfrequencyandtypefrequency,whichIfindtobedistincteffectsandthusprefertokeepseparate.

Toconcludethissection,bothtypeandtokenfrequencyareimportanttoourunderstandingofconstructionsas

Page 61: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 10 of 14

theyaffectcategoryformationforslotsinconstructions,productivityoftheseslots,aswellasthedegreesofanalyzabilityofconstructionsandparticularexemplarsofconstructions.Asfrequencyrepresentationsareanintegralpartofexemplarrepresentation,thesefactssupportexemplarmodelsasagoodchoiceforConstructionGrammars.

4.7.HowNewConstructionsDevelopoutofOldConstructions

Afurtherimportantargumentfortheexemplarrepresentationofconstructionsisthefactthatnewconstructionsdevelopoutofexistingconstructions(Bybee2003,2006).Insection4.3wediscussedthefamousWhat'sXdoingY?constructionandthefactthatithasspecialpragmaticimplicationsconventionallyassociatedwithit.Themorphosyntaxofthisconstructioniscompletelyunchangedfromitssource,whichisanormalquestionwithwhatandtheprogressiveformoftheverbdo,yetithasenteredthecatalogofconstructionsofEnglish,witharepresentationthatisatleastpartiallyindependentoftheconstructionfromwhichitarose.Exemplarrepresentationisnecessarytomodelthecreationofnewconstructions:aninstanceofaconstructionwithspecialimplicationsisrecordedinmemoryrepresentation.Ifthesameorsimilarinstancesoccurwiththesameorsimilarimplications,theexemplarswillbegintoformaclusterofrepresentationssuchasthatcharacteristicofaconstruction.

Itisimportanttonotethatthereisnothinginthemeaningofawhatquestionthatcanexplaintheimplicationsofthenewconstruction.Ratheritisthespecificcontextsinwhichtheconstructionisusedthataddthepragmaticimplications.AsIarguedabove,ifthesespecialimplicationsfoundintheparticularcontextswerenotmadepartoftherepresentationfortheinstanceoftheconstruction,nonewconstructionwouldarise.Thus,weseethatspeakersandhearerschangelanguageastheyuseit,andthesechangesarepermanentlyregisteredinrepresentation.

Onceanewconstructionisformed,itcanstillbecloselyrelatedtotheconstructionfromwhichitarose.However,overtimeandmanyusageevents,anewconstructioncanbecomemoreandmoreautonomous.Thisoccursviathechunkingprocessdiscussedabove.Whenasequenceisaccesseddirectlywithoutnecessarilyactivatingrelatedlexicalitemsandconstructions,itstrengthenstheinternalsequentialrelations,butdoesnotstrengthenanyassociationswithotheritems(Hay2001;BecknerandBybee2009;Bybee2010).Overmanyrepetitionsofthisprocess,thenewconstructioncanloseitsanalyzability.Itisthisprocessthatoccursingrammaticalization.Thebegoingto+verbconstructionforfutureinEnglishwasjustoneinstanceofageneralconstructionforexpressingmovementinspaceforapurpose,butoverthelastfewcenturiesithaslostitsassociationwiththepurposeconstruction,aswellaswiththemovementverbgo.Ithasbecomeanautonomousconstructionusedforadistinctfunction,asintheexampleAdealthisbigisgoingtocarryapricetag(TimeMagazineCorpus,2000),wherethespeakerismakingapredictionandnomovementinspaceisexpressed.Inaddition,agrammaticalizingconstructionincreasesintypefrequencyasitcomestobeusedwithmoreandmorelexicalitems(Bybee2003;Himmelmann2004),asforinstance,whenthebegoingtoconstructioncomestobeusedwithinanimatesubjectsandstativeverbs(seealsoFried,thisvolume).

Thus,anexemplarmodelprovidestherepresentationalbasisuponwhichaparticularinstanceofaconstructioncanbeestablishedasanewconstructionandcanbecomeautonomousfromitssource.

4.8.ExemplarSemanticsandPragmaticsforConstructions

Therichmemoryrepresentationthatischaracteristicofexemplarmodelsiscompatiblewiththeideasexpressedincognitivelinguisticsthatlinguisticmeaningishighlydependentuponandnotseparablefromencyclopedicknowledge,orknowledgeofone'sworld(Fillmore1982,1985b;Langacker1987;Lakoff1987;CroftandCruse2004).Therepresentationofencyclopedicknowledgealongwithlinguisticknowledgewouldemergenaturallyfromanexemplarmodelinwhichtokensofwordsorconstructionsarerepresentedinmemoryalongwiththesituationstheyhavebeenassociatedwithinexperience.

Giventhevastarrayofdetailthatmemorycanrecord,apressingtheoreticalissueishowsemanticcategoriesareabstractedfromtherepresentationofexperienceandhowabstractthesecategoriesreallyare.Notethatthisisnotaquestioncreatedinanywaybyexemplartheory.Itisaquestionwemustanswerinanytheoryunlesswewantto

Page 62: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 11 of 14

postulatethatlinguisticcategoriesareinnatelyspecified,foranytheorymusteventuallyaddressthenatureoftheabstractionprocessthattakesplaceinlanguagedevelopmentinthechildandtheresultingrelationshipbetweenexemplarsderivedfromexperienceandabstractcategories(see,forinstance,Abbot-SmithandTomasello2006).

Thisquestionhasbeenaddressedforlexicalmeaningthroughresearchinprototypecategorization,whichhasshownthatthenaturalcategoriesthathumansconstructarenotbasedonhighlyabstractnecessaryandsufficientconditions,butratherconsistofmembersthataremoreorlesssimilartoacentralmemberorprototype.Thequestionconcerningsuchcategoriesishowthefeaturesonwhichsimilarityisbasedarechosen.RoschandMervisarguethatprototype“categoriesformtomaximizetheinformation-richclustersofattributesintheenvironment”(RoschandMervis1975a:458).Tothisstatementonemightadd,“astheenvironmentisexperiencedbyahumanagent.”

Grammaticalmeaning,asmoreabstract,ismuchlessdependentuponencyclopedicknowledge,suchasthatrepresentedinsemanticframes(Fillmore1982,1985b).Grammaticalmeaningseemstomakereferencetothespeechsituationitself(aswithdeicticelements)ortointernalrelationswithintheutterance,bymeansofconstructions.Yet,themeaningofconstructionsandgrammaticalmorphemesmustalsobebuiltupthroughspecifictokensofuse,sothequestionoftherelationbetweenthespecificandtheabstractstillarises.Justaswemustbeabletoidentifyattributesresidingindifferententitiesintheenvironment,andjustasweareabletofindsimilaritiesintheformoflinguisticelements,sowearealsoabletofindattributesthataresharedbysituationsthatweencounterinassociationwithparticularconstructions.Whensharedattributesareidentified,theirrepresentationsarestrengthenedwhilethenonsharedpropertiesofsituationsarenotreinforced.Thus,abstractionoccursbecauseonlycertainfeaturesareassociatedwithaconstruction.

Theextentandnatureofabstractionforgrammaticalmeaningisstillsomewhatcontroversial,asithasbeentraditionaltoattempttoidentifyonlyoneabstractfeatureforeachgrammaticalmorpheme(Jakobson1971andmanyothers).Suchanalysesoftenrunintodifficulty,aswewillseebelow.Asexemplarmodelsdonotinsistonthereductionofacategorytoonlyonefeature,andastheresearchonnaturalcategorizationturnsupprototypeeffects,itisimportanttoexaminethequestionofjusthowabstractgrammaticalmeaningis.Thisquestioncanbeaddressedthroughthenaturalexperimentoflanguagechange,especiallywhenviewedasongoing.InthefollowingIreviewwhatlanguagechangecantellusaboutgrammaticalmeaningandarguethatexemplarrepresentationisnecessaryforourunderstandingofhowthesechangesoccur.

Justasrecentresearchonidioms,prefabs,andothertypesofformulaiclanguage(Nunbergetal.1994;Wray2002a;Corriganetal.2009)hasemphasizedthediversityandspecificityofconstructionsandcollocationsinlanguageuse,sostudiesofgrammaticalizationhaveturnedupmanycasesofusesofgrammaticalmorphemesthatarenotasabstractaswewereledtoexpect.

Undertheheadingof‘retention’(BybeeandPagliuca1987)or‘persistence’(Hopper1991),wefindexamplesinwhichthemeaningordistributionofgrammaticalmorphemesreflectstheirearlierpatternsofuseandmeaning.Forinstance,whilethereisnoquestionthatEnglishwillexpressestheabstractmeaningoffuture,asinexample(11),therearecontextsinwhichameaningof‘willingness’comesthroughasin(12)and(13),reflectingtheearliermeaningofwillas‘desire.’Thesecondwillin(13)canbeinterpretedaseitherpredictionorwillingness(negatedto‘refusal’).

(11)OffofpubliclandsinNevadaaloneoverthenextfiveyears,morethan$10billionworthofgoldisgoingtoberemoved.AndyouandIwillnotgetapennyforthat.(COCA,1990)(12)Thishackerisofferingme$1,000ifIwillgivehimadeclinedcreditcardthathasamother'smaidenname(COCASpoken,2005)(13)andnowhegotsomeonewhowillstandupandsay,‘Well,thisjurydidaterriblejob,becauseIknowthecasebetter,butgee,nooneinlawenforcementwilllistentome.’(COCA,1990)

Thus,despitethefactthatgrammaticalizationyieldsanincreasingabstractnessorgeneralizationofmeaningleadingfrom‘desire’to‘future,’insomecontextsamorespecificmeaningisretained.Thesecontextsareif-clausesasin(12),clauseswithanindefinitesubject,asinthefirstusein(13),andofteninnegativeclauses,asinthesecondinstancein(13).

Inorderforsuchasituationtobeestablishedandmaintainedovertime,itcannotbethecasethatalltokensofwill

Page 63: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 12 of 14

initsmodalusearemappedontoonemeaning.Itmustratherbethecasewhensomeusesofwillweregeneralizingtoexpressonlyfuture,exemplarsofwillinothercontextswereunaffected,belongingtoexemplarclusterspartiallyindependentfromthemorefrequent,generalizingexemplars.Thisphenomenonissimilartotheonewediscussedearlierinconnectionwithwhat'sXdoingY?,wherebyanexemplarofaconstructiontakesonspecialimplicationsandthusbecomesanew,quasi-autonomousconstruction.

Anotherkindofexampleinwhichaveryabstractgrammaticalmorphemeispolysemousandthuswouldberepresentedbymultipleexemplarclustersiscaseswherethemeaningoftheconstructionoverridesthemeaningoftheindividualelementsinit.Asthemeaningoftheconstructionisassignedtothewholeconstruction,elementswithinitcanlosetheircompositionalmeaning.Forinstance,doingintheWhat'sXdoingY?constructionisnolongerinterpretedcompositionally.Similarly,withgrammaticalmorphemeswesometimesfindthattheirgeneralmeaningisobscuredwithincertainconstructions.Forinstance,whentheEnglishPastTenseoccursinanif-clause,itconveysthemeaning‘hypothetical’ratherthan‘beforethemomentofspeech.’Constructiongrammarmatchedwithexemplarrepresentationhandlesthisfactnicely:theconstructionasawholehasameaningandtheoccurrenceofpasttenseusedhereisnotmapped,exceptinform,ontothegeneralsemanticexemplarclusterforpasttense.

Thesetwoexamplesshowthatdespitetheextremeabstractnessofthemeaningoffutureandpast,thereisnorequirementthateverytokenofthesemorphemeshavethesameabstractmeaning.Ratherautonomousexemplarclusterscandevelopforusesinparticularconstructions.

Anothertypeofrelatedcase,alsoamenabletoaconstructioncumexemplaraccount,occurswhenitappearsthatthemeaningofagrammaticalmorphemeorconstructionhasbeen‘absorbed’fromthecontextsinwhichitisoftenused.TheFrenchnegativeconstructionofne+verb+pasissuchacase.Thenounpas‘step’hascompletelylostitearliermeaninginthiscontextandinaddition,ithasabsorbedthenegativemeaningfromtheconstructionsothatitcanbeusedinotherconstructionstoindicatenegative,asinpasdutout‘notatall’orpasbeaucoup‘notmuch.’

Thesemanticprosodiesfoundincorpuswork(Sinclair1996b)alsodemonstratethatexemplarsofwordsorconstructionsareassociatedwithcontextsthatcanatoncebeveryspecificandcoverlongstretchesoftext.SmithandNordquist(2012)arguethattheassociationofnegativeaffectwiththecausetoconstructioncameaboutovertimeasthepercentageofnegativecollocatesincreased.InEarlyModernEnglish,causewasaslikelytooccurwithpositiveorneutralcollocatesaswithnegativeones,thoughtheneutralcollocatesgreatlyoutnumberedthepositiveones,withtheresultthatcausewasassociatedwithnegativeorneutralcollocates.SmithandNordquist(2012)findthatinPresentDayEnglish(PDE),thenegativecollocateshavecometodominate,representing81%ofthetokenstheyextractedfromaPDEcorpus.Note,however,thatitisnotthelexicalitemcauseitselfthathasabsorbedthenegativeaffect,butrathertheparticularconstructionitoccursin.Inthetransitiveconstruction,cause+NP(asincauseanaccident,causedamage),anegativecollocateisfoundin90%ofthecases,whileintheinfinitivalconstruction(asincausetheboattosink),thecollocateisnegative60%ofthetime.Inbothcases,theotherusesareneutralratherthanpositive.Asmightbeexpected,thegeneralmeaningoftheconstructionishighlyaffectedbythelexicaltypesthatoccurinit,anditistheconstructionalunitthataffectsthemeaningofitscomponentparts.

Theseexamplesdemonstratethateventhoughgrammaticalandconstructionalmeaningcanbeveryabstractincertaincontexts,itisalsothecasethatuseinspecificcontextscanaffectthemeaningofgrammaticalandlexicalmorphemes,sometimesprovidingthemwithmorespecificmeaningnotfoundinallcontexts.Thesespecificmeaningscanbeaddedinbyinference,aswediscussedinsection4.7,theycanberetainedfromearlierusesincontext,theycanbeabsorbedfromtheoverarchingmeaningoftheconstructiontheyoccurin,ortheycanbeaddedbyassociationwithcertainlexicalitems.Allofthesescenariosrequirecognitiverepresentationsinwhichagreatdealofcontextisassociatedwithindividualtokens,whichdetermineshowtheyaremappedontoexemplars.

4.9.TheLimitsofExemplarModels

Anexemplarmodelisnot,onitsown,atheoryoflanguage;itisamodelofmemoryrepresentation.Itaddressesonlythemechanismofmemorystorageandtheresultingstructuresorcategoriesanditsaysnothingaboutthecontentoflinguisticcategories,suchaswhylanguagestendtohavepastorfuturemarkersorcategoriessuchas

Page 64: The Oxford Handbook of Construction Grammar

Usage-based Theory and Exemplar Representations of Constructions

Page 13 of 14

‘subject.’Exemplarmodelsspeaktothedegreeofabstractnessofmemoryrepresentations,positingarich,detailedmemoryforexperiencethatspansmanydomains—linguisticform(phonetics),context,affect,andmeaning.Muchofthecomplexityneededforacompletemodeloflanguageisnotprovidedbytheexemplarmodelsthathavedevelopedforotherdomains.ThelevelsofabstractionforconstructionsarguedforinCroft(2001,2003a)andBarðdal(2008,2011a),forinstance,arenotgivenbyexemplartheory,butmustbediscoveredempiricallyforlinguisticconstructions.Also,whileexemplarmodelswouldallowexemplarcloudscontainingawordorotherelementthatoccursinmanydifferentcontexts,thefactorsthatdeterminewhetherawordisanalyzableinacertaincontext,suchasspiteininspiteof—thatis,semantictransparencyandfrequencyofuse—areoutsideofexemplartheory.

Atthesametimeonemightarguethatexemplarmodelsaretoounconstrained,astheyallowanytypeofcategorytoformandanypropertytoprovidethebasisofasimilaritymatching.Idonotconsiderthisaproblem,however,becauseitallowsthesubstanceoflinguisticexperience(asfilteredthroughhumancognitionandattention)tobetheprimedeterminantofthenatureofthecategories.Thus,itleavesmanylinguisticquestionstobeaddressedempirically,throughthecomparisonoflanguagesandthestudyoflanguagechange.

4.10.Conclusion

Usage-basedtheorycoversawiderangeofresearchquestionsarisingfromthepremisethatuseoflanguagefigurescriticallyindeterminingthenatureofcognitiverepresentationsoflanguage,orputanotherway,usageeventscreatelinguisticstructure.Thepremiseofusage-basedgrammarleadsdirectlytoconstructionswiththeirdirectassociationofmeaningwithformasthebasicunitsofgrammar.Exemplarrepresentationiscentraltoausage-basedapproachtolanguage,sinceitreflectsthewaythatlinguisticstructureemergeswhenhumancognitiondealswithtokensofexperiencewiththeworld.Ihavearguedherethatexemplarrepresentationhelpsusexplainmanyofthedynamicpropertiesofconstructions,suchashowtheyarisefromotherconstructionsandhowtheychangeovertime.

Afurtherimportantadvantageofexemplarmodelsistheirabilitytoregistertokenandtypefrequencyforconstructions.Frequencyeffectsraisemanyinterestingquestionsthatarestillunderinvestigation,suchasquestionsconcerningtheinteractionoftokenfrequencywithtypefrequency,thegradualdevelopmentofautonomy,theinteractionofschematicityandtypefrequencyindeterminingproductivity,andthequestionoftheeffectsofdifferentlevelsoftokenfrequency.

Twopropertiesoftheusage-basedapproachareworthemphasizing.First,representationsaredynamicandchangewithusageevents,notjustacrossgenerationsbutwithintheindividualasusagepatternschange.Theabilitytorepresentchangeprovidesinvestigatorswithexplanationsforhowandwhyconstructionsdevelop.Second,theprocesseswehaveexaminedherethatcreateconstructionsandthereforegrammararedomain-general.Theprocessesthatcreateconstructions—cross-modalassociation(forlinkingsoundandmeaningorcontext),chunking,andcategorization—areallprocessesthatapplyinothercognitivedomains,suchasvisionormotorproduction.Thesetwopropertiestogetherprovidealinguistictheorywithpowerfulexplanatorypossibilities.

Notes:

(1.)TimeMagazineCorpus,http://corpus.byu.edu/time/.

(2.)CorpusofContemporaryAmericanEnglish,http://www.americancorpus.org.

JoanL.BybeeJoanL.Bybee(B.A.UniversityofTexas,1966;M.A.SanDiegoStateUniversity,1970;Ph.D.,Linguistics,UCLA,1973)wasonthefacultyattheStateUniversityofNewYorkatBuffalofrom1973to1989andisnowDistinguishedProfessorEmeritaoftheDepartmentofLinguisticsattheUniversityofNewMexico,whereshehasservedasdepartmentchair,associatedean,anddirectorofthe1995LSALinguisticInstitute.Bybee’sresearchinterestsincludetheoreticalissuesinphonologyandmorphology,languageuniversals,andlinguisticchange.Herworkutilizinglargecrosslinguisticdatabases,e.g.Morphology:AStudyoftheRelationbetweenMeaningandForm(1985),TheEvolutionofGrammar:Tense,AspectandModalityintheLanguagesoftheWorld(withReverePerkinsandWilliamPagliuca,1994),providesdiachronicexplanationsfortypologicalphenomena.Inaddition,herbookspresentingausage-basedperspectiveonsynchronyanddiachronyincludePhonologyandLanguageUse(2001),FrequencyofUseandtheOrganizationofLanguage(2007),andLanguage,UsageandCognition(2010).

Page 65: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 1 of 17

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntax,LanguageandCognition

OnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0005

ConstructionsintheParallelArchitectureRayJackendoffTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterdiscusseswhattheParallelArchitecturehastakenfromConstructionGrammarandwhatitmightcontributetoConstructionGrammar.Afteroutliningthefundamentalsofthearchitecture,itexplainswhyrulesofgrammarshouldbeformulatedaslexicalitemsencodedaspiecesofstructure:thereisnohardlinebetweenwords,constructions,andstandardrules.Thechapteralsoarguesfora“heterogeneous”varietyofConstructionGrammar,whichdoesnotinsistthateverysyntacticconstructionisinvestedwithmeaning.Finally,itdiscussesthecrucialissueofsemiproductivity,usuallythoughttobeapropertyofmorphology,showingthatconstructionstoocanbeeitherproductiveorsemiproductive.

Keywords:ParallelArchitecture,ConstructionGrammar,lexicon,grammar,syntax,semantics,phrasalconstructions,lexicalitems

Overthepasttwodecades,Ihavebeendevelopingatheoreticalandformalconceptionofthelanguagefaculty,theParallelArchitecture,withaneyetoincorporatingtheinsightsofmanydistincttheoreticalframeworks(Jackendoff1997a,2002a;CulicoverandJackendoff2005).ThepresentchapterdiscusseswhattheParallelArchitecturehastakenfromConstructionGrammarandwhatitmightcontributetoConstructionGrammar.Section5.1laysoutthefundamentalsofthearchitecture;section5.2discussestherelationbetweenlexiconandgrammar,anareawheretheParallelArchitecturesharesagreatdealwithcertainversionsofConstructionGrammar.Section5.3brieflydescribeshowthenotionofphrasalconstructionsfounditswayintotheParallelArchitectureataboutthesametimethatConstructionGrammarwasbeginningtodevelop.Section5.4addressesthedegreeofindependenceofsyntaxfromsemantics,arguingthatcertainvariantsofConstructionGrammarlendthemselvestobetterintegrationwiththeParallelArchitecture—andbettertheoreticalgenerality—thanothers.Section5.5returnstothestructureofthelexiconwithintheParallelArchitecture;itdiscussesproductivityandsemiproductivityandtheirbearingoninheritancehierarchies,animportanttheoreticalconstructinConstructionGrammarandotherlexicalistframeworks.

5.1.TheParallelArchitecture

ThebasicpremiseoftheParallelArchitectureisthatphonology,syntax,andsemanticsareindependentgenerativecomponentswithinthelanguagecapacity,eachwithitsownprimitivesandprinciplesofcombination.Thetheorybuildsoninsightsaboutphonologyandsemanticsthataroseinthemiddleandlate1970s.Atthattime,itbecameclearthatphonologyhashighlyarticulatedstructurethatcannotbederiveddirectlyfromsyntax.Inparticular,hierarchicallystructuredunitssuchassyllables,feet,andprosodicconstituentsdonotcorrespondone-to-onewithsyntacticunits.Moreover,phonologicalstructureincludesseveralindependentsubstructuresor‘tiers,’eachwithitsowntypeofgenerativestructure:segmental-syllabicstructure,themetricalgrid,intonationcontour,and(intonelanguages)thetonetier.Thetiersarecorrelatedwitheachotherby‘interfacerules’:principlesthatestablishoptimalcorrespondencebetweenstructuresoftwoindependenttypes.Forinstance,stressrulesare

Page 66: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 2 of 17

treatednotasalgorithmicrulesthatderivestresspatternsdenovo,butasinterfacerulesthatnegotiatethebestcompromisebetweenasegmental-syllabicpatternandanoptimalmetricalgrid.Inturn,sincethesemultitierphonologicalstructurescannotbederivedfromsyntacticstructures,theconnectionbetweensyntaxandphonologyasawholehastobemediatednotbyderivations,butbyacomponentofinterfacerules.

The1970salsowitnessedthedevelopmentofmanyapproachestomeaning,rangingfromformalsemanticstoCognitiveGrammar,withmyConceptualSemantics(Jackendoff1983,1990,2002a,2010)somewhereinbetween.Despitetheirdifferences,alloftheseapproachesseemeaningstructuresascombinatorialandpotentiallyunlimitedincomplexity.ThesestructuresarenotbuiltoutoforderivedfromsyntacticunitssuchasNPsandVPs,astheywereinGenerativeSemantics.Rather,theyarebuiltofcharacteristicsemanticunitssuchasconceptualizedobjects,events,times,properties,andquantifiers,whichdonotcorrespondone-to-onewithsyntacticcategories.Asaconsequence,semanticstoomustbetreatedasagenerativecomponentoflanguagewhichcannotbederivedfromsyntax,butwhichratheriscorrelatedwithsyntaxbyacomponentofinterfacerules.Moreover,semanticstructure,likephonology,demonstrablyhasanarticulationintotiers,includingatleastpropositionalstructure(whodidwhattowhom)andanorthogonaldimensionofinformationstructure(topic/focus/commonground;oldvs.newinformation).Wordmeaningscontributeprimarilytopropositionalstructure;butwordssuchasonlyandeventhatassociatesemanticallywithtopicand/orfocuscontributetoinformationstructure,asdovariousprosodiccontoursandsyntacticconstructionssuchascleftandpseudocleft.

Syntaxtoomaybearticulatedintoindependenttiers.ThehallmarkofLexicalFunctionalGrammar(LFG)(Bresnan1982,2001)isanarticulationofsyntaxintoconstituentstructure(thestandardtree)andfunctionalstructure(grammaticalfunctions);astripped-downversionofthelatterappearsasthegrammaticalfunctiontierofCulicoverandJackendoff(2005).AutolexicalSyntax(Sadock1991)andRoleandReferenceGrammar(VanValinandLaPolla1997)proposeanarticulationintophrasalandmorphosyntactictiers—syntacticprinciplesthatoperateaboveandbelowthewordlevel,respectively.ThisdivisionhasalsobeenincorporatedintotheParallelArchitecture.Again,thetiersofsyntaxmustbebroughtintoregistrationwitheachotherbyinterfacerules.

TheresultisanarchitectureoftheoverallforminFigure5.1;phonology,syntax,andsemanticsalsohavefurtherinternalarticulationofasimilarsort.Inadditiontothetraditionalsyntax-to-phonologyandsyntax-to-semanticsinterfaces,Figure5.1alsoincludesaphonology-to-semanticsinterface,makingpossibleadirectcorrelationofprosodiccontourswithinformationstructure.Inthismodel,awell-formedsentenceisatripletconsistingofwell-formedphonological,syntactic,andsemanticstructures,pluslinksbetweencorrespondingconstituentsofthethree,establishedbytheinterfacecomponents.

OnecanregardMainstreamGenerativeGrammar(Chomsky1965,1981,1995)asavariantofthisarchitectureinwhichthecombinatorialpropertiesofphonologyandsemanticsarederivedfromsyntax;hence,therearenoindependentformationrulesforphonologyandsemantics,onlyderivationalinterfacesfromsyntaxintothesetwocomponents.Bycontrast,CognitiveGrammar(e.g.,Langacker1987)claimsthatall(oratleastmost)syntacticstructureissemanticallymotivated,soiteliminatesorminimizesthesyntacticformationrules(Broccias,thisvolume).WithintheParallelArchitectureframework,afundamentalempiricalissueistheproperbalancebetweenthesetwoextremes.

Figure5.1. TheParallelArchitecture

Animportantconstraintonthebalancecomesfromthefactthatsemanticstructureultimatelyhastoberichenoughtosupportreasoning(inference)andconnectiontotheperceivedworld(reference).Inthemainstreamarchitecture,whichderivessemanticstructure,allthisrichnesshastocomefromsyntax.Thisputssyntactictheoryunderconstantpressureforgreaterarticulationandcomplexity.Bycontrast,theParallelArchitecturegrants

Page 67: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 3 of 17

semanticsitsowngenerativecapacity,andthereforesyntaxhastobeonlyrichenoughtomodulatethemappingbetweensemanticsandphonology—aSimplerSyntax(CulicoverandJackendoff2005).However,thisdoesnotentailthatsyntaxcanbeeliminatedorentirelypredictedfromsemanticprinciples;thereisstillinterestingworktobedonebythesyntacticcomponentofthegrammar.

AcrucialadvantageoftheParallelArchitectureoverthemainstreamarchitectureisthatitextendsnaturallytotherelationbetweenlanguageandothercognitivecapacitiessuchasvision.Obviously,thecombinatorialstructureofthevisualfieldcannotbederivedfromsyntacticNPsandVPs.Nevertheless,speakersreferfreelytothestructureoftheworldasunderstoodthroughvision,forexample,inreportingthatThecatisonthemat.FollowingtheoverallconceptionoftheParallelArchitecture,therelationbetweenlanguageandvisioncanbetreatedasyetanotherinterfacecomponent,thistimelinkingthesemanticstructureoflanguagetothecombinatorialstructuresresponsibleforvisualunderstanding(Jackendoff1987,2010:chapters3,4).Hence,therelationshipsamongfacultiessuchaslanguageandvisionaremediatedbyexactlythesamesortsofformalcomponents—interfacerules—astherelationshipsamongthecomponentsoflanguageand,withinthesecomponents,thesubcomponentsortiers.Inotherwords,theParallelArchitectureisapplicabletoallscalesofmentalorganization.

5.2.TheLexiconandtheGrammar

Everytheoryofgrammartreatswordsastriplesofphonological,(morpho)syntactic,andsemanticinformation.InMainstreamGenerativeGrammar,thesetriplesareinsertedintosyntacticstructuresbylexicalinsertion(or,intheMinimalistProgram,byMerge).Atsomepointinthederivation,theirphonologicalpartsaresenttophonology(PF)wheretheyarepronounced;theirmeaningsaresenttothesemanticswheretheyarecombinedwitheachotherandinterpreted.IntheParallelArchitecture,however,awordfunctionsasasmall-scaleinterfacerule:itstipulatesthatitsphonological,(morpho)syntactic,andsemanticstructurescanbelinkedaspartofawell-formedsentence.

Forinstance,hereare(oversimplified)lexicalentriesforthewordsgoandinto.

(1)a.Phonology:ɡoʊSyntax:V –PPSemantics:[ GO ([ x],[ y] )]b.Phonology:ˡıntuːSyntax:P –NPSemantics:[ INTO ([ z] )]

Thesubscriptsin(1)stipulatepartsofthethreestructuresthataretobelinked:thephonology/ɡoʊ/linkstoasyntacticVandtothesemanticfunctionGO;thephonology/ιntuː/linkstoasyntacticPandtothesemanticfunctionINTO.Theitalicizedpartsof(1)aretypedvariablesthataretobeinstantiatedbyothermaterialinthephrase.Forinstance,inthesyntax,goistobefollowedbyaPP;inthesemanticsithastwoarguments,aThingandaPath,thelatterofwhichislinkedtothePPinsyntax.

Forafirstapproximation,theformalismin(1)canbetranslatedintotheConstructionGrammarnotionofa‘sign,’wherephonologyandsyntaxcomprise‘form’andsemanticscomprises‘function.’ However,differenceswillarisebelow.

Itiscrucialtodistinguishthenotionofa‘word’fromthatofa‘lexicalitem.’DiSciulloandWilliams(1987)showthatthereareimportantgrammaticaldistinctionsbetweenwordsandmorphemes,ontheonehand,andphrases,ontheother;thatis,thereisagrammaticalnotionofwordthatseparatesmorphology(combinatorialitywithinwords)fromphrasalsyntax(combinationsofwords).However,theyalsoassertthattheissueofwhatislistedinlong-termmemoryis“ofnointeresttothegrammarian.”Incontrast,theParallelArchitecturetakesthisdistinctiontobecrucialtodefiningthelexicon:whatDiSciulloandWilliamshaveshownisthatthenotionof‘grammaticalword’andthenotionof‘lexicalitem’arenotcoextensive.

Accordingly,Jackendoff(2002a:chapter6)comparesthreepossiblecriteriafordefiningalexicalitem.

(2)Lexicalitemsare:a.Thewords(andmorphemes)ofthelanguage.

1

1

1 2

Event 1 Thing Path 2

3

3 4

Path 3 Thing 4

2

Page 68: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 4 of 17

b.Theexceptional,idiosyncraticpartsofthelanguagethatcannotbepredictedbyrule.c.Thepartsoflanguagethatarelistedinlong-termmemory.

Criteria(2a)and(2b)aretraditionalviewsofthelexicon,inwhichitisacomponentoflanguagealtogetherdistinctfromrulesofgrammar.TheParallelArchitecture,however,adoptsthemorepsycholinguisticallybaseddefinition(2c).Thethreecriteriacoincideinthecaseofastereotypicallexicalitemsuchasdog:itisawordofEnglish,itisnotpredictablebyrule,anditisstoredinspeakers’long-termmemory.Butthecriteriadivergeinlessstereotypicalcircumstances.Forinstance,thepronunciation,syntax,andmeaningofthewordunhappyarecompletelypredictablefromthewordhappyplustheregularprincipleforthestructureandmeaningofun-adjectives.Thus,althoughunhappyisawordofEnglishandthereforesatisfies(2a),itdoesnotsatisfy(2b);anditsstatusvis-à-vis(2c)isanempiricalquestion(myguessisyesforunhappy,thoughperhapsnotfor,say,unzippable).

Foradifferentcase,theidiomskickthebucketandchewthefatarenotwordsormorphemesofthelanguage,sotheydonotsatisfy(2a).Theirphonologyandsyntaxarepredictablefromthephonologyandsyntaxoftheconstituentwords,justlikethrowthebucketandchewthegum;buttheirmeaningisnotpredictable.Sothelanguagemustnotetheiridiosyncrasy(2b),andthisidiosyncrasymustbestoredinspeakers’long-termmemory(2c).Usingtheformalismin(1),kickthebucketcanbeencodedas(3)(wherethesemanticsisapproximate).

(3)Phonology:[ kιk] [ ðə] [b∧kιt]Syntax:[ V [ Det N ]]Semantics:[DIE([ x])]

Whatmakesthisanidiomisthatthelinkingbetweensyntaxandsemantics(subscript4)isindependentofthelinkingbetweenphonologyandsyntax(subscripts1,2,3):intraditionalterms,theindividualwordsdonothaveindependentmeanings.ThistreatmentofidiomsfindsparallelsinHPSG,CognitiveGrammar,andConstructionGrammar.

Clichésandfixedexpressionssuchasmoneycan’tbuymeloveandIdon’tcarealsoconsistofmultiplewords.Butunlikeidioms,suchphrasesretain(muchmoreof)theirliteralpredictablemeanings.Hence,theysatisfyneithercriterion(2a)norcriterion(2b).Yettheydosatisfy(2c):whatmakesthemclichésispreciselythattheyarephraseseveryoneknows.

Theformalismforlexicalentriesalsogeneralizesreadilytoregularaffixes.Forinstance,theregularpasttenseverbaffixofEnglishcanbewrittenas(4).

(4)Phonology:[ [ x] [ -d] ]Syntax:[ V +[past] ]Semantics:[ PAST [ y] ]

Thissaysthatthephonologicalsuffix-dattachedtoaphonologicalwordcorrespondstothepasttenseformoftheverbwhosepronunciationisthatphonologicalword, anditexpressespasttime(amongotherthings)oftheeventorstatetypeexpressedbytheverb.Theupshotisthat,intermsofcriterion(2c),(4)isalexicalitem,butnotaword;andanovelinflectedverbsuchasskypedisaword—butnotalexicalitem.

Oneconsequenceofthisapproachisthatregularmorphologyistreatedanalytically(ordeclaratively)ratherthanprocedurally.Forinstance,theEnglishregularpasttenseisnotarulethatsays“add-dtoaverbtoformitspasttense.”Rather,thepasttensemorphemeisthelexicalentry(4),whichcombineswithverbsinexactlythesamewaytransitiveverbscombinewiththeirdirectobjects,namelythroughtheinstantiationofavariable.Inprinciple,thisapproachextendstoallregularaffixation—anissuethatbecomesfarmoresignificantinlanguagessuchasTurkishthathavemassiveregularmorphology.Wereturntomorphologyinsection5.5(seealsoBooij,thisvolume).

Alexicalitemneednotbeafulltripleofphonology,syntax,andsemantics.Wordssuchasyes,hello,wow,ouch,andallakazamdonotparticipateinphrasalcombination,andthereforetheyarebestthoughtofashavingphonologicalandsemanticfeaturesbutnosyntacticfeatures. Wordsthatfunctionas‘grammaticalglue,’suchasexpletiveitandthedoofdo-supportforinversionandnegation,havephonologicalandsyntacticfeaturesbutnosemantics.Similarly,nonsensephraseslikefa-la-la,heynonnynonny,doodahdoodah,andinka-dinka-doo,usedtofillupspaceinsonglyrics,havephonologybutneithercombinatorialsyntaxnorcombinatorialsemantics

Wd 1 Cl 2 3

VP 1 NP 2 3 4

Thing 4

Wd Wd 2 Aff 1

V 2 1

Event/State 1 Event/State 2

3

4

Page 69: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 5 of 17

(thoughtheymayconveyaffect).Soalthoughastereotypicallexicalitemconsistsofafulltripleofthreestructures—aSaussureansign—therealsoexistitemswithlessthanthefullcomplement.Inotherwords,likeeveryotherhumancategory,lexicalitemshavemanynonstereotypicalcaseswithpartiallydivergentproperties.

Next,considerthesyntacticstructureoftheidiomkickthebucketin(3).ThisisnothingbutarealizationofthenormalphrasestructurerulesforVPandNPshownin(5).

(5)a.[ VNP]

b.[ DetN]

Ifthesestructuresareadmissibleaspartofalexicalitem,thereisnoreasontheycannotalsostandaslexicalitemsontheirown.TheParallelArchitecturethereforeviewsphrasestructurerulesaslexicalitemsthathaveonlysyntacticfeatures.Sincetheylackphonologyandsemantics,theyalsolacklinkingsubscripts—andhencetheyarenotSaussureansigns.Inthesameway,theregularrulesthatlinkgrammaticalfunctionssuchassubjectandobjecttothematicrolessuchasAgentandPatientaretreatedaslexicalitemsthatlinksyntaxandsemanticsbutlackphonologicalfeatures.(Goldberg,thisvolume,andothersinConstructionGrammarcalltheselinkingrules‘abstractconstructions,’preciselybecausetheylackphonology.)

Inshort,theoverallconceptionis:

•Lexicalitemscancontainanycombinationofphonological,syntactic,andsemanticstructures,pluslinksamongthem.

•Lexicalwordssuchas(1),idiomssuchas(3),regularaffixessuchas(4),andrulesofgrammarsuchas(5)areencodedinacommonformat(concurringwithLangacker1987andCroft2001).

•Thegrammarofalanguageconsistsofstoredpiecesofstructureinthethreecomponentsandstoredlinksamongthem.

•Thecombinatorial‘engine’thatbuildsand/orlicensesphrasesandsentencesistheoperationofunification(Shieber1986),which‘clipstogether’storedpiecesintoafullutterance.

5.3.ConstructionsintheParallelArchitecture

Thenotionofconstructionaroseinmyownworkinthecourseofstudyingunusualcasesofverbargumentstructure(Jackendoff1990).OnephenomenonthatparticularlyattractedmyattentionwastheWayconstruction:

(6)a.Billbelchedhiswayoutoftherestaurant.b.Joejokedhiswayintothemeeting.c.BabeRuthhomeredhiswayintotheheartsofAmerica.

AlthoughtheVPsinthesesentenceshavecanonicalsyntacticconstituency(V-NP-PP),theNPandPPcomplementsarenotlicensedbytheverbs,whicharenormallyintransitive.Moreover,thephrasehiswayapparentlyoccupiesdirectobjectposition,asitisimpossibleeithertoaddadirectobject(7a)ortosubstituteanotherdirectobjectforhisway(7b).

(7)a.*Joetoldjokeshiswayintothemeeting.b.*BabeRuthhithomerunsintotheheartsofAmerica.

Thesentencesin(6)areunderstoodasexpressingmotiononthepartofthesubject,eitherliteral(6a,b)ormetaphorical(6c),andthePPexpressesthepathofmotion.Theverbisunderstoodasamannerormeansmodifierofthemotion,asshownintheapproximateparaphrasesin(8).

(8)a.Billcameoutoftherestaurant,belching(allthewhile).b.Joewentintothemeeting,joking(allthewhile).

VP

NP

Page 70: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 6 of 17

orJoegotintothemeetingbyjoking.c.BabeRuthenteredtheheartsofAmericaby/throughhomering.

Jackendoff(1990)exploredthreepossibleapproachestothisphenomenon.Thefirstderived(6)fromanunderlyingsyntacticstructurethatmoredirectlyreflectsthesemanticroleoftheverb,followingstandardmainstreampractice.Essentially,theverbstartedoutinanadjunct,asin(8),thenmovedtothemainverbpositionofthesentence.Itwaseasytoshowtheempiricalinadequacyofsuchaderivation,andIabandoneditwithoutregret.Thesecondapproach(lateradvocatedbyLevinandHovav1995)wastotreattheconstructionalongthelinesofthepassiveinLFGandHPSG:thereisa‘lexicalrule’thatconvertsintransitiveverbsoftheappropriatesemanticsortintoidiomsoftheformVX'swayPP,whichthenundergolexicalinsertionandargumentsatisfaction.Thiscanbemadetowork,butithastheeffectoflitteringthelexiconwithidiomssuchasbelchX'swayandhomerX'swaywhoseexistenceisentirelypredictable.Surelythesecombinationsneednotbestored.

ThethirdapproachwastotreatVX'swayPPitselfasanidiom—ameaningfulconstruction—thatcombinesproductivelyinsyntaxwithintransitiveverbsofaparticularsemantictype.Fromthepointofviewofdescription,thisisanotationalvariantofthesecondapproach,inthatallthesameconditionsontheverb,thePP,andtheinterpretationobtain.However,thisapproachalsoaccountsforthefactthattheruleisproductive,thewaysyntaxissupposedtobe.Inparticular,itdoesnotrequireonetosaythatthelexiconcontainsallthepossiblecombinationsofX'swaywithverbs.Thepriceisthatonehastoabandonthenotionthattheverb,asheadoftheverbphrase,isinvariablyalsoheadintheinterpretation—theelementthatdeterminestheVP'sargumentstructure.Rather,intheWayconstruction,theconstructionitselfdeterminestheargumentstructureoftheVP,andtheverbfunctionssemanticallyasameansormannermodifier.

Jackendoff1990discussedanumberofsuchVPconstructionsandleftopenthequestionofwhichofthesetwolatterapproachesiscorrect.However,ataboutthesametime,anumberofpapersbyAdeleGoldberg(eventuallyleadingtoher1995book)dealtwithmanyofthesameconstructionsIhadbeenaddressing.Thisandotherearlyworkonconstructionspersuadedmethattheconstructionalapproachwasthemorepromisingofthetwo.

WithintheParallelArchitectureformalism,theWayconstructioncanbetreatedasalexicalitemwithapproximatelythefollowingstructure:

(9)Phonology:weιSyntax:[ V [ pro +genN ]PP ]Semantics:[GO([ x],[ y] );WHILE/BY[F([ x])] ]

Theformalismin(9)isexactlythesameasforwords,idioms,affixes,andphrasestructurerules.Thedifferenceisinthecomplexityofthephonology-syntax-semanticcorrespondence,indicatedbythesubscripts.Inparticular,theonlypartofthesyntaxthatislinkedtophonologyisway,thedistinctivemarkoftheconstruction,andthesemanticfunctionsGOandBY/WHILEarenotlinkedtoanyparticularsyntacticorphonologicalconstituent.Thevirtueofthiscomplexityisthatitcompressestwoconceptualclauses—motionandmanner/means—intoasinglesyntacticclause.

Sincethen,alargenumberofconstructionshavebeeninvestigatedintheParallelArchitectureframework;alistappearsintheAppendix.

5.4.IntegratingConstructionGrammarandtheParallelArchitecture

ItiseasytoseethatvariousversionsofConstructionGrammarcanbeconstruedasrealizationsoftheParallelArchitecture.Inparticular,giventhattheParallelArchitecture'slexicondisplaysacontinuumamongwords,multiwordexpressions,andphrasestructurerules,constructionssuchastheWayconstructionfitintothelexiconaltogethernaturally.ThemainquestioniswhichversionofConstructionGrammarismostappropriate.

Ifirsthavetoclearawayaterminologicalissue.Croft(2001,thisvolume)andGoldberg(2006a,thisvolume)usethetermconstructionforallstoredpiecesofstructure,includingwords,idioms,andphrasestructurerules.Ithinkthisuseofthetermconstructioniscoextensivewiththetermlexicalitemasusedhere—itisapieceoflinguisticstructurestoredinlong-termmemory.Formypart,Ifinditconvenienttoretainaterminologicaldistinctionbetweenwords(whichtheycall‘lexicalconstructions’),idioms,phrasestructurerules,linkingrules(or‘abstract

1

VP 2 NP bound 1 3

Thing Path 3 Thing 2

Page 71: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 7 of 17

constructions’),andmeaningfulconstructions,thelatterbeingcanonicalornoncanonicalpiecesofsyntacticstructurelinkedtononcanonicalmeanings,suchastheWayconstruction.Butthereisnotheoreticaldistinctionamongthem—theyareallencodedinacommonformat,withnosharpdividinglines.

Amoresubstantiveissueconcernsthecontentofconstructions.Ononeapproach,whichIwillcallhomogeneousConstructionGrammar(e.g.,Goldberg1995,2006a,thisvolume;Croft2001,thisvolume;Barðdal2008;Michaelis,thisvolume,Broccias,thisvolume),allsyntacticformistakentobedeterminedbymeaningfulconstructions;therearenoautonomoussyntacticprinciples,freeofsemanticimplications;alllinguisticunitsaretakentobefullSaussureansigns.Onanotherapproach,whichIwillcallheterogeneousConstructionGrammar(e.g.,myreadingofFillmoreandKay1993),meaningfulconstructionsarejustonekindofabstractstoredstructure.Thegrammarofalanguagecanalsocontainindependentprinciplesofsyntacticformsuchas(5),aswellasindependentprinciplesofsemanticstructurethathavenosyntacticeffect,whichIwilldiscussshortly.Iwouldliketoarguethattheheterogeneousapproachismoresatisfactory.Onthelatterview,thetypicallexicalitemisaSaussureansign,but,assuggestedinsection5.2,therearealsonumerousatypicallexicalitemsthatarenotsigns.

Tosetthestage,wefirstnotethatthereareindependentprinciplesofstructureindomainsotherthansyntaxandsemantics.Phonotacticsareprinciplesofphonologicalstructurethathavenothingtodowithanyotherlinguisticdomain;theymightbeconsidered‘purelyphonologicalconstructions,’withnosyntacticorsemanticeffects.Atalargerscaleinphonology,poeticmeterssuchasiambicpentameterandpoeticformssuchasthehaikuandthesonnetcanbethoughtofasprosodicconstructionstowhichrelevantphonologicalstructuresmustconform,butwithnoimplicationsforthemeaningofthetext.Evenfurtherafield,musicalformssuchasthe12-barbluesandminuetformimposeametricalandharmonicstructureoncompositions,withoutanyimplicationsforinterfaceswithlanguage.Ifwedesiretointegrateourtheoryofsyntax/semanticswithanoverallviewofthearchitectureoflanguageandofthemind,theseconsiderationsshouldplayarole.

Turningbacktosyntax,therearesyntacticstructuresthatdonotsupportmeaningdistinctionsandarepresentjusttomarshaltheform.Oneobviouscaseistheprincipleofinvariableheadpositionwithinaphrase:itmakesnodifferenceformeaningthattheEnglishverbisphrase-initialandtheJapaneseverbisphrase-final.(Notethatdiachronicand/orprocessingmotivationsforfixedheadpositionhavenobearingonthesynchronicformofthegrammar,whichstillhastosaywheretheverbgoesintheclause.)Similarly,Englishdo-supportispresentininvertedandnegatedclausesonlytoprovideabearerfortense,nottoexpressadifferenceinmeaning,andexpletiveitispresentonlytosatisfytherequirementthatatensedclausemusthaveasubject.Inlanguageswithgrammaticalgender,principlesofgenderagreementmakenosemanticdifference(exceptwhereagenderedpronounorotheritemcanbefortuitouslyusedfordisambiguation).

Othersyntacticstructuresarecorrelatedwithmeaning,butnotuniformly.ConsiderforinstancetheEnglishTransitiveVerbconstruction.Dependingonthechoiceofverb,thepostverbalNPmaybeatheme,agoal,anexperiencer,astimulus,andsoon.InordertomaintainahomogeneousConstructionGrammar,onemightsaythatthegeneralmeaningoftheconstructionisthattheNP'smeaningisasemanticargumentoftheverb'smeaning;Goldberg(thisvolume)expressesthisintuitionbysayingthattheVNPconstructionexpressespredication.Inturn,eachoftheparticularthematicrolesforthedirectobjectmightbeitsownsubconstruction(Predicate-Themeconstruction,Predicate-Goalconstruction,etc.),fallinginaninheritancehierarchyunderthegeneralTransitiveVerbconstruction.

However,suchaconstrualofthetransitiveverbconstructionwouldexcludeasizablenumberofconstructionsinwhichthedirectobjectisnotanargumentoftheverb.First,itwouldexcludethemanyidiomssuchaskickthebucketthathavetheformoftransitiveVPs,butwhosedirectobjecthasnodiscretemeaning.Second,itwouldexcludetheWayconstructionin(10a)andtheTime-awayconstructionin(10b),inwhichthepostverbalNPislicensedbytheconstructionandtheverbfunctionsasamannerormeans.

(10)a.Bobdrankhiswaythroughthewholemeeting.b.Bobdranktheafternoonaway.

Third,itwouldexcludeobligatorilyreflexiveverbssuchasthosein(11),wherethereflexiveisasyntacticargumentbutnotasemanticargument.

Page 72: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 8 of 17

(11)a.Dickperjured/assertedhimself.b.Sueavailedherselfoftheopportunity.

Fourth,itwouldexcludeso-calledobjectraisingverbssuchasexpect,whichcanhaveobjectsthathavethematicrolesonlywithrespecttoasubordinateverb,ornothematicroleatall,asin(12).

(12)a.Bobbelievestheshittohavehitthefan.b.Bobexpectsittorain.

Fifth,itwouldexcludeLightVerbconstructionssuchas(13),inwhichthesyntacticdirectobjectcarriesmostofthesemanticcontentofthepredicate.

(13)a.Billtookawalk/ashower/anap.b.Suemadeadecision/avow.c.FredputtheblameonBillfortheaccident.d.AmygaveTomakissontheear/apunchinthearm.

Thus,aTransitiveVerbconstructionwiththesemanticfunction‘predication’isnotgeneralenoughtoencompassallthecasesoftransitiveverbsinEnglish.

ThetransitiveVPidiomsandtheconstructionsin(10)–(13)havethesyntaxofthenormalTransitiveVerbconstruction,buttheyusetheobjectforsomethingotherthanitsnormalthematicfunction—eachinadifferentway.Sincethereisnopointofsemanticgenerality,ahomogeneousConstructionGrammarmusttreatitasanaccidentthatthesealldisplaythesamesyntax.Abettersolutionistosetthesyntacticconfiguration[ VNP]atthetopoftheinheritancehierarchy,independentofsemantics.ThenthePredicationconstruction,aswellasalloftheconstructionsin(10)–(13),caninherittheirsyntaxfromthisconfiguration,whiledifferingintheirsemantics.Thisamountstocalling[ VNP]a‘form-only’construction,whichispossibleonlyifoneadoptsaheterogeneousconceptionofConstructionGrammar.

AsimilarsituationariseswiththeEnglishVerb-Particleconstruction,whichhasatleastfivedifferentinterpretations,eachwithadifferentrelationbetweensyntacticandsemanticconstituency.

(14)a.Particleisadirectional(path)argumentoftheverb,alternatingwithpathPPs;postverbalNPandParticleareindependentsemanticconstituents.JohnpushedBillin/out/through/intothehole/outthewindow/throughthepipe.b.Particleformsanidiomwiththeverb;postverbalNPisanindependentsemanticconstituent.Johnlookedtheanswerup.JohnfreakedBillout.c.Particlemarksaspect;particleisanindependentsemanticconstituent,butnotaverbalargument.Johnatethesandwichup.Johnreadthebookthrough.Thebandplayedon.d.Time-awayconstruction;particlemarksconstructioninwhichverbissemanticallysubordinate;NPisanargumentoftheconstructionbutnotanargumentoftheverb.We’retwistin’thenightaway.(≈‘We’respendingthenighttwistin’)e.ParticleformsanidiomwiththeNP;NPandParticleformasemanticconstituentfunctioningasdegreeadverbial.Johnsanghisheartout.Johnprogrammedhisbuttoff.

Becauseoftheheterogeneityofmeanings,inparticularthedifferentcombinationsofelementsthatformtheconstruction(V+Prtin(14b),NP+Prtin(14e)),thereisnowaytocombinetheseallunderasingleabstractform-meaningpairing.Thus,inahomogeneousConstructionGrammar,eachoftheseconstructionshastobestipulatedseparately,anditisacoincidencethattheyallhappentoconvergeontheformV-NP-Prt.

AheterogeneousConstructionGrammar,incontrast,cansaythatEnglishhasaVerb-Particleconstructionwithcertainsyntacticproperties,forexample,theparticlecanoccurbeforeorafterthedirectobject,butitmustappearafterapronominalobjectandbeforefullPPs.Inturn,thissyntacticconfigurationcanbeusedtoexpressfive

5

VP

VP

6

Page 73: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 9 of 17

independentsemanticconfigurations.Mostlikely,the‘basic’semanticconfigurationis(14a),inwhichtheparticleisagenuineargumentoftheverb,transparentlyrelatedtodirectionalprepositions.Butthesamesyntaxhasbeenutilizedbytheotherfourconstructions,eachinadifferentway.

Inotherwords,allthedifferentmeaningsassociatedwith[ VNPPrt]areindeeddaughtersofacommonnodeinaninheritancehierarchy.However,thiscommonnodeisthepuresyntacticstructureitself:thisispreciselywhattheyallhaveincommon.Andsincenodesintheinheritancehierarchyarethemselveslexicalitems,thepurelysyntacticstructureisalsoalexicalitem.Inturn,thissyntacticstructureinheritspropertiesfromthemoreabstractitem[ V…],whichexpressesthepurelysyntacticfactthatEnglishhasasyntacticcategoryVPthatisverb-initial.

Intheexamplessofar,asinglesyntacticstructuremapsintomultipleunrelatedmeanings.Buttherearealsocaseswherethemappingbetweensyntaxandsemanticsismany-to-many.Forinstance,manyofthesamesemanticrelationscanbeexpressedbyeitherNofNP,NP'sN,oranoun-noun(NN)compound.Example(15)offersasample.

(15)

NofNP NP'sN NNcompound

Part legofachair thechair'sleg chairleg

Unit grainofrice *rice'sgrain ricegrain

Aggregation pileofsand *sand'spile sandpile

Agent attackofthezombies thezombies’attack zombieattack

AhomogeneousConstructionGrammarapproachforcesustosaythatEnglishhastendistinctconstructionsthathappentorelatethesamefoursemanticrelationstothesamethreesyntacticstructures.Thereisnosinglemostabstractform-functionschemaunderwhichtheyallfall.Thisleavesitapparentlyanaccidentthatthesyntacticstructuresarethesameacrossallthecases.Incontrast,aheterogeneousConstructionGrammarallowsforalexiconthatincludesthreeautonomous(form-only)syntacticschemasandfourautonomous(function-only)semanticschemas.Eachoftheparticularconstructionsin(15)theninheritsitsformfromoneofthesyntacticschemasanditsfunctionfromoneofthesemanticschemas.Inotherwords,(15)isarathergoodrepresentationofthemultipleinheritancestructureoftheseconstructions.Hence,thegrammargainsgeneralitythroughadmittingpurelysyntacticconstructions.

OnemightcontendthatthereisacommonelementofmeaningamongtheinstancesofNofNPin(15),namelythattheNissemanticheadandthecomplementissomesortofargumentormodifier.However,afurtherNofNPconstruction,illustratedin(16),reversesthisrelation:thecomplementofofisthesemantichead,andthesyntacticheadservesasasemanticmodifier(Asaka2002;Booij2002).

(16)a. thatscoundrelofadoctor(=‘thatdoctor,whoisascoundrel’)b.thattravestyofatheory(=‘thattheory,whichisatravesty’)c.agemofaplumber(=‘aplumberwhoisagem’)

Thesyntacticstructureisthegarden-varietyNofNP,andwewouldlikethegrammartobeabletoexpressthisgeneralization.ButahomogeneousConstructionGrammarapproachdoesnotletussaythis,sincethesemanticsoftheconstructionisentirelydifferentfromotherusesofNofNP.Again,aheterogeneousCxGletsusseetheconstructionin(16)aspouringanewmeaningrelationintoanoldform.

AheterogeneousConstructionGrammaralsoallowsafurtherpossibility:theremaybeautonomoussemanticconstructionsthatarenotlinkeddirectlytosyntax.Thisisonewayofunderstandingprinciplesof‘coercion,’whichpermit(partly)conventionalizedalternativeinterpretationsofphrases,withnosyntacticreflex.Atypicalcaseis

VP

VP

Page 74: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 10 of 17

‘referencetransfer,’illustratedin(17).

(17)a.[Onewaitresstoanother:]Thehamsandwichinthecornerwantssomecoffee.(hamsandwich=‘thepersonwhoordered/whoiseatingahamsandwich’)b.Platoisupthereontheshelf,nexttoChomsky.(Plato=‘bookbyPlato/bustofPlato’)c.I’mparkedoutback.Igothitinthefender.(I=‘mycar’)

IntheParallelArchitecture,referencetransferistreatedasasemanticoperatorthathappenstohavenorealizationinsyntax.Sincereferencetransfersareconventionalized,theyhavetobelistedinthelexiconasconstructionsapproximatelyalongthelinesof(18)(Jackendoff1991,1997a).

(18)Syntax:NPSemantics:a.[ PERSONCONTEXTUALLYASSOCIATEDWITH[ x] ]b.[ REPRESENTATIONOF[ x] ]c.[ BOOKWRITTENBY[ x] ]d.[ VEHICLEBELONGINGTO[ x] ]

TheeffectoftheseconstructionsistopermitanNPtoreferatthesametimetoitsnormalreferent(thesandwich)andtotheassociatedindividual(thepersonwiththesandwich),withnosyntacticmarkingatall.Fromthepointofviewofthespeaker,theseextrapiecesofmeaningarethingsthatoneneednotsayovertly;fromthepointofviewofthehearer,thesearethingsthatonemayhavetoaddintoone'sinterpretationtorecoverthespeaker'sintendedmeaning.Thatis,liketheWayconstruction,thesecoercionsaffordeconomyofphonologicalandsyntacticexpression.

InahomogeneousConstructionGrammar,coercionsmustbetreatedasanothersortofphenomenonaltogether,sincetheyarenotSaussureansigns—theyhavenoform.Atbesttheyareoperatorsthatchangeonesignintoanother(seeMichaelis,thisvolumeforoneaccount),orpurelysemanticprocessesthathavenothingtodowithconstructions.ButinaheterogeneousConstructionGrammar,theyarejustanothersortofstoredstructure.

ThispartialindependenceofcomponentsischaracteristicoftheParallelArchitecture.ItmakesitpossibletoincorporatetheinsightsofConstructionGrammarwithoutcommittingtoalockstepcorrespondencebetweensemanticandsyntacticcombinatoriality.

5.5.Productivityvs.SemiproductivityinConstructions

Aslaidoutinsection5.2,theParallelArchitectureconceivesofthelexiconasthelong-termmemorystoreofpiecesoflinguisticstructure:words,idioms,affixes,phrasestructurerules,abstractlinkingrules,meaningfulconstructions,andsoon.Therefore,acentralquestionis:Whatpiecesofstructuremustbestoredinthelexicon,andwhatpiecescanbebuiltfromstoredpartsbyonlineprocessing?Noticethedifferenceinmodalitybetweenthetwopartsofthequestion.Anythingcanbestoredinmemory;afterall,peoplememorizethewholeKoran.Sotheissueiswhatmustbestoredversuswhatneednotbe.

ThisquestionhasanimportantbearingonacrucialissueforConstructionGrammarandalllexicalisttheories:thenatureofproductivityandthestatusofinheritancehierarchies.IfinditusefultomakeadistinctionbetweenwhatIwillcall‘fullproductivity’and‘semiproductivity.’Kay(thisvolume)addressesthesamedistinctioninadifferentway,whichIwilldiscussbelow.

AtypicalfullyproductiveconstructionisthephrasestructureoftheEnglishVP:onecanlearnnewtransitiveverbsandcomposethemwithoutlimitwithnoveldirectobjectsandnoveladjuncts.Withinmorphology,atypicalfullyproductiveconstructionistheEnglishregularplural.Thiscanbeappliedtonewnounsonehasneverheardbefore,andtherearenosurprises.Theclassicillustrationofspeakers’knowledgeofitsfullproductivityisthe‘wugstest’(Berko1958),inwhich6-year-oldchildrenreliablyproducepluralsfornonsensenouns(and4-and5-year-oldchildrendososomewhatlesscomprehensively).Anotherwell-knownfullyproductivemorphologicalconstructionisEnglishexpletiveinfixation(McCarthy1982),which,givenproperprosodicconstraints,isperfectly

1

Thing Thing 1 1

Thing Thing 1 1

Thing Thing 1 1

Thing Thing 1 1

7

8

Page 75: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 11 of 17

reliableinlicensingformssuchaskanga-fuckin-roo.

Forfullyproductiveconstructions,then,itisunnecessarytostoreinstances,sincetheycanbebuiltonline.Atthesametime,thisdoesnotprecludestoringparticularinstances,forexample,VPidiomssuchaskickthebucket,pluralslackingasingularsuchasscissors,and(forlinguists)well-knownexemplarssuchasmanu-fuckin-facturer.Thesewillfallunderthefullyproductiveconstructioninaninheritancehierarchy.

Experimentalresearchhasfurthershownthathigh-frequencyderivedinstancesoffullyproductiveschemasmaybestored(Baayen,Levelt,andHaveman1993;Pinker1999;Nooteboom,Weerman,andWijnen2002).Moreover,sincethelearnerhastoinducetheexistenceandformofaproductiveconstructionfromheardandstoredinstances,anumberofstoredinstancesofsuchconstructionsarelikelytoremainafteracquisition,atleastforsomeperiodoftime.

Afullyproductiveconstructioncanbesubjecttolistedexceptions,whichmaybeeitherinstancesormorespecializedconstructions.Inthiscase,thefullyproductiveconstructionfunctionsasadefault.Thisis,ofcourse,thestatusoftheEnglishregularplural,whichhasvariousexceptions,bothidiosyncraticinstanceslikewomenandsemiproductiveclasseslikesheep/deer/fish.

Theoverallpoint,however,isthattheinheritancehierarchyforafullyproductiveconstructionis‘open’;itdoesnotexhausttheformsthatspeakersarereadytoproduceandunderstandondemand.

Semiproductivephenomenatoodisplayaregularitythatcanbestatedasaruleorconstruction.However,acceptableinstancesmustbelearnedandstoredindividually.Oftenspeakershavetheintuitionthatanewformsoundsunusual(acoinage,inKay'sterms);theymayobservethattheyhaveneverhearditbefore,and/ortheymayfinditamusing.Inotherwords,theinheritancehierarchyforsemiproductiveconstructionsis(relatively)‘closed.’

AtypicalcaseofsemiproductivityistheclassofEnglishdenominalverbssuchasbutter(thebread),pocket(themoney),weed(thegarden),hammer(anail),father(achild),mother(one'sstudents),waitress(inarestaurant),andsoon,pluscasesusingaffixessuchasdisrobe,de-claw(acat)andbehead(aperson).Alongsidebutter(thebread),wemightexpecttofind*mustard(thebread),butwedonot(except,ifGooglecanbetrusted,veryrarelyinthecontextofrecipes);alongsidecarpetthefloorwemightexpect*linoleumthefloor;alongsidepocketthemoneywemightexpect*envelopetheletter(whereenvelopmeanssomethingratherdifferentsynchronically).

Anothercaseofsemiproductivity,wheremultipleformscompeteforthesamegrammaticalslot,istheclassofEnglishdeverbalnounssuchascontribution,destruction,appraisal,andharassment.Tosomeextenttheformofthenominalisphonologicallypredictable;forexample,verbsendingin-ateinvariablyhaveanominalin-ation.Butnoteverycaseispredictable.Recitehasthetwonominalsrecitalandrecitation,withdifferentmeanings.Thephonologicallyparallelincitehasincitementratherthan*incitalor*incitation,andexcitehasexcitationandexcitementbutnot*excital.

Inaddition,themeaningsoftheexistingformsarenotentirelypredictable.Forinstance,arecitationisonlyvaguelyrelatedtoreciting,andexcitation,unlikeexcite,canpertaintoelectronsbutnottopeople.Anddespitethesemanticsimilarityofthenounsfatherandmother,thecorrespondingverbshaveradicallydifferentmeanings(Imightmothermystudents,butAdelecouldn’tpossiblyfatherhers).ManyotherexamplesappearinBooij(thisvolume).

Withinasemiproductivepattern,theremaybesubconstructionswithfullproductivity.Forinstance,theEnglishdenominalverbconstructionhasasubconstructionthatmeans‘fastenwithN,’suchasnail,screw,tape,andglue.Thiscaseseemsfullyproductive,inthatthenewfastenervelcro,unlikemustard,occasionsnohesitationinformingarelatedverb.Anotherfullyproductivesubconstructionis‘communicateviaN,’whichhasgivenusthenewverbsemail,text(-message),andskype.Thereisnoevidentmorphologicalorsemanticreasonforonesubclasstobefullyproductiveandthenextsemiproductive.Ifanything,thereasonsarehistorical,sotheyhavenobearingonthesynchronicgrammar.

Thedistinctionbetweenfullproductivityandsemiproductivityhasalwaysbeenastapleofmorphology.Whathasbeenlessnoticedisthatthesamedistinctionappearsinphrasalconstructions.Forinstance,theWayconstruction,theTime-awayconstruction,andtheSound+Motionconstructionarefullyproductivewithintheconstraints

Page 76: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 12 of 17

imposedbytheirselectionalrestrictions.Oneneednotstoreinstancesoftheseconstructionsforeachverbthatoccursinthem—onlyfortheidiomaticones,suchaswendandwormintheWayconstructionandfritterintheTime-awayconstruction.Ifweinventanewverbofsoundproduction,saygrimble,weautomaticallyknowthatthesentencesin(19)arepossible,andweknowexactlywhattheymean(Zwicky1971).

(19)a.Wayconstruction:Joegrimbledhiswayalongtheriver.(=‘Joewentalongtherivermakinggrimblingsounds’)b.Time-awayconstruction:Joegrimbledtheafternoonaway.(=‘Joespenttheafternoonmakinggrimblingsounds’)c.Sound+Motionconstruction:Thecargrimbledoverthebridge.(=‘thecarmadegrimblingsoundsasaresultofitsmovingoverthebridge’)

However,someconstructionsaresemiproductive.OnecaseisSluice-stranding(Culicover1999;CulicoverandJackendoff2005).InordinarySluicing,whichisfullyproductive,awh-phraseappearsinplaceofafullindirectquestion(20a),andthiswh-phrasecanincludeapied-pipedpreposition(20b).InSluice-stranding,however,aprepositionappearsafterthewh-phrase,asthoughitwereleftoverfromtheelliptedclause(20c).

(20)a.Johnleftthemeeting,butIdon’tknowwhen.[Sluicing]b.Johnleftthemeeting,butIdon’tknowwithwhom.[Sluicingwithpied-pipedP]c.Johnleftthemeeting,butIdon’tknowwhowith.[Sluice-stranding]

InordinarySluicingwithpied-piping,anycombinationofwh-wordandpied-pipableprepositionispossible,consistentwiththesemantics;thecombinationsneednotbestored.However,inSluice-stranding,itturnsoutthatonlycertainwh-wordsarepossible,andeachonepermitsadifferentcollectionofprepositions.

(21)…butIdon’tknowa.whowith/to/from/for/*on/*nextto/*about/*besideb.whatwith/for/from/of/on/in/about/at/*before/*into/*near/*besidec.howmuchfor/*by/*with(notealso*howmanyfor)d.whereto/from/*neare.*which(book)with/to/from/nextto/about/beside

CulicoverandJackendoffconcludethattheacceptablecasesofSluice-strandingmustbelearnedindividually,despitetheirconformingtoacommonconstructionaltemplate.Hence,theconstructionissemiproductive.

AmorecomplexcaseofasemiproductiveconstructionistheNPNconstruction(Jackendoff2008b).Thisstructureappearsinanumberofidiomssuchashandoverfist,handinglove,tongueincheek,andtitfortat.Itappearsmoreproductivelywithfiveprepositions,allofwhichrequirethetwonounstobeidentical:

(22)a.daybyday,paragraphbyparagraph,countrybycountryb.dollarfordollar,studentforstudent,pointforpointc.facetoface,bumpertobumperd.termpaperaftertermpaper,pictureafterpicturee.book(up)onbook,argument(up)onargument

Eachofthesehasitsownidiosyncrasiesofmeaning(whosedetailsIwillnotgointohere).Syntactically,theycanallbeusedassententialadjuncts(23a)andasprenominalmodifiers(23b).However,NafterNandN(up)onNaremoreflexiblethantheotherthree,inthattheycanalsobeusedinNPpositions(23c).

(23)a.Westudiedtheworldcountrybycountry.Studentforstudent,wehavethebestrecordinthestate.

Page 77: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 13 of 17

b.aday-by-dayimprovementaface-to-facemeetingc.Studentafter/*bystudentturnedinterribletermpapers.Shepiledupargumentupon/*forargumentagainstmyposition.

By,for,after,anduponarefullyproductiveastothechoiceofnoun,withinthesemanticconstraintsoftheconstruction.ButatleastonesenseofNtoN,thesensethatdenotesjuxtapositionofpartsoftwoobjects,isonlysemiproductive.Twopeoplecanstandbacktobackbutnotfronttofront—rathertheyarefacetoface.Handtohandoccursmostlyonlyinthecontextofcombat;cheektocheekinthecontextofdancing;eyetoeyeinthecontextofseeingeyetoeye.Thejuxtapositionmeaningofsidetosideisusuallyexpressedinsteadbytheidiomaticsidebyside;and,unexpectedly,sidetosidecanalsodenoteoscillatorymotion,paralleltobackandforth,upanddown,androundandround.

Inshort,theNPNconstructionasawholeissemiproductive:itconsistsofalistofidiomsplusthefivesubconstructionsillustratedin(22),eachofwhichhasitsownsemanticandsyntacticpeculiarities.Inturn,fourofthefivesubconstructionsarefullyproductive,buttheonewithtoissemiproductive.(24)isapartialinheritancehierarchyfortheconstruction.Thefullyproductivesubconstructions,whoseinstancesneednotbestoredinmemory,arenotatedinboldface.

Tosumup,bothfullyproductiveandsemiproductiveconstructionsarefoundinbothmorphosyntaxandphrasalsyntax.Bothsortsofconstructionscanbeencodedinexactlythesameform:expressionsconsistingofsomecombinationofvariablesandconstants.Bothtypesofconstructionscanstoreinstancesasdaughtersinaninheritancehierarchy.Fullyproductiverulesaredifferentonlyinthatanindefinitenumberoffurtherinstancescanalsobecreatedonlinewithoutanyeffortorsenseof‘coining’andthereforeneednotbestoredinmemory.

Howshouldthegrammarencodethedifferencebetweenfullyproductiveandsemiproductiveconstructions?Thatis,whatdoesthedistinctionbetweenboldandordinarytypein(24)mean?Thesimilaritiesbetweenthetwotypesofconstructionssuggestthatweshouldnotsegregatethemformally,forinstance,byputtingfullyproductiveconstructionsinthegrammarbutsemiproductiveconstructionsina‘meta-grammar,’asurgedbyKay(thisvolume).Rather,wewantanaccountthatsaysthetwokindsofrulesareformallythesame,exceptforsomethingthatmarkswhethertheyarefullyproductiveornot.

One'sfirstimpulsemightbetolocalizethedifferenceasadiacriticonconstructions,ineffecttakingthedifferenceshowninboldin(24)astheoreticallysignificant.However,abettersolutionistomarkthedistinctionwithadiacriticonthevariableintheconstruction.Thereasonisthatthereexistconstructionsthatarefullyproductiveononevariableandsemiproductiveonanothervariable.ConsiderthefourconstructionsofEnglishthatyieldnamesofgeographicalfeatures.

(25)a.Beaver/WissahickonCreek(alsolake,bay,island,pond,mountain,hill,street)b.LakeMichigan/Superior/Geneva(alsomount)c.theAtlantic/Pacific/ArcticOcean(alsosea,river,desert,fault,turnpike)d.theBayofFundy/Biscay(alsogulf,isle)

Thevariableforthenameisfullyproductive:ifyouarenamingacreekafterDickCheney,youknowitwillbeCheneyCreek(orDickCreek).Butyouhavetolearnwhichpatternaparticulartypeofgeographicalfeature(creek,crater,col,ridge,etc.)fitsinto,andthereforethisvariableissemiproductive(though(25a)mightbea

9

Page 78: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 14 of 17

default). Example(26)showshowthefourpatternsmightbeencoded,notatingfullyproductivevariablesinbold.Theirsharedsemanticsplacesthemincompetitionwitheachother.

(26)Syntax:a.[ X Y ]b.[ Y X ]c.[ theX Y ]d.[ theY [ ofX ]]

Semantics:

GEOGRAPHICAL-FEATURE NAMEDX

Inotherwords,itisnotconstructionspersethatarefullyproductiveorsemiproductive;itisthevariablesintheconstructions.

Thereisafurtherproblem.Childrenacquiringalanguagehavetopickouttheregularitiesintheinputandtherebyaddconstructionswithvariablestotheirlexicons.Buthowdoesachilddiscoverwhetheraparticularvariableisfullyproductiveoronlysemiproductive?Itisnotgiveninthedata.Howisitthateveryonecomesupwithessentiallythesameanswer?

Inmorphology,sometimesthereiscompetitionbetweenregularities,forinstance,intheEnglishpasttense,betweenthe-daffix,ablaut,zeromarking,andendingin-ought(caught,bought,etc.).Atmostoneofthesecanbeafullyproductivedefault,sotheothersmustbetreatedassemiproductive.Onemightguessthatfullyproductivecombinationsaregenerallymorefrequentinprimarylinguisticdatathansemiproductivecombinations.Forinstance,thisispresumablyaclueforthechildthatthe-dpastisthedefaultform.

However,notallsemiproductivephenomenaareincompetitionwithadefault.Forinstance,NtoNformssuchasfacetofaceandzerodenominalverbssuchasbutterarenotincompetitionwithanything,yettheyarestillsemiproductive.Hence,therelativefrequencyofcompetingformscannotbetheonlyfactordistinguishingfullyproductivefromsemiproductiveconstructionsinacquisition.

Norarefullyproductiveconstructionsalwaysfrequent.Considertheconstructionillustratedin(27).

(27)(all)X-edout=‘exhaustedfromexperiencingXtoexcess’a.I’mOlympic’dout.[recordedinconversationwithsomeonewhoworkedattheOlympics]b.He'sallknittedout.[afterknittingforthreedayssolid]c.I’mBogartedout.[afterwatchingaHumphreyBogartmarathon]

Thisconstruction,presumablyacalqueontiredout/wornout,freelyallowseithernounsorverbsinthevariableposition,andoneclearlydoesnotlearnallpossibleinstances.Hence,theconstructionisfullyproductive.Yetitisexceedinglyinfrequent—Iwouldguessonehearsitperhapsthreetimesayearinconversation.Howdolanguagelearnerscometotheconclusionthatitisfullyproductiveonthebasisofsolittleevidence?

Itisplausiblethatchildrenobservingaregularityinitiallyencodeitassemiproductive,andlaterupgradeittofullyproductiveifevidencewarrants.Thiswouldcomportwiththeevidencethatonthewholechildrenarecautiousaboutextendinggeneralizationstonewitems,butdoeventually‘goproductive,’goingbeyonditem-basedlearning(Tomasello2003).Ontheotherhand,itisnotclearhowsuchanaccountappliesto(27),forwhichthereissuchskimpyinput.Moreover,childrenareknowntoovergeneralizesemiproductiveconstructions(e.g.,Bowerman's(1982a)exampleMommy,giggleme,overgeneralizingtheEnglishcausative),soitisnotclearthatthereisauniformone-waymovefromsemiproductivetoproductive.

Ideally,wewouldlikeanaccountinwhichchildrendonothaveto‘decide’whetheraphenomenonisproductiveornot.Rather,wewouldhopethatthecourseofacquisitionwouldfollowfromthedistributionofthedatainspeechandindependentpropertiesofhowthebrainlearns.

Thisisnottheplacetosolvetheseproblems,onlytolaythemonthetable.Insofarasproductivityisanissuenotonlyforphonologyandmorphologybutalsophrasalsyntaxandthetheoryofconstructions,inmyopinionitmust

9

ProperN 1 2

ProperN 2 1

NP 1 2

NP 2 PP 1

2 1

Page 79: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 15 of 17

takeaplaceasoneofthecentralissuesoflinguistictheoryforthecomingyears(asemphasizedalsobyBarðdal2008).

5.6.Summary

ThemotivatinginsightofConstructionGrammaristhattherelationofsyntaxtosemanticsisestablishednotmerelythroughword-by-wordcompositionbutalsothroughalargecollectionofconstructions,inwhichthemeaningofthewholegoesbeyondthemeaningsofthewords,andinwhichsyntacticstructurecanitselfcarryidiosyncraticmeaning.ThisinsightfindsacomfortablehomeintheParallelArchitecture,inwhichphonology,syntax,andsemanticsareeachautonomouscombinatorialcomponents,linkedbyinterfacecomponents.WithintheParallelArchitecture,theconstructionsofConstructionGrammarareparticularsortsofinterfacerules.Bothapproachesemphasizethecontributionofmeaningtosyntacticform;bothapproachesargueforthecontinuitybetweenwords,idioms,morphologicalaffixes,andmeaningfulsyntacticconstructions;bothapproachesconceiveofthelexiconasorganizedintermsofinheritancehierarchies.

IhavetriedtoshowherethattherelativelyformalperspectiveoftheParallelArchitecturehelpssharpentwoissueswithinConstructionGrammar.First,anumberofconsiderationsfromtheParallelArchitecture,somewithintheimmediateambitofthesyntax-semanticsmappingandsomefromfurtherafieldinlanguageandcognition,arguethataheterogeneousConstructionGrammarismorepromisingthanahomogeneousConstructionGrammar.Second,theParallelArchitecture'spsycholinguisticallybasedcriterionforinclusioninthelexicon,namelybeingstoredinlong-termmemory,helpsclarifytheissueofproductivity.Thedistinctionbetweenfullproductivityandsemiproductivityisnotduetothesephenomenabeingindifferentpartsofthegrammar,norisitduetosomedifferenceintheoverallformoftheconstructionsinquestion.Rather,thedistinctionappearstobelocalizedinadiacriticonthevariableswithintheconstruction.Itistobehopedthattheclosecompatibilityofthetwoframeworksleadstofurtherusefulinterchange.

ArgumentStructureconstructions

Wayconstruction(belchhiswaydowntheroad)(Jackendoff1990,1997b)With-Theme(butterthebreadwithmargarine)(Jackendoff1990)RecipientNP(throwBilltheball)(Jackendoff1990)BeneficiaryNP(bakeBillacake)(Jackendoff1990)Resultative(hammerthemetalflat)(Jackendoff1990;GoldbergandJackendoff2004)Time-away(drinktheafternoonaway)(Jackendoff1997b)Sound+Motion(rumblearoundthecorner)(GoldbergandJackendoff2004)Disappearance+Motion(vanishintothewoodwork)(GoldbergandJackendoff2004)Verb-Particleconstructions(Jackendoff2002b)

Phrasalconstructions

X+else(someoneelse,anythingelse)(CulicoverandJackendoff1995)theNE(thesoundp )(Jackendoff1984)FocusReduplication(youmakethesalad-salad)(Ghomeshietal.2004)NPN(dayafterday)(Jackendoff2008b)Nouncompounds(doghouse)(Jackendoff2010)Nomatter(Culicover1999)

Sententialconstructions

NPandS(OnemorebeerandI’mleaving)(Culicover1972)Implicationaland(youcoughandI’llhityou)(CulicoverandJackendoff1997)ComparativeCorrelative(themoreS,themoreS)(CulicoverandJackendoff1999)Not-topics(Notinmycaryouwon’t)(Culicover1999)

Ellipsisconstructions(allinCulicoverandJackendoff2005,CulicoverandJackendoff2012)

one-anaphora

h

Page 80: The Oxford Handbook of Construction Grammar

Constructions in the Parallel Architecture

Page 16 of 17

doit/soanaphoraBareArgumentEllipsisSluicingGappingVPellipsisPseudogappingComparativeellipsisSluice-stranding(Iwonderwhowith)(Culicover1999;CulicoverandJackendoff2005)

MuchofthischapterisadaptedfromJackendoff2010(chapter1andremarksprecedingchapter7).IamgratefultoAdeleGoldberg,LauraMichaelis,andJóhannaBarðdalforusefulcommentsonearlierversions.

Notes:

(1.)DetailedcomparisonsoftheParallelArchitecturewithMainstreamGenerativeGrammarappearinJackendoff(1997a,2002a(especiallychapters5–8),2007,2008a,2011);CulicoverandJackendoff(2005);PinkerandJackendoff(2005).Thesereferencesalsoofferargumentsthatsomeaspectsofthelanguagefacultyaretheconsequenceoffaculty-specificcognitivespecializations,incontrastwiththeconsensusamongConstructionGrammarians(Goldberg,thisvolume)thatallcrosslinguisticgeneralizationscanbeexplainedbydomain-generalcognitiveprocesses.

(2.)Interestingly,Head-DrivenPhrase-StructureGrammar(HPSG)groupsthesecomponentsdifferently:syntaxandsemanticstogetherformsynsem,whichistakentobethe‘function’partofasign,inoppositiontophonology,whichconstitutes‘form.’

(3.)Iglossovertheissueofhowtheaffix'sallomorphyisencoded.

(4.)Theycanhoweverbeparatacticallycombined,asinYes,youmaygoorMichaelis'sexample(thisvolume),Damn,I’mtired.Theycanalsoappearinquotativecontextssuchas“Yes,”shesaidandwordssuchas“yes.”Suchcontextshavenosyntacticrestrictionsandsodonotdependonsyntacticfeaturesfortheirinstantiation.

(5.)ThisexamplefurthermoreusesthesyntaxoftheDitransitiveconstructiontoexpresssomethingnormallycouchedasasimpletransitive:AmygaveTomakissparaphrasesAmykissedTom.

(6.)SeeJackendoff(2002b,2010:249notef.).Notealsothatin(14d,e)thesyntacticdirectobjectisnotanargumentoftheverb.

(7.)IamunawareofevidenceinfavorofhomogeneousConstructionGrammar.Rather,theassumptionofhomogeneityseemstobeheldoverfromtherootsoftheseparticularversionsofConstructionGrammarinCognitiveGrammar.InbothCognitiveGrammarandConstructionGrammar,ithasbeencorrectlyarguedthatsome—evenmany—syntacticstructuresbearmeaning.Butforthereasonsadducedinthissection,Ibelieveitisinappropriatetoextendtheexistentialquantifiertoauniversal.

(8.)Barðdal(2008)regimentsalargenumberofsensesoftheterm‘productivity’intheliterature,butdoesnotappeartopickouttheexactcontrastIamconcernedwithhere.Afulldiscussionofwhereherapproachandmineoverlapandwheretheydivergeisbeyondthescopeofthischapter.However,her‘clineofproductivity,’whichdependsontypefrequencyanddegreeofcoherence,appearstometobeacharacteristicofwhatIamcallingsemiproductivity;itissemiproductiveconstructionsinmysensethatcanbe‘moreorlessproductive’inhersense.

(9.)Afurthercomplicationisthatlakeandbayappearintwodifferentpatterns,soonehastolearninstancesoflakeandbayonebyone.

RayJackendoffRayJackendoffisSethMerrinProfessorofPhilosophyandCo-DirectoroftheCenterforCognitiveStudiesatTuftsUniversity.Hewasthe2003recipientoftheJeanNicodPrizeinCognitivePhilosophyandhasbeenPresidentofboththeLinguisticSocietyofAmericaandtheSocietyforPhilosophyandPsychology.HismostrecentbooksareFoundationsofLanguage(Oxford,2002),SimplerSyntax(withPeterCulicover,Oxford,2005),Language,Consciousness,Culture(MITPress,2007),MeaningandtheLexicon(Oxford,2010),andAUser’sGuidetoThoughtandMeaning(Oxford,2011).

Page 81: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 1 of 11

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0006

DatainConstructionGrammarStefanTh.GriesTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterexaminesthetypesofdatausedinconstructionistapproachesandtheparametersalongwhichdatatypescanbeclassified.Itdiscussesdifferentkindsofquantitativeobservational/corpusdata(frequencies,probabilities,associationmeasures)andtheirstatisticalanalysis.Inaddition,itprovidesasurveyofavarietyofdifferentexperimentaldata(novelword/constructionlearning,priming,sorting,etc.).Finally,thechapterdiscussescomputational-linguistic/machine-learningmethodsaswellasnewdirectionsforthedevelopmentofnewdataandmethodsinConstructionGrammar.

Keywords:judgementdata,observationaldata,experimentaldata,computationalapproaches,statistics

6.1.Introduction

Overthelastapproximatelytwenty-fivetothirtyyears,anewfamilyoflinguistictheorieshasestablisheditselfasapowerfulalternativetothethendominantgenerativeapproachtolanguageingeneralandgrammarinparticular,thefamilyofConstructionGrammars.ManyofthesetheoriessharemostoftheirassumptionswiththeapproachofCognitiveLinguistics,ofwhichsomeare,insomesense,thegrammaticalpart.However,thecommonalitiesdonotendthere,andIwanttopointouttwoadditionalonesthatbearonthedataandmethodology.

First,justlikesemanticcategoriesaredescribedinCognitiveLinguisticsasradialcategories—categorieswhosemembersmaybelinkednotdirectlybutviafamilyresemblances—differentinstantiationsofConstructionGrammarconstitutearadialcategoryofthesamekind.TheseConstructionGrammarsincludeGoldbergandLakoff'sCognitiveConstructionGrammar(cf.Boas,thisvolume),Bergen'sEmbodiedConstructionGrammar(cf.BergenandChang,thisvolume),Croft'sRadicalConstructionGrammar(cf.Croft,thisvolume),andothers.

Thus,althoughConstructionGrammarssharemanyassumptions,theyalsodifferinvariousways.However,andthisisthesecondpoint,giventheaffinitytoCognitiveLinguistics,many(inparticularmanyConstructionGrammarssuchasCognitiveConstructionGrammar,EmbodiedConstructionGrammar),butnotall(e.g.,BerkeleyConstructionGrammarorSign-BasedConstructionGrammar),explicitlycommittowhatLakoff(1990:40)hasreferredtoasthecognitivecommitment,namely“acommitmenttoprovidingacharacterizationofgeneralprinciplesforlanguagethataccordswithwhatisknownaboutthemindandbrainfromotherdisciplines.”Giventhiscommitment,itisnotsurprisingthat,empiricallyspeaking,ConstructionGrammarasafamilyofcloselyrelatedgrammarsisprobablyoneofthemethodologicallymostpluralisticfields,asitutilizesalargenumberofdifferentdataandmethodologies.

Averywidespreadclassificationoftypesoflinguisticdatahasbeentodistinguishintrospective,observational,andexperimentaldata,whereintrospectivedataresultfromaspeaker'ssecond-levelattentionorsecond-levelconsciousness(touseTalmy's2007terminology),whereobservationaldataresultfromrecordingsorcorporaoflinguisticproductionin(often)noisynaturalisticsettings,andwhereexperimentaldataresultfromsubjects’

Page 82: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 2 of 11

behaviorindesignedcontrolledexperimentalsituationsfacingcarefullydevelopedexperimentalstimuli.However,eventhoughthisclassificationiswidelyused,itisalsoabittoosimplistic,sinceavarietyofdata-gatheringtypesexhibitscharacteristicsofmorethanoneofthesegroups.Inanattempttodevelopamorefine-grainedclassification,GilquinandGries(2009)developacontinuumofkindsofobservationalandexperimentallinguisticdata.While,forthesakeofsimplicity,theypresenttheircontinuumasone-dimensional,itactuallyintegratesseveraldifferentdimensions:

•Hownaturaldoesthesubjectperceivehis(experimental)setting?

−mostnatural,e.g.,speakerswhoknoweachothertalktoeachotherinunpromptedauthenticdialog;

−intermediatelynatural,e.g.,aspeakerdescribespictureshandedtohimbyanexperimenter;

−leastnatural,e.g.,speakerliesinanfMRIunitundergoingabrainactivityscanwhilehavingtopressofoneofthreebuttonsinresponsestodigitallypresentedblack-and-whitepictorialstimuli.

•What(linguistic)stimulusdoesthesubjectacton?

−mostnatural,e.g.,speakersarepresentedwithnaturalutterancesandturnsinauthenticdialog;

−intermediatelynatural,e.g.,speakersarepresentedisolatedwordsbyanexperimenterinanassociationtask;

−leastnatural,e.g.,speakersarepresentedwithisolatedvowelphones.

•What(linguistic)units/responsesdoesthesubjectproduce?

−mostnatural,e.g.,subjectsproducenaturalandunconstrainedresponsestoquestions;

−intermediatelynatural,e.g.,speakersrespondwithisolatedwords(e.g.,toadefinition);

−leastnatural,e.g.,speakersrespondwithaphoneoutofcontext.

Forexample,asituationinwhichasubjectsitsinfrontofacomputerscreenwithaneyetracker,isauditorilypresentedwithaword,isvisuallypresentedwithtwopictures(oneofwhichrepresentsaninstanceofwhattheauditorilypresentedwordmeans),andrespondsbysaying“yes”or“no”tothequestionwhetherheseesaninstanceofwhatthewordrefersto,canbeclassifiedas

−afairlyunnaturalexperimentalsetting:sittinginfrontofaneye-tracker;

−anintermediatelynaturalstimulus:isolatedwordsandpictureinput;

−anintermediatelynaturalresponse:anisolated“yes”or“no”.

Asanotherexample,considerthecaseofanalyzingathere-constructionincorpusdatafromauthenticconversations,whichcouldbeclassifiedas

−averynatural(experimental)setting:anauthenticconversation;

−averynaturalstimulus:thepreviousturn;

−averynaturalresponse:aspeaker's(response)turn.

Whiletheaboveclassificationbymeansofthethreedimensionsisneithercompletelyexhaustivenoruncontroversial,itallowsforaheuristicallyvaluableclassificationofmostempiricalapproachesinConstructionGrammarinparticularandprobablyinlinguisticsingeneral. Aswillbeshown,linguistsworkinginConstructionGrammarhaveuseddataandmethodsfromallsixextremesofthesethreedimensions.Infact,theamountofrigorousempiricalworkinthefieldisveryhigh,asisindicatedbythefactthat,forinstance,thevastmajorityofstudiesinConstructionsandinthefirstvolumeofthenewjournalConstructionsandFramesusecorpusdata.

Inthefollowingsections,Iwilldiscussarangeofdataandmethodswithaneyetoexemplifyinghowdifferentmethodshavegivenrisetodifferentdata,andhowthesehaveadvanceddifferentsubfields,orareasofapplication,ofConstructionGrammar.Thestructureoftheexpositionbelowisasfollows:Section6.2willverybrieflydiscussafewclassicstudiesthatarebasedonintrospectivedatabutthathavestillhelpedfoundanddevelopConstructionGrammar.Section6.3willdiscussdataandmethodstraditionallyreferredtoasobservational,

1

Page 83: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 3 of 11

thatis,dataandmethodsthatscorehighlyonallthreenaturalnessdimensionsunderlyingtheabovecontinuumoflinguisticdata.Sincethesecorpus-basedapproachesdonotdiffermuchwithregardtothesedimensions,Iwillinsteaddividethemaccordingtothewaysinwhichthefrequencydatafromcorporaareused.Section6.4willthendealwithwhataretraditionallycalledexperimentalmethods,whereIwilldistinguishdifferentexperimentalapproachesonthebasisoftheabovethreedimensions.Section6.5willverybrieflyaddresscomputational-linguistic/machine-learningtypesofapproaches.Section6.6willconcludeandpresentafewdirectionsforfurtherevolutionandmaturationofdataandmethodsinConstructionGrammar.

6.2.FromIntrospectiveJudgmentstoOtherData

ItisprobablyfairtodatetheemergenceofConstructionGrammartothelate1980s,whenproponentsofwhatarenowregardedasdifferentmembersofthefamilyofConstructionGrammarspublishedhighlyinfluentialstudies.ForwhatisnowoftenreferredtoasCognitive,orGoldbergian,ConstructionGrammar(cf.Boas,thisvolume),Lakoff's(1987)studyofthere-constructionsbrokeimportantgroundinthewayitshowedhowdifferentthere-constructions(typesofdeicticandexistentialconstructions)formaradialcategoryofthesametypeasthesensesofpolysemouswordsformradialcategories.Similarly,forwhatisnowoftenreferredtoas(Berkeley)ConstructionGrammar,Fillmore,Kay,andO’Connor's(1988;Fillmore,thisvolume)studyofletalonepavedthewayformanyimportantlaterstudies.Crucially,theseandmanyotherground-breakingstudieswereultimatelybasedonintrospectivejudgmentsaboutwhatonecanandcannotsayindifferentcircumstancesorcontexts,withveryrarereferencestoauthentic/naturalexamples(e.g.,Fillmore,Kay,andO’Connor1988:524).Thesameistrueofotherearlierinfluentialpublications.AquickglanceattheearlyvolumesofCognitiveLinguisticsindicate,forexample,thatthefirstexplicitlyconstruction-basedstudies(e.g.,Goldberg1992;Smith1994;KemmerandVerhagen1994;DancygierandSweetser1997;Morgan1997)wereallintrospection-basedand,iftheyusedtheworddata,theyuseditreferringtointrospectivejudgmentsand/orexamplesentences(justlikeLakoff1987).

WhilethisdoesinnowaydiminishthewaytheseandmanyotherpublicationsfromthattimegaverisetoanewfieldwithinCognitiveLinguistics,itdidneverthelessnottakelongformanyscholarstoalsousemoreandmorediversemethods.In1998,CroftandSandradebatedthedegreetowhichlinguistsandtheirmethodscancontributetomattersofmentalrepresentation,withCroftarguingthat“evidencefromactualusageorpsycholinguisticexperiments”isneededtoaddressmentalrepresentation,and1998/1999sawthefirstconstruction-basedpublicationsinCognitiveLinguisticsthatuseddatafromexperiments,orcorpusdata(ordatafromboth):TomaselloandBrooks's(1998)experimentalstudyoftheearlyacquisitionofTransitiveandIntransitiveconstructions(whosediscussionalsoinvolvesTomasello'searlierdiary-baseddata),Palancar's(1999)comprehensionexperimentinvolvinghittingconstructions,andGries's(1999)corpusanalysisandacceptabilityratingexperimentofverb-particleconstructions.ThefollowingtwosectionswillthereforebeconcernedwithobservationalandexperimentalapproachesthathavebeenusedinConstructionGrammarstudies.

6.3.ObservationalApproaches

Asmentionedabove,observationaldataintheformofcorpusdatahavebeenplayingaveryimportantroleinConstructionGrammarformanyyearsnow.Thesecorpusdatadonotdiffermuchintermsoftheabovethreedimensionsoflinguisticdata,andcorpus-basedstudiesinConstructionGrammarhaveusedmanydifferentkindsofcorporaortextualdatabases(spaceonlypermitsmentionofmaximallytwoauthors):

−intermsoflanguages:Czech(cf.Fried2009b),Danish(cf.Hilpert2008),Dutch(cf.Colleman2009a),English(cf.Gries2003a,b;T.Hoffmann2006),Finnish(cf.KolehmainenandLarjavaara2004),French(cf.ChenuandJisa2006;Marandin2006),German(cf.Diewald2006;Hilpert2009),Germanvs.English(Boas2003),Greek(cf.KatisandStampouliadou2009),Hindi(Budwig,Narasimhan,andSrivastava2006),Mandarin(cf.Chen2006),Russian(cf.Eckhoff2006),Spanish(cf.Gonzálvez-García2006),andSwedish(cf.Hilpert2008);

−intermsofmodes/registers:journalese(cf.Croft2009c),internetdata(Stefanowitsch2011b),butprobablymoststudiesarebasedonamixtureofspokenandwrittendatathatcharacterizesmostcontemporarycorpora;

−intermsofdialects/varieties:AsianEnglishes(cf.MukherjeeandGries2009),Belgianvs.NetherlandicDutch(cf.Grondelaers,Speelman,andGeeraerts2007),Britishvs.AmericanEnglish(GriesandStefanowitsch2010),Lancashiredialect(cf.HollmannandSiewierska2007),…;

Page 84: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 4 of 11

−intermsofsynchroniccorpora(cf.Gries2003b)vs.languageacquisitioncorpora(cf.Goldberg1999;DiesselandTomasello2005b)vs.diachronic/historicalcorpora(cf.Hilpert2008;Fried2009b),…

Whilemostcorpusstudies’dataarefromthemostnaturalendoneachdimension,suchstudiesexhibitconsiderablevariationintermsofhowthecorpusdataareused,andsincecorpusdataprovidenothingbutfrequenciesof(co-)occurrence,thesecorpusstudiescanbelocatedona(nonevaluative)clineofstatisticalcomplexity.Thenextfewsubsectionswilldiscussdifferentlyquantitativecorpus-basedapproacheswithindifferentareasofConstructionGrammar.

6.3.1Frequenciesof(Co-)occurrence

Thesimplestapproachtoincludecorpusfrequenciesinvolvesmerelycheckingwhetheraparticularconstruction,acombinationofconstructions,oraparticularlexicaliteminaconstructionisattestedornot.Inotherwords,therelevantfrequencydistinctionisbetweenzeroandoneormore.Whilethisapproachmaynotseemparticularlyexciting,itcanhaveimportantimplications.Onekindofsuchimplicationshastodowiththefactthatcorporamayprovidecounterexamplesforhithertowidelyacceptedclaims.Forexample,theprobablymostwidelystudiedArgumentStructureconstruction,theEnglishDitransitiveconstructionVNP NP ,hasoftenbeenclaimedtoexhibitsomepuzzlinglexical(dis)preferences,forexample,thatitdoesnotoccurwiththeverbdonateintheditransitiveslot.However,thisassessmenthasnearlyexclusivelybeenbasedonlinguists’armchairjudgments.Stefanowitsch(2011b),ontheotherhand,showsthatnotonlydowebsiteswithukastheirtop-leveldomainnamecontaina“substantialnumber”ofditransitiveswithdonate,thesematchesalsoexhibitanoteworthysemanticpatterningthatfitsintoamoregeneralaccountofthedative‘alternation’anditsinformationstructure.SeeStefanowitsch(2011b)formoreexamplesinvolvingotherverbsintheditransitive.

Inspiteoftheirstatisticalsimplicity,rawfrequenciescanalsobehighlyrevealinginfirst-language(L1)acquisitionandsecond/foreign-language(L2/FLA)learningcontexts.Regardingtheformer,L1acquisitioncontexts,Goldberg(1999)discussesfrequenciesofdifferentverbsfromL1acquisitioncorporafromtheCHILDESdatabasewithaneyetohowthehighfrequenciesofparticular(oftensemanticallylight)verbsfacilitatestheacquisitionofargumentstructureconstructionswhosemeaningsarecompatiblewithsomeoftheseverbs.ObservingessentiallyZipfiandistributionsofverbsinparticularconstructions,sheargues,forexample,thatthehighfrequenciesofgo,put,andgivefacilitatetheacquisitionoftheIntransitive-motion,theCausedMotion,andtheDitransitiveconstructions,respectively(cf.alsoGoldberg,Casenhiser,andSethuraman2004;Goldberg2006a);cf.Tomasello(2003)forabook-lengthtreatmentwithmanydifferentinsightfulcasestudies.Regardingthelatter,L2/FLAcontexts,similarobservationsweremadeinEllisandFerreira-Junior's(2009a)studyoflongitudinaldatafromtheESLdataoftheEuropeanScienceFoundation(ESF)corpuswithregardtoputinCausedMotionconstructionsandgiveinDitransitiveconstructions(cf.alsoEllisandFerreira-Junior2009b).

6.3.2ConditionalProbabilities(Unidirectional)

Thenextsteponaclineofstatisticalcomplexityleadstoapproachesinvolvingthecomputationofconditionalprobabilitiesorotherunidirectionalmeasuresthatarebasedonthem. Again,thesearewidelyusedinstudiesinL1acquisitionandL2/FLlearning,andIwillagainuseexamplesfromtwoprominentfiguresinthesefields.Asfortheformer,Goldberg,Casenhiser,andSethuraman(2004)addressesthequestionofhowreliableconstructionsareaspredictorsofsentencemeaning(foranexperimentalapproachtowardthatquestion,cf.below).Theycountedallinstancesofcaused-motionmeaningsintheBatescorpusfromtheCHILDESdatabaseandthencomputedthecuevalidityofthepatternV-Obj-Locforcaused-motionmeaningsp(‘caused-motion’|V-Obj-Loc).Theyfindahighcuevalidity(between0.63and0.83,dependingonhowinclusiveadefinitionofthemeaningofcaused-motionisadopted),whichshowsthattheV-Obj-LocpatternisagoodcueforthemeaningitisassociatedwithinConstructionGrammaraccounts.However,themorecrucialimplicationofthisfindingonlyariseswhenthepattern'scuevalidityfor‘caused-motion’iscomparedwiththe(weighted)cuevalidityofverbsforthesamemeaning,0.68.Hence,usingcorpusdata,Goldberg,Casenhiser,andSethuraman(2004)showedthatsyntacticpatternsarejustasreliableascuestosententialmeaningasverbs.

Asforthelatter,EllisandFerreira-Junior(2009b)studytheeffectoftype/tokenfrequenciesofwordsinslotsoftheIntransitiveMotion,theCausedMotionconstruction,andtheDitransitiveconstructioninL2/FLAintheESFcorpus.

1 2

2

Page 85: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 5 of 11

Toquantifytheunidirectionalassociationofthewordstotheconstructions,theyuse—unlikemostsuchstudies,whichusedabidirectionalcollostructionalanalysismeasure(cf.thefollowingsection)—averyinterestingunidirectionalmeasurecalledΔP,whichiscomputedonthebasisofconditionalprobabilities.Morespecifically,ΔPisthedifferenceoftheprobabilityofanoutcomeOgivenacueCminustheprobabilityofOgiventheabsenceofC.Theyfindthatthefirst-learnedtypesineachslotofeachoftheconstructions—esp.theverbs—arehighlydistinctivefortheirconstructionalslots(bothintermsofΔPandthecollostructionalmeasuresdiscussedbelow).Thisfindinginturnsupportsanunderstandingofconstructionalacquisitionasdependentonalargervarietyoffactorsthanareoftendiscussed:whiletypeandtokenfrequenciesdoplayimportantroles,thedistributionsoffrequencies,aswellasthedistinctivenessofelementsforthepositionsinwhichtheyareusedandthedegreetowhichtheyformchunksarealsohighlyrelevant.(Cf.BybeeandScheibman1999andBybeeandThompson2000onhowhigh(type/token)frequenciesofwordsinparticularcontexts/withparticularmeaningsarerelatedtochunking,grammaticalization,andphonologicalreductionprocesses.)

6.3.3AssociationStrengths(Bidirectional)

Oneofthemostwidespreadcorpus-basedmethodologicalapproachesinConstructionGrammarisreferredtoascollostructionalanalysis,afamilyofseveraldifferentmethods.Sincethisapproachisdealtwithinaseparatechapter(cf.Stefanowitsch,thisvolume),Iwillnotdiscussitingreatdetail,butinorderforthischaptertobesufficientlyself-contained,afew,moregeneralremarksaboutthisfamilyofmethodsareinorder.

JustlikeEllisandFerreira-Junior's(2009b)measureofΔP,the(earlier)approachofcollostructionalanalysisisawaytoquantifyassociationstrength,whichisultimatelybasedoncollocationalapproachesfromcorpuslinguistics.ButunlikeΔP,itisanapproachtocomputeabidirectionalassociationmeasure.(Itisworthpointingoutthoughthatcollostructionalanalysisdoesnotrequirebidirectionalmeasures;fromthatpointofview,EllisandFerreira-Junior'sapproachisaparticularimplementationofcollostructionalanalysis.)Threedifferentmethodsaredistinguished,mostofwhichcanbecomputedwithColl.analysis3.2a(Gries2007),ascriptavailablefromtheauthor'swebsite:

−collexemeanalysis,whichcomputesfornwordshowstronglythesewordsareattractedtoaslotinaconstruction(cf.StefanowitschandGries2003);

−(multiple)distinctivecollexemeanalysis,whichcomputesfornwordshowstronglythesewordsareattractedtotwoormorefunctionallysimilarconstructions(cf.GriesandStefanowitsch2004a);

−co-varyingcollexemeanalysis(item-basedandsystem-based),whichcomputesfornwordsinoneslotofaconstructionhowstronglythesewordsareattractedtotheywordsinanotherslotofthesameconstruction(cf.GriesandStefanowitsch2004b;StefanowitschandGries2005).

Allofthesemethodsproviderankingsofhowmuchwordsandparticularslotsofconstructionsattracteachother(viz.,thebidirectionalnatureoftheseassociationmeasures)andwhatthatrevealsaboutconstructionalsemantics,andtheyhavegivenrisetomanystudies:Gilquin(2006)onEnglishperiphrasticcausatives,Wulff(2006,2008a)ongo(and)Vandgo/come/try(and)V,Hilpert(2006a)onthediachronicdevelopmentofverbalcomplementsofshall,Hilpert(2008)onthediachronicdevelopmentofFutureconstructionsinGermaniclanguages,andmanymore.Thisapproachalsohassomepsycholinguisticrelevance,sincethepreferenceofverbstooccurinparticular(ArgumentStructure)constructions(i.e.,theverbs’subcategorizationpreferences)areknowntostronglycorrelatewithlinguisticprocessing(cf.Garnseyetal.1997;Stallings,MacDonald,andO'seaghdha1998;Hare,McRae,andElman2003;MelingerandDobel2005).Whilethismustsufficehereforadiscussionofcollostructionalanalysis,thetopicwillberevisitedinthenextsectiontodiscussexperimentsthattriedtovalidatethisapproachexperimentally.

6.3.4MultifactorialandMultivariateApproaches

Given(1)theobviouslymultifacetednatureoflanguageanditsrelationto,orinteractionwith,cognitiveprocessingand(2)thecomplexityandnoisinessofdataobtainedfromcorpora,itisoftennecessarytoresorttostatisticalmethodsthatcandobetterjusticetotheobservedfacts.Intheory,ofcourse,nearlyeveryphenomenonstudiedcorpus-linguisticallycan,andprobablyshould,bestudiedmultifactorially,sotherangeofpossibilitiesthatcouldbesurveyedisextremelylarge.Iwillmentiontwokindsofapproaches,whicharenotonlymultifactorial/-variatebutalsomethodologicallypluralisticinhowtheycombinedatafromcorporaanddatafromexperiments.

Page 86: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 6 of 11

ThefirstoftheseisconcernedwithanotionfromtheverybeginningofConstructionGrammar,idiomaticity.Asmentionedabove,earlystudiesinConstructionGrammarweredevotedtothestudyofdifferentkindsofidiomsandtohowthestudyoftheseitemsthatareoftenconsidered‘marginal’illuminatesthestudyofmoreregularconstructions.However,ashasbeenwellknown,idiomaticityisaperplexinglymultidimensionalnotion,hardtooperationalizeorevenjustrank-orderonthebasisofintrospectionalone.Wulff(2009)isastudythatapproachesidiomaticityonthebasisofexperimentalandcorpusdataforthirty-nineV-NPidiomsfromtheBritishNationalCorpus.First,shecollectedidiomaticityjudgmentdatafromsubjectsusingthemethodofmagnitudeestimation.Second,andmoreimportantlyhere,sheusedtwodifferentcorpus-basedwaystooperationalizedifferentdimensionsofidiomaticity:collocationaloverlapandameasureofformalflexibilitythatwasinturnedbasedontwentyidiomaticvariationparameters(describingmorphologicalandsyntacticparametersoftheidioms’use).Usingmultivariateandmultifactorialmethods—principalcomponentsanalysisandmultipleregression—shethenidentifiedwhichidiomvariationparameterscluster(andarethuslikelytounderlieperceptionsofidiomaticity)andvalidatedthesefactors/clustersonthebasisofthespeakerjudgments.Thisshowed,amongotherthings,thatcompositionalitywasnotasstrongapredictoraswascommonlyheld;cf.Wulff(2008b)formoredetaileddiscussion.

Thesecondmultidimensionalstudytobediscussedherewasconcernedwithidentifyingprototypicalinstancesofconstructions.Gries(2003b)retrievedexamplesofthedativealternationfromtheBritishNationalCorpusandcodedthemforalargenumberofmorphological,syntactic,semantic,anddiscourse-pragmaticcharacteristics.Hethenusedalineardiscriminantanalysistodeterminewhichofthesecharacteristics,ifany,weregoodpredictorsoftheconstructionalchoicesinthecorpusdata.Heshowedthatnearly89%ofallconstructionalchoicescouldbeclassifiedcorrectly(andhowimportanteachvariable'scontributiontothatwas),butmoreimportantly,eachcorpusinstancewasassignedadiscriminantscorethatrevealshowgood,orprototypical,anexampleoftheditransitiveandtheprepositionaldativeis(intermsofhowsureandcorrecttheanalysiswasinassigningaconstructionalchoice).Asafirstattemptatvalidationofthesecorpus-basedfindings,hediscussedseveralsalientconstructions—prototypicalexamplesthatwerepredictedcorrectlyandexampleswherethemodelwaswrong—butthemoreforcefulvalidationwasanacceptabilityjudgmentexperiment,inwhichspeakersratedsentenceswellwhentheyoccurredintheconstructionthatthecorpus-baseddiscriminantanalysispredictedforthem.Amongotherthings,Griesargued,therefore,thatthiskindofcorpus-basedmultifactorialapproachisavalidandusefultooltoobtaingoodness-exampleinformationfordatathatcanbeusefulfor,say,acquisitionapproachesorthestudyofalternations,etc.Forothermultifactorialcorpus-basedapplications,cf.Gries(2003a),BrenierandMichaelis(2005),Bresnanetal.(2007),orT.Hoffmann(2006,2011).

6.4.ExperimentalApproaches

Apartfromalargeandgrowingnumberofcorpus-basedapproaches,studiesinConstructionGrammarhavealsoemployedmanydifferentkindsofexperiments.Whilethesearetypicallynotfromthemosttechnical/artificialtypeofsettings,theyneverthelessexhibitquitesomevariation.Thissectiondiscussesseveralexperimentalapproacheswithaneyetosurveyingthekindsofmethodsanddatathatwereused,anditdoessobymovingroughlyfrommorenatural/lessartificialsettings,stimuli,andresponsestolessnatural/moreartificialones.

Theexperimentalapproachesthatareamongthemostnaturalonalldimensionsarethoseinvolvingyoungchildren.Forexample,inoneofthefirstnonintrospectiveConstructionGrammarpapersinCognitiveLinguistics,TomaselloandBrooks(1998)performedanexperimentinwhichchildren(meanage2;3)weretaughtnovelverbs,onewith/inatransitive,theotherwith/inanintransitivemeaning/scenario.Theythenencouragedthechildren'suseofthenovelverbsinaconstructionby,forexample,saying“Thisiscalledmeeking.Canyousaymeeking?Saymeeking?”andbyaskingdescriptivequestionsinanelicitationtask.Whilethesetaskswouldbesomewhatartificialinanadult-onlycontext,interactionsliketheseare,ofcourse,notatallrareindiscoursesbetweenchildrenandtheircaretakers,whichiswhythiskindofexperimentalapproachiscategorizedasnaturalonalldimensions.TomaselloandBrooksfindthat“youngchildrenlearntheirfirstsentence-levelconstructionsonaverb-specificbasis”(1998:391),supportingtheitem-basedapproachforwhichtheTomaselloandhisresearchgrouphavebecomesowell-known(cf.Abbot-SmithandTomasello2006,2010;Kidd,Lieven,andTomasello2010,tonamebutafewmoreexamples).

AsimilarexampleisCasenhiserandGoldberg's(2005)studyofnovelverblearning.JustlikeTomaselloandBrooks,

Page 87: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 7 of 11

theyexposed(somewhatolder)children(meanage:6;4)tononceverbs,but,unlikeTomaselloandBrooks,theyexposedthemtononceverbsinaphrasalpatternthatdoesnotexistinEnglishandtheysystematicallyvariedthetokenfrequencieswithwhichthenonceverbsoccurredinthenovelpattern.Aftertheshorttrainingsession(lessthanthreeminutes),thechildrenparticipatedinaforced-choicecomprehensiontask;thedependentvariablewaswhethertheycouldunderstandsentencesusingthenovelpatterncorrectly,especiallywhentheirtraininginvolvedatokenfrequencydistributionthatwasskewedinawaythatisskewedsimilarlytotheZipfiandistributionsofverbsinconstructionsdiscussedabove,showingthatchildrenareveryfastatidentifyingprobabilisticpatternsinskeweddistributionsandassociatingameaningwiththem.

AnotherrangeofexperimentalapproachesusedinConstructionGrammarinvolvesseveralpaper-and-penciltasks,whichinvolveintermediatelyunnaturalexperimentalsettingsbutthatdifferwithregardtothenaturalnessofthestimuliandthe‘output’producedbythesubjects.

Onesetofexperimentsthathasprovideddifferentkindsofusefulfindingsinvolvesprimingeffects.Manyprimingstudieshaverestrictedthemselvestoapurelysyntactic/structuralviewofpriming,butinanimportantstudyHareandGoldberg(1999)extendedapreviousstudybyBockandLoebelltodeterminetowhatdegree,ifany,primingmaynotjustbesyntacticbutalsoinfluenced,orreinforced,bysemanticfactors.HareandGoldbergusedapicture-descriptiontaskinwhichsubjectsdescribeditransitivescenariosafterhavingheardoneofthreedifferentprimesentences(oranintransitivecontrolsentence).Importantly,theyfoundthat“theorderofexpressionofcoarsesemanticroles”and“thelevelofthemappingbetweensemanticsandsyntax”influencedsubjects’reaction(cf.alsoF.Chang,Bock,andGoldberg2003).

Otherprimingstudiesinvolve,forexample,experimentaldesignswherethesubjectsdonotproduceafull‘normal’sentencebutaslightlylessnaturalresponse,namelywheretheycompleteasentencefragment.Onesuchexampleinvolvesforeignlanguagelearners’knowledgeofconstructions.GriesandWulff(2005)conductedasentence-completionexperimentwithadvancedGermanlearnersofEnglish.Inthisstudy,primesweresetuptobiassubjectsintoproducingeitherditransitivesorprepositionaldativestodeterminewhether(1)Germanlearnersexhibitthesamekindsofprimingeffectsasnativespeakersand(2),justasimportantly,whetherGermanlearnersexhibitthesamekindsofverb-constructionpreferencesasnativespeakersofEnglish.GriesandWulfffoundbothoftheseeffects:thelearnersexhibitedconstructionalprimingeffectsandverb-constructionpreferencesthatwereverysimilartonativespeakers,buttheyalsoshowedthattheverb-constructionpreferencestheyfoundwerenotduetotranslationalequivalents’transfereffects.GriesandWulff(2009)thenconductedasimilarstudy,thistimetestingforwhetherprimingcanbeobtainedfortwocomplementationpatterns—to-vs.ing-complementationafterverbssuchaslikeortry—and,ifso,whatthesourceoftheprimingeffectis.Theyagainfoundstrongandsignificantprimingeffectsforbothconstructions,mainlyfromtheverbintheprimebutalsothesubjects’ownlastcompletion.Bothstudiesthereforelentsupporttoexemplar-basedapproachestowardlinguisticknowledgeingeneralandcollostructionalknowledge/subcategorizationpreferencesingeneral.

AnotherexampleisGries,Hampe,andSchönefeld's(2005)studyofas-predicatives.Theywereconcernedwiththequestionofwhatkindsoffrequencydataaremostusefultothestudyofverbsinconstructions.Theyfirstundertookacorpusstudyoftheas-predicative(asin,e.g.,Heregardedthatasabigmistake)todetermineverbsthatarefrequentornotsofrequentinthatconstruction,aswellasverbsthatarehighlyattractedorbarelyattractedtotheconstruction(intermsofcollostructionalattraction).Then,theypresentedsubjectswithsentencefragmentsfeaturingverbsfromeachofthefourgroupsthatresultedfromcrossingthefrequencyandtheattractionconditions.Thedependentvariablewasthereforewhethersubjectswoulduseanas-predicativeornot,andtheyfoundthatthecollostructionalmeasurehadaverylargeeffectonthesubjects’completionpatterns(ashadthevoiceofthesentencefragment)whereasrawfrequencydidnot,whichlendsexperimentalsupportforcorpusstudiesofconstructionsusinguni-orbidirectionalmeasuresofassociation.

Studiesinwhichsubjectswererequestedtodosomethinglessnaturalthanproduceorcompletesentences,includecaseswheresubjectsfillgapsorsortsentences.Asfortheformer,Dąbrowska(2009)presentednativespeakersubjectswithsentencesfromdictionarydefinitionsofverbsofwalkingfromwhichtheseverbshavebeenomitted.Sheasksthemtofillthegap,anotparticularlynaturalresponsetype,andfindsthatsubjectsarequitegoodatfindingstherightverbonthebasisofthecollocationalknowledgetheyhaveaccumulatedovertime.

Asforthelatter,BenciniandGoldberg(2000)usedasortingparadigmtostudywhichcomponentsofasentence—

Page 88: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 8 of 11

themainverbortheArgumentStructureconstruction—aremostcentraltothesentence'soverallmeaning.NativespeakersofEnglishreceivedsixteencards,eachwithadifferentsentencethatusedoneoffourverbsinoneoffourArgumentStructureconstructions(Ditransitive,Transitive,CausedMotion,andResultativeconstruction);thestimulicanthereforebecategorizedasrathernatural.Thesubjectswerethenaskedtosortthesixteensentencesintopilesdependingonoverallsimilarityofmeaning(i.e.,performanot-so-naturallinguistictask).Thedependentvariableandthequestioninpointwaswhetherthesubjectswouldproducepilesbasedontheverbsorontheconstructions.Itturnedoutthatthesubjectsproducedsignificantlystrongerconstruction-basedclusters,whichunderscoredtherelevanceofArgumentStructureconstructionsforsentencemeaning.

Areplicationofthisstudyprovidedadditionalresults.GriesandWulff(2005)replicatedthisexperimentwithadvancedGermanlearnersofEnglish,withadditionalfindings.TheGermanlearnersalsoexhibitedasignificantpreferenceforconstruction-basedsortings—infactanevenstrongereffectinthisdirectionthanthenativespeakers,butGriesandWulffalsoanalyzedthesortingdatabymeansofexploratorydataanalysismethods,ahierarchicalagglomerativeclusteranalysisandaprincipalcomponentsanalysis.Bothofthesemethodsnotonlysupportedthefindingsthatthesortedsentencescameinconstruction-basedpiles(accountingformorethan90%ofthevarianceinthedata)butalsoresultedinadendrogramthatreflectshowsimilartheconstructionsaretoeachotherintheeyesofthesubjects.Interestinglyenough,theclusteringoftheconstructionsisperfectlycompatiblewiththeirtheoreticaltreatmentinConstructionGrammarsuchthat,forexample,ResultativeandCausedMotionconstructionsarerelatedmoststrongly,reflectingGoldberg's(1995)analysis.Thisisthereforeacasewhereamorecomprehensivestatisticalanalysisofthedatacouldyieldresultsthatevengobeyondtheoriginalquestion.

Anexperimentalapproachthatissimilarintermsofexperimentalsettingandinput,butinvolvesthesubjects’producingmaybemoreunnaturalresponsesareexperimentsinvolvingjudgmentdata.Dąbrowska(2008)studiedquestionswithlong-distancedependenciesandfinds,usingacceptabilityjudgments,thatthesequestionsexhibitverystrongprototypeeffectssuchthatquestionsthatcorrespondcloselytooneofseveralcorpus-derivedtemplatesreceivesignificantlybetterratings.Asmentionedabove,GriesandWulff(2005)determinedthattheirlearnersofEnglishhadthesameverb-constructionpreferencesasnativespeakersbycomparingtheirexperimentalbehaviortotheverbs’andtheirtranslationalequivalents’preferencesinEnglishandGermancorpora.GriesandWulff(2009)testedtheGermanlearners’preferencesmoredirectly.SubjectswerepresentedwithsentencesofthetwocomplementationpatternsV toV andV V -ing,butthesentencesweredesignedtocontainV 'sthatcollostructionallypreferredthefirstorthesecondpatternincorpusdata.Thedependentvariablewasthesubjects’acceptabilityratingstothefourcombinations(oftwoconstructionalpreferencesandtwoconstructionalstimuli).Theyfoundagainthatthelearnerswereverysensitivetotheverbs’constructionalpreference,givinghighratingstostimuliwhereverbswereusedintheconstructiontheyprefer,andlowratingsotherwise,alsolendingsupporttothekindofassumptionmadeinexemplar/usage-basedmodelsthatevenlearnersareabletokeeptrackofthefrequencieswithwhichverbsareusedinconstruction.

Thefinalexperimentalmethodtobediscussedhereinvolvesadesignwitharatherartificialdesign(onallthreelevels).Gries,Hampe,andSchönefeld(2010)conductedafollow-upstudyoftheir2005sentence-completionexperiment,whichinvolvedaself-pacedreadingtask.Onthebasisofalargercorpussample,theyagaincrossedfrequencyofco-occurrence(highvs.low),collostructionalattraction(highvs.low),andvoiceandpresentedsubjectswithsentencesfromtheBritishNationalCorpusthatcontainedtheseverbsbutwerealteredtorendertheirlengthsandcomplexitiescomparable,aswellasreplacingcontext-dependentexpressionssuchaspropernamesbymoregenericexpressions.Thesubjectsreadthesentencesword-by-wordsuchthattheyhadtopressabuttontorequestandobtainthenextword.Thedependentvariablewasthetimefromthepresentationofonewordtilltherequestofthenextword.Withonlyfewsubjects,theyobtained254readingtimes,butwhentheyanalyzedthereadingtimeofthewordfollowing—thewordthatshouldrevealtothesubjectswhethertheirinitialparseexpectationbasedontheverbwascorrectornot—theyfoundthatagainfrequencyhadnosignificanteffectatall(p=0.293),whereascollostructionalattractionexhibitedamarginallysignificanteffectinthepredicteddirection(ptwo-tailed=0.065),againsupportingtheimportanceofassociationstrengthsoverrawfrequencies.

Spacedoesnotpermitdiscussionofmoreexperimentalparadigmsthatwoulddomerittotheircomplexityandpotential,butafinalgroupofexperimentsmustneverthelessnotgocompletelyunmentioned,namelythelargebodyofworkthathasbeendoneintheareasofSimulationSemanticsandEmbodiedConstructionGrammar(cf.BergenandChang,thisvolume).Considerasacaserepresentativeformuchworkinthesefieldsaveryinteresting

1 2 1 2

1

Page 89: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 9 of 11

studybyBergenandWheeler.Startingoutfromtheviewthatunderstandinglanguageofteninvolvesmentalperceptualandmotorsimulations(asindicatedbyactivationofareasinthebrainresponsibleformotoraction),theytestaction-sentencecompatibilityeffects,thatis,whetherthedirectionofmotionrepresentedinasentenceiscompatiblewiththehandmovementthesubjectshavetomaketopressaresponsebutton(andthusspeedsupreactiontimes)ornot(andthusslowsdownreactiontimes).Usingthisparadigm,theyfindthatprogressiveaspectandperfectiveaspectresultinverydifferentAction-sentencecompatibilityeffects,whichsuggests(1)thatthedifferentaspectsresultindifferentmentalsimulationsoftheactionsdescribedand(2)thatgrammaticalfeaturessuchasaspectmodulate“second-orderpropertiesofthementalsimulationtobeperformed”and“whatpartofanevokedsimulationanunderstanderfocuseson,orthegrainofdetailwithwhichthesimulationisperformed”(BergenandWheeler2010:155).Studieslikethisarestillratherrarebutpointtoveryintriguingpossibilitiesforfutureresearchalongtheselines;cf.Bergen(2007)foranexcellentsummaryofdifferentexperimentalparadigmsinthesearea,whichareevolvingquicklyandbecomingmoreandmorerelevanttothefieldofConstructionGrammar.

6.5.Computational-Linguistic/Machine-LearningApproaches

Thekindofdataandmethodologythatareleastusedinconstruction-basedapproachesarecomputational-linguisticapproachesinvolving,forexample,machine-learningorsimulation-basedapproaches,andmuchoftheworkintheseareasthatwouldinfactberelevanttoconstruction-basedapproachesdoesnotestablishadirectconnectiontoConstructionGrammar.OneexampleisF.Changetal.(2000),whodevelopedaconnectionistmodeltotestwhetherstructural/syntacticpriming—which,asdiscussedabove,isseenbysomeasconstructionalpriming—canbeconsideredasresultingfromimplicitlearning(ratherthan,say,fromresidualactivationofnodesinaspreadingactivationnetwork).Theyfindthattheirtypeofsimplerecurrentnetworksuggeststhatprimingmayindeedresultfromtheverysamemechanismsthatunderlielanguagelearninginthefirstplace,andthat,amongotherthings,amodelthatinvolves/simulatesmessagecomprehensionyieldedmoreprimingeffectsofthetypethathumansexhibit.

Anapproachthatislesscomputationaland,thus,moretransparenttothetraditionallinguististheTracebackapproachdevelopedbyDąbrowska,Lieven,andcolleagues(cf.,e.g.,DąbrowskaandLieven2005;Lievenetal.2009;VogtandLieven2010forrecentexamples).Inthisapproach,aprogramcalledAutotraceridentifiesallmultiwordutterancetypesinatestcorpus,typicallythelasttwohoursofrecordedspeechofachild,andthenidentifiesall(continuousanddiscontinuous)stringsthatoccuratleasttwiceinthepriorrecordingsandthatcontainedoverlappinglexicalmaterialwiththetargetutterancetypesinthetestcorpus.Afterallpotentialcomponentunitswereidentified,theprogramattemptstobuildupallutterancetypesinthetestcorpusfromthepotentialcomponentunitswithsuperimpositions,substitutions,andadditions.Theobjectiveistodeterminehowmanyofthenovelutterancesofachildcanactuallybetracedbacktoonlyslightlychangedpreviousutterances,andfindingsindicatethat,inspiteofthesparsityofeventhedensestlanguageacquisitioncorpora,oftenthevastmajorityofchildren'snovelutterancescanbeaccountedforasexactrepetitionsorwithoneoperation.Inaddition,resultssuggestthatchildrenareinfactlearningchunks—anddonotfreelyassembleutterancesfromparts.

Methodssuchasthese,orthosediscussedinDominey(2006),arenotyetparticularlyfrequentinConstructionGrammar,buttheycanbeextremelyusefuladditionaltools,sincetheyallowtheresearchertoidentifypatternsinuse,aswellasdevelopmentaltrendsinacquisitionandlearning,thatarevirtuallyimpossibletodetectotherwise.

6.6.FutureDevelopments

Astheprevioussectionshaveillustrated,ConstructionGrammarisanempiricallyandmethodologicallyvibrantfield,usingdifferentdataanddifferentcutting-edgetechniques,whichisprobablyinnosmallpartduetotheclosenessofthefield's‘superfield,’CognitiveLinguistics,toCognitiveScience,afieldinwhichdiscussionsofdataandmethodshavebeenprominentevenduringthetimeduringwhichmuchoftheoreticallinguisticswasdominatedbyintrospectivedata.However,thisgenerallyfortunatestatedoesnotmeanthattherearenowaysinwhichConstructionGrammarcanevolvefurtherintermsofdataandmethods,andthefollowingbriefdiscussionmentionsafewdirectionsinwhichthefieldismostlikelytomove.

Withregardtoobservationalapproaches,recentdevelopmentsinCorpusLinguisticshaveresultedinmany

Page 90: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

Page 10 of 11

methodologicalinnovationsandtechniquesthatarerelevanttoConstructionGrammarinparticularandCognitiveLinguisticsingeneral.Corpuslinguistsareexploring:

−moreandmorediverseassociationmeasurestoquantifyifandhowmuchdifferentlinguisticelementsareattractedtoeachother(cf.section6.3.3above);

−waystoidentifyuninterruptedandinterruptedn-grams,whichcaninformlanguageacquisitionresearchonmultiwordunits(cf.section6.5above)butalsotheprobabilisticidentificationofpartsofspeech(cf.Redington,Chater,andFinch1998;Mintz,Newport,andBever2002);

−waysofquantifyingthedispersion/distributionoflinguisticelements,whichcanhelpexplainthelearnabilityoftheseelements(cf.Ambridgeetal.2006);etc.

Obviously,ConstructionGrammarianshavemuchtogainfromstayingup-to-datewithregardtosuchdevelopments.Similarcommentsapplytopsycholinguisticandbroadercognitive-linguisticfields,whereexperimentalmethodsarecontinuouslydevelopedand/orrefined,and,forexample,thecollectionofpapersinGonzalez-Marquezetal.(2007)discussesmanypotentiallyinterestingapplicationssuchaseyemovementresearchexperimentsonlanguageandspace,mostofwhichshouldbeapplicableandusefulinConstructionGrammarcontexts,too.Forexample,itisprobablyonlyamatterof(little)timeuntilthekindofimagingtechniquesdiscussedbyCoulson(2007)willbeappliedtomorespecificallyconstruction-basedquestions(cf.Pulvermüller,thisvolume).

Finally,withtheimportancethatusageplaysinmostcontemporaryincarnationsofConstructionGrammar,computationalsimulationsoffirst-languageacquisitionordiachronicchangewillassumeamorecentralrolethantheyhavedonesofar,andEdelman(2007)surveyssomenotionsrelevantinthiscontext.

AfinaldevelopmentrelevanttoempiricalConstructionGrammarianstranscendsthe(somewhattenuous)distinctionbetweenobservationalandexperimentalapproaches:howdataareanalyzedstatistically.Therearestillmanystudiesinwhichdataarenotanalyzedwiththenecessarydegreeofrigor.Inaddition,therearerelativelynewdevelopmentsinthefieldofstatisticsthatareverypromising.Oneoftheseisthemethodofmixed-effectsmodels,ormultilevelmodels,afamilyofapproachesofgeneralizedlinearmodelsthatisextremelypowerfulinhowithandlesrandomeffects(suchassubject-/stimulus-specificvariation),unequalcellfrequencies,andmissingdataandinhowtheseadvantagesmakestatisticalestimatesmuchmoreprecise(cf.GelmanandHill2008).Whilestandardsarestillemerginginthisdomain,thisisamethodologicaltrendthatConstructionGrammariansshouldbeandremainawareof.

Toconclude,truetothespiritofthecognitivecommitmentmentionedabove,researchersworkingin/onConstructionGrammaralreadymakeuseofavastarrayofdataandmethodsthathaveprovenusefulandyieldedveryinformativeresultsinmanyneighboringdisciplines.Itseemsthat,overtime,thetrendtowardmethodsthataremorerigorousandreplicablethanintrospectivejudgmentshasonlybecomestronger,anditremainstobehopedthattheabovedesiderataandtheadoptionofsomeofthemorerecentdevelopmentsinCognitiveLinguistics,Psycholinguistics,andCorpusLinguisticsalsofindtheirwayintotheConstructionGrammarian'stoolbox.

Notes:

(1.)Eventhislengthiercharacterizationisstillasimplification,sincesubjectsmaybepresentedwithdifferentkindsofstimuliatthesametime,etc.

(2.)Conditionalprobabilitiesarewrittenasp(E|F),whichmeans‘theprobabilityofaneventE,giventhatanothereventFhasoccurred.’Anexamplewouldbetheprobabilityp(ditransitive|recipient=animate),i.e.,theprobabilitythataspeakerwilluseaDitransitiveconstruction(asopposedtoaprepositionaldativewithto)whentherecipientisanimate(asopposedtoinanimate).

StefanTh.GriesStefanTh.GriesisProfessorofLinguisticsattheUniversityofCalifornia,SantaBarbara,aswellasHonoraryLiebigProfessorattheJustus-Liebig-UniversitätGiessen.Theoretically,heisacognitivelyandConstructionGrammarorientedlinguistinthewidersenseofseekingexplanationsintermsofcognitiveprocesses;methodologically,heisaquantitativecorpuslinguistattheintersectionofcorpus,cognitive,computational,andquantitativelinguistics.Heisfoundingeditor-in-chiefoftheinternationalpeer-reviewedjournal

Page 91: The Oxford Handbook of Construction Grammar

Data in Construction Grammar

CorpusLinguisticsandLinguisticTheory,associateeditorofCognitiveLinguistics,andontheeditorialboardsofCogniTextes,ConstructionsandFrames,Corpora,LanguageandCognition,andothers.

Page 92: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 1 of 17

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0007

BerkeleyConstructionGrammarCharlesJ.FillmoreTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterdiscussestheBerkeleyConstructionGrammar(BCG),theframeworkthroughwhichgrammaticalphenomenaofEnglishwereorganizedanddescribedinthatclass.Itexplainsthebasicconceptsoffeaturestructuresandunification,andintroducesthevalencerequirementsofpredicatingwordsandthedevicesfortheirsatisfaction.ThechapteralsopresentsBCGanalysesofvariousconstructions,suchastheSubject-Predicateconstruction,Head-Complementconstruction,andInversionconstructionorLeft-Isolationconstruction.

Keywords:BerkeleyConstructionGrammar,grammaticalphenomena,valencerequirements,Subject-Predicateconstruction,Head-Complementconstruction,Inversionconstruction

7.1.Introduction

Atatimewhenmuchof‘mainstream’grammaticaltheorizingintheUnitedStateswasseekingsimpleanduniversalgeneralizationsaboutthecutofhumanlanguages(Chomsky1995;cf.Pesetsky1999),grammaticalstudiesinBerkeleyinthe1980sand1990s,influencedbyworkelsewherebyJamesMcCawley(1988a),HajRoss(undergroundsquibs),andArnoldZwicky(especially1994),werecenteredondiscoveringtheidiomaticand‘irregular’partsoflanguage,demonstratingtheirfrequencyintextandtheircentralityinthelinguisticknowledgeofspeakers.IncoursestaughtbyPaulKay,GeorgeLakoff,andCharlesFillmore,studentresearchpaperstendedtoexamineselected‘noncanonical’grammaticalpatternsandexploretheirconstraintsandrangeofvariation—structureslikethosein(1).

(1)a.What,me,getmarried?b.Boy,wasIhot!c.Isn’tit(about)timeyouwenttobed?d.TheharderItry,thelessIaccomplish.e.Howbig(of)aboxdoyouneed?f.Iwouldn’ttouchit,letaloneeatit.g.Mayherestinpeace/rotinhell.h.JustbecauseIliveinBerkeleydoesn’tmeanI’marevolutionary.i.What'sthatflagdoingathalfmast?

AcoursetaughtbyKayandFillmoretriedtocorrectanemergingviewthatwhatconstructionistsstudiedwererandomlycollectedlinguisticmarginaliaindependentof,oratbestparasiticon,therealgrammar.Wetriedtoshowthatthesameanalytictoolsaccountforbothmostbasicstructuresandthese‘special’cases.

ByBerkeleyConstructionGrammar(BCG)IwillmeantheframeworkthroughwhichgrammaticalphenomenaofEnglishwereorganizedanddescribedinthatclass.ImportantworkreshapingormovingoutfromtheBCG

Page 93: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 2 of 17

frameworkhasbeencarriedonbyMirjamFried,SeikoFujii,AdeleGoldberg,Jean-PierreKoenig,KnudLambrecht,YoshikoMatsumoto,LauraMichaelis,KyokoOhara,ToshioOhori,Jan-OlaÖstman,EveSweetser,andseveralothers.

Section7.2developsthebasicconceptsoffeaturestructuresandunification,illustratedwithsomesimplenominalstructures.Section7.3introducesthevalencerequirementsofpredicatingwordsandthedevicesfortheirsatisfaction.Section7.4reviewsthefeaturesofBCGastheyfigureinthedescriptionofcomplexstructures.Section7.5laysoutwaysinwhichthesystemhasbeenorcouldbeappliedtoidiomaticconstructions.

7.2.Constructions,FeatureStructures,andUnification

InBCG,thegrammarofalanguageisthesetofitsgrammaticalconstructions,therulesthatuniteformalandsemanticinformationintovariouskindsoflinguisticobjects, togetherwiththeprinciplesthatconstrainandconnectthem.Aswithanyintendedgenerativegrammar,thegrammaristobeevaluatedbywhetheritisformallycoherentenoughto‘generate’somevirtuallanguage,whetherthatlanguageiscloseenoughtothetargetlanguage,andwhetheritcanbebroughtcloserbycontinuingwiththesameconceptualtoolsormodifyingtheminwell-definedways.

Anywell-formedlinguisticentitycanbeinterpretedasanassemblyoftheconstructionsthatjointlylicenseit,andanambiguousexpressionpermitsmorethanonesuchassembly.InBCG,boththerulesandtheobjectstheyaccountforarerepresentedascomplexfeaturestructuresoflargelythesameform.Theconstructionsthemselvesarepartialdescriptionsofthoseobjects.Themainoperationis(naive)unification,sothegrammarhasnodeepstructure,notransformations,andnoemptycategories.Whatyouseeiswhatyouget.

FormallyBCGisaphrase-structuregrammarwhosenodesarecomplexfeatures.Inearlywork(Fillmore1986;seealsoFriedandÖstman2004a)constituentstructurewasrepresentedasleft-to-rightarrangementsof“boxeswithinboxes”;informationaboutanylinguisticentityappearedasanattribute-value-matrix(AVM)associatedwiththeboxthatrepresentedthatentity,andpropertiesoftheconstituentsofsuchanentitywererecordedasAVMsinsidesmallercontainedboxes.Aphraselikethedogcouldberepresentedwithalargerboxcontainingtwosmallerboxes,theboxontheleftrepresentingtheandtheoneontherightrepresentingdog.Thephrasethedogisawell-formedlinguisticobjectbecausethereisaconstructioninEnglishthatallowsawordofthekindthattheistobeadjoinedtotheleftofthekindofentitythatdogis,andthephrasethatresultsfromtheirjoininghasthepropertiesithasbyvirtueofitsparticipationinthatconstruction.Thegrammarian'sjobistoshowwhatkindofwordtheis,andwhatkindoflinguisticentitydogrepresentsinthiscontext,andtopositthemostgeneralformulationoftherulethatputsthemtogether.

7.2.1Unification

Themainformaloperationisunification,welcomingorrejectingcandidateconstituentsinparticularphrasalpositionsaccordingtowhetherthevaluesassignedtotheirattributesarecompatiblewithvaluesrequiredbythosepositions.Something,ofcourse,hastodefinethe‘positions’forwhichsuchstipulationsarestated.Patternsanalogoustophrase-structurerulesdefinethepositionsthatsatisfytheneedsoftheentitytheycombinetoform;andspecificgoverningwordscandeterminethekindsofentitiesthatcanormustaccompanytheminthephrasestheyparticipatein.

Therepresentationsinthissectionshowtwokindsofbracketing:squarebrackets“[]”blockofflinguisticconstituents,andanglebrackets“<>”followingthenameofanattributeenclosethenameofthevalueofthatattribute. Maximalityhaspolarvalues(max<+>,max<–>;categoryhasvaluesnoun,verb,preposition,adjective,adverb,ordeterminer(cat<n>,cat<v>,cat<p>,cat<a>,cat<av>,cat<det>);numberhasthevaluespluralandsingular(num<pl>ornum<sg>);grammaticalfunctionincludessubject,object,oblique,complement(gf<subj>,gf<obj>,gf<obl>,gf<comp>);phrasetypeencompassesNP,PP,infinitivalVP,finitesentence,etc.(pt<np>,pt<pp>,pt<vpinf>,pt<sfin>,etc.);theattributelexemehasanunlimitedsetofpossiblevalues,representingthelanguage'slexicalresources(lxm<the>,lxm<dog>,lxm<people>,lxm<Chucho>,etc).

7.2.2Constituency

1

2

Page 94: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 3 of 17

Figure7.1. PrepComp

Figure7.2. DetHead

Thewell-formednessofcompleteorpartialdescriptionsofphrasesisdeterminedbyconstraintsthatlicensethejuxtapositionoftwoormoreconstituentswithinaphrase.WestartwithPrepComp(prepositionpluscomplement,Figure7.1)thatcreatesprepositionalphrasesbyjuxtaposingaprepositiontoamaximalnominal,andDetHead(determinerplusnominalhead,Fig.7.2)thatcreatesamaximalnominalbyjuxtaposingadeterminertoanonmaximalnominal.Amaximalnominaliscapableofoccurringasanargument—assubjectorobjectofaverb,or,ashere,theobjectofapreposition.Anonmaximalnominalisanoun-headedconstituentthatcanparticipateinargumentstatusonlyifitoccursinsideaphraselicensedbyaconstructionthatcreatesmaximalnominals—suchasDetHead.

Ineachcase,theright-handpositioncanbefilledbyaphraseorbyasinglelexeme.Wenoticein(2)certain(in)compatibilitiesforourinitialvocabulary,andwedeclare,as(3),partiallexicaldescriptionstoaccountforthephenomenaof(2).

(2)PrepComp:*fordog,forwater,forpeople,forChuchoDetHead:thedog,thewater,thepeople,*theChucho(3)Chucho:[cat<n>,max<+>,lxm<Chucho>]dog:[cat<n>,max<–>,lxm<dog>].water:[cat<n>,max<>,lxm<water>]people:[cat<n>,max<>,lxm<people>]

Tryingtosubstitutetheselexemesintothecat<n>positionsinthetwostructures,wefindthatChucho,asapropernoun,startsoutasmax<+>andfitsthesecondpositionforPrepCompbutclasheswiththemax<–>stipulationinDetHead;thenoundogstartsoutasmax<–>andhencefitsthesecondpositioninDetHeadbutclasheswiththemax<+>stipulationinPrepComp;sincebothpeopleandwaterhaveunspecifiedmaximality,eachunifiesintoeitherposition.SinceaproductoftheDetHeadconstructionisitself[cat<n>,max<+>],ittoocansatisfythecomplementpositioninPrepComp,yieldingforthedog,forthepeople,forthewater.

TheleftconstituentinDetHeadisnotlimitedtothe,butinordertoincludethefullsetanewnotationaltoolisneeded.Thepromiscuouswordthefreelyoccurswithanynonmaximalnominal,butmostoftheothersaremoreselective.Weintroduceanewtool,theunificationindex,shownasaGreeklettervalueofaparticularattribute,interpretableonlywhenfoundmorethanonceinasingledescription:themeaningisthatwhatevertheattribute'svaluemaybe,itisthesameforeveryoccurrenceofthatindex.

Followingthetraditionofusingfamilymetaphorsfornamingtree-structuralrelations,wecanrecognizeNP-internalagreementrequirementsbetweensiblings(phrasemates),betweenaparentandagivenchild,andbetweenaparentandallchildren:

Page 95: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 4 of 17

•Siblingagreement:somedeterminersarecompatibleonlywithcertainkindsofnouns.Dogisacountnounandwaterisanoncount(‘mass’)noun,adistinctionrepresentedbyvaluesfortheattributeconfiguration:config<ct>andconfig<nct>.

(4)

a.dog:[cat<n>,max<–>,config<ct>,lxm<dog>]b.water:[cat<n>,max<>,config<nct>,lxm<water>]

TheindefinitesingulardeterminersinEnglishincludeaandsome:a(inourtentativeanalysis)hastheattributeconfig<ct>andsomehastheattributeconfig<nct>.

•Parentandspecifiedchildagreement:somedeterminersshowapolardefinitenessfeature,asdef<+>anddef<–>.Theagreementthistimeisbetweentheparent(thewholeNP)andthedeterminer.Definitedeterminers,includingthe,this,that,these,andthose,makedefiniteNPs.

(5)a.the:[cat<d>,def<+>,config<>,lxm<the>]b.a:[cat<d>,def<–>,config<ct>,lxm<a>]

•Agreementamongtheparentandallthechildren:intheEnglishNP, thisincludesnumber.Somedeterminershavemarkednumbervalues:a,this,andthatarenum<sg>,theseandthosearenum<pl>,theisnum<>.

(6)a.these:[cat<d>,def<+>,config<ct>,num<pl>,lxm<these>]b.some:[cat<d>,def<–>,config<nct>,num<sg>,lxm<some>]

Figure7.3showsthreeunificationindices:definiteness(as“α”)unifiesbetweenthedeterminerandtheNP;number(as“β”)unifieseverywhere;configurationality(as“γ”)unifiesbetweenthedeterminerandthenoun.

WenoticeinFigure7.3thatdefinitenessandnumberareprojectedtotheparentconstituent.ThisagreementfeatureisnecessarybecausedefiniteandindefiniteNPsparticipateindifferentsyntacticpositions.ThereareexistentialcontextsthatclearlypreferindefiniteNPs(Isthereany/*theketchuponmytie?);theformsoffinitebeandpresent-tenseverbsaresensitivetothenumber(amongotherthings)oftheirsubject.Theconfigurationfeaturemustmatchbetweensiblings,butitcontributesnothingtothesyntacticpotentialoftheparentNP.

Figure7.3. Determiner-Headwithagreementindices

Whiledeterminersaffectthesyntacticpotentialsofanonmaximalnominal,modifiers(inthesimplestcase)donot:thephrasesolddog,hotwater,friendlypeoplehavethesamepositionalrolewithintheDetHeadconstructionasthenounsbythemselves. Therecanalsobeselectionalissuesbetweenprepositionsandtheircomplements,butthatisamatterofvalence,treatedinsection7.3.Examplesofsuchdependenciesincludethefactthatbetweenandamongrequirepluralcomplements,thelocationaluseofthespatialprepositionsin,at,andonselectnounsthatnameentitiesthatdifferin‘dimensionality’;andsoon.

3

4

5

Page 96: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 5 of 17

7.2.3UnaryConstructions,a.k.a.Pumping

Inarguingforthedistinctionsjustreviewed,wesaidthatmassnounscouldnotbepluralized,thatcountnounscouldoccuronlywithcertaindeterminers,andwemighthavenotedthatpropernouns,bybeingmax<+>anddef<+>atbirth,aresingularinnumberandcannottakedeterminers.Caseswheretheseobservationsdonotholdareaccountedforbyspecialconstructions.Sometimesawordthatcomes‘offtheshelf’withcertainpropertiescanneverthelessbe‘usedas’awordwithdifferentproperties,andtoaccountforsuchphenomenaweallownonbranchingconstructions—weborrowthewordpumpingfromSign-basedConstructionGrammar(SBCG)(Sag2011;Michaelis,thisvolume)—which‘create’lexemesfromlexemes.Suchconstructionsassumeanasymmetrybetweenthetwouses,thejudgmentthatonesenseofthewordisbasicandtheothersomehowexploitsthatbasicsenseasasemanticcomponent.Intuitivelyitismorenaturaltothinkthatthemeatwecallchickenisbasedonthenameoftheanimalthatprovidesit,thantobelievethattheanimalnameisderivedfromthenameofthefoodstuff.

Thereareunaryconstructionsthatderivecountnounsfrommassnouns:oneofthesetakesnamesofsubstancesandproducesnounsthatdesignateportionsorservingsofthese(Figure7.4,example(7a);anothertakesnamesofsubstancesandproducesnounsthatnamevarietiesofthese(example(7b)).

(7)a.Twobeersplease.b.Proudtobeservingsomefinelocalbeers.

Conversely,thereareunaryconstructionsthatderivemassnounsfromcountnouns:oneoftheseproducesnounsthatnameunboundedquantitiesofthesubstancecomprisingthecountableobject(8),andonespecificallyusesthenamesofanimalswhosefleshcanbe(orcanbeconceivedofas)edible(Figure7.5,example(9)).

(8)You’vegottomatoonyourbeard.(9)Haveyouevertastedbeaver?

Figure7.4. Mass-to-count

Clicktoviewlarger

Figure7.5. Count-to-mass

Thedifferencebetweenactualpumpingconstructionsandetymologicalmodelsisoftenunclear.TheFrench-derivedmeatnames(pork,beef,mutton)showtheresultofjustthiskindofprocessinanotherlanguage,andthemeatsensesofchicken,turkey,fishandlambcancertainlyberegardedasalreadylexicalized;itisonlywhenweusethenamesofseldom-eatenanimalsthatwerecognizethereusabilityofthemodel(beaver,snake,dog).

Aviewthattheseinterpretationsare‘coerced’bythesyntaxisnotvalid,asshownbythefactthateventheunrestricteddeterminertheallowsambiguousinterpretations.Wethinkofcat-substanceinadeterminerlesscontextlike(10a)asaninstanceofcount-to-massconstrual,butthatreadingisallowedevenifthereisno

6

Page 97: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 6 of 17

coercingcontext,asin(10b).

(10)a.Therewascatalloverthedriveway.(cf.Therewasmudalloverthekitchenfloor.)b.Canyouhelpmecleanupthiscat?

Therearealsotwowaystousepropernamesascommonnouns.Oneofthese(Figure7.6,example(11))isbasedonthesemanticsofnaming.Theinnerlexemeisthenameofaspecificcontext-givenentity(e.g.,aperson);theouterlexemedesignatesanythingoftheexpectedtypebearingthesamename.Asecondwaytouseapropernameasacommonnounistousestereotypicalinformationaboutaknownindividualofthenameandrefertoothersassharingthoseproperties(12).

Figure7.6. Proper-to-common

(11)HowmanyMadisonsarethereinthisyear'sfifth-gradeclass?(Ans.:Threegirlsandoneboy.)(12)

a.MyalmamaterisknownastheHarvardofWestTexas.b.There’llneverbeanotherChucho.

7.3.ValenceandComplementation

Grammarsthatusefeaturestructurestendtoorganizethemintohierarchiesornetworks,distinguishingforexample,syntactic,semantic,andphonologicalfeaturesfromeachother,anddistinguishingheadfeatures(thosewhicharesharedbyaphraseanditshead)fromlevelfeatures.Maximalityisalevelfeature,andtwootherpolarlevelattributeshavebeenproposed,lexicality(lex<>)andsubject-requirement-satisfied(srs<>,thelatterdistinguishingverbsorverb-headedphrasesforwhichagf<subj>hasorhasnotbeenaccountedfor).

Inourfirstphrasalconstructions,itwasstipulatedforPrepComp(Figure7.1)thatboththeleftmemberandthephraseasawholeneededtobecat<p>,andforDetHead(Figure7.2)thattherightmemberandthephraseasawholeneededtobecat<n>.Thiscanbestatedmoregenerallybystipulatingthatthecategoryfeatureofalexemeisaheadfeature,automaticallyrequiringthatphrasesbuiltaroundthemareofthesamecategory.ThehierarchyoffeaturesismadepossiblebyallowinganAVMitselftobethevalueofanattribute:thuslevel<max<+>>,head<cat<n>,num<pl>>.Maximality,lexicality,andsrsareobviouslynotheadfeatures,sincetheirpurposeistoshowconstructionsinwhichonevalueisrequiredofachildandanothervalueisrequiredoftheparent.

TheclassofconstructionstobecalledHeadComphavelexicalheads,thephrasesthataccompanythemaretheircomplements,andthepresenceofparticular(kindsof)complementsisdeterminedbyvalencedescriptions;ourtentativePrepCompisaninstanceofsuchaconstruction,inwhichtheprepositionisthelexicalhead.

7.3.1Valence

Thewordvalence,takenfromthetraditionofdependencygrammarsassociatedwiththeFrenchlinguistLucien

Page 98: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 7 of 17

Tesnière(1959),asusedinBCG,includedboththeword'sconceptuallycentralcomplementsandthosethataspeakermaychoosetointroduce.BCGdidnotdistinguishbetweenargumentsandadjunctsinrespecttosyntacticorganization,butinsteadmadeacontrastbetweencoreandperipheralelementsofavalence.

Werefertotheindividualelementsofavalencedescriptionasvalents,anontraditionalbutconvenientterm.Itisusefultorefertosemanticvalents,havinginmindonlythesemanticrolesthataccompanyalexeme,andsyntacticvalents,roughlyanalogoustothesubcategorizationframesofgenerativetheories,butmostofthetimethewordwillrefertothewholepackage.Thecompletevalenceofalexemehasallvalentsfullyspecifiedintermsof(i)semanticrole,labeledhereassr<>, (ii)grammaticalfunction,labeledgf<>,and(iii)phrasetype,labeledpt<>. Insomecasestheoff-the-shelfvalencesetofalexemeentryincludesonlysemanticroles,andthegrammaticalpossibilitiesaredefinedbylexicallydeterminedvalencecompletionprinciples. Averblikegive,forexample,identifiesitsthreecoreargumentsasgiver,gift,andreceiver,or,inthemoregeneralvocabularyusedinBCG,agent,theme,andgoal.Somevalentscanhavesyntacticpropertiesbutnoassignedsemanticrole,liketheitin(13);somevalentscanbesemanticallyrelevantbutunexpressed,aswiththemissingfactin(14).

(13)Ihateitthatyoueatwithyourmouthopen.(14)Shealreadyknows.

Thevalentdescriptionswillcontainsr<null>inthefirstcase,pt<null>inthesecondcase.

Therepresentationofvalenceinformationusestwonewkindsofbracketing.First,thevalenceisrepresentedasasetofelementsenclosedbetweenwavybraces“()”;second,theAVMsfortheindividualvalentsareenclosedinparentheses“()”ratherthansquarebrackets“[]”,sincethelatterrepresentactualconstituents.Thus,usinglate-alphabetcapitalletterstorefertoAVMsofvalents,alexicalheadthatrequiresvalentsXandYwouldcontainthefeatureval{(X),(Y)}.Thevalueofavalenceattributeisasetofelements,andvalenceitselfisaheadfeature:itscontentisrecognizednotonlyatthelexicallevelbutalsoatthe(endocentric)phrasallevel.

Everyverb-valencecontainsonemember,thedistinguishedargument(da<+>),withaspecialsyntacticfunction;allothersareda<–>.Briefly,da<+>marksthevalentrealizableasthesubjectofasimpleactivesentence. Asaconsequence,thedaattributehasamutualentailmentrelationwiththeverb-featurevoice:thevalentcontainingbothgf<subj>andda<+>entailstheverb-formcategoryvoi<act>(‘activevoice’),whereasagf<subj>,da<–>valentrequiresvoi<pass>(‘passivevoice’).

7.3.2ValenceCompletion

Wedistinguishalexicon-internalvalencecompletionprocess,thelinkingofgrammaticalinformationtosemanticvalents,fromasyntacticprocessofvalentrealization,determiningwherephrasesrealizinggivenvalentsoccurintheword'senvironment.Onekindofphrasalconstructioncontainsalexemefollowedbyphrasesthatrealizeoneormorevalentsofthatlexeme.Oneinstanceofsuchaconstructioncouldlooklike[[val{(X),(Y)}][Y]],realizable,forexample,asatethesausage,aVPinwhichoneofthevalentsoftheverbeatappearsasitsdirectobject:theYinparenthesesstandsforspecificationsoftheneededvalent;theYinbracketsisthephrasethatinstantiatesthatrequirement.

Theoff-the-shelfvalencesofsomeverbsspecifyonlysemanticroleproperties;valencecompletionruleswillassignsyntacticformsandfunctionstothem.Forexample,thelexicallyprovidedvalencesofcertainAnglo-Saxonverbsofgiving-showing-tellingspecifyonlythesemanticroles,leavingtheircompletiontothekindsofprinciplestowhichwenowturn.Asampleofsuchaminimalvalence,forshow,isshowninFigure7.7;whereleft-to-rightorderofthevalentsisirrelevant.

Figure7.7tellsusthatthevalenceofshowcomeswithatrioofsemantic-roles,theassociatedcontentofwhichwillbelinkedtothemeaningofthesentence:

agent(sr<agt>),theonewhosetssomethingupforsomeoneelsetosee,theme(sr<thm>),theobjectputondisplay,andgoal(sr<goal>),theindividual(s)intendedtoseethetheme.

7

8

9

10

11

12

13

14

Page 99: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 8 of 17

Example(15)showsrealizationsofthesevalentsintheorderagent,goal,andtheme.

(15)Thesheriffshowedmethephotographs.

Subjectandobject,uniquegfsinanyvalence,arenuclear,andtheothersareoblique. Theformalrealizationofobliquearguments,typicallyprepositional,canbeseparatelydeterminedbyspecificlexemes(dependon,objectto,preventfrom,coaxinto,etc.),butinmanycasestheyareassignedbygeneralprinciples.Inthecaseofthisclassofagt+thm+goalverbs(thegive-show-tellverbs),theobliquerealizationpatternsaregivenin(16).

Clicktoviewlarger

Figure7.7. Minimalvalenceforshow

(16)ObliquenessPrinciplesa.sr<agt>,gf<obl>,pt<p:‘by’>b.sr<thm>,gf<obl>,pt<n>c.sr<goal>,gf<obl>,pt<p:‘to’>

Thenominalobliquetheme(16b)isourinterpretationofthe‘secondobject’intheso-calledditransitivepattern;BCGdidnothaveaconceptofindirectobjectbutinsteadregardedthegoalvalentascapableofbeingsubjectorobject,thethemeasbeingcapableofbeingsubject,object,oroblique. Whenaptvalueisindicatedwithaprepositioninquotes,theinterpretationisthatitisaprepositionalphraseheadedbythecitedpreposition.

Figure7.8representsonesetofpossibilitiesforfillingoutthevalents’values.Everyverbalvalencemusthavejustonesubject,andsinceinFigure7.7allthreegfslotsareempty,itfollowsthatgf<subj>maybeunifiedwithanyoneofthem.Nowifgf<subj>unifiesintoavalentwithda<+>,thevoiceoftheverbmustbeactive.WetakeitasafactofEnglishthatavalentwithsr<agt>isda<+>andallothersareda<–>.AconsequenceoftheassignmentisthatthephrasetypeisNP(abbreviatedaspt<n>).Furtherspecificationsallowgf<obl>tobeassignedtosr<goal>,andgf<obj>,pt<n>tobeassignedtosr<thm>.ThisisshowninFigure7.8:

Whenthesimplestvalent-realizationprocessesareappliedtothepossibilitiesjustoutlined,wefind,usingA,T,andGtorefertotheparticipatingNPsandexpressingtheoptionalityoftherealizationofobliqueagent,thefollowingpatterns:AshowedTtoG;AshowedGT;TwasshowntoG(byA);GwasshownT(byA).

InthecaseofcertainLatinateverbsinthesamesemanticclass,theobliquegoalisobligatorilyprepositional,markedwithto;thatleavessimpleactiveandsimplepassiveastheonlypossibilities.Thissubclassincludesdonateandcontribute,exhibitanddemonstrate,explainandreport.Thepredictedresultsareseenin(17),withoneexamplefromeachofthegive-show-tellclasses;passivesarenotexemplified.

(17)a.‘giving’:Icontributedsomemoneytothecause.(*Icontributedthecausesomemoney.)b.‘showing’:Idemonstratedthetechniquetomyassistants.(*Idemonstratedmyassistantsthetechnique.)c.‘telling’:Ireportedtheincidenttothesheriff.(*Ireportedthesherifftheincident.)

15

16

17

Page 100: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 9 of 17

Clicktoviewlarger

Figure7.8. Fully-specifiedvalenceforshow

Someotherverbsinthissemanticclasspre-assignwithtothethemebutleavegf<>unspecified.Anexamplewithplyisgivenin(18a).Theverbblame–letusforceitintothisclass–hasagrammaticallyemptyminimalvalencebutithasuniqueobliquenessassignments,onandfor,seenin(18b–c).(Passivesarenotexemplified.)

(18)a.Theypliedmewithgifts.(*Theypliedgiftstome.)b.Theyblameditonme.(Itwasblamedonme.)c.Theyblamedmeforit.(Iwasblamedforit.)

Thesequencesofoperationsjustsuggestedwereforeaseofexpositiononly.Itwouldhavebeenpossibletobeginbyassigninggf<obj>toanythingbutda<+>,requiringvoi<act>,leavingonlytheagenttobecomethesubject.Itwouldnotbepossibletoassigngf<obl>toeachvalent,becauseeveryvalencehastohaveasubject. Thepointisthatallvaluesmustbeprovidedandallcompatibilityconstraintssatisfied:everygrammaticallyinterpretedandpronounceablelinguisticentitymusthaveallvaluesfilledin.

7.3.3ValenceModification:PumpingRulesforVerbs

Inthenonbranchingconstructionsshownfornouns,achildlexemehadameaningwhichwasacomponentoftheparent'smeaning.Asimilarsituationexistsforverbs,inwhichthemeaningofthechildisasubeventoftheeventdesignatedbytheparent.Thephenomenonisoftenthoughtofasattributingtoaparticularsyntactictemplateameaningofitsownandthenclaimingthatwhenaverbofaparticularsemanticclassappearsintheverbalslotinthattemplate,thesurroundingstructureitselfinducesorcoercesthereinterpretation.TheBCGapproachwastoassignsuchphenomenatothelexiconandproductivevalencealternations.Theremaybecasesinwhichthemotherlexemefitsonlyinsidearigidfixedexpression,butweneedaprocessthatallowsforthosecasesinwhich,say,passivization,extraction,interruption,etc.,arealsopossible.Paradeexamplesare(19)–(21):

(19)Shesneezedthefoamoffthecappuccino.(sneezeisnotatransitiveverbmeaningcause-to-move)(20)Islippedthewaiterafifty-dollarbill.(slipisnotaverbofgiving)(21)Heranhisshoesragged.(runisnotacausativetransitiveverb)

Thesrpatternsthataremostlikelytooccurinthevalenceoftheresultingverbincludeagent+goal+theme,withthe‘recipient’interpretationofgoalandwiththemeas(gf<obl>,pt<n>),oragent+theme+goal/source/result,withgoalasgf<obl>(forcingthethemetobesubjectorobject).

Figure7.9isanexampleofaverb-pumpingconstructionthatassignsasenseofcompletenessforthesemanticsof‘placing’verbsthattakegoalobjects.

Clicktoviewlarger

Figure7.9. AVerbPumpingconstruction

Theparentconstructionhasanewminimalvalenceandisnowavailableforvalencecompletion.Fortheoriginal(inside)verb,thegoalwasoblique,andthelinkingpossibilitiesareavailableonlyfortheagentandthetheme;fortheparent,thethemeisoblique,markedwiththeprepositionwith.

18

19

Page 101: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 10 of 17

7.4.ValentRealization

Thevalencecompletionprocessesassignedsyntacticexpectationstoverbvalents.Knowingthatsomethingisthesubjectofaverbisnotthesameasknowingwhereitwillappearinasentencebuiltaroundtheverb.Oneexample,illustratedin(15),isthattheverbcouldbefinite,andthesubjectconstituentcouldprecedeitinthesentence.Anotheristhatthesubjectcouldbeunexpressedintheverb'simmediateVPandaccountedforinsomeotherway,byacontrolrelationorbybeingextractedbeforeanembeddingexpression.Ortheverbcouldbeapartofaconstructioninwhichthesubjectisobligatorilyunexpressed,asinanimperativeconstruction.

7.4.1OvertValentRealization

Figure7.10. Head-complementconstruction

Figure7.11. Inversionconstruction.Theconstructionsthatinheritthispatternmightspecifythattheverbisanyauxiliary(foryes-noquestions),aparticularauxiliary(mayincursesandblessings,should,had,orwereininversionconditionals),specificnonauxiliariesinmoreidiomaticpatterns(ComestheRapture),andsoon.

Overtvalentrealizationcanbeaccountedforwithintwogeneralpatterns.Inoneofthese,aframe-bearinglexicalitemisontheleft,andoneormoreconstituentstoitsrightareitsvalent-realizingrightsiblings.Onesubcaseissimplecomplementation,availableforallpartsofspeech:adjectives(fond—ofchildren,eager—tosucceed);relationalnouns(top—ofthecolumn,overture—totheopera),eventnouns(decision—toretire,argument—overmoney),prepositions(into—thecave,by—eatingit),andverbs(decide—toretire,persuade—hisparents—totakehimback),andsoon(seeFigure7.10).Inthecaseoffiniteinstancesoftheclassofverbscalledauxiliaries,thevalence-bearingelementcanbeamodal,tense,oraspectauxiliaryandthecomplementscanbethesubjectalone(did—he?),thesubjectpluscopularcomplements(was—he—here—thismorning?),orthesubjectoftheauxiliaryanditsVPcomplement(have—they—finishedtheirwork?)(seeFigure7.11).

Inasecondpattern,whichalsohastwosubcases,aconstituentontheleftsatisfiesavalence-bearingphraseontheright(Figure7.12).Thefirstcanbethesubjectdeterminedbythefinitelexicalheadofthesecondconstituent(thechildren—disappeared,I—opposedyournomination);thesecondcanbeaconstituentthatsatisfiesavalencerequirementoftherightsister,or,recursively,avalencerequirementof(avalencerequirementof)*therightsister.

ThecollectionoffeaturesX satisfiestheneedsofbeingthesubjectofthevalenceofitsrightsibling.Thatrightsiblingcouldbeasingleverb(sheleft),oraVP(sheleftmeforasociologist),sincevalenceisaheadfeature.Therightsiblingneedsasubject(srs<–>)andthephrasehasitssubject(srs<+>).Beingsrs<+>doesnotmeanthateveryvalentissatisfied,soaconstituentresultingfromthisconstructionisnotyetnecessarilyreadytostandonitsown:thatrequireseveryAVMtobecomplete.

Subject-verbagreementishandledatthislevel;theverbform(forBEandallpresent-tensenonmodalverbs)hascompatibilityrelationsbetweencertainpropertiesofthesubjectvalent:personandnumber.Thecaseformofthesubjectisdependentonthegfvalue.

20

i

Page 102: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 11 of 17

Clicktoviewlarger

Figure7.12. Subject-predicateconstruction

Clicktoviewlarger

Figure7.13. LeftIsolationconstruction

Figure7.13isthegeneraltemplateforafamilyofconstructions,eachofwhichcanhaveadditionalconstraints.Theleftsistercanbeonetypeofwh-wordinthecaseofinterrogativeclauses,anothertypeinthecaseofrelativeclauses;oritcanbeanykindoftopicalizedorfrontedconstituentinavarietyoftopicalizedsentences.Thenotationwithval*isanexcuseforaformalwayofidentifyingarecursiverelationofavalentof(avalentof)*,allowingtheleftsiblingtobeanargumentofthehighestverb,orofanyvalence-licensedembeddedverb.Constraintsonlong-distancedependenciesare(partially)accountedforbyafeaturesealed,whichisequivalentto“AVMcomplete”:allinternalfeaturesaresatisfied.

7.4.2CovertAccountingforValents

Syntacticinstantiationofvalentsinapositiondefinedbythelexemewhosevalencetheybelongtoistheworkofthekindsofrulesjustseen;buttherearealsosituationsinwhichspecificvalentsarenotformallyexpressedatall.Inonetype,calledcoinstantiation,aspecificvalentinanembeddedcontextisco-indexedwithaspecificvalentoftheimmediatelyhighercontext.Inthesecondtype,nullinstantiation,theomissionislicensedbyagivenconstructionorbyitshostvalence.

Withcoinstantiation,orargument-sharing,theindexingisbetweentwovalentdescriptions,andthevalentofthedependentstructureistheonethatbecomespt<null>.ThegeneralstructurelookssomethinglikeFigure7.14.

Thevalenceofagivenlexemeincludesamemberthatisitselfavalence-bearingentity.Thesymbol“A”inthefiguredenotesthehostlexeme,“x”isthevalencememberthatparticipatesinthecoindexing,“B”standsforthephrasetypeoftheembeddedconstituent,and“y”isthevalentthatisunexpressed.Inallcases,thevalentxofAisindexedtotheunexpressedvalentyofacomplementofphrase-typeB.AsmallcollectionofexamplesisfoundinTable7.1.

Clicktoviewlarger

Figure7.14. Co-instantiationconstruction

Thesecondpossibilityfornoninstantiationofavalentisforits“zero”realization,licensedbecauseoftheneedsofagrammaticalconstruction(constructionalnullinstantiation),orbyagivenlexeme;forthelatterwedistinguishdefiniteoranaphoricnullinstantiation(DNI),andindefiniteorexistentialnullinstantiation(INI).ExamplesofeachtypeareshowninTable7.2.Therelevantvalencerepresentationwillshow[…,pt<null>],whichmeansthatthevalenthasbeensatisfiedandnothingisneededtolinkittoanythinginthesurroundingtext.Thesemanticsassociatedwiththe“null”valuewilldependonthekindofwordandsense.

Page 103: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 12 of 17

Table7.1.Sampleswithcoinstantiation

1.Itriedtoquit. SubjectxoftryAisindexedtosubjectyofitsVinfcomp.

2.Hekeptmewaiting. ObjectofkeepisindexedtothesubjectofitsVingcomp.

3.Shemademequit. ObjectofmakeisindexedtothesubjectofitsV-comp.

4.Itwaseasytoread. ThesubjectofeasyisindexedtoanonsubjectofitsVinfcomp.TheexperiencerofeasyisindexedtothesubjectofitsVinfcomp.

5.Itisfuntoswim. TheexperienceroffunisindexedtothesubjectofitsVinfcomp.

6.Ifounditscary. Thesubjectoffindisindexedtotheexperiencerofscary.

7.Itmeritsconsideration. Thesubjectofmeritisindexedtothe“object”ofitsNPcomp.

8.Imadeapromise. Thesubjectofmakeisindexedtothe“subject”ofitsNPcomp.

Note:Vinf:infinitivalVP;V-:bareVP;Ving:gerundialVP;pseudo-gfsarenamedhereforthevalentsassociatedwitheventnouns;unexpressed‘experiencers’arevalentiallyassociatedwithfun,easy,interesting,andsoon.Eachdecisionhereisquestionable,buttheintendedsemanticvalent-chaininglinksthesamereferent,inTheyexpectedmetofinditfuntoreadthis,totheobjectofexpect,thesubjectoffind,theexperienceroffun,andthesubjectofread.ThetraditionaldistinctionsbetweenEquiandRaisingarerecognizedintermsofthepresenceorabsenceofsemanticrolesinthehighervalent.

7.5.Synthesis

Aconstructionisapartialdescriptionofasetoflinguisticexpressions,uponwhichmeaningsarebuilt,whoseformorinterpretationcannotbeexplainedintermsoftheotherthingsweknowaboutthelanguage.AsequenceofwordslikeShelovesmecannotbetreatedasaconstructiondeservingaseparatedescriptionbecauseeverythingweknowaboutsuchasentence,includingthecaseformsofthetwopronouns,canbeexplainedbywhatweknowaboutvalence,twopatternsforvalentrealization,grammaticalfunctionandthecaseformsofpersonalpronouns,etc.(thoughseeBybee,thisvolumeforadifferentviewofconstructionalstorage).Thesemanticinterpretation—unfortunatelyignoredinthischapter—islinkedtothemeaningoftheverbandthereferentialpossibilitiesofthearguments.

Thiswasbehindtheideathatidiomsandevenlexicalitemsareconstructions:wecannotfigureoutwhatwatermeansbyexaminingitscomponents:ithasto

Page 104: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 13 of 17

Table7.2.Examplesillustratingnullinstantiation

Example Comment

1.Trytoliftthis.(Imperative)

Apartofthedescriptionofthemainimperativeconstructionisthespecificationofavalentas“sr<agt>,gf<subj>,pt<null>”.

2.Holdawayfromyouandopencarefully.

Inrecipesandinstructionaltextsobjectsthatareinfocusmaybeomitted.

3.Stuffwithbreadcrumbs.4.Stuffintoamediumbellpepper.

Verbswithvariablevalenceselecttheversionwithobjectasomissibleinrecipes,etc.

5.Welost?Whowon?

Theomissibleargumentwithwinandloseconcernsthecontest,nottheprize.

6.Sheknows. Theverbknowthatmeans‘knowafact’(savoir,wissen)hasanomissiblecomplement;theknowthatmeans‘knowaperson’(connaître,kennen)cannotloseitsobject.

7.We’vealreadycontributed.

Withcontribute,donate,etc.,thesr<thm>isINI-omissible,thesr<goal>isDNI-omissible.

8.WehadtoreplaceHarry.9.WehadtosubstituteLouise.

LouisetookoverHarry'splaceontheteam.ThemissingwithLouisewithreplaceisINI;themissingforHarrywithsubstituteisDNI.

10.Somemembersobjected.11.Werethereanywitnesses?

Withmanyrelationalnouns,therelatumisDNIomissible.Membersofanorganizationweknowabout;witnessestoanincidentwe’vebeentalkingabout

12.Othermembershadasimilarobjection.

Members(oftheorganizationweallknowabout)(whoarenotthememberswe’vebeentalkingabout)hadobjections(totheproposalwe’vebeentalkingabout)whicharesimilar(totheobjectionthathasjustbeenmentioned)

beknowninitself;wedonotknowthatletaloneisevensayable,letalonehasaninterpretation,onthebasisofwhatweknowabouttheindividualwordsortheircombination.Butthestructurethatallowsthejuxtapositionofsubjectandpredicate—NPandVP—inthesentenceShelovesmedoeshavetobeseparatelystatedindependentlyofeverythingelse.InthefollowingparagraphsIwilltrytoshowhowsomeEnglishnoncanonicalconstructionsmightbehandledusingthetoolsandprinciplesintroducedinthischapter.

1.Wehaveseensituationsinwhichpositionsthatwelcomeorrequireparticularkindsoflinguisticexpressionsaredefined,andsituationsthatimposeconstraintsoncertainkindsofentitiesforwhich

Page 105: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 14 of 17

appearanceinsomestructuralpositionisindependentlyprovided(e.g.,valenceandvalencecompletion).Manyphenomenathatatfirstappeartobe‘position-defining’haveturnedouttobevalence-defining,allowingtheovertrealizationofelementstobehandledbyphrasalconstructions,andthisispossiblebyexploitingnonspecification.Considermaximalitywithprepositions.Prepositionsthattakeobjectsaremax<–>inthatcontext;whenapreposition'sobjectisleft-isolated(whodidyougiveitto?)itneedstobemax<+>sothatitcanbeasyntacticvalentforitsverb.Aprepositionthatallowsbothpossibilitiesis,inthelexicon,max<>.Nothingpreventsusfromlettingcat<p>,max<+>,withnovalence,represent“particles,”suchastheawayofputitaway.Someidiomswillallowamaximalitydistinctionintheprepositions.RegardingthelexemicidentityofanNPtobeaheadfeaturemeansthatidiomslikeatrisk,indanger,andinahurrycanbeseenasexhibitingthesamecollocationsasatgreatrisk,inconsiderabledanger,andinagreathurry.Assuchtheidiomscanbeexpressedvalentially,intermsofthelexicalvalueoftheheadsoftheircomplement:animportantdifferenceisthatinwithatrisktheprepositioncannotbestranded,soithastobemax<–>butthementionedidiomswithincanbemax<>:whatdangerarewein?,howbigahurryareyouin?aregrammatical;*whatkindofriskareweat?isnot.2.Therearealsocertainphrasalconstructionsthatallowthelexemicidentityofadependentconstituenttobeprojectedtotheparentphrase.Thisistrueofnominalstructuresoftheformnoun +of+noun .Thenounsthatappearinfirstposition,calledtransparentnouns,tendtorepresenttypes(type,kind,sort,variety,etc.),ormeasures(pound,stack,stick,etc.),orepithets(idiot,fool,jackass,etc.),asevidencedbywhatsortoftroubleareyouin?,whocouldwreakthatamountofhavoc?,myidiotofabrother,honoringtheexpectedpairingsofintrouble,wreakhavoc,mybrother.3.Byregardingvalencesasheadfeatures,wecandescribeverbo-nominalstructures,withsupportverbs,asacquiringsemanticvalentsfromthevalencesoftheirdependentnouns,andtheseinturnfromverbalsources.Thedependentelementsintakeadecisiontojointhearmy,haveafightovermoney,sayaprayerforpeace,etc.,arenotvalentsfortheirverbalcomponents.4.WhatKayandFillmore(1999)calledthe“What'sXdoingY?”construction(Whatwereyoudoingwalkingintherain?)looksatfirstlikeaphrasalortemplaticconstruction,andthejuxtapositionofthesecondarypredicate(thingslikewithmyshoeson,inmykitchen,walkingintherain)totheinterrogativeclausecanprobablyonlybedescribedinsuchmanner.Butthefirstpartcanbeamatterofvalence,leavingthesubject-verbordertosyntax(whatwasshedoingvs.Iwonderwhatshewasdoing),aswellastheinitialpositionofthewordwhat.Theheadverbisbe,itscomplementmustbetheverbdowithpt<Ving>,doingmusttake(cat<n>,gf<obj>,pt<n>,lxm<what>),andmustbemarkedmax<+>.Thismeansthatdoingcannothavealocalobject,soleftisolationoftheWHelementistheonlypossibility.Independentconstructionsallowinterrogativewordstobeextendedwithachangingsetofexclamatives,allowingwhatonearth,whatthehell,whatinheaven'sname,etc.5.Unificationindicesareusefulforidentifyingreferencesacrossconstituents,foridentifyingfeaturesinagreementpatterns,andforlinkingvalentdescriptionswithrealizedvalents;buttherearekindsofdependenciesotherthanidentitythatcomeinsystematicallyrelatedpairs.InBCG,auxiliariesaretreatedasembeddingverbs,andtheirvalencessharethesubjectvalentoftheircomplements:thesubjectsofprogressivebe,passivebe,haveandthemodalsarerespectivelythesubjectsoftheircomplementverbinshewasrevisingthemenu,themenuwasrevised,shehasrevisedthemenu,shemustrevisethemenu.Thisidentitycanextendthroughchainedauxiliaryverbs,asinshemusthavebeenrevisingthemenu.Thesemanticintentioncanusually bebuiltupcompositionally,butinsomecasesparticular‘chainings’ofsubject-sharingauxiliary-verbcomplexesdefinesyntactic-semanticpropertiesthatmustbeseenasunitarywholes.Thisisespeciallyobviousinthetreatmentofcorrespondingprotasisandapodosisclausesinconditionalsentences(Fillmore1986).Examplescanbeseenin(22),withsuggestionsontheirinterpretation:(22)

a.simplecondition:IfIsithere,Iam(always)abletoseetheparade.b.conditionalprediction:Ifyousithere,youwillbeabletoseetheparade.c.hypotheticalcondition:Ifyousathere,youwouldbeabletoseetheparade.d.counterfactualcondition:Ifyouhadsathere,youwouldhavebeenabletoseetheparade.

Thesystemisrifewithcomplexities,relatingtopragmaticintention,verbsemantics,morphologicalvariations,andconstructionalirregularities, butitisneverthelessclearthatwhentheprotasisisleft-isolatedover

1 2

21

22

Page 106: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 15 of 17

embeddingverbsofthinkingandspeaking,thedependenciesremain,suggesting,perhaps,thattheif-clausemustbeenseenasaconstruction-definedvalentofthemainclause.Insentence(23),theif-clauseisnotsubordinatetotheverbtell.(23)Ifyouhadsathere,theytellmeyouwouldhavebeenabletoseetheparade.6.Thereare,ofcourse,phrasalconstructionsthatcannotreadilybeseenasshowingalexicaldependencyrelationbetweentheparts.OneoftheseistheRateConstruction,builtoftwoadjacentNPs,thefirstofwhichexpressesaquantityofonekindofmeasuringunit,thesecondofwhichidentifiesameasuringunitofadifferentkind,andthemeaningofthewholeisderivedfromaratioofthetwoquantities,buildingconceptslikespeed(twentymilesanhour),wages(twentydollarsanhour),mileage(twentymilesagallon),frequency(twentytimesanhour),growth(twoinchesayear),andsoon.Thestructuresproducedbytheseserveaspredicatesoradverbialmodifiersinclauseswheretheirsemanticimportunifieswiththesemanticsoftheirhost.Thetwocomponentsmustactuallybedefinedaccordingtotheirsemanticpurposesratherthanbytemplatesoftheirown:thenountimes,inthe“frequency”example,ispartofamultiplesfamilyofexpressionsthatincludesthesinglewordsonceandtwice;thesecondphraseisfrequentlyanindefinitesingularNP,butoccasionallyitsfirstpartcanbe(to)theorper,andthephrasecanbeoccupiedbythesinglewordseachorapiece.Inshort,itisatwo-partphrasaltemplateinwhicheachcomponentissemanticallydefined.Importantly,itisdistinctfromcompositionalwaysofsettingupthesameratio:insteadofregardingIearntwohundreddollarseveryweekasaninstanceofthisconstruction,theeveryweekshouldbeanalyzedasatemporalvalentforearn,onethatallowsfronting,asin(24a)comparedwith(24b).(24)

a.EveryweekIearntwohundreddollars.b.*AweekIearntwohundreddollars.

7.Unaryconstructionsarenotonlyforlexeme-to-lexemedefinitions.ThereareseveralconstructionsthatofferwaysofcreatingmaximalNPsbycombiningthedefinitedeterminerwithanadjective.Oneoftheseisillustratedwiththepoor,therich,theyoung,theobese.Thesecondelementisclearlyanadjective,ratherthananounthatlookslikeanadjective,sinceitcanbemodified:considertheveryrich,thepatheticallyobese.TheconstructionhasthesinglelexicalitemtheintheleftpositionandtherightpositionhasanAPcapableofdefiningaclassofhumans.TheparentisagenericNPthataddstothemeaningoftheAPthefeatureshumanandplural.Thepoormeans‘poorpeopleingeneral’.8.Anotherunaryconstructioncreatesasymmetricpredicatefromapluralizednonmaximalnominalifitnamesasymmetricinterpersonalrelation,likefriendandneighbor,andtheresult,likeothersymmetricpredicates,allowsboththejointanddisjointrealizationoftheparticipants,asseenin(25).(25)

a.Iwasfriendswithhimincollege.b.HeandIwerefriendsincollege.

Thechildconstituentdoesnotneedtobeasingleword:anyphrasethatnamessucharelationshipispermitted,asseenin(26).(26)

a.Iambestfriendswiththenewmayor.b.Iwasoncenext-doorneighborswithShirleyTemple.

Thenewunithasavalencepatternsimilartoanadjectivelikesimilar,butitspecifieswithastheprepositionmarkingthesecondparticipant,independentlyofthevalenceofthenounitself.9.BCGsawdegreequalifiersasawayofcreatingaphrasalunitthatacceptsthevalenceofitsheadbutaddsanewone:thedegreequalifierstoo,enough,andsoaugmentthevalenceoftheadjectivestheyjoin.Ifapredicateadjectivecombinedwithtooorso,toitsleft,orwithenoughtoitsright,theresultwasacomplexunitthataddsanewvalent:[tooold]tojointhemarines,[oldenough]toknowbetter,[soeager]tofightthatthemarineswon’ttakehim.

Thedescriptionspresentedinthischapter,theproductoftheauthor'sattemptstorememberfromscatterednotesandwritingswhatwewereworkingtowardsomedecadesago,andtofigureoutonceagainhowitmightwork,willsurelynotmapwelltothenotebooksofanybodywhotookthecourse.Thephotocopiedtextbookusedduringtheperiodofthiscoursewasconstantlychanging,soItriedtolimitmyselftowhataparticularsetoftheoreticalcommitmentsmighthaveledto,imaginingateachpointhowmycolleaguesandstudentsmighthaveobjected,and

23

Page 107: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

Page 16 of 17

regrettingtheinabilitytoworkitthroughwithinacommunity.Thischapterfailedtoincludethereshapingoffeaturestructureswiththeuseoftypedfeatures,andsubstantivelyithadtoignoreconjunction,ellipsis,andconstituentorderissues.IhavenotbeenworkingonConstructionGrammarforsometime,butwithonegoodyouthpillIwillgladlyturnbacktoit.

TheauthorisindebtedtoRussellLee-Goldmanformuch-neededhelpinthefinalstagesofcompletingthischapter,and,morelongterm,toPaulKay,LauraMichaelis,andIvanSagfordozensofsplendidandfrustratinglunch-timeconversationsoverthedecadesaboutwhataConstructionGrammaroughttolooklike.

Notes:

(1.)TheauthorisindebtedtoRussellLee-Goldmanformuch-neededhelpinthefinalstagesofcompletingthischapter,and,morelongterm,toPaulKay,LauraMichaelis,andIvanSagfordozensofsplendidandfrustratinglunch-timeconversationsoverthedecadesaboutwhataConstructionGrammaroughttolooklike.

‘Formal’hereincludessyntactic,morphological,andphonologicalform;‘semantic’includespragmaticsandconventionsofusage.

(2.)VerticallystretchedsquarebracketsareouralternativetotheclosedrectanglesusedinearlyBCGwork.NotethatAVMsinsideaparagraphwillbewritteninsmallcaps;infiguresorwhenisolatedfromthetexttheywillbeinnormalfont.

(3.)Thefullstoryofsomecallsonacumbersomepolarfeaturecalledboundednessthatincludednoncountsingularsandcountplurals,eachbeingbnd<–>,allowingbothsomepeopleandsomewater.

(4.)Inseriouscase-markinglanguages,thiswouldincludecase.

(5.)Bothmodifiedandunmodifiednounsagreeinthemaximalityvalueofthenoun;BCGdidnotsucceedindefiningalogicfornonspecificationthatpermitsthisgeneralization.

(6.)Ineachcaseofpumping,bothfornounsandforverbs,thereisthequestionofwhethertheyareabsolutelyfullyproductiveconstructions,operatingfreelyonanythingthatmeetstheinputconditions,orwhethertheyare‘merely’aconventionalmetonymicpatternwhosehistorywerecognizeinthecreationofexistinglexicalitems—the‘patternofcoining’discussedbyKay(thisvolume).Forourpurposeswecanrecognizetheexistenceofthepattern,as‘motivating’thederivedlexeme,independentlyofmakingthisdecision.

(7.)Inadditionto‘core’and‘periphery’(Tesnière'sactantandcirconstant),therearealsoconstructionsthatintroduceintoclausesvarious‘extrathematic’subordinatestructuresthatarenotdirectlyapartofthesemanticframeofthesyntacticheadoftheclause.

(8.)InearlierworkthiswastheGreekletterθ.

(9.)Theptattributeisintendedtocovereveryaspectofthesyntacticpropertiesofthevalent;inthesimplestcasesthesearesimplymaximalXPs,butinthecaseofverbalandclausalcomplements,theconditionscanbeextremelycomplex.Thereaderisinvitedtoworkthroughthedifferencesintheclausalcomplementsofknow,think,demand,hope,andwish.

(10.)Thesearethe‘linkingrules’insometraditions.

(11.)InSBCGthenotation"<>"denotestheemptylist;itdoesnotacknowledgeunspecifiedvalues.InBCGablankbetweenanglebracketsmeansthatthevaluecanunifywithanylicensedvalueinthatposition,soanexplicitlyabsentvaluemustbespecifiedas<null>.

(12.)SBCG,bycontrast,projectstoahigherconstituentinformationaboutthevalentsthatneedtobephrase-externallyinstantiated.

(13.)TheBCGtraditionincludedfamiliargeneralizationsontheassignmentofda<+>status,basedonasemanticrolehierarchy,butthatwillnotbediscussedhere.

Page 108: The Oxford Handbook of Construction Grammar

Berkeley Construction Grammar

(14.)ThenotationinthetoplineofFigure7.7,withss|syn|head<…>,presupposesthehierarchicalarrangementsoffeatures,ignoredinthischapter.Onebranchof“syntaxandsemantics”(ss)is“syntax”;onebranchofsyntaxis“category”;theverticalbarleadsasinglepaththroughthenetwork,ignoringalternatepaths.

(15.)Weuse“comp”asthegfforverbalandclausalcomplements.

(16.)Othersituationswherenominalobliquearecalledforofferprepositionalalternatives.We’retalking(about)bigmoney,fly(on)United,shop(at)Macys.Iworked(for)twohours,Wepresentedthem(with)theirmedals.

(17.)Detailsnotdiscussedherewillincludespecificationoftheverbformsassociatedwithpassivevoice,theassignmentoffinite<–>tothepassiveVP(requiringthatthephraseoccurinapositionsupportedbyalicensingcontext,oneofthembeingtheverbbe),caseassignmentsthataffectpronominalvalents,andsoon.The‘optionality’oftherealizationofanobliqueagentalsofallsunderthetopicsdiscussedin7.3.2.

(18.)Inanearlyformulation,greatlyfavoredbyme,thegeneralizationthateveryEnglishverbhadtohaveasubjectwasexpressedbyautomaticallyincludingineveryverbalvalencesetamemberwiththeproperty“(sr<>,gf<subj>,pt<>)”.Inorderforfeaturecompletiontooccur,thismemberofthesetsimplyhadtoblendwithsomeothermemberoftheset,unifyingitselfwithavalentthatlackedwhatithadandhadwhatitlacked.Settheoryspecialistsshouted“Impossible!”

(19.)Itisconceivablethatthealternationsin7.3.1couldalsobedealtwithaspumpingratherthanmerelyvalencecompletion,shouldthegoal-as-objectversionbeunderstoodashavingadifferentsensefromtheoriginalmeaning.

(20.)BCGdidnothave“slashcategories”;further,BCGdidnothaveadeterminativewaytocheckthewell-formednessofananalysis.TheambiguityofasentencelikeWhendidyoudecidetoretire?istobeunderstoodasthedifferencebetweenindexingthewhentoaperipheralsr<time>valentofeitherretireorofdecide.

(21.)‘Usually’becausetherearecasesofinteractionsbetweenmodalmeaningandnegation,distinguishing,forexample,mustn’tandneedn’t.

(22.)For(22b)considershouldsit;for(22c)weretosit;for(22d),would’vesat,had’vesat.

(23.)ThisanalysisisexplicitlyrejectedinKayandSag2009.

CharlesJ.FillmoreCharlesJ.FillmoreisProfessorEmeritusofLinguisticsattheUniversityofCaliforniaatBerkeley,anddirectoroftheFrameNetprojectattheInternationalComputerScienceInstitute,alsoinBerkeley.Hiswritingshaveincludedcontributionstogenerativesyntax(1963),theintroductionofagrammaticalframeworkinwhichsemanticroles(‘deepcases’)werecentral(1968),therecognitionoftheimportanceofdeicticelementsinlinguisticdescriptions(1971/1997),suggestionsforfoundinglexicalmeaningson‘semanticframes’(1982,1985),andassociatedcontributionstopracticallexicography(1992,withB.T.S.Atkins),andcontributionstoagrammaticaltheorybasedonconstructions(1988andpresentvolume),withseveralcolleagues.HewasnamedFellowoftheAmericanAcademyofArtsandSciencesin1984,andheservedasPresidentoftheLinguisticSocietyofAmericain1990.

Page 109: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0008

Sign-basedConstructionGrammarLauraA.MichaelisTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterdiscussestheconceptofSign-BasedConstructionGrammar(SBCG),whichevolvedoutofideasfromBerkeleyConstructionGrammarandconstruction-basedHead-DrivenPhrase-StructureGrammar(HPSG).TheleadinginsightofSBCGisthatthelexiconprovidesamodelforthesyntax-semanticsinterface.ThechapterexplainsthatthoughSBCGcannotbedivorcedfromtheformalconventionsitusestorepresentlexemes,constructions,andthehierarchicalrelationsamongtypes,itoffersinsightstoconstructiongrammarianswhoseworkisnotprimarilyformal.ItalsoconsidersthestrictlocalityconstraintofSBCG,theavoidanceofovergeneralization,inheritance,aswellasthetreatmentofinflectionalandderivationprocesses.

Keywords:SBCG,BerkeleyConstructionGrammar,HPSG,syntax-semanticsinterface,lexemes,hierarchicalrelations,inflectionalprocess,derivationprocess

8.1.Introduction

Afoundationalassumptionoftraditionalgenerativegrammaristhatagrammarisorganizedinthemindofthespeakerasanumberofhermeticallysealedmodules,whichinthecourseofaderivationhandoffdataonetotheother.Sign-BasedConstructionGrammar(SBCG)(Sag2010,2012;Michaelis2010)liketheframeworksthatinspireit,BerkeleyConstructionGrammar(BCG)(FillmoreandKay1993;MichaelisandLambrecht1996;KayandFillmore1999;MichaelisandRuppenhofer2001;Fillmore,thisvolume),andconstruction-basedHead-DrivenPhrase-StructureGrammar(HPSG)(GinzburgandSag2001),assumenosuchmodules,butratherthatgrammarisaninventoryofsigns—complexesoflinguisticinformationthatcontainconstraintsonform,meaning,anduse—andthatconstructionsarethemeansbywhichsimplersignsarecombinedintomorecomplexsigns.Thenotionofconstruction,onthisview,isaformalization,inaconstraint-basedarchitecture,ofthenotionofconstructionintraditionalgrammar.Asimpleillustrationofaconstructionisthesubjectlesstaggedsentenceshownin(1):

(1)Kindahasaniceringtoit,doesn’tit?

Inasentencelike(1),asobservedbyKay(2002a),themissingsubjectofthemainclausecanonlybeinterpretedonthebasisofthereferenceofthetag'ssubject.Thatis,theaddressee(s)of(1)mustdeterminewhatitreferstoinordertoreconstructthemissingfirstargumentofthemainclause.Whileidiosyncratic,thebiclausalconstructionlicensing(1)sharespropertieswithmoregeneralconstructions.Thatis,thisconstructionevokessignsthatarelicensedbyotherconstructions,includingtheconstructionthatlicensesaquestiontagofoppositepolaritytothemainclauseandispronouncedwithrisingintonation.Theparticularcombination,arrangement,andinterpretationoftheseinherited,construction-licensedsignsis,however,uniquetosentencesofthisform:amainclausemissingasubject,followedbyaquestiontagwhosepronominalsubjectprovidesthereferenceofthemissingsubjectofthemainclause(Kay2002a).Generalizationsaboutconstructionsarecapturedthroughtheinteractionofahierarchicalclassificationoftypesandthetype-basedinheritanceofgrammaticalconstraints(Sag2010).

Page 110: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 2 of 14

Theapproachdescribedhereisdistinctfromtransformational-generativegrammaranditsextensions,asitisadeclarative,constraint-basedmodelratherthanaderivationalone,anditdoesnotassumemovement,underlyingstructure,oremptycategories.Inaddition,SBCGassociatessemanticconstraintsanduseconditionsdirectlywiththephrase-structurerulesthatdefineconstructions,ratherthanrequiringinterpretationstobe‘readoff’syntacticrepresentationorrelegatedtoa‘pragmaticcomponent.’Thatis,phrase-structurerulesinSBCGmeanthingsinthesamewaythatthewordsdo—asamatteroflinguisticconvention.Likewords,thephrase-structurerulesofSBCGcombineconstraintsonmeaning,use,andform.Finally,SBCGmakesonlylimiteduseofsyntactictrees,asrecordsofthederivationalhistoriesofphrases(or,equivalently,therecursiveapplicationofsyntacticrules).TreesarenotobjectsofgrammaticaldescriptioninSBCG,noraremappingsbetweentrees(‘transformations’).Instead,anSBCGgrammardescribesthesignconfigurationsthatthegrammarpermits—constructionsthatbuildwordsfromoneormorelexemesandconstructionsthatbuildphrases(phrasalsigns)fromoneormoreexpressions.Aswewilldiscuss,thesesignconfigurationsareakintolocaltrees(i.e.,structuresconsistingofamothernodeandoneormoredaughternodes).TheSBCGprogramdoes,however,qualifyasagenerativegrammarinthehistoricallybroadersensethatitaimstoprovideafullyexplicitaccountofthesentencesofeachlanguageunderstudy.Thisundertakingrequiresthattherepresentationalconventionsofone'sgrammarcoverbothregularandidiomaticphrasetypes,asthetwotypesinteractinthelicensingofsentences.Acaseinpointissentence(1):whiletheSubjectlesstagconstructionitselfisidiomatic(notcountenancedbythecanonicalEnglishphrase-structurerules)thetagthat(1)containsispresumablylicensedbythesamerulethatlicensesothertags.Thus,onecanviewSBCG,aswellasotherconstruction-basedtheoriesofsyntax,asretainingdescriptivegoalsthattheoriginalproponentsofgenerativegrammarabandonedinfavorofamaximallygeneraltheory(seeChomsky1995:435).

WhatisasigninSBCG?TheconceptissimilartothatofdeSaussure(1916).However,whileSaussure'ssignsrelateonlyformandmeaning,thesignpropertiesrecognizedinSBCGincludephonology,(morphological)form,syntax(e.g.,aword'ssyntacticcategoryandcombinatoricpotential),semantics(e.g.,theframesthatcollectivelydefinethemeaningofaword,aword'sreferentialindex),anduseconditions(e.g.,theinformation-structurearticulationofaphrasaltype).Further,thesignsofSBCGincludenotonlywordsandlexemesbutalsophrases,includingsentences;theformvalueofaphrasalsignisalistofwords.FollowingthetraditionofUnificationGrammar(Shieber1986),GeneralizedPhrase-StructureGrammar(Gazdaretal.1985),Head-DrivenPhraseStructureGrammar(PollardandSag1987,1994),andalliedtheories,SBCGmodelssignsasfeaturestructures.Featurevaluesincludewordclasseslikenoun,casevalueslikeaccusative,referentialindices,lexemeclasseslikeintr(ansitive)-verb-lexeme,wordformslikekickandthebinaryvalues+and–.Inthissystem,grammaticalcategoriesaremodelednotbymeansofatomicsymbolslikeV,N′,orPPbutratherascomplexesofproperties:forexample,nounsspecifyvaluesforthefeaturescase,number,andgenderandverbsspecifyvaluesforinflectionclass(e.g.,as[vformfinite]or[vformpresent-participle]). Featurestructuresarerecursive;thatis,thevalueofafeaturemaybeanotherfeaturestructure;forexample,thevalueofthesyn(tax)featureisafeaturestructurecontainingthefeaturescat(egory)andval(ence).Afeature-structurevaluemayalsobealistoffeaturestructures;forexample,boththeframesfeatureandthevalfeaturearelist-valued.

Allsignsaremodeledbyfeaturestructures,butnotallfeaturestructuresmodelsigns.FeaturestructuresarealsousedinSBCGtomodelconstructs.Constructscanbethoughtofaslocaltreeswithsignsatthenodes;allconstructsmustbelicensedbyaconstructionofthegrammar(wewilllearninamomentwhataconstructionis).Formally,aconstructisafeaturestructurecontainingamother(mtr)featureandadaughters(dtrs)feature.Thatis,constructsaremodeledasfeaturestructuresofthekinddescribedin(2):

In(2),whichrepresentsanappropriatenessdeclaration,weseethataconstruct(abbreviatedascxt)mustcontainbothamothersignandanonemptylistofdaughtersigns.Intuitivelyspeaking,amothersignisconstructedfromasequenceofdaughtersigns.Asfeaturestructures,constructs,likesigns,havetypes,notatedbyitaliclabelsatthetopofthesquarebracketsenclosingafeaturestructure.AsintheHPSGtradition,typesareorganizedhierarchically:everyfeaturestructureofatypetmustsatisfytheconstraintsonallthesupertypesoft,plusanyadditionalconstraintsimposedoninstancesoftitself.TheSBCGmodelofconstructionsisinessenceatheoryoftypes.Constructionsaretypeconstraints:

1

Page 111: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 3 of 14

(3)x-cxt⇒[…]

Statement(3)saysthatifaconstructisofxtype,itmusthavewhateverpropertiesarerepresentedbythethreedots,inadditiontothepropertiesexpressedin(2).Thetypeconstraintin(3)describesanycombinatoryconstruction:itspecifiesthepropertiesthatdefineaclassofconstructs(i.e.,featurestructurescontainingboththemtranddtrsfeatures).Inparticular,(3)isacombinatoryconstruction:itdescribesaparticularwayofconstructingamothersignfromoneormoredaughtersigns. Inadditiontocombinatoryconstructions,SBCGrecognizeslexical-classconstructions.Constructionsofthelattertypedescribeclassesoffeaturestructuresthatcorrespondtowordsorlexemes.Thatis,lexical-classconstructionsdescribeclassesofsignsratherthanclassesofconstructs(signcombinations).Lexical-classconstructionsaredistinctfromthelexicalentries(lexemedescriptions)withwhichtheycombine.Bothcombinatoryandlexical-classconstructionswillbediscussedfurtherinsection8.2.1.

Becausebothlexicalclassesandconstructshavetypes,theyaresubjecttotaxonomicorganization.Sag(2012:68–69)describesthisorganizationasfollows:

Thespaceoftypesishierarchicallystructured.Infact,thetypesareinterrelatedintermsofamultipleinheritancehierarchy.IfatypeBisasubtypeofanothertypeA,then[featurestructures]oftypeBmustsatisfyallconstraintsthatthegrammarimposesonobjectsoftypeA,aswellasthegrammaticalconstraintsimposedontypeB.

Aswewillsee,typehierarchiesinSBCGtakeoverfunctionsservedbyconstructionalinheritanceinBCG.Asasimpleexample,consideragainthesubjectlesstaggedsentenceexemplifiedin(1).Thetypeconstraint(construction)thatdefinesthisconstructtypewillmentionthesupertype(taggedsentence)initsconsequentclause,asthefeature-structuretype.Thisensuresthatthesubjectlesstaggedsentencewillsharesyntactic,semantic,pragmatic,andphonological(intonational)propertieswithitsdominatingtype(e.g.,oppositepolarityofmainclauseandtag).Anyadditionalpropertiesthatappearinthefeaturestructuresotyped(inparticular,thecovertmain-clausesubject)willrepresentpropertiesuniquetosubjectlesstagsentences.Inthiswaywerepresentbothidiosyncraticandgeneralpropertiesofthegrammaticaltype.Suchatypehierarchy,aswewillsee,enablesSBCGtocaptureconstraintsatalllevelsofgranularity.

Thenextsectionofthischapter,section8.2,willfocusontheformalarchitectureofSBCG,itshistoricaloriginsandmotivations,itsdescriptiveapplications,andthemannerinwhichitdivergesfromBCG.Iwillbeginsection8.2byaddressingthehistoricalconnectionbetweenSBCGandBCG.Iwillthenfocusonspecificcomponentsofitsformalism:locality(section8.2.1),variable-graindescription(section8.2.2),inheritance(section8.2.3),andtheuseofnonbranchingconstructionstocaptureinflectionalandderivationalprocesses(section8.2.4).Theconcludingsection,section8.3,willaskwhatinsightsSBCGofferstodescriptivelinguistswhoarenotformalists.

8.2.TheFormalArchitectureofSBCG

Formanyyears,theprinciplereferenceworkavailabletoConstructionGrammariansworkingintheBCGtraditionhasbeenanunpublished(butwidelycirculated)coursereader,FillmoreandKay(1995).Amongthenotablefeaturesofthisworkisitstreatmentoftheinteractionbetweenverb-levelargument-structureconstructions(e.g.,passiveandditransitive)andpartiallyspecifiedlexicalentries(i.e.,lexicalentriesinwhichthematicroleslackgrammatical-functionassignments).Thisanalysisisstreamlined,inthatitdoesnotusegrammatical-function-assignmentoverridesofthekindthatGoldberg(1995)requirestomodelvoicealternations.FillmoreandKaypropose,forexample,thatthesemanticsoftransferassociatedbyGoldberg(1995:chapter6)withaverbvalencecontaininganagentivesubjectandtwoobjectarguments(recipientandtheme,respectively)—inotherwords,theactive-voiceditransitivevalence—isinsteadassociatedwithasingleconstraintonargumentexpression:thethemeargumentisrealizedasanominaloblique(anobliqueargumentwithdirectratherthanprepositionalcoding).This‘strippeddown’ditransitiveconstruction(referredbyFillmoreandKayastheNominalObliqueThemeconstruction)iscapableofunifyingdirectlywitheithertheActiveconstruction(inwhichcasetherecipientargumentislinkedtothegrammaticalfunctionobject)orthePassiveconstruction(inwhichcasetherecipientargumentislinkedtothegrammaticalfunctionsubject). TheFillmoreandKayworkalsodemonstratesthatthemechanismoflexeme-constructionunificationallowsonetodescribeEnglishnominalandverbalsyntaxwithoutrecoursetotheunarybranchingphrasesand‘inaudible’determinersof‘syntactocentric,’X′-basedapproaches.A

2

3

Page 112: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 4 of 14

clearexampleofthismodeofanalysisisfoundinthetreatmentofnominals(cf.alsoFillmore,thisvolume).Nounsdenotingunboundedsubstancesandaggregates(i.e.,massnouns(likebutter)andpluralnouns(likehorses))areanalyzedasunmarkedwithregardtothebinaryfeaturemax(imal);thismoveensuresthatsuchnounscaneither(a)unifywithoneofvariousdeterminationconstructions—allofwhichrequirea[max-]headdaughter—thuslicensingexampleslikethebutterorsomebutter,or(b)withoneofvarious-headcomplementconstructions(e.g.,theVerbPhraseconstruction,thePrepositionPhraseconstruction)—allofwhichrequirea[max+]complement—thuslicensingphraseslikeIavoidbutterandinbutter.ThisanalysisprovidesanelegantalternativetotheprevalentX′-basedtreatment,inwhichcertainlexicalnounsareviewedasphrases(NPs)basedontheirexternalratherthaninternalsyntax.TheBCGmodelofnominalsyntaxismoreintuitivethanthisone,inthatitreservesthelabel‘phrase’forbranchingstructures(constructs).

Despiteitseleganceandintuitiveappeal,however,theBCGmodelalsohascertaindysfunctionalproperties:itrepresentsphrasalconstructionsusingacumbersomenested-boxnotationandpermitsanundefineddegreeofrecursion,asinthetreatmentoflong-distancedependencies.Further,therepertoireoffeaturesthatappearinconstructiondescriptionsislooselyorganizedandapparentlyopen-ended.Inaddition,whileFillmoreandKay(1995)persuasivelyarguethatformalandsemanticcommonalitiesamongconstructionscanbecapturedbymeansofinheritancerelations(ratherthan,say,transformations),theworkdoesnotprovideanymeansofrepresentingsuchtaxonomicrelationships,otherthannotationsthatappearsporadicallyinconstructiondiagrams.Constructiongrammariansseekingamorecomprehensiveandprincipledsystemofformalrepresentationwereinclinedtolooktoanallieddeclarativemodel,HPSG(PollardandSag1987,1994).LikeBCG,HPSGtreatswordsandphrasalpatternsassimilarkindsofform-meaningpairings,usesfeaturestructurestomodelsemanticandsyntacticclassesofgrammarobjectsandspecifiesastructuredinventoryofsigntypesthatconstitutesamultiple-inheritancenetwork.

SBCGrepresentsanattemptto“expandtheempiricalcoverageofHPSG,whileatthesametimeputtingBCGonafirmertheoreticalfooting”(Sag2012:62).LikeBCG,andtheCognitiveLinguistics-influencedversionofConstructionGrammarproposedbyGoldberg(1995,2006a;cf.alsoBoas,thisvolume),SBCGisatheoryofconstructionalmeaning:itassumesthatrulesofsyntacticcombination(descriptionsoflocaltrees)aredirectlyassociatedwithinterpretiveanduseconditions,expressedbysemanticandpragmaticfeaturesthatattachtothemotherordaughternodesinthesedescriptions(KayandMichaelis,forthcoming;Sag2012).Thisamountstotheclaimthatsyntacticrulescanhavemeanings.ThisclaimsetsConstructionGrammarapartfromprevailingmodelsofmeaningcomposition.SuchtheoriesarebasedonaprinciplethatJackendoff(1997a:48)describesasthe“doctrineofsyntacticallytransparentcomposition.”Accordingtothisdoctrine,“[a]llelementsofcontentinthemeaningofasentencearefoundinthelexicalconceptualstructures…ofthelexicalitemscomposingthesentence.”Toproposeaconstruction-basedmodelofsemanticcompositionlikeSBCGisnot,however,todenytheexistenceofsyntacticallytransparentcomposition.Itisinsteadtotreatit,inaccordancewithJackendoff(1997a:49),asa“defaultinawiderarrayofoptions.”Thatis,wheneveraclassofexpressionscanbeviewedaslicensedbyacontext-freephrasestructureruleaccompaniedbyarulecomposingthesemanticsofthemotherfromthesemanticsofthedaughter,aconstruction-basedapproachwouldproposeaconstructionthatisfunctionallyequivalenttosucharule-to-rulepair.Buttheconstructionalapproachalsoenablesusrepresentlinguisticstructuresinwhichthesemanticsofthemotherdoesnotfollowentirelyfromthesemanticsofthedaughters,asinthecaseofidiomaticexpressionslikethrowinthetowel.Inthefollowingfoursubsections,wewilldiscussfourfeaturesthatmakeSBCGarobustformalismforconstruction-basedsyntax.Thefeaturesare:(1)locality,(2)variable-graindescription,(3)amodelofinheritancethatoffersreducedstipulationandenhanceddatacoverageincomparisontoBCG,and(4)theuseofunarybranchingconstructionstocapturederivationalandinflectionalmorphology.

8.2.1Locality

InSBCG,thephrasetypesinthetargetlanguagearedescribedbymeansofcombinatoryconstructions.Combinatoryconstructionsdescribeconstructs—signsthatarebuiltfromoneormoredistinctsigns.ConstructionsinSBCGtaketheformoftypeconstraints.Atypeconstraint,asmentionedinsection8.1,isaconditionalstatementthattellsuswhatpropertiesafeaturestructurewillhaveifitisaninstanceofthetypeinquestion.Intuitively,constructsarelocaltrees(mother-daughterconfigurations)withfeaturestructures(specifically,signs)atthenodes.Constructionscandescribeonlysuchmother-daughterdependenciesandnot,forexample,mother-

Page 113: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 5 of 14

granddaughterdependencies(Sag2007,2012).

Asmentionedinsection8.1,alocaltreeisdistinctfromaderivationtree.Derivationtreesrecordtheprocessofphraseconstructionthroughtherecursiveexpansionofphrasalnodes,andcan,ofcourse,havemanymorethantwolevels.WhilederivationtreescanbeusedtodescribetherecursivelicensingofsignsinSBCG,suchtreesarenotobjectsofgrammaticaldescriptioninSBCG.Onlythosetreesthatqualifyasconstructsareinthelanguagemodel.Asdiscussedinsection8.1,aconstructismodeledinSBCGasafeaturestructurethatcontainsamother(mtr)featureandadaughters(dtrs)feature.Thevalueofthemtrfeatureisasignandthevalueofthedtrsfeaturealistofoneormoresigns.Letusnowrecallthecharacterizationofsignsfromsection8.1.Asign,asintheSaussureantradition,isaform-meaningpairing,buttheconcepthasarigorousformalimplementationinSBCG.Asignisafeaturestructure.Asignassignsvaluestothefollowingfeatures,amongothers:

•phon:aphonologicalphrase•form:alistoftheformatives(wordsormorphemes)thatcomprisetheexpression•arg-st:arankedlistofalexicalexpression'sarguments(i.e.,onlylexicalsignshavethisfeature)•syn:catandval(ence)•sem:ind(ex)andframes•cntxt:background(bckgrnd,includingthesetofpresuppositionsassociatedwithaconstructiontype),contextual-indices(c-inds;identitiesofspeakerandaddressee),topicandfocus(pragmaticrolessharingreferentialindiceswithelementsonthearg-stlist).

Severalofthesefeaturesdeservefurtherdiscussionhere.Letusbeginwiththefeaturesthatpertaintoapredicator'scombinatoricpotential,valandarg-st.Thefeaturevalliststhoseargumentsignsthatanexpressionhasyettocombinewithsyntactically,whetherthatexpressionisawordoraphrase.Thefeaturevaliscloselyrelatedtothefeaturearg-st—bothofthesefeaturescantakeanemptylistofsignsastheirvalues.Thevallistofalexemeisidenticaltoitsarg-stlist,providedthatthelexeme'sargumentsarelocallyinstantiated(i.e.,notextracted),andovertlyexpressed(seesection8.2.4foradiscussionoftheSBCGtreatmentofnull-instantiatedargumentsandSag(2010)foratreatmentofextraction).However,whileonlylexemescarrythearg-stfeature,bothlexemesandphraseshavevalsets,whosemembersaresubtractedateachmothernode,astheheadexpression'ssyntacticsistersaresupplied.Thus,whiletheverballexemedevourhastwoNPsonitsvallist(i.e.,<NP,NP>)theVPdevouredthecompetition(licensedbythecombinatoryconstructionthatcombinesheadswiththeircomplements)hasasingletonvallist,<NP>,signalingthatonlythesubjectrequirementremainstobesatisfied.

Figure8.1. Alexemeentry

Thesemanticfeaturesindandframesrequirecommenthereaswell.Thevalueoftheindfeatureisthereferentialindexofanexpression,representedasavariable.ThisindexisassignedtoanindividualwhentheexpressionbearingtheindexisaNPortoasituationwhentheexpressionbearingtheindexisaclause,verb,orVP.Thevalueoftheframesfeatureisthesetofpredicationsthatjointlyexpressthemeaningofasign.ThetermframeisunderstoodasitisinFillmore'sframesemantics(Bakeretal.2003):ascene(whetherdynamicorstatic)involvingoneormoreparticipants.Eachpredicateinthelistofframesisrepresentedbyatypedfeaturestructure,whosefeaturesarethesemanticrolesrequiredbythatframe.Forexample,thefeaturestructurethatrepresentsthemeaningofthelexemeeatbearsthelabeleat-fr(ame).Thefeaturesinsideeachfeaturestructureofthetypeframe(e.g.,eater,experiencer,orsimplyarg)takeindicesastheirvalues.ThemajorfeaturesdescribedhereareexemplifiedinthelexemeentryshowninFigure8.1.

Page 114: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 6 of 14

ThelexemerepresentedinFigure8.1istheverbbreak;itsarg-standvalvaluesareidentical(althoughwewillseeacaseinwhichtheydivergeinsection8.2.4).Themeaningofthisverbisrepresentedbytwoframe-typefeaturestructures.Thefirstislabeledbreak-fr.Thefeaturesinsidethisfeaturestructure,agentandtheme,representthesemanticrolesassignedbythebreak-frame.Thesecondislabeledrigid-object-fr.Thisframecapturesanentailedpropertyofthethemeargumentofbreak.FollowingthetraditionofMinimalRecursionSemantics(Copestakeetal.2005),SBCGusesa‘flat’representationoftheframesthatjointlycomprisethemeaningofanexpression.Thatis,therearenoembeddingrelationshipsamongtheframesintheframeslist.Theframesinthislistareboundtogethersolelybymeansofcoreferencerelationshipsamongtheirarguments.Forexample,inFigure8.1,thethemeargumentofthebreak-frameandthesoleargumentoftherigid-object-framearecoindexed.

TherepresentationinFigure8.1isofasinglesign;recall,however,fromabovethatbothsignsandsigncombinations(constructs)arerepresentedbyfeaturestructuresinSBCG.Bymodelingphrasetypesasfeaturestructures,SBCGcapturespropertiescommontolexemesandphrasetypesinawaythatBCGdidnot.Asmentioned,accordingtotheBCGvision,thegrammarisaninventoryoftrees(nestedboxes)ofarbitrarydepth.Bycontrast,argument-structureconstructionsliketheTransitiveconstructionarerepresentedinBCGbyfeaturestructures,asinFigure8.2.

Figure8.2. TheTransitiveconstructionasperBCG(FillmoreandKay1995)

TheconstructionshowninFigure8.2expressesaconstraintontransitivelexemes:eachsuchlexemeassignsthegrammaticalfunctionobjecttooneargumentinitsvalenceset,providedthatthisargumentisnotthehighestrankingor‘distinguished’argument.TheTransitiveconstructionpresumablyrepresentsaclassoflexemes(thosethattakedirectobjects),butitisintuitivelyunclearwhyalexemedescriptionlikethatinFigure8.2shouldqualifyasaconstruction,asitdoesnotcontainnestedboxes.SBCG,bycontrast,proposestwotypesofconstructions:theaforementionedcombinatoryconstructions,whichdescribepropertiesofphrasetypes,andlexical-classconstructions,whichdescribepropertiessharedbyclassesoflexemes(likedevour)andwords(likedevoured).Theonlydifferencebetweenlexical-classconstructionsandcombinatoryconstructionsisthetypenameintheantecedentofthetypeconstraint.Becausebothwordsandphrasesaresigns,thetwocanbedescribedinthesameway.ThisisshownbyFigures8.3–8.4,whichillustrate,respectively,alexical-classconstructionandacombinatoryconstruction.

TheApplicativeconstruction,showninFigure8.3,describesthelexemeclasstowhichtheverbsfillandcoverbelong,asillustratedby(4–5):

(4)Shefilledthebathtubwithchampagne.(5)Theycoveredthewallwithasheet.

Figure8.3. TheApplicativelexical-classconstruction

ThelexemeclassdescribedbytheconstructioninFigure8.3isasubtypeofthetransitive-lexemeclass,asindicatedbythetypingofthefeaturestructuretotherightofthearrow.Asshownbythearg-stlist,verbsofthislexemeclassexpressthethemeargumentasaPPheadedbywith.Thesemanticconstraintsassociatedwiththislexemeclassareasindicatedbytheframelabeledsaturation-frintheframeslist.Thisframeisintendedtocapturetheresultant-stateentailmentthatthethemeoccupiesacriticalmassofpointswithinaplanarregion(MichaelisandRuppenhofer2001).TheApplicativeconstructionalsodescribesoneoftheclassestowhichtheverbssprayandloadbelong:thelexicalentriesofthesevariableverbslackaspecifiedarg-stlist,makingthemcompatiblewithboththeApplicativeandOblique-Goallexical-classconstructions.

Page 115: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 7 of 14

Figure8.4. TheSubject-Predicatecombinatoryconstruction

TheSubject-Predicateconstruction,showninFigure8.4,licensesbasicdeclarativeclauses.Theconstructioncontainsafeaturenotpreviouslydiscussed:marking(mrkg).Thevalueofthisfeatureisthetypeofthegrammaticalmarkerthataccompaniesasign(e.g.,thatinthecaseofphrasalverbalsignwhosedtrsarethecomplementizerthatandaclause).InthecaseoftheSubject-Predicateconstruction,thevalueunmkindicatesthatneithertheverbalheadnoritsmtrhasagrammaticalmarker.AsdescribedinFigure8.4,asubject-predicateconstructconsistsoftwodaughtersigns,thesecondofwhichisafiniteverbalsignthatselectsforthefirstsignbymeansofitsvalfeature.Asshowninthisfigure,themotherofasubject-predicateconstructhasanemptyvallist,indicatingthatitisacompletepredication.

Insum,SBCGcapturespropertiescommontolexicalitemsandphrasesbydescribingbothasfeaturestructures.Combinatoryconstructionsdescribesignconfigurations(viathemtranddtrsfeatures),whilelexical-classconstructionsdescribesinglesigns.Butsignsandsignconfigurationsarebothtypesoffeaturestructuresandtheirdescriptionsthusdenoteclassesoffeaturestructures.

TheprinciplethatgovernsthelicensingoflanguageobjectsinSBCGistheSignPrinciple.AccordingtotheSignPrincipleofSBCG(Sag2012:97)asignislexicallylicensedifitsatisfies(correspondsto)alexicalentryandconstructionallylicensedifitisthemothersignofsomeconstructtype. Thismeansthatonecanverifythegrammaticalityofaphrasebasedonlyonthepropertiesofitstopmost(mtr)featurestructure,sincethesepropertiesincludeidentifyinginformationaboutthatnode'sdaughters(e.g.,theframesonthemtr'sframeslist).Theanalysistreeportraysgraphicallyhowtheconstructionsofthegrammarcollectivelylicensethissign,butthemodelofthephraseisjustthesinglesign—thatis,itistherootfeaturestructureoftheanalysistree.

8.2.2VariableGranularity

AsiswidelyrecognizedbyproponentsofConstructionGrammarandexemplar-basedapproaches(e.g.,Bybee2007,thisvolume),manygrammaticalgeneralizationsarenotverygeneral.TheISISconstruction,exemplifiedin(6)below,isacaseinpoint(smallcapsindicatepointsofprosodicprominence;|indicatesapause):

(6)SeeI—Iagreewiththat,butmywholeproblemis|isthatIreallydon’tlikeBush.

AsdiscussedbyBrenierandMichaelis(2005),ISISisoneofseveralstrategiesthatspeakersusetoannounceaforthcomingproposition(e.g.,Ireallydon’tlikeBush)bymeansofa‘setup’clause(e.g.,mywholeproblemis)whosesubjectisaninformationallylightnounphraselikethemainthing,theproblem,therealissue,orthequestion.AnalyzedasasyntacticamalgambyBrenierandMichaelis,ISIScontainstwoadjacenttensedformsoftheverbbe,thefirstofwhichistypicallyaccented.BrenierandMichaelisarguethatthisidiomaticpatternisdesignedtosolveasyntax-to-phonologymappingproblem:unlikethestandardalternativestructure(7),ISIScontainsanintonationallyunbrokenverbphraseandanunaccentedcopula:

(7)Thethingis|Ialwayscarrymycheckbook.

Butinsolvingaphonologicalproblem,ISIScreatesasyntacticone:theISIS‘setup’hasnoidentifiablesyntacticcategory—itismorethanaverbphrasebutlessthanafullclause.Whilefunctionallymotivated,theISISconstructionfeaturesseveralfine-grainedconstraintsthatappearinexplicableonsemanticorsyntacticgrounds.Oneoftheseconcernspossibletensecombinations:whiletheBE1slotmaybefilledbythewordswas,is,orbeing,theBE2slotisinvariantlyfilledbythewordis.

(8)Therealquestionwasisarewegettingareasonablereturnonourinvestment.

Independent-clause(IC)exclamativesprovideanotherexampleofafine-grainedconstraint:

(9)God,*(Ican’tbelieve)whotheyhired/wheretheywent!

4

5

Page 116: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 8 of 14

As(9)shows,ICandsubordinate-clauseexclamativesdifferwithregardtothesyntacticcategoryofthefillerdaughter:whoandwherearenotpossiblefillerdaughtersofICexclamativesinEnglish,althoughtheyareinsomeotherlanguages(Michaelis2001).Itisessentialtorepresentsuchcategoryrestrictionsifagrammaristobepreventedfromovergenerating.

Clicktoviewlarger

Figure8.5. TheFiller-Headconstruction(basedonSag2012)

HowdoesSBCGavoidovergeneration?Ittreatsnodes,andinparticulartheMTRnodesofconstructs,asfeaturestructures—notcategorylabels. Adescriptionofafeaturestructureisasetofproperties.Aspropertysets,feature-structuredescriptionsfollowthelogicofsetinclusion:themorepropertiesinthedescription,thesmallertheclassoflanguageobjectsthatdescriptionpicksout.Forexample,thefeaturesetthatdescribesanICexclamative(e.g.,Whatfools!)includesthatwhichdefinesthefiller-headconstruction,showninFigure8.5.Inclusionrelationsamongfeature-structuredescriptionsallowustomodelconstructsateachstepalongtheidiomaticitycontinuum,withanarrayofconstructionsofcorrespondinglygradedgenerality.

8.2.3AGeneralizedAccountofInheritance

AleadinginsightofConstructionGrammarfromitsinceptionisthatgrammarrulesarenotproceduresbutcategorydescriptions,andassuch,subjecttotaxonomicorganization.Suchtaxonomies,whichcametobeknownintheConstructionGrammarliteratureasinheritancenetworks(see,e.g.,Goldberg1995),provideforcross-cuttinggeneralizationsaboutconstructions.Theidea,simplyput,isthataconstructcanbeaninstanceofmultipletypesatonce.Goldberg(1995)positedtwomajorinheritancerelations:theinstancerelationandthesubpartrelation.BothrelationsareillustratedbytheExtraposedExclamativeconstruction,asin(10):

(10)It'samazingwhatshesaid.

Figure8.6. TheInvertedExclamativeconstruction

UnderGoldberg'sconceptionofinheritance,theExtraposedExclamativewouldberegardedasaninstanceoftheExtrapositionconstructionthatcontainsasasubpartthewh-InterrogativeClauseconstruction.Whileinheritancenetworksofthisnatureofferawaytodescribesemanticandsyntacticcommonalitiesamongsignsandconstructswithoutrecoursetoderivations,ithasremainedunclearjusthowsuchcross-constructionalgeneralizationsaretoberepresented.Shouldtheyberepresentedbyastipulationinaconstructionx,‘inheritconstructiony,’asperFillmoreandKay(1995)?Orshouldtheyberepresentedbytypedlinksinradial-categorydiagrams,asperLakoff(1987),Goldberg(1995),andMichaelisandLambrecht(1996)?Bothstrategieshaveanadhocflavor.The‘inherit’stipulationlookslikeafeature,butitisnevermadeclearwhattypesoffeaturestructurescontainit.Typedlinksdonotappearinthefeaturestructuresusedtorepresentconstructions,soonecannotknowfromlookingatagivenconstructionwhatits‘relatives’are.InSBCG,bycontrast,constraintinheritanceisdescribedbythehierarchyoftypes.Grammaticalobjectsofallkinds,includingphrasetypes,arerepresentedbyfeaturestructures,andthesefeaturestructuresbelongtoataxonomyoftypes.Onecanthereforedeterminewhatconstraintsaresharedbywhatconstructtypesbyconsultingthedescriptionsofthemtranddtrsignsofconstructions,wherethecommonpropertiesarecapturedbyfeaturesorfeature-structuretypes,asappropriate.Ratherthanbeingstipulated,inheritancerelationsareencodedineachconstruction's‘DNA.’Tounderstandhowthisworks,wemustrecallthataconstructionisaconditionalstatementdefiningthepropertiesthatarecommontoallinstancesofagivenfeature-structuretype.Thegeneralschemaforacombinatoryconstructionisshownin(11),repeatedfrom(3)above:

6

Page 117: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 9 of 14

(11)x-cxtX[…]

Thus,ratherthanpositinginstancelinks,SBCGallowsaconstructiontodefineaconstructtypeA,andanotherconstructiontodefineasubtypeofAbymentioningAinitsconsequentclause. AnexampleisgiveninFigure8.6.

TheconstructiondescribedinFigure8.6istheInvertedExclamativeconstruction,exemplifiedby(12–13):

(12)Man,wasIeverwrongaboutthatone!(13)Jesus,canthatguytalkorwhat?

AsindicatedinFigure8.6,theMTRsignofthisconstructionisthatofanAuxiliary-Initialconstruct.ThatconstructtypeisdescribedbythetypeconstraintinFigure8.7.

AccordingtotheconstructioninFigure8.7,anAuxiliary-Initialconstructisaheadedconstructthatqualifiesasclause(i.e.,aphrasewhoseheaddaughterisaverbandwhosevalencesetisempty).Theheaddaughterofthisheadedconstructisaverbthatappearsininitialposition(asindicatedbyitsplaceintheDTRSlist)andismarkedbythefeature[INV+](adistinctconstraintensuresthat[INV+]wordsarefiniteauxiliaries).BecausetheInvertedExclamativeconstructdescribedinFigure8.6isofthetypeAuxiliary-Initialconstruct,theformerwillinheritallofthepropertiesofthefeaturestructure(atypeofconstruct)describedinFigure8.7,aswellasadditionalpropertiesrepresentedbytheellipsesinFigure8.6.

Havingconsideredthetreatmentof‘instancelinks’inSBCG,letuslookattheSBCGanalogtoGoldberg'ssubpartlink.BecauseSBCGisalocalisttheoryofsyntax,asdescribedinsection8.2.1,itstreatmentofsubpartrelationswillnecessarilydivergesignificantlyfromthatfoundinBCGworks.Constructionsareconfigurationsofsignsratherthanconfigurationsofconstructs;therefore,aconstructioncannotincludeanotherconstructioninitsDTRSlistandaconstruction—adescriptionofaclassofconstructs—canmakenoreferencetothedaughtersofaconstruct'sdaughters.SuchapracticewouldbenomoreacceptableinSBCGthanplacingaphrase-structureruleintheexpansionofanotherphrase-structureruleinacontext-freegrammar,forexample,*VP→V(PP→PPP).Ifwecannotrepresentoneconstructionascontaininganother,howthenare‘subpart’relationstoberepresentedinSBCG?Theexamplein(14)willservetoillustrateSBCG'sapproachto‘subpart’relations:

(14)NeverhaveIseenone.

Figure8.7. AuxiliaryInitialconstruction(basedonSag2012:(9))

Lookingattheconstructtypeillustratedin(14),InvertedNegativeAdverbPreposing,wemightintuitivelysaythatitcontainstwodaughters,thefirstofwhichisanegativeadverbandthesecondofwhichistheMTRofanAuxiliary-Initialconstructtype.However,asdiscussedabove,constructionscannotcontainotherconstructions.Instead,wewouldsaythattheheaddaughterofaparticularconstructtypesharesoneormorefeatureswiththeMTRofsomeotherconstructtype.Toseehowthisworks,letuslookattheInvertedNegativeAdverbPreposingconstruction,showninFigure8.8.

TheconstructtypeshowninFigure8.8hasahead-daughtersignwiththeproperty[INV+].ThisfeatureissharedbytheAuxiliaryInitialconstructtype,showninFigure8.6:itsheaddaughterisalso[INV+].Thus,SBCGcapturesdaughterpropertiessharedbyconstructionsinamannerthatisconsistentwithlocalistassumptions.

Butadheringtolocalityisnottheonlyreasontoreplaceaconceptionofinheritancebasedonsubpartrelationswithonebasedonhead-daughterfeatures.AsI.Sagobserves(personalcommunication),onecannotapplythesubpart-basedconceptionofinheritancetothe(bracketed)headdaughtersin(15–17):

(15)Howmanybooks[youreadandIbuy]!

7

Page 118: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 10 of 14

(16)Neverbefore[havesomanypeopleattendedthatweranoutofroom].(17)She[fortunatelyalmostnevercomplains].

Clicktoviewlarger

Figure8.8. TheInvertedNegativeAdverbPreposingconstruction

Underasubpart-basedconceptionofinheritance,theInterrogativeExclamativeconstruction(e.g.,Howniceitwas!)wouldinherititsseconddaughterfromtheSubject-Predicateconstruction.Thisanalysiswouldnot,however,coverexampleslike(15),inwhichtheseconddaughterconsistsofconjoinedclauses.HowistheheaddaughteroftheInterrogativeExclamativecharacterizedinSBCG?Itsimplyhasthefeatures[VFORMfin]and[INV-].Similarly,underaBCGconceptionofinheritance,theInvertedNegativeAdverbPreposingconstruction,exemplifiedby(14)above,wouldinherititsseconddaughterfromtheAuxiliary-Initialconstruction,depictedinFigure8.7.Suchananalysiswouldnot,however,workfor(16),whoserightdaughterislicensednotbytheAuxiliary-InitialconstructionbutbytheHead-Extrapositionconstruction(KayandSag2009,2012).Exampleslike(16)arenotproblematicforanalysesofthesortproposedbySBCG.ThesesimplyrequirethattheseconddaughteroftheNegativeAdverbPreposingconstructionisaclausespecifiedas[INV+].Finally,example(17)underminestheBCGassumptionthattheSubject-PredicateconstructioninheritstheHead-Complementconstructionasitsheaddaughter.Becausein(17)thereareadverbialexpressions(fortunatelyalmostnever)precedingthematrixverb(complains),theheaddaughterwouldbelicensedbyamodificationconstruction,andnottheHead-Complementconstruction.If,however,weassume,inlinewithSBCG,thattheSubject-Predicateconstructionmerelyconstrainsitsseconddaughtertobe[VFORMfin],(17)isunproblematic.Insum,whileitmaybeconceptuallyappealingtorefertoclauseslike(15–17)asinheriting(oreven‘containing’)otherconstructionsastheirheaddaughters,accountingforthefullarrayofhead-daughterphrasesthatweactuallyencounterrequiresafeature-basedformalization.

8.2.4UnaryBranchingConstructions

SBCGmakesextensiveuseofunarybranchingconstructions—descriptionsofconstructsinwhichaMTRfeaturestructuredominatesasingledaughterfeaturestructure—tomodelinflectionalandderivationalprocesses.Thismightatfirstappeartobeastepbackwards,sincetheprogenitorofSBCG,BCG,stringentlyavoidedunarybranchinganalysesineithernominalorverbalsyntax.TheBCGtreatmentofnominalsyntaxwasdiscussedintheintroductiontosection8.2(seealsoFillmore,thisvolume):recallthatBCGpresumesbothmassandpluralnounlexemestobeunmarkedforthemaximalfeature,allowingthesenounstoeitherunifywiththeheaddaughterofadeterminationconstruction,inwhichcasethenounassumesthevalue[max-],orcombinedirectlywiththeHead-Complementconstruction,inwhichcasethenounassumesthevalue[max+].Inthelattercase,amass-orplural-nouncomplement(e.g.,inwater,gathercoins)isnotconsideredtobeanNexhaustivelydominatedbyaNP,asinthetraditionalX′approach,butratheranounthatislexicallyunmarkedformaximality,andthusabletotakeonthefeature[max+]whenunifyingwiththeHead-Complementconstruction(whosecomplementdaughterisrequiredtobe[max+]).Similarly,BCGtreatsintransitiveverblexemeslikedisappearas[max+];thismoveallowssuchverbstounifydirectlywiththeSubject-Predicateconstruction,whichrequiresitsheaddaughtertobe[max+]. ThistreatmentthuseschewsthetraditionalX′analysisinwhichanintransitive-verbpredicate(e.g.,Thestaindisappeared)isrepresentedbyaVPexhaustivelydominatingaV.

Atthesametime,certainBCGanalysesdorequirenonbranchingdomination.Inparticular,BCGemploysderivational(‘boxoverbox’)constructionstomodelnominalcoercionphenomena,exemplifiedby(18–19):

(18)Givemeabutter.(mass→countcoercion)(19)Addsometomato.(count→masscoercion)

Inthecaseof(18),FillmoreandKay(1995:chapter4)proposeaderivationalconstruction(theMass-to-Count

8

Page 119: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 11 of 14

construction)inwhichanounlexemetypebearinganegativevalueforthefeatureboundedisdominatedbyanounlexemetypebearingtheoppositevalue.ThisderivednounisthenabletounifydirectlywiththeIndefiniteDeterminationconstruction,whichrequiresa[bounded+]headdaughter.Correspondingly,aCount-to-Massconstructionispresumedtolicensethemass-nounlexemetomato,whichcanthenunifydirectlywiththe[bounded-]headdaughterofthePartitiveDeterminationconstructionin(19).Thus,whilecoercionphenomenaareelsewheredescribedascasesofsemantic-conflictresolution(see,e.g.,DeSwart1998),inwhichtheconstructionoverridesfeature-valuesofthelexemeintheinterestofverb-constructionunification(Michaelis2004),theBCGanalysesof(18–19)presumenosuchconflict.Instead,derivationalconstructionsgeneratenewlexemesbearingboundednessvaluesappropriatetothe‘coercive’context,e.g.,theindefinitearticlein(18),thepartitivearticlesomein(19). Thus,theuseofunarybranchingconstructionstomodelderivationalprocessesinSBCGfindsaclearprecedentinBCG.

Clicktoviewlarger

Figure8.9. AninflectionalconstructlicensedbythePreteritconstruction

Asmentioned,unarybranchingconstructionsareusedtomodelbothinflectionalandderivationalprocessesinSBCG:Thepropertiesofaword,e.g.,averb,arejointlydeterminedby:alexemedescriptioninthelexicon,alexical-classconstruction,andbyoneormorederivationalandinflectionalconstructions.Derivationalconstructionsallowlexemestobeformedfromotherlexemeswhileinflectionalconstructionsallowwordstobeconstructedfromlexemes.Asanexampleofaconstructionthatlicensesthe‘construction’ofawordfromalexeme,considerthePreteritconstruction(describedbySag2012:109−10).AconstructlicensedbythePreteritconstructionisgiveninFigure8.9(certaindetailsofSag'srepresentationhavebeeneliminatedforeaseofexposition).

ThePreteritconstructionconstructsawordfromalexeme,wherethemtrwordhasthefollowingproperties:first,theverbform(vform)isfinite;second,thesubjectvalencemember'scasevalueisnominative(asrequiredofallsubjectsoffiniteverbs);third,theframeslistcontains,inadditiontothedaughterlexeme'sframe(i.e.,thelaugh-frame),a‘pastness’frameandanexistentialquantifierframe(exist-fr).Thebound-variable(bv)argumentofthequantifierframeisidentifiedwiththesituationspecifiedinthedaughterlexeme'sframe,asindicatedbytheindexs.Thissituationisalsothesoleargumentofthepast-frame.

DerivationalconstructionsareusedinSBCGtomodellexeme-lexemerelations.TheMTRinaderivationalconstructisalexeme;thisMTRhasoneormorelexemeDTRS.Derivationalconstructionsareusedtomodelnotonlyderivationalmorphology(e.g.,compounding,morphemesthatchangethesyntacticcategoriesofwords)butalsovalence-changingconstructions.Valence-changingconstructionsincludetheCausedMotionconstruction,asdescribedbyGoldberg(1995:chapter7).GoldbergviewstheCausedMotionconstructionasaverballexemetypethatlicensesanagent,atheme,andapath.Shepointsoutthattheconstructioncombineswithtwotypesofverbs:

9

Page 120: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 12 of 14

thoseliketransitivemove,whichmatchthevalenceoftheconstruction,asin(20),andthoselikesneeze,whichdonot,andthereforemustundergovalenceaugmentation,asin(21):

(20)Shemovedthepianointothefoyer.(21)Shesneezedthefoamoffthebeer.

Ratherthanassumingaverb-constructionvalence-reconciliationmechanismincaseslike(21),Sag(2012:134),inspiredbyKay(2005),positsaderivationalconstructionwithaneffectsimilartothatofGoldberg'sCausedMotionconstruction.ThisderivationalconstructionisshowninFigure8.10.

Clicktoviewlarger

Figure8.10. TheCausedMotionconstruction

InFigure8.10,weseethatthemtrlexemeaddsthecaused-motion-frametotheframeslistofthedaughterlexeme,aswellasaugmentingthearg-stlistofthedaughterlexeme.Thethreemembersofthemtr'sarg-stlistarecoindexedwiththeappropriatemembersofitsframeslist.Noticethatonlyonesignisrequiredtoappearonthedaughterlexeme'sarg-stlist:thatwhichrepresentstheeffectorargument.Thisensuresthatbothtransitiveverbslikemoveandintransitiveverbslikesneezecanunifywiththedaughter-lexemedescription.Thus,thisconstructionlicensesboth‘valence-match’tokenslike(20)and‘valence-mismatch’tokenslike(21).

Inadditiontovalence-augmentingderivationalconstructionsliketheCausedMotionconstruction,SBCGpositsvalence-reducingderivationalconstructions,likePassive.Anadditionalexampleofavalence-reducingderivationalconstruction,istheNullComplementationconstruction,whichallowsformismatchesbetweenapredicator'sargument-structurelistanditsvalencelist,asin,e.g.,Iate(food)atnoonandMakeacopy(ofit)forme(fordetails,seeMichaelis2012andKay2004b).

ThisbriefexplorationofinflectionalandderivationalconstructionshasshownusthatconstructionsinSBCGarenotlimitedtothosethatlicenseconstructs(combinatoryconstructions)andlexicalclasses(lexical-classconstructions),butincludeaswellthosethatlicenselexeme-wordrelations(inflectionalconstructions)andlexeme-lexemerelations(derivationalconstructions).

8.3.Conclusion

AlthoughSBCGcannotbedivorcedfromtheformalconventionsthatitusestorepresentlexemes,constructions,andthehierarchicalrelationsamongtypes,italsooffersinsightstoConstructionGrammarianswhoseworkisnotprimarilyformal.TheleadinginsightofSBCG,simplyput,isthatthelexiconprovidesamodelforthesyntax-semanticsinterface.Lexical-classconstructions,whichdefineclassesoflexemesorwords,andcombinatoryconstructions,whichdefineclassesofphrases,arebothconstraintsonfeaturestructures.Inphrasalconstructions,alist-valuedfeatureofthemotherisusedtorepresentthepropertyofhavingthedaughtersitdoes.Further,theconstructionsthatembodyderivationalandinflectionalprocessesarenotdistinctinkindfromtheconstructionsthatbuildphrases.Thus,ratherthanseeingsyntax,semantics,andlexiconasindependentmodules,withthelexiconcharacterizedasasetofidiosyncraticform-meaningassociations,SBCGproposesalexiconstructuredbyhierarchicallyorganizedlexicalclassesandextendsthismodeltorelationsamongphrasalclasses.

Taxonomicorganizationisnottheonlythingthatwordsandconstructionshaveincommon:constructionsmeanwhattheymeaninthesamewaythatwordsdo.Likewords,constructionsmayinvokesemantic,pragmatic,andphonologicalconditionssimultaneously.Asanexampleofanidiomaticpatternwithhighlyparticularintonational

10

11

Page 121: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

Page 13 of 14

phonology,considertheExclamatoryconstructionthatMichaelisandLambrecht(1996)refertoastheAntitopicExclamative.Inthisconstruction,apreclausalinterjectionreceivesprosodicprominenceandthefollowingclausereceivestheintonationalcontourofaright-dislocatedphrase,asin(29–31):

(22)GODit'shot.(23)MANI’mtired.(24)DAMNyou’regood.

Thepointhereisthat,asCroftandCruse(2004:247)putit,“[c]onstructions,likethelexicalitemsinthelexicon,are‘vertical’structuresthatcombinesyntactic,semanticandevenphonologicalinformation(forthespecificwordsinaconstruction,aswellasanyuniqueprosodicfeaturesthatmaybeassociatedwithaconstruction).”Themoregeneralpoint,asexpressedbyCulicoverandJackendoff(2005:15)isthat“[t]hereisacontinuumofgrammaticalphenomenafromidiosyncratic(includingwords)togeneralrulesofgrammar.”

PortionsofthischapterappearasMichaelis(2012).ThischapteralsoreliesheavilyonSag(2012).MyunderstandingofthematerialpresentedherewasgreatlyenhancedbydiscussionswithIvanSag,PaulKay,andCharlesFillmore,whosecontributionsIgratefullyacknowledge.

Notes:

(1.)Admittedly,however,traditionallabelslikeNPandVPareoftenusedinSBCGtoabbreviateparticularfeaturestructures.

(2.)Itistemptingtoassumethatthemothersignsofcombinatoryconstructionsarenecessarilyphrases,andthatallmothersignsdominatetwoormoredaughtersigns.Infact,themajorityofcombinatoryconstructionsdescribesuchconstructtypes.However,aswewillseeinsection8.2.4,SBCGalsopermitsunary-branchingcombinatoryconstructions(e.g.,derivationalconstructions),inwhichthemtrsignisalexemeratherthanaphrase,andthismtrsignhasonlyonedaughtersign.

(3.)AlthoughbothGoldberg(1995)andFillmoreandKay(1995)describelinkingconstraintsusinggrammaticalfunctionslikesubjectandobjectratherthancaseroleslikenominativeandaccusative,itshouldbenotedthattheroleplayedbygrammaticalfunctionsinthesemodelsisservedinsteadbycaserolesinSBCG,whicheschewsgrammaticalfunctionsinfavorofanarg-st(argument-structure)list.Thearg-stlistencapsulatesarankingakintoKeenanandComrie's(1977)AccessibilityHierarchy.ThisrankingplaysaroleintheSBCGbindingconstraints,whicharestatedasconstraintsonarg-stlists.

(4.)IhaveslightlyalteredSag'sstatementoftheSignPrincipleforthesakeofclarity.WhileSagreferstothelistemiclicensingofsigns,toencompasslicensingofbothlexicalsignsandmulti-wordexpressions(thatis,listemes),mystatementoftheprinciplerefersonlytolexicallicensing.ThissimplifyingmoveseemsjustifiedasthepresentchapterdoesnotattempttodescribetheSBCGapproachtoidioms,uponwhichSagexpatiatesinSag2012.Itshouldbenotedthat,accordingtotheSignPrinciple,alexicalsigncanbeconstructionallylicensed,ifitcorrespondstotheMTRsignofaderivationalorinflectionalconstruction(suchconstructionswillbediscussedinsection8.2.4).Infact,theonlylexicalsignsthatarelicensedexclusivelybylexicalentriesarethosethatarenot‘produced’by(thatis,notthemothersof)derivationalorinflectionalconstructions.

(5.)BrenierandMichaelis'sobservationsareconfinedtotheSwitchboardcorpusofAmericanEnglishtelephoneconversations.

(6.)Certainly,X′approachesoccasionallyusefeaturestructuresinplaceofcategorylabels,buttherelevantfeatureinventoriesdonotoffermuchdescriptivepower:thefeaturesarelimitedtosyntacticones,andthefeaturevaluesareallatomic—thereisnoembeddingoffeaturestructuresinotherfeaturestructures.

(7.)Strictlyspeaking,mentioningatypeasanantecedentinadescriptiondoesnotcreateit.Rather,itisthetypehierarchythatcreatestypes.

(8.)ItshouldbenotedthatSBCGtooavoidsunarybranchinganalysesinVPsyntax,sincewhatiscalledaVPinothertheoriesisinSBCGsimplyaverbthathasonlyasubjectargumentleftonitsvallist.However,theunderspecificationanalysisproposedwithinBCGfornominalsyntaxisnotavailableinSBCG,whichdisallows

Page 122: The Oxford Handbook of Construction Grammar

Sign-based Construction Grammar

unmarkedfeature-values.Instead,SBCGpresumesaderivationalconstructionthatconverts(or‘pumps’)massandcountnounstodet+(i.e.,determined)nouns.

(9.)Thisanalysismightberegardedastheconstructionalequivalentoftheinterpolated-operatormodelproposedbyJackendoff(1997a:chapter3)withinamodularistframework.

(10.)Sag(2012)doesnotprovideaformalrepresentationoftheCausedMotionconstruction;Figure8.10isbasedonhisrepresentationoftheVerb-Wayconstruction(Sag2012:182).

(11.)NotethattheCausedMotionconstructionproducesanactive-voicelexeme.Anadditionalderivationalconstruction,Passive,isrequiredtolicensepassivepredicationslikeThefoamwassneezedoffthebeer.Asmentioned,Passiveisavalence-reducingderivationalconstruction.

LauraA.MichaelisLauraA.MichaelisisanAssociateProfessorintheDepartmentofLinguisticsandafacultyfellowintheInstituteofCognitiveScienceattheUniversityofColoradoatBoulder.SheisalsoafoundingmemberoftheeditorialstaffofthenewMoutonjournalLanguageandCognition.SheistheauthorofAspectualGrammarandPast-TimeReference(1998)andBeyondAlternations:AConstructionalModeloftheGermanApplicativePattern(2001),withJosefRuppenhofer.Sheistheco-editor,withElaineFrancis,ofMismatch:Form-FunctionIncongruityandtheArchitectureofGrammar(2003).SheiscurrentlyatworkonabookthatexploresnonstandardgrammaticalpatternsinEnglishconversationalspeech.Thiswork,tentativelyentitled,ConstructionGrammarandSyntacticInnovation,willbepublishedbyOxfordUniversityPress.Alongwithnumeroushandbooksandvolumes,herworkhasappearedinthejournalsLanguage,JournalofSemantics,JournalofPragmatics,JournalofLinguistics,Lingua,LinguisticsandPhilosophy,LinguisticsandStudiesinLanguage.

Page 123: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 1 of 11

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0009

FluidConstructionGrammarLucSteelsTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterfocusesonFluidConstructionGrammar(FCG),aformalismthatallowsConstructionGrammarresearcherstoformulatetheirfindingsinaprecisemannerandtotesttheimplicationsoftheirtheoriesforlanguageparsing,production,andlearning.ItexplainsthatFCGisnotintendedtodisplaceotherlinguisticproposalsforConstructionGrammarbuttobeanopeninstrumentwhichcanbeusedbyconstructiongrammarianswhowanttoformulatetheirintuitionsandanalysesinaprecisewayandwhowanttotesttheimplicationsoftheirgrammardesignsforlanguageparsing,production,andlearning.Thechapterfurthermoreshowsthattheconstruction-basedapproachisalsorelevantfortheinvestigationoflanguageprocessing,anddiscussesthemethodsandtechniquesadoptedfortheimplementationofcomplexlinguisticphenomena.

Keywords:FCG,ConstructionGrammar,languageparsing,linguisticproposals,languageprocessing,linguisticphenomena

9.1.Introduction

Aconstructionalapproachtolanguagehasabundantlyprovenitsworthindescriptivelinguisticsandisalsousedalmostuniversallyinsecondlanguageteaching.Moreoverempiricalevidencefromchildlanguageacquisitionshowsthatlanguagelearningcanbeunderstoodbytheprogressiveusage-basedacquisitionofconstructions(seeDiessel,thisvolume).Theconstructionalperspectivehasalsobeenveryproductiveforhistoricallinguists,andthereisnowalargebodyofclearexamplesshowinghownewconstructionstypicallydevelopfromthecreativeextensionofexistingconstructionsbyafewindividualstoaproductivecommonpatternthatisadoptedbythelinguisticcommunityasawhole(Fried2009b,thisvolume).

Thischaptershowsthattheconstruction-basedapproachisalsorelevantfortheinvestigationoflanguageprocessing.ItlooksinparticularatFluidConstructionGrammar(FCG),aformalismthatattemptstocaptureintuitionsandtheoreticalnotionsfromcognitivelinguisticsingeneralandConstructionGrammarinparticular.FluidConstructionGrammarisnotintendedtodisplaceotherlinguisticproposalsforConstructionGrammarbuttobeanopeninstrumentthatcanbeusedbyConstructionGrammarianswhowanttoformulatetheirintuitionsandanalysesinaprecisewayandwhowanttotesttheimplicationsoftheirgrammardesignsforlanguageparsing,production,andlearning.

TherearemanywaystoimplementConstructionGrammars,dependingonwhatrepresentationalandcomputationalmechanismsareadoptedasunderlyingfoundations.FCGusestechniquesnowcommoninformalandcomputationallinguistics,suchastherepresentationoflinguisticstructureswithfeaturestructures(Copestake2002),andtheuseofunificationforapplyingconstructionstoexpandlinguisticstructuresinlanguageparsingandproduction,aspioneeredinFunctionalUnificationGrammar(Kay1986),andalsousedinLexicalFunctionalGrammar(Dalrympleetal.1995),andHead-drivenPhraseStructureGrammar(PollardandSag1994).Likemanyothercomputationallinguisticsefforts,theFCG-systemisembeddedwithinacontemporaryCommonLISP-based

Page 124: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 2 of 11

programmingenvironmentfromwhichitinheritswell-testedmechanismsforrepresentingandprocessingcomplexsymbolicstructures.OtherproposalsforoperationalizingConstructionGrammar,suchasEmbodiedConstructionGrammar(BergenandChang2005,thisvolume)andSign-BasedConstructionGrammar(Michaelis2010a,thisvolume)drawonmechanismsarisingfromthesamecomputationaltraditionbutusethemindifferentways.Giventhecurrentstateofthefield,itishighlybeneficialthatmanyapproachesareexploredinordertodiscoverthebestwaytoformalizeandimplementConstructionGrammars.

FluidConstructionGrammarhasbeenfullyimplementedinasystemcalledtheFCG-system,whichismadeavailableforfreetotheresearchcommunity(http://www.fcg-net.org/).TheFCG-systemcontainsacorecomponent(calledtheFCG-interpreter)thatperformsbasicoperationsneededforparsingandproduction,aswellasvarioustoolstoaidlinguisticresearch,suchasatoolforbrowsingthroughlinguisticstructuresandatoolformonitoringthesuccessrateofagrammarwhenprocessingasetoftestcases.TheFCG-systemhasbeenunderdevelopmentfromaround1998inordertosupportexperimentsinmodelinglanguageevolutionusinglanguagegamesplayedbyautonomousrobots(Steels1998).Sincethen,ithasundergonemajorrevisionsandenhancementsandthesystemisstillcontinuouslybeingadaptedandrevisedtocopewithnewlinguisticphenomenaandnewprocessingchallenges.Nevertheless,thesystemisalreadysufficientlystablethatitcanbeusedtotacklesophisticatedissuesintherepresentationandprocessingoflanguage.AfulloverviewandmanyexamplesaregiveninSteels(2011a)andcomputationalissuesaretreatedinmoredepthinSteels(2011b).

TherearetwodifferentlevelsfromwhichFluidConstructionGrammarcanbeapproached.Thefirstlevelistheprocessinglevel.ItconcernsthefundamentalprimitivedatastructuresandoperationsthatareavailableforwritingConstructionGrammarsandthemachineryforcomputingparsetreesorproductiontrees.Itturnsoutthatevensimpleconstructionsarequitecomplicated.Verbalaccountsofconstructionsusuallysweepprocessingissues,suchasheuristicsfordampingcombinatorialsearch,underthecarpetbuttheyneedtobemadefullyexplicithere.Thesecondlevelisthedesignlevel.Itconcernsmethodsandtechniquesthathavebeendevelopedforcopingwiththecomplexityofwritingrealgrammars.Thischapterdiscussesbrieflyeachoftheselevels.

9.2.TheProcessingLevel

FCGusestransientstructurestorepresentalltheinformationaboutthesentencebeingparsedorproduced.Transientstructuresconsistofasetofunits,roughlycorrespondingtomorphemes,words,orphrases,andinformationattachedtoeachoftheseunitsintheformoffeaturesandvalues.TheexampleinFigure9.1showstheoutlineofthetransientstructurefortheGermanphrasederBlocklinksvonmir(‘theblockleftofme’)(seeSprangerandLoetzsch2011fordetailsofthegrammarunderlyingthisexample)asitisdisplayedintheFCG-interface.Atfirstsightthisstructurelookslikethekindoftreesfoundinallparsingandproductionsystems,exceptthatitisvisualizedfromlefttorightforthesemanticandfromrighttoleftforthesyntacticstructure.Thenamesoftheunits(suchas‘left-unit-4’or‘von-unit-1’)arepurelyconventional.Thenameshavebeenchosentomakeitpossibletoseewhatisgoingon.Theindicesareaddedbecausetheremaybeseveralvon-unitsorleft-units.Theseindicesareautomaticallycomputedwhennewunitsarecreatedduringprocessing.

Clicktoviewlarger

Figure9.1. TransientstructurecreatedduringtheproductionorparsingoftheGermanphrasederBlock

Page 125: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 3 of 11

linksvonmir.Topfigureshowsthesemanticpoleandbottomonethesyntacticpole.

Whenweclickononeoftheseboxesthefeaturesassociatedwitheachunitrevealthemselves.Thesefeaturesmayconcernanyleveloflanguage:pragmaticandsemanticinformationaregroupedinthesemanticpoleoftheunitandsyntactic,morphological,andphonologicalfeaturesinthesyntacticpole.Forexample,ifweclickon‘left-unit-4’weseethesemantic(left)andsyntacticpole(right)associatedwiththisunitasshowninFigure9.2.Whichfeaturesareadoptedforaparticulargrammarisentirelyopentothegrammardesigner.Weseehereonthesemanticpoleinformationaboutthemeaningofthewordlinksanditssemanticcategories(namelythatitisanangularlateralspatialcategory).Thesyntacticpolecontainsaformfeaturewithinformationonwhatisthestemforthisunit(namely‘link’),thestringitcovers(namely‘links’)andasyn-cat(syntacticcategory)featurecontaininginformationrelevantformorphology(case/gender)andthelexicalcategoryorpartofspeech(lex-cat).TheFCGprocessinglevelisentirelyagnosticaboutwhichfeaturesandvaluesareusedinthegrammar.Agrammardesignercanintroducenewonesatanytimebyjustusingthem.

Transientstructuresalsorepresentthehierarchyofunitsandsubunits,asshowninFigure9.3,whichshowsthesemanticandsyntacticpoleofspeaker-unit-1(coveringthewordmir)andpronoun-unit-2,whichisitsparent-unit.Theparent-unitwascreatedbytheapplicationofaconstruction(seelater).Noticethefeaturessem-subunitsandsyn-subunitswhichareeachfilledwithalistofsubunits,inthiscaseonlyspeaker-unit-1.Theexplicitrepresentationofsubunitsmakesitpossibletorepresenttreeswhicharenotstrictlyhierarchical,andtheseparationbetweensemanticandsyntacticsubunitsmakesitpossibletohavedifferencesinsemanticorsyntacticstructure.

Clicktoviewlarger

Figure9.2. Detailsofleft-unit-1inthetransientstructurecoveringthewordlinks(ontheleftthesemanticpoleandontherightthesyntacticpole)

Constructionshavethesamestructureastransientstructures.Theyalsoconsistofunits,features,andvaluesforthesefeaturesandtheinformationisagainorganizedintotwopoles.ConstructionsinFCGareconsideredtobebi-directionalassociations,inthesensethattheyestablishbi-directionalmappingsbetweenmeaningandformthroughtheintermediaryofsyntacticandsemanticcategorizationsandtheyareconsequentlyusablebothaspartofaproductionprocessthattranslatesmeaningintoformoraparsingprocessthattranslatesformintomeaning.AnexampleofaconstructionisshowninFigure9.4.Itisalexicalconstructionthatdefinesthemeaningcontributedbythestem‘link-.’Theconstructionintroducestheinformationthatthisstemintroducesthelateralcategory‘left’whichcanfunctionasaspatialadjective,alateral-adverb,orapreposition.

Page 126: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 4 of 11

Clicktoviewlarger

Figure9.3. Smallsectionofthehierarchicalstructurefortheword“mir”(toppartgivesthesemanticpolesandbottompartgivesthesyntacticpoles)

Clicktoviewlarger

Figure9.4. Exampleofaverysimplelexicalconstructionforthestem“link-”.Thetoppartdescribesthesemantic(left)andsyntactic(right)constraintsontheapplicationoftheconstruction.Thebottompartdescribeswhatiscontributedbytheconstrainttothetransientstructureforthesemanticpole(left)andthesyntacticpole(right).

Noticethattherearetwopartstoconstructions(seeFigure9.5).Thetoppartistheconditionalpartanddefineswhathastobethereinthetransientstructurebeforetheconstructioncanapply.Thereisbothasemanticandasyntacticconditionalpart.Thebottompartisthecontributingpartofaconstruction;itdefineswhatwillbeaddedtothetransientstructure.Againthereisbothasemanticandasyntacticcontributingpart.SoFCGrulesarenotlikegenerativegrammarrulesthatrewriteanonterminalsymbol,theyalwaysassociatesemanticwithsyntacticstructure.Moreovereachconstructionnotonlydefinesasyntacticstructurebutalsohowthatsyntacticstructureistobeinterpretedsemantically,whichmakesitpossibletohaveatighterintegrationofsyntaxandsemantics,ascomparedtoproposals(asinMontaguegrammarforexample)wheresyntaxandsemanticsiskeptseparate.

Figure9.5. GeneralstructureofaconstructioninFCG.Thereisaconditionalpartandacontributingpart.Inparsingtheconditionalpartofthesyntacticpoleismatchedfirstandtherestismergedintothetransientstructureifsuccessful.Inproductiontheconditionalpartofthesemanticpoleismatchedfirst

Page 127: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 5 of 11

andtherestismergedintothetransientstructureifsuccessful.

Constructionsareappliedinaprocessofmatchingandmerging:

•MatchingmeansthattheconditionalpartCofonepoleoftheconstruction(thesemanticpoleinproductionorthesyntacticpoleinparsing)iscomparedwiththecorrespondingpoleTinthetransientstructuretoseewhethercorrespondentscanbefoundforeveryunit,feature,andvalue.CandTmaybothcontainvariables.Variablesaredenotedbynamesprecededbyaquestionmark.Aspartofthematchingprocess,thesevariablesgetboundwithspecificvalues(orothervariables)inthetargetusingaunificationoperationfamiliarfromlogicprogrammingorotherfeature-basedformalisms(forfurtherdetails,seeSteelsandDeBeule2006).Forexample,iftheconstructioncontainsthevalue(stem?left-unit‘link’)andthetransientstructurecontainsthevalue(stemleft-unit-14‘link’)forthesamefeature,thenthesetwovaluesmatchif?left-unitisassumedtobeboundtoleft-unit-14.

•Mergingmeansthattheconditionalpartoftheotherpoleoftheconstruction(thesyntacticpoleinproductionorthesemanticpoleinparsing)iscombinedwiththecorrespondingpoleinthetransientstructure,inthesensethateverythingmissinginthetargetpoleofatransientstructureisaddedunlesstheyareinconflict.Inaddition,thecontributingpartsofbothpolesfromtheconstructionareaddedtothecorrespondingpolesofthetransientstructure.

AnexampleoftheapplicationoftheconstructionshowninFigure9.4duringaproductionprocessisshowninFigure9.6.Theinitialtransientstructure(beforeconstructionapplication)isshownatthetop.Therearealreadysomeunitswhichcoverpartofthemeaning.Theconstructionmatcheswithanotherpartofthemeaning(namely‘(bindlateral-category?cat-5left)’)andcreatesanewsubunit(calledleft-unit-14),whichcontainsthecoveredmeaninginthesemanticpole,aswellassemanticcategoriescontributedbytheconstruction.Thesyntacticpolecontainsinformationabouttheformandthesyntacticpropertiesthatcouldbededucedsofar.

AcrucialpropertyofFCGisthatexactlythesameconstructionappliesbothinparsingandinproduction.ThismeansfortheconstructionshowninFigure9.4thatifduringparsingaunitisfoundwiththestem‘link-‘thenthistriggersthesameconstructionandbuildsthesamestructure.Thismirrorpropertyhasalargenumberofadvantages.Forexample,duringparsingitispossibletostartapplyingconstructionsinatop-downmannertopredictpropertiesofwordsorphrasesthathavenotbeenpronouncedyetortofillininformationaboutfragmentsoftheutterancethathavenotbeenunderstoodproperlyorareunknown.Duringproductionitispossibletoengageinself-monitoring,becausepartialstructuresconstructedbyapplicationofconstructionsfrommeaningtoformcanbere-enteredinaparsingprocess,inordertosee,forexample,whetheranycombinatorialsearchisoccurringorwhethertheutterancetobepronouncedindeedexpressestheintendedmeaningasaccuratelyaspossible.

Clicktoviewlarger

Figure9.6. Anexampleofconstructionapplication.Thetopshowsthetransientstructurebeforeapplicationandthebottomshowstheexpandedtransientstructureaftertheoperationofmatchingandmerging.Thetransientstructurehasanadditionalunitwithasemanticpole(left)andsyntacticpole(right).

Page 128: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 6 of 11

Clicktoviewlarger

Figure9.7. PartofthesearchspaceforthephrasederBlocklinksvonmir.Thesearchbranchesoutwhentherearemultiplepossibilitiesthatcanbeexplored.Abestfirstsearchiscarriedoutbasedonvariousheuristiccriteria.

Languageusersoftenhavestoredseveralalternativeconstructionsintheirmemorybecausethereisunavoidablyalotofvariationinthelanguagetheyencounterandbecauseofconstructionalsynonymy,syncretism,homonymy,andambiguity.Theconstructionshaveanassociatedscorethatreflectstheirpastsuccessinutilization.Usuallymorethanoneconstructioncanpossiblyapplytoagiventransientstructureandconsequentlytheexplorationofasearchspaceisunavoidable.PartofsearchspacefortheexampleofderBlocklinksvonmirisshowninFigure9.7.Thedifferentnodesinthesearchspacearerepresentedinatree.Atthemomentofparsing‘links,’itisstillunclearwhetherthephrasewillendwith‘links’(asin‘derBlocklinks’),orwhetheritwillcontinue,asitdoesinthiscase.FCGsupportsstandardheuristicbest-firstsearch.Ascoreiscomputedforeverypossibleexpansionandthenthetransientstructurewiththehighestscoreispursuedfurther.Thisscoremight,forexample,takeintoaccountwhatconstructionshadmostsuccessinthepastandarethereforeprobablymoreentrenchedinthepopulation.Anothercomponentthatdeterminesthescore,particularlyoffinalnodesinthesearchspace,aregoaltests.Theyexamineatransientstructuretoseewhetheritsatisfiesdesiredcriteria.TheFCG-interpreterperformsbacktrackingtoearliernodesinthesearchspaceifaparticulartransientstructurereachesadead-endorifanunsatisfactoryendstatewasreached.

Languageuserspresumablyhavehundredsofthousandsofconstructionsinmemoryandtheyareapplyingthematincrediblespeeds.Increasingefficiencyanddampingcombinatorialexplosionsinsearchisthereforeoneofthekeychallengesinbuildingoperationalparsingandproductionsystems.Thischallengecanpartlybetackledbywritinggrammarsinsuchawaythattheyminimizesearch(seesection9.3below).Inaddition,FCGhastwovariousmechanismstohelpgrammardesignerscontrolatamorefine-grainedleveltheselectionandapplicationofconstructions:

•Constructionscanaddfootprintstotransientstructures,inasensetaggingthepointswheretheyhavemadeachange,sothatlateronwhentheconstructionistriedagainonthesametransientstructuretheconstructiondoesnotre-applyendlessly.Byconventionthenameofthefootprintisequaltothenameoftheconstruction.FeaturefootprintscanbeseenforexampleinFigure9.2.Weseethatleft-unit-14hasinthesyntacticpoletwofootprints:left-lexandlinks-morph,whichhavebeenaddedbytheapplicationoftwoearlierconstructions.WecanalsoseetheuseoffootprintsintheexampleconstructioninFigure9.4.Wherethetop,conditionalpolesarelookingforthepresenceoffootprintsanddonotproceedwhentheyarethere(==0left-lex)andthebottomaddsthefootprintstotherelevantunit.

Clicktoviewlarger

Figure9.8. Exampleofdependencyrelationsbetweenconstructions.Suchnetworkscanbeusedtosupportpriming,whichaidsconsiderablytoreducesearchandspeedupconstructionaccess.

•Constructionsareorganizedinsetsandnetworks.Setsareusefultoensurethatacertainsetofconstructions(e.g.,allmorphologicalconstructions)haveoperatedbeforeanotherset(e.g.,thephrasestructureconstructions).Networksgiveamorefine-grainedwaytoprioritizetheexecutionofconstructions.Forexample,

Page 129: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 7 of 11

onenetworkcouldbebasedonthegenerality/specificityrelationsbetweenconstructionsused,forexample,torepresentfamiliesofconstructions.Anothernetworkisbasedonconditionaldependencies(seeFigure9.8fromWellensandDeBeule2010).OneconstructionC-1(e.g.,adeterminer-nominalconstruction)conditionallydependsonanotherconstructionC-2(e.g.,anominalconstruction)ifthetriggeringofC-2createssomeoftheconditionsunderwhichC-1couldpotentiallytrigger.Conditionaldependenciescanbeusedforpriming.Ifanominalconstructionwasabletooperate,itwillmakeitmorelikelythatthedeterminer-nominalconstructionisapplicable,andconverselyifnonominalconstructionwasabletoapplyitdoesnotmakesensetotryadeterminer-nominalconstructioneither.

•Networksofconstructionsthathaveproventobeusefulcanbechunked.Chunkingmeansthattheinformationrequiredorsuppliedbyindividualconstructionsiscombinedintoasingleconstructionwhichcanthusbematchedandmergedmoreefficiently.Thischunkingprocessdoesnotonlyreduce.Intermediarystepsthatarenolongernecessarycanberemoved(Stadler2011).Thederivationofconstructionnetworksandthechunkingprocesshasbeenautomatedandhappensbasedonwhichconstructionstriggereffectively.

9.3.TheDesignLevel

Writingoperationalconstructionsisaverydifficultexercisefortworeasons.Firstofall,manyfactorsnormallyinterveneinasingleconstruction;indeed,itisoneofthemaintenetsofConstructionGrammarthatlinguisticknowledgeshouldbepackagedinsuchawaythatasmanyconstraintsaspossiblegetincorporatedineachconstruction.Soaconstructioncouldincludeaphonologicalconstraint(e.g.,vowelharmonytoselectamorphologicalaffix)orapragmaticconstraint(e.g.,whichconstituentisbeingemphasizedinthesentence).Thismakesprocessingmuchmoreefficientcomparedtohorizontallystructuredgrammarswherethesyntacticlevelisviewedasautonomousandcannotincorporateissuesofmeaningorphonology.Second,therearemanyinteractionsbetweenconstructionsthatmaynotbeimmediatelyobvioustothegrammardesigner:constructionscanbeincompetitionwitheachotherbecausetheycoverthesamemeaningorthesameform,andtheimplicationsofthebestchoicecouldonlybecomevisiblemuchlater.

Inthedesignorinvestigationofhighlycomplexsystemsitisoftenusefultointroducehigherlevelabstractionsthatthentranslateintodetailedstructuresandprocesses.Forexample,computerprogramsareusuallywritteninahighlevelprogramminglanguage(likeLISPorPython)andcodewrittenatthislevelisthentranslatedautomaticallybycompilersorinterpretersintoahugenumberofdetailedoperationsthatcanbeexecutedatthemachinelevel.Thegrammarofahumanlanguageiscertainlyahighlycomplexsystemanditthereforemakessensetousethesameapproach.

ThroughvariouscasestudiesinFCG,asetofdesignpatternsisgraduallybeingdiscovered,someofwhichhaveinfactalreadyalongtraditioninlinguistics.Thenotionofadesignpatterncomesfromarchitectureandisalsowidelyusedincomputerscience.Anarchitecturaldesignpatternisforinstancetheuseofadomestructure(suchastheSantaMariadelFioreDuomoinFlorencebuiltbyBruneschelli).Therearegeneralprinciplesofdomedesignbutspecificdetailsdependontherequiredsizeandheightofthespacethatneedstobecovered,aswellasonaestheticconsiderations.Inthecontextofgrammar,adesignpatterncircumscribesthecoresolutiontoaparticularaspectofgrammar,notjustinadescriptivewaybutalsointermsofprocessingandlearningoperations.Thespecificdetailshowthedesignpatternisinstantiatedinaparticularlanguagestillneedtobeworkedoutandthedetailswillbesignificantlydifferentfromonelanguagetoanotherone.Somelanguagesmayevenusecertaindesignpatternsthatareentirelyabsentfromothers.

Herearetwoexamplesofdesignpatterns:

•Manylanguagesfeaturecomplexmorphologicalandagreementsystemswhichgroupasetoffeatures(suchasnumber,case,andgender).Butthereisalmostneverasimplemapping.Oftenthesamewordormorphememayexpressdifferentcompetingfeaturebundles(e.g.,theGermanarticledieexpressesthenominativeandaccusativefemininesingular,aswellasallpluralnominativeandaccusativecases).Adesignpatterntoefficientlyhandletheprocessingoftheseambiguitiesisafeaturematrix,reminiscentofthedistinctivefeaturematricesinphonology(vanTrijp2011).Itcontainsrowsandcolumnsforthedifferentdimensionsandeither+or−ifthereisaknownvalueoravariableifthevalueisnotyetknown(seeFigure9.9).Puttingthesamevariableindifferentslotsofthematrixcanbeusedtocaptureconstraintsbetweenvalues.Givensuchfeature

Page 130: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 8 of 11

matrices,agreementandpercolationphenomenacanbehandledbythestandardmatchingandmergingoperationsofunification-basedgrammars.Ambiguitydoesnotneedtotranslateintoexplodingbranchesinthesearchtreebuttranslatesintoopenvariables,whichgetboundwhenevertheinformationbecomesavailable.

Clicktoviewlarger

Figure9.9. ThemorphologicalconstructionfortheGermanarticledie.Thistypeofconstructionhastwosyntacticpolesinsteadoftheusualsemantic/syntacticpoles.Onseeingthestring‘die,’theconstructioncreatesaunitforit,whichhasafeaturematrixforcaseandgender.‘Die’iseithernominativeoraccusativefeminine(singular/pluralnotrepresentedtosimplifythisexample).Butallotherpossibilitiesareexcluded.

•Allhumanlanguagestendtoreusethesamewordformsfordifferentpurposes.Forexample,theEnglishwordslowmaybeusedasanadjective(Theslowtrain),averb(Theyslowdown),apredicate(Thetrainwasslow)oranoun(Theslowgofirst).Itwouldbehighlycostlytocreatenewnodesinthesearchspaceeachtimethewordslowisencountered.Analternativeistouseadesignpatternbasedonadistinctionbetweenactualandpotential.Aparticularwordhasthepotentialtobelongtoseveralwordclassesbuttheninthecourseofprocessingitbecomesclearwhichoftheseistheactualvalue.Theconstructionsdiscussedearlierallusethisdesignpattern.Forexample,thesyntacticpoleofleft-unit-14inFigure9.7(bottom)showsthatthelex-cathasthepotentialtobespatial-adjectiveorlateral-adverb/preposition.Thesamefeaturealsohasaslottosaywhattheactualvalueisgoingtobe.Thisvaluewillbesetbyotherconstructionsthatmakeuseofthisunit.Byexplicitlyrepresentingthepotential,itispossibleforconstructionstoeliminatesomepossibilitiesoraddothers(e.g.,throughcoercion)beforeafinaldecisionismade.

TheFCG-systemcomeswithacollectionoftemplatesforsupportingsuchdesignpatternsandwithfacilitiesforintroducingnewtemplatesifthegrammardesignerwishestodoso.Thesetemplatesprovidewaystoleaveoutmanydetails,particularlydetailsrelatedtotheoperationalizationofconstructions.Theyhelptofocusonthelinguisticaspectsofconstructionsandbridgethegapbetweenthedetailedoperationallevelandthelevelatwhichlinguistsusuallywork.Forexample,thereisatemplatefordefiningtheskeletonoflexicalconstructionsthathasslotstospecifythemeaningandthewordstem.Thistemplatethencreatesthenecessaryunits,features,andvalues,includingthestructurebuildingoperationsthatarerequired.Therearealsotemplatesfordefiningthecomponentsofphrasalconstructions,includingtemplatesforspecifyingwhatagreementrelationsholdbetweentheconstituents,howinformationfromconstituentspercolatestotheparentunit,howthemeaningsofthedifferentconstituentsgetslinked,whatmeaningsareaddedbytheconstructiontothemeaningssuppliedbythedifferentunits,andwhatadditionalformconstraintsareimposedbytheconstruction.Othertemplatesareavailablefordefiningmorecomplexfeaturematricesandthegrammaticalparadigmsonwhichtheyarebased,andforusingthesefeaturematricestoestablishagreementandpercolation,fordefiningthetopologyoffieldsandtheconstraintsunderwhichaconstituentcanbe‘grabbed’byafield,andsoon.TheinventoryofpossibletemplatesnowusedinFCGiscertainlynotclaimedtobeauniversalorcompleteset,onthecontrary,wemustexpectthatthisinventorykeepsbeingexpandedandelaboratedasnewdesignpatternsareuncovered.

9.4.Learning,Fluidity,andRobustness

FCGcontainsmanycomponentsthatarealsofoundinotherformalandcomputationalapproachestogrammar,simplybecausethesecomponentsarenecessarytoachievelanguageprocessing.FCGdiffersfromotherapproachesbecausealllevelsoflinguisticanalysiscanbetakenintoaccountinaconstruction,notjustsyntax,andbecauseconstructionsareseenasbi-directionalassociationsthatcanbeusedunchangedinparsingandproduction.Besidesitsuseinconstructinggrammarsandininvestigatingparsingandproductionprocesses,FCGhasbeendevelopedspecificallytosupportresearchonlanguagelearningandlanguageevolution.Thesetwo

Page 131: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

Page 9 of 11

topicsareintimatelyrelatedbecauselanguageevolutionisonlypossibleiflanguageusersarenotonlyabletoinventnewlinguisticforms(e.g.,newwordsornewmeaningsforexistingwords),ontheonehand,butalsotoacquireinventionsthathavebeenadoptedbyothers.Moreovertherehastobeawaytocoordinatetheunavoidablevariationinlanguageusesothatasharedsetofconventionsarisesandismaintainedbythepopulation,eventhoughthereisnocentralcoordinator,norapriorinnategrammarortelepathy.

FCGproposesanumberofstepsforhandlinglearningandevolution.Thefirststepistoembedtheproductionorcomprehensionofsentencesinacompletesemioticcycle.Speakingandunderstandingindeedinvolvesmuchmorethanonlytheparsingorproductionofsemanticandsyntacticstructures:

•Thespeakermustdevelopandmaintainaninternalworldmodelbasedonperceptionandaction.Thegroundingprocessesneededforthisincludesegmentation,featureextraction,objectrecognition,eventclassification,objecttracking,objectmanipulation,etc.

•Next,thespeakermustconceptualizewhattosay,whichmeansthathemustcategorizerealitytoachievethecommunicativefunctionsthathewantstoachieve.Thisprocessyieldsthemeaningofthesentence.

•Thethirdstepistotranslatethismeaningintoinformationforspeecharticulationbyapplyingtheinventoryoflexicalandgrammaticalconstructions.Thiswillusuallyrequiretheconstructionofelaboratesyntacticandsemanticstructures,andtheFCG-systemisintendedtomodelthisparticularstep.

•Finally,speechandgesturesmustbearticulatedtoproducetheactualutterance.

Thehearerneedstogothroughanothersetofprocessesthatmirrorthoseofthespeaker:

•Hemustfirstofallprocessthespeechsignalbyapplyingabatteryofsignalprocessingandpatternrecognitionprocessestogetareasonablesetofhypothesesaboutthespeechelementsthatmightbepresentandtheotherphoneticfeatures.

•Thehearermustthenapplyhisowninventoryoflexicalandgrammaticalconstructionstoreconstructaswellaspossiblethemeaningoftheutterance.ThisisagainthepartthattheFCG-systemisintendedtocover.Constructionsarenowappliedintheotherdirection,fromsyntaxtosemantics.

•Finally,thehearermustconfrontthemeaningresultingfromtheparsingprocesswithhisowninternalworldmemoryandunderstandingoftheongoingdialoguecontextinordertofindaninterpretationoftheutterancethatfitswithhisorherownexpectationsandobservations.

Atpresentwehavecomputationalmodelsofallaspectsofthissemioticcycleandevenembedtheminoperationalsystemsworkingonautonomousrobots.FluidConstructionGrammarconcernsinprincipleonlyonepartofthesemioticcycle,however,byembeddingparsingandproductioninthislargercycle,itispossibletogetmorerobustlanguageprocessingandestablishtheconditionsforlearning.Forexample,evenifthemeaningreconstructedbytheparserisincomplete,thehearer'sinterpretationcomponentmaystillbesufficientlypowerfultofillinmissingelementsbasedonconsideringthecurrentsharedcontextofthedialog.Or,evenifconstructionsaremissingtoexpresseverythingthatneedstobeexpressed,thespeakermayalreadyproducefragmentsthatthencouldbepotentiallyunderstoodbyanintelligentlistener.

FluidConstructionGrammarputsalotofemphasisonreversibility.Everyconstructioniscarefullydesignedinsuchawaythatitcanbeusedunchangedbothinparsingandproduction.Apartfromleadingtoamoreparsimoniouslanguagesystem,thismakesitpossibletorunatanymomentalanguageprocessinreversedirection.Duringparsing,partiallyobtainedstructuresandmeaningsalreadyderivedfromthecontextorfromlanguagefragmentscanbeproducedinordertopredictwordsthatstillhavetocome,orinordertofillindetailsforwordsthathavenotbeenunderstoodproperly.Duringproduction,thespeakercancontinuouslymonitorhislanguagebyre-enteringsentencesorfragmentsofsentences,thussimulatinghowwellhemightbeunderstood.

Thenextsteptowardmorerobustnessandlanguagelearningistomaketheprocessingofconstructionsitselfmoreflexible.First,allconstructionsshouldbeappliedthatcanbeapplied,evenifthisdoesnotleadtoacompletelycoherentparsetree.Manysentencesinordinarydiscourseareungrammaticalorincomplete,andtheparsershouldnotgetstuckwhenithitsaproblem.FCGisnotconcernedwithgrammaticalityjudgmentsbutwithtryingtocarrytheparsingprocessasfaraspossiblewiththeavailableinventoryofconstructions,sothatitmaybepossibletointerpretwhatwassaid.Second,whenaconstructioncannotapplyfully,apartialmatchmaybe

Page 132: The Oxford Handbook of Construction Grammar

Fluid Construction Grammar

attempted.Thematchingstepintheapplicationofaconstructioncanbeskippedandallunitsandtheirfeaturesandvaluesaremergedwiththetransientstructure.Themergingprocesswouldstillgetblockedwhenthereareincompatiblefeaturevalues,butifitcangothrough,itwillimposevariousfeaturesandevencreateunitsthatshouldhavebeenthere.Forexample,ifaphrase‘a<unknown>mouse’isreceivedasinput,aflexiblematchwouldbeabletoimposethatthe<unknown>wordisanadjectivebasedonaflexiblematchoftheadjectival-nominalconstruction.Thisisalsothewayinwhichcoercioncanbemodeled,suchasofsentenceslikeAdelesneezedthenapkinoffthetableinwhichanintransitiveverbiscoercedtobecomeaverbexpressingacausedmotionbecausethepatternoftheCausedMotionconstruction(asinAdelepushedthebookoffthetable)hasbeenusedhere.Coercionisachievedbyimposingtheapplicationofaconstructioneventhoughnotallitscriteriaaresatisfied,intheexampletheneedforatransitiveverbisoverruledsothattheCausedMotionconstructioncanapply.

ThefinalfacilityavailableinFCGforhandlingfluidityandrobustnessallowsformonitoringtheoutcomeofastepinaparticularprocessusingdiagnostics.Thesediagnosticsrunatameta-levelinparallelwithroutineprocessing.Forexample,theremaybeadiagnosticthatistestingwhetherallthewordsintheutterancewerehandledbylexicalorgrammaticalconstructions,orwhetherthemeaningssuppliedbyallindividualwordscouldbelinkedtogethertoformthemeaningoftheutteranceasawhole.Thesediagnosticscanalsotriggerontheresultsofre-entrance,forexample,beforesendinganutteranceoftothespeechsystem,thespeakercouldre-entertheutteranceandgothroughparsingandinterpretation,thussimulatingthelistener.Adiagnosticcouldthencomparethemeaningthatresultsfromre-entrancewiththemeaningthatthespeakeroriginallywantedtoconvey.

Theproblemssignaledbyadiagnosticarethenhandedtooneofthepossiblerepairstrategies,whichwilltrytodealwiththeproblem.Oftenmorethanonerepairstrategyispossibleandtheonewiththehighestpotentialforleadingtoasuccessfulinteractionispreferred.Thediagnosticsignalingthatasentencecouldnotbeparsedmight,forexample,triggerthecoercionofaconstructionthatwaspartiallymatchingtostillproduceacompleteinterpretation,oradiagnosticsignalingthatsomemeaningcouldnotbeexpressedcouldleadtothesearchforalexicalconstructionthatisalreadypartiallymatchingandcouldbeexpanded.

9.5.Conclusions

FCGshowsthatConstructionGrammarneednotstayataverballevelonly.ItisentirelypossibletoformalizenotionsofConstructionGrammarandusethisformalismforachievingtheparsingandproductionofsentences.Itiseventhecasethatsuchformalizationshaveanumberofadvantagescomparedtothoseusedtraditionallyincomputationallinguistics,inparticularfromtheviewpointofefficiency(becauseconstructionscanstretchalllevelsoflinguisticanalysis),fromtheviewpointofrobustness(becauseconstructionscanbeflexiblyapplied),andfromtheviewpointofmodelinglanguageasanopen,adaptivesystem,whichitcertainlyis.

LucSteelsLucSteelsisICREAResearchProfessorattheInstituteforEvolutionaryBiology(CSIC,UPF).HestudiedlinguisticsattheUniversityofAntwerp(Belgium)andcomputerscienceattheMassachusettsInstituteofTechnology(USA).In1983hebecameaprofessorofcomputerscienceattheUniversityofBrussels(VUB)andco-foundedtheVUBComputerScienceDepartment(FacultyofSciences).HealsofoundedtheSonyComputerScienceLaboratoryinParisin1996andservedasitsfirstdirector.HismainresearchfieldisArtificialIntelligence,andhecurrentlyfocusesontheoriesoftheoriginsandevolutionoflanguageusingcomputersimulationsandroboticexperimentstodiscoverandtestthem.

Page 133: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 1 of 16

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0010

EmbodiedConstructionGrammarBenjaminBergenandNancyChangTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterfocusesonEmbodiedConstructionGrammar(ECG),anothercomputationalimplementationofConstructionGrammar.Itpointsoutthatthedrivingquestionofthisframeworkishowlanguageisusedinactualphysicalandsocialcontexts,andexplainsthatECGisanattempttocomputationallymodelthecognitiveandneuralmechanismsthatunderliehumanlinguisticbehavior.Thechapterevaluatestheroleofmentalsimulationinprocessingandoutlineshowlanguagecanbeseenasininterfacetosimulation.ItalsoshowshowconstructionsarerepresentedinECGanddescribesanECG-basedmodeloflanguagecomprehension.

Keywords:ECG,ConstructionGrammar,linguisticbehavior,mentalsimulation,languagecomprehension

10.1.Introduction

10.1.1ACognitiveScienceApproachtoGrammar

Theoriesofgrammararedefinedbythequestionstheyask.Intheanalyticallinguistictradition,thisquestionisoftensomethinglike:

Whatisthemostparsimoniousformalaccountthatpredictsallgrammatical(oracceptable)utterancesinalanguageanddoesnotpredictungrammatical(orunacceptable)ones?

Construction-basedapproachestogrammararedrivenbytheinsightthatmeaningplaysacrucialroleinansweringthisquestion:themeaningsofbothwordsandgrammaticalstructurescanaffectgrammaticality,andtheyshouldthereforebeincorporatedintogrammaticalrepresentations.Towit,thevariousapproachespresentedinthisvolumetakethistackinpredictingthegrammaticalityofutterancesofallsorts,fromthecreativetotheconventional.

Butfocusingonutterancegrammaticalityremovesutterancesfromtheirphysicalandsocialcontextsofuse.Theresultisanapproachtolanguagethatresemblestraditionalanatomy,inwhichutterancesaredissected,categorized,andexplainedintermsoftheirstructuralanddistributionalproperties.Insightsfromsuchanapproachcanbebothusefulandinformativeinaccountingforwhatutterancesareorpotentiallycouldbe.Theydonot,however,explainhumanlanguageuse—thatis,themechanismsandrepresentationsthathumanbeingsusetolearn,produce,andunderstandutterances.

Ifweareinterestedintheselatterissues,issuesmoreproperlysituatedwithincognitivesciencethananalyticallinguistics,thenwefindourinvestigationsofgrammar(andlanguagemoregenerally)guidedbyadifferentsortofquestion:

Whatcognitiveandneuralmechanismsdopeopleengagewhileusinghumanlanguage?

Page 134: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 2 of 16

EmbodiedConstructionGrammar(ECG)isanapproachtolanguagedefinedbythisquestion,onethatisnot,ofcourse,uniqueinadoptingthisorientation(see,e.g.,Bybee2006,thisvolume).Workinthisparadigmfollowstheconventionalscientificcyclethathasprovedfruitfulinthestudyofotheraspectsofhumancognition.Namely,observationsabouttheacquisition,production,orcomprehensionoflanguage(whichwewillbundletogetheras‘use’inthischapter)spurhypothesesaboutmechanismsthatmightaccountfortheobservations.Thentheoristsdevelopmodels,whichareformallyexpressedsoastocommunicateclaimsexplicitlytootherresearchers.Theyareoftenalsocomputationallyimplemented,whichallowsassumptionstobevalidatedandnewphenomenatobepredicted.Themodels’predictionsaretestedagainstnewobservations,andthemodelsarechangedorrejectedwhenobservationscontravenethem.

ECGshareswithotherconstruction-basedapproachesaninterestinhowconstructionsofvaryingtypescontributemeaningandfunctiontoutterances.Butitsemphasisisnotjustonwhatconstructionslooklikebutonhowtheyareused.Thatis,constructionsareincorporatedintomodelsoflanguageuse;ratherthanbeingjustdescriptiveobjects,astheyareinmostconstruction-basedapproachestogrammar,theseform-meaningpairingsarecomponentsofthehypothesizedsetofmechanismsengagedbylanguageusers.Anyconstructionproposedtoparticipateintheproductionorcomprehensionofagivenutteranceshouldthereforehaveobservableconsequencesinthatlanguageusageevent;thereshouldbenobehaviorallyvacuousconstructions.

Inthissense,eachconstructionalform-meaningpairrepresentsahypothesistobevalidatedthroughobservationsofbehaviorinnaturalandexperimentalsettings.Tofacilitatethebuildingofmodelsthatcanvalidatesuchhypotheses,constructionsinECGareexpressedinaformalnotationthathasastraightforwardcomputationalimplementation.Inshort,ECGtakestheinsightthatpeopleusegrammarmeaningfullyandfunctionally,andusesittobuildanempiricallydriven,computationallyimplemented,predictivetheoryoflanguageuse.

10.1.2TheComputationalLevel

ECGaskswhatcognitiveandneuralmechanismspeopleengagewhenusinglanguage.Thekindofansweroneseeksdependsontheintendedlevelofanalysis.Itmightbepossibleinprincipletoanswerthequestionatarelativelylowlevel—entirelyintermsofthebiology,chemistry,andphysicsofthehumansinvolvedinlanguageeventsandthesocialandphysicalcontextsinwhichtheyoccur.Thissortofreductioniscertainlyanattractiveultimateoutcome;cognitivescienceaimstoexplainahostofcognitivefunctionsintermsoftheirmaterial(includingbiological)underpinnings.Butforlanguage,wedonotyethavethemeanstofindanswersatthislevelofdetail;single-cellrecordingfromlivinghumansisonlyrarelypossible,andanimalmodelsprovideonlylimitedevidenceaboutspecificallyhumanlanguage.Moreover,itisnotobviousthatsuchalow-levelexplanationwouldbeespeciallyinformativetotheanalyst;knowinghowbillionsofcellsindividuallybehavemightnotshedmuchlightonthehigher-orderprocessesinwhichthosecellsparticipate,theircauses,effects,properties,andsoon.Infact,afullreductiontobiologymightnotbeanymorepredictiveofnewobservationsthanahigh-level,functionalcomputationalmodelofthemechanismsinvolved;itmightevenbelesscapableofmeaningfulpredictions.

Forthesereasons,ECGisarticulatedatahigher,functionallevel.Wedevelopcomputationalmodels,inwhichmechanismsaredescribedintermsofstructuresandprocesses:whatpeopleappeartoknowandhowtheyappeartolearnandusethatknowledge.Atthesametime,thesemodelsareproposedwithaneyetowardseekingconnectionstotheirunderlyingbiologicalsubstrate;towarddevelopingalinkingtheorybetweencomputationalandbiologicallevelsofdescription(Marr1982;Feldman2006).

ThecomputationallevelofdescriptionusedinECGisconsiderablymoredetailed,however,thanthekindsofdescriptionfoundinmostconstruction-basedaccounts(thoughseeMichaelis,thisvolume;Steels,thisvolume)andotherworkincognitivelinguistics.Constructionalaccounts,thoughinsightful,tendtostopshortofthelevelofformalprecisionneededforbuildingdetailedcomputationalmodelsofhowconstructionsarerepresented,used,andlearned.TheECGformalism(asintroducedinBergenandChang2005)isintendedasameansofbridgingthegapbetweentheselevels,usingasetofnotationsforthekindsofhigh-levelrelationsandconstraintsexpressed(typicallydiscursively)intheliterature.Theformalismitselfisthusnotalinguistictheoryperse;rather,itisatheoryofwhatconceptualdistinctionsarenecessaryandsufficienttoaccountforlanguagephenomena.

Theexpositioninthischapterwillmakeuseofformalnotationandwillmentioncomputationquiteabit,butitisnotintendedasathoroughintroductiontoeithertheECGformalismoritsassociatedcomputationalmodels.Infact,we

1

Page 135: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 3 of 16

willpresentrelativelysimpleanalysesandavoidmosttechnicaldetails;theseareavailableelsewhere(e.g.,Bryant2004,2008;FeldmanandNarayanan2004;Mok,Bryant,andFeldman2004;BergenandChang2005;Feldman2006;MokandBryant2006;N.Chang2008;Mok2008;Dodge2010).FormalnotationwillbeusedonlyattheserviceofhighlightingorexplicatingpropertiesoftheoreticalcomponentsofECG.Likewise,wewilldescribethecomputationalmodelsexpressingthesetheoreticalideasbutsayrelativelylittleabouttheirconcreteimplementations(thesystemsthatrealizetheproposedstructuresandprocesses).Thisisbecauseitisthetheoryitselfandthepropertiesofcomputationalmodelsdesignedtoexpressthattheorythatareofbroadestrelevancetocognitivescientistsoflanguage.

10.1.3EmbodiedSimulation

SinceECGaimstoaccountforthemechanismsoflanguageuse,thestructuresandprocessesweproposeshouldnaturallybeconstrainedbyevidenceabouthowpeopleactuallyproduceandcomprehendlanguage.AndbecauseECG,likeotherconstructiongrammars,hypothesizesthatmeaningplaysanimportantroleingrammar,weneedtotakeseriouslyresearchonhowmeaningworks.Thepastdecadehasseenanexplosionofbehavioralandbrainimagingresearchonlanguagecomprehension,focusingonmeaning.Onefoundationalfindingisthatpeopleunderstandutterancesbyperformingembodiedsimulations:theyengageperceptual,motor,andotherbrainsystemstocreateinternalexperiencessimilartothosetheywouldhavewhenexperiencingthedescribedscenes(Bergen2012).Toaccommodatethisfinding,ECGincludesembodiedsimulationasacomponentoflanguageuse,wherewordsandotherconstructionsarerepresentedasknowledgestructuresdrivingthespecificperceptualandmotorexperiencesthatthelanguageusersimulates.Section10.2reviewsembodiedsimulationresults,introducesthenotionoflanguageasaninterfacetosimulation,andbrieflyillustrateshowform,meaning,andconstructionalknowledgearerepresentedintheECGformalism.Section10.3describesgrammaticalconstructionsandthekindsofcontributionstheycanmaketoembodiedsimulation.Insection10.4wefocusmoredirectlyonlanguageprocessing,inparticularonamodeloflanguagecomprehensiondevelopedwithintheECGframeworkthatiscompatiblewithboththeideaofembodiedsimulationandrelevantevidencefromthesentenceprocessingliterature.

10.2.EmbodiedSimulationandLanguage

10.2.1MeaningisGroundedinEmbodiedSimulation

Overthepastcentury,thegradualspreadofthestudyoflanguageintothebehavioralandphysicalscienceshasbeendrivenpredominantlybyinvestigationsoflinguisticform(i.e.,observablecharacteristicsoflinguisticbehavior).Thereisapracticalreasonforthis.Itisrelativelystraightforwardtoobservewordorderormeasuretheamplitudeofastopconsonant'sburst.Bycontrast,meaningisinvisible,internal,andsubjective.Thus,whilespeakers’intuitionscanprovidesomecluestothenatureofsemanticandpragmaticstructure,thestudyofhowpeopleaccessandconstructmeaningduringonlinelanguageusehaslaggedfarbehindthestudyoflinguisticform.Thepastdecade,however,hasseensubstantialprogresstowardunderstandinghowmeaningworks,especiallyinlanguagecomprehension.Thisworkhasattackedtheproblemfromseveralangles,basedonindirectmeasuresofcognition(suchasreactiontimes,eyemovements,changesinfMRIsignal,etc.)takenwhilepeopleareprocessinglanguageundertightlyconstrainedexperimentalconditions.Crucially,avarietyofdesignsandmeasureshavebeenconvergingontheincreasinglystableconclusionthatwhenpeopleprocesswordsorsentences,theyactivateembodiedsimulationsofthedescribedscenes.

Embodiedsimulationistheinternalengagementofmodality-specificbrainsystemstocreateorre-createnonpresentexperiences(Barsalou1999).Embodiedsimulationcanbevisual,inwhichcasepartsofthebrain'svisualsystembecomeactiveinwayssimilartohowtheywouldreactifexternalstimuliwereactuallypresent(Kosslynetal.2001).Embodiedsimulationcanalsobemotor;peopleactivatepartsoftheirbrainsdedicatedtomotorcontrolasaninternalmodelofphysicalaction,usuallywithoutactuallyengagingtheirskeletalmuscles.Peopleautomatically(andmostlyunconsciously)engagemotorsimulationduringavarietyofcognitivebehaviors,suchasintentionalmentalimagery(Ehrsson,Geyer,andNaito2003),recall(Nybergetal.2001),andothertasks.

Recentworkonlanguagecomprehensionpointstoaroleforembodiedsimulation;whenpeopleprocesslanguageaboutperceptiblethingsandevents,theyactivateperceptualrepresentationsofwhatthementioneditemswould

Page 136: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 4 of 16

looklike(StanfieldandZwaan2001),soundlike(WinterandBergen2012),andsoon.Theseperceptualsimulationsencodeimplieddetailsnotexplicitlymentionedinthelanguage.Forinstance,peoplerepresentmentionedobjectswiththemostlikelyshapes(Zwaan,Stanfield,andYaxley2002),orientations(StanfieldandZwaan2001),orcolors(Connell2007)theywouldhave,dependingonthecontext.Thesesimulationsmayadoptaparticularperspective,orviewpoint,fromwhichthecomprehenderssimulateperceivingtheevent,andgrammaticalperson(e.g.youvs.he)affectswhichperspectiveisadopted(Brunyéetal.2009).Similarlytolanguageaboutperceivablethings,whenpeopleprocesslanguageaboutactions,theyengagemotorrepresentationsofthoseactions(GlenbergandKashak2002).Thesemotorrepresentationsarequitespecific,downtothebodypartusedtoperformtheaction(Bergenetal.2010;Pulvermülleretal.2005),orthemostlikelyhandshape(BergenandWheeler2005;Masson,Bub,andWarren2008).Roughly,then,understandingseemstoinvolveactivatinginternalrepresentationsofperceivablethingsorperformableactions,evenwithoutthethingsbeingpresentortheactionsbeingperformed.(Bycomparisonwiththegrowingbodyofworkonembodiedsimulationincomprehension,therehasbeenrelativelylittleworkonembodiedsimulationinlanguageproduction,thoughHostetterandAlibali(2008)andSato(2010)provideevidencefromgestureandreactiontimestudies,respectively,thatitplaysarolethere,too.)

Experimentalevidenceleaveslittledoubtthatpeopleperformembodiedsimulationsduringcomprehension.Itremainslessclearpreciselywhatrolesuchsimulationsplayinlanguageuse,andfurtherinvestigationisneededtoascertaintowhatdegreeandunderwhatconditionssimulationisnecessaryforparticularaspectsoflanguageunderstanding.Inprinciple,however,embodiedsimulationcouldofferelegantexplanationsforavarietyofphenomenaassociatedwithmeaning.Inparticular,wehypothesizethatcomprehendersmaybeabletounderstandlanguageaboutnonpresententitiesandeventsbecausetheprocessingofsuchlanguagetriggersbrainstatessimilartothosethatresultfromexperiencingthoseentitiesandevents(aclaimcompatiblewithargumentsbyBarsalou1999;GlenbergandRobertson2000,Pulvermülleretal.2005,andothers).Further,duringsimulation,thesebrainstatesmayactivatevariousrelatedprocesses,inmuchthesamewayasdotheactualperceptionortheperformanceofrealactions,thusallowingcomprehenderstoenrichtheirunderstandingbydrawingdetailed,relevantinferencesthataregroundedinsensorimotorexperienceandsensitivetocontextualconditions.Andfinally,attentiontoperceptualandmotorsimulationcaneffectthecomprehender'ssubjectivefeelingofunderstandingthetopicbeingdiscussed.Insum,simulationhasthepotentialtoexplainhowwegroundmeaninginthebrain'sperceptualandmotorsystems.

ECGisintendedtobeconsistentwiththeevidenceabove.Itisatheoryinwhichthepotentialfunctionsofsimulationcanbeproposedandmodeled,suchthatdetailedpredictionscanbegeneratedfortesting.Morespecifically,itprovidesthemeansforexploringthehypothesisthatlinguisticconstructionsserveasaninterfacetoembodiedsimulation.TheremainderofthissectionoutlinesminimalrequirementsforsuchaninterfaceandbrieflydescribeshowECGsatisfiesthem,beginningwithwordsandtheirmeaningrepresentations.

10.2.2AnInterfacetoEmbodiedSimulation

Theideathatlanguageunderstandinginvolvesembodiedsimulationcallsforaradicalrethinkingofwhatwordsandotherlinguisticconstructionsmean,aswellashowthismeaningshouldberepresented.Traditionally,linguistsusequick-and-dirtyapproximationsofmeaning,oftendodgingtheissueofwhatmeaningisbymerelylabelingwordmeanings(forinstance,themeaningofthewordcatmightberepresentedasCATorasacollectionofungroundedfeatures,like[+FELINE,+DOMESTIC]).Butworkonembodiedsimulationsuggeststhatwhatlanguageusersknowishowtotakealinguisticform(suchasawordorgrammaticalstructure)andactivateperceptualormotorrepresentationscorrespondingtothedenotedentitiesandevents,andviceversa.Thatis,linguisticunitsdriveembodiedsimulations:insteadoffizzlingoutinstatic,formal,arbitraryrepresentations(likeCAT),meaningfulwordsandconstructionsserveasinterfacesthatactivateandconstrainmodality-richembodiedsimulations.

Statedmoreconcretely,inECG,wordsandotherconstructionsserveaspathwaysconnectingdetailed,modality-specificknowledgeabouttheirformswithdetailed,modality-specificknowledgeabouttheirmeanings.Critically,however,thesepathwaysaremediatedinbothdomainsbycategorization.Therangeofpossiblephoneticandgraphicalrealizationsofawordvariesdramatically,notjustacrossbutevenwithinmodalities(considerthedifferencebetweencursivecatandblockprintstencilcat).Yetwecategorizethesedifferenttokens,whichvarycontinuouslywithinvisualandauditoryspace,asinstancesofthesamecategory—instancesofasinglewordform

Page 137: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 5 of 16

(Dahaene2009).Thepossibledenotationsofcatvarysimilarly;evenrestrictingourselvestothedomesticcat(feliscatus),theperceptualandmotorcomponentsofrelevantsimulationsvarywidelynotonlyinvisualfeatures(color,size,speed,andadorableness)butalsoinhaptic,auditory,andperhapsevenolfactoryandgustatoryfeatures.Justastheformalrealizationofthewordformcatvariesacrossuses,sodoestherealizationinsimulationoftheword'smeaning.Andcritically,forthemostpart,thetwodonotco-vary:itisnotthecasethatcatwrittenincursivedenotesaparticulartypeacat(say,acalico),whilestencilblocklettersdenoteanother(say,atabby).Rather,themultifariousformalrealizationsofawordactivateamentallyrepresentedcategoryofexperienceswiththingsintheworld.

Followingthecognitivelinguisticliterature,wecallcategoricalmentalrepresentationsthatgeneralizeoverinstances(liketheformormeaninggeneralizationsrelevanttocat)‘schemas’(Johnson1987;LakoffandJohnson1980).Andfollowingconstruction-basedapproachestogrammar,wecalltheentirestructurethatlinkssuchschemasa‘construction.’AssuggestedbytheimageinFigure10.1,constructionsare‘bipolar’:theyservetoconnectschemasinthetwodomainsof‘form’(phonological,orthographic,gestural,etc.)and‘meaning’(conceptual,ontological,situational,etc.).Aswewillseebelow,theECGformalismprovidesnotationsfordefiningthesetermsthataredesignedtobeconsistentwiththeliterature.

Clicktoviewlarger

Figure10.1. Wordsasbipolarconstructionslinkingschematicrepresentationsinthedomainsofformandmeaning

Clicktoviewlarger

Figure10.2. Anx-schemaforjumping,showingdifferentstagesofthemotoraction

Crucially,however,theapproachpursuedherefurtherrequiresanactive,groundednotionofmeaningcapableofsupportingembodiedsimulation.Forthisweturnto‘simulationsemantics’(Bailey1997;Naranayan1998;Feldman2006),anapproachtosemanticrepresentationdevelopedwithpreciselythisaiminmind.Broadly,thesamedynamiccontrolsystemsthatpeopleusetoperformactionsorperceiveeventsarehypothesizedtoalsobeusedtosimulateactionsandevents.Thoughafullexpositiongoesbeyondthescopeofthischapter,Figure10.2givesatasteofhowactions(inthiscase,theactionofjumping)mightberepresentedusingthe‘executingschema’(or‘x-schema’)formalism(Narayanan1998).Thex-schemashowncapturestheflowofactivationthroughamotorcontrolsystemresponsiblefortheperformanceorsimulationofjumping.First,assumingallenablingconditionsaremet(e.g.,thejumperispositionedonastablesurfaceandhassufficientenergytoexpend),thejumperpreparesbycoilingitslegs.Thencomestheexplosivephaseinwhichthelegs(andbody)arerapidlyextendedandthejumperispropelledupwardsfromtheground,resultinginaballisticperiod(correspondingtothenodemarked“done”)inwhichtheairbornejumperissubjecttotheforceofgravity.Likethecatcategoryabove,thisjumpingactionschemaisintendedtosupportconsiderablevariation:thejumpermightbeahuman,acat,orsomeotheranimateagent;theactionmayvaryinforceexpended,launchheightattained,directionofmotion,orbodilyorientation;anditmightprecedeahostofotheractions(landinginthesameplace,landingonahigherorlowersurface,grabbingatrapeze,flyingthroughtheair,falling,etc.).

Page 138: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 6 of 16

Languagedoesnotusuallyneedtoreachintosensorimotorknowledgeatthelowestlevelofdetail,ofcourse.Infact,thesimulation-basedframeworkhypothesizesthatspecificallylinguisticknowledgeconcernsonlylimitedsubpartsoftherichunderlyingactionschema,enoughtodescribe(or“parameterize”)therelevantsimulation.Forexample,alinguisticunitliketheverbjumpmightevokethismotoractionasagestalt,perhapsallowingafewfeatures(e.g.,identityofjumper,rate,direction,goal)tobecontributedbyotherlinguisticunits.Inotherwords,linguisticmeaningprovidesalimitedinterfacetoembodiedsimulation,whichinturnactivatesdynamic,embodiedschemasthatcombinewitheachotherandthecurrentcontext.Therunningofthesimulationandtheresultingpropagationofinferencestogetherallowthecomprehendertounderstandanutteranceincontext.

10.2.3Form,Meaning,andConstructions

NowthatwehavedrawnthebroadoutlinesofhowECGusesconstructionsandschemasastheinterfacetoembodiedsimulation,weturnbrieflytohowECGrepresentsthislinguisticknowledge.Incomprehension,schemasserveasthemeansbywhicharecognizedwordactivatesperceptualormotorcomponentsofasimulation.Thesimplestschemasdolittlemorethanthis,servingessentiallyaspointerstoexperientiallygroundedgestalts,ineithertheformormeaningdomain.AsimpleconceptualschemaillustratinghowthemeaningcategoryforcatisrepresentedintheECGformalismappearsbelow(1).Itincludesanameoftheschema(Cat),alongwithanotationindicatinganinheritancerelationship:partofwhatyouknowaboutthecategoryCatisthatitisasubtypeofasuperordinatecategoryAnimal.TheCatschemadoesnotexplicitlylistvariousothersalientfeaturesthatarebydefaultassociatedwiththecategory(e.g.,thatitisananimatephysicalentity),sincetheseareinheritedfromtheAnimalschema.Whilerepresentationslikethisonemayappearsimilartoclassicalstand-insformeaning(likeCAT),itiscrucialtorecallthattheyaremeanttoserveaspointerstodetailed,modality-specificstructuresthatareactivatedasaseparatepartoflanguagecomprehensionorproduction.

(1)schemaCat

subcaseofAnimal

Formschemasarerepresentedinasimilarlysimpleway,whereagain,schemasserveaspointerstomorecomplexandvariablephonological,articulatory,andauditoryrepresentations.TheCat-Formschemabelow(2)statesthatitisasubcaseofamoregeneralWord-Formschema(thecategoryofformsofwordsinthelanguage),andthatamongits‘roles’(orfeatures)isitsorthography,whichisthestring“cat”.

(2)schemaCat-Form

subcaseofWord-Form

roles

orth←“cat”

Theseformandmeaningschemasarelinkedbyaconstruction,Cat-Cxn(3),whoseformisdefinedasaninstanceoftheCat-Formschemaabove,andwhosemeaningisdefinedasaninstanceoftheCatschema,alsoabove.(ThisconstructioncorrespondstothesolidlineconnectingtheformandmeaningschemasinFigure10.1.)TheCat-Cxnisalsoasubcaseofaclassofsimilarconstructions,whichforcurrentpurposeswewillcallNoun.TheNounconstruction,notshown,pairsanunspecifiedWord-Formontheformsidewithsomeentityonthemeaningside.

(3)constructionCat-Cxn

subcaseofNoun

form:Cat-Form

meaning:Cat

Otherschemasandconstructionsaremorecomplex.Wordslikeprepositions,verbs,oradjectives,whosemeaningsarenecessarilycomputedincombinationwiththemeaningsofotherwords,involverelationsamongmultiplestructures,someofwhoseinternalcomponentsmustbesuppliedduringlanguageusebythelinguisticorphysicalcontext.Thewordjumped,forexample,invokestheactionofjumping,whichinherentlyincludessome

2

Page 139: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 7 of 16

animateentitydoingthejumping.Asnotedearlier,thex-schemaassociatedwiththisaction(Figure10.2)involvesrichdetailsaboutvariousstagesandsubactions,buttheconceptualJumpschemamightbemoresimplydefinedintermsofitskeyparameters(orroles).BelowwerepresenttheJumpschema(4)asasubcaseofAction(notshown),whichisaconceptualcategoryofeventsinwhichsomeanimateagentactsintentionally.Itisdefinedhereashavingoneparticipant,specifiedasthejumperundertherolesblock,whichisboundto(or‘identifiedwith’)theagentiveactor(inheritedfromtheActionschema),asindicatedbythedouble-headedarrow.

(4)schemaJump

subcaseofAction

roles

jumper↔actor

TheJumpschemaappearsinthemeaningpoleoftheJumpedconstruction(5),pairedwithaformspecifiedbytheJumped-Formschema(notshown,butsimilartotheCat-Formschema,mutatismutandis).TheJumpedconstructionisalsodefinedasakindofverb;theVerbconstruction(notshown)pairsanunspecifiedWord-Formwithaperceptualormotorschema.Verbsmayalsospecifyconstraintsonthetimeofthereportedevent(relativetospeechtime),asshowninthemeaningpoleoftheconstructionbelow,wherethetimeroleisinheritedfromtheVerbconstruction.IntheJumpedconstruction,forexample,thetimeofthejumpingeventisconstrainedtobeinthepast(i.e.,ittookplacepriortospeechtime).

(5)constructionJumped

subcaseofVerb

form:Jumped-Form

meaning:Jump

time←past

TheseexamplesillustratethegeneralapproachtodefiningECGschemasandlexicalconstructions:together,theformalnotationsaboveprovidewaysofspecifyingwhichformandmeaningschemasarelinked,aswellaswhichembodiedschemasshouldbesimulatedwithwhatparameters.

Itisfairtoask,however,howexactlyonedecideswhatschemasandconstructionstohypothesize—whatparticipantrolesmightappear,howmanydifferentschemasthereare,howrelatedconceptsmightbeorganized,etc.Afterall,thespaceofpossibleconceptsrangesoveravarietyofmodalitiesanddomains(motor,perceptual,affective,social),whichinvolvedifferentspecializedneuralandcognitivestructures.Justasthedynamicsofthecomplexmotoractionofjumpingcanbesummarized,orschematized,intermsofseveralparametersrelevanttotheverbjumpandotherrelatedconstructions,soweassumethatthespaceofpossibleconceptscanbeschematized.Indeed,someschemasmayberelevantformultiplewords(variantsoftheJumpschemamightbeused,forexample,forwordslikejump,leap,andhop).Atthesametime,somewordsmayevokemultipleschemas.Forinstance,inthecontextofplayingcheckers,jumpingyouropponent'spiecesdoesnotinvolveanythinglikethesamemotorcontrolasthetypeofjumpingacatdoes;moreover,therelevantschemainvariablyincludesaparticipantthatundergoesthejumping,withimplicationsfordifferencesinhowthissenseoftheverbmayberealized(e.g.,itistransitive).Tomakemattersworse,someschemasmaynotcorresponddirectlytoanyparticularconstruction;wemayhaveschematicknowledgeaboutproprioceptive,affective,gustatory,social,andothercategoriesofexperiencethatweneverlearntoputintowords.

Infact,speakersofdifferentlanguagesconvergeondifferentorganizationsofconceptualspaceintoschemasandconstructions,andevendifferentspeakersofthesamelanguagemayvaryintheparticularrepresentationstheyacquire.Previousworkinsimulationsemanticshasshownhowmachinelearningtechniquescanbeusedtolearnmappingsbetweenactionschemasandwordsenses(see,e.g.,Bailey1997),modelinghowthesameverbmayhavemultiplerelated(embodied)senses(e.g.,pushingablockonatableorabuttononakeyboarddifferdistinctlyinthedirectionofmovementanddefaulthandposture).Ideally,suchlearningmodelswouldprovideadynamic,

Page 140: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 8 of 16

experientiallydrivenbasisfordefiningschemasandconstructions.

Moregenerally,however,schemasandconstructionsmayalsobedefinedbytheanalysttoaccountforaparticularlinguisticphenomenonorsetofdata.Insuchcases,ourprocedureisconservative:wehypothesizeaschemaforeachsenseofeachmorphemeinalanguage(whereenumeratingsensesisadmittedlyanenterprisefraughtwithuncertainty),exceptwhereanotherhypothesizedschemacanalreadyaccountfortheprocessingofthatmorpheme.Inaddition,wehypothesizehigher-levelschemaswherelinguisticpatterningsuggeststhem;suchschemasoverlapwithwhathavebeencalled‘imageschemas’intheliterature,suchasContainment,Source-Path-Goal,andTrajector-Landmark(Johnson1987).Theseschemasareoftenrelativelysimple,andwhennotatedintheECGformalismalongwiththeirassociatedrolestheybearastrong(anddeliberate)resemblancetoinformaldescriptionsappearingintheliterature.LikeotherECGschemas,theyarepositedtoserveasinterfacestothericherstructuressupportingembodiedsimulation.Theyarealso,however,moregeneralintheirapplicability;theyrecuracrossawiderangeofeventtypesandarelexicalizedandgrammaticizedcrosslinguistically.

10.3.GrammaticalConstructionsandtheSpecificationofSimulation

Consistentwithaconstructionalview,ECGtreatsalllinguisticunits,includinglexicalitemslikethoseillustratedabove,aswellaslargerphrasalconstructionsandothertraditionalgrammaticalnotions,asvariantsofthesamekindofthing—mappingsbetweenformandmeaning.Unlikesimplewords,however,manygrammaticalconstructionsexhibitcomplexinternalstructurewithmultipleconstituents,eachofwhichinstantiatesaconstructionalform-meaningmapping.Fromasimulation-basedperspective,thismeansthattheyareparticularlywell-suitedforparameterizingcomplexembodiedsimulationsinvolvingmultiple,variableconceptualschemas.

ToillustratesomefeaturesofgrammaticalconstructionsinECG,wegiveasimplifieddefinitionforonecommonconstruction,theDirectedMotionconstruction(6).Thisisanargumentstructureconstruction(asnotedinthesubcaseline)licensingsentenceslikeThehorseracedpastthebarnorThecatjumpeddown,inwhichamover(designatedbysomeReferring-Expression)movesalongsomepath(designatedbyaPath-Expression)bymeansofsomeaction(designatedbytheverb).TheDirectedMotionconstructionspecifiesthesethreeconstituents,andinthenotationshownbelow,eachisgivenalocalalias(r,v,orp)bywhichtheconstructionmayrefertothem.Eachoftheseconstituentsisalsorestrictedtobeofaparticularconstructiontype.(Thoughnotshownhere,Referring-Expressionsareroughlyanalogoustonounphrases,whilePath-Expressionsexpressapath,andincludeparticlesandprepositionalphrases.)Similarconstructionsaredefinedin,forexample,BergenandChang(2005)andN.Chang(2008).

(6)constructionDirectedMotion

subcaseofArgumentStructure

constituents

r:Referring-Expression

v:Verb

p:Path-Expression

form

rbeforev

v beforep

meaning:Motion

mover↔r

action↔v

path↔p

f f

f f

m

m

m

Page 141: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 9 of 16

Theformofthisconstruction,specifiedintheformblockoftheconstruction,ismorecomplexthantheformsofmostlexicalitems:itstatesthattheformofthereferringexpressionprecedestheformoftheverb,andthattheformoftheverbprecedesthatofthepathexpression;thisismeanttoholdwhethertheconstructionisusedinwritingorspeech,orwhetherinproducingorcomprehendinganutterance.Initsmeaning,theDirectedMotionconstructionabovedenotesamotion,specifiedbytheMotionschema(7).IntheDirectedMotionconstruction,therolesoftheMotionschemaareexplicitlyboundtogetherinthemeaningblocktothemeaningpolesoftheReferring-Expression,Verb,andPath-Expressionconstituents,respectively.Thatis,themeaningoftheReferring-Expression(e.g.,astructureinvolvingtheCatschemainthecaseofthecat)willbeboundtothemoveroftheMotionschema.(TheconstraintintheMotionschemafurtherbindsthistotheactorofitsassociatedactionschema.)

(7)schemaMotion

roles

mover:Animate

action:Action

path:SPG

constraints

mover↔action.actor

Ofcourse,theDirectedMotionconstructiondefinedherehasbeensimplifiedforexpositionalpurposes,anditelidesmanyissuesthatwouldariseindefiningalarger-scalegrammar.Forsuchgrammars,onemaywishtohandlenotonlyThecatjumpedbutalsosentenceslikeIntowhichbucketdidthecatjump?andWhichbucketdidthecatjumpinto?,whichviolatetheorderingconstraintsaboveandinvolveadifferentpredicationtype,aninterrogative.TwopotentialsolutionstoissuesofconstructionalvariabilityhavebeenadoptedintheECGliterature.Thefirstistoproliferateargumentstructure(andother)constructionsintofamilies,suchthattheconstructionshownabove(whichmightberenamedDeclarativeDirectedMotion)isdefinedasasubcaseofamoreschematicDirectedMotionconstruction,whichitselfdoesnotspecifyconstituentorder.OthersubcasesofthisgeneralconstructionidentifythespecificconstituentorderingsandpredicationtypesrequiredforinterrogativeandotherusesofDirectedMotion.Asecondsolutionistofactoroutwordorderandpredicationtypeentirelyfromargumentstructureconstructionsandspecifythisinformationusingothersentence-levelconstructions,suchasaSubject-Auxiliary-InversionConstruction.Thesetwosolutions(familiesofargumentstructureconstructionsversusfactoringoutofwordorder)areexploredinmoredetailinDodge(2010).

MuchoftheworkinECGhasfocusedongrammaticalconstructionslikethese,specifyingtheirformsandmeanings,inheritancerelations,andinteractionswithoneanother(see,forinstance,Bryant2008).Whilesuchanalysesvaryinpreciselyhowtheypartitionthespaceofconstructionsandschemasandwhatlevelofspecificitytheyaimfor,theymakeacommonsetofpredictionsabouthowgrammaticalconstructionscanhelpspecifyembodiedsimulationsandtherebyaffectmeaningprocessing.Thesepredictionsfallintothreebroadclasses.

First,agrammaticalconstructionmayalign(or‘identify’or‘bind’)differentaspectsofthemeaningsofitsconstituents,bringingtogethertheirvariouscontributionstoembodiedsimulation.Forinstance,anargumentstructureconstructionliketheDirectedMotionconstructiondesignatesoneofitsconstituents(thereferringexpression)asthemoverengagedinthemotiondenotedbytheverbalongapathdenotedbythepathexpression.ThereisnothingneworspecifictoECGortoconstructionalapproachestogrammarinthisclaim;theonlynewcontributionECGmakesistopredictthatmanipulationsofsuchconstituent-aligningconstructionswillaffecttheembodiedsimulationscomprehendersgenerateinwayspredictablefromconstructionalmeaning.Forinstance,switchingthesubjectandprepositionalobject(Thefencejumpedoverthecat)shouldresultina(potentiallyimplausible)embodiedsimulationinwhichthemoverandlandmarkareswitched.Toourknowledge,thisclaimhasnotbeentestedexperimentally,perhapsbecauseitislargelysharedwithotheraccounts,thoughthereissomeevidence(GlenbergandKaschak2002)thatswitchingsubjectandobjectaffectssimulateddirectionofmotion.

Page 142: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 10 of 16

Second,agrammaticalconstructionmaycontributecontentdirectlytoembodiedsimulation,byevokingcategoriesofexperience(schemas)thatsupplypartofthecomprehender'smodalrepresentationofthedescribedentitiesandevents.Thiscontributioncorrespondsmorecloselytotheclaimthatmeaningisnotstrictlycompositional,adefiningassumptionofconstructionalapproachestogrammar.Forinstance,theDirectedMotionconstruction,describedabove,mightleadcomprehenderstosimulateeventsasthoughtheyinvolveananimatemovermovingalongapath,evenwhentheverbdoesnotexplicitlyspecifymotion(asinThecatmeoweddownthestreet).OtherexamplesincludetheDouble-ObjectorDitransitiveconstruction,whichhasbeenarguedtoactivateaTransfer-of-Possessionschema,ascontrastedwiththePrepositionalDative,whichmayactivateaCaused-Motionschema(Goldberg1995).TheECGviewproposesthattheseargumentstructureconstructionsactivatetherelevantschemas,bindthemwiththemeaningcontributionsoftheconstructionalconstituents,andthentogetherdrivetheembodiedsimulationoftheunifiedscene.Therehasbeenabitofexperimentalworkaddressingconstruction-specifichypothesesofthissort,comparingthetransitiveandditransitive(KaschakandGlenberg2000),whichshowscomprehendersaccessingdifferentmeaningsforsentencesusingthedifferentconstructions.Butthisexperimentalworkisinitsinfancy.

Third,grammaticalconstructionsmaymodulatesecond-orderpropertiesofsimulation,likeperspectiveorlocusofattentionalfocus.Forinstance,activesentencesmightleadthecomprehendertosimulateaneventfromtheperceptualormotorperspectiveoftheagent,whilepassivesentencesmightleadtosimulationfromtheperspectiveofthepatient.Likewise,aspectualconstructions(liketheEnglishprogressiveversusperfect)mightmodulatewhatpartofascenecomprehenderssimulateingreaterdetail.Suchpredictionsappeartobespecifictosimulation-basedapproachestolanguageuse,likeECG,andhavebeenborneoutexperimentally(e.g.,MaddenandZwaan2003;BergenandWheeler2010).

Insum,grammaticalconstructionsinECGarehypothesizedtocontributetoembodiedsimulationbybindingtogethertheconceptualschemasevokedbytheirconstituents,bydirectlycontributingconceptualschemastosimulate,andbyimposingsecond-orderconstraintsonsimulation.AtheoryofhowpeopleusegrammarintheECGframeworkshouldincludebothanaccountofhowgrammarconstrainssimulationinthesebroadestterms,aswellascomputationallyimplementedmodelsfleshingoutempiricallytestableclaimsabouthowspecificconstructionsaffectembodiedsimulationinoneof(orcombinationsof)thesethreeways.Thenextsectionprovidesfurtherdetailonthesetwoaims,withspecificreferencetolanguageprocessing.

10.4.LanguageProcessing

HowdoesECGsupportprocessesoflanguageuse?InthissectionwepresenttheprocessingsideofECG:amodeloflanguagecomprehensionthatusesECGconstructionsasaninterfacetoembodiedsimulation.(SeealsoBryant2004,2008;BergenandChang2005;N.Chang2008;andMok2008.)Ourmodelsoflanguagecomprehensionarecontinuallyevolvingtoaccommodatenewfindingsabouthumanlanguagecomprehensionandaddressawiderarrayoflanguagephenomena.Thecoreideas,however,arerelativelystable.Inthesectionsbelow,wefirstdiscusshowaspecificsentencemightbeunderstoodandwhatgeneralrequirementssuchaprocessmightimpose;wethendescribeacomputationalmodeloflanguagecomprehensionthatsatisfiesthoserequirementsinacognitivelyplausibleway.

10.4.1ComprehensionasConstraintSatisfaction

Thefoundational‘simulationhypothesis’ofourlanguagecomprehensionmodelisthatlinguisticconstructionsofallkindsparameterizeembodiedsimulations:theyevoketheexperientialschemas(correspondingtoevents,actions,objects,etc.)involvedinaparticularutteranceandspecifyhowthesearecombined.Theresultinggestaltissimulatedtoyieldtheinferencesthateffectusefulcomprehensionoftheutterance.Thenotationalmechanismsdescribedinsection10.3provideameansofspecifyingwhichformandconceptualcategoriesareinvolved,andwhatrelationshipsintheformandmeaningdomainsmustbesatisfied.Considerthesentencebelow:

(8)Thecatjumpeddown.

Restrictingourselvesforsimplicitytothesentenceinwrittenform(wewillassumeforpresentpurposesthattheinputisjustanorthographicstring,pre-segmentedintowords),anumberofsurfacecuesarealreadyavailable:

3

Page 143: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 11 of 16

theindividualwordforms,theorderinwhichtheyappear,andapunctuationmark.Understandingthissentenceinvolvesusingthesesurfacecuestoactivatevariouspiecesoflinguisticknowledge—constructionsforthewordsthe,cat,jumped,anddown,aswellaslargergrammaticalconstructions,likeDirectedMotion.Theseinturnevokevariousconceptualschemas,includingsomewehavealreadyseen(CatandJump)andotherslikeaDownschemacapturingthespatialdirectionevokedbytheworddown.Asbefore,theseschemasarenotmeresymbols,butratheraregeneralizationsassociatedwithexperientiallygroundedattributesandfunctions:thenotionofDownevokesvisual,motor,andproprioceptiveknowledgeaboutorientationandmotionofsometrajectortowardtheearth.

Duringsentenceprocessing,lexicalconstructionsactivatetheseconcepts,whicharecombinedbymorecomplexgrammaticalconstructions,asdescribedabove.Forinstance,thecomprehendermayrecognizethestringThecatjumpeddownasaninstanceoftheclause-levelDirectedMotionconstruction,definedinsection10.3.TheDirectedMotionconstructionincludesaPathExpressionconstituentinthepostverbalposition,whichisfilledinthiscasebytheDownlexicalconstruction(sinceitsassociatedschemasatisfiesthetypeconstraintsimposedbyDirectedMotion).ThecomprehenderalsobindstheActioneventinthemeaningofDirectedMotiontotheJumpschema,andtheMoverroletotheJumperoftheJumpschema,aswellaswiththeCatschema.Moreover,thejumpingeventisconstrainedtotakeplacepriortospeechtime(asindicatedbythepasttenseverbjumped).Notefinallythattheuseofthepicksoutaspecificcatofwhichthisjumpingeventisasserted.Thehearermayemployamoregeneraldeterminednounphraseconstruction(notshownhere,butseeBergenandChang2005)underlyingphraseslikethecatandacat,whichspecifiesthewordorderofthedeterminerandnoun,andcombinestheconstraintseachimposesontheidentityofthereferent.Intheexample,weassumethedeterminertherestrictsthereferenttoaparticularmemberoftherelevantcategory(i.e.,thatspecifiedbythenouncat)thatisidentifiable(orgiven)tothehearerinthecurrentcontext.

Clicktoviewlarger

Figure10.3. A(simplified)constructionalanalysisoftheexamplesentence,withconstructs(instancesofconstructions),showninthecenter.Theylinktheirassociatedformintheinputsentenceonthelefttotheirmeaninginthesemanticspecificationontheright.

Theresultingconfigurationofschemasispreciselytheparameterizationwehaveclaimedisusedtodriveanembodiedsimulation;wecallthisconfigurationa‘semanticspecification,’or‘semspec.’Thesemspecdictateswhatscenariotosimulateandhow,andthecomprehenderusesperceptualandmotorsystemstoperformthissimulation.AsemspeccorrespondingtoapossibleinterpretationoftheexamplesentenceusingsomeoftheconstructionsdefinedearlierisshowninFigure10.3.Theformsintheinpututteranceareshownintheboxontheleft;someconstructionsinstantiatedintheanalysisareshowninthecentralarea;andtheirassociatedsemspec(i.e.,collectionofboundmeaningschemas)isshownintheboxontheright.Anotherwaytothinkaboutthefigureisasdepictingthelinearstructureoftheinputsentencestringontheleft;theconstituentstructureoftheconstructionsinvolvedinthemiddle;andthesituationalstructureoftheidentityandrole-fillerrelationsamongthemeaningschemasevokedbythesentenceontheright.(Notethatwehaveleftouttheconstructionalandmeaningdetailsassociatedwiththeforthepurposesofthisexposition.)

Thesemspeccapturesmuchofwhatmightbeconsideredthetraditionalmeaningofthesentence,suchasbasicthematicrelations(whoisdoingwhattowhom),spatialandtemporalrelations,anddiscourseconstraintsabouttheinformationstatusofthevariousreferents(e.g.,thecat'sstatusasgiven).Butthesemspecalonedoesnotprovideeverythingacomprehenderneedstoknowtoperformtherelevantsimulationandmakeappropriateinferences.Infact,manyspecificinferencesmaydependontheparticularjumperandpath,whichmaybebased,amongotherthings,onthecurrentcontext,aswellasongeneralknowledgeabouttherelevantcategories.Specificparticipantsidentified(orinstantiated)incontextleadtosimulationsthatcanproducemuchfiner-grainedinferencesaboutthejumper'spositionbefore,during,andaftertheevent;theamountofenergyexpendedandtheinvolvementoffeline

Page 144: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 12 of 16

femursandpaws;thedefaultheightofjumping;andsoon.Thatis,knowinghowtosimulateinvolvesmorethanknowingmerelywhatcategoriesthelanguageevokes;itinvolvesresolvingreferencesmadeusingthosecategoriestospecificinstancesofthem.

Wethereforedistinguishthesemspecabove—whichcapturesthegeneralbindingsandconstraintsimposedbytherelevantconstructions—fromafully‘resolvedsemspec’thatfurtherconstrainsthesebasedonthehearer'sknowledgeandthecurrentsituationalanddiscoursecontext.Onlyaftersuchreferenceshavebeenresolveddoesthecomprehenderhaveenoughinformationtoengageinacontextuallyappropriatesimulation;thehearercanthenusetheresultsofsimulationtoupdateherbeliefsaboutthesituation,orpreparetorespondappropriatelytotheutterance.Forourexample,theresolvedsemspeccorrespondingtothesemspecabovemightinstantiatethegeneralCatcategorywithaspecificcatsalientinthephysicalorlinguisticcontext,or(intheabsenceofanysuchinstance)withaninstanceselectedarbitrarilyorbasedonprototypicalfeatures.

Itisworthnotingthatdifferentindividualsarelikelytobringdiverseconstructionalknowledgetobearonlanguageuse.Thedifferentconstructionsthateachlanguageuserhaslearnedoverthecourseofhisorheruniqueexposuretolanguagemayleadtotheconstructionofsubtlyorsubstantiallydifferent(unresolved)semspecswhenprocessingthesameutterance.Totheextentthatthesedifferentsemspecsleadtodifferentresolvedsemspecsanddifferentsimulation-basedinferences,communicationwillfail.Butbecauseanunresolvedsemspecservesasonlyanintermediatestepincomprehension,itispossible,onanECGview,fordyadstomeetwithsomecommunicativesuccessevenwithnonidenticallinguisticknowledge—again,totheextentthatthevariousprocessesengagedbythecomprehenderproduceinferencescompatiblewiththespeaker'sintentions.

10.4.2AComputationalModelofLanguageComprehension

TheaccountofcomprehensiondescribedaboveissummarizedinFigure10.4,whichdistinguishesthreemainprocessesinvolvedinunderstandinganutteranceincontext:

•Constructional‘analysis’:theidentificationofwhichconstructionsareinstantiatedbyagivenutteranceandhowtheyarerelated,alongwiththeassemblyofanassociated“semanticspecification”(or“semspec”)identifyingwhatmeaningschemasareevokedandhowtheyarerelated

•Contextual‘resolution’:themappingofobjectsandeventsinthesemspectothecurrentcommunicativecontext,producinga‘resolvedsemspec’

•Embodied‘simulation’:theinvocationofdynamicembodiedstructuresinthe(resolved)semspectoyieldcontextuallyappropriateinferences

Thefigurealsoidentifiesseveralkindsofinformationcomprehendersbringtobearinunderstandinganutterance:

•Linguisticknowledge:whattheyknowabouttheirlanguage,thatis,asetoflexicalandgrammaticalconstructions,eachastructuredassociationbetweenknowledgeaboutform(suchasphonologicalschemas)andknowledgeaboutmeaning(suchasconceptualschemas)

•Theutteranceitself:whattheyperceive,thatis,thewritten/spoken/signed/gesturedsignal•Thecommunicativecontext:boththephysicalsituationandtheprecedingdiscoursetheyhaveproducedandapprehended

•Worldknowledge:whattheyknowabouttheentitiesandeventualitiesoftheworld

Page 145: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 13 of 16

Clicktoviewlarger

Figure10.4. Overviewofstructuresandprocessesinvolvedinsimulation-basedlanguagecomprehension

Essentially,theanalysisprocessistheconstruction-basedanaloguetotraditionalsyntacticparsing,exceptthatallstructures,includingtheinputsetofconstructions(or‘constructicon’)andtheoutputsemspec,includemeaningalongwithform.Indeed,thisanalogymaybeevidentinFigure10.3:thestructuresinthecentercolumncaptureconstructionalconstituencyrelationsandthuscorrespondtoasyntacticparsetree.(Wecallthiscollectionofstructuresa‘constructgraph’or‘analysisgraph’.)Contextualresolutionslightlybroadensthetraditionalnotionofreferenceresolutiontoapplytoanystructuresinthesemspec,groundingmeaninginaspecificcontext.Bothofthesefirsttwoprocessescanbeseenasconstraint-basedsearches:acomprehenderperceivinganutterancemustfindthesubsetofherconstructiconthatbestexplainswhatsheisperceiving—thatis,theonethatbestsatisfiesallrelevantconstructionalconstraints,accountsfortheformsoftheutterance,andyieldsacoherentsemspec.Shemustthenfindtheparticularmapping(or‘resolution’)ofthesemspectothecurrentcontextthatmakesthemostsense(broadlyspeaking,inwhichthestructuresinthesemspecaremosteasilyidentifiable).Bycontrast,thesimulationprocessisnotoneofsearchbutinvolvesdynamicactivationofstructuresandpotentiallyunboundedinferentialreasoningbasedonsemanticallyandpragmaticallyrichsourcesofinformation;theresultinginferencescorrespondtothebroadestanddeepestsenseofthe“meaning”thatthecomprehenderconstructsfortheutterance.

10.4.3SolvingtheComprehensionPuzzle:BasicRequirementsandPsycholinguisticConstraints

Inthediscussionsofar,wehaveoutlinedtheECGviewofthecomprehensionprocess.Butboththeanalysisandresolutionprocessesareinfactagooddealmorecomplexthanwehavesuggested.Analysis,forinstance,appliesconstraintstodeterminewhichparticularconstructionsandschemasarerelevanttoagivenutterance,andhowtheyarerelated.Herewemightdrawan(imperfect)analogytosolvingajigsawpuzzle(oroneregionofapuzzle),whereconstructionsarelikeindividualpuzzlepiecesthatfittogetherinspecificways,accordingtothe(relativelyhard)constraintsimposedbytheirspecificshapes,aswellasthe(oftensofter)constraintsimposedbyfeaturesoftheirrespectiveimages.Theseconstraintsapplybothlocally(betweenneighboringpieces)andtoprogressivelylargersetsofpieces.Fromthisperspective,Figure10.3showsasolutiontothepuzzlethatsatisfiesthevariousapplicableconstraints,butitdoesnotshowpreciselyhowacomprehendermightarriveatthatsolution.

Thereareseveralminimalcapabilitiesrequiredforamodeloflanguagecomprehension:

•Givenaninpututteranceandasetofconstructions,themodelmustselectasubsetofconstructionsthatcancovertheinput.

•Givensuchasubsetofconstructions,itmustcomposetheseconstructionsintoacoherentconstructgraph.•Givensuchaconstructgraph,themodelmustverifythat(1)theutteranceformsareaccountedfor;and(2)theassociatedsemspecmakessenseincontext.

Amodelequippedwiththeaboveminimalrequirementscould,weclaim,performthebasictaskofactivatingacollectionofconstructionstobeusedtobuildaconstructgraphthatcoverstheinputstringandmakessense.Itcouldalso,incidentally,beusedtodeterminewhethertheinputgrammarcanaccountforagiveninpututterance.

4

Page 146: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 14 of 16

Butthisisonlyasimplifiedversionofthestory.Whenhumanlanguageusersapprehendutterances,theyexploitanumberofheuristicsforprioritizingsearchandpruningirrelevantconstructionsfromconsideration.Theseheuristicsworkbothbottom-upandtop-down.Wordswhoseconcreteformsarepresentintheutteranceplaythemostdirectroleintriggeringconstructionalselection.Atthesametime,conceptualanddistributionalexpectationscausesomeconstructionstobeprimedorprioritizedforearlierconsideration.Andtheseprocessesinteract.Forexample,recognizingaconstructionthatcanserveastheconstituentofalargerconstructionmayfacilitatetherecognitionofanyremainingconstituentsandofthewholeconstruction.

Humanlanguagecomprehendersalsohavetodealwithsubstantialinherentuncertainty,ambiguity,andnoisepresentduringallstagesoflanguagecomprehension.Foranyapprehendedutterance,theremaybemultiplepossibleconstructionalanalysesandmultiplewaysofresolvingreferencestothecontext.Theremayalsobeerrorsintheperceivedinpututteranceorcommunicativecontext.Moreover,someinferencesmaybeonlyprobabilisticallylicensed.Andinmanycases(notablyinmodelingchildlanguagelearning),theinputmaysimplynotbecoveredbythecurrentgrammar.Ingeneral,allthecapabilitiesidentifiedinthissectionaremorerealisticallyseenasproducingoutputsthatarenotcategoricalbutstochastic;interpretationscanbejudgedasrelativelymoreorlesscomplete,coherent,andconsistentwiththecontext.Amorerealisticmodelofcomprehensionthusrequiresthemeansofcombiningpartialanduncertaininformationandevaluatingcandidatestructuresateverystagetochoosethosethatbestcontributetoeffectiveandefficientcomprehension(e.g.,bymaximizingutteranceinterpretabilityincontext,orminimizingconstructionalandcontextualambiguity).

Lastbutfarfromleast,studiesofhumanlanguageprocessingprovidestrongevidencethatthecomprehensionprocessisincremental:itbeginsassoonasthehearerbeginstoperceiveanutterance,andevenbeforeagivenutterance(orclause,orphrase)iscompleted,andresultsofintermediateprocessingcanaffecttheprocessingoftheentireutterance(AltmannandSteedman1988).

10.4.4ACognitivelyPlausibleModelofConstructionalAnalysis

WenowbrieflysummarizeanimplementationofconstructionalanalysiswithintheECGframeworkthatsatisfiesthebasicconstraintsofcomprehensionwhileaddressingsomecomplicationsofcognitivelyplausiblelanguageprocessing.TheconstructionanalyzerdescribedandimplementedbyBryant(2008)usesunificationasthebasicmechanismforcomposingconstructionsandverifyingthattheirconstraintsareconsistent,wherebothconstructionsandschemasarerepresentedastypedfeaturestructureswithunificationconstraintsasspecifiedbytheECGformalism.Butthesearchforthebestanalysisalsoexploitsmanyheuristicstoimproveefficiency,limitsearch,andapproximatetheuncertain,incremental,robustnatureofhumanlanguageprocessing,including:

•Incrementalinterpretation:theanalyzerallowsincrementalleft-to-rightinterpretationoftheutterance.Todothis,itemploysleft-cornerparsingtechniques(ManningandCarpenter1997)tokeeptrackofthecurrentsetofcompetinganalysesandupdatetheirscores,wherepartiallymatchedsubportionsofcomplexconstructionsprovidetop-downexpectationsaboutwhichconstructionsmaynextbeencountered.

•Best-fitinterpretation:theanalyzerdefinesaquantitativeheuristicforcombininginformationfromdisparatedomainsandrankingcandidateinterpretations(andthusguidingparsingdecisions).ThetechnicalimplementationisaBayesianprobabilisticmodelthatintegratesanyavailableinformationaffectingthelikelihoodoftheanalysis,drawnwherepossiblefromcorpusdata.Suchinformationincludes,forexample,lexicalandconstructionalfrequencies;thelikelihoodthatoneconstructionhasanotherasaconstituent;andthelikelihoodthatagivenschemahasaparticularkindoffillerinagivenrole.

•Partialinterpretation:theanalyzerproducespartialanalysesevenwhentheinpututteranceisnotcoveredbythegrammar,orwhenpartsoftheutterancehavebeenomitted.Anextensiontotheanalyzerpermitsanalyseswithomittedconstituents(asoftenencounteredin,forexample,Mandarin)byintegratingthescoreofaninterpretationwiththeresultsofthecontextualresolutionprocess.

Insum,theanalyzerisconsistentwiththeconstructionalview,drawingonallavailableinformationateverysteptoensurethatsyntactic,semantic,andconstructionalconstraintsaresatisfied.Crucially,theearlyincorporationofsemantic,pragmatic,andstatisticalconstraintscandramaticallyreducethesearchspacethatmayresultfrompurelysyntacticapproaches.

Page 147: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

Page 15 of 16

Theconstructionalanalyzerhasbeenappliedtoavarietyoflinguisticphenomena,includingmodelingfamiliesofrelatedargumentstructureconstructions(Dodge2010),earlyMandarinconstructions(Mok2008),andHebrewmorphologicalconstructions(Schneider2010).Besidesservingasaplatformforlinguisticanalysis,ithasalsobeenappliedasapsycholinguisticmodelofreadingtimedata(Bryant2008),andversionsoftheanalyzerhavebeenintegratedinmodelsofchildlanguageacquisition(N.Chang2008;Mok2008).OngoingresearchhasintegratedECGrepresentationsofmentalspacesandmetaphorintotheconstructionalanalysisprocess(FeldmanandGilardiforthcoming),similartoearlierproposals(BergenandChang2005;N.Changetal.2002;Moketal.2004).

10.5.ConclusionandFutureDirections

Thediscoverythatlanguagecomprehensionusesembodiedsimulationhasdramaticramificationsfortheoriesoflanguageknowledgeanduse.Inthischapter,wehavelaidoutaconstruction-basedapproachtogrammarconsistentwiththisfinding.EmbodiedConstructionGrammarprovidesbothformalnotationsandcomputationalmodelsdesignedtosupporttheexplorationandvalidation(orfalsification)ofsuchsimulation-basedapproaches,whilealsoaccountingforthekindsofphenomenaaddressedbyconstructionalapproachesmoregenerally.Wehavefocusedinthischapterongivingabroadoverviewofhowconstructions,bothlexicalandgrammatical,cancontributetoandparameterizeembodiedsimulationsbasedonperceptual,motor,andothercognitivelyandneurallygroundedstructures.ThestructuresandnotationsusedintheECGformalismaremorethanjustawayofexpressinglinguisticanalyses;theyaredesignedtosupportprocessesoflanguageuse,asillustratedbythemodeloflanguagecomprehensiondescribedabove.

Thereremain,however,ahostofdetailsthatthisoverviewhaselided,inthehopethatreadersinclinedtodigdeeperwilldoso.Asforewarned,wehaveomitteddetailedspecificationoftheECGformalism,oranyalgorithmicorimplementationaldetailofthecomputationalmodelsassociatedwithlanguagecomprehension.Amajorstrandofrelatedresearchhasfocusedontheoriesandcomputationalmodelsofchildlanguagelearning(N.Chang2008;N.ChangandGurevich2004;N.ChangandMok2006;Mok2008);thesehaveshownhowaconstruction-basedgrammarformalism,amodelof(partial)languagecomprehensionlikethatdescribedabove,andusage-basedlearningstrategiescanbeintegratedintoacognitivelyanddevelopmentallyplausiblemodelofhowchildrenlearntheirearliestlexicalandgrammaticalconstructions.Also,boththex-schemaformalismandthesimulation-basedapproachtosemanticrepresentationwereoriginallymotivatedbyNarayanan's(1997)modeloflinguisticaspectandconceptualmetaphor(LakoffandJohnson1980);morerecentresearchhasbeenextendingtheECGformalismandimplementationstoaccommodatesuchcross-domainmappings.

AlloftheaboveprovideencouragingevidenceofEmbodiedConstructionGrammar'spotentialforsupportingtheoriesoflanguagestructure,learning,andusage.Manyareasneedmoreattention,ofcourse;mostnotably,amodeloflanguageproductionwouldprovideacomplementaryandcrucialcomponentoflanguageuse.Butwehopethebasicoutlooksetoutbyinvestigationstodatehavedemonstratedthebenefits(andchallenges)oftakingseriouslytheenterpriseofunderstandinggrammaraspartofcognitivescience.Thiscommitmentbindsustoinvestigatemuchmorethanjustthepossiblenatureofrepresentationoflinguisticstructures:wemustalsoaskhowtheyareused,andhowtheyinteractwithothercognitiveandneuralstructures.Onlybyasking—andanswering—therightquestionscanthestudyofgrammarcontributetoabroadercognitivescienceoflanguage.

TheauthorswouldliketoexpresstheirgratitudetoresearchersintheNeuralTheoryofLanguageprojectatUCBerkeley/InternationalComputerScienceInstitutefortheirsubstantialcontributionstothiswork,aswellastoananonymousreviewerforinsightfulcomments,andtocolleaguesattheSonyComputerScienceLaboratoryParisforhelpfuldiscussion.

Notes:

(1.)Someadditionalnotationshavebeenintroducedsinceouroriginalformulation;theseallowoverridingofinheritedconstraintsbysubcases,explicitmarkingofoptionalconstituents,andtheincorporationofdistributionalinformationessentialfortheprobabilisticversionoftheECGanalyzerdescribedbelow.Thebasicstructuralcomponentsremainstable,however.

Page 148: The Oxford Handbook of Construction Grammar

Embodied Construction Grammar

(2.)ThisversionoftheCat-Cxnisamassivesimplificationofboththeformandcontentthatwouldbehypothesizedinafulleranalysis,butitcapturesthegeneralflavoroftheformalization.

(3.)WorkinlanguageprocessingintheECGframeworkhasfocusedonlanguagecomprehension,inparttoexploitthelargerbodyofexperimentalandcomputationalworkoncomprehensionthanonproduction.Acorrespondingmodelofproductionshouldbesimilarlyconstrainedbyempiricalfindings.

(4.)ThisviewofcomprehensionisconsistentwiththatpresentedinBergenandChang(2005),thoughitdrawsaconceptualdistinctionbetweentheprocessesofcontextualresolutionandconstructionalanalysis.Thelineardepictionofthethreeprocessesis,ofcourse,asimplification;intheprocessingofagivenutterancetheymaytakeplaceatleastpartlyinparallel.Thatis,partialutterancesmaybeanalyzed,resolved,andsimulatedevenbeforetheentireutteranceisrecognized,andresultsoftheseprocessesmayfeedbackintoandconstraintheongoinganalysis.

BenjaminBergenBenjaminBergen(Ph.D.2001,UCBerkeley)isAssociateProfessorintheCognitiveScienceDepartmentattheUniversityofCalifornia,SanDiego,wherehedirectstheLanguageandCognitionLab.Hisresearchfocusesonlanguagecomprehension,production,andlearning,andhasappearedinjournalssuchasLanguage,CognitiveLinguistics,CognitiveScience,BrainandLanguage,MemoryandCognition,andtheJournalofChildLanguage.Heisalsotheauthorofaforthcomingbookonmentalsimulationandmeaning,tobepublishedbyBasicBooks.

NancyChangNancyChangisaPost-doctoralResearcherattheUniversitéSorbonneNouvelle(Paris3),combiningcomputationalanddevelopmentalperspectivesontheemergenceofgrammar.ShehaspreviouslybeenaresearcherattheUniversityofCaliforniaatBerkeley,wheresheearnedherdoctoratein2008,andSonyComputerScienceLaboratoryinParis.Incurrentworksheisextendingpreviousmodelsofcomputationalconstructiongrammar,simulationsemanticsandusage-basedgrammarlearningtoencompassmorecomplexargumentstructureandaspectualconstructions.

Page 149: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0011

CognitiveGrammarCristianoBrocciasTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapteroffersanoverviewofRonaldLangacker'sCognitiveGrammar(CG),withspecialreferencetotherelationbetweenCGandconstructionistapproaches.ItexplainsthatalthoughCGwasdevelopedpriortoconstructionistapproaches,itsharesmanyassumptionswiththem.CGviewslanguageasbeinggroundedinembodiedhumanexperienceandlanguage-independentcognitiveprocesses,anditassumesgrammartobeinherentlymeaningful,andthatlanguageconsistsofform-meaningpairingsorassembliesofsymbolicstructures.ThechapteralsoaddressestherelationbetweenlexemesandconstructionsanddiscussessemanticandgrammaticalrolesinCG.

Keywords:CognitiveGrammar,RonaldLangacker,ConstructionGrammar,cognitiveprocesses,form-meaningpairings,grammaticalroles

11.1.Introduction

Takenliterally,CognitiveGrammaris,quitetrivially,atheoryofgrammar.Ithasbeendevelopedoverthelastthirty-fiveyearsbyRonaldLangackerandhasbeendetailedsofarinsixmajorworks(Langacker1987,1991,1999,2000a,2008a,and2009a), butseealsoTaylor(2002)foranintroductoryoverview.Thetheoryhaschangedlittleovertime,althoughtherearesometerminologicaladjustmentsandrefinementsinhowcertaincentralnotionsareunderstoodthatdeservetobepointedout(e.g.,thedefinitionofnoun,theuseoftraditionallabelslike‘grammar’and‘syntax,’theanalysisoftrajectorandlandmark).Further,theriseofConstructionGrammarshasledLangackertodevoteconsiderableattentioninrecentyearstothenotionof“construction”anditsrelationtoCognitiveGrammar(see,e.g.,Langacker2005,2009a:chapter1,2009b,2009c).

CognitiveGrammarhasbeenappliedtovariousphenomenaacrossdifferentlanguages.AnupdatedlistofsomeofthemostrelevantworkinCognitiveGrammarcanbefoundinLangacker(2007)andIwillthereforerefrainfromrepeatingithere.Rather,thepresentoverviewaimstoillustratesomeofthemajorassumptionsCognitiveGrammarrestsuponwithspecialreferencetoLangacker'slatestthinking,asdetailed,inparticular,inLangacker's(2008a)monographentitledCognitiveGrammar(CGforshort).GiventhebreadthofcoverageofCognitiveGrammar,Iwilltouch,especiallyinthefinalsectionsofthischapter,uponthosenotionswhichIthinkareofparticularinteresttopractitionersofconstructionistapproachestolanguage.

11.2.GrammarinCognitiveGrammar

GrammarinCognitiveGrammarisunderstood,rathertraditionally,toinvolvethedescriptionofhowlinguisticelementscombineintocomplexexpressions.Theadjective‘Cognitive’doesnotsimplyrefertothefactthatlanguageisviewedasamanifestationofthehumanmind—inthissenseformalapproacheslikeGenerativeGrammararealsocognitive—butismeanttohighlightthatlanguageisgroundedinlanguage-independent

1

Page 150: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 2 of 14

cognitiveprocessessuchasassociation(establishingpsychologicalconnections),automatization(usingstructureswithoutmuchconstructiveeffort),schematization(extractingageneralstructureorschemaoutofthecommonalityofspecificexperiences),andcategorization(usingstoredstructurestointerpretnewexperience;seeCG:16–18andalsobelow).Thisviewdoesnotexcludethepossibilitythatourcapacityforlanguage(cf.Chomsky'slanguagefaculty)couldbehardwiredintoourgenes.Itsimplystressesthatlanguagecannotbethoughtofasaself-containedmoduleasintraditionalMainstreamGenerativeGrammar(see,e.g.,Taylor2007foranextensivediscussion).

CognitiveGrammaralsostressesthedynamicandopen-endednatureoflanguage.LanguageisultimatelyreducibletopatternsofneuralactivationandthusCognitiveGrammarexcludes,incontrasttoformalapproaches,thepossibilityofanexhaustiveformaldescription,whichpresupposestheconceptualizationoflanguageasaself-containedunit.Rather,CognitiveGrammarbelongstothefunctionalisttradition(seeNuyts2007)inthatitregardsthesemiologicalfunction(i.e.,theexpressionofmeaningthroughsoundsandgestures)andtheinteractivefunctionoflanguage(i.e.,socialinteractioninaculturalcontextthroughlanguage)asfoundational.

PivotaltoCognitiveGrammaristheassumptionthatlanguageconsistsofassembliesofsymbolicstructures,thatis,whatmostConstructionGrammarapproachescall‘constructions’(seealsosection11.5.1).Inparticular,CognitiveGrammarclaimsthatonlythreestructuresareneededtoservethesemiologicalfunctionoflanguage,namelysemanticstructures,phonologicalstructures,andsymbolicstructures.Linguisticexpressionsconsistofsymbolicstructures.Theyarebipolarstructurespairingasemanticpole(the‘meaning’ofanexpression)withaphonologicalpole(its‘form’).Forexample,thelexemeflowerisanalyzedasthepairingofthecomplexconceptualization(ormeaning)constitutingitssemanticpole(conventionallyabbreviatedas[FLOWER])withaphonologicalpole(arepresentationofhowthelexemeispronounced,abbreviated,e.g.,as[flower]),resultinginthesymbolicstructuregivenas[[FLOWER]/[flower]].Crucially,thesemanticpolealsoincludesinformationwhichistraditionallyregardedas‘encyclopedic,’forexample,theknowledgethatflowerscanbetransferredastokensofaffection.Similarly,thelabel‘phonologicalpole’isconstruedbroadlysoastoalsoencompassotherbodilymanifestationssuchasgesture.

Theclaimthatlanguageconsistsofassembliesofsymbolicstructureshastwoimportantconsequences.Oneistheviewoflexiconandgrammarasacontinuum(seealsoCroft,thisvolume,andGoldberg,thisvolume),alsoreferredtoasthe‘constructicon’inCxG.AlthoughCognitiveGrammarretainstraditionallabelslikemorphology,lexicon,andsyntax,itassumesthatnosharpboundariesexistbetweenthem.(Infact,oneofthemostimportantclaimsinCognitiveGrammaristhat,inlanguage,muchisamatterofdegree.)Theotheristhatgrammarismeaningful.Evenabstractelementslikegrammaticalwords(e.g.,prepositions,auxiliaries,determiners)andgrammaticalconstructions(e.g.,theDitransitiveconstruction,asexemplifiedbyHegaveherapresent)aremeaningful,sincetheyinvolveasemanticpolebydefinition.Inmoredetail,linguisticexpressionscanbeanalyzedalongthreedimensions:symboliccomplexity,specificity(or,conversely,schematicity),andentrenchment/conventionality.Higher-levelsymbolicstructuresrestonlower-levelones,event<eventful<eventfulday<aneventfulday<onaneventfulday,thusachievingahigherdegreeofsymboliccomplexity.Structurescanalsovaryintermsofspecificityorschematicity.Forexample,thestructureonaXday—whereXstandsforaslotintheconstruction—ismoreschematic(lessspecific)thanonaneventfulday.Entrenchmentreferstoalinguisticexpressionhavingunitstatus—i.e.,beingaccessedautomatically,withoutmuchconstructiveeffort—foraparticularspeaker,whileconventionalitypertainstotheentrenchmentofalinguisticexpressionwithinaspeechcommunity.Novelexpressionsarethusthosewithalowdegreeofentrenchment/conventionalitywhereasfixedexpressionsorlexicalitemshaveahighdegreeofentrenchment/conventionality,thatis,theyhaveachievedthestatusofconventionalunits.ItmustbestressedthatlexicalitemsdonotonlyincludesinglewordsbutalsomultiwordexpressionslikeIloveyou,withbatedbreath,inthemiddleofnowhere,seeyoulater,whatareyoudoingtonight?andsoon.AsisshowninFigure11.1,whicharrangeslinguisticexpressionsintermsofthetwodimensions‘symboliccomplexity’and‘schematicity,’prototypicallexicalitems(e.g.,cat)tendtobecharacterizedbyalowerdegreeofsymboliccomplexity(theytendtobesinglewords)andtobefairlyspecificsemantically(itiseasytodescribewhatacatis).Symbolicallymorecomplexitemssuchasseeyoulaterarealsopartofthelexicon,sincetheyarefairlyspecificsemantically(aswellasphonologically).Grammar,bycontrast,hastodowithschematicsymbolicassembliesorconstructionalschemas(CG:167).Whataretraditionallyknownasrulesareunderstoodasschemasorpatternsusedtoformcomplexexpressions(constructions)inCognitiveGrammar.PartlyspecificstructuressuchasonaXdaycannotbeassignedexclusivelytoeitherthe

Page 151: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 3 of 14

lexiconorgrammar,sincetheyareneitherfullyspecificnorfullyschematicandthusillustratethecontinuousnatureoflexiconandgrammar.

Clicktoviewlarger

Figure11.1. Thegrammar–lexiconcontinuum

Evengrammaticalmarkerslikeauxiliaries,agreementmarkers,andderivationalaffixes,whichareallspecificphonologically,areanalyzedasmeaningfulelementsinCognitiveGrammar.Theyinvolveasemanticpolebyvirtueofbeinglinguisticexpressionsalthoughtheirsemanticpoleisschematicandthereforemaybedifficulttodescribe.Grammaticalclassesdifferfromgrammaticalmarkersinthattheyarealsoschematicphonologically.However,likegrammaticalmarkerstheyareclaimedtobemeaningful.Thenounclass,forexample,isabbreviatedas[[THING]/[…]]:anounreferstoathing(atechnicalterminCognitiveGrammar,seesection11.4)andismaximallyschematicfromaphonologicalpointofview.

CognitiveGrammaralsoretainsthetraditionallabels‘morphology’and‘syntax.’Withingrammar,morphologyinvolvesschematicsymbolicassemblieswhoseinstantiationsaresinglewords(e.g.,N+ful,asineventful),whilesyntaxdealswithschematicsymbolicassembliesthathavemultiwordinstantiations(e.g.,N +fulN ,asineventfulday).

Insum,languagecanbedescribedas“astructuredinventoryofconventionallinguisticunits”(CG:222;seealsoBybee1985).Aswaspointedoutabove,aunitisastructurethataspeakeremploysinlargelyautomaticfashion.Linguisticunitsconsistofsymbolicassociationsbetweenasemanticandaphonologicalstructurewhichhaveunitstatus.Conventionalityimpliesthatsomethingissharedbyasubstantialnumberofindividuals.Bytheterm‘inventory,’itismeantthata“grammarisnota‘generative’description,providingaformalenumerationofallandonlythewell-formedsentencesofalanguage”(Langacker1987:63).Still,suchaninventoryisstructuredbecauselinguisticelementsarerelatedtooneanotherasifinanetwork.

ThereisnoroomfortransformationsorabstractelementsliketracesinCognitiveGrammarbecauseCognitiveGrammarmakesonlythreeprovisions,collectivelyknownasthecontentrequirement,forthelinguisticsystem,namely:(1)semantic,phonological,andsymbolicstructures(whichIhaveconsideredsofar);(2)schematization;and(3)categorization,whichIreferredtoattheveryoutset.

Schematizationreferstotheprocessbywhichthecommonalityinherentinasetofexpressionsisabstractedawayandstoredinschemas. Thisappliestobothlexiconandgrammar.Alexicalitemlikeflowerismoreschematicthantheconceptionsevokedinspecificusageevents(e.g.,whenreferringtoaflowersomeonewasgivenonaspecificoccasion),nottomentionthefactthatitislessspecificphonologically.Similarly,grammarisemergentbecauseitconsistsinsymbolicassembliesthatemergeoutofconcreteusageeventsbywayofschematization(e.g.,theDitransitiveconstructionemergesoutofrepeatedexposuretoexpressionslikeHegaveherapresent,Ihandedhimaslipofpaper,Ithrewhertheball,etc.).Moregenerally,sincelinguisticunitsemergeoutofspecificusageevents,CognitiveGrammarcanbedescribedasausage-basedmodeloflanguage(seealsoCroft2001andBybee,thisvolume).

1 2

2

3

Page 152: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 4 of 14

Storedschemasareusedtocategorize(orsanction)newexperience,whichconstitutesthetargetofcategorization,whereasthesanctioningschemaiscalledthestandardinthecategorizingrelationship.Twodifferenttypesofcategorizationarerecognized:elaboration(orinstantiation)andextension.UponhearingtheutteranceShehurledhimtheball,wemaycategorizeitasanelaborationorinstantiationoftheDitransitiveconstruction.TheutteranceisfullycompatiblewiththeDitransitiveconstruction,whichisschematicwithrespecttoit,sincetheutterancespecifies(amongotherthings)thataforcefultransfer,thatoftheball,waseffected.Categorizationishoweveramatterofconstrual.Forinstance,Goldberg's(1995)famousexampleShesneezedthenapkinoffthetableissanctionedbytheCausedMotionconstruction(cf.Sheknockedthevaseoffthetable)onlyinsofarassneezeisconstruedasaforcefulactioncapableofdisplacingobjects.Infact,thisexample(seeLangacker2009b)maybeviewedasaninstanceofextension.Thistypeofcategorizingrelationshipoccurswhenthetargetofcategorizationisonlypartlycompatiblewiththesanctioningschema.Lexicalexamplesareeasytocomeby.Mouseasacomputerdeviceis,forinstance,anextensionofmouseasananimalbyvirtueofsomeperceivedsimilarity(e.g.,inshape)althoughtheirreferentsconflictintermsofanimacy.Crucially,extensiongivesrisetoprototypes.Bothmouse‘computerdevice’andmouse‘animal’areprobablyperceivedasbelongingtothesamecategorywithinwhichmouse‘animal’is(possiblystill?)aprototype.Similarly,insofarassneezeisconstruedasaforcedynamicverb,knockisintuitivelymoreprototypicalforthiscategorythansneeze.

11.3.MeaningasConceptualization

Unlikeotherconstructionistapproaches,CognitiveGrammarrestsonaveryelaborateconceptualistviewofmeaning.Meaningisviewedasgroundedinembodiedhumanexperience(seealsoBergenandChang,thisvolume,forasimilarpositioninEmbodiedConstructionGrammar)andasresidinginthemindsofspeakers(CG:28–29)asconceptualization,atermwhichismeanttohighlightitsdynamic(processing)nature.Conceptualizationisimagistic,ratherthanpropositional,becauseitinvolvesbasicnotionslikeminimalandconfigurationalconcepts(e.g.,thenotionsofcolorandcontact,respectively)andconceptualarchetypes(e.g.,anobjectmovingthroughspace).Further,conceptualizationisinteractivebecauseitinvolvessomedegreeofnegotiationbetweeninterlocutors.Finally,conceptualizationmakespervasiveuseofimaginativeabilitieslikementalspaceconstructionandmanagement,blending(seeFauconnierandTurner2002)andvirtualityorfictivity(weoftentalkaboutimaginaryentities).

Giventhedynamicnatureofconceptualization,CognitiveGrammarrefutesthedictionaryviewoflinguisticmeaningprevalentinformallinguistics,whichregardslexicalitemsasclearlydefinablebuildingblocksthatcanbeusedcompositionallyfortheconstructionofcomplexmeanings.Langackerstressesthatlexicalmeaningsareencyclopedicinthattheyareconventionalpathsofaccesstoopen-endeddomainsofknowledge(CG:41)andcanbethoughtofasscaffoldingorpropsforthecreationofmeaningsofcomplexexpressions.Forexample,despitetheirstructuralidentity,thebeachissafeandthechildissafecanmeantwoverydifferentthings.Intheformer,thebeachcanbeinterpretedasnotbeinglikelytocauseanyharmtoachild(thebeachisagent-like),whileinthelatterthechildisnotlikelytobeendangered(thechildispatient-like).Thedifferenceininterpretationdependsonhowweaccessourknowledgeconcerningbeachesandchildrenrelativetothenotionofdanger.Thisalsoimpliesthatthemeaningofacomposite(orcomplex)expressionincludesthewayinwhichitscomponentstructuresleadtothecompositeexpression'ssemanticpole,thatis,itscompositionalpath(seealsosection11.5.1).

CognitiveGrammarcontendsthatlinguisticmeaninginvolvescontentandconstrual.Conceptualcontentisbasedonasetofdomains(collectivelyknownasamatrix),wheretheterm‘domain’refersto“anykindofconceptionorrealmofexperience”(CG:44)andisroughlyequivalenttoFauconnierandTurner'snotionof‘mentalspace’(CG:51).Forexample,petalcanonlybeunderstoodwithreferencetothedomainflower.Domainsarenotallequalbecause,givenacertainexpression,somedomainsaremorecentralorlikelytobeactivatedthanothers(e.g.,stabbingisprobablymoreprominentwithdaggerthanknife).Rankingofdomainsforcentralityisaninstanceofconstrual,bywhichconceptualcontentcanbeportrayed(orconstrued)invariousdifferentways.Forexample,acontainerwithwaterinit(conceptualcontent)couldbedescribed,underappropriatecircumstances,ashalfemptyratherthanhalffull.Construalinvolvesavarietyofdimensions,includingscope.Anexpression'sscopeisdefinedas“theconceptualcontentappearinginthesubjectiveviewingframeinherent”intheapprehensionofanexpression(CG:63).Inordertomakesenseofthis,itisusefulfirstofalltodistinguishbetweenmaximalandimmediatescope,sincethetwodonotalwayscoincide.Althoughthelexicalitemfingernail,

Page 153: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 5 of 14

forexample,evokestheconception(domain)ofthehumanbodyasitsmaximalscope(thefullextentofthesubjectiveviewingframeorscopeofawareness),theexpressionisapprehendedwithreferenceto(foregrounds)aspecificsubpartofthebody,namelythefinger,whichthusconstitutestheexpression'simmediatescope,whichisalsoreferredto(metaphorically)astheonstageregionor,alternatively,base.Withinthatbase,fingernailreferstoorprofilesaspecificsubstructure,whichisthefocusofattention.

Clicktoviewlarger

Figure11.2. Theinteractionofcontent,profile,andtrajector/landmarkalignment

InCognitiveGrammar,anylinguisticexpressionprofileseitherathingorarelationship.Forexample,parentandchild(seeFigure11.2)evokethesamerelationalcontent(an‘offspring’domain)butdonotrefertoarelation;theybothrefertothings.However,theyprofiledifferententities(aparentandachildarenotthesamething).Bycontrast,haveaparentandhaveachild,whichalsoevoketherelationaloffspringcontent,bothprofilearelationship.Inthiscase,theprofile(orreference)isidenticalbecausebothhaveaparentandhaveachildprofilearelationshipbetweenaparentandachild.Theexpressionsaredistinguishedbyvirtueoftrajectory(tr)/landmark(lm)alignment.Thetrajectororprimaryfocusisthemostprominentelement,whilethelandmark,ifitexists,isasecondaryfocus.Inhaveaparent,thetrajector(e.g.,TominTomstillhasaparent)isachildandthelandmarkisaparent,whileinhaveachildthetrajectorisaparentandthelandmarkisachild.

11.4.GrammaticalClasses

OneofthemostinterestingclaimsmadebyCognitiveGrammaristhatatleastsomegrammaticalclasses(‘partsofspeech’or‘wordclasses’)suchasnounandverbmaybeuniversalandmaybedefinedsemantically,sinceschematiccharacterizationsvalidforalltheirmembersareviable.Itisperhapsnotcontroversialtostatethattheprototypeforthenounclassistheconceptionofphysicalobjectbuttoclaimthatnounscanbedefinedschematicallycontrastswithmuchcurrentlinguisticthinking,evenwithintheconstructionistcamp.RadicalConstructionGrammarandCognitiveConstructionGrammar,forinstance,donotmakethisassumption(seealsoBoas,thisvolume;Croft,thisvolume;Goldberg2006a:220–25,thisvolume;Langacker2009c).

CognitiveGrammarproposesthattheprototypesforgrammaticalclassesinvolveexperientiallygroundedarchetypessuchastheconceptionofaphysicalobjectforthenounclassandtheconceptionofaforce-dynamiceventfortheverbclass.Schemasforthememergethroughbasiccognitiveabilities“initiallymanifestedinthosearchetypesandlaterextendedtoothercases”(CG:103).Attheschematiclevel,thenounclassinvolvesourcapacityforgroupingandconceptualreification(seeCG:105),thatis,ourabilitytoconstrueunitaryentitiesatasufficientlyhighleveloforganization,andtheverbclassinvolvesourcapacityforapprehendingrelationshipsandtrackingthemthroughtime(i.e.,scanning,seealsoBrocciasandHollmann2007).

TheCognitiveGrammaranalysisof(thesemanticpoleof)wordclassesissummarizeddiagrammaticallyinFigure11.3.Shadedboxesindicateconceptionsthatarethesemanticpolesoftraditionalwordclasses;forexample,non-processualrelationships,unlikethings,donotcorrespondtoanyparticularwordclass.ThediagramsusuallyemployedinCognitiveGrammartodepictmostoftheconceptionsinFigure11.3areprovidedinFigure11.4.

Westartwiththeverygeneralcategoryofentities,whichcanbedividedintothingsandrelation(ship)s.Athing,atechnicalterminCognitiveGrammar,isanyproductofgroupingandreificationandistheschematicsemanticpoleofthenounclass.Forexample,thenounteam(seeLangacker1987:197)profilesasetofentitiesratherthansinglingoutanyconstitutivemember.Relationships,bycontrast,profileinterconnectionsbetweenentities.Relationscanbeeitherprocessesornon-processual(oratemporal)relationships.Aprocessprofiles

Page 154: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 6 of 14

arelationshipmadeupofvariouscomponentstates(acomplexrelationship)scannedsequentiallythroughprocessingtime(CG:111and122),aswhenwatchingaballfallinamotionpicture.Forthisreason,aprocessissaidtohavea‘positive’temporalprofile(i.e.,timeisfocused,asisshownbytheheavytimelineinFigure11.4(c)).Processesconstitutethesemanticpoleofverbs(e.g.,enter).

Clicktoviewlarger

Figure11.3. TheCognitiveGrammaranalysisofgrammaticalclasses

Clicktoviewlarger

Figure11.4. CognitiveGrammardiagramsforgrammaticalword-relatedconceptions

Inanimportantdeparturefromearlierwork,CGnolongeranalyzesallrelationshipsasinvolvingatrajectorandalandmarkbecausetrajectorandlandmarkarenowdefinedintermsoffocalprominence(seealsosection11.6).Forexample,theverbarrivedepictsarelationshipbetweenamoverandspacebutonlythemoverisregardedasafocalelementand,hence,beingunique,theprimaryfocusofattentionortrajector.Thespatialpositionsoccupiedbythemover,includingthelocationshearrivesat(thegoal),aresaidtobebackgroundelementsandthuslackfocalprominence(CG:71–72and113).Consequently,theyarenottreatedaslandmarks.Itisalsoworthpointingoutthatarelationshipisconceptuallydependentonitsparticipantsbecausetheyareevokedasanintrinsicpartofitsownconceptualization.Bycontrast,athingcanbeconceivedofindependentlyofanyrelationshipsinwhichitmayparticipateandisthussaidtobeconceptuallyautonomous.Non-processualrelationshipshavea‘null’temporalprofile(timeisbackgrounded)andcomeintwotypes:simplexnon-processual(orstative)relationshipsandcomplexnon-processual(oratemporal)relationships.Simplexnon-processualrelationsinvolveasingleconfigurationthroughtimeandcorrespondtothesemanticpoleofadjectives(simplenon-processualrelationswithonlyonefocalparticipant,atrajector,correspondingtoathing),stativeprepositions(suchasin,asopposedtodynamicprepositioninto)andadverbs(simplenon-processualrelationswithonlyonefocalparticipant,atrajector,correspondingtoarelationship,butseealsoBroccias2011).Complexnon-processualrelations(e.g.,into)aremadeupofmorethanonecomponentstateovertimebutsuchcomponentstatesarescannedinsummaryfashion,thatis,thecomponentstatesaresuperimposeduponeachothersothatasingle,complexgestaltbecomesavailable(asinamultiple-exposurephotoofaball'sfall).Prepositions,bothofthesimplexandcomplextypes,differfromadjectivesandadverbsinthattheyinvolvetwo,ratherthanone,focalparticipants,atrajector(whichisanentity,eitherathingorarelationship)andalandmark(whichisathing).Observethatscanningiscrucialindistinguishingbetweenexpressionswhichhavethesameconceptualcontentlikeenterandinto.Theybothinvolvemotiontowardandattainmentofagoalbuttheydifferintermsofconstrual.Enterprofilesarelationshipscannedsequentiallywhileintoprofilesarelationshipscannedsummarily.Summaryscanningisalsopostulatedtobeinvolvedinto-infinitives(e.g.,toenter),presentparticiples(e.g.,entering)andpastparticiples(e.g.,enteredinhaveentered).

11.5.ConstructionsinCognitiveGrammar

Page 155: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 7 of 14

Sincethenotionofconstructionhasbecomeincreasinglyimportantoverthelasttwodecades,Langacker'srecentworkhasalsodealtextensivelywithitanditsstatusinCognitiveGrammar.HereIwilloverviewwhatCognitiveGrammarmeansby‘construction’andhowittacklestherelationbetweenlexemes,inthetraditionalsenseof‘singlewords,’andconstructions.

11.5.1AssembliesofSymbolicStructures

Langackerdefinesconstructionsasassembliesofsymbolicstructures(seesection11.1above).Mosquitonet,forexample,isacompositesymbolicstructureresultingfromtheintegrationofthecomponentsymbolicstructuresmosquitoandnet.ObservethatinCognitiveGrammarevenpluralnounslikeflowersareanalyzedasassembliesofsymbolicstructures.Plural-sisregardedasasomewhatschematicnoun,whosesemanticpoleprofilesamassofthings,intheCognitiveGrammarsenseoftheterm,anditsphonologicalpoleanunspecifiedsequenceofsoundsendingwith,forexample,/z/.Thesymbolicstructurecustomarilycalledplural-sisintegratedwiththesymbolicstructure[[FLOWER]/[flower]]resultinginthepluralnounflowers.

Itisimportanttostressthataconstructiondoesnotonlyresideinitscomponentstructuresbutalsointheirmodeofintegrationandtheresultingcompositestructure.Ifwedidnotpayattentiontohowintegrationiseffected,wewouldbeatalosswhentryingtoexplainwhymosquitonetandbutterflynet,despitetheirstructuralidentity,mean(i.e.,profile)twoverydifferentthings.Inordertoillustratethispointfurther,letusconsideronepossiblecompositionalpathforthenominalthecatunderthetable(seeFigure11.5),whereforthesakeofsimplicityIwillignorethecontributionofthedefinitearticlethe(hencetheparenthesesarounditinFigure11.5).

Clicktoviewlarger

Figure11.5. Acompositionalpathforthecatunderthetable

Compositionreliesoncorrespondence,theconceptualoverlapbetweencomponentstructures(‘horizontalcorrespondences,’shownasdashedhorizontallinesinFigure11.5),aswellasbetweencomponentstructuresandcompositestructures(‘verticalcorrespondences,’shownasdashedverticallinesinFigure11.5).Thesemanticpoleofthenominalthetable,forexample,profilesathingwhichcorrespondstothelandmarkoftheprepositionunder,sothatintegrationbetweenthenominalandtheprepositioncanbeeffected.AtthenextlevelofthecompositionalpathinFigure11.5,thetrajectorofunderisputincorrespondencewiththeprofileofthenominalthecat.Inverticalterms,thelandmarkofunderatthelowestlevelofthecompositionalpathcorrespondstothemoreelaboratelandmarkoftherelationalpredicateunderthetable(i.e.,thetable).Thatis,fromaverticalperspective,(vertical)correspondenceshowswhichfacetsofthecompositestructurearesymbolizedbythecomponentstructures.Componentandcompositestructuresarealsolinkedbyrelationshipsofcategorization.Thelandmarkofunder,forexample,isschematicwithrespecttothetable.Hence,thetableelaboratesorinstantiatesthelandmarkofunder.Thelandmarkofunderconstitutesanelaborationsite(ore-site),whichisrepresenteddiagrammaticallythroughhatching.Thesolidarrow(→)connectingthee-sitetothenominalthetablerepresents

4

Page 156: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 8 of 14

thecategorizingrelationshipofelaboration.Similarly,underthetableisanelaborationofundersinceitspecifiesitinmoredetail.Bycontrast,therelationbetweenthetableandunderthetableisoneofextension,asindicatedbythedashedarrow(––>),becausethetableprofilesathingwhileunderthetableprofilesarelationship.Thisisalsoimplicitinthechoiceofaheavy-lineboxfortheprepositionvis-à-visthenominalthecatatthelowestlevelinthecompositionalpathinFigure11.5.Theheavy-lineboxindicatesthattheprepositionfunctionsasprofiledeterminant.Thatis,thecompositestructureunderthetableinheritstheprofileofthepreposition,sinceitdesignatesarelationshipofspatialorientation,notatable.Similarly,atthenextlevelofintegration,thenominalthecatfunctionsastheprofiledeterminantoftheoverallconstructionthecatunderthetable,sincethisconstructionprofilesacat,notarelationshipofspatialorientation.Insum,Figure11.5showsthatcomponentstructuresdefineacompositionalpathtoward,thatis,theyarestepping-stonesforarrivingat,themeaningoftheoverallexpression.Thecompositestructurestandsintheforeground,butitscompositionalpathisalsoanimportantpartofitsformandmeaning(CG:166–67).

AfewmoreobservationscanbemadeinconnectionwithFigure11.5.Firstofall,itturnsoutthatcorrespondencebetweencomponentstructuresdoesnotalwaysinvolveawholeprofiledentity.InThedogbitthecat,forexample,theintegrationoftheprocessbitwiththenominalthecatapparentlyreliesonthecorrespondencebetweentheverb'slandmarkandthenominal'sprofile.Butitisverylikelythatthedogonlybitapartofthecat,forexample,itstail.Athigherresolution,thecorrespondencethereforeobtainsbetweentheverb'slandmarkandasubpartoractivezoneofthenominal'sprofile(thetailisanactivezonewithrespecttothecat).Moregenerally,anactivezoneistheentity(orentities)associatedwithaprofiledelementwhichistargetedinordertoeffectarelationofcorrespondence(seeLangacker2009a:chapter2).

Asecondimportantpointregardstherelationbetweenthenotionsofprofiledeterminantandelaboration,ontheonehand,andtraditionalgrammaticalnotionslike‘head,’‘complement,’and‘modifier,’ontheother. CognitiveGrammarcontendsthatthenotionsofprofiledeterminantandelaborationcanbeappealedtoinordertoprovideaconceptualcharacterizationofthesetraditionalgrammaticalnotions.Thenotionofprofiledeterminantallowsustorecastthetraditionalnotionofheadinconceptualtermsbyequatingheadwithprofiledeterminant(CG:194).Sometimes,however,thingsarenotsosimple.InappositioncaseslikemynieceMelanie,theprofileofthecompositestructureissaidtocorrespondtobothcomponentstructureprofiles,thatis,thecomponentstructuresbothprofilethingsandtheirprofilescorrespond(butseeAcuña-Fariña2009foradifferentanalysis),soitisimmaterialtosaywhetherbothornoneareprofiledeterminants(seeCG:195). Further,anexpressionmayhaveaprofilewhichdoesnotcorrespondtoanyofitscomponentstructures.Anexocentricexpressionsuchaspickpocket,forinstance,designatesathingbutitsprofiledoesnotcorrespondtoeitherofitscomponentstructures.Pickprofilesaprocess,notathing;pocketprofilesathingbutitdesignateswhatispicked,notthepersonwhopickspockets,whichiswhatthecompositestructureprofiles.

Thenotionofelaborationcanberelatedtothatofautonomous/dependent(A/D)alignment.Aswaspointedoutabove(seesection11.4),relationshipsarenecessarilydependentontheentitiestheyinterconnect,whilethingsaremuchmoreautonomous.Under,forexample,profilesarelationshipbetweentwofocusedentitiesandisthusdependentonatrajectorandalandmark.Thenominalsthecatandthetableelaboratethemandaremuchmoreautonomous.Whileitispossibletothinkofcatsandtablesinrelationtolocations,spatialconcernsareprobablynotasalientpartoftheirbases,andnominals,ingeneral,canbedescribedasrelativelyautonomous(ormoreautonomousthanrelationships).A/Dalignmentcanbeusedtoprovideaconceptualcharacterizationofthetraditionalterms‘complement’and‘modifier.’Acomplementisdefinedas“acomponentstructurethatelaboratesasalientsubstructureofthehead”(CG:203).Amodifierisdefinedas“acomponentstructurethatcontainsasalientsubstructureelaboratedbythehead”(CG:203).Inourexample,thetableisacomplementbecauseitelaboratesthelandmarkoftheheadunder.Underthetableisamodifierbecauseitstrajectoriselaboratedbythecompositeexpression'sheadthecat.

OnefinalpointconcernsthecompositionalpathillustratedinFigure11.5,whichincludesthreelevels,andmayresembletheclassicconstituencyhierarchyrepresentedinFigure11.6.

CognitiveGrammaralsoapproachesconstituencyinconceptualterms.Itidentifiesitwiththeorderofcompositionofthecomponentstructures,whichcanthusdefinevariousdifferentconceptualgroupings.Thesamecompositeconceptioncanbearrivedatviadifferentpaths,twoofwhich,inadditiontotheoneillustratedinFigure11.5,areofferedinFigure11.7fortheconstructionthecatunderthetable—butnoticethatLangackerdoes

5

6

7

Page 157: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 9 of 14

notaddresstheissueoftherelativelikelihoodofdifferentcompositionalpaths.Bydefiningconstituencyasconceptualgrouping,CognitiveGrammarisable,amongotherthings,todispensewiththeneedformovementrulespositedinderivationaltheoriesofgrammar.Forexample,thesentenceTheclaimwasdismissedthatshewasaspycontainsadiscontinuousnominal,whichisanalyzed,insomegenerativeapproaches,astheresultofthemovementofthethat-clausefromitsunderlyingpositionadjacenttoclaim(cf.theclaimthatshewasaspy)totheendofthesentence.Bycontrast,CognitiveGrammarclaimsthatthediscontinuousnominalresultsfromadifferenceinconceptualgroupingwithrespecttoTheclaimthatshewasaspywasdismissed.Inthediscontinuousversion,thenominaltheclaimcombineswithwasdismissedand,athigherleveloforganization,theclaimwasdismissedisintegratedwiththatshewasaspythankstoacorrespondencerelationshipbetweenthenominaltheclaimandthelatternominal(thenominalthatshewasaspyelaboratesasubstructureofclaim,namelyitslandmark).

11.5.2TheInteractionBetweenLexemesandConstructions

Clicktoviewlarger

Figure11.6. Constituency-basedanalysisoftheconstructionthecatunderthetable

Clicktoviewlarger

Page 158: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 10 of 14

Figure11.7. Alternativecompositionalpathsfortheconstructionthecatunderthetable

Clicktoviewlarger

Figure11.8. Therelationbetweenlexemesandconstructions

Sincelexiconandgrammarmakeupacontinuum,CognitiveGrammar,likeRadicalConstructionGrammar,claimsthatitisnotcorrecttothinkoflexemesindependentlyofthestructuralframesinwhichtheyoccur(seeCG:240–41andTaylor2006).Rather,theissueisoneofdecontextualization,thatis,towhatextentlexicalitemsareaccessedindependentlyofthestructuralframesinwhichtheyappear.HowtheinteractionbetweenlexemesandstructuralframesishandledinCognitiveGrammarcanbeillustratedbyconsideringtheCausedMotionconstruction,asinPeterkickedtheballintothestands(seeLangacker2009bandalsoCG:244–50forasimilaranalysisbasedontheDitransitiveconstruction).Letusassume,forthesakeoftheargument,thatthebasic(prototypical)senseofkickisthetransitiveone(cf.Peterkickedthedog),thatis,kickprofilesaforcedynamicinteraction(seethedoublearrowinFigure11.8a)betweentwoentitiesbutnotthepotential,ensuingmovementofthelandmark(simplearrowinFigure11.8a).ThismeansthatkickisnotcongruentwiththeCausedMotionconstruction,whichprototypicallyprofilesmotionasaresultoftheapplicationofforce(seetheupperboxinFigure11.8a).Kickwouldthenillustrateaskewinguse,“whereaverbisnon-congruentwiththeconstructionitappearsin”(Langacker2009c:171).TheintegrationofkickwiththeCausedMotionconstructionispossiblethroughthecategorizingrelationshipofextension,asindicatedbytheverticaldashedarrowinFigure11.8a.Kick(thetargetof

Page 159: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 11 of 14

categorization,TinFigure11.8a)isapprehendedasacausedmotionverb(thestandardofcategorization,SinFigure11.8a)byblendingwiththeconstructionalschema,whichresultsinthetargetT′,aninstantiationoftheCausedMotionconstructionwithkickintheverbal‘slot.’Throughrepeateduse,thepatterninFigure11.8acanbecomeconventionalized,seeFigure11.8b(wherethestraightangledLshapeinplaceoftheroundedangledLshapeismeanttovisualizeexactlythat)sothatkickcannowbesaidtohaveanew(causedmotion)sense,thatis,T′inFigure11.8b.Fromtheverb'spointofview,Figure11.8brepresentsapatternofsemanticextension,sincetheverbisnowapprehendedasacausedmotionverbinthecausedmotionstructuralframe.Fromtheconstruction'spointofview,theconventionalizationoftheblendedmeaningimpliesthatanewschemafortheCausedMotionconstruction—theschematizationofthecategorizationrepresentedinFigure11.8aresultinginahigher-ordergeneralization(CG:249)—emerges,whichcountsasanextensionfromtheprototype.Ifinterpretedasaschema,thestructureinFigure11.8bcanbeviewedaswhatLangackercallsanaugmentedconstructionalschema(seeCG:249).Sincethecausedmotionsenseofkicknowhasunitstatus,itcanprogressivelyloseanalyzability:itscausedmotionsensecanprogressivelybeaccessedwithoutreferencetoitsbasictransitivesense,asisshowninFigures11.8c–d.

11.6.SemanticRolesandGrammaticalRoles

Asisthecasewithgrammaticalclassessuchasnounandverbandgrammaticalnotionssuchasheadandcomplement,CognitiveGrammaralsostrivestoprovideaconceptualdescriptionofsemanticrolesandgrammaticalroles(orgrammaticalrelations).Pivotaltothisendeavoristheassumptionthatclausestructureisgroundedinbasichumanexperience.Thatis,clausestructurecanbedescribedwithreferencetovariousconceptualarchetypessuchasthebilliard-ballmodel(whichbasicallydescribesanagent–patientinteraction)andthestagemodel(whichinvolvesthedirectingandfocusingofattentionbydistinguishingbetween,metaphoricallyspeaking,the‘viewers’andwhatisbeing‘viewed’orputon‘stage’).Semanticroles,ofwhichanexhaustivelistisnotpossibleinLangacker'sview,includeagent,patient,instrument,experiencer,mover,andzero(thelatterroughlycorrespondingtowhatistraditionallyknownastheme). Theseareviewedasarchetypalroles,thatis,rolesdefinedwithrespecttoconceptualarchetypes.

CognitiveGrammaralsooffersaconceptualcharacterizationofgrammaticalrolesorrelations.Theirdescriptionisnotonlybasedonconceptualcontent,asisthecasewithsemanticroles,butalsoonprominence(CG:437).Inparticular,CognitiveGrammarclaimsthat,asisthecasewithnounsandverbs,itispossibletodefinesubjectandobjectbothprototypicallyandschematically.Importantly,theirschematiccharacterizationsmayturnouttobeuniversal(i.e.,applicabletoalllanguages),becausetheyaregroundedinconstrual(inparticular,thefocusingofattention;seeCG:382).Asubjectisidentifiedprototypically—atleastinEnglish—withanagent(i.e.,anenergysource)andisdefinedschematicallyas“anominalthatcodesthetrajectorofaprofiledrelationship”(CG:364). Thereferentofasubjectisthereforeaprimaryfocalrelationalelement.Anobjectisidentifiedprototypically,atleastinEnglish,withapatient(i.e.,anenergysink)andisdefinedschematicallyasanominalthatcodesthelandmarkofaprofiledrelationship(seeCG:364).Thereferentofanobjectisthereforeasecondaryfocalrelationalelement.If(thereferentof)anobjectisconstruableasapatient,Langackerusesthetermdirectobjecttodescribeit.Inotherwords,Langackerusesthelabel‘directobject’(and,hence,transitivity)restrictively,torefertothosenominalsthatallowpassivization,sincepassivizationistakentobesymptomaticofpatient-likeconstrual.Further,thebroadernotionof‘object’isnotonlylimitedtoparticipants(seeCG:370andCG:388n.19).Examplesofobjectsthatarenotparticipantsarepaths(Wehikedanewtrail),locations(Thetrainapproachedthestation),andmeasurements(Itweighstenkilos).Similarly,anominaltrajector(i.e.,subject)canbeasettingoralocationratherthanaparticipant,asinThegardenisswarmingwithbeesandThisbookcontainsalotofinformationonConstructionGrammars,wherethegardenisasettingandthisbookisa(metaphorical)location(see,e.g.,CG:361,374n.19,and387).

AnothertypeofobjectrecognizedinCognitiveGrammaristheindirectobject,whoseprototypeisidentifiedwithanexperiencerandwhichismarkedbyapreposition(typicallymeaning‘to’or‘at’)ordativecase.IncaseslikeItalianASallypiaceilcioccolato(lit.‘at/toSallylikesthechocolate’;i.e.,‘Sallylikeschocolate’),theindirectobject,Sally,isalandmark. WhenanindirectobjectoccurswithadativeverblikeItaliandare(‘give’),asinHadatoillibroaSally(lit.‘(s/he)hasgiventhebooktoSally’),Langackerdoesnotcommithimselftoauniqueanalysis.Hesaysthat“[p]erhapstheindirectobjectshouldbeconsideredasecondarylandmark.Ifnot,itisatleastquitesalientasaprofiledparticipant”(CG:393).LikeItaliandare,Englishgivetakesthreeparticipants(an

8

9

10

11

Page 160: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 12 of 14

agent,arecipient,andamover).InEnglish,itcanusuallyoccurintheDitransitiveconstruction(Shegavehimthepresent)andintheCausedMotionconstruction(Shegavethepresenttohim).Inthe(active)Ditransitiveconstruction,Langackeranalyzestheagentasthetrajectorandtherecipientasthelandmark.Thatis,him,intheexampleathand,istreatedasadirect,ratherthan,indirectobjectbecauseitcanbecomethesubjectinthecorrespondingpassivesentence(Hewasgiventhepresent).Theanalysisofthemovercomplement(theentitytransferred)ishoweveruncertain.Langacker(CG:360)claimsthatwhiletheagentandtherecipientarefocalparticipants,themoverisaparticipantwhichisnotfocusedastrajectororlandmarkandcallsitacentralparticipant.(Thiswouldimplythatthepresentisnotanobjectvis-à-visthedefinitionabove.)IntheCausedMotionconstruction,thelandmarkisidentifiedwiththemoverandthewholeconstructionissaidtohighlightmotion—whichmaybemetaphorical—sincetherecipientisthelandmarkofthepathprepositionto(seeCG:242and393–94).

Insum,itshouldbeclearfromthisshortsynopsishowdifferentthetreatmentofgrammaticalrelationsinCognitiveGrammariscomparedtootherconstructionistapproaches.Langackerisalwaysatpainstoshowthatgrammarismeaningful,henceconstructionsshouldnotbeviewedaspairingsofsemanticsandsyntax,asin(Radical)ConstructionGrammar,butratherpairingsofsemanticsandphonology(broadlyconstrued),sincesyntaxitself,asapartofgrammar,ismeaning.

11.7.Conclusion:WhatLiesAhead?

AttheverybeginningofthischapterIwrotethat“CognitiveGrammaris,quitetrivially,atheoryofgrammar,”butafterthisveryshortjourneythroughLangacker'stheory,whichbynecessityonlytoucheduponsomeofhisanalysesandcompletelyignoredtopicssuchasphonology(see,e.g.,Nesset2008),grounding(see,e.g.,CG:chapter9),anddiscourse(see,e.g.,CG:chapter13)tonameacouple,thereadermaywonderifthatisreallythecase.Ifgrammarisunderstoodasdealingwithschematicsymbolicassemblies,IwoulddaretosaythatCognitiveGrammarisnot(only)atheoryofgrammar.Rather,CognitiveGrammarisprobablybestdescribedasatheoryoflanguage.Infact,Langacker(2007)definesit(inprinciple)asatheoryofénonciationinCulioli's(1990)sense.

Apartfrom(partly)terminologicalissues,thereisonecrucialaspectofCognitiveGrammarthatshoulddeservemoreattentioninthefuture.LangackerinsiststhattheclaimsmadebyCognitiveGrammararepsychologicallyplausible(CG:15).Hence,oneofthechallengesforthistheorywillbetovalidatetheadequacyofsuchclaimsempirically(i.e.,throughpsycholinguisticexperiments).Forexample,atthetimeofwriting,noempiricalevidencehasbeengatheredtosupportthedifferencebetweensequentialandsummaryscanning,aswellasitsrelevancetolanguage,nomatterhowintuitiveand/ortheoreticallyappealingthedifferencemightbe(seeBrocciasandHollmann2007andLangacker2008bforextensivediscussion).

Morespecifically,futureresearchshouldfocusonprocessing(i.e.,whatLangackerreferstoasdynamicity).Forexample,whatarecompositionalpaths(seesection11.5.1)meanttodescribe?Aretheyintendedtocapturelanguageproduction,comprehension,both,orsomethingelse?Thewayinwhichtheyaredetailedseemstosuggestprocessingonthepartofthespeakerbutsomeclarificationinthisareaismuchneeded.Infact,onecouldarguethattheissueofprocessingshouldperhapsbeputonstagenotonlyinCognitiveGrammarbutalsoinotherconstructionistapproaches,whichtendtoconcentrateonlanguagerepresentation/storage(e.g.,languagenetworks)ratherthanlanguageproduction/comprehension.

FurtherasispointedoutbyBrocciasandHollmann(2007),thereisalsotheissueoftheapparentlydualnatureofCognitiveGrammar.Ontheonehand,CognitiveGrammarisausage-basedmodel(seesection11.2).AsLangackerputsit,“[w]hatisessentialisthateverylanguageandeveryconstructionbecharacterizedinitsownterms”(CG:423),whichis,ofcourse,fullycompatiblewith,forexample,RadicalConstructionGrammar.Ontheotherhand,CognitiveGrammarisalsoasemioticmodelinthateverylinguisticexpression,evenapparentlymeaninglesselementssuchas‘grammatical’morphemes(inthetraditionalsense),isregardedasmeaningful.Yetthisispotentiallyproblematic.Ifmuchinlanguageisaccessedautomaticallyandcanundergolossofanalyzability(see,e.g.,Figure11.8aboveandLangacker2009a:26–28),towhatextentcanwemapspecificmeaningsontoindividualforms?Langackeris,ofcourse,consciousofthisissuesinceheexplicitlyacknowledgestheimportanceofthenotionofdecontextualization(seesection11.5.2)andalsopointsout(seenote3)thatschematicmeaningultimatelyresidesinimmanent,recurringprocessingactivityratherthanabstractstructures(hierarchically)separatefromtheirmultifariousinstantiations.However,itmustbedemonstratedempiricallythatfullyschematic

Page 161: The Oxford Handbook of Construction Grammar

Cognitive Grammar

Page 13 of 14

meanings,ratherthanmorespecificschemas,arepsychologically‘real’andthusarenotjustausefuldescriptivetoolresortedtobylinguists.

Havingsaidthat,CognitiveGrammarremains,undeniably,oneofthemostinnovativeandcomprehensivetheoriesofgrammar(orshouldonesaylanguage?),thankstoitsrejectionofformalmachineryandmodularity,itsinsistenceongradienceandmeaningasconceptualization,whichinvolvesbothembodiedcognitionanddisengagedcognition,thatis,ourabilitytodealwithfictive(orvirtual)entities,seeCG:14.2.Ultimately,evenifanalysessuchastheschematiccharacterizationsofsubjectandobjectturnedouttobewrong,wewouldstillhavetofacethequestionofwhysuchandrelatedgrammaticalnotionsseemtoemergewhenstudyinglanguage,thatis,wewillalwaysandinevitablyhavetofocusourattentiononthecognitivefoundationsofgrammar.

Notes:

(1.)UnlikeLangacker(1987,1991,and2008a),theotherbooksmentionedarecollectionsofessayswhichhaveoftenappearedelsewhere.AcompletelistofLangacker'spublicationscanbefoundathttp://idiom.ucsd.edu/~rwl/(lastaccessedJanuary30,2011).

(2.)Itshouldbepointedoutthattheuseoftheterm‘grammar’inLangacker'srecentworkismorerestrictive(andtraditional)thaninthepast.Langacker(1987:57),forexample,assignsthedefinition“astructuredinventoryofconventionallinguisticunits”to‘grammar,’ratherthan‘language,’ashedoesinCG.

(3.)CG:219pointsoutthatschemasshouldnotbethoughtofasseparatefromtheirinstantiationsbut,rather,asimmanentinthem.

(4.)Singlewords(e.g.,flower)andmorphemeswouldcountasdegenerateconstructionsinthatthesymbolicassemblycontainsonlyonesymbolicelement,see,e.g.,Langacker(2009a:17)onmorphemes.

(5.)HereIomitLangacker'sconceptualanalysisofadjuncts,seeCG:204–5andLangacker(2009a:24–25).

(6.)Byconvention,Langackersaysthatneithercomponentisaprofiledeterminantinappositivecases(seeCG:195).

(7.)Langacker(CG:205)stressesthatnotionslikehead,complement,andmodifierare“notthemselvesthebasicunitsofCG[CognitiveGrammar]description.Theyaremoreaccuratelythoughtofasconvenientlabelsforcertainkindsofconfigurationscommonlyobservableatthesemanticpoleofsymbolicassemblies.”Thismeansthatnotallconstructionswillhaveaheadandthatitwillnotalwaysbepossibletodistinguishbetweencomplementsandmodifiers.

(8.)Langackerusestheterm‘theme’torefercollectivelytopatients,experiencers,movers,andzeros.

(9.)Noticethattrajectorandlandmarkareconceptualentitieswhilesubjectandobjectnominalsaresymbolicstructures.

(10.)Sincefocalprominencecanfallonentitiesotherthanaparticipantsuchasasetting,location,path,ormeasurement,Ihaveusedthemoreaccuratedefinition“focalrelationalelement”overthelooserdefinition“focalrelationalparticipant”forthereferentsofsubjectsandobjectsinthetext,seeCG:388n.33.

(11.)Sinceanobjectisdefinedasanominallandmark(see,e.g.,CG:432)andaprepositionalphraselikeaSallyisobviouslynotanominal,wehavetoconcludethatCognitiveGrammarviewstheobjectofthepreposition(Sally)astheindirectobject.

CristianoBrocciasCristianoBrocciasisanAssociateProfessorofEnglishLanguageattheUniversityofGenoa(Italy).Hisresearchfocusesoncognitivetheoriesofgrammar,Englishsyntaxandphonology,bothsynchronicanddiachronic.HispublicationsincludeamonographonEnglishchangeconstructions(TheEnglishChangeNetwork:ForcingChangesintoSchemas,2003)andvariouspapersonresultativeconstructions,simultaneityconstructions,-lyadverbs,andcognitiveapproachestogrammar.

Page 162: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 1 of 15

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0012

RadicalConstructionGrammarWilliamCroftTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterdiscussesthetheoryofRadicalConstructionGrammar(RCG).Thetypologicaldiversityoflanguagesleadstothehypothesisthatallgrammaticalcategoriesarelanguagespecificandconstructionspecificandsoconstructionsarebasicunitsofsyntacticrepresentation.Italsoleadstothehypothesisthatthereisnoformalsyntacticstructureotherthanthepart/wholestructureofconstructionsandthegrammaticalrolesthatoccurinconstructions,andthatconstructionsarelanguagespecific.Thechapteroffersinnovativeapproachestogrammaticalcategories,generalizationsanduniversals,andintegratesbothlanguage-internalandcrosslinguisticvariationintoconstructiongrammar.

Keywords:RadicalConstructionGrammar,grammaticalcategories,distributionalanalysis,typology,language-internalvariation,roles(inconstructions)

12.1.Introduction

RadicalConstructionGrammarisamodelofmorphosyntacticrepresentationthatemergedfrombringingtogethertheresultsoftypologicalresearchandtheissuesaddressedbycontemporarytheoriesofsyntax.TheproductofthismarriageoftypologyandsyntactictheoryisbasicallyavarietyofConstructionGrammar,butonethatisquitedifferentfromsomevarietiesofConstructionGrammardescribedinthishandbook,nottomentionnonconstructionalsyntactictheories.

Mostcontemporarysyntactictheoriestendtofocusontwodimensionsofascientifictheory(Laudan1977;Croft1999).Ascientifictheoryisnotamonolithicentity.Amajordivisionisbetweenthoseelementsofthetheorythatarepartoftheresearchtraditionofthescientistsworkinginthattheory(Laudan1977:78–81),andtheanalysesofspecificphenomena.Aresearchtraditionsimplyrepresents‘thewaywedoscience’andisnotdirectlyfalsifiable:“researchtraditionsareneitherexplanatory,norpredictive,nordirectlytestable”(Laudan1977:81–82).Inlinguistics,formalismandfunctionalismaretwomajorresearchtraditions(or‘approaches,’astheyareusuallycalledinlinguistics).ResearchtraditionsarecharacterizedbywhatLaudancallsmetaphysicalcommitments,thatis,commitmentsonthepartofthescientistaboutwhatthescientificphenomenonis.Examplesofmetaphysicalcommitmentsinlinguisticsincludewhethersyntacticconstraintsareinnate,whethersyntacticstructureisautonomous(self-contained)relativetomeaning,whethergrammarisafixedmentalobjectorevolvesthroughusage,andsoon.

Incontrast,analysesofparticularphenomena,suchasergativity,grammaticalvoice,orcomplexpredicates,aremuchclosertotheempiricalevidence.Often,alternativeanalysesofthesamephenomenonexistinasingleresearchtraditionandonemaybediscardedforanotherbasedonempiricaldata.Althoughanyparticularanalysisiscouchedinthemetaphysicalcommitmentsofaparticularresearchtradition,particularanalysesmaybetranslatedintoanotherresearchtradition(Croft1999:96–97).

Page 163: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 2 of 15

Asecondaspectofaresearchtraditionthatisspecifictolinguistictheoriesisrepresentationalcommitments(Croft1999:91–96).Representationalcommitmentsarethestructuresandnotationsusedtorepresentthegrammaticalanalysesofparticularsentencesorutterances.Thesystemsofrepresentationalcommitmentsareessentiallyformallanguagesforthedescriptionofgrammaticalstructuresofsentences;theyconstitutewhatHaspelmath(2010b:343)callsaframework.Thesearethemostsalientpropertiesofsyntactictheories,especiallyforstudentslearningthetheory.Onecharacteristicfeatureofformalsyntacticrepresentationsishowmuchtheyvaryacrosssyntactictheoriesandchangeovertime.Thesyntacticstructurespositedby,forinstance,theMinimalistProgram(MP),LexicalFunctionalGrammar(LFG),Head-drivenPhraseStructureGrammar(HPSG),andRoleandReferenceGrammar(RRG)forthesamesentenceareverydifferent.Thesefourmodelsareonlyasubsetofthemodelsthathavebeenproposed.Likewise,thesyntacticstructuresinthevariousversionsofChomskyangenerativegrammaroverthepastfiftyyears,andintheversionsoftheotheraforementionedmodelsreferredtoovertheirthirty-oddyearlifetimes,alsodifferconsiderably.

RadicalConstructionGrammar'scritiqueofothersyntacticmodelsfocusesonanother,muchlessdiscussedpartoftheresearchtraditionsinsyntax:themethodsofargumentationforasyntacticanalysis.ThisispartofwhatLaudancallsthemethodologicalcommitmentsofascientistinaparticularresearchtradition.Methodsofsyntacticargumentationarefarlessdiscussedintheliteraturecomparingformalismandfunctionalism,althoughothermethodologicalcommitments,suchastheuseofinventedexamplesvs.naturallyoccurringutterances,haveattractedmoreattention(e.g.,Sinclair1991;Fillmore1992;Stubbs1996).

ThecritiqueofsyntacticargumentationinRadicalConstructionGrammar,combinedwiththeempiricalresultsoftypologicalresearch,resultsintheeliminationofalmostallrepresentationalcommitments.RadicalConstructionGrammarisessentiallyaframework-freegrammaticaltheory(Haspelmath2010b),perhapstheonlyoneexistingatpresent.Thisisbecausethegrammaticalcategoriesthatmakeupthebuildingblocksoftherepresentationallanguagesofothersyntactictheoriesarearguedtobelanguage-specificandconstruction-specificandhencecannotformthebasisforasyntacticrepresentationlanguage(sections12.2−12.3).Theonlyuniversalrepresentationalcommitmentforformalsyntacticstructureisthepart–wholestructureofconstructionsandthegrammaticalrolesthatoccurinconstructions.

Instead,thesubstantiveuniversalsofgrammarinRadicalConstructionGrammararefoundinthemappingbetweenlinguisticformandlinguisticmeaninginconstructions.AsinmostothervarietiesofConstructionGrammardescribedinthishandbook,constructionsareapairingofsyntacticstructureswithsemanticstructures;thatis,theyaresignsorsymbolicunits.Themaindifferencebetweensigns/symbolsinConstructionGrammarsandsigns/symbolsinnonconstructionaltheoriesisthatsigns/symbolsinConstructionGrammarsmaybecomplexformalstructures,not(just)atomicunitssuchaswordsormorphemes.Syntacticuniversalsarethereforeuniversalsofthemappingbetweensuchcomplexsyntacticstructuresandthecorrespondingsemanticstructuresacrosslanguages.Thismappingiscomplex,partlyarbitrary,andprobabilistic(asacrosslinguisticgeneralization).

Thevariabilitythatisobservedacrosslanguagesrequiresinnovationsintherepresentationofsyntacticcategoriesandsyntacticconstructions.Theinnovationsinclude:acontinuousorsemicontinuousconceptualspaceforrepresentingcategorystructure;recastingtheform-meaningmappingasafrequencydistributionofformsacrossthatconceptualspace;abandoningabstractsyntacticrelations;andasyntacticspacewhosedimensionsaredefinedbyamodeloftheverbalizationofexperience.Developingthesenovelrepresentationsisanevolvingprocess.

12.2.DistributionalAnalysis,Constructions,andtheBuildingBlockModelofGrammar

Thebasicmethodforsyntacticargumentationisdistributionalanalysis.Thename‘distributionalanalysis’wasusedbytheAmericanstructuraliststodescribethemethod,butitgoesunderdifferentnamesincontemporarytheories:applyingmorphosyntactic(grammatical)tests,satisfyinggrammaticalcriteria,possessinggrammaticalproperties,displayinggrammaticalbehavior,presentingaseriesofsyntacticarguments,andsoon.Allofthesetermsdescribeessentiallythesamemethod.

Indistributionalanalysis,ahypothesisabouttheproperanalysisofasyntacticunitisformedbyexaminingtheoccurrenceornonoccurrenceofthatunitinarangeofsyntacticstructures.ArelativelysimpleexampleistheanalysisofcertainargumentphrasesinEnglish(moreprecisely,certainphrasesfunctioningasargumentsofa

Page 164: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 3 of 15

verbwithwhichtheyarecombined).CertainargumentphrasesofverbsoccurpostverballyandwithoutaprepositionintheEnglishActiveVoiceconstruction,asin(1):

(1)a.Thewindknockedthepottedplantsover.b.Thepolicetappedmyphone.

Thisisdescribedasthedistribution(orapropertyorbehavior)ofthephrasesinquestioninaparticularposition,orrole,intheActiveconstruction.Alternatively,occurrenceofthephrasesinquestionintheparticularroleiscalledatestorcriterionforthephrasetobelongtoagrammaticalcategory.Onthebasisofthisdistribution,wemayhypothesizethatthephrasesinboldfaceformasinglegrammaticalcategory,directobject.

Twoobservationscanbemadeaboutdistributionalanalysis.First,thedistributionofasyntacticunitisactuallydefinedoverasetofconstructions.The‘tests,’‘criteria,’‘properties,’or‘behavior’areconstructions,whicharesyntacticstructuresintheirownright.Constructiongrammarsofallvarietiesexplicitlyrecognizeconstructions(seeanyoftheothercontributionstothisvolume).Nonconstructionalgrammarsdonotexplicitlyrecognizetheexistenceofconstructions(indeed,sometheoriesdenytheirexistence;e.g.,Chomsky1981:121,1993:4);yettheirmethodofsyntacticargumentationpresupposestheexistenceandidentifiabilityofconstructions.RadicalConstructionGrammar,likeotherConstructionGrammars,explicitlypositstheexistenceofconstructions.

Thedistributionalmethoddefinesamappingbetweenonesetofsyntacticstructures—thewords,morphemes,phrases,andsooninquestion—andanothersetofsyntacticstructures,namelytheconstructionsusedastests,criteria,andsoon(moreprecisely,therelevantgrammaticalroleintheconstructionsused).Forexample,example(1)illustratesamappingbetweenthesetofphrasesinboldfaceandtheActiveVoiceconstruction,ormorepreciselythepostverbalNProleintheActiveVoiceconstruction.

Distributionalanalysisasoriginallyconceivedisapatternofrelationsamongformalstructures.However,distributionalanalysismaybe,andis,extendedtoincludemeaning.Forexample,inexaminingthedistributionofelementsthatfunctionintheroleofcomplementsoftheEnglishauxiliarymust,onemustdistinguishbetweenepistemicanddeonticmeanings.Epistemicmeaningsallowcomplementsoftheformhave+PastParticiplethatdenotepastevents(Hemusthavetakenhisjacket),whereasdeonticmeaningsonlyallowcomplementsintheBareInfinitiveformthatdenotepotentialfutureevents(Youmustwearajackettothedinner).

Thesecondobservationisthatthedistributionalmethodisgenerallyusedtoformhypothesesaboutthesyntacticunitthatoccurs/doesnotoccurintherelevantrolesoftheconstructionsusedas“tests,criteriaandsoon.”Again,theidentityoftheconstructionsusedas“tests,criteriaandsoon”istakenforgranted,andthedistributionalmethodistakentorevealsomethingabouttheidentityanddistinctnessofthesyntacticunits,ratherthananythingabouttheconstructionsinwhichtheyoccur/donotoccur.Forexample,thedistributionpatternin(1)istakentodefineacategoryDirectObject,nottodescribeafactabouttheTransitiveconstruction.

Thissecondobservationisbasedonametaphysicalandrepresentationalcommitmenttowhatmaybecalledthebuildingblockmodelofgrammar.Grammarisseenasbeingmadeupofminimalunits(wordsormorphemes)belongingtogrammaticalcategories,andconstructionsaredefinedasstructuredcombinationsoftheseunits.Thepurposeofthedistributionalmethod,therefore,istoidentifythegrammaticalcategoriesthatarethebuildingblocks,andtheunitsthatbelongtothosecategories.Itisalsoassumedthatthegrammaticalcategoriesarepartofaspeaker'smentalrepresentationofhergrammaticalknowledge.Thekaleidoscopeof‘grammaticaltheories’(representationalcommitments,frameworks)reflectsdifferentandevolvingideasofwhatthebuildingblocksofgrammarareandhowtheyareputtogether.

12.3.EmpiricalVariationandRadicalConstructionGrammar

Whenthedistributionalmethodandthebuildingblockmodelmeetempiricalfact,however,aproblemarises:distributionalpatternsdonotmatch.Forexample,wemayalsocomparethedistributionofthephrasesin(1)inaparticularroleinanothersyntacticconstruction,theEnglishPassiveVoiceconstruction,asin(2):

(2)a.Thepottedplantswereknockedoverbythewind.b.Myphonewastappedbythepolice.

Page 165: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 4 of 15

ItappearsthatthePassiveSubjectcategorycorrespondsexactlywiththeActiveDirectObjectcategory,buttheexamplesin(3)–(4)demonstratesthatthisisinfactnotcorrect:

(3)a.Theroadextendstenmilesintothemountains.b.2010sawthefirsthungparliamentinBritainforoverthirtyyears.

(4)a.  *Tenmilesisextendedbytheroadintothemountains.b.*ThefirsthungparliamentinBritainforoverthirtyyearswasseenby2010.

Thedistribution(property,behavior,etc.)oftheboldfacephrasesdiffersinthePassiveconstruction:theexamplesin(3)–-(4)areungrammaticalbyintrospectivejudgment.Hence,thereisadifferenceindistributionbetweenthepottedplantsandmyphonein(1)–(2)andtenmilesandthefirsthungparliament…in(3)–(4).

Acrosslanguagesaswell,constructionsappeartodefinedifferentcategories(Dryer1997):Absolutive-ErgativedoesnotmatchSubject-Object;theVerbcategorycanbeeithermoreinclusive(including‘Adjectives’aswell)orlessinclusive(whereasmallsetof‘Verbs’combineswithotherelementstoproducethetranslationequivalentsofEnglish‘Verbs’).

Inotherwordsthefactsoflanguagesrevealaconflictinthedistributionalmethod,ontheonehand,andthebuildingblockassumption,ontheother.Thewidelyacceptedresponsetotheempiricalfactsistoselectcertainconstructionsascriterialindefiningagrammaticalcategory.Forexample,thePassiveisgenerallytakenascriterialindefiningtheActiveDirectObjectcategory,anditisconcludedthattenmilesandthefirsthungparliament…arenotDirectObjects.Incrosslinguisticcomparison,ineachlanguagebeingcomparedaconstructionisused—notnecessarilythesameconstructionsacrosslanguages—toidentifyacategoryandtoidentifythatcategoryasthesameas,say,EnglishSubjectorEnglishVerb.Iftheconstructionchoseninalanguagedoesnotyieldthedesiredcategory,thenanotherconstructionmaybeusedinstead(foracritiqueofthisstrategyincrosslinguisticcomparison,seeCroft2009a).Anotherapproachtoconflictsindistributionistoconsiderthecategoryinquestiontobedefinedbyamajorityoftheconstructionsusedinthedistributionalanalysis.

Thisstrategy—arbitrarilyselectingoneconstructionorasubsetofconstructionsascriterial—canbetracedbacktotheAmericanstructuralists;itiscalledmethodologicalopportunisminRadicalConstructionGrammar.Methodologicalopportunismineffectabandonsthedistributionalmethod,sinceitisselectiveintheemploymentofdistributionalfacts.Forexample,iftheDirectObjectcategoryisdefinedbyoccurrenceinthePassiveconstruction,thenthereisnoexplanationastowhytenmilesandthefirsthungparliament…in(3)havethesamedistributionasthepottedplantsandmyphonein(1).Thepurposeofmethodologicalopportunismisessentiallytopreservethebuildingblockontologicalassumption:ifthereisasmallfinitesetofbuildingblocks,categorieshavetobeidentifiedasthesamefromoneconstructiontothenext(e.g.,ActiveDirectObject=PassiveSubject).

However,mismatchesindistributionalpatternsarepervasiveinallhumanlanguages,afactthatisevidentinanycrosslinguisticcomparisonandinanysignificantlydetailedanalysisofasinglelanguage.Muchoflinguisticanalysisaimstodevelophypothesesthatwillaccountforthemismatchesindistributionpatterns.However,anyanalysiscommittedtoaparticularsetofbuildingblocksisforcedtodeviseanessentiallyadhocexplanationforthedistributionpatternsthatdonotmatchthoseoftheconstructionsusedtodefinethebuildingblocks(seeCroft2005b,2009aforcritiquesofcrosslinguisticanalysesandCroft2007a,2010aforcritiquesofasingle-languageanalysis).Dryer(1997)essentiallyarguesthatthebuildingblocksmustbedifferentforeverylanguage(seealsoHaspelmath2007,2010a;Cristofaro2009;Barðdal2011a).Inaddition,within-languagevariationindistributionpatternsindicatesthatthe‘buildingblocks’mustactuallybedifferentforeveryconstructioninasinglelanguage.Inotherwords,ifthemethodologicalcommitmenttodistributionalanalysisisnotabandoned,thenonemustabandonthemetaphysicalcommitmenttothebuildingblockmodelofgrammar.ThisisthebasicstartingpointofRadicalConstructionGrammar.

InRadicalConstructionGrammar,therearenogrammaticalcategoriesindependentofconstructions,sinceeachconstructiondefinesitsowndistribution,whichmaybe(andusuallyis)distinctfromthedistributionofotherconstructionscontainingthesamewordsorphrases.Thisargumenthasbeenmadeforawidevarietyoffundamentalgrammaticalcategories,includingthemajorpartsofspeech(Croft2001:chapter2;vanCanegem-Ardijns2006;BrocciasandHollmann2007;Hollmann,forthcoming);thecoregrammaticalrelations(VanValinandLaPolla1997:chapter6;Croft2001:chapter4;Pedersen2005;Barðdal2006);heads(Croft2001:chapter7);

Page 166: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 5 of 15

transitivity(LaPollaetal.2011:487);finiteness(Cristofaro2003,2007);andevenconstituency(Croft2001:chapter5),embeddedness(Cristofaro,forthcoming),andmorphosyntacticwordhood(Haspelmath2011).Forexample,inthesentencesin(1)–(4),themismatchesindistributiontellussomethingabouttherangeofActiveandPassiveVoiceconstructions,notthecategoriesoftherolefillers;inparticular,thePassiveismorerestrictiveinwhatcanserveasPassiveSubjectthantheActiveisregardingtheActiveDirectObject.(Followingtypologicalpractice,thenamesoflanguage-specificgrammaticalcategoriesandconstructionsarecapitalized.)Inotherwords,RadicalConstructionGrammarismethodologicallyfullycommittedtodistributionalanalysisinallitscomplexity.(Thereareotherimportantmethodologicalcommitmentsthatcannotbediscussedhereforreasonsofspace;seeCroft2010b.)

Asaresult,constructionsarethebasicunitsofgrammarinRadicalConstructionGrammar(thefirst‘radical’hypothesis).RadicalConstructionGrammarisanonreductionisttheoryinthesensethatcomplexentitiesarebasic,asinthegestalttheoryofperception.Afrequentlyaskedquestionis:Howareconstructionsidentified?Actually,nonconstructionaltheories,andconstructionaltheoriesthatremaincommittedtothebuildingblockmodel,alsotakeconstructionsforgrantedbecausetheyarepresupposedindistributionalanalysis(section12.2).Constructionscanbedescribedintermsofpropertiesofformandespeciallymeaning.Constructionsformcategories,andlikeothercategoriestheymayhaveinternalstructure(e.g.,aprototypeandextensions)andtheboundariesbetweenconstructionsmaybedifficulttodefine(moreprecisely,aresubjecttoconstrual;CroftandCruse2004:93–95).

AnotherexampleoftheshifttoconstructionalprimacyinRadicalConstructionGrammaristhestatusofthedebatebetweenlexicalrulesandconstructionsintheanalysisofargumentstructurealternations.Thedebateistypicallyframedintermsofverbalpolysemyderivedbyrules(lexicalrules)vs.underspecifiedverbmeaningsthatarefilledoutwhencombinedwithanargumentstructureconstruction.Croft(2003a,2012:chapter9)arguesthatthereissufficientidiosyncrasyinverb-argumentstructurecombinationsthatanunderspecifiedverbmeaningisempiricallyinadequate,butonemustpositverb-class-specificandevenverb-specificargumentstructureconstructionsbecausetheverbhasitsmeaningonlyinparticularargumentstructureconstructions.ThesameargumenthasbeenmadebyBoas(2003)fortheEnglishresultativeconstruction,byCristofaro(2008)forpredicate+complementconstructionsinAncientGreek,byIwata(2008)fortheEnglishlocativealternation,andbyBarðdal(2008;Barðdaletal.2011a)forditransitivesandotherargumentstructureconstructionsinWestScandinavianlanguages.

Thenonreductionistapproachto(morpho)syntaxmayalsobeextendedtophonology;atentativeproposalisfoundinVihmanandCroft(2007).Thereisampleevidencethatphonologicalpatternsdonotjustifyabuildingblockmodelforphonology:forexample,thecategoriesofconsonantandvoweldifferdependingonphonotacticorprosodicposition(VihmanandCroft2007),andthecriteriaforphonologicalwordhooddonotmatch(Bickel,Hildebrandt,andSchiering2009;Schiering,Hildebrandt,andBickel2010).InaRadicalTemplaticPhonology,thebasiccomplexunitisaphonologicaltemplate.Thetemplatesarethemselvesgeneralizationsoverphonologicalwordstructures(ormoreconcretely,thearticulatoryscoreandacousticpatternofthewordasawhole).VihmanandCroftprovideextensiveevidencethatphonologicalacquisitionisguidedbytheformationofsuchtemplatesandarguethatadultphonologyshouldalsoberecastintemplaticform.

12.4.GrammaticalCategoriesandGrammaticalGeneralizationsinRadicalConstructionGrammar:NewModels

RadicalConstructionGrammarrejectsgrammaticalcategoriesindependentoftheconstructionsthatdefinethem(i.e.,itrejectsthebuildingblockmodel).OneconsequenceofthisrejectionisthatRadicalConstructionGrammardoesnotconsistofanelaborateframeworkofallthetheory-specificbuildingblocksandrulesofcombinationofthoseblocks.InRadicalConstructionGrammar,theonlyrepresentationalcommitmentsarethecommondenominatorofallConstructionGrammars:aconventionalsymbolicunit,consistingofaformcontainingoneormoreformalmorphosyntacticelements,pairedwithameaningthatcontainsoneormoresemanticcomponents.

Butthemetaphysicalandrepresentationalcommitmentofthebuildingblockmodelservesafundamentalpracticalandtheoreticalpurpose.Thebuildingblockmodelservesthepracticalpurposeofprovidingaframeworkforcapturinggrammaticalpatterns,thatis,grammaticalgeneralizations,forlinguisticdescription,aswellasforconstructinglinguistichypotheses.Andthebuildingblockmodelprovidesasimplemodelofthegrammatical

Page 167: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 6 of 15

knowledgeofaspeaker,compatibleforexamplewiththecomputationalmodelofthehumanmind.Atheorythatrejectsthebuildingblockmodelmustprovidealternativewaystorepresentgrammaticalpatternsandalternativehypothesesfortherepresentationofgrammaticalknowledge.

ARadicalConstructionGrammardescriptionofthegrammarofaparticularlanguage,forexample,shouldbeentirelyconstruction-based.Sectionsdevotedtogrammaticalcategories—acommonwaytoorganizegrammaticaldescriptions—shouldinsteadbedevotedtotheconstructionsthatareusedtodefinethosecategories:sectionsonpartsofspeechshouldbereplacedbysectionsonconstructionsexpressingpropositionalacts(referringexpressions,predicationconstructions,modifying/attributiveconstructions),sectionsongrammaticalrelationssuchasDirectObjectshouldbereplacedbysectionsonargumentstructureconstructions(includingvoiceconstructions),sectionsondifferenttypesofmodifiers(suchasAdjectives,Numerals,etc.)shouldbereplacedbysectionsonattributiveconstructions,andsoon(seeCroft,inprep.).Generalizationsdefinedintermsofgrammaticalcategoriesarereplacedbydistributionalanalysisoftheunitswhosedistributionisdeterminedbytheconstructionsinquestion.

Ahypothesizedgrammaticalcategoryisawayofrepresentingagrammaticalgeneralization.Forexample,thecategoryDirectObjectinEnglishisintendedtorepresentageneralizationoftheoccurrenceofaclassofphrasesacrossdifferenttypesofconstructions(Active,Passive,Information[WH]Question,etc.).InmanyvarietiesofConstructionGrammar(cf.,e.g.,alltheotherchaptersinPartII.Constructionistapproaches),grammaticalgeneralizationsarealsocapturedbytaxonomicrelationsamongconstructionsandtheinheritancerelation.Forexample,theTransitiveVerbconstructioncapturesgrammaticalgeneralizationsforalargeclassofEnglishwords,andamoregeneralVerbalPredicateconstructionsubsumingtheTransitive,Intransitive,andotherconstructionscapturesgrammaticalgeneralizationsforanevenlargerclassofEnglishwords.

However,neithergrammaticalcategoriesnorconstructionaltaxonomiesaregoodatcapturingthefullrangeofgrammaticalgeneralizations.First,asarguedinsection12.3,differentconstructionsonthewholedonotdescribethesamegrammaticalcategory,thatis,thedistributionsdefinedbytherelevantrolesintheconstructionsarenotthesame.Thesamecanbeobservedinthetaxonomichierarchyofconstructions.Forexample,tickleisaTransitiveVerb,whiledieisanIntransitiveVerb;thatis,[SbjtickleObj]and[Sbjdie]areinstancesoftheTransitive[SbjTrVerbObj]andIntransitive[SbjIntrVerb]constructions,respectively.ButbreakcanoccurineithertheTransitiveorIntransitiveConstruction,andweighoccursintheTransitive(Active)constructionbutnotthePassiveconstruction.WhyaretheseallcalledVerbs,despitetheirdistributiondifferences?Becausetheyallalsooccurinanotherconstruction,theMorphologicalVerbconstruction,definedbyoccurrenceintheTense-Agreementinflectionalconstruction(i.e.,thePresent-Ø/-salternationandthePast-edmorpheme).

Theserelationshipscouldberepresentedintermsoftaxonomicrelationsbetweenconstructionsandpartsofconstructions,asisdoneinFigure12.1,aslongasweallowforalternativeinstantiations(e.g.,forbreak).Thisisthecommonlyusedmultiple-inheritancemodelofConstructionGrammar.Butthereareothermethodsforrepresentingtheserelationships.Forexample,Figure12.2isageometricalrepresentationofthedistributionalpatterns.Bybee(1985)proposesanotherwayofrepresentinggeneralizations,illustratedinFigure12.3.Sherejectsthenotionofanabstractmorphemesuchas[Past],andinsteadarguesthatageneralizationsuchas[Past]emergesfromanetworkofsimilarityrelationsinform(finalphoneme/d/)andmeaning(Past)acrossindividualwords.

Clicktoviewlarger

Figure12.1. Taxonomicrepresentationofgrammaticalgeneralizations

Page 168: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 7 of 15

Clicktoviewlarger

Figure12.2. Geometric(Venndiagram)representationofgrammaticalgeneralizations

Figure12.3. Networkrepresentationofgrammaticalgeneralizations

Allofthesearedifferentmeansforcapturinggrammaticalgeneralizationsandrepresentingtheminanaccessiblewaytothelinguist.Somearemoreadequatethanothers.Inparticular,thenetworkandgeometricmodelsareabletocapturegrammaticalpatternsthataredifficultorimpossibletocaptureusingtaxonomiesorconstruction-independentgrammaticalcategories.Thesemodelsarebrieflydescribed,andthenwereturntothesecondgoal,namely,discussinghowgrammaticalknowledgeisrepresented.

Thenetworkmodelisusedalsoinmorphosyntacticanalysisintypology,whereitisknownasthesemanticmapmodel.Thesemanticmapmodelisahighlyflexiblemethodforcapturingcrosslinguisticgeneralizationswhenthereisahighdegreeofvariabilityingrammaticalcategoriesacrosslanguages.Totakeaverysimpleexample,inergativelanguages,theIntransitiveSubjectrole(usuallylabeledS)isgrammaticallygroupedwiththeTransitiveObjectrole(P)incasemarkingand/orindexation(agreement)constructions(asabsolutive),whiletheTransitiveSubject(A)roleiscategorizeddistinctly(asergative).TheergativepatternAvs.S+Pcontrastswiththenominative-accusativepatternA+Svs.P.ButAandParenotgroupedtogethervs.S;theyareonlygroupedtogetherifSisalsoincluded(i.e.,A+S+P;theneutralalignment).ThispatternofvariationandlimitsonvariationincoreargumentcategoriescanberepresentedbytheconceptualspaceinFigure12.4.Theconceptualspaceisanetwork(graph)structure,representedbythesolidboldfacelinksinFigure12.4.A,S,andPareunitsintheconceptualspace;moreprecisely,theyaresemanticcategories(categoriesofparticipantrolesinevents).Languagesmayvaryastotheirgrammaticalcategoriesforcoreargumentroles.Forexample,theergative-absolutivegroupingisrepresentedbythedotted-linesemanticmapsinFigure12.4(withlabelsinitalics),whilethenominative-accusativegroupingisrepresentedbythedashed-linesemanticmaps(withlabelsinroman).However,thetotalpossiblecrosslinguisticvariationinalignmentsystemsislimitedbythenetworkstructureofFigure12.4:inparticular,AandPcannotbegroupedtogetherwithoutalsoincludingS.Otherpossiblesystemsaretheneutralsystem,withnodifferentmarkingforanyofA,S,orP;andthetripartitesystemwithdifferentcodingforeachofA,S,andP.ThissemanticmapwaspresentedbyComrie(1978:332)andrepresentsoneoftheearliestusesofthesemanticmapmodelintypology.

Page 169: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 8 of 15

Figure12.4. Conceptualspaceforcoreparticipantroles(A,S,P),andsemanticmapsforergative-absolutiveandnominative-accusativealignmentsystems

Thesemanticmapmodelcapturesgeneralizationsbasedonsimilarity:ifalanguage-specificconstructiongroupstogethertwofunctions,forexample,AandS,orSandP,thenspeakersconceivesomesimilaritybetweenthosetwofunctions.Theconceptualspacerepresentssimilarityrelationsaslinksofvaryingdegreesofdirectnessbetweenthenodesinthenetwork:AissimilartoS,andSissimilartoP,butAissimilartoPonlyatalevelofgeneralitythatincludesS.Thesemanticmapmodelreplacesgrammaticalcategoriesbecausegrammaticalcategories,likecategoriesingeneral,arebasedonsimilarityrelationshipsamongtheirmembers.Traditionalgrammaticalcategoriesarelikeclassicalnecessary-and-sufficientconditioncategories,however,whichdonotcapturedegreesofsimilarityamongmembers,whichisnecessaryfordescribingpsychologicalcategories(Murphy2002).

Thesemanticmapmodelhasbeenwidelyusedintypology.Thesemanticmapmodelhasbeenappliedtogrammaticalcategoriessuchastheperfectandtheevidential(Anderson1982,1986),themiddlevoice(Croftetal.1987;Kemmer1993),indefinitepronouns(Haspelmath1997),intransitivepredicationtypes(Stassen1997),modality(vanderAuweraandPlungian1998),anddepictiveadjectivals(vanderAuweraandMalchukov2005),amongmanyothercategories.GeneralintroductionstothesemanticmapmodelcanbefoundinCroft(2003b)andHaspelmath(2003).Croft(2001,2003b)arguesthattheconceptualspace—thenetworkofsimilarityrelationsamongtheunitsrepresentedinthenodes—isuniversal,whilegrammaticalcategoriesarelanguage-specific.Botharepartoftherepresentationofaspeaker'sgrammaticalknowledge(butseeHaspelmath2004andCristofaro2009foranominalistandagnosticview,respectively,onthestatusoflanguageuniversalswithrespecttoaspeaker'sgrammaticalknowledge).

However,thesemanticmapmodelisdifficulttoscaleuptodescribecrosslinguisticvariation,forgrammaticalconstructionsdescribedoveramuchlargernumberofsituationtypes,forexample,the71spatialpicturesusedtoelicitspatialconstructionsinLevinsonetal.(2003)orthe250questionnairecontextsusedtoelicittense-aspectconstructionsinDahl(1985).CroftandPoole(2008)useageometrictechnique,multidimensionalscaling,torepresentthesimilarityrelationshipsamongthesituationtypesinthesemorecomplexdomains.Theprincipleisthesame:constructionsthatcategorizetwosituationtypesindicatethatthosesituationtypesareconstruedassimilartosomedegree.Multidimensionalscalingallowsustosimultaneouslyrepresentvastnumbersofsimilarityrelationsamongmanysituationtypesinageometricmodel(seeCroftandPoole2008;Croft2010c).

Oneempiricalresultfromsemanticmapandmultidimensionalscalingstudiesisthatthecrosslinguisticvariationobservedimpliesthatspeakersaresensitivetosubtlesemanticdifferencesbetweenalargenumberoffine-grainedsituationtypes:smalldifferencesindistanceandangleinthegeometricrepresentationsarelinguisticallysignificant.Theimplicationforgrammaticalknowledgeisthatspeakersknowveryspecific,richlydetailedsituationtypesarrangedinauniversalconceptualspace;thisformsthesemanticrepresentationonwhichthelanguage-specificgrammaticalcategories(rolesinconstructions)arebased.

Thissemanticrepresentationbasedoncrosslinguisticevidenceisconfirmedbyananalysisofthevariationin

Page 170: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 9 of 15

verbalizationinasinglelanguage(Croft2010d).Speakersvaryintheirverbalizationofthesameexperience(inthiscase,thePearfilm;Chafe1980)undersimilarcircumstances.Insemanticallysimilarscenes,speakersusedthesamerangeofverbalizations(e.g.,specificverbs,ordefinitearticlevs.possessivepronouns).However,thefrequencydistributionofthevariantsdependedonsubtlesemanticdifferencesbetweenthescenes;forexample,thePossessivePronounwaspreferredovertheDefiniteArticledependingonthelikelihoodthattheobjectwouldbepossessed(inalienability).Thesemanticfactorsthatinfluencedthefrequencydistributionofvariantverbalizationsarethesamesemanticfactorsthatinfluencecrosslinguisticvariationingrammaticalexpression(seealsoCroft2001:107;Hollmann2007).

Thesecrosslinguisticandlanguage-internalpatternsfitwithausage-based,exemplarmodelofthesyntax-semanticsrelation.Intheusage-basedmodel(Bybee1985,2010,thisvolume;Langacker1988b;BarlowandKemmer2000;Croft2000),speakersstoretokensoflanguageuseasexemplars.Inthecaseofconstructions,speakersstoretokensofconstructionsfromutterancesandtheirmeaningsinthecontextofuse.Grammaticalknowledgeisformedasgeneralizationsoverthepatternsofuse.Muchresearchinusage-basedlanguageacquisitiondemonstratesthatacquisitionofgrammaticalknowledgebeginswithexemplarsofconstructionsoverwhichgeneralizationsaregraduallyformed(seeTomasello2003,2006,andreferencescitedtherein;andalsoDiessel,thisvolume).DeucharandVihman(2005)applyRadicalConstructionGrammartomixedutterancesinbilingualacquisition.Theyarguethatchildren'suseofwordsinveryunadultlikeways,includinginalanguageinappropriatetotheconversationalcontext,suchasEnglish-Spanishhatdos‘wearingonehatandlookingforanother,’resultsfromchildrendevelopingconstructionsholistically,withwordclassesdefinedbyworduseinthechildren'sconstructions.

RadicalConstructionGrammarleadsultimatelytoanexemplarsemanticsmodelofthesyntax-semanticsmapping.Associatedwithspecificsituationtypesisthefrequencydistributionofthevariousconstructionsusedforthatsituationtype(seeCroft2010d).Thesituationtypes(thefine-grainedmeanings)areorganizedinamultidimensionalconceptualspace.Theformalconstructiontypeshaveafrequencydistributionoverthatconceptualspace.Thesetwostructures(theconceptualspaceofsituationtypesandthefrequencydistributionoftheformalconstructiontypesoverthatspace)arepartoftheorganizationofgrammaticalknowledge.

Anotherexampleofusage-basedorganizationofgrammaticalknowledgeisthelevelofschematicityofconstructions.Structuralistandgenerativegrammarians,andsomeConstructionGrammarians,posittheexistenceofthemaximallyschematicconstructionanalyzableinthedata,orevenonlyinthemorphosyntactic(notsemantic)patternsinthedata.Theusage-basedmodelpurportstobeamodeloftheactualgrammaticalknowledgeofaspeaker,notanabstractmodeltogeneratesentences.Intheusage-basedmodel,schematicconstructionsarepositedonlywhenjustifiedbysufficientlyhightypefrequencyanddegreeofmorphosyntacticandsemanticsimilarity;morespecificconstructionsmayalsobepositediftheyareofsufficientlyhightokenfrequency(seeBybeeandThompson2000;CroftandCruse2004:chapter11;Barðdaletal.2011).

Stillanotherwayinwhichgrammaticalknowledgeisorganizedisintermsofthesyntacticspaceofmorphosyntactictypes,andintheanalysisofconstructiontypesbasedontheverbalizationprocess;thesearebothdescribedinsection12.6.Insum,grammaticalknowledgeisorganizedinmultipleintersectingdimensionsofmorphosyntacticstructure,conceptualspace,andtheirsymbolicpairing.

12.5.TheInternalFormalStructureofConstructions

Theempiricalcrosslinguisticandlanguage-internalevidencepresentedintheprecedingsectionssupportstheabsenceofrepresentationalcommitmentstospecificuniversalgrammaticalcategoriessuchasVerborDirectObjectinRadicalConstructionGrammar.Thegeneralizationsacrossconstructionswithinandacrosslanguagesthatuniversalgrammaticalcategoriesareintendedtocaptureareinfacttoocomplexforthemtocapture,somoresophisticatedmeansofinferringgrammaticalpatternsmustbeemployed,suchasthesemanticmapmodelandmultidimensionalscaling.Thesegrammaticalpatternsareformedoverpopulationsofstoredexemplarsoftokensofconstructionalformandmeaninginutterances.

Hence,theparticularelementsinagrammaticalconstructionarelanguage-specific,construction-specificinferencesfromlanguageuseinRadicalConstructionGrammar.Whatabouttheinternalformalstructureofconstructions?AllConstructionGrammarshaveconstructionsascomplexmorphosyntacticunits,madeupof

Page 171: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 10 of 15

multipleelements,forexample,theEnglishAgentivePassiveconstructionillustratedin(2)consistsofthePassiveSubject,anAuxiliaryverb(be,get),theVerbinPastParticipleform,theprepositionbyandtheAgentphrase.Theelementsofaconstructionaredefinedbytherolethateachhasintheconstructionasawhole;thatis,thereisapart–wholerelationshipbetweenelementsandtheconstruction.Grammaticaltheoriesalsopositsyntacticrelationsbetweenelementsofaconstruction.Forexample,thePassiveSubjectisheldtobeinaSubjectrelationtotheVerbortheAuxiliary,theAgentphraseisheldtobegovernedbytheprepositionby,theprepositionbyplusitsgovernedphraseareheldtobeinanObliquerelationtotheVerb,andtheVerbisheldtobeinaComplementrelationtotheAuxiliary.

InRadicalConstructionGrammar,nosuchsyntacticrelationsarepositedtoexist.Syntacticrelationssufferfromthesameempiricalandmethodologicalproblemsasuniversalsyntacticorgrammaticalcategories.Thepatternsthatexistareexplainableintermsoftheexistenceofsemanticrelations,thatis,relationsbetweenthecomponentsofthesemanticstructureofaconstruction,whoseexistenceRadicalConstructionGrammarassumes,alongwithotherConstructionGrammars.

Syntacticrelationsaretypicallymodeledaseitherconstituentrelations(ingenerativegrammaranditsoffshoots),dependencyrelations(insomeEuropeanmodelssuchasWordGrammar),orboth(inLexicalFunctionalGrammar).Theexistenceofconstituentsordependenciesisarguedforbyusingdistributionalanalysis.Aswithargumentsforsyntacticcategories,distributionalanalysisgivesconflictingresultsforconstituencyanddependency,somethodologicalopportunismisemployedinordertopositasingleconstituenttreeordependencynetworkforaparticularsentence.Aswithcategories,onecanarguethatthegroupingofelementsisconstruction-specific,motivatedbythefunctionofthatconstruction,andtheymaynotbethesamefromoneconstructiontothenext(Croft2001:chapter5).

Syntacticrelationscanbedividedintotwobroadtypes,collocationalrelationsandcodedrelations.Collocationalrelationsarethosethatlinktheconventionalcombinationsofparticularexpressions,rangingfromselectionalconstraintstophrasalidiomaticcombinations.Collocationalrelationshavefrequentlybeenusedtopositsyntacticrelations,particularly‘deep’syntacticrelations,butNunberg,Sag,andWasow(1994)arguepersuasivelythatcollocationalrelationsarefundamentallysemantic.Forexample,intheidiomaticcollocationpullstrings‘exerciseinfluence,’pullmeans‘exercise’onlyincombinationwithstrings,andstringsmeans‘influence’onlyincombinationwithpull.ThisobservationhasanaturalinterpretationinConstructionGrammar:thereisa[pullstrings]constructionwithaconstruction-specificsemanticinterpretationforitselements,withanoverallsemanticstructureof[EXERCISEINFLUENCE](seealsoWulff,thisvolume,onidiomsinConstructionGrammar).

Clicktoviewlarger

Figure12.5. TheinternalstructureofaconstructioninRadicalConstructionGrammar

Codedrelations,suchasindexation(agreement)andcasemarking,aregenerallytakentobeevidenceof‘surface’syntacticrelations,inmanydifferentgrammaticaltheories.However,syntacticrelationsarenotnecessaryinaConstructionGrammar.InallConstructionGrammars,therearerepresentationalcommitmentstothestructuresinFigure12.5:(1)themorphosyntacticstructureasawhole;(2)theelementsofthatstructureandtheirroleinthewholeconstruction;(3)themeaningasawhole;(4)thecomponentsofthatmeaningandtheirsemanticroleinthemeaningasawhole;(5)thesemanticrelationsholdingbetweenthecomponentsofmeaning;and(6)thesymboliclinksbetweentheformanditsmeaning,includingbetweenformalelementsandsemanticcomponents.Incomprehensionofanutterance,aheareridentifiestheconstructionasawholeanditselements;thehearercanidentifythemeaning,sinceaconstructionisapairingofformandmeaning;thehearercanthenidentifythe

Page 172: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 11 of 15

semanticcomponentsandtheirrelations;andfinally,thehearercanidentifywhichsyntacticelementdenoteswhichsemanticcomponent,sincetheelementsoftheconstructionhavesymboliclinkswiththeirsemanticcomponents.Thehearercanidentifythesemanticrelationsbetweenthecorrespondingsemanticcomponentswithoutinvokingsyntacticrelationsofanytype,byvirtueofunderstandingthemeaningasawholeandrecognizingthesymbolicrelationsbetweenconstructionalelementsandthecorrespondingsemanticcomponents.

Thereisalsoempiricalevidencethatsyntacticrelationsarenotdesirableaspartofaconstruction'sformalstructure.Ifconstruction-independentsyntacticrelationsexist,thentheremustbegeneralprinciplesofsemanticinterpretationthatlinkthosesyntacticrelationstosemanticrelations.Nonconstructionaltheoriesalwayspositsomesortoficonicallybasedgeneralprinciples(projectionprinciples,linkingrules,etc.)inordertoachievethisgoal,becausethereisnopairingofformandmeaninginsyntacticstructuresinthesetheories.Butcrosslinguistically,therearefartoomanynoniconicmappingsbetweenputativesyntacticrelationsandsemanticrelationsforgeneralmappingprinciplestohold(Croft2001:206–20).Forexample,althoughwellasamanneradverbdescribesapropertyoftheactionexpressedbythemainpredicate,insomelanguagesmanneradverbsagreewithoneoftheparticipants,suchastheobjectasinthefollowingMarathiexample(HookandChauhan1988:179;citedinCroft2001:210).

(5)ti  haa  bhaag  tsaangLaa  vaatselshe  this  part.Msg  good.Msg  will.read‘Shewillreadthispartwell.’

Instead,onemustpositconstruction-specificmappingsbetweenconstructionalelementsandthecorrespondingsemanticcomponents—thatis,oneisbacktoConstructionGrammar,butwithoutsyntacticrelations(seealsovanCanegem-Ardijns2006:447).

Seriousproblemsarisealsoifstructuressuchasindexationandcasemarkingarehypothesizedtoencodesyntacticrelations(Croft2001:220–33).Bothindexationandcasemarkingarevariableintheirappearance;iftheyflagsyntacticrelations,thenthesyntacticrelationsimplausiblycomeandgo.Worse,thesyntacticelementlinkedbyindexationisoftensimplyabsent(e.g.,so-callednullanaphorainclauses,andso-calledheadlessnounphrasesinphrases).Hence,thereisnoelementforindexationtorelatetosyntactically.InRadicalConstructionGrammar,theelementsotherwisetakentoindicatesyntacticrelationsareinsteadarguedtobesymbolic,thatis,encodearelationbetweenformandmeaning:indexationdenotesareferent,casemarkingdenotesasemanticrelation,andsoon.Thesemanticstructureindirectlyprovidestherelationshipbetweentheelementssupposedlyrelatedsyntacticallybyindexationandcasemarking.Phenomenathatarecommonlyexplainedintermsofsyntacticrelations,suchasthehead-dependentcontrastandthedistinctionamongdependentsbetweenargumentsandadjuncts,canbereanalyzedaspropertiesofthesymbolicrelationbetweenformandmeaningamongelementsinconstructions(Croft2001:chapter7).

Thus,inRadicalConstructionGrammartheonlyrepresentationalcommitmentsfortheformalstructureofconstructionsare:theconstructionitselfasacomplexgestalt,itselements,andtherolethattheelementhasintheconstructionasawhole.Ontheotherhand,inRadicalConstructionGrammar,thereisacommitmenttoarichsemanticrepresentation,whichmayincludepartsofasemanticframeorscenethatareevokedbytheconstructionbutdonothaveformalelementsthatdenotethemdirectly;andarichsymbolicstructurethatlinksnotonlytheformalstructureasawholetothemeaningasawholebutalsolinksmostifnotallofthesyntacticelementstocorrespondingsemanticelements.

ThereishoweveranotherimportantrepresentationalcommitmentinRadicalConstructionGrammar,sharedwithotherConstructionGrammars,namelythatcomplexconstructionsmayfilltherolesofothercomplexconstructions.Forexample,thePassiveSubjectroleinthePassiveVoiceconstructionisfilledbyaSubjectPhraseconstruction,whichmayitselfbecomplex:forexample,consistingofaNouncombinedwithaRelativeClause,whichinturnisacomplexconstructioninitsownright.Thenestingofcomplexconstructionsissimilartotheconstituentstructureofnonconstructionaltheories.However,itdiffersfromthoseinthattheexistenceofanestedconstructionisjustifiedonlybytheautonomyoftheconstruction(i.e.,itsconventionalizationindependentofthelargerconstruction).InaConstructionGrammar,some‘constituents’maynotbeautonomousconstructions.Forexample,intheidiomaticconstructionshootthebreeze,thebreezeisnotanautonomousphrasalconstituentanymore,sincethechoiceofDefiniteArticle,aswellasthechoiceofNoun,arefixedforthisconstruction.

Page 173: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 12 of 15

12.6.TheLanguage-SpecificityofConstructionsandtheOrganizationofConstructions

RadicalConstructionGrammar'sthirdmajorhypothesisisthatconstructionsthemselves,ormorepreciselytheformalstructureofconstructions,arealsolanguage-specific.Therearenodiscreteuniversalconstructiontypessuchaspassiveorcoordination.Thishypothesisisagainbasedoncrosslinguisticevidence.Crosslinguisticcomparisonrevealsthattheparticularcombinationofsyntacticpropertiesthatdefineparticularconstructionsinbetter-knownlanguagesarenotalwaysfoundtogetherinfunctionallysimilarorequivalentconstructionsinotherlanguages.Forexample,thereisahugevarietyofvoiceconstructions,farricherthanadivisioninto‘active,’‘passive,’and‘inverse’implies(Croft2001:chapter8).‘Passive’and‘inverse’representtwoidealizedconstructiontypesthatdifferfromthecanonical‘active’transitive,theformertypifiedbyEnglishandotherwesternEuropeanlanguagesandthelattertypifiedbyCreeandotherAlgonquianlanguages.Butaconstructionsuchastheonein(6)fromBellaCooladoesnotfiteithertype(Forrest1994:151–52;citedinCroft2001:294).Theverbisinthebasic(‘active’)form;thereisaspecial“passivesubject”agreementsuffixontheverb,buttheagentphraseisoblique,governedbyapreposition.TheBellaCoolaconstructionisrequiredif3rdpersonactson1stor2ndperson,andisusedwhenthepatientisamoretopical3rdpersonthantheagent,notunliketotheAlgonquian‘inverse.’

(6)k’x-im  ci-xnas  -cx  x-  ti-  ?imlk  -txsee  -3sg.pass  art-woman  -art  prep-  art-  man  -art‘Thewomanisseenbytheman.’

Inasimilarvein,Cristofaro(2003,2007)showsthatthetraditionaldichotomybetween‘finite’and‘nonfinite’isinadequatetocharacterizethefullrangeofsubordinateclauseconstructionsfoundintheworld'slanguages.

Aswithgrammaticalcategoriesandtherelationsamongsyntacticelements,thereareconstraintsonthiscrosslinguisticvariation.Thesyntacticpropertiesforconstructionssuchasvoiceconstructionscanbeusedtodefineasyntacticspaceinwhichthevoiceconstructionsofparticularlanguagescanbemapped.Inparticular,therangeofconstructionstowhichthenames‘active,’‘passive,’and‘inverse’havebeenappliedcanbemappedontoatwo-dimensionalsyntacticspaceinwhichtheAargumentismoreorlesssubjectlikeandthePargumentismoreorlesssubjectlike,basedoncasemarking,indexation,andsyntacticpositionoftheAandParguments.Thissyntacticspaceinturncovarieswithatwo-dimensionalconceptualspaceofthedegreeoftopicalityoftheAandParguments(Croft2001:chapter8).Likewise,thesyntacticspaceofsubordinateclauseconstructionscovarieswiththefunctionalparametersofthecognitivestatusandrecoverabilityofsemanticinformationfromthedependentstateofaffairs,anddegreeofsemanticintegrationofthemainanddependentstateofaffairs(Cristofaro2003,2007).

Insum,constructionsquagestaltcombinationsofmorphosyntacticpropertiesdonotcomeindiscreteuniversaltypes.However,therangeofcombinationsofmorphosyntacticpropertiesisnotunconstrainedandismotivatedbythefunctionsthatthoseconstructionsperform.

Anotherquestionpertainingtothecombinationsofmorphosyntacticpropertiesthatoccurtogetherinconstructionsis:howareconstructionscombinedtoyieldsentences,orconversely,whatwaysarethemorphosyntacticpropertiesofutterancesdividedintodistinctconstructions?Forexample,inEnglish,argumentstructureconstructions(thecombinationsofpredicates,arguments,andtheirassociatedcasemarking,andindexationmarkers)areindependentoftheTense-Aspect-Mood(TAM)constructions(theinflectionandauxiliarycombinations),althoughbothoverlapincontainingaVerbelement.ThisseparationofargumentstructureandTAMconstructionsiswidespreadthoughbynomeansuniversal.Whyisthisso?

Onepossibleexplanationisfoundintheverbalizationprocess.Chafe(1977a,b)essentiallyattackstheproblemofhowaspeakerstartswithanunanalyzed,uniqueexperienceandproducesalinguisticutterancemadeupofreusedparts—wordsthathavebeenusedinpriorutterancesdescribingotherexperiencesdifferentfromthecurrentone.Hearguesthatthreeprocessesareinvolved:subchunking,propositionalizing,andcategorizing.Insubchunking,thespeakerbreaksuptheexperienceintosmallerunitsofthesametype.Roughly,subchunksare‘clause-sized’experiences.(Chafe1994treatstheprocessnotasdividingupalargerexperienceintoasetofparts,butasshiftingthefocusofconsciousnessfromone‘clause-sized’subchunktoanother.)Subchunksarepropositionalized,thatis,dividedintounlikeparts:thosepartswhich(may)recurinothersubchunks(i.e.,referents)andtherest,whichrepresentsthepartsoftheexperiencegenerallyspecifictothatsubchunk.Finally,

Page 174: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 13 of 15

thepartsofapropositionalizedsubchunk(objects,properties,actions)arecategorizedasrecurrententitiesthathavebeenverbalizedinpriorutterancesaboutotherexperiences.

Thisverbalizationmodelprovidesatheoryofhowanexperientialgestaltisbrokenupintopartsandhowthosepartsarecategorizedasthesameaspriorexperiences—ascontentwords.Anutterancealsoreestablishestheuniquenessoftheparticularobjects,properties,andactionsintheexperience,andreconstitutestheexperienceasawhole(Croft2007b).Thereestablishmentoftheuniquenessoftheexperienceandtheentitiesthatmakeitupisachievedbygrammaticalconstructionsthatparticularizetheentityintheexperience:individuatingandquantifyingtheentity,specifyingitslocationinspace,time,andthementalspacesoftheinterlocutors.Thesearegenerallyvarioustypesofinflectionalconstructionsandmodifyingconstructions(modifyingpredicates,aswellasreferringexpressions).Reconstitutingthewholesubchunkoutofitspartsisachievedbyargumentstructureandrelatedclausalstructuringconstructions.Relatingsubchunkstooneanotherisachievedbyclause-linking,referencetracking,andothercoheringconstructions.Forinstance,intheutteranceAndonekidtakestherockthathetrippedonandhethrowsittothesideoftheroad,theexperience(fromthePearFilmnarratives)issubchunkedintotwoevents,andtheeventsarecategorizedastakingandthrowing,withrecurringparticipantsbeingthekid,therock,andtheroad.Theeventsareparticularizedbysituatingtheminthenarrativepresenttense,inwhichtheentirefilmnarrativeisplacedbythisspeaker.Theparticipantsareparticularizedbyanumeraloneselectingonekidfromthegroupofthree,therelativeclausespecifyingtherelevantrock,andthepartphrasethesideoffortheparticularplacewithrespecttotheroad.Theparticipants’rolesintheeventarestructuredbytheTransitiveandCausedMotionargumentstructureconstructionsthatindicatewhodoeswhattowhom.Finally,coherencebetweentheclausesisprovidedbythecoordinatingconjunctionand,andthepronounsheandittrackingthekidandtherockacrossthetwoclauses.

Thismodelofverbalizationprovidesafunctionallymotivatedframeworkfordistinguishingthedifferentkindsofconstructionsthatcombineinanutterance.Ingeneral,theparticularizing,structuring,andcoheringfunctionsareencodedbyindependentconstructionsthatarecombinedintheproductionofutterances.Forexample,argumentstructureconstructionsstructurethepropositionalizedsubchunk,whileTAMconstructionsparticularizetheevent.Thesetwoverbalizationfunctionsareusuallycarriedoutbyindependentconstructions.Itisalsopossible,thoughlesscommon,thatparticularcombinationsofverbalizationprocessesareconventionallycombined.Forexample,inlanguageswithaspect-basedsplitergativepatterns,suchasHindi,oneargumentstructureconstruction(anergativeone)isusedinperfectivesituations,whileanotherargumentstructureconstruction(anaccusativeone)isfoundinimperfectivesituations.

BoththestructureofconceptualspaceandtheverbalizationprocessesprovideafunctionalbasisfortheorganizationofconstructionsinaConstructionGrammar.Inaddition,constructionsarealsoorganizedintermsoftheformal(morphosyntactic)propertiestheypossess.Aswiththerepresentationofgrammaticalgeneralizations,RadicalConstructionGrammarexploresnewmodelsforrepresentingconstructionalorganization,suchastheverbalizationprocessandsyntacticspacesdefinedbyformalproperties,tosupplementandpossiblyreplacemoretraditionalmodelssuchastaxonomichierarchies.

12.7.Conclusion:RadicalConstructionGrammarandTypologicalTheory

RadicalConstructionGrammarisarecenttheoryofgrammarevenbytheyouthfulstandardofcontemporaryConstructionGrammars.RadicalConstructionGrammaradherestorigorousmethodologicalcommitments,tothedistributionalmethodandtotypologicalpatternsdiscoveredusingcrosslinguisticallyvalidcriteria.Asaconsequence,comparedtootherConstructionGrammarmodels,RadicalConstructionGrammarhasaverysparesetofrepresentationalcommitmentstotheformalstructureofconstructions.Ontheotherhand,RadicalConstructionGrammarappealstoarich,fine-grainedmodelofsemanticstructure,andplacesmuchoftheexplanatorypowerinsemanticstructureandthesymbolicmappingbetweencomponentsofsemanticstructureandelementsofsyntacticstructureinconstructions.

RadicalConstructionGrammardrawsonanumberofconceptsfromtypologicaltheory(Croft2003b),aswellasprovidingawayinwhichConstructionGrammarcancontributetotypology.Typology'smostsalientmethodologicalprincipleisthecomparisonofageneticallyandgeographicallydistributedsampleoflanguagesoftheworld.Thebasisofcrosslinguisticcomparisonforgrammaticalstructuresmustbetheirfunction,becauseofthegreat

Page 175: The Oxford Handbook of Construction Grammar

Radical Construction Grammar

Page 14 of 15

structuraldiversityoflanguages(thestructuralpropertiesareessentiallylanguage-specific).Inotherwords,typologistscomparehowlanguagesencodefunctioninmorphosyntacticform.Hence,morphosyntactictypologyisthecrosslinguisticstudyofform-functionpairings,whichisdirectlycomparabletoConstructionGrammar.

RadicalConstructionGrammaradheresstrictlytothedistributionalmethod,recognizingthatthedistributionalmethodisbasedontheidentificationofconstructionsthatserveasthecontextsfordistributionalanalysis.Typologyrepresentsanextensionofthedistributionalmethodacrosslanguages.Justasfunctionsprovidethebasisforthecrosslinguisticcomparisonofconstructionsusedtoencodethosefunctions,typologiststhencomparethedistributionofsimilarconstructionsacrossthosefunctions.

Theintegrationoflanguage-specificandcrosslinguisticdistributionalanalysismaybeillustratedwithoneoftheclassicstudiesintypology,KeenanandComrie'stypologyofrelativeclausesandtheNPAccessibilityHierarchy(KeenanandComrie1977,1979).Relativeclauseconstructionsdefineadistributionintermsofthesyntacticroleplayedbytheheadintherelativeclausethatmodifiesit(subject,object,etc.).Alanguagemayhavemorethanonerelativeclauseconstruction,andthedistribution(occurrence)ofsyntacticrolesmaydifferfromonerelativeclauseconstructiontoanotherinasinglelanguage.KeenanandComrieaddacrosslinguisticdimensiontothisdistributionpattern,thatis,comparingrelativeclauseconstructionsacross,aswellaswithin,languages.KeenanandComrieproposecrosslinguisticallyuniversalconstraintsonthedistributionofrelativeclausesacrosssyntacticroles,basedonahierarchicalrankingofthoseroles(Subject<Object<IndirectObject<Oblique<Genitive<ObjectofComparison).Thus,thebasictypologicalmethodcombineslanguage-internalandcrosslinguisticdistributionalanalysis.

KeenanandComrie'sanalysis,likeotherclassic,methodologicallyrigoroustypologicalanalyses(e.g.,Greenberg1966/1990;Stassen1985,1997,2009;Koptjevskaja-Tamm1993;Haspelmath1997;Cristofaro2003),providesasoundbasisforsyntacticargumentationinidentifyinggrammaticaluniversals,thatis,universalsofgrammaticalconstructions(Croft2009a).First,universalsmustbebasedonthesimultaneouscomparisonofabalancedsampleoflanguages.Thesameconstructionsmustbecomparedacrosslanguages(i.e.,nomethodologicalopportunism).Theconstructionsmustbeidentifiedusingcriteriathatarevalidacrosslanguages(i.e.,nolanguage-specificcriteria).Thesecriteriaareultimatelybasedinfunction,asnotedabove.Distributionalpatternsmustbeexaminedindetail(i.e.,morethanoneortwofillersofaconstructionalrole),orelseextrapolationfromtheexamplesusedmustbedonecautiouslyandwithappropriatequalifications.Ifanomalouspatternsappeartoberelatedtootherdistributionalfacts—thatis,otherconstructionsappeartointeractwiththedistributionoftheconstructionbeingexamined—thenthecorrelationofconstructionsmustalsobecrosslinguisticallyvalidatedinthewaydescribedinthisparagraph.

TheremarkablediversityofmorphosyntacticstructuresandofdistributionalpatternsrevealedinrigoroustypologicalanalysisunderliesthebasictenetsofRadicalConstructionGrammar.Therejectionofthebuilding-blockmodelofsyntacticstructureisgroundedinthepervasivemismatchesofdistributionalpatternsacrossconstructionswithinasinglelanguageandwiththesameconstructionacrosslanguages.Asnotedinsection12.4,usefultechniquestoidentifyandrepresentlanguageuniversalswithoutsyntacticbuildingblocks—thesemanticmapmodelandmultidimensionalscaling—havetheiroriginsintypology.ThesimplificationofsyntacticstructureproposedinRadicalConstructionGrammarrepresentsthemarriageofatheoreticalargumentfromConstructionGrammar—thepresenceofsymbolicrelationsinagrammaticalconstruction—andempiricalevidencechallengingsyntacticrelationsfromtypology.Finally,thenonuniversalityofconstructionsisalsoaresultfromtypologicalresearch;researchinconstructionalanalysis,framesemantics,andverbalizationprocessesisbeginningtoprovideaframeworkforidentifyingconstructionswhoseuniversalpropertiescanbediscoveredincrosslinguisticcomparison.

WilliamCroftWilliamCroftisProfessorofLinguisticsattheUniversityofNewMexico.HereceivedhisPh.D.fromStanfordUniversityin1986and—besidesnumerousvisitingpositionsacrosstheworld—workedattheUniversityofMichigan,AnnArbor,aswellastheUniversityofManchester.Hismainresearchinterestsaretypology,semantics,cognitivelinguistics,constructiongrammar,andlanguagechange.Hehaspublishedwidelyonthesetopicsininternationally-renownedpeer-reviewedjournalsandhasauthoredseminalbookssuchasExplainingLanguageChange:AnEvolutionaryApproach(2000),RadicalConstructionGrammar:SyntacticTheoryinTypologicalPerspective(2001),TypologyandUniversals(2003),CognitiveLinguistics(withD.A.Cruse2004).

Page 176: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0013

CognitiveConstructionGrammarHansC.BoasTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterfocusesonCognitiveConstructionGrammar(CCG),whichaimsatprovidingapsychologicallyplausibleaccountoflanguagebyinvestigatingthegeneralcognitiveprinciplesthatservetostructurethenetworkoflanguage-specificconstructions.IttracesthefoundationsofCCG,discussesthemajororganizingprinciplesandthearchitectureofCCG,anddescribestheorganizationofconstructionalknowledgeinCCG.ThechapteralsocomparesCCGwithotherstrandsofConstructionGrammartoshowwhatideastheyshareandwheretheydiffer,andlooksattheinteractionofmultipleconstructions,theroleofnetworks,andinheritancehierarchies,aswellasfrequencyandproductivityfromaCCGperspective.

Keywords:CCG,language,cognitiveprinciples,constructionalknowledge,ConstructionGrammar,multipleconstructions,inheritancehierarchies

13.1.Introduction

SincethepublicationofGoldberg's(1995)seminalbookConstructions,ConstructionGrammar(CxG)hasbecomeincreasinglypopularandinspiredanalysesofawiderangeofgrammaticalconstructionsindifferentlanguages(e.g.,MichaelisandRuppenhofer2001;Barðdal2008;Iwata2008;Boas2010a).Inaddition,manyofGoldberg'sinsightshaveinfluencedresearchinfirstlanguageacquisition(e.g.,Tomasello2003;Diessel,thisvolume),secondlanguageacquisition(e.g.,GriesandWulff2005;Haberzettl2007;Ellis,thisvolume),andlanguagechange(e.g.,Diewald2007;Traugott2008a,b,c,d;LeinoandÖstman2008;Fried,thisvolume).ThischapteroutlinesthemainideasandorganizingprinciplesofGoldberg's(1995)constructionalapproach,whichhascometobeknownasCognitiveConstructionGrammar(CCG)sincethepublicationofher(2006a)bookConstructionsatWork.Thechapterisstructuredasfollows.Section13.2discussesthemajororganizingprinciplesandthearchitectureofCCG.Section13.3addressestheorganizationofconstructionalknowledgeinCCG.Section13.4comparesCCGwithotherstrandsofConstructionGrammartoshowwhatideastheyshareandwheretheydiffer.

13.2.WhatareConstructions?

ThemostbasicideathatCCGshareswithotherconstructionalapproachesisthatalinguisticmodelshouldinprinciplebeabletoaccountforallfacetsofaspeaker'sknowledgeabouttheirlanguage.Anotherbasicideaisthatgrammaticalconstructionsarethefundamentalbuildingblocksoflanguage. ThisviewisinstarkcontrasttotheChomskyanconceptionofconstructionsasmeretaxonomicartifacts,usefulfordescription,butwithoutanytheoreticalstatusandnoexplanatorypower.Ontheconstructionalview,constructionsarelearnedpairingsofformandmeaning(function),asGoldberg's(2006a:5)definitionillustrates.

Anylinguisticpatternisrecognizedasaconstructionaslongassomeaspectofitsformorfunctionisnotstrictlypredictablefromitscomponentpartsorfromotherconstructionsrecognizedtoexist.Inaddition,

1

Page 177: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 2 of 14

patternsarestoredasconstructionseveniftheyarefullypredictableaslongastheyoccurwithsufficientfrequency.

Inthisview,alllevelsofgrammaticalanalysisinvolveconstructions:learnedpairingsofformwithsemanticordiscoursefunction,includingmorphemesorwords,idioms,partiallyfilledandfullylexicalpatterns.Thismeansthateventhemostgeneralsyntacticconstructionshavecorrespondinggeneralrulesofsemanticinterpretation(theyaresymbolicunits).Thearchitectureofaconstruction,couplingaparticularformwithaspecific(conventional)meaning,isasfollows.

Clicktoviewlarger

Figure13.1. Thesymbolicstructureofaconstruction(Croft2001:18)

AsFigure13.1illustrates,theformofaconstructioncanbeassociatedwithdifferentkindsoflinguisticallyrelevantinformation(syntactic,morphological,orphonological).Theformsideofaconstructionislinkedtoitsmeaningsideviaasymboliccorrespondencelink.Theterm‘meaning’isunderstoodtoincludealloftheconventionalizedaspectsassociatedwithaconstruction'sfunction,forexample,thatcertainobligatoryargumentscanbeomitted,giventheproperdiscoursecontext,asinsentenceslikeThetigerkilledagain(Goldberg2002),orthatparticulartypesofconstructionscanbeemployedtoexpresssurpriseinacertainpragmaticsituation,asinsentenceslikeWhat'sthatflydoinginmysoup?(KayandFillmore1999).Theideathatconstructionsareregardedaslearnedpairingsofformandmeaninghasseveralimportantimplicationsforthearchitectureofgrammar.

13.2.1TypesofConstructions

Goldberg's(2006a:18)proposalthat“it'sconstructionsallthewaydown”bestcharacterizestheprimarystatusofconstructionsinCCG:wheneveritisnotpossibletopredictallofthefactsabouttheuse,internalcomposition,combinatorypotential,ormeaningofthepatternunderstudytosomeindependentlymotivatedprinciplesoralreadyknownconstruction(s),itmaybecomenecessarytoproposeaseparateconstruction. Whenanewconstructionispositeditisimportanttokeepinmindthatitfollowsthesamegeneralarchitectureofconstructions,thatis,acombinationofaparticularformwithaspecific(conventional)meaning,asinFigure13.1.

Notonlydoconstructionsdifferintheirsizeandcomplexitybutalsointhetypeofmeaningthattheyconvey.Forexample,themeaningsofcontentwords(averyspecifictypeofconstruction)aretypicallyespeciallyrichastheytendtodescribeverydetailedobjectsorsituations,whichcanbedescribedandanalyzedwithFrameSemantics(Fillmore1982).Incontrast,amoreabstractconstructionsuchasthepassiveiscomparativelylessrichinmeaningasitonlypresentsadifferentperspectiveofaneventandassuchencodesarelativelyabstractmeaning(i.e.,ashiftinperspectivefromtheactive).Thesebriefexamplesfocusonconstructionsthatdifferintheirsize,complexity,andproductivity,butdonotaddressthegroupofconstructionsmostwidelydiscussedinCCG,namelyArgumentStructureconstructions.

13.2.2ArgumentStructureConstructions

ResearchinCCGisperhapsbestknownforitsnovelthesisthatpatternsofargumentstructure(so-calledArgumentStructureconstructions)existindependentlyoflexicalargument-takingpredicates.Inthisview,proposedinGoldberg(1995),constructionssuchastheDitransitive,CausedMotion,ortheWayconstruction

2

3

Page 178: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 3 of 14

Table13.1.Examplesofconstructions,varyinginsizeandcomplexity

Morpheme pre-,-ing

Word avocado,anaconda,and

Complexword daredevil,shoo-in

Complexword(partiallyfilled) [N-s](forregularplurals)

Idiom(filled) goinggreatguns,givetheDevilhisdue

Idiom(partiallyfilled) jog<someone's>memory,send<someone>tothecleaners

CovariationalConditional TheXertheYer(themoreyouthinkaboutit,thelessyouunderstand)

Ditransitive(doubleobject) SubjVObj Obj (hegaveherafishtaco;hebakedheramuffin)

Passive SubjauxVP (PP )(thearmadillowashitbyacar)

Source:Goldberg2006a:5. are

capableofsupplyingaverb'ssemanticswithadditionalarguments.Oneofthecentralargumentsforpositingmeaningfulconstructionsthatexistindependentlyofthewordsthatinstantiatethemstemsfromthewishtoavoidtheclaimthatthesyntaxandsemanticsoftheclauseisprojectedexclusivelyfromthespecificationsofthemainverb(seeGoldberg1995:224).Thisviewhastheadvantageofnothavingtopositimplausibleverbsensesforcasesinwhichverbsoccurinanunusualenvironmentasinthefollowingexamples.

(1)a.Theylaughedthepoorguyoutoftheroom.b.Franksneezedthetissueoffthetable.c.MaryurgedBillintothehouse.(Goldberg1995:152)

Thesesentencescannotbeeasilyexplainedcompositionallybecauseverbslikelaugh,sneeze,orurgedonotindependentlyencodecaused-motionsemantics.InCCGtheverbsin(1)areassociatedwithspecificlexicalsemanticinformationthatallowsthemtointegrate(or‘fuse’)withthesemanticsofanArgumentStructureconstruction,namelytheCausedMotionconstruction.Thisfusioninturnlicensesthepostverbalconstituentssuchasthepoorguyandoutoftheroomin(1a).Inotherwords,althoughtheverbsin(1)contributetheirbasicmeanings,itistheCausedMotionconstructionthatisitselfassociatedwithmeaningandthereforecontributestheadditionalargumentsprovidingthefinalinterpretationofcaused-motion.Figure13.2illustrateshowtheconstructionalsemantics(constructionalroles)oftheCausedMotionconstructionandtheverbalsemantics(participantroles)ofanintransitivematrixverbarefusedinCCGinordertoformthecaused-motioninterpretation.

Clicktoviewlarger

Figure13.2. CausedMotionconstruction(Goldberg1995:163)

TherepresentationoftheCausedMotionconstructioninFigure13.2consistsofthreedifferentlayers.Thetoplineoftheboxcontainstheconstruction'sownmeaning(Sem),whichliststhesemanticargumentsoftheconstruction(theconstructionalroles)andrepresentstheirsemanticrelationswithrespecttoeachother.Thus,theCaused

1 2

PP by

Page 179: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 4 of 14

Motionconstructionisassociatedwiththesemantics‘XCAUSESYTOMOVEZ.’Solidlinesbetweenthesemanticrolesandrolesinthepredicate'srolearrayindicatethatthesemanticrolemustbefusedwithanindependentlyexistingverbalparticipantrole.Dottedlinesindicatethattheconstructionisabletoprovideadditionalparticipantroles.Themiddlelineoftheconstructioncontainsopenslotsintowhichtheverb'sparticipantrolesfuseandthebottomliststheovertsyntacticrealizationsofthesemanticargumentsofthecombinedverb-constructionsemantics.Rolesrepresentedinboldare‘profiled’arguments,thatis,entitiesinaverb'ssemanticsthatareobligatorilyaccessedandfunctionasfocalpointswithinthescene,achievingaspecialdegreeofprominence(Langacker1987;cf.Goldberg1995:44).

(2)a.sneeze:<sneezer>b.Marysneezedthenapkinoffthetable.

Example(2a)showsthestructureoflexicalentriesinCCG,inthiscasecapturingthefactthatsneezeisassociatedwithasingleparticipantargumentrole,thesneezer.Theparticipantroleisrepresentedinboldprinttoreflecttheobservationthatthesneezerroleislexicallyprofiled.InCCG,asinotherconstructionalapproaches,verbsareassociatedwithspecificsemanticframes(Fillmore1982).Framesemanticinformationcapturestherichnessofthevariousmeaningsassociatedwithalexicalitem,suchasreferencestoworldandculturalknowledge,experiences,andbeliefs(seePetruck1996;Ziem2008;FillmoreandBaker2010fordetails).Thisisinstarkcontrasttotheoriesthatplaceheavyemphasisonderivingthesemanticsofutterancesfromobjectivetruthconditions(seeFillmore1975).InCCG,thelexicalentriesofverbscontainparticipantrolesrepresentingtheframesemanticinformation.Forexample,whensneezefuseswiththeCausedMotionconstructioninFigure13.2,theverbsneezecontributesthesneezerrole(sneeze:<sneezer>),whereastheconstructioncontributesbothathemeandagoalroletotheverb'ssemantics. Inotherwords,sneezespecifiesthemeansbywhichtheCAUSE-MOVErelationisachieved,whereastheconstructionprovidestherestofthesemanticswhichthenin(2b)yieldstheinterpretationofMarycausedthenapkintomoveoffthetablebysneezing.

ThegeneralmechanismoffusingverbswithconstructionsoutlinedinthepreviousparagraphsisthesameforothertypesofArgumentStructureconstructionssuchastheWayConstruction(e.g.,Theylaughedtheirwayintothemeeting),theDitransitiveconstruction(e.g.,JoebakedMiriamacake),andtheResultativeconstruction(e.g.,Kimpaintedthebrushtopieces).Inordertoavoidunacceptablefusionsofverbswithconstructions,CCGpositsbothconstruction-specificandmoregeneralconstraints.TolimittheapplicationoftheCausedMotionconstruction,Goldbergproposesanumberofsemanticconstraints,forexample,thatthecauserargumentcanonlybeanagentornaturalforce,notaninstrument(Goldberg1995:165).Crucialtotheapplicationofconstraintsisthenotionthatexpressionsarelicensedbydifferentconstructions(andtheirconstraints)aslongastheycanbeconstruedasnotbeinginconflict(seeMichaelis2004onhowcoercioninfluencesconstrual).Inaddition,moregeneralconstraintsregulatethefusionofverbswithconstructions,especiallytheSemanticCoherencePrincipleandtheCorrespondencePrinciple:

TheSemanticCoherencePrinciple:Onlyroleswhicharesemanticallycompatiblecanbefused.Tworolesr andr aresemanticallycompatibleifeitherr canbeconstruedasaninstanceofr ,orr canbeconstruedasaninstanceofr .Forexample,thekickerparticipantofthekickframemaybefusedwiththeagentroleoftheditransitiveconstructionbecausethekickerrolecanbeconstruedasaninstanceoftheagentrole.Whetherarolecanbeconstruedasaninstanceofanotherroleisdeterminedbygeneralcategorizationprinciples.(Goldberg1995:50)

TheCorrespondencePrinciple:Eachparticipantrolethatislexicallyprofiledandexpressedmustbefusedwithaprofiledargumentroleoftheconstruction….Ifaverbhasthreeprofiledparticipantroles,thenoneofthemmaybefusedwithanonprofiledargumentroleofaconstruction.(Goldberg1995:50)

Goldberg'sapproachhasbeenverysuccessfulinaccountingforabroadrangeofArgumentStructureconstructionsinEnglishandhasalsoinspiredothercognitivelyorientedanalysesofconstructionsinlanguagessuchasFinnish(LeinoandÖstman2008;Leino2010),French(BergenandPlauché2001;LambrechtandLemoine2005),Icelandic(Barðdal2008),Japanese(Fujii2004),German(MichaelisandRuppenhofer2001;Boas2003;Hilpert2008),Spanish(Gonzálvez-García2010),andThai(TimyamandBergen2010).

Goldberg'sresearchonArgumentStructureconstructionshasattractedagreatdealofattentionbecauseoftheclaimthat“phrasalpatternsarenotdeterminedbyverbsalone.”Thisclaimwasmadewhenattentionwaspaid

4

1 2 1 2 2

1

Page 180: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 5 of 14

exclusivelytoverbsalone,inordertoaccountforwhyoneandthesameverbmayappearinmorethanonesyntacticframe(i.e.,ArgumentStructureconstructionslikeresultatives).Inthegenerativeframework,lexicalruleapproacheslikeLevinandRappaport(1988)andPinker(1989)arevirtuallytheonlypossibilityforaccountingforArgumentStructureconstructions.However,aftertheemergenceofGoldberg'stheory,evengenerativistshadtoconcedethatGoldberg'saccountisappealing,andsomescholarsevenintroducedtheoreticaldevicesthatareclearly‘constructional’:Jackendoff's(1990)AdjunctrulesandLevinandHovav's(1998/2001)lexicaltemplatesarevirtuallyconstructions.

Atthesametime,anumberofstudiessuchasKay(1996,2005),Nemoto(1998),vanderLeek(2000),Boas(2003),andIwata(2008)arguethatthetypesofabstractmeaningfulArgumentStructureconstructionsareoftentoopowerfulandhavethepotentialtoovergenerateunattestedsentences.Twopointshavebeenshowntobelargelyresponsiblefortheseissues.First,theconstraintsonthefusionofverbsandconstructionsarenotalwayssufficienttopreventconstructionsfromfusingwithcertaintypesoflexicalentries.Second,thestatusoflexicalentriesisproblematicasinmostcasesthemeaningsofverbsarerepresentedintermsofrelativelysparseframe-semanticinformationasin(2a)above(seealsoBoas2008).Toovercometheseproblems,Nemoto(2005),Boas(2005,2008),andIwata(2008)proposetopaycloserattentiontotheindividualsensesofverbsastheseareoftenconventionalizedinidiosyncraticwaysthatdefygeneralconstructionalgeneralizations.Inthislexical-constructionalview,individualverbsensesshouldberegardedasmini-constructionswiththeirownframe-semantic,pragmatic,andsyntacticspecificationswheneverabstractmeaningfulconstructionsovergenerate.ThesealternativeanalysesdonoteliminatetheneedforabstractmeaningfulconstructionsaspostulatedbyGoldberg,buttheylimittheirpowersubstantially.Inthisview,mini-constructionsmayformclasseswithothermini-constructions,establishinginheritancehierarchiescontainingmoreandlessgeneralpatternswithdifferentlevelsofsemanticabstraction.ThismeansthatwhileverybroadgeneralizationsarecapturedbyGoldberg-typeabstractmeaningfulconstructions,morelimitedconventionalizedpatternsarecapturedbymoreconcreteconstructionsatvariousmidpointsofthehierarchicalnetwork(seealsoCroft2003).

13.2.3OtherTypesofConstructions

Followingher(1995)book,Goldberg'sconstructionalinsightshavealsobeenappliedtothedescriptionandanalysisofotherlinguisticphenomenathatlieoutsideofthescopeof‘traditional’ArgumentStructureconstructions,suchasconstructionsthatincorporatediscourse-relevantinformation.Onesuchexampleistheso-calledDeprofiledObjectconstruction(Goldberg2000),adiscourseconstructionwithdirectbearingonargumentstructure.Thisconstructionlicensescasesinwhichanargumentthatisnormallyassociatedwiththeverbisunexpressedduetoacombinationofitslowdiscourseprominencetogetherwithanincreasedemphasisontheaction.InsentencessuchasTigersonlykillatnight,theargumentofthetransitiveverbkillcanbeomittedbecausetheargumentisnotprominentindiscourseandthereforeneedsnottobeexpressed.Inotherwords,thepatientargumentofthechangeofstateverbkillisneitherfocalnortopical(cf.Lambrecht1994),whileatthesametimetheactionisemphasized.TheDeprofiledObjectconstructionservesacommunicativefunctionbyshiftingdiscourseprominenceawayfromthepatientargument,effectivelylicensingitsomission.Otherexamplesofconstructionsincorporatingdiscourse-relevantinformationincludetheNominalExtrapositionconstruction(It'samazingthepeopleyouseehere;MichaelisandLambrecht1996),whichintegratescategoriesofinformationstructureintogrammaticaldescription,theImplicitThemeconstruction,whichallowsthemeargumentsofemissionandcontributionverbstobeomittedincertaincontexts(ShedonatedtotheUnitedWay;Goldberg2004a),andtheFrenchContextFocusingParceQueConstruction(CFPC)(e.g.,DeulofeuandDebaisieux2009),whosemeaningtranscendsregularpropositionalmeaningasitbelongstothedomainofinformationprocessingandthestructuringofcommonground.Closelyrelatedtothislineofresearcharestudiesofconstructionsandspeechacts,whereparticularpragmaticconditionsinfluencethelicensingofspecificconstructionssuchastheLetAloneconstruction(Fredwon’tordershrimp,letaloneLouise,squid;Fillmore,Kay,andO’Connor1988),theWXDYconstruction(What'sthatflydoinginmysoup?;KayandFillmore1999),andtheN-be-thatconstruction(thethingis/thepointis…;Günthner2008)(seealsoKay2004a;Leino,thisvolume).

Whileconstructionalresearchhasfocusedprimarilyontheroleofsemantic,pragmatic,andsyntacticfactorsinlicensingconstructions,veryfewstudiesaddresstheinfluenceofphonologicalfactors.Historicallyspeaking,thistendencymayperhapsbeexplainedbytheprimaryfocusofconstructionalresearchonshowingthatthereexistnostrictseparationbetweenthelexiconandsyntax,semantics,andpragmatics,therebyofferingatheoretical

5

Page 181: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 6 of 14

alternativetophrase-structuretreesinthegenerative-transformationalparadigm.Assuch,onlyfewaccountshavereliedonphonologicalfactorsforexplainingthedistributionofconstructions.OnesuchexampleisLambrecht(1990),whodemonstrateshowtheinterpretationofaparticularlinguisticexpressionsuchasHimbeadoctor!?dependsontheprosodiccontourconventionallyassociatedwithit.Inthiscase,theunusualmorpho-syntaxofaccusativesubjectandbarestemverbphrasecannotbeaccountedforbyotherexistingconstructions.Instead,itisonlyacceptablewhenpairedwithaparticularintonationalcontour,therebyexpressingthespeaker'sincredulitywithregardtosomepropositionthathasjustbeenproposed.Otherstudieshighlightingtheinfluenceofphonologicalfactorsforlicensingconstructionsinclude:MichaelisandLambrecht(1996),whoillustratethatexamplesofnominalextrapositionnecessarilycontainanactivationaccentfallingatsomepointwithinthepostpredicateNP;Boas(2004),whoshowsthatwanna-contractioninEnglishisduetophonologicalreductioninfastspeech,whichcanbeaccountedforbyanonmodulararchitectureallowingforsimultaneousinteractionofsyntactic,semantic,pragmatic,andphonologicalinformation;andFriedandÖstman's(2004b)discussionofpatternssuchasIsasaunahot,whoseinterpretationdependsontheintonationusedinpronouncingthestringofwords,whichinturnreflectsthelicensingbydistinctgrammaticalconstructions,thatis,eitheraquestionconstructionoranexclamativeconstruction.Theseexamplesshowthatphonologicalinformationmaysometimesalsoberelevantforthelicensingofconstructions,andassuchneedtobeincludedinformalrepresentationswhennecessary.

13.2.4InteractionofMultipleConstructions

Unlikegenerative-transformationalanalysesthatassumevariouslevelsofrepresentation,constructionalapproachessuchasCCGdonotemployderivationstorelaterepresentationsatdifferentlevelstoeachother.Instead,thereisonlyonelevelofrepresentationatwhichsentencesarelicensedbydifferentsetsofconstructions,ineffecta‘whatyouseeiswhatyouget’modeloflanguage.Constructionscanbecombinedfreelytoformactualexpressionsaslongastheyarenotinconflict.ConsiderthesentenceWhatdidMichaelsendMiriam?,whichinvolvesthecombinationofanumberofdifferentconstructions:first,eachofthefivewordsareconstructions;thentherearetheVPconstruction,theNPconstruction,aSubject-AuxiliaryInversionconstruction,thewh-construction,andtheDitransitiveconstruction.Whencombiningdifferentconstructions,itbecomesobviousthattheyeachfulfilldifferenttasksinconstructingsentences.Forexample,whiletheVPandNPConstructionscombinelargerphrasesoutofindividualwords(constructions),thewh-constructionlicensestheargumentoftheverbwhatinsentence-initialposition,andtheDitransitiveconstructionisunderstoodtoencodethegrammaticalrelationsbypairingaparticularform/functionwithaparticularmeaning.Theimportantpointhereisthatallconstructionsinvolvedcombinetoformsentencesatonlyonelevel,andtheycombinefreelybecausetheyarenotinconflict.

Goldberg'sviewofconstructionalinteractionhastheadvantageofbeingabletorelatethosesurfaceformssystematicallytoeachotherwhichsharecertainmeanings.Forexample,Goldberg's(2002)analysisofthelocativealternationproposesanalternativetoLevin's(1993)account.WhereLevin(1993)reliedonstrictlysyntacticcriteriatoidentifysemanticverbclassesthatexhibitsimilaralternatingbehavior,GoldbergproposestotakeacloserlookathowsemanticallyrelatedverbsinteractwithdifferentsetsofArgumentStructureconstructions.Tothisend,Goldberg(2002:343–44)claimsthattheoverlapinmeaningbetweenthealternantsin(3a)and(3b)isaccountedforbyrecognizingthattherearetwosetsofconstructionsthatsharethemeaningofload.

(3)a.CausedMotionconstruction(e.g.,Patloadedthehayontothetruck)CAUSE-MOVE(causethemepath/location)load(loaderloaded-themecontainer)b.Causativeconstruction+Withconstruction(e.g.,Patloadedthetruckwithhay)CAUSE(causepatient)+INTERMEDIARY(instrument)load(loadercontainerloaded-theme)(cf.Goldberg2002:344)

Theverbload,whoselexicalentryconsistsoftheparticipantrolesloader,loaded-theme,andcontainer,iscapableoffusingwithtwosetsofconstructions,namelytheCausedMotionconstructionortheCausative+Withconstruction.Crucially,thedifferentconstrualsoftheverb'sparticipantrolesallowforloadtofusewithdifferentconstructions:theloaded-themerolecanbeconstruedaseitheratypeofthemeroleasin(3a),orasatypeof

6

Page 182: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 7 of 14

intermediaryasin(3b),andthecontainerrolecanbeconstruedasapath/locationasin(3a)orasapatientroleasin(3b)(cf.Goldberg2002:344).

AlthoughanalysescouchedwithinCCGprovidestimulatingalternativestononconstructionalaccountsassumingamodulararchitectureofgrammar(e.g.,Chomsky1965,1981),theyoftenremainsilentwhenitcomestodetailingthespecificsofhowdifferentconstructionsinteract.Morespecifically,sentencessuchasWhatdidMichaelsendMiriam?arelicensedbyanumberofdifferentconstructionsbecausetheseconstructionsdonotconflictandcanhenceunify(or,‘fuse’or‘combine’).Asalreadydiscussed,fusionbetweenverbsandArgumentStructureconstructionsispossibleonceconstruction-specificandmoregeneralconstraintsaremet.However,comparativelylittleworkhasbeendoneinCCGonthedetailedconditionsunderwhichothertypesofconstructionscancombinetolicensemorecomplexsentencesinvolvingnotonlyArgumentStructureconstructionsbutalsoothertypesofconstructions.SpecifyingthedifferentrolesofunificationinconstructionalinteractionshasbeenoneofthemainfociofadifferentstrandofConstructionGrammarcalledBerkeleyConstructionGrammar(BCG)(Fillmore,thisvolume)orUnificationConstructionGrammar(Goldberg2006a:213).Withinthistheory,forexample,FriedandÖstman(2004b:71)pointoutthatunificationphenomenafallintodifferentcategories,dependingonthetypesoflinguisticrelationstheyrepresent.Theyidentifyfivedifferenttypesofunification,dependingonthetypeoflinguisticprocessinvolved:agreement(matchininherentpropertiesbetweenstructuralsisters),government(matchinrelationalpropertiesbetweenheadanddependents),semanticlinking(matchbetweenframeelementsandvalenceelements),semanticintegration(semanticunificationbetweenstructuralmotheranddaughter(s)),andvalenceexpansion(incorporationofnonargumentvalenceelements(adjuncts)betweenstructuralmotheranddaughter(s)). Sincedifferentconstructionalapproachesshareagreatdealofimportantinsightsandmechanisms,itshouldinprinciplebefeasibletoadopttheproposalsaboutconstructionalinteractionmadebyproponentsofBCGintoCCG.

13.3.OrganizationofConstructionalKnowledgeinCCG

Allconstructionalapproachestolanguageregardgrammarasnonderivationalandnonmodular,representingknowledgeoflanguageinauniformway.However,therearesomeimportantdifferencesinhowthevariousconstructionalapproachesviewconstructionalorganization.ThefollowingsectionspresentmorespecificsabouthowconstructionalknowledgeisorganizedinCCG,focusingspecificallyonmotivation,constructionaltaxonomies,andproductivity.

13.3.1Motivation

OnecentralpointthatsetsCCGapartfromotherconstructionalapproachesisthatitaimstoofferapsychologicallyrealisticaccountoflanguagebydetermininghowdifferentmoregeneralcognitiveprinciplesservetostructuretheinventoriesofconstructions.InCCG,theexistenceofanyconstructioninthegrammaristhoughttobebyandlargemotivatedbypropertiesofhumaninteractionandcognition,asmanyfacetsofgrammaticalformemergefromsocialinteractionbetweenspeakers.Thisideacomesfromseveralcommonprinciplesofinteractionthatarearguedtohaveinfluencedgrammaticalstructures,suchasiconicity(Haiman1983),reasoningthroughmetaphorandmetonymy(Lakoff1987),categorizationintermsofprototypes(Lakoff1987),categorizationbasedonbasicexperientialpatterns(Johnson1987),andtheperceptionoffigureandground(Talmy2000).WhiletheideathatinteractionbetweenspeakersshapesgrammarhasalsobecomeincreasinglypopularinthegeneralCognitiveLinguisticscommunity(see,e.g.,Cuyckensetal.2003;RaddenandPanther2004),itisimportanttorememberthatmotivationpersedoesnothaveanypredictivepower,butisinsteademployedtoarriveatplausiblescenariosabouthowaparticularlinguisticpatterncametobethatway.Whenitcomestogrammaticalconstructions,motivationisoftenusedtoaccountforthefactthatformallysimilarconstructionsarealsooftensemanticallysimilar.BesidesthePrincipleofMaximizedEconomy,thePrincipleofMaximizedExpressivePower,andthePrincipleofNoSynonymy,thePrincipleofMaximizedMotivationisperhapsthemostinfluentialwhenitcomestomodelinghowconstructionsareorganized:“IfconstructionAisrelatedtoconstructionBsyntactically,thenthesystemofconstructionAismotivatedtothedegreethatitisrelatedtoconstructionBsemantically….Suchmotivationismaximized”(Goldberg1995:67).

Toillustratetheroleofmotivationinstructuringgrammar,considerGoldberg's(2006a:166–82)analysisofSubject-AuxiliaryInversion(SAI)constructions,whichincludeyes/noquestions,(nonsubject)wh-questions,counterfactual

7

8

Page 183: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 8 of 14

conditionals,initialnegativeadverbs,wishes/curses,exclamatives,comparatives,negativeconjuncts,andpositiverejoinders.PreviousaccountssuchasFillmore(1999)andNewmeyer(2000)focusedexclusivelyonformalpropertiessharedbytheseconstructions,withoutpayingmuchattentiontotheirsemanticandpragmaticproperties.GoldbergclaimsthatsuchanalysesonlystipulatetheformofSAIwithoutbeingabletoarriveatfurtherpredictionsorgeneralizations.Tosupportherargument,shepointsoutthatthereexistsasystematicdifferenceinform(subject-auxiliaryinversion)whichsignalsasystematicdifferenceinfunction(adistinctionfromprototypical(positive)sentences).ThisleadshertoarguethattheconstructionsexhibitingSAInaturallyformacoherentfunctionalcategorythathasconventionalizedextensionsradiatingoutfromacentralcore,asillustratedinFigure13.3,wherethepartialsemanticoverlapmetonymicallymotivatesthesyntacticcorrespondencesbetweenthevariousSAIconstructions.

Clicktoviewlarger

Figure13.3. FunctionalcategoryofSAIconstructionswithprototypicalsentenceasitsprototypeandmarkednesslinksmotivatingeachoftheextensionsfromtheprototype(Goldberg2006a:179)

TheprototypeofSAIconstructionsisassumedtobea‘nonprototypicalsentence,’ageneralizationthatisnotdirectlyinstantiatedbySAIconstructions.InGoldberg'sview,SAIconstructionsaredifferentfromothertypesofconstructionsinthattheyshareadominantfeatureofbeingnonpositive(andnondeclarative).InFigure13.3,thisnonpositivefeatureissharedbyeachoftheextensionsfromtheprototype(indicatedbymarkednesslinks(curvedlines)),therebymotivatingthem.Figure13.3alsoshowsthatconstructionswhichdonotsharethisnonpositivefeaturesuchaswh-questions(G)aremotivatedbyconstructionsthatshareit,suchasyes/noquestions(A).Goldberg'sdiscussionofSAIconstructionssuggests“thatfunctionalmotivationsoftenunderlieseeminglyidiosyncraticfactsofgrammar”(2006a:181). Byappealingtomotivationasanexplanatoryfactor(bothinthepsychologicalandthediachronicdomain),CCGthusmakesitpossibletoaccountforsemanticandformaloverlapsofconstructions.CloselyrelatedtotheconceptofmotivationisanotherorganizationalprincipleofCCG,namelythatconstructionswithrelatedformsandfunctionsarelinkedtoeachotherinadefaultinheritancehierarchy.

13.3.2NetworksandInheritanceHierarchies

FollowingLangacker(1987:63–76),CCGregardsconstructionsasformingastructuredinventoryofaspeaker'sknowledgeoftheconventionsoftheirlanguage,andnotarandomcollectionofexceptionsandirregularities.Tomodeltherelationsbetweenconstructions,taxonomicnetworksarepositedwhereeachconstructionconstitutesanodeinthenetworkthatformsacontinuumfromthefullyconcretetothehighlyschematic.InheritancehierarchiesareonecrucialfeatureoftaxonomicnetworksinCCGinthattheyallowbroadgeneralizationstobecapturedbyhigher-levelconstructionswhichareinheritedbyotherconstructions.Atthesametime,subregularitiesarecapturedbypositingconstructionsthatareatvariousmidpointsofthehierarchicalnetwork.Theexistenceofsuchnetworksisassumedtobetheresultofcategorizationwherebothgeneralizationsandmorespecificconventionalinstancesarestoredinanetworkduringlanguageprocessing(fordetails,seeGoldberg2006a:54–64).Figure13.4illustratesapartialtaxonomichierarchyrangingfromveryspecifictoveryschematic.

Atthebottomofthehierarchywefindtwopartiallyfilledidiomaticconstructions,kickthebucketandkickthehabit,whichtypicallyexhibitthesameargumentstructurepatternasgeneralusesoftransitivekick,whichislocatedonelevelhigherinthehierarchy.Theidiomaticconstructionsatthebottomaresaidtoinheritmoregeneralpropertiessuchasverbinflection,phonologicalrealization,andcertainspecificationsregardingthesubject(itshouldbeanimate)fromitsmotherconstruction.Transitivekick,inturn,inheritsitsargumentstructurepatternfromthemoreschematicTransitiveVerbPhraseconstruction,whichinturninheritspropertiesfromthemoregeneralClauseconstruction.

9

Page 184: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 9 of 14

Clicktoviewlarger

Figure13.4. TaxonomicHierarchy(CroftandCruse2004:264).Foreaseofexposition,onlytheconstructions’syntactic(form)sideisrepresented.

CCGassumesthatinheritanceinnetworkscanbepartial,whileotherstrandsofConstructionGrammarassumethatitmustbecomplete(seeLakoff1987:492;Goldberg1995:73;Kay2000:20).Goldberg'snotionofpartialinheritancehastheadvantagethatitallowsforsystematicexceptionsamongspecificinstancesofacategory.Forexample,allverbsinEnglishshareacertainregularsetofmorphologicalproperties:theregularwayofformingpasttenseformsinEnglishinvolvesaconstructionthatcombines-edwithanotherconstruction,namelytherootofverbs.Thisimpliesthateverytimeweencounteraverbwewouldexpectitspasttenseformtohavean-edsuffix,justlikewewouldexpectitsthirdpersonsingularformtohavean-sattachedtoitsroot.Thus,theverbnodeatthetopofthetaxonomichierarchyofverbsislinkedtoallconstructionsthatcanfusewithverbs.Furtherdowninthehierarchywefindsubclassesofverbsthatarelinkedtothetopnodeviainstancelinks.However,sinceirregularverbsmaynotoccurwith-edintheirpasttenseform,theirentries(representedintheformofconstructions)blockinheritancebecauseitconflictswithmorespecificinformation(i.e.,theirregularpastparticipleform).Inthisconnection,Lakoff(1987:483–87)developsarelatedproposalthatregardsgrammaringeneralasaradialcategory,inwhichtheschematicandmoreregularconstructionsconstituteprototypeswhilemorespecificandidiosyncraticconstructionsarelocatedattheperipheryandinheritpropertiesfrommorecentralinstancesofconstructions.

AnotherwayinwhichCCGdiffersfromotherconstructionalapproachesishowitregardstherelationsbetweenconstructionsintaxonomichierarchies.Basedontheideathatconstructionalorganizationfollowssimilarprinciplesasconceptualcategories,CCGpositsavarietyoflinksbetweenconstructions.Onesuchlinkistheso-called‘subpartlink,’whichshowsthatoneconstructionisapropersubpartofanotherconstructionandexistsindependently.AnexampleistheIntransitiveMotionconstruction(Kimran),whichisrelatedtotheCausedMotionconstruction(KimranPatoffthestreet)byasubpartlink(Goldberg1995:78).Anotherlinkisthe‘instancelink,’whichshowsthataconstructionisaspecialcaseofanotherconstructioninthesensethatitisamorefullyspecifiedversionoftheotherconstruction.AnexamplethereofisaparticularsenseofdriveasinKimdroveFredcrazy/mad/bonkers/upthewall,whichoccursonlyintheResultativeconstruction,andwhichislinkedtoitbyaninstancelink(Goldberg1995:79–81;seealsoGoldbergandJackendoff2004:535–36).

BasedonkeyinsightsfromLakoff's(1987)analysisofThere-constructions,CCGalsopositsaparticulartypeofinheritancelink,namelymetaphoricalextensioninheritancelinksthatrepresentparticularmetaphoricalextensionsbetweenconstructions.Forexample,Goldberg(1995:81–89)arguesthattheResultativeconstruction(JoekickedBobblackandblue)isametaphoricalextensionoftheCausedMotionconstruction(Joekickedthebottleintotheyard),wherethemetaphor‘ChangeofStateasChangeofLocation’accountsfortherelationbetweenthesemanticsofthetwoconstructions.WhiletheResultativeconstructioninheritsthesyntacticspecificationsofthemetaphoricalextensionfromtheCausedMotionconstruction,itisstilladistinctconstructionwithitsownsetofconstraints(seeGoldberg1995:87–99,164–74,193–97).

AnotherimportantconstructionlinkinCCGistheso-calledpolysemylink,whichrepresentsrelationsbetweensubtypesofconstructionsthatexhibitthesamesyntacticspecificationsbutdifferintheirsemantics.AnexampleisGoldberg's(1995:75)analysisoftheDitransitiveconstruction,whichhasacentralsenseof‘XCAUSESYTORECEIVEZ’(JoegaveSallytheball)associatedwithparticularverbclassessuchasverbsthatinherentlysignifyactsofgiving(suchasgive,pass,hand),verbsofinstantaneouscausationofballisticmotion(suchasthrow,toss,slap),andverbsofcontinuouscausationinadeicticallyspecifieddirection(suchasbring,take,etc.).Inaddition,theDitransitiveconstructionhasatotaloffiveextendedsensesthatareeachlinkedtothecentralsensebypolysemylinks,whichinheritthesyntacticconstructionschemafromtheprototype,andwherethesenseextensionsarealsoassociatedwithspecificverbclasses:(1)conditionsofsatisfactionimply‘XCAUSESYTO

10

11

Page 185: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 10 of 14

RECEIVEZ’(verbsofgivingwithassociatedsatisfactionconditions,e.g.,JoepromisedBobacar),(2)‘XCAUSESYNOTTORECEIVEZ’(verbsofrefusal,e.g.,JoerefusedBobacookie),(3)‘XACTSTOCAUSEYTORECEIVEZatsomefuturepointintime’(verbsoffuturetransfer,e.g.,JoebequeathedBobafortune),(4)‘XENABLESYTORECEIVEZ’(verbsofpermission,e.g.,JoepermittedChrisanapple),and(5)‘XINTENDSTOCAUSEYTORECEIVEZ’(verbsinvolvedinscenesofcreationandverbsofobtaining,e.g.,JoebakedBobacake).Positingconstructionalpolysemyhasanumberofadvantages,suchasnothavingtopositlexicalrulesinordertoaccountforsenseextensionsofverbswhosevarioussensesarenotpredictableongeneralgroundsandmustbeconventionallyassociatedwiththeconstruction.Putdifferently,insteadofpostulatingverbsenseshiftsintermsoflexicalrules(Pinker1989)oreventstructureaugmentations(HovavandLevin1998),thedifferenttypesofDitransitiveconstructionsexistindependentlyoftheparticularlexicalitemsthatinstantiatethem.Theconstructionalviewreflectsthegeneralideathatasetofconstructionsdoesnotconsistofindependententitiesthatexhibitirregularorganizationalpatterns,butisinsteada“highlystructuredlatticeofinterrelatedinformation”that“displayprototypestructuresandformnetworksofassociations”(Goldberg1995:5).

13.3.3FrequencyandProductivity

Clicktoviewlarger

Figure13.5. Theproductivitycline(Barðdal2008:172)

CCGisusage-basedinthatitallowsbothinstancesandgeneralizationstobecapturedintermsoffullyarticulatedschematicnetworksincludinglow-levelschemas(Langacker1987;BarlowandKemmer2000).Inthisview,itisalsopossiblethatpatternsoccurringwithsufficientfrequencyarestoredasconstructionsalongsidemoregenerallinguisticgeneralizationsevenwhentheyarefullycompositionalandpredictable(BybeeandHopper2001;Goldberg2006a;Bybee,thisvolume).InCCG,frequencyisalsoimportantforaccountingforaconstruction'sproductivity,thatis,thespeaker'sabilitytoextendArgumentStructureconstructionstonewverbsandtoavoidovergeneralizations.Theideatousefrequencytoaccountforaconstruction's(potential)distributioncomesfromresearchinmorphologyshowingthattypefrequencydeterminestheproductivityofaschema(cf.Bybee1985).Itismeasuredinthecontextofaconstructiontodeterminehowmanydifferentitemsoccurinthevariousschematicslotsofaconstruction.Determiningaconstruction'stypefrequencyisimportantbecauseincreasedtypefrequencyhasbeenshowntodirectlycorrelatewithaconstruction'sabilitytooccurwithnovelitems.Forexample,theWayconstruction(Goldberg1995:199–218)isratherproductiveasitappearswithanextensivenumberofverbs,whiletheResultativeconstructionisconsiderablymorelimitedwithrespecttothetypesofverbswithwhichitcanoccur(GoldbergandJackendoff2004;Boas2005,2011a;Jackendoff,thisvolume).Tocapturethisstateofaffairs,constructiondescriptionswouldthusalsoincludeinformationaboutthetypefrequencyfortheverbslot,inadditiontomoregeneralsemanticconstraints(see,e.g.,Goldberg1995:129–38,2006a:98–99).Otherimportantaspectsinfluencingaconstruction'sproductivityaretokenfrequency,whichdeterminesthedegreeofentrenchmentofindividualsubstantivewordforms(seeCroftandCruse2004;Bybee,thisvolume),degreeofopenness(i.e.,thevariabilityoftheitemsoccurringinaparticularpattern),andstatisticalpre-emption(therepeatedwitnessingofthewordinacompetingpattern)(Goldberg2006a:93).

Morerecently,Barðdal's(2008)studyofcaseandArgumentStructureconstructionsinIcelandichasshownthatproductivityisbestregardedasafunctionoftypefrequency,semanticcoherence,andtheinversecorrelationbetweenthetwo.Basedonhistoricalandpsycholinguisticdata,BarðdalproposestheproductivityclineinFigure13.5whereconstructionslocatedatthetoparenotonlythemostproductiveconstructions(andoccurwithhigh

12

13

Page 186: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 11 of 14

typefrequency)butarealsothemostgeneralandregular.Constructionslowestintypefrequencycaneitherexhibitaloworahighdegreeofsemanticcoherence.Barðdal'sproposalsareinspiringbecausetheymakeitpossibletoregardbothfullproductivityandanalogyas“twosidesofthesamecoin,notdifferentinontologicalstatus,onlydifferentindegree”(Barðdal2008:173).

13.4.DifferencesbetweenCCGandOtherStrandsofConstructionGrammar

CCGsharesagreatdealwithotherstrandsofConstructionGrammar,suchasthecentralroleofconstructions,theideathatthearchitectureoflanguageisnonmodularandnonderivational,andthatconstructionsarelearnedonthebasisofinput.Atthesametime,therearesomeimportantdifferencesthatareindicatedbythelabelcognitive,signalingthecloseconnectiontoCognitiveGrammar(CG)(cf.Broccias,thisvolume)andCognitiveLinguisticsingeneral,aswellasRadicalConstructionGrammar(RCG)(Croft,thisvolume).Theseapproachesareallusage-based,placeheavyemphasisonpsychologicalplausibility,employCognitiveSemantics(construal,adynamicviewofmeaningincontext,etc.),regardmotivationasacentralconcept,anddeveloponlyrelativelysparseformalizations.

OnemajoraspectthatsetsCCGapartfromSign-basedConstructionGrammar(SBCG)(Sag2010,2011;BoasandSag2012;Michaelis,thisvolume)andtheframeworkinspiringit,BerkeleyConstructionGrammar(BCG)(FillmoreandKay1995;Fillmore,thisvolume),concernsthegoalofofferingapsychologicallyplausibleaccountoflanguagebydetermininghowvariousgeneralcognitiveprinciplesservetostructuretheinventoriesofconstructions.WhereasinCCGtheexistenceofconstructionsinthegrammararethoughttobemotivatedbymoregeneralpropertiesofhumaninteractionandcognition,BCGandSBCGdonotemphasizetheroleofmotivation.Inthisconnection,mostworkinBCGandSBCGisaimedatfindingmaximalgeneralizationswithoutanyredundancy.Thismeansthatifaparticularexpressioncanbeaccountedforonthebasisofconstructionsalreadyknowntoexist,thenthereisnoneedtopostulateaseparateconstruction.Similarly,BCGandSBCGarenotconcernedwithfrequenciesofconstructions.Incontrast,CCGtakesastrongusage-basedviewoftheroleoffrequencyandthestatusofitem-specificinstances,leadingtotheideathatevenfullyregularpatternsmaybestoredalongsidemoreabstractschematicconstructionswhentheyoccurwithsufficientfrequency(fordetails,seeCroftandCruse2004:308–18;Goldberg2006a:45–65;Bybee,thisvolume).

Anotherkeydifferenceistheroleofformalexplicitnessandmaximalgeneralizations.CCGhasnotfocusedonrigidformalizationsoflinguisticinsights,asitseekstorepresentlinguisticknowledgeinsuchawaythatitcaninterfacetransparentlywiththeoriesofprocessing,acquisition,andhistoricalchange(Goldberg2006a:215).Tothisend,formalizationinCCGiskepttoaminimumandtypicallytakestheformofboxeddiagramsrepresentingArgumentStructureconstructionsasinFigure13.3above,togetherwithconstraintsinprosespecifyingthesemanticandpragmaticconditionsunderwhichaconstructionmayfusewithaverb.Theadvantageofthisapproachisthatitisrelativelyflexibleanddoesnotimposeanystringentformalization.Apossibledrawbackisthatsuchatypeofformalizationmaysometimesbeabitvagueaboutsomeissues.Incontrast,BCGhastraditionallybeenconcernedwithdetailedunification-basedformalisms.EachconstructionisrepresentedintermsofanAttribute-ValueMatrix(AVM)whereeachattributecanhaveatmostonevalue(fordetails,seeKayandFillmore1999;FriedandÖstman2004b;Fillmore,thisvolume).Sometimessuchdetailedformalizationsarethoughtofastoorigidbecausetheymakeitdifficulttocapturedetailedlexicalsemanticproperties(cf.Goldberg2006a:216).However,SBCG,combiningkeyinsightsfromBCGandHead-drivenPhraseStructureGrammar(PollardandSag1994),offersatheoryofconstructionalmeaningbecauseitassumesthatcombinatoricconstructionsaredirectlyassociatedwithinterpretiveanduseconditions,expressedbysemanticandpragmaticfeatures(Michaelis,thisvolume).Morerecently,FillmoreandhisassociateshavebeendevelopingaunifiedanalysisthatregardsconstructionsasanaturalextensionofthelexiconinSBCG.UsingthesameformalismemployedforrepresentinglexicalentriesinFrameNet(Fillmoreetal.2003;FillmoreandBaker2010;http://framenet.icsi.berkeley.edu),constructionsalsohavetheabilitytoevokesemanticframestodifferentextentsandatdifferentlevelsofschematicity(Fillmore,Lee-GoldmanandRhomieux2012).

13.5.Conclusions

ThischapterpresentedthemainorganizingprinciplesofCCGasdevelopedbyGoldberg.WhileresearchinCCGismostprominentlyknownforthecentralroleofArgumentStructureconstructions,thischapterhasshownthatother

Page 187: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Page 12 of 14

typesofconstructionsalsoplayapivotalroleinGoldberg'sframework.Whatisperhapsmostintriguingaboutconstructionalapproachesingeneralisthatwhiletheyshareagreatdealofgeneralconceptsandprinciples,theyalsohavequitedistinctiveviewsregardingpsychologicalplausibility,theroleofmotivation,andtheimportanceofrigidformalizations(orthelackthereof).Whileonthesurfacesuchdifferencesmayseemrathersignificant,itisimportanttorememberthatthedifferentconstructionalapproachesdonotpresentanyinherentandsystematiccontradictionswhenitcomestotheorganizingprinciplesoflanguage(Goldberg,thisvolume).Mostoften,researcherssimplydifferinwhattheyseeastheimportanceofagivenaspectofconstructionalorganization.Suchpreferencesareoftenamatterofpersonalchoiceandinterest.Forexample,whenitcomestoformalizationwehaveseenthatCCG(likeRCG)doesnotgointotoomanydetailswhileBCGandSBCGgotogreatlengthstoprovideadetailedandrigidnotationthatleavesrelativelylittlespaceforanyotherinterpretationsthantheonecapturedbytheirparticularformalism.Itisexactlythesedifferencesthatkeepconstructionalresearchsovibrant:ConstructionGrammarianschoosetheirpreferredmethodforfindingandexpressingtheirlinguisticinsights.Assuch,manyusefulinsightsfromCCGcaninprinciplebe‘translated’intothenotationallymorestringentformalismsemployedbyBCGandSBCG(andviceversa)becausethesedifferentconstructionalapproachesshareacommonbaseofconceptsandprinciples.Tothisend,researchinCCGiscertaintocontinuegeneratingexcitinginsights.

ThankstoFranciscoGonzálvez-García,SeiziIwata,JaakkoLeino,MarcPierce,ananonymousreviewer,andtheeditorsforextensivecommentsonearlierversionsofthischapter.IthanktheAlexandervonHumboldtfoundationforafellowshipforexperiencedresearchers,whichsupportedthewritingofthispaper.Theusualdisclaimersapply.

Notes:

(1.)ConstructionGrammarevolvedoutofFillmore'searlierworkonCaseGrammar(1968,1977a)andFrameSemantics(1982,1985b).Inthe1980s,FillmoreandKaycoinedtheterm“ConstructionGrammar”(Fillmore1988,1989;Fillmore,Kay,andO’Connor1988;FillmoreandKay1995;Fillmore,thisvolume),which—togetherwithLakoff's(1987)accountofexistentialconstructions—canberegardedasthefoundationforthedifferentversionsofConstructionGrammarfoundtoday.

(2.)Forotherdefinitionsofconstructions,seeGoldberg(1995:4),Croft(2001:17–21),andFriedandÖstman(2004b:18–23).

(3.)ConstructionGrammarsharesthisinterestinaccountingforallfacetsofthelexiconandsyntax/semanticswithotherframeworks,suchasValencyTheory(HelbigandSchenkel1971;Helbig1992;Welke1988,2009;HerbstandKötz-Votteler2007),PatternGrammar(Sinclair1987;HunstonandFrancis1999),andWordGrammar(Hudson1990;HolmesandHudson2005).ForparallelsbetweenCCGandEuropeanfunctionallinguistics,seeGonzálvez-GarcíaandButler(2006).

(4.)Whenaverbislexicallyassociatedwithtwoparticipantroles(inthecaseofpush:<push,pushee>),theCausedMotionconstructiononlycontributesoneargumentrole,namelythegoalrole.Incasesinwhichaverbislexicallyassociatedwiththreeparticipantroles(put:<putter,put.place,puttee>;cf.Goldberg1995:52),“theconstructionalmeaningisentirelyredundantwiththeverb'smeaningandtheverbmerelyaddsinformationtotheeventdesignatedbytheconstruction”(1995:51).

(5.)Fordiscussionoftheroleofabstractschematicconstructions,seealsoCroft(2003),GoldbergandJackendoff(2004),Iwata(2008),Traugott(2008d),Zeschel(2008a),andBoas(2011a).SeealsoFillmoreetal.(2012)oncurrenteffortstousedescriptiveandorganizationalprinciplesoftheFrameNetlexicon(http://framenet.icsi.berkeley.edu)toarriveatasystematicdescriptionandanalysisofgrammaticalconstructionsofdifferenttypesandlevelsofabstraction.

(6.)Constructionsarenotcapableofcombiningfreelywhentheyareinconflict,forexample,theconstraintsonindividualconstructionsareviolated.Consider,forexample,*WhatdidMichaelsendthepeace?,whichisunacceptablebecausethepeacecannotbeconstruedasananimaterecipientargumentandhenceviolatesacentralconstraintoftheDitransitiveConstruction.

(7.)Foralternativeanalyseshighlightingtheimportanceofmoredetailedframe-semanticinformationtoaccountforalternatingbehavioramonglargernumberofsemanticallyrelatedverbs,seeNemoto(2005),Iwata(2008),and

Page 188: The Oxford Handbook of Construction Grammar

Cognitive Construction Grammar

Boas(2011b).

(8.)Fordetails,seeZwicky(1994),FillmoreandKay(1995),KayandFillmore(1999),Kay(2002b),andFriedandÖstman(2004b).

(9.)AnalternativewayofaccountingfortheapparentrelatednessoftheSAIconstructionswouldnotinvolvetheconceptofmotivationbutinsteadaccidentalhistoricalfactors:theseconstructionsallhadtheverbbeforethesubjectforpragmaticreasonsinearlierversionsofEnglish,orhadtheverbinsecondposition.Allthemainverbsmovedoutofthepositionasdodevelopedasanauxiliary.Suchanalternativeexplanationwouldnotpositanysemantic/pragmaticorpsychologicalrelationamongtheseconstructions,thatis,noneoftheseconstructionsdevelopedonthebasisofanyoftheothers.Assuch,theywouldnotberegardedas‘extensions’inanydiachronicsense.Ithankananonymousreviewerforthisimportantobservation.

(10.)Itisalsopossiblethatsomeconstructionsexhibitmultipleinheritance,whichhelpscapturethefactthatinstancesofsomeconstructiontypesseemtoresistbeinguniquelycategorizedinanaturalway(Goldberg1995:97).SeealsoMichaelisandLambrecht(1996:237–38),CroftandCruse(2004:276),andGoldbergandJackendoff(2004:563).

(11.)Boas(2003:94–97)arguesthatthereisnoneedforametaphoricalextensionlinkbetweentheCausedMotionandtheResultativeconstructions,becauseaverb'sabilitytooccurwitheitherpatternisapropertythatislexicallyassociatedwitheachindividualverbwithoutaconstructionhavingtoaddadditionalargumentstoitssemantics.SeealsoKay(2005)forsimilararguments.

(12.)Boas(2003:97–104)proposestodoawaywithconstructionalpolysemybecause—leavingthenotionofmotivationaside—thereisnoneedforitastherelevantsyntacticframesandpolysemypatternsaredirectlyassociatedwithverbclasses,individualverbs,andtheirlexicalsemanticnetworks(seealsoCroft2003andBoas2008,2010b).Onthisview,constructionaldistributioncanbeaccountedforonthebasisofverbsalone,withouthavingtorelyonavarietyofrelatedArgumentStructureconstructions.Kay(2005)makesasimilarpointbyproposingthatthenotionofconstructionalpolysemyislargelyredundantbecausemanyofthedistinctionsinentailmentsfollowfromthesemanticsoftheverbalone.Instead,Kay(2005:76–87)developsthreemaximalRecipientconstructionstoreplaceGoldberg'ssixsensesoftheDitransitiveconstructions.Foradditionalpositionsonconstructionalpolysemy,seeWierzbicka(1988:chapter5),Jackendoff(1997a:100),Croft(2001:117),Goddard(2000:132–40),andCroftandCruse(2004:274).

(13.)SeealsoStefanowitschandGries(2003)andStefanowitsch(thisvolume)forquantitativecorpus-linguisticmethodsforsystematicallyanalyzingtherelationshipsbetweenwordsandthegrammaticalpatternsinwhichtheyoccur.

HansC.BoasHansC.BoasisanAssociateProfessorintheDepartmentofLinguisticsandtheDepartmentofGermanicStudiesattheUniversityofTexasatAustin.BeforecomingtoAustin,hewasapostdoctoralresearcherwiththeFrameNetproject(http://framenet.icsi.berkeley.edu)attheInternationalComputerScienceInstituteinBerkeley,California.Hispublicationsincludetwomonographs:AConstructionalApproachtoResultatives(CSLIPublications,2003)andTheLifeandDeathofTexasGerman(DukeUniversityPress,2009),aswellasfour(co-)editedvolumesandvariousarticlesonConstructionGrammarandFrameSemantics.HeisthedirectoroftheTexasGermanDialectProject(http://www.tgdp.org)andtherecipientofthe2007HugoMoserPrizeforGermanicLinguistics(InstitutefortheGermanLanguage,Mannheim),aswellasthe2011LeonardBloomfieldBookAward(LinguisticSocietyofAmerica).

Page 189: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0014

MorphologyinConstructionGrammarGeertBooijTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterpresentsawholerangeofargumentsforaConstructionGrammarapproachtomorphology.Itshowsthatthelexiconcontainsboth(simplexandcomplex)wordsandwordformationschemasofvariousdegreesofabstraction,andprovidesevidencesupportingtheviewthatmorphologicalconstrucitonshaveholisticproperties.Thechapterconsidersbothwordformationandinflectionalmorphologyanddiscussestherelationshipbetweenmorphologicalandsyntacticconstructions.Thefindingsconfirmthatthelexiconistobereinterpretedasthe"constructicon,"astructuredandhierarchicallyorderedarrayofconstructionsandconstructswithphrasalorwordstatus.

Keywords:ConstructionGrammar,morphology,lexicalitems,constructionalschemas,wordformation,inflectionalmorphology,syntacticconstructions,constructicon

14.1Booij.Introduction:MorphologicalConstructions

Morphologyisthesubdisciplineoflinguisticsthatdealswiththesystematicrelationshipsbetweenformandmeaningoflinguisticconstructsatthewordlevel.TheframeworkofConstructionGrammarmaybeexpectedtoprovideanadequateframeworkformorphologicalanalysis,sinceConstructionGrammartakesasitsstartingpointthat“thegrammarrepresentsaninventoryofform-meaning-functioncomplexes,inwhichwordsaredistinguishedfromgrammaticalconstructionsonlywithregardtotheirinternalcomplexity”(MichaelisandLambrecht1996:216).ThisapproachtomorphologycanbereferredtoasConstructionMorphology,whichisalsothetitleofarecentmonograph(Booij2010)thatprovidesarangeofargumentsforthiswayofdoingmorphology.

Letmefirstgiveanexampleofamorphologicalconstruction.Englishfeaturesalargesetofdeverbaladjectivesendinginthesuffix-able,suchasaccept-able,afford-able,approach-able,believ-able,anddo-able.Themeaningoftheseadjectivescanbecircumscribedinformallyas‘canbeV-ed,’whereVstandsforthemeaningofthebaseverboftheadjective. Theverbsintheseadjectivesaretransitiveverbs,andtheobjectargumentsoftheseverbscorrespondtothesubjectargumentsoftherespectivederivedadjectives.Onthebasisofthesystematicform-meaningcorrespondencesbetweenasetofsuchadjectivesandthesetofcorrespondingverbs,theEnglishlanguageusercanformanabstractschemathatexpressesageneralizationconcerningtheform-meaningcorrespondencesthatholdfortheseadjectives:

(1)[Vtr -able] ↔[[CANBESEM-ed] ] (Vtr=transitiveverb,SEMstandsforthemeaningoftherelevantconstituent)

ThenotationthatIusehereisinessencethatproposedbyJackendoff(2002a,thisvolume).Thedoublearrowexpressesthattherearecorrespondencesbetween(partsof)theformalandthesemanticrepresentation,expressedbyco-indexation.Thesemanticrepresentationisapartiallyinformalone.Therepresentationontheleftofthedoublearrowistheformspecificationofthedeverbaladjectives.Theformalrepresentationofaword

1

i Aj i PROPERTY j

Page 190: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 2 of 14

comprisestwosubrepresentations,amorphosyntacticrepresentationandaphonologicalone(Jackendoff2010,2011),whichareconflatedhereintoonerepresentationforeaseofexposition.

Inthischapter,IwillprovidearangeofargumentsforaConstructionGrammarapproachtomorphologyasintroducedbythebriefcharacterizationofEnglishadjectivesin-ableabove.Insection14.2,itisarguedthatbothcomplexwordsandtheabstractschemasthattheyinstantiatemustbespecifiedinthelexicon.Therearedegreesofschematicity,whichimpliestheideaofahierarchicallexicon.InthisrespectmorphologyisstrikinglysimilartothewaythatConstructionGrammardealswithsyntacticconstructsandconstructions.Section14.3showsthatmorphologicalconstructionsmayhaveholisticproperties.Thismeansthatthemeaningofamorphologicalconstructionmustindeedbespecifiedasapropertyoftheconstructionasawhole.

Adesirableimplicationofthisviewofthelexiconasarichlystructuredcomponentofthegrammar(insteadofjustalistofidiosyncraticinformation)isthatitispossibletospecifyandrefertoparadigmaticrelationsbetweencomplexwordsandbetweenmorphologicalconstructions.Insection14.4,itisshownthatweneedthistypeofrelationshipforaproperaccountofwordformation.Thenotionofparadigmaticrelationshipalsoplaysanessentialroleininflectionalmorphology(section14.5),andinflectionalmorphologyisshowntostronglysupportthebasicconceptionsofConstructionMorphology.

Therelationshipbetweenmorphologyandsyntacticconstructionsisthetopicofsections14.6and14.7.Insection14.6,itisarguedthatthedistributionofboundmorphemesmaydependontheiroccurrenceinspecificsyntacticconstructions.Thatis,boundmorphemesmayfunctionasmorphologicalmarkersofconstructions.Section14.7showsthatmorphologicalandphrasalconstructionsmayhaveidenticalfunctionsinthegrammar,andhencecompete,sincethereareproductiveclassesofphrasallexicalunits.

Finally,section14.8presentsasummaryoffindingsandconclusionsconcerningmorphologyinConstructionGrammar.

14.2.TheHierarchicalLexicon

Schema(1)summarizesthepredictablecommonpropertiesoftheadjectivesendingin-ablementionedabove.Hence,schema(1)dominatestheindividuallistedexistingadjectivesin-able.

Atthesametime,theschemaspecifieshownewadjectivesofthistypecanbederivedfromtransitiveverbs.Therequiredtransitivityoftheverbalbaseisconfirmedbyadjectiveslikeplayableanddebatable,whichselectatransitivesenseofthepolysemousverbsplayanddebate.Moreover,theV-ableconstructioncancoerceatransitiveinterpretationofintransitives,asinlaughableandunlistenable.

Byspecifyingboththeindividualexisting-ableadjectivesandtheproductiveschema,weavoidthe‘ruleversuslistfallacy,’themistakenideathatalinguisticconstructthatiscompletelyregularandhencepredictablecannotatthesametimebelisted.Listingisawayofspecifyingthelexicalconventionsofalanguage,whereasschemasexpressthegenerativepowerofthegrammar.Languageusersacquireabstractmorphologicalschemasonthebasisoftheindividualcomplexwordsthattheyhaveacquiredinthefirstplace.

Therelationbetweenaschemaanditsinstantiationscanbemodeledinahierarchicallexiconbymakinguseofthemechanismofdefaultinheritance:theindividual-ableadjectivesinheritalltheirpredictablepropertiesfromtheirdominatingschemabydefault.Byusinginheritanceasadefaultmechanism,weallowforindividual-ableadjectivestohaveidiosyncraticpropertiesthatdifferfromwhatispredictedbytheschema.Forinstance,theadjectiveagreeableisderivedfromtheverbagree,whichisnottransitive,butrequiresaprepositionalobject(agreewith/on/to).Inthecaseofapplicable,theformoftheverbstem,applic-isnotidenticaltothatoftheverbapply.Theadjectiveclubbable‘worthyofmembershipofaclub’isderivedfromthenounclub,notfromaverb.Finally,therearealsoadjectivesin-ablewiththerelevantmeaningtype,butwithoutabaseverbbeingavailable,adjectivessuchasamenable,ineluctable,andtractable.Thisimpliesthatthepartbefore-ablecannotbeco-indexedwithanindependentverb.Hence,therepresentationsoftheseadjectivescontainlesspredictable,redundantinformation,sincetheydonotinheritpartoftheirmeaningfromthemeaningofacorrespondingbaseverb.

Defaultinheritancecanbeinterpretedintwoways.Intheimpoverishedentrytheory(Sagetal.2003:chapter8)all

Page 191: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 3 of 14

predictableinformationconcerningalexicalentryisomittedfromthatentryitself,andspecifiedintheschemasorconstraintsthatdominateit.Thelexicalentryinheritsthesepropertiesfromitsdominatingnodes.Inthefullentrytheory(advocatedinJackendoff1975)lexicalentriesarefullyspecified,andinheritanceisamechanismforcomputingtheamountofnonindependentinformation.Thislattervariantistobepreferred,becauseabstractmorphologicalschemasareconstructedonthebasisoflexicallystoredcomplexwords.Thereisnoevidencethat,oncealanguageuserhasacquiredtheabstractschema,informationinthelexicalentriesforindividualcomplexwordsisdeleted.

Schema(1)accountsforaproductivewordformationprocess.Forinstance,sincethenounSkypehasbeenconvertedintoanewverbtoskype,wemaynowalsosaythatweareskypable,asaGooglesearchwillconfirm.Hence,ithastobespecifiedthatthisschemaisnotjustageneralizationoverasetofexistingwords,butcanbeusedproductivelyaswell(seeJackendoff2010:28–34fordiscussion).Thisschemathusrepresentsamorphologicalconstruction,andtheindividual-ableadjectivesaremorphologicalconstructs.Inthismorphologicalconstruction,oneconstituentisspecified(-able),whereastheotherconstituentisanopenposition,avariablefor‘transitiveverb.’Hence,thismorphologicalschemaisa‘constructionalidiom’atthewordlevel.Aconstructionalidiomisa(syntacticormorphological)schemainwhichatleastonepositionislexicallyfixed,andatleastonepositionisvariable.Themorpheme-ableisaboundmorpheme.Itdoesnotoccurasalexicalitembyitself,butonlyaspartofthelistedadjectivesofthistype,andofthewordformationschema(1).

Thepsychologicalrealityofmorphologicalschemasisshownbytheproductiveuseofthecorrespondingwordformationprocesses.Childrenareabletodiscoverabstractpatternsinsetsofcomplexwords,andusethesepatternsproductively(Mos2010).Thisdoesnotmeanthatallwordformationprocesseshavethesamedegreeofproductivity:thereisclinefromunrestrictedproductivitytoextensionofapatternbasedonaspecificwordasamodel,theprototypicalcaseofanalogy.Productivitycorrelateswithtypefrequency,andthuswithdegreeofentrenchmentofaschema(Baayen2003;HayandBaayen2005;Barðdal2008,2011a).

Sincemorphologicalschemasdefinehownewcomplexwordscanbecoined,onemaywonderwhethertheassumptionofanabstractschemanecessarilyimpliesproductivityoftherelevantmorphologicalprocess.Thisisnotthecase.Forinstance,weshouldbeabletoexpressthatEnglishcompoundsareallright-headed,includingtherestrictedsetofverbalcompoundssuchastobabysitandtohead-hunt,coinedbymeansofbackformationfrombabysitterandhead-hunter,respectively(verbalcompoundingisnotproductiveinEnglish;Lieber2009).Similarly,thesuffix-shipisclearlydiscernibleinwordslikefriendshipandleadership,yetitcannotbeextendedtomanagership.Hence,wemustbeabletostateasanindependentpropertytowhatextentandinwhichsubdomainsaparticularschemacanbeusedproductively.

Themostcommontypeofwordformationinthelanguagesoftheworldiscompounding,thejuxtapositionoftwowordstoformanewone(Bauer2009).English,likeallGermaniclanguages,hasaproductiveprocessformakingright-headedcompounds,suchasNNcompounds.ThemostabstractschemaforEnglishcompoundsmaythereforelookasfollows:

(2)[X Y ] ↔[SEM withrelationRtoSEM ]

Inthisschemathestructureontheleftofthearrowexpressesthatthecategoryofthewordasawholeisidenticaltothatoftherightconstituent,sincetheysharethecategoryvariableY.Theschemaexpressesthroughco-indexationthattheformalheadofthecompoundisalsoitssemantichead.Awindmill,forinstance,isatypeofmill,notatypeofwind,anditdenotesamillwhichhassomesemanticrelationwiththeleftconstituentwind,namely‘poweredby.’ThisspecificinterpretationoftherelationvariableRisprovidedbyconceptualandencyclopedicknowledge,andisconventionalizedforexisting,listedcompounds.

Schema(2)dominatesvarioussubschemas.Forinstance,thereisasubschemaforNVcompoundsthatdominatesverbslikehead-hunt,andhastobequalifiedasunproductive.Ontheotherhand,thesubschemaforNNcompoundsisproductive.Subschemasmaythusdifferintheirdegreeofextensibility(Barðdal2008,2011a).

Additionalevidencefortheclaimthatweneedsubschemasforsubclassesofcompoundinginwhichoneoftheconstituentsislexicallyspecified,isfoundintheexistenceof‘affixoids,’thatis,wordsthatareaffix-likeinthattheyhaveaspecificboundmeaningwhenformingpartofacompound(Booij2005;VanGoethem2008).InDutchNNandNAcompounds,whichareright-headed,themodifierconstituentmaydevelopamoreabstractmeaningof

i j Yk Yj Xi k

Page 192: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 4 of 14

intensification.Forinstance,thenounreuze(citationformreus‘giant,’herefollowedbythelinkingelement-e),combineswithbothnounsandadjectives,andhasacquiredthemoreabstractmeanings‘great’and‘very,’respectively:

(3)a.reuze-Nreuze-idee‘greatidea’reuze-kerel‘greatguy’reuze-mop‘greatjoke’reuze-plan‘greatplan’b.reuze-Areuze-aardig‘verykind’reuze-leuk‘verynice’reuze-gemeen‘verynasty’

ThisimpliesthattheDutchlexiconcontainsconstructionalidioms,subcasesofNNandNAcompounding,withthefirstconstituentlexicallyspecified:

(4)a.[[reuze] [x] ] ↔[great SEM]b.[[reuze] [x] ] ↔[very SEM]

AsdiscussedindetailinVanGoethem(2008),theDutchadjectiveoud‘old’hasacquiredthemeaning‘former’whenusedaspartofacompound,asin:

(5)oud-docent‘lit.old-teacher,formerteacher’oud-student‘lit.old-student,formerstudent’oud-strijder‘lit.old-warrior,warveteran’

Thisuseofoudrequiresaconstructionalidiom,asubcaseofANcompounding:

(6)[[oud] [x] ] ↔[former SEM]

AsimilartypeofsemanticdevelopmentcanbeobservedforMandarinChinese(Arcodia2011).ArcodiaobservesthattheEnglishloanmorphemebā‘bar’hasdevelopedamoregeneralmeaninginChineseright-headedcompounds,asillustratedin(7):

(7)bōli-bā‘glass-bar,aglassworkshopwherecustomersmaycreatetheirownproducts’bù-bā‘cloth-bar,atextileworkshopwherecustomersmaycreatetheirownproducts’wàiyuˇ-bā‘foreignlanguage-bar,a(virtual)placeforlearningforeignlanguages’yóuxì-bā‘game-bar,amusementarcade’

Arcodiaalsopointsoutthatwhenthewordbāisusedastheleft,non-headconstituentofacompound,ithastheregularmeaningofEnglishbar.Therefore,hegeneralizesthefollowing,boundgeneralizedmeaningofbā:‘place(actualorvirtual)wheresomekindofservicerelatedtoN (theleftconstituent)isofferedorwhereinformationrelatedtoN maybeexchanged.’Thismeaningmustbespecifiedinaconstructionalidiom,withtheheadN oftheNN compoundspecifiedasbā.

Insum,thepervasivephenomenonofgeneralizationsforsubclassesofcomplexwordsrequiresahierarchicallexicon.Moreover,suchsubschemasexplainwhypartsofcompoundscandevelopintorealaffixes(Booij2005):wordsreceiveaspecial‘bound,’oftenmoreabstractinterpretation,withincompounds.Whentheactuallexemegetslost,itscorrespondingboundversionmaysurviveasanaffix.Forinstance,theEnglishaffixes-hoodand-shipderivefromnounsthatwereoncetheheadconstituentsofcompounds.

Theriseofsuchaffixoidsmeansthatalexicalworddevelopsintoawordwithamoregeneralizedmeaningandexhibitscontextexpansion:awordattachestoalargerclassofwordsthanitsoriginalsemanticswouldjustify.TheDutchwordreuze‘giant,’forinstance,notonlyattachestoconcreteentitieswithaphysicalsize,buttoallkindsofabstractnouns,astheexamples(3)illustrate.Thispatternofchangemayalsoleadtothedevelopmentofgrammaticalmeaningsforsuchaffixoids,andhencetogrammaticalization(cf.Himmelmann2004fordiscussionoftherelationbetweenlexicalizationandgrammaticalization).

Ni Nj Nk i j k

Ni Aj Ak i j k

Ai Nj Nk i j k

i

i j

i j

Page 193: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 5 of 14

14.3.HolisticProperties

AnessentialargumentforConstructionGrammaristhataconstructionmayhaveholisticpropertiesthatarenotderivablefromthepropertiesofitsconstituentsand/oritsstructure.Thisargumentalsoappliestomorphologicalconstructions.Anexampleofatypeofmorphologicalconstructionwithholisticpropertiesisreduplication.Inthecaseoffullreduplication,awordiscopied,andthereduplicatedstructurecanexpressallsortsofmeaningssuchasincreaseinnumber,quality,orintensity.InIndonesian,reduplicationisusedforexpressingpluralityonnouns(Sneddon1996:16):

(8)meja‘table’  meja-meja‘tables’rumah‘house’  rumah-rumah‘houses’singkatan‘abbreviation’  singkatan-singkatan‘abbreviations’

Thissemanticpropertyofpluralitycannotbederivedfromoneoftheconstituentsofthepluralnounorfromitsformalstructure(asequenceoftwoidenticalconstituents).Itisthecopyingconfigurationassuchthatevokesthesemanticnotionofplurality.Hence,wemightassumethefollowingconstructionalschemafortheformationofpluralnounsinIndonesian(wheretheco-indexationofthetwonounswithiindicatesthattheyhavetobeidentical;PLURisaquantifier,thesemanticoperatorofplurality):

(9)[N N ] ↔[PLUR[SEM]]

AninterestingcaseforaconstructionalanalysisofcompoundingisformedbytheclassofresultativeV VcompoundsinMandarinChinese,whereV representstheactionandfunctionsastheheadofthecompound,andV mentionsthestatebroughtaboutbytheactionspecifiedbyV .Thefollowingexamplesillustratethistypeofcompounding(Basciano2010:278–79):

(10)a.Zhāngsānyáo-xĭngleLĭsìZhangsanshake-awakeASPLisi‘ZhangsanshookLisiawake’b.Tākū-shīleshŏupàHecry-wetASPhandkerchief‘Hecriedthehandkerchiefwet’(ASP=aspectualmarker)

TheseVVcompoundsaresemanticallysimilartotheResultativeconstructioninEnglish,instantiatedbytheglossesofthesentences(10).Notethatin(10b),themainverbkū‘cried’inisolationisintransitive,andcannottakethedirectobjectshŏupà‘handkerchief,’aso-calledunselectedobject.ItistheResultativeconstructionasawholethatmakestheoccurrenceofadirectobjectpossible(Goldberg1995).ThisChineseVVCompoundconstructionmakesthesamesemanticcontributionofresultativity.AsinthecaseoftheEnglishResultativeconstruction,theresultativemeaningisnotderivablefromoneoftheV-constituents,butisevokedbytheVV-compoundassuch.Theresultativecompoundisnotnecessarilytransitive,asshownin(11),butthestatementionedbyV in(11)isagainaresultoftheeventdenotedbyV :

(11)Píngzi  pò-suè  leVase break-smash  ASP‘Thevasebrokeintopieces’

ThisimpliesaconstructionalschemaforVVcompoundsinMandarinChinesewiththisresultativemeaningspecified:theeventorpropertyspecifiedbythesecondverbisbroughtaboutbytheeventspecifiedbythefirstverb:

(12)[V V ] ↔[SEMCAUSE[SEM]]

TheargumentstructureofV whichispartofitsSEMmightintroduceanunselectedobjectforV .

AsubclassoftheseVVcompoundsisformedbycompoundsinwhichthefirstVisalightverbwithableachedmeaning.Themeaningcontributionoftheseverbsisthatofcausation;theyarecomparabletocausativeverbsoraffixesinotherlanguages.Theseareverbslikedă‘beat,strike,’nòng‘make,handle,’andgăo‘do,’whichmaynot

2

i i Nj i j

1 2

1

2 1

2

1

i j Vk i j k

2 1

Page 194: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 6 of 14

denoteanyparticularaction.Basciano(2010)pointsoutthatsomeofthecompoundswiththesethreeverbsareinfactambiguous.Acompoundlikedă-huàicaneithermean‘hit/strikeandruin/breakasaresult,’orsimply‘break’(Basciano2010:209).Similarly,thisverbwhencombinedwiththeverbsĭ‘todie’caneithermean‘causetodie=kill’or,withthenonbleachedmeaningofdă:‘beatandkillasaresult.’

ThissubclassofcausativeverbswithalightverbinV positioncanbecharacterizedbymeansofsubschemas,dominatedbyschema(12),inwhichtheselightverbsarelexicallyspecified:

(13)[[dă] V ] ↔[CAUSE [SEM]][[nòng] V ] ↔[CAUSE [SEM]][[găo] V ] ↔[CAUSE [SEM]]

Theseschemasexpressthattheselightverbscanbeusedascausativemarkerswithoutcontributingaspecificlexicalmeaning:inthesemanticpartsofthesesubschemasSEM(thelexicalmeaningofV ),hasbeenomitted,andV onlyexpressesthemeaningCAUSE.Thisspecificationwilloverrulethemeaningasspecifiedin(12).Theother,moreliteralinterpretationoftheseVVcompoundsisstillavailableaswell,sincetheycanbegeneratedbymeansofschema(12).

Theschemas(13)areconstructionalidioms,becauseoneofthepositionsislexicallyspecified.Thisconceptisalsorelevantinrelationtosuru-compoundsinJapanese(Kageyama1982,1999).Theseareright-headedcompoundswiththeverbsuru‘do’astheirhead;thenonheadisaverbalnoun,anditisonlywiththisverbthatverbalnounscombineintosuchcompounds.Thefollowingexamplesillustratethistypeofcompounding:

(14)a.yama-nobori-surumountain-climbing-do‘todomountainclimbing’b.kenkyuu-suru‘todoresearch’c.saikuringu-surucycling-do‘todocycling’

Insteadofsuru,anumberofsuppletiveformswiththemeaning‘todo’canbeused:thepotentialformdekiru,thehonorificformnasaru,andthehumbleformitasimasu(Kageyama1999:313).Thatthistypeofcompoundingisonlyproductivewiththeverbsuruanditssuppletiveformsmentionedabove,isexpressedbythefollowingconstructionalschema:

(15)[[x] [suru] ] ↔[DO SEM]

wheresurustandsforthelexemesuruanditssuppletiveforms.

Theholisticinterpretationofcompoundsisfurtherillustratedbytheclassofexocentriccompounds,thatis,compoundsforwhichtheirlexicalclasscannotbeidentifiedwiththatofoneoftheirconstituents.

Awell-knownclassofexocentriccompoundsinRomancelanguagesisVNcompoundsthatdenotetheagentorinstrumentoftheactionexpressedbytheV;theNdenotestheobjectoftheaction,asinItalian[[lava] [piatti] ]‘lit.wash-dishes,dish-washer,’[[porta] [lettere] ] ‘lit.carry-letters,postman,’and[[ruba]V[cuori]N]N,‘lit.break-hearts,heartbreaker.’Thereisnosubpartinthesecompoundsthatexpressesthenotionof‘agent’or‘instrument.’Onemighttrytosolvethisproblembyassuminganominalizingzero-suffix.Forinstance,onemightassumethatthemorphologicalstructureofItalianportalettere‘postman’is[[[porta] [lettere] ] -Ø] ,withazero-suffixthatiscategory-determiningandcarriesthemeaningofagent/instrument.However,theassumptionofsuchazero-suffixisacompletelyarbitrarywayofmakingthesecompoundsendocentric,sincethereisnoindependentevidenceforsuchazero-morpheme.Instead,inaconstructionalanalysisthewordclassandmeaningofthesecompoundsarespecifiedaspropertiesofthemorphologicalconstructionassuch:

(16)[V N ] ↔[x[xSEMy],y=SEM]

Thatis,suchcompoundsdenotethex-argument(agentorinstrument)ofthepredicateexpressedbytheV-

1

Vi j Vk i j k

Vi j Vk i j k

Vi j Vk i j k

i 1

1

3

VNi V0j V0k j i k

V N N

V N N

V N V N

i j Nk i j k

Page 195: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 7 of 14

constituent,andtheNfunctionsasthey-argument(patient)ofthatpredicate.

ExocentriccompoundingisnotrestrictedtoRomancelanguages.SimilarVNcompoundsoccurinPolish(Szymanek2009:474):

(17)[VN] łam-i-strajk‘lit.break-strike,strikebreaker’[NV] list-o-nosz‘lit.letter-carry,postman’

CompoundsofthetypeN+nominalizedVare“frequentlyexploitedinJapanesetonametheagents,instruments,products,andotherconcreteentitiesthatfigureprominentlyintheeventsorstatesdescribedbythecompounds”(Kageyama2009:515):

(18)e-kaki‘lit.picture-drawing,painter’tume-kiri‘lit.nail-cutting,nailcutter’tamago-yaki‘lit.egg-frying,omelette’

Again,thesemeaningsdonotderivecompletelyfromtheNandVconstituentsofthesecompounds,andarethustobeseenasbeingevokedbythemorphologicalconstruction.

14.4.ParadigmaticMorphology

Whenconventionalizedcomplexwordsarelistedinthelexicon,asassumedinConstructionMorphology,theymaybeparadigmaticallyrelated.Forinstance,whentheEnglishwordssocial,socialist,andsocialismarealllisted,thelanguageusermayconsidersocialistandsocialismastheconcatenationofsocialand-ist/-ism(syntagmaticrelationships),butalsoestablishadirect,paradigmaticrelationshipbetweensocialismandsocialistintermsofsuffixsubstitution,sincetheysharetheirbasewordsocial.Suchparadigmaticrelationshipsmayleadtothecoiningofnewwordsbymeansofsubstitution.Indeed,thecreationofderivedcomplexwordsisnotjustamatterofattachingaffixestobasewords.Thereareclearcasesofaffixsubstitution.ConsiderthefollowingrelatedpairsofEnglishwords:

(19)alpin-ismalpin-istcommun-ismcommun-istde-ismde-istfasc-ismfasc-istsolips-ismsolips-ist

Semantically,themeaningofthenounsontherightcanbecircumscribedasacompositionalfunctionofthemeaningofthenounsontheleft.Forinstance,acommunistmaybedefinedasan‘adherentofcommunism.’Yet,fromaformalpointofviewthewordin-istisascomplexasthecorrespondingwordin-ism.Forthatreason,Aronoff(1976)proposedtruncationrulesthatdeletesuffixesbeforeothersuffixes:-ismisdeletedbefore-istinacomplexwordlikealpin-istderivedfromunderlyingalpin-ism-ist.Theassumptionoftruncationrulescanbeavoidedoncewemakeuseofschemasinsteadofrules:schemasforcomplexwordswiththesamedegreeofmorphologicalcomplexitycanberepresentedasparadigmaticallyrelated(symbolizedhereby≈):

(20)<[x-ism] ↔[SEM] >≈<[x-ist] ↔[PERSONinvolvedinSEM] >(angledbracketsmarktheboundariesoftheindividualconstructionalschemas).

In(20)itisexpressedthatweinterpretnounsin-istasdenotingpersonsinvolvedinsomeideology,belief,propensity,oractivity.Notethatwecanlearnandunderstandthemeaningof,forinstance,alpinistwithouthavingthewordalpinisminourlexicalmemory.Inthatcase,thereisnoco-indexationoftherelevantsemanticpart(SEM)ofthewordin-istwithalexemewiththatmeaningSEM.Anadditionaladvantageofthisanalysisisthatwedonothavetoarbitrarilydecidewhetherthewordin-istisformallyderivedfromthewordin-ism,orviceversa.Theschemapair(20)allowsustogoinbothdirectionsincoininganewword.Yet,themeaningofthewordin-istwillbeacompositionalfunctionofthemeaningofthewordin-ismifsuchawordin-ismisavailable.

Suchschematiccorrespondencescanalsobeusedincasesofnonconcatenativemorphologysuchasvowelalternation,themechanismusedformakingthepasttenseformsofthestrongverbsinGermaniclanguages.

N

N

Ni i Nj i j

Page 196: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 8 of 14

However,inthelattercase,thepairofparadigmaticallyrelatedschemasisnotproductive,sincethesealternationscannotbeextendedtonewverbs.

14.5.Inflection

Inflectionalphenomenaprovideclearargumentsfortheconstructionalapproachtomorphologyintroducedintheprevioussections.Thebasicargumentsconcerntheholisticpropertiesofinflectionalforms,andtherelevanceofparadigmaticrelations.Thisleadstoaword-basedapproachtoinflection,theWord-and-Paradigmmodel.

Theclassicalproblemofinflectionalmorphologyisthecomplicatedrelationbetweenformandmeaning.Itisoftenimpossibletoassignaspecificmeaningtoaninflectionalaffix,becauseitsactualvaluedependsonthekindofstemitcombineswith,andthepropertiesofthatstem,unlessoneallowsforlargesetsofhomonymousinflectionalaffixes.Consider,forinstance,theparadigmofmasculineandneuternouns(declensionI)inRussian(Gurevich2006),exemplifiedbystol‘table’andbljudo‘dish’:

(21)masculine  neuter  sg  pl  sg  plnom  stol  stol-y  bljud-o  bljud-aacc  stol-a  stol-y  bljud-a  bljud-agen  stol-a  stol-ov  bljud-a  bljuddat  stol-u  stol-am  bljud-u  bljud-aminst  stol-om  stol-ami  bljud-om  bljud-amiloc  stol-e  stol-ax  bljud-e  bljud-ax

Astheseparadigmsillustrate,thesameending,forinstance-a,mayhavedifferentinterpretationsdependingontheclassofthenoun.Moreover,theparticularvalueexpressedisacombinationofproperties,suchas[acc.sg]or[nom.pl].Thatis,thereisnoone-to-onecorrespondencebetweenformandmorphosyntacticproperties.OnealsofindselementsininflectionalformssuchasthethematicvowelsofverbalconjugationinLatinandtheRomancelanguagesthatdonotcontributebythemselvestothemeaningoftheinflectedforms;theyare‘morphomic’properties(Aronoff1994).ThesearefamiliarobservationssupportingaWord-and-Paradigmbasedapproachtoinflectionalmorphology(Blevins2006;Ackermanetal.2009).FromaConstructionMorphologyperspective,thismeansthatthemorphosyntacticpropertiesofeachwordforminaparadigmarebestconsideredasconstructionalproperties,thatis,aspropertiesofthewordformasawhole.Thiscanbeexpressedbymorphologicalschemas,forexample,schemasthatabstractoverwordformsofthesamedeclensionclass(cf.Gurevich2006forsuchananalysisofGeorgian).Theschemafortheacc/gen.sgwordformsstolaandbljudaisgivenin(22):

(22)(x-a) ↔[N] ↔SEM

wherexisaphonologicalvariablefornominalstems,andωisaphonologicalword.ThemeaningSEMmentionedhereisthatofthelexeme.Thesemanticinterpretationofthemorphosyntacticfeaturesisnotspecifiedhere,sincethisinterpretationdepends—asfarasthecasefeaturesareconcerned—onthesyntacticcontextsinwhichawordoccurs.

AsarguedbyHarris(2009),theproblemof‘exuberantexponence,’thephenomenonthataparticularmorphosyntacticpropertyisexpressedmorethanoncebythesamemorphemewithinonegrammaticalword(andhenceasubcaseofextendedexponence),isaproblemformorpheme-basedmorphologicaltheories.HarrisshowsthatforaproperaccountoftherecurrentclassagreementmarkersinverbformsofBatsbi,aNakh-Dagestanianlanguage,itisessentialtomakeuseofschemasandsubschemasinahierarchicallexicon.Forexample,thefollowingschemascanbeassumedtocharacterizespecificsetsofverbalformswiththeclassmarkerd(Harris2009:298;LEXstandsforalexicalstem).Intheexamples,theyarefollowedbytheendings-i andiy-e ,respectively:

(23)a.Schemaforverbstems[d-LEX] example:d-ek’-i ‘theyfell’b.Schemawithextensionfortransitiveverbs[V-d-i]

4

ωi acc/gensg i

i

n n

V n

V

Page 197: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 9 of 14

c.Unifiedschemaa+b[[d-LEX] d-i] example:d-ek’-d-iy-e ‘threwitoff’

Theseschemasspecifytheformalmake-upofthevariousverbforms.Thus,thephenomenonofextendedexponencesupportsaconstructionistapproachtoinflectionalmorphology.

Formanylanguages,theconstructionofaparticularinflectionalformcannotbeconceivedofasamatterofconcatenationofastemandoneormoreinflectionalmorphemes.Instead,aparticularinflectionalformmayhavetobecomputedonthebasisofanotherformofitsinflectionalparadigm.Aparticularinflectionalformmayplaytwodifferentrolesinaccountingfortheconstructionofinflectionalforms.First,particularinflectionalformsmaybeusedtoidentifytheinflectionalclasstowhichawordbelongs(FinkelandStump2007,2009).Forinstance,thegenitivesingularformoftheLatinnoungenus‘type,’gener-is,identifiesthisnounasbelongingtothe3rddeclension.Thatis,gener-isisaprincipalpartoftheinflectionalparadigmofgenus.Second,aninflectionalformmaybeusedtocomputetheformofothercellsinthesameinflectionalparadigm(Blevins2006;Ackermanetal.2009).Forinstance,inSaamifirstdeclensionnounsexhibitapatternbasedontwoprincipalparts,thegenitivesingularandthenominativesingular.Thesenounformsaresubjecttogradation.Ifthenominativesingularformisstrong,andhencehasageminate,theillativesingularandtheessiveformarealsostrong.Inthatcase,thegenitivesingularhasaweakform,withasingleconsonant(asinbihttávs.bihta‘piece,nom.sg./gen.sg.’).Conversely,ifthenominativesingularformisweak,thecorrespondingillativesg.andtheessiveareweakaswell,whereasinthatcasethegenitivesingularformisstrong(asinbarguvs.barggu‘work,nom.sg./gen.sg.’;Blevins2006:546).Inotherwords,morphologicalgeneralizationsaboutsuchparadigmscanonlybemadeintermsofsystematicparadigmaticrelationshipsbetweencellsoftheseparadigms.Therelationsbetweenthenominativesg.,theillativesg.,andtheessivecanbeexpressedasparadigmaticcorrespondencerelationsbetweenmorphologicalschemas:

(24)[x-á] ≈[x-ái] ≈[x-án]

Ifthevariablexstandsforastrongstemwithageminateconsonant,asinbihttá,thisgeminateconsonantwillbepredictedtorecurinallthreeforms.Inversely,ifxstandsforaweakstemandáisreplacedwithu,asinbargu,itispredictedthatthisweakstemshowsupinthesethreeinflectionalforms.Thatis,thesemutuallyimplicativerelationshipsbetweenparadigmcellscanbeexpressedstraightforwardlybymakinguseofschemasforfullyspecifiedinflectionalformsandparadigmaticrelationshipsbetweensuchschemas.

SinceConstructionMorphologyallowsforbothmorphologicalschemasandtheirinstantiationstobelisted,itispossibletolistinflectionalformsthatarecompletelyregular,becausetheinflectionalschemaswillindicatetheirpredictableproperties.Thisisnecessarybecauseinflectionalformsfunctionasprincipalparts,andbecausetheformofoneparadigmcellmaybepredictedfromanother.

Storageofcertaininflectionalformsisawelcomeresultfromapsycholinguisticpointofview.Ithasbeenshown,forinstance,thatregularpluralnounsinItalianexhibitafrequencyeffectinlexicaldecisiontasks.Thisimpliesthelexicalstorageofthesepluralnouns(Baayenetal.1997).Moreover,inflectedformsmayberegularfromthemorphologicalpointofview,butexhibitidiosyncraticpropertiesinotherdomains.ThepluralformofItalianamico[amiko]‘friend’isamici[amitši],withpalatalizationofthestem-final/k/to[tš].However,thisruledoesnotapplytoallItaliannouns(comparebuco[buko]‘hole,’pluralbuchi[buki]),andhenceonehastomemorizewhichpluralnounshavepalatalization.Therefore,apluralnounsuchasamicihastobelexicallystoredforphonologicalreasons.Yet,theinflectionalschemaforpluralizationofnounswillindicatethatitismorphologicallyregular.

Anotherargumentfortheconstructionalapproachtoinflectionisformedbyperiphrastictenseforms.Forinstance,inEnglishthewordcombination‘have+pastparticiple’isusedtoexpresstheperfecttenseofverbs.Inthisconstruction,theverbhavedoesnotexpressthemeaning‘topossess,’butagrammaticalmeaningofperfectivity,incombinationwiththepastparticiple.Thegrammaticalmeaningofperfecttenseisapropertyofthisconstructionasawhole(SadlerandSpencer2001;AckermanandStump2004;Spencer2004).Hence,periphrasticinflectionalformsaretobetreatedasconstructionalidiomsinwhichtheauxiliaryislexicallyfixed,whereastheslotfortheparticipleisavariable.Thesemanticpropertiesofperiphrasticforms,suchasperfectivity,areholisticpropertiesthatarespecifiedaspropertiesoftheconstruction.Wordcombinationsmaythereforefillcellsofmorphologicalparadigms.Thisshowsthatmorphologicalandsyntacticconstructsmayhavethesamemorphosyntacticfunction.Thiscanbehandledadequatelyinaconstruction-basedapproachtomorphosyntacticphenomena.

V V n

nomsg illativesg essive

Page 198: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 10 of 14

14.6.Construction-DependentMorphology

Theconceptof‘construction’isnotonlyimportantforapropercharacterizationofcomplexwords:syntacticconstructionsmaybeessentialforapropercharacterizationofthedistributionofboundmorphemesaswell.AnexampleisthedistributionoftherelicsoftheoldGermanicgenitivesuffix-sinvariousDutchconstructions,discussedinBooij(2010).Originally,allwordsinNPsinspecifierpositionweremarkedbymeansofthegenitivecase,asillustratedhereforMiddleDutch:

(25)a.Symon-sDankard-sson-shuseSymon-genDankard-genson-genhouse.nom.sg‘thehouseofthesonofSymonDankard’b.De-rduechd-enmoederethe-gen.plvirtue-gen.plmother.sg.nom‘themotherofthevirtues’

LanguageslikeEnglish,Swedish,andDutchhavelosttheircasesystem.Yet,theboundmorpheme-s,sometimesreferredtoasthePossessorsuffix,isstillusedinparticularconstructionsintheselanguages. Nordereferstothischangeoftheroleofthesuffix-sasdeinflectionalization(Norde2009).The-shasbecomeamarkerofspecificconstructions.InthecaseofDutch,propernames—butonlythese—canbeusedasprenominalspecifiers,andhence(26c)isill-formed,unlikeitsEnglishglosswiththesamestructure:

(26)a.Jan-sboek‘John'sbook’b.PrinsBernhard-szesdochters‘PrinceBernhard'ssixdaughters’c.*dedirecteur-spet‘thedirector'scap’

AnadditionalsemanticpropertyofthisconstructionisthatitmakestheNounPhrasedefinite,eventhoughthereisnoovertgrammaticalmorphemetoexpressthisdefiniteness.Hence,wehavetoassumeaspecificconstructionforDutchNPswithsuchpropernamesasspecifiers,ofwhichtheboundmorpheme-sispartandparcel:

(27)TheDefinite-sConstruction(Dutch)[[….[x-s] ] …N ] ↔[the…N ofNP ] Condition:NP isapropername

Apparently,thegenitivecasemarkerhasdevelopedintothemorphologicalmarkerofaparticularconstruction.Thisshowsthatmorphologicalmarkersmaygettiedtospecificconstructions.Thatis,thedistributionofboundmorphemesmaydependonspecificconstructionsduetoreanalysisofsyntacticconstructions.Croftqualifiesthiskindofreanalysisashypoanalysis:“Inhypoanalysis,thelistenerreanalyzesacontextualsemantic/functionalpropertyasaninherentpropertyofthesyntacticunit.Inthereanalysis,theinherentpropertyofthecontext…isthenattributedtothesyntacticunit,andsothesyntacticunitinquestiongainsanewmeaningorfunction”(Croft2000:126–27).

AslightlydifferentexampleoftheuseofaparticularmorphologicalformbeingtiedtoaparticularconstructionisthatofthesuperlativeformofDutchadjectives,whichoccursintwospecificconstructions,exemplifiedin(28):

(28)a.  Hijzingthetmooi-st(e)vanallemannenHesingsthebeautiful-estofallmen‘Hesingsmorebeautifullythananyotherman’b.Dieboomisopzijnmooi-stinmeiThattreeisonhisbeautifull-estinMay‘ThattreeismostbeautifulinMay’

In(28a)thesuperlativeformoftheadjectiveisprecededbythedeterminerhet,andoptionallyfollowedbytheending-e.Thepositionofthesuperlativeafteradeterminerimpliesthatitfunctionsasanoun.Hence,thesuperlativemusthavebeenconvertedfromadjectivetonouninthisconstruction.Dutchadjectivescanbeconvertedtonouns,eitherbymeansofconversion(‘zero-derivation’),orbymeansofsuffixationwith-e.Thesetwoprocessesleadbothtoneuternounsthatselectthedefinitedeterminerhetthatshowsupintheexamplesin(28).However,onlytheoptionwithovertmorphologicalmarkingisproductiveinDutchoutsidethecontextofthisspecificconstruction.Forinstance,hetmooi-e‘thebeautifulthing’isgrammatical,but*hetmooiisungrammatical,unlikehetgeel‘theyellowcolor,’aconversionoftheadjectivegeel‘yellow.’ConversionofAtoNonlyappliestoa

5

6

N NPi j NPk j i k i

Page 199: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 11 of 14

restrictedsetofadjectivessuchascoloradjectives.Soitisthesetwospecificconstructionsthatenableproductiveconversionofthesuperlativeadjective(A-sup)toanoun.Moreover,NPsofthetype[hetA-sup]areonlyusedadverbially:wecannotusetheNPhetmooist‘themostbeautifulthings’asasubject-NPoranobject-NP,whereastheformwithfinal-eleadstoagrammaticalresult,sinceitisacaseofregular,productivederivationofnounsfromadjectivesbymeansofsuffixationwith-e:

(29)a.Het{*mooi-st/mooi-st-e}isnognietgoedgenoegThe{beautifull-est/beautifull-est-nom}isyetnotgoodenough‘Themostbeautifulstuffisnotgoodenoughyet’b.Hijverkochthet{*mooi-st/mooi-st-e}aanzijnvriendenHesoldthe{beautifull-est/beautifull-est-nom}tohisfriends‘Hesoldthemostbeautifulstufftohisfriends’

Thesuperlativeconstructionexemplifiedin(28b)alsohasspecialsyntacticproperties.ThePP[opzijn+A-sup]in(28b)containsthefixedprepositionop‘on’andthefixed3.sgpossessivepronounzijn/zεin/‘his,’obligatorilyrealizedwiththeweakform[zƏn](spelledasz’n),butthechoiceofpossessivepronounoptionallyco-varieswiththepersonoftheantecedent.Forallpersons,theweakformofthepossessivepronounmustbechosen.Hence,bothofthefollowingsentencevariantsaregrammatical:

(30)Indiejurkbenjeop{je/zijn}mooistInthatdressareyouat{your/his}beautifull-est‘Inthatdressyoulookmostbeautiful’

Bothconstructionsareconstructionalidioms,withonevariableposition,thatforthesuperlativeadjective.Theymayberepresentedasfollows:

(31)a. [[het] [A ](e)] ] ↔[withhighestdegreeofSEM]b.[[op] [POSS[A ] ] ] ↔[withhighestdegreeofSEM](wherePOSSisaweakpossessivepronounandmaybefixedasz’n)

Theschemasin(31)expressthattheconversionofthesuperlativeformofanadjectivewithoutovertmorphologicalmarkingispossibleinthisspecificsyntacticcontext.Thisisthereforeanotherillustrationofhowmorphologicaloperationscanbeconstruction-dependent.

Inthiscase,thereisnoreferenceintheconstructionalschematothesuperlativesuffixassuch,onlytothemorphologicalfeature[+superlative].ThiscorrectlypredictsthatDutchadjectiveswithirregularsuperlativeformssuchasbest‘best’andmeest‘most,’thesuperlativeformofgoed‘good’andveel‘much,’respectively,canalsooccurintheseconstructions:opzijnbest‘atbest,’opzijnmeest‘atmost.’

14.7.PhrasalLexicalUnits

Thenotion‘word’isnotcoextensivewiththenotion‘lexicalunit,’sincevarioustypesofphrasalconstructsmayfunctionaslexicalunits(seealsoJackendoff,thisvolume).ClassicexamplesareEnglishANphrasessuchasredtape,bluecheese,andmodernart(Jackendoff1997a).Itisoftenthecasethatwhatisexpressedinonelanguagebymeansofcompoundingisexpressedinanother,relatedone,bymeansofphrasalconstructs,asisillustratedhereforDutchandGerman.Germanappearstohaveapreferenceforcompounding,whereDutchusesphrases(Booij2009):

(32)DutchANphraseGermanANcompoundglossbijzonder-ezittingSonder-sitzung‘specialsession’gebruikt-ebatterijenAlt-batterien‘usedbatteries’geheim-enummerGeheim-nummer‘secretnumber’mobiel-etelefoonMobil-telefon‘mobilephone’nieuw-eautoNeu-wagen‘newcar’

ThephrasalnatureoftheDutchconstructsisrevealedbytheovertinflectionalendingontheprenominaladjective.Bothtypesofconstructsarelexicalizedinthesensethattheybelongtothelexicalconventionofthe

Det [+sup]i N NPj i j

P [+sup]i N NP PPj i j

Page 200: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 12 of 14

languageinvolved.Hence,thelexiconmustcontainphrasalschemasandisthustobeconceivedasa‘constructicon.’

Aninterestingclassofphrasallexicalconstructsisformedbyparticleverbs,whicharefeaturedby,amongmanyotherlanguages,Hungarian(33),English(34),andGerman(35):

(33)a.Pétertegnapki-mentazerdöbePeteryesterdayout-wenttheforest.into‘Yesterday,Peterwentintotheforest’b.PétertegnapmentkiazerdöbePeteryesterdaywentouttheforest.into‘Yesterday,Peterwentintotheforest’(KieferandHonti2003:142)

(34)a.Johnlookeduptheinformationb.Johnlookedtheinformationup

(35)a.dassHansseineMutteran-ruftthatHanshismotherat-calls‘thatHansphonesuphismother’b.HansruftseineMutteranHanscallshismotherat‘Hansphonesuphismother’

Thereisanextensivediscussioninthelinguisticliteratureashowtoanalyzeparticleverbs:aretheywords(andhenceacaseofmorphology),oraretheyphrasalconstructs,andhencetobeaccountedforbysyntax(Booij2010:chapter5;Losetal.2012)?Theirwordlikestatusrelatestotheobservationthattheyoftenhaveanidiosyncraticmeaningandcanfeedwordformation;ontheotherhand,theyareseparableinsomesyntacticcontexts.ThisisillustratedherebymeansoftheDutchparticleverbaan-vallen‘lit.toat-fall,toattack’:ithasderivativeslikeaan-vall-er‘attacker,’anditissplitinrootclauses:

(36)a.datHanszijnvijandaan-valtthatHanshisenemyat-falls‘thatHansattackshisenemy’b.HansvaltzijnvijandaanHansfallshisenemyat‘Hansattackshisenemy’

Particlesarealsousedasadpositionsorasadverbs.Asparticles,theymayexhibitaspecificmeaning,tiedtotheparticleconstruction.Particleverbformationisanalternativetocreatingcomplexverbsbymeansofprefixation.InEnglish,forinstance,farmorecomplexpredicatesarecoinedbymeansofparticleverbformationthanbyprefixation.Hence,thereiscompetitionbetweenthesetwoprocesses,whichreceivesanaturalaccountiftheyarebothaccountedforbyconstructionalschemasthatformpartofthesamelexicon/constructicon.ComparetheuseoftheDutchadpositions/adverbsdoor,om,onder,over,andvoorasaparticleandasaprefix:

(37)particleverb  prefixedverbdóor-boren‘togoondrilling’  door-bóren‘toperforate’óm-blazen‘toblowdown’  om-blázen‘toblowaround’ónder-gaan‘togodown’  onder-gáan‘toundergo’óver-komen‘tocomeover’  over-kómen‘tohappento’vóor-komen‘tooccur’  voor-kómen‘toprevent’

Theseminimalpairsdifferinmeaning,inthelocationofmainstress,andintheirseparability(theparticlesareseparatedfromtheverbsinrootclauses,unliketheprefixes).Eventhoughtheparticleshavealexicalizedmeaningthatisnotthesameasthatofthecorrespondingadposition,theiruseisatthesametimeproductive.Forinstance,theparticledoorexpressescontinuativeaspectandcanbeusedproductivelytocoinnewparticleverbswithsuchanaspectualmeaning.Theprefixdoor-,ontheotherhand,expressesthattheactiondenotedbytheverbgoescompletelythroughtheobjectoftheaction.Hence,wemayassumethefollowingtwoconstructional

Page 201: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

Page 13 of 14

schemasfordoor,oneforparticleverbs,withthesyntacticstatusofbeingaverbalprojection(V′),andoneforprefixedverbs(V):

(38)a.[[door] V ] ↔[continueSEM]b.[[door] V ] ↔[SEMcompletelythroughobject]

Theseschemasareconstructionalidioms,sinceonepositionislexicallyfixed,whereastheslotfortheverbisopen.Thus,itisexpressedthattheselexicalizedmeaningsofdoorareproductiveinspecificstructuralconfigurations.

Particlesmaychangeintoprefixes.ThereareverbsthatwereparticleverbsinMiddleDutch,butbecameprefixedverbsinModernDutch,sometimeswithaconcomitantbleachingofthemeaningofthepreverbalelement.MiddleDutchover-bruggen‘tobridge,’forinstance,isaparticleverb,withtheliteralmeaningof‘puttingabridgeacross,’whereasinmodernDutchitisaprefixed,inseparableverb,withthemoreabstractmeaningof‘reconcile.’However,theparticleverbsystemisremarkablystable,andparticlescanalsohaveamoreabstract,bleachedmeaning.

Particleverbs,eventhoughtheyarenotmorphologicalconstructs,thussupportthemodelofConstructionMorphologyoutlinedabove,sincethismodelenablesustoexpressthecommonalitiesofparticleverbsandprefixedverbswithoutobliteratingtheirformaldifferences.

14.8.Conclusions

InConstructionGrammarcomplexwordscanbeseenasconstructsatthewordlevel,instantiatingmorphologicalconstructions.Theseconstructionsarespecifiedbyschemasthatexpresssystematicform-meaningcorrespondences.Constructionalschemasformpartofahierarchicallexicon,withdifferentlayersofabstraction,andwithindividualcomplexwordsatthelowestlevelofabstraction.Thisconceptionofmorphologyenablesustomakegeneralizationsaboutsubclassesofcomplexwords,andtospecifyholisticpropertiesofmorphologicalconstructions.InConstructionMorphology,wecanalsospecifysystematicparadigmaticrelationshipsbetweensetsofcomplexwordsandsetsofinflectionalforms.Weneedsuchrelationshipsinordertoaccountforparadigmaticwordformation,andforsystematicpatternsininflectionalparadigms.

InConstructionMorphology,morphologyandsyntaxrelateinvariousways.Syntacticconstructionsmaycarryspecificmorphologicalmarkers,periphrasticformsmayfillthecellsofparadigmsofwords,andmorphologicalandsyntacticconstructsmaybothfunctionasnamesofentities.

ThesefindingsconfirmtheinsightofConstructionGrammarthatthelexiconistobereinterpretedastheconstructicon,astructuredandhierarchicallyorderedarrayofconstructionsandconstructswithphrasalorwordstatus.

Notes:

(1.)Exceptionstothissemanticcharacterizationarehonorable‘mustbehonored’anddespicable‘mustbedespised.’

(2.)Forasimilarconstructionalanalysisofreduplication,butwithextensiontoreduplicationofphrases,seeGhomeshietal.(2004).

(3.)Suru-compoundsareprobablynotmorphologicalcompounds,butsyntacticcompoundsinwhichtheverbalnounisChomsky-adjoinedtoaV ,formingacomplexV .SeeBooij(2010:chapter4)foradiscussionofthisissue.

(4.)Theideaofschemasandcorrespondencerelationbetweenschemasisbynomeansanewproposal.OnecanfinditinpublicationssuchasJackendoff(1975),Bybee(1988a),andBochner(1993).ThepointhereisthatmorphologicalschemasareanimportantbuildingblockofConstructionMorphology.

(5.)TheEnglishfactsandthetheoreticalissuesarewellpresentedanddiscussedinRosenbach(2002).

(6.)Cf.Booij(2010:chapter9)foradetailedanalysisofthisDutchconstruction,acomparisonwiththeEnglish

P j V’k j k

P j Vk j k

7

0 0

Page 202: The Oxford Handbook of Construction Grammar

Morphology in Construction Grammar

constructions,andanumberofothercasesofconstruction-dependentmorphology.

(7.)Cf.BlomandBooij(2003),Blom(2004,2005a,b),Booij(2010:chapter5),andLosetal.(2012)fordetailedstudiesofparticleverbsandmoreextensiveargumentation.

GeertBooijGeertBooijisProfessorofLinguisticsattheUniversityofLeiden.HeobtainedhisPh.D.fromtheUniversityofAmsterdam,wherehelaterworkedasProfessorofGeneralLinguistics.Hismainresearchinterestismorphologyanditsinterfaceswithphonology,syntax,semantics,andthelexicon.HismajorpublicationsincludeThePhonologyofDutch(1995),Morphology:AnInternationalHandbookofInflectionandWordFormation,2vols.(withCh.LehmannandJ.Mugdan,eds.,2000‒2),TheMorphologyofDutch(2002),TheGrammarofWords:AnIntroductiontoLinguisticMorphology(2007)andConstructionMorphology(2010).Moreover,heisoneofthefoundersandeditorsoftheYearbookofMorphology(1988‒2005)anditssuccessor,thejournalMorphology.

Page 203: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 1 of 13

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0015

WordsandIdiomsStefanieWulffTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterpresentsaconstructionistanalysisofwordsandidioms.Itprovidesasummaryofearlyconstructionistresearchthatarguedinfavorofviewingidiomsnotasanomalies,andaddressestheproblemofhowthedegreeofsemanticandsyntacticirregularityofdifferentconstructionscanbemeasuredquantitatively.Thechapterproposesaquantitativemeasurethatnumericallyexpressesthisdegreeofsemanticsimilarityandwhichcanbeconsideredanumericalmeasureofcompositionality.

Keywords:words,idioms,constructionistanalysis,semanticirregularity,syntacticirregularity,quantitativemeasure,semanticsimilarity,compositionality

15.1.IdiomsareConstructions

15.1.1IdiomsintheChomskyanParadigm

CroftandCruse(2004:225)describetheroleofidiomsinConstructionGrammarasfollows:“ItisnotanexaggerationtosaythatConstructionGrammargrewoutofaconcerntofindaplaceforidiomaticexpressionsinthespeaker'sknowledgeofagrammaroftheirlanguage.”Thereasonforthisfocusonidiomsinearlyconstructionistresearchisthatidiomsareincompatiblewithatheoryoflanguagethatassumesastrictseparationofgrammarandthelexicon,suchasearlyversionsoftransformational-generativegrammar(MainstreamGenerativeGrammar;Chomsky1965),thepredominantlinguisticframeworkinthetwentiethcentury.Chafe(1968)accordinglyreferredtoidiomsasan“anomalyintheChomskyanparadigm.”

Ontheonehand,idiomsarenotjustfixedwordcombinationsbecausetheycanbemodifiedbothlexicallyandsyntactically.Theidiomwalkatightrope(‘toactverycarefullysothatyouavoideitheroftwoequallybadbutcontrastingsituations’)isanexampleinquestion.ThefollowingattestationfromtheCorpusofContemporaryAmericanEnglish(COCA) illustratesthatanadjectivelikelegalcanbeinsertedbeforethenounandthatthesyntacticarrangementofverbandobjectisnotfixedeither.

(1)Aswellnessprogramsandemployeesurchargesmultiply,employersareacutelyawareofthelegaltightropetheymustwalk.

Atthesametime,manyidiomsdefythestandardrulesofgrammar,sotheyarenotassembledinthesamefashionasregularphrases.Example(2)isanidiomsoformallyfixedasnottolicensetensevariation;(3)isanidiomthatcannotundergosyntactictransformationslikepassivization;conversely,idiomsliketheonein(4)arerestrictedtowhatwouldbeconsideredatransformationallyderivedsurfacestructureintransformational-generativegrammars(example(2)takenfromCroftandCruse2004:233;example(4)takenfromNunbergetal.1994:516;glossesadded).

1

Page 204: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 2 of 13

(2)a.Ittakesonetoknowone.(‘Onlyapersonwithidenticalcharactertraitsisabletorecognizethosetraitsinsomeoneelse.’)

b.*Ittookonetoknowone.

(3)a.Heshotthebreeze.(‘Heengagedinidleconversation.’)b.*Thebreezewasshot(byhim).

(4)a.Thediceiscast.(‘Anirrevocabledecisionhasbeenmade.’)b.*Xcastthedice.

15.1.2IdiomsintheConstructicon

Fillmore,Kay,andO’Connor(1988)addressedthisissuebysuggestingthatidiomsshouldbeseenasunitsofsyntacticrepresentationthatareassociatedwithuniquefunctional(semantic/pragmatic)properties.Tomaketheircase,theylaidoutthesemanticandsyntacticirregularitiesoftheLetalone-constructionasin(5)and(6).Syntactically,letalonebasicallyfunctionslikeacoordinatingconjunction,butitdoesnotlicensethesamesyntacticarrangements(examplesfromFillmoreetal.1988:515–16):

(5)a.ShrimpandsquidMoishewon’teat.b.*ShrimpletalonesquidMoishewon’teat.

(6)a.*ShrimpMoishewon’teatandsquid.b.ShrimpMoishewon’teat,letalone,squid.

Similarly,letalonesharessomecontextswithcomparativethan,butdoesnotlicenseVPellipsislikethandoes(examplestakenfromFillmore1988etal.:517,516):

(7)a.JohnhardlyspeaksRussianletaloneBulgarian.b.JohnspeaksbetterRussianthanBulgarian.

(8)a.MaxwilleatshrimpmorewillinglythanMinniewill.b.Maxwon’teatshrimpbutMinniewill.c.*Maxwon’teatshrimpletaloneMinniewill.

Moreover,letalonecanbeunderstoodasanegativepolarityitem,whichmakesitsimilartoany,butletaloneisallowedincontextsinwhichanyandothernegativepolarityitemscannotoccur(examplestakenfromFillmoreetal.1988:518–19):

(9)a.Hedidn’treachDenver,letaloneChicago.b.Hedidn’treachanymajorcity.

(10)a.You’vegotenoughmaterialthereforawholesemester,letaloneaweek.b.*You’vegotenoughmaterialforanysemester.

Withregardtoitssemantics,letalonerequiresacomplexchainofinterpretativeprocessesonthepartofthehearerthatCroftandCruse(2004)summarizeasfollows:

Firsttheinterpretermustrecognizeorconstructasemanticpropositioninthefragmentarysecondconjunctthatisparalleltothepropositioninthefirstfullconjunct….Morespecifically…[t]heinterpretermustconstructascalarmodel,whichrankspropositionsonascale—forexample,thedistastefulnessofeatingseafood…theinitial,fullconjunctdenotesthepropositionthatisstrongerormoreinformativeonthescale….Thiswholesemanticapparatusisrequiredfortheinterpretationoftheletaloneconstruction,andisnotnecessary(asawhole)forotherconstructions.(CroftandCruse2004:239)

Goingbeyondtheindividualexampleofletalone,Fillmoreetal.(1988:506–10)classifiedidiomsintodifferenttypesdependingontheextenttowhichtheydeviatefromregularsyntacticexpressionsintermsoftheirlexical,semantic,andsyntacticirregularity.Table15.1providesanoverview.

Whenpartsofanexpressionarenotfoundoutsidetheidiomtheyoccurinandtheidiomhasasyntactically

Page 205: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 3 of 13

irregularconfiguration,Fillmoreetal.classifytheexpressionas‘unfamiliarpiecesunfamiliarlyarranged’;examplesherearekithandkin(‘familyandfriends’)andwithmightandmain(‘withalotofstrength’).AsCroftandCruse(2004:235)note,“[u]nfamiliarwordsarebydefinitionunfamiliarlyarranged:ifthewordsdonotexistoutsidetheidiom,thentheycannotbeassignedtoasyntacticcategoryintermsofaregularsyntacticrule.”Atthesametime,itisimportanttonotethatwhiletheseexpressionsmaybequiteirregularacrossallthreerelevantdomains,theyarenotnecessarilyentirelynoncompositional:partsoftheexpressioncanstillbemappedontopartsofitsmeaning.Wemustconcludethatevenwithinthisgroupofhighlyirregularexpressions,thereisconsiderablevariationasfarasthedegreeofnoncompositionalityisconcerned,withthemajorityofexpressionsbeingpartiallycompositional.Thatis,mostexpressionseveninthiscategoryarenotdecodingidioms(inFillmoreetal.'sterminology)thatneedtobelearnedaswholes,butencodingidioms.

Table15.1.Typesofidiomscomparedtoregularsyntacticexpressions.

Lexis Syntax Semantics

unfamiliarpiecesunfamiliarlyarranged irregular irregular irregular

familiarpiecesunfamiliarlyarranged regular irregular irregular

familiarpiecesfamiliarlyarranged regular regular irregular

regularsyntacticexpressions regular regular regular

Thecategoryof‘familiarpiecesunfamiliarlyarranged’ishometoexpressionslikeallofasudden(suddendoesnotfunctionasanounoutsideofthisidiom)andinpointoffact(inothercontexts,factrequiresadeterminer).Theydifferfromthefirstgrouponlyinthatalltheircomponentwordsarefamiliar,thatis,alsooccuroutsideoftheexpression.Theymayalsovaryintermsoftheirdegreeofschematization(Fillmoreetal.refertothelexicallyfullyspecifiedidiomsassubstantiveandpartiallylexicallyspecifiedidiomsasformalidioms),andtowhatextentthecontributionsthatthecomponentwordsmaketotheexpressionoverlapwiththeirmeaningsoutsideofthatexpression.

Thethirdclass,‘familiarpiecesfamiliarlyarranged,’covers(againbothsubstantiveandformal)expressionslikepullX'sleg(withXbeingalexicallyunspecifiedslotthatcanbefilledwithahumanreferent)ortickletheivories(‘playthepiano’).Thesedifferfromtheothertypesofidiomaticexpressionsonlyinthattheircomponentwords“arearrangedinawaythatreflectstheregulargrammaticalpatternsofthelanguage”(EvansandGreen2006:645).

AseriesofinfluentialstudiesonschematicconstructionsthataresemanticallyandsyntacticallymoreorlessirregularfollowedFillmoreetal.'s(1988)casestudyonletalone.Moreover,variousearlierstudieswithinthelargerfamilyofcognitive-functionallinguisticswerereassessedfromaspecificallyconstructionistperspective. Table15.2providesanoverviewofsomeoftheseconstructions,examplesentences,andreferences(inchronologicalorderofpublication).Theselectionofstudieslistedhereisbynomeansrepresentativeofthevast(andsteadilygrowing)numberofstudiespublishedfrom,oratleastcompatiblewith,aconstructionistperspective,butillustratestherangeofconstructionsthatcanbedescribedby

2

Page 206: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 4 of 13

Table15.2.Examplesofsemanticallyand/orsyntacticallyirregularconstructions.

Constructionlabel Example(reference)

omitteddeterminers Idon’tthink,Mac/*cabby,thatthisisthebestwaytogo.(Zwicky1974)

it-clefts Itisagainstpardoningthesethatmanyprotest.(Prince1978)

deicticThereconstruction Theregoesthebellnow!(Lakoff1987)

tautologies Boyswillbeboys.(Wierzbicka1987)

syntacticamalgams Therewasafarmerhadadog(Lambrecht1988)

have/give/takeaV haveadrink;givetheropeapull;takealookat(Wierzbicka1988)

MadMagazineconstruction Him,adoctor?!(Lambrecht1990)

NPNconstruction housebyhouse;dayafterday(Williams1994)

nominalextraposition It'samazingthedifference!(MichaelisandLambrecht1996)

TIMEawayconstruction Twistin’thenightaway(Jackendoff1997b)

preposing It'sverydelicate,thelawn.(BirnerandWard1998)

What'sXdoingY? What'sthatflydoinginmysoup?(KayandFillmore1999)

means

ofparametersofcomplexity,lexicalspecification,andsemanticandsyntacticirregularity.

FollowingFillmoreetal.'sreconceptualizationofidiomsassymbolicunitsandtheensuingwealthofstudiesonconstructions,“ConstructionGrammarianscametoarguethat,infact,grammaticalorganizationisentirelyvertical”(CroftandCruse2004:247–48).Foraschematicrepresentationofthis“verticallyorganized”extendedmentallexicon,orso-called“constructicon,”pleaseseeGoldberg(thisvolume).AsGoldbergexplains,thedifferencebetweenwordsandgrammaticalframesisoneofdegreeratherthanquality,andonecandescribeitalongtwoparameters,complexityandschematization.First,constructionsdifferintermsoftheircomplexity:morphemesandwordsaresimpleconstructions,whereasidiomsandgrammaticalframesareincreasinglycomplex.Second,constructionsdifferintheirdegreeofschematizationorlexicalspecification:wordsarefullylexicallyspecified,whereasgrammaticalframesaremaximallyunspecifiedwithregardtothelexicalmaterialthatcanbeinserted.Idiomsoccupythespaceinbetweenthesetwoextremes,withsomelikeshootthebreezebeingfullylexicallyfilledandotherslikepullX'slegbeingonlypartiallyspecified.

15.1.3IdiomsinUsage-basedConstructionGrammar

Aswehaveseeninthesectionabove,theintegrationofsemanticallyandsyntacticallyirregularexpressions,firstandforemostidioms,hasbeenamajorimpetusforthedevelopmentofConstructionGrammar.Itmustbeemphasizedhere,however,thatformaland/orfunctionalunpredictabilityaresufficient,butnotnecessaryconditionsforconstructionstatus:evenhighlyregularexpressionsthatareusedsufficientlyoftentobecomeentrenchedinthespeaker'smentallexiconqualifyasconstructions(Goldberg2006a:64;seealsoBybee,thisvolume).Thisqualificationofearlierdefinitionsofconstructions(Goldberg1995:4)asnecessarilyirregularisbeingpromotedmoststronglyinusage-basedConstructionGrammar(seeBybee,thisvolume;Goldberg,thisvolume).Inusage-basedConstructionGrammar,theacquisition,representation,andprocessingoflanguageareshapedbyusage.Methodologically,ausage-basedperspectiveentails(amongotherthings)adecidedfocuson(ideally

3

Page 207: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 5 of 13

representativesamplesof)authenticlanguagedata.Assuch,usage-basedConstructionGrammarreflectsthegeneralempiricalturninlinguistics(seeGries,thisvolume).

Theinclusionofexpressionsintheconstructiconbasedonfrequencyalonedoesnotcontradictthehypothesisthatitisprimarilyverticallyarranged.However,itstronglyemphasizestheideathatdifferentfactorslikelexicalspecification,semanticirregularity,syntacticirregularity,andcognitiveentrenchmentprevailatalllevelsoftheconstructiconindifferentshadesofprominenceandrelativeimportance.Itappearsthatatthelevelofidioms(inparticularnotfullylexicallyspecifiedones),wenotonlyfindallofthesefactorsinstantiated,whichmakesschematicidiomsparticularlyinteresting;whatismore,anygiven(schematic)idiomcanbecharacterizedindividuallyalongeverysinglefactor,resultingina‘multi-dimensionalcontinuum’ofdifferentlyformallyandsemanticallyirregularandcognitivelyentrenchedexpressionsthatultimatelyblurstheboundariesofidiomtypesasdescribedinFillmoreetal.(1988)andvariousother,nonconstructionistidiomtypologies(e.g.,CacciariandGlucksberg1991;Fernando1996).Moreover,suchanunderstandingrenderstheterm‘irregularity’rathermisleading;instead,Iwillhenceforthrefertothesemanticandformal‘behavior’ofconstructions.

Fromausage-basedConstructionGrammarperspective,itbegsthequestionhowtodescribethiscontinuummostadequately.Inwhatfollows,IwillpresentmeasuresderivedfromcorpusdataforthesemanticandsyntacticirregularityofdifferentlyfrequentsubstantiveVNPconstructionsthatstandinaccordwithbasicconstructionistpremises.Iwillthendiscusswhichofthesefactors,andtowhatextent,underlienativespeakerjudgmentsoftheperceivedidiomaticityoftheseconstructions.

Thedatainquestionarethirty-nineVNPconstructions(13,141tokenstotal)retrievedfromtheBritishNationalCorpus(BNC).Thirty-threeoftheseconstructionsarelistedintheCollinsCobuildIdiomDictionary,andsixconstructionswereaddedbasedonfrequencyalone(seeWulff2008b:25–27fordetails).Example(11)listsallthirty-nineVNPconstructionswiththeirfrequencies(inalltheirvariantforms)inparentheses.

(11)bearDET fruit(90),begDETquestion(163),breakDETground(133),breakDETheart(183),callDETpolice(325),carryDETweight(157),catchDETeye(491),changeDEThand(212),closeDETdoor(827),crossDETfinger(150),crossDETmind(140),deliverDETgood(145),doDETtrick(155),drawDETline(310),fightDETbattle(192),fill/fitDETbill(116),followDETsuit(135),footDETbill(109),getDETacttogether (142),gritDETtooth(164),haveDETclue(232),haveDETlaugh(98),holdDETbreath(292),leaveDETmark(145),makeDETheadway(136),makeDETmark(213),makeDETpoint(1,005),make/pullDETface(371),meetDETeye(365),paveDETway(269),playDETgame(290),scratchDEThead(100),seeDETpoint(278),takeDETcourse(294),takeDETpiss(121),takeDETplunge(115),takeDETroot(113),tellDETstory(1,942),writeDETletter(1,370)

15.2.TheContributionofWordstoIdioms:MeasuringSemanticBehavior

Asdiscussedabove,constructioniststudiesofidiomsagreethat“themeaningsofmostidiomshaveidentifiableparts,whichareassociatedwiththeconstituentsoftheidioms”(Nunbergetal.1994:531;seealsoLakoff(1987:448ff.)foradetailedaccountofhowmetaphortheorycanbeemployedtoaccountforthelinkagesbetweenliteralandidiomaticmeaning).Attheoppositeendofthecompositionalityspectrum,welikewisefindthat“[s]trictcompositionalityisrarely,ifever,encountered.Mostexpressions(Iamtemptedtosay:allexpressions),wheninterpretedinthecontextinwhichtheyareuttered,arenon-compositionaltosomedegree”(Taylor2002:550).Inotherwords,acognitive-linguisticperspectiveentailsthatcompositionalityisnotabinary,butascalarconcept.

Thisviewissupportedbyfindingsfromvariouspsycholinguisticstudieswhichsuggestthattheliteralmeaningsthatareactivatedduringprocessingfacilitateidiomaticconstructioncomprehensiontotheextentthattheyoverlapwiththeidiomaticmeaning.Forinstance,Gibbsandcolleagues(GibbsandNayak1989;Gibbsetal.1989)demonstratedthatsubjectscandistinguishbetweenatleastthreeclassesofidiomaticconstructionsintermsoftheircompositionality,andthatsentencescontainingdecomposableconstructionsarereadfasterthanthosecontainingnondecomposableconstructions(seealsoGlucksberg1993;PetersonandBurgess1993;McGloneetal.1994;TitoneandConnine1994).

Aconstructionistperspectivefurthermoreentailsanumberofworkingassumptionsthatanadequatecompositionalitymeasureshouldbeabletoincorporate.First,anycomplexconstructionisassumedtocomprisea

4

5

Page 208: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 6 of 13

numberofsmallerconstructions,allofwhichmakeasemanticcontributiontothatcomplexconstruction(inotherwords,constructionsfurtherupintheconstructiconfeedintothesemanticsofconstructionsfurtherdownofwhichtheyarepart;seeGoldberg2006a:10).InthecaseofVNPconstructions,boththeverbandthenounphraseofaVNPconstructionareexpectedtomakeacontribution.

Second,asmentionedabove,constructionsareassumedtobedifferentlyentrenchedintheconstructicondependingon(amongotherthings)theirfrequencyofuse.Accordingly,atheoreticallyinformedmeasureshouldlicensethepossibilitythatcomponentwordsmakevariablylargecontributions(sincethereisnoreasontoassumethatverbsandnounphrasesnecessarilymakeequallylargecontributionstoVNPconstructions).Moreover,themeasureshouldbeitem-specificinthesensethatthecontributionofanycomponentwordcanbedifferentlylargedependingontheconstructioninwhichitoccurs.Forinstance,itisdesirabletohaveameasurethatlicensesthepossibilitythatthecontributionmadebypointinseeapointishigherorlowerthaninmakeapoint.

Anumberofcompositionalitymeasureshavebeenproposedthatstartoutfromtheassumptionthatsemanticregularityisafunctionofthesemanticsimilarityoftheconstituentwordsandthephrasalexpression.Somemeasurecompositionalityviatheabilitytoreplacecomponentwordswithoutlosingtheidiomaticinterpretationoftheconstruction.Forinstance,McCarthy,Keller,andCarroll(2003)automaticallyextractedverb-particleconstructionsfromtheBNC.Theythenretrievedthesemanticneighborsoftheseconstructionsfromanautomaticallyacquiredthesaurus.Therelativecompositionalityofagivenverb-particleconstructionwasdefinedastheoverlapofthesemanticneighborsandthecomponentwordsoftheconstruction(seealsoLin1999foranothersubstitution-basedmeasure).

Otherapproaches,includingtheonepresentedhere,measurecompositionalityviathesemanticsimilarityofthecontextsoftheconstructionscomparedwiththoseofitscomponentwords.Morespecifically,theworkinghypothesisisthatthesemanticsimilarityoftwowordsorconstructionsisreflectedintheextenttowhichtheysharecollocates.Collocatesofwordsare‘thecompanytheykeep,’thatis,wordsthatoccurina(usuallyuser-defined)contextwindowleftorrightofthewordmoreoftenthanwouldbepredictedonthebasisoftheword'sgeneralfrequency.Themoresemanticallysimilartwowordsorconstructionsare,themoresimilartheircontextswillbe.Togiveanexampleofastatisticallysophisticatedmeasurealongthoselines,SchoneandJurafksy(2001)extractedmultiword-expressions(MWEs)fromcorporausingLatentSemanticAnalysis.ThecompositionalityoftheseMWEswasmeasuredasthecosinebetweenthevectorrepresentation(containingcollocationfrequencies)oftheMWEandaweightedvectorsumofitscomponentwords,theassumptionbeingthatsmallcosinesindicatecompositionality.(Forothermeasuresbasedonsemanticsimilarity,see,e.g.,Bannard,Baldwin,andLascarides2003;Bannard2005.)

ThemeasuretobepresentedhereinalittlemoredetailisanextensionofBerry-Rogghe's(1974)work.Inafirststep,thesetsofsignificantcollocatesforeachcomponentwordWandtheconstructionCtheyarepartofaredetermined. Twoindexvaluesarethencombinedtoarriveataword'soverallcontributiontotheconstruction:R(whichisthenumberofcollocatessharedbetweenthewordandtheconstructiondividedbythetotalnumberofcollocatesoftheconstruction);andtheword's‘share’(whichisthenumberofcollocatessharedbetweenthewordandtheconstructiondividedbythetotalnumberofcollocatesoftheword).Bothindicescanrangebetween0(ifthereisnooverlapatall)and1(ifthecollocatesetsmatchperfectly).

Bycombiningthetwoindexvalues,weightedRassessestheoverallcompositionalityofaconstructionfromtwocomplementaryperspectives:Rreflectshowmuchofthesemanticsoftheconstructionisaccountedforbythecomponentword;conversely,thesharereflectshowmuchofitselfeachcomponentwordbringsintotheconstructionalmeaning.

Inordertoillustratethemotivationforthisapproach,considertheVNPconstructionmakeDETmark.Obviously,makeisahigh-frequencyverb,andsincethenumberofsignificantcollocatesawordwillattractisnaturallycorrelatedwithitsoverallfrequency,makehasmanysignificantcollocates.Thenounmark,onthecontrary,ismuchlessfrequent,andthereforeattractsfewersignificantcollocates.Insum,thecollocatesetsofmakeandmarkdifferconsiderablyinsize.Consequently,makestandsamuchhigherchancetocontributetoany

6

Page 209: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 7 of 13

construction'ssemanticsthanmark;whatismore,sincelexicallyfullyspecifiedcomplexconstructionscannotbemorefrequentthantheircomponentwordsandaccordinglyalwayshavecomparativelysmallercollocatesets,theresultingoverlapbetweenahighlyfrequentcomponentword'scollocatesandtheconstructionitispartofwillbequitehighbydefault.InthecaseofmakeDETmark,Ractuallyamountsto1.0:makeDETmarkattracts33significantcollocates,allofwhichitshareswithmake'scollocateset.Inotherwords,thesemanticcontributionofmaketomakeDETmarkmaybeconsideredextremelyhighwhenlookingonlyathowmuchoftheconstruction'ssemanticsisaccountedfor.Whenwelookintotheoppositedirection,however,weseethatmakecontributesonlyafractionofitsmeaningpotential:the33collocatesitshareswithmakeDETmarkconstituteonlyasmallshareofitstotalcollocatesetof4,234collocates.ThiscallsforareevaluationofthesemanticsimilaritybetweenmakeandmakeDETmark.BymultiplyingtheR-valuewiththesharevalue,weachieveexactlythat.

Formark,averydifferentpictureemerges:theoverlapbetweenmark'scollocatesetandthatofmakeDETmarkis31,whichagainindicatesahighsemanticcontribution,andsincemarkattracts298collocatesoverall,theshareof31outof298isrelativelyhigh.Thatis,markissemanticallymuchmoresimilartomakeDETmarkthanmakeisinthesensethatitismuchmoresemanticallytiedtothisconstruction,whilemakeoccursinsomanydifferentcontextssomuchmoreoftenthatonecannotspeakofaparticularlytightsemanticassociationbetweenmakeandmakeDETmark.

Clicktoviewlarger

Figure15.1. VNPconstructionsandtheirweightedR-values

TheoverallcompositionalityvalueofaconstructionCisdefinedasthesumoftheweightedcontributionsofallitscomponentwordsW(inthecaseofVNPconstructions,theverbandthenounphrase).Figure15.1providesanoverviewoftheresultsfortheVNPconstructions(theexactvaluescanbefoundinWulff2009:137).

AscanbeseeninFigure15.1,theweightedR-measureneatlyreproduceswhatwewouldexpectfromestablishedidiomtypologies:idiomslikemakeDETheadwayandtakeDETplungeranklowestincompositionality;metaphoricalexpressionslikemakeDETmarkandmeetDETeyeoccupythemiddleranks;quasi-metaphoricalconstructions,theliteralreferentofwhichisitselfaninstanceoftheidiomaticmeaning,likecrossDETfinger,holdDETbreath,andscratchDEThead,tendtorankevenhigherincompositionality;andmostoftheconstructionsthatwerenotpickedfromtheidiomdictionaryrankhighest,withwriteDETletteryieldingthehighestweightedR-value.

Notealsothatthemajorityofitemsisassignedafairlynoncompositionalvalueonthescalefrom0to1,whichtiesinnicelywiththefactthatmostofthemwereinfactobtainedfromanidiomdictionary.ItemssuchaswriteDETletterandtellDETstory,ontheotherhand,wereselectedtotestifitemsthatareintuitivelyassessedas(nearlyperfectly)compositionalareactuallytreatedaccordinglybythemeasure—soweightedRprovesveryaccurate,sincetheseitemsnotonlyrankhighest,butmoreover,theircompositionalityvaluesareveryhighinabsoluteterms(.73fortellDETstoryand.84forwriteDETletter).

WeightedRmaybeusednotonlyforassessingthecompositionalityofVNPconstructions,butisprincipallyapplicabletoanykindofconstruction.Beyondthat,weightedRmayproveausefultoolforquantitativeapproachestorelatedissuesinConstructionGrammarandcognitivesemantics.Forinstance,itcouldbeemployedtoquantifythedegreeofsemanticbleachingofverbsasafunctionofincipientgrammaticalizationprocesses.

Page 210: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 8 of 13

15.3.MeasuringtheFormalBehaviorofIdioms

Thereiscomparativelylittleresearchtodateontheformalbehaviorofidioms,particularlyempiricalstudies(exceptionsareNicolas1995;Moon1998).Moreover,moststudieshavebeenconcernedspecificallywiththesyntacticflexibilityofidiomsandmostlydisregardedotheraspectsofformalbehavior.Maybethebest-knownstudyonsyntacticflexibilityisFraser(1970),whoproposedanimplicational‘frozennesshierarchy’foridiomsthatcomprisedsixlevelsofsyntactictransformationssuchasnominalizationorparticlemovement.Thevalidityofthisfrozennesshierarchywastakenintoquestionbyvariousscholars(McCawley[Dong](1971)providedavarietyof(vulgar)counterexamples,Makkai(1972)evenmore,ifonlynotvulgar).

However,variouspsycholinguisticstudieshaveestablishedaconnectionbetweenassessmentsofanidiom'scompositionalityanditsformalproperties,includingpositivecorrelationsbetweenanidiom'scompositionalityanditsratedsyntacticflexibility(GibbsandNayak1989),lexicalflexibility(thatis,regardingthequestionifandtowhatextentmaterialcanbeinserted;Gibbsetal.1989),andratedsemanticproductivity(thatis,thecreationofvariantmeaningthroughwordsubstitutions;McGloneetal.1994).

Fromausage-basedperspective,ameasureofformalbehaviorshouldbemaximallydata-driveninthesensethattherearenoaprioriexpectationsastothesyntacticconfigurations,lexicalinsertions,ormorphologicalflexibilityofagivenconstruction;instead,thesecategoriesemergebottom-upbylookingatalargesampleofcorpusdata.Intheirtotality,thesecategoriesconstituteaconstruction'sformalbehavioralprofile(seeBarkema1994forsuchanapproachtonominalcompounds).

Wulff(2008b)developedsuchaformalbehavioralprofilebycodingall13,141instantiationsoftheafore-mentioned39VNPconstructionsaccordingtoanymorphological,lexical,orsyntacticvariationinstantiatedinthedata.Fromthisdata-drivenapproach,eighteenvariablesemergedthatmeasuredifferentaspectsofformalbehavior,includingthesyntacticconfigurationrealized,variousaspectsoflexico-syntacticbehavior(suchasthepresenceofadverbsorattributiveadjectivesprecedingthenoun),andthemorphologicalvariationspossibleinEnglish,includingtense,aspect,person,andsoon;Wulff(2009:151–53)providesanoverviewofthesevariables.Eachconstruction'sformalbehavioralprofilewasthencomparedwiththatofarandomsampleof1,151VNPconstructionsservingasabaseline.Morespecifically,foreveryoneoftheeighteenaspectsofformalbehavior,theinformationobtainedforeachvariablelevelwasconvertedintoanoverallindexvaluebydeterminingthegivenconstruction'sdeviationfromtheVNP-baseline.Toillustratehowthedifferentvariablelevelswereweighted,considerTable15.3,whichsummarizesthebehavioralprofileoffootDETbillregardingthemorphologicalvariableTense.Theobservedfrequencies(n )foreachvariablelevelarecomparedwiththefrequenciesexpectedfromtheVNPbaseline(n );deviationsbetweenobservedandexpectedfrequenciesaresquared(SSD),summed,andnormalized(NSSD).

Notehowbysquaringthedeviationsoftheobservedandexpectedfrequencies,weagainyieldaweightingofthecontributionofthedifferentvariablelevelsthatstandsinaccordwithconstructionalpremises:smalldeviationscontributeonlylittletotheoverallvalue,whilebigdeviationswillcontributemuchmore.Spacedoesnotpermitapresentationoftheresultsforallthirty-nineconstructionsandalleighteenaspectsofformalbehavior;anoverviewtableisgiveninWulff(2009:154–55),andWulff(2008b)discussesallresultsindetail.

Table15.3.FootDETbill'sformalbehavioralprofileforMF_Tense.

Tense n (%) n (%) n -n (%) SSD SummedSSD NSSD

past 9.17 25.97 −16.80 282.24 1,700.625 0.234

present 41.28 61.94 −20.66 426.836

future 8.26 1.57 6.69 44.756

nonfinite 41.28 10.51 30.77 946.793

obs

exp

obs exp obs exp

Page 211: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 9 of 13

15.4.ConstructionsareIdioms

15.4.1TheRelationshipBetweenIdiomaticVariationandIdiomaticity

Anotherquestionarisingfromausage-basedperspectiveonidiomsconcernstherelationshipbetweenaconstruction'ssemanticandformalbehaviorasitmanifestsitselfincorpusdata(insum,theconstruction's‘idiomaticvariation’)andnativespeakers’assessmentofaconstruction's‘idiomaticity,’whichisaninherentlypsychologicalconstruct.Ausage-basedapproachpredictsthatidiomaticityjudgmentswillbebased(atleastinpart)onthespeaker'sprocessingofhisorherlinguisticenvironmentwithregardtotheconstruction'sbehavior.Sowhichoftheseaspectsareassociatedwithidiomaticityjudgmentsandhowstrongly?

Inordertoaddressthisquestion,amultipleregressionanalysiswascomputedwiththeweightedR-valuesofeachVNPconstructionrepresentingitssemanticbehaviorandtheNSSDvaluesfortheeighteenaspectsofformalbehaviorrepresentingtheconstructions’formalbehavioralprofileastheindependentvariables,andnormedidiomaticityjudgmentsoftheseconstructionsasthedependentvariable. Takingallvariablesintoaccount,nearly57%ofthevarianceintheaverageidiomaticityjudgmentsisaccountedfor,ahighlysignificantresultthattestifiestoasolidrelationshipbetweenthevariablesandthejudgments(adjustedR =.565,p=.005**). Morespecifically,theregressionanalysisprovidesso-calledbetaweights forallvariables;thecloserabetaweightisto1,themoreimportant(inthesenseofcoveringvariance)itis.Variableswithbetaweights+.22canbeconsideredrelevantbecausetheyaccountforatleast5%ofthevariance.ConsiderTable15.4foranoverview.

AsTable15.4shows,themostimportantvariablesarethemorphologicalvariablesencodingthenumberoftheverb(MF_NumV)andthemoodoftheverb(MF_Mood),followedbytwolexico-syntacticvariables,LF_KindAdv(whichassesseswhatkindofadverbispresent,ifany)andLF_NoAdv(acountofthenumberofadverbsrealized,ifany).Nextinlinearecompositionalityandthesyntacticconfigurationinwhichtheconstructionisrealized(SF).ThemorphologicalflexibilityparametersMF_Voice(encodingthevoiceoftheverb)andMF_Neg(encodingtheverbaseithernegatedornot)alsoyieldsufficientlyhighbetaweightstobeconsideredrelevant.Thelastvariablewithavaluehigherthan+.22isthelexico-syntacticvariableLF_Addition(.265),whichcountsifthegivenconstructionwaslexicallymodifiedinanyway.Theconstruction'scorpusfrequency(CorpFreq)obtainsabetaweightof.209.TheseresultscoincideinastrikingfashionwiththoseofaPrincipalComponentsAnalysis(PCA)ofthecorpus-basedresults(seeWulff2008b:150–56).

Insum,speakersseemtorelyonavarietyofparameterswhenjudgingtheoverallidiomaticityofVNP-constructions,withaparticularfocusonthe

7

2 8

9

Page 212: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 10 of 13

Table15.4.Betaweightsforvariationparametersasdeterminedbyamultipleregressionofcorpusandjudgmentdata.

Variable Abbreviation Absolutebetaweight

numberofverb MF_NumV .757

moodofverb MF_Mood .695

kindofadverbinserted LF_KindAdv .651

numberofadverbsinserted LF_NoAdv .632

compositionality Comp .578

syntacticconfiguration SF .573

voiceofverb MF_Voice .351

negationofverb MF_Neg .275

materialinserted LF_Addition .265

corpusfrequency CorpFreq .209

personofverb MF_Person .197

gerundialverb MF_Gerund .16

tenseofverb MF_Tense .125

numberofNP MF_NumNP .109

insertedattributiveNP LF_AttrNP .083

realizationdeterminer MF_Det .055

aspectofverb MF_Aspect .046

insertedPP LF_PP .043

insertedrelativeclause LF_RelCl .038

insertedattributiveadjective LF_AttrAdj .032

morphologicalandlexico-syntacticvariabilityoftheverb.Furthermore,themultifactorialanalysissuggeststhattree-syntacticandsemanticfeaturesofthephraseplayanimportant,yetsecondaryrole—aresultthatstandsinoppositiontothewidelyheldassumptionthatcompositionalityisthemostdecisiveparametercontributingtoidiomaticity.Moregenerally,thesolidcorrelationsbetweenthecorpus-basedmeasuresandthejudgmentdatasupportaperformance-basedapproachtolanguage.

15.4.2TheExtendedConstructicon

Page 213: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 11 of 13

Howcouldthisprobabilisticandcomplexinformationbeimplementedintoexistingschematicmodelsoftheconstructicon?Asoutlinedabove,theconstructiconismainlyspecifiedwithregardtoitsverticalaxis,withdelexicalization(orschematization)beingtheprimaryprocesscreatingdiversificationalongthisaxis.Toacertainextent,idiomatizationcanbeconceptualizedasbeingdiametricallyopposedtodelexicalization:themoreidiomaticaconstructionis,thatis,themoreformallyandsemanticallyirregular,thelesslikelyitisthatthisconstructionwilldelexicalize.Inotherwords,onacontinuumofidiomaticphrasesrangingfromcollocationstoidioms,themoreidiomaticthephrase,thelessdelexicalizationpotentialithas.

Accordingly,theconstructiconcouldbeextendedbyahorizontalaxisasshowninFigure15.2,whichcutsacrosstherangeoftheverticalaxiswherefullylexicallyspecifiedcomplexconstructionsarelocated.Moreprecisely,onecanthinkoftheconstructiconasbifurcatingbeyondthelevelofwords,openingaquadrantspaceinwhichconstructionscanbepositionedaccordingtotheirdegreeofschematizationandidiomaticity.Thecloseraphrasalconstructionislocatedonthehorizontalaxistotheverticalaxis,themoresemanticallyandsyntacticallyregularitis(e.g.,writealetter);themoreformallyfrozenandsemanticallyopaqueaconstructionis(suchastaketheplunge),thefurtherawayfromtheverticalaxisthatconstructionispositionedonthehorizontalaxis.

Clicktoviewlarger

Figure15.2. Extendedschematicrepresentationoftheconstructicon

Thehighertheoverallidiomaticityofaconstruction,thelessitsrepresentationisconnectedtoitsconstitutinglexicalconstructions.Forinstance,taketheplungeisbothsemanticallyhighlyirregularandformallyrestricted,soitsoverallidiomaticityishigh;accordingly,taketheplungeisonlyweaklyconnectedwiththelexicalrepresentationsoftakeandplunge.Writealetter,ontheotherhand,isbothformallyandsemanticallyregular,soitsconnectionwiththelexicalrepresentationsofwriteandletterfurtheruptheverticalaxisiscomparativelystronger.Likewise,itismorestronglyconnectedwithotherlexicalconstructionsthatareassociatedwithwriteandletter,suchastype/composeoremail/paper,whichinturnmakeswritealetteralikelycandidateforsubsequentschematization.

15.5.ConclusionsandDesiderata

Thischapterstartedoutwithasummaryofearlyconstructionistresearchthatarguedinfavorofviewingidiomsnotasanomalies,butconstructionsthatareessentiallyonaparwithallotherkindsofconstructions.Fromausage-basedConstructionGrammarperspective,inparticular,wemaywanttotakethisstatementonestepfurther:allconstructionsareidiomatic.Whatmaylicensereferringtosomeconstructionsasidiomsandnotothersismerelyareflectionofthefactthateffectsofidiomaticvariationarebestobservableinpartiallyschematiccomplexconstructions—however,thisdoesnotmakethemfundamentallydifferentinnaturefromotherconstructions.

Viewingallcomplexconstructionsasmoreorlessidiomaticemphasizestheneedformoreempiricalresearchintothevariablesthatconstituteidiomaticvariationandtheirmeasurement.Speelmanetal.(2009)describetheagendaasfollows:

[A]tleastinsomecases,constructionsaremorestronglycharacterizedbythe(construction-internal)collocationsthatinstantiatethemthanbythesingleitemsthatinstantiatethem.Consequently,thesyntagmaticaxisshouldbecomeaconstitutivedimensioninacomprehensiveConstructionGrammarmodel.(Speelmanetal.2009:87)

Page 214: The Oxford Handbook of Construction Grammar

Words and Idioms

Page 12 of 13

AnadequatemodelingofthissyntagmaticaxisinConstructionGrammar(orwhatIreferredtoasthehorizontalaxisintheconstructicon)callsforfutureresearchgoingbeyondthestudiesandmeasuresreviewedhereinvariousregards.Astothesemanticbehaviorofconstructions,futurestudiesshouldsystematicallyexplorewhatkind(s)ofmeasure(s)aresuitedbesttomodeltheassociationstrengthbetweenconstructionsatdifferentlevelsofschematization.WeightedRisjustoneofmanycollocation-basedassociationmeasuresthatarecompatiblewithconstructionistpremisesthathighlightdifferentaspectsofthesyntagmaticandparadigmaticdimensionsofconstructionalinteraction.AfirstexampleofsuchacontrastiveanalysisispresentedbySpeelmanetal.(2009),whocontrastGriesandStefanowitsch's‘collostructions’(StefanowitschandGries2003)and‘co-varyingcollexemes’(GriesandStefanowitsch2004b;Stefanowitsch,thisvolume)formeasuringtheassociationstrengthbetweendifferentinflectionalvariantsofDutchattributiveadjectivesandtheirheadnouns.Anotherdesirablestrandoffutureresearchcouldaddresstherelationshipbetweenidiomaticvariationandidiomaticityinmoredepthbycomparingandvalidatingcorpus-derivedmeasuresofidiomaticvariationincontrolledexperimentalsettings.

Notes:

(1.)http://www.americancorpus.org/.

(2.)Itisalsoimportanttopointoutherethatoutsideof(directpredecessorsof)ConstructionGrammar,thereisavastamountofliteratureonidiomaticlanguage,someofitconsiderablypre-datingconstructionistwork,yetresonatinginessencewiththefundamentalclaimsmadeinConstructionGrammar.Exampleshereinclude(mainlyEuropean)phraseologicalresearch(CowieandHowarth(1996)provideaselectedbibliography),discourse-analyticalapproaches(Hymes's(1962)‘conversationalroutines,’PawleyandSyder's(1983)‘formulas,’orNattingerandDeCarrico's(1992)‘formulaicsequences’arebutthreeexamplesofdifferentnotionsofidiomaticstructures),andaplethoraofpsycholinguisticwork(CacciariandTabossi(1995)maybeagoodstartingpointhere).ThereadershouldfurthermorebereferredtoWray(2002a),whogivesanexcellentoverviewofphraseologicalresearchanditsimplicationsfortheoriesofthementallexicon.

(3.)Theconceptof“constructionalidioms”hasalsobeenappliedtolanguagesotherthanEnglish.ForatreatmentofparticleverbsinGerman,forinstance,seeBooij(2010,thisvolume),whichprovidesplentyofexamples.

(4.)DETstandsforanykindofdeterminer,includingazerodeterminer.

(5.)WhilethisisnotaVNPconstruction,itwasincludedinseveralpre-testsandisreportedalongsidetheVNPconstructions.

(6.)SignificantcollocateswerecalculatedusingaFisherYatesexact(FYE)test(seeStefanowitsch,thisvolume).CollocateshadtoyieldanassociationstrengthofFYE≥100toqualifyasmembersofacollocateset.

(7.)Thirty-ninefirst-yearstudentsofEnglishattheUniversityofSheffieldwereaskedtoassesstheoverallidiomaticityofthethirty-nineVNPconstructions.Eachparticipantwasgivenadifferentconstructionasareferenceconstructionandthenaskedtojudgetheidiomaticityoftheother38constructionsrelativetothisreferenceconstruction(fordetails,seeWulff2008b:28–33).

(8.)WhiletheadjustedR -valueonlyamountsto.565,ithastobeborneinmindthatthisvalueisloweredbytheoverallnumberofvariablesenteringintothecomputation:themorevariablesarerequiredtoaccountforallthevarianceinthedata,thelowertheadjustedR willbe.

(9.)Betaweightsquantifythecontributionofeachindividualindependentvariabletotheoverallcorrelationobserved.

StefanieWulffStefanieWulffisAssistantProfessorattheUniversityofFlorida.AfterherPh.D.(UniversityofBremen),shefirstworkedasavisitingscholarandlaterasalecturerattheUniversityofCaliforniaatSantaBarbara,aswellasaPostdoctoralResearchFellowattheUniversityofMichiganandanAssistantProfessorattheUniversityofNorthTexas.Hermainresearchinterestsareusage-basedcognitivelinguisticvariation(includingissuessuchaslexico-syntacticvs.constructional,aswellasdialectalvariation)andsecondlanguageacquisition.ShehaspublishedwidelyininternationaljournalssuchasCorpusLinguisticsandLinguisticTheory,CorporaandInternationalJournalofCorpusLinguistics,co-editedthevolumeCorpusLinguisticApplications:CurrentStudies,NewDirections(withStefanGriesandMarkDavis,2009)andistheauthorofRethinkingIdiomaticity:AUsage-basedApproach(2008).

2

2

Page 215: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 1 of 16

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0016

CollostructionalAnalysisAnatolStefanowitschTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterdiscussestheconceptofcollostructionalanalysis,afamilyofquantitativecorpus-linguisticmethodsthatallowresearcherstoexpressthestrengthoftherelationshipbetweenwordconstructionsandthegrammaticalstructurestheyoccurin.Itprovidesseveralcasestudiesandshowshowvaryingcollostructionalmeasurescanenlightenconstructionistanalysesoflexicalandgrammaticalconstructions.ThechapterexplainsthatalthoughadoptionofcollostructionalanalysisisacomparativelyrecentdevelopmentinConstructionGrammar,ithasalreadybeenappliedtoafairlywiderangeofconstructionsinthecontextofresearchquestionsrangingfromsystemicdescriptionoverlanguagevariationandchangetolanguageacquisitionandprocessing.Italsoaddressesimportantmethodologicalissuesofcollostructionalanalysissuchastheuseofinferentialstatistics,thecognitivemechanismsassumed,aswellasthechoiceofstatisticaltests.

Keywords:collostructionalanalysis,corpus-linguisticmethods,wordconstructions,grammaticalstructures,grammaticalconstructions,methodologicalissues,inferentialstatistics,cognitivemechanisms

16.1.Overview

Collostructionalanalysisisafamilyofquantitativecorpus-linguisticmethodsforstudyingtherelationshipsbetweenwordsandthegrammaticalstructurestheyoccurin.ItwasdevelopedbyStefanTh.GriesandAnatolStefanowitschasanextensionofexistingcollocation-basedmethods(itsnameisablendofthewordscollocationandconstruction).

Collostructionalanalysisdiffersfromothercollocation-basedmethodsinthatitisspecificallygearedtowardatheory-driveninvestigationofthelexis-grammarinterfaceandintheemphasisitplacesonstrictquantificationandinferentialstatistics.Itsrelationshiptoothercollocation-basedmethodsisdiscussedindetailinStefanowitschandGries(2009),itsrelationshiptoothercorpus-basedapproachestothesyntax-lexisinterfaceisdiscussedinRömer(2009).

Collostructionalanalysisiscompatiblewithanygrammaticaltheorythatseesgrammaticalstructuresas(potentially)independentfromlexicalitemsandthatattributesatleastsomerelevancetothefrequencydistributionsoflinguisticitems.Itisverynaturallyappliedinthecontextofconstructivistapproachestolanguage(seeGoldberg,thisvolume),especiallyusage-basedversionsofsuchapproaches(foritsmethodologicalplacewithincognitiveapproachestolanguage,seeTummersetal.2005;Gries,thisvolume).

Thefocusofmostexistingworkusingcollostructionalanalysisistherelationshipbetweenverbsandconstructionsencodingargumentstructure,tenseandaspect,andmoodandmodality.DuetothebroadviewonlanguagecharacteristicofConstructionGrammar,collostructionalanalysishasnotbeenconfinedtoproblemsofgrammaticaltheory,buthasalsobeenappliedtoresearchquestionsfromfirstandsecondlanguageacquisition,languagechange,andlanguageprocessing.

16.2.VariantsofCollostructionalAnalysis

Page 216: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 2 of 16

Therearecurrentlythreemajorvariantsofcollostructionalanalysis:simplecollexemeanalysis,distinctivecollexemeanalysis,andcovaryingcollexemeanalysis.Whileallofthesemethodsarebasedonthesameunderlyingmethod,eachofthemisgearedtowardanalyzingaparticularsubsetofquestionsconcerningtherelationshipbetweenwordsandgrammaticalconstructions.

16.2.1SimpleCollexemeAnalysis

Simplecollexemeanalysis(StefanowitschandGries2003)isastraightforwardreinterpretationofcollocationalanalysisfromagrammaticalperspective:insteadofinvestigatingtheco-occurrencerelationofwordpairs,itinvestigatesthatbetweenpairsofconstructions.Typically,oneoftheseconstructionsisalexicalconstruction(i.e.,aword),andtheotherisagrammaticalconstruction(e.g.,anArgumentStructureconstructionoraTense-Aspectconstruction).

Inordertodeterminetheassociationofaparticularlexicalconstructionl fromawordclassLtoaparticulargrammaticalconstructioncfromaclassofconstructionsC,thefollowingfrequenciesmustbedetermined:thefrequencyofl inc,thefrequencyofl inotherconstructionsofclassC,thefrequencyofothermembersofLinc,andthefrequencyofothermembersofLinotherconstructionsofclassC.Thesefrequenciesarethenenteredintoatwo-by-twocontingencytableasshowninTable16.1.

Next,itisdeterminedwhethertheobservedfrequencyofl incdeviatesfromitsexpectedchancefrequency,andifso,whetherthatdeviationisstatisticallysignificant.Theexpectedfrequenciesarecalculatedaccordingtothestandardmethodwidelyknownfromthechi-squaretest,butforstatisticalsignificancetestingtheFisher-Yatesexacttestistypicallychoseninsteadofthechi-squaretestandthep-valueofthistestisinterpreteddirectlyasameasureofassociationstrength.Ofcourse,othercontingencytestsmaybeusedinadditionorinstead(seesection16.4foradiscussionofstatisticalmeasuresandtheirroleincollostructionalanalysis).ThisprocedureisrepeatedforeachlofLinthecorpus,andthecollexemesarethenrankedbyassociationstrength.Thisrankingtypicallyconstitutesthedataonwhichtheresearcherbasestheirinterpretation.Lexicalconstructions

Table16.1.Frequencyinformationneededforasimplecollexemeanalysis

Wordl ofClassL OtherWordsofClassL Total

ConstructioncofClassC Freq.ofL(l )inC(c) FrequencyofL(¬l )inC(c) TotalfrequencyofC(c)

OtherConstructionsofClassC

FrequencyofL(l )inC(¬c)

FrequencyofL(¬l )inC(¬c)

TotalfrequencyofC(¬c)

Total TotalfrequencyofL(l ) TotalfrequencyofL(¬l ) TotalfrequencyofC

thatare(significantly)morefrequentthanexpectedarereferredtoas(significantly)attractedcollexemes,andthosethataresignificantlylessfrequentas(significantly)repelledcollexemes.

Asanexample,considerthequestionofwhichverbsarestronglyattractedtoorrepelledbytheDitransitiveconstruction.Thisconstructionoccurs1,824timesintheInternationalCorpusofEnglish(ICE-GB;Nelsonetal.2002).Thetotalnumberoftokensofalllexicalverbs,andthustheoverallnumberofverbalArgumentStructureconstructions,is136,551.Ifonenowdeterminesforeachverbitstotalnumberofoccurrencesintheditransitiveandinthecorpusasawhole,acontingencytablelikethatinTable16.1canbederivedfromthesefourfrequenciesandsubmittedtoacontingencytest.ThisproducesthelistofsignificantlyattractedandrepelledcollexemesshowninTable16.2.

Thetotalnumberofditransitivesinthecorpusis1,824,thetotalnumberofArgumentStructureconstructionsis136,551.ThefiguresdifferslightlyfromthoseusedinStefanowitschandGries(2003)becausetheDitransitiveconstructionhereincludescaseswithasententialsecondobject,particleverbsaretreatedasdistinctlemmas,andthetotalnumberofArgumentStructureconstructionsisestimatedmoreprecisely.Thep-valuesforthefirsttwoverbsareshownaszerobecausetheyaretoosmalltobecalculatedbythefloating-pointprocessorsfoundincurrentstandard-issuecomputers.

ThelistofattractedcollexemesinTable16.2isfairlytypicalforasimplecollexemeanalysisinseveralways.First,itshowsthatstatisticalassociationsbetweenverbsandArgumentStructureconstructions(andwordsandconstructions

i

i i

i

1…n

i

i i

i i

i i

Page 217: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 3 of 16

ingeneral)existandthatverbsdifferquitedrasticallywithrespecttothedegreeofthisassociation.Thisinitselfisanimportantinsightwithimplicationsforareassuchaslanguageacquisition(cf.Diessel,thisvolume,section19.4),languageprocessing(cf.Bencini,thisvolume),andlanguagechange(cf.Fried,thisvolume).Second,theverbsthataremoststronglyassociatedwiththeconstructionarethosethataretraditionallythoughtofasgoodexemplarsofverbsthat‘have’thecorrespondingcomplementationpattern.Thus,collostructionalanalysiscanuncoverhighlyconventionalizedverb-constructionpairingsevenundertheassumptionthatverbsandArgumentStructureconstructionsareindependentofeachotherinthe

Page 218: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 4 of 16

Table16.2.Significantlyattractedandrepelledcollexemesoftheditransitive(ICE-GB)

Collexeme FrequencyintheCorpus

FrequencyintheDitransitive

ExpectedFreq.IntheDitransitive

Fisher-Yatesp

ATTRACTED

Give 1,091 560 14.57 0

Tell 792 493 10.58 0

Send 295 78 3.94 4.13E-076

Ask 504 92 6.73 9.65E-074

Show 628 84 8.39 5.15E-056

Offer 196 54 2.62 3.73E-054

Convince 32 23 0.43 1.70E-036

Cost 65 23 0.87 9.04E-027

Inform 55 20 0.73 9.57E-024

Teach 92 23 1.23 7.94E-023

Assure 19 13 0.25 1.04E-020

Remind 41 16 0.55 7.25E-020

Lend 31 12 0.41 3.48E-015

Promise 43 12 0.57 3.26E-013

Owe 25 9 0.33 2.24E-011

(Thirtymorenotshown)

REPELLED

Make 1,865 3 24.91 3.39E-008

Do 2,937 12 39.23 2.56E-007

Find 854 2 11.41 7.96E-004

Call 616 1 8.23 2.32E-003

Keep 374 1 5 3.95E-002

Note:DatafromStefanowitsch(2006a).

constructicon.Third,theattractedcollexemesallhavea‘transfer’meaning,eitherinaliteralorinametaphorical

Page 219: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 5 of 16

sense(mostlyfallingundertheCOMMUNICATIONISTRANSFERmetaphor;cf.Reddy1979).ThisisthemeaningthatConstructionGrammaraccounts(Goldberg1995:38)assigntotheDitransitiveconstruction;theoccurrenceofverbswiththismeaningcanbeattributedtoasemanticcompatibilityeffect,andunderthisassumption,collostructionalanalysiscanbeusedtoidentifythemeaning(orrangeofmeanings)ofconstructions(cf.StefanowitschandGries2003).

Therepelledcollexemesarealsointerestinginseveralways.First,noneoftheminvolveanotionof‘transfer,’providingfurtherevidenceforthesemanticcompatibilityeffect.Second,theyaremostlyhighfrequencyverbsthatoccurditransitivelyonlyinasmallnumberofrelativelyfixedexpressions(dos.o.afavor,keeps.o.company).Assuch,theyareunlikelytohaveastrongimpactsemanticallyontheDitransitiveconstruction.Finally,theyareevidenceforthefactthatverbsandconstructionsareindeedindependentofeachothertoadegreethatallowsforverbstooccurinagivenconstructioneventhoughtheirdominantassociationsaretootherconstructions(cf.Stefanowitsch2006a:65).

Thenotionofrepelledcollexemescanbeextendedwithinasimplecollexemeanalysistoverbsthatoccurinthecorpusbutdonotoccurintheconstructioninquestionatall.Nonoccurrenceissimplythelimitingcaseoffrequencyofoccurrence,andassuch,theschemainTable16.1canbeusedtodeterminewhetherthenonoccurrenceofagivenwordinagivenconstructionislikelytobeduetochanceornot;(significantly)repelledcollexemeswithafrequencyofzeroinaconstructionarereferredtoas(significant)zerocollexemes(Stefanowitsch2008a).Inotherwordscollostructionalanalysisallowsustoderivenegativeevidencefromcorpora(Stefanowitsch2006a),contrarytoreceivedwisdomthatgoesbacktoChomsky(1957:15ff.)andwaswidelyacceptedevenamongcorpuslinguists(e.g.,McEneryandWilson2001:11–12).

Asanexample,taketheditransitiveagain.IntheICE-GB,thereare4,778verbtypesthatdonotoccurintheditransitiveatall.Ifonesubmitsthesetoasimplecollexemeanalysis,itturnsoutthatthisnon-occurrenceisstatisticallysignificantfor67ofthem.ThetopfifteenareshowninTable16.3.

Thelemmabe|havecapturescasesoftheclitic's,whichcanstandforisorhas;itwouldhavecontributedverylittletotheanalysistogothroughthe6,261tokensofthiscliticmanuallytocategorizethemappropriately.

Suchzerocollexemesareinterestinginseveralways.First,theirexistenceshowsthatcorporacontainnegativeevidencenotonlyintheorybutalsoinpractice.Suchnegativeevidencecouldprovidelanguagelearnerswithcrucialinformationaboutthesemanticandsyntacticboundariesofaconstruction.Second,forsomeofthemtheirnonoccurrenceisclearlymotivatedsemantically:itisdifficulttoimaginewhatitwouldmeantouse,forexample,be,have,orknowditransitively.Third,however,thereareverbsthatfallstraightforwardlyintothecategoryofliteralormetaphoricaltransfer:sayinTable16.3isoneexample,othersamongthefifty-threesignificantzerocollexemesnotshownhereincludeput,move,provide,speak,talk,andsuggest.TherewouldbenothingwrongsemanticallywithusingtheseverbsintheDitransitiveconstruction(theycanallbeusedmoreorlessfreelyinthesemanticallyverysimilarprepositionaldative,andmanyofthemoccurfreelyintheDitransitiveconstructioninotherlanguages).ThisisevidenceforarbitraryrestrictionsontheproductivityofconstructionsthatmustbeintegratedintoanyrealisticConstructionGrammar(mostcurrentversionsofConstructionGrammarcandosowithoutproblems,asconstructionsareseenasSaussureansignsandthusasinherentlyarbitrarytosomedegree).

Itisimportanttonotethatwhilesignificantzerocollexemesprovideimportantdatatotheresearcher,noconclusionscanbedrawnfromthefactthatazerocollexemeisnotsignificant.Amongthe4,711nonsignificantzerocollexemesof

Page 220: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 6 of 16

Table16.3.SignificantzerocollexemesoftheDitransitiveconstruction(ICE-GB)

Collexeme FrequencyintheCorpus

FrequencyintheDitransitive

ExpectedFreq.intheDitransitive

Fisher-Yatesp

Be 25,416 0 339.5 4.29E-165

be|have 6,261 0 83.63 3.66E-038

Have 4,303 0 57.48 2.90E-026

Think 3,335 0 44.55 1.90E-020

Say 3,333 0 44.52 1.96E-020

Know 2,120 0 28.32 3.32E-013

See 1,971 0 26.33 2.54E-012

Go 1,900 0 25.38 6.69E-012

Want 1,256 0 16.78 4.27E-008

Use 1,222 0 16.32 6.77E-008

Come 1,140 0 15.23 2.06E-007

Look 1,099 0 14.68 3.59E-007

Try 749 0 10 4.11E-005

Mean 669 0 8.94 1.21E-004

Work 646 0 8.63 1.65E-004

(Fifty-threemorenotshown)

Note:DataarefromStefanowitsch(2006a).SeeTable16.2foradditionalinformation.

theditransitiveintheICE-GB,therearesomethatclearlycannotoccurditransitivelyinEnglish(e.g.,destroy:*Theweedsdestroyedmetheflowerbeds;cf.thefullygrammaticalGermanequivalentDasUnkrauthatmirdasBlumenbeetzerstört).Otherscanbefoundoccasionallyincreativeusesoftheditransitive(e.g.,airlift:Therefugeeswerecutoff,sothemilitaryairliftedthemfoodandwater).Stillothersmayberarebutfullyconventionalintheditransitive(e.g.,postinBritishEnglish:Theydon’ttakecreditcards,soIpostedthemachequeinstead).TheseverbsaretooinfrequentandtheICE-GBissimplytoosmalltodecidewhethertheirnonoccurrenceintheditransitiveissignificantornot.

16.2.2DistinctiveCollexemeAnalysis

Distinctivecollexemeanalysis(GriesandStefanowitsch2004a)isanextensionofcollostructionalanalysisalongthelinesofacollocationalmethodforthedifferentiationofsynonymsdevelopedinChurchetal.(1991).Thismethodtakesallcollocatesoftwonear-synonymouswordsandcomparesthefrequencieswithwhicheachofthesecollocatesoccurswitheachofthesewords.Inotherwords,theassociationstrengthofacollocatetooneofthesesynonymsiscalculatedbycomparingitnottoitsoverallnumberofoccurrencesinthecorpus,buttoitsnumberofoccurrences

Page 221: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 7 of 16

withtheothersynonym.Thiswillpickoutthosewordsthatbestdistinguishthetwosynonymsinquestionfromeachother,whilesayingnothingdefinitiveabouttheirgeneralassociationtoeitherofthem.Distinctivecollexemeanalysisfollowsthesameprinciple,comparingallwordsthatoccurinagivenslotoftwosemanticallyorfunctionallysimilarconstructionsacrosstheseconstructions(e.g.,semanticalternativeslikethewill-futureandthegoing-to-futureorconstructionsinvolvedinso-called“alternations,”likeactivevs.passive,ditransitivevs.dative,orParticle-Firstvs.Object-FirstVerb-Particleconstructions).

Thecalculationoftheassociationstrengthisessentiallythesameasforasimplecollexemeanalysis,exceptthatthesecondconstructiontakestheplaceofthecorpus.Thatis,inordertodeterminetheassociationofaparticularlexicalconstructionl belongingtoawordclassLtotwogrammaticalconstructionsc andc ,bothbelongingtoaclassofconstructionsS,thefollowingfrequenciesmustbedetermined:thefrequencyofl inc ,thefrequencyofl inc ,thefrequencyofothermembersofLinc ,andthefrequencyofothermembersofLinc .Thesefrequenciesarethenenteredintoatwo-by-twocontingencytableasshowninTable16.4.

IfthistableissubmittedtoacontingencytestliketheFisher-Yatesexacttest,theresultwillbethatthelexicalconstructioninquestioniseithersignificantlymorefrequentinconstructionc ,orinconstructionc ,orthatitdoesnotdeviatesignificantlyfromitsexpectedchancedistribution.Thisprocedureisrepeatedforeachl ofLoccurringinc orc (or,ifoneisinterestedsolelyinalternatinglexemes,inbothc andc )andthecollexemesarethenrankedbyassociationstrength.Alexicalconstructionthatis(significantly)morefrequentinoneofthetwoconstructionsisreferredtoasa(significant)distinctivecollexemeofthatconstruction.

Asanexample,considertheditransitiveandthesemanticallyandfunctionallysimilarprepositionaldative.TakingallverbsthatoccurinbothoftheseconstructionsintheICE-GBandsubmittingthemtoadistinctivecollexemeanalysisproducesthelistsinTable16.5(asinthecaseofasimplecollexemeanalysis,theverbsarerankedbytheirp-value,whichistakentorepresentassociationstrength).

Ditransitivesanddativesincludeonlythosewithtwonominalobjects.Thetotalnumberofoccurrencesinthecorpusis1,035fortheditransitiveand1,919for

Table16.4.Frequencyinformationneededforadistinctivecollexemeanalysis

Wordl ofClassL OtherWordsofClassL Total

Constructionc ofClassC

FrequencyofL(l )inC(c ) FrequencyofL(¬l )inC(c ) TotalfrequencyofC(c )

Constructionc ofClassC

FrequencyofL(l )inC(c ) FrequencyofL(¬l )inC(c ) TotalfrequencyofC(c )

Total TotalfrequencyofL(l )inC(c ,c )

TotalfrequencyofL(¬l )inC(c ,c )

TotalfrequencyofC(c ,c )

theprepositionaldative.Theoverallnumberofconstructionsinthecorpusisnotrelevantforthistypeofanalysis.

ThelistinTable16.5isatypicalresultofadistinctivecollexemeanalysis.First,itshowsthatthetwoconstructionsinquestiondiffersemanticallyorfunctionallyinawaythatcanbereflectedintheverbsthatspeakersselecttheseconstructionsfor.Thisisnotthecaseforallfunctionaldifferences,ofcourse;apurelyinformation-structuraldifference,forexample,wouldnotbecomevisibleasaresultofadistinctivecollexemeanalysis.Second,thespecificverbsandverb-classesattractedtoeachofthetwoconstructionsgiveaclearindicationofwhatthosedifferencesare:thedistinctivecollexemesoftheditransitive(whichoverlapsubstantiallywiththosepickedoutbythesimplecollexemeanalysis)mostlyrefertoliteralormetaphoricaltransfersituationswheretheagentandtherecipientareincloseproximitytoeachother,whilethedistinctivecollexemesoftheprepositionaldative

i 1 2

i 1 i 2

1 2

1 2

1…n

1 2 1 2

i

1 i 1 i 1

1

2 i 2 i 2

2

i

1 2

i

1 2 1 2

Page 222: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 8 of 16

Table16.5.DistinctivecollexemesintheDitransitiveandthePrepositionalDativeconstructionintheICE-GB

Collexeme FrequencyintheDitransitive FrequencyintheDative Fisher-Yatesp

DistinctivefortheDitransitive

Give 461 146 1.84E-120

Tell 128 2 8.77E-058

Show 49 15 8.32E-012

Offer 43 15 9.95E-010

Cost 20 1 9.71E-009

Teach 15 1 1.49E-006

Wish 9 1 5.00E-004

Ask 12 4 1.30E-003

Promise 7 1 3.60E-003

Deny 8 3 1.22E-002

Award 7 3 2.60E-002

DistinctiveforthePrepositionalDative

Bring 7 82 1.47E-009

Play 1 37 1.46E-006T

Take 12 63 2.00E-004

Pass 2 29 2.00E-004

Make 3 23 6.80E-003

Sell 1 14 1.39E-002

Do 10 40 1.51E-002

Supply 1 12 2.91E-002

Note:DatafromGriesandStefanowitsch(2004a). mostly

refertotransfersituationswheretheagentmustovercomesomedistanceinordertoeffectthetransfer.Thisistrueevenofplay,whichintheICE-GBisusedintheconstructionsinquestionalmostexclusivelyinitssportssense(playtheballtos.o.).Acorrespondingsemanticdifferencebetweentheditransitiveandtheprepositionaldativehasbeenpositedindependentlybyotherresearchers(ThompsonandKoide1987;Goldberg1995:89ff.).

Itisimportanttokeepinmindthatdistinctivecollexemeanalysismeasuresdifferencesbetweenconstructionsand

Page 223: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 9 of 16

thereforehidessimilarities:wordsthatoccurequallyfrequentlyinbothconstructionswillnotappearamongtheresults,regardlessofwhethertheyarestronglyassociatedwiththeconstructionsinquestionornot;wordsthataresignificantlyassociatedwithbothconstructionsbutaresignificantlymorefrequentinoneofthemwillbecomedistinctiveforthatconstruction.Thus,astudythatisaimedatuncoveringboththesimilaritiesandthedifferencesbetweentwoconstructionshastocombineadistinctivecollexemeanalysiswithseparatesimplecollexemeanalysesforeachconstruction.

16.2.3CovaryingCollexemeAnalysis

Covaryingcollexemeanalysis(GriesandStefanowitsch2004a;StefanowitschandGries2005)isessentiallyanextensionofwhatStefanowitschandGries(2009)callstructure-sensitivecollocateanalysis.Inmanycollocation-basedstudies,co-occurrencesofwordsareinvestigatedonapurelylinearbasis:twowordsareconsideredpotentialcollocatesiftheyoccurnexttoeachotherinthecorpus.Whilethisprocedureoftenyieldsimportantinsights,ithaslongbeenrecognizedthatitissometimesadvantageoustoincludeintheanalysisonlywordsthatstandinaparticularsyntacticrelationtoeachother(cf.JustesonandKatz1995,wholookatwordsoccurringinthestructureAdj+N).Covaryingcollexemeanalysisappliesthisideatoconstructions:itaimstoidentifypairsofwordsthatoccurwitheachothermoreorlessfrequentlythanexpectedintwoslotsofthesameconstruction.Inotherwords,theconstructionprovidesaframeinwhichpotentialassociationsofpairsofwordsarestudied;asinthecaseofdistinctivecollexemeanalysis,theoverallcorpusfrequencyofthewordsisignored.

Inordertodeterminetheassociationoftwolexicalconstructionsl andl occurringinslotss ands ofaconstructionC,thefollowingfrequenciesmustbeknown:thefrequencywithwhichl occursins whenl occursins ,thefrequencywithwhichl occursins whenlexemesotherthanl occurins ,thefrequencywithwhichlexemesotherthanl occurins whenl occursins ,andthefrequencyofCwhenneitherl norl occurins ands ,respectively.Thesefrequenciesarethenenteredintoatwo-by-twotableasshowninTable16.6.

Thistableisthensubmittedtoacontingencytest;theprocedureisrepeatedforeverywordpairoccurringintheconstructioninquestion.Pairsofwordsthatco-occursignificantlymorefrequentlythanexpectedarereferredtoascovaryingcollexemes.

Table16.6.Frequencyinformationneededforacovaryingcollexemeanalysis

Wordl inSlots ofConstructionC

OtherWordsinSlots ofConstructionC

Total

Wordl inSlots ofConstructionC

Frequencyofs (l )&s (l )inC

Frequencyofs (¬l )&s (l )inC

Totalfrequencyofs (l )inC

OtherWordsinSlots ofConstructionC

Frequencyofs (l )&s (¬l )inC

Frequencyofs (¬l )&s (¬l )inC

Totalfrequencyofs (¬l )inC

Total Totalfrequencyofs (l )inC

Totalfrequencyofs (¬l )inC TotalfrequencyofC

Asanexample,consideronceagaintheDitransitiveconstruction.Therecipientargumentdoesnotvarymuch,itistypicallyahumanbeingoraninstitution,buttheverbandthethemeargumentcanvaryquitefreely.ApplyingacovaryingcollexemeanalysistotheverbnouninthethemeslotproducesthelistinTable16.7.

Allditransitiveswithnominalorsententialsecondobjectsareincluded,clauseswithincompletesecondobjects(falsestarts,etc.)wereremoved.

Casestudiesusingcovaryingcollexemetypicallyfindoneofthreefactorsgoverningtherelationbetweenthecollexemesinagivenconstruction:frames(e.g.,GriesandStefanowitsch2004b),image-schematiccoherence(e.g.,GriesandStefanowitsch2010),orprototypes(StefanowitschandGries2005:14ff.).TheresultsinTable16.7mostlyfallintotheformercategory:asks.o.aquestion,offers.o.ajob,writes.o.aletter,andotherverb-objectpairsrevealculture-specificknowledgeaboutwhatpeopledowithquestions,jobs,andlettersinsituationswherethereisapotentialrecipientinvolved.Table16.7revealsanadditionalfactor:anumberofwordpairsareevidenceforfixedor

1 2 1 2

1 1 2 2

1 1 2 2

1 1 2 2 1 2 1 2

1 1 1

2 2 1 1

2 2

1 1 2 2

2 2

2 1 1

2 2

1 1 2 2

2 2

1 1 1 1

Page 224: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 10 of 16

semi-fixedexpressions,forexample,thepragmaticallymotivatedexpressionstellyouwhatorwishs.o.(all)thebestortheidiomtakeaminute/anhour/andsoon.

Theexistenceofframe-basedcovaryingcollexemesprovidesfurtherevidenceforthesemanticsideofconstructions.Additionally,suchcollexemes,aswellasthosebasedon(semi-)fixedexpressions,provideevidenceforagrammaticalsystemthatdoesnotsimplystoreandprocessconstructionsatthemostgenerallevelbutthatisalsocapableofstoringexemplarsevenifthoseexemplarsinheritalltheirformalpropertiesfrommoregeneralconstructions.

16.2.4Extensions

Therearetwoobviousextensionsofthecollostructionalmethodsintroducedintheprecedingthreesubsections.Thefirstoftheseisconcernedwithincreasingthenumberofvaluesofthevariableconstructioninadistinctivecollexemeanalysis;thesecondisconcernedwithincreasingthenumberofvariablesinasimpleordistinctivecollexemeanalysis.

Page 225: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 11 of 16

Table16.7.CovaryingcollexemesintheverbandthemeslotoftheDitransitiveconstruction(ICE-GB)

CollexemeLinSlotA

CollexemeMinSlotB

FrequencyA(L)

FrequencyB(M)

Obs.Freq.A(L)&B(M)

Exp.Freq.A(L)&B(M)

FisherYatesp

ATTRACTED

tell what 122 37 25 3.75 8.56E-18

ask question 20 9 9 0.15 1.17E-17

tell that 122 24 19 2.43 9.19E-16

do good 12 4 4 0.04 5.66E-09

offer job 53 9 7 0.40 7.21E-09

take minute 11 5 4 0.05 1.88E-08

write letter 5 13 4 0.05 4.07E-08

guarantee place 7 9 4 0.05 4.99E-08

tell story 122 7 7 0.71 9.31E-08

send copy 72 10 7 0.60 2.12E-07

wish best 9 3 3 0.02 2.89E-07

wish success 9 3 3 0.02 2.89E-07

send cheque 72 5 5 0.30 6.66E-07

set deadline 3 2 2 0.00 4.14E-06

take hour 11 5 3 0.05 5.62E-06

(315morenotshown)

REPELLED

give what 551 37 2 16.92 5.26E-08

give that 551 24 3 10.97 5.49E-04

give pound 551 15 2 6.86 8.87E-03

give one 551 15 3 6.86 3.67E-02

Note:DataarefromStefanowitschandGries(2009).

Distinctivecollexemeanalysiswasinitiallyimplementedtocomparethefrequencydistributionofwordsacrosstwosemanticallyorfunctionallysimilarconstructions,butaspointedoutinGriesandStefanowitsch(2004a:123),itmayalsobeappliedinsituationswheretherearemorethantwosuchconstructions(e.g.,EnglishAnalyticCausative

Page 226: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 12 of 16

constructions(Gilquin2006),orTense-Aspectconstructions(Wulffetal.2009);cf.alsotheapplicationofthismethodinstudyingdiachrony(Hilpert2008)andvariation(MukherjeeandGries2009)).Suchamultipledistinctivecollexemeanalysisisastraightforwardextensionofdistinctivecollexemeanalysis,therequiredinformationisshowninTable16.8.

Multipledistinctivecollexemeanalysisraisesthequestionofwhichcontingencytesttouse,asexacttestsfortableslargerthantwo-by-twoarecomputationallytoocostlytouse(cf.section16.4).

Clicktoviewlarger

Figure16.1. Frequencyinformationneededforadistinctivecollexemeanalysisincludingchannel

Sincecollostructionalanalysiswasoriginallydevelopedtoanswerresearchquestionspertainingtogrammaticaltheory,itsbasicvariantsincludeonlytwovariables(lexicalconstructionandgrammaticalconstruction).However,ConstructionGrammardoesnotviewgrammarasanencapsulatedaspectof(linguistic)knowledge,whichmeansthatextra-grammaticalvariables(suchaschannel(spokenvs.written),register,variety,etc.)mayneedtobeintroduced.Inordertodealwithsuchadditionalvariables,StefanowitschandGries(2005)introduceamultidimensionalversionofcollostructionalanalysisthatsimplyaddsoneormorevariablestothedesign,producingamultidimensionalcontingencytablethatisthensubmittedtoanappropriatestatisticalproceduresuchasHierarchicalConfiguralFrequencyanalysis.Forexample,theinformationneededtoincludechannelisshownin16.1.(fromStefanowitschandGries2008).

Table16.8.Frequencyinformationneededforamultipledistinctivecollexemeanalysis

Wordl ofClassL OtherWordsofClassL Total

Constructionc Freq.ofL(l )inC(c ) Freq.ofL(¬l )inC(c ) TotalfrequencyofC(c )

Constructionc Freq.ofL(l )inC(c ) Freq.ofL(¬l )inC(c ) TotalfrequencyofC(c )

… … … …

Constructionc Freq.ofL(l )inC(c ) Freq.ofL(¬l )inC(c ) TotalfrequencyofC(c )

Total Totalfreq.ofL(l )inC(c ) Totalfreq.ofL(¬l )inC(c ) Totalfreq.ofC(c )

Thisvariantofcollostructionalanalysishasbeenusedtoincludethevariablelanguage(Englishvs.Russianvs.Germaninastudyofpostureverbsandprepositions;Schönefeld2006),dialect(Britishvs.AmericanEnglishinastudyofcovaryingcollexemesintheverbsslotsoftheinto-causative;Wulffetal.2007),andchannel(spokenvs.writteninadistinctivecollexemeanalysisofactiveandpassivevoice;StefanowitschandGries2008).

16.3.Applications

i

1 i 1 i 1 1

2 i 2 i 2 2

n i n i n n

i 1,2,…n i 1,2,…n 1,2,…n

Page 227: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 13 of 16

Theareaofgrammarresearchedmostintenselyfromacollostructionalperspectiveisargumentstructure,withstudiesontheDitransitiveconstructioninEnglish(StefanowitschandGries2003)andDutch(Colleman2009a),thedativealternationinEnglish(GriesandStefanowitsch2004a)andDutch(Colleman2009b),Englishcomplex-transitives(HampeandSchönefeld2006),CognateObjectconstructions(Höche2009),Causativeconstructions(GriesandStefanowitsch2003,2004b;StefanowitschandGries2005;Gilquin2006),Verb-Particleconstructions(GriesandStefanowitsch2004a;StefanowitschandGries2008),theWayconstruction(StefanowitschandGries2005),theAs-Predicativeconstruction(Griesetal.2005,2010a),Germanmotionverbs(GoschlerandStefanowitsch2010),andpostureverbsinEnglish,German,andRussian(Schönefeld2006).

Anotherrelativelyintenselyresearchedareaistense/aspect,withstudiesonprogressiveaspectinEnglish(StefanowitschandGries2003;WulffandRömer2009;Wulffetal.2009),theEnglishpasttense(StefanowitschandGries2003),differencesbetweentheEnglishWill-andGoing-to-Futureconstructions(GriesandStefanowitsch2004a;StefanowitschandGries2008),andtheequivalentSwedishfutureconstructions(Hilpert2006b).

Asmallernumberofstudiesdealswithmoodandmodality(theEnglishimperative,GriesandStefanowitsch2003;andGermanmodalinfinitives,Stefanowitsch2006b,2009),voice(EnglishActiveandPassiveconstructions,GriesandStefanowitsch2004a;StefanowitschandGries2008),andEnglishPossessiveconstructions(GriesandStefanowitsch2004a;StefanowitschandGries2005).

Insum,collostructionalanalysishassofarmainlybeenappliedtocoreareasofgrammar,buttherearesomecasestudiesofmore‘minor’constructions,suchasEnglish[go-V]and[go-and-V](Wulff2006),thealternationbetween[trytoV]and[tryandV](GriesandStefanowitsch2004a),[thinknothingofV-ing](GriesandStefanowitsch2003),[NPwaitingtohappen](GriesandStefanowitsch2003),andvariousmetaphoricalpatternssuchas[attheheartofNP](Stefanowitsch2005).

Allthesestudiesareconcernedtosomeextentwithasemanticand/orfunctionalcharacterizationoftheconstructionsinquestion;additionally,theseandotherstudiesdealwithlargerissuesingrammaticaltheory.

Therearerelativelyfewstudiessofardealingwithfirstlanguageacquisition(butcf.StefanowitschandGries2005ondifferencesbetweencaretakerlanguageandnormaladultlanguage).Secondlanguageacquisitioniscurrentlyamoreactivefocusofresearch,withstudiesshowingtheexistenceofconstruction–collexemeassociationsinlearnerlanguage(GriesandWulff2005,2009),changesinsuchassociationsinsuccessivestagesoflearnerlanguage(WulffandRömer2009;Wulffetal.2009),theusethatsecond-languagelearnersmakeofdistinctivecollexemesinidentifyingprototypicalinstancesofaconstruction(EllisandFerreira-Junior2009),andthenativizationofconventionalizedL2varieties(MukherjeeandGries2009)(seefurtherEllis,thisvolume,esp.section20.3.4).

Anotherfocusofmoretheoreticallyorientedstudiesislanguagechange,withstudiestrackingthedevelopmentofconstruction–collexemeassociationsacrosstime,specificallyfortheEnglishshall-future(Hilpert2006a),SwedishFutureconstructions(Hilpert2006b)andGermanicFutureconstructionsingeneral(Hilpert2008),andthedevelopmentoffairlyasacompromiserinEnglish(Margerie2008)(cf.Hilpert,thisvolume,esp.section25.2.12).Someofthesestudieshavebeencriticizedfortheirarbitrarydelineationofdiachronicstages(Stefanowitsch2006b;cf.theproposalforaninductiveidentificationofsuchstagesinGriesandHilpert2008)andforapplyingdistinctivecollexemeanalysisinawaythatisinconsistentwiththetheoreticalstatusofthismethod(Stefanowitsch2006c;seefurthersection16.5).

Synchronicvariationhasbeenlessintenselystudied,buttherearestudiesontheinfluenceofchannel(spokenvs.written,StefanowitschandGries2008)anddialect(Britishvs.AmericanEnglish,Wulffetal.2007)onconstruction–collexemeassociations,aswellasonfactorsinfluencinginflectionalvariationinDutch(Speelmanetal.2009).

ThecloseassociationbetweencollostructionalanalysisandtheUsage-basedModelhasalsoinspiredanumberofpsycholinguisticstudiesshowingthatcollostructionalassociationshaveaninfluenceonlanguageprocessing.Ithasbeenshownthatcollostructionsarerelevanttosyntacticpriming(Gries2005andSzmrecsanyi2005)andtoparsing(Griesetal.2005,2010a;Wiechmann2008b;Zeschel2008b)andthatthenegativeassociationstrengthofzerocollexemesinfluencesacceptabilityjudgments(Stefanowitsch2006a,2008a).

Finally,collostructionalanalysishasbeenappliedtotheinductiveidentificationofsemanticclasses(GriesandStefanowitsch2010),diachronicstages(GriesandHilpert2008)andrelationshipsbetweendialects(MukherjeeandGries(2009)byapplyingsomeformofclusteranalysistotheoutputofcollostructionalanalyses).

Page 228: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 14 of 16

16.4.MethodologicalIssues

16.4.1UseofInferentialStatistics

Collostructionalanalysisissometimescriticizedforusingmeasuresofassociationstrengthderivedfrominferentialstatisticsinsteadofsimplefrequencies(cf.,e.g.,Bybee2010:97ff.;Schmid2010).Thiscriticismisnotbasedongeneraltheoreticalormethodologicalfoundations;instead,itisclaimedthatinferentialstatisticsistoocomplextoprovideanadequatemodelofhowspeakersprocessandrepresentdistributionalpropertiesofthelanguageinput(Bybee2010:97);additionally,itisclaimedthatthereisempiricalevidencethatfavorsmodelsbuiltonfrequencyovermodelsbuiltonprobabilitiesorinferentialstatistics(Bybee2010:97ff.).However,thereisnoevidencetosuggestthatthefirstclaimistrue(seealsosection16.4.2.).Astothesecondclaim,therearestudiesthathavetreatedthisissueasanempiricalone;allofthesehavefoundthat,whilefrequencyisagoodapproximationofassociationstrength,itisoutperformedbymoresophisticatedstatisticalmeasureswhenitcomestopredictingtaskbehaviorinpsycholinguisticexperiments(Griesetal.2005,2010;Wiechmann2008a;cf.alsothecomparisonoffrequencyanddistinctivityinEllisandFerreira-Junior2009andtheremarksonpredictivepowerinGriesandStefanowitsch2004a:119ff.).

Aseparateissueiswhetherp-valuesfrominferentialstatisticaltestscanbeuseddirectlyasmeasuresofassociationstrength.Incollostructionalanalysis,itisassumedthatthisisindeedpossible,sincep-valuesassessthejointinfluenceofcorpussizeanddistributionandarethereforeanaturalmeasureofcognitiveentrenchment(StefanowitschandGries2003:239;seealsoPedersen1996).However,thisisonlytruewithinagivencorpus;wheredatafromdifferentlysizedcorporaarecompared,effectsizesmustbeusedinsteadofp-values(Wiechmann2008a).

16.4.2CognitiveMechanismsAssumedbyCollostructionalAnalysis

Asecond,relatedcriticismaimedatcollostructionalanalysisistheclaimthatnocognitivemechanismisproposedonwhichpsycholinguisticinterpretationsofthemethodcanbebased(Bybee2010:100).Thiscriticismisfalse:proponentsofcollostructionalanalysisspecifythatmechanismexplicitlyascuevalidityinthesensepopularizedwithinlinguisticsbyRosch(e.g.,1978;cf.GriesandStefanowitsch2003:273;Wiechmann2008a:257;EllisandFerreira-Junior2009).Specifically,collostructionalanalysismaybeunderstoodasacombinationofcueavailabilityofagivenwordforagivenconstruction(whichcanbecalculated

Table16.9.Collostructionalanalysisandcuevalidity

mostsimplybydividingthefrequencyofthewordintheconstructionbythetotalfrequencyoftheconstruction),andthecuereliabilityofagivenwordforagivenconstruction(whichcanbecalculatedmoststraightforwardlybydividingthefrequencyofthewordintheconstructionbythetotalfrequencyofthewordinthecorpus).Table16.9showsthisschematically.

Cuevalidityisawell-understoodmechanism;crucially,itisbasedatleastonrelativefrequenciesandoftenonacomparisonofobservedandexpectedfrequencies(e.g.,KruschkeandJohansen1999),lendingfurthersupporttothe

Page 229: The Oxford Handbook of Construction Grammar

Collostructional Analysis

Page 15 of 16

useofinferentialstatisticsratherthanrawfrequencies.Itmustbenotedthatnotallusesofcollostructionalanalysisareintendedtobeinterpretedinapsycholinguisticframework,butwheretheyarenot(e.g.,inthedomainofdiachronicorregionalvariation),itremainsunclearhowtheresultsareplausiblyinterpreted(cf.Stefanowitsch2006c).

16.4.3ChoiceofStatistic

Althoughthelogicofcollostructionalanalysisiscompatiblewithanystatisticaltestsuitablefortheevaluationof(2-by-2)contingencytables,almostallstudiessurveyedinsection16.3usetheFisher-Yatesexacttest,forthesimplereasonthatitmakesnodistributionalassumptionsandthatitcanbeappliedtosamplesofanysize(cf.Pedersen1996).Versionsofcollostructionalanalysisthatusecontingencytableslargerthan2-by-2typicallyuseindividualbinomialtestsforeachcell(oreachrelevantcell)ofthetable,asexacttestsarenotavailableorarecurrentlystillcomputationallytoocostlytoapplytheminarealistictimeframe.

Itisanopenempiricalquestionwhichstatisticaltestismostplausiblefromapsycholinguisticview.Wiechmann(2008a)comparesanumberofmeasures,usingreading-timedatafromHareetal.(2003)asabenchmarkandfindsthattheFisher-YatesexacttestperformsbetterthanotherwidelyusedstatisticaltestsandisoutperformedonlybyPedersenandBruce's‘minimumsensitivity’measure.

16.5.Conclusion

Theadoptionofcollostructionalanalysis(andotherquantitativecorpus-linguisticmethods)isacomparativelyrecentdevelopmentinConstructionGrammar,butthestudiesdiscussedinthischaptershowthatithasalreadybeenappliedtoafairlywiderangeofconstructionsinthecontextofresearchquestionsrangingfromsystemicdescriptionoverlanguagevariationandchangetolanguageacquisitionandprocessing.

Manyoftheseresearchquestionshavesofaronlybeentoucheduponandadditionalresearchisneededtogainaclearerpictureofthedescriptiveandexplanatoryscopeofcollostructionalanalysis.Clearly,themethodsmustbeappliedtoawidertypologicalrangeoflanguagesandtoawiderrangeofphenomenaatdifferentlevelsoflinguisticstructure.Inparticular,theclaimstopsycholinguisticrelevancemustbesubjectedtointensescrutiny;thiswillnotonlyhelptoanswerthequestionwhich—ifany—associationmeasurebestmodelslexeme-grammarassociationsinthemindsofspeakers,butitwillalsohelpresearchersgainabetterunderstandingoftherelationshipbetweenwordsandconstructionsinlanguageprocessing(Bencini,thisvolume,esp.section21.2.5)andlanguageacquisition(Diessel,thisvolume;Ellis,thisvolume).

Nevertheless,theresultsofthebodyofcollostructionalliteraturecurrentlyavailablealreadyhaveimportantramificationsforgrammaticaltheoryatamoreabstractlevel:theyshowthatthecombinatorycapacityofconcretelexicalconstructions(words)andmoreabstractgrammaticalconstructions(phrasalandclausalstructuralschemas)isnot(ornotexclusively)acategoricalissue,butonethatmustbeframedinprobabilisticterms.

Presumably,allversionsofConstructionGrammararecapableofintegratingthisinsightinprinciple.However,itseemsobvioussuchanintegrationismorenaturalandmoreeasilyaccomplishedinthemoreusage-basedversionssuchasCognitiveGrammar(Broccias,thisvolume),CognitiveConstructionGrammar(Boas,thisvolume),andRadicalConstructionGrammar(Croft,thisvolume)thaninSign-BasedConstructionGrammar(Michaelis,thisvolume)oreventraditionalBerkeleyConstructionGrammar(Fillmore,thisvolume),astheformer,butnotthelatter,accordacentraltheoreticalstatustodegreesofentrenchment(cf.alsoGoldberg2006a:215).

Thus,whileallversionsofConstructionGrammarcanprofitfromcollostructionalanalysisasatoolintheory,notallofthemwillbeabletoaccommodatetheresultswithoutchangingtheirtheoreticalframeworktosomeextent.Whetherthistheory-changingpotentialiseveractualizedisimpossibletopredict,especiallyintherapidlychanginganddiversifyingfieldofConstructionGrammar.

AnatolStefanowitschAnatolStefanowitsch(Ph.D.2001)isProfessorofEnglishLinguisticsattheUniversityofHamburg.HereceivedhisPh.D.fromRiceUniversity,Houston,Texas,andwasProfessorofEnglishandGeneralLinguisticsattheUniversityofBremenbeforetakinghiscurrentposition.Hismajorresearchinterestsincludeconstructiongrammar,metaphor,andquantitativecorpuslinguistics,withspecialemphasisonassociationsbetweenlexicalitemsandotherelementsofsyntacticandsemanticstructure.Hehaspublishedanumberofpapersoncollostructionalanalysis(manyofthemco-authoredwithStefanTh.Gries)andwasafoundingeditorofthejournalCorpusLinguisticsandLinguisticTheory.

Page 230: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 1 of 16

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0017

AbstractPhrasalandClausalConstructionsThomasHoffmannTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterexaminesabstractphrasalandclausalconstructions,themostcomplexandschematicendoftheconstructiconcline.Itoutlineshowconstructionistapproachescandescribeandmodeleventhemostabstractofsyntacticstructures.Thechapterdiscussesthearrangementofdeclarative,interrogative,imperative,andrelativeclausesindefaultinheritancenetworksandpointsoutdifferencesbetweenthoseConstructionGrammarframeworksthattakeausage-basedapproachandthosewhichdonot.ItalsoanalyzesEnglishcomparativecorrelativeconstructionandprovidesempiricalevidenceforaspecificintonationalsignatureofthisconstruction.

Keywords:abstractphrasalconstruction,clausalconstruction,constructiconcline,syntacticstructures,ConstructionGrammar,usage-basedapproach,correlativeconstruction,intonationalsignature

17.1.Introduction

OneofthemaintenetsofallConstructionGrammarapproachesistheclaimthatalllevelsofgrammaticaldescriptioninvolveconstructions,thatis,conventionalizedform-meaningpairings(cf.Croft,thisvolume;Goldbergthisvolume;TrousdaleandHoffmann,thisvolume).Thus,insteadofassumingaclear-cutdivisionoflexiconandsyntax,ConstructionGrammariansassumethatallconstructionscanbeplacedonalexicon-syntaxcontinuum(a‘constructicon,’Fillmore1988;seealsoGoldberg,thisvolume).Animportantparameteralongwhichtheentriesofthisconstructiconaresupposedtovaryistheirdegreeofschematicity(Jackendoff2002a:176;Goldberg2003:220;CroftandCruse2004:255):thelexicalendofthisclineincludessubstantiveconstructionswhicharefullyspecifiedwithrespecttotheirphonology,suchasthewordgreen(whichhasaspecifiedform/ɡɹiː n/).Abstractconstructions,ontheotherhand,suchastheschematicDitransitiveconstruction[SBJVerb-TNSOBJ OBJ ],whichonlycontainslotsthatcanbefilledbyvariouselements(suchasBradbakedhiswifeacakeorGraemesentTomanemail),constitutethemoresyntacticendofthiscontinuum.Inadditiontoschematicity,CroftandCruse(2004:255)pointoutthatconstructionscanalsobeclassifiedwithrespecttotheircomplexity:constructionsaresaidtobeatomicifthey“cannotbefurtherdividedintomeaningfulparts”(CroftandCruse2004:255),otherwisetheyareconsideredcomplex.Accordingtothisclassification,boththesubstantiveconstructiongreenaswellastheschematicwordclassconstructions[N]or[ADJ]areatomicconstructions.TheschematicDitransitiveconstruction[SBJVerb-TNSOBJ OBJ ],aswellasthemostlysubstantiveidiomspill-TNSthebeans,ontheotherhand,arecomplexsincetheycanbedividedintosmallermeaningfulparts(inthelatterexamplethemeaningofspillcanbeparaphrasedas‘divulge’andthebeansmeans‘information’;cf.Nunbergetal.1994;CroftandCruse2004:252).Morphemes(cf.Booij,thisvolume),idioms(Wulff,thisvolume),aswellaspartiallylexicallyfilledgeneralphrasalpatterns(forvariousexamples,cf.Boas,thisvolume;Fillmore,thisvolume;Jackendoff,thisvolume;Kay,thisvolume)thenconstituteconstructionsthatliesomewherein-betweenatomicandsubstantiveconstructions(traditionallylabeled‘lexical’items)andcomplexandschematicconstructions(phenomenathatwouldusuallybecalled‘syntactic’).

1 2

1 2

1

Page 231: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 2 of 16

Theaimofthepresentchapteristofocusonthemostcomplexandschematicendoftheconstructiconcline.Inparticular,IwilllookatAbstractClausal(declarative,interrogative,imperative,andrelativeclauses;cf.section17.1)andvariousFiller-Gapconstructions(suchastopicalizedclauses,[Mybagels] ,sheloves ,orwh-interrogatives,[What(books)] dotheylove ?,inwhichaconstituent(here[Mybagels] and[What(books)] )appearsinanonargumentposition;cf.section17.3).Moreover,IwillshowhowinEnglishtheFiller-Gapconstructioninteractswithprepositionstranding(Whodidshetalkto?)andpied-piping(Towhomdidshetalk?;cf.section17.3.1),andhowthecomplexitiesoftheEnglishrelativeclauseconstructionscanbeaccountedfor(section17.3.2).Ontopofthat,Iwillalsoillustratehowintonationcanbeanintegralformfeatureofconstructions(e.g.,theEnglishComparativeCorrelativeconstructionTheX-ertheY-er;cf.section17.3.3).Besides,contrarytotheprejudicethatConstructionGrammaronlyfocusesonperipheralandidiosyncraticexceptions,throughoutthechapter,Iwillnotonlyshowthatconstructionistapproachescandescribeandmodelthemostabstractandschematicofsyntacticstructures,butthatsuchanalysesalsofurtherourunderstandingofthephenomenainquestion.

Finally,notethatinordertokeepthechaptermaximallyaccessibletoallreaders,IwillnotemployaparticularConstructionGrammarformalismbutwillinsteadshowhowconstructionalapproachesingeneralanalyzethesephenomena(thoughoccasionallyIwillillustratecompetinganalysesofcertainphenomena;cf.thediscussionofsubject-verbagreementinsection17.2.).Moreover,throughoutthechapter,Iwillpointouthowcertainanalysesareaffectedbythedifferentviewsofcompleteinheritanceapproaches(suchasSign-BasedConstructionGrammar,Michaelis,thisvolume,orBerkeleyConstructionGrammar,Fillmore,thisvolume)andusage-basedanalyses(suchasRadicalConstructionGrammar,Croft,thisvolume,orCognitiveConstructionGrammar,Boasthisvolume).

17.2.AbstractClausalConstructions

(1)a.Shelovesyou(yeah,yeah,yeah)(songbyTheBeatles)b.Doyouloveme?(songbyTheContours)c.Loveme(fromthesongLovefoolbyTheCardigans)d.Thespywholovedme.(JamesBondmovietitle)

Theclausesin(1)clearlydifferwithrespecttotheirprototypicalsemanticfunction(cf.,e.g.,GinzburgandSag2000:6–10):

•thedeclarativeclausein(1a)isastatement,

•theinterrogativeclausein(1b)isaquestion,•theimperativeclausein(1c)isacommand,and

•therelativeclausein(1d),wholovedme,contributestotheidentificationoftheantecedentnoun's(spy)reference.

Moreover,ontopofthesesemanticdifferencestheclausesalsoexhibitdistinctsyntacticproperties:

•in(1a)bothoftheverb'sarguments(theloveraswellasthelovedone)arerealizedsyntacticallywiththeloverfunctioningasthesubjectoftheclauseandprecedingtheverb,whilethelovedoneisrealizedasapostverbalelement;

•inadditiontothesefeatures,theinterrogativein(1b)alsocontainsaclause-initialauxiliarydo;•incontrasttothis,themainverbisininitialpositionin(1c),whiletheloverargumentissuppressed;

•finally,intherelativeclause(1d)aclause-initialrelativepronoun(who)functionsasaclause-internalargument(thelover)whilealsobeingco-referentialwiththeclause-externalantecedentnoun.

Furthermore,ontopoftheseproperties,relativeclausesalsodifferfromtheotherthreeclausetypesinnotbeingabletofunctionasmainclauses(cf.thenoninterrogative*Wholovedme.).

Alltheseproperties,especiallythefactthattheclausesaboveconstitutepairingsofformandmeaning,allowthem

i i

i i i i

Page 232: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 3 of 16

tobeinterpretedasabstractclausalconstructions.NotethatthisviewisinstarkcontrasttoPrinciple-and-Parameterapproaches(cf.Chomsky1995,2000b,interalia),whichclaimthatconstructionsareonlyepiphenomena,thatis,merelyacollectionofstructuresthataretheresultoftheinteractionofuniversalprinciplesandparametersettings(cf.Chomsky1995:129).Yet,evidencefrompsycholinguistic(Bencini,thisvolume),neurolinguistic(Pulvermüller,Cappelle,andShtyrov,thisvolume),aswellasfirstandsecondlanguageacquisitionstudies(Diessel,thisvolume;Ellis,thisvolume)showsthatconstructionsarenotjustepiphenomena,butactuallythecentralbuildingblocksofmentalgrammars.Moreover,asBorsley(2006,2007)pointedout,inordertoachieveanalysesthatcancapturethedifferencesbetweenclausetypes(aswellasotherabstractconstructions)Principle-and-Parameterapproachesheavilyrelyoninvisibleelementsthatheadfunctionalprojections(suchasvP,orC;cf.Chomsky1995,2000b)andwhichhavespecificpropertiesandinterpretations—ananalysis“thatislittledifferentfromaconstruction-basedaccountthatassociatesthesamesetofpropertiesandinterpretationsdirectly”(Sag2010:488).

Onceclausetypesareidentifiedasstoredabstractconstructions,followingGinzburgandSag(2001:39;cf.alsoSag2010:501),itbecomespossibletocapturetheirpropertiesinataxonomicnetwork, whichtakesintoaccountthedifferencebetweenCoreClauseconstructions(i.e.,thosethatcanappearasindependentsentences,namelyDeclarativeClause,ImperativeClause,andInterrogativeClauseconstructions)andthosethatobligatorilyneedtobeembeddedinlargersyntacticcontexts(i.e.,RelativeClauseconstructions;cf.SagandGinzburg2000:38).Alltheseconstructionsthenconstitutethemostabstractclausaltypeswhichinmanycaseswillhavemorespecificsubordinateconstructionalinstantiations(with,e.g.,interrogative-clausesinEnglishincludingyes/no-questionssuchas(1b),aswellaswh-questionssuchasWhodoyoulove?orrelativeclausesincludingthat-andØ-relative-clauseslikethemanthatshelovedandthemansheloved;cf.below).Moreover,theseAbstractClausalconstructionswillalsointeractwithotherschematicconstructionsthatregulatesubject-verbagreementortherealizationofobligatorycomplements:

(2)a.Iloveyou./Helovesyou.Theyloveyou./etc.b.DoIloveyou?/Doesheloveme?/Dotheyloveyou?/etc.c.Themanwholovesyou/Themenwholoveyou

Astheexamplesaboveshow,Standard(BritishandAmerican)Englishhassubject-verbagreement(ofthegrammaticalfeaturesnumberandperson)in,forexample,presenttensedeclarativeclauses(2a),interrogativeclauses(2b),aswellasrelativeclauses(2c).Notethatdespitethefactthatdifferentconstructionalapproachesmightstateorformalizethisphenomenoninvariousways(cf.,e.g.,3a–d),theywouldallagreethatsubject-verbagreementishandledbyanabstractschematicconstruction(thoughapproachesalsodifferastowhetheritexistsasanindependentconstructionoronlyaspartofalargerconstruction;cf.below):

(3)a.[GF (>…)] ⇔[ …T+agr …] (CulicoverandJackendoff2005:192)b.MVerb-TA(CroftandCruse2004:288;Croft,thisvolume)

c.

(SBCGanalysisImodeledonSag,Wasow,andBender2003:107;Sag,personalcommunication)

c′.

(SBCGanalysisII;Sag,personalcommunication;cf.alsoMichaelis,thisvolume)

2

i k S i k

Page 233: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 4 of 16

d.

AstheParallelArchitecture(cf.Jackendoff,thisvolume)representationbyCulicoverandJackendoffin(3a)shows,thisapproachmodelssubject-verbagreementasanabstractconstructionthatlinksthefirstgrammaticalfunctionGF(i.e.,thesubject)withthetense-carryingverbalelement(T)andviacoindexationofsubjectandagr(eement)suffixensuresfeatureconcord.Whatdistinguishestheirapproachfrommostotherconstructionalapproachesisthattheytreattheconstructionin(3a)asapurelyformalonewithoutasemanticpole(i.e.,ameaninglessone).Thisispossiblesincetheyassumethatprototypicalconstructionsareclassicform-meaningpairingsbutthattherealsoexistotherswhichcanbe‘defective,’thatis,withoutaphonological,syntactic,orsemanticpole(cf.Jackendoff2002a:180–81,thisvolume).

Incontrasttothis,inCroft'sRadicalConstructionGrammar(RCG,thisvolume)theconstructionin(3b)isausage-basedmorphologicalconstructionthatarisesbecauseofthehightypefrequencyofnumerousdifferenttokenssuchas(2)ortobemoreprecise,accordingtoRCG'susage-based,bottom-upperspective,fromevenmorespecificvalenceconstructionssuchastheIntransitiveconstruction(IntrSbjIntrV)ortheTransitiveconstruction(TrSbjTrVTrObj).Besides,incontrasttotheParallelArchitecturemodel,RCGassumesthatagreementdoesinfacthavealanguage-specific,semanticfunctioninthatitallowstheidentificationoftherolesthatparticipantsplayinaspecificconstruction(withdifferentlanguagesmarkingdifferentparticipantsinsemanticallyequivalentconstructions;cf.Croft2001:209–13;cf.alsoBarðdal2006).

Next,(3c)and(3c′)showtwocompetingSign-basedConstructionGrammar(cf.Michaelis,thisvolume)approachestosubject-verbagreement(for3rdpersonsingularsubjects;non-3rdpersonsubjectswouldbemodeledaccordingly).Bothpartialfeaturestructuresarepartsofverballexicalentries(andthuspartofacompletesigns,i.e.,formandmeaningpairings).In(3c)theverb'sCAT(egory)featureincludesinformationonitsAGR(eement)features(whichforEnglisharegoingtoincludethefeaturesPERSONandNUMBER;cf.Sag,Wasow,Bender2003:107).As(3c)shows,thevalueofAGRhastobeidentical totheAGRfeaturesoftheverb'ssubject(ortobemorepreciseitsXARG,thedistinguishedargumentoftheverbthatistoberealizedassubject).Incontrasttothis,followinganHPSGanalysisofPollardandSag(1994:chapter2),in(3c′)thereisnoAGRfeatureintheverb'sCATentrybuttheverbsimplyrequiresitsdistinguishedargumentXARGtohavetheAGRfeatures3rdpersonsingular(thusagreementisinessencemodeledsimilarlytogovernment;cf.Kathol1999:229).

Finally,(3d)givesthepartofthesyntacticpoleoftheFluidConstructionGrammar(FCG)Subject-Predicateconstructionwhichregulatessubject-verbagreement(forthefullconstruction;cf.SteelsanddeBeule2006:220).Similarly,to(3c),bothsubject(thesubject-unit)andtheverbalelement(predicate-unit)havefeaturesthatexpresspersonandnumber.Theirconcordisensuredbytokenidenticalvariables(cf.thevariables?nand?pinbothsubjectandpredicate).

Whiletheexamplesin(3)illustratethewiderangeofpossibleconstructionalanalysesofsubject-verbagreement,theonethingtheyshareisthefactthattheyeschewtransformationsorfeature-drivenmovementtomodelthis(oranyotherphenomenon).Instead,theconstructionsin(3)arestaticconstraintsthatlicensewell-formedsyntacticstructures(makingallConstructionGrammarstypesofconstraint-basedgrammars).

Inthesamevein,theobligatoryrealizationofsyntacticcomplementsofaverb(orso-calledArgumentStructureconstruction,cf.Boas,thisvolume)isalsoconsideredtobelicensedbyanabstractconstruction.Infact,thegenerationofanyactuallinguisticexpression(‘construct’)isassumedtoinvolvethecombinationofvarious

3

4

Page 234: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 5 of 16

schematicconstructions.AsillustratedinFigure17.1, inSBCGthesentences(1a,b,d)would,forexample,partlybecreatedbythecombinationoftheSubject-Predicateconstruction(whichregulatessubjectandpredicatecombinations)andtheHead-Complementconstruction(whichsanctionstheappearanceofallnonsubjectarguments).Again,variousConstructionGrammarapproachesdifferslightlyintheiranalysesofthesephenomena:theParallelArchitectureModel,BerkeleyConstructionGrammar(BCG),andFCGalsoassumetwosimilarconstructions.IntheParallelArchitectureModel,thesewouldbetwomeaninglessconstructions,suchasthe‘PositionofSubject’constructionandthe‘VPconstruction’(cf.CulicoverandJackendoff2005:192,225–27);inBCGtheconstructionsarecalled‘Subject-Predicateconstruction’and‘HeadplusComplementconstruction’(cf.Fillmore,thisvolume;KayandFillmore1999:7,13);FCGemploystheSubject-PredicateconstructionandPredicate-Unitconstructionmentionedabove.Incontrasttothis,RCGwouldpostulateonlyasingleconstructionthatlicensesalltheverb'sarguments(heretheTransitiveconstructionTrSbjTrVTrObj;cf.CroftandCruse2004:287).

Clicktoviewlarger

Figure17.1. TheinteractionofAbstractClause,Subject-Predicate,andHead-Complementconstructionsinataxonomicconstructionnetwork(thepartslicensedbytheHead-Complementconstructionhavebeenitalicized/underliningmarksthepurveyoftheSubject-Predicateconstruction)(simplifiedversionmodeledonGinzburgandSag2000:39andSag2012:142)

Yet,whiledifferentConstructionGrammarschoolsmightpostulatedifferentconstructionsandtaxonomiesfortheabovephenomena,allschoolsassumethatalllanguage-specificgeneralizationscanbecapturedbymultipleinheritancenetworkssuchasFigure17.1(cf.,e.g.,GinzburgandSag2000:5–8;Goldberg2003:222–23;CroftandCruse2004:262–65).Theseinheritancenetworksallowconstructstobeformedfreelyaslongastheconstructionstheyconsistofdonotconflict(cf.Goldberg2006a:22;moreonthisbelow). Consequently,“[c]onstructionalapproachessharewithmainstreamgenerativegrammarthegoalofaccountingforthecreativepotentialoflanguage(Chomsky1957,1965).Thatis,itisclearthatlanguageisnotasetofsentencesthatcanbefixedinadvance.Allowingconstructionstocombinefreelyaslongastherearenoconflicts,allowsfortheinfinitelycreativepotentialoflanguage”(Goldberg2006a:22).AllConstructionGrammarapproachesarethereforetypesofgenerativegrammars.However,incontrasttoPrincipleandParameterapproaches,ConstructionGrammarsdonotemploytransformationstogeneratestructures,but,aspointedoutabove,employinteractingconstraints/templatesinstead.

ThetaxonomyinFigure17.1mightseemtobelargelymotivatedbysyntacticphenomena.AsCroftandCruse(2004:320–21)pointout,however,constructionaltaxonomiescanalsobeinfluencedbysemanticsimilarity.Inparticular,theycitethefollowingpieceofevidencefromthepositiveandnegativepredicationphrasesinEnglish:

(4)positivedeclarativea.FORM:BE +ADJ b.FORM:VERBHeishappy.Hesleeps.(5)negativedeclarativea.FORM:BE +NOT+ADJb.FORM:DO+NOT+VERBHeisn’thappy.Hedoesn’tsleep.(6)positiveimperativea.FORM:be +ADJb.FORM:VERBBehappy!Sleep!(7)negativeimperativea.FORM:be +NOT+ADJb.FORM:DO+NOT+VERBa′.FORM:DO+NOT+BE+ADJDon’tsleep!Don’tbehappy!

As(4a)shows,adjectivesrequireaformofthecopulaBEinEnglishtocreatewell-formedpositivedeclarativeclauses.SincethenegativeadverbNOTneedstoattachtoanauxiliaryverbinEnglish,negativedeclarativeclauseswithadjectivescanthensimplybeformedbyattachingNOTtoBE(cf.5a).Incontrasttothis,fullverbsneedadditionaldo-supportinnegativepredicationphrases(cf.5b).Movingontopositiveimperatives,(6)showsthatthesepatternwiththeirdeclarativecounterpartsinthatadjectivesneedtoco-occurwithbe(6a),whilefullverbscanappearontheirown(6b).Thenegativeimperatives,however,differfromtheirpositivealternatives:sinceNOTcouldattachtobe,theconstructionthatmightbeexpectedwouldbebe+NOTADJ.Yet,ascanbeseenin(7a),theconstructioninsteadexhibitsdo-support.ThisleadsCroftandCruse(2004)toassumethe

4

5

AUX 6

AUX

AUX

AUX

Page 235: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 6 of 16

taxonomyshowninFigure17.2forEnglishimperativepredicationphrases.

AsCroftandCruseargue(2004:320),illocutionaryforceseemstobemoreimportantfortheimperativetaxonomyinEnglishthandifferencesinpredicatetype.InFigure17.2thisisrepresentedbyverbalandadjectivalpredicatesnotbeingsubnodesofanindependentPREDICATIONconstructionthatthensimplycombineswithindependentnegationandimperativeconstructions(sincethiswouldleadtotheungrammaticalstructurebe +NOT+ADJin(7a)).Instead,negativecommandsinEnglishformtheirownnodeintheimperativenetworkandincludetheconstraintthattheyhavetobeintroducedbyDon’t_—regardlessofwhetherthepredicationphraseinvolvesanadjectiveorverb.

Yet,whiletheexistenceofidiosyncraticpropertiessuchasexhibitedby(4)–(7)wouldleadtothepostulationofanindependentconstructioninmostconstructionistapproaches,thereisdisagreementastotheroleofthefrequencyintheemergenceoftaxonomicconstructionalnetworkssuchasFigure17.2(cf.CroftandCruse2004:276–78):ontheonehand,“completeinheritance(constructional)models” (suchasSBCGorBCG)assumethatonly“idiosyncraticmorphological,syntactic,lexical,semantic,pragmaticordiscourse-functionalpropertiesmustberepresentedasanindependentnodeintheconstructionalnetworkinordertocaptureaspeaker'sknowledgeoftheirlanguage”(CroftandCruse2004:263).Incontrasttothis,usage-based(KemmerandBarlow2000;Bybee,thisvolume)constructionalapproaches—includingCognitiveGrammar(Langacker1987,2005;Broccias,thisvolume),RadicalConstructionGrammar(Croft2001,thisvolume),andCognitiveConstructionGrammar(e.g.,Lakoff1987;Goldberg2003,2006a;Boas,thisvolume)—advocatethatthefrequentuseofaconstructioncanleadtoitbeingcognitivelyentrenched,evenifitspropertiescanbecompletelyderivedcompositionallybymoreabstractsuperordinateconstructions(Goldberg2006a).

Clicktoviewlarger

Figure17.2. ThetaxonomyofEnglishImperativeconstructions(CroftandCruse2004:321)

Inessence,usage-basedapproachesidentifytwotypesofinputfrequencyeffects:hightokenfrequencyleadstotheentrenchmentofphonologicallyfilledconstructions(Langacker1987:59–60;CroftandCruse2004:292–93).Thus,highlyfrequentimperativeconstructssuchasDon’tworry!orBehappy!(oreventheircombinationasinthesongDon’tworry,behappy!)canbecomeentrenchedinthementalconstructicon.Ontheotherhand,hightypefrequency,thatis,patternsthatareobservedwithmanydifferentitemscanleadtotheemergenceofmoreabstractschematicsuperordinateconstructions(thustheabstractVERB!nodeinFigure17.2wouldbeanentrenchedabstractionoveragreatnumberofspecificconstructssuchasJump!,Eat!,Pray!,Love!,Sleep!,etc.).Inordertodistinguishthesevariouslevelsofschematicityinusage-basedapproaches,Traugott(2008a,2008b)introducedthefollowingterminologythathasbeenwidelyadopted,particularlyinconstructionistworkonlanguagechange(cf.,e.g.,Bergs2008;Trousdale2008c;TraugottandTrousdale2010onEnglish,orFried2008onCzech): entrenched,substantiveconstructionswhichappearatthebottomofthementalconstructicon(suchasBehappy!orDon’tworry!)arecalled‘micro-constructions.’Ahightypefrequencyofmicro-constructionscanthenleadtothegeneralizationofmoreabstract‘meso-constructions’suchasBeADJ!orDon’tV!(oratanevenmoreabstractlevel,theDon’tPREDPHRASEconstruction,sincetherecanbemorethanonelevelofmeso-constructions).Finally,themostschematicconstructionssuchastheImperativeconstructionatthetopnodeofFigure17.2arelabeled‘macro-constructions.’Sincemostconstructscouldalsobegeneratedbydrawingonexistingmicro-andmeso-constructions,formanyphenomenathequestionofwhetheramacro-constructionhastobepostulatedatallisanempiricalonefromausage-basedperspective.Barðdal(2008,2011a),forexample,hasclaimedthattheproductivityofabstractconstructionscanbeseenasaninversecorrelationoftypefrequencyandsemanticcoherence,withhighlyabstractmacro-constructionsonlyarisingiftheunderlyingmeso-constructionshaveahightypefrequencyandahighdegreeofvarianceinsemanticdistribution.

Asrecentresearchhasshown,thereisstrongempiricalevidencethatalllevelsoflinguisticform—phonology(Bybee2000,2001;Pierrehumbert2001),morphology(Bybee1985,1995;HayandBaayen2005),andsyntax(Saffran,2001,2002;CasenhiserandGoldberg2005;StefanowitschandGries2005)—areheavilyaffectedby

AUX

7

8

Page 236: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 7 of 16

inputfrequencyeffects.Suchfindingsthereforesupporttheinterpretationofconstructionaltaxonomiesasusage-basednetworksofmicro-,meso-,andmacro-constructionsthatariseinabottom-upfashion.Consequently,taxonomicnetworkssuchasFigure17.2canalsobeinterpretedasprobabilisticactivationmodelsinwhichhigher(tokenandtype)frequencycorrelateswithcue-strength(inthecognitiveconnectionistsenseofinteractionactivationmodelssuchas,e.g.,Rumelhart,McClelland,andthePDPresearchgroup1986;BatesandMacWhinney1989;cf.alsoGries2003a:157–84;T.Hoffmann2011:271–72):thehighertheinputfrequencyofaparticularconstruction,thestrongeritisgoingtobeentrenchedintheneuralnetwork.

Thisimportanceofinputeffectsnaturallyledtoanincreasedinterestofusage-basedConstructionGrammariansinstatisticalcorpusapproaches(fordetails,samplestudies,andfurtherreferences,cf.Gries,thisvolume;Hilpert,thisvolume).Forpurposesofillustration,takethedistributionoftheImperativeconstructionintheBritishcomponentoftheInternationalCorpusofEnglish(ICE-GB;cf.Nelson,Wallis,andAarts2002)showninTable17.1.

IftheICE-GBdataaretakentoberepresentativeoftheinputofImperativeconstructionsthatBritishEnglishspeakersareexposedto,thenTable17.1implies

Table17.1.TheImperativeconstructionintheICE-GBcorpus

Positive Negative(Don’t…)

V! 1,430 87 1,517

BeADJ 29 10 39

1,459 97 1,556

Note:ThedataofTable17.1wereretrievedusingICE-GB'sFuzzyTreeFragmentFTFoption.Allinall,fourquerieswerespecifiedsothattheprogramretrievedallimperativeclauseswithacopular(foradjectivalimperatives)andnoncopularmainverb(forfullverbimperatives)bothwithnegative,aswellasnonnegativeverbphrases.NotethattheseresultsdifferslightlyfromasimpleFTF‘PU|CL(imp)’searchthatlooksforallunitstaggedas‘imperative’andyields1,598hits.Thissimplersearch,however,alsoincludesfragments(suchasWellImeandon’t<-><.>f<ICE-GB:S1A-005#089:1:A>),whichwereexcludedbythefourindividualFTFsearches.

thatpositivefullverbimperatives(V!)arebyfarthemostfrequenttypeandshouldthereforebefarmorestronglyentrenchedinthementalconstructiconthantheothermeso-constructions.Inaddition,asaFisherExactTestofthedatainTable17.1shows,therearefarmorenegativeadjectiveimperativesthanexpectedbychance(p<0.001,oddsratio5.66),makingthosefewinstancesofDon’tbeADJtokensmoresalient,whichmightalsoaddtothisstructurebeingmoredeeplyentrenched(cf.alsoEllis,thisvolume).

Moreover,asacollostructionalanalysisoftheverbalimperativesintheICE-GBcorpushasshown(StefanowitschandGries2003;cf.alsoStefanowitsch,thisvolume),therealsoseemtobespecificlexicalizationsthataresignificantlyassociatedwiththeimperativeconstruction(and,consequently,moredeeplyentrenched):thefivemoststronglyattractedverbsinStefanowitschandGries'sstudywerelet,see,look,listen,worry(2003:231–33),withworryexclusivelyappearinginnegativeimperatives(thusimplyingamorestronglyentrenchedDon’tworrymicro-construction).

Besidestheissueofwhich(more)substantiveconstructionsareentrenchedinthespeakers’mentalgrammars,theConstructionGrammar-inspiredanalysisoftheirdataalsoallowedStefanowitschandGriestodrawimportantconclusionsaboutthesemanticmeaningandpragmaticforceofImperativeconstructionsinEnglish.Previousstudieshademphasizedthatimperativesentencesusuallyhaveadirectivefunctionofaspeakerwantingthehearertoperformarequestedactionthatisdesirabletothespeaker(cf.StefanowitschandGries2003:231).Incontrasttothis,themoststronglyattractedverbsinStefanowitschandGries'sstudyimpliedameaningthatcouldbeseenasmorebeneficialtothehearer,withthespeakersuggestingandadvisingratherthancommanding.Moreover,theactionsrequestedbytheseverbsseemedtobemoredirectedatensuringfuturecooperationand

9

Page 237: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 8 of 16

interactionbetweenspeakerandhearerthanbeingpurelyforthebenefitofthespeaker(StefanowitschandGries2003:233).

WhilethediscussionsofarillustratedhowConstructionGrammarcanhandlestandardimperativestructuresinEnglish,theinterpretationofabstractsyntacticstructuresasform-meaningpairingsobviouslyalsoallowsananalysisofmoreidiosyncratic,noncanonicalstructuressuchasparticipialimperativesinDutch(OosterhofandCoussé2010:136),insubordinationeffects(Evans2007),thatis,nonfiniteimperativesusedasmainclauses(Gras2010:84)orconditionalusesoftheimperativeinEnglish(8)(cf.Stefanowitsch2003;Takahashi2004;DancygierandSweetser2005),Dutch(9),andRussian(10)(cf.FortuinandBoogaart2009):

(8)MoveandI’llknockyouout(9)Leen hem20.000Eurouitenuzulthetnietterugkrijgen.lend-IMP him20,000euroout andyou willit notget.back‘Lend him20,000euros,andyouwon’tgetit back’.(Trouw[Dutchnewspaper],July14,2007;citedinFortuinandBoogaart2009:644)(10)[A]sprosiunego,kakprojtikfabrike—ontebjabutask-IMP-SGathim,howgotofactory,heyou-SGobol’etprezreniemsnogdogolovy.pour.overwith.contemptfromfeettillhead‘Butaskhimhowtogettothefactory,andhewilllookatyoucontemptuouslyfromheadtotoe.’(Strugackie,Gadkielebedi;citedinFortuinandBoogaart2009:646)

Insteadoftheirprototypicaluseasdirectivesorsuggestions/advice(cf.above),theimperativepartofthesentenceintroducesaconditionin(8)–(10).AsFortuinandBoogaart(2009)pointout,theidiosyncraticpropertiesofsuchstructurescanbeexplainedbytheinteractionoftwomoreabstractconstructionswhichtheconditionalimperativeconstructioninheritsfrom:namely,theAbstractImperativeconstructionandthe‘Conditional-and-Construction(cf.YoudrinkonemorecanofbeerandIamleaving;CulicoverandJackendoff1997;formoredetails,cf.Stefanowitsch2003,aswellasFortuinandBoogaart2009:648–70).

17.3.Filler-GapConstructions

Besidesabstractclausalpatterns,onephenomenonthathasreceivedconsiderableattentioninthesyntacticliteratureare‘Filler-Gap’constructionssuchas(11)(examplesadaptedfromSag2010:491;cf.alsoPollardandSag1994:157)

(11)a.[Mybagels] ,sheloves_ .[topicalizedclause]b.[What(books)] dotheylove_ ?[wh-interrogative]c.(theman)[who(sebooks)] sheloves_ .[wh-relative]d.[Whataplay] ,hewrote_ ![wh-exclamative]e.[themorebooks] theyread_ ,[the-clause][themoreintelligent] theyget_ .

Inallexamplesof(11)aconstituentinanonargumentposition(the“filler,”i.e.,[Mybagels] in(11a),[What(books)] in(11b),etc.)appearsinadifferentpositionthanitwouldnormallybeexpectedtooccurinadeclarativesentence(cf.Shelovesmybagels.).This‘normal’position,withwhichthefillerisstilllogicallyassociated,iscalleda“gap”(indicatedby“_ ”in(11)).

Filler-GapconstructionshaveplayedaparticularlyimportantroleinthevariousversionsofMainstreamGenerativeGrammar(MGG)(cf.,e.g.,Chomsky1973,1977,1986,1995,2000b):sincethedisplacementofthefillersacrossalloftheaboveclausetypesseemstobeunbounded(cf.,e.g.,Mybagelstheysaidshelikes.orThemorebooksJohnclaimedtheysaidtheyread…)andalsosubjecttothesamekindofrestrictions(theso-called‘islandconstraints’; cf.Ross1986),thedifferentstructuresin(11)arealltreatedasbeinggeneratedbyonesimplemechanism(variouslycalled‘wh-movement,’‘A-bar-movement,’etc.)inMGG.

Whilethisaccountmightappearappealinginthatitallowsonetocapturethevariousphenomenain(11)byasinglegeneralization,itturnsoutthatsuchananalysisglossesoverimportantdifferencesbetweenthesestructures.AsSag(2010)hasshown,despitetheirsharedproperties,thereareseveralparametersofvariation

i i

i i

i i

i i i i

i i

i

i

i

9

Page 238: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 9 of 16

thatdistinguishthevariousFiller-Gapconstructionsfromeachother,namely(cf.Sag2010:490–96):

•whetherthefillerdaughterisadistinguishedwh-element(cf.,e.g.,[What(books)]in(11b))ornot(cf.,e.g.,[Mybagels]in(11a));•thepossiblesyntacticcategoryofthefiller(cf.,e.g.,Whatbookdidsheread?vs.*Whatbooksheread!)andtheheaddaughter(cf.,e.g.,interrogativeswhichdonotallowathatcomplementizer*Whothatlikesyou?vs.correlative-comparativessuchasthemorethatisconsumedthebetter.LOB L-J42);

•whethertheclauseinquestionexhibitsobligatory(Whatdoesshelove?vs.*Whatsheloves?),optionalsubject-auxiliaryinversion(Themoremyheadhasached,themorehaveI/Ihaveindulgedinhumour;fromCulicoverandJackendoff1999:559)ornoinversion(cf.therelative*themanwhodidshelove);

•whethertheclausecanbenonfinite(cf.,e.g.,therelativetheroominwhichtomeet)ornot(cf.topicalizedclausessuchas*Mybagels(forthem)tolike);

•whethertheclauseinquestioncanbeanindependentclause(e.g.,interrogatives)ornot(e.g.,relativeclauses;cf.above).

Foradetailed(Sign-Based)ConstructionGrammaranalysisthatcapturesalloftheseproperties,IreferthereadertoSag(2010).Inthefollowing,Iwillinsteadfocusonthreephenomenathatillustratehowconstructionistapproachescanprovideastraightforwardexplanationof

•awidelydiscussedphenomenonofEnglishgrammar(namelyprepositionplacement)thatinteractswithmanyoftheaboveFiller-Gapconstructions,

•thecomplexitiesdisplayedbythesetofrelativeconstructions,

•thecrucialrolethatintonationplaysasanintegralformfeatureofconstructions(asshown,e.g.,bythecomparativecorrelativeconstructions).

17.3.1Filler-GapConstructionsandPrepositionPlacement

Besidestypeandtokenfrequency,anotherusage-basedfactorleadingtotheentrenchmentofabstractschemataiseaseofprocessing(Hawkins2004):ifthesamecontentcanbeexpressedbytwocompetingstructuresandoneoftheseiseasiertoprocessthantheother,thenthesimplerstructurewillbepreferredinperformance.Consequently,itwillbeusedmoreoftenwithagreaterrangeoflexicalizations,whichincreasesitstypefrequencyandultimatelyleadstoitbeingmorecognitivelyentrenchedthanitsalternative(Hawkins2004:6).

Take,forexample,thecaseofprepositionplacement,avariablefeatureofEnglishsyntax(T.Hoffmann2011):inseveralsyntacticcontextssuchasquestions,aprepositioncaneitherappearwithoutanadjacentNPcomplementinfrontofthegap“_ ”(‘prepositionstranding,’asin12a),oritcanbepartofthefiller(‘prepositionpied-piping,’asin12b):

(12)a.[Whichstudent] didyouask(_ )Maryabout_ ?b.[Aboutwhichstudent] didyouaskMary_ ?(examplestakenfromHawkins1999:277)

Crosslinguisticallyprepositionpied-pipingisfarmorecommonthanstranding(Hawkins1999:277).Hawkins(1999,2004)arguesthatthisisduetothefactthatfromaprocessingperspectiveprepositionstrandingisfarmorecomplexthanpied-piping.Firstofall,prepositionstrandingcanleadtogardenpatheffects:in(12a)afterencounteringaskthehumanparsermighterroneouslytrytointegratewhichstudentasthefillerofanobjectgap(_ ),anincorrectparsethatisnotpossibleinthepied-pipedversion(12b);seefurtherHawkins's(2004)‘AvoidCompetingSubcategorizers’principle.Second,thepied-pipedalternativeallowsthefillertobeinterpretedafterprocessingthemainverb,whichactsasthemainsubcategorizeroftheclause,whilethegapinthestrandedconstructionismoredeeplyembeddedwithinaPPthatitselfisembeddedintheVP(ageneralprocessingprinciplewhichHawkins(2004:210–15)calls‘ValencyCompleteness’).Whilethesefactorsexplainthetypologicalrarityofprepositionstranding,thesituationis,ofcourse,muchmorecomplexin(differentvarietiesof)English,whereprepositionalverbssuchasrelyonexist,inwhichtheprepositionisobligatoryandthecomplexlexicalitemhasasingle,noncompositionalmeaning.Inthesecases,prepositionstrandingdoesnotresultinagreatincreaseinprocessingcost,sincetheprepositiononisalreadyautomaticallyexpectedbytheprocessorafterencountering

10

i

i i i

i i

i

Page 239: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 10 of 16

rely(Hawkins1999:260n.15).AsIhaveargued(2011;cf.alsoT.HoffmannandTrousdale2011),theseorthogonalcognitiveconstraintsonprepositionplacementhelptoexplainthecomplexstructuralvariationbetweenprepositionstrandingandpied-pipinginEnglish.

Investigatingthestructuralvariationofprepositionstrandingandpied-piping(12)inL1BritishandL2KenyanEnglishusingcorpusdata,aswellaspsycholinguisticexperiments,I(T.Hoffmann2011)havelookedathowtheabovegeneralcognitiveprinciplesaffectthestrandingandpied-pipingalternationandhowtheseinteractwithinputfrequencyeffects.ForwhilegeneralprocessingprinciplesshouldaffectL1andL2speakersalike(cf.Diessel,thisvolume;Ellis,thisvolume),inputfrequencyinL2smightbelimitedduetoarestrictionofthevarietyinquestiontocertainfunctionaldomainsandtheavailabilityoflocalL1sotherthanEnglish(cf.Schmied2004:923–24).Onemajorstatisticallysignificantresultofmystudywasthatprepositionstrandingisstronglyfavoredinquestions(Whodidyoutalkto?),whilerelativeclausesexhibitapreferenceforpied-piping(e.g.,themantowhomyoutalked)inbothBritishandKenyanEnglish.Whilethiseffectoftheclausetypeonprepositionplacementmightatfirstappearsomewhatrandom,itcaninfactbeexplainedastheeffectofseveralcognitiveconstraints:firstofall,asTrotta(2000:55)pointsout,onereasonwhyinterrogativeclausesstronglyfavorprepositionstrandingistheirdiscoursefunction:ininterrogativeclauses“thewh-wordrepresentsunspecifiedinformationwhichcharacteristicallyhasnotpreviouslybeenintroducedintothediscourse.Inthetypicalcommunicativefunctionofinterrogativesasquestions,itisthewh-wordwhichsignalsinterrogationandshouldlogicallycomeearlytosuccessfullyfulfillthatpurpose”(Trotta2000:55).Withrespecttoprepositionplacement,theeasiestwaytoachievethiseffectistostrandthepreposition.Inboundrelativeclauses,ontheotherhand,“thewh-worddoesnotrepresentunspecifiedinformation:theantecedentprecedesthewh-XPandineffect‘signposts’thatsomethingelseiscoming,which,sincesomethingelseisknown,maybedelayedoveralongerstretchoflanguage”(Trotta2000:55).Thus,pied-pipingisamoreviableoptionwithrelativeclauses(aneffectthatisfurthermoreenforcedbyHawkins's(2004)‘AvoidCompetingSubcategorizers’and‘Valencycompleteness’principles,asdiscussed).Fromausage-basedperspective,suchprocessingeffectswillhaveadirecteffectonthementalconstructionnetworkofspeakers:ifthesamecontentcanbeexpressedbytwocompetingstructuresandoneoftheseiseasiertoprocessthantheother,thenthesimplerstructurewillbepreferredinperformance.Consequently,itwillbeusedmoreoftenwithagreaterrangeoflexicalizations,whichincreasesitstypefrequencyandultimatelyleadstoitbeingmorecognitivelyentrenchedthanitsalternative(cf.the‘Performance-GrammarCorrespondenceHypothesis’;Hawkins2004:6).

Basedonthefindingsofhisempiricalstudies,IhavearguedthatanindependentStrandedQuestionandPied-Pipedwh-RelativeClauseconstructionmustbepartoftheconstructionalnetworkofbothBritishandKenyanEnglish.Incontrasttothis,apied-pipedquestionsuchasTowhomdidyoutalk?hastobeconstructedonlinebycombiningtheQuestionconstructionwithanindependentPied-Pipingconstruction(fordetails,seeT.Hoffmann2011).Duetothecompetingstrandedalternative,whichismuchmoredeeplyentrenched,however,thiscombinatoriallyconstructedpied-pipedquestionconstructoccursonlyveryinfrequently.Anotherresultofmydataanalysiswasthattheeffectofthelevelofformalityismuchmorecomplicatedthanusuallyassumed.Normally,itisclaimedthatprepositionstrandingisassociatedwithspeechandinformalwrittencontexts,whereaspied-pipingispreferredinformalwriting(see,e.g.,Biberetal.1999:107).Yet,asthestatisticalanalysisunveiled,formalityonlyaffectedrelativeclausesinBritishEnglish:inbothKenyanandBritishEnglishprepositionstrandingwasfavoredinquestions,regardlessofthelevelofformalityofthetexttype(whichrangedfromprivatedialoguestopublishedwritings).Moreover,inKenyanEnglishrelativesexhibitedapreferenceforpied-pipingacrossalltexttypes.ItwasonlyinBritishEnglishthatrelativeclausesfavoredpied-pipinginmoreformaltextsandstrandinginmoreinformaltexts.ExtendingananalysissuggestedbyWilcock(1999:382–84),thisleadmetoclaimthatinBritishEnglishthePied-Pipedwh-RelativeClauseconstructionismarkedas[formal]andcompeteswithanentrenched[informal]Strandedwh-RelativeClauseconstruction.Incontrasttothis,inKenyanEnglishthelatterconstructionisnotpartoftheconstructiconandinsteadhastobeassembledcombinatorially.NotethatthefactthatL1speakerspossessmoreentrenchedconstructionsthanL2speakersisactuallysomethingthatispredictedbyusage-basedapproaches:sinceL2speakersnormallyreceivelessinputofthetargetlanguagethananativespeaker,theywillalsodevelopfewerandlessdeeplyentrenchedsubstantiveandschematicconstructions.Moreover,oncetwopotentiallysynonymousconstructionsareentrenchedintheconstructicon(suchasthePied-Pipedwh-RelativeClauseandStrandedwh-RelativeClauseconstructioninBritishEnglish),preemptionwillplayanimportantrole(Tomasello2003:300;Goldberg2006a:94–98;seealsoGoldberg2011a;Stefanowitsch2011a):ifonaparticularoccasiononeconstructionisusedinsteadofapossiblealternative,thenthehearerwillassumethatthischoice

Page 240: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 11 of 16

reflectsafunctionaldifferencebetweenthetwostructures.Ultimately,thiswillleadtothefunctionaldifferentiationofthetwoalternativeswith,inthiscase,oneconstructionbeingassociatedwithmore[formal]andtheotherwithmore[informal]contexts(forotherconstructionpairsthatexhibitpreemptioneffects;cf.,e.g.,Goldberg2011aandStefanowitsch2011aontheditransitivevs.prepositionaldativealternationorGries2003onvariableparticleplacementinphrasalverbs).

Finally,drawingonStefanowitschandGries'scovarying-collexemeanalysis(2005:9–11;Stefanowitsch,thisvolume),IalsofoundstatisticalsupportfortheclaimthatL2varietiesexhibitfewersubstantiveconstructions:whiletheBritishEnglishdatayieldedthreesignificantantecedentnoun+P manneradjunctrelativeclauselexicalizations(i.e.,thewayinwhichhekilledthecat;theeasewithwhichhewon;thespeedwithwhichhedidit),theKenyanEnglishdatashowednosignificantlexicalizationinthesamecontext.

Astheabovedemonstrates,anin-depthconstructionistanalysisofdifferentlanguagevarietiescanprovidenewinsightsforthestudyofvariationevenofwell-knownabstractsyntacticphenomenafromwell-documentedlanguages.Ontheotherhand,suchvariationistdataandtheiranalysiscanobviouslyalsohelptorefineConstructionGrammartheories(cf.alsoHollmann,thisvolume).Yet,whileagreatnumberofstudiesonvarietiesofEnglishexistthatchartnon-L1-likefeaturesofsyntacticstructure(e.g.,BresnanandHay(2008)ontheditransitivevs.prepositionaldativealternationoftheverbgiveinNewZealandandAmericanEnglish;Schneider(2003)onparticleverbsandHundt(2006)onverbalconcordpatternswithcollectivenounsinvariousvarieties;BaoandWee(1999)onthepassiveinSingaporeEnglish;tosingleoutafewstudiesthatshowthewiderangeofsyntacticphenomenafromthelexis-syntaxclinecovered),thenumberofstudiesthattakeaconstructionistapproacharestilllimited(notableexceptionsareMukherjeeandHoffmann(2006)onverb-complementationinIndianEnglish;MukherjeeandGries(2009)onvarietyspecificverb-constructionassociationsinNewEnglishes,aswellasHollmann(2007)andHollmannandSiewierska(2007)onDitransitiveandPossessiveconstructionsinBritishdialects).Thisisthereforeclearlyanareainwhichconsiderablefurtherresearchisneeded.

17.3.2EnglishRelativeClausesandConstructionGrammar

Asdiscussedabove,therearemanypropertiesthatdistinguishthedifferenttypesofFiller-Gapconstructions.However,evenwithinasinglecategory,suchasrelativeclauses,severalidiosyncraticpropertiescanbefoundthatwarrantacarefulconstructionalanalysis.NotethatinthefollowingIwillparticularlyfocusonSag's(1997)constructionistaccountoftheidiosyncratic,aswellassystematic,propertiesofEnglishrelativeclauses.Yet,detailedConstructionGrammaranalysesofrelativeclausescanalsobefoundforotherlanguages(suchas,e.g.,Japanese(Ohara1994,1996,2005;Kikuta1998,2002);German(WebelhuthandHoller2010);French(Lambrecht1997,2002);orforatypologicalaccount,cf.Croft(2003:322–51)).Besides,ausage-basedaccountofrelativeclauseacquisitionisprovidedbyDiesselandTomasello(2000).

WorkingwithanHPSG-basedConstructionGrammarapproach,Sag(1997:464,473)assumesthefollowingnetwork.

AsFigure17.3illustrates,allinallSagpostulatessixRestrictiveRelativeClauseconstructionsforEnglish,whichinheritfeaturesfromvariousabstracth(eade)d-ph(rase)constructions(cf.thelowerboxofFigure17.3).Besides,therelativeclausesetissubdividedintotwogroups(cf.themiddlesectionofFigure17.3,whichalsocontainsexampleconstructslicensedbytheseconstructions).

Ontheonehand,Sagpostulatesthreetypesofwh-relative-clauseconstructions:11

•thewh-subject-relative-clauseconstruction(e.g.,themanwholeft),

•thefinite-wh-filler-relative-clauseconstruction(e.g.,themanwhotheylike),

•andthenonfinite-wh-filler-relative-clauseconstruction(e.g.,onwhichtodepend),

whichallsharetheconstraintthatthenon-head-daughterisawh-relativizer(forfurtherdetails;cf.Sag1997:451).Ontopofthese,heassumestheexistenceofthreenon-wh-relative-clauseconstructions:

•thebare-relative-clauseconstruction(e.g.,themanSandylikes),

•thesimple-nonfinite-relative-clauseconstruction(e.g.,themantovisit),and

piedpiped

Page 241: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 12 of 16

•thereduced-relative-clauseconstruction(e.g.,themanstandinghere),

Clicktoviewlarger

Figure17.3. (Standard)EnglishRelativeClauseconstructionnetwork(Hoffmann2011:248,originallyadaptedfromSag1997:464,473)

whichareofthetypenon-wh-rel(ative)-clause,aconstructionwithaconstraintthatensuresthatthesecanmodifytheirnominalantecedentevenwithoutthepresenceofanovertwh-filler(Sag1997:468)andthattheargumentthatislocallymissingintheVPinthebare-relative-clauseandsimple-nonfinite-relative-clauseconstructionsisidentifiedwiththisantecedent(cf.Sag1997:468).Finally,non-wh-rel(ative)-clauseareconstrainedtocombinewithanominalheadthatstillrequiresadeterminer(licensingstructureslike[the][man[Sandylikes]]).ThisaccountsfortheempiricalfactthatØ-relativeclausesmustprecedewh-relativeclauses,cf.Theman[Johnsaw][wholivesinLondon]…vs.*Theman[wholivesinLondon][Johnsaw]…(aneffectthatisfurthermodeledbyspecifyingthelatterconstructionsascombiningwithfullantecedentNPsonly;cf.Sag1997:465–68).

Inadditiontothis,inFigure17.3thehead-filler-phrasehastwosubconstructions:inf(inite)-h(ea)d-fill(er)-ph(rase)andfin(ite)-h(ea)d-fill(er)-ph(rase).Sagmotivatesthisdistinctionbypointingoutthatfinitehead-fillerstructuresalwaysrequireanovertsubject(cf.*Thesebagels,Ølikesor*thebaker[[whosebagels]Ølikes];1997:454).Incontrasttothis,nonfinitehead-fillerclausesdonotallowovertsubjects(cf.*thebaker[[inwhom]](for)youtoplaceyourtrustor*Iwonder[[inwhom](for)themtoplacetheirtrust],Sag1997:461).

ReturningtothespecificRelativeClauseConstructionsinFigure17.3,itturnsoutthatSag'sclassificationincorporatesonetypeofobligatorypied-pipedprepositionconstruction:theinfinite-wh-filler-relative-clauseconstruction.Allotherpropertiesofpied-pipingandstrandingareassumedtofollowfromtheinteractionoftheabovesetofrelativeclauseconstructionsandthegeneralconstructionslicensingprepositionplacement(thoughseesection17.3.1aboveontheissueofformalityandtheneedforadditionalconstructions).Thepostulationoftheinfinite-wh-filler-relative-clauseconstruction,whichisofthetypeinfinite-head-filler-ph(rase)andwh-relative-clause(cf.Figure17.3),isjustifiedbytheidiosyncraticobligatorypied-pipingrequirementoftheseclauses:

(13)a.themanonwhomtorelyb.*themanwhomtorelyon(14)a.Iwonderonwhomtorelyb.Iwonderwho(m)torelyon

Whilenonfinitewh-interrogativeclausesallowbothstrandingandpied-piping(cf.14),prepositionstrandingisprohibitedinnonfinitewh-relativeclauses(cf.13b;alsoT.Hoffmann2011:37–38).Sagcapturesthispropertybyaconstraintwhichrequiresthefillerininfinite-wh-filler-relative-clauseconstructiontobeofthetypePP(cf.Sag1997:462).

Finite-wh-filler-relative-clauseconstructionsareofthetypefinite-head-filler-phraseandwh-relative-clause.Apartfromtheconstraintsthattheyinheritfromthesetwomoregeneralconstructions,finite-wh-filler-relative-clauseshavetheadditionalrestrictionthattheirfillermusteitherbeanNPoraPP(Sag1997:454).Moreover,Saganalyzesrelativethatasawh-relativeword(i.e.,carryingaREL-feature;cf.Sag1997:463).Accordingly,heconsidersthat-relativeclausesinstantiationsofthefinite-wh-filler-relative-clausesconstructions.This,however,meansthatthat-relativeclausesshouldalsoallowpied-piping,sinceasfinite-wh-filler-relative-clausesconstructionstheyareeligibleforthewh-pied-pipingconstruction.Yet,structureslike*themantothathetalkedareclearlyungrammatical.

Sagexplainstheungrammaticalityofpied-pipingwiththatastheresultofcasemismatchand/orregisterrestrictions:forhim,thatinPresent-dayEnglishisarelativepronounwhichcarriesnominativecaseandislimitedtoinformalregistersjustlikewho.Prepositions,however,requiretheircomplementstocarryobliquecase.

12

Page 242: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 13 of 16

Consequently,theimpossibilityofP+thatwouldbeacaseassignmentviolationsimilartopied-pipingwithnominativewho.Besidesthis,accordingtohisaccount,bothwhoandthatareinformalandpied-pipingisformal,whichmightalsoleadtoobligatorystrandingwiththatandwho(cf.1997:463).Whilethisexplanationalsoaccountsforthefactthatthatcannotoccurinnonfiniterelativeclauses,wherepied-pipingisobligatory(SagthusformalizesanideaalreadyadvocatedinVanderAuwera1985),theevidencefrommyexperimentalMagnitudeEstimationstudies(T.Hoffmann2011:175–225)clearlyindicatesthatsuchananalysiscannotbeentertainedforeitherBritishorKenyanEnglish,sincespeakersofbothvarietiesjudgepied-pipingwiththatandwithØasbadaswordorderviolations.Moreover,pied-pipingwithwhoisjudgedsimilartocasemismatcheffectsbutsignificantlybetterthanpiedpipingwiththatorØ.AllofthisindicatesthatthatisnotonparwithwhoineitherBritishorKenyanEnglish.Instead,inbothvarietiesthatbehaveslikeØwithrespecttoprepositionplacement,whichledmetopostulateanextrathat-finite-relative-clauseconstructionforEnglish,whichisaninstanceofthemoreabstractnon-wh-relative-clausesconstruction(fordetailsandfurtherrepercussionsofthisanalysis;cf.T.Hoffmann2011:250–52).Notethatthisconclusionissupportedbyvariousvariationiststudiesonfactorsinfluencingthechoiceofrelativizer:boththatandØarerestrictedtorestrictiverelativeclausesandbotharepreferredinmoreinformalcontexts(cf.,e.g.,GuyandBayley1995;Ball1996).

17.3.3TheComparativeCorrelativeConstruction

WhilethelastsectionoutlinedacompleteinheritanceconstructionistanalysisofaparticularsubsetofcanonicalFiller-Gapconstructions,Iwillnowclosethischapterwithausage-basedlookatamoreidiosyncraticFiller-GapconstructionofEnglish(thoughtheconstructionisnottypologicallyuncommon;cf.DenDikken2005):the“ComparativeCorrelative(CC)”construction.

Thisconstructionhasattractedconsiderableattentioninthesyntacticliterature(cf.,e.g.,CulicoverandJackendoff1999;DenDikken2005);alternatively,itisalsoknownas“covariational-conditionalconstruction”(Fillmore,Kay,andO’Connor1988),“comparative-conditionalconstruction”(McCawley1988b),or“the-clauses”(Sag2010):

(15)a.[the[moremoney] wecomeacross][the[moreproblems] wesee](NotoriousB.I.G.—MoMoneyMoProblems)b.but[the[more] Igrow][the[less] Iknow](NellyFurtado—Try)c.[the[more] ,] [the[merrier] ]

Astheexamplesin(15a–c)show,theconstructionsconsistoftwoparts(C1andC2),inwhichthesecondclauseC2canbeinterpretedasthedependentvariablefortheindependentvariablespecifiedbyC1(cf.Goldberg2003:220;e.g.,themoremoneywecomeacross→themoreproblemswesee;foramoredetaileddiscussionoftheconstruction'ssemantics;cf.Beck1997;Cappelle2011).Ontopofthis,theconstructionconsistsoffixedsubstantive,phonologicallyspecifiedmaterial([ð…] [ð…] ),aswellasschematic,openslots(whichincludesanobligatorycomparativephrase;cf.,e.g.,[more] and[merrier] in(15c)with[moremoney] and[moreproblems] in(15a)).

PreviousConstructionGrammaraccountshaveprovideddetailedaccountsoftheseidiosyncraticfeaturesoftheconstruction(aswellasotherssuchasthetypesofpossiblecomparativephrasefillers,be-omission,thepresenceofthat-complementizers,optionalsubject-auxiliaryinversion,etc.;cf.inparticularCulicoverandJackendoff(1999)andSag(2010)).Ontopofthis,however,theconstructionalsoexhibitsaninterestingintonationalcontour.

AsCulicoverandJackendoff(1999:552)pointout,CCconstructionshaveameaningsimilartotheparatacticconstructionin(16),whichismarkedbyarisingintonationonthefirstclauseandafallonthesecondclause(notethat,followingCulicoverandJackendoff,↑=rising/↓=fallingintonation):

(16)MarylistenstoGratefulDead↑,shegetsdepressed↓.(fromCulicoverandJackendoff1999:552)

Yet,whileCulicoverandJackendoffdonotcommentontheintonationoftheCC,ananalysisofthesoundfilesof

comparativephrase1 C1

comparativephrase2 C2

comparativephrase1 C1

comparativephrase2 C2

comparativephrase1 C1 comparativephrase2 C2

C1 C2

comparativephrase1 comparativephrase2

comparativephrase1 comparativephrase2

Page 243: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 14 of 16

theICE-GBcorpus(T.Hoffmann2010)showsthattheconstructionalsodisplaysaspecificprosodicmarking:

(17)[the[moremistakes↑] youmake↓][the[moreinteresting↓] itbecomes↓](ICE-GB:S1B-002-F48)

Incontrastto(16),allofthethirteeninstancesoftheCCconstructionintheICE-GBhaveafallingintonationatbothclauseboundaries.Yet,therisingandfallingintonationareneverthelessalsousedtomarktheindependentanddependentvariable,respectively.However,as(17)shows,theseintonationalcontoursarenotlocatedattheclauseboundaries,butinsteadarerealizedonthecomparativephrases.

Sinceconstructionsasform-meaningpairingscanalsoincludeinformationontheirintonation(cf.,e.g.,Lambrecht1990;MichaelisandLambrecht1996;Marandin2006;Croft2007c;Fried2009b),thisfindingcaneasilybeincorporatedintoconstructionalaccounts(take,e.g.,thefollowingamendedversionofCulicoverandJackendoff's(1999:567)representationoftheconstruction'sform:[[themoreXP↑] (that)IP↓] [[themoreXP↓] (that)IP↓] ).Moreover,oncetheintonationalcontourisseenasanintegralpartoftheconstruction,itbecomesobviouswhystackingofmorethantwoclausesisinfactpossibleinEnglish(contradenDikken2005:503):

(18)[the[moreopaque↑]thatatmosphereis↓] ,[the[lessconductive↑]itis↓][the[biggerthetemperaturedifference↓]youneedtocrossit↓](ICE-GB:S2A-043)

In(18)thesecondclauseC2simultaneouslyactsasthedependentvariableofC1(themoreopaque→thelessconductive)andtheindependentvariableofC3(thelessconductive→thebiggerthetemperaturedifference).Thisstackingofthreeclausesispossiblesincetheriseonthesecondcomparativephrasesignalstheindependentvariableuseofthisclause.

17.4.Conclusion

OneofthemainaimsofthepresentvolumeistoshowthatConstructionGrammardoesnotonlyfocusontheidiosyncraticperipheryoflanguage,butthatitisafull-fledgedgrammaticaltheorythatisobservationally,descriptivelyaswellasexplanatorilyadequate(inthesenseofChomsky1965:24–26).Inthepresentchapter,Ihaveoutlinedhowconstructionistapproachescandescribeandmodeleventhemostabstractofsyntacticstructures,butalsothatsuchanalysesfurtherourunderstandingofthephenomenainquestion.Inparticular,IhavefocusedonAbstractClausalconstructions,Filler-Gapconstructions(especiallyrelativeclausesandcomparativecorrelatives)andprepositionplacement.Ontopofthese,manyotherabstractschematicconstructionshavealsoalreadyreceivedconsiderableattentionintheConstructionGrammarliterature(suchaspassives,cf.,e.g.,BrooksandTomasello1999c;BarðdalandMolnár2003;Fried2006;TenseandAspectconstructions,cf.,e.g.,Bergs2008,2010;Hilpert2008;Croft2012,orArgumentStructureconstructions,fordetails,cf.Boas,thisvolume).Nevertheless,therearestillmanyinterestingissuesthatneedtobeaddressedbyfutureresearch(andotherConstructionGrammarresearcherswouldundoubtedlyaddseveralmore):

•anin-depthanalysisofsynchronicregionalandsocialvariation(evenofwell-documentedlanguages;cf.T.HoffmannandTrousdale2011,aswellasHollmann,thisvolumeandOstmanandTrousdale,thisvolume);

•ontheonehand,aclosercollaborationofusage-basedandcompleteinheritanceapproacheswithrespecttothequestionofhowabstractconstructionalnetworksneedtobe(i.e.,whethermacro-constructionsarenecessaryatall:cf.Barðdal2011a;Bybee,thisvolume,aswellasPulvermüller,Cappelle,andShtyrov,thisvolume,forneurolinguisticevidencethatisofrelevanceforthisquestion);and

•ontheotherhand,amorethoroughinvestigationofabstractphenomenabeyondtheclause-level,suchastext-embeddedclause-combining(cf.Nir2008;NirandBerman2010b).

IamgratefultoGraemeTrousdale,StefanMüller,andJóhannaBarðdalfortheircommentsonearlierdraftsofthispaper.OntopofthatIwouldliketothankMirjamFried,IvanSag,andLucSteelsfortheirhelpwiththeanalysisofsubject-verbagreementinvariousConstructionGrammarapproaches.Allremaininginaccuraciesaremine.

comparativephrase1

comparativephrase2

13

i C1 i

C2

C1

C2

C3

Page 244: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 15 of 16

Notes:

(1.)Barðdal,Kristoffersen,andSveen(2011)adda‘lexicality–schematicity’clinetotheconstructicontodistinguishstoredschematicconstructions(suchasTheX-er,theY-er;Fillmore,Kay,O’Connor1988;seealsosection17.2.3)fromlexicalizedvariantsofsuchpatternsthatmustalsobestoredinthementallexiconduetotheiridiosyncraticidiomaticmeaningorhighfrequency(cf.Themore,themerrier).WhilethereaderisreferredtoBarðdal,Kristoffersen,andSveen(2011)forfurtherdetails,thisclinewillnotbeadoptedforthiscontribution,sincetheschematicityandcomplexitycategorieswillsufficetoexplainthephenomenadiscussedinthefollowing.

(2.)Foralternativerepresentationsofthementalconstructionnetwork;cf.Hudson(2007b)andCroft(thisvolume).

(3.)Notethattheidentityhereisduetothefactthat‘3sing’isatypedfeaturestructure(Sag,Wasow,andBender2003:107)thatoccurstwicein(3c).Inordertoensureidentityinmorecomplexagreementsituations(whenlanguagesexhibitagreaterrangeofconcordmarkingthanEnglish),onewouldhavetoemployvariables(e.g.,byusingavariable‘Z’thatindicatesthatthevalueofSYN|CAT|AGRZandSYN|CAT|XARG|AGRZalwayshavetobeidenticalinvalue).

(4.)TheactualSBCGanalysisofthesephenomenahasbeenslightlysimplifiedforexpositorypurposes.Foramoredetailedanalysis,cf.Michaelis(thisvolume)andSag(2012).

(5.)ButseeMichaelis(2004b,thisvolume)forcoercion,thatis,structureswhereapparentlyconflictingconstructionsareresolvedintoacceptableinterpretations:take,forexample,thestructureGivemeabutter(Michaelis,thisvolume),inwhichamassnounisprecededbyanindefinitearticle(whichisusuallyconsideredungrammatical;cf.*Ilikeabutter).Yetsincetheditransitiveherestronglyimpliesthetransferofaboundedentity,butteriscoerced(conceptualized)asaboundednominal,i.e.acountnoun,renderingthesentenceacceptable.(ThoughalsoseeTraugott2007forproblemsassociatedwiththisnotionofcoercion.)

(6.)ForthepurposesofillustrationandinordertomaketheanalysiscomparabletothatofCroftandCruse(2004),Ilimitmyselftosentenceswithjustasingleadjectiveorverb.Amorepreciseconstructionalanalysiswill,however,havetotakeintoaccountthattheconstructionsactuallyhaveslotsthataccommodatefulladjectivephrases(Heisproudofhisson)andverbphrases(Hekissedthebride;StefanMüller,personalcommunication).

(7.)Itshouldbepointedoutthatearliercompleteinheritancemodelsadvocatedmonotonicinheritancerelationshipsamongconstructions,whilecurrentlyallconstructionistapproaches,includingSBCG,assumedefaultinheritance(i.e.,thatamoregeneral,schematicconstructioncontributesallitsinformationtoamorespecificoneunlessthelatterconstructioncontainsspecificinformationwhichoverridesthemoregeneralone;cf.Goldberg2006a:215).

(8.)Croft(2003)suggestsasimilardistinctionforverbalconstructionsthatrangesfrom‘verb-specificconstructions’(suchas[SBJrefuseOBJ1OBJ2]/‘negativeXPossbyrefusing’or[SBJdenyOBJ1OBJ2]/‘negativeXPossbydenying’)to‘verb-class-specificconstructions’(suchas[SBJREFUSE.VERBOBJ1OBJ2]/‘negativeXPoss,’whichwouldbeanabstractionoverthemorespecificrelyanddenyverb-specificconstructions).Forarecentrevisionofthisclassification,seeBarðdal(2006,2008,2011a).

(9.)InMGGislandeffectsareusuallyexplainedbythepostulationofsomeabstract,clear-cutsyntacticconstraint,suchasthe‘A-over-Acondition’(Chomsky1962),the‘subjacencycondition’(Chomsky1973),‘barriers’(Chomsky1986),orthe‘minimallinkcondition’(Chomsky1995,2000b).However,asKluender(1992,1998),Sag,Hofmeister,andSnider(2007),orHofmeisterandSag(2010),interalia,haveshown,islandeffectsseemtobemuchmoregradientinnatureandthereforereceiveamuchmorestraightforwardexplanationasgradientcognitiveprocessingconstraints.WhileIreferthereadertoHofmeisterandSag(2010)foramuchmoredetaileddiscussionofthisissue,itisnoteworthythattheacceptabilityofnon-islandFiller-GapconstructionsthenfollowsfromHawkins’Performance-GrammarCorrespondenceHypothesis,thatis,that“[g]rammarshaveconventionalizedsyntacticstructuresinproportiontothedegreeofpreferenceinperformance”(Hawkins2004:3).Inotherwords,sincenon-islandFiller-Gapconstructionsareeasiertoprocessthanonescontainingahigherprocessingload(alsoknownasislands),itfollowsnaturallythatmanylanguageshaveentrenchedtheformerandnotthelatteraspartoftheirgrammars.Asthisillustrates,processingeasecanthusalsobesaidtoplayasignificantroleintheusage-basedentrenchmentofabstractconstructions(cf.alsosection17.3.1).

Page 245: The Oxford Handbook of Construction Grammar

Abstract Phrasal and Clausal Constructions

Page 16 of 16

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). (c) Oxford University Press, 2014. All RightsReserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of a title in OxfordHandbooks Online for personal use (for details see

(10.)Lancaster-Oslo-BergenCorpus(accessibleon-lineviahttp://icame.uib.no).

11.NotethatinthefollowingIadoptSag'sterminology,whichincludesitalicfontandhyphensforconstructionnames.

(12.)Thisisnecessarysinceotherfinite-head-filler-phraseconstructionsarelessrestrictivewithrespecttofillerstheylicense.Questions,forexample,alsoallowAdjPfillers(cf.HowcoolamI?;cf.Sag1997:454).

(13.)Thosefamiliarwiththesongsin(15a,b)willnoticethattheconstruction'sintonationalcontouristhusmirroredbythesongs’melodycontour.

ThomasHoffmannThomasHoffmannisAssistantProfessorattheUniversityofOsnabrück.Hismainresearchinterestsareusage-basedConstructionGrammar,synchronicsyntacticvariationandWorldEnglishes.HehaspublishedarticlesininternationaljournalssuchasCognitiveLinguistics,EnglishWorld-WideandCorpusLinguisticsandLinguisticTheory.Ontopofthat,hismonographPrepositionPlacementinEnglish(2011)waspublishedbyCambridgeUniversityPress.Currently,heiswritingthetextbookConstructionGrammar:TheStructureofEnglishfortheCambridgeTextbooksinLinguisticsseries.

Page 246: The Oxford Handbook of Construction Grammar

Information Structure

Page 1 of 11

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0018

InformationStructureJaakkoLeinoTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterexaminestheroleofinformationstructureinconstructionistapproaches.ItevaluatesthecentralnotionsofinformationstructureandhowthesefactorscanbeincorporatedintoaConstructionGrammarviewofmentalgrammar.Thechapterexplainsthatinformationstructureisanimportantelementofsentencegrammarbecauseitinfluencesbywhichconstructionaparticularmeaningisexpressedandwhyspeakersthereforechooseoneconstructionoveralternativeonesinspecificsituations

Keywords:informationstructure,constructionistapproaches,ConstructionGrammar,mentalgrammar,sentencegrammar,meaning

18.1.Introduction

TheterminformationstructurewasintroducedbyHalliday(1967).Hallidaydistinguishesbetweenthree‘areasofsyntacticchoice’thataffecttherealizationofclauses:transitivity,mood,andtheme.Theme,hestates,“isconcernedwiththeinformationstructureoftheclause;withthestatusoftheelementsnotasparticipantsinextralinguisticprocessesbutascomponentsofamessage;withtherelationofwhatisbeingsaidtowhathasgonebeforeinthediscourse,anditsinternalorganizationintoanactofcommunication”(Halliday1967:199).Thus,thestudyofinformationstructurefocusesonthepositioningofclauseelementsindiscourse,ratherthan,forexample,thepositioningoftheirreferentsintheeventsdescribedbyclauses.

ManyofthecentralideasbehindworkoninformationstructurebyHalliday,aswellasmorerecentworks,owemuchtopreviousworkbymembersofthePragueSchool(notablyDaneš1964andFirbas1962,1964).Notalllinguistsemploytheterminformationstructure,however;DanešspeaksofFunctionalSentencePerspective,othermembersofthePragueSchool(e.g.,FirbasandMathesius)ofCommunicativeDynamismandorganizationofutterance,andChafe(1976,1994)andPrince(1981)ofinformationpackaging.

InformationstructureiswithoutanydoubtthepragmaticallyorientedphenomenonorsubjectareathathasreceivedthemostattentioninthecontextofConstructionGrammar.InthecontextofandinrelationtoConstructionGrammar,InformationStructurehasbeenmoststudiedbyKnudLambrecht(e.g.,Lambrecht1981,1984,1994,2000,2001,2004;MichaelisandLambrecht1994,1996).Lambrecht,likeanumberofotherlinguistsworkingwithinformationstructure,hasworkedextensivelyoncleftedanddislocatedstructuresandtheirfunctions.

Lambrechtdefinestheconceptofinformationstructureasfollows:

Informationstructure:Thatcomponentofsentencegrammarinwhichpropositionsasconceptualrepresentationsofstatesofaffairsarepairedwithlexicogrammaticalstructuresinaccordancewiththementalstatesofinterlocutorswhouseandinterpretthesestructuresasunitsofinformationingivendiscoursecontexts.(Lambrecht1994:5)

Page 247: The Oxford Handbook of Construction Grammar

Information Structure

Page 2 of 11

Lambrecht'sownparaphraseforthisdefinitionisthattheinformationstructureofasentenceistheformalexpressionofthepragmaticstructuringofapropositioninadiscourse.Lambrechtliststhefollowingas“themostimportantcategoriesofinformationstructure”(1994:6):

1.Presuppositionandassertion,whichhavetodowiththestructuringofpropositionsintoportionswhichaspeakerassumesanaddresseealreadyknowsordoesnotyetknow.2.Identifiabilityandactivation,whichhavetodowithaspeaker'sassumptionsaboutthestatusofthementalrepresentationsofdiscoursereferentsintheaddressee'smindatthetimeofanutterance.3.Topicandfocus,whichhavetodowithaspeaker'sassessmentoftherelativepredictabilityvs.unpredictabilityoftherelationsbetweenpropositionsandtheirelementsingivendiscoursesituations.

Thesecharacterizationsareveryabstract,andapplyingthemtoauthenticsentencesandutterancesisnotalwaysunproblematic.Iwillreturntoeachoneinmoredetailinsections18.3.1–18.3.3.Putsimply,whatLambrechtmeansby(pragmatic)presuppositionisthe“old”informationcontainedorevokedbythesentence,andwhathemeansby(pragmatic)assertionisthe“new”informationexpressedorconveyedbythesentence(Lambrecht1994:52).Thus,forexample,thesentencein(1)presupposesseveralpropositions:

(1)Maryheardyesterdaythatherbrotherhasfinallyquitsmoking.

Thesentenceevokes(atleast)thefollowingpropositions:

1.theaddresseeisabletoidentifythepersonreferredtoasMary2.thepersonreferredtoasMaryhasabrother3.thepersonreferredtoasMary'sbrotherpreviouslyhadahabitofsmoking(tobacco)4.thepersonreferredtoasMaryhasexpected(orwished)thatherbrotherwillquitsmoking5.thequittingeventtookplacebeforethepersonreferredtoasMarywasinformedofit.

Theassertionmadeinthesentencein(1)canbeexpressedasfollows:‘propositions1–4aretakenforgranted;thepersonreferredtoasMaryhaslearned,onthedaybeforethespeechevent,thatherbrotherhasquitsmoking,’

Onethingthatremainsopeniswhetherthepropositionexpressedbythecomplementclausethatherbrotherhasquitsmokingispresupposedorasserted.Thetraditionalanalysisisthatitispresupposedandnotasserted.However,itmayalsobearguedthatwhetherthepropositionispresupposedorassertedmaydependonthediscoursecontext.Iwillreturntothispointinsection18.3.1.

Whilepresuppositionandassertionrelatetopropositionsand,forexample,theirgivennessandacceptability,identifiabilityandactivationrelatetoreferentsandtheirdiscourseproperties.AsLambrecht(1994:77–78)pointsout,thedistinctionbetweenidentifiableandunidentifiablereferentsissimilartothatbetweenpragmaticallypresupposedandpragmaticallyassertedpropositions.Thespeakerandthehearerhavesharedknowledgeofapresupposedpropositionpriortotheutterance,whileonlythespeakerhasknowledgeofanassertedpropositionbeforetheutterance.Correspondingly,thespeakerandthehearerhaveasharedrepresentationofanidentifiablereferentatthetimeoftheutterance,whileonlythespeakerhasarepresentationofanunidentifiablereferentatthetimeoftheutterance.

Thus,acentralquestionbehindidentifiabilityiswhetherthespeakerexpectsthehearertounderstand,atthetimeoftheutterance,whichobjectorpropositionthespeakerhasinmindwhentheyrefertothatobjectorpropositioninthesentence.Activation,inturn,concernsthequestionastowhetherthereferent—objectorproposition—ispresentintheuniverseofdiscourse: isitpresentinthespeechsituation,hasitbeenalreadymentionedinthediscourse,isit“reachable”throughsomeotheralreadyactivereferent,andsoon.

Overall,asKuningas(2007:26)pointsout,weneedinformationstructureasthe‘thirdcomponent’ofgrammarinordertoexplainwhy,inagivencontext,thespeakerchoosestouseagivensyntacticstructureorexpressiontype(say,acleftsentenceorpassivevoice)ratherthansomeotherone(e.g.,thecanonicalsentencestructureoractivevoice).Typically,ifnotalways,thereareseveralpossiblewaysofexpressingthesameproposition—labeledallosentencesbyLambrecht(1994)—buttheirsuitabilitytoanygivencontextvariesandisdependent,crucially,onsuchfactorsaswhattheinterlocutorsexpecteachothertoknowbeforehand,whathasandwhathasnotbeenalreadymentionedinthesamediscourse,andthelike.

1

2

3

4

Page 248: The Oxford Handbook of Construction Grammar

Information Structure

Page 3 of 11

SeveralConstructionGrammarians—including,notably,Lambrechthimself—haveincorporatedinformationstructurenotionsaspartsoftheirframeworks.Categorieslikethosepresentedabovehaveprovenusefulinthedescriptionsofvariousconstructionsandclause-levelphenomena.Asanearlyexample,Lambrecht(1990)discusseswhathecallsthe“MadMagazinesentence,”illustratedbythesentenceWhat,meworry?andshowsconvincinglythataproperunderstandingofthefunctionsofthissentencetyperequiresanunderstandingofitsinformationstructureproperties.Briefly,thesentence(or‘conventionalizedsentencefragment,’asLambrechtcallsit)presentstopicalinformationandisoftenfollowedbya‘follow-upproposition’thatisfocal.

Lambrecht,likeseveralothers,alsospellsoutinformationstructurerequirementsorconstraintsinfeaturematrixdescriptionsofconstructions.Forinstance,Lambrecht(2004)presentsananalysisoftheFrenchcomme-Nconstruction(e.g.,c’estpasmarrant,commehistoire‘that'snotafunnystory’)inwhichthedescriptionofthefirstnominalelement(inthepreviousexample,thepronominalce,truncatedasc’)includesthefeature[ACTactive].Inotherwords,thissentencetyperequiresthatthefirstnominalelementofanyactualsentenceinstantiatingthissentencetypebeactiveinthepresentdiscourse.Whatthismeansinactualpracticewillbediscussedinsection18.3.2below.

Informationstructurephenomenaareoftenassociatedwithsomethingakintosentencetypes,orexpressiontypes.TheEnglish“MadMagazinesentence”andtheFrenchcomme-Nconstructionarecasesinpoint,andalsosuchphenomenaascleftsentences,dislocations,andextrapositionstructureshaveoftenbeenstudiedwithaninformationstructuralpointofview.ConstructionGrammarapproachesgrammarinamannerthatmakesitrelativelyeasytodescribethepropertiesof,forexample,singleNPsasdependentontheirpositioninalargerconventionalizedstructure.ItseemslikenosurprisethatinformationstructurehasattractedalotofinterestwithinConstructionGrammar—and,conversely,thatConstructionGrammarseemslikeanattractiveframeworkforinformationstructureresearchers.

18.2.InformationStructureandGrammar

Researchoninformationstructurefocusesonissueslikethegivennessandaccessibilityofinformation,theorganizationofinformationinsentencesandindiscourse,andvariousrelatedtopics.What,then,istheplaceofinformationstructureinthelanguagesystem,orgrammar?Howdoesitrelatetosyntaxorsemantics?Areinformationstructurephenomenasemanticbynature,oraretheyprimarilyconnectedtothestructureandorganizationofsentences,andtherebytosyntax?

Lambrecht(1994:6–13)presentsinformationstructureasthethirdpartofatripartitelanguagesystem,onaparwithsyntaxandsemantics.Hecitesseverallinguistswhohaveproposedfundamentallysimilararchitecturesofgrammar,comprisingthreepartsorlevels.Accordingtothislineofthought,grammarconsistsofform(orformalstructure),or(morpho)syntax;meaning(anditsstructure),orsemantics;andathirdpart,whichdoesnothaveanequallyconventionalname,thoughinallcasescitedbyLambrechtthisthirdpartofgrammarhasastrongpragmaticflavortoit.

Whileasteadilyincreasingnumberoflinguistsconsiderinformationstructureanintegralpartofthelanguagesystem,itisquitecommonthatinactualpractice,comparedto(morpho)syntaxandsemantics,informationstructureisthe‘oddoneout,’andisoftenleftoutofgrammarsanddescriptionsofgrammaticalphenomena.ÖstmanandVirtanen(1999:92–93)pointthisoutasfollows:

Itistruethat,inverygeneralterms,linguistsdotakeInformationStructuring,or(followingCaseandConstructionGrammarterminology)‘rhetoricalrelations’intoaccountonaparwith—andasbeinginprincipleequalinimportanceto—investigationsofthesituationalframe-roles(like‘buyer’and‘seller’inacommercialtransaction),semantic,orthetaroles(likeAgent,Patient,andExperiencer),andgrammaticalrelationsorfunctions(likeSubjectandObject).Butthespecificnatureoftheserhetoricalrolesisoftenleftvague.Inmanytheories—includingmostversionsofConstructionGrammar(cf.Fillmore&Kay,forthcoming;butseeLambrecht1994)—theyseemsofartohavebeenrelegatedtothestatusof‘extras’thatmighthavetobereferredtoatsomepoint,inordertounderstandandbeabletodescribesomeconstructions,butmostofthetime,formostlanguages,linguistsseemtobequitehappynotreferringtothem.

Page 249: The Oxford Handbook of Construction Grammar

Information Structure

Page 4 of 11

Andyet,InformationStructureinaveryobvioussenseformsthebridgebetweengrammaranduse,aninterfacewitha‘missinglink’statusforunderstandinglanguage.

Informationstructureis,thus,anelementofsentencegrammar,supplementarytomorphosyntaxandsemantics.Itisconcernedwiththemannerinwhichthemessageisconveyed,orwiththequestionastowhythespeakermakestheparticularsyntacticandsemanticchoicesandusestheparticularexpressiontypesratherthansomeotherones(cf.,e.g.,Kuningas2007:26),or,morebroadly,“whytherearesomanykindsofsentencestructures”(Lambrecht1994:9).ThislastpointmaybeillustratedwithaparallelpointedoutbyLambrecht(1994:9),namely“[i]fecologyfocusesontheinteractionoforganismsandtheirenvironments,thestudyofinformationstructurefocusesontheinteractionofsentencesandtheircontexts.”Inotherwords,whiletheecologistaskswhytherearesomanyspeciesoflivingorganisms,the(functional)linguistaskswhytherearesomanycompetinggrammaticalstructures.Wemaypushthisparallelonestepfurtherbythinkingofdifferentcommunicationsituationsasthe‘livingconditions’oflinguisticexpressionsandexpressiontypes.Biologyandecologyspeakofecologicalniches.Thosespeciessurvivewhichhavea‘suitableslot’intheecosystem:sufficientnutrition,nottoomanyenemies,andsoon.Similarly,linguisticscould(andsometimesdoes)speakofgrammaticalniches.Thoseexpressiontypessurvive—grammaticalize,orremainpartsoftheexpressiveapparatusofthelanguage—whichhaveaslotinthelanguagesystem:sufficientexpressiveneed,nottoomanycompetingexpressiontypes,andsoon.

18.3.CentralConceptsofInformationStructure

Inordertodiscussinformationstructureinanydetailandtoillustratethetopicwithsomeconcreteexamples,somebasicconceptsneedtobeintroduced.Thisdiscussioncannotpossiblydojusticetothewidespectrumofapproachesandplethoraofliteratureontheseandrelatedconcepts.Duetoavastoversupplyofcompetingtheoriesofandapproachestoinformationstructure,andduetolimitationsofspace,IshallfocusmainlyontheideasandanalyticaltoolspresentedinKnudLambrecht'swork,notablyinhis1994bookInformationStructureandSentenceForm,whichisbyfarthesinglemostinfluentialworkoninformationstructurewritteninthespiritofConstructionGrammar.

18.3.1PresuppositionandAssertion

Thenotionspragmaticpresuppositionandpragmaticassertion,asdefinedbyLambrecht(1994),werediscussedbrieflyinsection18.1.ItwaspointedoutthatthesentenceMaryheardyesterdaythatherbrotherhasfinallyquitsmokingpresupposesanumberofthings(seethediscussionabove)andasserts,essentially,that‘thepersonreferredtoasMaryhasnowlearnedthatherbrotherhasquitsmoking.’However,itwasleftopenastowhetherthepropositionexpressedbythecomplementclausethatherbrotherhasquitsmokingispresupposedorasserted.Theanswertothisquestionwillultimatelydependonhowexactlyweunderstandthenotionspresuppositionandassertion.

Lambrechtspeaks,specifically,ofpragmaticpresuppositionandassertion.Hedefinesthesenotionsasfollows(1994:52):

–Pragmaticpresupposition:Thesetofpropositionslexicogrammaticallyevokedinasentencewhichthespeakerassumestheheareralreadyknowsorisreadytotakeforgrantedatthetimethesentenceisuttered.

–Pragmaticassertion:Thepropositionexpressedbyasentencewhichthehearerisexpectedtoknowortakeforgrantedasaresultofhearingthesentenceuttered.

Atraditionalanalysiswouldstatethatthecomplement-takingverbhearisfactive(inthesenseofKiparskyandKiparsky1971)and,therefore,thepropositionexpressedbythecomplementclauseis,bydefinition,presupposed.However,itisnotobviousthatthetraditionalanalysisiscoherentwithLambrecht'sdefinitionsofpragmaticpresuppositionandassertion.

Firstofall,thefactthatMary'sbrotherhasquitsmokingmayormaynothavebeendiscussedearlierduringthesamediscussion.Ifithas,thenthespeakermayrightfullyassumethat“theheareralreadyknowsorisreadytotakeforgranted”thisproposition“atthetimethesentenceisuttered”—hence,thepropositionwouldbepragmaticallypresupposed.Ifithasnot,thespeakermayexpectthehearer“toknowortakeforgranted”this

Page 250: The Oxford Handbook of Construction Grammar

Information Structure

Page 5 of 11

proposition“asaresultofhearingthesentenceuttered”—andhenceitwouldbepragmaticallyasserted,despitethefactthatthepropositioncomesupinacomplementclause.

Second,however,itisperhapsproblematictosaythatthespeakerexpectsthehearer“toknowortakeforgranted”thepropositionexpressedbyacomplementclause“asaresultofhearingthesentenceuttered.”Traditionally, assertinghasbeentakentobeapropertyofmainclauses,andthefactthatLambrecht'sdefinitionofpragmaticpresuppositiondoesnot,strictlyspeaking,fittogetherwithourexampledoesnotentailthathisdefinitionofpragmaticassertiondoes.Furthermore,Lambrechtdefinespragmaticpresuppositionasasetofpropositions,butpragmaticassertionasthepropositionexpressedbyasentence.Ifthisistobetakenliterally,thenpragmaticassertionwouldapparentlyonlyincludethepropositionexpressedbythemainclause.

Insteadofpushingthisdiscussiontoofar,itmaybemoreusefultothinkofthepropositionsexpressedbycomplementclausesasdiscoursereferents.Thismakesitmorerelevanttospeakoftheiractivationstatusandtheirroleasthetopicorfocusofthesentenceratherthantheirpresupposednessorassertedness.

18.3.2Activation

Anothercentralconceptininformationstructurebesidespresuppositionandassertionistheactivationstateofareferent.Arelativelywidespreadaccountofactivationstates,particularlywithinConstructionGrammar,isthatformulatedbyChafe(1987)andsubsequentlyusedbyLambrecht(1994),aswellasanumberofothers.Thenotionofactivationstatereferstothepresenceorabsenceofthereferentinthespeaker'sandthehearer'sconsciousness.ThethreestatesproposedbyChafeareactive,semi-active(alsooftencalledaccessible),andinactive.Thisthree-stepdivisioniscloselyrelatedtotheclassicaldivisionofgivenvs.new,supplementedbythethirdstep,previouslyproposed,forexample,byPrince(1981)bythenameevoked.Thethreestatescanbecharacterizedasfollows,inthespiritofChafe(1987:25)andLambrecht(1994:94):

–Anactivereferentisconceptuallypresentintheperson'sfocusofconsciousness.Itcanbesaidtobeinthelimelight(or“currentlylitup,”asChafeputsit).

–Asemi-active(oraccessible)referentisintheperson'sconsciousnessonlyperipherallybutisavailabletoattentiondespitethefactthatitisnotinfocus.

–Aninactivereferentisnotevenperipherallyintheperson'sconsciousnessatthemomentbut,rather,storedinthelong-termmemory.

Areferentcanbesemi-activefortwogeneralreasons.First,itmayhavelostpartofitspreviousactiveness(“typicallybyhavingbeenactiveatanearlierpointinthediscourse”accordingtoChafe).Second,areferentmaybeactivebecauseitisapartofaframeevokedbythediscourse. Thus,speakingofarighttriangle(touseaclassicexampleusedinLangacker1986:7andinGoldberg1995:26)evokestheconceptofhypotenuse:itbecomesaccessible,orsemi-active.However,merelyspeakingofarighttriangledoesnotbringthehypotenusetothefocusofattentiontotheextentthatitwouldbecomeafullyactivereferent.

Theactivationstatusofareferentmaybearelevantfeaturewithregardtothepragmaticallyadequateuseofagivensentencestructure.TheFrenchcomme-Nconstructionmentionedinsection18.1isacaseinpoint,andotherswillbediscussedinsection18.4.Inalimitedsetofinstances,activationstatusmaybesetevenforasinglelexeme,apronounoradeterminerinparticular(e.g.,GundelandFretheim(2009:148)pointoutthat“theproximaldemonstrativedeterminerthis/thesecodesthecognitivestatus‘activated’”),butitismuchmorecommonthatactivationstatusisdeterminedbyaclause-levelconstruction,wordorder,orsomeothernonlexicalfactor.Therefore,lookingatconstructionsinaholisticmannerisnecessaryforunderstandingthemechanismsinwhichtheactivationstatusofagivenreferentiscodedinthelinguisticexpressionassociatedwiththatreferent.

Therecognizability,accessibility,andactivationofreferentsrelatesfirstandforemosttotheimagethattheinterlocutorshaveofoneanother'sconsciousness,andtohowspeakersstructuretheirutterancestofitwhatthehearersdoanddonotknow.AsTomlinetal.(1997:80)putit,“[i]fthespeakerassumes,priortoutteringanintonationunit,thataconceptisalreadyactiveinthelistener'smind,hewillverbalizethatconceptinanattenuatedmanner,mostprobablyproniminalizingit.Ifheassumesthataconceptisnotpresentlyactivatedinthelistener'sconsciousness,hewillverbalizethatconceptinalessattenuatedmanner,mostprobablynominalizingit.”

5

6

7

Page 251: The Oxford Handbook of Construction Grammar

Information Structure

Page 6 of 11

Inorderforthespeakertobeabletoaccommodatehisutterancetothespeaker'sconsciousness,henaturallyneedstohaveknowledgeoratleastassumptionsaboutwhatthehearerknowsbeforehandandwhatisandwhatisnotcurrentlyactiveinthehearer'smind.Toputitsimply,thebasicideaisthatboththespeakerandthehearerhaveamindwiththoughtsandconceptsinit.Amongthesethoughtsandconceptsareassumptionsaboutwhatkindsofthoughtsandconceptsexistintheotherperson'smind;wemightcalltheseprojectionsoftheotherperson'smind.Thus,thespeakerhasinhismindaprojectionofthehearer'smind,andviceversa,asshowninFigure18.1.

Differentlinguisticunitsmaycontaininformationaboutwhattheseprojectionsarelike.Forinstance,theuseofindefiniteanddefinitearticleshastodowithwhetherthespeakerexpectsthehearertorecognizethereferent.Similarly,whatChafe,Lambrecht,andothershavewrittenconcerningtheactivationstatusesofconceptsand/orreferentsalsorelatestothissetting:inactivestatusappliestoreferentswhichthespeakerdoesnotassumeto‘exist’inthehearer'smind,thatis,whicharenotincludedinthespeaker'sprojectionofthehearer'smind.Thedifferencebetweenactiveandsemi-activereferents,inturn,dependsonthestatusofthereferentinthespeaker'sprojectionofthehearer'smindandwhetherthespeakerassumesthehearertobeabletoaccessthereferent.

Figure18.1. Thespeaker–hearersetting

Thisoverallsettingissimpletothepointofbeingsomewhatnaïve.Yet,thereisplentyofevidencetosupportit.Onepieceofevidenceisthetheoryofmindapproachtoautism(originallyBaron-Cohenetal.1985,inmoredetailandillustration,e.g.,Happé1994:38–44).Themainideabehindthislineofresearchis,roughly,thatautisticindividualslack(tosomeextent)thecapabilitytoconceiveofthefactthatotherindividualshaveknowledgeandbeliefsthatdifferfromtheirownknowledgeandbeliefs.Thismaybereformulatedbysayingthatautisticindividualseitherlackthecapabilityorhaveadifferentwayofformingprojectionsofotherpeople'sminds.

Happé(1994)illustratesthetheory-of-mind-relateddeficitwithasimpletestsettingknownastheSmartiestask(Perneretal.1989).IntheSmartiestask,thesubjectisfirstshownaclosedcandyboxthatcontainsapencil,andaskedwhatisinthebox.Typically,thesubject—whetherautisticornot—repliesthattheboxcontainscandy.Next,thesubjectisshownthattheboxinfactcontainsapencil.Then,thesubjectistoldsomethinglikethefollowing:“Inaminute,Billywillcomein.IwillshowtheboxtohimlikeIshowedittoyou,andaskwhatisinit.Whatwillhesay?”A4-year-oldnonautisticchildisabletoreplytothisquestiontakingintoaccountthefactthatBillydoesnotknowthecontentsofthebox.Anautisticchild,incontrast,assumesBillytoreplythattheboxcontainsapencil.

Ipointedoutinsection18.1that,inLambrecht'sview,anidentifiablereferentisoneforwhichthereisasharedrepresentationinthemindsofthespeakerandthehearer,whereasforanunidentifiablereferent,thereonlyisarepresentationinthespeaker'smind.ThewholeconceptofidentifiabilityisthusdependentonthesettingshowninFigure18.1,thatis,onhowtheinterlocutorsformprojectionsorideasofoneanother'smindandconsciousness.

18.3.3TopicandFocus

Lambrechtbuildsmostofhisinformationstructuretheoryonthenotionsoftopicandfocus.Hedevotesmostofhis1994booktoadiscussionofthesetwoconceptsandanalyseswhichillustratedifferentkindsoftopicandfocusstructures.Hedefinestopicandfocusasfollows:

Page 252: The Oxford Handbook of Construction Grammar

Information Structure

Page 7 of 11

Topic:thatwhichthesentenceisabout.(Lambrecht1994:118)Focus:Thesemanticcomponentofapragmaticallystructuredpropositionwherebytheassertiondiffersfromthepresupposition.(Lambrecht1994:213)

Asvagueasthisdefinitionoftopicmaysound,itisfairlywidelyused(fordiscussiononthenotionanditsdefinition,see,e.g.,Lambrecht1994:117–31;Erteschik-Shir2007:7–27).

Arangeofphenomenatypicallyusedforillustratingtheconceptoftopicistopicalization,orleftdislocation(thoughthetwoarenotcompletelysynonymous).Erteschik-Shir(2007:7–8)givesthefollowingDanishexamples:

(2)a.HunhilstepåOle.Hamhavdehunikkemødtfør…ShegreetedOle.Himhadshenotmetbeforeb.HunhilstepåOle.Hunhavdeikkemødthamfør…ShegreetedOle.Shehadnotmethimbefore

Theexamples(2a)and(b)differwithregardtothetopic.In(2a),theobjectpronounhamistopicalizedandoccursinthesentenceinitialposition.Correspondingly,thesentenceisinterpretedassayingsomethingaboutthereferentofthepronoun,thatis,aboutOle.In(2b),theobjectpronounisinitsunmarkedpositionaftertheverb,andthesentenceisconstrued(mostprobably)asbeingaboutthepersonreferredtoasshe.

Erteschik-Shir(2007:27)interpretsLambrecht'sdefinitionoffocusas“thenon-presupposedinformationinthesentence.”Sheillustratestheconceptwiththefollowingexample(Erteschik-Shir2007:28),which,inheranalysis,showsthepresuppositioninitalicsandthefocusinCAPITALS:

(3)a.Q:WhatdidJohndo?A:HeWASHEDTHEDISHES.b.Q:WhatdidJohnwash?A:HewashedTHEDISHES.c.Q:Whowashedthedishes?A:JOHNwashedthem.d.Q:Whathappenedtothedishes?A:JOHNWASHEDthem.e.Q:Whathappened?A:JOHNWASHEDTHEDISHES.f.Q:WhatdidJohndowiththedishes?A:HeWASHEDthem.

In(3a),thesubjectpronounheistheonlyelementinthesentencethatreferstoapresupposedentity.Therestofthesentence(i.e.,thewholeVP)isthusthefocus.In(3b),notonlythesubjectbutalsotheeventtypeispresupposed:inotherwords,itisestablishedinthediscoursethatJohnwashedsomething.Thenonpresupposedpart,andhencethefocus,iswhatJohnwashed(i.e.,thedishes).In(3c–f),thepresupposedpartsofthesentencevary,andthefocusvariesaccordingly,oneextremebeing(3e)whichhasnopresupposedelements.Hence,itisasentencefocusstructureinLambrecht'sterms(1994:233–35).

18.4.InformationStructureandSentenceForm

WithregardtoConstructionGrammar,themostimportantaspectofinformationstructurephenomenais,perhaps,theinfluencewhichtheyhaveontheformalstructureoflinguisticexpressions.Totheextentthatinformationstructurephenomenaareconventionalizedinalanguage,theymustbecodedaspropertiesofconstructions(atleastifweadopttheviewthatthegrammarofalanguageconsistssolelyofconstructions).Therefore,aproperandcomprehensiveaccountofthegrammarofanygivenlanguagewillincludeinformationstructure,andnotasaseparateaddendumorapragmaticcomponentbut,rather,asasetoffeaturesandpropertieswhichanyofthegrammaticalconstructionsinthatlanguagemayhave.

Tothisend,togetherwithLauraMichaelis,LambrechthasproceededtoimplementhisinformationstructuretheoryinthefeaturematrixnotationusedinConstructionGrammar,notablyinthebranchofConstructionGrammarreferredtoasBerkeleyConstructionGrammarbyFillmore(thisvolume)andUnificationConstructionGrammarbyGoldberg(2006a:213–17).TheyhaveincorporatedintotheConstructionGrammarfeaturematricesinformationconcerning,forexample,theactivationstatusesofeachpartofasentence,andthetopicandfocusofasentence(seeMichaelisandLambrecht1996;LambrechtandMichaelis1998).

8

Page 253: The Oxford Handbook of Construction Grammar

Information Structure

Page 8 of 11

CentraltotheworkofMichaelisandLambrechtisthefactthatinformationstructureproperties,includingactivationstatus,topicvs.focusrole,andpresupposedness,tendtobeconventionalized—or,indeed,grammaticalized—aspartsofsentence-sizedconstructions.This,andmoregenerallyintegratinginformationstructurefeaturesinsentencegrammar,maybeillustratedwiththeiraccountoftheEnglishnominalextrapositionconstruction—or,moreaccurately,thesentence(or“construct”)It'samazing,thedifferencewhichinstantiatestheNominalExtrapositionconstruction(MichaelisandLambrecht1996:227),asshowninFigure18.2.

Asthefigureshows,thesubjectpositionofthesentenceisoccupiedbythepronounit.Yet,thesentencealsoincludesthecoreferentialNPthedifference.Theconstructioninstantiatedbythissentenceincludestheinformationthatthesetworepresentthesamethetarole(#2)andthattheircommonreferentisactiveinthediscourse.Thesamereferentisalsothetopicofthesentence.Inotherwords,thespeakerassumesthereferentoftheNPthedifferencetobeactiveinthehearer'sconsciousnessatthetimeoftheutterance,andheexpresses,byutteringthissentence,apropositionthatconcernsthatparticularreferent(i.e.,theideathatthereferentisamazing).

Clicktoviewlarger

Figure18.2. NominalExtrapositionconstructionasillustratedbytheconstructIt'samazing,thedifference(MichaelisandLambrecht1996:227)

Oneofthecentralmotivatingforcesbehindinformationstructureresearchistheobservationthatanygivenlanguagemakesitpossibletoexpresstheverysamepropositionalmeaningoftenwithquiteanumberofformallydifferentsentences.Inthiscontext,Lambrechtusesthetermallosentence(presentedinDaneš1964).Thetermhasitsrootsinthestructuralisttradition:morphologyspeaksofmorphs,morphemes,andallomorphs,andphonologyspeaksofphones,phonemes,andallophones.Similarly,sentencegrammarmayspeakofallosentences.Theideaisthatinthesentencegrammarcontext,propositionscorrespondtothe‘emic’units.Thus,differentsentencesexpressingthesamepropositionareconsideredallosentencesofthatproposition.

Asphonologydescribesthecontextsofoccurrenceforeachoftheallophonesofagivenphoneme,agrammarwhichtakesinformationstructureseriouslydescribesthecontextsofoccurrencefordifferentallosentencesofagivenproposition.ThisideamaybeillustratedbyasetofsentencesexemplifyingwordordervariationsinFinnish.Forexample,thesentencesin(4a–d)allexpressthesamepropositionandareitsallosentences:

(4)a.KalleantoiomenanVillelle.name-nomgive-pst-3sgapple-accname-all‘Kallegavethe/anappletoVille.’b.VillelleKalleomenanantoi.name-allname-nomapple-accgive-pst-3sg‘ItwastoVillethatKallegavethe/anapple.’c.AntoiKalleVillelleomenan.give-pst-3sgname-nomname-allapple-acc‘Kalledidgivethe/anappletoVille.’d.OmenanVilleKallelleantoi.apple-accname-allname-nomgive-pst-3sg‘Itwasthe/anapplethatKallegavetoVille.’

Eachoneofthesesentencesisnaturalinadifferentsetofcontexts,andtheirnaturalnessandsuitabilitytocontextisregulatedfirstandforemostbyinformationstructure:whatthespeakerandthehearerknowbeforehand,whathasandhasnotbeendiscussedpreviouslyinthesamediscourse,whatisitthatisbeingtalkedabout,andsoon.

9

10

Page 254: The Oxford Handbook of Construction Grammar

Information Structure

Page 9 of 11

Asasidenote,Lambrechtspecificallywarnsofthefallacyofa‘pragmaticallyneutral’sentence.Therecanbenopragmaticallyneutralallosentence,sinceeveryallosentencehasitspragmaticusageconditions:

Justastherearenosentenceswithoutmorphosyntacticandphonologicalstructure,therearenosentenceswithoutinformationstructure.Sayingthatsomesyntacticorprosodicstructures‘haveaspecialpragmaticfunction’whileothersdonotissomewhatlikesayingthatsomemechanicaltoolshaveaspecialfunctionwhileothersarefunctionallyneutral.Accordingtothislogic,ascrewdriverforexamplewouldbesaidtohavea‘specialfunction’becausetheobjectsmanipulatedwithit(i.e.screws)musthaveaspecialshape,whileahammerwouldbesaidtobefunctionallyneutralbecauseitmaybeusedtodriveinvariouskindsofobjectsincludingnails,fencepoles,andifneedbeevenscrews.(Lambrecht1994:16–17)

Thus,theexamples(4a–d)allhavetheirspecificinformationstructuresandpossiblediscoursefunctions,whichmaybepartiallyoverlappingandpartiallyseparate.Onemightclaimthatthesentence(4a)isunmarkedascomparedto(4b–d)inthesensethatitapparentlyhasabroadersetofpossiblecontexts. However,itiscertainlynotwithoutapragmaticfunction,constraintsonpossibleusagecontexts,andinformationstructure.

ThetermallosentencewasfirstintroducedbyDaneštocapturewordordervariationslikethese,andwordordervariationisoftenthestartingpointofinformationstructureaccounts(e.g.,Erteschik-Shir2007:1).However,intheConstructionGrammarframework,wordorderhasreceivedsurprisinglylittleattention.Some(mostlypreliminary)suggestionshavebeenmade(e.g.,FriedandÖstman2004a:69–71;LeinoandKuningas2005;KuningasandLeino2006),butonecanhardlyspeakofatradition,letaloneconvention,ofconstruction-basedwordorderresearch.

Whilethetoolsofinformationstructureanalysispresentedaboveareessentialtodescribingwordordervariations,andwhileithardlymakessensetodiscusswordordervariationswithoutmakingreferencetoinformationstructure,onecannotequatethetwo.Forthepurposesofconstructionalanalysis,wordorderphenomenaandinformationstructurefeatures(i.e.,notionsliketopicvs.focus,presuppositionvs.assertion,activation,etc.)areseparatefeaturesorpropertiesofconstructions.Whilethereareobviouscorrelationsbetweenwordorderandinformationstructure,itisequallyobviousthatwordorderhasotherfunctionsbesidesmarkinginformationstructure(forabriefandinformativeoverview,seee.g.,Fried2003).Furthermore,wordorderisbynomeanstheonlyformalwayofcodinginformationstructurefeatures.

Ratherthanthinkingofwordorderandinformationstructureasinherentlyinterdependent,itisoftenmorefruitfultopostulateconstructionsthatcombineoneorbothofthetwowithotherlinguisticfeaturesandphenomena.Michaelis'sandLambrecht'saccountoftheEnglishNominalExtrapositionconstruction,showninFigure18.2,isacaseinpoint:itspecifiesthewordorderandthetopicandthefocusofthesentence,butitalsoincludesquitealotofotherinformationconcerningthesubjectandthepredicateofthesentence.Theconstructionalsospecifiesthatthesentence-initialpronounandthesentence-finalNParecoreferential,andsoon.

Also,theworkofMettouchi(e.g.,2003,2008)andKuningas(2007)onKabyle(Berber/Afro-Asiatic,spokeninAlgeria)sentencetypesandinformationstructureillustratestheinterconnectednessofinformationstructure,ontheonehand,andwordorderphenomena,ontheother,withsuchissuesasgrammaticalfunctions,casemarking,sentencemeaning,andthelike.Considerthefollowingexamples,takenfromKuningas(2007:79–80;Englishglossesadded.):

(5)a.teggenttlawintiγrifinthey.makewomenpancakes‘Thewomenmakepancakes.’b.tilawin,teggenttiγrifinwomenthey.makepancakes‘Thewomen,theymakepancakes.’c.dtiγrifiniteggenttlawinAUXpancakesRELthey.makewomen‘It'spancakesthatthewomenmake.’

Thesentence(5a)exemplifieswhatKuningascallsthecanonicalsentence(‘phrasecanonique’),byandlarge

11

Page 255: The Oxford Handbook of Construction Grammar

Information Structure

Page 10 of 11

correspondingtotheEnglishunmarkedSVOsentence;(5b)exemplifiestheleftdislocationsentence(‘phrasedisloquéeàgauche’),which,forthepurposesofthisillustration,correspondstotheEnglishLeftDislocationconstructiondiscussedabove;(5c)exemplifiesthecleftsentence(‘phraseclivée’),whichmoreorlesscorrespondstotheEnglishcleftsentenceusedhereasitstranslation.

AstheEnglishtranslationssuggest,thesesentencetypesdifferwithregardtotheirinformationstructure,notablywithregardtotheirchoiceoftopicandfocus,andwithregardtowhatistakentobepresupposedand/oractive.Formally,(5a)and(5b)differfromeachotherwithregardtowordorder,butalsowithregardtocasemarking:thewordfor‘women’isintheannexedcase(‘étatd’annexion’)in(5a)(i.e.,tlawin).In(5b),itisinthefreecase(‘étatlibre’)(i.e.,tilawin).Thecleftsentence(5c)resembles(5a)withregardtocasemarkingbutdiffersfromboth(5a)and(5b)withregardtowordorderandinotherrespectsaswell:takenliterally,itisapredicativeclausewhichessentiallycorrespondstoitsEnglishcounterpart.

Thisratherquicklookatasmallsetofexamplesissufficienttoshowthatinformationstructurephenomenatypicallyassociate,andgrammaticalizetogether,withsentencetypes:conventionalpairingsofmorphosyntacticandsemanticfeatures.Thus,overall,wordorderandinformationstructuredorelatetoeachotherveryclosely,buttheyarealsointrinsicallyintertwinedwithotheraspectsandphenomenaofgrammar.

18.5.Conclusion

Informationstructureisasplendidlywidetopic,anditisimpossibletogiveacomprehensivepictureofsucharangeofdifferent,thoughinterrelated,phenomenawithinthelimitsofabriefbookchapter.Theaimofthischapterhasbeentoprovidethereaderwithaselectionofreferencesandaroughmapfornavigatingtowardthoseandotherrelatedsources.Foranin-depthintroductiontoinformationstructure,especiallyintheConstructionGrammartradition,theinterestedreaderisreferredfirstandforemosttotheworkofKnudLambrecht,inparticulartohis1994monograph.

Asfortheplaceofinformationstructurein(construction)grammar,wemightconcludethatinformationstructuredependsongrammarandgrammardependsoninformationstructure.Itmaysometimesbeusefultotemporarilyseparateinformationstructure,morphosyntax,andsemanticsfromoneanotherforthesakeofsimplicityandtofocusononeovertheothers.Moreoftenthannot,however,givingareallyusefulandcompletedescriptionofaconstruction,orsentencetype,orasingleutterance,requirestakingallthreeaspectsofgrammarintoaccountandintegratingthemintothesameconstructionaldescription.Here,asinsomanyothercases,theexceptionalstrengthofConstructionGrammarisitscapabilityofintegratingseeminglydifferentbutintrinsicallyinterrelatedphenomenaintoacoherentdescriptiveapparatusandintocredibleandcomprehensiveanalyses.

Notes:

(1.)IshallreturntoLambrecht'snotionsofpragmaticpresuppositionandpragmaticassertioninsomemoredetailinsection18.3.1.

(2.)Foramorethoroughdiscussionofasimilarexample,seeLambrecht(1994:55–56).

(3.)Thisratherblack-and-whitedistinctionisanovergeneralization.Forabroaderpointofviewconcerningtheidentifiabilityofareferent,see,forexample,Dubois(1980),Chafe(1994:93–107),aswellasthediscussioninLambrecht(1994:77–92)andthereferencescitedintheseworks.

(4.)Bytheuniverseofdiscourse,Lambrechtreferstoanabstractionoftheparticipants,objects,propositions,andsoon,which‘exist’forthepurposesofthediscourse(cf.Lambrecht1994:36–43).Itconsistsoftheinterlocutors(or‘speechparticipants’),thespeechsetting,andthe‘text-internalworld’(which,inturn,consistsoflinguisticexpressionsandtheirmeanings).Theuniverseofdiscourseregulates,forexample,whatisandwhatisnot‘present’intermsofspeech.

(5.)Itmustbepointedout,however,thatspeakingof‘traditional’analysesofpresuppositionsandassertionsisagrossoversimplification.AsalreadyLevinson(1983:167)pointsout,“thereismoreliteratureonpresuppositionthanonalmostanyothertopicinpragmatics,”andtheamountofrelevantliteraturehasmultipliedsincehemade

12

Page 256: The Oxford Handbook of Construction Grammar

Information Structure

thisobservation.

(6.)Chafespeaksofactivationstatesaspropertiesofconcepts.IprefertofollowLambrechtininterpretingactivationstatusesaspropertiesofreferents(ortheirmentalrepresentations).

(7.)BothLambrechtandChafespeakofaschemainthiscontext.However,asLambrechthimselfpointsout(1994:99),“Chafe'snotionofaschemaanditsassociatedexpectations,whichhetakesfromcognitivepsychology,iscloselyrelatedtotheFillmoreannotionofasemanticframe.”

(8.)Foradiscussionofthistermanditsrelationshiptotheperhapsmorewidespreadnotionoftheticsentence(asopposedtocategoricalsentence),seeLambrecht(1987).

(9.)AsMukherjee(2001:89)pointsout,LambrechtusesthistermmorebroadlythanDaneš,whoonlyusesthetermfordifferentwordordervariationsofthesame“sentencepattern”(cf.Daneš1964:233).

(10.)Nom=nominativecase,pst=pasttense,3sg=3rdpersonsingular,acc=accusativecase,all=allativecase.

(11.)Lambrecht(1994:17)presentsthepragmaticmarkednessofgrammaticalstructuresasarelativenotion:giventwoallosentences,oneisunmarkedifitservestwodiscoursefunctionsandtheotheroneonlyservesoneofthetwo.

(12.)Thedbeforethewordtiγrifinin(5c)isanauxiliaryofpredicationwhichmarksthesentenceasapredicativestructure—likeroughlyhalfoftheworld'slanguages,Kabyledoesnotuseacopulahere—andtheimarkstherestasarelativeclause.ForafullConstructionGrammardescriptionofthesesentencetypes,theinterestedreaderisreferredtoKuningas(2007).AcomparisonoftheLeftDislocationconstructionexemplifiedin(5b)withsimilarstructuresin(spoken)FrenchandFinnish,seeLeinoandKuningas(2005).

JaakkoLeinoJaakkoLeinoisProfessorofFinnishattheUniversityofHelsinki.HehasworkedascoordinatoroftheFinnishgraduateschoolforlanguagestudies,Langnet,asProfessorofFinnishattheUniversityofJyväskyläandattheÅboAkademiUniversity,asLecturerofGeneralLinguisticsattheUniversityofHelsinki,andasresearcherattheResearchInstitutefortheLanguagesofFinland.HehaspublishedonFinnishhistoricalsyntax,cognitivedimensionsoftheConstructionGrammarframework,diachronyandvariationinConstructionGrammar,nonfiniteconstructions,wordorder,analyticalcausatives,andthesyntaxofspokenlanguage,Finnishdialectsyntaxinparticular.

Page 257: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,LanguageAcquisition,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0019

ConstructionGrammarandFirstLanguageAcquisitionHolgerDiesselTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapteranalyzesfirstlanguageacquisitioninthecontextofConstructionGrammar.Itprovidesanoverviewofusage-basedresearchinfirstlanguageacquisitionthatisbasedonassumptionsofConstructionGrammarandconsiderssomebasicassumptionsofConstructionGrammarthatunderlietheusage-basedapproachtofirstlanguageacquisition.Thechapteralsoconsidersthelexicalspecificityofearlypivotschemasthatchildrenuse,andtheirrelationshiptotheverb-argumentconstructionswhichtypifyadultlanguage.

Keywords:firstlanguageacquisition,ConstructionGrammar,usage-basedresearch,lexicalspecificity,children,verb-argumentconstructions,adultlanguage

19.1.Introduction

Therearetwomajortheoreticalapproachestothestudyofgrammaticaldevelopmentinfirstlanguageacquisition:thenativistapproach,whichrestsoncentralassumptionsof(Mainstream)GenerativeGrammar(cf.Chomsky1972),andtheusage-basedapproach,whichiscloselyassociatedwithConstructionGrammar(cf.Tomasello2003).Thetwogrammaticaltheoriesmakeradicallydifferentassumptionsaboutthenatureofgrammaticalelementsandtheoverallorganizationofthegrammaticalsystem.

GenerativeGrammarisaformalsyntactictheorythathascruciallyinfluencedresearchonfirstlanguageacquisitionforseveraldecades(seeO’Grady1997foranoverview).Inthisapproach,thecoreofgrammarconsistsofinvariableconceptsandconstraintsthatarepredeterminedbyaninnate‘languagefaculty’(seePinkerandJackendoff2005forarecentdiscussion).Therearetwocentralassumptionsthatunderlietheanalysisofsyntacticstructureinthisapproach.

First,GenerativeGrammarisbasedontheassumptionthatthelanguagefacultyconsistsofmodules.AccordingtoChomsky(1965),(mental)grammarcanbedividedintothreebasiccomponents:syntax,semantics,andphonology.Eachcomponent(ormodule)hasitsowncategoriesandrulesthatareinprincipleindependentofeachother.Onthisaccount,syntacticrepresentationsareautonomousinthesensethattheycanbeanalyzedwithoutreferencetomeaning.

Second,GenerativeGrammarisbasedontheassumptionthatsyntacticrepresentationsarederivedfromauniversalsetofsyntacticcategories.AlthoughthereisnogeneralconsensusamongGenerativeGrammariansastowhichcategoriesareuniversal(andinnate),researchersagreethatgrammaticalcategorieshavetobedefinedpriortoandindependentlyofparticularsyntacticconfigurations.Onthisaccount,syntacticrepresentationsareformedfromprimitivecategoriesprovidinga‘toolkit’(Jackendoff2002:75)fortheanalysisofsyntacticstructureinalllanguages.

Bothassumptions(i.e.,theassumptionthatsyntaxisautonomousandthatsyntacticstructuresarederivedfrom

Page 258: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 2 of 14

primitivecategories)arebasedontheinnatenesshypothesisofGenerativeGrammar.Accordingtothishypothesis,childrenarebornwithauniversalsetofformalsyntacticcategories,towhichGenerativeGrammariansreferas‘UniversalGrammar(UG)’orthe‘languagefaculty’(cf.PinkerandJackendoff2005).WhatchildrenhavetolearninthisapproachishowwordsandstructuresoftheambientlanguagearerelatedtoelementsofUG(cf.Pinker1984).AccordingtoChomsky(1999),grammaticaldevelopmentisaparticularcognitivephenomenonthatmustbedistinguishedfromlearning—atermthatChomskyconsidersinappropriatefortheanalysisoflanguageacquisition(cf.Chomsky1999:43).Learningisagradualprocessinwhichcategoriesareacquiredinapiecemealfashionfromexperience,whereasgrammaticaldevelopmentisaninstantaneousprocesswherebyelementsoftheambientlanguageare‘hookedup’tocategoriesofUG.Ifthecoreinventoryofgrammarisinnate,asinglelinguistic‘trigger’intheinputisinprinciplesufficienttoacquireaparticularlinguisticcategory(e.g.,Meisel1994:20).

Challengingthenativistaccount,usage-basedresearchershavedevelopedaverydifferentframeworkfortheanalysisofgrammaticaldevelopmentthatiscruciallyinfluencedbyassumptionsofConstructionGrammar.Inthisapproach,grammarisseenasadynamicsystemofconventionalizedform-functionunits(i.e.,constructions)thatchildrenacquirebasedondomain-generallearningmechanismssuchasanalogy,entrenchment,andautomatization.Onthisaccount,syntacticcategoriesarefluidentitiesthatemergefromprocessinglargeamountsoflinguisticdata(cf.Tomasello2003;Dąbrowska2004;Diessel2004;Goldberg2006a;Bybee2010).

Althoughusage-basedlinguistsemphasizetheimportanceofexperienceforlanguageacquisition,theydonotgenerallydenytheroleofinnateconstraintsforgrammaticaldevelopment.AccordingtoElmanetal.(1996),wehavetodistinguishbetweendifferenttypesofinnateknowledge.Specifically,Elmanetal.suggestedthatassumptionsaboutinnaterepresentationsmustbedistinguishedfromassumptionsabouttheinnatearchitectureofgrammar(orhumancognition).Generativelinguistsemphasizetheimportanceofinnatelinguisticrepresentations(i.e.,theimportanceofinnatecategoriessuchNPorCPforgrammarlearning);whereasusage-basedlinguistsemphasizetheimportanceofthegrammaticalarchitectureforsyntacticdevelopment(anddiachroniclanguagechange).ThearchitecturalconstraintsarecommonlycharacterizedbyassumptionsofConstructionGrammar,whichhascruciallyinfluencedthedevelopmentoftheusage-basedapproach.Infact,usage-basedlinguistshavedrawnsofrequentlyonconceptsofConstructionGrammarthatConstructionGrammariscommonlyseenasanintegralpartoftheusage-basedapproach(cf.Tomasello2003;Diessel2004;Goldberg2006a;Langacker2008a;Bybee2010).

Inwhatfollows,Iprovideanoverviewofusage-basedresearchinfirstlanguageacquisitionthatisbasedongeneralassumptionsofConstructionGrammar.Thechapterisdividedintofoursections.Section19.2considerssomebasicassumptionsofConstructionGrammarthatunderlietheusage-basedapproachtofirstlanguageacquisition,section19.3providesanoverviewofresearchonchildren'searlyitem-basedconstructions,section19.4isconcernedwiththeemergenceofconstructionalschemas,andsection19.5considersthedevelopmentofcomplexsentenceconstructions.

19.2.TheoreticalFoundations

Incontrastto(Mainstream)GenerativeGrammar,ConstructionGrammardoesnotpresupposeapredefinedsetofgrammaticalcategories.However,likeallgrammaticaltheoriesConstructionGrammarmakesparticularassumptionsaboutthenatureofgrammaticalentitiesandtheoverallorganizationofthegrammaticalsystem.TwoassumptionsofConstructionGrammarareessentialtounderstandtheusage-basedanalysisofgrammarlearning.

■First,ConstructionGrammarassumesthatsyntacticstructuresaresymbolicunits(i.e.,constructions)thatcombineaparticularformwithaparticularmeaning(cf.Goldberg1995:4).

■Second,ConstructionGrammarassumesthatconstructionsareassociatedwitheachotherbyvarioustypesoflinksthatconstituteacomplexnetworkofsymbolicexpressions(cf.Langacker1987).

Clicktoviewlarger

Figure19.1. Examplesoflexicalandconstructionalsymbols

Page 259: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 3 of 14

Ifgrammarconsistsofconstructions,thereisnoprincipleddifferencebetweenwordsandgrammaticalassemblies(cf.Goldberg1995;Langacker1987).Likewords,constructionsaresymbolicentitiesinwhichaparticularmeaningorfunctionismappedontoaparticularform(i.e.,aparticularconfigurationofstructuraland/orlexicalelements).Forinstance,animperativesentencesuchasOpenthedoor!canbeseenasaconstruction(i.e.,acomplexlinguisticsign),inwhichaparticularstructuralpatternisassociatedwithaparticularillocutionaryforce:animperativesentenceconsistsofanuninflectedverbform,lacksanovertsubject,andfunctionsasdirectivespeechact(seealsoHoffmann,thisvolume).Figure19.1showstheparallelstructuresofwordsandconstructionsthatunderlietheusage-basedapproach.

Theparallelismbetweenlexicalandgrammaticalsymbolshasledsomeresearcherstoapplythenotionofconstructiontoalllinguisticexpressionsincludingwordsandboundmorphemes(cf.Goldberg1995:4),ortocharacterizeconstructionsas‘bigwords’(cf.Dąbrowska2000).However,inthischapter,wordsandconstructionsarekeptseparate.Althoughbothcanbeseenassymbols,thenotionofconstructionisreservedforlarger(andschematic)grammaticalunits,notablyforphrases,clauses,andsentences.

Theassumptionthatgrammarconsistsofsymbolsisconsistentwiththeusage-basedhypothesisthatlinguisticknowledgeisdomain-general.Cognitivepsychologistshaveemphasizedtheimportanceofsymbolsforreasoningandabstractthought.AccordingtoDeacon(1997),symbolsenablepeopletotalkaboutentitiesthatarenotimmediatelypresentinthespeechsituation,providingacognitivefoundationforhigher-levelcognitiveprocesses.Moreover,symbolsencodeparticularperspectivesonentitiesandsituations,whichTomasello(1999)relatestotheuniquelyhumanabilitytounderstandthatotherpeopleviewtheworldfromadifferentperspective.Nootherspeciesisabletoconsiderthementalrepresentationsandviewpointsofotherbeings,whichisreflectedinthefactthatnootherspeciesisabletodealwithsymbols—onlyhumansare(cf.Tomasello1999).Thus,ifweassumethatgrammarconsistsofsymbols(i.e.,constructions),wearenotmakingfar-reachingassumptionsaboutinnatelinguisticrepresentationsasinGenerativeGrammar;theonlythingweclaimisthatgrammarissymbolic,whichcanbeseenasadomain-generalaspectofhumanthought.

Clicktoviewlarger

Figure19.2. NetworkofArgumentStructureconstructions(Goldberg1995:109)

ThesecondgeneralassumptionofConstructionGrammar(i.e.,thatgrammarconstitutesanetwork)isequallybroad(i.e.,domain-general).Ifgrammarconsistsofconstructions(i.e.,ofcomplexlinguisticsigns),itisjustconsequenttoassumethatthegrammaticalsystemisorganizedinasimilarwayasthementallexicon(cf.Langacker1987)orasnonlinguisticconceptsinmemory(cf.Harley2001).Thereisabundantevidencefrompsycholinguisticresearchthatpeopleassociatewordsandmorphemeswitheachotherbasedonoverlappingandcontrastingfeatures,whichhasledpsychologiststocharacterizethementallexiconasanetworkofsymbols(cf.CollinsandLoftus1975).Inanalogytothementallexicon,grammarcanbeseenasanetworkofcomplexlinguisticsignsthatareassociatedwitheachotherbyvarioustypesoflinks(cf.Langacker1987,2008a;Goldberg1995,2006a;Bybee2010).Figure19.2showsanexampleofthenetworkarchitectureofconstructions,whichGoldberg(1995:109)usedtorepresentthesemanticandstructuralrelationshipsbetweenVerb-ArgumentconstructionsinEnglish.

ThenetworkmodelofConstructionGrammarprovidesahighlyflexibleframeworkforthesyntacticanalysisoflinguisticstructurethatisconsistentwiththeusage-basedhypothesisthatgrammarconsistsoffluidcategoriesandvariableconstraints.Bothindividualconstructionsandtheparticularorganizationofthenetworkareemergentphenomenathatchildrenacquireinapiecemealbottom-upfashionbasedongenerallearningmechanismsthatarealsorelevantfortheacquisitionofknowledgeinothercognitivedomains(cf.Elmanetal.1996;Tomasello2003;

Page 260: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 4 of 14

Goldberg2006a;Bybee2010).

Inwhatfollows,Iprovideanoverviewofempiricalresearchonthedevelopmentofconstructionsinearlychildlanguage(forearlierreviewsofconstruction-basedresearchonfirstlanguageacquisition,seeTomasello2000b,2003;seealsoDąbrowska2004andGoldberg2006a).

19.3.EarlyItem-BasedConstructions

Theearliestutterancesthatchildrenproduceconsistofisolatedwordsandholophrases(i.e.,unanalyzedchunksofmultiwordexpressionsthatarelearnedasfrozenunits)(cf.1–3).

(1)Get-it(2)All-gone(3)What-s-that?

Theseearlywordsandholophrasesareoftentiedtoparticularcommunicativesituations.Theycanbeseenasholisticsymbolsthatchildrenuseasspeechactstoaccomplishparticularcommunicativegoals(cf.Tomasello2003:36–40).Afewmonthsafterchildrenproducetheirfirstone-wordutterancesandholophrases,theybegintousemorecomplexutterancesconsistingoftwoorthreewords.Theemergenceofchildren'searlymultiwordconstructionsinvolvestwocomplementarystrategies.Insomecases,childrencombinetwoormorewordswhichtheypreviouslyusedinisolationunderasingleintonationcontour(cf.Clark2003:162–65).Thisstrategyisconsistentwiththeclassicscenarioofgrammarlearninginwhichsyntacticstructuresarecompiledfromsmallerunits(cf.Pinker1984).Inothercases,children'searlymultiwordconstructionsarenotderivedfromtwoseparatewordsbutfromfrozenmultiwordexpressionsthataresegmentedintotheircomponents(cf.Lievenetal.1997;PineandLieven1993).Forinstance,manychildrenbegintoproducecontentquestionsinunanalyzedexpressionssuchasWhassis?orWhatchadoing?,whichareonlylaterdividedintoseparatewords(seebelow).

Oneofthemoststrikingpropertiesofchildren'searlymultiwordconstructionsisthattheyareoftenorganizedaroundparticularwords(seeTomasello2000aforareview).Inaclassicstudy,MartinBraine(1976)characterizedtheseearlyitem-basedconstructionsas‘pivotschemas’(or‘pivotconstructions’).Analyzingchildlanguagedatafromseveralunrelatedlanguages(e.g.,English,Finnish,Samoan),Braineobservedthatchildren'searlymultiwordutterancesarecommonlycomposedofaspecific‘pivotword,’thatis,arelationalterm(e.g.,averb),andan‘openslot’thatcanbefilledbyvariousexpressionsaslongastheseexpressionsaresemanticallyappropriateinaparticularposition(cf.Bowerman1973).Forinstance,ascanbeseeninTable19.1,English-speakingchildrenmakecommonuseofpivotschemasthatcombineaparticularquantifier(ornegativeterm)withareferentialexpression.

Althoughpivotschemascanoccurwithalargevarietyoflexicalexpressionsinthevariableposition,Braineemphasizedthateachpivotschemaisanisolatedconstructionthatchildrenacquireindependentlyofotherpivotschemas,challengingthegenerativehypothesisthatgrammaticaldevelopmentinvolvestheacquisitionofgeneralrulesandcategoriesthatareindependentofparticularlexicalexpressions.

Inaccordancewiththisanalysis,Tomasello(1992)observedthatchildren'searlyVerb-Argumentconstructionsarecommonlyorganizedaroundspecific

Page 261: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 5 of 14

Table19.1.Examplesofpivotschemas(adoptedfromBraine1976)

More__ All__ No__

Morecar Allbroke Nobed

Morecereal Allclean Nodown

Morecookie Alldone Nofix

Morefish Alldressed Nohome

Morehot Alldry Nomama

Morejuice Allshut Nopee

Moresing Allwet Noplug

verbs.Analyzingdiarydatafromhis2-year-

olddaughterTravis,hefoundthatasubstantialnumberofTravis'searlytwo-andthree-wordutterancesarebasedonparticularverbs,whichsheassociatedwithspecificstructuralpatterns.AlthoughverbsarecommonlyusedacrossawiderangeofVerb-Argumentconstructions(seebelow),Tomasello'sdaughterusedmostofherearlyverbsinparticularframes.Eachverbwastiedtoaparticularstructuralpattern,whichTomasellocharacterizedasa‘constructionalisland’(cf.Table19.2).

LikeVerb-Argumentconstructions,questionsareinitiallyorganizedaroundparticularwords;theyexhibitthesame‘pivotlook’asotherlexicallyspecificconstructions.Inthegenerativeliterature,theacquisitionofquestionsinvolvestheacquisitionofhighlygeneralsyntacticoperationssuchaswh-movementandsubject-auxiliaryinversion(cf.Radford1990).However,thereiscompellingevidencethatchildren'searlyquestionsarelexicallyspecificformulasconsistingofaparticularquestionwordand/oraparticularauxiliarythataretiedtospecificstructuralpositions(theclassicanalysisofchildren'searlyquestionsisKlimaandBellugi1966;foraconstruction-basedanalysisofchildren'squestions,seeDąbrowska2000;RowlandandPine2000;DąbrowskaandLieven2005;Rowland2007;andDąbrowskaetal.2009).Forinstance,children'searlyyes-noquestionsareusuallyformedfromaparticularauxiliaryandaparticularsubjectpronoun,whichtogetherconstituteaframe(e.g.,CanI__?orWillyou__?).Similarly,wh-questionsarecommonlyderivedfromlexicallyspecificformulas(cf.Dąbrowska2000).Theearliestwh-questionsthatEnglish-speakingchildrenproducearefrozenexpressionssuchasWhat-s-that?orHow-do-you-know?Aschildrengrowolder,theseearlyholisticexpressionsaresegmentedintotheircomponentsandtheuseofwh-questionsbecomesgraduallymoreflexible.Considerforinstancetheexamplesin(4)to(12),whichDąbrowska(2000)providedtoillustratethedevelopmentofaparticulartypeofquestioninthespeechofa2-year-oldchildnamedNaomi.

Page 262: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 6 of 14

Table19.2.Examplesofverb-islandconstructions(adoptedfromTomasello1992)

Findit__ __getit __gone

Find-itfunny Blockget-it PeterPangone

Find-itbird Bottleget-it Raisinsgone

Find-itchess Phoneget-it Doo-doogone

Find-itbricks Towelget-it Cherrygone

Find-itWeezer Bedusget-it Foxgone

Find-itball Coffeeget-it Hammergone

Finditstick Mamaget-it Frenchfriesgone

(4)Whatdoing?(manytimes)1;11.11(5)WhatsMommydoing?(manytimes)1;11.21(6)Whatsdonkeydoing?(4times)2;0.18(7)WhatsNomidoing?(2times)2;0.18(8)Whatstoydoing?2;0.18(9)WhatsMommyholding?2;0.26(10)WhatsGeorgiesaying?2;1.19(11)Whatistheboymaking?2;11.17(12)WhatisAndydoing?2;11.18

Ascanbeseenintheseexamples,thedevelopmentoriginatesfromaverysimplepatternconsistingofthequestionwordwhatandtheverbdoing,whichNaomiusedmanytimesbeforewhatappearedinanyothercontext.Later,thechildinsertedthenounMommyintothispattern,whichagainwasusedmanytimesbeforeNaomibegantoproducequestionswithdifferenttypesofnounsandabitlateralsowithdifferenttypesofverbs.Attheendoftheanalyzedperiod,Naomirecognizedthatthequestionwordwhatandtheauxiliaryisareseparatewordsandabandonedthecontractedformwhats,whichonlyrecurredinNaomi'suseofwh-questionsafterafewmonths.

Obviously,thispathwayofdevelopmentisdifficulttoreconcilewiththehypothesisthattheacquisitionofquestionsinvolvestheacquisitionofgeneralsyntacticoperations.Thereisnoevidencethatchildren'searlyquestionsinvolvewh-movementorsubject-auxiliaryinversion.However,whenchildrengetoldertheyoftenproducewh-questionswithuninvertedword-order(cf.13)andothersyntacticerrors(cf.14–17),whichsomeresearchersinterpretedasevidencefortheemergenceofsubject-auxiliaryinversioninearlychildlanguage(cf.Santelmannetal.2002).

(13)Whyhecango?(Non-inversionerror)(14)Whatcanhecando?(Doublemarkingerror)(15)Whatdoeshelikes?(Agreementerror)(16)Wheredoeshergo?(Caseerror)(17)Doeshegoingtotheshops?(Auxiliary-verbmismatch)

1

Page 263: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 7 of 14

Clicktoviewlarger

Figure19.3. Errorrateinchildren'squestionswithandwithoutapriorframe(basedondatafromRowland2007:123,Table6)

However,ifwelookattheseerrorsmorecloselywefindthattheyareconsistentwithaframe-basedanalysisofchildren'searlyquestions.ExaminingseveralthousandinterrogativesentencesfromtenEnglish-speakingchildrenaged2to5,Rowland(2007)foundthatchildren'serrorswithbothyes-noquestionsandwh-questionscorrelatewiththefrequencyofparticularwordcombinationsatthebeginningofthesentence(seealsoRowlandandPine2000).Specifically,sheshowedthatchildrenarepronetoproduceinversionerrorsandothertypesofmistakesinlexicallyinnovativequestions(i.e.,inquestionswithinnovativecombinationsofauxiliaries,subjects,andinterrogativepronouns).AscanbeseeninFigure19.3,questionsthatarebasedonapriorlexicalframe(i.e.,aspecificlexicalpatternsuchasWhat's__orCanI__,whichthechildrenofRowland'sstudyhadusedinearliertranscripts)aremuchlesslikelytoincludeanerrorthanquestionsthatarenotbasedonapriorframe(e.g.,DoesNP__,Whocan__,Whatmust__),suggestingthatchildren'slackofexperiencewithspecificpatterns,ratherthananabstractsyntacticoperation,accountsfortheirproblemswithproducingquestions.Ifchildrenhaddifficultieswiththeacquisitionofsubject-auxiliaryinversion,onewouldexpectanequalnumberoferrorsacrossallquestiontypes.Thelexicallyspecificnatureoftheirerrors,however,providesstrongevidenceforthehypothesisthatchildrenacquirequestionslikeothersentence-levelconstructionsbasedonearlyformulasorlexicallyspecificframes.

19.4.TheEmergenceofConstructionalSchemas

Likechildren'searlypivotschemas,theconstructionsofadultlanguageareoftenlexicallyparticular.Manyadultconstructionsincludespecificwordsoraffixesthatcannotbereplacedbyotherlexicalexpressions.Forinstance,theComparativeCorrelativeconstruction(e.g.,Thefasterthebetter)consistsoftwoparticulargrammaticalmarkers,thatis,twoinstancesofthe,andtwocomparativephrases,whichtogetherconstituteauniqueframe(cf.Fillmoreetal.1988;Hoffmann,thisvolume).Similarly,theHortativeconstruction(e.g.,Let'sgotothemovies)includesaparticularlexicalframeconsistingoftheauxiliaryletandacontractedpronoun,whichtogethermarkaparticularsentencetype(cf.HopperandTraugott2003:10–13).Thus,likechildlanguage,adultlanguageincludeslexicallyparticularconstructionsthatrevolvearoundspecificwords.Infact,itisacentralassumptionoftheusage-basedapproachthatgrammaticalpatternsarecommonlyassociatedwithparticularlexicalexpressions(cf.Langacker1987),eithercategoricallyasintheaboveexamples,inwhichparticularconstructionsaremarkedbyspecificwords(e.g.,The__the__;Lets__),orprobabilisticallyasinthecaseofVerb-Argumentconstructionsthatoftenstronglypreferparticularnounsandverbs(cf.StefanowitschandGries2003).

Nevertheless,therelationshipbetweenwordsandconstructionsismuchmorevariableinadultlanguagethaninchildren'searlypivotschemas.Withage,children'sconstructionsbecomeincreasinglymoreabstractandindependentofparticularlexicalexpressions.Therearetwoimportantaspectsofgrammaticaldevelopmentthatreflecttheincreasingvariabilityandabstractnessofchildren'sVerb-Argumentconstructions.

First,the‘slots’ofchildren'searlypivotschemasarecommonlyextendedfromafewwordstoawholeclassoflexicalexpressions.Forinstance,intheprevioussectionwesawthatchildren'searlyquestionsareinitiallyoftentiedtospecificsubjectpronounsthatarelaterreplacedbyotherreferentialexpressions(e.g.,CanI→CanMommy).Similarextensionsoflexicalexpressionsinaparticular‘slot’havebeenobservedinthedevelopmentofmanyotheritem-basedconstructions.Consider,forinstance,thechildutterancesin(18)to(22),whichillustratethedevelopmentofthe‘objectslot’aftertheverbwantinspontaneouschildlanguage(dataandanalysisfromDiessel2004:68–72).

Page 264: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 8 of 14

(18)Iwannabag.(Sarah2;3)(19)Iwannaride(myhorsie).(Sarah2;3)(20)Iwanticecreamintherefrigerator.(Sarah2;10)(21)Wantmeopenit?(Adam2;9)(22)Dowanthewalk?(Adam2;10)

Atfirst,childrenusetheverbwantanditsphoneticvariantswanandwannaincombinationwithafirstpersonpronounassubjectandanominalcomplement,asinexample(18).However,soontheybegintousethesameverbforms(i.e.,want,wan,andwanna)alsowithinfinitivalcomplements(cf.19),butthenittakesseveralmonthsuntiltheyproducemorecomplexconstructionsinwhichanominalcomplementiselaboratedtoanonverbalclause(cf.20)ortoacomplexinfinitiveinwhichthenominalobjectofwantservesasthesemanticsubjectofanonfinitecomplementclause(cf.21).Interestingly,ascanbeseeninexample(22),insomecaseschildrenrealizethesyntacticobjectofwantbyasubjectpronoun,suggestingthattheyoverextendtheuseofinfinitivalcomplementstofinitecomplementclauses.Whatthisexampledemonstratesisthattheslotsofchildren'searlypivotschemasbecomeincreasinglymoreflexibleandschematic,thusmakingtheuseofchildren'searlyitem-basedconstructionsgraduallymoreproductive.

Thesecondaspectofsyntacticdevelopmentthatincreasesthevariabilityandproductivityofchildren'searlyconstructionsconcernstheroleofverbs,whichcommonlyemergeasthepivotwordsofitem-specificconstructions.Asnotedabove,children'searlyverbsareusuallytiedtospecificsyntacticpatterns(i.e.,particularconstructions),whereastheverbsofadultlanguageareoftenusedacrossseveralconstructions.Forinstance,ascanbeseeninexamples(23)to(26),theverbbreakcanoccurintheTransitiveandIntransitiveconstructions(23–24),intheCausedMotionconstruction(25),andinthePassiveconstruction(26)withminimaldifferenceinmeaning.

(23)Hebrokehisarm.(Transitiveconstruction)(24)Thewindowbroke.(Intransitiveconstruction)(25)Shebrokethevaseintopieces.(CausedMotionconstruction)(26)Themirrorisbroken.(Passiveconstruction)

Inchildlanguage,theuseofbreakisverydifferent.Forinstance,Tomaselloobservedthathis2-year-olddaughterTravisusedbreakonlyintransitiveclauses(orinone-wordutterances);thereisnotasingleexampleofbreakintheIntransitiveconstructionortheCausedMotionconstructioninTravis'sdataandthefewutterancessheproducedwithbroken(e.g.,Brokenglass—afterTravisbrokeit)donotseemtobepassivesentencesandmayhavebeenproducedindependentlyoftheuseofbreakintheTransitiveconstruction(cf.Tomasello1992:108–9).Thus,sincechildren'searlyverbsaretiedtoparticularconstructions,wehavetoaskwhenandhowdotheylearntoextendtheuseofverbsacrossconstructions?

Thisisoneofthemostfundamentalquestionsintheconstruction-basedapproachtofirstlanguageacquisitionthathasbeenaddressedinawiderangeofstudies(seeTomasello2000bforareview).ThequestionisimportantbecausetheextensionofverbsacrossconstructionsmarksamilestoneinthedevelopmentofmoreschematicVerb-Argumentconstructions,whichcruciallyincreasethechild'sabilitytousesyntacticpatternsproductively.Aslongaseachverb(oreachpivotword)isusedastheconstantpartofaconstructionalisland,children'sproductivitywithgrammarislimited.Apivotschemaallowsthechildtovarytheelementsinaparticularslot;butatthesametimeitforcesthechildtopresentaneventfromaparticularperspective.Aslongaschildren'sgrammaticalknowledgeisbasedonpivotschemas(orconstructionalislands),theyhaveonlyonewayofpresentingaparticularevent.However,whenchildrenbegintoextendtheuseofindividualverbsacrossconstructionstheygainthepossibilityofchoosingbetweenalternativeconstructionsfortheexpressionofthesameevent(orthesamescene),makingtheuseofgrammarmuchmoreproductive.Whatchildreneventuallylearnisanetworkofrelatedconstructionsinwhichthesameeventisconstruedfromdifferentperspectivessothatspeakerscanchoosetheconstructionthatismostappropriatetorealizetheircommunicativeintentioninaparticularsituation.

TherearetwomethodologiesthatchildlanguageresearchershaveusedtoinvestigatetheemergenceoflinguisticproductivityinthedomainofVerb-Argumentconstructions.Eithertheyhaveanalyzedtheoccurrenceofoverextensionerrorsinspontaneouschildlanguage(cf.Bowerman1982b,1988;Clark1987;Pinkeretal.1987),

Page 265: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 9 of 14

ortheyhavedesignedexperimentstotestifchildrenarewillingtoextendanovelverbtoanotherconstruction(cf.Pinkeretal.1987;TomaselloandBrooks1998;Brooksetal.1999;BrooksandTomasello1999a).

Thereisabundantevidencethatchildrenoverextendtheuseofverbsinspontaneouslanguage.Atfirst,children'suseofverbsisveryconservative.EachverbisassociatedwithaparticularsyntacticpatternthatreflectsthefrequencyofindividualVerb-Argumentconstructionsintheambientlanguage.However,whenchildrengetoldertheyrealizethestructuraloverlapbetweendifferentitem-basedconstructionsandbegintogeneralizeacrossthem,creatingschematicrepresentationsthatarenolongertiedtoparticularlexicalexpressions.Theemergenceofsuchschematicconstructionsisreflectedintheoccurrenceofoverextensionerrors,whichhavebeeninvestigatedbyBowerman(1982b,1988),Clark(1987),andothers.Forinstance,Bowerman(1982b)observedthathertwodaughtersextendedtheuseofvariousintransitiveverbstotheTransitiveconstructioninwaysthatarenotlicensedbyadultgrammar.

(27)Kendallfallthattoy.2;3(28)Whodeadedmykittycat?2;6(29)Theyjustcoughme.2;8(30)Don’tgiggleme.3;0(31)Iamgonnaputthewashraginanddisappearsomethingunderthewashrag.3;7

ChildrenalsooverextendtheuseoftransitiveverbstotheIntransitiveconstruction,buttheseerrorsaremuchlessfrequent(cf.Bowerman1988).IfchildrenextendatransitiveverbtotheIntransitiveconstruction,theresultingstructureusuallyhasapassive(i.e.,unaccusative)meaning(cf.examples(32)–(34)fromClark2003:235).

(32)Theflowercuts.[=Theflowercanbecut]2;8(33)Bertknockeddown.[=Bertgotknockeddown]3;0(34)Theydon’tseemtosee.[=Theycannotbeseen]3;8

OtheroverextensionerrorsinvolvetheuseofintransitiveverbsinthePassiveconstruction(35–37),theuseofsimpletransitiveverbsintheResultativeconstruction(38–40),andtheuseoftransitiveverbsintheDitransitiveconstruction(41–42)(examplesfromPinkeretal.1987andBowerman1988):

(35)Hegetdied.3;8(36)Idon’tlikebeingfalleddownon.ageunclear(37)Idon’twanttogetwaded(on).ageunclear(38)Ipulleditunstapled.3;8(39)Iampattingherwet.4;0(40)Areyouwashingmeblind?5;6(41)I’llbrushhimhishair.2;3(42)Isaidherno.3;1(43)Buttonmetherest.3;4

Notethatmostoftheerrorsinexamples(27)to(43)occurafterthethirdbirthday;thereareafewerrorsfromanearlierage,buttheyarerare,suggestingthattheproductivityofVerbArgumentconstructionsemergesonlygraduallyduringthepreschoolyears.Thishypothesisissupportedbyfindingsfromexperimentalresearchwithpreschoolchildren.

Pinker,Lebeaux,andFrost(1987)conductedanexperimentinwhich4-year-oldEnglish-speakingchildrenweretaughtnovelverbsinActiveTransitiveconstructionsandPassiveconstructions.Thenovelverbsreferredtoatransitivescenewithtwoanimatenounsthatbothcouldfunctionasagentorpatient.Forinstance,oneoftheverbsthatthechildrenhadtolearnreferredtoasituationinwhichatoydogrubbedtheneckofatoyelephant.Participantsweredividedintotwogroups.OnegroupofchildrenlearnedthenovelverbsinanActiveTransitiveconstruction(e.g.,Thedogispellingtheelephant),andtheothergrouplearnedtheminaPassiveconstruction(e.g.,Theelephantisbeingpelledbythedog).Attest,childrenhadtoanswertwotypesofquestionsthatweredesignedtopulleitherforapassivesentence(e.g.,Whatishappeningtothepatient?)orforanactivesentence(e.g.,Whatistheagentdoing?).

Figure19.4showstheproportionsofactiveandpassivesentencesthechildrenproducedinresponsetothetwo

Page 266: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 10 of 14

questions.Ascanbeseen,4-year-oldchildrenusebothconstructionsproductively;thatis,theyproducedpassivesentenceswithverbstheyhadonlyheardinactivevoice,andtheyproducedactivesentenceswithverbstheyhadonlyheardinpassivevoice.Notethatthechildrenweremorelikelytogeneralizeverbsfrompassivesentencestoactivesentencesthanviceversa:86%ofthechildrenwhohadheardaverbonlyinthepassiveusedthisverbalsoinactivesentences,butonly64%ofthechildrenwhohadheardaverbonlyinactivevoiceusedthisverbalsoinpassivesentences,suggestingthattheTransitiveSVOconstructionisastrongerattractorthanthePassiveconstruction(cf.Brooksetal.1999).Notealsothatchildren'sexperiencewiththenovelverbsinfluencedtheirperformance.Althoughthechildrenwereabletoextendanovelverbtoconstructionsinwhichtheyhadneverheardtheverbbefore,theyusedthenovelverbsmorefrequentlyinconstructionstheyhadlearnedinthecourseoftheexperiment.

Clicktoviewlarger

Figure19.4. Proportionofsuccessfullyelicitedactiveandpassiveconstructions(basedondatafromPinker,Lebeaux,andFrost1987:125,Table4)

Inasimilarstudy,BrooksandTomasello(1999a)showedthatevenyoungerchildrenareabletoextendanovelverbtheyhadonlyheardinpassivevoicetoconstructionsinactivevoiceandviceversa.However,theabilitytoextendanovelverbfromoneconstructiontoanotherincreaseswithage.Comparingchildrenfromtwodifferentagegroups,BrooksandTomaselloshowedthatyoungerchildren(meanage34.6months)aresomewhatlesslikelytoextendanovelverbfromactivetopassivesentences(andviceversa)thanolderchildren(meanage40.9).However,thedifferencebetweenthetwoagegroupswasrelativelysmall.Inanotherstudy,TomaselloandBrooks(1998)exposedchildrenoftwoyoungeragegroups(meanage:24monthsvs.30months)tonovelverbsinTransitiveandIntransitiveconstructions.Inthiscase,therewasacleardifferenceinperformancebetweenthetwoagegroups.Aboutone-thirdoftheolderchildrenshowedatleastsomeproductivitywithnovelsverbs,whereasthevastmajorityoftheyoungerchildrenusedthenovelverbsonlyinconstructionstheyhadexperiencedduringtraining,suggestingthatveryyoungchildrenarereluctanttoextendverbsacrossconstructions.Similarage-relateddifferencesintheproductiveuseofVerb-Argumentconstructionswereobservedinotherexperimentalstudiesusingothermethodologiesandfocusingonotherconstructions(cf.AktharandTomasello1997;Akthar1999;Brooksetal.1999).

Whiletheextensionofverbsacrossconstructionscanbeseenasamilestoneinthedevelopmentofschematicconstructions,itmustbeemphasizedthattheverbsofadultlanguagecannotbefreelyextendedacrossconstructions.Insomecases,particularverbsaretiedtospecificconstructions.Forinstance,incontrasttobreak,whichcanappearinbothIntransitiveandTransitiveconstructions(seeabove),theverbfallcanonlyoccurintheIntransitiveconstructionbecauseifaspeakerwantstoexpressthatanagentmadeanobjectfallheorshecanusetheverbdrop,whichcanbeseenasthetransitive(orcausative)counterpartoftheintransitivefall(i.e.,‘drop’=‘makefall’).Inotherwords,theexistenceofdroppreemptstheoccurrenceoffallintheTransitiveconstruction.BrooksandTomasello(1999b)showedthatwhenchildrenareawareofapre-emptingverbformsuchasdrop,theyarereluctanttoextendanovelintransitiveverbtotheTransitiveconstructionandviceversa,suggestingthatpreemptionisanimportantfactorforconstrainingchildren'sVerb-Argumentconstructions(cf.Clark1987;Bowerman1988;BraineandBrooks1995).Arelatedfactorthatconstrainschildren'sproductiveuseofverbsisfrequencyorentrenchment.AsBrooksetal.(1999)havedemonstrated,afrequentverbthatchildrenonlyexperienceinoneconstructionislesslikelytobeextendedtoanotherconstructionthananinfrequentone.Forinstance,childrenaremorelikelytoextendtheinfrequentintransitiveverbvanishtotheTransitiveconstruction(e.g.,Hevanishedit)thanthefrequentintransitiveverbdisappear(e.g.,Idisappearedit).Brooksetal.suggestthatfrequencyofoccurrenceprovidesanimportantconstraintforchildren'sproductiveuseofverbsacrossconstructionsbecausefrequencydeterminestheactivationstrengthofthelinkbetweenverbsandconstructions,

Page 267: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 11 of 14

whichchildrenseemtointerpretas‘indirectnegativeevidence’fortheconstruction-specificuseofaparticularverbifthelinkisstronglyactivated,thatis,deeplyentrenchedinmemory(cf.Brooksetal.1999;Clark1987;Stefanowitsch2008b).

19.5.TheEmergenceofComplexSentenceConstructions

Havingoutlinedthedevelopmentofsimplesentences,Inowturntotheacquisitionofcomplexsentences(seeDiessel2004forareviewandnewdata;seealsoDiessel2007).Complexsentencesareparticularconstructionsconsistingoftwoormoreclauses.Theyarecommonlydividedintothreebasictypes:(1)sentencesincludingcomplementclauses,(2)sentencesincludingrelativeclauses,and(3)sentencesincludingadverbialclauses.

Thedevelopmentofcomplexsentencesoriginatesfromsimplenonembeddedsentencesthataregraduallytransformedintobiclausalconstructions.Twogeneraldevelopmentalpathwayscanbedistinguished:complexsentencesincludingcomplementclausesandrelativeclausesevolvefromsemanticallysimplesentencesthroughaprocessof‘clauseexpansion,’whereascomplexsentencesincludingadverbialclauses(orcoordinateclauses)arecommonlyderivedfromtwoindependentsentencesthatare‘integrated’intoaspecificbiclausalunit(cf.Diessel2004).

Theearlieststructuresthatonecouldanalyzeascomplexsentencesemergeataroundthesecondbirthday;theyoftenincludeanonfinitecomplementclausethatoccursinplaceofanominalcomplement.AcaseinpointisthedevelopmentofWantconstructions,whichwasconsideredbrieflyintheprevioussection.Wantisacomplement-takingverbthatcanoccurwithnominalandinfinitivalcomplements.Aspointedoutabove,theearliestusesofwantoccurwithanominalcomplement(e.g.,Iwannabag),butshortlythereafterchildrenbegintousewantalsowithsimpleinfinitives(e.g.,Iwannaswim),whicharegraduallyexpandedtocomplexinfinitivalconstructions(e.g.,Wantmeopenit?),whichsomechildrenoverextendtofullydevelopedcomplementclauses(e.g.,Dowanthewalk?).ThedevelopmentofWantconstructionsischaracteristicofthedevelopmentofnonfinitecomplementclauses.Itcommonlyoriginatesfromstructuresthatareonlylittledifferentfromsimplesentences,whicharegraduallyexpandedtobiclausalconstructions.

Thedevelopmentinvolvesbothsemanticandstructuralchanges.Semantically,itoriginatesfromstructuresdenotingasingleeventthatarereplacedbystructuresdenotingtworelatedevents,andstructurallythedevelopmentinvolvesaseriesofmorphosyntacticchangeswherebyunmarkedinfinitivesaregraduallyreplacedbymoreexplicittypesofnonfinitecomplementclausesincludingtheirownsemanticsubject(e.g.,Shemademecry)and/oraninfinitivemarkerorwh-pronoun(e.g.,Iknowhowtodrive).Paralleltothisdevelopment,thematrixclausesbecomeincreasinglymorecomplexandschematic.TheearliestnonfinitecomplementclausesoccurwithformulaicmainclausessuchasIwannaorIhafta,whichcanbeseenaspivotschemas;butaschildrengrowoldertheyproducematrixclauseswithothertypesofsubjectsandothertypesofverbsthataresemanticallymoreindependentoftheeventinthecomplementclausethanchildren'searlymatrixclauses,inwhichtheverboftenfunctionsasasemi-modal.Thefirstnonfinitecomplementclausesoccurwithverbssuchaswantto,stop,like,andtry,whichelaborateanaspectoftheinfinitiveorparticipleinthecomplementclause,whereaslatercomplement-takingpredicatesarecommonlyusedtodenoteindependentactivities(e.g.,Thedoctorsaidtostayinbedallday).Table19.3summarizesthechangesthatareinvolvedintheacquisitionofnonfinitecomplementclauses.

Likenonfinitecomplementclauses,finitecomplementclausesoriginatefromstructuresthatdenoteasinglesituation.TheearliestnonfinitecomplementclausesthatEnglish-speakingchildrenproduceincludeformulaicmainclausessuchasIthink__,(Do)youknow__,Iknow__,Howdoyouknow__,Iguess__,Remember__,andSeeif__,whichcanbeseenaspivotschemas(cf.DiesselandTomasello2001;Diessel2004:77–115).However,incontrasttotheconstantpartsofotherpivotschemas,theformulaicmainclausesofchildren'searlycomplementclausesarenonreferential.AmainclausesuchasIthinkorSeeifdoesnotdenoteamentalactivityoranactofperception,ratheritfunctionsasanepistemicmarkerormarkeroftheillocutionaryforcethatisattachedtoacomplementclause,whichisreallyanindependentassertion.Inaccordancewiththisanalysis,DiesselandTomasello(1999)observedthatchildren'searlycomplementclausesaregenerallyproducedwithoutacomplementizerandthatthematrixclausesareoftenplacedattheendofthesentence(e.g.,It'sacrazybone…Ithink).Boththeabsenceofacomplementizerandthefinaloccurrenceofthematrixclausereflectthefactthat

Page 268: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 12 of 14

children'searlycomplementclausesarenonembeddedassertions.However,aschildrengrowolder,theformulaicmainclausesaregraduallyreanalyzedasreferentialexpressionsdenotingatruementalstateoracommunicativeactivity.Thedevelopmentisreflectedinthegradualincreaseofinflectedverbformsandalargervarietyofsubjectsinthemainclause(e.g.,Paulthought),theoccurrenceofmainclausenegation(e.g.,Mommydon’tknow),andtheemergenceofspeechactverbssuchassayandtellwhichfromtheverybeginningareusedtodenoteanindependentstateofaffairs(e.g.,Thekittysayshewantstocomein).

Ingeneral,thedevelopmentofcomplexsentenceswithfiniteandnonfinitecomplementclausesexhibitssomestrikingparallelstothedevelopmentofchildren's

Table19.3.Thedevelopmentofnonfinitecomplementclauses(basedonDiessel2004:49–76)

Featuresofchildren'searlynonfiniteCOMP-clauses

Featuresofchildren'slaternonfiniteCOMP-clauses

Meaningofthesentence

S denotesasingleevent S denotestwoevents

Mainclause formulaic nonformulaic

Matrixverb semi-modal semanticallyindependentverb

Argumentstructure

NP-V-VP NP-V-NP-VP

Grammaticalmarking

bareinfinitive to-infinitive/wh-infinitive

earlyVerb-Argumentconstructions.Itoriginatesfromitem-specificconstructionsthatareorganizedaroundapivotexpression(i.e.,aformulaicmainclause)andanopenslotthatcanbefilledbyaninfinitiveorafullydevelopedclause.Aschildrengrowolder,theseconstructionsbecomeincreasinglymorecomplexanddiverse,resultinginbiclausalschemasinwhichboththemainclauseandthesubordinateclausedenotesomestateofaffairs.

Likecomplementclauses,relativeclausesdevelopfromsemanticallysimplesentencesthataregraduallyexpandedtobiclausalconstructions(Diessel2004:116–48;seealsoDiesselandTomasello2000andDiessel2009).TheearliestrelativeclausesEnglish-speakingchildrenproduceoccurintwocontexts.Eithertheyareattachedtoanisolatedheadnounthatisusedtoanswerapreviousquestion(44)ortheymodifythepredicatenominalofacopularclausethatfunctionstodrawthehearer'sattentiontoanobjectinthesurroundingsituation(45).

(44)Adult:Nowhatdidyoueat?Child:Someapplesthatweresweet.(Abe3;6)(45)Child:Here'shisboxthathe'sgonnagoin.(Nina3;0)

Bothtypesofconstructionsdenoteasinglestateofaffairsthatisexpressedintherelativeclause,whereasthemainclauseservestoestablishareferentintopicorfocuspositionwithoutdenotinganindependentstateofaffairs.Withage,childrenbegintoproducemorecomplexrelativeconstructionsinwhichthemainclausesaregraduallyexpandedtofullydevelopedclausesdenotinganeventthatissemanticallyindependentoftheeventintherelativeclause(e.g.,OnceIsawapersonthatshotafirearrow).

IfwedisregardthetopicalizedNPsandcopularclausesthatprecedechildren'searlyrelativeclauses,thesesentenceshavethesamestructureassimpleSV(O)clauses.TheearliestrelativeclausesthatEnglish-speakingchildrenproducearesubject-relatives,whichinvolvethesamesequenceofgrammaticalrolesas(in)transitiveclauses,thatis,agent-verb-(patient).Otherstructuraltypesofrelativeclausesinwhichtheagent(orsubject)followsanotherreferent(e.g.,ThepictureImade)tendtooccurlaterandcausecomprehensiondifficultiesin

complex complex

Page 269: The Oxford Handbook of Construction Grammar

Construction Grammar and First Language Acquisition

Page 13 of 14

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). (c) Oxford University Press, 2014. All RightsReserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of a title in OxfordHandbooks Online for personal use (for details see

experiments,whichiscommonlyexplainedbythefactthatthesestructuresdeviatefromtheSV(O)wordorderofsimple(in)transitiveclauses(cf.Bever1970;SlobinandBever1982),suggestingthatwhatchildreneventuallylearnisanetworkofrelatedstructuralpatternswithoverlappingsemanticandsyntacticfeatures(cf.DiesselandTomasello2005a;Diessel2009).

Finally,thedevelopmentofcomplexsentencesincludingadverbialandcoordinateclausestakesadifferentpath.Incontrasttocomplementandrelativeclauses,whicharecommonlyderivedfromsimplesentencesthroughclauseexpansion,adverbialandcoordinateclausesarederivedfromindependentsentencesthatarelinkedtoaprecedingsentencebyaconjunction.TheearliestadverbialclausesthatEnglish-speakingchildrenproduceareintonationallyunboundsentences,theyalwaysfollowthesemanticallyassociatedclause,andfunctionasindependentspeechactsthatchildrenlearninthecontextofparticularcommunicativesituations.Forinstance,causalbecause-clausesareatfirstcommonlyusedinresponsetoacausalquestion(46),andbut-clausesareinitiallyalwaysusedbythechildtoobjecttoaprioradultutterance(47).

(46)Child:Youcan’thavethis.Adult:Why?Child:CauseI’musingit.(47)Adult:Itiscalledtheskinofthepeanut.Child:Butthisisn’ttheskin.

Aschildrengrowolder,theycombineadverbialandcoordinateclauseswiththesemanticallyassociated(main)clauseunderasingleintonationcontourandbegintoplaceadverbialclausesbeforethemainclause,whichpresupposestheabilitytoanticipatethelinkbetweentwoclausesbeforethefirstclausehasbeenproduced.Whilefinaladverbialclausescanbeplannedafterthecompletionoftheprecedingmainclause,aninitialadverbialclausecanonlybeproducedifthespeakerisabletoplantheoccurrenceoftwoconjoinedclauses(cf.Diessel2005).Thedevelopmentcanbecharacterizedasaprocessofclauseintegrationwherebytwoindependentsentencesarecombinedto(orintegratedinto)aparticularbiclausalunit.

Inconclusion,thischapterhassurveyedconstruction-basedresearchonfirstlanguageacquisition.Ithasbeenarguedthatgrammaticaldevelopmentstartswithlexicallyspecificformulasthatchildrengraduallydecomposeandelaboratetomorecomplexandschematicunits.Thedevelopmentisdrivenbygenerallearningmechanismssuchasanalogyandcategorizationthatarenotrestrictedtolanguageacquisition.Whatchildreneventuallylearnisanetworkofconstructionsthatisimmediatelygroundedintheirlinguisticexperience.

Notes:

(1.)Thenumbersindicatethechild'sagewhenaparticularquestionfirstappearedinthecorpus,e.g.1;11.11=oneyear,elevenmonths,andelevendays.

HolgerDiesselHolgerDiesselisProfessorforEnglishLinguisticsattheUniversityofJena.AfterhisPh.D.fromtheStateUniversityofNewYorkatBuffaloin1998,heworkedasajuniorresearcherattheMax-Planck-InstituteforEvolutionaryAnthropologyinLeipziginthedepartmentofcomparativeanddevelopmentalpsychologyuntilhewasofferedtheprofessorshipattheUniversityofJena.Heisanexpertonlanguageacquisitionandhismainresearchinterestisintherelationshipbetweentheuseandstructureofcomplexsentencesandthepsychologicalandlinguisticfoundationsforausage-basedtheoryofgrammar.HeisontheeditorialboardofCognitiveLinguisticsandCorpusLinguisticsandLinguisticTheoryandaconsultingeditorofStudiesinLanguage.

Page 270: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 1 of 9

PrintPublicationDate: Feb2013 Subject: Linguistics,LanguageAcquisition,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0020

ConstructionGrammarandSecondLanguageAcquisitionNickEllisTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapteranalyzessecondlanguageacquisitioninthecontextofConstructionGrammar.Itprovidesevidenceforthepsychologicalrealityofconstructionsinasecondlanguageandpresentsapsychologicalanalysisoftheeffectsofform,function,frequency,andcontingencythatarecommontobothfirstandsecondlanguageconstructionlearningfollowingstatisticallearningprocesseswhichrelateinputandlearnercognition.Thechapteralsoconsiderscrosslinguistictransfereffectsandpossiblefuturedirectionsforresearchintoconstructionalapproachestosecondlanguageacquisition.

Keywords:secondlanguageacquisition,ConstructionGrammar,statisticallearningprocesses,learnercognition,crosslinguistictransfer,constructionalapproaches

20.1.Introduction

Usage-basedapproachesholdthatwelearnlinguisticconstructionswhileengagingincommunication,the“interpersonalcommunicativeandcognitiveprocessesthateverywhereandalwaysshapelanguage”(Slobin1997:267).Constructionsareform–meaningmappings,conventionalizedinthespeechcommunity,andentrenchedaslanguageknowledgeinthelearner'smind.Theyarethesymbolicunitsoflanguagerelatingthedefiningpropertiesoftheirmorphological,syntactic,andlexicalformwithparticularsemantic,pragmatic,anddiscoursefunctions(BatesandMacWhinney1987;Lakoff1987;Langacker1987;Goldberg1995,2003,2006a;Croft2001;CroftandCruse2004;Tomasello2003;RobinsonandEllis2008a,b;Bybee2008).Broadly,ConstructionGrammararguesthatallgrammaticalphenomenacanbeunderstoodaslearnedpairingsofform(frommorphemes,words,andidioms,topartiallylexicallyfilledandfullygeneralphrasalpatterns)andtheirassociatedsemanticordiscoursefunctions.Suchbeliefs,increasinglyinfluentialinthestudyofchildlanguageacquisition,haveturnedupsidedowngenerativeassumptionsofinnatelanguageacquisitiondevices,thecontinuityhypothesis,andtop-down,rule-governed,processing,bringingbackdata-driven,emergentaccountsoflinguisticsystematicities.Constructionisttheoriesofchildfirstlanguageacquisition(L1A)usedenselongitudinalcorporatocharttheemergenceofcreativelinguisticcompetencefromchildren'sanalysesoftheutterancesintheirusagehistoryandfromtheirabstractionofregularitieswithinthem(Goldberg1995,2006a,2003;Tomasello1998b;Tomasello2003;Diessel,thisvolume).

Secondlanguage(L2)learnerssharethegoalofunderstandinglanguageandhowitworks.Sincetheyachievethisbasedupontheirexperienceoflanguageusage,therearemanycommonalitiesbetweenfirstandsecondlanguageacquisitionthatcanbeunderstoodfromcorpusanalysesofinputandfromcognitiveandpsycholinguisticanalysesofconstructionacquisitionfollowingassociativeandcognitiveprinciplesoflearningandcategorization.Thereforeusage-basedapproaches,CognitiveLinguistics,andCorpusLinguisticsareincreasinglyinfluentialinsecondlanguageacquisition(L2A)researchtoo(Ellis1998,2003;RobinsonandEllis2008a,b;EllisandCadierno2009;CollinsandEllis2009),albeitwiththetwistthatsincetheyhavepreviouslydevoted

Page 271: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 2 of 9

considerableresourcestotheestimationofthecharacteristicsofanotherlanguage—thenativetongueinwhichtheyhaveconsiderablefluency—L2learners’computationsandinductionsareoftenaffectedbytransfer,withL1-tunedexpectationsandselectiveattention(Ellis2006c;EllisandSagarra2010a)blindingtheacquisitionsystemtoaspectsoftheL2sample,thusbiasingtheirestimationfromnaturalisticusageandproducingthelimitedattainmentthatistypicalofadultL2A.Thus,L2AisdifferentfromL1Ainthatitinvolvesprocessesofconstructionandreconstruction.

Theorganizationoftheremainderofchapterisasfollows.Section20.2providesevidenceforthepsychologicalrealityofconstructionsinL2.Section20.3presentsapsychologicalanalysisoftheeffectsofform,function,frequency,andcontingencythatarecommontobothL1andL2constructionlearningfollowingstatisticallearningprocesseswhichrelateinputandlearnercognition.ItillustrateseachpointwithempiricaldemonstrationsoftheseeffectsseparatelyforL1andforL2.Section20.4considersL1⇒L2transferasitaffectstheconceptualunderpinningsofconstructionsandtheirunderstanding.Section20.5considersL1⇒L2transferor‘learnedattention’andhowthisaffectslearners’sensitivitytodifferentaspectsofthelinguisticformofconstructions.Finally,section20.6presentssomeprioritiesforfutureresearch.

20.2.L2Constructions

Demonstrationsofthepsychologicalrealityofconstructionsinnativespeakers’language(e.g.,Goldberg,Casenhiser,andSethuraman2004;Pickering2006)promptedresearchinvestigatingwhetherconstructionsalsounderpinsecondlanguagelearners’linguisticcompetence,andhowL2learnersimplicitly‘tally’(Ellis2002)andtunetheirconstructionalknowledgetoconstruction-specificpreferencesintermsofthewordsthatpreferablyoccurinthoseconstructions.Forexample,GriesandWulff(2005)showed(1)thatadvancedL2learnersofEnglishwhowerenativespeakersofGermanshowedsyntacticprimingforditransitive(e.g.,Theracingdrivershowedthehelpfulmechanic…)andprepositionaldative(e.g.,Theracingdrivershowedthetornoverall…)ArgumentStructureconstructionsinanEnglishsentencecompletiontask,(2)thattheirsemanticknowledgeofArgumentStructureconstructionsaffectedtheirgroupingofsentencesinasortingtask,and(3)thattheirprimingeffectscloselyresembledthoseofnativespeakersofEnglishinthattheywerehighlycorrelatedwithnativespeakers’verbalsubcategorizationpreferenceswhileuncorrelatedwiththesubcategorizationpreferencesoftheGermantranslationequivalentsoftheseverbs.ThereisnowagrowingbodyofresearchdemonstratingsuchL2syntacticprimingeffects(McDonough2006;McDonoughandMackey2006;McDonoughandTrofimovich2008;GriesandWulff2009).

ThisrecentresearchwithinaCognitiveLinguisticsframeworkechoessomeoftheearliestworkonL2AwithintheStructuralisttradition.CharlesFries,thefounderoftheEnglishLanguageInstituteattheUniversityofMichigan,distinguishedbetweenlexicalandstructuralmeaning,withstructuralmeaningconcerningthepatternsrelatingaparticulararrangementofformclassestoparticularstructuralmeanings.Inhisview,languageacquisitionisthelearningofaninventoryofpatternsasarrangementsofwordswiththeirassociatedstructuralmeanings.Fries's(1952)StructureofEnglishpresentedananalysisofthesepatterns,Roberts's(1956)PatternsofEnglishwasatextbookpresentationofFries'ssystemforclassroomuse,andEnglishPatternPractices:EstablishingthePatternsasHabits(Fries,Lado,andtheStaffoftheMichiganEnglishLanguageInstitute1958)taughtbeginningandintermediateEFLstudentsEnglishaspatternsusingaudiolingualdrills.

SecondLanguageAcquisition(SLA)theoryhascontinuedtorecognizetheimportanceofphraseologysince:asholophrases(Corder1973),prefabricatedroutinesandpatterns(Hakuta1974),formulaicspeech(Wong-Fillmore1976),memorizedsentencesandlexicalizedstems(PawleyandSyder1983),lexicalphrases(Nattinger1980),formulas(R..Ellis1994;McLaughlin1995),chunks(Ellis1996),andconstructions(Ellis2003,2006a).

EverygenreofEnglishforAcademicPurposesandEnglishforSpecialPurposeshasitsownphraseology,andlearningtobeeffectiveinthegenreinvolveslearningthis(Swales1990).Lexicographersdeveloptheirlearnerdictionariesuponlargecorpora(HunstonandFrancis1996;Ooi1998)anddictionariesfocusuponexamplesofusageasmuchasdefinitions,orevenmoreso.NattingerandDeCarrico(1992)argueforthe‘lexicalphrase’asthepedagogicallyapplicableunitofprefabricatedlanguage,“foragreatdealofthetimeanyway,languageproductionconsistsofpiecingtogethertheready-madeunitsappropriateforaparticularsituationand…comprehensionreliesonknowingwhichofthesepatternstopredictinthesesituations.Ourteachingtherefore

Page 272: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 3 of 9

wouldcenteronthesepatternsandthewaystheycanbepiecedtogether,alongwiththewaystheyvaryandthesituationsinwhichtheyoccur”(Nattinger1980:341).TheLexicalApproach(Lewis1993),similarlypredicatedupontheidiomprinciple(Sinclair1991),focusesinstructiononrelativelyfixedexpressionsthatoccurfrequentlyinspokenlanguage.Corporanowplaycentralrolesinlanguageteaching(Sinclair1996a;Cobb2007;Römer2008).Therehasneverbeenmoreinterestinsecondlanguagephraseology,asrecentreviewsinappliedlinguistics(Cowie2001;Wray2002a;Schmitt2004;GrangerandMeunier2008)andcognitivelinguistics(RobinsonandEllis2008a,b)attest.

20.3.Form,Function,andFrequencyinL1andL2LearningofConstructions

Iftheunitsoflanguageareconstructions,thenlanguageacquisitionisthelearningofconstructions.SoL2Adependsuponlearners’experienceoflanguageusageanduponwhattheycanmakeofit.Psychologicalanalysesofthelearningofconstructionsasform-meaningpairsisinformedbytheliteratureontheassociativelearningofcue-outcomecontingencieswheretheusualdeterminantsinclude:factorsrelatingtotheformsuchasfrequencyandsalience;factorsrelatingtotheinterpretationsuchassignificanceinthecomprehensionoftheoverallutterance,prototypicality,generality,redundancy,andsurprisevalue;factorsrelatingtothecontingencyofformandfunction;andfactorsrelatingtolearnerattention,suchasautomaticity,transfer,overshadowing,andblocking(Ellis2002,2003,2006c,2008a,b).Thesevariouspsycholinguisticfactorsconspireintheacquisitionanduseofanylinguisticconstruction.Constructionistaccountsoflanguageacquisitionthusinvolvethedistributionalanalysisofthelanguagestreamandtheparallelanalysisofcontingentperceptualactivity,withabstractconstructionsbeinglearnedfromtheconspiracyofconcreteexemplarsofusagefollowingstatisticallearningmechanisms(ChristiansenandChater2001)relatinginputandlearnercognition.

Thedeterminantsoflearninginclude(1)inputfrequency(type-tokenfrequency,Zipfiandistribution,recency),(2)form(salienceandperception),(3)function(prototypicalityofmeaning,importanceofformformessagecomprehension,redundancy),and(4)interactionsbetweenthese(contingencyofform-functionmapping).Iconsidereachinturn.

20.3.1InputFrequency

20.3.1.1ConstructionFrequencyFrequencyofexposurepromoteslearningandentrenchment—frequentlyexperiencedconstructionsareprocessedwithgreaterfacility.Psycholinguisticresearchshowshowlanguageprocessingisintimatelytunedtoinputfrequencyatalllevelsofgrain:inputfrequencyaffectstheprocessingofphonologyandphonotactics,reading,spelling,lexis,morphosyntax,formulaiclanguage,languagecomprehension,grammaticality,sentenceproduction,andsyntax(Ellis2002).Thatlanguageusersaresensitivetotheinputfrequenciesofthesepatternsentailsthattheymusthaveregisteredtheiroccurrenceinprocessing.Thesefrequencyeffectsarethuscompellingevidenceforusage-basedmodelsoflanguageacquisition,whichemphasizetheroleofinput.

20.3.1.2TypeandTokenFrequencyTokenfrequencycountshowoftenaparticularformappearsintheinput.Typefrequency,ontheotherhand,referstothenumberofdistinctlexicalitemsthatcanbesubstitutedinagivenslotinaconstruction,whetheritisaword-levelconstructionforinflectionorasyntacticconstructionspecifyingtherelationamongwords.Forexample,the‘regular’Englishpasttense-edhasaveryhightypefrequencybecauseitappliestothousandsofdifferenttypesofverbs,whereasthevowelchangeexemplifiedinswamandranghasmuchlowertypefrequency.Theproductivityofphonological,morphological,andsyntacticpatternsisafunctionoftyperatherthantokenfrequency(BybeeandHopper2001).Itissobecause:(a)themorelexicalitemsthatareheardinacertainpositioninaconstruction,thelesslikelyitisthattheconstructionisassociatedwithaparticularlexicalitemandthemorelikelyitisthatageneralcategoryisformedovertheitemsthatoccurinthatposition;(b)themoreitemsthecategorymustcover,themoregeneralareitscriterialfeaturesandthemorelikelyitistoextendtonewitems;and(c)hightypefrequencyensuresthataconstructionisusedfrequently,thusstrengtheningitsrepresentationalschemaandmakingitmoreaccessibleforfurtherusewithnewitems(BybeeandThompson2000).Incontrast,hightokenfrequencypromotestheentrenchmentorconservationofirregularformsandidioms;theirregularforms

Page 273: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 4 of 9

onlysurvivebecausetheyarehighfrequency.Thesefindingssupportlanguage'splaceatthecenterofcognitiveresearchintohumancategorization,whichalsoemphasizestheimportanceoftypefrequencyinclassification.

20.3.1.3ZipfianDistributionIntheearlystagesoflearningcategoriesfromexemplars,acquisitionisoptimizedbytheintroductionofaninitial,low-variancesamplecentereduponprototypicalexemplars(ElioandAnderson1981,1984).Thislowvariancesampleallowslearnerstogetafixonwhatwillaccountformostofthecategorymembers.Theboundsofthecategoryaredefinedlaterbyexperienceofthefullbreadthofexemplartypes.Goldberg,Casenhiser,andSethuraman(2004)demonstratedthatinsamplesofchildlanguageacquisition,foravarietyofVerb-Argumentconstructions(VACs:VerbLocativeconstruction(VL),VerbObjectLocativeconstruction(VOL),andtheVerbObjectObjectDitransitiveconstruction(VOO)),thereisastrongtendencyforonesingleverbtooccurwithveryhighfrequencyincomparisontootherverbsused,aprofilethatcloselymirrorsthatofthemothers’speechtothesechildren.Innaturallanguage,Zipf'slaw(Zipf1935)describeshowthehighestfrequencywordsaccountforadisproportionatelyhighamountoflinguistictokens—themostfrequentwordoccursapproximatelytwiceasoftenasthesecondmostfrequentword,threetimesasoftenasthethirdmostfrequentword,andsoon.Thusthe,themostfrequentlyoccurringword,byitselfaccountsfornearly7%ofallwordoccurrences.Goldbergetal.(2004)showthatZipf'slawapplieswithintheseVACstoo,andtheyarguethatthispromotesacquisition:tokensofoneparticularverbaccountforthelion'sshareofinstancesofeachparticularargumentframe;thispathbreakingverbalsoistheonewiththeprototypicalmeaningfromwhichtheconstructionisderived(seealsoNinio1999,2006).

EllisandFerreira-Junior(2009a,2009b)investigateeffectsoftype/tokendistributionsintheislandscomprisingthelinguisticformofthesameEnglishVerb-ArgumentconstructionsintheEuropeanScienceFoundation(ESF)corpusofnaturalisticsecondlanguageacquisition(Perdue1993).TheyshowthatinthenaturalisticL2AofEnglish,VACverbtype/tokendistributionintheinputisZipfianandlearnersfirstacquirethemostfrequent,prototypical,andgenericexemplaroftheverbisland(Tomasello1992)(e.g.,putinVOL,giveinVOO,etc.).Theirworkfurtherillustrateshow,intheacquisitionof,forexample,theCausedMotionconstruction(XcausesYtomoveZ[SubjVObjObl ]),thewholeframeasanarchipelagoofislandsisimportant.TheSubjislandhelpstoidentifythebeginningboundsoftheparse.Morefrequent,moregeneric,andmoreprototypicaloccupantsaremoreeasilyidentified.Pronouns,particularlythosethatrefertoanimateentities,readilyactivatetheschema(seelikewiseforL1;ChildersandTomasello2001).TheObjislandtooismorereadilyidentifiedwhenoccupiedbymorefrequent,moregeneric,andmoreprototypicallexicalitems(pronounslikeit,requiredbydiscourseconstraints,ratherthannounssuchasnapkin).So,too,thelocativeisactivatedmorereadilyifopenedbyaprepositionalislandpopulatedbyahighfrequency,prototypicalexemplarsuchasonorin(seelikewiseforL1;Tomasello2003:153).ActivationoftheVACschemaarisesfromtheconspiracyofallofthesefeatures,andargumentsaboutZipfiantype/tokendistributionsandprototypicalityofmembershipextendtoalloftheislandsoftheconstruction.EllisandLarsen-Freeman(2009a,b)describecomputational(Emergentconnectionist)serial-recurrentnetworkmodelsofthesevariousfactorsastheyplayoutinsyntacticandsemanticbootstrappingandtheemergenceofconstructionsasgeneralizedlinguisticschemafromtheirfrequencydistributionsintheinput.

20.3.1.4RecencyCognitivepsychologicalresearchshowsthatthreekeyfactorsdeterminetheactivationofmemoryschemata—frequency,recency,andcontext(Anderson1989;AndersonandSchooler2000).Languageprocessingalsoreflectsrecencyeffects.Thisphenomenonisknownasprimingandmaybeobservedinphonology,conceptualrepresentations,lexicalchoice,andsyntax(McDonoughandTrofimovich2008).Syntacticprimingreferstothephenomenonofpreferentiallyusingorprocessingaparticularsyntacticstructuregivenpriorexposuretothesamestructure.Thisbehavioroccursinhearing,speaking,reading,orwriting.

Section20.2introducedearlyresearchintoL2syntacticprimingeffects(GriesandWulff2005;McDonough2006;McDonoughandMackey2006;McDonoughandTrofimovich2008).AmorerecentdemonstrationisthatofGriesandWulff(2009),whofocusedonwhetherEnglishgerundandinfinitivalcomplementconstructionsarestoredassymbolicunitsbyGermanlanguagelearnersofEnglish.AcorpusanalysisoftheseconstructionsintheInternationalCorpusofEnglishidentifiedtheverbsdistinguishingbestbetweenthetwoconstructions,andthesewerethenusedasexperimentalstimuliinsentencecompletionandsentenceacceptabilityratingexperiments.GriesandWulffinvestigatedtwokindsofshort-distanceprimingeffects:howoftensubjectsproduceaning-/to-

path/loc

path/loc

Page 274: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 5 of 9

/‘other’constructionafterratinganing-orto-construction,andhowoftentheyproduceaning-/to-/‘other’constructionafterproducinganing-orto-constructioninthedirectlyprecedingcompletion,aswellasameasureoflongertermwithin-subjectaccumulativepriming.Boththegerundandinfinitivalcomplementspatternsexhibitedverb-specificconstructionalpreferencesandprimingeffects,confirmingtheirstatusasconstructions.

20.3.2Form(SalienceandPerception)

Thegeneralperceivedstrengthofstimuliiscommonlyreferredtoastheirsalience.Lowsaliencecuestendtobelessreadilylearned.Ellis(2006b,2006c)summarizedassociativelearningresearchdemonstratingthatselectiveattention,salience,expectation,andsurprisearekeyelementsintheanalysisofalllearning,animalandhumanalike.AstheRescorla-Wagner(1972)modelofassociativelearningencapsulates,theamountoflearninginducedfromanexperienceofacue-outcomeassociationdependscruciallyuponthesalienceofthecueandtheimportanceoftheoutcome.

Manygrammaticalmeaning-formrelationships,particularlythosethatarenotoriouslydifficultforsecondlanguagelearnerslikegrammaticalparticlesandinflectionssuchasthethirdpersonsingular-sofEnglish,areoflowsalienceinthelanguagestream.Forexample,someformsaremoresalient:‘today’isastrongerpsychophysicalformintheinputthanisthemorpheme‘-s’marking3rdpersonsingularpresenttense,thuswhilebothprovidecuestopresenttime,todayismuchmorelikelytobeperceived,andscanthusbecomeovershadowedandblocked,makingitdifficultforsecondlanguagelearnersofEnglishtoacquire(GoldschneiderandDeKeyser2001;Ellis2006b,2008b;EllisandSagarra,2010b,2011).

20.3.3Function

20.3.3.1PrototypicalityofMeaningCategorieshavegradedstructure,withsomemembersbeingbetterexemplarsthanothers.Intheprototypetheoryofconcepts(RoschandMervis1975b;Roschetal.1976),theprototypeasanidealizedcentraldescriptionisthebestexampleofthecategory,appropriatelysummarizingthemostrepresentativeattributesofacategory.Asthetypicalinstanceofacategory,itservesasthebenchmarkagainstwhichsurrounding,lessrepresentativeinstancesareclassified—peoplemorequicklyclassifyasbirdssparrows(orotheraveragesized,averagecolored,averagebeaked,averagefeaturedspecimens)thantheydobirdswithlesscommonfeaturesorfeaturecombinationslikegeeseoralbatrosses(RoschandMervis1975b;Roschetal.1976).Prototypesarejudgedfasterandmoreaccurately,eveniftheythemselveshaveneverbeenseenbefore—someonewhohasneverseenasparrow,yetwhohasexperiencedtherestoftherunoftheavianmill,willstillbefastandaccurateinjudgingittobeabird(PosnerandKeele1970).Thegreaterthetokenfrequencyofanexemplar,themoreitcontributestodefiningthecategory,andthegreaterthelikelihooditwillbeconsideredtheprototype.Thebestwaytoteachaconceptistoshowanexampleofit.Sothebestwaytointroduceacategoryistoshowaprototypicalexample.EllisandFerreira-Junior(2009a)showthattheverbsthatsecondlanguagelearnersfirstusedinparticularVACsareprototypicalandgenericinfunction(goforVL,putforVOL,andgiveforVOO).Thesamehasbeenshownforchildlanguageacquisition,whereasmallgroupofsemanticallygeneralverbs,oftenreferredtoaslightverbs(e.g.,go,do,make,come)arelearnedearly(Clark1978;Pinker1989;Ninio1999).Ninio(1999)arguesthatbecausemostoftheirsemanticsconsistofsomeschematicnotionoftransitivitywiththeadditionofaminimumspecificelement,theyaresemanticallysuitable,salient,andfrequent;hence,learnersstarttransitivewordcombinationswiththesegenericverbs.Thereafter,asClarkdescribes,“manyusesoftheseverbsarereplaced,aschildrengetolder,bymorespecificterms….Generalpurposeverbs,ofcourse,continuetobeusedbutbecomeproportionatelylessfrequentaschildrenacquiremorewordsforspecificcategoriesofactions”(Clark1978:53).

3.3.2RedundancyTheRescorla-Wagnermodel(1972)alsosummarizeshowredundantcuestendnottobeacquired.Notonlyaremanygrammaticalmeaning-formrelationshipslowinsalience,buttheycanalsoberedundantintheunderstandingofthemeaningofanutterance.Forexample,itisoftenunnecessarytointerpretinflectionsmarkinggrammaticalmeaningssuchastensebecausetheyareusuallyaccompaniedbyadverbsthatindicatethetemporalreference.Secondlanguagelearners’relianceuponadverbialoverinflectionalcuestotensehasbeenextensivelydocumentedinlongitudinalstudiesofnaturalisticacquisition(Dietrich,Klein,andNoyau1995;Bardovi-Harlig

Page 275: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 6 of 9

2000),trainingexperiments(EllisandSagarra2010b,2011,andstudiesofL2languageprocessing(VanPatten2006;EllisandSagarra2010a).

20.3.4InteractionsBetweenThese(ContingencyofForm-FunctionMapping)

Psychologicalresearchintoassociativelearninghaslongrecognizedthatwhilefrequencyofformisimportant,sotooiscontingencyofmapping(Shanks1995).Considerhow,inthelearningofthecategoryofbirds,whileeyesandwingsareequallyfrequentlyexperiencedfeaturesintheexemplars,itiswingswhicharedistinctiveindifferentiatingbirdsfromotheranimals.Wingsareimportantfeaturesforlearningthecategoryofbirdsbecausetheyarereliablyassociatedwithclassmembership,whileeyesarenot.Rawfrequencyofoccurrenceislessimportantthanthecontingencybetweencueandinterpretation.Distinctivenessorreliabilityofform-functionmappingisadrivingforceofallassociativelearning,tothedegreethatthefieldofitsstudyhasbeenknownas‘contingencylearning’sinceRescorla(1968)showedthatforclassicalconditioning,ifoneremovedthecontingencybetweentheconditionedstimulus(CS)andtheunconditionedstimulus(UCS),preservingthetemporalpairingbetweenCSandUCSbutaddingadditionaltrialswheretheUCSappearedonitsown,thenanimalsdidnotdevelopaconditionedresponsetotheCS.Thisresultwasamilestoneinthedevelopmentoflearningtheorybecauseitimpliedthatitwascontingency,nottemporalpairing,thatgeneratedconditionedresponding.Contingency,anditsassociatedaspectsofpredictivevalue,informationgain,andstatisticalassociation,havebeenatthecoreoflearningtheoryeversince.Itiscentralinpsycholinguistictheoriesoflanguageacquisitiontoo(MacWhinney1987a,b;GriesandWulff2005;Ellis2006b,c,2008a,b;Gries,thisvolume),withthemostdevelopedaccountforsecondlanguageacquisitionbeingthatoftheCompetitionmodel(MacWhinney1987a,b,1997,2001).EllisandFerreira-Junior(2009b)useΔPandcollostructionalanalysismeasures(StefanowitschandGries2003;GriesandStefanowitsch2004a;Stefanowitsch,thisvolume)todemonstrateeffectsofform-functioncontingencyuponL2VACacquisition.Wulff,Ellis,Römer,Bardovi-Harlig,andLeBlanc(2009)usemultipledistinctivecollexemeanalysistoshoweffectsofreliabilityofform-functionmappinginthesecondlanguageacquisitionoftenseandaspect.BoydandGoldberg(2009)useconditionalprobabilitiestoanalyzecontingencyeffectsinVACacquisition.Thisisstillanactiveareaofinquiry,andmoreresearchisrequiredbeforeweknowwhichstatisticalmeasuresofform-functioncontingencyaremorepredictiveofacquisitionandprocessing.

20.3.5ConclusionsonConstructionAcquisition

Arangeoffactorsthusinfluencetheacquisitionoflinguisticconstructions,whetherinL1orL2:

a.thefrequency,thefrequencydistribution,andthesalienceoftheformtypes;b.thefrequency,thefrequencydistribution,theprototypicalityandgeneralityofthesemantictypes,theirimportanceininterpretingtheoverallconstruction;c.thereliabilitiesofthemappingbetweenaandb;d.thedegreetowhichthedifferentelementsintheislandsofaconstructionaremutuallyinformativeandformpredictablechunks.

20.4.ReconstructingMeaninginL2—CrosslinguisticTransfer

CognitiveLinguistics(Langacker1987,2000b;Taylor2002;CroftandCruse2004;RobinsonandEllis2008a,b)providesdetailedqualitativeanalysesofthewaysinwhichlanguageisgroundedinourexperienceandourphysicalembodimentwhichrepresentstheworldinaveryparticularway.Themeaningofthewordsofagivenlanguage,andhowtheycanbeusedincombination,dependsontheperceptionandcategorizationoftherealworldaroundus.Sinceweconstantlyobserveandplayanactiveroleinthisworld,weknowagreatdealabouttheentitiesofwhichitconsists.Thisexperienceandfamiliarityisreflectedinthenatureoflanguage.Ultimately,everythingweknowisorganizedandrelatedtoourotherknowledgeinsomemeaningfulway,andeverythingweperceiveisaffectedbyourperceptualapparatusandourperceptualhistory.

Languagereflectsthisembodimentandthisexperience.Consider,forexample,themeaningsofverbslikepush,poke,pull,hold,andsoon,andsimilarwordsfromotherlanguages.Theoreticalunderstandingofthedifferencesbetweenthesewordscannotbeforthcomingwithoutinclusionofamodelofhigh-levelmotorcontrol—handposture,jointmotions,force,aspect,andgoalsareallrelevanttotheselinguisticdistinctions(Bailey1997;Lakoff

Page 276: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 7 of 9

andJohnson1999;BergenandChang2005,thisvolume;Feldman2006).Thesesensori-motorfeaturesarepartofourembodiment,theystructureourconcepts,theyplayoutintime.Thus,CognitiveLinguisticsemphasizeshowlanguageislearnedfromparticipatoryexperienceofprocessinglanguageduringembodiedinteractioninsocialcontextswhereindividuallydesirednonlinguisticoutcomesaregoalstobeachievedbycommunicatingintentions,concepts,andmeaningwithothers.AnunderstandingofparticipationinsituatedactionisthusessentialtotheunderstandingofmeaningandtheacquisitionoflinguisticconstructionsinL1andL2.

Considertoothemeaningsofspatiallanguage.Thesearenotthesimplesumthatresultsfromadditionoffixedmeaningsgivenbyprepositionsfor‘where’anobjectis,tothemeaningsofotherelementsinthesentencedescribing‘what’isbeinglocated.Spatiallanguageunderstandingisfirmlygroundedinthevisualprocessingsystemasitrelatestomotoraction(RegierandCarlson2002;CoventryandGarrod2004),themultipleconstraintsrelatingtoobjectknowledge,dynamic-kinematicroutines,andfunctionalgeometricanalyses.Meaningsareembodiedanddynamic(Elman2004;Spivey2006;McRaeetal.2006);theyareflexiblyconstructedon-line.MeaningslikethiscannotsimplybetaughtbyL2rulesandlearnedbyrote;optimallytheyarelearnedinsituatedaction.

Constructionsareconventionalizedlinguisticmeansforpresentingdifferentinterpretationsorconstrualsofanevent.Theystructureconceptsandwindowattentiontoaspectsofexperiencethroughtheoptionsspecificlanguagesmakeavailabletospeakers(Talmy2000).Thedifferentdegreesofsalienceorprominenceofelementsinvolvedinsituationsthatwewishtodescribeaffecttheselectionofsubject,object,adverbials,andotherclausearrangements.Inlanguagecomprehension,abstractlinguisticconstructions(likesimplelocatives,datives,andpassives)serveasa‘zoomlens’forthelistener,guidinghisorherattentiontoaparticularperspectiveonascenewhilebackgroundingotheraspects(Langacker1987,1999;Croft2001;Taylor2002;CroftandCruse2004).Languagehasanextensivesystemthatassignsdifferentdegreesofsaliencetothepartsofanexpression,reference,orcontext.Talmy(2000)analyzeshowtheAttentionalSystemofLanguageincludessomefiftybasicfactors,its‘buildingblocks.’Eachfactorinvolvesaparticularlinguisticmechanismthatincreasesordecreasesattentiononacertaintypeoflinguisticentity.Learningalanguageinvolvesthelearningofthesevariousattention-directingmechanismsoflanguage,andthis,inturn,restsuponL1learners’developingattentionalsystemsandL2learners’attentionalbiases.

Languagesleadtheirspeakerstoexperiencedifferent‘thinkingforspeaking’andthustoconstrueexperienceindifferentways(Slobin1996).Crosslinguisticresearchshowshowdifferentlanguagesleadspeakerstoprioritizedifferentaspectsofeventsinnarrativediscourse(BermanandSlobin1994).Becauselanguagesachievetheseattention-directingoutcomesindifferentways,learninganotherlanguageinvolveslearninghowtoconstruetheworldlikenativesoftheL2,thatis,learningalternativewaysofthinkingforspeaking(Cadierno2008;BrownandGullberg2008;BrownandGullberg2010)orlearningto‘rethinkforspeaking’(RobinsonandEllis2008a,b).TransfertheoriessuchastheContrastiveAnalysisHypothesis(Lado1957,1964;James1980;GassandSelinker1983)holdthatL2learningcanbeeasierwherelanguagesusetheseattention-directingdevicesinthesameway,andmoredifficultwhentheyusethemdifferently.TotheextentthattheconstructionsinL2aresimilartothoseofL1,L1constructionscanserveasthebasisfortheL2constructions,but,becauseevensimilarconstructionsacrosslanguagesdifferindetail,theacquisitionoftheL2patterninallitsdetailishinderedbytheL1pattern(Odlin1989,2008;Cadierno2008;RobinsonandEllis2008a,b).

Achard(2008),Tyler(2008),andotherreadingsinRobinsonandEllis(2008b)showhowanunderstandingoftheitem-basednatureofconstructionlearninginspiresthecreationandevaluationofinstructionaltasks,materials,andsyllabi,andhowcognitivelinguisticanalysescanbeusedtoinformlearnershowconstructionsareconventionalizedwaysofmatchingcertainexpressionstospecificsituationsandtoguideinstructorsinpreciselyisolatingandclearlypresentingthevariousconditionsthatmotivatespeakerchoice.

20.5.ReconstructingForminL2—CrosslinguisticTransfer

AsSlobin(1993:242)notes,“[f]orthechild,theconstructionofthegrammarandtheconstructionofsemantic/pragmaticconceptsgohand-in-hand.Fortheadult,constructionofthegrammaroftenrequiresarevisionofsemantic/pragmaticconcepts,alongwithwhatmaywellbeamoredifficulttaskofperceptualidentificationoftherelevantmorphologicalelements.”L2learnersaredistinguishedfrominfantL1acquirersbythefactthattheyhave

Page 277: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 8 of 9

previouslydevotedconsiderableresourcestotheestimationofthecharacteristicsofanotherlanguage—thenativetongueinwhichtheyhaveconsiderablefluency(andanyotherssubsequentlyacquired).SincetheyareusingthesameapparatustosurveytheirL2too,theirinductionsareoftenaffectedbytransfer,withL1-tunedexpectationsandselectiveattention(Ellis2006c)blindingthecomputationalsystemtoaspectsofL2form,thusrenderingbiasedestimatesfromnaturalisticusageandtheconcomitantlimitedendstatetypicalofL2A.

Incaseswheretheformslackperceptualsalienceandsogounnoticed(Schmidt1990,2001)bylearners,orwherethesemantic/pragmaticconceptsavailabletobemappedontotheL2formsareunfamiliar,additional‘FocusonForm’(attentiontoformincommunicativecontext:Long1991;Lightbown,Spada,andWhite1993;DoughtyandWilliams1998;R.Ellis2001;Robinson2001;Ellis2005)islikelytobeneededinorderforthemappingprocesstobefacilitated.

InordertocounteracttheL1attentionalbiasestoallowestimationprocedurestooptimizeinduction,alloftheL2inputneedstobemadetocount(asitdoesinL1A),notjusttherestrictedsampletypicalofthebiasedintakeofL2A.Reviewsoftheexperimentalandquasi-experimentalinvestigationsintotheeffectivenessofinstruction(Lightbown,Spada,andWhite1993;EllisandLaporte1997;HulstijnandDeKeyser1997;Spada1997;DoughtyandWilliams1998;NorrisandOrtega2000)demonstratethatfocusedL2instructionresultsinsubstantialtarget-orientedgains,thatexplicittypesofinstructionaremoreeffectivethanimplicittypes,andthattheeffectivenessofL2instructionisdurable.Form-focusedinstructioncanhelptoachievethisbyrecruitinglearners’explicit,consciousprocessingtoallowthemtoconsolidateunitizedform-functionbindingsofnovelL2constructions(Ellis2005).Onceaconstructionhasbeenrepresentedinthisway,itsuseinsubsequentimplicitprocessingcanupdatethestatisticaltallyingofitsfrequencyofusageandprobabilitiesofform-functionmapping.

20.6.FutureDirections

Somuchremainstobeunderstood.RobinsonandEllis(2008b)detailalonglistofissuesforresearchintoCognitiveLinguistics,ConstructionGrammar,andSLA.ForsakeofbrevityIhighlightherejustafew.

Thestudyofchildlanguageacquisitionhasmadesomuchprogressinthelastthreedecadesbecauseitundertookproperempiricalanalysesoflearnerlanguage.SLAresearchissorelyinneedofdenselongitudinalcorporaofadultlanguageacquisitiontoallowdetailedinvestigationofL2constructionacquisitionasafunctionofinputandlearnercognition(OrtegaandIberri-Shea2005;CollinsandEllis2009).

AlthoughmuchhasbeenlearnedaboutsyntacticandsemanticbootstrappingintheemergenceofafewparticularVACsfromusage,athoroughinvestigationofthetype-tokenfrequencyusagedistributionsofallEnglishgrammaticalconstructionsisrequired.LargecorporasuchastheBritishNationalCorpus(e.g.,BNC-BYU;Davies2004–)ortheCorpusofContemporaryAmericanEnglish(COCA;Davies2008–)arerevolutionizingthestudyoflexicalandphraseologicalform.ButtheprimarymotivationofConstructionGrammaristhatwemustbringtogetherlinguisticform,learnercognition,andusage.Animportantconsequenceisthatconstructionscannotbedefinedpurelyonthebasisoflinguisticform,orsemantics,orfrequencyofusagealone.Allthreefactorsarenecessaryintheiroperationalizationandmeasurement.Thisisatallorder.O’DonnellandEllis(2010)outlineaproposaltodescribetheverbalgrammarofEnglish,toanalyzethewayVACsmapformandmeaning,andtoprovideaninventoryoftheverbsthatexemplifyconstructionsandtheirfrequency.Thislaststepisnecessarybecausethetype-tokenfrequencydistributionoftheirverbsdeterminesVACacquisitionasabstractschematicconstructions,andbecauseusagefrequencydeterminestheirentrenchmentandprocessing.NLPtechniqueshelpwiththeparsing,buttheanalysisofconstructionsemanticsremainseverdifficult.

Theresearchreviewedinsection20.3demonstrateseffectsofawiderangeoffrequency-relatedfactorsunderpinningeaseordifficultyofconstructionacquisition.Researchtodatehastendedtolookateachhypothesisbyhypothesis,variablebyvariable,oneatatime.Buttheyinteract.Andwhatisreallyneededisamodelofusageanditseffectsuponacquisition.Wecanmeasurethesefactorsindividually.Butsuchcountsarevagueindicatorsofhowthedemandsofhumaninteractionaffectthecontentandongoingcoadaptationofdiscourse,howthisisperceivedandinterpreted,howusageepisodesareassimilatedintothelearner'ssystem,andhowthesystemreactsaccordingly.Weneedtodevelopmodelsoflearning,development,andemergencethattakethesefactorsintoaccountdynamically.EllisandLarsen-Freeman(2009b)illustratehowthismightbedone,butonlyfortheusualsuspectsofVL,VOL,andVOO.Itisuncertainhowwellsuchmodelsmightscaleup.Andagain,properly

Page 278: The Oxford Handbook of Construction Grammar

Construction Grammar and Second Language Acquisition

Page 9 of 9

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). (c) Oxford University Press, 2014. All RightsReserved. Under the terms of the l icence agreement, an individual user may print out a PDF of a single chapter of a title in OxfordHandbooks Online for personal use (for details see

representingsemanticsinthesemodelsremainsamajorproblem.

Finally,weneedevertorememberthatlanguageisallaboutinteractions.Cognition,consciousness,experience,embodiment,brain,self,andhumaninteraction,society,culture,andhistoryareallinextricablyintertwinedinrich,complex,anddynamicwaysinlanguage.Yetdespitethiscomplexity,despiteitslackofovertgovernment,insteadofanarchyandchaos,therearepatternseverywhere.LinguisticpatternsarenotpreordainedbyGod,genes,schoolcurriculum,orotherhumanpolicy.Insteadtheyareemergent(Hopper1987;Ellis1998,2006b;MacWhinney1998)—synchronicpatternsoflinguisticconstructionatnumerouslevels(phonology,lexis,syntax,semantics,pragmatics,discourse,genre,etc.),dynamicpatternsofusage,diachronicpatternsoflanguagechange(linguisticcyclesofgrammaticalization,pidginization,creolization,etc.),ontogeneticdevelopmentalpatternsinchildlanguageacquisition,globalgeopoliticalpatternsoflanguagegrowthanddecline,dominanceandloss,andsoon.Wecannotunderstandthesephenomenaunlessweunderstandtheirinterplay.TheframeworkofComplexAdaptiveSystemscanusefullyguidefutureresearchandtheory(EllisandLarsenFreeman2006a,b;Ellis2008a,b;EllisandLarsen-Freeman2009a,b;Beckneretal.2009).

NickEllisNickEllisisProfessorofPsychology,ProfessorofLinguistics,ResearchScientistattheEnglishLanguageInstitute,andAssociateattheCenterfortheStudyofComplexSystems,UniversityofMichigan.Hisinterestsincludelanguageacquisition,cognition,emergentism,corpuslinguistics,cognitivelinguistics,andpsycholinguistics.Hislinguisticresearchconcerns(1)explicitandimplicitlanguagelearningandtheirinterface,(2)usage-basedacquisitionandstatisticallearning,(3)vocabularyandphraseology,and(4)learnedattentionandlanguagetransfer.Hisemergentistresearchconcernslanguageasacomplexadaptivesystem,networksanalysisoflanguage,scale-freelinguisticdistributionsandrobustlearning,andcomputationalmodeling.HeservesasGeneralEditorofLanguageLearning.

Page 279: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 1 of 13

PrintPublicationDate: Feb2013 Subject: Linguistics,Psycholinguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0021

PsycholinguisticsGiuliaM.L.BenciniTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterfocusesonpsycholinguisticsoflanguageproduction.Itprovidesempiricalevidenceforandagainstthetwo-stagemodeloflanguageproduction,whichassumesseparatelevelsforfunctional(semanticandsyntactic)processing,aswellasforpositionalprocessing.Thechapteralsodiscussestheresultsofstudiessupportingtheexistenceoflexicallyindependentstructurebuildingoperationsinlanguageproductioninadditiontolexicalrepresentations.ItalsocontendsthatlexicallyindependentstructuralprocessesoftenreceiveastraightforwardinterpretationasabstractconstructionsinaConstructionGrammarframework.

Keywords:psycholinguistics,languageproduction,functionalprocessing,positionalprocessing,lexicalrepresentations,abstractconstructions,ConstructionGrammar

21.1.Introduction

Psycholinguisticstraditionallycoversthreebroadareas:production,comprehension,andacquisition.First(L1)andsecond(L2)languageacquisitionaretreatedseparatelyinthisvolumeinthechaptersbyDiessel(L1)andEllis(L2).Thischapterfocusesonthepsycholinguisticsoflanguageproduction,whichisconcernedwithspecifyingtherepresentationsandprocessesrequiredtoturnfeaturesofthoughtorpreverbal‘messages’intolinguisticexpressionsthatcanbephonologicallyencodedandarticulated.Historically,totheextentthatlinguisticshastakenpsycholinguisticdatatoconstraintheorydevelopment,ithasdonesomorefromcomprehensionthanproduction.Therearegoodreasons,however,toexamineproductiondataandtoseekintegrationwiththeoriesoflanguageproduction.IfollowBock(1995),inlistingwhy.First,thereisanobservationmadebyGarrett(1980:216,quotedinBock1995:205)that“[t]heproductionsystemmustgetthedetailsoftheform‘right’ineveryinstance,whetherthosedetailsaregermanetosentencemeaningornot.”Bockelaborates,“Therecanbenoargumentaboutwhethersyntax,forexampleis‘important’inproduction,becausethespeakerasamatterofcoursecreatesthosefeaturesofutterancesthatwecalltheirsyntax”(1995:205).Second,“althoughvariationinaspeaker'suseofstructuresisoftenascribedtoaconcernformakingcomprehensioneasier,itislikelythatsomevariationsserveonlytomakeproductioneasier”(1995:207).

21.2.LanguageProduction

21.2.1AConsensusModel

SincetheseminalworkofGarrett(1975),languageproductionresearchershaveworkedwithinanoverallfunctionalarchitecturewhich,followingFerreiraandSlevc(2007)Irefertoasthe‘consensusmodel’forproduction.EarlierreviewscanbefoundinBockandLevelt(1994)andintheclassicvolumebyLevelt(1989).‘Consensus’shouldnotbetakentosignifyunanimousagreementonthemodel,andIwilladdresssomeofthepast

Page 280: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 2 of 13

andpresentlydebatedissueslaterinthischapter.Unlikeotherareasoflanguageresearch,however,thereappearstobegreateragreementamongresearchersbothontheexplanandaforatheoryoflanguageproductionaswellastheoverallarchitectureofthesystem.ToillustratetheoverallorganizationofthemodelIwilluseasanexampletheplanningofthesentenceThegirlisgivingflowerstoherteacher(Figure21.1).Themodeltakesasitsstartingpointarepresentationofanintentiontocommunicateameaning,calledthe‘message,’followedbyseveralprocessesof‘grammaticalencoding,’andendswithaphoneticcodeinterfacingwithmotorplansfororal/manualarticulation.Themessagespecifiesconceptualandsemanticinformationaboutentitiesandeventsandincludesperspective-takingandinformationstructure.Littleexperimentalresearchhasexaminedthenatureoftherepresentationsinthemessage,andassumptionsaboutthislevelareoftenderivedfromlinguistics.BenciniandGoldberg(2000;seealsoGriesandWulff2005)examinedthenatureofmessagesusingacategorizationtask(seesection21.2.4.3).Basedontheseresults,alongwithadditionalexperimentalfindingsreviewedlaterinthechapter,Iwillsuggestarevisionofthemessagelevelwithintheconsensusmodel.Ithenproposeanintegrationoftherevisedlanguageproductionmodelwithconstructionistapproachesinsection21.2.6.

Grammaticalencodingisthe‘linguisticengine’oflanguageproductionandisthoughttoinvolveseparateprocessesof‘selection’and‘retrieval,’referredtoas‘functional’and‘positional’processes,respectively(Garrett1975;Levelt1989;BockandLevelt1994).Functionalprocessesselectlinguisticrepresentationsunderspecifiedforcertaintypesofinformation(i.e.,phonology,inflectionalmorphology,andlinearorder).Positionalprocessesretrievetheselectedrepresentations‘fillingin’theunderspecifiedinformation.Bothfunctionalandpositionalprocessingcontainmechanismsthatdealwithcontent(lexicalprocesses)andmechanismsthatdealwiththeorderingofcontentunitsatdifferentlevelsofgranularity(structuralprocesses).Thelexicalprocessesatthefunctionallevelaretheselectionofabstractlexicalentriescalled‘lemmas,’suitableforconveyinglexicalmeanings.Lemmascontainsemanticandsyntacticinformation(butcruciallynophonologicalinformation).Thestructuralprocessesofthefunctionallevelassignlemmastogrammaticalrolessuitabletoexpressthesemanticandpragmaticdistinctionsspecifiedinthemessage.

Clicktoviewlarger

Figure21.1. Schematicrepresentationofconsensusmodelforlanguageproduction.Informationflowthroughthesystemisindicatedtobeoneway,from‘top’to‘bottom’;butbidirectionalflow,viafeedbackbetweenlevels,ispossible(dottedarrows).

Functionassignmentisinfluencedbymessagelevelvariables(conceptual,semantic/pragmatic)butalsobyproductionspecificpressuresrelatedtotheprocessingdynamicsoflanguageproduction,suchastheeasewithwhichreferentsareretrievedfrommemory(‘accessibility’).Itisacontentionofthischapterthat,inadditionto

Page 281: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 3 of 13

meaningandlistener-orientedconstraints,productionconstraintsalsoprovideavaluablewindowintounderstandinghowprocessing/usageshapesgrammars,forexample,intheareaofword-orderphenomena.Aswewillsee,productionmayinfactprovidethefunctionalmotivationfortheexistenceofconstructionsthatallowforwordorderflexibility.

Thelemmalevelisassumedtobeanabstract,modalityneutrallevelofrepresentation,sharedbetweentheprocessesofcomprehensionandproduction,bothwrittenandspoken.IntheexampleThegirlgivesflowerstoherteacherthefunctionallevelwouldselectlemmasforGIRL,TEACHER,FLOWER,GIVE,andTOandassigntheSubjectfunctiontoGIRL,theDirectObjectfunctiontoFLOWER,andObliquetoTEACHER.Thenextlevelispositionalprocessing.Lexicalprocessingatthepositionallevelinvolvestheretrievaloftheabstractphonologicalcodesofwordscalled‘lexemes’containinganoverallspecificationofaword'ssyllablestructureandabstractsegmentalinformation(CVstructure).Structuralprocessesatthepositionallevelcreatelinearizedsentenceframeswithopenslotsforclosedclassitemssuchasinflectionsanddeterminers.Theconsensusmodelassumesthatdominancerelationsarecomputedseparatelyfromlinearorderrelations.Itisonlyatthepositionallevelthatsurfacelinearizationofconstituents(wordorder)isassumedtotakeplace.

Alongdebatedissuewithintheconsensusmodelishowinformationisprocessedthroughthesystem,thatiswhetherprocessingis‘serial’and‘discrete’(Levelt1988;Leveltetal.1999),orwhetheritis‘interactive’or‘cascading,’allowingforlexemestobeaccessedbeforelemmasarefullyselected.Somevariantsofthemodelalsoallowforinformationtoflowbackwards(via‘feedback’)fromthephonologicallevelofthelexemetothelemmalevel(Dell1986;CuttingandFerreira1999;RappandGoldrick2000).Althoughtheissueofinformationflowthroughthesystemisimportant,itdoesnotsubstantiallychangetheoverallarchitectureoftheproductionsystem.Moresubstantialchallengestotheconsensusmodelcomefromchallengingthelemma/lexemedistinctionandtheexistenceofmodalityneutrallemmasinfavorofmodalityspecificrepresentationsthatcombinepropertiesoflemmasandlexemes(seesection21.2.4.1).

21.2.2EmpiricalEvidencefortheTwo-stageModel

Garrettproposedthetwo-stagearchitectureonthebasisofspeecherrordata(e.g.,Garrett1980).Therearequalitativelytwotypesoferrors:thosesensitivetothesemanticandsyntacticpropertiesofwords,andthosesensitivetoaword'sphonologicalform.Subsequentevidencecamefromexperimentssuchassentencerecallandpriming.Forexample,BockandWarren(1985)showedthat‘conceptualaccessibility’(theeasewithwhichthereferentofaNounPhrasecanberetrievedfrommemory)affectsthefunctionallevelandnotthepositionallevel,henceprovidingexperimentalevidenceforthedistinctionbetweenthesetwolevels.Inasentencerecallparadigm,theymanipulatedconceptualaccessibilitybymanipulatingtheimageabilityoftheNPreferents(high/lowimageability)inactive/passive,ditransitive/prepositionaldatives,andinactivesentencescontaininganobjectconjunctclause(e.g.,Thelosthikerfoughttimeandwinter).Speakersplacedaccessiblematerialinsubjectpositionforactivesandpassivesandindirectobjectpositionfordatives.Accessibility,however,didnotaffecttheorderingofconjuncts.BockandWarren(1985)deducedthatconceptualvariablesaffectthemappingfromthemessageleveltotheabstract(unordered)grammaticalleveloffunctionassignment,whereasphonologicalvariables(e.g.,length,phonologicalrelatednesstoaprime)affectlinearorderingandphonologicalencodingatthepositionallevel.

Bock(1986a)alsoprovidedevidenceforthedistinctionbetweenfunctionalandpositionalprocessingusingalexicalprimingprocedure.Speakersdescribedpicturesoftransitiveevents(e.g.,lightningstrikingachurch).Beforeeachpicture,participantswereprimedwithawordinoneoffourconditionsobtainedbycrossingrelatedness(semantic,phonological)andsemanticrole(agent,patient).So,forexample,thewordcouldbesemanticallyrelatedtotheagent(thunder)andthepatient(worship)orphonologicallyrelatedtotheagent(frightening)andthepatient(search).Resultsshowedthatsemantic,butnotphonologicalprimingincreasedtheprobabilityofassigningthelemmatosubjectrole.Moreactiveswereproducedaftersemanticallyprimingtheagentandmorepassiveswereproducedaftersemanticallyprimingthepatient;crucially,phonologicalprimingdidnothavethiseffect.

OneproblemwiththeseearlyexperimentalstudiesonEnglish,however,isthattheeffectofconceptualaccessibilityonfunctionassignmentalwaysresultsinearlierplacementoftheNPinthesentence(e.g.,aseitherthesubjectofanactiveorapassivesentence).ItisthusdifficultinEnglishtodisentanglewhetheraccessibility

Page 282: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 4 of 13

affectsNPplacementinthesentencedirectly(atthepositionallevel),orindirectly(atthefunctionallevel)byassigningthemoreaccessiblelemmatoagrammaticalfunctionthatthenhappenstooccurearlierinthesentence.Languageproductionresearchershaveincreasinglyaddedacrosslinguisticdimensiontoresearch,examiningthesequestionsinlanguageswheregrammaticalfunctionandlinearordercanbemoreeasilydecoupledthaninEnglish(seesection21.2.4.2).

21.2.3ProductionSpecificInfluencesonGrammar

InthissectionIpresenttwohighlystudiedproductionspecificprocessingphenomena:(1)‘incrementality’and(2)‘accessibility’andIsuggestthattheyhavearoleinshapinggrammars.Theyalsoservetoillustratehowgrammarscanbetheresultofaninteractionofdifferentfunctionalpressuresfrombothcomprehensionandproduction,alongsidestructural/typologicalvariationoperatingoverdiachronictime.Iproposethatwordordervariationscorrespondingtoclassic‘alternations,’inadditiontobeingassociatedwithsubtlemeaningdifferences(Goldberg1995),arealsomotivatedbythedynamicsoflexicalretrievalduringon-linespokenproduction.

21.2.3.1IncrementalityEarlymentionofreferentsinspeakingislinkedtothedynamicsoflanguageproduction,specificallytoincrementality—thenotionthatgrammaticalencodingproceedsefficientlybyprocessingmaterialasitbecomesavailable,thusreducingtheamountofadvancedplanningrequiredbeforearticulation.

Theextenttowhichproductionisincrementalisanactivetopicofresearch,withsomestudiessuggestingahigherdegreeofincrementality(KempenandHuijbers1983;Ferreira1996;Roelofs1998;FerreiraandDell2000;KempenandHarbush2003)andothersagreaterdegreeofforwardplanning(Bock1986a;FerreiraandSwets2002;vanNiceandDietrich2003).Incrementalitymayalsobecontextdependent.Forexample,FerreiraandSwets(2002)foundthatspeakersweremoreincrementalwhentheyweregivenadeadlinetoinitiateproduction.Thefindingthatincrementalitymaybecontextsensitivesuggeststhatthelanguageproductionsystemoperatesbyoptimizingitsresources(FerreiraandHenderson1998;FerreiraandSwets2002).Italsohighlightstheneedsforstudiesinavarietyoftestingconditions,aswellasmorenaturalisticcontexts(Gries,thisvolume).

21.2.3.2AccessibilityAccessibilityistheeasewithwhichrepresentationsareretrievedfrommemory.Alargenumberoffactorshasbeenfoundtoaffectaccessibility,including‘imageability’(BockandWarren1985),‘animacy/humanness’(Bock,Loebell,andMorey1992;Prat-SalaandBranigan2000),‘prototypicality’(Onishietal.2008),‘previousmention/priming’(BockandIrwin1980;Bock1986a;MacWhinneyandBates1978;Prat-SalaandBranigan2000;FerreiraandYoshita2003),and‘visualsalience’(Gleitmanetal.2007).

Accessibilityandincrementalityprovideanadditionalmotivationfortheexistenceofwordordervariations,includingclassicgrammatical‘alternations.’Wordordervariationallowsspeakerstobeincrementalandproducemoreaccessiblematerialearlier,increasingprocessingefficiency(Bock1982;FerreiraandDell2000).Itwillbeimportanttosystematicallyexaminetheinteractionbetweenproductiontaskswithdifferenttimingconstraints(deadline/nodeadline),complexity(e.g.,Ferreira1991),aswellascommunicativesetting(isolatedlanguageproductionvs.dialogue).Whateverthefindings,itisclearthatproductionspecificpressureshaveeffectsongrammarandthusneedtobetakenintoconsiderationinadditiontosemantic-conceptual(messagelevel)effectsandlistener-based,or‘audience-design’effects.

21.2.4DebatesinLanguageProduction

21.2.4.1AreLexicalRepresentationsModalityGeneralorModalitySpecific?Inrecentyearsthetwo-levelarchitecturehasbeenchallengedbothattheleveloflexicalandstructuralprocessing.Letusstartwithlexicalprocessing.Recallthatintheconsensusmodelthelemmaisamodalityneutralsemantic/syntacticrepresentationandthelexemeisamodalityspecificrepresentation(phonological,graphemic,gestural).Caramazzaandcolleagues,however,haverejectedthelemma/lexemedistinctiononneuropsychologicalandexperimentalgrounds(seeCaramazza1997,fortheinitialchallenge)arguinginsteadforamodelthatdoesawaywiththeabstractamodallemmainfavorofauniquerepresentationthat,inthecaseof

Page 283: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 5 of 13

languageproduction,combinessyntactic,semantic,andphonologicalinformation.Inthisapproach,allofaword'sproperties—syntactic,semantic,andphonological—operateinparalleltodetermineitspositioninalinearizedstring(e.g.,AlarioandCaramazza2002).Notethatalthoughthisviewmayprimafacieevoketherepresentationsassumedinparallelarchitecturelinguisticmodels(e.g.,Jackendoff,thisvolume)thismodelactuallydeniestheexistenceofamodalrepresentationssharedbetweentheprocessingsystemsandarguesforseparate/separableentriesforeachinput/outputmodality,includingseparateinputrepresentationsforlisteningandreading,andseparateoutputrepresentationsforspeakingandwriting.

Neuropsychologicalevidenceforthemodalityspecificnatureofgrammaticalandphonologicalrepresentationscomesfrombrain-damagedindividualswhoexhibitmodalitybygrammaticalcategorydissociations.Themoststrikingcaseistheexistenceofadoubledissociationwithinasinglesubjectwhoshowedanoun/verbdissociationwithmoreimpairednamingfornounsthanverbsinthespokenmodality,andtheoppositepatterninthewrittenmodality(HillisandCaramazza1995;RappandCaramazza2002).Thereisalsomorerecentexperimentalevidencefortheinteractionofgrammaticalcategoryandphonologicalrelatednessontheorderingofshortmultiwordstrings.JanssenandCaramazza(2009)foundevidencethatphonologicalrelatednessdifferentiallyaffectedthetimeittookspeakerstoproducethefirstwordinwordsequenceswithwordsfromsameversusdifferentgrammaticalclasses(noun–noun,noun–adjective,adjective–noun,adjective–adjective–nouncombinations).

Thedebateoverwhetherthelanguageprocessingsystemsoperateonmodalityspecificrepresentationsorwhetherthereareinstead(orinaddition)modalitygeneralrepresentationssharedbetweentheprocessingsystemshasconsequencesbeyondpsycholinguisticsandtherelationshipbetweentheprocessingsystems.Forexample,ithasconsequencesforacquisition,becauseitimpliesthatlearninginonemodality(e.g.,learningincomprehension)shouldnottransferdirectlyintolearningintheothermodality(e.g.,learninginproduction).Beyondpsycholinguistics,italsohasconsequencesfortherelationshipbetweenlinguisticsandpsycholinguistics.Astrongversionofthecognitivelyandneurallyseparaterepresentationsviewwouldunderminetherelevanceoflinguisticrepresentationsgleanedfromjudgmentsofacceptability/grammaticality(anotherformofcomputation/representation)totheirinvolvementinunderstandingandproducinglanguage.Iwillreturntothequestionofwhetherthereisevidenceformodalityneutralsentencerepresentationsinsection21.2.4.4.

21.2.4.2HowManyStepstoProduceWordsinOrder?Withrespecttostructuralprocesses,thedistinctionbetweenthetwolevelsoffunctionassignmentandlinearorderinghasalsobeenchallenged.Evidenceagainstthisdistinctioncomesfromaparticularlypowerfulexperimentaltechniquethathasbeenusedextensivelytoexaminethearchitectureandmechanismsoflanguageproduction:structuralpriming.Structuralprimingreferstothetendencyofspeakerstoproducepreviouslyexperiencedsentencestructures.IntheclassicdemonstrationbyBock(1986b),primingwasfoundwithActive/PassiveandDitransitive/PrepositionalDativeconstructions.Speakersweremorelikelytodescribetwo-participanttransitiveevents(e.g.,apictureofabeestingingaman)withapassivesentenceiftheypreviouslyheardandrepeatedapassive(e.g.,The747wasalertedbytheairportcontroltower).Asinothertypesofpriming,ifpreviousprocessingofastimulusinfluencestheprocessingofasubsequentstimulus,wecaninferthatthecognitivesystemissensitivetotheoverlappingdimensions(e.g.,structure,semantics,etc.)betweentheprimingstimulusandthesubsequenttargetstimulus.Byvaryingthedimensionsofoverlapbetweenprimingsentenceandtargetsentence(e.g.,Bock1986b,1990;BockandLoebell1990;PickeringandBranigan1998)priminghasallowedresearcherstostudythelinguisticrepresentationsinvolvedinlanguageproduction(Braniganetal.1995).Usingastructuralprimingparadigm,Pickeringetal.(2002)arguedforasinglemodelofproductioninwhichdominancerelationsandlinearorderarecomputedsimultaneously,basedonthefindingthatdominancerelationsalonedonotexhibitpriming:“shifted”prepositionaldativessuchasThedrivershowedtothemechanictheoverallswiththestainsdonotprimenonshiftedprepositionaldativessuchasThemechanicshowedtheinjurytothedoctor.

Inadditiontoevidencefromstructuralpriming,anotherapproachtodeterminewhetherthelanguageproductionsystemisstagedwithrespecttothecomputationofdominancerelationsandlinearorderingcomesfromstudiesonlanguageswithmoreflexiblewordordersthatallowlinearordertobedecoupledfromgrammaticalroles.Aswesawinsection21.2.3,theaccessibilityofareferentinfluencestheassignmentofthecorrespondingNPtosubjectpositioninEnglish(Bock1986a),inturnresultingintheplacementofthatlemmaearlierinthesentence.Thisraises

Page 284: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 6 of 13

thequestionastowhethertheaccessibilityofalemmaaffectsearlymentionofthatlemmadirectly,orindirectlyviaassignmenttosubject—andsubjecthappenstooccurearlierinEnglish.ThisquestionwasexaminedrecentlyinJapanese,whichallowsbothsubject–verb–object(SOV)andobject–verb–subject(OVS)sentencesinbothactiveandpassiveforms(distinguishedmorphologicallyontheverb).Braniganetal.(2008)foundthatgreaterconceptualaccessibilityhadaneffectbothonfunctionassignmentandonlinearordering.Inasentencerecallparadigm,Japanesespeakersassignedmoreconceptuallyaccessible(animateentities)tosubjectfunctionindependentofwordorder,andconversely,theypreferredtoplacemoreaccessibleentitiesearlierinsentences,independentofgrammaticalfunction.Basedontheseresults,theauthorsarguedthatdominanceandlinearorderarecomputedinonestage.

Itisbeyondthescopeofthischapter(andperhapstooearly)toresolvetheissueofwhetherfunctionalandpositionalprocessingshouldbeseparated.Ananswermaycomefromtheconvergencebetweenfunctionallyorientedlinguistictypologyandpsycholinguistics.Thereisagrowingbodyofcrosslinguisticpsycholinguisticresearchoveramoretypologicallydiverseempiricalbasethanwhatdominatedthefirstthirtyyearsofresearch.Thepicturethatseemstoemergeisoneinwhichprocessingconstraints(fromproductionandcomprehension)interactwiththeindividualgrammaticalpropertiesofdifferentlanguagesbothinnormal(BatesandDevoscovi1989;BatesandMcWhinney1982;YamashitaandChang2001,2006)andcognitivelyimpairedspeakers(e.g.,seeBencinietal.2011).Generally,itisbecomingapparentthatcrosslinguisticpsycholinguisticsisneededinordertodeterminebothwhichaspectsoftheprocessingsystemsmaybeuniversalandwhichonesresultfromtheinteractionbetweenprocessingconstraintsandgrammaticalvariation.

21.2.4.3StructuralInfluencesinSentenceProductionWithinalargerdebateincognitivescience(RumelhardtandMcClelland1986;FodorandPylyshyn1988),questionsabouttheexistenceofabstract,structuralrepresentationsandprocessesoperatingindependentlyofspecificcontenthavedominatedmuchresearchinallareasofpsycholinguistics.Theuseoftheterm‘abstract,’however,needstobequalifiedrelativetothetheoryoneassumes.Inlinguistics,whatis‘abstract’inoneframeworkmaybetooconcreteinanothertheoreticalapproach.Inpsycholinguisticstheterms‘structural’and‘abstract’representations,however,aretobeunderstoodincontrasttolexicallyspecificrepresentations,andmoregenerallytheassociatedviewthatthelexicalrequirementsofspecificwords(typically,verbs)arethedrivingforceinsentencecomprehension(e.g.,MacDonaldetal.1994),production(e.g.,BockandLevelt1994),andacquisition(e.g.,Tomasello1992;2000a).Withinlanguageproduction,therepresentationsassumedaregenerallysurfaceoriented,nonderivational(butseeF.Ferreira2000;FerreiraandEngelhardt2006),andphilosophicallycompatiblewiththetypesofrepresentationsassumedintheconstructionisttradition(Goldberg1995,2006a;includingSign-BasedConstructionGrammar,seeMichaelis,thisvolume;theparallelarchitecturemodelofJackendoff2002a,thisvolume)andin‘simplersyntax’approaches(CulicoverandJackendoff2005).Thestructuralviewisneutralastowhethersentencestructuresarestoredorwhethertheyareassembledon-line(Jackendoff2002a).Thecrucialpointiswhethertheproductionofsentenceformreliesontheretrievalofspecificwords,orwhetheritisindependentofit(KonopkaandBock2008).

Clearly,producingwell-formedsentencesrequiresknowledgeofindividualwords/lemmasandtheirsemanticandsyntacticproperties.Thereisalargebodyofevidencepointingtotheexistenceoflexicallyspecificrepresentationsinpsycholinguistics(BockandLevelt1994;McDonaldetal.1994;Lievenetal.1997;Tomasello2000a;2000b,interalia).Atissueistheextenttowhichstructuralinformation(e.g.,whetheraspeakerproducesaditransitiveorprepositionaldativesentencestructure)necessarilyrequirespriorselectionofalexicalentry(e.g.,averb)orwhether,alongsidelexicalrepresentations/processes,therearelexicallyindependentmappingsfromrelationalmeaningsinthemessagetogrammaticalencoding.

Accordingtolexicalistviewsofsentenceproduction(e.g.,BockandLevelt1994;Ferreira2000)thesentencelevelrepresentationsthatorganizespeakers’utterancesareintrinsicallyboundtotheverblemmasthatareassumedtolicensethem.Thisviewamountstoaprojectionistviewoftherelationshipbetweenverb,sentencestructure,andsentencemeaning,aviewthathasdominatedmuchofpsycholinguisticresearchnotjustinproductionbutalsoincomprehensionandinacquisition.Accordingtoabstract,structural(orframe-based)viewsofsentenceproduction(e.g.,WardlowLaneandFerreira2010;KonopkaandBock2008;seeBenciniandGoldberg2000;GoldbergandBencini2005foraconstructionistperspective),sentenceproductionoperatesonrepresentationsthatsegregatelexicalandstructuralinformation.Thatis,lexicalitemsandsentencestructuresareseparately

Page 285: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 7 of 13

retrieved/assembledonthebasisofmessagerepresentations.Intheexampleabove,theditransitivesentenceframeisnotactivatedsolelybytheselectionoftheverbgive,ratheritresultsfrom‘give’andanabstract(lexicallyindependent)mappingfromthespeaker'sintentiontodescribeathreeparticipantencodingofaneventtoanallowablesentencestructuresuitabletoexpressthespeaker'smessage.

Theclassiclexicalist/projectionistviewoftherelationshipbetweenverb,sentencestructure,andsentencemeaningisonethat,startingwithGoldberg's(1995)workinlinguistics,waschallengedbothontheoreticalandexperimentalgrounds(Goldberg1995;BenciniandGoldberg2000;GoldbergandBencini2005).IreferthereadertoGoldberg'sseminalworkforthelinguisticarguments;hereIbrieflyreviewsomeoftheexperimentalevidence.BenciniandGoldberg(2000)usedacategorizationtaskinwhichtheypittedverbsagainstconstructionsindeterminingoverallsentencemeaning.TheyfoundthatnativeEnglishspeakerscategorizedsentencesbasedonoverallmeaningbytakingintoaccounttheoverallargumentstructuresofsentences,inadditiontoverbs.Theseresultssuggestacontributionofsentencestructuretosentencemeaning,independentofverbmeaning.Returningtotheconsensusmodeloflanguageproduction,theseresultsshedlightintothemessagelevelofencoding,undertheassumptionthatthemessageissharedbetweentheprocessingsystems.

Inlanguageproductionresearchproper,therearethreeexperimentalsetsofresultsthatprovideevidencefortheexistenceoflexicallyindependentstructurebuildingoperationsinlanguageproductioninadditiontolexicalrepresentations.Iwillstartwiththeclassicone,basedonevidencefromstructuralpriming.TheoriginalstructuralprimingstudiesbyBockandcolleagues(e.g.,Bock1986b;BockandLoebell1990)demonstratedthatspeakersrepeatedsentencestructuresfromprimingsentencetotargetsentenceevenwhentheprimeandthetargetdidnotusethesamecontentwords,andimportantlyforourpurposes,didnotshareverbs.Theexistenceofstructuralprimingintheabsenceofsharedlexicalcontentindicatesthattheprocessorissensitivetosharedsemantic/syntacticstructuresindependentoflexicalcontent.PickeringandBranigan(1998)extendedthesefindingsbymanipulatingtheamountoflexicalandmorphologicaloverlapbetweentheverbsintheprimeandinthetarget.Theyfoundanenhancedprimingeffectwhentheverbwasrepeatedbetweentheprimeandthetarget,whichisnowreferredtoasthe‘lexicalboost’effectinstructuralpriming.Theyalsofoundthatprimingwasunaffectedbywhethertense,aspect,ornumberoftheverbstayedthesameordifferedbetweenprimeandtarget.

PickeringandBranigan(1998)proposedamodelofverblemmasthatincludeslinkstoaverb'smultipleargumentstructures,called‘combinatorialnodes.’Inthismodel,whengiveisusedinaditransitiveconstructionthecombinatorialnodesNPNPareactivated.Whengiveisusedintheprepositionaldative,theNPPPnodesareactivated.Combinatorialinformationisnotlinkedtoaparticularmorphologicalinstantiationofaverbform(itslexeme)buttotheverblemma,henceitcanbeactivatedcross-modally.Althoughthismodelaccountsbothforlexicallyindependentprimingandthelexicalboosteffect,itisstillalexicallydrivenmodeloflanguageproductioninthattheverblemmaspecifiesthesubcategorizationpreferences(combinatorialnodes)aspartofitsstoredlexicalrepresentation.

AmorerecentexperimentbyKonopkaandBock(2009)alsousedaprimingparadigmtopitlexicalversusstructuralguidanceviewsofsentenceproduction.Theyexaminedprimingbetweenactivetransitivesentencescontainingidiomatic(i.e.,semanticallynoncompositional)phrasalverbs(e.g.,pulloffarobbery/pullarobberyoff)andnonidiomaticphrasalverbs(e.g.,flipoverapancake/flipapancakeover).Thequestionwaswhetherparticleplacementcouldbeprimedinidiomsandnonidiomsalike.Theyalsoexaminedwhetherprimingwouldbeinfluencedbythedegreeofstructuralflexibilityofthephrasalverb(e.g.,frozenidiomslikeMary'sgrandpafinallygaveuptheghost/*gavetheghostupvs.flexibleidiomslikeThehikersbroketheirnewbootsin/brokeintheirnewboots;frozennonidiomslikeThenewmaterialgaveoffaweirdsmell/*gaveaweirdsmelloffvs.flexiblenonidiomslikeJudysnappedonherearrings/snappedherearringson).Onalexicalistmodelofproduction,sentenceswithidiomaticparticleverbsshouldbestoredandretrievedaslexicalentries(moreakintolexicalaccess),andnotproducedviatheregularrouteassumedforsentencegeneration(functionassignment/constituentassembly).Onastructuralmodel,incontrast,thestructurebuildingmechanismsshouldoperateforidiomaticandnonidiomaticphrasalverbconstructionsalike.Onalexicalview,therefore,idiomaticprimesarepredictedtoexhibitreducedprimingeffectivenesscomparedtononidiomaticprimes.Onastructuralview,primingispredictedtobequalitativelyandquantitativelysimilarforidiomsandnonidioms.Consistentwithastructuralview,theresultsshowedidenticalprimingpatternsforidiomsandnonidioms(bothasprimesandastargets).Forbothidiomsandnonidiomsalike,structuralinflexibilityreducedtheeffectivenessofpriming.Theseresultsargueforaviewofproductioninwhichtheoperationsresponsibleforgeneratingsurfacestructureare

Page 286: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 8 of 13

lexicallyindependentevenformultiwordexpressionsthataresemanticallynoncompositional.

Athirdsetofresultsbearingontheissueoflexicallyversusstructurallyguidedsentencegenerationcomesfromanerror-elicitationparadigminwhichspeakersareaskedtorapidlyrepeatshortsentencesorphrasesunderconditionsthatmakeerrorslikely.WardlowLaneandFerreira(2010;seealsoFerreiraandHumphreys2001)elicitedso-called‘stemexchange’errorsinwhichwordsofdifferentsyntacticcategoriesexchangepositions,oftenstrandinganinflection.Forexample,aspeakerwhoerroneouslyuttersIroastedacookhasexchangedanintendedverbcookwithanounroast,producingwhatappearstobeanill-formedutterancewithanounerroneouslyproducedwhereaverbwasintendedandviceversa.FerreiraandHumphreys(2001),however,notedthatmanystemexchangeerrorsinvolvewordsthatareformallyambiguousastocategorymembership,raisingthepossibilitythat,infact,stemexchangeerrorsaresyntacticallywell-formed(i.e.,thatroastedisreallyaverb).Becausethestemsfortheverbandnounformsforcookandroastareindistinguishable,WardlowLaneandFerreiraselectednounandverbformsthatdifferinstressplacement,orstress-shiftingstems(thenounformREcord,vs.theverbformreCORD).Ifspeakerscanbeinducedtomakestemexchangeerrorswithstress-shiftingstems,itispossibletodeterminethesyntacticcategoryofthestem.Theauthorselicitederrorsundertwostructuralconditions:asyntacticallyconstrainingcondition,whichforcedcategorymembership,andasyntacticallyneutralconditionthatdidnot.Participantsheardpairsofwords(e.g.,REcord,hate)andwereinstructedtoproducethewordsineitheroneoftwoframes,promptedonthescreen:eitherinasyntacticallyconstrainingtransitiveframe(Use:______the______)orsyntacticallyneutralconjunctclause(Use:______and______).Participantswerealsoinstructedtoeitherrepeatthewordsinthesequencegiven,ortoswaptheirposition.Swappingwasincludedtomaximizethechancesofproducingerrors,butitonlyoccurredonfillertrials,neverexperimentalones.Thequestionwaswhetherspeakerswouldbemoreorlesslikelytoproduceerrorsasafunctionoftheproductiontemplateprovided.Onastructuralview,speakersshouldbeconstrainedbythetransitiveframetoproduceaverbforminthefirstslot,andanounforminthesecondslot.Moreerrorsshouldoccurresultinginastress-shiftedstemwithstressonthesecondsyllable(i.e.,resultinginaverbstem)inthesyntacticallyconstrainingframethaninthesyntacticallyneutralframe.Onalexicalview,therateofsucherrorsshouldnotdifferacrossframes.Resultsconfirmedthepredictionsofthestructuralview,suggestingthateveninthegenerationofspeecherrors,constraintsfromlexicallyindependentsentencetemplatesoperateinsentenceproduction.

FerreiraandWardlowLanemaketheinterestingproposalthatthemorphologicalconversionmechanismthatisrequiredtoaccountfortheseresults(changingthenounstemintoaverbstem)hasafunctionallybasedmotivationwithintheprocessesofproduction.Inordertomaintainfluencyandpreservewell-formednessinthefaceofthechallengesoflexicalretrieval,speakersmayexploitmorphologicalconversiononthefly.AspeakerwhostartstosayIamwaitingforthe…buterroneouslyretrievestheverbformdeliver,ratherthanproducingtheill-formedIamwaitingforthedelivercanresorttomorphologicalconversionandproduceIamwaitingforthedelivery.Asinthecaseofwordorderflexibility,theexistenceofmorphologicalrelationsandprocessesthatprovidethesystemwithflexibilityinlexicalretrievalduringsentenceproductionwillbeadvantageoustothecognitivesystem(Bock1982;FerreiraandDell2000).

Tosummarize,threeindependentlinesofexperimentalevidenceinlanguageproductionsuggestthatproducingsentencesinvolvesbothlexicalrepresentationsandprocessesandlexicallyindependentstructuralprocesses.Thisisinnowaytodenytheimportanceoflexicalprocessesinlanguageuse;examplesoflexicaleffectsaboundinallthreeareasofpsycholinguistics.Inlanguageproduction,lexicalandstructuralprocessingmustconvergetoproducewell-formedutterances;butcrucially,structuralprocessesdonotrestuponpriorlexicalaccessandretrieval.

Theevidencepointstotheexistenceinlanguageproductionofstructurallymediatedmappingsfrommessagelevelrelationalrepresentations(e.g.,‘Twoparticipanteventstructure,’‘Threeparticipanteventstructure’)tostructuralrepresentations.AsIunderstandthem,thesemappingscorrespondtotheabstractArgumentStructureconstructionsasdevelopedinGoldberg(1995;2006a).

21.2.4.4ThePCProblems:Performance/CompetenceandProduction/Comprehension.AreThereSharedStructuralRepresentations?Linguisticsisatheoryoftheknowledgethatnativespeakershaveoftheirlanguage.Acontentious,yet

Page 287: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 9 of 13

foundationalquestioniswhethertherepresentationsdiscoveredbylinguistsarealsodirectlyengagedintheprocessesofunderstandingandproducinglanguage.Theansweris‘yes’accordingtotheStrongCompetenceHypothesis(BresnanandKaplan1984)and‘no’accordingtowhatBocketal.(2007)calltheWeakCompetenceHypothesis.Linguists,irrespectiveoftheoreticalinclination,fallonbothsidesofthedebate—expressionofacommitmenttoconvergencewithpsycholinguisticscanbefoundbothinexponentsofmainstreamgenerativelinguists(e.g.,Marantz2006),aswellaswithinconstructionistandusage-basedapproachessuchasthoserepresentedbymanyofthechaptersinthisvolume.Arelatedquestioniswhetherproductionandcomprehensionengageoneuniquerepresentationalsystem,orwhetherlinguisticrepresentationsaresplitalongperformancelines:therepresentationsforcomprehensionnotneedingtobeidenticaltothoserequiredforproduction(ClarkandMalt1984).Theweakcompetenceviewfindsmoreadherentswithinpsychologyandneuropsychologythaninlinguistics;itseemssafetosaythatlinguists,irrespectiveoftheoreticalinclination,viewlinguisticknowledgeasauniquesystemofrepresentations.Evidenceforseparateknowledgerepresentationsunderlyingperformanceindifferentmodalitiescomesfromintuitiveobservations,suchasthefactthatone'scomprehensionofdifferentlinguisticvariantsanddialectstypicallysurpassesone'sproductionabilities(ClarkandMalt1984).Neuropsychologicalevidence(seesection21.2.4.1)alsoprovidessupportfortheviewthatdifferentknowledgerepresentationsforlexicalitemsunderliecomprehending,reading,speaking,andwriting.

Thefirstquestion(Performance/Competence)maybemoreepistemologicalinnature;whethertherepresentationspositedwithtraditionallinguistmethodsconvergewiththoseofpsycho/neurolinguisticswillbeamatterofscientificsuccessofthetheoriesthatcoverthelargestnumberofphenomena(Marantz2006).Thesecondquestion,whetherproductionandcomprehensionsharethesameamodalsentencerepresentations,hasrecentlybeeninvestigateddirectly,usingaspokencomprehensiontospokenproductionstructuralprimingtechnique(Bocketal.2007).Speakersdescribedpicturesafterhavingbeenexposedtospokensentencesindifferentstructures(active/passive;ditransitive/prepositionaldative),whichcrucially,theydidnotrepeat.Becausepriminginvolvesthereusingofpreviousrepresentations/processes,WeakCompetencepredictsthatprocessingaprimesentencethroughthecomprehensionsystemalone(listeningwithoutrepeating),shouldnotengagethesamerepresentations/processesasdoesactuallyproducingtheprimes(listeningandrepeating),andhencecomprehensiontoproductionprimingshouldnotmirrorproductiontoproductionpriming.StrongCompetence,ontheotherhand,predictssimilaramountsofprimingacrossmodality.ConsistentwiththepredictionsofStrongCompetence,resultsshowedremarkablysimilarcross-modalpriming,bothintermsofmagnitudeandtemporalduration,topreviouswithin-modalityprimingresults(BockandGriffin2000).Thesefindingsarguestronglyfortheexistenceofmodalitygeneralsharedsentencelevelrepresentationsbetweencomprehensionandproduction.Becauseprimingwasnotcontingentonlexicaloverlap,therepresentationsarealsostructuralandabstract.Similarcross-modalprimingresultsalsocanbefoundforspokentowrittenpriming(PickeringandBranigan1998),arguingforsharedrepresentationsirrespectiveofoutputmodality(speaking/writing).

Themechanismthattiesrepresentationsinonemodalitytoothermodalitiesisnotyetunderstood.Onepossibilityisthatthelinksaretiedtothemechanismsoflanguageacquisition(F.Chang2002;F.Changetal.2006).Inthisview,structuralpriming(whetherwithinoracrossmodality)supportslearningandgeneralizationinlanguageproductionbycreatingnewutterancesfromoldstructures.Cross-modalprimingtakesthisfurther,andallowsgeneralizedandabstractenoughlanguagelearningtooccurfromlisteningtospeaking.Evidencethatabstract(verbindependent)structuralprimingoccursinveryyoungchildrencanbefoundinBenciniandValian(2008)andwillbereviewedinsection21.2.5.Shimpietal.(2007)alsofoundcross-modalabstractpriming,inslightlyolderchildren.Thesefindingscorroboratethefunctionofprimingasamechanismforlanguagelearning.

Anotherintriguingpossibilityisthatcross-modalprimingistiedtoperception/actioncouplingaspartofageneraltendencyforrepresentationalalignmentduringdialogue,withstructuralprimingbeingthemostabstractformofalignmenttodate(PickeringandGarrod2004).Thisplacesstructuralprimingwithintherealmofsocialimitationandmorespeculativelymirrorsystems(Galleseetal.1996).Itisanopenquestionastowhetherthesetwooptionsareindependent,whetherthecomprehension/productionlinkexploitspre-existingmirrorcircuitry,orwhethertheexistenceofwithin-systemmodalitygeneralrepresentationslinkinganindividual'sprocessingsystemsaffordssocialimitation.Eitherway,bothmirrorsystemsandwithinindividualperception/actionlinksrequirethatthesystemhavetherepresentationalvocabularyinabstractenoughformattoallowgeneralizationovereventparticipants/semanticroles,inthecaseofmirrorsystems,andgeneralizationovereventparticipantsandpredicates,inthecaseofstructuralpriming.Inallofthesecases,therepresentationalvocabularyinConstruction

Page 288: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 10 of 13

Grammar,withslotsforparticipantsandpredicates(e.g.,Goldberg1995)isboththenecessaryandsufficientamountofabstractiontoallowforlearningandgeneralizationinlanguage(F.Chang2002;DomineyandHoen2006).

21.2.5TheDevelopingLanguageProductionSystem

Inthissection,Iexaminelanguageacquisitionfromaprocessingperspectivechosingtofocusonthedevelopmentofthelanguageproductionsystem,andaquestionthathasbeenatthecenterofrecentdebatesintheacquisitionliterature.IreferthereadertothechaptersbyDiesselandEllis(thisvolume)foramoreextensivetreatmentoflanguageacquisition.Thequestioniswhetherchildren'searlymultiwordutterancesrelyonanabstractvocabularywithgeneralizationsoverclassesofwords(e.g.,N,V,NP,VP)andgeneralizationsoverargumentstructures,suchasActive,Passive(EarlyAbstractionHypothesis),orwhethertheyaremoreconcretelyorganizedaroundlexicalknowledge(LexicalSpecificityHypothesis).

Untilrecently,comprehensionandproductiondatainchildlanguagepointedtoa‘paradox’inwhichchildrenappearedtorelyonabstractrepresentationsincomprehension(Naigles1990;Fisheretal.1994;Naiglesetal.2005;Gertneretal.2006),andlexicallydependentonesinproduction(Lievenetal.1997;ChildersandTomasello2001;seeTomasello2000b,forareview).HereIreviewevidenceforabstract(verb-independent)representationsinyoungchildren'slanguageproduction.Theevidenceonceagaincomesfromstructuralpriming.BenciniandValian(2008)examinedpriminginyoung3-year-olds(agerange2;11–3;6)intheabsenceofverboverlapandcontrollingforanimacy(seeHuttenlocheretal.2004forsimilarfindingswitholderchildren;Shimpietal.2007;andSavageetal.2003,forcontrastingresults).Becauselexicalretrievalisacomputationallydemandingproductionoperation(Bock1982),reasonablyallthemoresoforyoungchildren,theprimingphasewasprecededbyalexicalwarm-upphasefornounsandverbs.Verbswerepresentedinthegerund(e.g.,“look,hereisstirring”).Duringpriming,theexperimenterdescribedapicture(e.g.,themilkisstirredbythespoon)andthenthechildrepeated.Thiswasfollowedbya‘YourTurn’trial,inwhichthechilddescribedatargetpicture(e.g.,apictureofahammercrackinganegg).Usingstrictercriteriaforwhatcountedasapassivethantypicalinchildstudies,theresultsshowedabstractprimingofpassivesentences.Thisfindingsupportstheexistenceofabstract(verb-independent)representationsinyoung3-year-olds’languagesystems.Shimpietal.(2007)foundsimilarresults(withthecaveatthatanimacywasnotproperlycontrolled,andscoringcriteriawerelaxer)withactive/passiveandditransitive/prepositionaldatives.Theyalsofoundthatabstractprimingoperatedcross-modally(fromcomprehensiontoproduction)for4-and5-year-olds,butnotfor3-year-olds.Thereasonforthisisyetunknown,butimportanttopursue.Oneseriesofissuesiswhetheritisaconsequenceofthelearningmechanism:doesproductiontosomeextentproceedseparatelyfromcomprehension?Ordorepresentationshavetobestrongenoughtobeactivatedcross-modally?

Manyquestionsremaintobeansweredaboutthenatureofpriminginacquisition,forexample,whether,assuggestedforadults(Changetal.2006),itisaformofimplicitlearning(Savageetal.2006).BenciniandValian(2008)foundevidenceforlearningovertrialswhenscoringcriteriawererelaxedandlearningoccurredrapidly,overatotalof8trials.Priminghasprovedsuccessfulinansweringbasicquestionsinacquisitionwhoserelevanceextendsbeyondacquisitiontomodelsoflanguageproduction(Changetal.2006)andtolinguistics.Importantlyforconstructionistapproaches,theyseemtoofferthe‘right’levelofgeneralizationtodescribeyoungchildren'ssentenceproduction.Constructionsmayplayanimportantroleinlanguagelearningastheyprovideameansforchildrentogeneralizebeyondtheinput.Ontheotherhand,thislevelofgeneralizationisnotsoabstractthatitcannotbelearnedfromtheinputviagenerallearningmechanisms(e.g.,Goldberg2006b).

21.2.6IntegratingConstructionsintoaTheoryofSentenceProduction

Theevidencereviewedintheprevioussectionssupportstheexistenceoflexicallyindependent,modalityindependentstructuralrepresentationsrepresentedinthemessage(BenciniandGoldberg2000)andusedduringsentenceproduction(e.g.,WardlowLaneandFerreira2010).Theacquisitionstudiesshowtheserepresentationsaredevelopedearly(e.g.,BenciniandValian2008).

Inthissection,IofferanaturalintegrationoftheseresultswiththemodelproposedinGoldberg(1995)onthebasisoflinguisticdata.Twodifferentscenarioscanoccurinproduction.Thefirstinvolvestheintegrationofaconstruction(ditransitive)withapredicatewhosenumberofparticipantrolesisequaltothenumberofargument

Page 289: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 11 of 13

rolesrequiredbytheconstruction,usingasanexamplethesentenceThegirlgivesherteacherflowers.TherevisedrepresentationisshowninFigure21.2.

Themessagelevelrepresentsageneralizedtrivalentrelationbetweenanagent,arecipient,andatheme(Ileaveitasanopenquestionastowhatexactlytherepresentationcontains;theimportantpointisthattherelationbenottiedtoaparticularpredicate).Ingrammaticalencoding,therelationalmeaningmapsdirectlyontotheappropriatesyntacticrelationalstructure(functionalrepresentation)<subject,object1,object2>.Themessagealsocontainslexicallyspecificinformation,inthisexampletiedtogive.Themessagespecifiesalinkbetweenthethreepredicateparticipantroles<giver,givee,given>andthethreeditransitiveargumentroles<agent,recipient,patient>.Functionalprocessinglinksthesegeneralrolestothethreefunctionalroles<subject,object1,object2>ingrammaticalencoding.Asecondpossibility(tomyknowledgenotaddressedinanypsycholinguisticmodel)isoneinwhichaconstructionintegrateswithapredicatewhosenumberofparticipantrolesisasubsetofthenumberofargumentroles,asinthe‘classic’sentenceShesneezedthefoamoffthecappuccino(Ahrens1995;Goldberg1995).Inthiscasetheconstructioncontributesthetwoadditionalargumentrolesandtheoverallcause-to-receiveinterpretation,whilethepredicatecontributesthemeanscomponentoftheevent(Figure21.3).

Clicktoviewlarger

Figure21.2 Constructionsintheproductionmodel.Themessagelevelincludesverbindependentconstructionalmeaning,inadditiontoverbmeaning.Theconstructionandtheverbspecifythesamenumberofparticipantroles.

Thecombinationofverbsandconstructionsisbynomeansanunregulatedrelationship(Goldberg1995,1997).Thistopicisnotaddressedhere,butIreferthereadertothechaptersbyGries(thisvolume)andStefanowitsch(thisvolume)forasurveyofstudiesandmethodsexaminingtheinteractionofconstructionswithparticularlexicalitems.Thereisnoresearchinlanguageproductionexaminingwhethertheproductionofverbsinsentencestructurestheydoordonotsubcategorizeforisqualitativelysimilar.Bothstructural(KonopkaandBock2008;WardlowLaneandFerreira2010)andconstructionistviewspredictthatitshouldnotdifferandonewaytoexaminethiswouldbeinaprimingparadigmasinKonopkaandBock(2008).

Page 290: The Oxford Handbook of Construction Grammar

Psycholinguistics

Page 12 of 13

Clicktoviewlarger

Figure21.3 Constructionsintheproductionmodel.Themessagelevelincludesverbindependentconstructionalmeaning,inadditiontoverbmeaning.Theconstructionspecifiesadifferentnumberofrolesthantheparticipantrolesspecifiedbytheverb.Ingrammaticalencoding,theconstructionassignsthetwoadditionalrolestogrammaticalfunctions(shadedroles/functions).

Toconclude,Itakeitasasuccessforresearchinlanguagerepresentationandusethattwoveryindependentlinesofresearchwithindisparatedisciplinesandtheoreticaltraditionshaveconvergedonpositingremarkablysimilarrepresentations.

21.5.Conclusion

Thischapterprovidedareviewofsomecurrentissuesinpsycholinguisticsfocusingonlanguageproduction.Muchresearchinlanguageproductionhasconvergedwithrepresentationsthatappearconstructionalinspirit,thatis,surfacegeneralizationswithasufficientdegreeofabstractiontocapturelinguisticcreativityinthemappingfrom‘thought’to‘talk.’Additionalfindingsinlanguageacquisitionwithregardstothedevelopmentofthelanguageproductionsystemcorroboratetheutilityofaconstructionistframeworktoexplainlanguagerepresentationandprocessing,fromacquisitiontoadultlanguage.

GiuliaM.L.BenciniGiuliaM.L.BenciniisAssistantProfessorintheCommunicationSciencesProgramattheHunterCollegeBrookdaleHealthSciencescampus.SheobtainedherPh.D.inlinguisticsfromtheUniversityofIllinoisin2002andafterthatworkedasaNationalInstituteofHealthfundedpostdoctoralresearchfellowattheLanguageAcquisitionResearchCenter,atHunterCollege-CUNY.Sheisanexpertinpsycholinguisticsandcombinesinsightsandmethodsfromlinguistictheory,psycholinguistics,andlanguagepathologyinherwork.

Page 291: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,LanguageandCognition,MorphologyandSyntax

OnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0022

BrainBasisofMeaning,Words,Constructions,andGrammarFriedemannPulvermüller,BertCappelle,andYuryShtyrovTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterexaminestheneurophysiologicalplausibilityofsomeoftheclaimsofConstructionGrammarwithregardtosyntacticstructures.Itsuggeststhatevidencefromneurosciencehashighlyimportantrepercussionsforlinguistictheorybuildingingeneralandarguesthattheconstructionistenterprisereceivesconsiderableempiricalsupportfromneurolinguisticstudies.Thechapterexaminesviewsontheembodimentofgrammarinneuronalcircuitryandcontendsthatneurologicalevidenceindicatesthatitmakessensetopostulateflexibleconstructionaltemplatesasdistinctfromlexicalconstructionstorage.

Keywords:neurophysiologicalplausibility,syntacticstructures,neuroscience,linguistictheory,grammar,neuronalcircuitry,lexicalconstruction

22.1.Introduction

Howislanguagerealizedinthebrain?Thisisoneofthemajorscientificquestions,but,oddly,linguistshavebeenformulating,developing,andusingtheoriesoflanguagewhilepayinglittleattentiontotheissueofneuronalmechanismsthataretheultimatesubstrateofanylinguisticmechanisms.Infairness,someoftheavailablelinguisticapproachesmakeassumptions,atsomelevel,aboutthewaylanguageisinterlinkedwiththebrain.Averygeneralclaimhasbeenthatthereis,residinginthebrain,aninbornmodule—similartoaseparateorgan—responsiblefortheacquisitionandprocessingoflanguage(Chomsky1965;Fodor1983).Thisstanceistakenbygenerativelinguists,whoascribehumans’languagecapacitiestoaninnateandthereforeuniversalgrammar,conceivedsimilartoanalgebraicsystemgeneratingwell-formedsyntacticstructureswhichdonothavemeaning.Inrecentyears,thispositionhasbeenweakened,asneuroscienceandbehavioralexperimentsprovidedstrongevidenceforfunctionalinteractionsbetweenthebrainsystemsforlanguage,action,andperception,emergingevenattheearlieststagesoflinguisticprocessing,thusquestioningthepostulateofapartiallyautonomous‘linguisticmodule’(Pulvermüller1999;PulvermüllerandFadiga2010).Ithasbeensuggestedthat,insteadofanautonomous,self-containedlanguagemodule,humanlanguagemechanismsmayheavilydrawupongeneralcognitiveabilitiestoacquireandunderstandlanguage:theabilitytograduallyabstractstableform-meaningpatternsbytuningintorecurrentpropertiesoftheutterancestheyareexposedtoandtheabilitytousethecognitiveroutinestheyhavethusmasteredtocategorizenewlinguisticusageevents(Langacker1987,1991;Tomasello2003;cf.alsoBybee,thisvolume).Thispositionisembracedbycognitivelinguists,whoclaimthatsyntacticstructures,justlikewords,areinherentlymeaningfulandthatourknowledgeoflanguageemergespiecemealandinausage-basedfashion.Still,admittedly,thereisalongwaytogofromtheseclaimstoneuromechanisticexplanation.

Whereaslinguistshavesometimesnotbeenveryconcernedwithneurobiologicalmechanisms,researchersintheneurosciencearenamayfrequentlynothavesufficientlyfocusedonthekindsoffundamentalissueswhichgototheheartofdebatesintheoreticallinguistics.Althoughneurosciencehasmadegreatprogressinunderstanding

Page 292: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 2 of 14

thebrainbasisoflexicalandsemanticprocesses,ithas,untilrecently,remainedlargelysilentaboutthemechanismsunderlyingtheonedomainthatismostcentraltoanylinguistictheory:syntacticstructure.Smallwonder,then,thatthislacunahasallowedlinguiststochurnouttheoriesofgrammarpracticallyunhinderedbyconcernsofneurophysiologicalplausibility.Whatismore,manylinguistsareoftheopinionthatsyntaxisfartoocomplextobecapturedintermsofneuronsandconnectionsbetweenthem.Linguistsmaybereadytoacceptthattheneuronalstructureunderlyingwordsandtheconceptstheyrepresentcanbespelledoutasassembliesofneurons(cf.section22.2.1).Buthowaboutabstractlinguisticconstructions,suchastheonelicensingthesequenceofanounphraseandaverbphrase,forexample?Thesewouldseemtobetoofarremovedfromtheworldofconcreteevents,toomuchreliantonhighlysophisticatedprocessesofcognitiveabstraction,toallowadescriptionintermsofthebricksandmortarofnervecellsandcircuits.Indeed,Chomsky(1963)aimedtoprovethatneuronalalgorithmsofaquitebasicnatureareincapableofrepresentingandprocessingessentialsyntacticphenomena,andmostlinguistshavelonggivenupthehopeofbuildingaunifyingtheoryofthebrainandlinguisticcompetence.Asaresult,veryfewlanguageresearcherstakeaseriousinterestintheputativeneuronalmechanismsunderlyinggrammar.Thegenerallinguisticpositionseemstobethatneuroscientistshavenotyetmanagedtograpplewiththequestionsthatreallymattertolinguists.TociteChomsky(2000a:25):“TheERP(event-relatedpotential)observationsarejustcuriosities,lackingatheoreticalmatrix.”Similarly,Jackendoff(2002a:58)writes:“[A]lthoughagreatdealisknownaboutfunctionallocalizationofvariousaspectsoflanguageinthebrain,Ithinkitisfairtosaythatnothingatallisknownabouthowneuronsinstantiatethedetailsofrulesofgrammar.”Webelieve,however,thatsuchviewsarenotwell-founded,asbraintheoriesofgrammarhavebeenofferedandexperimentaldatahavebeenreportedthathelptoevaluatethese.Critically,braintheoryhasevenofferedperspectivesonaddressingissuessuchasabstractruleformationandrecursionintermsofneuronalautomata(forreview,see,e.g.,Pulvermüller2003a,2010a).Wethereforesubmitthatagroundingoflinguisticproposalsinneuromechanisticcircuitryispossible,and,asweaimtoshowhere,evenfruitful.

Jackendoff'squoterecognizesthatneuroimagingandalsoneuropsychologyhavebeenimmenselysuccessfulindefiningthebrainregionsactivatedduringsyntacticprocessingalongwiththebrainregions(andtimeintervals)ofsyntacticprocesses(Ullman2001;Friederici2002;Hagoort2005;Caplan2006;GrodzinskyandFriederici2006).However,thisworkhashadsofarlittleimpactonlinguistics.Onereasonforthismaybethatmuchneuroscienceresearchfocusedonthequestionof‘where’inthebrainsyntaxislocalized.The‘where’-questionassuchisnotconsideredtobeextremelyexcitingbymanylinguists,andtheirskepticismtowardmodernvariantsofphrenology(thenineteenth-centuryscienceofwhereaspectsofthemindarelocated,insomesenserevitalizedbyafloodofneuroimagingdatainthelasttwodecades)issharedbyneuroscientistswithatheoreticalinterest(e.g.,Page2006;Knight2007).Incontrast,thequestionof‘when’physiologicalindicatorsofsyntaxwouldfirstoccurdoessparkmoreinterest,especiallyinpsycholinguists.Theprecisetimingofelectro-andmagnetoencephalography(EEG,MEG)responsesprovidescluesaboutthetimecourseofsyntacticinformationaccessinthecontextoflexicalandsemanticprocessing;thishasimplicationsforpsycholinguistictheory(Friederici2002;Hagoort2005;PulvermüllerandShtyrov2006).However,themostinterestingfunctionalquestionof‘how’syntaxisrepresentedandcomputedinbraincircuits,had,foralongtime,notbeenaddressedatall,letalonethequestionof‘why’suchrepresentationsandprocessesarebeingsetupinthehealthyhumanbrain(Pulvermüller2010b).

ConstructionGrammarians,likecognitivelinguists,believethatlexiconandsyntaxarenotdevoidof,orencapsulatedfrom,meaning.Moreover,theyrejecttheexistenceofastrictdistinctionbetweenspecific(wordlike)lexicalitemsandmoreabstractrulelikegrammaticalpatterns:bothwordsandlargerconstructionscanbedescribedasform-functionpairings(cf.,e.g.,Goldberg1995,2006a;Boas2010b;HoffmannandTrousdale,thisvolume).Accordingly,thefollowingfundamentalquestionshavebeenbroughtup:Aretheneuronalcounterpartsofwordsandwordmeaningsinfactsimilartothoseofconstructionsandconstructionalsemantics?Doestheconceptof‘ruleofgrammar’becomeentirelyredundantinthecontextofconstructionstorageandprocessing,oristhereamechanisticbasisfordiscretecombinatorialrules?Canlinguisticquestionsofsuchfundamentalnaturebeaddressedonthebasisofneuroscientificevidence?Andcan,potentially,datadirectlytakenfromthebrainprovideneuroscientificjustificationforspecificlinguisticstatements?Weargueherethatworkintheneuroscienceoflanguagecanindeed,onthebasisofassumptionsaboutcorrespondencesbetweenlinguisticandneuronalmechanisms,addresscruciallinguisticquestions.Consequently,weproposethatthereisgoodreasonforanyexplanatorylinguistictheorytotakeintoaccountbothneurosciencetheoryanddata.

Therefore,inwhatfollowswewillfocusonthequestionofwhatweknowandcanreasonablyassumeaboutthe

Page 293: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 3 of 14

mechanisticleveloflanguageinthebrain.Wewillreviewviewsontheembodimentofgrammarinneuronalcircuitry(section22.2)anddiscussimplicationsofneuroscienceexperimentsforlinguistictheory,inparticularforConstructionGrammar(section22.3).Wewillthenprovideasummaryofthemainissues(section22.4).

22.2.ABriefOverviewofSomeBrainMechanismsUnderlyingLanguage

Beforeweturntosentences,wewilltakealookatwhatisknownaboutthewaythesmallestlinguisticelementsrelevanttomeaning—phonemes,morphemes,andwords—areprocessedinthebrain(section22.2.1).Wewillthenmentionsomefindingsthatarerelevantinlightofourgoaltoassesswhether,withinaneurobiologicalperspective,thelexicon-syntaxcontinuumproposedbyConstructionGrammarianshasaplausibleneuronalcircuitequivalent(section22.2.2),athreadwhichwewillpickupinsection22.3.4.Subsequently,wewilldiscussthebrainbasisofconstructionsabovethelevelofwords(section22.2.3)andof‘rulesofgrammar,’asopposedtofrequentpatternsofco-occurrence(section22.2.4).

22.2.1BrainIndexesforPhonologicalandLexicosemanticProcessing

Inneuroscience,recentprogresshasledtoabetterunderstandingofthebrainmechanismsofphonologyandsemantics.Oldmodels—accordingtowhichspeechproductionengages‘Broca'sarea’(theleftinferiorfrontalareainthecortexofthedominantlefthemisphere)while‘Wernicke'sarea’(theposteriorlanguageareainthesuperiortemporalcortex)actsasauniquecenterforlanguagecomprehension(Wernicke1874;Lichtheim1885)—havebeenoverthrownandthepictureemergingnowisthatphonemesandmorphemes,thesmallestunitsthatdistinguishorcarrymeaning,arerepresentedandprocessedinthebrainbyinteractivedistributedneuronalensemblesspreadoutoverarangeofareas,includingbothfrontalandtemporallanguageareas(Pulvermüller1999).Approachesstillclaimingarelativeindependenceofspeechcomprehensionfrominferior-frontalandmotormechanisms(see,e.g.,HickokandPoeppel2007)areconfrontedwiththeevidencefromlesionandmagneticstimulationexperimentsthatdemonstrateaninfluenceoffunctionalchangesintheseinferior-frontalareasonspeechcomprehensionandspeechperceptionprocesses(forreview,seePulvermüllerandFadiga2010).

Wordsarelinkedtotheirmeaningandtheneurobiologicalbasisforsuchlinkagehasbeenproposedtolieinneuronalconnectionsbetweenphonologicalandlexicalcircuits,ontheonehand,andsemanticcircuits,ontheother.Theseneuronallinksexploitcortico-corticalfiberbundlesbetweenareasthatmaybedistantfromeachother.Arangeofdataindicatesthatlexicalitemsreferringtoobjectsoractionsareorganizedinthebrainaslateralizedfronto-temporalneuronalcircuitslinkingupwithobjectoractionrepresentationsinvisualandothersensoryareasormotorsystemsofthebrain(Pulvermüller2005;Martin2007;Barsalou2008).Evidenceforthisviewcomesfromneuroimagingstudiesandfromworkwithpatients.Wordssemanticallyrelatedtoactions(e.g.,actionverbssuchastalk,write,orwalk)activatespecificallythosepartsofthemotorsystemthatareinvolvedinperformingtypicalreferentactions(articulator,hand,andlegmovements).Wordssemanticallyrelatedtovisuallydefinedobjectsactivatespecificpartsoftheinferiortemporalstreamofobjectprocessing(Chao,Haxby,andMartin1999;PulvermüllerandHauk2006;Simmonsetal.2007)andevenwordsreferringtosoundsorodorsmayactivatespecificallycorticalareasinvolvedingeneralhearingorolfaction(Gonzálezetal.2006;Kieferetal.2008).Hence,thebrainappearstoclassifywordsaccordingtotheirreferentialsemanticfeaturesandtothemodalities(movement,acousticsense,senseofsight,senseofsmell,etc.)involvedinprocessingthisinformation.Interestingly,similarextensionstosensoryandmotorregionsofthecortexmayevencontributetosemanticprocessingoffigurativeratherthanliterallanguageuse(GlenbergandKaschak2002;Glenberg,Sato,andCattaneo2008;Boulenger,Hauk,andPulvermüller2009).Forexample,whenunderstandingidiomaticsentencessuchasMarygraspedtheideaorPaulkickedthebucket,Boulengerandcolleaguesfoundsomatotopicactivationofthemotorsystemlongafteractionwordshadalreadybeenprocessedandsentencemeaningisnormallycomputed(Boulenger,Hauk,andPulvermüller2009).Theauthorsinterpretthisword-relatedactivationatastageofabstractsentencemeaningprocessingasevidenceforcompositionalsemanticcomputationsrecurringtoword-relatedmotorschemas.Abstractsemanticprocessesappeartodrawupontransmodalareas,especiallydorsolateralprefrontalcortexandtemporalpole(Binderetal.2005;Patterson,Nestor,andRogers2007).

22.2.2EarlyBrainSignaturesofLexicalandSyntacticProcessing

Thebrainrespondsextremelyfasttoincomingwords.Withinone-fifthofasecondaftertheinformationfor

1

Page 294: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 4 of 14

identifyingawordisavailableintheinput(eitheraftertheprintedwordisflashedonthescreenorafterthespokenwordseparatesfromalternativewords),specificbrainresponsesindicatewordrecognition.Thiscanbeobserved,forexample,intheEEGorMEGresponsetowordsandmatchedmeaninglessnovelsoundorlettersequencesthatrespectthephonotacticandorthographicrulesofthelanguage,so-called‘pseudowords’(Serenoetal.1998;Pulvermülleretal.2010). Theearlybrainactivationdifferencebetweenspokenwordsandpseudowordsevenemergeswhensubjectsareinstructedtoignoretheincomingspeechandhearwordsandpseudowordsasrareso-called‘oddballstimuli’embeddedinasequenceoffrequentlyreoccurring‘standardstimuli.’Ifthestandardstimulusisbi-,theraredeviantstimulusbitegivesrisetoalargerbrainresponsethanthedeviantpseudowordstimulus*piteinasequenceofpi-standards(Pulvermülleretal.2001;ShtyrovandPulvermüller2002;Garagnanietal.2009).Suchenhancedbrainactivitytowordscanonlybeexplainedasreflectingthelexicalaccessofthewordinmemory,or,putdifferently,theactivationofapreviouslyestablishednervecellassemblythatneuronallyinstantiatesthefamiliarword(PulvermüllerandShtyrov2006;Näätänenetal.2007).Thislexicalenhancementeffectforwords,comparedtopseudowords,contrastscruciallywithanenhanced‘syntactic’effectseenequallyearlyforsyntacticallyincorrectstringsofwords,comparedtofelicitousstringsofwords.Whereasbothoftheseeffectsoccurwithsimilarearlylatencies,theirpolarityisdifferent(seesection22.2.4)andthisfeaturemightbeofspecialinterestforlinguisticresearch(seesection22.3.4).Aneurobiologicalperspectiveexplainsthedifferentdynamicsbydifferentmechanisms:amplificationofthebrainresponsebytheignitionofcircuitsunifyingthepartsofastoredsingleword(or,moregenerally,storedlexicalitem)inthecaseofmemorizedwithin-wordlinksversusprimingbycontextandreductionofthebrainresponseinthecaseofsyntacticabove-wordlinks(Pulvermüller2003a,2010a;seealsoFeldmanandNarayanan2004;Feldman2006).

22.2.3.TheBrainBasisofConstructions:FromSequenceDetectorsinAnimalstoSyntacticLinkage

Notwithstandingtheprogressmadeindecipheringaspectsofthebraincodeofwordmeaning(section22.2.1),onemayjudgethatneurosciencehassofarfailedtoenlightenwhatweconsidertobethecoreofourlinguisticpowers,namelythespecificallyhumanabilitytovariablycombinemeaningfulunitsintorule-conforminghigher-orderstructures(cf.ourintroductionandthequotesthereinfromChomsky2000a;Jackendoff2002a).Indeed,atthemechanisticlevelofnervecellsandtheirconnections,theoreticalworkonsyntaxassociatedwithempiricaleffortsisstillrelativelysparse.Substantialworkfocusesonneuralnetworkarchitecturesthatmaysolvespecificsyntacticproblems(e.g.,BotvinickandPlaut2006;F.Chang,Dell,andBock2006;ChaterandManning2006)andsomestudiesofneuralnetworksevenrelatesyntacticandserialordercomputationstocorticalareasandsubcorticalstructures(e.g.,Dominey,Hoen,Blanc,andLelekov-Boissard2003;VoegtlinandDominey2005).However,onlyafewproposalsexistthatrelatesyntaxtodefinedmechanisms,forexample,tocellpopulationssequentiallylinkedintosynchronouslyfiring‘synfire’chainsestablishingserialorder,discreteneuronalassembliesthatstructurethesequentialactivationofotherneuronalpopulations,andtheordereddynamicsofwholebrainareasbroughtaboutbythevariousexcitatoryandinhibitoryneuronalpopulationstheyhouse(e.g.,Pulvermüller1993,2003a;Smolensky1999;Hayon,Abeles,andLehmann2005;Wennekers,Garagnani,andPulvermüller2006).Despitetherelativescarcityofsuchproposals,theprogressthathasbeenmadeshouldnotbeleftunnoticedbylinguists.

AsimplesentencelikeKateglamsitup,depictedinFigure22.1,canbeusedtoillustratesomelocalanddistantlinksbetweentheconstituentsofasentence.Notethatthelinks,whichareallindicatedbyarcsinFigure22.1(a),maybeofdifferentnature.Especially,therearelinksbetweenwords(betweensubjectandverb)andothersthatlinkmorphemeswithinonelargerlexicalitem(theverbstemanditsaffix).Notefurthermorethatthelinkbetweentheverbanditsparticle(glam…up)maybelexicalorsyntactic(seesection22.3.4).Therearenext-neighborsyntacticlinksbetweenthetransitiveverbanditscomplements,theheadnounofthesubjectandtheobjectrealizedaspronoun.Also,theverbanditsaffixcanbeanalyzedasmorphologicallylinked.Nonlocalrelationshipsincludeagreementbetweensubjectnounandverbsuffix(Kate…-s)andverb-particlelinkage(glam…up).

2

Page 295: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 5 of 14

Figure22.1. MorphosyntacticlinksbetweenpartsofthesentenceKateglamsitupforwhichthereislinguisticjustificationandneedforneurobiologicalimplementation.Diagram(a)representslinksimplementedbysyntacticbindingcircuits(DCNAs).Diagram(b)showsasyntactictreestructurerepresenting(simplified)constituentstructureresultingfromcombiningconstructions,plusmorphosyntacticorlexicalrelationshipsnoteasilycapturedbytrees(brokenlines).

Describingnonlocalrelationshipsinatwo-dimensionaltreestructureisnoteasy,becauseprojectionlineswouldcrossifthelinksbetweenconnectedbutdistantconstituentswereaddedtothetree(cf.thedashedlinesinFigure22.1(b)).Linguistictheoriesthereforeintroducedadditionalmechanisms,forexample,percolationofgrammaticalfeaturesthroughthetree(foranintroduction,seeHaegeman1991)tocapturenoun-verbagreement,andmovementorfiller-gaprelationstoaccountfor‘displaced’constituents(thatis,elementsforwhichitisassumedthattheyhavebeenshiftedfromtheircanonicalposition).However,afteraslightrevisionofthemechanismforsyntacticlinkage,theseadditionalmechanismsmaynotbenecessary.

Syntacticlinksbetweenmeaningfullanguageunits(thatis,wordsorlargerconstituents),agreementandothernonlocalsyntacticrelationships,havebeenproposedtobeneurobiologicallygroundedindiscretecombinatorialneuronalassemblies(DCNAs)thatlinktogetherpairsofconstituents(Pulvermüller2003a).DCNAsarediscreteinthesensethattheiractivationtakesoneofseveraldiscretestates.Thesestatesincludeinactivity,fullaccess-activation(orignition),memory-activation,andpriming.DCNAsarecombinatorialastheylinktogether,orbind,specificphrasal,lexical,andmorphologicalcategories.TheneuronpooloftheDCNAwouldlinktotwosetsoflexicalrepresentations,forexample,nouns(oraspecificsubsetthereof)andverbs.DCNAscanalsolinklargerconstructions.Inaddition,DCNAsareneuronalastheyarenotdescribedasabstractconstructsofthemindbutfunctionatthelevelofcorticalcells.Finally,theyareassembliesofcorticalcells,whichmeansthatthecategorieslinkedtogetherjointlyformahigher-order(composite)neuronalstructure.Apartfromlinkingcategoriestogether,typicalDCNAsestablishatemporalorderbetweenthecategorymemberstheybindto.DCNAsthatdonotimposetemporalorder(thusacting,inprinciple,asANDunitsfortwoconstituents)arethoughttojointogetherconstituentswhosesequentialorderisfreeorallowforscrambling(Pulvermüller2003a).SeparatespecificDCNAsarethoughttounderlieproductiveconstructionswithnoncanonicalwordorder(topicalization,wh-questions,etc.).Insuchaneuronalframework,eachofthe(morpho)syntacticlinksindicatedbyarcsinFigure22.1(a)iscarriedbyoneDCNA. Thenonlocalnatureofthesyntacticbindingmechanism,whichmayapplytoadjacentelementsorunitsthatareseparatedintime,allowsDCNAstodealbothwithphrasestructurerelationshipsbetweenadjacentwordsinasentence,ascapturedbytreestructures,andwithagreementmechanismsbetweendistantunits.Italsorenderspossibletheprocessingandrepresentationoflong-distancedependencies.Thelong-distancelinksandthelocalonesarethoughttobemechanisticallyunderpinnedbyexactlythesamemechanism.

3

Page 296: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 6 of 14

IstherereasonforassumingsuchstructuresasDCNAs?Thecentralnervoussystemofhighermammals(BarlowandLevick1965;Hubel1995)andeventhatofinvertebrates(ReichardtandVarú1959;VarjúandReichardt1967)includesalargenumberofneuronalunitscalledmotiondetectorsthatrespondspecificallytoorderlysequencesofsensoryevents.Thesesequencedetectorslinktotwootherneuronalunits,AandB,andrespondspecificallyifthese‘input’unitsareactivatedinagivenorder,A–then–B,butnot,ormuchless,tothereverseorderofactivations,B–then–A.Assequence-sensitiveunitsaresocommoninnervoussystems,itwouldbesurprisingiftheywereunavailableforprocessingcombinatorialinformationinsentences.Sequencedetectorsformorphemesandwordsmayprovideamechanismfordynamicallylinkingconstituentspair-wise,thusrealizingsyntacticlinksthatareeffectiveinprocessingasentence(Pulvermüller2003b).Thelinkageincludesthepreviousactivationofthefirstandsecondwordinastring(e.g.,noun–verb),theresultantignitionoftheorder-sensitiveDCNAandabindingprocessfunctionallylinkingthelatterwiththeactivelexicalrepresentations.Thebindingmechanismisbestthoughtofassynchronizedoscillatoryactivityathighfrequencies(gammaband),asoscillatorydynamicsappeartoreflectlexical(Pulvermüller,Birbaumer,Lutzenberger,andMohr1997;Canoltyetal.2007)andsentenceprocessing(Weissetal.2005;BastiaansenandHagoort2006).

Acriticalquestionremainshowneuronalcorrelatesofmoreschematicconstructionaltemplates(i.e.,abstractsyntacticprocessingunitsoperatingonclassesoflexicalitems)canbebuiltfromsequencedetectors.Inasimulationprojectexploringsyntacticlearning,PulvermüllerandKnoblauch(2009)usedaprestructuredauto-associativememorywithbuilt-insequencedetectorsforeverypossiblenoun-verbpairthatcanbeformedwithasetoftennounsandasetoftenverbs(e.g.,balloonrises;childsleeps;…).The‘grammararea’ofthenetworkdidnotonlyincludetheseprewiredsequencedetectors,but,inaddition,initiallyveryweaklinksbetweenallpairsofsequencedetectors.Bindingofsequencedetectorsintocircuitsoperatingonclassesoflexicalitems(e.g.,nounsandverbs),inotherwords,theemergenceofDCNAs,wasobservedinthesesimulations.Interestingly,dependingonthethresholdofactivation,theseDCNAseitherboundallnounstoallverbs,orprovidedspecificsyntactic-semanticlinkagebetweennounsrelatedtoliving,consciousentities(child,woman,…)andverbsthat,bytheirinherentsemantics,requiresuchsubjects(e.g.,verbsofcognition/preference/…,believes,hates),andbetweennounsreferringtoflyingentities(balloon,bird,…)andflight-relatedverbs(rise,land,…).Thisbindingwascriticallyrelatedtotheavailabilityofsequencedetectorsandauto-associativelinksbetweenthem,whichcouldbestrengthenedbyassociativelearning.Co-activationofsequencedetectorsledtotheemergenceofneuronalaggregatescomprisingsequencedetectorssensitivetosimilarcontexts(seePulvermüllerandKnoblauch2009fordetailsofthesimplemechanismunderlyingthisprocessofgeneralization).Becauseofthestronginternalconnectionswithintheneuronalaggregates,theyactedasahigher-orderdiscretefunctionalunitforsyntacticbinding.Thus,ausage-basedaccountoflanguageacquisition,accordingtowhichgeneralizationsofvariablespecificity(eitherlow-levelschemasormoregeneralrules)emergefromandco-existwithstoredexemplarsofwordstrings(cf.Lakoff1987;Langacker1987,1991;Tomasello2003),hasamechanisticcorrelateinthisneuronalsimulation.

Althoughnotallsyntacticphenomenaareeasilymodeledbytheseemergingaggregatesofsequencedetectors,theydoprovideacandidateneuronalmechanismforoneverygeneraltypeofconstruction:the‘binaryconstruction’[XY],whichcombinestwodistinct(lexicalorsyntactic)categories(Pulvermüller2003b;KnoblauchandPulvermüller2005;PulvermüllerandKnoblauch2009).Suchaneuronalaggregate,byvirtueofthestrongconnectionswhich‘holdittogether,’hasthestatusofanentrenchedunitinthegrammarandisinthatsensenotqualitativelydifferentfromtheentrenchedlower-levelunitsitbinds.Itisalsoimportanttonotethatthetwounits,XandY,thatarelinkedtogetherareform-meaningpairs,withtheimplicationthatsemanticcircuitpartsareinvolvedinthecomputations.CriticalforthedevelopmentofDCNAsisahighdegreeofauto-associativeconnectivityina‘grammararea,’betweensequence-specificunits.Modelsanddataseemtoconvergeontheconclusionthatahighdegreeofauto-associativeconnectivityinleft-perisylvianlanguagecortexmighthavebeenacrucialevolutionarysteptowardhumanlanguage(forfurtherdiscussion,PulvermüllerandFadiga2010).

22.2.4EvidencethattheBrainIndexes‘RulesofGrammar’RatherthanJustFrequencyofCo-occurrence

Intheneurophysiologicaldomain,considerableefforthasbeenspenttofindbrainindexesthatpossiblyindicategrammaticalorsyntacticprocesses.Agoodnumberofstudiescomparebrainresponsestoungrammaticalstringswiththosetogrammatical,well-formedones.Thiscomparisonisparticularlyimportanttolinguistictheory,asmost

4

Page 297: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 7 of 14

theoriesmakeafundamentaldistinctionbetweenstringsrepresentedandsuccessfullyprocessedbythesystemandthosethatarenot,adifferencesometimesthoughttobebestassessableempiricallywiththeuseofthegrammaticalityjudgmenttask.Neurophysiologymayprovidethetoolforlookingatthebrain'sgrammaticalityjudgmentsdirectly.

Arangeofcomponentsoftheevent-relatedbrainpotentialandfield,includingtheearlyleft-anteriornegativity(ELAN)(Neville,Nicol,Barss,Forster,andGarrett1991;Friederici,Pfeifer,andHahne1993),alateposteriorpositivity(P600)(OsterhoutandHolcomb1992;Hagoort,Brown,andGroothusen1993),andtheearlyleft-lateralizedsyntacticMismatchNegativity(sMMN)(Shtyrov,Pulvermüller,Näätänen,andIlmoniemi,2003),areenhancedtoungrammaticalascomparedwithgrammaticalwordstrings(forreview,seeFriederici2002;Hagoort2005;Kuperberg2007;Pulvermüller,Shtyrov,andHauk2009).Corticallocalizationindicatesthattwoareas,theinferiorfrontalgyrusandthesuperiortemporalcortex,housetheneuronpopulationsthatcontributetothesephysiologicalresponsespotentiallylinkedtogrammaticalsequencecomputation(Friederici,Wang,Herrmann,Maess,andOertel2000;Shtyrov,Pulvermüller,Näätänen,andIlmoniemi2003).

However,itisdifficulttodecidewhetherthebrainresponsetoanungrammaticalstringreflectsitssyntacticfeaturesorrathertheconditionalprobabilitywithwhichwordsfolloweachother.Aseveryungrammaticalstringisalwaysalsoaveryrarely(ornever)occurringitem,itisalsopossiblethatthedegreeofexpectedness—orstringprobability—isreflected.Thispoint,whichequallyappliestocomparisonsbetweensimple(andthereforefrequent)andmorecomplex(rarer)sentencetypes(seediscussionbelow),isespeciallyvirulent,asthemajorbrainresponsesdiscussedaspossiblereflectionsofungrammaticalityandsyntacticprocessingload—ELAN,P600,andsMMN—aresimilarinlatenciesanddistributionstowell-knowncomponentsoftheevent-relatedpotentialthatreflecttheprobabilitywithwhichstimulioccurwithinanexperiment.Respectableargumentshavebeenbroughtuptodefendthepositionthatcomponentselicitedbyungrammaticalsentencesaredifferentfromtheirnonlinguisticsisters—forexample,thatthesyntacticP600isdifferentfromtheP300(OsterhoutandHagoort1999)—butthisdoesnotproveaspecificallysyntacticroleoftherespectivecomponent.Itmaystillbethatsequentialprobabilitiesinlanguageuse,ratherthanthepresenceorabsenceofanunderlyingdiscretecombinatorialrule,determinetheemergenceofasyntacticELAN,P600,orMMN. Directevidenceforaneuronalcorrelateofrulesisthereforenotprovidedbytheworkreviewedsofarasgrammaticalhigh-probabilitystringswerecomparedwithungrammaticallow-probabilityones.Althoughthenumberofoccurrencesofgrammaticalandungrammaticalstringscanbebalancedexactlywithinagivenexperiment,thegeneraloccurrenceprobabilityofasentence,sentencetype,orrelevantsectionthereof,isusuallyhigherthanthatofasimilarasyntacticstring.Asbothgrammaticalityandsequentialprobabilitymayhaveacorrelateinthebrainresponse,itisnecessarytoaskwhethertheseeffectscanbeseparated.

Importantinformationaboutgrammaticalprocessesinthebrainmaycomefromstudiesofdifferenttypesofcorrectsentencesdifferingintheirgrammaticalstructure,wordorder,ormemoryrequirements.However,totheextentthatoneiswillingtoacceptthatperformanceshapesgrammar(Hawkins1994),itisnaturallydifficulttoseparateeffectsofgrammaticalstructurepersefromthoseofstringfrequency.Asalreadymentioned,ifsentenceswith‘canonical’and‘noncanonical’wordorderarecontrasted,theinfluenceofstringtypefrequenciesmayplayarole.Forexample,noncanonicalpassiveorobject-relativesentencesactivatedleftinferiorfrontal/premotorandsuperiortemporalareasmorestronglythancanonicalactiveorsubject-relativesentences(Just,Carpenter,Keller,Eddy,andThulborn1996;Kinno,Kawamura,Shioda,andSakai2008).Similarly,theposteriorinferiorfrontalareawasmorestronglyactiveforsubordinateclauseswithaninitialaccusativenounphrase(NP)referringtononanimateobjectsthanforsimilarstructureswithaninitialdativenounphraseincludingananimatenoun(Greweetal.2006);notethattheformerarerarerthanthelatterand,interestingly,grammaticalityratingsreflectthisfrequencydifference(Keller2000).Togetherwithdataongrammaticalityviolationsmentionedabove,theseresultsongrammaticalsentencetypesarecompatiblewiththeideathatsyntacticcircuitsarehousedininferiorfrontalcortex,withanadditionalpotentialcontributionofsuperiortemporalareas.Interestingly,asKeller'sdatademonstrate,alinkbetweengraduallymoreorlesscanonical(andthuscommon)syntacticstructureandgrammaticalityratingsexists(Keller2000).However,thisfactstillleavesuswiththecriticalquestionofhowtoseparatestringprobabilityandgrammaticalityperse.

InarecentexperimentusingthesyntacticMismatchNegativity,sMMN,asthedependentvariable,anattemptwasmadetodissociategrammaticality-andprobability-relatedprocesses.Allcriticalsequencesofwordswerepresentedwithequalprobabilityintheexperiment,butbothgrammaticalityandtheprobabilitywithwhichthese

5

Page 298: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 8 of 14

stringsareusedinlanguagevaried.Criticalwordswereidenticalindifferentconditionssothatfrequency-relatedorlexical-semanticfactorscouldnotaccountforanydifferences.Determiner-nounagreementinGermanwasusedasthesyntacticfeatureofinterest,asitrepresentsasyntacticlinkwithoutanyobvioussemanticimplication(i.e.,inGermangrammaticalgenderisformostlexicalitemsnotsemanticallymotivated).Asinpreviouswork,grammaticalstringsfrequentlyusedinlanguagewerecomparedwithrareungrammaticalones.However,inaddition,grammaticallycorrectbutextremelyrarestringswerealsopresented.Theoppositelogicalpossibility—ungrammaticalstringsthatoccurfrequently—isnotavailableinthelanguage;descriptivegrammarclassifiesfrequentlyoccurringstringsasregular.Forthecriticalcondition—rarebutgrammaticalstrings—twocompetingpredictionsareavailable.Cognitivescientistsandcomputationalmodelershaveclaimedthatrules(inthesenseofdiscretecombinatorialunitsor‘constructions’)donotexistinneuralsystems,brainsincluded,andthatprobabilitymappingisthemechanismthatexplainsthehumanabilitytostringwordsinsentences(Elmanetal.1996;McClellandandPatterson2002).Aprobabilitymappingapproachwithoutdiscreteneuronalunitspredictsthatararegrammaticalstringelicitsabrainresponsecomparabletothatofsimilarlyrareungrammaticalstrings.However,ifthebrainusesadiscreterulemechanismtodistinguishbetweengrammaticalandungrammaticalstrings,thismechanismshouldbeinvolvedtothesamedegreeintheprocessingofrareandcommongrammaticalwordstringswiththesamesyntacticstructure,butitshouldfailtoprocessungrammaticalwordsequences.Therefore,probabilitymappingandruleaccesstheoriesleadtodifferentpredictionsonthebrainresponsetoraregrammaticalstringsofwords.

Intheexperiment,thesyntacticenhancementoftheMMNwasfoundtorareungrammaticalstringscomparedwithcommongrammaticalones,ashadbeenreportedinarangeofpreviousexperiments(foranoverview,seePulvermüller,Shtyrov,Hasting,andCarlyon2008).Relevantcorticalgeneratorsarebeingsparkedearly(60–200ms)ininferiorfrontalandsuperiortemporalcortex.Critically,grammaticalbutrarephrasesledtoanMMNsimilartothatofcommongrammaticalstrings(Figure22.2)clearlysupportingtherule-based(butnotprobability-mapping)account.

Clicktoviewlarger

Figure22.2. Event-relatedmagneticbrainresponses,heretheMismatchNegativity(MMN),reflectgrammaticalityofwordstringsbutnotsequentialprobability.ThesyntacticenhancementoftheMMNoccurredtorareungrammaticalstrings(*dieMut),butnottosimilarlyraregrammaticalsequences(derWut)ortofrequentlyoccurringgrammaticalstrings(dieWutandderMut).Statisticaldata—meansandstandarderrors—ofthegrammaticalityeffectaredisplayed(PulvermüllerandAssadollahi2007).

ThesMMNseemstobehaveasifitwereanautomaticgrammaticalityindex(Pulvermüller,Shtyrov,Hasting,andCarlyon2008).TheneurophysiologicalrulecorrelaterevealedbythesMMNprovidessupportfortheruleaccesstheory(PulvermüllerandAssadollahi2007)andisconsistentwiththeexistenceofdiscretecombinatorialneuronalprocessorsgeneratingthediscreteneurophysiologicalresponses.Rules,interpretableinthesenseofschematicgrammaticalconstructions(e.g.,Langacker1987,1991;Goldberg2003),appeartobeeffectiveinthebrain'sgrammaticalityjudgments.

22.3.IsConstructionGrammarNeuroscientificallyPlausible?

Wearenowatapointwherewecanevaluatetheneuronalplausibilityofexistinggrammarframeworks.Ourfocuswillbeonthebroadfamilyofcognitive-linguisticandconstructionistframeworksandsomeoftheircentraltenets.

6

Page 299: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 9 of 14

22.3.1AreLinguisticAbilitiesanExtensionofOtherCognitiveAbilities?

Immanenttotheideaofcognitivelinguisticsistheassumptionthatourlinguisticendowmentisanaturalextensionofmoregeneralcognitiveabilities.Thisassumptioncontrastswithsuggestionsingenerativetheorythatanencapsulated,tosomedegreefunctionallyindependentmoduleunderlieslanguageuse.Wealreadysummarizedneuroscienceevidencethatgivessubstancetothefunctionallyinterwoven,nonmodularnatureoflanguage,especiallythefunctionallinksbetweenthebrainsystemsforlanguage,perception,and,critically,action.Inthiscontext,itappearscorrecttosaythatlanguageis‘wovenintoaction,’therebyconfirminganoldpostulate,whichactuallyemergedfromthephilosophyoflanguage(Wittgenstein1953)andlaysthegroundforcognitive-linguisticperspectives.

Furthersupportforacognitive-linguisticperspectiveisprovidedbytheputativerelationshipbetweenbiologicalmechanismsofsequence-sensitivityinarangeofdifferentnervoussystemsandcombinatoriallinguisticmechanisms.Eventhoughthelatterwouldconstituteafurtherdevelopmentoftheformer,theycanbeviewedasbeingbuiltupon,andthereforegroundedin,sequencingmechanismssharedbymanyanimals.Asafurtherimportantpoint,wehavementionedneurobiologicalmechanismsforabstractingrulelikecombinatorialmechanismsfromlinguisticutterancesand,therefore,wesubmitthat‘syntactic’structuresneednotbepre-installedinthebrain;theycanarisegraduallyonthebasisofcombinationsofwordsthatthelanguageuserencounters.Thissupportstheusage-basedoutlookonlanguagelearningsharedbycognitive-linguisticandconstructionistapproaches.Necessaryforbuildinghigher-ordercombinatorialunitsinthebrainareassociativelearningmechanismsofaspecialkind(whichmay,asmentionedinsection22.2.3,requiretheirownspecificbrainendowment).Finally,mechanismsofrecursionarealsoanecessaryingredientoflanguage.Thesemechanisms,whichwemerelytouchedupon(cf.footnote4),havebeenpostulatedtoemergefromgeneralmemorymechanisms(Pulvermüller1993,2010a).Weimmediatelyseethatarangeofassumptionscruciallyunderlyingcognitive-linguistictheoryappearwell-supportedbyaneuroscienceperspective.

22.3.2AreSyntaxandSemanticsTwoSidesoftheSameCoin?

Grammartheoriesinthecognitive-linguistictradition,incontrasttomostotherlinguistictheories,viewgrammarasanintegrationmachineryforformandmeaning(e.g.,Lakoff1987;Langacker1987,1991;Bresnan2001).Suchaviewisinlinewith,andfindscross-validationfrom,thebrain-mechanisticmodelandcorroborativeevidenceoutlinedabove.

Aswehaveseen,lexical-semanticcategories(e.g.,moving-objectnouns,verbsofmotion),arenotonlygroundedinactionandperceptionsystemsofthebrain,overandabovesuchdirectneurobiologicalembodimenttheyarealsomanifestinthehigher-orderbindingcircuits(DCNAs)theyconnectwith.Theseunitsprovideacombinatorialembodimentoflexicosemanticcategoriesinneuronalcircuits.Inturn,thesecombinatorialunitsdeterminethegrammaticalstringsinwhichthelexicalunitsregularlyoccur.Syntaxandsemanticsarethereforeintricatelyconnectedwitheachother:the‘syntactic’DCNAs,afterlicensingthetogethernessofconstituents(inagivenorderornot),canregulatewhichkindsoflexicalitemscancombinewitheachother,notjustinasyntacticallybutalsosemanticallyfelicitousway.Clearly,then,theamalgamatednatureofsyntaxandsemantics,whichistypicalofcognitive-linguisticandconstructionistapproachestogrammar,canbemodeledsuccessfullyinaneurobiologicalframework.

Asafurtherfeature,weshouldre-emphasizethatDCNAsconnectlexicosemanticcellassemblies,whichare,inturn,conceptualizedasform-meaningrepresentations.ThemechanismsoutlinedheremayhelptoneurobiologicallyunderpinthetightlinkbetweensyntaxandsemanticspostulatedinthecontextofCognitiveandConstructionGrammar.

22.3.3CanConstructionsChangeValency?

Constructionistframeworksmakeverystrongclaimsaboutthefusionofsyntaxandsemantics:constructionaltemplatesarestoredjustlikewordsandtheycanhavemeaningthemselves,upandabovethemeaningcontributedbytheircomponentparts.AfrequentlydiscussedexampleisprovidedbytheCausedMotionconstructionanditscapacitytoinvolveanotherwiseintransitiveverb,forexample,Shesneezedthefoamoffthecappuccino(Goldberg2006a:42;Boas,thisvolume):themessagethatanobjectwascausedtomovealonga

Page 300: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 10 of 14

certainpathisthenmostconvincinglyanalyzedasasemanticcontributionmadebytheconstructionitself.Thequestion,now,iswhethertheclaimthatconstructionscanbeinherentlyandindependentlymeaningfulisviablewithinaneuromechanisticapproachtolanguage.

Syntacticbindingcircuits(DCNAs),theneurobiologicalanaloguetoconstructionaltemplates,havebeenshowntobe‘suffused’withmeaning:theyemergefromclassesoflexicalitemswithsimilarcombinatorialproperties,whosemembers,infact,cansharesemanticpropertiesapartfrompurelysyntacticones(PulvermüllerandKnoblauch2009).Assneezeissemanticallysimilartoblow,hit,andpushinsofarastherelevantactions(somemorereadilythanothers)cancausemotion,semanticprimingbetweenthesemanticallyrelatedwordsmayleadtocoactivationofthesequencedetectorfortheaccusativecomplement(thecappuccino)withtheone-placeverb(sneeze);thewordsneezemaythereforeoccasionallybeplacedinathree-argumentcausedmotioncontextandbindtotherespectivesyntactic-semanticsequencedetector(s).Therefore,DCNAsandsemanticprimingcanprovideamechanisticaccountforvalencechangeofverbsdrivenbysemanticfactors.

Inessence,thebrainequivalentofaCausedMotionconstructaroundsneezewouldbeasyntacticbindingcircuitwhichisstilllargelyconfinedtosemanticallyspecificsubstitutionclasses(verbsofforceexertiononanobject,movableobjectnouns,andpath-expressingitems)butwhichcanbeseentooccasionallyconnecttoelementsoutsidetheseclasses.Intheneurobiologicalframeworkintroducedhere(section22.2),amechanismcanbeenvisagedbywhichsemanticaspectsoftheverbsneeze(especiallytheideaofforcefullyblowingoutair),togetherwithaspectsoftheconversationcontext(especiallythoseevokingaforce-dynamicscenarioinwhichtheblown-outairsetssomethingelseinmotion)mayprimethree-placeverbssuchasblow,push,orshoot,whichmayactivatetherelevantthree-placeconstructionalframe.Asexplainedabove,thisconstellationofbrainactivitiesmayinitiallyleadtothecoactivationoftheverbsneezewiththeDCNAsforblowandthustothesentencementioned.Ultimately,suchcoactivationofaone-placeverbandDCNAsassociatedwithotherverbsmayresultintheformerone-placeverbbeingsubsumedintoathree-placeverbcategoryandDCNAset,aprocesswhicharguablyhasbeenaccomplishedfortheverblaughasusedinthesequencelaughNPoffthestage(fordiscussionofthisexample,seeKay,thisvolume).

22.3.4DoLexiconandSyntaxFormaContinuum?

Arecurringclaimofconstructionistapproachestogrammaristhatsinglelexicalitemsandlarger,moreabstractconstructionsarenotqualitativelydifferent.Beingform-meaningpairsinasinglelarge‘constructicon,’theyonlydifferfromeachotherintheirdegreeofphonologicalspecificityandinternalcomplexity.Again,weshouldaskthequestionwhetherabandoningasharpdistinctionbetweenlexiconandsyntaxisneurophysiologicallyjustified.

Withintheneurobiologicalframeworkpresentedhere,bothindividuallexicalitemsandtemplatesfortheircombinationarestoredinthebrain.Inadditiontothese,thebrainalsostoresrepresentationsforspecificwordsequencesbywayofsequencedetectors,aswearguedinsection22.2.3,whichfitswellwithusage-based,constructionistproposalsthatspecificwordsequencesarestoredalongsidemoreschematicpatternsemergingfromthem.Inthissection,however,weareparticularlyconcernedwiththeequaltheoreticaltreatment,asconstructions,ofindividualwordsandlargertemplatescontainingslotsintowhichthesecanbefitted.Ofcourse,noConstructionGrammarianwouldexpectthesetwoextremecasestohaveidenticalinstantiationsinthebrain,buttheconstructionisttenetthattheyarenonethelessontologicallysimilarobjects(namely,thattheyareform-meaningunits)leadstotheassumptionthat,intermsofthebrain,theymakeuseofasimilarmechanism.

Clicktoviewlarger

Figure22.3. Earlybrainresponsetoexistingcomparedwithnonexistingphrasalverbs.Theoddball

Page 301: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 11 of 14

responseisenhancedtoparticlesembeddedinexistingphrasalverbs(e.g.,upinheatupordownincooldown)comparedwiththesameparticlesembeddedinnonexistingphrasalverbs(e.g.,upin*coolupordownin*heatdown).Thedivergencebetweenresponsestorealandinfelicitousverb-particlecombinationsishighlightedarounditspeak.Theenhancedresponsetoparticlesinthecontextofa‘legal’verbissimilartothatofasoundorsyllablepresentedafterasyllablewithwhichitformsanexistingword.Itisopposedtotheresponseofawordwhichfollowsanotherwordmerelyinconformitywithrulesofsyntacticandsemanticwell-formedness(Cappelle,Shtyrov,andPulvermüller2010).

Intheneurobiologicalperspectiveadoptedhere,theunderlyingmechanismsofsinglelexicalitemsandlargerconstructionsare,however,fundamentallydifferent.Whereaslexicalunitsarestoredinthebrainasword-relatedneuronalcircuitslinkingformandmeaning,DCNAsforcombinatorialbindinglinkgroupsoflexicosemanticunitstoeachother.Crucially,theneurophysiologicalindexesforthesetwotypesoflinguisticbinding,lexicalandsyntactic,aredistinct(PulvermüllerandShtyrov2006).Asregardslexicalbinding,thebrainindexknownastheMismatchNegativityisenhancedforstoredpreviouslyencounteredlexicalitems,butrelativelyreducedforpseudowords(i.e.,for‘incorrect’syllableorphonemesequences),aswesawinsection22.2.2.Bycontrast,thesyntacticMismatchNegativityisenhancedforincorrect,ungrammaticalstrings,butrelativelysmallforcorrectsentences.Inotherwords,neurophysiologicalresearchcouldobjectifythelexicon-syntaxdistinction,thusconfirmingtheassumptionthata‘lexicon’ofobjectsanda‘grammar’ofbetween-objectlinksaremechanisticallydifferentinourmindsand,importantly,inthebrain.Thisneuroscienceevidencecannotbeexplainedbyapproachesthatpostulateonlyonetypeoflinkbetweenmeaningfullanguageunits(beitcalled‘lexical,’‘constructional,’or‘symbolic’).Theneurophysiologicalevidencesitsmorenaturallywithapproachesdistinguishingbetweenlexiconandsyntax,or,intheconstructionistframework,withviewsongrammarthatacknowledgefundamentaldifferencesbetweenconstructiontypes,especially(1)word-level,lexicallyspecificconstructionsvs.(2)phrasalorclausalconstructionaltemplatesandtheirsubstantiveinstantiations(seeJackendoff2002a,thisvolume,forarareexampleofaconstructionistwhopointsoutdifferencesingrammaticalorganizationatthemorphologicalvs.syntacticlevels).

Theopposinglexical(‘word-level’)andsyntactic(‘above-word-level’)signaturesoftheMismatchNegativityandrelatedso-called‘linguisticoddballresponses’openanewfieldforneurophysiologicalresearchinlinguistics.Iftheseindexesdoindeedreliablyrevealtwotypesoflinkagebetweenmeaningfullanguageelements,theycanbeusedtoaddresslong-standinglinguisticdebates.Oneexcitingdebatesurroundsthestatusofverb-particlecombinations(e.g.,carryout,makeup,showoff,etc.)aseithersyntacticallyconnectedsequencesofwordsormorphologicallylinkedsinglewords.Inthelattercase,thesomewhatparadoxicalconceptemergesofawordthatcanappearinseparateparts,distributedasadiscontinuousconstituentoverasentence(e.g.,Hemadetheentirestoryup).Arecentstudyaddressingthisissuefoundthelexicalpatternofneurophysiologicalresponses:highlyexpectedsequences(heatup,cooldown)elicitedenhancedbrainresponsescomparedwithinfelicitousones(*heatdown,*coolup).Similartosinglewordscomposedoftwomorphemes,theMismatchNegativityandoddballresponsewereenhancedtothewell-formed,expectedstringtypes(Figure22.3)(Cappelle,Shtyrov,andPulvermüller2010).

Thisresultsuggeststhat,fromabrainperspective,phrasalverbs(atleast,frequentlyoccurringones)seembestanalyzedaswordlikestoreditems.Thismightnotbesuchasurprisingfinding,giventhatphrasalverbsarenotoriousfortheiridiomaticity,whichnecessitateslexicalstorage.However,wefoundlexical-responsepatternforcombinationswhicharesemanticallymotivated(e.g.,heatup)orevenfullytransparent(e.g.,riseup).Thisfindingisinlinewiththecognitive-linguisticview(e.g.,Langacker1987)thatcommonsequencescanbestoredeveniftheyarefullycompositional,thatis,eveniftheirformandmeaningfollowsfromwhatisavailableintherestofthegrammar.

Tosumupandconcludethissection,word-levelunits(‘lexicalitems’),whichcanconsistofmorethanonegrammaticalword(definedasastringoflettersseparatedbyspacesinwriting),areverydifferentthings,inneuromechanisticterms,fromabove-word-levelunits.Therefore,ConstructionGrammarianswoulddowelltomoreexplicitlyacknowledgethedistinctionbetweenthesetwotypesofunit.Infact,thisdistinctionisimmanenttoConstructionGrammar,totheextentthatsubstantiveconstructionsandschematictemplatesaresituatedatoppositeendsofacontinuum.Ourpointis,however,thatspeakingofacontinuumisquitemisleading,asitdoesnotdojusticetotheradicallydifferentanddistinctunderlyingbrainmechanisms.

22.4.SummaryandConclusion

7

8

Page 302: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 12 of 14

Discussingaspectsoftheputativebrainbasisofmeaning-bearingunits,rangingfromphonemes,morphemes,andwordsoverwordsequencesandschematicsyntacticstructuresto‘rulesofgrammar,’wehavehereaddressedtheneurobiologicalfoundationoflinguisticsandtheputativebeneficialimpactofresearchbringingtogetherlinguisticsandneuroscience.Lexicalentriesareinstantiatedinthebrainasdistributedneuronalassembliesspreadoutoverfronto-temporallanguageareasfurtherextendingintomotorandsensoryregions.Thisisnotjustthecaseforwordsreferringtophysicalobjects,concreteactions,andbodilysensationsbutalsoforthoseusedtocommunicateaboutmoreabstractideas.Recurrentwordsequencesandmoreabstractconstructionsgeneralizingoversuchspecificsequencesarealsostoredinthebrain,possiblybyprocessesdistinctfromwordstorageasweargued.Inabrain-inspiredmodelofgrammar,theirstoredrepresentationstaketheformofsequencedetectorsandsequencedetectoraggregates,DCNAs,respectively.Rulesofgrammarcanthenbeviewedasemergentpropertiesofmultiplestoredsequencesthatareboundtogetherinassembliesasaconsequenceofstringsegmentsbeingabletosubstituteforoneanotherinthesamestructuralslotonthebasisoftheirsemanticcommonality.Suchsemanticallycoloredcombinatorialassembliescanfurthermergetogethertoformbraincorrelatesofhighlyabstractconstructions,suchasthoselicensingthemeresequenceof(any)nounand(any)verb.

Apartfromdiscussingthepossibleemergenceofgrammarrules(or‘constructions’)inthebrain,wearguedthatpositingtheminthefirstplaceisnecessarytoaccountfortheexperimentalobservationthatthebrainrespondsdifferentlytogrammaticallyacceptablephrasesthantoungrammaticalones;crucially,unlikewhatmanyneuralnetworkmodelerssuggest,thebrain's(relativelyreduced)immediateresponsesignalingacceptanceisnotdependentonstringprobability,sincethisresponsetoagrammaticalandcommonsequenceisnotsignificantlydifferentinstrengthfromtheresponsetoanequallygrammaticalbutveryuncommonsequence.Rule-conformityratherthanstringprobabilityiswhatexplainstheobservedbrainresponses.

WethenprovidedananswertothequestionwhetherCognitiveandConstructionGrammarframeworksareneuroscientificallyplausible.Ouranswercanbesummedas‘Yes,withabut.’Morespecifically,itseemscorrecttotreathumanlanguageasanextensionofother(andolder)neurocognitivefeatures,suchasthebrain'sgeneralabilitytobuildandestablishlinksbetweenobjectandactionrepresentationsandregistersequentialevents.Italsoseemscorrecttotreatsyntaxandsemanticsasinterwoven,assyntacticcircuitsinthebrainemergefromsequenceswhosesegmentsbelongtogethersemantically.Itmayalsobethattherearebraincorrelatesofschematicbutstillinherentlymeaningfulconstructionalschematawithsemanticfeaturesboundtothem,sothatasemanticcontributionofaconstructionaltemplatecouldbeexplained.Finally,althoughthebraincorrelatesofsyntacticconstructionscanrangefrommedium-leveltohighlyabstract(therebysupportinginpartthelexicon-syntaxcontinuumtenet),andinspiteoftheexistenceofcertainmultiwordlexicalitems,wewouldwarnagainstatotalabolitionofalexicon-syntaxdistinction,aswordsand(evencommon)syntacticsequencesofwordstriggerdifferent,infactopposite,brainresponsesrelativetotheirerroneouscounterparts(pseudowordsandillegitimatewordcombinations,respectively).Itmaythereforebeadvantageoustokeepobservingthetraditionaldistinctionbetween,ontheonehand,word-level(‘lexical’)itemsand,ontheotherhand,instantiationsofabove-word(‘syntactic’or‘phrasal’)templatesinthegrammar.Interestingly,theneurophysiologicalmanifestationsofthelexicon/syntaxdistinctionmayhelpaddressspecificallylinguisticquestionswithneurosciencetools,aswehereexemplifiedbyhighlightingthecaseofverb-particlecombinations.Needlesstosaythatthislatterstrategyrequiresafirmlinkbetweenthelanguageofneuronsandthatusedtospeakaboutlanguage,thatis,aneurobiologicallinguistictheorythatprovidesthetranslationandtherebyconnectionofthetwoworlds.

Thistextislargelybasedonanarticlebythefirstauthor,whichappearedinBrainandLanguage112(2010),167–79.WewouldliketothankValentinoBraitenberg,AdeleGoldberg,AndreasKnoblauch,MarionOrmandy,MarkSteedman,KambizTavabi,aswellastheeditorsofthecurrentvolume,fortheircomments,input,andhelpatdifferentstagesofthiswork.SupportedbytheMedicalResearchCouncil(UK)(MC_US_A060_0034,U1055.04.003.00003.01).

Notes:

(1.)OuruseofembodimentcouldeasilybemisinterpretedgiventhevarietyofConstructionGrammarknownasEmbodiedConstructionGrammar(foranintroductionofwhich,seeBergenandChang,thisvolume;Broccias,thisvolume).Ifthetermembrainmentexisted,wewouldhaveusedit.

Page 303: The Oxford Handbook of Construction Grammar

Brain Basis of Meaning, Words, Constructions, and Grammar

Page 13 of 14

(2.)Brainactivationdifferencesariseearly,within200ms,notonlybetweenexisting(hencemeaningful)wordsandmeaninglesspseudowordsbutalsobetweenwordsfromdifferentsemanticcategories,forexample,betweenthealreadymentionedarm-andleg-relatedactionwordsorbetweencolor-andform-relatedwords(Pulvermüller,Hummel,andHärle2001;MoscosoDelPradoMartín,Hauk,andPulvermüller2006).

(3.)Differencesbetweenlexicalandsyntacticlinkagearediscussedinsection22.3.4.

(4.)SyntacticlinkagebywayofDCNAsisbutoneimportantgrammarmechanism.Thehumanbrainmayincludeadditionalmechanisms,forexample,toallowforrecursionofrulesofgrammar,morespecificallyofthetypethatleadstonestedstructuresandlong-distancedependencies(center-embedding),whichhavebeenarguedtorepresentahallmarkofhumanlanguage.Putativeneuronalmechanismsimplementingsuchapushdownstoreinthebrainhavebeendiscussedelsewhere(Pulvermüller2003a,2010a;Buzsáki2010).AneuronalgrammarmodelintegratingneuronalpushdownstoreandsyntacticbindingbymeansofDCNAshasbeenlaidoutinPulvermüller(2003a).Inthismodel,theelementaryclause-andsentence-internallevelofprocessingisdealtwithbysyntacticsequencedetectorsandthemorecomplexsimultaneousandrecursiveprocessingofmorethanoneclause,forexample,inthecaseofembedding,isdealtwithbythepushdownmemorycomponent.

(5.)Inthiscontext,‘discrete’meansthattherelevantmechanismlinkingtogetherlexicalitemsintoalargerconstructionisonethateitherappliesornot,notallowingformuchgradualvariationbetweenthebinaryyesornodecisions.Incontrast,aprobabilisticapproachimpliesgradualvariationofstringprobabilitiesand,correspondingly,acceptability.

(6.)Thisdoesnotruleoutapossiblephysiologicaldifferencebetweenbrainprocessessparkedbyrareandcommongrammaticalstrings.Asthebrainissosensitivetoconditionalprobabilities(Näätänen,Gaillard,andMäntysalo1978;Donchin1981;KutasandHillyard1984;Näätänen,Tervaniemi,Sussman,Paavilainen,andWinkler2001),itwouldnotbesurprisingtofindaphysiologicalcorrelatefortheminanexperimentinvestigatingwordsequences.

(7.)Thisdistinctionisnottobeequatedwithonebetween‘atomic’and‘complex’constructionsinCroftandCruse's(2004)sense,becausenotallword-levelconstructionsareatomic,manyofthemdisplayinginternalmorphologicalstructuring.Notethatthe‘lexical’responserelatedtotheMismatchNegativityisalsoenhancedwhenanaffixislinkedwithanothermorphemeinamorphologicallycomplexword,relativetoitssizetothesameaffixincludedinapseudoword(Shtyrov,Pihko,andPulvermüller2005).

(8.)Aninterestingquestion,ofcourse,relatestotheneuralmechanism(s)underlyingso-called‘constructionalidioms,’whichincludebothlexicalitemsandopensyntacticpositions(e.g.,verbone'sheadoff).Wesurmisethattheseinvolvebothlexicalandsyntacticlinks,whichaswehavearguedhereareofadifferentnature.

FriedemannPulvermüllerFriedemannPulvermüllerworkedasProgrammeLeaderattheMRCCognitionandBrainSciencesUnit,Cambridge;recentlyheacceptedthepositionofProfessorofNeuroscienceofLanguageandPragmaticsattheFreieUniversitätBerlin.Hediscoveredthatthebraindiscriminatesearlybetweenmeaningfulwordsandsenselesspseudowords,andbetweengrammaticalandsemanticwordkinds;healsoreportedearlybrainactivationpatternsindicatingthemeaningofwordsandsentences.Hehaspublishedfourbooksandover200articles,puttingforwardthefirstneurobiologicalmodeloflanguageintegratingneurobiologicalprincipleswithlinguistictheoryandofferingmechanisticnervecellcircuitsunderpinninglanguageinthehumanbrain.Neuroscienceinsightswerebroughttofruitbyhisworkinthedevelopmentofnewtreatmentproceduresforpatientswithpost-strokeaphasia.

BertCappelleBertCappelleisalecturerofEnglishlinguisticsattheUniversityofLilleNorddeFrance.Hehaspublishedarangeofjournalarticlesandbookchaptersonverb-particleconstructionsinEnglish.Inaddition,hehascollaboratedonresearchprojectsinthecoregrammarareasoftense,aspectandmodality.Hislonger-standingresearchinterestsincludethelinguisticrepresentationofmotionandchangeofstate,andthetensionbetweenconventionandinnovationinlanguageuse.

YuryShtyrovYuryShtyrovisProfessor,SeniorScientistandtheHeadofMEGLaboratoryattheMedicalResearchCouncil’sCognitionandBrainSciencesUnitinCambridgeandcurrentlyalsoaco-directoroftheCognitiveBrainResearchUnitattheUniversityofHelsinki.Hisresearchinthefieldofcognitiveneuroscienceiscentredoninvestigatingneurophysiologicaldynamicsoflanguageprocessinginthehumanbrain.Hisparticularcontributiontothisareaisindescribingearlyandautomaticstagesofneurallanguageprocessingandinestablishingfunctionalparallelismfortheneuralprocessingofdifferentlinguisticinformationtypes.Hehasauthoredandco-authoreddozensofarticlesinleadingneuroscientificjournalsandbookchapters.

Page 304: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 1 of 14

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0023

PrinciplesofConstructionalChangeMirjamFriedTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

ThischapterconsiderstheapplicationoftheprinciplesofConstructionGrammartolanguagechange.ItdescribesaparticularchangeinamorphologicalconstructionofOldCzechanddiscussessomeofthewaysinwhichconstructionsmaychangeinternally.ThechapterexplainstheconceptofconstructionalizationandestablishesitsconnectionwithConstructionGrammar.Ithighlightsthegradualnatureofconstructionalchange,themicro-stepsinvolvedatdifferentconstructionallevels,andtheimportanceofcontext.

Keywords:languagechange,ConstructionGrammar,morphologicalconstruction,constructionalization,constructionalchange,OldCzech

23.1.Introduction

Amajorchallengeinlinguisticanalysisconcernsthepossibilityofcapturingtheinherentlydynamicnatureoflinguisticstructureandthegradualnessofgrammaticalchange,whilesatisfyingthegeneralrequirementofsystematicityanddescriptiveandexplanatoryadequacy.Partofthechallengeisrootedinthebasic,obvious,butanalyticallynontrivialfactthatlanguageisalwayssituatedincontext,whichalsoimpliesaconnectionbetweenlinguisticchangeandlanguageuse.Awarenessofthisinterdependencegoeshandinhandwithamarkedshiftintheorientationofpresent-daydiachronic(and,morebroadly,variationist)research:insteadoffocusingondescribinginstancesof(inherent)change,seenasaffectingmerelydiscreteunitsofacontext-independentgrammaticalsystem,greateremphasisisnowputonsearchingforgeneralizationsoverthoseinstancessothatwecanbegintoexploreandexplainrecurrenttypesofchange.Thisshiftinginteresthighlightsatleasttwomutuallydependentpoints:(1)theneedforestablishingaworkableandplausibleexplanatorymodelthatcanaccommodatethegradientnatureoflanguagechange,butalso(2)toacknowledgetherelevanceofdiachronicanalysesanddiachronicevidencefordevelopinganadequatetheoryoflanguageingeneral.

Anintersectionofalltheseperspectivesandresearchagendasismostprominentlyassociatedwithgrammaticalizationresearch,which,inturn,hasbeenincreasinglydrawingattentiontoConstructionGrammarasanapproachthatmightbeparticularlywellsuitedtothegoalofanalyzingandrepresentingthecontextuallygroundeddevelopmentofnewgrammaticalfunctionsandpatterns.SincesystematicandsufficientlydetailedexplorationsoftherelationshipbetweengrammaticalizationresearchandConstructionGrammar(andhowtheycaninformeachother)areonlyintheirbeginningstages,thereisarichinventoryofissuestobeworkedout,withnewonesstilllikelytoemergeasourappreciationofthechallengesdeepens.Atpresent,thefollowingsubsetofinterrelatedquestionstendstoattractthemostfocusedattention:theroleofcontextingrammaticalizationaccounts;theintegrationofsemanticandpragmatictriggersinconstructionalrepresentations;thesenseinwhichitmightbetruethatconstructionsarethelocusofchange;thestatusofanalogyinconstructionalanalyses;addressingthegradualnessofchangeinallitscomplexity;andtheproblemofcapturingpolysemyrelations,asaninevitableby-productofgrammaticalizationprocesses.

Page 305: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 2 of 14

Theseissuesalsoframethediscussioninthischapter.Mypurposeisnottosurveythetypologyoflinguisticchangesvis-à-visconstructions,nortoreviewmetaphor-basedaccountsofgrammaticalization,nortoaddressargument-structurerelatedchanges(which,sofar,havemotivatedthebulkofconstructionaldiachrony).Instead,thegoalistofocusonarticulatingthestructureandmechanicsofanalogy-basedprocesses,whichiswheretheconceptof(relatively)schematicgrammaticalconstructionsbecomesparticularlysalient.IbelievethatthisisalsoadirectioninwhichfuturegrammaticalizationresearchislikelytogrowandadirectionthatwillcontributetowarddevelopingConstructionGrammarintoareliabletoolfordiachronicanalysis,includingitspotentialusefulnessforintegratingtheconcernsofevolutionaryapproachestoculturallygroundedlinguisticchange,aslaidoutpersuasivelyinSteels(2012,cf.alsoSteels2007).

Thegrammaticalizationframeworktracesitsconceptualoriginstocertainobservationsabouttheroleofcollocationalpatternsingrammaticalchange(cf.Humboldt1825),thussuggestingquiteearlyontheimportanceofsyntagmaticrelationsinlanguagechange.ThisideawaseventuallyexplicitlyarticulatedbyLehmann(1995:406),whostatesthat“grammaticalizationdoesnotmerelyseizeawordormorpheme…butthewholeconstructionformedbythesyntagmaticrelationsoftheelementinquestion.” Thisconceptionformsthefoundationofthefunctionallyorientedapproachtogrammaticalization,namely,onethatbringstheprinciplesoflanguageuseintothestudyofmeaningchangesthataccompanygrammaticalization(e.g.,Traugott1982,1988,2003a;Bisang1998;Hopper1998;Harris2003;WiemerandBisang2004).Suchalineofthinking,whichrejectstheideathatameaningfulinvestigationofgrammaticalchangecouldbereducedtodescribingindividual,context-freegrammaticalitems,seemsnaturallycompatiblewiththenotion‘construction’asdefinedinConstructionGrammar.

Fromthehistoricallinguist'sperspective,thepotentialofasystematicallyappliedconstructionalanalysisinthecontextofgrammaticalchangeshasbeenactivelyexploredespeciallyinTraugott'swork(Traugott2003a,2008a,2008b,2008c),firstinspiredbyCroft's(2001)constructionalapproach.Conversely,Fried(2008,2009a,2009b,2010a)examinesthesamequestionwiththegoaltotestthemodel'spotentialforrepresentingaspreciselyaspossiblethegradualnatureofgrammaticalizationandtheresultinglayeringeffects,allofwhichrequiresasufficientlevelofdetail.Butthepotentialofabroadlyunderstoodconstructionalapproachforaddressingagoodrangeofdiachronicissueshasbeenexploredalsobymanyotherscholars,asevidencedinthepapersinBergsandDiewald(2008,2009),Leino(2008),TrousdaleandGisborne(2008),orBarðdalandChelliah(2009),aswellasnumerousindividualstudies(Noël2007a,2007b,2008;Trousdale2008a,2008b;NoëlandvanderAuwera2009;NoëlandColleman2009;Bisang2010;NirandBerman2010a;Barðdal2011b,etc.),includingdiachroniccollostructionalwork(e.g.,GriesandHilpert2008;Hilpert2008).

Theconstructionalapproachisalsoprovingitselffruitfulingrapplingwithvariousbroaderanalyticchallenges,suchasaccountingforseeminglyunmotivatedsyntacticpatternsthatdonoteasilyfitinasynchronicallyattestedgrammaticalnetworkforagivenlanguage,orthatpresentatypologicallyoddandinexplicablepattern.Ithasbeenshownthatwiththehelpofconceptualizingthepuzzlingpatternsintermsofconstructionalchange,wecanarriveatinspiringandpenetratinganalyses,forexample,incasesofconstructionalborrowing(cf.Mithun's2008accountofaborrowedrhetoricalstrategythatgrammaticalizedintoanunusualwayofmarkinggrammaticalrelationsintheWakashanandThimshianicfamilies)or‘constructionaldisharmony’involvinganisolatedremnantofanoldersyntacticnetwork(cf.Ross's2008studyofthePuyumasystemofverbalnegation).AratheruniqueperspectiveinexploringthedynamicnatureoflinguisticstructureiscontributedbyaprojectknownasFluidConstructionGrammar(e.g.,Steels1998,2004,2011a,andthisvolume),whichextendsthechallengeofmodelingthegradualandinteraction-basedemergenceoflinguisticstructureintothedomainofartificialintelligenceandroboticsimulations.

Thechapterisstructuredasfollows.AfterestablishingtheconnectionbetweenConstructionGrammarandgrammaticalization(section23.2)andexplicatingthenotionofconstructionalization(section23.3),IwillpresentapracticalexampleofrepresentingtheseprocessesintheConstructionGrammarnotation(section23.4)andconcludebysuggestingareasforfurtherresearch(section23.5).

23.2.ConstructionGrammarintheServiceofGrammaticalizationTheory

Theaffinitybetweengrammaticalizationresearchandtheconstructionalapproachisnotcoincidental.ThedevelopmentofConstructionGrammarwasmotivated,amongotherthings,bytherealizationthattrue

1

Page 306: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 3 of 14

understandingofspeakers’linguisticknowledgecannotbereducedtotrackingdownmerelythestructuralpropertiesoflinguisticexpressionsandthemeaningofwords,butmustincorporatealsoprinciplesthatgovernthewaysinwhichlinguisticunitsareusedandinterpretedinactualcommunication(Fillmore[1974]1981).ConstructionGrammardoesnotdivorcelinguisticformfromitsmeaning,function,andprinciplesofusagebut,instead,takesthesedimensionstoformanintegratedwhole—agrammaticalconstruction,thatis,aconventionalpatternofspeakers’understanding.Grammaticalization,inturn,asaninherentlysyntagmaticandcontext-sensitivephenomenon,isconcernedwithidentifyingchangesintherelationshipbetweenformandfunctionwithinaparticularlinguisticpattern.Inthisrespect,theideaofaconstructionasaninternallycomplex,multidimensionalsignsuggestsitselfquitenaturallyasausefuldescriptiveandexplanatoryconcept.Moreover,theideathatintracingameaningchangeofaconcreteelementwemustrefertotheentiresyntagmaticsequenceinwhichtheelementoccursiscompatiblewiththeconstructionists’claimthatconstructionsthemselveshavemeanings.

Thesearerathergeneralcontoursofwhatmaybebringingthetworesearchagendastogether;theydonotanswerthemorespecific(andpractical)questionsofhowtheconstructionalmodelhelpsaddresstheconcernsofdiachronicresearchinactualanalyses.Inordertoexaminethemodel'sviabilityfordiachronicpurposeswehavetobringforwardseveralspecificfeaturesthatarecentraltoConstructionGrammarandthatappearparticularlyrelevantingrammaticalizationresearch.Thesefeaturesaresummarizedbelow,drawingmostlyontheexpositionofConstructionGrammar(CxG)inFriedandÖstman(2004a)andFried(forthcoming),whichiscloselyassociatedwithFillmore'soriginalconceptionoftheframework(Fillmore1988,1989,andthisvolume,whereitislabeledBCG;Fillmore,Kay,andO’Connor1988;Lambrecht2004)andfurtherenrichedbycertaininsightsofCroft's(2001)RadicalConstructionGrammar(RCG),particularlybyincorporatingexplicitlythenotionoffunctionalprototype.

Firstofall,bydescribingconventionalassociationsbetweenformandmeaning/function,constructionsprovidegeneralblueprints,understoodasfunctionalprototypes,forlicensingwell-formedlinguisticexpressions,applyingtounitsofanysizeorinternalcomplexity(morphologicalunits,words,phrases,clauses,turns,etc.).Constructionsaremeanttocapturegeneralizationsaboutspeakers’linguisticknowledge,byidentifyingclustersofco-occurringfeaturesthatfacilitatetheproductionanduptakeinactualcommunication.Themultidimensionalcharacterofconstructionsreflectsthemodel'soriginalinterestinidentifyingthecombinatorialpropertiesofwords,thusconceptualizinggrammarians’workprimarilyasthestudyofwordsincontext.Thisalsoprovidesforaprincipledandsystematicinclusionofrecurring,conventionallyexpectedsemanticandpragmaticpropertiesoflinguisticexpressions,whethertheymanifestthemselvesas(relatively)stablefeaturesofagrammaticalpattern,orastriggersofnovelinterpretations.

AcrucialfeatureofconstructionalrepresentationsfollowsfromthefactthatCxGmakesatheoreticaldistinctionbetweenwhatconventionallyidentifiesaconstructionasawholeandwhatischaracteristicofitsconstituents.Theformerisreferredtoastheexternal(or‘constructional’)properties,whichisasetofconstraintsonhowagivenexpressionfitsinandinteractswithlargergrammaticalpatterns.Thelatterrepresentstheinternal(or‘constituent-level’)properties,whicharetherequirementsplacedontheconstruction'sconstituents.Thisdistinctionallowsustoarticulatesystematicgeneralizationsaboutsyntagmaticconstraints,whilealsoprovidingaprincipledaccountoftheinternalstructureoflinguisticsignsinwhateverdetailmaybenecessary.Indiachronicanalyses,thisdistinctionisparticularlyuseful:grammaticalizationtypicallyconsistsofaseriesofsmall-scale,feature-basedadjustments,whichmay,collectively,leadtoaperceptiblechangeintheshapeandgrammaticalstatusofthewholepatterninwhichagivenunitoccurs.CxGgivesusawaytocapturetheincrementalnature,includingthepotentialmismatchesbetweengrammaticalpatternsandtheitems(words,morphemes)thatfillthemandwhosemeaningorgrammaticalstatuschangesovertime.

Theexternal/internalcontrastisrelatedtoanotherconstitutivefeatureofconstructions,namely,theirnoncompositionalcharacter:aconstructionhasitsownfunction(ormeaning),unpredictablefromsimplyaddingthepropertiesofitsconstituents.Asexplicatedinoneoftheearliestdefinitions,constructionsarerepresentationalobjectsthat“areassignedoneormoreconventionalfunctions…togetherwithwhateverisconventionalizedaboutitscontributiontothemeaningortheuseofstructurecontainingit”(Fillmore1988:36,emphasismine). Inadiachroniccontext,thismeansthataconstructionalanalysisprovidesawayofcapturingthetransitionsbetweencompositionalandnoncompositionalpatterns,astheinevitableeffectoftheconstanttensionbetweencreatingnewcombinationsofunits(withafullytransparentmeaningorfunction)andconventionalizingexistingcombinationsinnewinterpretations(leadingtolossoftransparentinternalstructure).Itiscrucialtoemphasize,though,thatnoncompositionalityattheconstructionalleveldoesnotmeanthatwecannot‘lookinside’and

2

Page 307: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 4 of 14

analyzetheconstruction'sinternalstructureorthepropertiesofindividualpieces.ThepointofnoncompositionalityasunderstoodinCxGismerelythis:whenweaddthepiecestogether(i.e.,theirinherentmeanings,functions,orcategorialstatus),wedonotgettheholistic,constructionalmeaning/functionthatisconventionallyassociatedwiththepatternasawhole(forspecificexamplesandanexplicationofvariousless‘direct’manifestationsoffunctionalnoncompositionalityinCxG,cf.Fried,forthcoming).

ItisalsoimportanttostressthatCxGmakesadistinctionbetweenconstructionsandconstructs.Constructionsare“piecesofgrammar”(KayandFillmore1999:2),whileconstructsareactualphysicalrealizationsofconstructions,thatis,utterance-tokens(words,phrases,sentences)thatinstantiateconstructionsindiscourse.Aconstructionisthusageneralizationoverconstructs.Maintainingthisdistinctionisparticularlyrelevantinestablishingtheroleofconstructionsindiachronicshifts:aseriesofpartialchangesinanexpressionmaygiverisetoanewconstructionorleadtoareorganizationofanexistingone,butthechangesthemselvesoriginateinactualutterances,notinconstructionsthemselves.

Finally,diachronicanalysisdrawsattentiontotheissueofcategoriality.SinceCxGdoesnotassumeacategoricaldistinctionbetweenlexiconandgrammaritoffersthenecessaryflexibilityinaccommodatingthepervasivegradienceinlinguisticcategorization(cf.Aarts2007).Thismatters,amongotherthings,inthecontextofdeterminingtheboundariesbetweengrammaticalizationandlexicalization,ifwewishtomovebeyondtheoverlyreductionistapproachwhichassumesthatchangesresultinginnewgrammaticalentitiescanbetracedtosomethingdistinctlynongrammaticalandviceversa.TheconceptualbasisandthearchitectureofCxGdoesnotrequireustoimposeanyarbitraryboundariesandcaneasilyaccommodatecategorialunderspecificationorindeterminacy.

23.3.TheNotionofConstructionalization

Theinteractionbetweenaconstructionalanalysisandcertainconcreteobservationsandhypothesesmadeaboutvariousaspectsofgrammaticalizationprocessesallbringupserioustheoreticalissuesthatrequiresomeexplication:thehypothesisthatconstructionsarethedomainofchange;thegradualnatureofgrammaticalization;theemergenceoffunctionalpolysemies;andthequestionofcontextingrammaticalization.Takingallthisintoaccount,thediscussioninthissectionwillbebasedonthegeneralthesisthatgrammaticalizationprocessesaremostaccuratelyconceptualizedasinstancesof‘constructionalization’:aprocessthatleadsto(1)theemergenceofanewgrammaticalpattern(construction)outofpreviouslyindependentmaterialor(2)areorganizationofanexistingconstruction,leadingtoanincreasinglymoreopaquemeaningofthepattern.Thecatalystforchangeinsuchaprocessisalwaysaparticularlocalcontext,whichischaracterizedbyaconfluenceoffactors(semantic,pragmatic,syntagmatic,etc.)thattogetherfacilitateameaningshiftanditssubsequentconventionalization.

Thecontributionofaconstructionalapproachtowardcapturingthetruenatureofgrammaticalizationcanbeassessedfromtwodifferentperspectives,ultimatelyrelatedtotheexternal/internalcontrastdescribedabove.Thereadilyobviousadvantageofaconstructionalanalysishastodowiththe‘holistic’dimensionofchange,thatis,thefocusonthepatternasawhole.TheholisticbiasisparticularlyprominentinRadicalConstructionGrammar(Croft2001,thisvolume),whichwasoriginallymotivated,atleastinpart,bytheneedtoaccountforgrammaticalization.RCG-basedanalysesareprimarilyinformedbytheclaimthat“theconstructionasawholechangesmeaning”(Croft2001:261);thishighlightsthesyntagmaticnatureofgrammaticalization,aswellastheerosionofcompositionality,whichisanecessaryby-productoftheseprocessesandadefinitionalfeatureofconstructions.

Wecanillustratethisperspectiveonthewell-known,bynowclassic,exampleofgrammaticalization:thedevelopmentofbegoingtoinEnglishasafuturemarker.Theholisticapproachdrawsourattentiontothefactthataverbofmotiondevelopedaparticulargrammaticalmeaningthatis,inthefinalstage,associatedwiththesyntacticallycomplexform[BEgoingtoV ]. Thisbracketedsequencehastocountasagrammaticalconstruction(inthetechnicalsense)inthatitisaconventionalcombinationofseveralelementsanditistheentirecombinationthatservestoexpressaparticularmeaning;themeaningisnotpredictablefromsimplyaddinguptheinherentmeanings/functionsofthoseelements.Thisconstructionfallssomewhereinthemiddleofthecontinuumofconstructionalschematicity:itispartiallylexicallyfilled(theverbBE,theformgoingandthewordtoareallfixed)butitsopenslots(theperson/tense/aspectformoftheverbBEandtheverbintheinfinitivalcomplement)provideforfullproductivityoftheconstructionotherwise.Thecrucialpointisthatitisnotenoughtosaythattheverbof

inf 3

Page 308: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 5 of 14

motionGO(inits-ingform)hasdevelopedagrammaticalfunctionmarkingfuturetense.Instead,theverb(-form)developedthisfunctioninaspecificstructuralenvironmentusedinaparticularsemanticandpragmaticcontext.Theabilitytoincorporatethisimportantinsightisuniquetotheconstructionalapproach,sincenoothersyntactictheorycancoherentlyintegratethecontextualprerequisitesofsuchchanges.

Thepurelyholisticfocushasalsoledtothehypothesisthatconstructionsarethedomainofchangebecauseitistheentireconstruction,notjustoneitem,thatchangesmeaning(e.g.,Croft2001;Bybeeetal.1994).Itisnotclear,however,whatexactlythismeans.Theformulationintheabovequotethat“theconstructionasawholechangesmeaning”ispotentiallymisleadingandmayobscurewhatitmeansforconstructionstobethedomainofchange.Thisisatheoreticalproblem.TheformulationimpliesthatthereisaconstructionwithmeaningX,whichovertimechangesitsmeaningintoY;putdifferently,thechangewouldalwaysbeamatterofreorganizingexistingconstructions,byremappingcertainform-functionassociations.Leavingasideforthemomenttheproblemthatsuchaninterpretationmakesitdifficulttoconceptualizehownewconstructionscomeintoexistencetobeginwithorhowanitemchangesaconstructionalaffiliation,sotospeak(showninsection23.4.1),itisproblematicevenforunderstandingwhathappenedinthedevelopmentofcasessuchasbegoingto,namely,whatconstitutestheoriginal(nonfuture)construction.

Fornowletussimplysuggestthatemphasisontheholisticdimensiontendstodrawattentionprimarilytotheobservableresultofgrammaticalizationandislessconcernedwithexaminingtheparticularsofthediachronicprocess.Yet,understandingtheparticularsisnecessaryifwewishtoanswerthequestionsofhowagivenchangemayhavedevelopedandwhyittakestheshapeitdoes.Consistentwiththewhyandhowquestions,theprimaryfocusofgrammaticalizationresearchhasbeenthestudyoftheinternalmechanicsofandmotivationsforagivenchange;thisperspectiveoperateswithsmall-scale,feature-based,andnotimmediatelyperceptiblediscretechangesthatarepartoftheon-lineproductionanduptakeinthecourseoflinguisticinteractionandthatmayormaynotleadtoanobservablechangethatbecomesconventionalizedasanewpattern;thishasbeenarguedinmanydiachronicaccounts(e.g.,Timberlake1977;Andersen1987,2001;Lichtenberk1991;Harris2003;Traugott2003a;Fried2008,2009b,2010a).Thepurelyholistic,construction-levelanalysisismuchtoogeneralforcapturingthepartialandhighlylocaltransitionsand,therefore,foraddressingthegradualnessissue.

Inordertoexemplifytheconcernsofthe‘process-oriented’perspective,wecanagainconsidertheEnglishbegoingtodevelopment.AninformalschematizationofthepartialtransitionswouldhavetoincludeatleasttheshiftsidentifiedinTable23.1,whichisaslightadaptationofTraugott's2010apresentationoftherelevantdetails.Theaffectedfeaturesareorganizedintotwosubsets,oneconcerningthecontext(i.e.,externalproperties,whichconstraintheitem'srelationshiptoitsenvironment)andtheotherconcerningthepropertiesofthe(formofthe)verbGOitself(i.e.,propertiesinternaltotheiteminquestion).InTable23.1,thepairsofemptybrackets[]standfor‘underspecifiedforthatvalue’andthedashedlinesindicatethefluidnatureofthetransitionsbetweenidentifiablestages.

Whatisimportanttoadd,though,isalsotheparticularlinearizationpatternassociatedwiththechange(theverbGOmustbeimmediatelyfollowedbythepurposivetoVcomplement)andtheabsenceofadirectionalcomplement,whichisotherwisepossiblewithmotionverbs.Bothoftheseconditionshavetodowithspecificusageinspecificcommunicativecontextsandbothareinstrumentalinfacilitatingthereanalysisthatisnecessaryfortheholisticchangetobecomemanifest.Evenwithoutdiscussingallthe(mostlyquitefamiliar)detailsofthisdevelopment,theschematizationinTable23.1helpsusappreciatetherelevanceofgreatergranularity,inwhichwecanseehowcertainfeaturesgraduallydepartedfromtheiroriginalspecification,thusopeningthepathtowardthereorganizationweobserveattheendoftheprocess.Theoverallchangeconsistsofanumberofconcretesmallshiftsthataffectboththeformitself(i.e.,theverbGO)anditsrelationshiptotheimmediatesyntagmaticcontext.Noteinparticulartheexternal-to-internaltransitioninmodelingthemotion>futurechange:theprocessstartsatthecontextual(external)levelofapragmaticallyconditioned‘later-oriented’inference(left-mostcolumn),movingtowardtheinternallymarkedfuturityasasemanticfeatureofthebegoingtoformitself(right-mostcolumn),withbothofthesepossibilitiespresentinthetransitionalstageinthemiddle.

Page 309: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 6 of 14

Table23.1.Featuresinvolvedintheriseofbegoingto

“GOmotion”(c.1400)

ambiguous(c.1475)

“begoingtofuture”(c.1640)

External syn biclausal bi-~[] auxV

sem purposiveimperfective

purp.~[]imperf.~[]

prag later-oriented later-oriented~[]

Internal syn mainVclause-final

mainVclause-final~medial

complexauxV

sem motion motion~futureimminent~[]

futureimminent

morph -ing -ing

lex GO GO~begoingto begoingto

Itfollowsthatifwewishtomodelthefullcomplexityofthedevelopmentandtoexplaintheprocess,wecannotoperatesolelywiththeideathatonewholeconstructionchangesitsmeaning.Rather,itistheinterplayofvarioussubsetsoffeaturesthathappentoco-occurinactualutterance-tokens(i.e.,constructs)andthatgiverise,throughinferencesthataremadeavailableandsalientinthoseconstructs,toanewpatternwithitsownmeaning.Tobemorespecific,itwouldbedifficulttoarguethatutterancessuchasIamgoingtotakeanap,whicharepotentiallyambiguousbetweenapurposivedirectionalandfuturereadings(‘I’monmywaytoacouch/bed/bedroominordertosleep’vs.‘I’mabouttofallasleep’),areinstantiationsofagrammaticalconstruction,inthesenseofaconventionalpieceofgrammaranddistinctfromaconstructionthatwouldbeinstantiatedbytheconstructIamgoinghometotakeanap.Bothutterancesmaybeinstantiationsofthesamesemanticargumentstructure(contributedbytheverbGO),whichalsodeterminesthefactthatthedirectionalcomplementmayormaynotbeexpressedinsyntaxandbothcanbetakenasbiclausaltokenscontainingapurposiveclause.Moreover,itisnotclearwhatconstructionalmeaningwewouldplausiblyattachtoanyconstructionthatwouldserveastheinitialstageofthemotion>futurityshift.Itisstillthecasethat,strictlyspeaking,theonlypiecesthatchangetheirmeaning/functionaretheformgoingandthemarkerto,buttheydosoinaspecificcombinationwithotherlinguisticelementsinconcretecommunicativetokens.Thetokensmayformcoherentclustersofsyntagmatic,semantic,pragmatic,andmorphologicalfactors,whichthenallowtherelevantinferencesandsubsequentreanalysis,butthetrulyconstructionaldimensioncomesinonlyatthestagewhenthenewmeaningoftheitem(s)inquestionisobligatorilyassociatedwithaparticularsequenceofelementswhichalsorepresentaparticularsyntacticconfiguration,differentfromtheconfigurationinthetrulydirectionaltokens.Putdifferently,onlythelastcolumninTable23.1hasthestatusofafullyestablished,newconstruction,whiletheothertwocorrespondtocertaincombinationsoffeaturesandlinguisticpiecesfoundincertainkindsofconstructscontainingtheverbGOnotonlyinitsfullmeaningofamotionverbbutalsoinanytense/aspectform,notjustthe-ingform.

ThesummaryinTable23.1pointstoanotherimportantobservationthatjustifiestheprocess-orientedanalysis.Asiswellknown,theoriginalmeaningdoesnotnecessarilydisappearwhenanewonestartsemerging,orevenwhenitbecomesfullyconventional.ThisleadstoformingpolyfunctionalnetworksofcoexistingmeaningsandthiseffectcanbecapturedonlythroughreferencetosubsetsoffeaturesatthelevelofdetailsuchassuggestedinTable23.1,notmerelyattheholisticlevel.Thetableisnotintendedasafullyworkedoutrepresentationofthenetworkoftherelevantpatterns,ofcourse,butitwouldbethenecessarystartingpointtowardconstructingsuchanetwork.

Tobesure,thisfeature-basedconceptualizationofgrammaticalization,whichismotivatedbythefocusonpartial

4

Page 310: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 7 of 14

transitions,doesnotrejectthenotionofconstructionortheholisticdimensionasrelevant;onthecontrary.Butitcallsforexaminingtheinternalmake-upofconstructionsandforacknowledgingtheroleofverylocalcontextsandparticular,lexicallyspecificsubtypes(‘micro-constructions’inTraugott's(2008a:32)terminology),whichmaygraduallyacquirethestatusofmoregeneralizedcollocationalcombinations,asakindofintermediatestageofconventionalization. ThiselusiveandsofarnotverysystematicallyaddressedstageisperhapsthemotivationbehindTraugott's(2008a)ideaof‘meso-constructions’orthenotionof‘coiningpatterns’inotherconstructionalaccounts(e.g.,Fillmore1997;NirandBerman2010a).Inthisview,emphasisisputonthedetailsofusageascrucialsourcesofexplanationsforhownewmeanings(andyes,constructions)actuallyemergeincommunicationandwhatmotivatesthechange.Consequently,itismoreaccuratetotreatthediachronicprocessasacaseof‘constructionalization,’reflectingthefactthattheresultoftheprocessistheformationofaconstructionasanewpieceofgrammar,ratherthananewgrammaticalstatusofanindividualitem.Moreover,itaffordsanexplicitaccountoftheshiftfromacompositionaltoanoncompositionalmeaningofsuperficiallyidentical-lookingstringsofwords.

Tosummarize,diachronicprocessesthatleadtoformingnewgrammaticalunitscanbemostaccuratelycapturedandexplainedbymakingreferencetoboththeexternal(holistic)andtheinternaldimensionsofconstructions,thereby‘unpacking’theprimarilyholisticapproachandconceptualizingtheprocessasadevelopmentinwhichmeaningXofanitem(lexicalorgrammatical)changesintomeaning/functionYinalargercontextC.Thismeanstreatingtheholisticandtheprocess-orientedanalysesnotonlyassimplycomplementaryintheirperspective,butasequallycrucialinthedescriptionandexplanationofanygrammaticalchange.

23.4.Constructionsat(Diachronic)Work

ThepurposeofthissectionistoconsiderhowCxGcanbeappliedinanalyzingaspecificdiachronicprocess,therebydemonstratingthreethings:(1)lookinginsideconstructionsinordertopindownthemechanicsofchange;(2)theuseoftheschematicnotationasawaytobemorepreciseinarticulatingthegradualnessofgrammaticalchangeanditsanalogicalnature;and(3)thewaysuchananalysiscancontributetowardestablishingrelationshipsacrossstagesofgrammaticaldevelopmentthroughtheconceptof‘constructionalmap’(borrowingtheconceptoffunctionalmapsusedintypologicalresearch).

Theillustrativeexampleconcernstheerosionofacompositionalmorphosemanticstructureofaninflectionalform,withtheattendantlossofsyntacticfreedomandrecategorialization,asitispulledintoanexistingsyntacticconstruction.Thedevelopmentinquestion(roughly,participle>adjective)isalexicogrammaticalchangethatstraddlesthederivation/inflectiondistinctionandmayappearlesstransparentthanthebegoingtocase,becauseatissueisnotjustachangeinthemeaningofanitemvis-à-visitslocalenvironment.Theitemitselfisamorphologicallycomplexformandwhatchangesisitsinternalstructuretogetherwithits(external)categorialstatusandsyntacticbehavior.Atthesametime,thisaddedcomplexitymakesitaninstructivecaseinshowingtheusefulnessofconstructionsasthecrucialconceptandthedevelopmentcanstillbeclassifiedasaninstanceofconstructionalization,thatis,emergenceofanewconstructionoutofspecificconfigurationsoffeatures(syntactic,morphological,semantic)inconcreteutterances.

23.4.1CapturingaChangeinProgress

Theso-called‘long’participleinSlavic(withsomeparallelsinGerman),hereexemplifiedbyOldCzech(OCz)material,isaschematicmorphologicalconstruction(anexpositionofconstructionalmorphologycanbefoundinBooij2010andthisvolume).Itsconstituentsaremorphemes,eachofwhichcontributesparticularsemanticcontent,asindicatedinthetemplatein(1);theabbreviationPAPstandsfor‘presentactiveparticiple’andCNGstandsforthecase-number-genderportmanteausuffixthatisaddedtotheparticiple;thewholeformislabeled‘participialadjective’(PA),inreferencetoitsmorphologicalshape:

(1)[[[Vroot—pres.stem]—ppl] —CNG] ‘(theone)V-ing’e.g.,[[[chod—ie]—c] —í] ‘(theone)walking’

Theinternalstructureofthiscategorialhybridshowsthatitpreservesitsverbaloriginbymarkingpresenttenseandvoice(active)aspartofthepresentstem(asopposedtopastorpassivestems).Therootalsobringsalongvalencepropertiesthathavebothsyntacticandsemanticmanifestations(expressingcompletepropositions,

5

6

PAP PA

PAP PA

Page 311: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 8 of 14

expectingthepresenceofaNPthatinfiniteclauseswouldbethesubjectandmaintainingverbalgovernmentinmarkingnonsubjectarguments).Externally,thePAiscategoriallyindeterminate:theCNGsuffixisadjectivalinform,butevidentlyvariableinfunction,asthePAspannedthespectrumofexpressingpredication,modification,andactor-nounreference.Overtime,though,thecategorialstatusofthePAsbecamefixedbyconventionalizingparticularcontextuallymotivatedpreferences.HereIconsideronlythepredicate>attributedevelopment,illustratedbelow;eachexamplerepresentsthemosttypicaltokenofagiveninterpretation(i.e.,acentralmemberinitscategory):‘event-profiling’in(2)andmoreorlessclearly‘participant-profiling’in(3)–(4).(Foreasierorientation,thePAanditstranslationequivalentareprintedinboldandthenounitco-occurswithisunderlined.)

(2)uslyšelžáčkadřéveřečenýveršzpievajícieho [PovOl255]‘(andwhenheagainsecretlyenteredthechurchonFriday,)heheardayouthsingthataforementionedsong’(3)počeliobětovatikaždýzlatýpeniezmající nasoběobrazanjelský[PovOl276b]‘[everyone]startedofferingagoldcoin,whichhadonitapictureofanangel’(4)žádajúcího vítězstvítrojzvuksčastněpřijal[ŘádKor42b]‘hejoyouslyacceptedthesoundofthedesirable/welcomevictory’

Thetokenin(2)exemplifiesamorphosemanticallytransparentmemberoftheverbalparadigm,usedasadepictivesecondarypredicate.ThePAistruetotheverbalnatureofpresentactiveparticiplesbothsemanticallyandstructurallyinthatitexpressesafullpropositionconcurrentwiththemaineventandissyntacticallyrelativelyindependentofitssubjectcomplement(‘youth’);thePA'spositioninthesentenceisnottiedtoanyparticularslot.Moreover,therootexpressesanactionandthesubjectargumentisanimate.ThePA'smeaningiscompositionallyderivablefromitsmorphologicalstructure(1)andcanbeglossedas‘[aperson]whodoessomethingatthetimeofthemainevent.’Incontrast,themeaningsin(3)and(4)moveprogressivelyclosertowardtargetingaparticipantinthesecondarypropositionandascribingasalientpropertytothatparticipant.In(3),thePAcanbeglossedas‘[one]who'spronetoV-ing’:ithasahabitualflavorandhighlightsitstemporalautonomyrelativetotheeventofthemainclause.Thisexamplealsoshowsadeparturefromtheoriginalrestrictiononverbsemantics(mít‘have’isnotaverbofaction)andananimateagent(peniez‘coin’).Thefullclusteroffeaturesthatparticipateintheshiftcanberepresentedschematically(Figure23.1),showingexplicitlywhichoftheverbalfeaturesbecomeweakened(ingray).Theinsideboxesrepresenttheverbalstem(ontheleft)andtheCNGsuffix(ontheright):thestemshowsitsverbalproperties(voice,tense,meaningoftherootbyreferencetothesemanticallyappropriateclassofframes)andthesuffixspecifiesonlyitsintrinsicagreementfeatures.TheouterboxspecifiesthePA'sexternalcharacteristics,thatis,howthisformmanifestsitselfrelativetolargerstructuresinwhichitisused:itrequiresanominalthatwillinstantiatetheagentargumentoftheroot.

Finally,inthereadingexemplifiedin(4),thePAbearsclearsignsofatrulyadjective-likestatus,bothsemanticallyandsyntactically:theexpectedassociationbetweensemanticargumentsoftheverbalrootandtheircanonicalexpressioninsyntaxiscompletelysevered,leadingtoanoncompositionalmeaning(‘desired/desirable’<lit.‘desiring’),whichcomesinvariousflavors(passive,modal,resultative,purposive,augmentative,etc.).Allthatremainsoftheverbaloriginistherootwithitssemanticframe,whoseparticipantroles,however,arereconfiguredinanewandotherwiseunpredictablemeaningoftheconstructionasawhole.

Clicktoviewlarger

Figure23.1. PAinhabitualusage

Wecouldnoteasilyaccountforthefunctionalandsemanticdifferencesbetweenthesethreestagesofdevelopmentifwelimitedourselvestoaholisticobservationthattheconstruction(i.e.,thePA)changeditsfunctionfrombeingapredicatetobeinganattribute.Thiswouldbesimplyarestatementofthetraditional(andinaccurate)sweepingclaimthatparticiplesturnintoadjectives.Suchageneralizationwouldnottellusanythingabouthowandwhythischangetakesplace,letaloneaddressanotherknownfact,namely,thatactiveparticiplesactuallytendto

PA.ACC.SG

PA.ACC.SG

PA.GEN.SG.N

Page 312: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 9 of 14

berelativelyresistanttoafullcategorialshift(participle>adjective),incontrasttopassiveorpastparticiples.Nor,however,canwemotivatethedevelopmentifwefocusonlyontheinternalstructureofthePAitself(whichisthetraditionalsourceofexplanationindiachronicanalysesoftheseforms),withoutconsideringthecontextsinwhichitwasusedatthetimewhentheformwasstillcategoriallyfluidandthefunctionandmeaningdependedonthelocalsurroundings.Inparticular,thedevelopmenttowardattributivenessandevenafull-fledgedadjectivecanbetracedtotwocrucialfactorsoutsideoftheformitself:theanimacyofthePA'ssubjectandthelinearizationpatternsinwhichthePAoccurred.

Thepredicativeusagetypicallyandoverwhelminglyinvolvesanimateandhighlyindividuatedentities(realagents),whiletheerosioninpredicativenesscorrelatesstronglywithlowerindividuationoftheagentandwithaccommodatinginanimateandabstractentities.Wordorderasafactorconsistsinseveralsubtleandinterrelatedmodulations(analyzedindetailinFried2008),buttwoaspectsaredominant:theadjacencyofthePAanditssubject,shownin(3–4)incontrastto(2),andalsotheirrelativeposition,thatis,thedifferencebetweentheNP-PAorder(3)andthePA-NPorder(4).Inaninformalschematization,theclustersoffeaturesthatcorrelatewiththesetwolinearizationsaresummarizedinFigure23.2.Bothsequences(NP-PAandPA-NP)representregularlyoccurringcombinationsinactualtextsandeachisassociatedwithanoverwhelmingpreferenceforacertainclusteroffeaturesintheattestedconstructs.Thegraycolorindicatestheverbalfeaturesthattendtofadeineachlinearizationpattern;theboldfaceisusedforfeaturesthatarenewlyassociatedwiththePAsineachpattern.

Clicktoviewlarger

Figure23.2. ContextualfactorsinPAinterpretation

ItisstrikingthatacompletelossofthePA'sverbalpropertiesissostronglyassociatedwiththeimmediatelyprenominalposition,which,inturn,isalsotheneutralorderinothermodificationalstructuresinOCz,[Modifier-Head].ThetextualmaterialshowsthattheNP-PAorderisstillarelativelylooseconfigurationandtheadjacencyismuchlessofarequirement,ascomparedtotheprenominalplacement;thethreedotsintheparenthesesinthediagramindicatethatinterveningmaterialisstillpossible(andnotthatuncommon)inthepostnominalplacementofthePA.Itiscognitivelyplausibletohypothesizethatthealmostexclusiveadjacencyintheprenominallinearizationfacilitatestheemergenceofatighterconceptualunitinthemindofthespeakers(indicatedbythedashed-lineenclosureinFigure23.2)andthestructuralsimilaritywitharegularmodificationconstructioncanonlyreinforcesuchaconceptualization.However,theattractionbetweenthePAandtheinherentlyPA-independentschematicPrenominalModificationconstructioncomesataprice,asthe(semanticallymorerestricted)PAadjuststothe(moregeneral)requirementsofthemodificationconstruction.Crucially,themodificationpatternimposesanattributiveinterpretation,favoringatemporallyascribed

Page 313: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 10 of 14

Table23.2.Capturingchangeinprogress:partialshiftsinthePA'sdevelopment

SecondaryPredicate

Habitual Passive,Modal,etc.

External cat [] [] A

sem anim.subj.(agent) anim.subj.(agent) functionallyandsemant.unrestricted‘headN’

prag contrastivefocus

Internal cat V V V

syn non-subj.complementsverbalgovernmentactivevoice

non-subj.compl.verbalgovernmentactivevoice

sem Vsofactiontense(contemporaneous)

Vsofaction/statestense(contemp.)/habit.

Vsofaction/statesatemporal

propertiesthatsimplyrestrictreferencefortheheadnounandthenounisalsodevoidofsemanticconstraintsastoanimacyordegreeofindividuation.Putdifferently,thereisapotentialmatchbetweenthePAandthemodifierslotinthisconstructionandpullingthePAintotheslotresultsinminimizingthePA'spredicativepotentialinfavorofattribution,analogicallytoothermodifiers.

TherelevantpropertiesandtheirreorganizationindifferentfunctionalandcategorialclustersaresummarizedinTable23.2;thedownplayingorlossoffeaturesisingray,newlydevelopedfeaturesareinboldandthedashedlinerepresentsafluidtransitionbetweentwointerpretations.Thetablerevealsageneralpatternofdevelopment:thegradualerosionandlossofthetransparentmorphosemanticstructureofthePAasamemberoftheinflectionalverbalparadigmstartswiththesemantics,bothinternalbutparticularlyexternal(subject)tothePAitself,whilethesyntacticmanifestationsofthechangeandafullcategorialshift(ifitoccursatall)becomeconventionalizedmoreslowly.Inpresent-dayCzech,thePA'splacementimmediatelynexttoitssubjectisobligatory,regardlessofitsinterpretation(predicativeorattributive)anditsoriginalsyntacticautonomyinrelationtootherpartsofthesentenceisthuscompletelyobliterated.

ThecontextualconditionsthatfacilitatedthereinterpretationsresideinseveraldomainsandareatleastpartlyreinforcedbyananalogytothePrenominalModificationconstruction.Thepropertiesofthetransitionalcontextsarelistedin(5);anysubsetoftheconditionsin(5a–d)hadasignificantpotentialtotriggeranovelinterpretationandconceptualization.

(5)a.structural:adjacencyofPAanditssubjectb.semantic:generalizations,classificationsc.pragmatic:lowerreferentialstrengthoftheanimatesubjectNPd.textual:descriptiveandnarrativetexts

ThedevelopmentcanbeclassifiedasacaseofconstructionalizationbothatthelevelofthePAitself(amorphologicalconstructionchangeditsmeaning/functionthroughreorganizationofitsdefiningproperties)andatthelevelofitssyntacticbehavior(attractionintoaparticularslotinamodificationconstruction).Takentogether,theprocess,whichappliestoalleligiblemembersofthecategory,resultedinestablishingasyntacticallyrestricted

Page 314: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 11 of 14

andsemanticallygeneralizedcategoryoutofasyntacticallyrelativelyautonomous,semanticallyconstrained,andfunctionallyfluid,context-dependentword-form.

Thedetailsoftheconstructionalanalysiscanbearrangedinanonlinearfashionaswell,inanetworkofrelationshipsacrossthedifferentstages.Theresultisaconstructionalmap(Figure23.3),whichisbasedonthefeatures(inbold)usedintheconstructionalrepresentations.Eachconstructionisenclosedinarectangleinsuchawaythatweseeexactlywhichofthefeaturesaresharedacrosswhichmembersofthenetwork.Thefeaturesthatappeartobethemostprominenttriggersoftheoverallchangeareenclosedinthegrayareaandwenotethat,significantly,theyareallexternaltothePAform.Theconflictingpressuresoftheconstruction-internalandtheconstruction-externalpropertiesevidentlyformahierarchysuchthattheinternalpropertieshavetheeffectofa“backwardpull”(Traugott2008a:34),againsttheexternallymotivatedanalogicaladjustment.Notealsothatthetransitionscanbeonlytreatedastendencies,notasabsolutevalues(thedirectionandrelativestrengthofpreferenceisindicatedbythesymbols‘</<<’and‘>/>>’).Finally,thedashed-linerectangleindicatesthatthefullyadjectivalconstructionsdidnotsurviveintothemodernlanguage.

23.4.2ConstructionsastheDomainofChange,Revisited

ThemotivatingfactorsandtheoverallprocesssketchedabovearequitecomparabletothosethathavebeenlaidoutforthedevelopmentofthebegoingtoconstructioninEnglish,eventhoughatfirstblushitisafairlydifferentkindofchange.Wecanthusgeneralizeevenfurtheraboutthecrucialingredientsofconstructionalization,listedin(6):

(6)

Clicktoviewlarger

Figure23.3. ConstructionalmapoftheOCzPAdevelopment(Fried2008:73)

a.theimmediateneighborhoodinwhichaformfindsitselfindiscourse(i.e.,theco-text);b.attractionandpartialadjustmenttoanother,alreadyexistingpattern,throughanalogy;c.backwardpullexertedbytheoriginalmeaning.

Thepropertyin(6b)isrelevantinaccountingfortheavailabilityofnewinferences(cf.Traugott's(2008a:33)ideaofconstructionsservingas“attractorsets”).Thefactor(6c)notonlycontributestoformingapolysemyinwhichtheoldmeaningmaintainsitspresencebutmayalsolimittheextenttowhichapotentialchangecanactuallyprogresstocompletion.Needlesstosay,therelativestrengthofthebackwardpullindifferentkindsofchangesisoneofthemanyissuesthatwillrequiremorefocusedattentionandfurtherresearch.

Itfollowsthattheideaofconstructionsbeingthedomainofchangeshouldbeformulatedwithsomewhatgreaterprecision,inordertoproperlyunderstandwhatrolegrammaticalconstructionsareexpectedtoplay.Constructionsarethedomainofchangeinsofarasconcreteconstructs(C1,C2,C3…Cn)ofacertainshaperesultintheestablishmentofaconstructionX,oftenwiththehelpoftheexistenceofanother,inherentlyindependentbutinsomewayssimilar,constructionY.Putdifferently,aparticularcombinationofelementsinanutterancemaybecomeaunitwithitsown,previouslynonexistentandnotpredictablemeaning/function.Thecrucialpoint,though,isthattheoriginal(‘triggering’)combinationcanbearelativelyfreesequenceofsyntacticallyindependentpieces(oneofwhichistheforminquestion)andnotnecessarilyaninstanceofanestablishedconstruction(recallFigure23.2).Thus,withregardtothedomainofchange,weneedthenotionofconstructionintwosenses:(1)asasourceofanalogicalmotivationand(2)astheendpointofthegrammaticalizationprocess.Whetherwecanalsoidentifyaspecificconstructionasthestartingpointwilldependonthenatureoftheformandchangeunderstudy;theonsetofachangeiscruciallyconnectedonlywithconstructs,notconstructions.

Page 315: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 12 of 14

Wealsoneedtokeepinfocusthatconstructionsparticipateinthechangeattwolevelssimultaneously.Constructionallyinformeddiachronicanalysescantakeadvantageoftheabilitytoanalyzesystematicallytheinternalpropertiesofconstructions.Atthesametime,referencetotheexternalpropertiesofconstructionscannotbelimitedtoestablishingconstructionalmeaningbutprovidesalsoawaytotrackthechangingrelationship(s)betweentheforminquestionanditscontextofuse.Thecontextisunderstoodintermsofasyntagmatic/structuralorganization,aswellasintermsofpragmaticpreferencesandconstraints,whethertheseconcerndiscourseorganization,speech-actfunctions,interactionalfeatures,orotherpragmaticissues.Thelatterhasnotbeenprominentlyaddressedinthischapter,butitwillbeparticularlysalientintracingcasesofpragmaticization,althoughpragmatictriggersarenotlimitedtothem,ofcourse.AfirstattempttomapachangeinvolvingpragmaticrelationsintermsofconstructionalreorganizationcanbefoundinFried(2009b)butalotmoreresearchconcerningthisconnectionwillbeneeded,onavarietyofdata.

Finally,theideathatdifferentstagesofdevelopment(orsynchronicpolysemiesinthecaseofcoexistingvariants)canbetiedtogetherinconstructionalmaps(forotherimplementations,seealsoFried2007b,2009c)seemsconceptuallyclosetotheconcernsofculturallygroundedevolutionarylinguistics(e.g.,Steels2007,2012),wherethecentralissueistounderstandtheprocessesbehindtheevolutionandconstant(re)aligningofthegrammaticalsystemsshared,alwaysimperfectly,acrossmembersofaspeechcommunity.Themapsareagoodvehiclenotonlyfordealingwithfluidboundariesofneighboringlinguisticcategoriesbutalsofororganizingthepropertiesofrelatedconstructionsinawaythatshowspossibleconstraintsondevelopingnewgrammaticalpatterns,possiblyilluminatingprocessesofself-organizationandselectioninrealigninggrammaticalsystems.Itisclearthatthefeaturesparticipatinginthepartialtransitionsleadingtoacomplexdiachronicchangearenotallinvolvedequally:somearemoreresistanttochange,othersarereadilyshiftedandstillothersareinstrumentalininitiatingand/orfacilitatingtheprocessinthefirstplace.Itisthuspossibletodrawatleastpartialhypothesesaboutwhichvariantsinanetworkaremorelikelytosurviveandbecomedominantandwhicharemorelikelytobeshort-livedandwhy.Themapscanthusbeanothertoolformodelingandtestingmechanismsthatseemtounderliegrammaticalizationprocessesandforwhichconstructionsareacrucialconcept.

23.5.ConclusionsandOutlook

Ifthegoalistouncoverthemotivationforchange,thentheintegrationofgrammaticalizationresearchandconstructionalanalysisappearstobeahighlypromisingapproach.AlthoughCxGhasonlyrecentlystartedtobeappliedtodiachronicanalyses,itisrapidlygainingcurrencyamonghistoricallinguists,asitisbecomingevidentthattheconstructionalapproachcanbehelpfulincapturingtheemergenceofgrammaticalstructure,therebygoingbeyondsimplycomparingdiscretesynchronicstagesandtowardidentifyingdiachronicrelationshipsatanappropriatelevelofdetail.

Aboveall,CxGisausefultoolforanalyzingandrepresentingthedirectrelationshipbetweenlanguageuseandlanguagechange,whichisthesinequanonofgrammaticalizationprocesses.Thetaskofmappingouttheintricatewebofmotivationsandpartialshiftsfromwhichanobservablechangemaygraduallyariserequiresanapproachinwhichthechangingformcanbestudiedinrelationtoitsusageenvironment.Thisrequirementisservedwellbythenotionofconstruction,asitallowsanalysisfromboththeholisticandthe‘inside-out’perspective.Finally,itgoeswithoutsayingthatCxGisinherentlyconsistentwiththeco-evolutionhypothesis(Bybeeetal.1994),thatis,formandmeaningchangingsimultaneously,althoughthisappearstobetrueonlyattheconstructional,notthefeature-based,level.Itisclearweneedtocombinethesetwoperspectives,ratherthaninterprettheideaofa‘constructional’analysisonlyinthenarrowsenseofnoncompositionality,concerningmainlytheresultofchange.Whiletheholisticaspectisanindispensablepartoftheenterprise,therangeofissuesthatneedtobeincorporatedgoesfarbeyondthequestionofconstructionalmeaning.

TheconceptualunderpinningsofCxGarethusnaturallycompatiblewiththegoalsoffunctionallyorienteddiachronicanalyses,primarilyduetothefollowingfeatures:

•allowingthepossibilitythatconstructionsmaintain(atleastacertaindegreeof)internalcomplexity,regardlessoftheirnoncompositionalmeaning/function;

•incorporatingfeaturesthatconstrainaform'sbehaviorinlargerstructures;•allowingunderspecificationatanylevelofrepresentation,thusaccommodatingthedescriptivechallengesin

Page 316: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

Page 13 of 14

dealingwithincompletelyattestedlanguages,aswellasgradientcategoriality;

•treatingconstructionsasmultilayeredfunctionalprototypes(i.e.,‘blueprints’definedbyclustersoffeatures)thatcanbestretchedandtheirshapenegotiated,throughnovelusesinactualcommunication;

•incorporatingrecurrentpragmaticfeaturesandcontextualconstraintsifwarrantedbythedata.

ItisthesefeaturesthatmakeCxGhelpfulinconceptualizingthegradualnessofchangeandformulatingmorepreciselyourhypothesesaboutit,sothatwecanbesystematicwithoutsacrificingthecomplexityoftheprocesses.Inparticular,wecanmoreeasilyaddressthefollowing:

•theincremental(feature-by-feature)natureofvariationandchange;•therelationshipbetweenpartialtransitionsandalargerdiachronicshiftatasufficientlevelofgranularity;•potentialmismatchesbetweenschematicgrammaticalpatternsandthewordsthatfillthem(thusallowingfornoncompositionality);

•theroleofpragmaticandsemantictriggersofnovelinterpretations.

TheprinciplesofconstructionalchangecapitalizeonthefactthatCxGtreatsgrammaticalknowledgeastheresultofagradualconventionalizationofpatternsofunderstanding,inwhichmorphosemanticstructure,syntacticfunction,communicativefunction,andlexicalmeaningformanintegratedwholeandlinguisticchangecaninvolveanysubsetoftheseaspects.Thenatureanddetailsoftheintegrationanditsmanifestationsintheemergenceofnewlinguisticstructureareonlybeginningtobesystematicallyaddressed,butafirmdirectionforfurtherresearchhasbeencharted.Moreover,CxG-basedanalysesholdpromiseforgivingsufficientprominencealsototopicsthathavenotattractedasmuchattentionyet(comparedtothepreoccupationwithissuessurroundingtheevolutionofgrammaticalmarkers),forboththeoreticalandmethodologicalreasons:focusonthediscoursaloriginoflinguisticchangeandtheevolutionofnonpropositional(pragmatic)meanings,suchastheemergenceofpragmaticparticles.Allofthisstillawaitsseriousresearch.

Notes:

(1.)ItbearsstressingthatLehmann'suseofthewordconstructionreflectsthetraditional,nontheoreticalsenseof‘syntagmaticstring.’Itisnotmeantinthetechnicalsenseof‘form-meaningpairing’asitisunderstoodanddefinedinConstructionGrammar.

(2.)ThisislaterechoedalsoinCroft's(2001:18–19)formulationthatconstructionshave“(conventional)meaning,”wheremeaningisdescribedas“alloftheconventionalaspectsofaconstruction'sfunction.”Stressingthefunctionaldimensionsisimportantespeciallyfordealingwithhighlyschematicgrammaticalconstructions,wheretheterm‘meaning’canbemisleading.

(3.)Theuseofthecapitallettersrepresentslexemesandindicatesthattheprocessesinquestionapplytoallmorphologicalformsofagivenlexeme.

(4.)Thisaccountmaydifferfromthewayconstructionalstatusistreatedandassignedinexemplar-basedmodelsofchange,butsuchdifferencesdonotinvalidatethebasicclaimthatchangeoriginatesinspecificusage(constructs),notinconstructions(abstractpiecesofgrammar).Sortingoutthedetailsofmodel-specificalternativesisnot(andcannotbe)theconcernofthischapter,butIsuspecttheapparentdifferencesmaybemoreterminologicalthanconceptual.

(5.)Itseemsthatmicro-constructionsmightcorrespondroughlytowhatotheraccountsrefertoas“substantiveconstructions”(cf.T.Hoffmann2008)andwhatinfrequency-basedmodelsisdescribedasconstructionsthatarefullylexicallyfilled(e.g.,Bybee2006),thatis,repeatedlyco-occurringsequencesofwords.However,therehasnotbeenmuchdiscussionconcerningtheterminologicalpracticesassociatedwithdifferentaccounts(whichalsoreflecttheoreticaldifferencesacrossvariousschoolsofthought)anditisnotmygoalheretoengageincomparingtheircontentorinevaluatingtheirrelativemeritandmutualcompatibility.

(6.)Thepresentdiscussionisbasedonseveralpartialstudiesofthematerial(e.g.,Fried2007a,2008,2010a);theinterestedreaderisreferredtothesestudiesforfullanalysisandargumentation.

Page 317: The Oxford Handbook of Construction Grammar

Principles of Constructional Change

MirjamFriedMirjamFriedisProfessorofLinguisticsattheCharlesUniversityPrague.SheobtainedherPh.D.inGeneralLinguisticsfromtheUniversityofCaliforniaatBerkeleyandhastaughtatseveralAmericanuniversities(UCBerkeley,UniversityofOregon,PrincetonUniversity).Herresearchfocusesonthecognitiveandfunctionalgroundingoflinguisticstructure,particularlyinmorphosyntax(e.g.casemarkingalternations,subordination,gradientcategoriality,interactionbetweenmorphologicalstructureandsyntaxinlanguagechange).

Page 318: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 1 of 16

PrintPublicationDate: Feb2013 Subject: Linguistics,HistoricalLinguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0024

Construction-BasedHistorical-ComparativeReconstructionJóhannaBarðdalTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterexaminestheapplicationoftheconstructionalapproachtosyntacticreconstruction.Itarguesthataconstructionalapproachtolanguageisanoptimaltheoreticalframeworkforreconstructingsyntax,andexplainsthataconstructionalapproachovercomessomeoftheallegeddifficultieswithsyntacticreconstruction.Thechapteroutlinesthebasicpremisesofhistorical-comparativereconstructionandhowthecomparativemethodworksinpractice.ItalsoshowshowConstructionGrammarmaycontributetohistorical-comparativesyntacticreconstructionbyreconstructingoneparticularargumentstructureconstructionforProto-Indo-Europeanlanguage.

Keywords:syntacticreconstruction,constructionalapproach,historical-comparativereconstruction,comparativemethod,ConstructionGrammar,Proto-Indo-European

24.1.Introduction

Inthehistoricallinguisticenterprise,differentgoalsareassociatedwithhistorical-comparativereconstruction.Onegoalistocontributetoincreasedknowledgeofpossibleorrealprehistoricstagesoflanguages.Anotheristoaidintheinvestigationofmechanismsoflanguagechange(cf.FerraresiandGoldbach2008).Athirdoneistothrowlightonpossibledevelopmentalpathsofcertainsynchronicstructures,notnecessarilytorevealanyprincipledmechanismsoflanguagechangebutratheroutofinterestfortherelevantsynchronicstructuresandtheirevolutionaryhistory(cf.BarðdalandEythórsson2012a).

Historical-comparativereconstructionhastraditionallyfocusedonlexical,morphological,andphonologicalcomparisons,whilesyntacticreconstructionhaseitherbeensystematicallyleftunattended,regardedasfruitlessoruninteresting,orevenrebuked(cf.,interalia,Watkins1964;Jeffers1976;Lightfoot1979,2006;Harrison2003;PiresandThomason2008;vonMengden2008).Thereasonforthisisthatsyntacticstructureshavebeenregardedasfundamentallydifferentfrom,forinstance,morphologicalstructures,inseveralrespects.Thatis,syntacticstructuresarelargerandmorecomplexunitsthanmorphologicalunits.Semanticallytheyhavenotbeenregardedonparwithmorphologicalunitseither,inthattheirmeaningisregardedasthesumofthemeaningofthelexicalpartsthatinstantiatethem,andbecauseofthissemanticcompositionalitytheyarenottreatableasarbitraryform–meaningpairingslikewords(cf.Klein2010).Ithasalsobeenarguedintheliteraturethatsyntacticstructuresarenotinheritedinthesamewayasthevocabulary(Lightfoot1979andlaterwork),thatthereisnocognatematerialtocomparewhencomparingsentencesacrossdaughterlanguages(Jeffers1976),thereisnoregularityofsyntacticchange,asopposedtotheregularityofphonologicalchange(Lightfoot2002;PiresandThomason2008),andthatthereisnoarbitrarinessfoundinsyntax(Harrison2003),allofwhichrendersyntacticreconstructionfundamentallydifferentfromphonologicalreconstruction(seefurtherdiscussioninsection24.3below).

ThegoalofthischapteristodiscusswhatConstructionGrammarhastooffertotheenterpriseofhistorical-comparativereconstruction.IwillarguebelowthatthebasicpremiseofConstructionGrammar,thatthelinguistic

Page 319: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 2 of 16

systemconsistsofform–meaningpairings,namelyconstructions,arrangedinasystematicwayinaConstructicon,makesConstructionGrammaranoptimaltheoreticalframeworkforreconstructingsyntax.BecausewordsandlargersyntacticstructuresaretreatedalikeinConstructionGrammar,thatis,asform–meaningpairings,theframeworkofConstructionGrammarcaneasilybeextendedfromitscurrentareaofcoveragetohistorical-comparativereconstructionthroughtheComparativeMethod.Thereisanaturalleapfromsynchronicform–meaningpairingstohistoricalreconstruction,basedonform–meaningpairings.Iarguethatthisleapismorecredibleforframeworkswhichassumethatmeaningandformareintegratedwitheachotherthanforframeworkswhichseparatemeaningfromformandviewthetwoasdistinctautonomoussystems,asinthe(early)generativetradition(Chomsky1981).Inthatsense,ConstructionGrammarisadynamiclinguisticframework,easilyextensibletonewlinguisticareas.

Ibeginbygivinganoutlineofthebasicpremisesofhistorical-comparativereconstructionandhowtheComparativeMethodworksinpractice(section24.2).ThenIdiscussandrefutesomeofthemisconceptionsofwhysyntacticreconstructionisuntenable(section24.3).Finally,insection24.4,IshowhowConstructionGrammarmaycontributetohistorical-comparativesyntacticreconstructionbyreconstructingoneparticularArgumentStructureconstructionforProto-Indo-European,namelyasubconstructionofargumentstructureswithnoncanonicalsubjectmarking,showinghowsuchstructuresmustbeassumedtohaveexistedintheIndo-Europeanproto-language.

24.2.TraditionalHistorical-ComparativeReconstruction

Historical-comparativereconstructionhasitsrootsinthenineteenth-centuryNeogrammariantraditionandithasmostlyfocusedoncomparinglexicalitems,morphemes,andsoundsacrossrelatedlanguages.Thegoalhasbeentoreconstructcommonproto-forms,fromwhichthedescendentformscanbederived.

Thefirsttaskistosetupcorrespondencesets,thatis,equivalentmaterialfrommorelanguages.Theinputforthecorrespondencesetscanconsistofsounds,morphemes,orwords.ConsiderTable24.1wherethreeIndo-Europeancognatesmeaning‘Icarry,’‘brother,’and‘brow’arelisted(MalloryandAdams2006:41).SeveralcorrespondencesetscanbededucedfromthetablebutforthepresentpurposesIwillletitsufficetodiscussthestemvowelthatalternatesbetween-e-and-a-for‘Icarry,’asisshowninTable24.2.

Inthiscasethemajorityofthelanguagesshows-e-andaminorityshows-a-intheirstems.AsithasbeenpresumedthatSanskrithaschangedleastfromthecommonIndo-Europeanproto-language,theearlyIndo-Europeanistsreconstructed*ahereonthebasisoftheSanskritform.Lateron,however,acomparisonoftheIEvelarsandtheirfollowingvowelsrevealedadistinctionofvelarsinto-k-and-c-inSanskrit,correlatingwiththequalityofthevowelinLatinandGreek.Thatis,when

Page 320: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 3 of 16

Table24.1.ComparisonofthreeIndo-Europeanwords

‘Icarry’ ‘brother’ ‘brow’

OldIrish beru brāthair forbrū

Latin ferō frāter ––

OldEnglish bere brōðor brū

Lithuanian –– broterė̃lis bruvìs

OldChurchSlavic bero̧ bratrŭ brŭvĭ

Greek phérō phrḗtēr ophrûs

Sanskrit bhárāmi bhrā́tar- bhrū̃-

TocharianB parau procer pärwāne

Table24.2.Correspondencesetsforthestemvowelof‘Icarry’

Alt.1 Alt.2

OldIrish Sanskrit

OldLatin e TocharianB a

OldEnglish

OldChurchSlavic

Greek

abackvowelfollowedavelarinLatinand

Greek,thevelarwas-k-inSanskrit,whereaswhenafrontvowelfollowedavelarinLatinandGreek,thevelarshowedupas-c-.ThisdifferentiationofvelarsinSanskritsuggestsanearlydistinctionbetweenfrontandbackvowels,orbetween-e-and-a-,duringanearlierprehistoricperiodofSanskrit,supportedbythecomparativematerial,asthecombinatorychangeofvelarsfrom-k-to-c-withfrontvowelsiswellknownfromthehistoryofotherlanguages.ThevowelsinSanskritwouldthenhavemerged,leavingthevelarsintact.

ThiscomparisonoffrontandbackvowelswithvelarssuggeststhatitisinfactSanskritthathaschangedwhiletheremainingIndo-Europeanlanguageshavenot.AfurthercomparisonrevealsthatSanskrithas-a-systematicallywhereLatinandGreekhave-e-and-o-.TheresultofthisisthatnowadaysIndo-Europeanistsreconstruct*eforLatinandGreek-e-andSanskrit-a-,*oisreconstructedwhereLatinandGreekhave-o-andSanskrit-a-,while*aispreservedforwordswhereallthreelanguageshave-a-(MalloryandAdams2006:43–44).

Themethodthatisusedtoarriveatthesecommonproto-vowelsandproto-forms,theComparativeMethod,presupposeslinguisticentitiesthathaveaformsideandameaningside,thatis,entitiesthatarearbitraryform-functioncorrespondences,likeinthiscase‘Icarry.’Thisisbecausebeforereconstructioncanbecarriedout,cognatesmustbedeterminedandgeneticrelatednessestablished.Cognatelexicalitemsareitemsthatareinheritedfromaparentlanguage,andthusshowsimilarityinformandmeaningacrossthedaughterlanguages.Generally,lexicalitemsintwoormorelanguagesthatshowsimilarityinformandmeaningmusteitherbeinheritedfromanancestorlanguageorborrowed.Inotherwords,twoormorelanguagesthatsharecognatesareeither

Page 321: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 4 of 16

geneticallyorareallyrelated.ThismeansthatifLatinferō,Greekphérō,andSanskritbhárāmididnotallmean‘Icarry,’despitethesimilarityinform,theywouldnotformanybasisforreconstruction.

Theargumentationaboveisbasedontheassumptionthatitispossibletodistinguishbetweeninheritedandborrowedlexicalitems.Thiscanbedoneinpartonthebasisoftheirform,ascognateitemsaretheresultofphonologicaldevelopments,whichmaybedifferentacrossdifferentbranchesofgeneticallyrelatedlanguages,whileborroweditemsarephonologicalreflectionsoftheirsourceitemsinthegiverlanguage.Compare,forinstance,Englishshirtandskirt,ofwhichtheformerisinheritedandconformstoknownphonologicalrulesinthehistoryofEnglish,[sk]>[ʃ],whilethelatterwasborrowedfromOldNorseskyrta.ThelexemeskirtwasfirstdocumentedinEnglishtextsduringthelatterpartofthethirteenthcentury(OED),whilethechangefrom[sk]to[ʃ]tookplaceintenth-centuryEnglish(Minkova2003:130ff.).

Moreover,lexicalitemsthatarecognatesareexpectedtoshowupwithsharedsemantics,inheritedfromanearlierstageofthelanguage,resultinginsemanticsimilaritiesofcognatesacrossrelatedlanguages,orsomerecognizedsemanticextensionsorrelations.Thismaybecontrastedwithalexicalitemthathasbeenborrowedandincidentallyshowsphonologicalsimilaritieswithanalreadyexistinginheritedvocabularyitem.Insuchacase,onecanlooktothesemantics,asthereisnoreasontoassumethatphonologicalsimilarityisconcomitantwithsemanticsimilarity,unlessofcoursethephonologicalsimilarityisnotaccidentalandweare,infact,dealingwithinheritedvocabularyandnotborrowing.Ifthelexicalitemhasbeenrecentlyborrowed,itshouldalsocloselyreflecttheoriginalsemanticcontentofthecorrespondingiteminthesourcelanguage.

Evenborrowingsbetweengeneticallyrelatedlanguagescanbeidentified,astheexamplewithshirtandskirtaboveshows.However,ifsuchborrowingsareancient,itmaybemoredifficulttoteaseinheritanceandgeneticborrowingapart,astheborroweditemswillhaveundergonephonologicalchangesthatmaymakeitmoredifficulttodistinguishbetweenthemandinheriteditems.Also,borrowingbetweengeneticallyrelatedlanguagesmayinvolvecognatematerial,asinthecaseofshirtandskirt,whichagainmakesitdifficulttousesemanticsascriterialwhendistinguishingbetweengeneticborrowingsandinheritedmaterial(seealsoBowern2008oncriteriafordistinguishingbetweensyntacticchangeandsyntacticborrowing).

24.3.IsSyntacticReconstructionPossible?

Thedebateonthereconstructabilityofsyntaxintheliteraturehasfirstandforemostbeenfocusedonparticularphenomenainthedaughterlanguageswherethereisdiscrepancybetweenthedaughters.ThetextbookexampleisthereconstructionofwordorderinIndo-European(cf.Delbrück1893–1900;Watkins1964),wherethedaughtersvarybetweenOVandVOwordorder.AnotherexampleisfoundinHarrisandCampbell(1995:352–53)wheretheFinnishPassiveConstructioniscomparedbetweenStandardFinnish,inwhichtheagentcannotbeexpressed,andAmericanFinnish,wheretheagentispermitted,inparticularifitisaninstitution.Similarly,PiresandThomason(2008:59–66)discusstheRomancefuture,whichissyntheticinasubsetoftheRomancelanguages,butperiphrasticinothers.Alltheseexamplesentailsomelanguagechange.

However,syntaxandsyntacticstructuresaremuchlesspronetochangethanphoneticandphonologicalstructures,asdiscussedbyBlust(1996),basedontheideasofHermannPaul(1886).Thesamepoint,thatthereisagreatamountofstabilityinsyntaxovertime,hasalsobeenexplicitlymadebyBarðdalandEythórsson(2003),JandaandJoseph(2003),Keenan(2003),andNichols(2003).Iwillbearguingbelowthatreconstructingstability,orstructuresthatarestablethroughtime,isnotabanalitybutagenuinecontributiontohistoricallinguisticsandacontributiontoourknowledgeaboutearlierprehistoricstagesofthelanguagesinvestigatedandinparticularthehistoryofthestructuresunderinvestigation.

Therearefirstandforemostthreeissuesthathavebeendiscussedintheliteratureaspertainingtoaprincipleddistinctionbetweensyntaxandphonology,highlyrelevantfortheperceivedlackofreconstructabilityoftheformer,asopposedtothelatter.Thesethreeissuesrelatetotheconceptsofcognatestatus,arbitrariness,andregularity.Iwillnowdiscusseachinturn.

Ithasbeenarguedbyseveralhistoricallinguistsandsyntacticiansthatsyntacticstructurescannotbereconstructedforproto-stagesbecauseofaprincipleddistinctionbetweensyntaxandphonology/morphology,amongotherthingsbecausetherearenocognatesinvolvedinsyntacticcomparisons(interalia,Jeffers1976;

Page 322: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 5 of 16

Winter1984).Thisissimplywrong.ThetermcognateinLatinmeans‘ofcommondescent,bloodrelative’andhasinhistoricallinguisticsbeenusedtomeanthataunitfoundinmorethanonedaughterlanguagesis‘descendentfrom’orinheritedfromacommonproto-stage.Clearlynotonlywordformsandmorphemesmaybeinheritedfromanearlierlanguagestagebutalsolargerandmorecomplexstructureslikesentencestructures.Thisapplies,forinstance,toArgumentStructureconstructions,asthesemaywellbeinheritedfromacommonproto-stageandthusbefoundinmorethanoneofthedaughterlanguages.

Giventhatargumentstructuresmaybeinheritedfromearlierlanguagestages,suchargumentstructureconstructionsqualifyascognateargumentstructureconstructions.Kikusawa(2003)arguesfortheappropriatenessofthetermcognatestructureswhendealingwithlargerandmorecomplexunitsthanmorphologicalunitsinthehistoryoftheIndonesianlanguages.Sheshowsinacarefulmannerwhichsyntacticstructures,inheritedfromanearlierstage,correspondtowhichsyntacticstructuresinthemodernlanguages.BarðdalandEythórsson(2012a)introducethetermcognateargumentstructureconstructionsandclaimthatsuchcognateconstructionsintheIndo-Europeandaughterlanguagescaneasilybeidentifiedthroughthecasemarkingofthedirectarguments,aswellasthroughacomparisonofthelexicalpredicatesinstantiatingthem.

Tables24.3–24.5listsomeoftheArgumentStructureconstructionsofoneandtwoplacepredicatesfoundinthreeancientIndo-Europeanlanguages,OldNorse-Icelandic,Latin,andAncientGreek.Allthreelanguages,forinstance,haveNom-Acc,Nom-Dat,andNom-Genconstructions.TwolanguageshaveNom-PP/S,Acc-PP/S,Dat-PP/S,andGen-PP/S.AllthreehaveAcc-Nom,Dat-Nom,andGen-Nomconstructions.And,finally,allthreelanguageshaveaDat-Genconstruction.ThemaindifferencesbetweenthethreelanguagesarethatonlyOldNorse-IcelandichasanAcc-Accconstruction,notfoundinLatinandAncientGreek,andAncientGreekdoesnotsharetheAcc-GenconstructionwithOldNorse-IcelandicandLatin.

Table24.3.CaseandargumentstructureconstructionsinOldNorse-Icelandic

Nom Acc Dat Gen

Nom Acc Dat Gen

Nom-Acc Acc-Nom Dat-Nom Gen-Nom

Nom-Dat Acc-Acc Dat-Gen Gen-PP

Nom-Gen Acc-Gen Dat-PP Gen-S

Nom-PP Acc-PP Dat-S

Nom-S Acc-S

Toillustratethenatureofthesedatatotheuninitiatedreader,examplesoftheDat-NomconstructioninOldNorse-Icelandic,Latin,andAncientGreekaregivenin(1–3)below:

(1)hvártsemmérangrarreykreðabruniOldNorse-Icelandicwhetherorme.datbotherssmoke.nomorfire.nom‘irrespectiveofwhetherIambotheredbysmokeorfire’(Njála130)(2)numtibistultitia  accessitLatinptcyou.datstupidity.nomcomes.to.3sg‘haveyoubecomeafool?’(Plautus)(3)outegarmoiPolykratēsēreskedespozōnAncientGreekand.notforI.dat  Polykrates.nom pleased.3sgruling.nom‘forIdidnotlikePolykrateswhenhewasruling’(Hdt.3.142)

LetusnowcomparethecaseandargumentstructureconstructionsfromOldNorse-Icelandic,AncientGreek,and

Page 323: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 6 of 16

LatininTables24.3–24.5withanequivalenttablefromJapanese,whichisalsoaNominative–Accusativelanguagewheredativesubjectsarefound.AsTable24.6shows,therearenocaseandargumentstructureconstructionsinJapanesewithaccusativeorgenitivesubjects,butonlynominativeanddativesubjects.Inaddition,JapanesedoesnothavethesamearrayofDative-SubjectconstructionsfoundinOldNorse-Icelandic,Latin,andAncientGreek.

Table24.4.CaseandargumentstructureconstructionsinLatin

Nom Acc Dat Gen

Nom Acc Dat Gen

Nom-Acc Acc-Nom Dat-Nom Gen-Nom

Nom-Dat Acc-Acc Dat-Gen Gen-PP

Nom-Gen Acc-Gen Dat-PP Gen-S

Nom-PP Acc-PP Dat-S

Nom-S Acc-S

Table24.5.CaseandargumentstructureconstructionsinAncientGreek

Nom Acc Dat Gen

Nom Acc Dat Gen

Nom-Acc Acc-Nom Dat-Nom (Gen-Nom)

Nom-Dat Acc-Acc Dat-Gen (Gen-PP)

Nom-Gen Acc-Gen Dat-PP (Gen-S)

Nom-PP (Acc-PP) Dat-S

Nom-S Acc-S

Page 324: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 7 of 16

Table24.6.CaseandargumentstructureconstructionsinJapanese

Nom Acc Dat Gen

Nom

Nom-Acc Dat-Nom

Nom-Dat

Dat-PP

Nom-PP Dat-S

Nom-S

Thatis,intransitivestructureswithadative

subjectandDat-GenaretotallyabsentinJapanese.

Thelargedegreeofconformityofargumentstructureconstructionswithtwo-placepredicatesbetweenOldNorse-Icelandic,AncientGreek,andLatinisofcoursenotsurprisinggiventhattheselanguagesaregeneticallyrelated.Observe,however,thatifonedoesnotfindthisconformitysurprising,thatmustbebecauseoneassumesthattheseargumentstructureconstructionsareinfactinheritedfromacommonproto-stage.Howelsewouldoneexpectsuchconformityofargumentstructureconstructionsacrossgeneticallyrelatedlanguages?ThequestionisparticularlyprotrudingsincemanyoftheseargumentstructureconstructionsseemtobeconfinedtotheIndo-Europeanlanguages.Letmeaddherethatthecasemarkersarealsogeneticallyrelatedacrosstheselanguages,thatis,theyarealsocognates.

Thereis,moreover,nodoubtthatnotonlytheverbstembutalsothecaseframeofcognateverbsin,forinstance,GermanicaresharedacrosstheGermaniclanguages.Forinstance,theverb‘help,’hjálpainIcelandicandhelfeninGerman,selectsfortheNom-DatconstructioninbothmodernlanguagesandselectedforthiscaseframeinalltheOldGermaniclanguages.

Theverb‘like’selectedfortheDat-NomcaseframeinOldNorse-Icelandic,OldEnglish,OldHighGerman,andGothic(cf.EythórssonandBarðdal2005;BarðdalandEythórsson2012a).IthasbeendocumentedintheliteraturethatArgumentStructureconstructionsmaybeborrowedacrosslanguages,especiallytogetherwiththeirrelevantverbs(seeBarðdal1999,2008:chapter3),butinthiscase,itisclearfromtheformoftheverb(i.e.,líkainOldNorse-Icelandic,licianinOldEnglish,lîchêninOldHighGerman,andleikaninGothic)thatwearedealingwithinheritedvocabularyandnotborrowings.

SimilarclaimscanbemadeaboutseveralpredicatesthatarefoundacrosstheearliestGermaniclanguages,includingpredicateswherethesubjectlikeargumentisnotinthenominativecase,like‘hunger’,‘thirst’,‘lust’andmanyothers(seeBarðdalandEythórsson2012b).

Theseexampleswith‘help’,‘like’,‘hunger’,etc.showbeyonddoubtthatnotonlymorphemesandlexicalitemshavecognatesinrelatedlanguages,butthatcaseframesorargumentstructureconstructionsarealsoinheritedfromtheirparentlanguageandcanbeidentifiedassuch.Hence,theargumentationthatsyntacticreconstructionisruledoutbecauseoflackofcognatemateriallacksforceentirely.

Turningnowtoarbitrariness,itisgenerallyassumedthatreconstructioninvolvingarbitrariness,thatis,reconstructionbasedonarbitraryfeaturesisthehallmarkofgoodreconstruction.Thelogicbehindthisviewisthatanarbitraryform–meaningpairingisneededinordertoensurecognatestatusandruleoutchanceresemblances.Thisisbasicallyhowlexicalitemsarereconstructed.Therefore,ifarbitrarinessisnotfoundinsyntax,asisclaimedbyHarrison(2003:223),thatwouldweakenthequalityofthereconstruction,giventhisargumentation.However,thisassumptionisfaultyforthefollowingtworeasons.Tobeginwith,arbitrarinessisfirstandforemostneededtoestablishgeneticrelatedness,asHarrisonpointsouthimselfinadifferentsectionofhis(2003:216)article.This

Page 325: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 8 of 16

meansthatthearbitrarinessrequirementissimplynotneededforsyntacticreconstruction,becausesyntacticreconstructioniscarriedoutaftergeneticrelatednesshasalreadybeenestablishedanywaythroughphonological,morphological,andlexicalcomparisons.Withwordsorspecificmorphemes,however,theform–meaningpairingisentirelyarbitraryandthisisneededasafirmgroundforestablishinggeneticrelatedness.

Second,arbitraryform–meaningpairingsarealsofoundinsyntax,contraHarrison'sclaims.OneofthemajorcontributionsofConstructionGrammartocurrentlinguisticresearchhasbeentodocumentandshowthatsubstantialpartsofourgrammaticalpatternsarearbitraryinthesensethattheyaresemanticallynoncompositional,asopposedtosemanticallycompositional,wherethemeaningofthewholecanbederivedfromthemeaningoftheparts(cf.Fillmore,Kay,andO’Connor1988;Nunberg,Sag,andWasow1994;Goldberg1995;andothers).Thesehavealsobeenreferredtointheliteratureassemanticallyspecificvs.semanticallygeneralconstructions(Tomasello1998a;CroftandCruse2004;Barðdal2001,2008,2009;Barðdal,Kristoffersen,andSveen2011;BarðdalandEythórsson2012a).Oneexampleisthe“What'sXdoingY?”construction,exemplifiedbythestringWaiter,what'sthatflydoinginmysoup?,wherethespeaker'sintentionisnottoaskarealquestion,butrathertosignalsomeincongruity(KayandFillmore1999).Thisexampleshowsthatarbitraryform–meaningpairingsarecertainlyfoundinsyntaxandarenotconfinedtothelexicon.

Turningnowtothelastconcept,regularityofphonologicalchange,withintheIndo-Europeanparadigmthisregularityisgenerallyperceivedofintermsofsoundlaws,whichhavebeenregardedasregularprocessesaidinginphonologicalandmorphologicalreconstruction,notonlybytheNeogrammariansthemselvesbutalsobythecontemporaryhistoricallinguisticcommunity.Theregularityofthesoundlaws,however,isgrosslyoverstated.Firstofall,notallwordscontainingtherightmorphophonologicalstructureneededasinputforaparticularsoundlawhaveundergonethephonologicalprocess.IquoteHarrison(2003:220,emphasisinoriginal):

Iusetheterm[regularity]assumptionherequitepurposefully,becauseitisbynowwelldemonstratedthatsoundchangeisnotregular,intheusualintendedsense,butprecedesinaquasi-wavelikefashionalongthesocialandgeographicdimensionsofthespeech-community,andthroughthelinguisticsystemitself.Atanygivenpointintime,aparticularsoundchangemaybefeltonlyinapartofthespeech-communityand,ifitaffectslexicalsigns,onlythroughaportionofthelexicon.

Second,asobservedalreadybyHoenigswald(1987),thesoundlawsareonlyregularbydefinition.Thatis,allirregularitiesandlessregularprocessesaresystematicallyexcludedfromthisnotion.Thephonologicalprocesseswhereexceptionscanbefoundaresimplynotlabeled‘regular’ordefinedas‘soundlaws.’

Third,thedirectionalityofsoundchangeisnotagivenbutmustbedecideduponthroughfurtherexaminationofthelanguagedata.Forinstance,letusassumethatwewerestartingworkonalanguagefamily,previouslyunencumberedbytheinterferenceoffieldlinguistsandanthropologists,andhenceundescribed.Letusassumethatwefindasystematiccorrespondencebetween/a/and/o/inthedaughterlanguages,sothathalfofthelanguageshave/a/andtheotherhalfshowsupwith/o/intherelevantcognates.Itisimpossibletodecideonthebasisofsuchdatawhether/a/haschangedinto/o/or/o/into/a/inthehistoryofthislanguagefamily.Bothchangesareentirelypossible.

Thishypothesizedvariationbetween/a/and/o/isexactlyparalleltothesituationwithOVandVOwordordersinIndo-European.SomeoftheancientIndo-EuropeanlanguageshaveOVorder,whileothershaveVOorder,anditisnotclearonthebasisofthissimplecorrespondencesetwhichordertoreconstructforProto-Indo-European.Itisonlybyinvestigatingthecontext,theenvironment,andwhateverotherrelevantconditionsthatallowustouncoverwhichvarianttoreconstructandwhichnot,thatwemaybeabletoargueforaparticularchoice,andthisistrueforbothphonologicalandsyntacticchange.

Clearlywhenitcomestocombinatorychanges,thatis,changesofsoundsduetoothersoundsintheirimmediatephonologicalenvironment,likewiththeabove-mentionedvelarsinSanskrit,suchchangestakeplaceinacertaindirection,whilethedirectionalityofsystemicchangesmaybequiterandom.Therefore,theso-called‘regularity’ofphonologicalchangeisillusionaryandallchanges,betheyphonologicalorsyntactic,requireacarefulinvestigationoftherelevantlanguagedata,adetailedexaminationoftheconditionsforthechange,andawell-argued-foranalysisofthechoicesforthereconstruction.

Finally,andquiteimportantly,Harrison(2003)pointsoutthatthesoundlawsarebasicallystand-insforasimilarity

Page 326: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 9 of 16

metric,whichisneededtoestimatetherelativesimilarityofformsinordertodistinguishbetweencognatesandnoncognates.Asalreadydiscussedabove,theComparativeMethodistotallydependentoncognates,thatis,form–meaningpairings,inordertoestablishgeneticrelatedness.Ascognateformsareseldomcompletelyidenticalacrossrelatedlanguages,oneneedsanindependentlymotivatedsimilaritymetrictodecideuponwhentwoformsaresimilarenoughtocountascognatesandwhennot.Developingsuchasimilaritymetricisnotaneasytask.Thesoundlaws,however,dothisjobinstead.Whencomparingformsintworelatedlanguages,andthedifferencesbetweenthemcanbeexplainedintermsofanestablishedsoundlaw,thatinitselfcountsasavalidationoftheassumptionthatthetwoformsareinfactcognates.Therefore,becauseoftheperceivedexistenceofthesoundlawsandtheirroleinidentifyingcognates,anindependentlymotivatedsimilaritymetricisnotneeded.Thisisoneofthereasonsthatthesoundlawshavebecomesoimportantinphonologicalandmorphologicalreconstruction.

However,anylackofcorrespondingsyntacticlawsdoesnotmeanthatsoundlawsandtheperceivedregularityofsoundchangemakephonologicalandmorphologicalreconstructiondifferentinprinciplefromsyntacticreconstruction.Thereareothermeansofestablishingwhether,forinstance,ArgumentStructureconstructionsarecognates,likeforinstancewhethertheyshareacaseframewherethecasemarkersarecognates,thatis,thattheyarenotonlyfunctionalequivalentsbutalsomorphologicalcognates.Itisalsopossibletoidentifydevelopmentalpathsinsyntax,exactlylikeinphonology,likeforinstancethatpartitivecaseisknowntodevelopfromeithergenitivecaseoralocative/ablativecaseandnotviceversa(HarrisandCampbell1995:362–63).Anddiscoursemarkersoftenoriginateasadverbswithscopeoverlexicalitems,thentheygoontoacquirescopeoverlargerphrases,andeventuallywholeutterances(cf.BrintonandTraugott2005:136–40;Fischer2007:274ff.).

Yetanotherissuethatneedstobetakenintoaccountisstabilityovertime,whichismuchmoreprofoundinsyntaxthaninphonology.Infact,theroleofstabilityinsyntaxispresumablyavastlyunderestimatedfactor,whichshould,inturn,contributetomakingsyntacticreconstructioneasierthanphonologicalandlexicalreconstruction.Inthatsense,stabilitymayberegardedasafactorthatcompensatesforlackofdirectionality.Whatisunclearatthispoint,however,givenourpresentknowledge,iswhattheproportionisbetweenstructuresthathavechangedandstructuresthathaveremainedstable.Ifstabilityisthenorm,asopposedtochangebeingthenorm,thensyntacticreconstructionispossibleina‘normal’case,inspiteoflackofdirectionality.

Inconclusion,thequalityofanyreconstructionisalwaysentirelydependentonthequalityoftheinputdataandthedetailsofourknowledgeofthelanguagesinquestionandtheirhistory.Ifsyntacticreconstructionislesssuccessfulthanphonologicalormorphologicalreconstructions,thisisbecauseofthenatureofthesyntacticinputandthestatusofouradditionalknowledgeaboutthelanguagesinquestion,anddoesnotreflectanyprincipleddistinctionbetweenphonologyandsyntax,whichsupposedlymakessyntaxlessreconstructableinprinciple.

Tosummarizethecontentofthissection,Ihavearguedthatcognatesarealsofoundinsyntax,forinstance,cognateArgumentStructureconstructionswhichareinheritedfromacommonproto-stage,andareidentifiablethroughthecasemarkingofeachargumentstructureandthroughacomparisonofthelexicalpredicatesinstantiatingthem.Ihavealsoarguedthatarbitraryform–meaningpairingsarefoundinsyntax,exactlyasinmorphology,althoughnotallform–meaningpairingsinsyntaxarearbitrary.However,theissueofarbitrarinessisanonissue,asarbitraryform–meaningpairingsareonlyneededwhenestablishinggeneticrelatedness,andasgeneticrelatednesshasusuallyalreadybeenestablishedonothergroundswhensyntacticreconstructionstartsthereisnoneedforarbitrarinesshere.Finally,Ihavearguedthattheregularityofphonologicalchangeisgrosslyoverstated,asallirregularandlessregularphonologicalchangesareexcludedfromthenotionofsoundlawsbydefinition.Eventhoughnoequivalent‘syntacticlaws’havebeenidentified,researchonsyntacticchangeuncoversmoreandmorecrosslinguisticallyidentifiabledevelopmentalpaths,whichofcourseaidinreconstructingsyntax.Thelastfactorofimportancehereisstability,assyntacticstructurestendtobemuchmorestablediachronicallythanphonologicalstructures,afactorthatmaycompensatefortheperceivedlackofdirectionalityinsyntacticchange,andmakesyntacticreconstructionamoredesirableenterprise.

InowturntothequestionofwhatConstructionGrammarhastoofferonhistorical-comparativereconstruction.

24.4.ReconstructingGrammarinCxG

Intheirexaminationofthevalidityofsyntacticreconstruction,PiresandThomason(2008:43–44)discussHarrisandCampbell's(1995)andHarris's(2008)approachandcriticizethemfortheirimprecisedefinitionof‘syntactic

Page 327: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 10 of 16

patterns’and‘types.’WhileIhavenoproblemsunderstandingwhatHarrisandCampbellmeanwiththetermssyntacticpatternandtype(cf.alsoGildea1997,1998,2000),Iwouldliketopointoutthataconstructionalapproachtosyntacticreconstructiondoesnotencounterthatproblem,becauseofthevariousnotationalsystemsusedtoformalizethenotionofconstructionwithindifferentversionsofConstructionGrammar.Thatis,HarrisandCampbell'snotionsofasentencepatternwouldberegardedasanabstractform–meaningpairingofitsownonaconstructionalapproach,exactlylikealexicalitemisaform–meaningpairingofitsown,onlyitismorecomplexandperhapsonlypartiallylexicallyfilled.

TheclinesfromconcretelexicallyfilledtoschematicandfromatomictocomplexareshowninTable24.7fromCroftandCruse(2004:255),whereitisarguedthatthedifferencebetweenwordsandsyntaxisthatwordsareatomicform-functionpairings,whilesyntacticstructuresarecomplexandschematicform-functionpairings(cf.alsoGoldberg,thisvolume).Thedifferencebetweenthetwothusliesintheirdegreeofcomplexityandschematicity,aswellaswhethertheform–meaningpairingistotallyarbitraryornot,butnotinsyntacticstructuresbeingfundamentallydifferentlinguisticentitiesthanwords.Therefore,onaconstructionalapproachtosyntacticreconstruction,therecanbenoprincipleddistinctionbetweenlexicalitemsandmorecomplexsyntacticstructures,inthatbothqualifyasform–meaningpairings,whichistheunitofcomparandathattheComparative

Table24.7.Thesyntax–lexiconcontinuum(CroftandCruse2004:255)

Constructiontype Traditionalname Examples

Complexand(mostly)schematic syntax [SBJbe-TNSV-enbyOBL]

Complex,substantiveverb subcategorizationframe [SBJconsumeOBJ]

Complexand(mostly)substantive idiom [kick-TNSthebucket]

Complexbutbound morphology [NOUN-s],[VERB-TNS]

Atomicandschematic syntacticcategory [DEM],[ADJ]

Atomicandsubstantive word/lexicon [this],[green]

Method

requires.Inthatsense,ConstructionGrammariseasilyandstraightforwardlyextendedtotherealmofhistorical-comparativesyntaxandsyntacticreconstruction.

IntheremainderofthischapterIwillshowhowArgumentStructureconstructionsmaybereconstructed,focusinginparticularonArgumentStructureconstructionswithnoncanonicalsubjectmarkinginIndo-European(cf.alsoEythórssonandBarðdal2011;BarðdalandEythórsson2012a).Copiousamountsofinkhavebeenspilledoverthereconstructionofwordorder,whilethereconstructionofArgumentStructureconstructionsinIndo-Europeanhasnotfiguredinthisdebateatall,althoughsuchreconstructionshavesuccessfullybeenworkedoutforotherlanguagefamilies,likeKartvelianandNach-Daghestanian(HarrisandCampbell1995;Harris2008)andIndonesian(Kikusawa2003).ThisisallthemoresurprisingconsideringthefactthatthereconstructionofArgumentStructureconstructionsisbasedonmorphologicalconsiderations.Thatis,itmustinpartbebasedonthereconstructionoftherelevantpredicates,whichinstantiatetheconstruction,andinpartonthecaseframeoftheverb.Thecasemarkers,inturn,belongtoamorphologicalparadigmthatalsoconsistsofcognates.Therefore,thereconstructionofArgumentStructureconstructionsisalwayspartlybasedonmorphologicalconsiderations,whichiswhatbothWatkins(1964)andFox(1995:105)emphasizeasbeingimportantforreconstructioningeneral,andinparticularforsyntacticreconstructionifitistobesuccessful.

Observe,now,thatthereconstructionofArgumentStructureconstructionswouldaccountforasubstantialpartofthesyntax,orinotherwords,thegrammar,ofProto-Indo-European.Therefore,considerthefollowingattestedexamplesfromOldEnglish,OldNorse-Icelandic,andMiddleHighGerman:

(4)Wabiðþæmþesceal…OldEnglish

Page 328: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 11 of 16

woeisthem.datthatshall‘Woehavetheythatshall…’(Beowulf183)(5)núerossvei,ervérerumhjalplausarOldNorse-Icelandicnowisus.datwoe,aswearehelpless‘nowwehavewoe,aswearehelpless’(Ridd65 )(6)imniewürdewêMiddleHighGermanthem.datnotbecamewoe‘theydidnothavewoe’(Erec66,6693)

Thelexicalitemwoeisoriginallyaninterjectionoradverb,althoughithasdevelopednominalusesinmanyoftheGermaniclanguagesandadjectivalusesinEnglishandGerman(OED;Kluge2002).ThisinterjectionisfoundinallthemajorbranchesoftheIndo-Europeanlanguages,seeTable24.8,althoughtheArmenianvayandNewTestamentGreekòá,oùá,oùaíareassumedtobelaterdevelopments.

Theexamplesin(4–6)aboveconstitutetheinputforthecorrespondencesetgiveninTable24.9,onwhichbasisitispossibletoreconstructacompositionalpredicatewithadativemaleficiary,theverbbeandtheadverb/interjectionwoeforProto-Germanic.Themodernlanguagesthathavelosttheircasemarkinghavepreservedalexicalizedvariantoftheconstructionwithwoefocalizedinpreverbalpositionandtheexperiencerinpostverbalposition(ModernEnglishandModernNorwegian).Incontrast,ModernIcelandicandModernGerman,whichhavebothmaintainedtheircasemarking,havepreservedtheconstructionwiththedativecaseofthesubjectlikemaleficiary.Asthemaleficiaryisalwaysinthedativecasein

Table24.8.ThelexicalitemwoeinIndo-European

OldEnglish wā

OldSaxon wê

OldHighGerman wê

OldNorse-Icelandic vei

Gothic wai

Latin vae

OldChurchSlavic ouvy

Latvian vai

OldIrish fē

Avestan auuoi

34

1

Page 329: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 12 of 16

Table24.9.Germaniccorrespondencesetforthecollocation[dat-is-woe]

Alt.1 Alt.2

OldNorse-Icelandic [dat-is-woe]

OldEnglish [dat-is-woe]

MiddleHighGerman [dat-is-woe]

ModernIcelandic [dat-is-woe]

ModernEnglish [woe-is-me]

ModernNorwegian [ve-er-meg]

alltheOldGermaniclanguageswhere

thisconstructionisfound,thereisnodoubtthatthisconstructionmustbereconstructedforProto-Germanicwithdativecaseonthesubjectlikemaleficiary(foraProto-Indo-Europeanreconstruction,seeBarðdaletal.2011).

Figure24.1givesareconstructionofthisconstructionforProto-Germanic,usingSign-Basedformalism(Michaelis2012,thisvolume;Sag2012).Thefigurerepresentsareconstructedlexicalentryforthepredicatebewoeandtheentryisdividedintofourlevels.ThefirstlevelFORMgivesthemorphophonologicalformofthepredicate,thatis,theinfinitiveofbeandwoe.ThesecondlevelARG-STgivestheargumentstructureofthepredicate,andthethirdlevelSEMgivesthesemanticsofthepredicateintermsofframesemantics(Fillmore1982,1985b;Petruck1996).HeretherelevantFRAMEisthemalefactiveframewithonlyoneparticipant,themalefactiverolethatiscoindexedwiththedativeargument.Asthisconstructionwasgenerallyusedwhensomethingbadhappenedtopeople,afourthlevelexpressingthepragmaticuseoftheconstructionmaybeadded(followingFriedandÖstman2005;Fried2009b),labeledPRAG,withthevalueINVOLVEMENT.TheINVOLVEMENTfeatureisrealizedinformallyas‘speaker'sdismay.’

Thevariationinwordorderbetweenthedativeandwoe,seenin(4)vs.(5–6)above,wouldbecapturedwithapreposingconstructionofsomesort,thatis,topicalizationconstructionorfocusconstruction,dependingonouranalysisofthesemanticnatureoftheinvertedwordorderanddependingonourdefinitionsofthevariouspreposingconstructions.ItisalsoimplicitinmyanalysisthatProto-Germanicwasaconfigurationallanguagewithadefaultwordorderandsomenondefaultwordorders,contraHewsonandBubenik(2006)andLuraghi(2010),butfollowingRögnvaldsson(1995)andSmithermanandBarðdal(2009).

Clicktoviewlarger

Figure24.1. Areconstructionofthe[dat-is-woe]constructioninProto-Germanic

Page 330: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 13 of 16

Clicktoviewlarger

Figure24.2. ReconstructionofconstructionsandConstructicons(BarðdalandEythórsson2012a:19)

Asalreadystatedabove,Figure24.1givestheGermanicproto-construction[dat-is-woe],reconstructedonthebasisoftheearliestattestationsinGermanic.Whenotherproto-constructionsofProto-Germanichavealsobeenreconstructed,onthebasisoftheearliestattestationsineachlanguage,theywilltogethermakeupaProto-Constructicon,thatis,astructuredinventoryofconstructions,forProto-Germanic.WhensuchProto-ConstructiconshavebeenreconstructedforallthebranchesofIndo-European,theproto-constructionsintheProto-Constructiconsmaybeusedasinputforthecorrespondencesetsneededtoreconstructproto-proto-constructionsforProto-Indo-European.Atthatpoint,wewillhaveaProto-ConstructiconforProto-Indo-European(cf.BarðdalandEythórsson2012a).AsConstructionGrammarmodelsgrammarsoflanguagesintermsofConstructicons,whicharestructuredinventoriesofconstructions,reconstructionofthetypethatIhaveoutlinedherecountsasreconstructionofgrammar(seeFigure24.2).

Moreover,inthisparticularcase,itturnsoutthatthe[dat-is-woe]constructionisnotconfinedtotheGermaniclanguagesbutisfoundinotherIndo-Europeanlanguagesaswell(seeBarðdaletal.2011).TheexamplesbelowarefromLatinandAvestan:

(7)ImmovaetibisitLatinthen,woeyou.datbe.subj‘Then,woeshallyouhave’(Casina634)(8)aebiioanhausauuoianhatapamamAvestanthem.datlife.genwoebe.subjlast‘theywillhavewoe,theendoftheirexistence’(Y.45.3)

Moreresearchisofcourseneededonthetopic(cf.Barðdaletal.2011),butthefactthatthe[dat-is-woe]constructionexistsinLatinandAvestan,inadditiontoEarlyGermanic,allowsus,forthesakeofmethodologicalillustration,totakeashort-cuthereandjumpdirectlytothereconstructionofthisconstructioninProto-Indo-European.Table24.10givesthecorrespondencesetforthe[dat-is-woe]constructionfromthethreeancientIndo-Europeanlanguagebranchesdiscussed.

Ifallthevaluesofthecorrespondencesetarethesameacrossthelanguagebranchesbeingcompared,likeinthiscasewith[dat-is-woe]wherethereis

Table24.10.Indo-Europeancorrespondencesetforthecollocation[dat-is-woe]

Alt.1

OldGermanic [dat-is-woe]

Latin [dat-is-woe]

Avestan [dat-is-woe]

systematicallyadative

subjectlikeargumentpresentintheargumentstructure,thenanIndo-Europeanproto-constructionmaybe

Page 331: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 14 of 16

reconstructedonthebasisofthat,heregiveninFigure24.3.

ObservethatIhaveherereconstructedasyntacticconstructionconsistingofthreeelements,adativesubjectlikeargument,theverbbe,heregivenwithitsreconstructedroot,andtheadverb/interjectionwoeforProto-Indo-European.Thereis,moreover,nodiscrepancyfoundbetweenthedaughterlanguages,astheformoftheconstructionisthesameacrossallthreebranchesofIndo-Europeanwheretheconstructionisfoundandthesemanticsinvolvesamalefactive,thatis,anegativelyaffectedparticipant.Again,theconstructionisusedwhensomethingbadhappenstopeopleandexpressesthespeaker'sdismay.Thisreconstructionhasthereforenotinvolvedanychoicesbetweenvariantsinthecorrespondenceset,asthecorrespondencesethasshownaremarkableinternalconsistencyacrossitsmembers.ThisisinaccordancewithPaul's(1886)observationsthatsyntacticchangeismuchslowerinratethanphonologicalchange.Althoughthereconstructionofstablesyntacticandmorphologicalfeaturesdoesnotnecessarilyinvolveanylanguagechange,suchreconstruction,Iargue,isstillamajorcontributiontohistorical-comparativelinguistics,asitmayrevealwhatkindofstructurestheproto-languageconsistedof.Assuchitmaygenerateinsightsintospecificresearchquestions,asinthiscasetheevolutionaryhistoryofnoncanonicalsubjectmarkingintheIndo-Europeanlanguages.

Clicktoviewlarger

Figure24.3. Areconstructionofthe[dat-is-woe]constructioninProto-Indo-European

While,itisofcoursedifficulttogeneralizeaboutaprehistoriclanguagestageonthebasisofasinglepredicate-specificconstruction,likeinthiscasethe[dat-is-woe]construction,thepresentreconstructionmaybetakentosuggestthatArgumentStructureconstructionswherethesubjectlikeargumentisnoncanonicallycase-markedmaybereconstructedforProto-Indo-European.Itwouldatleastbeoddiftheproto-languageconsistedofonlyonesuchconstruction.Moreresearchisneeded,andisinfactbeingcarriedout,astowhethermorenoncanonicallycase-markedpredicatesarereconstructableforProto-Indo-European,andeventhoughthisresearchisinitsinfancy,theoutlookappearspromising(cf.BarðdalandSmitherman,2013).WhetherthisdativesubjectlikeargumentbehavessyntacticallyinthesamewayasordinarynominativesubjectsintheearlyIndo-Europeanlanguagesremainstobeinvestigated.ResearchonOldGermanic,however,suggeststhatthenon-nominativesubjectlikeargumentbehavedsyntacticallyasasubjectinOldandEarlyGermanic(BarðdalandEythórsson2003;EythórssonandBarðdal2005;BarðdalandEythórsson2012b),andresearchonOldFrenchhasyieldedsimilarresults(Mathieu2006).Noticethatanintellectualexerciseinvolvingthesyntacticbehaviorofthenon-nominativesubjectlikeargumentpresupposesreconstructionofgrammaticalrelations(cf.BarðdalandEythórsson2012b),anenterprisethatoughttobeevenmorefraughtwithperilthan‘ordinary’syntacticreconstruction,ifithadbeensuggestedintheearlierliteratureatall.

24.5SummaryandConclusions

Morphologicalandphonologicalreconstruction,basedontheComparativeMethod,presupposesarbitraryform-functionpairings,thatis,cognates,whenestablishinggeneticrelatednessbetweenlanguages.Syntacticreconstruction,incontrast,hasbeenfrowneduponbythelinguisticcommunity,becauseof(a)lackofcognates,(b)lackofarbitraryform-functionpairings,and(c)lackofregularityinsyntacticchange.Onafurtherinvestigation,itturnsoutthattheseobjectionsagainstsyntacticreconstructionareentirelylackinginforce.

EmployingthetheoreticalframeworkofConstructionGrammar,wherebotharbitraryandlessarbitraryform–meaningpairingsareregardedasthebasicunitsoflanguage,overcomesatleasttwooftheaforementionedobstacles.First,onaconstructionalapproach,complexandschematicconstructionsalsocountasform–meaningpairings,whichinturnmeansthatwhatisusuallyreferredtoas‘syntacticstructures’intheliteratureisheretreatedonparwithlexicalitems.ThismakescomplexandpartiallyschematicconstructionslegitimateobjectsoftheComparativeMethod,ascognateArgumentStructureconstructions,atleastincaselanguages,caneasilybeidentifiedasinheritedfromaparentlanguage,throughcognatemorphologicalcasemarkingandacomparisonof

Page 332: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

Page 15 of 16

thelexicalverbsinstantiatingtheconstruction.

Second,arbitrarinessisalsofoundinsyntax,asschematicandcomplexconstructionsarealsoform–meaningpairings.WithinConstructionGrammar,itisassumedthatthereisaclinefromtotallyarbitraryform–meaningpairingstosemanticallyfullyregularform–meaningpairings.However,arbitrarinessisonlyarequirementwhenestablishinggeneticrelatedness,andassyntacticreconstructionpresupposesthatgeneticrelatednesshasalreadybeenestablished,thearbitrarinessrequirementbecomesanonissueundersyntacticreconstruction.

Third,theperceivedregularityofsoundchangeisanillusionforseveralreasons;(a)soundchangesdonottakeplacewithallpossibletargetitems,(b)thereexistquasi-regularandirregularsoundchanges,whicharebydefinitionexcludedfromthenotionofsoundlaw,(c)onlycombinatorysoundchangesareregular,notthesystemicones,andfinally,(d)theimportanceofthesoundlawsisduetothefactthattheyprovideasimilaritymetricfordecidinguponwhetherlexicalitemsindifferentlanguagesarecognatesornot.Moreover,developmentalpathsarealsodetectableinsyntax,althoughtheyaremorelimitedinscopethanthesoundlawsbecausesyntacticchangetakesplaceatamuchslowerratethanphonologicalchange.Thefactthatsyntacticstructuresareconsiderablymorestablethansoundsindeedcompensatesfortheperceivedlackofregularityinsyntacticchange,whichinturnmaybeseenasafactorcontributingtomakingsyntacticreconstructioneasierthanphonologicalandlexicalreconstruction.

Finally,Ihaveshownherehowconstruction-basedreconstructionmaybecarriedout,illustratingthiswiththe[dat-is-woe]constructionintheearlyGermanicandtheearlyIndo-Europeanlanguages.Examplesofthisconstruction,withadativesubjectlikeargument,arereadilyfoundinOldEnglish,OldNorse-Icelandic,andMiddleHighGerman,andthesedatamakeuptheinputforacorrespondenceset,onwhichbasisaProto-Germanic[dat-is-woe]constructionmaybereconstructed,usingtheformalismofSign-BasedConstructionGrammar.Clearly,ifthe[dat-is-woe]constructionisalsofoundintheotherearliestattestedIndo-Europeanlanguages,thenaproto-[dat-is-woe]constructionmaybereconstructedformorebranches.Then,onthebasisoftheproto-constructionsforeachlanguagebranch,aproto-proto-constructionmaybereconstructedforProto-Indo-European.Inotherwords,whentheArgumentStructureconstructionsofeachlanguagebranchhavebeenidentifiedandmodeledintheirrespectiveConstructicons,theyconstitutetheinputdataforcorrespondencesetsforreconstructingproto-proto-constructionsforProto-Indo-European,alsomodeledinaProto-Constructicon.ThisishowtheComparativeMethodmaybeusedinconjunctionwithConstructionGrammartoreconstructlargerandmorecomplexunitsthanlexicalitems.SinceConstructiconsarestructuredinventoriesofconstructions,reconstructingconstructionsandConstructiconsentailsreconstructionofthegrammar.

AfurthercomparisonofyetmoreearlyIndo-Europeanlanguageshasrevealedinstancesofthe[dat-is-woe]constructioninLatinandAvestan,inadditiontotheinstancesintheearlyGermaniclanguages.Onthebasisofthesefindings,aProto-Indo-European[dat-is-woe]constructionhasbeenreconstructed.TheimplicationofsuchareconstructionisthatProto-Indo-EuropeanhadArgumentStructureconstructionswherethesubjectargumentwasnotinthenominativecase,butwasnon-canonicallymarked.FurtherresearchisneededtoverifythatamoreabstractobliquesubjectconstructionexistedinProto-Indo-European,althoughthedatapresentedinthischapterprovideclearinnuendosthatobliquesubjectsintheModernIndo-Europeanlanguageswheresuchstructuresarefound,likeinModernIcelandic,arenotaninnovationbutanIndo-Europeaninheritance.

IamgratefultoEysteinDahl,ThórhallurEythórsson,ThomasSmitherman,theeditors,andananonymousreviewerforcommentsanddiscussionsthathavesubstantiallyimprovedthequalityofthischapter.IthankMisumiSadlerforherhelpwiththeJapanesedatainsection24.3,andThórhallurEythórsson,EysteinDahl,ChiaraFedriani,IljaSeržant,andThomasSmithermanforhelpwiththedataon‘woe’insection24.4.

Notes:

(1.)ThisinterjectionisalsofoundoutsideofIndo-European,forinstance,inSemitic,whichmaysuggestthatitisevenmoreancientthanPIE.

JóhannaBarðdalJóhannaBarðdalisaResearchAssociateProfessorattheUniversityofBergen.Shehasworkedoncasemarking,obliquesubjects,grammaticalrelations,constructionalsemantics,andsyntacticproductivityinasynchronicanddiachronicperspective.

Page 333: The Oxford Handbook of Construction Grammar

Construction- Based Historical-Comparative Reconstruction

HerlastbookProductivity:EvidencefromCaseandArgumentStructureinIcelandicwaspublishedbyBenjaminsin2008.ShehaspublishedarticlesinNordicJournalofLinguistics,JournalofLinguistics,Language,Morphology,Linguistics,Lingua,andDiachronica.SheisafoundingcoeditoroftheJournalofHistoricalLinguistics.SheiscurrentlyrunningalargeresearchprojectonnoncanonicalsubjectmarkingintheearlyandarchaicIndo-Europeanlanguages,fundedbytheUniversityofBergen,BergenResearchFoundation,andtheNorwegianResearchCouncil.

Page 334: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 1 of 13

PrintPublicationDate: Feb2013 Subject: Linguistics,HistoricalLinguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0025

Corpus-basedApproachestoConstructionalChangeMartinHilpertTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapteranalyzeshowcorpuslinguisticshassignificantlyaffectedourunderstandingofconstructionalchange.Itdescribesthreedifferentkindsofconstructionalchange,includingchangesinaconstruction'sfrequency,initsform,andinitsfunction.Theanalysisrevealsthat,inallcases,thedataretrievedfromcorporacanrevealfine-graineddetailregardingthegradualnatureofconstructionalchangeandfacilitatethequantitativeanalysisthattypifiesmuchworkonlinguisticvariationandchange.

Keywords:corpuslinguistics,constructionalchange,quantitativeanalysis,linguisticvariation,linguisticchange

25.1.Introduction

TheuseofcorpusdataisinnowayuniquetoworkinConstructionGrammar,letalonetothestudyofconstructionalchange.Modernandhistoricalcorporaareusedacrosstheentirespectrumoflinguistics,withverydiversemethodsandstillmorediversegoals(LüdelingandKytö2008).Thepresentchapterfocusesonagrowingbodyofresearchthatusesthemethodsofhistoricalcorpuslinguisticsinawaythatisconsistentwithaconstructionalapproachtolanguagechange(Fried,thisvolume).NotallofthisworkexplicitlyalignsitselfwithaparticularbrandofConstructionGrammar,butnonethelessanimplicitcommitmenttotheconstructionistagendaisthere:thefocusofinvestigationliesonform-meaningpairingsanddevelopmentswithregardtotheirstructure,function,andfrequency.Showcasingsomeofthiswork,itisdiscussed(a)whydiachroniccorporaandsimilartextualresourcessuchastheOxfordEnglishDictionary(OED)areaparticularlyaptsourceofdataforthestudyofconstructionalchange,(b)howtheseresourcesareusedforthatpurpose,and(c)howtheconstructionalperspectivesetsthisworkapartfromotherapproachestolanguagechange.Thecasestudiesthatarechosenalsoillustratetherichvarietyofprocessesthatcanbegroupedtogetherundertheheadingofconstructionalchange,showingineachcasehowthephenomenoninquestioncanbefruitfullyinvestigatedandhowtheseinvestigationsadvancethetheoryofConstructionGrammar.

Inlinewithmostcurrentdefinitionsoftheterm,constructionsareunderstoodhereassymbolicpairingsofaformandameaningthatdisplaystructuralorsemantic/pragmaticidiosyncrasiesor,evenwithoutsuchidiosyncrasies,ahighlevelofentrenchment(Croft2001;Goldberg2006a;Langacker2005).Thisdefinitionisdeliberatelyverybroad,includingabstractsyntacticpatternssuchaspseudo-cleftsandtheDitransitiveconstruction,concretelexicalitemssuchastheEnglishnoundog,andidiomaticexpressionssuchasblowthewhistleonNP.Whiletheseformsdifferintermsofhowabstractandschematictheyare,eachofthemconnectsaparticularformwithaparticularmeaning,thusformingasymbolicunit.Abroaddefinitionoftheterm‘construction’isneededbecausethesumtotaloflinguisticknowledgeistakentobeahierarchicallyorderednetworkofsymbolicunits,inwhichmoreschematicconstructionssharesomeoralloftheircharacteristicswithmorespecificconstructions(Broccias,thisvolume).Eachlinguisticformthatismentallystoredbythespeaker(becauseofeitheritsidiosyncrasiesoritsfrequency)representsanodeinthisconstructionalnetwork.Diachronically,singlenodesorgroupsofnodesinthe

Page 335: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 2 of 13

networkmaychange.Howtheseprocessesofchangecanbestudiedonthebasisofcorpusdataisthemainconcernofthischapter.

Inlanguageuse,bothformandmeaningofaconstructionaresubjecttovariationandchange.Changesinformpertaintothephonologyormorphosyntaxofaconstruction,whereasmeaningchangesaffectitssemanticanddiscourse-pragmaticcharacteristics.Eachofthesetypesofchangeisamenabletocorpusanalysis.Changesinformandmeaningcanbestudiedthroughfrequencymeasurementsofvariantforms.Onevariantofaconstructionmaybecomemorefrequentovertime,thusalteringtheprototypeofthatconstruction.Evenintheabsenceofstructuralchange,aconstructionmayundergochangesinabsolutefrequency,relativefrequency,ortypefrequency.Thesechangesarenolessindicativeofdevelopmentthanchangesinthesubstanceofaconstruction.Inordertoestablishthatanobservedtrendconstitutesreliableevidence,frequencymeasurementsinquantitativecorpuslinguisticsaresubmittedtoinferentialstatistics.Anumberoftechniquesparticularlygearedtowardtheanalysisofdiachroniccorpusdatahavebeendevelopedrecently(GriesandHilpert2008andreferencestherein).Itisimportanttopointoutthatfrequencymeasurementsarenotmerelyusedtodocumentdifferencesbetweenhistoricalstages.Theartofapproachingconstructionalchangewithacorpus-basedmethodologyistoformulatetheoreticalquestionsinsuchawaythatfrequenciesdrawnfromcorpusdatacanilluminatethem.Morespecifically,hypotheseshavetobeconstructedsothatobservedfrequenciesincorporaprovideevidencefororagainstthem.

Beforegoingintoanumberofindividualcasestudies,itisnecessarytospecifythenotionofconstructionalchangeinsomemoredetail.Iftheterm‘construction’weretodenoteonlygrammaticalconstructions,theterm‘constructionalchange’wouldbelargelysynonymouswithgrammaticalization,thatis,“thechangewherebylexicaltermsandconstructionscomeincertainlinguisticcontextstoservegrammaticalfunctions,and,oncegrammaticalized,continuetodevelopnewgrammaticalfunctions”(HopperandTraugott2003:xv).However,constructionalchangeismoreencompassing.Asoutlinedabove,lexicalitemsarealsoconstructions,suchthatlexicalization(BrintonandTraugott2005)andallmannersoflexicalsemanticchangefallunderthepurviewofconstructionalchangeaswell.Inaddition,manydefinitionsofgrammaticalizationexcludewordorderchanges(HopperandTraugott2003:24),someofwhichinstantiateconstructionalchange.ThediachronicbranchofConstructionGrammaristhusabroaderenterprisethanthestudyofgrammaticalization,eventhoughthereissubstantialoverlapinbothsubjectmatterandtheoreticalorientation(Noël2007a).Giventhebroaddefinitionofconstructions,itisfairtoaskwhetherthenotionofconstructionalchangeisnotinfactcoextensivewithlanguagechange,whichwouldrenderthetermrathermeaningless.Thisisnotthecase.Manyprocessesoflanguagechangearenotspecifictoindividualconstructions,andarehencebestdescribedasaffectingalanguageoralanguagevarietyasawhole,acrossmanydifferentconstructions.OneexamplewouldbeaphonologicalchangesuchastheGreatVowelShift.Onthesyllabiclevel,thedevelopmentofacodaconditionwouldnotbeaconstructionalchange.Alsoagenerallossofinflectionalmorphologyduetolanguagecontactexemplifiesachangethatisnotconstruction-specific,butthatseizestheconstructionalnetworkofalanguageasawhole.WordorderreorganizationfromOVtoVOcommonlydoesnotonlychangetransitiveclauses,butaffectstheorderofheadandmodifierinothersyntacticconstructions.Likewise,phenomenasuchasthedevelopmentofdiglossia,theformationofsupra-regionalstandards,andlanguagedeatharenotconstructionalchanges.Manymoreexamplescouldbeadduced.Forthepurposesofthepresentchapter,constructionalchangewillbedefinedasfollows:

Constructionalchangeselectivelyseizesaconventionalizedform-meaningpairofalanguage,alteringitintermsofitsform,itsfunction,itsfrequency,itsdistributioninthelinguisticcommunity,oranycombinationofthese.

Themainpointofthatdefinitionisthatconstructionalchangestartsbyaffectingasinglenodeintheconstructionalnetwork.Overtime,suchachangemayspreadtowardothernodes,thusaffectinggroupsofconstructions.Thismeansthatthedistinctionbetweenconstructionalchangeandlanguagechangeisoneofdegree—evenhighlygeneralprocessesoflanguagechangemayhaveoriginatedbyaffectingasingleconstructionfirst.Withthisprovisionaldefinitioninplace,wecanillustratethenotionofconstructionalchangewithseveralconcreteexamplesanddiscusscorpus-basedmethodsthathavebeenusedfortheiranalysis.

25.2.StudyingConstructionalChangeThroughCorpora

Thischapterdiscussesthreedifferenttypesofconstructionalchangeonthebasisofillustratingcorpus-based

Page 336: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 3 of 13

studies.Thesetypesrepresentchangesinfrequency,form,andfunction.Thediscussionbeginswithfrequency,notonlybecausefrequencymeasurementsarethemainstayofcorpuslinguisticmethodologybutalsobecausecorpus-basedanalysesofformalandfunctionalchangetypicallyrelyonfrequencies.Thefirstsectionsoutlinefourkindsoffrequencychangethatdonotnecessarilyinvolvestructuralchange,butthatpertaintothequestionhowoftenaconstructionappearsinagivencontext.Thediscussionthenturnstochangesattheformalpoleofaconstruction,rangingfromthemorphophonemictothesyntactic.Thelastsectionshighlightdifferentmeaningchangesthatconstructionsmayundergo.Naturally,formal,functional,andfrequencychangesareanythingbutunrelated.Manyofthestudiesthatarediscussedhereillustratemultipletypesofchange.Still,eachofthesetypesdeservestobehighlightedseparately.Table25.1givesanoverviewofthechangesandcorrespondingstudiesthatareexaminedinthefollowingsections.

25.2.1ChangeinTextFrequency

Ofallchangesdiscussedhere,changeintextfrequencyisprobablythemostwidelyusedmeasureofconstructionalchange.Itiswellknownthatanincreaseinfrequencyisacommoncorollaryofgrammaticalization(Bybee2007),eventhough

Table25.1.Corpus-basedstudiesofconstructionalchange

Type Phenomenon Reference

Frequency (1) changeintextfrequency Mair(2004,2006);Hundt(2004)

(2) changeinrelativefrequency HilpertandGries(2009);Tagliamonte(2004)

(3) changeinproductivity Bauer(2001)

(4) changeingenreandvariety Noël(2008)

Form (5) morphophonemicchange GriesandHilpert(2010)

(6) morphosyntacticchange Nübling(2005))

(7) changeinargumentstructure HilpertandKoops(2008)

(8) hostclassexpansion Patten(2010)

Function (9) metaphorandmetonymy PetréandCuyckens(2008)

(10) analogicalextension Israel(1996)

(11) collocationalchange Hilpert(2008)

itremains

disputedwhetherincreasingfrequencyisacause,aneffect,oramereoptionalconcomitantofgrammaticalization.Supportingevidenceforthelatterviewcomesfromgrammaticalizedformsthatremainhighlyinfrequentthroughouttheirhistory(Hoffmann2005).Inordertointerpretanobservedfrequencychange,itisusefultocheckwhethersuchadevelopmentcoincideswithformalorfunctionalchanges,asdiscussedinlatersections.However,alookatrawfrequenciesisacommonstartingpointforanalysesofconstructionalchange.

Theeasiestdocumentationofadevelopmentisthecomparisonofaconstructionacrosstwocorporathatrepresentdifferentperiodsoftime.TheBROWNfamilyofcorporaaffordspreciselysuchcomparisons,offeringrepresentationsofBritishandAmericanEnglishfromthe1960sand1990s,respectively. Mair(2006:113)usestheseresourcestoshowthattheEnglishget-passiveincreasedinfrequencyinbothvarieties,thusestablishing

1

Page 337: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 4 of 13

itselfasanewgrammaticalconstructioninthedomainofthepassivevoice.Diachroniccorpusanalysescanbemeaningfullycomplementedbystudiesofapparenttime,inwhichthesynchronicusageofspeakersfromdifferentagegroupsiscompared.Mair(2006:115)adducesdatafromthespokenpartoftheBritishNationalCorpus(BNC),whichshowsthatspeakersbetween14and24yearsofageusetheget-passiveatasignificantlyhigherratethantheiroldercontemporaries.Diachronicandsynchroniccorporacanthusbecombinedtotestandrefinehypotheses.

Givenaresourcewithatemporallycontinuouscoverageoflanguageuse,frequencydevelopmentscanbeanalyzedinamorefine-grainedfashion.AveryfinelevelofdetailisofferedbythequotationsintheOxfordEnglishDictionary,whichindicatetheexactyearinwhichaparticularutterancewasproduced.HistoricalstudiesbasedontheOEDtypicallydividethedataintobinsofhalf-centuriesorsimilartimeslices.Usingtwenty-five-yearincrements,Mair(2004:129)showsthattheEnglishbegoingtoconstructionunderwentafrequencydevelopmentthatdescribesa“delayedincrease.”Despitebeingagrammaticalizedconstructionbytheendofthesixteenthcentury,begoingtoonlygentlyincreasedinfrequencyuntilthedawnofthetwentiethcentury,afterwhichitshowedrapidspread.Thisobservationraisesthetheoreticalquestionwhetherthedelayedincreasepatterngeneralizestothefrequencydevelopmentsofothergrammaticalizingconstructions.

Oftenitisinstructivetocomparethefrequencydevelopmentsoftworelatedconstructions.InastudyoftheEnglishvoicesystem,Hundt(2004)discussestheemergenceoftheprogressivepassive(Thehouseisbeingbuilt)intheearlynineteenthcenturyanditsrelationtothesemanticallysimilarpassivalconstruction(Thehouseisbuilding).ThelattersurvivesinmodernexpressionssuchasFlight780isnowboarding,butisrareotherwise.Doesthisallowtheconclusionthattheprogressivepassiveoustedthepassivalconstruction?DatafromtheARCHERcorpus showsthatbythetimetheprogressivepassiveappeared,thedecayofthepassivalwasalreadyunderway,whichcastsdoubtonacausalconnection(Hundt2004:112).ThisinterpretationisinlinewithHopperandTraugott(2003:126),whoargueagainsttheideathatgrammaticalconstructionsdeveloptofillfunctionalgaps.Quitetothecontrary,similarconstructionstendtoclustertogetherinvariousgrammaticaldomainssuchasdeonticmodalityorfuturetimereference.Hundt'scorpusanalysiscorroboratesthisidea.

25.2.2ChangeinRelativeFrequency

Inmanycases,constructionaldevelopmentsmanifestthemselvesmoreclearlyinchangingrelativefrequencies,ratherthaninchangingoveralltextfrequencies.Aslightincreaseinthefrequencyofoneconstructionmaynotappearsignificantifviewedinisolation,butifsimultaneouslyasimilarconstructiondecreasesinfrequency,thetwodevelopmentsmayadduptoaphenomenonworthyofinvestigation.Thissectiondiscussestwotypesofscenariosinwhichattentiontochangingrelativefrequenciesisparticularlyuseful.BothscenariosrelatetoprinciplesthatHopper(1991)hasidentifiedascharacteristicforincipientgrammaticalization.Thefirstscenarioistiedtotheprincipleofdivergence;thesecondtotheprincipleoflayering.

Divergencedescribesthephenomenonthatagrammaticalizingformanditscoexistinglexicalsourceundergoseparatedevelopments.Awell-wornexampleofthisisthatonlythegrammaticalizingconstructionbegoingtounderwentattritiontogonna,whereasthelexicalverbgoretaineditsfullphonologicalform.Similarly,grammaticalizingformsundergofrequencydevelopmentsthatsetthemapartfromtheirlexicalsources,whichtypicallydonotundergoanydramaticfrequencychanges.Hence,contrastingthefrequenciesofaputativecandidateforgrammaticalizationanditslexicalsourcecansubstantiatethatanapparentdevelopmentisactuallyunderway.

AnillustratingexamplefromEnglishistheverbkeep,whichhasanumberoflexicalsensesrelatingtotheideaofretainingastate(keepquiet),object(keepthemoney),orlocation(keepoff).Inconjunctionwithaning-formcomplement,keephascometoexpressthegrammaticalcategoryofcontinuativeaspect,asinHekeepstellingmeabouthisproblems.HilpertandGries(2009)usetheTIMEcorpus toinvestigatewhethertheratioofexamplesinwhichkeeptakesaning-formcomplementhasincreasedovertime,showingthatthisisindeedthecase.WhereastheformkeepV-ingonlyaccountsfor4%ofallusesofkeepinthe1920s,itsrelativefrequencycomparedtootherusesofkeeprisescontinually,rangingbetween12%and15%inthe1990sand2000s.Thisrelativefrequencychangesuggeststhateithertheconstructionisbecomingmorepopularinjournalisticwriting,orthattheconstructionisbecomingmorewidelyapplicable,andhencemoregrammatical.Ineithercase,therelativefrequenciesrevealachangeinprogress.

2

3

4

Page 338: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 5 of 13

Thesecondphenomenontobediscussedhereisrelativefrequencychangeinascenariooflayering.Layeringdescribesthecoexistenceofdifferentconstructionsfortheexpressionofsimilargrammaticalmeanings.Forinstance,thegrammarofEnglishcontrastspresentandpasttimereferenceeitherphonologicallythroughablaut(Iwritevs.Iwrote),morphologicallythroughadentalsuffix(Iworkvs.Iworked),orperiphrasticallythroughthePerfectconstruction(Iworkvs.Ihaveworked).Themetaphorof‘layering’indicatesthatthesealternativesrepresentdifferenthistoricalstages;theperiphrasticstrategyisnewerthanthetwoothers(Hopper1991:24).Changeintherelativefrequenciesofsuchalternativesmayrevealthatoneconstructionisbecomingthedefaultchoice,whileanotherisfallingoutofusage.Alternatively,itmaybecomeapparentthateachconstructionhasfounditsgrammaticalnicheandthattherelativefrequenciesstayconstant.

Tagliamonte(2004)studiesthegrammaticaldomainofdeonticmodalityinnorthernBritishEnglish,whichcontainsseveralconstructionsfortheexpressionofobligation,suchasmust,haveto,havegotto,andcontractedvariants.Ratherthananalyzingdiachroniccorpusdata,Tagliamontereliesonsynchronicdataandworkswithanapparenttimedesign.UsingtheYORKcorpus, shecomparestherelativefrequenciesoffourlayeredconstructionsacrossthreedifferentagegroups.Theanalysisrevealsthatmustisusedlessbyyoungerspeakers,whereashavetoemergesasthedefaultchoicefortheyoungestgroup.Baregottoorgottaisusedmorebyyoungerspeakers,butremainsaveryinfrequentchoiceacrossallagegroups.Thedifferingrelativefrequenciesoftheseconstructionspointtoongoingtransitionsinthedomainofdeonticmodality.Tagliamontetakesthisobservationasastartingpointforamultivariateanalysisthatpredictsthechoiceofagivenconstruction,takingintoaccountnotonlyage,butalsosex,education,anddifferentlinguisticvariables.Modelingconstructionalchangethroughsimultaneousmeasurementsofmultipleinteractingfactorsoffrequency,form,andfunctionisafruitfulavenueofresearchthatisfurtherdiscussedinsection25.2.5.

25.2.3ChangeinProductivity

Frequencydevelopmentsmaynotonlypertaintotokenfrequency(i.e.,thequestionofhowoftenaconstructionoccursinagivencorpus).Developmentsmayalsotakeplaceattheleveloftypefrequency.Toillustrate,aconstructionmaywidenitsrangeofco-occurringelementsbyhost-classexpansion(i.e.,theintegrationofdifferentsyntacticcategories)(section25.2.8).Aconstructionmayalsochangebyacceptingagreaternumberofdifferentelementsfromthesamesyntacticcategory,asisdiscussedinthesectionsonanalogicalextension(25.2.10)andcollocationalchange(25.2.11).Allofthesephenomenareflectthetypefrequency,andhencetheproductivityofaconstruction.Toillustrate,theEnglishsuffix-ableallowstheformationofadjectivesthroughattachmenttoaverbalstem,yieldingformssuchasacceptable,enjoyable,orpredictable.Fromadiachronicperspective,theV-ableconstructionhasundergoneseveralchangesregardingitsacceptanceofverbalstems:enteringtheEnglishlanguageasaFrenchimportduringtheMiddleEnglishperiod,itfirstoccurredwithRomanceverbalbases.Astheconstructiongeneralized,itcametobeusedwithGermanicstemsinformssuchasbreakableorwearable.TheconstructionremainsproductiveinPresent-DayEnglish,asisevidencedbyclickable,faxable,orcut-and-pasteable.AdiachroniccorpusanalysiscandeterminewhenspeakersofEnglishbeganformingadjectiveswithGermanicverbalstems,whethertherewereperiodsofenhancedproductivity,andwhetherproductivityremainsstronginPresent-DayEnglishorhasinfactstalled.

Inastudyofthiskind,Bauer(2001:8)usestheOEDtoanalyzediachronicvariationintheproductivityofthesuffix-ment,whichattachestoverbalstemsformingnounssuchasengagementorsettlement.InPresent-DayEnglish,theconstructionisnolongerproductive.BauerdividestheOEDintofifty-yearperiodsanddetermineshowmanynewformationswith-mentarefoundineachperiod.Heshowsthatwhiletheoverallrateofproductivitydiachronicallyvaries,newformskeepenteringtheEnglishlanguageuntilthenineteenthcentury,afterwhichtheconstructionbecomesmarkedlylessproductive.FromtheperspectiveofConstructionGrammar,thelossofproductivityisaninterestingphenomenonbecauseitsuggeststhatageneralconstructionalschema(V-ment)hasbeenlost,whileofcourseitslexicalizedforms(engagement,settlement,etc.)remaininuse.

25.2.4ChangeinGenreandVariety

Constructionsmayalsochangeinthewaytheyareassociatedwithacertainlinguisticgenreorvariety.Forinstance,aconstructionthatoriginatesininformalspeechmayovertimefinditswayintoelevatedwrittenregisters—manysuchexamplesofcolloquializationarediscussedinMair(2006).Conversely,constructionsmayalso

5

Page 339: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 6 of 13

undergoaprocessinwhichtheirrangeofgenresnarrowsovertime.

OneexampleofthiskindisdiscussedbyNoël(2008),whostudiestheEnglishNominativeandInfinitiveconstruction(NCI),whichinvolvesasubject,averbofperception,cognition,orutteranceinthepassive,andato-infinitive.Theconstructionfunctionsasamarkerofevidentiality,asshownin(4a)and(4b).

(4)a.Thedealisthoughttobeworth£200m.b.HalfofCanada'sacidrainissaidtocomefromUSpowerstations.

NoëlusestheCorpusofLateModernEnglishTexts(CLMET) andtheBNCtotrackthefrequencyoftheconstructionfrom1640tothe1990s,findinganonlinearpattern.ThroughoutthefourperiodsoftheCLMET(1640–1920),theconstructioncontinuallydecreasesinfrequency,onlytoberepresentedmuchmorestronglyintheBNC.NoëlsuggeststhattheU-shapeddevelopmentfindsanexplanationingenredifferencesbetweentheCLMETandtheBNC.Whereasthelatterisabalancedcorpus,theCLMETislimitedtoworksofliterature.ComplementingevidencefromtheARCHERcorpusshowsthathistorically,theNCIbecameincreasinglymorefrequentinjournalisticandscientificwriting(i.e.,genresinwhichthelinkageofevidencetoasourceisofgreatimportance),whereasothergenresreplicatethedecreasingtrendthatcanbeobservedintheCLMET.CloserexaminationoftheBNCdataconfirmsthatnewspapersandscientifictextsinPresent-DayEnglishshowahigherrateofNCIexamplesthanothergenres.Hence,thedevelopmentoftheNCIcanbeseenasaprocessofspecializationintoaparticulargenre.Noël'sanalysisfurtherrevealsthattheNCIhasdevelopeddifferentpreferencesforparticularverbalpatternsinjournalisticandscientificwriting,respectively.Frequentpatternsinnewspaperwritingarebesaidto,bereportedto,beunderstoodto,andbeallegedto,frequentpatternsinsciencetextsare,amongothers,beshownto,befoundto,beknownto,andbeestimatedto.ThestudythusdemonstrateshowthespecializationprocessoftheNCIproceedsevenbeyondparticulargenresatthelevelofindividualcollocations.Theroleofcollocationsinconstructionalchangeisthemainsubjectofsection25.2.11below.

25.2.5MorphophonemicChange

Regularsoundchangerepresentsaprimeexampleoflanguagechangethatdoesnotinstantiateconstructionalchange.Ifthelexiconasawholeisaffected,thefactthatasinglelexicalconstructionusedtohaveaninitialstopbutnowhasaninitialfricativefallsoutofageneralchangeanddoesnotneedtobestatedseparately.Still,therearechangesinvolvingalternationsofsoundsthatareconfinedtospecificconstructionsandthathencemeritinclusioninthisoverview.

GriesandHilpert(2010)studychangeintheEnglishpresenttense.Between1450and1680,theinterdentalthirdpersonsingularsuffix,asinhedoth,isreplacedbyasuffixwithanalveolarfricative,asinhedoes.Thisdevelopmentisnotparalleledinlexicalitems;thefricativesinbreathoryouthdonotchange.Acorpus-basedanalysiscantakemeasurementsofrelativefrequencies(section25.2.2)todeterminewhenthisconstructionalchangehappenedandforhowlongtherewasvariationbetweenthetwoforms.Beyondthat,corporaaffordananalysisofthelinguisticandnonlinguisticfactorsthatcorrelatewiththechoicesthatspeakersmake.Technicallyspeaking,theoutcomeofadependentvariable(-thvs.-s)canbepredictedonthebasisofseveralindependentvariables,whichconcernfrequency,form,function,andextra-linguisticfactors.Itcanthusbeinvestigatedwhichstructuralcontextsfavortheuseofthenewform,whichspeakersweredrivingthechange,andhowlinguisticandextra-linguisticfactorsinteract.

UsingtheCorpusofEarlyEnglishCorrespondence(CEEC), acorpusoflettersthatisannotatedforvarioussocialvariables,GriesandHilpertextractallrelevantverbforms.Eachexampleisannotatedforthedependentvariable(-thvs.-s)andforindependentvariablesincluding(1)whichverbisused,(2)whetherthatverbislexical(sing)orgrammatical(do,have),(3)whethertheverbstemendsinasibilant(wish),(4)whetherthefollowingwordbeginswithansorth,and(5)whichvariantofthethirdpersonsingularsuffixwasusedlast.Nonlinguisticfactorsinclude(6)thesuccessivetimeperiods,(7)theidentityoftheauthor,and(8)thegenderofauthorandintendedrecipient.Tomodelthedevelopmentfromdothtodoes,GriesandHilpertfitalogisticregression.Theanalysisrevealsfirstofallwhichoftheabovefactorsdonotinfluencethechoice.Forinstance,thegenderoftheintendedrecipientdoesnotplayarole.Bycontrast,thegenderoftheauthor,thegrammaticalityoftheverb,andthepreviouslyusedformeachshowasignificantinfluence.Crucially,someinfluencesdonotstayconstantovertime.Afinalsibilantmatters

6

7

Page 340: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 7 of 13

tothechoiceonlyduringoneearlyperiod,afterwhichtheeffectdisappears.

Multivariateanalysesofphonologicalchangesalreadyhavealonghistoryinquantitativesociolinguistics,andrecentyearshaveseenthedevelopmentofcorpus-basedhistoricalsociolinguistics(NevalainenandRaumolin-Brunberg2003).Corpus-basedapproachestoconstructionalchangestandtobenefitalotfromadoptingandrefiningthesetechniques.EspeciallyasConstructionGrammarendorsestheideathatconstructionsmaychangeatanylevelofstructure,meaning,oruse,thepossibilitytomonitortheselevelssimultaneouslyovertimeholdsagreatpotentialtodeepenourunderstandingofconstructionalchange.

25.2.6MorphosyntacticChange

Perhapsthemostsalientcharacteristicofconstructionalchangeisthecontinuousmorphosyntacticreductionofgrammaticalizingforms.StockexamplesofthisareconstructionssuchasEnglishgonnafrombegoingto,andtheFrenchmorphologicalfuturejeferai‘Iwilldo’fromtheperiphrasticLatinformfacerehabeo‘Ihavetodo’(HopperandTraugott2003).

AnexamplefromGermanstudiedbyNübling(2005)concernsthefusionofprepositionswithdefinitearticles,fromaufdas‘ontothe’toaufs,frombeidem‘atthe’tobeim.Thisprocess,whichisattestedsinceOldHighGermantimes,isnotfullyproductive,assomeprepositionsandarticlesdonotfuse.Formssuchas*gegenüberm‘acrossthe’arenotattested,necessitatingapiecemealaccountofindividualconstructionaldevelopments.Fusionisobligatoryinsomeusages(imHimmel‘inheaven,’*indemHimmel),highlyfrequentwithotherforms,andrarewithstillothers.Orthographicalvariation(für's/fürs‘forthe’)suggeststhatconstructionalchangeisinprogress,butinterestingly,modernvarietiesofGermanhavelostfusedformsthatwereattestedinearlierdialects;especiallyfusedformswithgenitivearticles.Thephenomenonunderinvestigationisthusaclusterofconstructionalchangesthatrepresentdevelopmentalstagesofsimilar,butnonidenticalprocesses.

NüblingusestheIDScorporaofmodernwrittenGermanasaresourcetostudythisprocessfromasynchronicperspective.Incurrentusage,approximatelytwenty-fivepreposition-articlepairsalternatebetweenthefusedformandtheanalyticalvariant.Theformsam‘atthe,’zum‘tothe,’andim‘inthe’vastlyoutnumbertheirnonfusedcounterparts;hinters‘behindthe,’ans‘tothe,’andums‘aroundthe’displayrelativelybalancedfrequencies;theformsunterm‘underthe,’vorm‘infrontofthe,’andüberm‘overthe’arerelativelyrare.Howcanthesefiguresbeexplained?Nüblingsuggestsseveralmotivatingfactors.First,fusionismorefrequentwithprepositionsthatgovernthedative,ratherthantheaccusativeorgenitive.Alsotheco-occurrencefrequencyofprepositionandarticleexplainssomeoftheobservedvariance.Developmentallyoldandshortprepositions(an,in,zu)aremorelikelytofusethanmorerecentlygrammaticalizedprepositions(anstatt,wegen,dank).Beyondthatthereismuchdialectalandgenre-specificvariation.Indeterminingtherelevantparametersofanongoingchange,Nübling'sanalysisofpresent-daycorpusdatalaysthegroundworkforthekindofdiachronicmultivariateanalysisthatwasdescribedintheprecedingsection.Heranalysisunderscoresthatthestudyofconstructionalchangerequiressimultaneousattentiontoform,frequency,andamultitudeofextra-linguisticvariables.Theworkalsoraisesquestionsthatshowthelimitsofcurrentcorpus-linguisticresources:atpresenttherearenodiachroniccorporaofGermanthatwouldallowtheresearchertocontrolthevariablesofgenre,variety,andseveralstructuralfactorswhiletracingthedevelopmentofaconstructionovertime.Corpus-basedapproachestoconstructionalchangethushavetofindwaystoaskquestionsthatcurrentlyavailableresourcescanilluminate.

25.2.7ChangeinArgumentStructure

Thedevelopmentofaconstructionmaybringalongachangeinthevalencyofanelementthatformspartofit;commonlythisconcernstheargumentstructureofverbs.Newargumentsmaybecomeobligatory,requiredargumentsmaybesubjecttochange,orevenloss.Diachroniccorpusdatacanshowwhenandhownewargumentstructurepatternsdeveloped.

HilpertandKoops(2008)usedatafromtheSwedishSPRÅKBANKENresource todocumentachangeintheargumentstructureofagrammaticalizingcomplexverbconstruction.Specifically,theydiscussthedevelopmentoftheSwedishPseudo-coordinationconstruction(SPC)withthepostureverbsitta‘sit.’Inthisconstruction,theverbsittaiscoordinatedwithasecondverbandtakesonanaspectualmeaningofdurativity.(Foraroughcomparison,considertheEnglishsentenceWesatandtalked.)Diachronically,Swedishsittaunderwentachangeinits

8

Page 341: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 8 of 13

argumentstructureasitslexicalmeaninggavewaytogrammaticalmeaning:themoresittacametobeusedasanaspectualmarker,thelessrelevantitslocativemeaningbecame.Thisisreflectedinthedegreetowhichtheverbisusedwithprepositionalphrasesoradverbialsthatspecifywherethesittingtakesplace.Whilelexicalsittaisoftenelaboratedspatially(7a),suchelaborationistypicallyabsentintheSPC(7b).Nevertheless,lexicalsittacanoccurwithoutspatialelaboration(7c),andcoordinatedsittamayoccurwithit(7d).

(7)a.Jagharsuttitvidskrivbordetnästanheladagen.Ihavesatatdesk.thealmostallday‘I’vesatatthedeskalmostallday.’b.Jagharsuttitochlästheladagen.Ihavesatandreadallday‘Ihavebeenreadingallday.’c.Hansattenstundigen.hesatawhileagain‘Hesat(somewhere)forawhileagain.’d.Jagharsuttitvidskrivbordetochlästheladagen.Ihavesatatdesk.theandreadallday‘I’vesatatthedeskandreadallday.’

Inordertodemonstratethatsittaunderwentachangeinitsargumentstructure,HilpertandKoops(2008:255)measuretherelativefrequencyofspatialelaborationwithlexicalsittaandtheSPCoverdifferentperiodsoftime,findingthattherateofspatialelaborationstaysconstantwithlexicalsitta,butdecreasessignificantlywithintheSPC.Theobservedchangeisthusaconstruction-specificchangeinargumentstructure.

25.2.8Host-classExpansion

Himmelmann(2004)suggeststhenotionofhost-classexpansionasausefulparameterofconstructionalchangeingrammaticalization.Asyntacticheadthatinitiallyonlyoccurswithaspecifictypeofcomplementmayincreaseitsrangeofpossiblecomplementstoincludemoreanddifferentsyntacticcategories.Diachroniccorpusanalysiscanbringtolightwhenandhowthisextensiontookplace,andhowthedifferentcomplementtypeschangedinfrequency.

AnillustratingexampleisofferedbyPatten(2010),whostudiesthedevelopmentofit-cleftsinthePennParsedCorporaofHistoricalEnglish(PPCME2,PPCEME). TheEnglishit-cleftconsistsofafocalelement,mostfrequentlyanounphrase,whichisfollowedbyarelativeclauseexpressingpresupposedinformation.Anexampleofthisstructuraltypeisgivenin(8a).InPresent-DayEnglish,it-cleftsallowseveraldifferentcategoriesascomplementsofthecopula,forinstance,prepositionalandadverbialphrases(8b,8c)orthat-clauses(8d).

(8)a.Itwasthetherapist thatkilledher.b.ItisinDecember thatshe'scoming.c.Itwashere thatwemet.d.It'sthathe'ssoannoying thatbothersme.

PattendemonstratesthroughcorpusanalysisthattheIt-cleftconstructionhasundergonehost-classexpansionovertime.Prepositionalphrasesappearasfocalelementsfromthefourteenthcenturyonward,adverbialandclausalelementsareaninnovationoftheEarlyModernEnglishperiod.Theconstructionassuchgainsinoverallfrequency;nounphrasesremainthemostcommonfocalelementsthroughouttheentirecorpus.Thegradualprocessofhost-classexpansionthatPattenobservesallowshertoargueagainstanalternativeaccount(Ball1994),whichderivesit-cleftswithnon-nominalfocalelementsfromanimpersonalconstruction.

25.2.9MetaphorandMetonymy

Metaphoricalandmetonymicextensionhasbeenstudiedextensivelyinthecontextofpolysemy(CroftandCruse2004).Inparticularthepolysemyofprepositionshasreceivedmuchattention,asmanyspatialmeaningsof

9

NP

PP

AdvP

THAT-CL

Page 342: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 9 of 13

prepositionsaremetaphoricallyextendedintotemporalmeaningsorothersemanticallymoreabstractdomains.Meanwhile,alsosyntacticallycomplexconstructionsdisplaysynchronicpolysemy,astheyaresubjecttothesamediachronicprocessesofextension.TheEnglishDitransitiveconstruction(Goldberg1995)commonlyencodesanintendedtransferofanobject,butisbeingusedmetaphoricallyinexamplessuchasThymegivesthesoupaniceflavor,whichdenotesacausalrelationshipratherthanatransfer.Thesemanticextensionofconstructionscanbestudiedthroughdiachroniccorpusanalysis,inparticularthequalitativeanalysisofambiguousexamplesthatrepresentbridgingcontextsbetweentwodifferentinterpretations.

PetréandCuyckens(2008)studythedevelopmentoftheverbalprefixbe-fromOldEnglishonward.OldEnglishhadavarietyofverbalprefixesindicatingspatialdirections,asina-drifan‘driveout/away,’be-rowan‘rowround,’orto-brecan‘breakasunder.’ThelossofthisprefixationpatterncanbeseenasalevelingprocessresultingfromtheOVtoVOshiftthatEnglishunderwent.OftheOEprefixes,onlybe-remainssporadicallyproductiveinexpressionssuchasbe-trainered‘wearingrunningshoes’orbe-wigged‘wearingawig.’Itisthusofinterestwhyasingleconstructionresistedthetideofasyntacticseachange,anditisthisquestionthatPetréandCuyckensinvestigate.BasedoncorpusdatafromtheYCOE andYPC, theyarguethatthedominantsenseoftheconstructioninOEwasthespatialmeaningofenclosure,asinbesiege.Thissenseservedasaprototypefromwhichother,moreabstractmeaningsdevelop.Thesemanticextensionswiththehighestfrequenciesarelabeled‘ExtensiveCoverage,’‘TotalAffectedness,’and‘Furnishing.’Thelatteristhemostrecentextensionandalsothemostproductiveinmodernusage(cf.be-trainered).

Themeaningextensionfromenclosuretoextensivecoverageismotivatedbymetonymy,specificallyapart–wholerelationship.Someusesofthebe-constructionextendedtheideaofenclosuretothecloselyrelatedideaofcompletecoverage.Examplessuchas(9a)illustratethisextension(PetréandCuyckens2008:152).

(9)a.Seapostolhinebegeatmidðamwæteretheapostlehimbe-pouredwiththewater‘Theapostledousedhimwiththewater.’

Thenextextensionabstractsawayfromthedomainofspaceandprojectstheideaofcoverageontoaffectedness.PetréandCuyckensarguethataneventofcoveringanentitywillinvitetheinferencethattheentityinquestionisalsostronglyaffected.Overtime,thespatialmeaningofthebe-constructionbecamesecondary,leadingtothepossibilityofexamplessuchas(9b),whichmakesnoreferencetoaspatiallocation(2008:154).

(9)a.Bewylþarameolceþriddandæl.be-boilof.themilkthirdpart‘Boilawayathirdofthemilk.’

Afurtherextensionofthebe-constructiontargetstheideaoffurnishinganentitywithanobjectoranattribute.Thisprocessinvolveshost-classexpansion(cf.section25.2.8),astheprefixattachestonominalandadjectivalhosts(bespouse,befoul),whicharetherebycoercedintoverbhood.Themeaningoffurnishingisthemainstayofthebe-constructioninPresent-DayEnglish,whereitoccursprimarilyintheformofparticipialadjectiveslikebespectacled.

PetréandCuyckensarguethatthesuccessivesemanticextensionsofthebe-constructionrendereditmoresalientthanotherOEprefixes.Becauseofthissalience,be-resistedtheprocessbywhichtheotherprefixessemanticallybleachedandsubsequentlyfelloutofusage.Thestudythusdemonstratestheimportanceofclose,qualitativeinvestigationsofconstructionaldevelopments,asthesemaygoagainstthegrainoflanguagechangesonalargerscale.

25.2.10AnalogicalExtension

InConstructionGrammar,grammaticalknowledgeisthoughttobeahierarchicalnetworkofsymbolicunitsthatdisplayvaryingdegreesofschematicity.Aconstructionthatishighlyspecific,forinstance,anidiomsuchasnotgiveadamn,maysproutoffshootssuchasnotgiveahootornotgiveamonkey's,inwhichspeakersreplaceonepartoftheidiomwithananalogouselement.Repeatedanalogicalextensionsmayovertimeleadtothe

10 11

Page 343: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 10 of 13

emergenceofageneralschemanotgiveaNP,whichinvitesfurtheradditionstotherangeofexpressionsoccurringinthisnowpartlyschematicidiom.Stepwiseanalogicalextensionmayofferadiachronicexplanationforconstraintsonthesynchronicuseofaconstruction,motivatingforinstance,whycertainlexicalitemscannotoccurwithit.Throughtheuseofdiachroniccorpora,thenotionofanalogicalextensioncanbeoperationalizedasanincreasingrangeofelementsthatoccurwithinagivenconstructionalslot(cf.section25.2.3onproductivity).

Takingsuchanapproach,Israel(1996)usestheOEDtostudytheEnglishWayconstruction,whichexpressesthecreationofa(commonlymetaphorical)paththroughaterrainwithsubstantialobstacles.ThisconstructionalmeaningisreflectedintwotypesofmainverbsthattheWayconstructionaccommodatesinmodernusage:verbsdenotingthemannerofamovement,suchascrawl,wind,orplod,andverbsdenotingthemeansofcreatingapath,suchasdig,cut,orforce.Thefocuswillbeonthelattertypehere.Israelreportsthatby1650,theOEDcontainsexampleswithmeansverbssuchaspave,smooth,andcut,whichdirectlyrepresentthedomainofroadbuildingandpathclearing(10a).Acenturylater,theOEDattestsmoreverbsfromthesamesemanticdomains,butadditionallyalsoverbsfromthedomainoffighting(10b).Sinceroadbuildingrequiresforce,Israelarguesthattheoriginalmeaninghasundergoneanalogicalextensiontotheuseofforceinordertomakeprogressonapath.Example(10c)demonstratesthattheconstructionhasbecomeevenmoreschematizedbythenineteenthcentury,duringwhichanevengreaterrangeofco-occurringverbsisattested.Semantically,theuseofforcetomakeprogresshasbeenextendedtoanypurposive,goal-directedactivity.

(10)a.ArminiuspavedhiswayfirstbyaspersingandsugillatingthefameandauthorityofCalvin.(1647)b.Everystepthathetakeshemustbattlehisway.(1794)c.Notonemaninfivehundredcouldhavespelledhiswaythroughapsalm.(1849)

25.2.11CollocationalChange

Theprecedingsectiondiscussedtheideathattheoccurrenceofaconstructionwithnewlexicalelementsisindicativeofanalogicalextension,andhencesemanticchange.Besidesqualitativechangesinthecollocatesetofaconstruction,therearealsoquantitativechanges.Certaincollocatesmaybecomemoreorlessfrequentovertime,andtheseshiftsmaybenolessindicativeofchangethantheoccurrenceofnewlexicalcollocates.Aquantitativeanalysisofshiftingcollocationalpreferencescanthusoffersomeinsightintothesemanticdevelopmentofagrammaticalconstruction.Withthisaiminmind,Hilpert(2006a;2008)appliesthemethodofdistinctivecollexemeanalysis(Stefanowitsch,thisvolume)todiachronicdata.Distinctivecollexemeanalysiscontraststwoormoreconstructionswithrespecttotheircollocationalpreferences.Unlikethesynchronicvariant,thediachronicapplicationdoesnotcontrastmultipleconstructions,butstudiesasingleconstructionacrossmultipleperiodsoftime.Onthebasisofdiachroniccorpora,themethodcandeterminewhattypesofco-occurringelementsweretypicalofagivenconstructionatdifferenthistoricalstages.

UsingtheexampleofEnglishshall,Hilpertshowsfirstthattheconstructionappearstoremainstablewithregardtoitslexicalcollocates.InthePPCEME,whichisdividedintothreeperiodsoftime,shalloccurswithremarkablysimilarsetsofcollocatesbetween1500and1710.Ineachofthethreeperiods,theverbsbe,have,see,andfindareamongthefivemostfrequentverbs.Thecollostructionalmethodabstractsawayfromitemsthatarecommonineachperiod,insteadhighlightingthosethataresignificantlymorefrequentthanexpected,giventheiroverallfrequenciesacrossthethreeperiods.Itemsarejudgedasdistinctiveiftheyoccurfrequentlyinoneperiodbutarerelativelysparseinthetwoothers.Inthisway,differencesbetweenthethreeperiodsareaccentuatedandsemanticdevelopmentsbecomevisible.Technically,themethoddeterminesforeachverbwhethertheobservedoccurrencesdeviatesignificantlyfromachancedistribution.Thiskindofcalculationdeterminesthemostdistinctiveelementsforeachofthethreeperiods,promotingtheranksofverbsthataremaximallyunevenlydistributed.ThetenmostdistinctiveelementsforeachperiodareshowninTable25.2.Allverbsthatareshownaresignificantlydistinctivefortheirrespectiveperiods;aCollStrvaluelargerthan1.3signalssignificanceatthelevelofp<0.05.

Themostdistinctiveverbofthefirstperiod,fortune,occursninetimesinthefirstperiod,butnotatallelsewhere.Itappearsasthemostdistinctiveelementofthefirstperiod.Themostdistinctiveverbsarethereforethosethatoccurwithsomefrequency,butareidiosyncraticforjustoneperiod.Changesinthemostattractedverbscanbeinterpretedasasymptomofongoingsemanticchange,andtheobservedtendenciescanbecomparedagainst

Page 344: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 11 of 13

pre-existinghypothesesaboutthedevelopmentoftheconstruction.

AtendencythatcanbeobservedinTable25.2isanincreaseofstativeverbsoverthethreeperiods.Thefirstperiodlistsnone,thesecondperiodhascontain,thethirdperiodlistsstay,stand,extend,andbe.Anothertendencythatemergeswhenactualexampleswithdistinctiveverbsareexaminedisthatshallinthethirdperiodtypicallyassumesatext-structuringfunction,asisillustratedinthefollowingexamples.

(11)a.Ishalladdalittlemoreevidence.b.Someproductsandplacesprovideafewexceptionstothispattern,asweshalldiscoverinthenextchapter.c.WeshallexaminethisfurtherinChapter7.d.Ishallrefertothiscomparativestudywhereappropriateinthisreport.

Whereastheauxiliaryshallhaslargelyfallenoutofusageasageneralfuturemarker,theidiomaticityoftheaboveexamplessuggeststhatishasfoundafunctionalnicheasatext-structuringdeviceinformalwriting.

Table25.2.Toptendistinctivecollexemesofshalloverthreeperiodsoftime

1500–1570 1570–1640 1640–1710

Verb N CollStr Verb N CollStr Verb N CollStr

fortune 9 3.80 incur 13 3.38 add 14 5.52

wage 8 3.38 assemble 7 2.68 direct 8 5.46

divide 8 2.60 feed 7 2.68 discover 5 3.41

appear 24 2.07 hear 43 2.56 examine 7 3.39

perceive 11 1.92 offend 12 2.17 stay 6 2.81

understand 26 1.81 fall 14 2.04 refer 4 2.73

beg 7 1.75 contain 7 1.97 stand 11 2.65

require 9 1.66 enjoy 5 1.92 endeavor 7 2.60

say 28 1.61 imagine 5 1.92 extend 15 2.58

attain 5 1.50 do 65 1.79 be 234 2.50

25.3.ConcludingRemarks

Theintroductionposedthequestionwhycorporaareusefulforthestudyofconstructionalchange.Astheprevioussectionshaveillustrated,constructionschangeinfrequency,form,andfunction,typicallyinagradualfashion.Theprimarymeanstodocumentchangeintheusageofaconstructionarefrequencymeasurementsinconjunctionwiththeuseofinferentialstatistics.Corporaholdthepotentialtodemonstratethatthereareindeeddifferencesbetween,say,theusagepatternsofaconstructionintwodifferenthistoricalperiods,oracrosstwodifferentagegroupsinsynchronicdata.Especiallywhennoncategoricalchangesareatstake,quantitativedatafromcorporamayprovidetheonlymeanstoestablishthatachangeisunderway.

Thecasestudiesdiscussedabovehavealsoshownthepracticalsideofcorpusworkinthestudyofconstructional

Page 345: The Oxford Handbook of Construction Grammar

Corpus-based Approaches to Constructional Change

Page 12 of 13

change.Insomecases,acloseexaminationofcorpusexamplesfromdifferenthistoricalperiodsalreadyrevealsthatagivenchangehasoccurred.Inmanycases,however,analysesrelyontheexhaustiveretrievalofagivenconstructionfromacorpus,becausetheentiresetoftokensisneededtoarriveatgeneralizationsabouthowtheconstructionchanged.Annotationofalltokensforformalandfunctionalfeaturesallowsquantitativeanalysesofthefactorsthatchangedaconstruction,andtherebyallowstheresearchertodecidebetweencompetingexplanationsofthesamephenomenon.

Finally,itneedstobebrieflydiscussedhowtheconstructionalperspectivesetsaparttheworkdescribedabovefromotherapproachesinhistoricallinguistics.Onatheoreticallevel,workonconstructionalchangesetsouttodocumentphenomenathatcannotbereducedtomoregeneralpatternsoflanguagechange(i.e.,itaimstofindprocessesthatareconstruction-specific).MuchasearlyworkinsynchronicConstructionGrammar(e.g.,Fillmore,Kay,andO’Connor1988)determinedtheaspectsofagivenconstructionthatwerenotreducibletomoregeneralgrammaticalpatterns,workonconstructionalchangestudiesthoseprocessesthatdonotreducetoabroaderdevelopmentofgrammaticalreorganization.Inordertodemonstratethatachangeisindeedidiosyncratic,notgeneral,corpusdataareagainofutmostimportance.Onlyquantitativecorpusdatacandocumentthat,forinstance,twoseeminglyrelatedprocessesareinfactindependent.Whiletheconstructionalperspectivethusplacesgreatimportanceontheparticularitiesoflanguagechange,itisalsocommittedtofindinggeneralizationswhereverpossible.Changesinhistoricalcorpusdatamayyieldevidenceforthedevelopmentofabstractconstructionalpatternsthatrepresentlinguisticgeneralizations.Determiningtheexactlevelofabstractnessatwhichagivenconstructionneedstobepositedisagainonlypossiblewithrecoursetocorpusdata.

Insum,corpus-basedapproachestoconstructionalchangeaddressawidearrayofphenomenawithasimilarlywidearrayofmethods,butconvergeonthetheoreticalaimofofferingthebestdescriptionsof—andexplanationsfor—thedevelopmentofform-meaningpairingsinlanguage.

Notes:

(1.)BROWNfamilycorpora,http://khnt.aksis.uib.no/icame/manuals/index.htm.

(2.)BNC(TheBritishNationalCorpus),http://www.natcorp.ox.ac.uk/,http://corpus.byu.edu/bnc/x.asp.

(3.)ARCHER(ARepresentativeCorpusofHistoricalEnglishRegisters),nowebsite.

(4.)TIMEcorpus,http://corpus.byu.edu/time/.

(5.)YORK(SaliTagliamonte'sCorpusofYorkEnglish).http://www-users.york.ac.uk/~st17/.

(6.)CLMET(CorpusofLateModernEnglishTexts),https://perswww.kuleuven.be/~u0044428/.

(7.)CEEC(CorpusofEarlyEnglishCorrespondence),http://www.helsinki.fi/varieng/domains/CEEC.html.

(8.)SPRÅKBANKEN(variousSwedishcorpora),http://spraakbanken.gu.se/.

(9.)PPCME2(ThePennParsedCorpusofMiddleEnglish),http://www.ling.upenn.edu/hist-corpora/.PPCEME(ThePennParsedCorpusofEarlyModernEnglish),http://www.ling.upenn.edu/hist-corpora/.

(10.)YCOE(TheYork-Toronto-HelsinkiParsedCorpusofOldEnglishProse),http://www.ling.upenn.edu/mideng/ppcme2dir/YCOE/doc/annotation/YcoeLite.htm.

(11.)YPC(TheYork-HelsinkiParsedCorpusofOldEnglishPoetry),http://www-users.york.ac.uk/~lang18/pcorpus.html.

MartinHilpertMartinHilpertisanAssistantProfessorofEnglishLinguisticsattheUniversityofNeuchâtel.HisresearchinterestsincludeGrammaticalization,ConstructionGrammar,andCognitiveLinguistics.Heapproachesthesetopicswiththetoolsandmethodsofcorpuslinguistics.HismonographonGermanicFutureConstructions(JohnBenjamins,2008)documentschangeinthedomainoffuturetimereferenceacrossseverallanguagesandconstructions.HisbookConstructionalChangeinEnglish(CUP,2013)worksouthowConstructionGrammarcanbeappliedtoissuesoflanguagechange.Martin.hilpert@unine.ch

Page 346: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 1 of 11

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntaxOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0026

Dialects,Discourse,andConstructionGrammarJan-olaÖstmanandGraemeTrousdaleTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterexaminesconstructionalapproachestolanguagevariationandaspectsofdiscourse,includingidiolectalandcommunityvariation.ItprovidesthreecasestudiestoillustratethemodelingofinherentvariabilityincognitivelinguisticsingeneralandinConstructionGrammarinparticular.Thechaptershowshowtheusage-basednatureofmuchresearchinConstructionGrammarmaybeappliedtoemergentvariationindiscoursestructures,particularlyindialogiccontexts.

Keywords:inherentvariability,modelingvariation,communityvariation,cognitivelinguistics,ConstructionGrammar,discoursestructures,dialogiccontext

26.1.Introduction

Thischapterisconcernedwithhowconstructionalapproachestolanguagemodel,andtherebyattempttoexplain,particularkindsofvariability—idiolectalandcommunityvariationwhichcorrelateswithgeographicalspace,andinteractionalanddiscoursevariability.Thenatureoflinguisticvariationisatopicofinteresttomanylinguistsandiscoveredinotherpartsofthishandbook(e.g.,Croft'schapterdealsinpartwithtypologicalvariation,andHollmann'schapterwithsociolinguisticvariation).HereweareprimarilyconcernedwithwhatHudson(1997),followingLabov(1969),hasdescribedas‘inherentvariability,’thatis,withthefactthatindividualspeakershaveavarietyofformswhichtheymayusetorefertothesamething,thefactthatnetworksofspeakersinacommunitybehaveinstatisticallyregularways,andthefactthatdiscoursecontextcorrelateswithvariablelinguisticbehavior.Butsinceconstructionalapproachesareoftenusage-based,wewillinourdiscussionofinteractionalanddiscoursevariabilityalsofindreasontogobeyondsociolinguisticvariabilityintotherealmofemergentvariability.

Inherentvariabilityisofconsiderableimportancetolinguistictheory,becauseanadequatelinguistictheorymustbeabletoaccountforthestructuralvariation,thecontextualvariation,andthevariationinfrequencyofuse,threetypesofvariationidentifiedbyHudson(2007a)inhisWordGrammaraccountofdialectsyntax.Bystructuralvariation,wemeanthecapacityforatheorytoaccountforpatternsofandconstraintsontheformalvariationwhichindividualspeakersofalanguagecommand.Bycontextualvariation,wemeantheabilityofthetheorytoaccountadequatelyforthefactthatspeakersadoptdifferentstructuresondifferentoccasions,withdifferentspeakers,fordifferentcommunicativepurposes.Byvariationinfrequencyofuse,wemeantheabilityofthetheorytoreflectaccuratelytheuseofdifferentproportionsofvariousstructuresbythesamespeaker.

Inadditiontothinkingofspeakervariationbyplace,wealsoconsiderthenatureofdiscoursevariation.Hereweseevariabilityascloselyconnectedtocontextandtakeabroadviewonwhataspectsofcontextinfluenceandconstrainlanguagestructure.Linguisticstructureisstillinfocus,butwithrespecttotheinfluenceofdiscourse,thelinguisticchoicesaffectedarenotonlythoseofmorphologyorsyntax;thechoiceofawordoraphraseorevenofaparticularactivity(manifested,forexample,inaparticularsequenceofutterances)isamatterofvariability.In

Page 347: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 2 of 11

thissphereofconstructionalanalysisscholarshaveseenitascruciallyimportanttostakeoutthediscoursalfactorsthatinfluencevariablechoices.Thiswillalsobeonefocusofthisoverview.

Intheconstructionalliteraturesofar,wecandistinguishthreemacro-approachestocontext;thesearecomplementarybuthaveslightlydifferentstartingpoints.First,therearefruitfulattemptsataddingcontextualinformationintheformof,forexample,apragmaticsattribute(cf.MichaelisandLambrecht1996;Lambrecht2004;Leino,thisvolume)andtherebyspecifyingissuesoftopicandfocus,andthecognitiveaccessibilityofalinguisticunit.Second,frequencyandstatisticalanalysesofdatahave,through,forexample,factoranalysis,providedcontextsforvariationthathaveprovedusefulbothfordiscourseandcollostructionalanalyses(cf.thearticlesbyStefanowitschandHilpertinthisvolume).Thethirdapproachisinductiveandqualitative,andusesdiscoursemethods,andmethodsfromsociopragmaticsandinteractionallinguisticstomodelvariability.Thisistheapproachwewillprimarilybeconcernedwithinthelatterpartofthischapter.

26.2.DialectVariationandConstructionGrammar

Inthispartofthechapter,wedealwithdialectal(regional)variationinlanguage,andhowthisfitsintoatheoryoflanguagewhichtakestheconstructionasthegrammaticalprimitive.Weaddressinparticularthefollowingtwotopics:

•thenotionofinherentvariability(Labov1969;Hudson1997,2007c)•modelingdialectvariationinaconstructionalframework.

Regionalvariationinalanguageintersectswithmanyotherkindsofvariation;anygivenidiolectisnotshapedsolelybythegeographicalareainwhichthespeakergrewuporresidesbutalsobythespeaker'ssocialcharacteristics(onwhich,seefurtherHollmann,thisvolume).Furthermore,regionallectsareinparttheproductofdifferenttrajectoriesofconstructionalchange(Fried,thisvolume).Nonetheless,itisclearthatparticularconstructionsmaybeassociatedwithparticularregionalvarietiesinthemindsofindividualspeakers.Thismaybeeasilyillustratedbyconsideringspeakers’knowledgeofatomic,substantiveconstructions(seefurtherthediscussioninHoffmann,thisvolume).SomespeakersofBritishEnglishareawareofthefactthattheconcept‘establishmentwhichsellsalcoholforconsumptionoff-premises’isassociatedwithtwodifferentforms,oneBritish(off-license)andoneAmerican(liquorstore).Suchknowledgemustbetheproductofparticularusageevents:itisimpossibleforBritishspeakerstoknow,forinstance,thatotherpeoplehaveawordotherthanoff-licensefortheconcept‘establishmentwhichsellsalcoholforconsumptionoff-premises’untilonehearsorreadsliquorstore,andoneistoldofthemeaningoftheexpression,oroneinfersitfromthecontext.Knowledgeofdialectvocabulary,therefore,isusage-based.Furthermore,whenconstructionsaredefinedasconventionalsymbolicunits(Langacker1987;Croft2005a),theyareassumedtobesharedbyanetworkofspeakers,anddifferentsubnetworksmayhavedifferentconventions.

Thehypothesisthatlanguageisaconstructionalnetworkwouldsuggestthatknowledgeofdialectvocabulary,andknowledgeofdialectsyntax,shouldbeorganizedaroundthesamegeneralprinciples.Inthefollowingsection,weexplorethenatureofthisknowledgeofvariation.

26.2.1InherentVariabilityandModelingDialectVariation

AsAdgerandTrousdale(2007:261)observe,“themodellingof(syntactic)variationinandacrossdialectsisacriticalissueinanytheoreticalframework,asvariationisubiquitousinlanguage,andthefactthatlanguagecanvaryraisesimportantquestionsregardingwhatthattheoryisactuallymodelling.”Differentmodelsofgrammarhavedifferentwaysofdealingwithvariability,thoughitisclearthatmanylinguistictheoriesareincreasinglytakinginherentvariability,andfrequencyeffects,moreandmoreseriously(inadditiontothecognitivelinguisticapproachofWordGrammardiscussedabove,seeforinstanceAdger(2006)onminimalistapproachestomorphosyntacticvariationinBuckieScots,orBresnanandFord's(2010)probabilisticaccountofdifferencesbetweenspeakersofAustralianandAmericanEnglishintheiruseofdativeconstructions).Evenwithinconstructionalapproachestolanguagestructure,differencesaswellassimilaritiesexist.LeinoandÖstman(2005),forinstance,pointoutthatdifferentwaysofformalizingconstructionswillleadtodifferentwaysofaccountingfordialectvariation:HPSG-typeConstructionGrammarsmaybeparticularlyconcernedwiththeproperformalizationofinheritancehierarchies,

Page 348: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 3 of 11

whilecognitiveConstructionGrammarsmaybemoreconcernedwithestablishingprototypeeffectsandthemodelingofpartiallysanctionedconstructsinagivendialect.AsLeinoandÖstman(2005)observe,however,suchconcernsarenotmutuallyexclusive;andallapproachestomicro-variationofthiskindmustbeconcernedwithappropriateconstraintsonvariation,howeverthoseconstraintsareformalized.

Theremainderofthissectionfocusesonusage-basedapproachestoconstructionalvariationwithinalanguage.LikeWordGrammar,usage-basedConstructionGrammarsareconcernedwiththeinteractionbetweenlinguisticstructure,andthedeploymentofthatstructuretomeetcommunicativeneeds.Principlesofusage-basedmodelsoflanguageprovideexplanationsforfrequencyeffects,andforanunderstandingoftheroleofcontextinlanguagevariationandchange,afeaturesharedwithvariationistsociolinguistics(Hollmann,thisvolume);themodelingofvariationinWordGrammarisparticularlyconcernedwithbothprototypeeffectsanddefaultinheritancehierarchies(see,forinstance,Hudson1997ont/ddeletion,andthedevelopmentofdo-periphrasisinEnglish);theseissues,asnotedabove,arealsofeaturesofLeinoandÖstman's(2005)construction-basedaccountofdialectanddiscoursevariationinFinnish.

Usage-basedmodels(BarlowandKemmer2000;Bybee2010,thisvolume)areconcernedprimarilywith:

•testingthehypothesisthatstructuralpropertiesofaparticularvarietyarelargelytheproductofinductionfromexperienceoflinguisticevents;

•accountingforstatisticalpatternsacrossalargenumberofusageevents(typicallyadducedfromcorporaofnaturallanguage);

•accountingforthenatureofregularitiesandirregularitieswithinandacrosslanguages.

Thesemodelsoperateonthebasisthatusersofalanguagewillbeexposedtoquantitativeandqualitativedifferencesinstructurewhentheyinteractwithotherspeakers.Furthermore,theyassumethatlanguageisdynamic,“amentalstructurethatisinconstantuseandfilteredthroughprocessingactivitiesthatchangeit”(Bybee2010:6).Thisuse,ofcourse,doesnotpreventtheestablishmentofregularpatterns:frequentlyusedpatternsaresubjecttogreaterentrenchment:“themoreoftenweexperienceapattern,themorefirmlyembeddedinourcompetenceitbecomes”(Hudson1997:83).Butthenatureofdifferentregionalvarietiesmeansthatforsomespeakers,exposuretoconstructionxwillbemorefrequentthanexposuretoconstructiony.Animportantissueforusage-basedConstructionGrammarsistoestablishhowknowledgeofvariationworks,since“instancesofuseimpactthecognitiverepresentationoflanguage”(Bybee2010:10).Inthefollowingsection,wedescribethreeexamplesofusage-basedapproachestoregionalvariation,showinghowthenotionofconstructionmaybeusedtoelucidatemorphosyntacticpatternsassociatedwithparticulardialects;conversely,weshowhowsuchdialectmaterialinformsourunderstandingofthenatureofgrammaticalconstructions.

26.2.2ThreeCaseStudies

Inthissection,webrieflyconsiderthreeexamplesofstudiesofdialectvariationwhichareamenableto,orexplicitlycouchedin,aconstructionalapproachtodialectvariation:theDitransitiveconstructioninLancashireEnglish(HollmannandSiewierska2011);theemergenceofanewintensifiermassa'sinFlemishvarietiesofDutch(DeClerckandCollemanforthcoming);andverb-constructionassociationsinnewEnglishes(MukherjeeandGries2009).

26.2.2.1TheDitransitiveinLancashireEnglishWhilethesyntaxandsemanticsofthestandardEnglishDitransitiveconstructionhavebeencentraltothedevelopmentofaspectsofthetheoryofConstructionGrammar,thepatternsassociatedwithnonstandardvarietiesofEnglishhavereceivedlessattention.Examplesofnonstandardditransitivesinclude:

(1)Shegaveabooktheman(HughesandTrudgill1996:16)(2)Shegaveittheman(SiewierskaandHollmann2007:86)

SiewierskaandHollmann(2007)identifyanumberoffactorsthatmaybeassociatedwithalanguageuser'spreferenceforonevariantoveranother.Theseinclude:

Page 349: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 4 of 11

•argumentweight(whetherthetheme/recipientisapronounorafullNP)•verb/argumenttype(whethertheindirectobjecthasthesemanticroleofrecipientorbeneficiary)•propertiesofthepronominalobjects(personandanimacy,forinstance)•typeofdiscourse(spokenvs.writtenEnglish;formalvs.informaldiscourse)•regionalprovenanceofthespeaker.

HerewefocusonHollmannandSiewierska'sdiscussionofregionalprovenance,withreferencetootherfactorswhereappropriate;wealsorestrictthediscussiontotheDitransitiveconstructionsexemplifiedby(1)and(2),withonlyoccasionalreferencetothePrepositionalDativeconstruction.

First,theliteraturesuggeststhatconstructssuchas(1),wherethethemeprecedesthebeneficiary/recipient,arepeculiartoBritishEnglish.Whethertheme-recipientpatternsaremorecommoninnorthernEnglishvarietiesorsouthernEnglishvarietiesisamatterofdebate:Kirk(1985)andKoopmanandvanderWurff(2000),forinstance,seethetheme-recipientorderasafeatureofsouthernEnglish,whileHughesandTrudgill(1996)considerthatordertobeverycommoninnorthernEnglish,evenamongeducatedspeakers(whoareassumedtobemoreexposedtothestandardEnglishpatterns).However,SiewierskaandHollmann(2007)maketheimportantobservationthatthedataonwhichthesejudgmentsaremadecomefromcorporawherefine-graineddistinctionsmaynotbeavailable.Thecorporainclude,forexample,theSurveyofEnglishDialects(SED),avaluableresourceforthestudyofsomeaspectsofEnglishdialectology,butnotsufficientlynuancedtoallowfordetailedevidenceofconstructionalvariation:onlyoneresponseperinformant,andthataproductofaparticularelicitationtest,wheretheinformantisaskedtocompleteaparticularsentence(SiewierskaandHollmann2007:89).

InordertoestablishtheparticularpatternsassociatedwithDitransitiveconstructionsinLancashireEnglish(inthenorth-westofEngland),SiewierskaandHollmann(2007)madeuseoffourcorpora:tenspokentextsfromtheBritishNationalCorpus(BNC), whichwerecategorizedas‘Lancashiredialect’;incidentalrecordingsmadeaspartoftheSEDprojectfromninelocationsinLancashire;twenty-threetextsfromtheFreiburgEnglishDialectCorpus(FRED); andmaterialfromthreelocationsinLancashire,takenfromtheHelsinkiCorpusofBritishEnglishDialects.Thecorpusyielded449instancesofbothDitransitiveandPrepositionalDativeconstructions(83%and17%,respectively;SiewierskaandHollmann2007:91).

AmongtheDitransitiveconstructions,recipient/beneficiary-themeorderingwassignificantlymorefrequentthanthereverse(94%and6%,respectively;SiewierskaandHollmann2007:91).ThisgenerallymorefrequentpatternwasinfactcategoricalwhenbothobjectswerefullNPs(SiewierskaandHollmann2007:91).Withpronominalobjects(taking‘pronominal’heretorefertopersonalpronounsonly,andexcludingformslikeeverything),thereis,however,adifferentpattern.Insuchcases,thereisaslightpreference(65%)fortheme-recipient/beneficiaryorder(asinhegaveitme);asSiewierskaandHollmann(2007:92)note,however,thetotalnumberofinstancesofDitransitiveconstructionswithtwopersonalpronounobjectsinthecorpusislow(N=23),sotheresultsmustbeinterpretedcautiously.ItseemsthatthisparticularsubconstructionofthemoregeneralDitransitiveconstruction(i.e.,wherebothobjectsarepronouns)patternsmorelikethePrepositionalDativeconstructioninthisdialect:“in81%oftheclauseswithbothapronominalthemeandrecipient/benefactive,theformerprecedesthelatter”(SiewierskaandHollmann2007:92).Inotherwords,ofinstanceswherebothobjectsarepersonalpronouns,theprepositionaldative(e.g.,hegaveittome)ismostfrequent;inthisrespect,LancashireEnglishislikestandardBritishEnglish;however,inLancashireEnglish,thesecondmostcommonpatternisnotthecanonicalditransitivepatternwithrecipient/beneficiaryfirst(e.g.,hegavemeit)asinthestandardvariety,butratherthe‘theme-first’variant(e.g.,hegaveitme).

InthespiritofCroft(2001),SiewierskaandHollmann(2007)arguethatsuchdatashowtheimportanceoflinguists’attentiontovariationintheory-buildingandnotethatGoldberg'sclaim(Goldberg1992:71n.4)regardingtheorderingofthemeandrecipient,namelythatthethemealwaysfollowstherecipient,holdstrueonlyfor(standard)AmericanEnglish.InCroft'sfunctional-typologicalapproachtolanguagestructure,theconstructionalprimitivesarevariety-specific.Asignificantquestionhere,ofcourse,iswheretheboundariesaroundaparticularvarietylie.BasedontheirLancashiredata,SiewierskaandHollmann(2007:98)arguethat“alanguage-specificdouble-object[i.e.,Ditransitive]constructionistoosimplistic,”whichimpliesthattheyseeaplacefor‘dialect-specific’constructions.Theproblemiswherethisnarrowingstops:thetrajectoryfromlanguage(English)tonation(AmericanEnglish)toregion(LancashireEnglish)hasasitsendpointtheindividualspeaker.Thismoreradical

1

2 3

Page 350: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 5 of 11

approachtovariationshiftsthefocusawayfromsocialandculturalcategorieslike‘language’and‘dialect’inordertoforegroundtheindividualspeaker,hisorherinternallinguisticnetwork,andthesocialandculturalknowledges/heassociateswithparticularforms(Hudson1996,2007c).Thatis,oneofthethingsthatspeakersofLancashireEnglishmayknowaboutDitransitiveconstructionsisthatconstructslikehegaveitmeareassociatedwiththeconcept‘Lancashire’(knowledgewhichisderivedfromthespeaker/hearer'sexposuretoparticularusageevents),thatis,that‘Lancashire’isassociatedwiththediscourse/pragmaticcomponentofthemeaningpoleofparticularconstructions(onwhich,seefurtherHollmannandSiewierska2011inrelationtodefinitearticleconstructionsinLancashireEnglish).This,ofcourse,wouldneedindependentverification.HollmannandSiewierska(2006)proposeamethodforgaugingspeakerawarenessofvariationbyestablishingpatternsofaccommodation.AsHollmann(personalcommunication)pointsout,formanyfeaturesofdialectsyntax,itmaybedifficulttoestablishwhatspeakersknowabouttheprovenanceofaparticularconstruction(thatis,whethertheconstructionishighlylocal,orsimplypartofasupralocalnonstandardvariety).

26.2.2.2ANewDegreeModifierinDutchDeClerckandColleman(forthcoming)provideanaccountofanewdegreemodifier(massa's)inFlemishvarietiesofDutch.Althoughnotexplicitlycouchedinaconstructionalframework,thepatternsassociatedwithFlemishmassa'sareclearlyamenabletoaconstructionalanalysis.Thedatausedcomeprimarilyfromacorpusofinformalwrittenlanguage,namelyasetofstudentblogpostings.

Thepatternsinquestionconcerntheuseofmassa'sinanexamplesuchas(3),takenfromDeClerckandColleman(forthcoming):

(3)Ikhadnietechtveelzinenmassa'sweinigtijd.‘Ididn’treallyfeellikeandIhad[lit.*masses/]verylittletime.’

Theuseoftheintensifiermassa'sseemstobesociolinguisticallyandregionallyrestrictedinDutch:itisprimarilyassociatedwiththeinformalspeechandwritingofyoungpeoplefromthenorth-westofBelgium,thatis,inwestFlandersandthewesternpartofeastFlanders.AsisthecaseinEnglish(see,interalia,Traugott2007;Brems2010),theevolutionofdegreemodifiersfromhistoricalbinominalconstructionsinvarietiesofDutchappearstobeaninstanceofgrammaticalconstructionalization,thegradualdevelopmentofnewform-meaningpairingsatvariouslevelsofschematicityandcomplexity.Inparticular,asubsetofsizenouns(suchaseenhoop‘aheap,’eenbeetje‘abit’)comestodevelopaquantifyingfunction;thismaythenleadtoafurtherdevelopment,wherebyquantifiersarereanalyzedasmoregeneralmarkersofdegree(Brems2007;DeClerckandCollemanforthcoming).

Theregionallyrestricteduseofmassa'sinconstructslinkedtotheDutchDegreeModifierconstructionillustrateshowdifferentpartsofaconstructionalnetworkmaygrowinparticularregionalvarieties.Forinstance,constructionsofthetype[[eenbeetjeAdj]–[‘slightlyAdj’]],whichsanctionsconstructssuchas[[eenbeetjedronken]–[‘slightlydrunk’]]appeartobemoreentrenched,andmorewidespreadinDutch,thaninstancessuchas[[massa'sAdj]–[‘veryAdj’]],whichsanctionconstructssuchas[[massa'sgrappigmens]–[‘veryfunnyperson’]],whicharelimitedtocertainspeakersfromcertaindialectareas,andareusedincertaindiscoursecontexts.Thisnotionofsanction(Langacker1987;seealsoBroccias,thisvolume)isparticularlyhelpfulforourunderstandingofregionalconstructionalvariation:forsomespeakers,themassa'sconstructionisnotsanctionedbythemoregeneralDegreeModifierconstruction—utterancessuchasmassa'sgrappigmenswouldbeconsideredungrammatical;forothersitispartiallysanctioned(asanextensionofthemoretypicalpattern,usedinveryrestrictedcontexts);foryetothers,itmightbefullysanctioned,andacceptableforuseinallcontexts,dependingonthedegreeofentrenchmentinaparticularidiolect.Regionalvarieties,understoodasanetworkofidiolectsassociatedwithaparticulargeographicalspace,willdisplaydifferentdegreesofentrenchmentofinnovativeconstructions,asthechangespreadsthroughdifferentnetworksatdifferentrates.

26.2.2.3Verb-constructionAssociationsinNewEnglishesAstudyusingthreesub-corporaoftheInternationalCorpusofEnglish(ICE) illustrateddifferentpatternsofassociationsbetweenverbsandconstructionsinSingaporean,Indian,andHongKongvarietiesofEnglish(MukherjeeandGries2009).Byusingthesedifferentsubcorpora,MukherjeeandGrieswereabletoidentifydifferencesbetweenBritishEnglishandnewEnglishes,andbetweenthethreenon-Britishvarieties.Theirstudyisparticularlyconcernedwiththedegreeof‘structuralnativization,’“theemergenceoflocallycharacteristic

4

Page 351: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 6 of 11

linguisticpatterns”(Schneider2007:5–6),measuredinthiscasebydifferencesinthefrequencyofcollostructionalalternations(seeStefanowitsch,thisvolume,foradiscussionofcollostructionalanalysismoregenerally).SingaporeEnglishisconsideredtobethemostdifferentiatedfromBritishEnglishinSchneider'sdynamicmodelofthedevelopmentofnewEnglishes,HongKongEnglishisconsideredtheleastdifferentiated,withIndianEnglishintermediary(MukherjeeandGries2007:31–34).

Byadoptingamultipledistinctivecollexemeanalysis,MukherjeeandGrieswereabletoidentifywhichofthreekindsoftransitiveconstructions(Intransitive,Monotransitive,andDitransitive)eitherattractedorrepelledaparticularverb,aswellasthestrengthofthatattraction/repulsion.Havingestablished(usingICE-GB,acorpusofBritishEnglish)whichverbsweremorelikelytoappearintheIntransitive,Monotransitive,andDitransitiveconstructionsinthe‘parent’variety,afurthersetofverbswasincluded;theseverbs,whilefrequentinICE-GB,werenotespeciallyassociatedwithanyofthetransitiveconstructions.ExamplesofverbsattractedtotheDitransitiveconstructioninBritishEnglishincludeverbsoftransfersuchasgive,lend,andsend;thoseattractedtotheIntransitiveincludestativeverbssuchassit,wait,andlive;thoseattractedtotheMonotransitiveincludeverbssuchashave,make,andinvolve;andthe‘neutral’verbsincludeserve,bet,and

Table26.1.Collostructionaldispreferencesofallverbsinpairwisecorpuscomparisons(fromMukherjeeandGries2009:46)

ICE-GBvs.ICE-HK ICE-GBvs.ICE-IND ICE-GBvs.ICE-SIN

same different NA same different NA same different NA

38 14 7 37 16 6 34 19 6

surprise.SomeoftheresultssuggestedthatBritishEnglishandthenewEnglisheswerealikeincollostructionalorganization:forinstance,itwasgenerallythecasethattheverbswhichpreferredtheDitransitiveconstructioninICE-GBalsopreferredtheDitransitiveinthecorporaofnewEnglishes,particularlytheSingaporeandHongKongcorpora.Thesamewastrueforverbswhich(dis)preferredtheIntransitiveconstructioninBritishEnglish;however,thesimilaritydidnotholdfortheMonotransitivecollostructionalpattern.Thebehavioroftheneutralverbswasparticularlystriking,sincethecollostructional(dis)preferencesseemedtovarysignificantlyacrossallfourvarietiesinvestigated(MukherjeeandGries2009:44–45).IntermsofdegreesofdifferenceofparticularvarietiesfromBritishEnglish,MukherjeeandGriesobservethatcollostructionalpatternsappeartocorrelatewiththestagetheyareatinSchneider'sdynamicmodelofthedevelopmentofnewEnglishes:“themoreadvancedthenewEnglishvarietyisintheevolutionarycycle,themoredissimilaritistopresent-dayBritishEnglishwithregardtocollostructionalpreferences”(MukherjeeandGries2009:46).ThisisillustratedinTable26.1.

26.3.DiscourseVariabilityandConstructionGrammar

Attemptsatexplicatingvariabilityindiscoursewithinconstructionalapproacheshavebeenguidedbytheideathatlanguageuseislinguisticactiontiedtothedifferent(communicative)practicesweareengagedin.Thismeansthatthelanguageweuseisnotonlyaffectedbycontextualrestraintsfromoutsidethatuse(i.e.,thecharacteristicsofthecommunityandculturewithinwhichsomethingissaidwillconstrainwhatformthatexpressiontakes),butthatwealsoconstrueactivitiesandsituationsbyouruseofparticularlinguisticexpressions.Constructionalapproachestodiscourse,context,andvariabilityattempttobetrulyusage-basedandtakeactuallyoccurringdiscourse,withits(inherentoremergent)variabilityasitsdata.Butalthoughrecentpromisingadvanceshavebeenmadetoaccountfortheseemingelusivenessofdiscourse(cf.,e.g.,Auer2011),thetemporalityandsequentialityoflinguisticinteractionisstilloneofthebiggestchallengesforanyapproachtogrammar,includingConstructionGrammar.

26.3.1ConstructionalDiscourse

Toillustratethedynamicityandcomplexityofdoingdiscourseconstructionalanalysis,wewillbrieflyhavealookatWellerisms.AWellerismisasayingoftheform“Itallcomesbacktomenow,”saidthecaptain,ashespatinto

Page 352: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 7 of 11

thewind.

InananalysisofWellerismsinSolf,Östman(2002)notedfirstthattheinternalstructureofWellerismsislicensedbyaveryparticularconstruction.Butmoreimportantlyforourpurposes,thereareveryspecificcircumstancesthathavetoholdinorderforaWellerismtobefelicitouslyused:theuseofaWellerismisaresponsiveact—itsequentiallyfollowsandisaresponsetowhatsomebodyelsehassaid;itisusedamongmembersofthesamecommunitytoenhancecamaraderie;itistypically(intheSolfcontext)usedtoanchorthespeakerstoajointculturalhistory;itisusedinanaffectivecontext;itcanbeuseddistributively,sothatseveralspeakerspartakeincoproducingtheWellerism;anditisusedasawaytoexpresstaboos,forexample,youcanswearorbeachauvinistorracistifyousaythatsomebodyelsehassaidwhatyouyourselfjustsaid—cf.(4),whereintheSolfsocietyswearingistaboo,unlessyouqualifyyourswearingbysayingthatsomebodyelsedidtheswearing.(Example(4)isafreetranslationofanactualoccurrence;Sanderisapersoninthevillagewhodiedsomefortyyearsago.)

(4)B.A:AndIseeyou’vegotalotofapplesthisyear.B:Awholedamnedlot,saidSander.

Thus,inadditiontotalkingabouttheinternalrelationsbetweenattributeslike‘wordclass’andvalueslike‘noun’inaconstruction,wealsoneedtobeabletospecifywhen,where,how,andwhysomethingcanbesaid—putdifferently,howitcanbefittedintoadiscoursecontext.IfwetakethevariableuseofexpressionslikeExcuseme,Sorry,Myapologies,andOuchtobeamatterofgrammar(asadiscoursegrammarianwould),thenwewillneedother(typesof)contextualattributesthanthosemadeuseofinsentence-grammaranalyses(e.g.,attributesspecifyingpoliteness,culturalcoherence,involvement,andgenre).

Thereisconsiderablecomplexityinhowtheexternalfeaturesaffectthevariabilityoflinguisticexpressions.InthecaseofWellerismswehaveafairlyrigidformulaictypeofexpressionthatislicensedby(i.e.,sanctionedtobeusedin)particular,verystrictlydefinedcontexts.UtteringaWellerismwillthereforealsocallforththatcontext.

Therearetwotypesoffeaturesthatspecifythecontextofanexpression.Thetraditionaltypeisonethatisdefinitionaloftheparticularconstruction.Togethertheinternalstructureandthesecontextualfeaturesmakeupthe(inthiscase,larger-than-sentence)constructionsasthelinguisticresourceswehaveatourdisposal.Thesefeatureshavetypicallybeenreferredtowiththesamecategorylabelsasconstruction-internalfeatures:syntax,semantics,pragmatics,andprosody.Butonthediscourselevelthesetakeslightlydifferentformsfromwhatweareusedtoonsentencelevel.Syntaxonthediscourselevelreferstotaxis(cf.Nir's2008notionof‘packaging’:isotaxis,parataxis,hypotaxis,endotaxis)andalsospecifieswhichelementsaconstructionneedstobemadeupof(e.g.,theprotasisandapodosisinanimplicationalconstruction,orthekindsofrelationswefindinMannandThompson's(1988)rhetoricalstructuretheory);semanticsrefersprimarilytoframesemanticcategories(cf.Fillmore1982,andworkscitedontheFrameNetsite:http://framenet.icsi.berkeley.edu);pragmaticstraditionallyreferstoinformationstructure(includingdegreesofactivity,ofaccessibility,andofnewness;cf.Leino,thisvolume);andprosodyreferstotextualanddiscourseprosody,bothintermsofepisodeintonationandintermsoffunction-carryingtunes(e.g.,inlullabies)(cf.Segerståhl2011).

Theothertypeofdiscoursefeaturesconstrainwhereandwhenanexpressioncanbeused,atthesametimeastheyprovidetherangeofthedifferentpossiblemanifestationstheexpressioncantake.Theelementsthatareconstrainedbysuchdiscoursefeaturesarenotonlypronunciations,oraspectsofmorphologyorsyntax,butcanbemultiwordandmulti-utteranceexpressions,co-constructedsequencesamongseveralspeakers,etc.

26.3.2FeaturesofDiscoursethatConstrainandInviteVariability

TherangeofapproachestodiscourseanddialogicalanalyseswithinaconstructionalframeworkiswellexemplifiedinthepapersinGünthnerandImo(2007),BergsandDiewald(2009),andinthe2010specialissueongrammarandinteractionofConstructionsandFrames:seeespeciallytheintroductionbyFried(2010b)tothelatter.TheseissuesarealsoaddressedfromtheperspectiveofdiachronicvariationandtheemergenceofconstructionsindialogiccontextsbyTraugott(2008b).

Thus,Linell(2009:99),inhisanalysisoftheresponsivex-och-xconstructioninSwedish,furtherdevelopsthe

Page 353: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 8 of 11

externalsyntaxofconstructionsinordertotakeintoaccount“whatconditionsonpriorsequences”agrammaticalconstructionsetsup,“whatconditionsonsubsequentsequences”itsetsup,andwhatare“itspreferredco-occurringresources.”AndWide(2009)elaboratessimilaraspectsinheranalysisofademonstrativeconstructioninaFinland-Swedishvariety,singlingoutfeatureslike‘sequentialposition,’‘activity,’‘coherence,’and‘functionalpotential.’

Whenweattempt‘tobuildin’thesetypesofcontextualfeaturesintoourconstructionalanalyses,werealizethatmanyquestionsarestilltoberesolved.Amostpertinentquestionisnaturallywhatfactorsofcontextshouldbetakenintoaccount,butanequallyimportantquestionishowthesefactorsaretoberelatedtoeachother.AnumberoftraditionalConstructionGrammarfeaturesmayalsoneedtobere-evaluatedwhentakenintothediscourserealm:constructions(asabstractgeneralizationsoverinstancesofusage)aremaybenotmerelyform-meaningpairs,buttripartiteform-meaning-functionconstellations;whenthecrucialConstructionGrammardistinctionbetweeninternalandexternalfeaturesistakenbeyondthesentenceintodiscourse,itisnotsoclear—albeitthatthedistinctionisarecursiverelation—whatis‘internal’anymore;theideathatnotallattributeshavetoreceivevalues(whichisasinequanonforanyusage-basedconstructiongrammaticalanalysis)willhavetobegivenamoreprominentplace:thenonassignmentofavalueisnotonlyalackofspecificationbutalsotheresultofacommunicativechoice;and,finally,traditional‘constituency’mayhavetobeproperlyre-evaluated(bybeingsocially,interactively,cognitively,andculturallyinformed)soastoallowanalysestogoagainsttheverybasicsofconstituencyintraditionalgrammar.

Workondiscourseaspectsofconstructionalanalysisisverytopicalatthemomentandthereareexcellentstudiesthatshowhowgrammarisinfluencedbyaspectsofdiscourse,buttheverynotionof‘discourse’isoftenpresupposed.Fromagrammaticalpointofviewthisisunderstandable,butifwewanttofindthecausesbehind‘discoursevariability,’weneedtoapproachdiscoursesystematically.

Inanattempttobuildafeasiblebridgebetweencontextandlinguisticstructure,specifyingthepossibilitiesandrestraintsthereareonwhatwe(can)say,sign,write,andsoon,when,where,how,andwhy,wehereofferalistofthekindsofexternalattributesthathaveprovenimportantinearlierstudiesinordertoadequatelyexplaintheconstraintsonavailablelinguisticresources.Thesearethusalsothekindsoffeaturesthatarerelevantforrestrainingdiscoursevariability(cf.alsoTraugott2010a).

Thefeaturesareherepresentedinaloosesenseasattributesinattribute-valuematrices,butthefeaturesthemselvesareindependentoftheformalism.Thecrucialaspectofthislist,whichisbeingconstantlydevelopedandrefined,isthataseeminglyinfinitenumberofcontextualcuescanbetieddowntoadozeninordertogetamoresystematicpictureofwhatisinvolved.Thus,onthebasisofresearchsofar,thissetoffeaturesisthesetaccordingtowhichexpressionsmayshowvariabilityandvariationduetodiscourseinfluence.(Asinallconstructionalresearch,theverticalorderofpresentationisarbitrary.Themainattributesaregiventotheleft,valuesofthesetotheright,andfurthervaluesfurthertotheright.)

ImplicitanchoringCoherence

Interaction

Involvement

DiscoursepatternD-Frame

Texttype

ActivitytypeGenre

SociolinguisticvariablesAge

Gender

StyleEducation

Lifestyle

Page 354: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 9 of 11

Role

SpeechcommunityPlace/Space

Size

Modernity

Nonverbal

Grounding

Aquickrun-throughoftheattributesabovewillnotdojusticetothecomplexityinvolvedbutmightgivethereadersomeideaofwhatthefactorsasattributesstandfor.Implicitanchoring(cf.FriedandÖstman2005,andreferencesthere)isthegeneralattributetocaterforthekindsoffeatureswealsofoundtoberelevantintheWellerismexamplein(4):culturalidentity(Coherence),responsiveness(Interaction;cf.alsoLinell2009;Wide2009),andaffect(Involvement).

UndertheattributeDiscoursepattern(cf.Östman1999,2005),whichisseenascognitivetext-levelcategorization,therearethreevalues,eachneeding(asvalues)furthervaluespecifications.D-Framestandsfor‘discourseframe’(cf.Fried2009b;NikiforidouandTorresCacoullos2010);cf.newsreport,lullaby,telephoneconversation.Genre,asonetypeofActivity,isusedinthesenseofSwales(1990,andlater)andBhatia(1993,andlater)todenotesocialactivitieslikeletterwriting,fairly-talereading;andTexttypefocusesonstructuralfeaturesandstartsfromthetraditionaldistinction(cf.Werlich1976)betweennarrative,instructive,descriptive,expository,andargumentativetexttypes,addingtoitvalueslikefragment.Agoodsuggestionofhowtofurthersubcategorize(i.e.,assignvaluestoattributes)isNir's(2008)categorizationofnarrativesintoeventive,interpretive,andillustrative.

TheattributesSpeechcommunityandSociolinguisticvariablesaretocaterforwhataretraditionallyreferredtoasdialectsandsociolects,respectively(cf.thediscussionin26.1.2above).TheattributesAgeandGenderarelargelycomparabletothetraditionalsociolinguisticvariables,whereasEducation,Lifestyle(cf.Bourdieu1979,Pedersen2010),andRole(bothintermsof,e.g.,occupationalrolesandintermsofwhatSarangi2010calls‘role-sets’)areseenasvaluesofStyle(inthesenseofEckertandRickford2002andAuer2009).ThevaluesofSpeechcommunity,Place/Space,Size,andModernity,suggestthatratherthantalkingaboutdialectsintermsofgeographyperse,characteristicshavingtodowithhowlargethecommunityis,themembers’senseofspace(cf.Cresswell2004),andatwhatperceivedstageof‘development’(traditional,modern,latemodern)itis,willbemoredecisiveasconstraintsandresourcesofusage.

TheattributeGroundingwillhavevaluesreferringbothtoissuesofforegroundandbackgroundandtoFigureandGround;Nonverbalwillspecifykinesicsintermsofgestures,position,gaze,andsoon,butalsorefertomultimodality.

Inadiscourseapproachtoconstructions,contextfeatureslikethesearenotoutsideofconstructions,butpartoftheconstructions.Togetherwiththeinternalfeatures,theyspecifyresourcesforlanguageusersinanordinaryconstructionalfashion.Whenlookedatinthismanner,‘contextualfeatures’thataffectvariabilityarenotseenasbeingoutsidegrammar,butasbeingpartofgrammar.

Averypertinentquestionishowfaronecanandshouldgoinsuchattemptsatsystematizingcontext.Ultimately,itsusefulnesswillnaturallybeseeninconcreteanddetailedanalysesoflinguisticexpressions.Eventhoughthefeaturesintheirpresentformmighttosomereadersatmostseemlikeacollectionofpotentiallyworthwhileaspectstodigdeeperinto,listingtheminthisfashionalsoattemptstorespondtoCharlesFillmore'srequirementthatConstructionGrammarbeconsistentwithwhatweknowaboutcognitionandsocialinteraction.Alltheattributessuggestedabovetouchonissuesofcognition(includingdiscoursecognition)andsocialinteraction(includingculturalinteraction).Inordertoapproachdiscoursevariabilitysystematically,asystematicaccountofthefeaturesthatconstrain(or,indeed,invite)variabilityneedstobethebasisfortheanalysis.

26.4.Conclusions

Page 355: The Oxford Handbook of Construction Grammar

Dialects, Discourse, and Construction Grammar

Page 10 of 11

Aswehaveseen,workonConstructionGrammar,dialects,anddiscourseispartofamoregeneralresurgenceofinterestintherelationshipbetweenthemodelingoflanguagestructureandthenatureoflinguisticvariation.Inthissense,theworkondialectanddiscoursevariationinConstructionGrammarjoinsagrowingbodyofworkinotherlinguisticframeworks,theaimofwhichistounderstandtheconstraintsonvariation,andthedistributionofdifferentgrammaticalstructuresindifferentvarietiesofthesamelanguage.Furthermore,workonmicro-variationindialectmorphosyntaxcomplementscrosslinguistictypologicalwork,anissueaddressedbySiewierskaandHollmann(2007).Theintersectionofconstructionalapproachestolanguageanddialectologicalresearchincorporatesotherdevelopmentstoo.Forinstance,asMukherjeeandGries(2009)show,collostructionalanalysisisahelpfultoolintestinghypothesesregardingquantifieddegreesofdistanceofonevarietyofalanguagefromanother.Thisregionaldialectologicalworkintersectswithsocialvariationtoo(cf.Hollmann,thisvolume),butalsowithresearchonsecondlanguageacquisition(see,forinstance,thediscussionofprepositionplacementinBritishandKenyanvarietiesofEnglishinHoffmann(2011)).Wehavealsoseenthatconstructionalapproachesformaviabletoolforapproachingemergentfeaturesinlanguage.

Sincelanguageisvariablebothintimeandspace,thesynchronicvariationandgradiencewhichweobserveinregionaldialectsisrelatedtothegradualchangewewitnessinprocessessuchasgrammaticalization(foraninsightfuldiscussionwhichbringstogetherdialectvariation,(de)grammaticalizationandconstructions,seeDeVogelaer(2008));dialectvariationmaythereforebeseenasaproductofthedifferenthistoricaltrajectoriesofconstructionalchangeinspeechcommunities.Nonetheless,weshouldalsorecognizetheplaceoftheidiolect,andthefactthatConstructionGrammarmayalsobeseenasatheoryofindividualknowledge.AsHoneybone(2011)observes,theproblemsfacedbyanylinguistictheorywhichtakesvariationseriouslyincludetheproblemofhowtobringtogetherinter-speakervariationatthelevelofdialect,andintra-speakervariationatthelevelofidiolect.Manychallengesremain,butitisclearfromtheresearchthathasalreadybeenpublishedthatConstructionGrammarsprovideaframeworkinwhichtoexplorethenatureofregionalvariationanddiscoursevariability,andtherelationofthesetoindividualknowledgeoflanguage.

Notes:

(1.)BNC(TheBritishNationalCorpus),http://www.natcorp.ox.ac.uk/,http://corpus.byu.edu/bnc/x.asp.

(2.)FRED(FreiburgEnglishDialectCorpus),http://www2.anglistik.uni-freiburg.de/institut/lskortmann/FRED/.

(3.)TheHelsinkiCorpusofEnglishTexts(nowpartofthePPCME/PPCEMEcorpora),http://khnt.aksis.uib.no/icame/manuals/HC/INDEX.HTM.

(4.)ICE(InternationalCorpusofEnglish),http://ice-corpora.net/ice/index.htm.

Jan-olaÖstmanJan-olaÖstmanisProfessorofScandinavianLanguagesintheDepartmentofFinnish,Finno-UgrianandScandinavianStudiesattheUniversityofHelsinki.HehasanM.A.inlinguisticsfromReadingUniversity(1976),andaPh.D.inlinguisticsfromUniversityofCalifornia,Berkeley(1986).Since1988hehasheldvariouspositionsattheUniversityofHelsinki(includingAssociateProfessorofEnglishLinguistics,ActingProfessorofGeneralLinguistics,andProfessorofEnglishPhilology).Since2006hehasalsoworkedasProfessorIIofScandinavianlanguagesattheUniversityofTromsø,Norway.HismainresearchinterestsaretheroleofdiscourseandgeneralpragmaticissuesinConstructionGrammar,aswellasvariabilityandcontact-inducedchange.

GraemeTrousdaleGraemeTrousdaleisaSeniorLecturerattheUniversityofEdinburgh.WithNikolasGisborne,heeditedConstructionalApproachestoEnglishGrammar(deGruyter,2008).HehaspublishedanumberofjournalarticlesandbookchaptersonconstructionalapproachestovariationandchangeinEnglish.HeistheauthorofAnIntroductiontoEnglishSociolinguistics(EUP,2010),andisco-editorofGradience,GradualnessandGrammaticalization(withE.C.Traugott,Benjamins2010),andTheOxfordHandbookofConstructionGrammar(withT.Hoffmann,OUP,2013).HislatestbookisConstructionalizationandConstructionalChanges(withE.C.Traugott,OUP,inpress)[email protected]

Page 356: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 1 of 13

PrintPublicationDate: Feb2013 Subject: Linguistics,MorphologyandSyntax,SociolinguisticsOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.013.0027

ConstructionsinCognitiveSociolinguisticsWillemB.HollmannTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

AbstractandKeywords

Thischapterexploressomeofthewaysinwhichcognitivelinguistshaveinterfacedwithsociolinguisticsinordertoenrichandrefinetheiranalyses.ItdescribescognitivelinguisticresearchonEnglishandDutchwheretheanalysisbenefitsfromtakingabroader,socialperspectivethanisusualforcognitivelinguists.Furthermore,itdiscussesworkonLancashiredialecttoillustratesomewaysinwhichacognitiveperspectivecanbeusedtoactuallyfeedbackintosociolinguistictheoryandpractice.Thechapteralsostressestheneedforgreaterdialoguebetweenconstructiongrammariansandsociolinguisticsinordertoachieveapsychologicallyandsociallyplausibleaccountoflinguisticvariation.

Keywords:cognitivelinguistics,ConstructionGrammar,sociolinguistics,linguisticvariation,frequency,schemas

27.1.Introduction

Inanimportantvolumeoncognitivesociolinguistics,theeditorsstarttheirintroductorychapterwiththefollowingquotefromLangacker:

Articulatingthedynamicnatureofconceptualandgrammaticalstructureleadsusinexorablytothedynamicsofdiscourseandsocialinteraction.WhilethesetoohavebeenpartofCognitiveGrammarfromtheveryonset,theyhavecertainlynotreceivedtheemphasistheydeserve.(1999:376;citedbyKristiansenandDirven2008:1)

KristiansenandDirvencharacterizethisquoteas“programmatic”(ibid.),bywhichtheymeanthatLangacker'sworkincognitivelinguistics—bothintheperiodleadinguptoandinthetimethathaselapsedsincehewrotethesewords—generallydoesnotdealwithlanguagevariationanditsrelationto(variouskindsof)socialstratification.

NotthatLangacker'sworkisuniqueinthisrespect.OtherpublicationsthatareconsideredfoundationalincognitivelinguisticsandConstructionGrammar(e.g.,Lakoff1987;Goldberg1995;Croft2001)donottakeasocialperspectiveeither.

TheworkofRichardHudson,architectofthecognitivelinguistictheoryofWordGrammar,issomethingofanexceptioninthisregard,inthatsocialmeaninghasalwayshadaplaceinhistheoryofgrammar(see,e.g.,1996,2007a,2007b).However,whetherHudson'sword-dependency-basedworkshouldbeseenasconstructionistincharacterisdebatable.Hudsonhimselfdoesnotexplicitlydoso(seealsoHolmesandHudson2005;Hudson2008),butonewonderswhetherthedifferencebetweendependencystructuresandgrammaticalconstructionsasdefinedintheConstructionGrammarliteratureismerelyadifferenceinemphasis(andformalism)ratherthaninkind.

Atanyrate,duringthepastfewyearscognitivesociolinguisticshasstartedtoemergeasaparadigm.Theverytitle

Page 357: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 2 of 13

ofKristiansenandDirven's(2008)volumesupportsthis,asdoes,forexample,theconfidentsuggestionbyHeylenetal.,that“CognitiveLinguisticsisideallysuitedtointegrateinternalandexternalvariablesintoacoherenttheoreticalframework:theframeworkofCognitiveSociolinguistics”(2008:123).

Cognitivesociolinguisticsmaybe,andindeedbyKristiansenandDirven(2008)is,broadlyconstrued,encompassingnotonlythestudy,fromasocialandcognitiveperspective,oflanguagestructure,includingphonology(e.g.,ClarkandTrousdale2009),lexis(e.g.,Geeraertsetal.1994;Speelmanetal.2008),andgrammar(e.g.,Gries2002,2003a;Grondelaers2000;Grondelaersetal.2007,2008;HollmannandSiewierska2006,2007,2011),butalsomoreappliedareassuchaslanguagepolicydebatesandotherpoliticalissues(e.g.,Berthele2008;Janicki2008),corporatebranding(e.g.,Koller2008),andissuestodowithlanguagelearningandlanguageteaching(e.g.,Fiksdal2008;Holme2009;Littlemore2009).

Muchofthisworkisnotconcernedwithgrammaticalconstructionsandsofallsoutsidetheremitofthischapter.Theexclusionofcognitivesociolinguisticworkonlexisperhapsdeservestobementionedexplicitly,inviewofthesyntax-lexiconcontinuum(e.g.,Langacker1987:25–27;thoughseePulvermülleretal.,thisvolume,forsomecriticalremarks)thatpractitionersofConstructionGrammargenerallysubscribeto.Inotherwords:sincewordsarealsoconstructions(albeitrathersimpleones)thischaptercouldhavecoveredthestudyoflexicalvariationaswell.ButtodosowouldhaverequiredconsiderablymorespaceandwouldhavedetractedfromthemoreusualfocusofConstructionGrammar,namely,morecomplexconstructions.Afurtherrestrictionimposedonthischapterwillbetheexclusionofworkinappliedareas,suchaslanguageteaching,whereconstructionsareactuallybeginningtobeconsidered(Holme2009:chapter8;Littlemore2009:chapter9;thecontributionsinBoersandLindstromberg2008onidiomsandfixedphrases;LindstrombergandBoers2008).Whatremains,‘cognitivesociolinguisticsnarrowlydefined,’isthestudyoflinguisticvariationfromthecombinedperspectiveofsocialandcognitiveconstraints.

Thischapterexploressomeofthewaysinwhichcognitivelinguistshaveinterfacedwithsociolinguisticsinordertoenrichandrefinetheiranalyses,andinsomecasesthetheory(ortheories)aswell.Inthecontextofthisinterfacesomecognitivelinguistshaveopenedtheireyestoconstraintsimposedbylanguage-externalvariables(e.g.,channel,register,region).Section27.3discussestworepresentativeexamplesofcognitivelinguisticresearchwheretheanalysisbenefitsfromtakingabroader,socialperspectivethanisusualforcognitivelinguists:theworkofGriesonparticleplacementinEnglishandthatofGrondelaersandcolleaguesoncertainpresentativeor“referencepoint”constructionsinBelgianandNetherlandicDutch.Section27.4illustratessomewaysinwhichacognitiveperspectivecanbeusedtoactuallyfeedbackintosociolinguistictheoryandpractice(andviceversaaswell),referringtoaseriesofstudiesbyHollmannandSiewierskaonLancashiredialect(2006,2007,2011).ButfirstIwillturntoanotherrecentdevelopmentthatisrelevanttoadiscussionof(constructionsin)cognitivesociolinguistics:Croft's(andsomeothers’)workinwhathecalls‘socialcognitivelinguistics’(2009b),anditsrelationtowhatis(comingtobe)knownascognitivesociolinguistics.

27.2.SocialCognitiveLinguisticsversusCognitiveSociolinguistics

Nextto‘cognitivesociolinguisticsnarrowlydefined’(seesection27.1,above),theemergingresearchparadigmconcernedwiththestudyofvariationfromasociocognitiveperspective,severalcognitivelinguistshaveofferedsuggestionstowardwhathasbeencalled‘asocialcognitivelinguistics.’WhilethetermitselfisduetoCroft(2009b),itmaymakesensetogroupunderthesamelabeltheworkonintersubjectivityandintersubjectificationby,forexample,Sinha(1999,2004),Verhagen(2005),andLópez-Couso(2010).OtherscholarswhohaveworkedonintersubjectificationprominentlyincludeTraugott(e.g.,2003b,2010b),butherresearchislessexplicitlycouchedincognitiveterms.

Croft'sargumentforasocialcognitivelinguisticsstartsfromthesuggestionthatcognitivelinguisticsoffers,oratleastsetsouttooffer,apsychologicallymorerealisticalternativetoformaltheoriesofgrammarandsemantics.Oneofthemainwaysinwhichitdoessoisbyrejectingthemodularityhypothesis,insteadstartingfromthenullhypothesisthatlanguageisanintegratedpartofhumancognition.InCroft'sownwords:“grammaticalstructuresandprocessesinthemindareinstancesofgeneralcognitiveabilities”(2009b:1,emphasisinoriginal).Croft,ofcourse,supportsthisbutgoesontonotethatthequestforcognitiverealismhashadanegativeside-effect,thatis,therelativelackofattentiontolanguageasasocialphenomenon.Asheputsit:

Page 358: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 3 of 13

[C]ognitivelinguisticsisindangerofconstruingitselftoonarrowlyasanapproachtolanguage,inthesamewaysthatformalsyntacticandsemantictheorieshavebeencriticizedastoonarrow.Thisisnottosaythatthefoundationsofcognitivelinguisticsareinvalid.Theydoofferamodeloflinguisticcognitionthathasgreaterpotentialthantheformalalternatives,inmyopinionatleast.Buttheyareincomplete.Inparticular,asmytitleimplies,theyaretoosolipsistic,thatis,toomuch‘insidethehead’.Inordertobesuccessful,cognitivelinguisticsmustgo‘outsidethehead’andincorporateasocial-interactionalperspectiveonthenatureoflanguage.Croft(2009b:1)

MostofCroft'ssubsequentdiscussionelaboratesonwaysinwhichcognitivelinguistsneedtotakeonboardworkinlinguisticandphilosophicalpragmatics,withspecialattentiontoClark's(1992,1996,1999)andBratman's(1992,1993,1997)researchonhumancommunicationasajointactivity(anapproachwhichissharedbySinha,Traugott,Verhagen,andsoon,intheirpublicationsonintersubjectivityandintersubjectification).Morespecifically,thisimpliesthatlanguagemustnotonlybestudiedasamentalphenomenonbutalsoasasocial-interactionalone.

Croftsubsequentlyshowshowvariousbasictenetsofcognitivelinguisticscanberevisedsoastoincludesocial-interactionalconsiderations.Forexample,thecrucialideathatlanguageisanintegratedpartofgeneralcognitionhasmeantthatcognitivelinguistshavebeenabletodrawonnotionsfromcognitivepsychology,especiallynotionsregardingperception(e.g.,ideasfromgestaltpsychologyconcerningthecognitiveorganizationofperceptualinput),categorization(e.g.,prototypes),andmemory(e.g.,networksandspreadingactivation).Thesenotionshavestrengthenedtheirexplanationsofaspectsoflinguisticstructureandmeaningconsiderably,notleastsincetheyhave‘independentgrounding’:theyhavenotbeendevisedspecificallytoaccountfortheexplanandabutwereestablishedindependentlyfromthose(apointnotmadebyCroft2009bbutnonethelessworthobserving).

Butinorderforcognitivelinguisticstoembracethesocial-interactionalfunctionoflanguage,Croftrephrasesthetenetcitedaboveas“grammaticalstructuresandprocessesinthemindareinstancesofgeneralsocialcognitiveabilities”(2009b:3).Itisactuallynotclearwhetherthisisthemostfortunateformulation,sincethepurelycognitiveabilitiestraditionallyusedtoanchorcognitivelinguisticresearchpresumablyshouldnotbereplacedbutcomplementedbysocialcognitiveones.Thus,itmaybemoreaccuratetoreferto“cognitiveandsocialcognitiveabilities.”ButthishardlydetractsfromCroft'spointthatourknowledgeofgeneralsocial-interactionalabilitiesshouldhelpshedlightonlinguisticbehavioraswell.

Themostimportantsocialabilitiesinrelationtolinguisticbehavior,Croftargues,arejointaction,coordination,andconvention(2009b:3).

Jointaction,inlanguage,remindsusofthefactthatwhenspeakerandhearerengageinconversation,theyarenotmerelyfocusedon(encodinganddecoding)themessagebutalsooneachother,asparticipantsinthecommunicativesituation,equippedwithknowledge,andharboringintentionswhichcanusuallybeinferred,albeitonlypartly(cf.theso-calledProblemofOtherMinds).

Inordertoachievejointaction,speakerandhearermustcoordinatetheiractionstooneanother.Oneofthemostimportantabilitieswehaveinthisregardisourcapacitytoestablishjointattention.Thisability,oratanyratethedegreetowhichwepossessit,setsus,humans,apartfromotheranimals,includingcloselyrelatedspeciessuchaschimpanzees(Tomasello1999:chapter3;citedbyCroft2009b:6).

Convention,finally,referstoourabilitytoshareasystemofsymbolsamongmembersofacommunity,includingaspeechcommunity.Withoutthisability,linguisticcommunicationwouldobviouslybefarlessefficient,assignswouldessentiallyneedtobemadeuponthespot.

ImportantthoughCroft'sdiscussionofClark's,Bratman's,andothers’researchis,hisargumentremains(inevitably)fairlyprogrammatic—thatis,untilthefinalsectionofhispaper.Here,Croftdiscussesacasestudy(basedontheanalysisthatwaseventuallypublishedas2010d,althoughinthatstudythefocusisdiachronicratherthansynchronic)inwhichtheimportanceofthissocial-interactionalperspectiveisillustratedinmoreconcreteterms.

Inthecasestudy,Croftisinterestedinthenatureofconstrual.Thetraditionalassumptionincognitivelinguistics(andactuallyfunctionallinguisticsmoregenerally)isthatadifferenceinlinguisticformcorrespondstoadifferenceinmeaning.Thus,iftwospeakersverbalizethesameobjectivesituationdifferentlyfromeachother,nomatterhowslightthatdifference,thentheassumptionisthatthisreflectsacontrastintheircommunicativeintentionand/or

Page 359: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 4 of 13

experienceofthesituation.Croftsetsouttoshowthatthisisactuallyfarfromcertain.HisdatabaseconsistsoftwentyEnglishnarrativestakenfromtheso-calledPearStoriesproject(Chafe1980).Forthisproject,asix-minutefilmnotcontaininganyspeechwasdeveloped.Itwasshowntospeakersofdifferentlanguages,whoweretheninstructedtoretellthestory.Theoriginalfocusoftheresearchersworkingontheprojectwasdiscourse-pragmaticandsociolinguistic,butCroftexploresthegrammarandsemanticsofcertainconstructionsinthenarratives.

Asanexample,heconsidershowdifferentspeakersverbalizeascenewherethepearpickerputssomepearsthathehaspickedintohisapron.Itturnsoutthateveninjustasinglescenethereisquiteabitofvariationacrossspeakers(1,11referstothe11thintonationunitofspeaker1,andsoon,cf.Croft2009b:20,andalsoCroft2010d):

1,11andhe’ddropthemintohis[.25]thing,

3,7[3.15um[2.35]um[.35]]picksthepearsandputsthemina[.45]inum[.4]thesebasketsthathehas.

5,10..filledwith[.8]andhe'sfillingthemwithpears.

6,8andputtingthemina..whiteapron,

8,12[.6[.1]A—nd]he[1.0]fillshis—..thingwithpears,

10,27andthenhe’ll..he’llstufftheminthat,

11,9[.95]putsthemin..hisapron,

12,21andheputtheminanapron/thathehad/,

13,9andputtingitinhis..apron,

15,11[.15]buthewasalsohealsohadanaprononandhewasfilling/those/up.

17,20[.45]puttingtheminhisapron,

18,14andputtingthemin[.7…{breath}…]hisapron.

Thetraditionalcognitive(functional)explanationwouldbethateachverbalizationcorrespondstoadifferentconstrualofthesamesituation(inotherwords,adifferentmeaning).Forexample,thestuffconstructionismorespecificthanput,andsospeaker10takesafiner-grainedperspectiveontheeventthanmostothers(cf.theconstrualoperationknownasqualitativescalaradjustment,CroftandCruse2004:53).ButCroftarguesthatinviewofthesocial-interactionalnatureoflanguagethisexplanationmaynotbewarranted.

ThefirstproblemCroftnotesisthatspeakersandhearersdonotsharethesameexperiencewiththeconstructionused(say,stuff),andpossiblealternatives(including,butnotlimitedto,putanddrop).Speaker10mayhavebeenexposedtoandhavepreviouslyusedthestuffconstructioninsituationssimilartothepresentone,increasingthelikelihoodofthisconstructionbeingactivatedasthemostsuitableoneinthiscontext.Butthehearer'sexposuretoanduseofthisconstructionmaywellhavebeendifferent,atleasttosomedegree.Therefore,Croftargues,“thehearercannotbecertainthattheconceptualizationhethinks[familiarwordsandconstructions]representistheconceptualizationthatthespeakerthinkssheisconveying”(2009b:21).Inotherwords,Crofthypothesizesthatinassigningmeaningstoverbalizations,hearerswillessentiallyleaveacertaindegreeofindeterminacy.Thiswouldimplythatthecorrelationbetweenverbalizationstrategies(constructionsselected)andconstrualsisnotasone-to-oneashasgenerallybeenassumedincognitive-functionallinguistics.

InevaluatingthisratherphilosophicalpartofCroft'sproposals,onerunsupagainstthequestionastohowthehypothesizeddegreeofindeterminacyleftbythehearercouldbeempiricallyverified.Mightnotanalternativehypothesisbethathearersdonotallowforthisindeterminacy,butinsteadinterpretaverbalizationexactlyinlinewiththeirownexperiencewithconstructions?Communicationwouldnotnecessarilybreakdownasaresult,becausetherewillnaturallystillbeahighdegreeofoverlapbetweenthehearer'sexposureanduseofagivenconstruction,andthespeaker's.Afterall,thatistheessenceofbelongingtothesamespeechcommunity,usingthesameconventionalsystemofsigns.Onthisalternativemodelofcommunicationwewouldthushaveconstant,

Page 360: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 5 of 13

butonlyslight,mismatchesbetweenthespeaker'sandthehearer'sconceptualizations.Communicationworksbecausethe‘bigpicture’isgenerallytransferredinarelativelylucidmanner.

Croft'ssecondargumentagainstthetraditionalviewofverbalizationandconceptualizationseemsmoreconvincing.Sincenotwosituationsareeveridentical,andsincethemeaningswestorearebasedonpastexperiences,thereisalwaysgoingtobeamismatch,howeverslight,betweenthestoredmeaningofsomeconstructionselectedbyaspeaker,andtheirpresentexperience.InCroft'sownwords:“anychoiceofwordsandconstructionswillnotpreciselycharacterizetheconstrualofthescenebeingcommunicatedanyway….Thus,alternativeconstrualsprovidedbyalternativeverbalizationscannotbeprecise”(2009b:22).Thissuggestion,itwouldseem,doesnotrequireindependentverification.Andtheindeterminacymustbeaggravatedbythefactthatthehearer'sexperiences(andthereforestoredmeanings)areinevitablydifferentfromthespeaker's.Therefore,intryingtocoordinateontheconceptualizationintendedbythespeakerthehearerisnotevenabletorelyonthesamebackgroundofexperiencesandverbalizations.

Whatthismeansinconcretetermsforacomparisonofthedifferentconstructionsusedinthe‘putting’scenereferredtoabove,isthatstuffconstructionsdonotnecessarilycorrespondtodifferentconstrualsofthesituationthanputconstructions,andlikewisefordropconstructions.Weseehere,then,thatasocial-interactionalperspectivecarriesimportantimplicationsforcognitivelinguistictheory,inthiscasethestandardaccountoflinguisticconstrual.

Buttheimplicationsdonotendthere.Croft(2009b,andespecially2010d)suggeststhatthereareimportantconsequencesforourunderstandingofgrammaticalchangeaswell,specificallyforgrammaticalizationtheory(e.g.,HopperandTraugott2003).Theriseofnewgrammaticalconstructionshastraditionallybeenassumedtoresultfromsomespecialmechanismsuchasspeakerstryingtoavoidbeingmisunderstood(e.g.,Keller1994:109;Croft2000:159–60;citedbyCroft2010d:2)ortryingtobeespeciallyexpressive(e.g.,Lehmann1985:314–17;Heineetal.1991:78;citedbyCroft2010d:2).However,Croftarguesthatthesenewvariantsemergeinsteadinordinaryeverydayspeech,astheunintendedresultoftheinherentindeterminacyinthemeaningofconstructionsdiscussedabove.Thisrenderstheriseofnewgrammaticalvariantsverysimilartothatofnewphoneticvariants,whicharealsoproducedsimplyasamatterofcourseinordinaryspeechproduction(Ohala1989:176;citedbyCroft2010d:2).

SupportforCroft'sargumentsinrelationtogrammaticalizationcomesfromtherangeofvariationdisplayedbyspeakersverbalizingaspectsofscenesbyusingconstructionswhosegrammaticalizationpathwaysarewellunderstoodanddocumented(see,e.g.,HeineandKuteva2002).ThereisawealthofexamplesinCroft(2010d),butforpresentpurposeswecansimplysticktoput.Manyofthealternative,morespecificverbsused(e.g.,stuff)arewell-knownsourcesforthemoregrammaticalizedlightverbput.

Tosumup,byopeningupcognitivelinguisticssoastoincludeasocial-interactionalperspectiveCrofthasshednewlightontheimportantissueofourconstrualofwordsandconstructions,andhasalsocontributedsignificantlytoourunderstandingoftheearlystagesofgrammaticalization(i.e.,theemergenceofnewvariants).

TheperspectivetakeninCroft(2009b,2010d)isrelevantinthecontextofthischapter,inthatthesestudiesarepioneeringexamplesofthecognitivelinguisticstudyofconstructionswherelanguageisnotseenasapurelymentalphenomenon,butasasocial-interactionalactivityinstead.Yetthescopeofwhatisbecomingestablishedascognitivesociolinguisticsisevenbroaderthanthat,takingintoaccountnotonlythecommunicativesituationofspeakerandhearerbutalsothewidersocialcontext.Thiswidercontextincludesvariablestraditionallystudiedinsociolinguisticssuchasregister,region,andsoon,aswellasthenotionofprestige(see,e.g.,Labov1972a,1972b,1994,2001;Trudgill1974).

WhereasCroft(2009b,2010d)focusesonthefirststageofgrammaticalization,arguingthatthisshouldbeexplainedinsocial-interactionalterms,inearlierwork(especially2000)heactuallyemphasizestheimportanceofsocialfactorsinlanguagechange(includinggrammaticalization).However,thisbroadersocialperspectiveappliestoadifferentaspectofchange,thatis,nottheemergenceofnewvariantsbuttheirpropagationacrossthespeechcommunity,withCroftarguingthatthisspreadshouldbeexplainedwithreferencetothesociolinguisticnotionofprestige.

Thenexttwosectionsdiscusssomefurthermajorcontributionsincognitivesociolinguistics.

Page 361: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 6 of 13

27.3.ConstructionalVariationinCognitivePerspective

27.3.1GriesonEnglishParticlePlacement

GrieshasconductedaseriesofstudiesonthevariationintheplacementofparticlesinEnglishtransitivephrasalverbconstructions.Thevariationmaybeillustratedinexamples(1–2)(fromGries1999:109):

(1)Fredpickedupthebook.(2)Fredpickedthebookup.

Initially,Gries's(1999)accountofthisvariationwaspurelyincognitive(-functional)terms,butmorerecently(2001,2002,2003a)theextralinguisticvariableofspokenvs.writtenlanguagewasalsoincluded.(Theserecentstudiesalsosawasmallbutsignificantrevisiontothecognitiveaccountproposedin1999,see,e.g.,2001:39fordetails.)

ThecognitivelinguisticdimensionofGries'sexplanationstartsfromacarefuloverviewofintralinguisticfactorsproposedintheliteratureontheconstructionsinquestion,forexample,thenewsworthiness/accessibilityofthedirectobject,itslengthandcomplexity,andwhetheritsreferentisconcreteorabstract(forthefulllistoffactors,see2001:35).Mostofthesevariablescanberelatedtoprocessingcost,whichleadsGriestoputforwardtheProcessingHypothesis(e.g.,1999:131,2001:37–39,2003a:48–66).Thebasicideaisthatspeakerstakeintoaccountthecommunicativeneedsofhearers(Siewierska1988;citedbyGries1999:133)andtendtominimizetheirprocessingcost.Inrelationtotheorderexemplifiedby(1–2),above,thisessentiallymeansthattheverbandparticlewilltendtobekepttogether(asin(1))ifthedirectobjectcarriesaheavyprocessingcost(e.g.,whenitisabrandnewdiscourseparticipant),withspeakersinclinedtoseparatethem(asin(2))ifthedirectobjectiscomparativelyeasytoprocess(forexample,whenitisanalreadyactiveparticipant).Thereasonforthehypothesistobeformulatedinthisparticulardirectionisthatintheorderinexample(2)theearlyavailabilityoftheobjecthelpshearersbuilduparepresentationoftheeventportrayedbytheutterancemorequickly.In(1),bycontrast,thespeakerrecognizes(subconsciously)thatthehearerwillneedahigheramountofconsciousnesstodecodetheobject,whichisthereforebestleftforthefinalposition,whichinEnglishisassociatedwithahighdegreeofnewsworthiness(seealsoSiewierska1988:84–85;citedbyGries1999:132;seefurtherCappelle2009formoredetails,aswellasinterestingdiscussionofdifferencesbetweenBritishandAmericanusages).

Toillustratehowlengthandcomplexitymayplayarole,comparethecontrastinacceptabilitybetween(3)and(4)(takenfromGries1999:110):

(3)Hebroughtbackthebooksthathehadleftathomeforsolong.(4)??Hebroughtthebooksthathehadleftathomeforsolongback.

Inlinewiththeexplanationconcerningnewsworthiness/accessibilitysummarizedabove,Griesaccountsforthisdifferenceinacceptabilitybypointingoutthatnewinformationoftenneedstobeintroducedinamoreelaboratefashionthanoldinformation(seealsoQuirketal.1985:1361;citedbyGries1999:120).

Griescollects403examplesofverb-particleconstructionsfromtheBritishNationalCorpus,selectedongroundsofthehighfrequencyofthephrasalverbsinquestion(see,e.g.,2001:40).Hethenfirstanalysesthecontributionofeachofhisvariablesseparately.TheresultsclearlysuggestthatthefactorsassociatedwiththeProcessingHypothesisareverysignificant.Fromtheperspectiveofthischapteritisimportanttonotethatthespokenvs.writtendimensionalsoplaysarole,withthepatternexemplifiedby(1)and(3)beingfavoredinthewrittenlanguage,andtheconfigurationillustratedby(2)andthemarginalexample(4)beingmorecloselyassociatedwithspeech.Table27.1presentstherawdataGriesobtainedforthisvariable,whereVerb-Particle-Objrepresentsthepatternof(1)and(3),withVerb-Obj-Particlecorrespondingto(2)and(4).

Subsequently,notingthatthevariablesidentifiedgenerallydonotoccurinisolationbuttogether,andinordertofindoutwhetherandtowhatextenttheymaybeconspiringwithoneanotherGriesalsosubjectsthedatatomultivariateanalysis.HehimselfcompareshisapproachtoBiber's(1988)workongrammaticalvariationacrosstexttypes.ThestatisticaldetailsofGries'sanalysisarenotsoimportantinthepresentcontext,butitisinterestingtonotethatthemultivariatetechniquesheemploysarealsoverysimilartoanalyticalmethodsoftenusedinsociolinguisticworkintheso-calledLanguageVariationandChange(LVC)tradition(cf.,e.g.,Tagliamonteetal.1997;CheshireandFox2009).ForGoldvarbX,thesoftwarefrequentlyemployedbyresearchersintheLVC

Page 362: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 7 of 13

paradigm,seeSankoffetal.(2005);foratextbookintroduction,seeTagliamonte(2006:chapter10).

27.3.2Grondelaers(andColleagues)onDutcher-presentatives

GrondelaersandcolleagueshavepublishedextensivelyonBelgianandNetherlandicDutcher-presentativeswithsentence-initialadjuncts(see,e.g.,Grondelaers2000;Grondelaersetal.2007,2008).Thevariationunderconsiderationinthiswork

Table27.1.ConstructionalvariationinEnglishparticleplacement(afterGries2001:40)

Verb-Particle-Obj Verb-Obj-Particle Rowtotals

Spoken 67 133 200

Written 127 76 203

Columntotals 194 209 403

concernsthepresence

orabsenceofer‘there’inbetweentheverbandthesubject.Thetwodifferentpatternsareillustratedinexamples(5–8)(alltakenfromGrondelaersetal.2008:158,whochoosenottoglosseras‘there’):

(5)In1977wasereenfusietussenMaterneenConfilux.in1977wasthereamergerbetweenMaterneandConfilux‘In1977therefollowedamergerbetweenMaterneandConfilux.’(6)Morgenvolgteenextraministerraad.tomorrowfollowsanadditionalcabinetmeeting‘Tomorrowthereisanextracabinetmeeting.’(7)Inonslandisernogaltijdgeenopenbaargolfterrein.inourcountryistherestillnopublicgolfcourse‘Inourcountrytherestillisnopublicgolfcourse.’(8)Inhetredactielokaalstaanenkeleflessenwijninthenewsroomstandsomebottlesofwineenwatborrelhapjes.andsomeappetizers‘Inthenewsroomtherearesomebottlesofwineandsomeappetizers.’

Theinitialadjunctsinexamples(5–6)aretemporal,whiletheonesin(7–8)arelocative.Thisisactuallyoneofthevariablesidentifiedinthetraditionalliteratureontheseconstructions:erissupposedtobeattractedmorebytemporaladjunct-initialsentencesthanbysentencesbeginningwithalocative(see,e.g.,thestandardreferencegrammarofDutchbyHaeserynetal.1997:477).Yetasexamples(5)and(7)show,thisrestrictionmerelyconstitutesatendency.Infact,Grondelaersetal.(2008:159)reportthatmorethanaquarterofthelocative-initialsentencesintheircorpusdofeatureer.

Grondelaersandhisco-workersrefertoanotherlinguisticconstraint,thatis,thespecificityoftheverb:“themorespecifictheverb,thelesser”(2008:159).Butinterestingly,fromtheperspectiveofthischapter,somefurtherknownrestrictionsareexternaltothegrammarperse:erisgenerallymorefrequentinthesesentencesinBelgianthaninNetherlandicDutch(cf.DeRooij1991;Haeserynetal.1997)anderisfoundmoreininformalthanformalregisters(Haeserynetal.1997).Again,thesearetendenciesratherthanabsoluteconstraints.

TheessenceofthecognitivelinguisticexplanationthatGrondelaersetal.offerforsomeaspectsofthevariationisthatersignalstothehearerthatthefollowingsubjectisnotgoingtobeeasilyaccessible.Inmoretraditional,discourse-functionalterms,postverbalerisan“inaccessibilitymarker”(Grondelaersetal.2008:170).TheauthorscouchtheiraccountinworkbyLangacker(1993)andTaylor(1996)onso-calledreferencepointconstructions.Thefunctionoftheseconstructionsistodirectthehearer'sattentioninthedirectionofsomeentitieswhichtheremainderoftheutterancewillreferto,thusfacilitatingtheprocessingofthoseentities.Grondelaersetal.(2008:168)usethefollowingexamplestoillustratethepoint:

Page 363: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 8 of 13

(9)Intherefrigeratorwasasausage.(10)Intherefrigeratorwasmilk.(11)?Intherefrigeratorwasbread.(12)??Intherefrigeratorwasatoothbrush.

Theideaisthatthelocativeadjunctin(9–10)functionsasagood“mentaladdress”(Grondelaersetal.2008:167)forthesubsequentactivationofthesubjects.Thisislessobviouslythecasein(11),asbreaddoesnotneedtobekeptintherefrigerator,whiletheadjunctin(12)clearlyoffersaverypoorreferencepointforactivationoftheconceptthatcorrespondstothesubjectatoothbrush.

InearlierworkGrondelaersandBrysbaert(1996)actuallyfoundpsycholinguisticevidencefortheprocessingadvantagesofferedbygoodreferencepointconstructions.Theyconductedaself-pacedreadingexperimentinwhichtheycomparedtheprocessingtime(ofthesubject)inthecontextoftwodifferenttypesofsentence-initialadjunctreferencepoints:veryspecific,concretelocativessuchasinhetkoffertje‘inthesmallbriefcase’asagainstvagueorabstractspacessuchasinhettoneelstuk‘intheplay’(seealsoGrondelaersetal.2008:172).Theresultssuggestedthattheeffectofthereferencepointisindeedsignificant(p<.05)inrelationtoprocessingspeed.

Withrespecttoer,then,giventheexistentialorappearance(‘thereis…’/‘therefollows…’)natureofmanyofthesesentences,locativeadjunctsaregenerallymuchbetterreferencepointsthantemporalones.This,Grondelaersetal.argue,explainsthetrendwefindinthedatatowarduseoferintemporaladjunctsentencesbutlesssoinlocative-initialones.Verbalspecificitycanalsobeconsideredinlightofreferencepointpotential,withmorespecificverbssuchasLevin's(1993)verbsofentity-specificmodesofbeing(e.g.,vloeien‘flow,’branden‘burn’)restrictingtherangeofpotentialsubjectsmuchmorethan,forexample,themaximallyvagueverbofexistencezijn‘be’(seefurtherGrondelaersetal.2008:163fordiscussion).

Grondelaersetal.'slogisticregressionanalysisofthecontributionofthevariablesadjuncttype(i.e.,locativevs.temporal)andverbspecificitysuggeststhatbothareextremelyimportant(forfurtherdetails,seeGrondelaersetal.2008:164).

Butasnotedabove,thereisalsoregionalandregistervariation,whichcannotbeaccountedforinthesecognitiveterms.Afterall,thelanguageuser'scognitivemachinerydoesnotvaryacrossregionsorregisters.

Intermsofregister,Grondelaersetal.distinguishbetweenthreelevels:examplestakenfromUseNet,popularnewspapers,andqualitynewspapers.AlloftheseincludeBelgianandNetherlandicdata.Wediscusstheregionaldimensioninmoredetailbelow,butitmustbenotedherethatalthoughtheeffectofregisterturnsouttobesignificant,thisonlyappliestoBelgianDutch:inNetherlandicDutchitisnotadeterminantfactor(Grondelaersetal.2008:183).Inaddition,Grondelaersetal.findthatdespiteitsrelevance,thecontributionofregisterisneverthelessconsiderablyweakerthanthatofthelinguisticvariablesadjuncttypeandverbalspecificity.Butatanyrate,inadditiontoplayingthediscourse-functionalroleofinaccessibilitymarkerandalsobeingregionallyconstrained(seebelow),inBelgianDutchermayalsobeseenasan“informalitymarker”(Heylenetal.2008:115).

Theregionvariablewasalsofoundtoberelevant,infactslightlymoresothanregister(Grondelaersetal.2008:164).IalreadyobservedabovethaterisgenerallymorefrequentintheBelgianthantheNetherlandicvarietyofDutch.Butthesituationisactuallymorecomplicatedthanthat.WhereasinNetherlandicDutchadjuncttypeandverbalspecificityaccountverywellforthevariation,thepictureforBelgianDutchisconsiderablymoremessy,withnotonlyregisterplayingarolebutalsothenondiscretevariablesofadjunctconcretenessandadjuncttopicalitycomingintoplay.Thecumulativepredictivepowerofthesefactorsishigh,buttheydonotallowforastraightforwarddescriptionoftheconstraintsonthedistributionoferinthisvarietyatall.

TheexplanationGrondelaersetal.offerforthefarhigherdegreeofpredictabilityinNetherlandiccomparedtoBelgianDutchissociohistoricalinnature.Theauthorssummarizethemainideaasfollows:

Whereas—inanutshell—NetherlandicDutchbenefittedfromanormalstandardizationwhichwasfinishedatthebeginningofthe20thC,thestandardizationofBelgianDutchwasblockedinthe16thCasaresultofpoliticalandsocialfactors,resumingitscourseonlyinthe20thC.(Grondelaersetal.2008:186)

Page 364: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 9 of 13

InNetherlandicDutch,then,postverbalerhasspecializedtoadegreethathasnotbeenreachedinBelgianDutchyet,hencetheobserveddifferencesinpredictability.Theauthorsgoontoobservethatthisdovetailswithfindingsregardingotherareasofthelanguage,whereNetherlandicDutchalsoshowsagreaterdegreeofstability(cf.Geeraertsetal.1999;Grondelaersetal.2001;citedbyGrondelaersetal.2008:186).

TheresearchbyGrondelaersandcolleagueshasshedmuchlightonaphenomenonforwhichthestandardreferencegrammarandotherstudiesremainedrelativelyvague.Toalargeextentthishasbeenachievedbymovingbeyondfactorsinternaltothelinguisticandcognitivesystems,incorporatingvariablesrelatedtostyleandregionaswell.Thisisasignificantachievement.However,thereisalongtraditioninsociolinguistics/LVC,goingbackatleastasfarasLabov'sseminal(1963)studyonMartha'sVineyard,whichsuggeststhatregionaldistributionissometimesalsorelatedtospeakerswishingtomarktheiridentity(i.e.,asmembersofacertainsocialgrouporcommunity).Inthisrelation,onemightperhapshaveexpectedGrondelaersandhisco-workerstodelveintothepossibilitythatpostverbalersomehowfunctionsasaregionalidentitymarker.Thisremainsaninterestingpossibleavenueforfutureresearch.

27.4.CognitiveLinguisticContributionstoSociolinguisticTheory(andviceversa)

HollmannandSiewierska's(2006)studyoftwogrammaticalphenomenainLancashiredialectfurnishesthefirstexampleofhowcognitivelinguisticworkmaynotonlydrawontheinsightsandmethodsfromsociolinguisticsbutcanalsofeedintoitstheoreticalmachinery.ThephenomenainquestionarepasttenseBE,exemplifiedby(13–15),below,andreductionofthearticleinthedefiniteNPconstruction,asin(16–18).AllexamplesarefromaLancashiredialectcorpusthatisbasedoninterviewswithspeakersfromtheregion,heldattheNorthWestSoundArchive(fordetails,see,e.g.,SiewierskaandHollmann2007).Theinitialsrepresentthespeaker,andtheslashsymbolsin(16–18)separatethestandardspellingfromthepronunciation:

(13)…I’dhavehaveaworkmanthenthatcouldn’tunderstandwhatyouwassayingbutthisotherPakistanihadlearnedhim(FB)(14)Iftheshipswascominginheusedtobewatchingthemcomeinandwatching’emgoingout.(TC)(15)NowBillywereaparttimerat-catcher.(JA)(16)Ohyesyestheywereaprimaryschool(.)MissRileyshewereer(.)erinthe/[θ]infantsyouseeandthenyouwentupintothe/[?]bigschool(ED)(17)gothroughTownleyPark(.)andMrMcKaywerethe/[t]erparkkeeperthen(ED)(18)Noitwereni—itwerenicebecausetheyhadthembigpipes(.)’coswehadthembigpipesinthe/[?]greenhousesupthe/∅smallholdingsyouknowthembig(ED)

Examples(13–16)showthatinLancashirebothwasandweremaybeusednonstandardly.Thisnonstandarduseisprobablymorecommoninthisdialectforwerethanforwas(seeHollmannandSiewierska2006:25–26).Definitearticlereduction—oftenabbreviatedasDARinthesociolinguisticliterature(e.g.,Jones1999,2002;TagliamonteandRoeder2009)—mayresultinavowel-lessform,forexample,[θ],[t],[?],asin(16–18).Butthearticlemayalsobeomittedaltogether,markedas∅in(18).YetDARisanoptionalprocess—speakershavethefullformattheirdisposaltoo(seealsoJones1999:103–4).

ThespecificquestionHollmannandSiewierska(2006)addressconcerningthesetwogrammaticalvariablesiswhetherthedistributionofthevariantsissubjecttowhatinthesociolinguisticliteratureisknownas‘accommodation.’Theirbriefdiscussionofthisphenomenonisworthciting:

Thestandardassumptionintheliteratureisthatspeakersaccommodatetotheirinterlocutors(seee.g.Bell1984;Gilesetal.1987;StreetandGiles1982;Thakerar,GilesandCheshire1982).Accommodationisacomplexphenomenon,encompassingnotonlyaccent/dialectbutalso,forinstance,speedandloudness.Intermsofaccent/dialect,however,PaulKerswill(p.c.)arguesthattherealitymaybemorecomplicatedthanthetraditionalassumption.Hepointsoutthatitmaybemoreaccuratetosaythatspeakersaccommodatetotheirperceptionoftheinterlocutor'svariety.SinceinourNWSAdatatheinterviewer'svarietyisusuallyclosertostandardEnglishthanthatoftheinterviewee,andbecauseoftheinherentlyformalcharacterofaninterviewsituation,theintervieweesmaybeadjustingtheirspeechtostandardEnglishortotheirperceptionofthestandard.Forourpurposesitisnotnecessarytotakeapositiononwhatexactlyisthetargetoftheaccommodation—thecrucialobservationisthatthegeneraldirectionis

1

Page 365: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 10 of 13

awayfromcasualspeech.(HollmannandSiewierska2006:26n.9)

Thequestionregardingaccommodationinthedataisrelatedtoanimportantnotioninsociolinguistictheory,namelythatof‘sociolinguisticsalience.’Sociolinguisticsaliencemaybedefinedasthedegreetowhichspeakersareconsciouslyawareofacertainfeature,andtheextenttowhichtheymay(therefore)adjusttheirlinguisticbehaviorwithrespecttothisfeature.Accordingtotheclassicdefinitionofsalience,asproposedbyLabov(e.g.,1972a,1994,2001),linguisticvariantscanbeplacedonthescaleindicator>marker>stereotype,withmarkerandstereotyperepresentinganincreaseinsaliencerelativetothecategorytotheirleft.Indicatorsrepresentdifferentvariantsthatbehaveidenticallyacrossstyles.Markersdodisplaystylisticstratification,suggestingagreaterdegreeofspeakerawareness.Stereotypes,finally,attractovertcommentsbyspeakers,whoalsocorrectorevenhypercorrectthem.

Importantthoughthisnotionofsalienceisinsociolinguisticresearch,itsdeterminantfactorshavenotbeenwelldefined.KerswillandWilliams(2002)provideausefuloverviewofthefactorsproposedintheliterature.Theycanberoughlydividedintosocialfactors,ontheonehand,andcognitive-perceptualones,ontheother,forexample,tokenfrequency(cf.Bardovi-Harlig1987)andprosodicsalience(seeYaeger-Dror1993).

BasedontheresearchongrammaticalizationandlanguagechangeingeneralbyCroft(e.g.,2000),HollmannandSiewierska(2006)suggestthatsocialfactorsmaycertainlycontributetosalienceatsomepointinthelifespanofavariant,butthatwhennewvariantsemerge,thefactorsdeterminingtheirdegreeofsaliencemustbecognitive-perceptualinnature.Thisisbecausenewvariants,bydefinition,havenotbeenaroundlongenoughtohaveacquiredanysocialvalue.RecallinthisconnectionCroft'stwo-stagemodeloflanguagechange(discussedinsection27.2,above),accordingtowhichsocialfactors(specifically,prestige)determineonlythesecondstageofagivenchange(i.e.,whetherornotavariantwillspreadthroughthespeechcommunity).

Thus,acognitivelinguisticperspectiveonlanguagevariationandchangemayimmediatelycontributetoourunderstandingofsociolinguisticsalience.Butitdoesnotendthere.ThesociolinguisticstudiescitedbyKerswillandWilliams(2002)alwaysfocusonjustasingledeterminantfactor(e.g.,tokenfrequencyorprosodicsalience).Yetinrealitythesefactorswillco-occur:agivenlinguisticvariantwilloccurwithacertainfrequency,willhavesomedegreeofprosodicsalience,andsoon.Itwouldthereforebeusefultobeabletodeterminetherelativeimportanceofthevariouscognitive-perceptualdeterminantsofsalience.

Inafirstattempttoachievethis,HollmannandSiewierskahypothesizethatstudyingthedegreetowhichtheirspeakersdisplayaccommodationintheirproductionofthelinguisticvariablesinquestionmayshedlightontheirdegreeofsalience:ifanonstandardvariantisverysalientoneexpectstoseeclearevidenceofaccommodation,whereasnonsalientnonstandardvariantsshouldnotbesubjecttoaccommodation.Ifadifferenceinaccommodation,andthereforesalience,isfoundbetweenthenonstandardvariantsinquestion,thevariantscanbeanalyzedintermsoftheircognitive-perceptualproperties.

Inordertooperationalizeaccommodation,HollmannandSiewierskamarkthepositionofeachtokenofpasttenseBEandthedefinitearticleintheinterviews.Thesuggestionisthataccommodationismorelikelytooccuratthebeginningofaninterview,whenthespeakermaybeassumedtostillbeconsciousoftherelativeformalityofthesituation.Forthatreason,aclusteringofstandardvariantsatthebeginningisinterpretedasevidenceofaccommodation,whereasamoreevendistributionacrosstheinterviewindicatesthatthespeakerisnotsufficientlyawareofthenonstandardnatureoftheregionalvariantstoadjusttheirlinguisticbehavior.

Theresultssuggestthatthedefinitearticleisnotsubjecttoanyaccommodationatall.Bycontrast,forwerethereissomeevidenceforaccommodation,atleastinoneofthetwospeakersinvestigated.

Movingontotherelativeimportanceofthedeterminantsofsalience,HollmannandSiewierskaconsiderthatinthespokenpartoftheBritishNationalCorpusthedefinitearticleoccursmorethanthreetimesasoftenaspasttenseBE.Thiswouldseemtoimplythatwhiletokenfrequencymayperhapscontributetosalience,itcannotbeenough.Assessingtheprosodicsalienceofthesetwovariablesinaveryprecisewayislessthanstraightforward.However,theonlyreceivesemphaticstressinspecialcircumstances(e.g.,wherethereissomekindofcontrastwithanearliernounphrase).Furthermore,itcannotappearattheendofatoneunit,norformoneonitsown.Forthisreason,itisclearthatingeneraltermspasttenseBEoutrankstheonprosodicsalience.Ifthiseffectwerefoundtoberobustacrossmoredata,preferablysubjectedtomultivariateanalysis,onewouldhavegroundstoconclude

Page 366: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 11 of 13

thatprosodicsalienceoutweighsfrequencyasadeterminantfactorofsalience.

TwofurtherstudiesonLancashiredialectbyHollmannandSiewierska(2007,2011)providefurtherillustrationofthefeedbackloopsexistingbetweencognitivelinguisticsandsociolinguistics.Bothdealwithreductionphenomena:whilethe(2007)paperinvestigatesreductionofmyin1Sgpossessive-nounconstructions,their(2011)studyfocusesonceagainonDAR.ForexamplesofDAR,see(16–18),above.Variationintherealizationofmyisillustratedby(19–22):

(19)…Icouldn’tplayforthembecausetheycouldn’taffordmy/[mрι]footballshoes(JA)(20)Iwassoyoungthenlikeanderme/[mi]brothertooktheopportunityandhewent.(HF)(21)whenIwasfourIusedtogoroundthishousewithmy/[ma]eyesclosed(RG)(22)Iremembermy/[mə]fathercomingoutasmallroom(CS)

Inordertoaccountforthevariationinthe1Sgpossessives,HollmannandSiewierskaconsiderseveralexplanationsthatfitwellwithcognitive,functional,andtypologicallinguistictheory,butwhichonedoesnotgenerallycomeacrossinsociolinguistics.Thefirstexplanationisintermsofthetypologicalnotionofalienability(e.g.,Nichols1988,1992),andthewayinwhichthisrelatestothecognitive-functionalnotionoficonicity(e.g.,Haiman1985b).Thealienabilityhierarchy,givenbelow,isacrosslinguisticgeneralizationthatcapturesthesemanticfactorswhichinmanylanguagesmotivatethestructuralproximitybetweenpossessorandpossessednouns.Theleft-handsideofthehierarchyrepresentspossessednounswhichareconceptuallyclosertothepossessor,andarealsooftencodedassomehowstructurallycloser.Thestandardtypologicalexplanationforthisobservationisintermsoficonicity(butseeHaspelmath2006,whichIreturntobelow).

bodypartsand/orkinshipterms>part-whole>

spatialrelations>culturallybasicpossesseditems>other

AlienabilityeffectshaveneverbeendemonstratedforEnglish,butHollmannandSiewierska(2007)showthatinLancashiredialectthehierarchydoesarathergoodjobofpredictingwhichpossessednounswilloccurwithreduced1Sgpossessives,withkinandbodypartnounsfeaturingreducedpossessivessignificantlymoreoftenthanothernouns.(SeeinthisrelationalsoCappelle2009:169,whoseanalysissuggeststhatalienabilitymayalsoplayaroleinparticleplacementinEnglish.)

Haspelmath(2006)hasarguedthatwhatunderliesalienabilityeffectsisactuallyfrequency.Heshowsthatinhiscrosslinguisticdatafrequencymakesbetterpredictionsthanalienability.Thereasonwhyalienabilityseemsareasonablygoodpredictor,then,isthatbodypartandkinshipnounshappentobemorefrequentthantheothersemanticcategories.

Thesuggestionthatfrequencydeterminesreductionhasalonghistory.OftenassociatedwithZipf(1935),itactuallygoesbackasfarasSchuchardt(1885).Itisalsoverywidelyacceptedamongcognitivelinguistsandisoftenusedbythoseworkingwithintheusage-basedmodel(e.g.,BybeeandScheibman1999;Berkenfield2001).HollmannandSiewierska(2007)discussbothabsolutefrequency(usedbymostlinguists)andrelativefrequency(advocatedbyHaspelmath2006).TheyfindthatabsolutefrequencymakesgoodpredictionsconcerningreductionintheLancashiredata,butthatrelativefrequencyissuperior.Thehighdegreeof1Sgpossessivereductioninconstructionswithanumberofrelativelyinfrequentkinshipterms(e.g.,stepfather,niece,child)remaindifficulttoaccountfor,however,althoughrelativefrequencystilldoesabetterjobthanabsolutefrequency.

Drawingonthenotionofconstructionalschemas,HollmannandSiewierska(2007)proposeanalternativeexplanationfortheseproblemcases.Theyarguethatinmorefrequent1Sgpossessive—kinshipnounconstructions(mybrother,myfather,mymother,etc.)thepossessivemayhavebeenreducedfirst.Thiswillhaveledtoaschemaof[myKIN]withareducedformofmy.Constructionssuchasmychildarequiteinfrequentbutbecauseoftheirsemanticsimilaritymayhavebeencategorizedbythesameschemanonetheless.Asaresulttheyappearwithreduced1Sgpossessivesmoreoftenthanonewouldexpectonfrequencygrounds.

HollmannandSiewierska's(2011)studyonDARtosomeextentechoesthevalue,forthestudyofreductioninnonstandarddata,ofthenotionsoftokenfrequencyandconstructionalschemas.ThephenomenonofDARhasreceivedquitealotofattentionintheLVCliterature(seeaboveforsomereferences),butsociolinguistshaveremainedoblivioustothepossibleroleoffrequencyorschemas.

Page 367: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

Page 12 of 13

Mostofthedetailsconcerningtheroleoffrequencyandschemasinaccountingforthereduction,orevencompleteomission,ofthedefinitearticleinthedefiniteNPconstructionarenotsorelevanthere.Whatisespeciallyimportantisthatthereisonenoun,smallholdings(seeexample(18),above),whosefrequentco-occurrencewithreducedorzeroformsofthecannoteasilybeexplainedeitherwithreferencetoitsfrequencyoritsstatusasaninstanceofsomeschemawithareduceddefinitearticle.

ThetentativeexplanationofferedbyHollmannandSiewierskaisasocialone.HavingestablishedtheimportanceoffarmingtoLancashireineconomicterms,theysuggestthatfarming,andperhapsespeciallythetypeofsmall-scalefarmingthatsmallholdingsrefersto,maybeanimportantaspectofLancashireidentityaswell.ThenextstepoftheirargumentisinspiredbytheworkonCardiffEnglishbythesociolinguistNikolasCoupland.Withregardtothephonologicalvariable(a),Couplandobservesthatitslocalpronunciationfeaturesin“anumberoflexicalformswhichrelatetofocalelementsofCardifflocalculture”(1988:27).OneofhisexamplesistheverywordCardiff,andheadds:

ThecontemporaryCardifffolk-hero,singerandbroadcasterFrankHennessey,hasasoneofhiscatchwordsremarkable([rilmæ͂kə:blן]),and,ashisradioprogrammejingle,asungfanfaredominatedbythevowel-quality[æ:]inHarkharkthelarkinCardiffArm'sPark.Atonelevel,theselexicalformsareavehicleforthephonologicalvariable,buttheiravailabilityinturnreinforcesandcompoundsthesocialsignificanceofthevariableitself.Inthisinstance,regionalpronunciationandlocalexperiencehaveamutuallyencouraging,wemightsaysymbiotic,relationship.(Coupland1988:27)

HollmannandSiewieskaproposethatasimilarexplanationmayunderlietheuseofDARwithsmallholdingsinLancashiredialect:alocalvariantisusedtomarkLancashireidentityinaspecificconstructionwhichdescribesafocalelementoflocalculture.

Totheextentthatthisaccountisvalid HollmannandSiewierskasuggestthatthiscallsforamoresubtleunderstandingofthecorrelationbetweentokenfrequencyandreduction.Inparticular:ifaconstructionisnotveryfrequentthenitmaystillundergoreduction,ifthekindofreductioninquestionisafeatureofsomevariety(regionalorotherwise)andiftheconstructioninquestiondescribessomefocalelementofthecommunityassociatedwiththatvariety.Ofcourse,comparedtoothervarietiesthesefocalelementswillbediscussedfairlyoften,butthesmallholdingsisnonethelesslessfrequentthanpresumablyuniversallyfrequentdefiniteNPssuchasthemanorthehouse,whichdonot,however,displaythesamedegreeofdefinitearticlereduction.Inotherwords,tokenfrequencywillnotalwaysleadtoreductioninapurelymechanisticmanner,buteachpotentialcaseshouldbeseenasbeingembeddedinacertainsocialcontext.

Tosumup,HollmannandSiewierska'sstudyonDARshowsthatnotonlydoescognitivelinguisticshaveimportantnotionstocontributetosociolinguistics/LVC(theimportanceoffrequencyandtheroleofconstructionalschemas),butcognitivelinguistsmustalsotakeonboardthesociolinguisticinsightthatlanguagedoesnotexistonlyinthemindsofspeakersbutalsoinsociety.ThistakesusbacktoCroft's(2009b)suggestion,discussedinsection27.2,above,thatcognitivelinguistsshouldbecarefulnottoconstruelanguagetoosolipsistically.

27.5.Conclusion

Thischapterhasdiscussedsomeofthewaysinwhichconstructionsmaybe,andhavebeen,studiedfromacognitivesociolinguisticperspective.

Croft(2009b)hasarguedstronglythatcognitivelinguistsshouldbewareofthetendencytocouchtheirexplanationsinpurelymentalterms.His(2009b,2010d)workofferssomefirstexcitinginsightsintotheimplicationsofopeningupourscopeofresearchsoastoincludethesocial-interactionaldimensionoflinguisticcommunication.Concerningboththebasiccognitivelinguisticnotionofconstrualandgrammaticalization,theimplicationsspanbothsynchronyanddiachrony.

TheworkbyGries(e.g.,2003a),Grondelaersetal.(e.g.,2007,2008),andHollmannandSiewierska(2006,2007,2011)inasenseopensupthescopeofthecognitivelinguisticsandConstructionGrammarenterpriseevenfurther.BothGriesandespeciallyGrondelaersetal.showthatcognitivelinguisticexplanationscanshedmuchvaluablelightondatathatwouldtraditionallyhavefallenwithinthejurisdictionofsociolinguists,prominently

2

Page 368: The Oxford Handbook of Construction Grammar

Constructions in Cognitive Sociolinguistics

includingscholarsworkingintheLVCparadigm.Butwithoutthecognitivelinguisticconceptualtoolkit,muchofthevariationobserved(inEnglishTransitivePhrasalVerbconstructionsandDutcher-presentativeconstructions,respectively)wouldremainunexplained.TheresearchbyHollmannandSiewierskashowsthatatthetheoreticallevel,too,cognitivelinguisticsandConstructionGrammarhavealottooffertosociolinguistics(theroleoftokenfrequencyinreductionprocesses,theimportanceofschemas),andthataspectsofsociolinguistictheoryshouldbeincorporatedincognitiveandconstructionistworkaswell(thepotentialforcertainconstructionstobeusedasidentitymarkersofsomesort,andtheconsequencesthismayhavefortheirstatusinthelinguisticsystem).

Thereadermayhavenotedwithsomeinterestthatthisbriefdiscussionofthecognitivesociolinguisticenterprisehasessentiallyonlyincludedestablishedcognitivelinguistsacknowledgingtheneedtotakeonboardissuesrelatedtosocialstructure.Rapprochementfromthesociolinguisticsideofthespectrumhasbeenmuchthinnerontheground.Significantly,despitewhatmightappeartobepromisedbythetitleofLabov's(2010)monographPrinciplesofLinguisticChange,Volume3:CognitiveandCulturalFactors,thestudydoesnotcontainevenasinglereferencetotheworkofthecognitivelinguistsmentionedhere.

Atleastfornow,then,cognitivesociolinguisticsismoreorlessaone-waystreet.Constructiongrammariansandcognitivelinguistsingeneralmostlikelystandtogainconsiderableprofitsfromincorporatingevenmoreinsightsfromsociolinguisticsintotheirworkthantheyalreadyhave.Buttheinterfacebetweenthetwoareasoflinguisticswillalmostcertainlydevelopfasterifandwhentrafficstartstoflowbothways:sociolinguists,too,woulddowelltostarttozoominonlanguageasacognitivephenomenonaswellasasocialone.

IamverygratefultoBertCappelleandtheeditorsofthisvolumefortheirelaborateandinsightfulcommentsonthischapter.Anyinadequaciesthatremainaresolelymyresponsibility.

Notes:

(1.)Grondelaersetal.(2008)actuallyincorrectlyrefertoGrondelaersetal.(2001)as(2002).

(2.)MindfulofJohnstoneandKiesling's(2008)warningagainstautomaticallyassumingthatcorrelationsbetweennonstandardvariantsandregionshouldbeinterpretedasevidenceforlocalidentitymarkerstatus,HollmannandSiewierskaareverycarefulnottostatetheircasetoostrongly.

WillemB.HollmannWillemB.HollmannhasbeenaLecturerinLinguisticsatLancasterUniversitysince2003.HispublicationsincludearticlesinCognitiveLinguisticsonsummaryvs.sequentialscanning(2007)andcognitivesociolinguisticsandLancashiredialect(2011).HehaspublishedonLancashiredialectgrammarelsewhereaswell,e.g.inEnglishLanguageandLinguistics(2007).Hehasedited(withAnnaSiewierska)aspecialissueofFunctionsofLanguageonditransitiveconstructions,towhichhecontributedapaperaswell(2007).Oneofthethingsheisworkingoncurrentlyistheroleofphonologicalanddistributionalcuesinlexicalcategorization.OnepaperonthistopicwillappearinaspecialissueofStudiesinLanguage,whichheiseditingwithNikolasGisborne.

Page 369: The Oxford Handbook of Construction Grammar

References

Page 1 of 88

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

DOI: 10.1093/oxfordhb/9780195396683.004.0001

ReferencesTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

References

Aarts,Bas(2007).SyntacticGradience:TheNatureofGrammaticalIndeterminacy.Oxford:OxfordUniversityPress.

Abbot-Smith,Kirsten,Dittmar,Miriam,andTomasello,Michael(2007).“GradedRepresentationsintheAcquisitionofEnglishandGermanTransitiveConstructions.”CognitiveDevelopment23:48–66.

Abbot-Smith,Kirsten,andTomasello,Michael(2006).“Exemplar-learningandSchematizationinaUsage-basedAccountofSyntacticAcquisition.”TheLinguisticReview23:275–90.

———.(2010).“TheInfluenceofFrequencyandSemanticSimilarityonHowChildrenLearnGrammar.”FirstLanguage30:79–101.

Achard,Michel(2008).“CognitivePedagogicalGrammar,”inPeterRobinsonandNickC.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge,432–55.

Ackerman,Farrell,Blevins,JamesP.,andMalouf,Robert(2009).“PartsandWholes:ImplicativePatternsinInflectionalParadigms,”inJamesP.BlevinsandJulietteBlevins(eds.),AnalogyinGrammar.Oxford:OxfordUniversityPress,54–82.

Ackerman,Farrell,andGoldberg,AdeleE.(1996).“ConstraintsonAdjectivalPastParticiples,”inAdeleE.Goldberg(ed.),ConceptualStructure,DiscourseandLanguage.Stanford,CA:CSLIPublications,17–30.

Ackerman,Farrell,andNikolaeva,Irina(2004).ComparativeGrammarandGrammaticalTheory:AConstruction-BasedStudyofMorphosyntax.Stanford,CA:CenterfortheStudyofLanguageandInformation.

Ackerman,Farrell,andStump,GregoryS.(2004).“ParadigmsandPeriphrasticExpression:A

Page 370: The Oxford Handbook of Construction Grammar

References

Page 2 of 88

StudyinRealization-basedLexicalism,”inLouisaSadlerandAndrewSpencer(eds.),ProjectingMorphology.Stanford,CA:CSLIPublications,111–57.

Acuña-Fariña,JuanCarlos(2009).“AspectsoftheGrammarofCloseAppositionandtheStructureoftheNounPhrase.”EnglishLanguageandLinguistics13:453–81.

Adger,David(2006).“CombinatorialVariability.”JournalofLinguistics42:503–30.

Adger,David,andTrousdale,Graeme(2007).“VariationinEnglishSyntax:TheoreticalImplications.”EnglishLanguageandLinguistics11:261–78.

Ahrens,Kathleen(1995).“TheMentalRepresentationofVerbs.”UCSDdissertation.

Akthar,Nameera(1999).“AcquiringBasicWordOrder:EvidenceforData-drivenLearningofSyntacticStructure.”JournalofChildLanguage26:339–56.

Akthar,Nameera,andTomasello,Michael(1997).“YoungChildren'sProductivitywithWordOrderandVerbMorphology.”DevelopmentalPsychology33:952–65.

Alario,Xavier,andCaramazza,AlfonsoA.(2002).“TheProductionofDeterminers:EvidencefromFrench.”Cognition82:179–223.

Alishahi,Afra,andStevenson,Suzanne(2008).“AComputationalModelofEarlyArgumentStructureAcquisition.”CognitiveScience32:789–834.

Allen,Cynthia.(2001).“TheDevelopmentofaNewPassiveinEnglish,”inMiriamButtandTracyHollowayKing(eds.),TimeoverMatter:DiachronicPerspectivesonMorphosyntax.Standford:CSLIPublications,43–72.

Allen,Kachina,Pereira,Francisco,Botvinick,Matthew,andGoldberg,AdeleE.(2012).“DistinguishingGrammaticalConstructionswithfMRIPatternAnalysis”.BrainandLanguage123:174–82.

Alsina,Alex,andMchombo,Sam.A.(1990).“TheSyntaxofApplicativesinChichewa:ProblemsforaThetaTheoreticAsymmetry.”NaturalLanguageandLinguisticTheory8:493–506.

Altmann,GerryT.M.,andSteedman,MarkJ.(1988).“InteractionwithContextduringHumanSentenceProcessing.”Cognition30/3:191–238.

Ambridge,Ben,andGoldberg,AdeleE.(2008).“TheIslandStatusofClausalComplements:EvidenceinFavorofanInformationStructureExplanation.”CognitiveLinguistics19:357–89.

Ambridge,Ben,andLieven,ElenaV.M.(2011).ChildLanguageAcquisition:ContrastingTheoreticalApproaches.Cambridge:CambridgeUniversityPress.

Ambridge,Ben,Theakston,Anna,Lieven,ElenaV.M.,andTomasello,Michael(2006).“TheDistributedLearningEffectforChildren'sAcquisitionofanAbstractGrammaticalConstruction.”CognitiveDevelopment21:174–93.

Page 371: The Oxford Handbook of Construction Grammar

References

Page 3 of 88

Andersen,Henning(1987).“FromAuxiliarytoDesinence,”inMartinHarrisandPaoloRamat(eds.),HistoricalDevelopmentofAuxiliaries.BerlinandNewYork:MoutondeGruyter,1–50.

———.(ed.)(2001).Actualization:LinguisticChangeinProgress.CurrentIssuesinLinguisticTheory219.Amsterdam:JohnBenjamins.

Anderson,JohnR.(1989).“ARationalAnalysisofHumanMemory,”inHenryL.I.RoedigerandFergusI.M.Craik(eds.),VarietiesofMemoryandConsciousness:EssaysinHonourofEndelTulving.Hillsdale,NJ:LawrenceErlbaumAssociates,195–210.

Anderson,JohnR.,andSchooler,LaelJ.(2000).“TheAdaptiveNatureofMemory,”inEndelTulvingandFergusI.M.Craik(eds.),TheOxfordHandbookofMemory.Oxford:OxfordUniversityPress,557–70.

Anderson,LloydB.(1982).“The‘Perfect’asaUniversalandasaLanguage-particularCategory,”inPaulHopper(ed.),Tense-Aspect:BetweenSemanticsandPragmatics.Amsterdam:JohnBenjamins,227–64.

———.(1986).“Evidentials,PathsofChangeandMentalMaps:TypologicallyRegularAsymmetries,”inWallaceChafeandJohannaNichols(eds.),Evidentiality:TheLinguisticEncodingofEpistemology.Norwood:Ablex,273–312.

ARCHER(ARepresentativeCorpusofHistoricalEnglishRegisters).nowebsite.

Arcodia,GiorgioF.(2011).“AConstructionMorphologyAccountofDerivationinMandarinChinese.”Morphology21:89–130.

Aronoff,Mark(1976).WordFormationinGenerativeGrammar.Cambridge,MA:MITPress.

———.(1983).“PotentialWords,ActualWords,ProductivityandFrequency.”PaperPresentedtotheProceedingsofthe13thInternationalCongressofLinguists,Oslo.

———.(1994).MorphologybyItself:StemsandInflectionalClasses.Cambridge,MA:MITPress.

Asaka,T.(2002).“ALexicalLicensingAnalysisoftheAdjectivalNounConstruction.”EnglishLinguistics19:113–41.

Auer,Peter(ed.)(2009).StyleandSocialIdentities:AlternativeApproachestoLinguisticHeterogeneity.Berlin:MoutondeGruyter.

———.(2011).“OnlineSyntaxandConstructionGrammar:APerfectMatch?SomeReflectionsonEllipticalExpansions.”PaperPresentedattheWorkshoponGrammarandInteractionRevisited,Helsinki,March10–12,2011.

Baayen,R.Harald(1993).“OnFrequency,TransparencyandProductivity,”inGeertE.BooijandJaapvanMarle(eds.),YearbookofMorphology1992.Dordrecht:KluwerAcademicPublishers,181–208.

Page 372: The Oxford Handbook of Construction Grammar

References

Page 4 of 88

———.(2003).“ProbabilisticApproachestoMorphology,”inRensBod,JenniferHay,andStefanieJannedy(eds.),ProbabilisticLinguistics.Cambridge,MA,andLondon:MITPress,229–87.

Baayen,R.Harald,Burani,Christina,andSchreuder,Rob(1997).“EffectsofSemanticMarkednessintheProcessingofRegularNominalSingularsandPluralsinItalian,”inGeertBooijandJaapVanMarle(eds.),YearbookofMorphology1996.Dordrecht:Kluwer,13–34.

Baayen,R.Harald,Levelt,Willem,andHaveman,Alette(1993).ResearchReportedinAnnualReportoftheMax-PlanckInstitutfürPsycholinguistik,6.Nijmegen:Max-PlanckInstitut.

Bailey,David(1997).“AComputationalModelofEmbodimentintheAcquisitionofActionVerbs.”Doctoraldissertation,ComputerScienceDivision,EECSDepartment,UniversityofCalifornia,Berkeley.

Bailey,David,Feldman,Jerome,Narayanan,Srini,andLakoff,George(1997).“ModellingEmbodiedLexicalDevelopment,”inMichaelG.ShaftoandPatLangley(eds.),Proceedingsofthe19thCognitiveScienceConference.Mahwah,NJ:LawrenceErlbaumAssociates,84–89.

Baker,CollinF.,Fillmore,CharlesJ.,andCronin,Beau(2003).“TheStructureoftheFramenetDatabase.”InternationalJournalofLexicography16:281–96.

Baker,Mark(1997).“ThematicRolesandSyntacticStructure,”inLilianeHaegeman(ed.),ElementsofGrammar.Kluwer:Dordrecht,73–137.

———.(2004).Verbs,NounsandAdjectives:TheirUniversalGrammar.Cambridge:CambridgeUniversityPress.

Ball,CatherineN.(1996).“ADiachronicStudyofRelativeMarkersinSpokenandWrittenEnglish.”LanguageVariationandChange8:227–58.

Bannard,Colin(2005).“LearningAbouttheMeaningofVerbParticleConstructionsfromCorpora.”JournalofComputerSpeechandLanguage19/4:467–78.

Bannard,Colin,Baldwin,Timothy,andLascarides,Alex(2003).“AStatisticalApproachtotheSemanticsofVerb-Particles.”ProceedingsoftheACL-WorkshoponMultiwordExpressions:Analysis,AcquisitionandTreatment,65–72.

Bannard,Colin,andMatthews,Danielle(2008).“StoredWordSequencesinLanguageLearning:TheEffectofFamiliarityonChildren'sRepetitionofFour-WordCombinations.”PsychologicalScience19:241–48.

Bao,Zhiming,andWee,Lionel(1999).“ThePassiveinSingaporeEnglish.”WorldEnglishes18:1–13.

Barðdal,Jóhanna(1999).“CaseandArgumentStructureofSomeLoanVerbsin15thCenturyIcelandic,”inIngerHaskåandCarinSandqvist(eds.),Allatidersspråk:EnVänskrifttillGertrudPetterssonNovember1999.LundastudieriNordiskspråkvetenskapA55.Lund:DepartmentofScandinavianLanguages,9–23.

Page 373: The Oxford Handbook of Construction Grammar

References

Page 5 of 88

———.(2001).“ThePerplexityofDat-NomVerbsinIcelandic.”NordicJournalofLinguistics24:47–70.

———.(2006).“Construction-specificPropertiesofSyntacticSubjectsinIcelandicandGerman.”CognitiveLinguistics17:39–106.

———.(2008).Productivity:EvidencefromCaseandArgumentStructureinIcelandic.Amsterdam:JohnBenjamins.

———.(2009).“TheDevelopmentofCaseinGermanic,”inJóhannaBarðdalandShobhanaChelliah(eds.),TheRoleofSemantic,PragmaticandDiscourseFactorsintheDevelopmentofCase.Amsterdam:JohnBenjamins,123–59.

———.(2011a).“Lexicalvs.StructuralCase:AFalseDichotomy.”Morphology21/1:619–54.

———.(2011b).“TheRiseofDativeSubstitutionintheHistoryofIcelandic:ADiachronicConstructionGrammarAnalysis.”Lingua121/1:60–79.

Barðdal,Jóhanna,Bjarnadóttir,Valgerður,Dahl,Eystein,Eythórsson,Thórhallur,Fedriani,Chiara,andSmitherman,Thomas(2011).“‘WoetotheRichandtheSordidFellows’:TheSyntax,SemanticsandInformationStructureof‘Woe’inIndo-European.”Ms.,UniversityofBergen.[UnderrevisionforTheJournalofIndo-EuropeanStudies].

Barðdal,Jóhanna,andChelliah,Shobhana(eds.)(2009).TheRoleofSemantic,PragmaticandDiscourseFactorsintheDevelopmentofCase.Amsterdam:JohnBenjamins.

Barðdal,Jóhanna,andEythórsson,Thórhallur(2003).“TheChangethatNeverHappened:TheStoryofObliqueSubjects.”JournalofLinguistics39:439–72.

———.(2012a).“ReconstructingSyntax:ConstructionGrammarandtheComparativeMethod,”inHansC.BoasandIvanA.Sag(eds.),Sign-BasedConstructionGrammar.Stanford,CA:CSLIPublications,261–312.

———.(2012b).“‘HungeringandLustingforWomenandFleshlyDelicacies’:ReconstructingGrammaticalRelationsforProto-Germanic.”TransactionsofthePhilologicalSociety110/3:363–93.

Barðdal,Jóhanna,Kristoffersen,KristianE.,andSveen,Andreas(2011).“WestScandinavianDitransitivesasaFamilyofConstructions:WithaSpecialAttentiontotheNorwegian‘V-REFL-NP’Construction.”Linguistics49/1:53–104.

Barðdal,Jóhanna,andMolnár,Valéria(2003).“ThePassiveinIcelandic—ComparedtoMainlandScandinavian,”inJorunnHetlandandValériaMolnár(eds.),StructuresofFocusandGrammaticalRelations.LinguistischeArbeiten477.Tübingen:MaxNiemeyerVerlag,231–60.

Barðdal,Jóhanna,andSmitherman,Thomas(2013).“TheQuestforCognates:AReconstructionofObliqueSubjectConstructionsinProto-Indo-European.”LanguageDynamicsandChange3/1.

Page 374: The Oxford Handbook of Construction Grammar

References

Page 6 of 88

Bardovi-Harlig,Kathleen(1987).“MarkednessandSalienceinSecond-LanguageAcquisition.”LanguageLearning37:385–407.

———.(2000).TenseandAspectinSecondLanguageAcquisition:Form,MeaningandUse.Oxford:Blackwell.

Barkema,Henk(1994).“DeterminingtheSyntacticFlexibilityofIdioms,”inUdoFries,GunnelTottie,andPeterSchneider(eds.),CreatingandUsingEnglishLanguageCorpora.Amsterdam:Rodopi,39–52.

Barlow,Horace,andLevick,WilliamR.(1965).“TheMechanismofDirectionallySelectiveUnitsinRabbit'sRetina.”JournalofPhysiology178:477–504.

Barlow,Michael,andKemmer,Suzanne(eds.)(2000).UsageBasedModelsofGrammar.Stanford,CA:CSLIPublications.

Baron-Cohen,Simon,Leslie,AlanM.,andFrith,Uta(1985).“DoestheAutisticChildhavea‘TheoryofMind’?”Cognition21:37–46.

Barsalou,LawrenceW.(1999).“PerceptualSymbolSystems.”BehavioralandBrainSciences22/4:577–609;discussion610–50.

———.(2008).“GroundedCognition.”AnnualReviewofPsychology59:617–45.

Basciano,Bianca(2010).VerbalCompoundingandCausativityinMandarinChinese.Verona:UniversitàdegliStudidiVerona.

Bastiaansen,Marcel,andHagoort,Peter(2006).“OscillatoryNeuronalDynamicsduringLanguageComprehension.”ProgressinBrainResearch159:179–96.

Bates,Elizabeth,andDevoscovi,Antonella(1989).“CrosslinguisticStudiesofSentenceProduction,”inBrianMacWhinneyandElizabethBates(eds.),TheCrosslinguisticStudyofSentenceProcessing.Cambridge,CambridgeUniversityPress,225–53.

Bates,Elizabeth,andMacWhinney,Brian(1982).“FunctionalistApproachestoGrammar,”inEricWannerandLilaGleitman(eds.),LanguageAcquisition:TheStateoftheArt.Cambridge:CambridgeUniversityPress,173–218.

———.(1987).“Competition,VariationandLanguageLearning,”inBrianMacWhinney(ed.),MechanismsofLanguageAcquisition.Hillsdale,NJ:LawrenceErlbaumAssociates,157–93.

———.(1989).“FunctionalismandtheCompetitionModel,”inElizabethBatesandBrianMacWhinney(eds.),TheCrosslinguisticStudyofSentenceProcessing.Cambridge:CambridgeUniversityPress,3–73.

Bauer,Laurie(2001).MorphologicalProductivity.Cambridge:CambridgeUniversityPress.

———.(2009).“TypologyofCompounds,”inRochelleLieberandPavolStekauer(eds.),TheOxfordHandbookofCompounding.Oxford:OxfordUniversityPress,343–56.

Page 375: The Oxford Handbook of Construction Grammar

References

Page 7 of 88

Beck,Sigrid(1997).“OntheSemanticsofComparativeConditionals.”LinguisticsandPhilosophy20/3:229–71.

Beckner,Clay,Blythe,Richard,Bybee,Joan,Christiansen,MortonH.,Croft,William,Ellis,NickC.,Holland,John,Ke,Jinyun,Larsen-Freeman,Diane,andSchoenemann,Thomas(2009).“LanguageisaComplexAdaptiveSystem.PositionPaper.”LanguageLearning59,Supplement1:1–26.

Beckner,Clay,andBybee,Joan(2009).“AUsage-basedAccountofConstituencyandReanalysis.”LanguageLearning59,Supplement1:29–48.

Bell,Allan(1984).“LanguageStyleasAudienceDesign.”LanguageinSociety13:145–204.

Bencini,GiuliaM.L.,andGoldberg,AdeleE.(2000).“TheContributionofArgumentStructureConstructionstoSentenceMeaning.”JournalofMemoryandLanguage43:640–51.

Bencini,GiuliaM.L.,Pozzan,Lucia,Biundo,Roberta,McGeowan,WilliamJ.,Venneri,Annalena,Valian,Virginia,andSemenza,Carlo(2011).“LanguageSpecificEffectsinAlzheimer'sDisease:SubjectOmissioninItalianandEnglish.”JournalofNeurolinguistics24:25–40.

Bencini,GiuliaM.L.,andValian,Virginia(2008).“AbstractSentenceRepresentationsin3-year-olds:EvidencefromComprehensionandProduction.”JournalofMemoryandLanguage59:97–113.

Bergen,BenjaminK.(2007).“ExperimentalMethodsforSimulationSemantics,”inMonicaGonzalez-Marquez,IreneMittelberg,SeanaCoulson,andMichaelJ.Spivey(eds.),MethodsinCognitiveLinguistics.Amsterdam:JohnBenjamins,277–301.

———.(2012).LouderThanWords:TheNewScienceofHowtheMindMakesMeaning.NewYork:BasicBooks.

Bergen,BenjaminK.,andChang,Nancy(2005).“EmbodiedConstructionGrammarinSimulation-BasedLanguageUnderstanding,”inJan-OlaOstmanandMirjamFried(eds.),ConstructionGrammars:CognitiveGroundingandTheoreticalExtensions.Amsterdam:JohnBenjamins,147–90.

Bergen,BenjaminK.,Lindsay,Shane,Matlock,Teenie,andNarayanan,Srini(2007).“SpatialandLinguisticAspectsofVisualImageryinSentenceComprehension.”CognitiveScience31:733–64.

Bergen,BenjaminK.,andPlauché,Madelaine(2001).“VoilàVoilà:ExtensionsofDeicticConstructionsinFrench,”inAlanCienki,BarbaraLuka,andMichaelSmith(eds.),ConceptualandDiscourseFactorsinLinguisticStructure.Stanford,CA:CSLIPublications,238–49.

Bergen,BenjaminK.,Ting-TingChanLau,Narayan,Shweta,Stojanovic,Diana,andWheeler,Kathryn(2010).“BodyPartRepresentationsinVerbalSemantics.”MemoryandCognition38/7:969–81.

Bergen,BenjaminK.,andWheeler,Kathryn(2005).“SentenceUnderstandingEngagesMotor

Page 376: The Oxford Handbook of Construction Grammar

References

Page 8 of 88

Processes.”ProceedingsoftheTwenty-SeventhAnnualConferenceoftheCognitiveScienceSociety,Stresa.

———.(2010).“GrammaticalAspectandMentalSimulation.”BrainandLanguage112:150–58.

Bergs,Alexander(2008).“Shallandshan’tinContemporaryEnglish,”inGraemeTrousdaleandNikolasGisborne(eds.),ConstructionalApproachestoEnglishGrammar.BerlinandNewYork:MoutondeGruyter,113–44.

———.(2010).“ExpressionsofFuturityinContemporaryEnglish:AConstructionGrammarPerspective.”EnglishLanguageandLinguistics14/2:217–38.

Bergs,Alexander,andDiewald,Gabriele(eds.)(2008).ConstructionsandLanguageChange.Berlin:MoutondeGruyter.

———.(eds.)(2009).ContextsandConstructions.Amsterdam:JohnBenjamins.

Berkenfield,Catie(2001).“TheRoleofFrequencyintheRealizationofEnglishthat,”inJoanL.BybeeandPaulJ.Hopper(eds.),FrequencyandtheEmergenceofLinguisticStructure.Amsterdam:JohnBenjamins,281–308.

Berko,Jean(1958).“TheChild'sLearningofEnglishMorphology.”Word14:150–77.

Berlin,Brent,andKay,Paul(1969).BasicColorTerms:TheirUniversalityandEvolution.Berkeley:UniversityofCaliforniaPress.

Berman,RuthA.,andSlobin,DanI.(eds.)(1994).RelatingEventsinNarrative:ACrosslinguisticDevelopmentalStudy.Hillsdale,NJ:LawrenceErlbaum.

Berry-Rogghe,GodelieveL.M.(1974).“AutomaticIdentificationofPhrasalVerbs,”inJohnL.Mitchell(ed.),ComputersintheHumanities.Edinburgh:EdinburghUniversityPress,16–26.

Berthele,Raphael(2008).“ANationisaTerritorywithOneCultureandOneLanguage:TheRoleofMetaphoricalFolkModelsinLanguagePolicyDebates,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,301–32.

Bever,ThomasG.(1970).“TheCognitiveBasisforLinguisticStructures,”inJohnR.Hayes(ed.),CognitionandDevelopmentofLanguage.NewYork:Wiley,279–352.

Bhatia,VijayK.(1993).AnalysingGenre—LanguageUseinProfessionalSettings.London:Longman.

Biber,Douglas(1988).VariationAcrossSpeechandWriting.Cambridge:CambridgeUniversityPress.

Biber,Douglas,Johansson,Stig,Leech,Geoffrey,Conrad,Susan,andFinegan,Edward(1999).LongmanGrammarofSpokenandWrittenEnglish.London:Longman.

Page 377: The Oxford Handbook of Construction Grammar

References

Page 9 of 88

Bickel,Balthasar,Hildebrandt,KristineA.,andSchiering,René(2009).“TheDistributionofPhonologicalWordDomains:AProbabilisticTypology,”inBarişKabakandJanetGrijzenhout(eds.),PhonologicalDomains:UniversalsandDeviations.Berlin:MoutondeGruyter,47–75.

Binder,JeffreyR.,Westbury,ChrisF.,McKiernan,KristenA.,Possing,EdwardT.,andMedler,DavidA.(2005).“DistinctBrainSystemsforProcessingConcreteandAbstractConcepts.”JournalofCognitiveNeuroscience17:905–17.

Birner,BettyJ.,andWard,Gregory(1998).InformationStatusandNoncanonicalWordOrderinEnglish.Amsterdam:JohnBenjamins.

Bisang,Walter(1998).“GrammaticalizationandLanguageContact,ConstructionsandPositions,”inAnnaGiacaloneRamatandPaulJ.Hopper(eds.),TheLimitsofGrammaticalization.Amsterdam:JohnBenjamins,13–58.

———.(2010).“GrammaticalizationinChinese:AConstruction-basedAccount,”inElizabethClossTraugottandGraemeTrousdale(eds.),Gradience,GradualnessandGrammaticalization.TypologicalStudiesinLanguage90.Amsterdam:JohnBenjamins,245–77.

Bjorvand,Harald,andLindeman,FredrikOtto(2000).Vårearveord:Etymologiskordbok.[‘OurInheritedVocabulary:AnEtymologicalDictionary.’]Oslo:Novusforlag.

Blevins,JamesP.(2001).“Realisation-basedLexicalism.”JournalofLinguistics37:355–65.

———.(2006).“Word-basedMorphology.”JournalofLinguistics42:531–73.

Blom,Corrien(2004).“OntheDiachronyofComplexPredicatesinDutch:PredicativeandNon-predicativePreverbs.”JournalofGermanicLinguistics16:1–75.

———.(2005a).ComplexPredicatesinDutch:SynchronyandDiachrony.Utrecht:LOT.

———.(2005b).“TheDemarcationofMorphologyandSyntax:ADiachronicPerspectiveonParticleVerbs,”inWolfgangU.Dressler,DieterKastovsky,OskarE.Pfeiffer,andFranzRainer(eds.),MorphologyanditsDemarcations.Amsterdam:JohnBenjamins,53–66.

Blom,Corrien,andBooij,Geert(2003).“TheDiachronyofComplexPredicatesinDutch:ACaseStudyinGrammaticalization.”ActaLinguisticaHungarica50:61–91.

Blust,Robert(1996).“TheNeogrammarianHypothesisandPandemicIrregularity,”inMarkDurieandMalcolmRoss(eds.),TheComparativeMethodReviewed:RegularityandIrregularityinLanguageChange.Oxford:OxfordUniversityPress,135–56.

BNC(TheBritishNationalCorpus).http://www.natcorp.ox.ac.uk/,http://corpus.byu.edu/bnc/x.asp.

Boas,HansC.(2003).AConstructionalApproachtoResultatives.Stanford,CA:CSLIPublications.

Page 378: The Oxford Handbook of Construction Grammar

References

Page 10 of 88

———.(2004).“YouWannaConsideraConstructionalApproachtoWanna-Contraction?”inMichaelAchardandSuzanneKemmer(eds.),Language,CultureandMind.Stanford,CA:CSLIPublications,471–91.

———.(2005).“DeterminingtheProductivityofResultativeConstructions:AReplytoGoldbergandJackendoff.”Language81/2:448–64.

———.(2006).“AFrame-semanticApproachtoIdentifyingSyntacticallyRelevantElementsofMeaning,”inPetraSteiner,HansC.Boas,andStefanSchierholz(eds.),ContrastiveStudiesandValency:StudiesinHonorofHansUlrichBoas.FrankfurtandNewYork:PeterLang,119–49.

———.(2008).“DeterminingtheStructureofLexicalEntriesandGrammaticalConstructionsinConstructionGrammar.”AnnualReviewofCognitiveLinguistics6:113–44.

———.(2009a).“VerbMeaningsattheCrossroadsbetweenHigher-levelandLower-levelConstructions.”Lingua120:22–34.

———.(ed.)(2009b).MultilingualFrameNetsinComputationalLexicography:MethodsandApplications.Berlin:MoutondeGruyter.

———.(ed.)(2010a).ContrastiveStudiesinConstructionGrammar.Amsterdam:JohnBenjamins.

———.(2010b).“TheSyntax-lexiconContinuuminConstructionGrammar:ACaseStudyofEnglishCommunicationVerbs.”BelgianJournalofLinguistics24:54–92.

———.(2011a).“ZumAbstraktionsgradvonResultativkonstruktionen,”inStefanEngelberg,AnkeHoller,andKristelProost(eds.),SprachlichesWissenzwischenLexikonundGrammatik.BerlinandNewYork:MoutondeGruyter,37–69.

———.(2011b).“AFrame-semanticApproachtoSyntacticAlternationswithBuild-verbs,”inPilarGuerreroMedina(ed.),MorphosyntacticAlternationsinEnglish.London:Equinox,207–34.

Boas,HansC.,andSag,Ivan(eds.)(2012).Sign-BasedConstructionGrammar.Stanford,CA:CSLIPublications.

Bochner,Harry(1993).SimplicityinGenerativeMorphology.BerlinandNewYork:MoutondeGruyter.

Bock,Kathryn(1982).“TowardaCognitivePsychologyofSyntax:InformationProcessingContributionstoSentenceFormulation.”PsychologicalReview89:1–47.

———.(1986a).“Meaning,SoundandSyntax:LexicalPriminginSentenceProduction.”JournalofExperimentalPsychology:Learning,MemoryandCognition12:575–86.

———.(1986b).“SyntacticPersistenceinLanguageProduction.”CognitivePsychology18:355–87.

Page 379: The Oxford Handbook of Construction Grammar

References

Page 11 of 88

———.(1990).“StructureinLanguage:CreatingForminTalk.”AmericanPsychologist45:1221–36.

———.(1995).“SentenceProduction:FromMindtoMouth,”inJoanneL.MillerandPeterD.Eimas(eds.),HandbookofPerceptionandCognition:Speech,LanguageandCommunication.Vol.11.Orlando,FL:AcademicPress,181–216.

Bock,Kathryn,Dell,Gary,Chang,Franklin,andOnishi,Kristine(2007).“PersistentStructuralPrimingfromLanguageComprehensiontoLanguageProduction.”Cognition104:437–58.

Bock,Kathryn,andGriffin,Zenzi(2000).“ThePersistenceofStructuralPriming:TransientActivationorImplicitLearning?”JournalofExperimentalPsychology:General129:177–92.

Bock,KathrynJ.,andIrwin,D.E.(1980).“SyntacticEffectsofInformationAvailabilityinSentenceProduction.”JournalofVerbalLearningandVerbalBehavior19:467–84.

Bock,Kathryn,andLevelt,Wilhelm(1994).“LanguageProduction:GrammaticalEncoding,”inMortonAnneGernsbacher(ed.),HandbookofPsycholinguistics.SanDiego:AcademicPress,945–84.

Bock,Kathryn,andLoebell,Helga(1990).“FramingSentences.”Cognition35:1–39.

Bock,KathrynJ.,Loebell,Helga,andMorey,Randel(1992).“FromConceptualRolestoStructuralRelations:BridgingtheSyntacticCleft.”PsychologicalReview99:150–71.

Bock,Kathryn,andWarren,Richard(1985).“ConceptualAccessibilityandSyntacticStructureinSentenceFormulation.”Cognition21:57–67.

Bod,Rens(1998).BeyondGrammar:AnExperience-basedTheoryofLanguage.Stanford,CA:CSLI.

Boers,Frank,andLindstromberg,Seth(eds.)(2008).CognitiveApproachestoTeachingVocabularyandPhraseology.BerlinandNewYork:MoutondeGruyter.

Bolinger,Dwight(1977).MeaningandForm.London:Longman.

Booij,Geert(2002).“ConstructionalIdioms,MorphologyandtheDutchLexicon.”JournalofGermanicLinguistics14:301–29.

———.(2005).“CompoundingandDerivation:EvidenceforConstructionMorphology,”inWolfgangU.Dressler,DieterKastovsky,OskarE.Pfeiffer,andFranzRainer(eds.),MorphologyanditsDemarcations.Amsterdam:JohnBenjamins,109–32.

———.(2009).“PhrasalNames:AConstructionistAnalysis.”WordStructure3:219–40.

———.(2010).ConstructionMorphology.Oxford:OxfordUniversityPress.

Borer,Hagit(1994).“TheProjectionofArguments.”UniversityofMassachusettsOccasionalPapersinLinguistics17:19–48.

Page 380: The Oxford Handbook of Construction Grammar

References

Page 12 of 88

Borsley,RobertD.(2006).“SyntacticandLexicalApproachestoUnboundedDependencies.”EssexResearchReportsinLinguistics49:31–57.

———.(2007).“Hangonagain!Arewe‘ontheRightTrack’?MartinAtkinsonTheMinimalistMuse.”EssexResearchReportsinLinguistics53:43–70.

Botvinick,MatthewM.,andPlaut,DavidC.(2006).“Short-termMemoryforSerialOrder:ARecurrentNeuralNetworkModel.”PsychologicalReview113/2:201–33.

Boulenger,Véronique,Hauk,Olaf,andPulvermüller,Friedemann(2009).“GraspingIdeaswiththeMotorSystem:SemanticSomatotopyinIdiomComprehension.”CerebralCortex19/8:1905–14.

Bourdieu,Pierre(1979).LaDistinction:Critiquesocialedujugement.Paris:Minuit.

Bowerman,Melissa(1973).EarlySyntacticDevelopment.Cambridge:CambridgeUniversityPress.

———.(1982a).“EvaluatingCompetingLinguisticModelswithLanguageAcquisitionData:ImplicationsofDevelopmentalErrorswithCausativeVerbs.”QuadernidiSemantica3:5–66.

———.(1982b).“ReorganizationalProcessesinLexicalandSyntacticDevelopment,”inEricWannerandLilaR.Gleitman(eds.),LanguageAcquisition:TheStateoftheArt.Cambridge:CambridgeUniversityPress,319–46.

———.(1988).“The‘NoNegativeEvidence’Problem:HowdoChildrenAvoidConstructinganOverlyGeneralGrammar,”inJohnA.Hawkins(ed.),ExplainingLanguageUniversals.NewYork:Blackwell,73–101.

Bowern,Claire(2008).“SyntacticChangeandSyntacticReconstructioninGenerativeGrammar,”inGisellaFerraresiandMariaGoldbach(eds.),PrinciplesofSyntacticReconstruction.Amsterdam:JohnBenjamins,187–216.

Boyd,JeremyK.,andGoldberg,AdeleE.(2009).“InputEffectswithinaConstructionistFramework.”ModernLanguageJournal93/2:418–29.

———.(2011).“LearningWhatNottoSay:TheRoleofStatisticalPreemptionandCategorizationina-adjectiveProduction.”Language87/1:55–83.

Braine,MartinD.S.(1976).“Children'sFirstWordCombinations.”MonographsoftheSocietyforResearchinChildDevelopment41/1:1–104.

Braine,MartinD.S.,andBrooks,Patricia(1995).“Verb-argumentStructureandtheProblemofAvoidinganOvergeneralGrammar,”inMichaelTomaselloandWilliamMerriman(eds.),BeyondNamesforThings:YoungChildren'sAcquisitionofVerbs.Hillsdale,NJ:Erlbaum,353–76.

Branigan,Holly,Pickering,Martin,Liversedge,Simon,Stewart,Andrew,andUrbach,Thomas(1995).“SyntacticPriming:InvestigatingtheMentalRepresentationofLanguage.”Journalof

Page 381: The Oxford Handbook of Construction Grammar

References

Page 13 of 88

PsycholinguisticResearch24:489–506.

Branigan,Holly,Pickering,Martin,andTanaka,Mikihiro(2008).“ContributionsofAnimacytoGrammaticalFunctionAssignmentandWordOrderduringProduction.”Lingua118:172–89.

Bratman,Michael(1992).“SharedCooperativeActivity.”PhilosophicalReview101:327–41.

———.(1993).“SharedIntention.”Ethics104:97–113.

———.(1997).“IIntendthatweJ,”inRaimoTuomelaandGhitaHolmstrom-Hintikka(eds.),ContemporaryActionTheory.Vol.II.Dordrecht:Kluwer,49–63.

Brems,Lieselotte(2007).“TheGrammaticalizationofSmallSizeNouns:ReconsideringFrequencyandAnalogy.”JournalofEnglishLinguistics35:294–324.

———.(2010).“SizeNounConstructionsasCollocationallyConstrainedConstructions:LexicalandGrammaticalizedUses.”EnglishLanguageandLinguistics14:83–109.

Brenier,JasonM.,andMichaelis,LauraA.(2005).“OptimizationviaSyntacticAmalgam:Syntax-prosodyMismatchandCopulaDoubling.”CorpusLinguisticsandLinguisticTheory1:45–88.

Bresnan,Joan(1982).TheMentalRepresentationofGrammaticalRelations.Cambridge,MA:MITPress.

———.(2001).LexicalFunctionalSyntax.Oxford:Blackwell.

Bresnan,Joan,Cueni,Anna,Nikitina,Tatiana,andBaayen,R.Harald(2007).“PredictingtheDativeAlternation,”inGerlofBouma,IreneKrämer,andJoostZwarts(eds.),CognitiveFoundationsofInterpretation.Amsterdam:RoyalNetherlandsAcademyofArtsandSciences,69–94.

Bresnan,Joan,andFord,Marilyn(2010).“PredictingSyntax:ProcessingDativeConstructionsinAmericanandAustralianVarietiesofEnglish.”Language86:186–213.

Bresnan,Joan,andHay,Jennifer(2008).“GradientGrammar:AnEffectofAnimacyontheSyntaxofGiveinNewZealandandAmericanEnglish.”Lingua118:245–59.

Bresnan,Joan,andKanerva,Jonni(1989).“LocativeInversioninChichewa.”LinguisticInquiry20:1–50.

Bresnan,Joan,andKaplan,Ronald(1984).“GrammarsasMentalRepresentationsofLanguage,”inWalterKintschandPeterPolson(eds.),MethodsandTacticsinCognitiveScience.Hillsdale,NJ:Erlbaum,103–35.

Brinton,LaurelJ.,andTraugott,ElizabethCloss(2005).LexicalizationandLanguageChange.Cambridge:CambridgeUniversityPress.

Broccias,Cristiano(2011).“MotivatingtheFlexibilityofOriented–lyAdverbs,”inKlaus-UwePantherandGünterRadden(eds.),MotivationinGrammarandtheLexicon.Amsterdam:John

Page 382: The Oxford Handbook of Construction Grammar

References

Page 14 of 88

Benjamins.71–88.

Broccias,Cristiano,andHollmann,WillemB.(2007).“DoWeNeedSummaryandSequentialScanningin(Cognitive)Grammar?”CognitiveLinguistics18:487–522.

Brooks,Patricia,andTomasello,Michael(1999a).“YoungChildrenLearntoProducePassiveswithNonceVerbs.”DevelopmentalPsychology35:29–44.

———.(1999b).“HowChildrenConstraintheirArgumentStructureConstructions.”Language75:720–38.

———.(1999c).“LearningtheEnglishPassiveConstruction,”inBarbaraFox,DanJurafsky,andLauraMichaelis(eds.),CognitionandFunctioninLanguage.Stanford,CA:CLSIPublications,84–98.

Brooks,Patricia,Tomasello,Michael,Dodson,Kelly,andLewis,LawrenceB.(1999).“YoungChildren'sOvergeneralizationswithFixedTransitivityVerbs.”ChildDevelopment70:1325–37.

BROWN(OnemillionwordcorpusofAmericanEnglish).http://khnt.aksis.uib.no/icame/manuals/index.htm.

Brown,Amanda,andGullberg,Marianne(2008).“BidirectionalCrosslinguisticInfluenceinL1-L2EncodingofMannerinSpeechandGesture.”StudiesinSecondLanguageAcquisition30:225–51.

———.(2010).“ChangesinEncodingofPathofMotionafterAcquisitionofaSecondLanguage.”CognitiveLinguistics21:263–86.

Brugman,Claudia(1988).TheStoryof“Over”:Polysemy,SemanticsandtheStructureoftheLexicon.NewYork:Garland.

Brunyé,Tad,Ditman,Tali,Mahoney,Caroline,Augustyn,Jason,andTaylor,Holly(2009).“WhenYouandISharePerspectives:PronounsModulatePerspectiveTakingDuringNarrative.”PsychologicalScience20:27–32.

Bryant,John(2004).“ScalableConstructionBasedParsingandSemanticAnalysis.”ProceedingsoftheSecondInternationalworkshoponScalableNaturalLanguageUnderstanding,Boston,33–44.

———.(2008).“Best-FitConstructionalAnalysis”.Ph.D.dissertation.ComputerScienceDepartment,UniversityofCalifornia,Berkeley.

Budwig,Nancy,Narasimhan,Bhuvana,andSrivastava,Smita(2006).“InterimSolutions:TheAcquisitionofEarlyVerbConstructionsinHindi,”inEveV.ClarkandBarbaraF.Kelly(eds.),ConstructionsinAcquisition.Stanford,CA:CSLIPublications,163–83.

Buzsáki,György(2010).“NeuralSyntax:CellAssemblies,SynapsemblesandReaders.”Neuron68/3:362–85.

Page 383: The Oxford Handbook of Construction Grammar

References

Page 15 of 88

Bybee,JoanL.(1985).Morphology:AStudyintotheRelationbetweenMeaningandForm.Amsterdam:JohnBenjamins.

———.(1988a).“MorphologyasLexicalOrganization,”inMichaelHammondandMichaelNoonan(eds.),TheoreticalMorphology.SanDiego:AcademicPress,119–41.

———.(1988b).“SemanticSubstancevs.ContrastintheDevelopmentofGrammaticalMeaning.”BerkeleyLinguisticSociety14:247–64.

———.(2000).“ThePhonologyoftheLexicon:EvidencefromLexicalDiffusion,”inMichaelBarlowandSuzanneKemmer(eds.),Usage-basedModelsofLanguage.Stanford,CA:CSLIPublications,65–85.

———.(2001).PhonologyandLanguageUse.Cambridge:CambridgeUniversityPress.

———.(2002).“SequentialityastheBasisofConstituentStructure,”inTalmyGivónandBertramMalle(eds.),TheEvolutionofLanguagefromPre-Language.Amsterdam:JohnBenjamins,109–32.

———.(2003).“MechanismsofChangeinGrammaticization:TheRoleofFrequency,”inBrianD.JosephandRichardD.Janda(eds.),TheHandbookofHistoricalLinguistics.Oxford:Blackwell,602–23.

———.(2006).“FromUsagetoGrammar:TheMind'sResponsetoRepetition.”Language82:711–33.

———.(2007).FrequencyofUseandtheOrganizationofLanguage.Oxford:OxfordUniversityPress.

———.(2008).“Usage-basedGrammarandSecondLanguageAcquisition,”inPeterRobinsonandNickC.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge,216–36.

———.(2010).Language,UsageandCognition.Cambridge:CambridgeUniversityPress.

Bybee,JoanL.,andBeckner,Clay(2009).“Usage-basedTheory,”inHeikoNarrogandBerndHeine(eds.),HandbookofLinguisticAnalysis.Oxford:OxfordUniversityPress,827–55.

Bybee,JoanL.,andEddington,David(2006).“AUsage-basedApproachtoSpanishVerbsof‘Becoming’.”Language82:323–55.

Bybee,JoanL.,andHopper,Paul(eds.)(2001).FrequencyandtheEmergenceofLinguisticStructure.Amsterdam:JohnBenjamins.

Bybee,JoanL.,andPagliuca,William(1987).“TheEvolutionofFutureMeaning,”inAnnaGiacaloneRamat,OnofrioCarruba,andGuilianoBernini(eds.),PapersfromtheVIIthInternationalConferenceonHistoricalLinguistics.Amsterdam:JohnBenjamins,109–22.

Bybee,JoanL.,Perkins,Revere,andPagliuca,William(1994).TheEvolutionofGrammar:

Page 384: The Oxford Handbook of Construction Grammar

References

Page 16 of 88

Tense,AspectandModalityintheLanguagesoftheWorld.Chicago:UniversityofChicagoPress.

Bybee,JoanL.,andScheibman,Joanne(1999).“TheEffectofUsageonDegreesofConstituency:TheReductionofdon’tinEnglish.”Linguistics37:575–96.

Bybee,JoanL.,andThompson,Sandra(2000).“ThreeFrequencyEffectsinSyntax.”BerkeleyLinguisticSociety23:65–85.

Cacciari,Cristina,andGlucksberg,Sam.(1991).“UnderstandingIdiomaticExpressions:TheContributionofWordMeanings,”inGregSimpson(ed.),UnderstandingWordandSentence.TheHague:NorthHolland,217–40.

Cacciari,Cristina,andTabossi,Patrizia.(eds.).(1995).Idioms:Processing,StructureandInterpretation.Hillsdale,NJ:LawrenceErlbaum.

Cadierno,Teresa(2008).“LearningtoTalkaboutMotioninaForeignLanguage,”inPeterRobinsonandNickC.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge.239–75.

Canolty,RyanT.,Soltani,Maryam,Dalal,SarangS.,Edwards,Erik,Dronkers,NinaF.,Nagarajan,SrikantanS.,Kirsch,HeidiE.,Barbaro,NicholasM.,andKnight,RobertT.(2007).“SpatiotemporalDynamicsofWordProcessingintheHumanBrain.”FrontiersinNeuroscience1/1:185–96.

Caplan,David(2006).“AphasicDeficitsinSyntacticProcessing.”Cortex42/6:797–804.

Cappelle,Bert(2009).“ContextualCuesforParticlePlacement:Multiplicity,Motivation,Modelling,”inAlexanderBergsandGabrieleDiewald(eds.),ContextsandConstructions.Amsterdam:JohnBenjamins,145–92.

———.(2011).“Thethe…the…Construction:MeaningandReadings.”JournalofPragmatics43/1:99–117.

Cappelle,Bert,Shtyrov,Yury,andPulvermüller,Friedemann(2010).“HeatinguporCoolinguptheBrain?MEGEvidencethatPhrasalverbsareLexicalUnits.”BrainandLanguage115/3:189–201.

Caramazza,Alfonso(1997).“HowManyLevelsofProcessingareThereinLexicalAccess?”CognitiveNeuropsychology14:177–208.

Carlson,G.(2006).“TheMeaningfulBoundsofIncorporation,”inSvetlanaVogeleerandLilianeTasmowski(eds.),Non-definitenessandPlurality.Amsterdam:JohnBenjamins,35–50.

Carpenter,Bob.(1992).TheLogicofTypedFeatureStructures.Cambridge:CambridgeUniversityPress.

Casenhiser,Devin,andGoldberg,AdeleE.(2005).“FastMappingbetweenaPhrasalFormandMeaning.”DevelopmentalScience8:500–508.

Page 385: The Oxford Handbook of Construction Grammar

References

Page 17 of 88

CEEC(CorpusofEarlyEnglishCorrespondence).http://www.helsinki.fi/varieng/domains/CEEC.html.

Chafe,Wallace(1968).“IdiomaticityasanAnomalyintheChomskyanParadigm.”FoundationsofLanguage4/2:109–27.

———.(1976).“Givenness,Contrastiveness,Definiteness,SubjectsandTopics,”inCharlesN.Li(ed.),SubjectandTopic.London:AcademicPress,27–55.

———.(1977a).“CreativityinVerbalizationanditsImplicationsfortheNatureofStoredKnowledge,”inRoyFreedle(ed.),DiscourseProductionandComprehension.Norwood,NJ:Ablex,41–55.

———.(1977b).“TheRecallandVerbalizationofPastExperience,”inPeterCole(ed.),CurrentIssuesinLinguisticTheory.Bloomington:IndianaUniversityPress,215–46.

———.(ed.)(1980).ThePearStories.NewYork:Ablex.

———.(1987).“CognitiveConstraintsonInformationFlow,”inRussellTomlin(ed.),CoherenceandGroundinginDiscourse.TypologicalStudiesinLanguage,Vol.XI.Amsterdam:JohnBenjamins,21–52.

———.(1994).Discourse,ConsciousnessandTime:TheFlowandDisplacementofConsciousExperienceinSpeakingandWriting.Chicago:UniversityofChicagoPress.

Chang,Franklin(2002).“SymbolicallySpeaking:AConnectionistModelofSentenceProduction.”CognitiveScience26:609–51.

Chang,Franklin,Bock,J.Kathryn,andGoldberg,AdeleE.(2003).“CanThematicRolesLeaveTracesofTheirPlaces?”Cognition90:29–49.

———.(2006).“BecomingSyntactic.”PsychologicalReview113/2:232–72.

Chang,Franklin,Dell,GaryS.,Bock,J.KathrynandGriffin,ZenziM.(2000).“StructuralPrimingasImplicitLearning:AComparisonofModelsofSentenceProduction.”JournalofPsycholinguisticResearch29:217–29.

Chang,Nancy(2008).“ConstructingGrammar:AComputationalModeloftheEmergenceofEarlyConstructions.”Ph.D.dissertation.ComputerScienceDivision,UniversityofCaliforniaatBerkeley.

Chang,Nancy,Feldman,Jerome,Porzel,Robert,andSanders,Keith(2002).“ScalingCognitiveLinguistics:FormalismsforLanguageUnderstanding,”inProc.1stInternationalWorkshoponScalableNaturalLanguageUnderstanding.Heidelberg,Germany,10–17.

Chang,Nancy,andGurevich,Olya(2004).“Context-drivenConstructionLearning,”inProceedingsofthe26thAnnualMeetingoftheCognitiveScienceSociety.Chicago,IL,204–9.

Chang,Nancy,andMok,Eva.(2006).“AStructuredContextModelforGrammarLearning,”in

Page 386: The Oxford Handbook of Construction Grammar

References

Page 18 of 88

Proceedingsofthe2006InternationalJointConferenceonNeuralNetworks.Vancouver,BC,1604–11.

Chao,LindaL.,Haxby,JamesV.,andMartin,Alex(1999).“Attribute-basedNeuralSubstratesinTemporalCortexforPerceivingandKnowingaboutObjects.”NatureNeuroscience2/10:913–19.

Chater,Nick,andManning,ChristopherD.(2006).“ProbabilisticModelsofLanguageProcessingandAcquisition.”TrendsinCognitiveSciences10/7:335–44.

Chen,Jidong(2006).“TheAcquisitionofVerbCompoundinginMandarin,”inEveV.ClarkandBarbaraF.Kelly(eds.),ConstructionsinAcquisition.Stanford,CA:CSLIPublications,111–35.

Chenu,Florence,andJisa,Harriet(2006).“CausedMotionConstructionsandSemanticGeneralityinEarlyAcquisitionofFrench,”inEveV.ClarkandBarbaraF.Kelly(eds.),ConstructionsinAcquisition.Stanford,CA:CSLIPublications,233–61.

Cheshire,Jenny,andFox,Sue(2009).“Was/wereVariation:APerspectivefromLondon.”LanguageVariationandChange21:1–23.

Childers,JaneB.,andTomasello,Michael(2001).“TheRoleofPronounsinYoungChildren'sAcquisitionoftheEnglishTransitiveConstruction.”DevelopmentalPsychology37:739–48.

Chomsky,Noam(1957).SyntacticStructures.TheHague:MoutonandCo.

———.(1962).“TheLogicalBasisofLinguisticTheory,”inMorrisHalle(ed.),ReprintsofthePapersoftheNinthInternationalCongressofLinguists,August27–31,1962,509–74.

———.(1963).“FormalPropertiesofGrammars,”inR.DuncanLuce,RobertR.Bush,andEugeneGalanter(eds.),HandbookofMathematicalPsychology.Vol.2.NewYorkandLondon:Wiley,323–418.

———.(1965).AspectsoftheTheoryofSyntax.Cambridge,MA:MITPress.

———.(1972).StudiesonSemanticsinGenerativeGrammar.TheHague:Princeton.

———.(1973).“ConditionsonTransformations,”inStephanAndersonandPaulKiparsky(eds.),AFestschriftforMorrisHalle.NewYork:Holt,ReinhartandWinston,232–86.

———.(1977).“OnWH-movement,”inPeterCulicover,ThomasWasow,andAdrianAkmajian(eds.),FormalSyntax.NewYork:AcademicPress,71–132.

———.(1981).LecturesonGovernmentandBinding.Dordrecht.Foris.

———.(1986).Barriers.Cambridge,MA:MITPress.

———.(1993).“AMinimalistProgramforLinguisticTheory,”inKennethHaleandSamuelJayKeyser(eds.),TheViewfromBuilding20.Cambridge,MA:MITPress,1–52.

———.(1995).TheMinimalistProgram.Cambridge,MA:MITPress.

Page 387: The Oxford Handbook of Construction Grammar

References

Page 19 of 88

———.(1999).“OntheNature,UseandAcquisitionoflanguage,”inWilliamC.RitchieandTejK.Bhatia(eds.),HandbookofChildLanguageAcquisition.Cambridge:CambridgeUniversityPress,33–54.

———.(2000a).NewHorizonsintheStudyofLanguageandMind.Cambridge:CambridgeUniversityPress.

———.(2000b).“MinimalistInquiries:TheFramework,”inRogerMartin,DavidMichaels,andJuanUriagereka(eds.),StepbyStep:EssaysonMinimalistSyntaxinHonorofHowardLasnik.Cambridge,MA:MITPress,89–155.

Christiansen,MortonH.,andChater,Nick(eds.)(2001).ConnectionistPsycholinguistics.Westport,CO:Ablex.

Church,Kenneth,Gale,William,Hanks,Patrick,andHindle,Donald(1991).“UsingStatisticsinLexicalAnalysis,”inUriZernik(ed.),LexicalAcquisition:ExploitingOn-LineResourcestoBuildaLexicon.Hillsdale,NJ:Erlbaum,115–64.

Clark,EveV.(1978).“DiscoveringWhatWordscando,”inDonkaFarkas,WesleyM.Jacobsen,andKarolW.Todrys(eds.),PapersfromtheParasessionontheLexicon,ChicagoLinguisticsSocietyApril14–15,1978.Chicago:ChicagoLinguisticsSociety,34–57.

———.(1987).“ThePrincipleofContrast:AConstraintonLanguageAcquisition,”inBrianMacWhinney(ed.),MechanismsofLanguageAcquisition.Hillsdale,NJ:Erlbaum,1–33.

———.(2003).FirstLanguageAcquisition.Cambridge:CambridgeUniversityPress.

Clark,HerbertH.(1992).ArenasofLanguageUse.Chicago:UniversityofChicagoPress.

———.(1996).UsingLanguage.Cambridge:CambridgeUniversityPress.

———.(1999).“OntheOriginsofConversation.”Verbum21:147–61.

Clark,HerbertH.,andMalt,Barbara(1984).“PsychologicalConstraintsonLanguage:ACommentaryonBresnanandKaplanandonGivón,”inWalterKintschandPeterPolson,(eds.),MethodsandTacticsinCognitiveScience.Hillsdale,NJ:Erlbaum,191–214.

Clark,Lynn,andTrousdale,Graeme(2009).“TheRoleofTokenFrequencyinPhonologicalChange:EvidencefromTH-frontinginEast-centralScotland.”EnglishLanguageandLinguistics13:33–56.

CLMET(CorpusofLateModernEnglishTexts).https://perswww.kuleuven.be/~u0044428/.

Cobb,Tom(2007).TheCompleatLexicalTutor(v.4.503/06)forData-drivenLanguageLearningontheWeb.http://132.208.224.131/.

Colleman,Timothy(2009a).“TheSemanticRangeoftheDutchDoubleObjectConstruction:ACollostructionalPerspective.”ConstructionsandFrames1:190–221.

Page 388: The Oxford Handbook of Construction Grammar

References

Page 20 of 88

———.(2009b).“VerbDispositioninArgumentStructureAlternations:ACorpusStudyoftheDativeAlternationinDutch.”LanguageSciences31:593–611.

Collins,AllanM.,andLoftus,ElizabethF.(1975).“TheSpreadingActivationTheoryofSemanticProcessing.”PsychologicalReview82:407–28.

CollinsCobuildDictionaryofIdioms(2002)(2ndedn.).London:HarperCollins.

Collins,Laura,andEllis,Nick.C.(2009).“InputandSecondLanguageConstructionLearning:Frequency,FormandFunction.”ModernLanguageJournal93/2:329–35.

Comrie,Bernard(1978).“Ergativity,”inWinfredLehmann(ed.),SyntacticTypology.Austin:UniversityofTexasPress,329–94.

Connell,Louise(2007).“RepresentingObjectColourinLanguageComprehension.”Cognition102:476–85.

Copestake,Ann(2002).ImplementingTypedFeatureStructureGrammars.Stanford,CA:CSLIPublications.

Copestake,Ann,Flickinger,Daniel,Pollard,CarlJ.,andSag,IvanA.(2005).“MinimalRecursionSemantics:AnIntroduction.”ResearchonLanguageandComputation3:281–332.

Corder,S.Pit(1973).IntroducingAppliedLinguistics.NewYork:Penguin.

Corrigan,Roberta,Moravcsik,Edith,Ouali,Hamid,andWheatley,Kathleen(eds.)(2009).FormulaicLanguage.TypologicalStudiesinLanguage82.Amsterdam:JohnBenjamins,

Coulson,Seana(2007).“ElectrifyingResults:ERPDataandCognitiveLinguistics,”inMonicaGonzalez-Marquez,IreneMittelberg,SeanaCoulson,andMichaelJ.Spivey(eds.),MethodsinCognitiveLinguistics.Amsterdam:JohnBenjamins,400–23.

Coupland,Nikolas(1988).StylesofDiscourse.London:CroomHelm.

Coventry,KennyR.,andGarrod,SimonC.(2004).Saying,SeeingandActing:ThePsychologicalSemanticsofSpatialPrepositions.HoveandNewYork:PsychologyPress.

Cowie,AnthonyP.(ed.)(2001).Phraseology:Theory,AnalysisandApplications.Oxford:OxfordUniversityPress.

Cowie,AnthonyP.,andHowarth,Peter.(1996).“Phraseology—ASelectBibliography.”InternationalJournalofLexicography9/1:38–51.

Cresswell,Tim(2004).Place—AShortIntroduction.Oxford:Blackwell.

Cristofaro,Sonia(2003).Subordination.Oxford:OxfordUniversityPress.

———.(2007).“DeconstructingCategories:FinitenessinaFunctional-TypologicalPerspective,”inIrinaNikolaeva(ed.),Finiteness:TheoreticalandEmpiricalFoundations.Oxford:OxfordUniversityPress,91–114.

Page 389: The Oxford Handbook of Construction Grammar

References

Page 21 of 88

———.(2008).“AConstructionalApproachtoComplementation:EvidencefromAncientGreek.”Linguistics46:571–606.

———.(2009).“GrammaticalCategoriesandRelations:Universalityvs.Language-specificityandConstruction-specificity.”LanguageandLinguisticsCompass3/1:441–79.

———.(forthcoming).“IsSubordinationaSyntacticCategoryofParticularLanguages?”inIlonaHerlin,JyrkiKalliokoski,andLauraVisapää(eds.),ContextsofSubordination:Cognitive,TypologicalandDiscoursePerspectives.Amsterdam:JohnBenjamins.

Croft,William(1991).SyntacticCategoriesandGrammaticalRelations:TheCognitiveOrganizationofInformation.Chicago:ChicagoUniversityPress.

———.(1998).“LinguisticEvidenceandMentalRepresentations.”CognitiveLinguistics9:151–73.

———.(1999).“What(some)FunctionalistscanLearnfrom(some)Formalists,”inMichaelDarnell,EdithMoravcsik,FrederickNewmeyer,MichaelNoonan,andKathleenWheatley(eds.),FunctionalismandFormalisminLinguistics,Vol.1:GeneralPapers.Amsterdam:JohnBenjamins,85–108.

———.(2000).ExplainingLanguageChange:AnEvolutionaryApproach.London:Longman.

———.(2001).RadicalConstructionGrammar:SyntacticTheoryinTypologicalPerspective.Oxford:OxfordUniversityPress.

———.(2003a).“LexicalRulesvs.Constructions:AFalseDichotomy,”inHubertCuyckens,ThomasBerg,ReneDirven,andKlaus-UwePanther(eds.),MotivationinLanguage:StudiesinHonourofGünterRadden.Amsterdam:JohnBenjamins,49–68.

———.(2003b).TypologyandUniversals(2ndedn.).Cambridge:CambridgeUniversityPress.

———.(2005a).“LogicalandTypologicalArgumentsforRadicalConstructionGrammar,”inJan-OlaÖstmanandMirjamFried(eds.),ConstructionGrammars:CognitiveGroundingandTheoreticalExtensions.Amsterdam:JohnBenjamins,273–314.

———.(2005b).“WordClasses,PartsofSpeechandSyntacticArgumentation:CommentaryonEvansandOsada,‘Mundari:TheMythofaLanguagewithoutWordClasses’.”LinguisticTypology9:431–41.

———.(2007a).“BeyondAristotleandGradience:AReplytoAarts.”StudiesinLanguage31:409–30.

———.(2007b).“TheOriginsofGrammarintheVerbalizationofExperience.”CognitiveLinguistics18:339–82.

———.(2007c).“IntonationUnitsandGrammaticalStructureinWardamanandinCrosslinguisticPerspective.”AustralianJournalofLinguistics27:1–39.

Page 390: The Oxford Handbook of Construction Grammar

References

Page 22 of 88

———.(2009a).“MethodsforFindingLanguageUniversalsinSyntax,”inSergioScalise,ElisabettaMagni,andAntoniettaBisetto(eds.),UniversalsofLanguageToday.Berlin:Springer,145–64.

———.(2009b).“TowardaSocialCognitiveLinguistics,”inVyvyanEvansandStéphaniePourcel(eds.),NewDirectionsinCognitiveScience.Amsterdam:JohnBenjamins,395–420.

———.(2009c).“ConnectingFramesandConstructions:ACaseStudyofeatandfeed.”ConstructionsandFrames1:7–28.

———.(2010a).“PragmaticFunction,SemanticClassesandLexicalCategories:CommentaryonSmith,‘PragmaticFunctionsandLexicalCategories’.”Linguistics48:787–96.

———.(2010b).“TenUnwarrantedAssumptionsinSyntacticArgumentation,”inKasperBøyeandElisabethEngberg-Pedersen(eds.),LanguageUsageandLanguageStructure.Berlin:MoutondeGruyter,313–50.

———.(2010c).“Relativity,LinguisticVariationandLanguageUniversals.”CogniTextes4.303,http://cognitextes.revues.org/303/.

———.(2010d).“TheOriginsofGrammaticalizationintheVerbalizationofExperience.”Linguistics48:1–48.

———.(2012).Verbs:AspectandCausalStructure.Oxford:OxfordUniversityPress.

———.Inpreparation.Morphosyntax:ConstructionsoftheWorld'sLanguages.

Croft,William,Barðdal,Jóhanna,Hollmann,Willem,Sotirova,Violeta,andTaoka,Chiaki(2010).“RevisingTalmy'sTypologicalClassificationofComplexEvents,”inHansBoas(ed.),ContrastiveConstructionGrammar.Amsterdam:JohnBenjamins,201–35.

Croft,William,andCruse,D.Alan(2004).CognitiveLinguistics.Cambridge:CambridgeUniversityPress.

Croft,William,andPoole,KeithT.(2008).“InferringUniversalsfromGrammaticalVariation:MultidimensionalScalingforTypologicalAnalysis.”TheoreticalLinguistics34:1–37.

Croft,William,Shyldkrot,HavaBat-Zeev,andKemmer,Suzanne(1987).“DiachronicSemanticProcessesintheMiddleVoice,”inAnnaGiacoloneRamat,OnofrioCarruba,andGuilianoBernini(eds.),Papersfromthe7thInternationalConferenceonHistoricalLinguistics.Amsterdam:JohnBenjamins,179–92.

Culicover,PeterW.(1972).“OM-sentences.”FoundationsofLanguage8:199–236.

———.(1999).SyntacticNuts:HardCases,SyntacticTheoryandLanguageAcquisition.Oxford:OxfordUniversityPress.

Culicover,PeterW.,andJackendoff,Ray(1995).“SomethingElsefortheBindingTheory.”LinguisticInquiry26:249–75.

Page 391: The Oxford Handbook of Construction Grammar

References

Page 23 of 88

———.(1997).“SemanticSubordinationDespiteSyntacticCoordination.”LinguisticInquiry28:195–217.

———.(1999).“TheViewfromthePeriphery:TheEnglishComparativeCorrelative.”LinguisticInquiry30:543–71.

———.(2005).SimplerSyntax.Oxford:OxfordUniversityPress.

———.(2012).“Same-Except:ADomain-generalCognitiveRelationandHowLanguageExpressesIt.”Language88:305–40.

Culioli,Antoine(1990).Pourunelinguistiquedel’énonciation.Vol.1:Opérationsetreprésentations.Paris:Ophrys.

Cutting,J.Cooper,andFerreira,VictorS.(1999).“SemanticandPhonologicalInformationFlowintheProductionLexicon.”JournalofExperimentalPsychology:Learning,MemoryandCognition25/2:318–44.

Cuyckens,Hubert,Berg,Thomas,Dirven,René,andPanther,Klaus-Uwe(eds.)(2003).MotivationinLanguage:StudiesinHonorofGüntherRadden.Amsterdam:JohnBenjamins.

Dąbrowska,Eva(2000).“FromFormulatoSchema:TheAcquisitionofEnglishQuestions.”CognitiveLinguistics11:83–102.

———.(2004).Language,MindandBrain:SomePsychologicalandNeurologicalConstraintsonTheoriesofGrammar.Edinburgh:EdinburghUniversityPress.

———.(2008).“QuestionswithLongDistanceDependencies:AUsage-basedPerspective.”CognitiveLinguistics19:391–425.

———.(2009).“WordsasConstructions,”inVyvyanEvansandStéphaniePourcel(eds.),NewDirectionsinCognitiveLinguistics.Amsterdam:JohnBenjamins,201–23.

Dąbrowska,Eva,andLieven,Elena(2005).“TowardsaLexicallySpecificGrammarofChildren'sQuestionConstructions.”CognitiveLinguistics16:437–74.

Dąbrowska,Eva,Rowland,Caroline,andTheakston,Anna(2009).“TheAcquisitionofQuestionswithLong-DistanceDependencies.”CognitiveLinguistics20:571–98.

Dahaene,Stanislaus(2009).ReadingintheBrain.NewYork:Penguin.

Dahl,Östen(1985).TenseandAspectSystems.OxfordandNewYork:Blackwell.

Dalrymple,Mary,Kaplan,Ron,Maxwell,John,andZaenen,Anne(1995).FormalIssuesinLexical-FunctionalGrammar.Stanford,CA:CSLIPublications.

Dancygier,Barbara,andSweetser,Eve(1997).“TheninConditionalConstructions.”CognitiveLinguistics8:109–36.

———.(2005).MentalSpacesinGrammar:ConditionalConstructionsCambridge:Cambridge

Page 392: The Oxford Handbook of Construction Grammar

References

Page 24 of 88

UniversityPress.

Daneš,František(1964).“AThreeLevelApproachtoSyntax,”inFrantišekDaneš(ed.),TravauxdeLinguistiquedePrague.Vol.1.Tuscaloosa:UniversityofAlabamaPress,225–40.

Davies,Mark(2004−).BYU-BNC:TheBritishNationalCorpus.http://corpus.byu.edu/bnc.

Davies,Mark(2008−).TheCorpusofContemporaryAmericanEnglish(COCA):400+MillionWords,1990-present.http://www.americancorpus.org.

Deacon,Terrance(1997).TheSymbolicSpecies:TheCo-evolutionofLanguageandtheHumanBrain.London:Penguin.

Deane,Paul(1991).“LimitstoAttention:ACognitiveTheoryofIslandPhenomena.”CognitiveLinguistics2:1–63.

DeClerck,Bernard,andColleman,Timothey(forthcoming).“Fromnountointensifier:massaandmassa'sinFlemishvarietiesofDutch.”LanguageSciences.

Delbrück,Berthold(1893–1900).VergleichendeSyntaxderIndogermanischenSprachen.Strasbourg:Trübner.

Dell,GarryS.(1986).“ASpreading-activationTheoryofRetrievalinSentenceProduction.”PsychologicalReview93:283–321.

DenDikken,Marcel.(2005).“ComparativeCorrelativesComparatively.”LinguisticInquiry36:497–532.

DeRooij,Jaap(1991).“RegionaleVariatieinhetGebruikvanerIII.”TaalenTongval43:113–36.

deSaussure,Ferdinand(1916).CourseinGeneralLinguistics.NewYork:PhilosophicalLibrary.

———.(2006).CourseinGeneralLinguistics.Ed.CharlesBallyandAlbertSechehaye.Trans.RoyHarris.LaSalle,IL:OpenCourt.

DeSwart,Henriëtte(1998).“AspectShiftandCoercion.”NaturalLanguageandLinguisticTheory16:347–85.

Deuchar,Margaret,andVihman,Marilyn(2005).“ARadicalApproachtoEarlyMixedUtterances.”InternationalJournalofBilingualism9:137–57.

Deulofeu,Henri-José,andDebaisieux,Jeanne-Marie(2009).“ConstructionsandContext:WhenaConstructionConstructstheContext,”inAlexanderBergsandGabrieleDiewald(eds.),ContextsandConstructions.Amsterdam:JohnBenjamins,43–62.

DeVogelaer,Gunther(2008).“(De)grammaticalizationasaSourceforNewConstructions:TheCaseofSubjectDoublinginDutch,”inAlexanderBergsandGabrieleDiewald(eds.),ConstructionsandLanguageChange.Berlin:MoutondeGruyter,229–57.

Page 393: The Oxford Handbook of Construction Grammar

References

Page 25 of 88

Diessel,Holger(2004).TheAcquisitionofComplexSentences.Cambridge:CambridgeUniversityPress.

———.(2005).“CompetingMotivationsfortheOrderingofMainandAdverbialclauses.”Linguistics43:449–70.

———.(2007).“FrequencyEffectsinLanguageAcquisition,LanguageUse,andDiachronicChange.”NewIdeasinPsychology25:108–27.

———.(2009).“OntheRoleofFrequencyandSimilarityintheAcquisitionofSubjectandNon-subjectRelativeClauses,”inTalmyGivónandMasayoshiShibatani(eds.),SyntacticComplexity.Amsterdam:JohnBenjamins,251–76.

Diessel,Holger,andTomasello,Michael(1999).“WhyComplementClausesdonotIncludeathat-ComplementizerinEarlyChildLanguage.”BerkeleyLinguisticsSociety25:86–97.

———.(2000).“TheDevelopmentofRelativeClausesinEnglish.”CognitiveLinguistics11:131–51.

———.(2001).“TheAcquisitionofFiniteComplementClausesinEnglish:ACorpus-basedAnalysis.”CognitiveLinguistics12:1–45.

———.(2005a).“ANewLookattheAcquisitionofRelativeClauses.”Language81:1–25.

———.(2005b).“ParticlePlacementinEarlyChildLanguage:AMultifactorialAnalysis.”CorpusLinguisticsandLinguisticTheory1:89–112.

Dietrich,Rainer,Klein,Wolfgang,andNoyau,Colette(eds.)(1995).TheAcquisitionofTemporalityinaSecondLanguage.Amsterdam:JohnBenjamins.

Diewald,Gabriele(2006).“ContextTypesinGrammaticalizationasConstructions.”ConstructionsSV1–9/2006.http://www.elanguage.net/journals/index.php/constructions/index.

———.(2007).“KonstruktioneninderDiachronenSprachwissenschaft,”inKerstinFischerandAnatolStefanowitsch(eds.),Konstruktionsgrammatik:VonderAnwendungzurTheorie.Tübingen:StauffenburgVerlag,79–104.

DiSciullo,AnnaMaria,andWilliams,Edwin(1987).OntheDefinitionofWord.Cambridge,MA:MITPress.

Dodge,Ellen(2010).“ConstructionalandConceptualComposition.”Ph.D.dissertation,LinguisticsDepartment,UniversityofCaliforniaatBerkeley.

Dominey,Peter(2006).“FromHolophrasestoAbstractGrammaticalConstructions:InsightsfromSimulationStudies,”inEveV.ClarkandBarbaraF.Kelly(eds.),ConstructionsinAcquisition.Stanford,CA:CSLIPublications,137–61.

Dominey,Peter,andHoen,Michael(2006).“StructureMappingandSemanticIntegrationina

Page 394: The Oxford Handbook of Construction Grammar

References

Page 26 of 88

Construction-basedNeurolinguisticModelofSentenceProcessing.”Cortex42:476–79.

Dominey,Peter,Hoen,Michael,Blanc,Jean-Marc,andLelekov-Boissard,Taïssia(2003).“NeurologicalBasisofLanguageandSequentialCognition:EvidencefromSimulation,AphasiaandERPStudies.”BrainandLanguage8/2:207–25.

Donchin,Emanuel(1981).“Surprise!…Surprise?”Psychophysiology18:493–513.

Dong,QuangPhuc(1971).“TheApplicabilityofTransformationtoIdioms.”ChicagoLinguisticSociety7:198–205.

Doughty,Cathy,andWilliams,Jennifer(eds.)(1998).FocusonForminClassroomSecondLanguageAcquisition.Cambridge:CambridgeUniversityPress.

Dowty,David(1991).“ThematicProto-rolesandArgumentSelection.”Language67:547–619.

Dryer,MatthewS.(1997).“AreGrammaticalRelationsUniversal?”inJoanBybee,JohnHaiman,andSandraA.Thompson(eds.),EssaysonLanguageFunctionandLanguageType.Amsterdam:JohnBenjamins,115–43.

Eckert,Penelope,andRickford,JohnR.(eds.)(2002).StyleandSociolinguisticVariation.Cambridge:CambridgeUniversityPress.

Eckhoff,HannePartine(2006).“OldRussianPossessiveConstructions:AConstructionGrammarAccount.”Ph.D.dissertation,UniversityofOslo.http://folk.uio.no/hanneme/EckhoffDissertation.pdf.

Eddington,David(2000).“StressAssignmentinSpanishwithintheAnalogicalModelingofLanguage.”Language76:92–109.

Edelman,Shimon(2007).“BridgingLanguagewiththeRestofCognition,”inMonicaGonzalez-Marquez,IreneMittelberg,SeanaCoulson,andMichaelJ.Spivey(eds.),MethodsinCognitiveLinguistics.Amsterdam:JohnBenjamins,424–45.

Ehrsson,HenrikH.,Geyer,Stefan,andNaito,Eiichi(2003).“ImageryofVoluntaryMovementofFingers,ToesandTongueActivatesCorrespondingBody-partSpecificMotorRepresentations.”JournalofNeurophysiology90:3304–16.

Elio,Renee,andAnderson,JohnR.(1981).“TheEffectsofCategoryGeneralizationsandInstanceSimilarityonSchemaAbstraction.”JournalofExperimentalPsychology:HumanLearningandMemory7/6:397–417.

———.(1984).“TheEffectsofInformationOrderandLearningModeonSchemaAbstraction.”MemoryandCognition12/1:20–30.

Ellis,NickC.(1996).“SequencinginSLA:PhonologicalMemory,ChunkingandPointsofOrder.”StudiesinSecondLanguageAcquisition18/1:91–126.

———.(1998).“Emergentism,ConnectionismandLanguageLearning.”LanguageLearning

Page 395: The Oxford Handbook of Construction Grammar

References

Page 27 of 88

48/4:631–64.

———.(2001).“MemoryforLanguage,”inPeterRobinson(ed.),CognitionandSecondLanguageInstruction.Cambridge:CambridgeUniversityPress,33–68.

———.(2002).“FrequencyEffectsinLanguageProcessing:AReviewwithImplicationsforTheoriesofImplicitandExplicitLanguageAcquisition.”StudiesinSecondLanguageAcquisition24/2:143–88.

———.(2003).“Constructions,ChunkingandConnectionism:TheEmergenceofSecondLanguageStructure,”inCatherineDoughtyandMichaelH.Long(eds.),HandbookofSecondLanguageAcquisition.Oxford:Blackwell,63–103.

———.(2005).“AttheInterface:DynamicInteractionsofExplicitandImplicitLanguageKnowledge.”StudiesinSecondLanguageAcquisition27:305–52.

———.(2006a).“CognitivePerspectivesonSLA:TheAssociativeCognitiveCREED.”AILAReview19:100–121.

———.(2006b).“LanguageAcquisitionasRationalContingencyLearning.”AppliedLinguistics27/1:1–24.

———.(2006c).“SelectiveAttentionandTransferPhenomenainSLA:Contingency,CueCompetition,Salience,Interference,Overshadowing,BlockingandPerceptualLearning.”AppliedLinguistics27/2:1–31.

———.(2008a).“TheDynamicsofLanguageUse,LanguageChangeandFirstandSecondLanguageAcquisition.”ModernLanguageJournal41/3:232–49.

———.(2008b).“Usage-basedandForm-focusedLanguageAcquisition:TheAssociativeLearningofConstructions,Learned-attentionandtheLimitedL2Endstate,”inPeterRobinsonandNickC.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge,372–405.

Ellis,NickC.,andCadierno,Teresa(2009).“ConstructingaSecondlanguage.”AnnualReviewofCognitiveLinguistics7(specialsection):111–39.

Ellis,NickC.,andFerreira,Fernando,Jr.(2009a).“ConstructionLearningasaFunctionofFrequency,FrequencyDistributionandFunction.”ModernLanguageJournal93:370–86.

———.(2009b).“ConstructionsandtheirAcquisition:IslandsandtheDistinctivenessoftheirOccupancy.”AnnualReviewofCognitiveLinguistics7:187–220.

Ellis,NickC.,andLaporte,Nadine(1997).“ContextsofAcquisition:EffectsofFormalInstructionandNaturalisticExposureonSecondLanguageAcquisition,”inAnnetteM.DeGrootandJudithF.Kroll(eds.),TutorialsinBilingualism:PsycholinguisticPerspectives.Mahwah,NJ:Erlbaum,53–83.

Ellis,NickC.,andLarsenFreeman,Diane(eds.)(2006a).“LanguageEmergence:Implications

Page 396: The Oxford Handbook of Construction Grammar

References

Page 28 of 88

forAppliedLinguistics.”AppliedLinguistics27/4(wholeissue).

———.(2006b).“LanguageEmergence:ImplicationsforAppliedLinguistics:IntroductiontotheSpecialIssue.”AppliedLinguistics27/4:558–89.

———.(2009a).“ConstructingaSecondlanguage:AnalysesandComputationalSimulationsoftheEmergenceofLinguisticConstructionsfromUsage.”LanguageLearning59,Supplement1:93–128.

———.(eds.)(2009b).“LanguageasaComplexAdaptiveSystem.”LanguageLearning59,Supplement1.

Ellis,NickC.,andSagarra,Nuria(2010a).“LearnedAttentionEffectsinL2TemporalReference:TheFirstHourandtheNextEightSemesters.”LanguageLearning60,Supplement2:85–108.

———.(2010b).“TheBoundsofAdultLanguageAcquisition:BlockingandLearnedAttention.”StudiesinSecondLanguageAcquisition32/4:553–80.

———.(2011).“LearnedAttentioninAdultLanguageAcquisition:AReplicationandGeneralizationStudyandMeta-analysis.”StudiesinSecondLanguageAcquisition33/4:589–624.

Ellis,Rod(1994).TheStudyofSecondLanguageAcquisition.Oxford:OxfordUniversityPress.

———.(2001).“Introduction:InvestigatingForm-focusedInstruction.”LanguageLearning51,Supplement1:1–46.

Elman,JeffreyL.(2004).“AnAlternativeViewoftheMentalLexicon.”TrendsinCognitiveScience8:301–6.

Elman,JeffreyL.,andBates,Elizabeth(1997).“ResponsetoLetters.”Science276:1180.

Elman,JeffreyL.,Bates,Elizabeth,Johnson,MarkH.,Karmiloff-Smith,Annette,Parisi,Domencio,andPlunckett,Kim(1996).RethinkingInnateness:AConnectionistPerspectiveonDevelopment.Cambridge,MA:BradfordBooks/MITPress.

Erteschik-Shir,Nomi(1979).“DiscourseConstraintsonDativeMovement,”inTalmyGivón(ed.),SyntaxandSemantics12:DiscourseandSyntax.NewYork:AcademicPress,441–67.

———.(1998).“TheSyntax-FocusStructureInterface,”inPeterCulicoverandLouiseMcNally(eds.),SyntaxandSemantics29:TheLimitsofSyntax.NewYork:AcademicPress.211–40.

———.(2007).InformationStructure:TheSyntax–DiscourseInterface.Oxford:OxfordUniversityPress.

Evans,Nicholas(2007).“InsubordinationanditsUses,”inIrinaNikolaeva(ed.),Finiteness:TheoreticalandEmpiricalFoundations.Oxford:OxfordUniversityPress,366–431.

Evans,Nicholas,andLevinson,Stephen(2009).“TheMythofLanguageUniversals.”Brain

Page 397: The Oxford Handbook of Construction Grammar

References

Page 29 of 88

andBehavioralSciences32:429–92.

Evans,Vyvyan,andGreen,Melanie(2006).CognitiveLinguistics:AnIntroduction.Edinburgh:EdinburghUniversityPress.

Eythórsson,Thórhallur,andBarðdal,Jóhanna(2005).“ObliqueSubjects:ACommonGermanicInheritance.”Language81/4:824–81.

———.(2011).“DieKonstruktionsgrammatikunddiekomparativeMethode,”inThomasKrischandThomasLindner(eds.),IndogermanistikundLinguistikimDialog.Wiesbaden:ReichertVerlag,148–56.

Family,Neiloufar(2006).“ExplorationofSemanticSpace:TheCaseofLightVerbConstructionsinPersian.”Ph.D.dissertation,EHESS,Paris.

Fauconnier,Gilles(1975).“PragmaticScalesandLogicalStructure.”LinguisticInquiry6:353–75.

———.(1985).MentalSpaces.Cambridge,MA:MITPress.

Fauconnier,Gilles,andTurner,Mark(2002).TheWayWeThink:ConceptualBlendingandtheMind'sHiddenComplexities.NewYork:BasicBooks.

Feldman,Jerome(2006).FromMoleculetoMetaphor:ANeuralTheoryofLanguage.Cambridge,MA:MITPress.

Feldman,Jerome,andL.Gilardi,Luca(forthcoming).“ExtendingECGtoCommunities,MentalSpacesandMaps,”inHansC.Boas(ed.),ComputationalApproachestoEmbodiedConstructionGrammar.Amsterdam:JohnBenjamins.

Feldman,Jerome,andNarayanan,Srinivas(2004).“EmbodiedMeaninginaNeuralTheoryofLanguage.”BrainandLanguage89/2:385–92.

Fernando,Chitra(1996).IdiomsandIdiomaticity.Oxford:OxfordUniversityPress.

Ferraresi,Gisella,andGoldbach,Maria(eds.)(2008).PrinciplesofSyntacticReconstruction.Amsterdam:JohnBenjamins.

Ferreira,Fernanda(1991).“EffectsofLengthandSyntacticComplexityonInitiationTimesforPreparedUtterances.”JournalofMemoryandLanguage30:210–33.

———.(2000).“SyntaxinLanguageProduction:AnApproachUsingTree-adjoiningGrammars,”inLindaWheeldon(ed.),AspectsofLanguageProduction.London:PsychologyPress,291–330.

Ferreira,Fernanda,andEngelhardt,Paul(2006).“SyntaxinProduction,”inMathewTraxlerandMortonAnnGernsbacher(eds.),HandbookofPsycholinguistics.Oxford:Elsevier,61–91.

Ferreira,Fernanda,andHenderson,John(1988).“LinearizationStrategiesduringLanguageProduction.”MemoryandCognition26:88–96.

Page 398: The Oxford Handbook of Construction Grammar

References

Page 30 of 88

Ferreira,Fernanda,andSwets,Benjamin(2002).“HowIncrementalisLanguageProduction?EvidencefromtheProductionofUtterancesRequiringtheComputationofArithmeticSums.”JournalofMemoryandLanguage46:57–84.

Ferreira,Victor(1996).“IsitBettertoGiveortoDonate?SyntacticFlexibilityinLanguageProduction.”JournalofMemoryandLanguage35:724–55.

———.(2000).“EffectofAmbiguityandLexicalAvailabilityonSyntacticandLexicalProduction.”CognitivePsychology40:296–340.

Ferreira,Victor,andDell,GaryS.(2000).“EffectofAmbiguityandLexicalAvailabilityonSyntacticandLexicalProduction.”CognitivePsychology40:296–340.

Ferreira,Victor,andHumphreys,Karin(2001).“SyntacticInfluencesonLexicalandMorphologicalProcessinginLanguageProduction.”JournalofMemoryandLanguage44:52–80.

Ferreira,Victor,andSlevc,Robert(2007).“GrammaticalEncoding,”inGarethGaskell(ed.),TheOxfordHandbookofPsycholinguistics.Oxford:OxfordUniversityPress,453–69.

Ferreira,Victor,andYoshita,Hiromi(2003).“Given-newOrderingEffectsontheProductionofScrambledSentencesinJapanese.”JournalofPsycholinguisticResearch32:669–92.

Fiksdal,Susan(2008).“MetaphoricallySpeaking:GenderandClassroomDiscourse,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,419–48.

Fillmore,CharlesJ.(1968).“TheCaseforCase,”inEmmonBachandRobertT.Harms(eds.),UniversalsinLinguisticTheory.NewYork:Holt,RinehartandWinston,1–88.

———.([1974]1981).“PragmaticsandtheDescriptionofDiscourse,”inPeterCole(ed.),RadicalPragmatics.NewYork:AcademicPress,143–66.[ReprintofBerkeleyStudiesinSyntaxandSemantics,1974,Berkeley,CA:UCBerkeley].

———.(1975).“AnAlternativetoChecklistTheoriesofMeaning.”BerkeleyLinguisticSociety1:123–31.

———.(1977a).“TheCaseforCaseReopened,”inPeterCole(ed.),GrammaticalRelations.SyntaxandSemantics8.NewYork:AcademicPress,59–81.

———.(1977b).“TopicsinLexicalSemantics,”inRogerCole(ed.),CurrentIssuesinLinguisticTheory.Bloomington:IndianaUniversityPress,76–138.

———.(1982)“FrameSemantics,”inLinguisticSocietyofKorea(ed.),LinguisticsintheMorningCalm.Seoul:Hanshin,111–38.

———.(1985a).“SyntacticIntrusionsandtheNotionofGrammaticalConstruction.”BerkeleyLinguisticSociety11:73–86.

Page 399: The Oxford Handbook of Construction Grammar

References

Page 31 of 88

———..(1985b)“FramesandtheSemanticsofUnderstanding.”QuaderniediSemantica6/2:222–54.

———.(1986).“VarietiesofConditionalSentences,”inFredMarshall,AnnMiller,andZheng-ShengZhang(eds.),ProceedingsoftheThirdEasternStatesConferenceonLinguistics.Columbus,Ohio:OhioStateUniversityDepartmentofLinguistics,163–82.

———.(1988).“TheMechanismsof‘ConstructionGrammar’.”BerkeleyLinguisticSociety14:35–55.

———.(1989).“GrammaticalConstructionTheoryandtheFamiliarDichotomies,”inRainerDietrichandCarlF.Graumann(eds.),LanguageProcessinginSocialContext.NorthHolland:ElsevierPublishers,17–38.

———.(1992).“CorpusLinguisticsvs.Computer-aidedArmchairLinguistics,”inJanSvartvik(ed.),DirectionsinCorpusLinguistics.Berlin:MoutondeGruyter,35–60.

———.(1997).ConstructionGrammarLectureNotes.Ms.http://www.icsi.berkeley.edu/~kay/bcg/lec02.html.

———.(1999).“InversionandConstructionalInheritance,”inGertWebelhuth,Jean-PierreKoenig,andAndrasKathol(eds.),LexicalandConstructionalAspectsofLinguisticExplanation.Stanford,CA:CSLIPublications,113–28.

Fillmore,CharlesJ.,andBaker,Colin(2010).“AFramesApproachtoSemanticAnalysis,”inBerndHeineandHeikoNarrog(eds.),TheOxfordHandbookofLinguisticAnalysis.Oxford:OxfordUniversityPress,313–40.

Fillmore,CharlesJ.,Johnson,ChristopherR.,andPetruck,MiriamR.L.(2003).“BackgroundtoFrameNet.”InternationalJournalofLexicography16/3:235–50.

Fillmore,CharlesJ.,andKay,Paul(1993).“ConstructionGrammarCoursebook.”Ms.UniversityofCaliforniaatBerkeleyDepartmentofLinguistics.

———.(1995).“ConstructionGrammar.”Ms.,DepartmentofLinguistics,UniversityofCalifornia,Berkeley.

———.(forthcoming).ConstructionGrammar.Stanford,CA:CSLIPublication.

Fillmore,CharlesJ.,Kay,Paul,andO’Connor,MaryC.(1988).“RegularityandIdiomaticityinGrammaticalConstructions:TheCaseofletalone.”Language64/3:501–38.

Fillmore,CharlesJ.,Lee-Goldman,Russell,andRhomieux,Russell(2012).“TheFrameNetConstructicon,”inIvanA.SagandHansC.Boas(eds.),Sign-BasedConstructionGrammar.Stanford,CA:CSLIPublications,283–99.

Finkel,Raphael,andStump,GregoryS.(2007).“PrincipalPartsandMorphologicalTypology.”Morphology17:39–75.

Page 400: The Oxford Handbook of Construction Grammar

References

Page 32 of 88

———.(2009).“PrincipalPartsandDegreesofParadigmaticTransparency,”inJamesP.BlevinsandJulietteBlevins(eds.),AnalogyinGrammar.Oxford:OxfordUniversityPress,13–53.

Firbas,Jan(1962).“NotesontheFunctionoftheSentenceintheActofCommunication.”JournalofStudiesofthePhilosophicalFacultyofBrno.SeriesA10:133–48.

———.(1964).“OnDefiningtheThemeinFunctionalSentenceAnalysis,”inFrantišekDaneš(ed.),TravauxdeLinguistiquedePrague.Vol.1.Tuscaloosa:UniversityofAlabamaPress,267–80.

Fischer,Olga(2007).MorphosyntacticChange:FunctionalandFormalPerspectives.Oxford:OxfordUniversityPress.

Fisher,Cynthia,Hall,DavidG.,Rakowitz,Susan,andGleitman,Lila(1994).“WhenItIsBettertoReceiveThantoGive—SyntacticandConceptualConstraintsonVocabularyGrowth.”Lingua92:333–75.

Fodor,JeroldA.(1983).TheModularityofMind:AnEssayonFacultyPsychology.Cambridge,MA:MITPress.

Fodor,JeroldA.,andPylyshyn,Zenon(1988).“ConnectionismandCognitiveArchitecture:ACriticalAnalysis.”Cognition28:3–71.

Forrest,LindaB.(1994).“TheDe-transitiveClausesinBellaCoola:Passivevs.Inverse,”inTalmyGivón(ed.),VoiceandInversion.Amsterdam:JohnBenjamins,147–68.

Fortuin,Egbert,andBoogaart,Ronny(2009).“ImperativeasConditional:FromConstructionaltoCompositionalSemantics.”CognitiveLinguistics20/4:641–73.

Fox,Anthony(1995).LinguisticReconstruction:AnIntroductiontoTheoryandMethod.Oxford:OxfordUniversityPress.

Fraser,Bruce.(1970).“IdiomswithinaTransformationalGrammar.”FoundationsofLanguage6/1:22–42.

Fried,Mirjam(2003).“WordOrder,”inJan-OlaÖstmanandJanBlommaert(eds.),HandbookofPragmatics.Amsterdam:JohnBenjamins,1–13.

———.(2006).“AgentBack-groundingasaFunctionalDomain:ReflexivizationandPassivizationinCzechandRussian,”inBenjaminLyngfeltandTorgrimSolstad(eds.),DemotingtheAgent:Passive,MiddleandOtherVoicePhenomena.Amsterdam:JohnBenjamins,83–109.

———.(2007a).“AFrameSemanticAccountofMorphosemanticChange:TheCaseofOldCzechvěřící,”inDagmarDivjakandAgataKochanska(eds.),CognitivePathsintotheSlavicDomain.BerlinandNewYork:MoutondeGruyter,283–315.

———.(2007b).“ConstructingGrammaticalMeaning:IsomorphismandPolysemyinCzech

Page 401: The Oxford Handbook of Construction Grammar

References

Page 33 of 88

Reflexivization.”StudiesinLanguage31/4:721–64.

———.(2008).”ConstructionsandConstructs:MappingaShiftbetweenPredicationandAttribution,”inAlexanderBergsandGabrieleDiewald(eds.),ConstructionsandLanguageChange.BerlinandNewYork:WalterdeGruyter,47–79.

———.(2009a).“RepresentingContextualFactorsinLanguageChange:BetweenFramesandConstructions,”inAlexanderBergsandGabrieleDiewald(eds.),ContextsandConstructions.Amsterdam:JohnBenjamins,63–94.

———.(2009b).“ConstructionGrammarasaToolforDiachronicAnalysis.”ConstructionsandFrames1/2:261–91.

———.(2009c).“Plainvs.SituatedPossessioninaNetworkofGrammaticalConstructions,”inWilliamMcGregor(ed.),ExpressionofPossession.BerlinandNewYork:WalterdeGruyter,213–48.

———.(2010a).“GrammaticalizationandLexicalizationEffectsinParticipialMorphology:AConstructionGrammarApproachtoLanguageChange,”inAnVanlinden,Jean-ChristopheVerstraete,andKristinDavidse(eds.),FormalEvidenceinGrammaticalizationResearch.TypologicalStudiesinLanguage94.Amsterdam:JohnBenjamins,191–224.

———.(2010b).“GrammarandInteraction:NewDirectionsinConstructionalResearch.”ConstructionsandFrames2/2:125–33.

———.(forthcoming).“ConstructionGrammar,”inArtemisAlexiadouandTiborKiss(eds.),HandbookofSyntax(2ndedn.)BerlinandNewYork:WalterdeGruyter.

Fried,Mirjam,andÖstman,Jan-Ola(2004a).“ConstructionGrammar:AThumbnailSketch,”inMirjamFriedandJan-OlaÖstman(eds.),ConstructionGrammarinaCross-languagePerspective.Amsterdam:JohnBenjamins,11–86.

———.(eds.)(2004b).ConstructionGrammarinaCross-languagePerspective.Amsterdam:JohnBenjamins.

———.(2005).“ConstructionGrammarandSpokenLanguage:TheCaseofPragmaticParticles.”JournalofPragmatics37/11:1752–78.

Friederici,AngelaD.(2002).“TowardsaNeuralBasisofAuditorySentenceProcessing.”TrendsinCognitiveSciences6/2:78–84.

Friederici,AngelaD.,Pfeifer,Erdmut,andHahne,Anja(1993).“Event-relatedBrainPotentialsduringNaturalSpeechProcessing:EffectsofSemantic,MorphologicalandSyntacticViolations.”CognitiveBrainResearch1/3:183–92.

Friederici,AngelaD.,Wang,Yunhua,Herrmann,ChristophS.,Maess,Burkhard,andOertel,Ulrich(2000).“LocalizationofEarlySyntacticProcessesinFrontalandTemporalCorticalAreas:AMagnetoencephalographicStudy.”HumanBrainMapping11/1:1–11.

Page 402: The Oxford Handbook of Construction Grammar

References

Page 34 of 88

Fries,CharlesC.(1952).TheStructureofEnglish.NewYork:Harcourt,BraceandCo.

Fries,CharlesC.,Lado,Robert,andtheStaffoftheMichiganEnglishLanguageInstitute(1958).EnglishPatternPractices:EstablishingthePatternsasHabits.AnnArbor:UniversityofMichiganPress.

Fujii,Seiko(2004).“Lexically(Un)filledConstructionalSchemesandConstructionTypes:TheCaseofJapaneseModalConditionalConstructions,”inMirjamFriedandJan-OlaÖstman(eds.),ConstructionGrammarinaCross-languagePerspective.Amsterdam:JohnBenjamins,121–56.

Gallese,Vittorio,Faddiga,Luciano,Fogassi,Leonardo,andRizzolatti,Giacomo(1996).“ActionRecognitioninthePremotorCortex.”Brain119:593–609.

Garagnani,Max,Shtyrov,Yury,andPulvermüller,Friedemann(2009).“EffectsofAttentiononWhatisKnownandWhatisnot:MEGEvidenceforFunctionallyDiscreteMemoryCircuits.”FrontiersinHumanNeuroscience3/10:1–12.

Garnsey,SusanM.,Pearlmutter,NealJ.,Myers,Elizabeth,andLotocky,MelanieA.(1997).“TheContributionsofVerbBiasandPlausibilitytotheComprehensionofTemporarilyAmbiguousSentences.”JournalofMemoryandLanguage37:58–93.

Garrett,Merrill(1975).“TheAnalysisofSentenceProduction,”inGordonH.Bower(ed.),ThePsychologyofLearningandMotivation.Vol.9.NewYork:AcademicPress,133–77.

———.(1980).“LevelsofProcessinginSentenceProduction,”inBrianButterworth(ed.),LanguageProduction.Vol.1.London:AcademicPress,177–220.

Gass,Susan,andSelinker,Larry(eds.)(1983).LanguageTransferinLanguageLearning.Rowley,MA:NewburyHouse.

Gawron,JeanMark(1986).“SituationsandPrepositions.”LinguisticsandPhilosophy9:327–82.

Gazdar,Gerald,Klein,Ewan,Pullum,Geoffrey,andSag,IvanA.(1985).GeneralizedPhraseStructureGrammar.Cambridge,MA:HarvardUniversityPress.

Geeraerts,Dirk,Grondelaers,Stefan,andBakema,Peter(1994).TheStructureofLexicalVariation:Meaning,NamingandContext.Berlin:MoutondeGruyter.

Geeraerts,Dirk,Grondelaers,Stefan,andSpeelman,Dirk(1999).ConvergentieenDivergentieindeNederlandseWoordenschat:EenOnderzoeknaarKleding-enVoetbaltermen.Amsterdam:MeertensInstituut.

Gelman,Andrew,andHill,Jennifer(2008).DataAnalysisUsingRegressionandMultilevel/HierarchicalModels.2ndprintingwithcorrections.Cambridge:CambridgeUniversityPress.

Gertner,Yael,Fisher,Cynthia,andEisengart,Julie(2006).“LearningWordsandRules:AbstractKnowledgeofWordOrderinEarlySentenceComprehension.”PsychologicalScience

Page 403: The Oxford Handbook of Construction Grammar

References

Page 35 of 88

17/8:684–91.

Ghomeshi,Jila,Jackendoff,Ray,Rosen,Nicole,andRussell,Kevin(2004).“ContrastiveFocusReduplicationinEnglish(thesalad-saladPaper).”NaturalLanguageandLinguisticTheory22:307–57.

Gibbs,RaymondW.,andNayak,Nandini.(1989).“PsycholinguisticStudiesontheSyntacticBehaviorofIdioms.”CognitivePsychology21/1:100–38.

Gibbs,RaymondW.,Nayak,Nandini,Bolton,John,andKeppel,Melissa(1989).“Speakers’AssumptionsabouttheLexicalFlexibilityofIdioms.”MemoryandCognition17/1:58–68.

Gildea,Spike(1997).“EvolutionofGrammaticalRelationsinCariban:HowFunctionalMotivationPrecedesSyntacticChange,”inTalmyGivón(ed.),GrammaticalRelations:AFunctionalistPerspective.Amsterdam:JohnBenjamins,155–98.

———.(ed.)(1998).OnReconstructingGrammar:ComparativeCaribanMorphosyntax.Oxford:OxfordUniversityPress.

———.(2000).“OntheGenesisoftheVerbPhraseinCaribanLanguages:DiversitythroughReanalysis,”inSpikeGildea(ed.),ReconstructingGrammar:ComparativeLinguisticsandGrammaticalizationTheory.Amsterdam:JohnBenjamins,65–105.

Giles,Howard,Mulac,Anthony,Bradac,JamesJ.,andJohnson,Patricia(1987).“SpeechAccommodationTheory:TheFirstDecadeandBeyond,”inMargaretL.McLaughlin(ed.),CommunicationYearbook10.BeverlyHills,CA:Sage,13–48.

Gilquin,Gaëtanelle(2006).“TheVerbSlotinCausativeConstructions.FindingtheBestFit.”ConstructionsSV1–3/2006.http://www.elanguage.net/journals/index.php/constructions/index.

Gilquin,Gaëtanelle,andGries,StefanTh.(2009).“CorporaandExperimentalMethods:AState-of-the-artReview.”CorpusLinguisticsandLinguisticTheory5:1–26.

Ginzburg,Jonathan,andSag,IvanA.(2000).InterrogativeInvestigations:TheForm,MeaningandUseofEnglishInterrogatives.Stanford,CA:CSLIPublications.

Givón,Talmy(1979).OnUnderstandingGrammar.NewYorkandSanFrancisco:AcademicPress.

———.(1991).“IsomorphismintheGrammaticalCode:CognitiveandBiologicalConsiderations.”StudiesinLanguage1:85–114.

Gleitman,Lila(1994).“TheStructuralSourcesofVerbMeanings,”inPaulBloom(ed.),LanguageAcquisition:CoreReadings.Cambridge,MA:MITPress,174–221.

Gleitman,Lila,January,David,Nappa,Rebecca,andTrueswell,John(2007).“OntheGiveandTakebetweenEventApprehensionandUtteranceFormulation.”JournalofMemoryandLanguage57:544–69.

Page 404: The Oxford Handbook of Construction Grammar

References

Page 36 of 88

Glenberg,ArthurM.,andKaschak,MichaelP.(2002).“GroundingLanguageinAction.”PsychonomicBulletinandReview9/3:558–65.

Glenberg,ArthurM.,andRobertson,David(2000).“SymbolGroundingandMeaning:AComparisonofHigh-dimensionalandEmbodiedTheoriesofMeaning.”JournalofMemoryandLanguage43/3:379–401.

Glenberg,ArthurM.,Sato,Marc,andCattaneo,Luigi(2008).“Use-inducedMotorPlasticityAffectstheProcessingofAbstractandConcreteLanguage.”CurrentBiology18/7:R290–91.

Glucksberg,Sam.(1993).“IdiomMeaningsandAllusionalContent,”inCristinaCacciariandPatriziaTabossi(eds.),Idioms:Processing,StructureandInterpretation.Hillsdale,NJ:Erlbaum,3–26.

Goddard,Cliff(2000).“AProblemofDefinition,”inYaelRavinandClaudiaLeacock(eds.),Polysemy.Oxford:OxfordUniversityPress,129–51.

Goldberg,AdeleE.(1989).“AUnifiedAccountoftheSemanticsoftheDitransitive.”BerkeleyLinguisticSociety15:79–90.

———.(1992).“TheInherentSemanticsofArgumentStructure:TheCaseoftheEnglishDitransitiveConstruction.”CognitiveLinguistics3:37–74.

———.(1995).Constructions:AConstructionGrammarApproachtoArgumentStructure.Chicago:UniversityofChicagoPress.

———.(1996).“JackendoffandConstruction-BasedGrammar.”CognitiveLinguistics7:3–20.

———.(1997).“RelationshipsbetweenVerbandConstruction,”inMarjolijnVerspoorandEveSweetser(eds.),LexiconandGrammar.Amsterdam:JohnBenjamins,383–98.

———.(1999).“TheEmergenceoftheSemanticsofArgumentStructureConstructions,”inBrianMacWhinney(ed.),TheEmergenceofLanguage.Mahwah,NJ:LawrenceErlbaum,197–212.

———.(2000).“PatientArgumentsofCausativeVerbscanbeOmitted:TheRoleofInformationStructureinArgumentDistribution.”LanguageSciences34:503–24.

———.(2002).“SurfaceGeneralizations:AnAlternativetoAlternations.”CognitiveLinguistics13:327–56.

———.(2003).“Constructions:ANewTheoreticalApproachtoLanguage.”TrendsinCognitiveSciences7/5:219–24.

———.(2004a).“ArgumentRealization:TheRoleofConstructions,LexicalSemanticsandDiscourseFactors,”inJan-OlaÖstmanandMirjamFried(eds.),ConstructionGrammars:CognitiveGroundingandTheoreticalExtensions.Amsterdam:JohnBenjamins,17–44.

———.(2004b).“ButdoweNeedUniversalGrammar?CommentonLidzetal.(2003).”

Page 405: The Oxford Handbook of Construction Grammar

References

Page 37 of 88

Cognition94:77–84.

———.(2006a).ConstructionsatWork:TheNatureofGeneralizationinLanguage.Oxford:OxfordUniversityPress.

———.(2006b).“LearningLinguisticPatterns,”inArthurMarkmanandBrianRoss(eds.),CategoriesinUse.PsychologyofLearningandMotivation47.SanDiego:AcademicPress/Elsevier,33–63.

———.(2010).“Verbs,FramesandConstructions”.InMalkaR.Hovav,EditDoron,andIvySichel(eds.),Syntax,LexicalSemanticsandEventStructure.Oxford:OxfordUniversityPress.

———.(2011a).“CorpusEvidenceoftheViabilityofStatisticalPreemption.”CognitiveLinguistics22/1:131–54.

———.(ed.)(2011b).CognitiveLinguistics,Vols.I–V.Oxford:Routledge.

Goldberg,AdeleE.,andBencini,Giulia(2005).“SupportfromProcessingforaConstructionalApproachtoGrammar,”inAndreaTyler,MariTakada,YiyoungKim,andDianaMarinova(eds.),LanguageinUse:CognitiveandDiscoursePerspectivesonLanguageandLanguageLearning.Washington,DC:GeorgetownUniversityPress,3–18.

Goldberg,AdeleE.,andCasenhiser,Devin(2006).“EnglishConstructions,”inBasAartsandAprilMcMahon(eds.),TheHandbookofEnglishLinguistics.Malden,MA:Wiley-Blackwell,343–55.

Goldberg,AdeleE.,Casenhiser,Devin,andSethuraman,Nitya(2004).“LearningArgumentStructureGeneralizations.”CognitiveLinguistics15:289–316.

———.(2005).“TheRoleofPredictioninConstruction-learning.”JournalofChildLanguage32:407–26.

Goldberg,AdeleE.,andDelGiudice,Alex(2005).“Subject-auxiliaryInversion:ANaturalCategory.”LinguisticReview22:411–28.

Goldberg,AdeleE.,andJackendoff,Ray(2004).“TheEnglishResultativeasaFamilyofConstructions.”Language80:532–68.

Goldschneider,JenniferM.,andDeKeyser,Robert(2001).“Explainingthe‘NaturalOrderofL2MorphemeAcquisition’inEnglish:AMeta-analysisofMultipleDeterminants.”LanguageLearning51:1–50.

Goldwater,MicahB.,andMarkman,ArthurB.(2009).“ConstructionalSourcesofImplicitAgentsinSentenceComprehension.”CognitiveLinguistics20:675–702.

González,Julio,Barros-Loscertales,Alfonso,Pulvermüller,Friedemann,Meseguer,Vanessa,Sanjuán,Ana,Belloch,Vicente,andÁvila,César(2006).“ReadingCinnamonActivatesOlfactoryBrainRegions.”Neuroimage32/2:906–12.

Page 406: The Oxford Handbook of Construction Grammar

References

Page 38 of 88

Gonzalez-Marquez,Mónica,Mittleberg,Irene,Coulson,Seana,andSpivey,Michael(eds.)(2007).MethodsinCognitiveLinguistics.Amsterdam:JohnBenjamins.

Gonzálvez-García,Francisco(2006).“PassivesWithoutActives:EvidencefromVerblessComplementClausesinSpanish.”ConstructionsSV1–5/2006.http://www.elanguage.net/journals/index.php/constructions/index.

———.(2010).“ContrastingConstructionsinEnglishandSpanish:TheInfluenceofSemantic,PragmaticandDiscoursefactors,”inHansC.Boas(ed.),ContrastiveStudiesinConstructionGrammar.Amsterdam:JohnBenjamins,43–86.

Gonzálvez-García,Francisco,andButler,Chris(2006).“MappingCognitive-functionalSpace.”AnnualReviewofCognitiveLinguistics4:39–96.

Goschler,Juliana,andStefanowitsch,Anatol(2010).“PfadundBewegungimGesprochenenDeutsch:EinKollostruktionalerAnsatz,”inThomasStolz,EstherRuigendijk,andJürgenTrabant(eds.),LinguistikimNordwesten.Bochum:Brockmeyer,103–15.

Granger,Sylvianne,andMeunier,Fanny(eds.)(2008).Phraseology:AnInterdisciplinaryPerspective.Amsterdam:JohnBenjamins.

Gras,Pedro(2010).“NonPrototypicalSentence-levelConstructions:ConstructionGrammarMeetsInsubordination.”Talkgivenatthe6thInternationalConferenceofConstructionGrammarICCG6,CharlesUniversity,Prague,September3–5,2010.

Green,GeorgiaM.(1974).SemanticsandSyntacticRegularity.Bloomington:IndianaUniversityPress.

Greenberg,JosephH.(1990).“SomeUniversalsofGrammarwithParticularReferencetotheOrderofMeaningfulElements,”inKeithDenningandSuzanneKemmer(eds.),OnLanguage:SelectedWritingsofJosephH.Greenberg.Stanford,CA:StanfordUniversityPress,40–70.

Grewe,Tanja,Bornkessel,Ina,Zysset,Stefan,Wiese,Richard,vonCramon,D.Yves,andSchlesewsky,Matthias(2006).“LinguisticProminenceandBroca'sArea:TheInfluenceofAnimacyasaLinearizationPrinciple.”Neuroimage32/3:1395–402.

Gries,StefanTh.(1999).“ParticleMovement:ACognitiveandFunctionalapproach.”CognitiveLinguistics10:105–45.

———.(2001).“AMultifactorialAnalysisofSyntacticVariation:ParticleMovementRevisited.”JournalofQuantitativeLinguistics8:33–50.

———.(2002).“TheInfluenceofProcessingonGrammaticalVariation:ParticlePlacementinEnglish,”inNicoleDehé,RayJackendoff,AndrewMcIntyre,andSilkeUrban(eds.),Verb-particleExplorations.BerlinandNewYork:MoutondeGruyter,269–88.

———.(2003a).MultifactorialAnalysisinCorpusLinguistics:AStudyofParticlePlacement.LondonandNewYork:ContinuumPress.

Page 407: The Oxford Handbook of Construction Grammar

References

Page 39 of 88

———.(2003b).“TowardsaCorpus-basedIdentificationofPrototypicalInstancesofConstructions.”AnnualReviewofCognitiveLinguistics1:1–27.

———.(2005).“SyntacticPriming:ACorpus-basedApproach.”JournalofPsycholinguisticResearch34/4:365–99.

———.(2007).“Coll.analysis3.2a.RScripttoComputeCollostructionalStrengths.”http://tinyurl.com/collostructions.

Gries,StefanTh.,Hampe,Beate,andSchönefeld,Doris(2005).“ConvergingEvidence:BringingTogetherExperimentalandCorpusDataontheAssociationofVerbsandConstructions.”CognitiveLinguistics16/4:635–76.

———.(2010).“ConvergingEvidenceII:MoreontheAssociationofVerbsandConstructions,”inJohnNewmanandSallyRice(eds.),EmpiricalandExperimentalMethodsinCognitive/FunctionalResearch.Stanford,CA:CSLIPublications,59–72.

Gries,StefanTh.,andHilpert,Martin(2008).“TheIdentificationofStagesinDiachronicData:Variability-basedNeighbourClustering.”Corpora3/1:59–81.

———.(2010).“ModelingDiachronicChangeintheThirdPersonSingular:AMultifactorial,Verb-andAuthor-specificExploratoryApproach.”EnglishLanguageandLinguistics14/3:293–320.

Gries,StefanTh.,andStefanowitsch,Anatol(2004a).“ExtendingCollostructionalAnalysis:ACorpus-basedPerspectiveon‘Alternations’.”InternationalJournalofCorpusLinguistics9/1:97–129.

———.(2004b).“Co-varyingCollexemesintheInto-causative,”inMichelAchardandSuzanneKemmer(eds.),Language,CultureandMind.Stanford,CA:CSLIPublications,225–36.

———.(2010).“ClusterAnalysisandtheIdentificationofCollexemeClasses,”inJohnNewmanandSallyRice(eds.),EmpiricalandExperimentalMethodsinCognitive/FunctionalResearch.Stanford,CA:CSLIPublications,73–90.

Gries,StefanTh.,andWulff,Stefanie(2005).“DoForeignLanguageLearnersAlsoHaveConstructions?EvidencefromPriming,SortingandCorpora.”AnnualReviewofCognitiveLinguistics3:182–200.

———.(2009).“PsycholinguisticandCorpus-linguisticEvidenceforL2constructions.”AnnualReviewofCognitiveLinguistics7/1:163–86.

Grimshaw,Jane.(1990).ArgumentStructure.Cambridge,MA:TheMITPress.

Grodzinsky,Yosef,andFriederici,AngelaD.(2006).“NeuroimagingofSyntaxandSyntacticProcessing.”CurrentOpinioninNeurobiology16/2:240–46.

Grondelaers,Stefan(2000).“DeDistributievanNietAnaforischerBuitendeEersteZinplaats:Sociolexicologische,FunctioneleenPsycholinguïstischeAspectedvaner'sStatusals

Page 408: The Oxford Handbook of Construction Grammar

References

Page 40 of 88

PresentatiefSignaal.”Ph.D.dissertation,Leuven,KatholiekeUniversiteitLeuven.

Grondelaers,Stefan,vanAken,Hilde,Speelman,Dirk,andGeeraerts,Dirk(2001).“InhoudswoordenenPrepositiesalsStandaardiserings-indicatoren.DeDiachroneemSynchroneStatusvanhetBelgischeNederlands.”NederlandseTaalkunde6:179–202.

Grondelaers,Stefan,andBrysbaert,Marc(1996).“DeDistributievanhetPresentatieveerBuitendeEersteZinsplaats.”NederlandseTaalkunde1:280–305.

Grondelaers,Stefan,Speelman,Dirk,andGeeraerts,Dirk(2007).“ACaseforaCognitiveCorpusLinguistics,”inMonicaGonzalez-Marquez,IreneMittelberg,SeanaCoulson,andMichaelJ.Spivey(eds.),MethodsinCognitiveLinguistics.Amsterdam:JohnBenjamins,149–69.

———.(2008).“NationalVariationintheUseofer‘there’.RegionalandDiachronicConstraintsonCognitiveExplanations,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,153–203.

Gundel,Jeanette,andFretheim,Thorstein(2009).“InformationStructure,”inFrankBrisard,Jan-OlaÖstman,andJefVerscheuren(eds.),Grammar,MeaningandPragmatics.Amsterdam:JohnBenjamins,146–60.

Günthner,Susanne(2008).“Die‘dieSache/dasDingist’-KonstruktionimGesprochenenDeutsch—eineInteraktionalePerspektiveaufKonstruktionenimGebrauch,”inAnatolStefanowitschandKerstinFischer(eds.),KonstruktionsgrammatikII—VonderKonstruktionzurGrammatik.Tübingen:StauffenburgVerlag,157–78.

Günthner,Susanne,andImo,Wolfgang(eds.)(2007).KonstruktioneninderInteraktion.Berlin:MoutondeGruyter.

Gurevich,Olga(2006).“ConstructionalMorphology:TheGeorgianVersion.”Ph.D.dissertation,StanfordUniversity,Stanford,CA.

Gurevich,Olga,Johnson,MattandGoldberg,AdeleE.(2010).“IncidentalVerbatimMemoryforLanguage”.LanguageandCognition2/1:45–78.

Guy,GregoryR.,andBayley,Robert(1995).“OntheChoiceofRelativePronounsinEnglish.”AmericanSpeech70:148–62.

Haberzettl,Stefanie(2007).“KonstruktionenimZweitspracherwerb,”inKerstinFischerandAnatolStefanowitsch(eds.),Konstruktionsgrammatik:VonderAnwendungzurTheorie.Tübingen:StauffenburgVerlag,55–78.

Haegeman,Liliane.(1991).IntroductiontoGovernmentandBindingTheory.Cambridge,MA:BasilBlackwell.

Haeseryn,Walter,Romijn,Kirsten,Geerts,Guido,deRooij,Jaap,andvanderToorn,MaartenC.(1977).AlgemeneNederlandseSpraakkunst(2ndedn.).Groningen/Deurne:MartinusNijhoff

Page 409: The Oxford Handbook of Construction Grammar

References

Page 41 of 88

Uitgevers/WoltersPlantyn.

Hagoort,Peter(2005).“OnBroca,BrainandBinding:ANewFramework.”TrendsinCognitiveSciences9/9:416–23.

Hagoort,Peter,Brown,Colin,andGroothusen,Jolanda(1993).“TheSyntacticPositiveShift(SPS)asanERP-MeasureofSyntacticProcessing.”LanguageandCognitiveProcesses8/4:439–83.

Haiman,John.(1980).“DictionariesandEncyclopedias.”Lingua50:329–57.

———.(1983).“IconicandEconomicMotivation.”Language59:781–819.

———.(1985a).IconicityinSyntax.Cambridge:CambridgeUniversityPress.

———.(1985b).NaturalSyntax:IconicityandErosion.Cambridge:CambridgeUniversityPress.

Hakuta,Kenji(1974).“PrefabricatedPatternsandtheEmergenceofStructureinSecondLanguageAcquisition.”LanguageLearning24:287–98.

Hale,Ken,andKeyser,Jay(1997).“OntheComplexNatureofSimplePredicators,”inAlexAlsina,JoanBresnan,andPeterSells(eds.),ComplexPredicates.Stanford,CA:CSLIPublications,29–65.

Halliday,M.A.K.(1967).“NotesonTransitivityandThemeinEnglish:Part2.”JournalofLinguistics3:199–244.

Hampe,Beate,andSchönefeld,Doris(2006).“SyntacticLeapsorLexicalVariation?—Moreon‘CreativeSyntax’,”inStefanTh.GriesandAnatolStefanowitsch(eds.),CorporainCognitiveLinguistics:Corpus-BasedApproachestoSyntaxandLexis.BerlinandNewYork:MoutondeGruyter,127–58.

Happé,Francesca(1994).Autism:AnIntroductiontoPsychologicalTheory.London:UCLPress.

Hare,MaryL.,andGoldberg,AdeleE.(1999).“StructuralPriming:PurelySyntactic?”inMartinHahnandScottC.Stones(eds.),Proceedingsofthe21stAnnualMeetingoftheCognitiveScienceSociety.Mahwah,NJ:LawrenceErlbaum,208–11.

Hare,MaryL.,McRae,Ken,andElman,JeffreyL.(2003).“SenseandStructure:MeaningasaDeterminantofVerbSubcategorizationPreferences.”JournalofMemoryandLanguage48:281–303.

Harley,Trevor(2001).ThePsychologyofLanguage:FromDatatoTheory(2ndedn.).Hove,UK:PsychologyPress.

Harris,AliceC.(2003).“Cross-linguisticPerspectivesonSyntacticChange,”inBrianD.JosephandRichardD.Janda(eds.),TheHandbookofHistoricalLinguistics.Malden,MA:Blackwell,

Page 410: The Oxford Handbook of Construction Grammar

References

Page 42 of 88

529–51.

———.(2008).“ReconstructioninSyntax:ReconstructionofPatterns,”inGisellaFerraresiandMariaGoldbach(eds.),PrinciplesofSyntacticReconstruction.Amsterdam:JohnBenjamins,73–95.

———.(2009).“ExuberantExponenceinBatsbi.”NaturalLanguageandLinguisticTheory27:267–303.

Harris,AliceC.,andCampbell,Lyle(1995).HistoricalSyntaxinCross-LinguisticPerspective.Cambridge:CambridgeUniversityPress.

Harrison,S.P.(2003).“OntheLimitsoftheComparativeMethod,”inBrianD.JosephandRichardD.Janda(eds.),TheHandbookofHistoricalLinguistics.Oxford:Blackwell,343–68.

Hasegawa,Yoko,Lee-Goldman,Russell,OharaHirose,Kyoko,Fujii,Seiko,andFillmore,CharlesJ.(2010).“OnExpressingMeasurementandComparisoninEnglishandJapanese,”inHansC.Boas(ed.),ContrastiveStudiesinConstructionGrammar.Amsterdam:JohnBenjamins,169–200.

Haspelmath,Martin(1997).IndefinitePronouns.Oxford:OxfordUniversityPress.

———.(2003).“TheGeometryofGrammaticalMeaning:SemanticMapsandCross-linguisticComparison,”inMichaelTomasello(ed.),TheNewPsychologyofLanguage.Vol.2.Mahwah,NJ:LawrenceErlbaumAssociates,211–42.

———.(2004).“Doeslinguisticexplanationpresupposelinguisticdescription?”StudiesinLanguage28/3:554–79.

———.(2006).“ExplainingAlienabilityContrastsinAdnominalPossession:Economyvs.Iconicity.”Paperpresentedatthe2ndConferenceontheSyntaxoftheWorld'sLanguages,LancasterUniversity.Handoutavailablehttp://email.eva.mpg.de/~haspelmt/2006swl.pdf,accessedMarch12,2010.

———.(2007).“Pre-establishedCategoriesDon’tExist:ConsequencesforLanguageDescriptionandTypology.”LinguisticTypology11:119–32.

———.(2008).“ParametricversusFunctionalExplanationofSyntacticUniversals,”inTheresaBiberauer(ed.),TheLimitsofSyntacticVariation.Amsterdam:Benjamins,75–107.

———.(2010a).“ComparativeConceptsandDescriptiveCategoriesinCrosslinguisticStudies.”Language86/3:663–87.

———.(2010b).“Framework-freeGrammaticalTheory,”inBerndHeineandHeikoNarrog(eds.),TheOxfordHandbookofLinguisticAnalysis.Oxford:OxfordUniversityPress,341–65.

———.(2011).“TheIndeterminacyofWordSegmentationandtheNatureofMorphologyandSyntax.”FoliaLinguistica45:31–80.

Page 411: The Oxford Handbook of Construction Grammar

References

Page 43 of 88

Hauser,MarkD.,Chomsky,Noam,andFitch,W.Tecumseh(2002).“TheFacultyofLanguage:Whatisit,WhohasitandHowdiditEvolve?”Science298:1569–79.

Hawkins,JohnA.(1994).APerformanceTheoryofOrderandConstituency.Cambridge:CambridgeUniversityPress.

———.(1999).“ProcessingComplexityandFiller-gapDependenciesacrossGrammars.”Language75:245–85.

———.(2004).EfficiencyandComplexityofGrammars.Oxford:OxfordUniversityPress.

Hay,Jennifer(2001).“LexicalFrequencyinMorphology:IsEverythingRelative?”Linguistics39:1041–70.

Hay,Jennifer,andBaayen,R.Harald(2005).“ShiftingParadigms:GradientStructureinMorphology.”TrendsinCognitiveSciences9:342–48.

Hay,Jennifer,andBresnan,Joan(2006).“SpokenSyntax:ThePhoneticsofGivingaHandinNewZealandEnglish.”LinguisticReview23:321–49.

Hayon,Gaby,Abeles,Moshe,andLehmann,Daniel(2005).“AModelforRepresentingtheDynamicsofaSystemofSynfireChains.”JournalofComputationalNeuroscience18/1:41–53.

Heath,Jeffrey(1977).“ChoctawCases.”BerkeleyLinguisticSociety3:204–13.

Heine,Bernd(1992).“GrammaticalizationChains.”StudiesinLanguage16:335–68.

Heine,Bernd,Claudi,Ulrike,andHünnemeyer,Friederieke(1991).Grammaticalization:AConceptualFramework.Chicago:UniversityofChicagoPress.

Heine,Bernd,andKuteva,Tania(2002).WorldLexiconofGrammaticalization.Cambridge:CambridgeUniversityPress.

Helbig,Gerhard(1992).ProblemederValenz-undKasustheorie.Tübingen:Niemeyer.

Helbig,Gerhard,andSchenkel,Wolfgang(1971).WörterbuchzurValenzundDistributionDeutscherVerben(2ndedn.).Leipzig:BibliographischesInstitut.

HELSINKI(TheHelsinkiCorpusofEnglishTexts/nowpartofthePPCME/PPCEMEcorpora).http://khnt.aksis.uib.no/icame/manuals/HC/INDEX.HTM.

Herbst,Thomas,andKötz-Votteler,Karin(2007).Valency:Theoretical,DescriptiveandCognitiveIssues.BerlinandNewYork:MoutondeGruyter.

Hewson,John,andBubenik,Vit(2006).FromCasetoAdposition:TheDevelopmentofConfigurationalSyntaxinIndo-EuropeanLanguages.Amsterdam:JohnBenjamins.

Heylen,Kris,Tummers,José,andGeeraerts,Dirk(2008).“MethodologicalIssuesinCorpus-basedCognitiveLinguistics,”inGitteKristiansenandRenéDirven(eds.),Cognitive

Page 412: The Oxford Handbook of Construction Grammar

References

Page 44 of 88

Sociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,91–128.

Hickok,Gregory,andPoeppel,David(2000).“TowardsaFunctionalNeuroanatomyofSpeechPerception.”TrendsinCognitiveSciences4/4:131–38.

Hillis,Argie,andCaramazza,Alfonso(1995).“RepresentationofGrammaticalKnowledgeintheBrain.”JournalofCognitiveNeuroscience7:396–407.

Hilpert,Martin(2006a).“DistinctiveCollexemeAnalysisandDiachrony.”CorpusLinguisticsandLinguisticTheory2/2:243–57.

———.(2006b).“ASynchronicPerspectiveontheGrammaticalizationofSwedishFutureConstructions.”NordicJournalofLinguistics29:151–72.

———.(2008).GermanicFutureConstructions.AUsage-basedApproachtoLanguageChange.Amsterdam:JohnBenjamins.

———.(2009).“TheGermanmit-predicativeConstruction.”ConstructionsandFrames1:29–55.

Hilpert,Martin,andGries,StefanTh.(2009).“AssessingFrequencyChangesinMulti-stageDiachronicCorpora:ApplicationsforHistoricalCorpusLinguisticsandtheStudyofLanguageAcquisition.”LiteraryandLinguisticComputing24/4:385–401.

Hilpert,Martin,andKoops,Christian(2008).“AQuantitativeApproachtotheDevelopmentofComplexPredicates:TheCaseofSwedishPseudo-Coordinationwithsitta‘sit’.”Diachronica25/2:242–61.

Himmelmann,NikolausP.(2004).“LexicalizationorGrammaticization:OppositeorOrthogonal?”inWalterBisang,NikolausP.Himmelmann,andBjörnWiemer(eds.),WhatMakesGrammaticalization?ALookfromitsFringesanditsComponents.Berlin:MoutondeGruyter,21–42.

Höche,Silke(2009).“ACorpus-basedInvestigationofCognateObjectConstructions,”inUteRömerandRainerSchulze(eds.),ExploringtheLexis-GrammarInterface.Amsterdam:JohnBenjamins,189–209.

Hoenigswald,HenryM.(1987).“TheAnnusMirabilis1876andPosterity.”TransactionsofthePhilologicalSociety76/1:17–35.

Hoffmann,Sebastian(2004).“AreLow-frequencyComplexPrepositionsGrammaticalized?OntheLimitsofCorpusData—andtheImportanceofIntuition,”inHansLindquistandChristianMair(eds.),CorpusApproachestoGrammaticalizationinEnglish.Amsterdam:JohnBenjamins,171–210.

———.(2005).GrammaticalizationandEnglishComplexPrepositions:ACorpus-BasedStudy.London:Routledge.

Page 413: The Oxford Handbook of Construction Grammar

References

Page 45 of 88

Hoffmann,Thomas(2006).“CorporaandIntrospectionasCorroboratingEvidence:TheCaseofPrepositionPlacementinEnglishRelativeClauses.”CorpusLinguisticsandLinguisticTheory2:165–95.

———.(2008).“EnglishRelativeClausesandConstructionGrammar:ATopicwhichPrepositionPlacementcanShedLighton?”inGraemeTrousdaleandNikolasGisborne(eds.),ConstructionalApproachestoEnglishGrammar.TopicsinEnglishLinguistics57.BerlinandNewYork:MoutondeGruyter,77–112.

———.(2010).“TheMoreData,theBetter:AUsage-basedAccountoftheEnglishComparativeCorrelativeConstruction.”Talkgivenatthe6thInternationalConferenceofConstructionGrammarICCG6,CharlesUniversity,Prague,September3–5,2010.

———.(2011).PrepositionPlacementinEnglish:AUsage-basedApproach.Cambridge:CambridgeUniversityPress.

Hoffmann,Thomas,andTrousdale,Graeme(2011).“Variation,ChangeandConstructionsinEnglish.”CognitiveLinguistics22/1:1–24.

Hofmeister,Philip,andSag,Ivan(2010).“CognitiveConstraintsandIslandEffects.”Language86:366–415.

Hollmann,Willem(2007).“FromLanguage-specificConstraintstoImplicationalUniversals:ACognitive-typologicalViewoftheDativeAlternation.”FunctionsofLanguage14:57–78.

———.(forthcoming).“WordClasses:TowardsaMoreComprehensiveUsage-basedAccount.”StudiesinLanguage.

Hollmann,Willem,andSiewierska,Anna(2006).“Corporaand(theNeedfor)otherMethodsinaStudyofLancashireDialect.”ZeitschriftfürAnglistikandAmerikanistik54:203–16.

———.(2007).“AConstructionGrammarAccountofPossessiveConstructionsinLancashireDialect:SomeAdvantagesandChallenges.”EnglishLanguageandLinguistics11:407–24.

———.(2011).“TheStatusofFrequency,SchemasandIdentityinCognitiveSociolinguistics:ACaseStudyonDefiniteArticleReduction.”CognitiveLinguistics22:25–54.

Holme,Randal(2009).CognitiveLinguisticsandLanguageTeaching.Basingstoke:PalgraveMacmillan.

Holmes,Jasper,andHudson,Richard(2005).“ConstructionsinWordGrammar,”inJan-OlaÖstmanandMirjamFried(eds.),ConstructionGrammars:CognitiveGroundingandTheoreticalExtensions.Amsterdam:JohnBenjamins,253–72.

Honeybone,Patrick(2011).“VariationandLinguisticTheory,”inWarrenMaguireandAprilMcMahon(eds.),AnalyzingVariationinEnglish.Cambridge:CambridgeUniversityPress,155–77.

Hook,PeterEdwin,andChauhan,MohabhatSinghManSingh(1988).“ThePerfectiveAdverbin

Page 414: The Oxford Handbook of Construction Grammar

References

Page 46 of 88

Bhitrauti.”Word39:177–86.

Hopper,PaulJ.(1987).“EmergentGrammar.”BerkeleyLinguisticsSociety13:139–57.

———.(1991).“OnSomePrinciplesofGrammaticalization,”inElizabethC.TraugottandBerndHeine(eds.),ApproachestoGrammaticalization.Vol.1.Amsterdam:JohnBenjamins,17–35.

———.(1998).“EmergentGrammar,”inMichaelTomasello(ed.),TheNewPsychologyofLanguage.Mahwah,NJ:LawrenceErlbaum,155–76.

Hopper,PaulJ.,andThompson,SandraA.(1980).“TransitivityinGrammarandDiscourse.”Language56:251–99.

———.(1984).“TheDiscourseBasisforLexicalCategoriesinUniversalGrammar.”Language60:703–52.

Hopper,PaulJ.,andTraugott,ElizabethCloss(2003).Grammaticalization(2ndedn.).Cambridge:CambridgeUniversityPress.

Hostetter,AutumnB.,andAlibali,MarthaW.(2008).“VisibleEmbodiment:GesturesasSimulatedAction.”PsychonomicBulletinandReview15/3:495–514.

Hovav,MalkaRappaport,andLevin,Beth(1998).“BuildingVerbMeaning,”inMiriamButtandWilhelmGeuder(eds.),TheProjectionofArguments,Stanford,CA:CSLIPublications,97–134.

———.(2001).“AnEventStructureAccountofEnglishResultatives.”Language77:766–97.

———.(2005).“AreDativeVerbsPolysemous?”PaperPresentedtotheLinguisticsColloquium,PrincetonUniversity,February17,2005.

Hubel,DavidH.(1995).Eye,BrainandVision.ScientificAmericanLibraryNo.22.NewYork:WHFreeman.

Hudson,RichardA.(1990).EnglishWordGrammar.Oxford:Blackwell.

———.(1996).Sociolinguistics(2ndedn.).Cambridge:CambridgeUniversityPress.

———.(1997).“InherentVariability.”CognitiveLinguistics8:73–108.

———.(2000).EnglishWordGrammar.Oxford:Blackwell.

———.(2007a).“EnglishDialectSyntaxinWordGrammar.”EnglishLanguageandLinguistics11:383–405.

———.(2007b).LanguageNetworks:TheNewWordGrammar.Oxford:OxfordUniversityPress.

———.(2007c).“InherentVariabilityandModularity:ACommentonAdger's‘CombinatorialVariability’.”JournalofLinguistics43:683–94.

———.(2008).“WordGrammarandConstructionGrammar,”inGraemeTrousdaleandNikolas

Page 415: The Oxford Handbook of Construction Grammar

References

Page 47 of 88

Gisborne(eds.),ConstructionalApproachestoEnglishGrammar.BerlinandNewYork:MoutondeGruyter,257–302.

Hughes,Arthur,andTrudgill,Peter(1996).EnglishAccentsandDialects:AnIntroductiontoSocialandRegionalVarietiesofEnglishintheBritishIsles(3rdedn.).London:Arnold.

Hulstijn,Jan,andDeKeyser,Robert(eds.)(1997).“TestingSLATheoryintheResearchLaboratory.”StudiesinSecondLanguageAcquisition19/2(specialissue).

Humboldt,Wilhelmvon(1825).ŰberdasEntstehenderGrammatikalischenFormenundihrenEinfluβaufdieIdeenentwicklung.AbhandlungenderKőniglichenAkademiederWisseschaftenzuBerlin,401–30.

Hundt,Marianne(2004).“ThePassivalandtheProgressivePassive—aCaseStudyofLayeringintheEnglishAspectandVoiceSystems,”inHansLindquistandChristianMair(eds.),CorpusApproachestoGrammaticalisationinEnglish.Amsterdam:JohnBenjamins,79–120.

———.(2006).“Thecommitteehas/havedecided…OnConcordPatternswithCollectiveNounsinInnerandOuterCircleVarietiesofEnglish.”JournalofEnglishLinguistics34:206–32.

Hunston,Susan,andFrancis,Gillian(1996).PatternGrammar:ACorpusDrivenApproachtotheLexicalGrammarofEnglish.Amsterdam:JohnBenjamins.

———.(1999).PatternGrammar.Amsterdam:JohnBenjamins.

Huttenlocher,Janellen,Vasilyeva,Marina,andShimpi,Priya(2004).“SyntacticPriminginYoungChildren.”JournalofMemoryandLanguage50:182–95.

Hymes,Dell(1962).“TheEthnographyofSpeaking,”inThomasGladwinandWilliamC.Sturtevant(eds.),AnthropologyandHumanBehavior.Washington,DC:AnthropologicalSocietyofWashington,13–53.

Ibbotson,Paul,Theakston,AnnaL.,Lieven,ElenaV.M.,andTomasello,Michael(2012).“SemanticsoftheTransitiveConstruction:PrototypeEffectsandDevelopmentalComparisons.”CognitiveScience36:1–21.

IDS(VariousGermanCorpora).http://www.ids-mannheim.de/cosmas2/.

Israel,Michael(1996).“TheWayConstructionsGrow,”inAdeleGoldberg(ed.),ConceptualStructure,DiscourseandLanguage.Stanford,CA:CSLIPublications,217–30.

Iwata,Seizi(2006).“ArgumentResultativesandAdjunctResulativesinaLexicalConstructionalAccount:TheCaseofResultativeswithAdjectivalResultPhrases.”LanguageSciences28:449–96.

———.(2008).LocativeAlternation:ALexical-constructionalApproach.ConstructionalApproachestoLanguage6.Amsterdam:JohnBenjamins.

Page 416: The Oxford Handbook of Construction Grammar

References

Page 48 of 88

Jackendoff,Ray(1975).“SemanticandMorphologicalRegularitiesintheLexicon.”Language51:639–71.

———.(1983).SemanticsandCognition.Cambridge,MA:MITPress.

———.(1984).“OnthePhrasethephrase‘thephrase.’”NaturalLanguageandLinguisticTheory2:25–37.

———.(1987).“OnBeyondZebra:TheRelationofLinguisticandVisualInformation.”Cognition26:89–114.

———.(1990).SemanticStructures.Cambridge,MA:MITPress.

———.(1991).“Mme.TussaudMeetstheBindingTheory.”NaturalLanguageandLinguisticTheory10:1–31.

———.(1997a).TheArchitectureoftheLanguageFaculty.Cambridge,MA:MITPress.

———.(1997b).“Twistin’theNightAway.”Language73:532–59.

———.(2002a).FoundationsofLanguage:Brain,Meaning,Grammar,Evolution.Oxford:OxfordUniversityPress.

———.(2002b).“EnglishParticleConstructions,theLexiconandtheAutonomyofSyntax,”inNicoleDehé,RayJackendoff,AndrewMcIntyre,andSilkeUrban(eds.),Verb-ParticleExplorations.Berlin:MoutondeGruyter,67–94.

———.(2004).“Categorization,FuzzinessandFamilyResemblances,”inBasAarts,DavidDenison,EvelienKeizer,andGerganaPopova(eds.),FuzzyGrammar:AReader.Oxford:OxfordUniversityPress,109–29.

———.(2007).“LinguisticsinCognitiveScience:TheStateoftheArt.”LinguisticReview24:347–401.

———.(2008a).“AlternativeMinimalistVisionsofLanguage.”ChicagoLinguisticSociety41(Thepanels):189–226.

———.(2008b).“ConstructionafterConstructionanditsTheoreticalChallenges.”Language84:8–28.

———.(2010).MeaningandtheLexicon:TheParallelArchitecture1975–2010.Oxford:OxfordUniversityPress.

———.(2011).“WhatistheHumanLanguageFaculty?TwoViews.”Language87/3:586–624.

Jackendoff,Ray,andPinker,Steven(2005).“TheNatureoftheLanguageFacultyanditsImplicationsforEvolutionofLanguage(ReplytoFitch,HauserandChomsky).”Cognition97/2:211–25.

Jakobson,Roman(1971).“Shifters,VerbalCategoriesandtheRussianVerb,”reprintedin

Page 417: The Oxford Handbook of Construction Grammar

References

Page 49 of 88

RomanJakobson,SelectedWritingsII.Mouton:TheHague,130–47.

James,Carl(1980).ContrastiveAnalysis.London:Longman.

Janda,RichardD.,andJoseph,BrianD.(2003).“Introduction:OnLanguage,ChangeandLanguageChange—or,ofHistory,LinguisticsandHistoricalLinguistics,”inBrianD.JosephandRichardD.Janda(eds.),TheHandbookofHistoricalLinguistics.Oxford:Blackwell,3–180.

Janicki,Karol(2008).“HowCognitiveLinguistscanHelpSolvePoliticalProblems,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,517–42.

Janssen,Niels,andCaramazza,Alfonso(2009).“GrammaticalandPhonologicalInfluencesonWordOrder.”PsychologicalScience20:1262–68.

Jeffers,RobertJ.(1976).“SyntacticChangeandSyntacticReconstruction,”inWilliamM.Christie,Jr.(ed.),CurrentProgressinHistoricalLinguistics:ProceedingsoftheSecondInternationalConferenceonHistoricalLinguistics.Amsterdam:North-HollandPublishingCompany,1–15.

Johnson,Marc(1987).TheBodyintheMind:TheBodilyBasisofMeaning,ImaginationandReason.Chicago:UniversityofChicagoPress.

Johnson,MatthewA.,andGoldberg,AdeleE.(2012).“EvidenceforAutomaticAccessingofConstructionalMeaning:Jabberwockysentencesprimeassociatedverbs.”LanguageandCognitiveProcesses:1–14.

Johnstone,Barbara,andKiesling,ScottE.(2008).“IndexicalityandExperience:ExploringtheMeaningsof/aw/-monophthongizationinPittsburgh.”JournalofSociolinguistics12:5–33.

Jones,Mark(1999).“ThePhonologyofDefiniteArticleReduction.”LeedsStudiesinEnglish30:103–21.

———.(2002).“TheOriginofDefiniteArticleReductioninNorthernEnglishDialects:EvidencefromDialectAllomorphy.”EnglishLanguageandLinguistics6:325–45.

Jurafsky,Daniel(1992).“AnOn-lineComputationalModelofHumanSentenceInterpretation,”inAmericanAssociationforArtificialIntelligence(eds.),ProceedingsoftheNationalConferenceonArtificialIntelligence(AAAI-92),Cambridge,MA:MITPress,302–8.

Just,MarcelA.,Carpenter,PatriciaA.,Keller,TimothyA.,Eddy,WilliamF.,andThulborn,KeithR.(1996).“BrainActivationModulatedbySentenceComprehension.”Science274/5284:114–16.

Justeson,JohnS.,andKatz,SlavaM.(1991).“Co-occurrencesofAntonymousAdjectivesandtheirContexts.”ComputationalLinguistics17/1:1–19.

———.(1995).“TechnicalTerminology:SomeLinguisticPropertiesandanAlgorithmforIdentificationinText.”NaturalLanguageEngineering1:9–27.

Page 418: The Oxford Handbook of Construction Grammar

References

Page 50 of 88

Kageyama,Taro(1982).“WordFormationinJapanese.”Lingua57:215–58.

———.(1999).“WordFormation,”inNatsukoTsujimura(ed.),TheHandbookofJapaneseLinguistics.Oxford:Blackwell,297–325.

———.(2009).“Isolate:Japanese,”inRochelleLieberandPavolStekauer(eds.),TheOxfordHandbookofCompounding.Oxford:OxfordUniversityPress,512–26.

Kaschak,MichaelP.,andGlenberg,ArthurM.(2000).“ConstructingMeaning:TheRoleofAffordancesandGrammaticalConstructionsinSentenceComprehension.”JournalofMemoryandLanguage43:508–29.

Kathol,Andreas(1999).“AgreementandtheSyntax-morphologyInterfaceinHPSG,”inRobertLevineandGeorgiaGreen(eds.),StudiesinCurrentPhraseStructureGrammar.Cambridge:CambridgeUniversityPress,223–74.

Katis,Demetra,andStampouliadou,Chrysoula(2009).“TheDevelopmentofSententialComplementConstructionsinGreek:EvidencefromaCaseStudy.”ConstructionsandFrames1/2:222–61.

Kay,Martin(1986).“ParsinginFunctionalUnificationGrammar,”inBarbaraGrosz,KarinSparkJones,andBarbaraWebber(eds.),ReadingsinNaturalLanguageProcessing.LosAngeles:MorganKaufmann,251–78.

Kay,Paul(1990).“EVEN.”LinguisticsandPhilosophy13:59–111.

———.(1996).“ArgumentStructure:CausativeABC-Constructions.”Ms.UniversityofCaliforniaatBerkeley.

———.(2000).“ComprehensionDeficitsofBroca'sAphasicsProvidenoEvidenceforTraces.”BehavioralandBrainSciences23/1:37–38.

———.(2002a).“EnglishSubjectlessTaggedSentences.”Language78:453–81.

———.(2002b).“AnInformalSketchofaFormalArchitectureforConstructionGrammar.”Grammars5:1–19.

———.(2004a).“PragmaticAspectsofConstructions,”inLaurenceHornandGregoryWard(eds.),TheHandbookofPragmatics.Oxford:Blackwell,675–700.

———.(2004b).“NullComplementationConstructions.”Ms.,UniversityofCalifornia,Berkeley.

———.(2005).“ArgumentStructureConstructionsandtheArgument-adjunctDistinction,”inMirjamFriedandHansC.Boas(eds.),GrammaticalConstructions:BacktotheRoots.Amsterdam:JohnBenjamins,71–100.

Kay,Paul,andFillmore,CharlesJ.(1999).“GrammaticalConstructionsandLinguisticGeneralizations:TheWhat'sXDoingY?Construction.”Language75:1–34.

Kay,Paul,andMichaelis,LauraA.(forthcoming).“ConstructionsandCompositionality,”in

Page 419: The Oxford Handbook of Construction Grammar

References

Page 51 of 88

ClaudiaMaienborn,KlausvonHeusinger,andPaulPortner(eds.),Semantics:AnInternationalHandbookofNaturalLanguageMeaning.HSKHandbooksofLinguisticsandCommunicationScienceSeries,23:SemanticsandComputerScience.Berlin:MoutondeGruyter.

Kay,Paul,andSag,IvanA.(2009).“HowHardaProblemWouldThisbetoSolve?”inStefanMüller(ed.),ProceedingsoftheHPSG09Conference.DepartmentofLinguistics,StanfordUniversity.CSLIPublications,171–91.http://csli-publications.stanford.edu.

———.(2012).“DiscontinuousDependenciesandComplexDeterminers,”inHansC.BoasandIvan.A.Sag(eds.),Sign-BasedConstructionGrammar.Stanford,CA:CSLIPublications,217–46.

Keenan,EdwardL.(2003).“AnHistoricalExplanationofSomeBindingTheoreticFactsinEnglish,”inJohnMooreandMariaPolinsky(eds.),TheNatureofExplanationinLinguisticTheory.Stanford,CA:CSLIPublications,153–89.

Keenan,EdwardL.,andComrie,Bernard(1977).“NounPhraseAccessibilityandUniversalGrammar.”LinguisticInquiry8:63–99.

———.(1979).“DataontheNounPhraseAccessibilityHierarchy.”Language55:333–51.

Keller,Frank(2000).“GradienceinGrammar:ExperimentalandComputationalAspectsofDegreesofGrammaticality.”Ph.D.dissertation.UniversityofEdinburgh.

Keller,Rudi(1994).OnLanguageChange:TheInvisibleHandinLanguage.London:Routledge.

Kemmer,Suzanne(1993).TheMiddleVoice.TypologicalStudiesinLanguage23.Amsterdam:JohnBenjamins.

Kemmer,Suzanne,andBarlow,Michael(2000).“Introduction:AUsage-basedConceptionofLanguage,”inMichaelBarlowandSuzanneKemmer(eds.),Usage-basedModelsofLanguage.Stanford,CA:CSLIPublications,vii–xxviii.

Kemmer,Suzanne,andVerhagen,Arie(1994).“TheGrammarofCausativesandtheConceptualStructureofEvents.”CognitiveLinguistics5:115–56.

Kempen,Gerard,andHarbusch,Karin(2003).“WordOrderScramblingasaConsequenceofIncrementalSentenceProduction,”inHoldenHärtlandHeikeTappe(eds.),MediatingbetweenConceptsandGrammar.Berlin:MoutondeGruyter,141–64.

Kempen,Gerard,andHuijbers,Pieter(1983).“TheLexicalizationProcessinSentenceProductionandNaming:IndirectElicitationofWords.”Cognition14:185–209.

Kerswill,Paul,andWilliams,Ann(2002).“‘Salience’asanExplanatoryFactorinLanguageChange:EvidencefromDialectLevellinginUrbanEngland,”inMariC.JonesandEdithEsch(eds.),LanguageChange:TheInterplayofInternal,ExternalandExtra-linguisticFactors.BerlinandNewYork:MoutondeGruyter,81–110.

Page 420: The Oxford Handbook of Construction Grammar

References

Page 52 of 88

Kidd,Evan,Lieven,ElenaV.M.,andTomasello,Michael(2010).“LexicalFrequencyandExemplar-basedLearningEffectsinLanguageAcquisition:EvidencefromSententialComplements.”LanguageSciences32:132–42.

Kiefer,Ferenc,andHonti,László(2003).“Verbal‘Prefixation’intheUralicLanguages.”ActaLinguisticaHungarica50:137–53.

Kiefer,Markus,Sim,Eun-Jin.,Herrnberger,Bärberl,Grothe,Jo,andHoenig,Klaus(2008).“TheSoundofConcepts:FourMarkersforaLinkbetweenAuditoryandConceptualBrainSystems.”JournalofNeurolinguistics28/47:12224–30.

Kikusawa,Ritsuko(2003).“TheDevelopmentofSomeIndonesianPronominalSystems,”inBarryJ.Blake,KateBurridge,andJoTaylor(eds.),HistoricalLinguistics2001:SelectedPapersfromthe15thInternationalConferenceonHistoricalLinguistics,Melbourne,13–17August2001.Amsterdam:JohnBenjamins,237–68.

Kikuta,ChiharuUda(1998).“AMultipleInheritanceAnalysisoftheInternally-HeadedRelativeClauseinJapanese.”Language,InformationandComputation(PACLIC12),18–20Feb,1998,82–93.

———.(2002).“ClausalComplementorAdverbialClause?:TowardanIntegralAccountoftheJapaneseInternally-HeadedRelativeClauses,”inFrankvanEynde,LarsHellan,andDorotheeBeermann(eds.),Proceedingsofthe8thInternationalHPSGConference,NorwegianUniversityofScienceandTechnology.CSLIPublications.http://csli-publications.stanford.edu.

Kinno,Ryuta,Kawamura,Mitsuru,Shioda,Seiji,andSakai,KuniyoshiL.(2008).“NeuralCorrelatesofNoncanonicalSyntacticProcessingRevealedbyaPicture-sentenceMatchingTask.”HumanBrainMapping29/9:1015–27.

Kiparsky,Paul,andKiparsky,Carol(1971).“Fact,”inDannyD.SteinbergandLeonA.Jakobovits(eds.),Semantics:AnInterdisciplinaryReaderinPhilosophy,LinguisticsandPsychology.Cambridge:CambridgeUniversityPress,345–69.

Kirk,John(1985).“LinguisticAtlasesandGrammar:TheInvestigationandDescriptionofRegionalVariationinEnglishSyntax,”inJohnKirk,StewartSanderson,andJohnD.A.Widdowson(eds.),StudiesinLinguisticGeography.London:CroomHelm,130–56.

Klein,JaredS.(2010).“ReviewofPrinciplesofReconstruction,ed.byGiselaFerraresiandMariaGoldbach(2008).”Language86:720–26.

Klima,EdwardS.,andBellugi,Ursula(1966).“SyntacticRegularitiesintheSpeechofChildren,”inJohnLyonsandRichardJ.Wales(eds.),PsycholinguisticPapers.Edinburgh:UniversityofEdinburghPress,183–208.

Kluender,Robert(1992).“DerivingIslandConstraintsfromPrinciplesofPredication,”inHelenGoodluckandMichaelRochemont(eds.),IslandConstraints:Theory,AcquisitionandProcessing.Dordrecht:Kluwer,241–79.

Page 421: The Oxford Handbook of Construction Grammar

References

Page 53 of 88

———.(1998).“OntheDistinctionbetweenStrongandWeakislands:AProcessingPerspective.”SyntaxandSemantics29:241–79.

Kluge,Friedrich(2002).EtymologischesWörterbuchderDeutschenSprache(24thedn.).Berlin:Gruyter.

Knight,RobertT.(2007).“Neuroscience:NeuralNetworksDebunkPhrenology.”Science316/5831:1578–79.

Knoblauch,Andreas,andPulvermüller,Friedemann(2005).“SequenceDetectorNetworksandAssociativeLearningofGrammaticalCategories,”inStefanWermter,GüntherPalm,andMarkElshaw(eds.),BiomimeticNeuralLearningforIntelligentRobots.Berlin:Springer,31–53.

Kolehmainen,Leena,andLarjavaara,Meri(2004).“The‘Bizarre’ValencyBehaviourofFinnishVerbs:HowaSpecificContextGivesRisetoValencyAlternationPatterns.”Constructions1/2004.http://www.elanguage.net/journals/index.php/constructions/index.

Koller,Veronika(2008).“CorporateBrandsasSocio-cognitiveRepresentations,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,389–418.

Konopka,AgnieszkaE.,andBock,Kathryn(2008).“LexicalorSyntacticControlofSentenceFormulation?StructuralGeneralizationsfromIdiomProduction.”CognitivePsychology58:68–101.

Koopman,WillemF.,andvanderWurff,Wim(2000).“TwoWordOrderPatternsintheHistoryofEnglish:Stability,VariationandChange,”inRosannaSornicola,ErichPoppe,andArielShisha-Halevy(eds.),Stability,VariationandChangeinWordOrderPatternsoverTime.Amsterdam:JohnBenjamins,259–84.

Koptjevskaja-Tamm,Maria(1993).Nominalizations.London:Routledge.

Kosslyn,StephenM.,Ganis,Giorgio,andThompson,WilliamL.(2001).“NeuralFoundationsofImagery.”NatureReviewsNeuroscience2:635–42.

Kristiansen,Gitte,andDirven,René(2008).“Introduction.CognitiveSociolinguistics:Rationale,MethodsandScope,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,1–17.

Kruschke,JohnK.,andJohansen,MarkK.(1999).“AModelofProbabilisticCategoryLearning.”JournalofExperimentalPsychology:Learning,MemoryandCognition25/5:1083–119.

Kuningas,Johanna(2007).StructureInformationelleetConstructionsduKabyle:EtudedeTroisTypesdePhrasedansleCadredelaGrammaireConstructionelle.Publications,vol.42.Helsinki:UniversityofHelsinkiDepartmentofGeneralLinguistics.

Kuningas,Johanna,andLeino,Jaakko(2006).“WordOrdersandConstructionGrammars,”inMickaelSuominen,AnttiArppe,AnuAirola,OrvokkiHeinämäki,MattiMiestamo,UrhoMäättä,

Page 422: The Oxford Handbook of Construction Grammar

References

Page 54 of 88

JussiNiemi,KariK.Pitkänen,andKaiusSinnemäki(eds.),AManofMeasure:FestschriftinHonourofFredKarlssononhis60thBirthday.AspecialsupplementtoSKYJournalofLinguistics19:301–9.

Kuperberg,GinaR.(2007).“NeuralMechanismsofLanguageComprehension:ChallengestoSyntax.”BrainResearch1146:23–49.

Kutas,Marta,andHillyard,StevenA.(1984).“BrainPotentialsduringReadingReflectWordExpectancyandSemanticAssociation.”Nature307:161–63.

Labov,William(1963).“TheSocialMotivationofaSoundChange.”Word19:273–309.

———.(1969).“Contraction,DeletionandInherentVariabilityoftheEnglishCopula.”Language45:715–62.

———.(1972a).LanguageintheInnerCity.Philadelphia:UniversityofPennsylvaniaPress.

———.(1972b).SociolinguisticPatterns.Philadelphia:UniversityofPennsylvaniaPress.

———.(1994).PrinciplesofLinguisticChange.Vol.1:InternalFactors.Oxford:BasilBlackwell.

———.(2001).PrinciplesofLinguisticChange.Vol.2:SocialFactors.Oxford:Blackwell.

———.(2010).PrinciplesofLinguisticChange.Vol.3:CognitiveandCulturalFactors.Oxford:Blackwell.

Lado,Robert(1957).LinguisticsAcrossCultures:AppliedLinguisticsforLanguageTeachers.AnnArbor:UniversityofMichiganPress.

———.(1964).LanguageTeaching:AScientificApproach.NewYork:McGraw-Hill.

Lakoff,George(1970).IrregularityinSyntax.NewYork,:HoltRinehartandWinston.

———.(1987).Women,FireandDangerousThings:WhatCategoriesRevealabouttheMind.Chicago:ChicagoUniversityPress.

———.(1990).“TheInvarianceHypothesis:IsAbstractReasonBasedonImage-schemas?”CognitiveLinguistics1:39–74.

Lakoff,George,andJohnson,MarkH.(1980).MetaphorsWeLiveby.Chicago:UniversityofChicagoPress.

———.(1999).PhilosophyintheFlesh:TheEmbodiedMindanditsChallengetoWesternThought.London:HarperCollins.

Lambrecht,Knud(1981).Topic,AntitopicandVerbAgreementinNon-standardFrench.PragmaticsandBeyondII6.Amsterdam:JohnBenjamins.

———.(1984).“Formulaicity,FrameSemanticsandPragmaticsinGermanBinominal

Page 423: The Oxford Handbook of Construction Grammar

References

Page 55 of 88

Expressions.”Language60/3:753–96.

———.(1987).“SentenceFocus,InformationStructureandtheThetic–categorialDistinction.”BerkeleyLinguisticsSociety13:366–82.

———.(1988).“ThereWasaFarmerHadaDog:SyntacticAmalgamsRevisited.”BerkeleyLinguisticSociety14:319–39.

———.(1990).“‘WhatmeWorry?’MadMagazineSentencesRevisited.”BerkeleyLinguisticSociety16:215–28.

———.(1994).InformationStructureandSentenceForm:Topic,FocusandtheMentalRepresentationsofDiscourseReferents.Cambridge:CambridgeUniversityPress.

———.(1997).“FrenchRelativeClausesasSecondaryPredicates,”inRalphC.BlightandMichelleJ.Moosally(eds.),TheSyntaxandSemanticsofPredication.Proceedingsofthe1997TexasLinguisticsSocietyConference.Austin,Texas:DepartmentofLinguistics,UTAustin,205–21.

———.(2000).“PrédicationSecondeetStructureInformationnelle:LaRelativedePerceptioncommeConstructionPrésentative.”LangueFrançaise127:49–66.

———.(2001).“Dislocation,”inMartinHaspelmath,EkkehardKönig,WulfOesterreicher,andWolfgangRaible(eds.),LanguageTypologyandLanguageUniversals:AnInternationalHandbook.Vol.2.Berlin:WalterdeGruyter,1056–78.

———.(2002).“Topic,FocusandSecondaryPredication:TheFrenchPresentationalRelativeConstruction,”inClaireBeyssade,ReinekeBok-Bennema,FrankDrijkoningen,andPaolaMonachesi(eds.),ProceedingsofGoingRomance2000.Amsterdam:JohnBenjamins,171–212.

———.(2004).“OntheInteractionofInformationStructureandFormalStructureinConstructions:TheCaseofFrenchRight-detachedcomme-N,”inMirjamFriedandJan-OlaÖstman(eds.),ConstructionGrammarinaCross-LanguagePerspective.ConstructionalApproachestoLanguage2.Amsterdam:JohnBenjamins,157–99.

Lambrecht,Knud,andLemoine,Kevin(2005).“DefiniteNullObjectsin(spoken)French:AConstruction-GrammarAccount,”inMirjamFriedandHansC.Boas(eds.),GrammaticalConstructions:BacktotheRoots.Amsterdam:JohnBenjamins,13–56.

Lambrecht,Knud,andMichaelis,LauraA.(1998).“SentenceAccentinInformationQuestions.”LinguisticsandPhilosophy21/5:477–544.

Langacker,RonaldW.(1986).“AnIntroductiontoCognitiveGrammar.”CognitiveScience10:1–40.

———.(1987).FoundationsofCognitiveGrammar.Vol.I:TheoreticalPrerequisites.Stanford,CA:StanfordUniversityPress.

Page 424: The Oxford Handbook of Construction Grammar

References

Page 56 of 88

———.(1988a).“AnOverviewofCognitiveGrammar,”inBrygidaRudzka-Ostyn(ed.),TopicsinCognitiveLinguistics.Amsterdam:JohnBenjamins,49–90.

———.(1988b).“Usage-BasedModel,”inBrygidaRudzka-Ostyn(ed.),TopicsinCognitiveLinguistics.Amsterdam:JohnBenjamins,127–61.

———.(1991).FoundationsofCognitiveGrammar.Vol.2:DescriptiveApplication.Stanford,CA:StanfordUniversityPress.

———.(1993).“ReferencePointConstructions.”CognitiveLinguistics4:1–38.

———.(1999).GrammarandConceptualization.Berlin:MoutondeGruyter.

———.(2000a).Concept,ImageandSymbol:TheCognitiveBasisofGrammar(2ndedn.).Berlin:MoutondeGruyter.

———.(2000b).“ADynamicUsage-basedModel,”inMichaelBarlowandSuzanneKemmer(eds.),Usage-basedModelsofLanguage.Stanford,CA:CSLIPublications,1–63.

———.(2005).“ConstructionGrammars:Cognitive,RadicalandLessSo,”inFranciscoJ.RuizdeMendozaIbáñezandM.SandraPeñaCervel(eds.),CognitiveLinguistics:InternalDynamicsandInterdisciplinaryInteraction.BerlinandNewYork:MoutondeGruyter,101–59.

———.(2007).“CognitiveGrammar,”inDirkGeeraertsandHubertCuyckens(eds.),TheOxfordHandbookofCognitiveLinguistics.Oxford:OxfordUniversityPress,421–62.

———.(2008a).CognitiveGrammar:ABasicIntroduction.Oxford:OxfordUniversityPress.

———.(2008b).“SequentialandSummaryScanning:AReply.”CognitiveLinguistics19:571–84.

———.(2009a).InvestigationsinCognitiveGrammar.Berlin:MoutondeGruyter.

———.(2009b).“ConstructionsandConstructionalMeaning,”inVyvyanEvansandStéphaniePourcel(eds.),NewDirectionsinCognitiveLinguistics.Amsterdam:JohnBenjamins,225–67.

———.(2009c).“Cognitive(Construction)Grammar.”CognitiveLinguistics20:167–76.

LaPolla,Randy,Kratochvíl,František,andCoupe,AlexanderR.(2011).“OnTransitivity.”StudiesinLanguage35/3:469–91.

Laudan,Larry(1977).ProgressanditsProblems:TowardsaTheoryofScientificGrowth.BerkeleyandLosAngeles:UniversityofCaliforniaPress.

Lehmann,Christian(1985).“Grammaticalization:SynchronicVariationandDiachronicChange.”LinguaeStile20:303–18.

———.(1995).ThoughtsonGrammaticalization.Munich:LincomEuropa.

Leino,Jaakko(ed.)(2008).ConstructionalReorganization.Amsterdam:JohnBenjamins.

Page 425: The Oxford Handbook of Construction Grammar

References

Page 57 of 88

———.(2010).“Results,CasesandConstructions:ArgumentStructureConstructionsinEnglishandFinnish,”inHansC.Boas(ed.),ContrastiveStudiesinConstructionGrammar.Amsterdam:JohnBenjamins,103–36.

Leino,Jaakko,andKuningas,Johanna(2005).“StructureInformationnelleetl’ordredesMotsaveclesExemplesduFrançais,KabyleetduFinnois,”inJocelyneFernandez-Vest(ed.),LesLanguesOuraliennesAujourd’hui—TheUralicLanguagesToday.Bibliothèquedel’EcoledesHautesEtudes.Paris:EditionsHonoréChampion,587–98.

Leino,Jaakko,andÖstman,Jan-Ola(2005).“ConstructionsandVariability,”inMirjamFriedandHansC.Boas(eds.),GrammaticalConstructions:BacktotheRoots.Amsterdam:JohnBenjamins,191–213.

Leino,Pentti,andÖstman,Jan-Ola(2008).“LanguageChange,VariabilityandFunctionalLoad:FinnishGenericityfromaConstructionalPointofView,”inJaakkoLeino(ed.),ConstructionalReorganization.Amsterdam:JohnBenjamins,37–54.

Levelt,Willem(1989).Speaking:FromIntentiontoArticulation.Cambridge,MA:MITPress.

Levelt,Willem,Roelofs,Ardie,andMeyer,Antje(1999).“ATheoryofLexicalAccessinSpeechProduction.”BehavioralandBrainSciences22:1–75.

Levin,Beth(1993).EnglishVerbClassesandAlternations.Chicago:UniversityofChicagoPress.

Levin,Beth,andHovav,MalkaRappaport(1995).Unaccusativity:AttheSyntax–LexicalSemanticsInterface.Cambridge,MA:MITPress.

Levin,Beth,andRapoport,TovaR.(1988).“LexicalSubordination.”ChicagoLinguisticSociety24:275–89.

Levinson,StephenA.(1997.)“YélîDnyeandtheTheoryofBasicColorTerms.”JournalofLinguisticAnthropology10:3–55.

Levinson,StephenC.(1983).Pragmatics.Cambridge:CambridgeUniversityPress.

Levinson,StephenC.,Meira,Sérgio,andtheLanguageandCognitionGroup(2003).“‘NaturalConcepts’intheSpatialTopologicalDomain—AdpositionalMeaningsinCrosslinguisticPerspective:AnExerciseinSemanticTypology.”Language79/3:485–516.

Lewis,Michael(1993).TheLexicalApproach:TheStateofELTandtheWayForward.Hove,UK:LanguageTeachingPublications.

Li,CharlesN.(1976).SubjectandTopic.NewYork:AcademicPress.

Lichtenberk,František(1991).“OntheGradualnessofGrammaticalization,”inElizabethClossTraugottandBerndHeine(eds.),ApproachestoGrammaticalization.Vol.1.Amsterdam:JohnBenjamins,37–80.

Page 426: The Oxford Handbook of Construction Grammar

References

Page 58 of 88

Lichtheim,Ludwig(1885).“OnAphasia.”Brain7:433–84.

Lidz,Jeffrey,Gleitman,Henry,andGleitman,Lila(2003).“UnderstandingHowInputMatters:VerbLearningandtheFootprintofUniversalGrammar.”Cognition87:151–78.

Lieber,Rochelle(2009).“IE,Germanic:English,”inRochelleLieberandPavolStekauer(eds.),TheOxfordHandbookofCompounding.Oxford:OxfordUniversityPress,357–69.

Lieven,ElenaV.M.,Behrens,Heike,Speakers,Jennifer,andTomasello,Michael(2003).“EarlySyntacticCreativity:AUsage-basedApproach.”JournalofChildLanguage30:333–70.

Lieven,ElenaV.M.,Pine,JulianM.,andBaldwin,Gillian(1997).“Lexically-basedLearningandEarlyGrammaticalDevelopment.”JournalofChildLanguage24:187–219.

Lieven,ElenaV.M.,Salomon,Dorothé,andMichaelTomasello(2009).“Two-yearoldChildren'sProductionofMultiwordUtterances:AUsage-basedAnalysis.”CognitiveLinguistics20/3:481–507.

Lightbown,PatsyM.,Spada,Nina,andWhite,Lydia(1993).“TheRoleofInstructioninSecondLanguageAcquisition.”StudiesinSecondLanguageAcquisition15/2:143–45.

Lightfoot,DavidW.(1979).PrinciplesofDiachronicSyntax.Cambridge:CambridgeUniversityPress.

———.(2002).“MythsandthePrehistoryofGrammars.”JournalofLinguistics38/1:113–36.

———.(2006).HowNewLanguagesEmerge.Cambridge:CambridgeUniversityPress.

Lin,Dekang(1999).“AutomaticIdentificationofNoncompositionalPhrases.”Proceedingsofthe37thAnnualMeetingoftheACL,CollegePark,USA,317–24.

Lindstromberg,Seth,andBoers,Frank(2008).“TheMnemonicEffectofNoticingAlliterationinLexicalChunks.”AppliedLinguistics29:200–222.

Linell,Per(2009).“GrammaticalConstructionsinDialogue,”inAlexanderBergsandGabrieleDiewald(eds.),ContextsandConstructions.Amsterdam:JohnBenjamins,97–110.

Littlemore,Jeannette(2009).ApplyingCognitiveLinguisticstoSecondLanguageLearningandTeaching.Basingstoke:PalgraveMacmillan.

Long,Michael.H.(1991).“FocusonForm:ADesignFeatureinLanguageTeachingMethodology,”inKeesd.Bot,RalphB.Ginsberg,andClaireKramsch(eds.),ForeignLanguageResearchinCross-culturalPerspective.Amsterdam:JohnBenjamins,39–52.

López-Couso,MariaJosé(2010).“SubjectificationandIntersubjectification,”inAndreasH.JuckerandIrmaTaavitsainen(eds.),HistoricalPragmatics.BerlinandNewYork:MoutondeGruyter,127–63.

Los,Bettelou,Blom,Corrien,Booij,Geert,Elenbaas,Marion,andVanKemenade,Ans(2012).Morpho-syntacticChange:AComparativeStudyofParticlesandPrefixes.Cambridge:

Page 427: The Oxford Handbook of Construction Grammar

References

Page 59 of 88

CambridgeUniversityPress.

Lüdeling,Anke,andKytö,Merja(eds.)(2008).CorpusLinguistics:AnInternationalHandbook.Berlin:MoutondeGruyter.

Luraghi,Silvia(2010).“TheRise(andPossibleDownfall)ofConfigurationality,”inSilviaLuraghiandVitBubenik(eds.),ACompaniontoHistoricalLinguistics.London:ContinuumBooks,212–29.

MacDonald,Maryellen,Perlmutter,Neal,andSeidenberg,Mark(1994).“TheLexicalNatureofSyntacticAmbiguityResolution.”PsychologicalReview101:676–703.

MacWhinney,Brian(1987a).“ApplyingtheCompetitionModeltoBilingualism.”AppliedPsycholinguistics8/4:315–27.

———.(1987b).“TheCompetitionModel,”inBrianMacWhinney(ed.),MechanismsofLanguageAcquisition.Hillsdale,NJ:Erlbaum,249–308.

———.(1997).“SecondLanguageAcquisitionandtheCompetitionModel,”inAnnetteM.B.DeGrootandJudithF.Kroll(eds.),TutorialsinBilingualism:PsycholinguisticPerspectives.Mahwah,NJ:LawrenceErlbaumAssociates,113–42.

———.(1998).“ModelsoftheEmergenceofLanguage.”AnnualReviewofPsychology49:199–227.

———.(2001).“TheCompetitionModel:TheInput,theContextandtheBrain,”inPeterRobinson(ed.),CognitionandSecondLanguageInstruction.Cambridge:CambridgeUniversityPress,69–90.

MacWhinney,Brian,andBates,Elizabeth(1978).“SententialDevicesforConveyingGivennessandNewness:ACross-culturalDevelopmentalStudy.”JournalofCommunicationDisorders17:539–58.

Madden,CarolJ.,andZwaan,RolfA.(2003).“HowDoesVerbAspectConstrainEventRepresentations?”MemoryandCognition31:663–72.

Mair,Christian(2004).“CorpusLinguisticsandGrammaticalisationTheory:Statistics,FrequenciesandBeyond,”inHansLindquistandChristianMair(eds.),CorpusApproachestoGrammaticalisationinEnglish.Amsterdam:JohnBenjamins,121–50.

———.(2006).Twentieth-CenturyEnglish:History,VariationandStandardization.Cambridge:CambridgeUniversityPress.

Makkai,Adam(1972).IdiomStructureinEnglish.TheHague:MoutondeGruyter.

Mallory,JamesP.,andAdams,DouglasQ.(2006).TheOxfordIntroductiontoProto-Indo-EuropeanandtheProto-Indo-EuropeanWorld.Oxford:OxfordUniversityPress.

Mann,WilliamC..andThompson,SandraA.(1988).“RhetoricalStructureTheory:Towarda

Page 428: The Oxford Handbook of Construction Grammar

References

Page 60 of 88

FunctionalTheoryofTextOrganization.”Text8:243–81.

Manning,Christopher,andCarpenter,Bob(1997).“ProbabilisticParsingUsingLeft-cornerLanguageModels,”inProceedingsofthe5thInternationalWorkshoponParsingTechnology.Cambridge,MA:MITPress,147–58.

Marandin,Jean-Marie(2006).“ContoursasConstructions.”ConstructionsSV1–10.http://www.elanguage.net/journals/index.php/constructions/index.

Marantz,Alec(1997).“NoEscapefromSyntax:Don’tTryMorphologicalAnalysisinthePrivacyofYourOwnLexicon.”PennsylvaniaWorkingPapersinLinguistics.4/2:201–25.

———.(2006).“GenerativeLinguisticswiththeCognitiveNeuroscienceofLanguage.”LinguisticReview22:429–45.

Margerie,Hélène(2008).“AHistoricalandCollexemeAnalysisoftheDevelopmentoftheCompromiserfairly.”JournalofHistoricalPragmatics9/2:288–314.

Marr,David(1982).Vision:AComputationalInvestigationintotheHumanRepresentationandProcessingofVisualInformation.SanFrancisco:W.H.FreemanandCompany.

Martin,Alex(2007).“TheRepresentationofObjectConceptsintheBrain.”AnnualReviewofPsychology58:25–45.

Masson,MichaelE.J.,Bub,DanielN.,andWarren,ChristopherM.(2008).“KickingCalculators:ContributionofEmbodiedRepresentationstoSentenceComprehension.”JournalofMemoryandLanguage59:256–65.

Mathieu,Eric(2006).“QuirkySubjectsinOldFrench.”StudiaLinguistica60/3:282–312.

Matlock,Teenie(2004).“FictiveMotionasCognitiveSimulation.”MemoryandCognition32:1389–400.

Matsumoto,Yo(2007).“TypologicalSplitinLinguisticExpressionsforMotionEvents:Motion,CausedMotion,andFictiveCausedMotion.”Presentationatthe3rdSeoulInternationalConferenceonDiscourseandCognitiveLinguistics.KoreaUniversity.

McCarthy,Diana,Keller,Bill,andCarroll,John(2003).“DetectingaContinuumofCompositionalityinPhrasalVerbs,”inAssociationforComputationalLinguistics(eds.),ProceedingsoftheACL-WorkshoponMultiwordExpressions:Analysis,AcquisitionandTreatment.Stroudsburg,PA:ACLPress,73–80.

McCarthy,John(1982).“ProsodicStructureandExpletiveInfixation.”Language58:574–90.

McCawley,JamesD.[Đông,QuangPhúc](1971).“EnglishSentenceswithoutOvertGrammaticalSubject.”http://douglemoine.com.

———.(1988a).TheSyntacticPhenomenaofEnglish.2vols.(2ndedn.1998).Chicago:UniversityofChicagoPress.

Page 429: The Oxford Handbook of Construction Grammar

References

Page 61 of 88

———.(1988b).“TheComparativeConditionalConstructioninEnglish,GermanandMandarinChinese.”BerkeleyLinguisticsSociety14:176–87.

McClelland,JamesL.,andPatterson,Karalyn(2002).“RulesorConnectionsinPast-tenseInflections:WhatdoestheEvidenceRuleout?”TrendsinCognitiveScience6/11:465–72.

McDonough,Kim(2006).“InteractionandSyntacticPriming:EnglishL2speakers’ProductionofDativeConstructions.”StudiesinSecondLanguageAcquisition28:179–207.

McDonough,Kim,andMackey,Alison(2006).“ResponsestoRecasts:Repetitions,PrimedProductionandLinguisticDevelopment.”LanguageLearning56:693–720.

McDonough,Kim,andTrofimovich,Pavel(2008).UsingPrimingMethodsinSecondLanguageResearch.London:Routledge.

McEnery,Tony,andWilson,Andrew(2001).CorpusLinguistics:AnIntroduction(2ndedn.).Edinburgh:EdinburghUniversityPress.

McGlone,MatthewS.,Glucksberg,Sam,andCacciari,Cristina(1994).“SemanticProductivityandIdiomComprehension.”DiscourseProcesses17/2:176–90.

McLaughlin,Barry(1995).“FosteringSecondLanguageDevelopmentinYoungChildren:PrinciplesandPractices.http://www.ncbe.gwu.edu/miscpubs/ncrcdsll/epr14.htm.

McRae,Ken,Hare,Mary,Elman,JeffreyL.,andFerretti,Todd(2006).“ABasisforGeneratingExpectanciesforVerbsfromNouns.”MemoryandCognition33:1174–84.

Meisel,JürgenM.(1994).“ParametersinAcquisition,”inPaulFletcherandBrainMacWhinney(eds.),HandbookofChildLanguage.Oxford:Blackwell,9–35.

Melinger,Alissa,andDobel,Christian(2005).“Lexically-drivenSyntacticPriming.”Cognition98:B11–B20.

Mettouchi,Amina(2003)“ContrastiveFocalizationonCleftsinTaqbaylitBerber.”ProceedingsofIP2003InterfacesProsodiques,Nantes,27–29Mars2003,143–48.

———.(2008)“Case-marking,SyntacticDomainsandInformationStructureinKabyle(Berber),”inZygmundFrajzyngierandErinShay(eds.),InteractionofMorphologyandSyntax:CaseStudiesinAfroasiatic.Amsterdam:JohnBenjamins,7–40.

Michaelis,LauraA.(1994).“ACaseofConstructionalPolysemyinLatin.”StudiesinLanguage18:45–70.

———.(2001).“ExclamativeConstructions,”inMartinHaspelmath,EkkehardKönig,WulfÖsterreicher,andWolfgangRaible(eds.),LanguageUniversalsandLanguageTypology:AnInternationalHandbook.Berlin:WalterdeGruyter,1038–50.

———.(2004).“TypeShiftinginConstructionGrammar:AnIntegratedApproachtoAspectualCoercion.”CognitiveLinguistics15/1:1–67.

Page 430: The Oxford Handbook of Construction Grammar

References

Page 62 of 88

———.(2010).“Sign-BasedConstructionGrammar,”inBerndHeineandHeikoNarrog(eds.),TheOxfordHandbookofLinguisticAnalysis.Oxford:OxfordUniversityPress,155–76.

———.(2012).“MakingtheCaseforConstructionGrammar,”inHansBoasandIvanSag(eds.),Sign-BasedConstructionGrammar.Stanford,CA:CSLIPublications,1–38.

Michaelis,LauraA.,andLambrecht,Knud(1994).“OnNominalExtraposition:AConstructionalAnalysis.”BerkeleyLinguisticsSociety20:262–73.

———.(1996).“TowardaConstruction-BasedModelofLanguageFunction:TheCaseofNominalExtraposition.”Language72:215–47.

Michaelis,LauraA.,andRuppenhofer,Josef(2001).BeyondAlternations:AConstructionalModeloftheGermanApplicativePattern.Stanford,CA:CSLIPublications.

Minkova,Donka(2003).AlliterationandSoundChangeinEarlyEnglish.Cambridge:CambridgeUniversityPress.

Mintz,TobenH.,Newport,ElissaL.,andBever,ThomasG.(2002).“TheDistributionalStructureofGrammaticalCategoriesintheSpeechtoYoungChildren.”CognitiveScience26:393–424.

Mithun,Mariane(2008).“BorrowedRhetoricalConstructionsasStartingPointsforGrammaticalization,”inAlexanderBergsandGabrieleDiewald(eds.),ConstructionsandLanguageChange.Berlin:MoutondeGruyter,195–230.

Mok,Eva(2008).“ContextualBootstrappingforGrammarLearning.”Ph.D.dissertation.ComputerScienceDepartment,UniversityofCalifornia,Berkeley.

Mok,Eva,andBryant,John(2006).“ABest-FitApproachtoProductiveOmissionofArguments.”BerkeleyLinguisticsSociety32:1–10.

Mok,Eva,Bryant,John,andFeldman,Jerome(2004).“ScalingUnderstandinguptoMentalSpaces.”ProceedingsoftheWorkshoponScalableNaturalLanguageUnderstanding.Boston,MA,41–8.

Moon,Rosamund(1998).FixedExpressionsandIdiomsinEnglish:ACorpus-basedApproach.Oxford:Clarendon.

Morgan,PamelaS.(1997).“Figuringoutfigureout:MetaphorandtheSemanticsoftheEnglishVerb-particleConstruction.”CognitiveLinguistics8:327–58.

Mos,Maria(2010).ComplexLexicalItems.Utrecht:LOT.

MoscosoDelPradoMartín,Fermin,Hauk,Olaf,andPulvermüller,Friedemann(2006).“CategorySpecificityintheProcessingofColor-relatedandForm-relatedwords:AnERPStudy.”Neuroimage29/1:29–37.

Mukherjee,Joybrato(2001).FormandFunctionofParasyntacticPresentationStructures:ACorpus-basedStudyofTalkUnitsinSpokenEnglish.Amsterdam:Rodopi.

Page 431: The Oxford Handbook of Construction Grammar

References

Page 63 of 88

Mukherjee,Joybrato,andGries,StefanTh.(2009).“CollostructionalNativisationinNewEnglishes:Verb-constructionAssociationsintheInternationalCorpusofEnglish.”EnglishWorld-Wide30/1:27–51.

Mukherjee,Joybrato,andHoffmann,Sebastian(2006).“DescribingVerb-complementationalProfilesofNewEnglishes:APilotStudyofIndianEnglish.”EnglishWorld-Wide27:147–73.

Müller,Stefan(2006).“PhrasalorLexicalConstructions?”Language82/4:850–83.

———.(2010).“PersianComplexPredicatesandtheLimitsofInheritance-basedAnalyses.”JournalofLinguistics46:601–55.

Murphy,Gregory(2002).TheBigBookofConcepts.Cambridge,MA:MITPress.

Näätänen,Risto,Gaillard,AnthonyW.,andMäntysalo,Sirkka(1978).“EarlySelective-attentionEffectonEvokedPotentialReinterpreted.”ActaPsychologica42:313–29.

Näätänen,Risto,Paavilainen,Petri,Rinne,Teemu,andAlho,Kimmo(2007).“TheMismatchNegativity(MMN)inBasicResearchofCentralAuditoryProcessing:AReview[Dec].”ClinicalNeurophysiology118:2544–90.

Näätänen,Risto,Tervaniemi,Mari,Sussman,Elyse,Paavilainen,Petri,andWinkler,István(2001).“‘PrimitiveIntelligence’intheAuditoryCortex.”TrendsinNeurosciences24/5:283–88.

Nader,Karim,Schafe,GlennE.,andleDoux,JosephE.(2000).“FearMemoriesRequireProteinSynthesisintheAmygdaleforReconsolidationafterRetrieval.”Nature406:722–26.

Naigles,Letitia(1990).“ChildrenUseSyntaxtoLearnVerbMeanings.”JournalofChildLanguage17:357–74.

Naigles,Letitia,Bavin,Edith,andSmith,Melissa(2005).“ToddlersRecognizeVerbsinNovelSituationsandSentences.”DevelopmentalScience8:424–31.

Narayanan,Srini(1997).“Knowledge-basedActionRepresentationsforMetaphorandAspect(KARMA).”Ph.D.dissertation,ComputerScienceDivision,UniversityofCaliforniaatBerkeley.

Nattinger,JamesR.(1980).“ALexicalPhraseGrammarforESL.”TESOLQuarterly14:337–44.

Nattinger,JeanetteR.,andDeCarrico,JeanetteS.(1992).LexicalPhrasesandLanguageTeaching.Oxford:OxfordUniversityPress.

Nelson,Gerald,Wallis,Sean,andAarts,Bas(2002).ExploringNaturalLanguage:WorkingwiththeBritishComponentoftheInternationalCorpusofEnglish.Amsterdam:JohnBenjamins.

Nemoto,Noriko(1998).“OnthePolysemyofDitransitivesave:TheRoleofFrameSemanticsinConstructionGrammar.”EnglishLinguistics15:219–42.

———.(2005).“VerbalPolysemyandFrameSemanticsinConstructionGrammar,”inMirjamFriedandHansC.Boas(eds.),GrammaticalConstructions:BacktotheRoots.Amsterdam:

Page 432: The Oxford Handbook of Construction Grammar

References

Page 64 of 88

JohnBenjamins,118–36.

Nesset,Tore(2008).AbstractPhonologyinaConcreteModel:CognitiveLinguisticsandtheMorphology-PhonologyInterface.Berlin:MoutondeGruyter.

Nevalainen,Terttu,andRaumolin-Brunberg,Helena(2003).HistoricalSociolinguistics:LanguageChangeinTudorandStuartEngland.London:PearsonEducation.

Neville,Helen,Nicol,JanetL.,Barss,Andrew,Forster,KennethI.,andGarrett,MerrillF.(1991).“SyntacticallyBasedSentenceProcessingClasses:EvidencefromEvent-relatedBrainPotentials.”JournalofCognitiveNeuroscience3:151–65.

Newell,Allen(1990).UnifiedTheoriesofCognition.Cambridge,MA:MITPress.

Newmeyer,FrederickJ.(2000).LanguageFormandLanguageFunction.Cambridge,MA:MITPress.

Nicolas,Tim(1995).“SemanticsofIdiomModification,”inMartinEveraert,Erik-JanvanderLinden,AndreSchenk,andRobSchreuder(eds.),Idioms:StructuralandPsychologicalPerspectives.Hillsdale,NJ:LawrenceErlbaum,233–52.

Nichols,Johanna(1986).“Head-MarkingandDependent-MarkingGrammar.”Language62:56–119.

———.(1988).“OnAlienableandInalienablePossession,”inWilliamShipley(ed.),InHonorofMaryHaas:FromtheHaasFestivalConferenceonNativeAmericanLinguistics.BerlinandNewYork:MoutondeGruyter,557–609.

———.(1992).LinguisticDiversityinSpaceandTime.Chicago:UniversityofChicagoPress.

———.(2003).“DiversityandStabilityinLanguage,”inBrianJosephandRichardJanda(eds.),TheHandbookofHistoricalLinguistics.Oxford:Blackwell,283–311.

Nikiforidou,Kiki,andCacoullos,RenaTorres(2010).“VariablyFuture-markedConditionalsinGreek:IntegratingDiscourseandGrammar.”ConstructionsandFrames2:90–123.

Ninio,Anat(1999).“PathbreakingVerbsinSyntacticDevelopmentandtheQuestionofPrototypicalTransitivity.”JournalofChildLanguage26:619–53.

———.(2006).LanguageandtheLearningCurve:ANewTheoryofSyntacticDevelopment.Oxford:OxfordUniversityPress.

Nir,Bracha(2008).“ClausePackagesasConstructionsinDevelopingNarrativeDiscourse.”Ph.D.dissertation,TelAvivUniversity.

Nir,Bracha,andBerman,RuthA.(2010a).“PartsofSpeechConstructions:TheCaseofHebrew‘Adverbs’.”ConstructionsandFrames2/2:242–74.

———.(2010b).“ComplexSyntaxasaWindowonContrastiveRhetoric.”JournalofPragmatics42:744–65.

Page 433: The Oxford Handbook of Construction Grammar

References

Page 65 of 88

Noël,Dirk(2007a).“DiachronicConstructionGrammarandGrammaticalizationTheory.”FunctionsofLanguage14/2:177–202.

———.(2007b).“VerbValencyPatterns,ConstructionsandGrammaticalization,”inThomasHerbstandKarinKötz-Votteler(eds.),Valency:Theoretical,DescriptiveandCognitiveIssues.TrendsinLinguistics,StudiesandMonographs187.Berlin:MoutondeGruyter,67–83.

———.(2008).“TheNominativeandInfinitiveinLateModernEnglish:ADiachronicConstructionistApproach.”JournalofEnglishLinguistics36/4:314–40.

Noël,Dirk,andColleman,Timothy(2009).“TheNominativeandInfinitiveinEnglishandDutch:AnExerciseinContrastiveDiachronicConstructionGrammar.”LanguagesinContrast9:144–81.

Noël,Dirk,andvanderAuwera,Johan(2009).“Revisiting<besupposedto>fromaDiachronicConstructionistPerspective.”EnglishStudies90/5:599–623.

Noonan,Michael(1998).“Non-structuralistSyntax,”inMichaelDarnell,EdithMoravcsik,FrederickNewmeyer,MichaelNoonan,andKathleenWheatley(eds.),FunctionalismandFormalisminLinguistics.StudiesinLanguageCompanionSeries41.Amsterdam:JohnBenjamins,11–31.

Nooteboom,Sieb,Weerman,Fred,andWijnen,Frank(eds.)(2002).StorageandComputationintheLanguageFaculty.Dordrecht:Kluwer.

Norde,Muriel(2009).Degrammaticalization.Oxford:OxfordUniversityPress.

Norris,John,andOrtega,Lourdes(2000).“EffectivenessofL2Instruction:AResearchSynthesisandQuantitativeMeta-analysis.”LanguageLearning50:417–528.

Nosofsky,RobertM.(1988).“Similarity,FrequencyandCategoryRepresentations.”JournalofExperimentalPsychology:Learning,MemoryandCognition14:54–65.

Nübling,Damaris(2005).“Vonindieüberin’nundinsbisim:DieKlitisierungvonPräpositionundArtikelals‘Grammatikalisierungsbaustelle’,”inTorstenLeuschner,TanjaMortelmans,andSarahDeGroodt(eds.),GrammatikalisierungimDeutschen.Berlin:MoutondeGruyter,105–31.

Nunberg,Geoffrey,Sag,IvanA.,andWasow,Thomas(1994).“Idioms.”Language70:491–538.

Nuyts,Jan(2007).“CognitiveLinguisticsandFunctionalLinguistics,”inDirkGeeraertsandHubertCuyckens(eds.),TheOxfordHandbookofCognitiveLinguistics.Oxford:OxfordUniversityPress,421–62.

Nyberg,Lars,Petersson,KarlMagnus,Nilsson,Lars-Göran,Sandblom,Johan,Åberg,Carola,andIngvar,Martin(2001).“ReactivationofMotorBrainAreasduringExplicitMemoryforActions.”NeuroImage14:521–28.

Page 434: The Oxford Handbook of Construction Grammar

References

Page 66 of 88

Odlin,Terence(1989).LanguageTransfer.NewYork:CambridgeUniversityPress.

———.(2008).“ConceptualTransferandMeaningExtensions,”inPeterRobinsonandNickC.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge,306–40.

O’Donnell,Matt,andEllis,NickC.(2010).“TowardsanInventoryofEnglishVerbArgumentConstructions.”Proceedingsofthe11thAnnualConferenceoftheNorthAmericanChapteroftheAssociationforComputationalLinguistics,LosAngeles,June1–6,2010,9–16.

OED(TheOxfordEnglishDictionary).http://www.oed.com/.

O’Grady,William(1997).SyntacticDevelopment.Chicago:UniversityofChicagoPress.

Ohala,John(1989).“SoundChangeisDrawnfromaPoolofSynchronicVariation,”inLeivEgilBreivikandErnstHåkonJahr(eds.),LanguageChange:ContributionstotheStudyofitsCauses.BerlinandNewYork:MoutondeGruyter,173–98.

Ohara,KyokoHirose(1994).“AnEvent-reportingRelativeConstructioninJapanese.”BerkeleyLinguisticsSociety20:260–72.

———.(1996).“AConstructionalApproachtoJapaneseInternally-HeadedRelativization.”Ph.D.dissertation,U.C.Berkeley.

———.(2005).“FromRelativizationtoClause-linkage:EvidencefromModernJapanese,”inMirjamFriedandHansC.Boas(eds.),GrammaticalConstructions:BacktotheRoots.Amsterdam:JohnBenjamins,57–70.

———.(2009).“Frame-basedContrastiveLexicalSemanticsinJapaneseFrameNet:TheCaseofriskandkakeru,”inHansC.Boas(ed.),MultilingualFrameNetsinComputationalLexicography:MethodsandApplications.Berlin:MoutondeGruyter,163–82.

———.(2011).“FullTextAnnotationsinJapaneseFrameNet:AnnotatingBCCWJwithSemanticFrames,”(inJapanese.)InAssociationforNaturalLanguageProcessing(eds.),Proceedingsofthe17thAnnualMeetingoftheAssociationforNaturalLanguageProcessing.Tokyo:NakanishiPrintingCompany,703–4.

Ohori,Toshio(2005).“ConstructionGrammarasaConceptualFrameworkforLinguisticTypology:ACasefromReferenceTracking,”inMirjamFriedandHansC.Boas(eds.),GrammaticalConstructions:BacktotheRoots.Amsterdam:JohnBenjamins,215–37.

Onishi,Kristine,Murphy,Gregory,andBock,Kathryn(2008).“PrototypicalityinSentenceProduction.”CognitivePsychology56:103–41.

Ooi,VincentB.Y.(1998).ComputerCorpusLexicography.Edinburgh:EdinburghUniversityPress.

Oosterhof,Albert,andCoussé,Evie(2010).“Form,MeaningandUseofParticipialImperativesinDutch:AConstructionalPerspective.”Papergivenatthe6thInternationalConferenceof

Page 435: The Oxford Handbook of Construction Grammar

References

Page 67 of 88

ConstructionGrammarICCG6,CharlesUniversity,Prague,September3–5,2010.

Ortega,Lourdes,andIberri-Shea,Gina(2005).“LongitudinalResearchinSecondLanguageAcquisition:RecentTrendsandFutureDirections.”AnnualReviewofAppliedLinguistics25:26–45.

Osterhout,Lee,andHagoort,Peter(1999).“ASuperficialResemblancedoesnotNecessarilyMeanyouarePartoftheFamily:CounterargumentstoCoulson,KingandKutas(1998)intheP600/SPS-P300Debate.”LanguageandCognitiveProcesses14:1–14.

Osterhout,Lee,andHolcomb,PhillipJ.(1992).“Event-relatedBrainPotentialsElicitedbySyntacticAnomaly.”JournalofMemoryandLanguage31:785–806.

Östman,Jan-Ola(1999).“CoherencethroughUnderstandingthroughDiscoursePatterns:FocusonNewsReports,”inWolframBublitz,UtaLenk,andEijaVentola(eds.),CoherenceinSpokenandWrittenDiscourse.Amsterdam:JohnBenjamins,77–100.

———.(2002).“SulvanKansanWellerismitKonstruktiona,”inIlonaHerlinetal.(eds.),ÄidinkielenMerkitykset.Helsinki:FinnishLiteratureSociety869,75–97.

———.(2005).“ConstructionDiscourse:AProlegomenon,”inJan-OlaÖstmanandMirjamFried(eds.),ConstructionGrammars:CognitiveGroundingandTheoreticalExtensions.Amsterdam:JohnBenjamins,121–44.

Östman,Jan-Ola,andVirtanen,Tuija(1999).“Theme,CommentandNewnessasFiguresinInformationStructuring,”inKarenvanHoek,AndrejA.Kibrik,andLeoNoordman(eds.),DiscourseStudiesinCognitiveLinguistics.CurrentIssuesinLinguisticTheory176.Amsterdam:JohnBenjamins,91–110.

OxfordEnglishDictionary(1989).http://www.oed.com.

Page,MikeP.A.(2006).“WhatCan’tFunctionalNeuroimagingTelltheCognitivePsychologist?”Cortex42/3:428–43.

Palancar,Enrique(1999).“WhatdoWeGiveinSpanishWhenWeHit?AConstructionistAccountofHittingExpressions.”CognitiveLinguistics10:57–91.

Patten,Amanda(2010).“Grammaticalizationandtheit-cleftConstruction,”inElizabethClossTraugottandGraemeTrousdale(eds.),Gradience,GradualnessandGrammaticalization.Amsterdam:JohnBenjamins,221–43.

Patterson,Karalyn,Nestor,PeterJ.,andRogers,TimothyT.(2007).“WheredoYouKnowWhatYouKnow?TheRepresentationofSemanticKnowledgeintheHumanBrain.”NatureReviewsNeuroscience8:976–87.

Paul,Hermann(1886).“OnSoundChange,”inPhilipBaldiandRonaldN.Werth(eds.),ReadingsinHistoricalPhonology:ChaptersintheTheoryofSoundChange.UniversityPark:PennsylvaniaStateUniversityPress,3–22.

Page 436: The Oxford Handbook of Construction Grammar

References

Page 68 of 88

Pawley,Andrew,andSyder,Frances(1983).“TwoPuzzlesforLinguisticTheory:NativelikeSelectionandNativelikeFluency,”inJackC.RichardsandRichardW.Schmidt(eds.),LanguageandCommunication.London:Longman,191–226.

Pedersen,IngeLise(2010).“TheRoleofSocialFactorsintheShapingofLanguageAttitudes—withanEvaluationoftheConceptofLifeStyle.”InternationalJournaloftheSociologyofLanguage204:129–50.

Pedersen,Johan(2005).“TheSpanishImpersonalse-constructions:ConstructionalVariationandChange.”Constructions1/2005.http://www.constructions-online.de/articles/145.

Pedersen,Ted(1996).“FishingforExactness.”ProceedingsoftheSouthCentralSASUserGroupConference,Austin,TX:SouthCentralSASUserGroup,188–200.

Perdue,Clive(1993).AdultLanguageAcquisition.2vols.Cambridge.CambridgeUniversityPress.

Perfors,Amy,Kemp,Charles,Tenenbaum,Josh,andWonnacott,Elizabeth(2007).“LearningInductiveConstraints:TheAcquisitionofVerbArgumentConstructions.”Papergivenatthe29thAnnualConferenceoftheCognitiveScienceSociety,Nashville,Tennessee,USA,1836.

PernerJosef,Frith,Uta,Leslie,AlanM.,andLeekam,SusanR.(1989).“ExplorationoftheAutisticChild'sTheoryofMind:Knowledge,BeliefandCommunication.”ChildDevelopment60:688–700.

Pesetsky,David(1999).“LinguisticUniversalsandUniversalGrammar,”inRobertA.WilsonandFrankG.Keil(eds.),TheMITEncyclopediaoftheCognitiveSciences.Cambridge,MA:MITPress,476–78.

Peterson,RobertR.,andBurgess,Curt(1993).“SyntacticandSemanticProcessingduringIdiomComprehension:NeurolinguisticandPsycholinguisticDissociations,”inCristinaCacciariandPatriziaTabossi(eds.),Idioms:Processing,StructureandInterpretation.Hillsdale,NJ:Erlbaum,201–25.

Petré,Peter,andCuyckens,Hubert(2008).“Bedusted,yetnotBeheaded:TheRoleofbe-“sConstructionalPropertiesinitsConservation,”inAlexanderBergsandGabrieleDiewald(eds.),ConstructionsandLanguageChange.Berlin:MoutondeGruyter,133–70.

Petruck,MiriamR.L.(1996).“FrameSemantics,”inJefVerschueren,Jan-OlaÖstman,JanBlommaert,andChrisBulcaen(eds.),HandbookofPragmatics1996.Amsterdam:JohnBenjamins,1–11.

Pickering,MartinJ.(2006).“TheDanceofDialogue.”ThePsychologist19:734–37.

Pickering,MartinJ.,andBranigan,Holly(1998).“TheRepresentationofVerbs:EvidencefromSyntacticPersistenceinWrittenLanguageProduction.”JournalofMemoryandLanguage39:633–51.

Pickering,MartinJ.,Branigan,Holly,Cleland,Alexandra,andStewart,Andrew(2000).

Page 437: The Oxford Handbook of Construction Grammar

References

Page 69 of 88

“ActivationofSyntacticInformationduringLanguageProduction.”JournalofPsycholinguisticResearch29:205–16.

Pickering,MartinJ.,Branigan,Holly,andMcLean,Janet(2002).“ConstituentStructureisFormulatedinOneStage.”JournalofMemoryandLanguage46:586–605.

Pickering,MartinJ.,andGarrod,Simon(2004).“TowardaMechanisticTheoryofDialogue.”BehavioralandBrainSciences27:169–226.

Pierrehumbert,JanetB.(2001).“ExemplarDynamics:WordFrequency,LenitionandContrast,”inJoanBybeeandPaulHopper(eds.),FrequencyandtheEmergenceofLinguisticStructure.Amsterdam:JohnBenjamins,137–57.

Pine,JulianM.,andLieven,ElenaV.M.(1993).“ReanalyzingRote-learnedPhrases:IndividualDifferencesintheTransitiontoMulti-wordspeech.”JournalofChildLanguage20:551–71.

Pinker,Steven(1984).LanguageLearnabilityandLanguageDevelopment.Cambridge,MA:HarvardUniversityPress.

———.(1989)LearnabilityandCognition:TheAcquisitionofArgumentStructure.Cambridge,MA:MITPress.

———.(1999).WordsandRules.NewYork:BasicBooks.

Pinker,Steven,andJackendoff,Ray(2005).“TheFacultyofLanguage:What'sSpecialaboutit?”Cognition95:201–36.

Pinker,Steven,Lebeaux,DavidS.,andFrost,LorenAnn(1987).“ProductivityandConstraintsintheAcquisitionofthePassive.”Cognition26:195–267.

Pires,Acrisio,andThomason,SarahG.(2008).“HowMuchSyntacticReconstructionisPossible?,”inGisellaFerraresiandMariaGoldbach(eds.),PrinciplesofSyntacticReconstruction.Amsterdam:JohnBenjamins,27–72.

Polinsky,Maria(1998).“ANon-SyntacticAccountofsomeAsymmetriesintheDoubleObjectConstruction,”inJean-PaulKoenig(ed.),ConceptualStructureandLanguage:BridgingtheGap.Stanford:CSLIPublications,403–23.

Pollard,Carl,andSag,IvanA.(1987).Information-basedSyntaxandSemantics.Vol.1:Fundamentals.Stanford,CA:CSLIPublications.

———.(1994).Head-drivenPhraseStructureGrammar.Chicago:ChicagoUniversityPress.

Posner,MichaelI.,andKeele,StevenW.(1970).“RetentionofAbstractIdeas.”JournalofExperimentalPsychology83:304–8.

PPCEME(ThePennParsedCorpusofEarlyModernEnglish).http://www.ling.upenn.edu/hist-corpora/.

PPCME2(ThePennParsedCorpusofMiddleEnglish).http://www.ling.upenn.edu/hist-

Page 438: The Oxford Handbook of Construction Grammar

References

Page 70 of 88

corpora/.

Prat-Sala,Mercè,andBranigan,Holly(2000).“DiscourseConstraintsonSyntacticProcessinginLanguageProduction:ACross-linguisticStudyinEnglishandSpanish.”JournalofMemoryandLanguage42:168–82.

Prince,EllenF.(1978).“AComparisonofWH-cleftsandIt-cleftsinDiscourse.”Language54/4:883–906.

———.(1981).“TowardaTaxonomyofGiven–newInformation,”inPeterCole(ed.),RadicalPragmatics.NewYork:AcademicPress,223–55.

Pulvermüller,Friedemann(1993).“OnConnectingSyntaxandtheBrain,”inAdAertsen(ed.),BrainTheory—Spatio-temporalAspectsofBrainFunction.NewYork:Elsevier,131–45.

———.(1999).“WordsintheBrain'sLanguage.”BehavioralandBrainSciences22:253–336.

———.(2003a).TheNeuroscienceofLanguage.Cambridge:CambridgeUniversityPress.

———.(2003b).“SequenceDetectorsasaBasisofGrammarintheBrain.”TheoryinBiosciences122:87–103.

———.(2005).“BrainMechanismsLinkingLanguageandAction.”NatureReviewsNeuroscience6/7:576–82.

———.(2010a).“BrainEmbodimentofSyntaxandGrammar:DiscreteCombinatorialMechanismsSpeltoutinNeuronalCircuits.”BrainandLanguage112/3:167–79.

———.(2010b).“BrainLanguageResearch:WhereistheProgress?.”Biolinguistics4/2–3:255–88.

Pulvermüller,Friedemann,andAssadollahi,Ramin(2007).“GrammarorSerialOrder?:DiscreteCombinatorialBrainMechanismsReflectedbytheSyntacticMismatchNegativity.”JournalofCognitiveNeuroscience19/6:971–80.

Pulvermüller,Friedemann,Birbaumer,Niels,Lutzenberger,Werner,andMohr,Bettina(1997).“High-frequencyBrainActivity:ItsPossibleRoleinAttention,PerceptionandLanguageProcessing.”ProgressinNeurobiology52/5:427–45.

Pulvermüller,Friedemann,andFadiga,Luciano(2010).“ActivePerception:SensorimotorCircuitsasaCorticalBasisforLanguage.”NatureReviewsNeuroscience11/5:351–60.

Pulvermüller,Friedemann,Hauk,Olaf,Nikulin,Vadim,andIlmoniemi,RistoJ.(2005).“FunctionalLinksbetweenMotorandLanguageSystems.”EuropeanJournalofNeuroscience21/3:793–97.

Pulvermüller,Friedemann,andHauk,Olaf(2006).“Category-specificProcessingofColorandFormWordsinLeftFronto-temporalCortex.”CerebralCortex16/8:1193–201.

Pulvermüller,Friedemann,Hummel,Friedhelm,andHärle,Markus(2001).“WalkingorTalking?:

Page 439: The Oxford Handbook of Construction Grammar

References

Page 71 of 88

BehavioralandNeurophysiologicalCorrelatesofActionVerbProcessing.”BrainandLanguage78/2:143–68.

Pulvermüller,Friedemann,andKnoblauch,Andreas(2009).“DiscreteCombinatorialCircuitsEmerginginNeuralNetworks:AMechanismforRulesofGrammarintheHumanBrain?”NeuralNetworks22/2:161–72.

Pulvermüller,Friedemann,Kujala,Teija,Shtyrov,Yury,Simola,Jaana,Tiitinen,Hannu,Alku,Paavo,Alku,Kimmo,Martinkauppi,Sami,Ilmoniemi,Risto,andNäätänen,Risto(2001).“MemoryTracesforWordsasRevealedbytheMismatchNegativity.”Neuroimage14/3:607–16.

Pulvermüller,Friedemann,andShtyrov,Yury(2006).“LanguageoutsidetheFocusofAttention:TheMismatchNegativityasaToolforStudyingHigherCognitiveProcesses.”ProgressinNeurobiology79/1:49–71.

Pulvermüller,Friedemann,Shtyrov,Yury,Hasting,Anna,andCarlyon,RobertP.(2008).“SyntaxasaReflex:NeurophysiologicalEvidenceforEarlyAutomaticityofGrammaticalProcessing.”BrainandLanguage104/1:244–53.

Pulvermüller,Friedemann,Shtyrov,Yury,andHauk,Olaf(2009).“UnderstandinginanInstant:NeurophysiologicalEvidenceforMechanisticLanguageCircuitsintheBrain.”BrainandLanguage110/2:81–94.

Pustejovsky,James(1995).TheGenerativeLexicon.Cambridge,MA:MITPress.

Quirk,Randolph,Greenbaum,Sidney,Leech,Geoffr,andSvartvik,Jan(1985).AComprehensiveGrammaroftheEnglishLanguage.London:Longman.

Radden,Günter,andPanther,Klaus-Uwe(2004).StudiesinLinguisticMotivation.Berlin:MoutondeGruyter.

Radford,Andrew(1990).SyntacticTheoryandtheAcquisitionofEnglishSyntax.Cambridge,MA:Blackwell.

Rapp,Brenda,andCaramazza,Alfonso(2002).“SelectiveDifficultieswithSpokenNounsandWrittenVerbs:ASingleCaseStudy.”JournalofNeurolinguistics15:373–402.

Rapp,Brenda,andGoldrick,Matthew(2000).“DiscretenessandInteractivityinSpokenWordProduction.”PsychologicalReview107:460–99.

Reali,Florencia,andChristensen,MortenH.(2005).“UncoveringtheRichnessoftheStimulus:StructureDependenceandIndirectStatisticalEvidence.”CognitiveScience29:1007–28.

Reddy,MichaelJ.(1979).“TheConduitMetaphor:ACaseofFrameConflictinourLanguageaboutLanguage,”inAndrewOrtony(ed.),MetaphorandThought.Cambridge:CambridgeUniversityPress,284–324.

Redington,Martin,Chater,Nick,andFinch,Steven(1998).“DistributionalInformation:APowerfulCueforAcquiringSyntacticCategories.”CognitiveScience22:425–69.

Page 440: The Oxford Handbook of Construction Grammar

References

Page 72 of 88

Regier,Terry,andCarlson,Laura(2002).“SpatialLanguage:PerceptualConstraintsandLinguisticVariation,”inNancyStein,PatriciaBauer,andMitchellRabinowitz(eds.),Representation,MemoryandDevelopment:EssaysinHonorofJeanMandler.Mahwah,NJ:Erlbaum,199–222.

Reichardt,Werner,andVarjú,Dezso(1959).“ÜbertragungseigenschaftenimAuswertesystemfürdasBewegungssehen.”ZeitschriftfürNaturforschung14b:674–89.

Rescorla,RobertA.1968.“ProbabilityofShockinthePresenceandAbsenceofCSinFearConditioning.”JournalofComparativeandPhysiologicalPsychology66:1–5.

Rescorla,RobertA.,andWagner,AllenR.(1972).“ATheoryofPavlovianConditioning:VariationsintheEffectivenessofReinforcementandNonreinforcement,”inAbrahamH.BlackandWilliamF.Prokasy(eds.),ClassicalConditioningII:CurrentTheoryandResearch.NewYork:Appleton-Century-Crofts,64–99.

Roberts,Paul(1956).PatternsofEnglish.NewYork:Harcourt,BraceandWorld.

Robinson,Peter(ed.)(2001).CognitionandSecondLanguageInstruction.Cambridge:CambridgeUniversityPress.

Robinson,Peter,andEllis,NickC.(2008a).“Conclusion:CognitiveLinguistics,SecondLanguageAcquisitionandL2Instruction—IssuesforResearch,”inPeterRobinsonandNickC.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge,489–528.

———.(eds.)(2008b).HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge.

Roelofs,Ardi(1998).“RightwardIncrementalityinEncodingSimplePhrasalFormsinSpeechProduction:Verb–participleCombinations.”JournalofExperimentalPsychology:Learning,MemoryandCognition94:904–21.

Rögnvaldsson,Eiríkur(1995).“OldIcelandic:ANon-ConfigurationalLanguage?”NorthWesternEuropeanLanguageEvolution26:3–29.

Römer,Ute(2008).“CorporaandLanguageTeaching,”inAnkeLüdelingandMerjaKytö.(eds.),CorpusLinguistics:AnInternationalHandbook.Vol.1.Berlin:MoutondeGruyter,112–30.

———.(2009).“TheInseparabilityofLexisandGrammar:CorpusLinguisticPerspectives.”AnnualReviewofCognitiveLinguistics7:141–63.

Rosch,Eleanor(1973).“NaturalCategories.”CognitivePsychology4:328–50.

———.(1978).“PrinciplesofCategorization,”inEleanorRoschandBarbaraB.Lloyd(eds.),CognitionandCategorization.Hillsdale,NJ:LawrenceErlbaum,27–48.

Rosch,Eleanor,andMervis,CarolynB.(1975a).“FamilyResemblances:StudiesintheInternal

Page 441: The Oxford Handbook of Construction Grammar

References

Page 73 of 88

StructureofCategories.”CognitivePsychology7:573–605.

———.(1975b).“CognitiveRepresentationsofSemanticCategories.”JournalofExperimentalPsychology:General104:192–233.

Rosch,Eleanor,Mervis,CarolynB.,Gray,WayneD.,Johnson,DavidM.,andBoyes-Braem,Penny(1976).“BasicObjectsinNaturalCategories.”CognitivePsychology8:382–439.

Rosenbach,Anette(2002).“TheEnglishs-genitive:ACaseofDegrammaticalization?”inOlgaFischer,MurielNorde,andHarryPerridon(eds.),UpandDowntheCline—TheNatureofGrammaticalization.Amsterdam:JohnBenjamins,73–96.

Ross,BrianH.,andMakin,ValerieS.(1999).“PrototypeversusExemplarModels,”inRobertJ.Sternberg(ed.),TheNatureofCognition.Cambridge,MA:MITPress,205–41.

Ross,JohnR.(1986).InfiniteSyntax!Norwood,NJ:AblexPublishingCorporation.

Ross,Malcolm(2008).“NegativeVerbalClauseConstructionsinPuyuma:ExploringConstructionalDisharmony,”inAlexanderBergsandGabrieleDiewald(eds.),ConstructionsandLanguageChange.Berlin,NewYork:MoutondeGruyter:171–93.

Rowland,CarolineF.(2007).“ExplainingErrorsinChildren'sQuestions:AuxiliaryDOandModalAuxiliaries.”Cognition104:106–34.

Rowland,CarolineF.,andPine,JulianM.(2000).“Subject-auxiliaryInversionErrorsandWh-questionAcquisition:WhatChildrendoKnow?”JournalofChildLanguage27:157–81.

Rumelhardt,David,andMcClelland,James(1986).“OnLearningthePastTensesofEnglishVerbs,”inJamesMcClellandandDavidRumelhardt(eds.),ParallelDistributedProcessing.Vol.2:PsychologicalandBiologicalModels.Cambridge,MA:MITPress,216–71.

Sadler,Louisa,andSpencer,Andrew(2001).“SyntaxasanExponentofMorphologicalFeatures,”inGeertBooijandJaapVanMarle(eds.),YearbookofMorphology2000.Dordrecht:Kluwer,71–96.

Sadock,Jerrold(1991).AutolexicalSyntax.Chicago:UniversityofChicagoPress.

Saffran,JennyR.(2001).“TheUseofPredictiveDependenciesinLanguageLearning.”JournalofMemoryandLanguage44:493–515.

———.(2002).“ConstraintsonStatisticalLanguageLearning.”JournalofMemoryandLanguage47:172–96.

Sag,IvanA.(1997).“EnglishRelativeClauseConstructions.”JournalofLinguistics33:431–84.

———.(2007).“RemarksonLocality,”inStefanMüller(ed.),ProceedingsoftheHPSG07Conference,StanfordUniversity.Stanford,CA:CSLIPublications,394–414.

———.(2010).“EnglishFiller-gapConstructions.”Language86/3:486–545.

Page 442: The Oxford Handbook of Construction Grammar

References

Page 74 of 88

———.(2012).“Sign-BasedConstructionGrammar:AnInformalSynopsis,”inHansC.BoasandIvanA.Sag(eds.),Sign-BasedConstructionGrammar.Stanford,CA:CSLIPublications,39–170.

Sag,IvanA.,Hofmeister,Philip,andSnider,Neal(2007).“ProcessingComplexityinSubjacencyViolations:TheComplexNounPhraseConstraint.”ChicagoLinguisticsSociety43/1:215–29.

Sag,IvanA.,Wasow,Thomas,andBender,EmilyM.(2003).SyntacticTheory:AFormalIntroduction.Stanford,CA:CSLIPublications.

Sankoff,David,Tagliamonte,Sali,andSmith,Eric(2005).GoldvarbX:AVariableRuleApplicationforMacintoshandWindows.DepartmentofLinguistics,UniversityofToronto.http://individual.utoronto.ca/tagliamonte/Goldvarb/GV_index.htm.

Santelmann,Lynn,Berk,Stephanie,Austin,Jennifer,Somashekar,ShamithaandLust,Barbara(2002).“ContinuityandDevelopmentintheAcquisitionofInversioninYes/NoQuestions:DissociatingMovementandInflection.”JournalofChildLanguage29:813–42.

Sarangi,Srikant(2010).“ReconfiguringSelf/Identity/Status/Role:TheCaseofProfessionalRolePerformanceinHealthcareEncounters,”inGiulianaGarzoneandJamesArchibald(eds.),Discourse,IdentitiesandRolesinSpecializedCommunication.Bern:PeterLang:33–57.

Sato,Manami(2010).“Messageinthe‘Body’:EffectsofSimulationinSentenceProduction.”Ph.D.dissertation.LinguisticsDepartment,UniversityofHawaii,Manoa.

Savage,Ceri,Lieven,Elena,Theakston,AnnaandTomasello,Michael(2003).“TestingtheAbstractnessofChildren'sLinguisticRepresentations:LexicalandStructuralPrimingofSyntacticConstructions.”DevelopmentalScience6:557–67.

———.(2006).“StructuralPrimingasImplicitLearninginLanguageAcquisition:ThePersistenceofLexicalandStructuralPrimingin4-year-olds.”LanguageLearningandDevelopment2:27–49.

Schiering,René,Bickel,Balthasar,andHildebrandt,KristineA.(2010).“TheProsodicWordisnotUniversal.”JournalofLinguistics46:657–709.

Schmid,Hans-Jörg(2010).“DoesFrequencyinTextInstantiateEntrenchmentintheCognitiveSystem?”inDylanGlynnandKerstinFischer(eds.),QuantitativeMethodsinCognitiveSemantics.BerlinandNewYork:MoutondeGruyter,101–34.

Schmidt,Richard(1990).“TheRoleofConsciousnessinSecondLanguageLearning.”AppliedLinguistics11:129–58.

———.(2001).“Attention,”inPeterRobinson(ed.),CognitionandSecondLanguageInstruction.Cambridge:CambridgeUniversityPress,3–32.

Schmied,Josef(2004).“EastAfricanEnglishPhonology,”inBerndKortmann,KateBurridge,RajendMesthrie,EdgarW.Schneider,andCliveUpton(eds.),AHandbookofVarietiesofEnglish.Vol.1:Phonology.BerlinandNewYork:MoutondeGruyter,918–30.

Page 443: The Oxford Handbook of Construction Grammar

References

Page 75 of 88

Schmitt,Norbert(ed.)(2004).FormulaicSequences.Amsterdam:JohnBenjamins.

Schneider,EdgarW.(2003).“HowtoTraceStructuralNativization:ParticleVerbsinWorldEnglishes.”WorldEnglishes23:227–49.

———.(2007).PostcolonialEnglish:VarietiesAroundtheWorld.Cambridge:CambridgeUniversityPress.

Schneider,Nathan(2010).“ComputationalCognitiveMorphosemantics:ModelingMorphologicalCompositionalityinHebrewVerbswithEmbodiedConstructionGrammar.”BerkeleyLinguisticsSociety36.

Schone,Patrick,andJurafsky,Daniel(2001).“IsKnowledge-freeInductionofMultiwordUnitDictionaryHeadwordsaSolvedProblem?”Proceedingsofthe6thConferenceonEmpiricalMethodsinNaturalLanguageProcessing:100’8.http://www.aclweb.org/anthology/W/W01/#0500.

Schönefeld,Doris(2006).“FromConceptualizationtoLinguisticExpression:WhereLanguagesDiversify,”inStefanTh.GriesandAnatolStefanowitsch(eds.),CorporainCognitiveLinguistics:TheSyntax-LexisInterface.BerlinandNewYork:MoutondeGruyter,297–344.

Schuchardt,Hugo(1885).ÜberdieLautgesetze:GegendieJunggrammatiker.Berlin:R.Oppenheim.

Segerståhl,Sinikka(2011).VaggvisoriKvevlax:SpråkligaStrukturerochKonstruktioner.Helsinki:NordicaHelsingiensia24.

Sereno,SaraC.,Rayner,Keith,andPosner,MichaelI.,1998.“EstablishingaTime-lineofWordRecognition:EvidencefromEyeMovementsandEvent-relatedPotentials.”NeuroReport9:2195–200.

Shanks,DavidR.(1995).ThePsychologyofAssociativeLearning.NewYork:CambridgeUniversityPress.

Shieber,Stuart(1986).AnIntroductiontoUnification-BasedApproachestoGrammar.Stanford,CA:CSLIPublications.

Shimpi,Priya,Gámez,PerlaB.,Huttenlocher,Janellen,andVasilyeva,Marina.(2007).“UsingSyntacticPrimingtoTrackEmergingLinguisticRepresentationsofTransitiveandDativeConstructions.”DevelopmentalPsychology43/6:1334–46.

Shtyrov,Yury,andPulvermüller,Friedemann(2002).“NeurophysiologicalEvidenceofMemoryTracesforWordsintheHumanBrain.”Neuroreport13:521–25.

Shtyrov,Yury,Pulvermüller,Friedemann,Näätänen,Risto,andIlmoniemi,RistoJ.(2003).“GrammarProcessingoutsidetheFocusofAttention:AnMEGStudy.”JournalofCognitiveNeuroscience15/8:1195–206.

Siewierska,Anna(1988).WordOrderRules.London:CroomHelm.

Page 444: The Oxford Handbook of Construction Grammar

References

Page 76 of 88

Siewierska,Anna,andHollmann,Willem(2007).“DitransitiveClausesinEnglish,withSpecialReferencetoLancashireEnglish,”inMichaelHannayandGerardJ.Steen(eds.),Structural-functionalStudiesinEnglishGrammar.Amsterdam:JohnBenjamins,83–102.

Simmons,W.Kyle,Ramjee,Vimal,Beauchamp,MichaelS.,McRae,Ken,Martin,Alex,andBarsalou,LawrenceW.(2007).“ACommonNeuralSubstrateforPerceivingandKnowingAboutColor.”Neuropsychologia45/12:2802–10.

Simpson,Jane(1983).“Resultatives,”inLoriLevin,MalkaRappaport,andAnnieZaenen(eds.),PapersinLexical-functionalGrammar.Bloomington:IndianaUniversityLinguisticsClub,143–57.

Sinclair,John(ed.)(1987).Lookingup:AnAccountoftheCOBUILDProjectinLexicalComputing.London:HarperCollins.

———.(1991).Corpus,Concordance,Collocation.Oxford:OxfordUniversityPress.

———.(ed.)(1996a).HowtoUseCorporainLanguageTeaching.Amsterdam:JohnBenjamins.

———.(1996b).“TheSearchforUnitsofMeaning.”Textus9:75–106.

Sinha,Chris(1999).“Grounding,MappingandActsofMeaning,”inTheoJanssenandGiselaRedeker(eds.),CognitiveLinguistics:Foundations,ScopeandMethodology.BerlinandNewYork:MoutondeGruyter,223–55.

———.(2004).“TheEvolutionofLanguage:FromSignalstoSymbolstoSystem,”inD.KimbroughOllerandUlrikeGriebel(eds.),EvolutionofCommunicationSystems:AComparativeApproach.Cambridge,MA:MITPress,217–35.

Skousen,Royal(1989).AnalogicalModelingofLanguage.Dordrecht:Kluwer.

Slobin,DanI.(1993).“AdultLanguageAcquisition:AViewfromChildLanguageStudy,”inClivePerdue(ed.),AdultLanguageAcquisition:Cross-linguisticPerspectives.Cambridge:CambridgeUniversityPress,239–52.

———.(1996).“From“ThoughtandLanguage”to“ThinkingforSpeaking”,”inJohnJ.GumperzandStephenC.Levinson(eds.),RethinkingLinguisticRelativity.Cambridge:CambridgeUniversityPress,70–96.

———.(1997).“TheOriginsofGrammaticizableNotions:BeyondtheIndividualMind,”inDanI.Slobin(ed.),TheCrosslinguisticStudyofLanguageAcquisition.Mahwah,NJ:Erlbaum,265–323.

Slobin,DanI.,andBever,ThomasG.(1982).“ChildrenUseCanonicalSentenceSchemas:ACross-LinguisticStudyofWordOrderandInflection.”Cognition12:229–65.

Smith,K.Aaron,andNordquist,Dawn(2012).“ACriticalandHistoricalInvestigationintoSemanticProsody.”JournalofHistoricalPragmatics13/2:291–312.

Page 445: The Oxford Handbook of Construction Grammar

References

Page 77 of 88

Smith,MichaelB.(1994).“AgreementandIconicityinRussianImpersonalConstructions.”CognitiveLinguistics5:5–56.

Smitherman,Thomas,andBarðdal,Jóhanna(2009).“TypologicalChangesintheEvolutionofIndo-EuropeanSyntax?”Diachronica26/2:253–73.

Smolensky,Paul(1999).“Grammar-basedConnectionistApproachestoLanguage.”CognitiveScience23/4:589–613.

Sneddon,JamesM.(1996).Indonesian:AComprehensiveGrammar.LondonandNewYork:Routledge.

Spada,Nina(1997).“Form-focusedInstructionandSecondLanguageAcquisition:AReviewofClassroomandLaboratoryResearch.”LanguageTeachingResearch30:73–87.

Speelman,Dirk,Grondelaers,Stefan,andGeeraerts,Dirk(2008).“VariationintheChoiceofAdjectivesintheVarietiesofDutch,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,205–33.

Speelman,Dirk,Tummers,Jose,andGeeraerts,Dirk(2009).“LexicalPatterninginaConstructionGrammar:TheEffectofLexicalCo-occurrencePatternsontheInflectionalVariationinDutchAttributiveAdjectives.”ConstructionsandFrames1/1:87–118.

Spencer,Andrew(2001).“TheParadigm-BasedModelofMorphosyntax.”TransactionsofthePhilologicalSociety99/2:279–314.

———.(2004).“Morphology—AnOverviewofCentralConcepts,”inLouisaSadlerandAndrewSpencer(eds.),ProjectingMorphology.Stanford,CA:CSLIPublications,67–109.

Spivey,Michael(2006).TheContinuityofMind.Oxford:OxfordUniversityPress.

SPRÅKBANKEN(variousSwedishcorpora).http://spraakbanken.gu.se/.

Spranger,Michael,andLoetzsch,Martin(2011).“SyntacticIndeterminacyandSemanticAmbiguity:ACaseStudyforGermanSpatialPhrases.”inLucSteels(ed.),DesignPatternsinFluidConstructionGrammar.Amsterdam:JohnBenjamins,265–98.

Stallings,LynneM.,MacDonald,MaryellenC.,andO'seaghdha,Padraig(1998).“PhrasalOrderingConstraintsinSentenceProduction:PhraseLengthandVerbDispositioninHeavy-NPShift.”JournalofMemoryandLanguage39:392–417.

Stanfield,RobertA.,andZwaan,RolfA.(2001).“TheEffectofImpliedOrientationDerivedfromVerbalContextonPictureRecognition.”PsychologicalScience12:153–56.

Stassen,Leon(1985).ComparisonandUniversalGrammar.Oxford:Blackwell.

———.(1997).IntransitivePredication.Oxford:OxfordUniversityPress.

———.(2009).PredicativePossession.Oxford:OxfordUniversityPress.

Page 446: The Oxford Handbook of Construction Grammar

References

Page 78 of 88

Steels,Luc(1998).“TheOriginsofSyntaxinVisuallyGroundedRoboticAgents.”ArtificialIntelligence103:133–56.

———.(2004).“ConstructivistDevelopmentofGroundedConstructionGrammars,”inDoniaScott,WalterDaelemans,andMarilynWalker(eds.),ProceedingsoftheAnnualMeetingoftheAssociationforComputationalLinguisticsConference,9–16.Barcelona:ACL.http://acl.ldc.upenn.edu/P/P04/.

———.(2007).“TheRecruitmentTheoryofLanguageOrigins,”inCarolineLyon,ChrystopherNehaniv,andAngeloCangelosi(eds.),TheEmergenceofCommunicationandLanguage.Berlin:SpringerVerlag,129–51.

———.(ed.)(2011a).DesignPatternsinFluidConstructionGrammar.Amsterdam:JohnBenjamins.

———.(ed.)(2011b).ComputationalIssuesinFluidConstructionGrammar.Berlin:SpringerVerlag.

———.(2012).“Introduction:Self-organizationandSelectioninLanguageEvolution,”inLucSteels(ed.),ExperimentsinLanguageEvolution.AdvancesinInteractionStudies3.Amsterdam:JohnBenjamins.

Steels,Luc,andDeBeule,Joachim(2006).“UnifyandMergeinFluidConstructionGrammar,”inPaulVogt,YuugaSugita,ElioTuci,andChrystopherNehaniv(eds.),SymbolGroundingandBeyond:ProceedingsoftheThirdInternationalWorkshopontheEmergenceandEvolutionofLinguisticCommunication,LNAI4211.Berlin:Springer,197–223.

Stefanowitsch,Anatol(2003).“TheEnglishImperative:AConstruction-basedApproach.”Ms.,UniversitätBremen.

———.(2005).“TheFunctionofMetaphor:DevelopingaCorpus-basedPerspective.”InternationalJournalofCorpusLinguistics10/2:161–98.

———.(2006a).“NegativeEvidenceandtheRawFrequencyFallacy.”CorpusLinguisticsandLinguisticTheory2/1:61–77.

———.(2006b).“KonstruktionsgrammatikundKorpuslinguistik,”inKerstinFischerandAnatolStefanowitsch(eds.),Konstruktionsgrammatik:VonderAnwendungzurTheorie.Tübingen:Stauffenburg,151–76.

———.(2006c).“DistinctiveCollexemeAnalysisandDiachrony:AComment.”CorpusLinguisticsandLinguisticTheory2/2:257–62.

———.(2008a).“NegativeEvidenceandPreemption:AConstructionalApproachtoUngrammaticality.”CognitiveLinguistics19/3:513–31.

———.(2008b).“NegativeEntrenchment:AUsage-basedApproachtoNegativeEvidence.”CognitiveLinguistics19:513–31.

Page 447: The Oxford Handbook of Construction Grammar

References

Page 79 of 88

———.(2009).“BedeutungundGebrauchinderKonstruktionsgrammatik:WieKompositionellsindModaleInfinitiveimDeutschen?.”ZeitschriftfürGermanistischeLinguistik37/3:565–92.

———.(2011a).“ConstructionalPreemptionbyContextualMismatch:ACorpus-linguisticInvestigation.”CognitiveLinguistics22/1:107–30.

———.(2011b).“CognitiveLinguisticsMeetstheCorpus,”inMarioBrdar,MilenaŽicFuchs,andStefanTh.Gries(eds.),ExpandingCognitiveLinguisticHorizons.Amsterdam:JohnBenjamins,257–90.

Stefanowitsch,Anatol,andGries,StefanTh.(2003).“Collostructions:InvestigatingtheInteractionofWordsandConstructions.”InternationalJournalofCorpusLinguistics8/2:209–43.

———.(2005).“Co-varyingCollexemes.”CorpusLinguisticsandLinguisticTheory1/1:1–43.

———.(2008).“ChannelandConstructionalMeaning:ACollostructionalCaseStudy,”inGitteKristiansenandRenéDirven(eds.),CognitiveSociolinguistics:LanguageVariation,CulturalModels,SocialSystems.BerlinandNewYork:MoutondeGruyter,129–52.

———.(2009).“CorporaandGrammar,”inAnkeLüdelingandMerjaKytö(eds.),CorpusLinguistics:AnInternationalHandbook.BerlinandNewYork:MoutondeGruyter,933–52.

Street,RichardL.,Jr.,andGiles,Howard(1982).“SpeechAccommodationTheory:ASocialCognitiveApproachtoLanguageandSpeechBehavior,”inMichaelE.RoloffandCharlesR.Berger(eds.),SocialCognitionandCommunication.BeverlyHills:SagePublications,193–226.

Stubbs,Michael(1996).TextandCorpusAnalysis.Oxford:Blackwell.

Suttle,Laura,andGoldberg,AdeleE.(2011).“ThePartialProductivityofConstructionsasInduction.”Linguistics49/6:1237–69.

Swales,JohnM.(1990).GenreAnalysis—EnglishinAcademicandResearchSettings.Cambridge:CambridgeUniversityPress.

Szmrecsanyi,Benedikt(2005).“LanguageUsersasCreaturesofHabit:ACorpus-basedAnalysisofPersistenceinSpokenEnglish.”CorpusLinguisticsandLinguisticsTheory1/1:113–49.

Szymanek,Bogdan(2009).“IE,Slavonic:Polish,”inRochelleLieberandPavolStekauer(eds.),TheOxfordHandbookofCompounding.Oxford:OxfordUniversityPress,464–77.

Tagliamonte,Sali(2004).“Haveto,gotta,must.Grammaticalisation,VariationandSpecializationinEnglishDeonticModality,”inHansLindquistandChristianMair(eds.),CorpusApproachestoGrammaticalisationinEnglish.Amsterdam:JohnBenjamins,33–55.

———.(2006).AnalysingSociolinguisticVariation.Cambridge:CambridgeUniversityPress.

Tagliamonte,Sali,Poplack,Shana,andEze,Ejike(1997).“PluralizationPatternsinNigerian

Page 448: The Oxford Handbook of Construction Grammar

References

Page 80 of 88

PidginEnglish.”JournalofPidginandCreoleLanguages12:103–29.

Tagliamonte,Sali,andRoeder,RebeccaV.(2009).“VariationintheEnglishDefiniteArticle:Socio-historicalLinguisticsint'speechCommunity.”JournalofSociolinguistics13:435–71.

Takahashi,Hidemitsu(2004).“TheEnglishImperative:ACognitiveandFunctionalAnalysis.”Hokaido:HokaidoUniversitydissertation.

Talmy,Leonard(1976).“SemanticCausativeTypes,”inMasayoshiShibatani(ed.),SyntaxandSemantics6:TheGrammarofCausativeConstructions.NewYork:AcademicPress,43–116.

———.(1988).“ForceDynamicsinLanguageandCognition.”CognitiveScience12:49–100.

———.(2000).TowardaCognitiveSemantics.Vol.1:ConceptStructuringSystems.Vol.2:TypologyandProcessinConceptStructuring.Cambridge,MA:MITPress.

———.(2007).“Foreword,”inandMichaelJ.Spivey(eds.),MethodsinCognitiveLinguistics.Amsterdam:JohnBenjamins,xi–xxi.

Taylor,JohnR.(1995).LinguisticCategorization.Oxford:OxfordUniversityPress.

———.(1996).PossessivesinEnglish:AnExplorationinCognitiveGrammar.Oxford:Clarendon.

———.(2002).CognitiveGrammar.Oxford:OxfordUniversityPress.

———.(2006).“PolysemyandtheLexicon,”inGitteKristiansen,MichelAchard,RenéDirven,andFranciscoJ.RuizdeMendozaIbáñez(eds.),CognitiveLinguistics:CurrentApplicationsandFuturePerspectives.Berlin:MoutondeGruyter,51–80.

———.(2007).“CognitiveLinguisticsandAutonomousLinguistics,”inDirkGeeraertsandHubertCuyckens(eds.),TheOxfordHandbookofCognitiveLinguistics.Oxford:OxfordUniversityPress,421–62.

Tesnière,Lucien(1959).ÉlémentsdeSyntaxeStructurale.Paris:ÉditionsKlincksieck.

Thakerar,JitendraN.,Giles,Howard,andCheshire,Jenny(1982).“PsychologicalandLinguisticParametersofSpeechAccommodationTheory,”inColinFraserandKlausR.Scherer(eds.),AdvancesintheSocialPsychologyofLanguage.NewYork:CambridgeUniversityPress,205–55.

Thompson,SandraA.,andKoide,Yuka(1987).“Iconicityand‘IndirectObjects’inEnglish.”JournalofPragmatics11/3:399–406.

Timberlake,Alan(1977).“ReanalysisandActualizationinSyntacticChange,”inCharlesN.Li(ed.),MechanismsofSyntacticChange.Austin:UniversityofTexasPress,141–80.

TIME(TheTimeMagazine–fromthe1920stothe2000s).http://corpus.byu.edu/time/.

Timyam,Napasri,andBergen,Benjamin(2010).“AContrastiveStudyoftheCaused-motion

Page 449: The Oxford Handbook of Construction Grammar

References

Page 81 of 88

andDitransitiveConstructionsinEnglishandThai:SemanticandPragmaticConstraints,”inHansC.Boas(ed.),ContrastiveStudiesinConstructionGrammar.Amsterdam:JohnBenjamins,137–68.

Titone,DebraA.,andConnine,CynthiaM.(1994).“TheComprehensionofIdiomaticExpressions:EffectsofPredictabilityandLiterality.”JournalofExperimentalPsychology:Learning,MemoryandCognition20/5:1126–38.

Tomasello,Michael(1992).FirstVerbs:ACaseStudyofEarlyGrammaticalDevelopment.Cambridge:CambridgeUniversityPress.

———.(1998a).“CognitiveLinguistics,”inWilliamBechtelandGeorgeGraham(eds.),ACompaniontoCognitiveScience.Oxford:Blackwell,477–87.

———.(ed.)(1998b).TheNewPsychologyofLanguage:CognitiveandFunctionalApproachestoLanguageStructure.Mahwah,NJ:Erlbaum.

———.(1999).TheCulturalOriginsofHumanCognition:AnEssay.Cambridge,MA:HarvardUniversityPress.

———.(2000a).“TheItem-basedNatureofChildren'sEarlySyntacticDevelopment.”TrendsinCognitiveSciences4:156–63.

———.(2000b).“DoYoungChildrenhaveAdultSyntacticCompetence?”Cognition74:209–53.

———.(2003).ConstructingaLanguage:AUsage-BasedTheoryofLanguageAcquisition.Cambridge,MA:HarvardUniversityPress.

———.(2006).“ConstructionGrammarforKids.”ConstructionsSpecialVolume2006/1.http://elanguage.net/journals/index.php/constructions.

———.(2008).OriginsofHumanCommunication.Cambridge,MA:MITPress.

Tomasello,Michael,Aktar,Nameera,Dodson,Kelly,andRekau,Laura(1997).“DifferentialProductivityinYoungChildren'sUseofNounsandVerbs.”JournalofChildLanguage24:373–87.

Tomasello,Michael,andBrooks,Patricia(1998).“YoungChildren'sEarliestTransitiveandIntransitiveConstructions.”CognitiveLinguistics9:379–95.

Tomlin,RussellS.,Forrest,Linda,Pu,MingMing,andKim,Myung-Hee(1997).“DiscourseSemantics,”inTeunA.VanDijk(ed.),DiscourseasStructureandProcess.London:Sage,63–111.

Traugott,ElizabethCloss(1982).“FromPropositionaltoTextualandExpressiveMeanings:SomeSemantic-pragmaticAspectsofGrammaticalization,”inWinfredP.LehmannandYakovMalkiel(eds.),PerspectivesonHistoricalLinguistics.Amsterdam:JohnBenjamins,245–71.

Page 450: The Oxford Handbook of Construction Grammar

References

Page 82 of 88

———.(1988).“PragmaticStrengtheningandGrammaticalization.”BerkeleyLinguisticSociety14:406–16.

———.(1989).“OntheRiseofEpistemicMeaningsinEnglish:AnExampleofSubjectificationinSemanticChange.”Language65:31–55.

———.(2003a).“ConstructionsinGrammaticalization,”inBrianD.JosephandRichardD.Janda(eds.),TheHandbookofHistoricalLinguistics.Oxford:Blackwell,624–47.

———.(2003b).“FromSubjectificationtoIntersubjectification,”inRaymondHickey(ed.),MotivesforLanguageChange.Cambridge:CambridgeUniversityPress,124–39.

———.(2007).“TheConceptsofConstructionalMismatchandType-shiftingfromthePerspectiveofGrammaticalization.”CognitiveLinguistics18:523–57.

———.(2008a).“TheGrammaticalizationofNPofNPPatterns,”inAlexanderBergsandGabrieleDiewald(eds.),ConstructionsandLanguageChange.Berlin:MoutondeGruyter,23–45.

———.(2008b).“‘AllThatheEndeavouredtoProvewas…’:OntheEmergenceofGrammaticalConstructionsinDialogualContexts,”inRuthKempsonandRobinCooper(eds.),LanguageChangeandEvolution.London:KingsCollegePublications,143–77.

———.(2008c).“Grammaticalization,ConstructionsandtheIncrementalDevelopmentofLanguage:SuggestionsfromtheDevelopmentofDegreeModifiersinEnglish,”inRegineEckardt,GerhardJäger,andTonjesVeenstra(eds.),Variation,Selection,Development:ProbingtheEvolutionaryModelofLanguageChange.BerlinandNewYork:MoutondeGruyter,219–50.

———.(2008d).“Grammatikalisierung,EmergenteKonstruktionenundderBegriffder‘Neuheit’,”inAnatolStefanowitschandKerstinFischer(eds.),KonstruktionsgrammatikII:VonderKonstruktionzurGrammatik.Tübingen:StauffenburgVerlag,5–32.

———.(2010a).“ContextsandGrammaticalConstructionalization.”PlenarylecturepresentedattheSixthInternationalConferenceonConstructionGrammar(ICCG-6),Prague,September2010.

———.(2010b).“(Inter)subjectivityand(Inter)subjectification:AReassessment,”inKristinDavidse,LievenVandelanotte,andHubertCuyckens(eds.),ContextsandConstructions.Amsterdam:JohnBenjamins,145–92.

Traugott,ElizabethCloss,andDasher,RichardB.(2002).RegularityinSemanticChange.Cambridge:CambridgeUniversityPress.

Traugott,ElizabethCloss,andTrousdale,Graeme(2010).“Gradience,GradualnessandGrammaticalization:HowdoTheyIntersect?,”inElizabethClossTraugottandGraemeTrousdale(eds.),Gradience,GradualnessandGrammaticalization.TypologicalStudiesinLanguage90.Amsterdam:JohnBenjamins,19–44.

Page 451: The Oxford Handbook of Construction Grammar

References

Page 83 of 88

Trotta,Joe(2000).Wh-clausesinEnglish:AspectsofTheoryandDescription.AmsterdamandPhiladelphia:Rodopi.

Trousdale,Graeme(2008a).“WordsandConstructionsinGrammaticalization:TheEndoftheEnglishImpersonalConstruction,”inSusanM.FitzmauriceandDonkaMinkova(eds.)StudiesintheHistoryoftheEnglishLanguageIV:EmpiricalandAnalyticalAdvancesintheStudyofEnglishLanguageChange.BerlinandNewYork:MoutondeGruyter,301–26.

———.(2008b).“AConstructionalApproachtoLexicalizationProcessesintheHistoryofEnglish:EvidencefromPossessiveConstructions.”WordStructure1:156–77.

———.(2008c).“ConstructionsinGrammaticalizationandLexicalization:EvidencefromtheHistoryofaCompositePredicateConstructionintheHistoryofEnglish,”inGraemeTrousdaleandNikolasGisborne(eds.),ConstructionalApproachestoEnglishGrammar.TopicsinEnglishLinguistics57.BerlinandNewYork:MoutondeGruyter,33–67.

Trousdale,Graeme,andGisborne,Nikolas(eds.)(2008).ConstructionalApproachestoEnglishGrammar.TopicsinEnglishLinguistics57.BerlinandNewYork:MoutondeGruyter.

Trudgill,Peter(1974).TheSocialDifferentiationofEnglishinNorwich.Cambridge:CambridgeUniversityPress.

Tummers,Jose,Heylen,Kris,andGeeraerts,Dirk(2005).“Usage-basedApproachesinCognitiveLinguistics:ATechnicalStateoftheArt.”CorpusLinguisticsandLinguisticTheory1/2:225–61.

Tyler,Andrea(2008).“CognitiveLinguisticsandSecondLanguageInstruction,”inPeterRobinsonandN.C.Ellis(eds.),HandbookofCognitiveLinguisticsandSecondLanguageAcquisition.London:Routledge,456–88.

Ullman,MichaelT.(2001).“ANeurocognitivePerspectiveonLanguage:TheDeclarative/ProceduralModel.”NatureReviewsNeuroscience2/10:717–26.

vanCanegem-Ardijns,Ingrid(2006).“TheExtrapositionofPrepositionalObjectsofAdjectivesinDutch.”Linguistics44:425–57.

vanderAuwera,Johan(1985).“Relativethat—aCentennialDispute.”JournalofLinguistics21:149–79.

vanderAuwera,Johan,andMalchukov,Andrej(2005).“ASemanticMapforDepictiveAdjectivals,”inNikolausP.HimmelmannandEvaF.Schultze-Berndt(eds.),SecondaryPredicationandAdverbialModification:TheTypologyofDepictives.Oxford:OxfordUniversityPress,393–421.

vanderAuwera,Johan,andPlungian,VladimirA.(1998).“Modality'sSemanticMap.”LinguisticTypology2:79–124.

VanderLeek,Frederike(2000).“Caused-motionandthe‘Bottom-up’RoleofGrammar,”inAdFoolenandFrederikevanderLeek(eds.),ConstructionsinCognitiveLinguistics.Amsterdam:

Page 452: The Oxford Handbook of Construction Grammar

References

Page 84 of 88

JohnBenjamins,301–31.

VanGoethem,Kristel(2008).“Oud-leerlingversusancienélève:AComparativeStudyofAdjectivesGrammaticalizingintoPrefixesinDutchandFrench.”Morphology18:27–49.

VanNice,KathyY.,andDietrich,Rainer.(2003).“Task-sensitivityofAnimacyEffects:EvidencefromGermanPictureDescriptions.”Linguistics41/5:825–49.

VanPatten,Bill(2006).“InputProcessinginAdultSecondLanguageAcquisition,”inBillVanPattenandJessicaWilliams(eds.),TheoriesinSecondLanguageAcquisition:AnIntroduction.Mahwah,NJ:LawrenceErlbaum,115–35.

VanTrijp,Remi(2011).“FeatureMatricesandAgreement,”inLucSteels(ed.),FluidConstructionGrammar.Amsterdam:JohnBenjamins,205–36.

VanValin,RobertD.,Jr.(1980).“OntheDistributionofPassiveandAntipassiveConstructionsinUniversalGrammar.”Lingua50:303–27.

———.(1998).“TheAcquisitionofWh-QuestionsandtheMechanismsofLanguageAcquisition,”inMichaelTomasello(ed.),TheNewPsychologyofLanguage:CognitiveandFunctionalApproachestoLanguageStructure.Hillsdale,NJ:LawrenceErlbaum,221–49.

VanValin,RobertD.,Jr.,andLaPolla,Randy(1997).Syntax:Structure,MeaningandFunction.Cambridge:CambridgeUniversityPress.

Varjú,Dezsö,andReichardt,Werner(1967).“ÜbertragungseigenschaftenimAuswertesystemfürdasBewegungssehenII.”ZeitschriftfürNaturforschung22b:1343–51.

Verhagen,Arie(2002).“FromPartstoWholesandBackAgain.”CognitiveLinguistics13:403–39.

———.(2005).ConstructionsofIntersubjectivity:Discourse,SyntaxandCognition.Oxford:OxfordUniversityPress.

Vihman,Marilyn,andCroft,William(2007).“PhonologicalDevelopment:Towarda‘Radical’TemplaticPhonology.”Linguistics45:683–725.

Voegtlin,Thomas,andDominey,PeterF.(2005).“LinearRecursiveDistributedRepresentations.”NeuralNetworks18/7:878–95.

Vogt,Paul,andLieven,ElenaV.M.(2010).“VerifyingTheoriesofLanguageAcquisitionUsingComputerModelsofLanguageEvolution.”AdaptiveBehavior18:21–35.

vonMengden,Ferdinand(2008).“ReconstructingComplexStructures:ATypologicalPerspective,”inGisellaFerraresiandMariaGoldbach(eds.),PrinciplesofSyntacticReconstruction.Amsterdam:JohnBenjamins,97–119.

WardlowLane,Liane,andFerreira,VictorS.(2010).“AbstractSyntaxinSentenceProduction:EvidencefromStem-exchangeErrors.”JournalofMemoryandLanguage62:151–65.

Page 453: The Oxford Handbook of Construction Grammar

References

Page 85 of 88

Watkins,Calvert(1964).“PreliminariestotheReconstructionofIndo-EuropeanSentenceStructure,”inHoraceG.Lunt(ed.),ProceedingsoftheIXInternationalCongressofLinguists.TheHague:Mouton,1035–45.

Webelhuth,Gert,andHoller,Anke(2010).“RelativeClauseConstructions.”Papergivenatthe6thInternationalConferenceofConstructionGrammarICCG6,CharlesUniversity,Prague,September3–5,2010.

Wechsler,Stephen(1997).“ResultativePredicatesandControl,”inRalphBlightandMichelleMoosally(eds.),TexasLinguisticForum38:TheSyntaxandSemanticsofPredication.Austin,Texas:UniversityofTexasDepartmentofLinguistics,307–21.

Weiss,Sabine,Mueller,HorstM.,Schack,Baerbel,King,JonathanW.,Kutas,Martha,andRappelsberger,Peter(2005).“IncreasedNeuronalCommunicationAccompanyingSentenceComprehension.”InternationalJournalofPsychophysiology57/2:129–41.

Welke,Klaus(1988).EinführungindieValenz-undKasustheorie.Leipzig:BibliographischesInstitutTübingen:Francke.

———.(2009).“Konstruktionsvererbung,ValenzvererbungunddieReichweitevonKonstruktionen.”ZeitschriftfürGermanistischeLinguistik37:514–43.

Wellens,Pieter,andDeBeule,Joachim(2010).“PrimingthroughConstructionalDependencies:ACaseStudyinFluidConstructionGrammar,”inAndrewSmith,MariekeSchouwstra,BartdeBoer,andKennySmith(eds.),TheEvolutionofLanguage(EVOLANG8).NewYork:WorldScientific,344–51.

Wennekers,Thomas,Garagnani,Max,andPulvermüller,Friedemann(2006).“LanguageModelsBasedonHebbianCellAssemblies.”JournalofPhysiologyParis100:16–30.

Werlich,Egon(1976).ATextGrammarofEnglish.Heidelberg:QuelleandMeyer.

Wernicke,Carl(1874).DerAphasischeSymptomencomplex:EinePsychologischeStudieaufAnatomischerBasis.Breslau:KohnundWeigert.

Wide,Camilla(2009).“InteractionalConstructionGrammar:ContextualFeaturesofDeterminationinDialectalSwedish,”inAlexanderBergsandGabrieleDiewald(eds.),ContextsandConstructions.Amsterdam:JohnBenjamins,111–41.

Wiechmann,Daniel(2008a).“OntheComputationofCollostructionStrength.”CorpusLinguisticsandLinguisticTheory4/2:253–90.

———.(2008b).“InitialParsingDecisionsandLexicalBias:CorpusEvidencefromLocalNP/S-Ambiguities.”CognitiveLinguistics19/3:447–63.

Wiemer,Björn,andBisang,Walter(eds.)(2004).WhatMakesGrammaticalization?ALookfromitsFringesanditsComponents.BerlinandNewYork:MoutondeGruyter.

Wierzbicka,Anna(1987).“BoyswillBeBoys.”Language63/1:95–114.

Page 454: The Oxford Handbook of Construction Grammar

References

Page 86 of 88

———.(1988).TheSemanticsofGrammar.Amsterdam:JohnBenjamins.

Wilcock,Graham(1999).“LexicalizationofContext,”inGertWebelhuth,Jean-PierreKönig,andAndreasKathol(eds.),LexicalandConstructionalAspectsofLinguisticExplanation.Stanford,CA:BerkeleyLinguisticSociety,373–87.

Williams,Edwin(1994).“RemarksonLexicalKnowledge.”Lingua92:7–34.

Wilson,DamiánVergara(2009).“From‘Remaining’to‘Becoming’inSpanish:TheRoleofPrefabsintheDevelopmentoftheConstructionquedar(se)+ADJECTIVE,”inRobertaCorrigan,EdithMoravcsik,HamidOuali,andKathleenWheatley(eds.),FormulaicLanguage.Vol.1.TypologicalStudiesinlanguage82.Amsterdam:JohnBenjamins,273–95.

Winter,Bodo,andBergen,Benjamin(2012).“LanguageComprehendersRepresentObjectDistancebothVisuallyandAuditorily:EvidencefortheImmersedExperiencerView.”LanguageandCognition4/1:1–16.

Winter,Werner(1984).“ReconstructionalComparativeLinguisticsandtheReconstructionoftheSyntaxofUndocumentedStagesintheDevelopmentofLanguagesandLanguageFamilies,”inJacekFisiak(ed.),HistoricalSyntax.TheHague:Mouton,613–25.

Wittgenstein,Ludwig(1953).PhilosophicalInvestigations.Oxford:BlackwellPublishers.

Wong-Fillmore,Lily(1976).TheSecondTimeAround.Stanford,CA:StanfordUniversity.

Wray,Alison(2002a).FormulaicLanguageandtheLexicon.Cambridge:CambridgeUniversityPress.

———.(2002b).TheTransitiontoLanguage.NewYork:OxfordUniversityPress.

Wulff,Stefanie(2006).“Go-Vvs.go-and-VinEnglish:ACaseofConstructionalSynonymy?”inStefanTh.GriesandAnatolStefanowitsch(eds.),CorporainCognitiveLinguistics:Corpus-basedApproachestoSyntaxandLexis.BerlinandNewYork:MoutondeGruyter,101–25.

———.(2008a).“V-and-VundV-VimEnglischen:eineKonstruktionsgrammatischeAnalyse.[V-and-VandV-VinEnglish:AConstructionistApproach.],”inKerstinFischerandAnatolStefanowitsch(eds.),KonstruktionsgrammatikII:VonderKonstruktionzurGrammatik.Tübingen:Stauffenburg,189–201.

———.(2008b).RethinkingIdiomaticity:AUsage-basedApproach.London/NewYork:ContinuumPress.

———.(2009).“ConvergingEvidencefromCorpusandExperimentalDatatoCaptureIdiomaticity.”CorpusLinguisticsandLinguisticTheory5:131–59.

Wulff,Stefanie,Ellis,NickC.,Römer,Ute,Bardovi-Harlig,Kathleen,andLeBlanc,Chelsea(2009).“TheAcquisitionofTense-Aspect:ConvergingEvidencefromCorpora,CognitionandLearnerConstructions.”ModernLanguageJournal93:354–69.

Page 455: The Oxford Handbook of Construction Grammar

References

Page 87 of 88

Wulff,Stefanie,andRömer,Ute(2009).“BecomingaProficientAcademicWriter:ShiftingLexicalPreferencesintheUseoftheProgressive.”Corpora4:115–33.

Wulff,Stefanie,Stefanowitsch,Anatol,andGries,StefanTh.(2007).“BrutalBritsandPersuasiveAmericans:Variety-specificMeaningConstructionintheinto-causative,”inGünterRadden,Klaus-MichaelKöpcke,ThomasBerg,andPeterSiemund(eds.),AspectsofMeaningConstructioninLexiconandGrammar.Amsterdam:JohnBenjamins,265–81.

Yaeger-Dror,Malcah(1993).“LinguisticAnalysisofDialect‘Correction’anditsInteractionwithCognitiveSalience.”LanguageVariationandChange5:189–224.

Yamashita,Hiroko,andChang,Franklin(2001).“‘LongbeforeShort’PreferenceintheProductionofaHead-finalLanguage.”Cognition81/2:B45–B55.

———.(2006).“SentenceProductioninJapanese,”inMineharuNakayama,ReikoMazuka,andYasuhiroShirai(eds.),HandbookofEastAsianPsycholinguistics.Vol.2:Japanese.Cambridge:CambridgeUniversityPress,291–97.

YCOE(TheYork-Toronto-HelsinkiParsedCorpusofOldEnglishProse).http://www.ling.upenn.edu/mideng/ppcme2dir/YCOE/doc/annotation/YcoeLite.htm.

YORK(SaliTagliamonte'sCorpusofYorkEnglish).http://www-users.york.ac.uk/~st17/.

YPC(TheYork-HelsinkiParsedCorpusofOldEnglishPoetry).http://www-users.york.ac.uk/~lang18/pcorpus.html.

Zeschel,Arne(2008a).“FunktionsverbgefügealsIdiomverbände,”inAnatolStefanowitschandKerstinFischer(eds.),KonstruktionsgrammatikII:VonderKonstruktionzurGrammatik.Tübingen:StauffenburgVerlag,263–78.

———.(2008b).“LexicalChunkingEffectsinSyntacticProcessing.”CognitiveLinguistics19/3:419–38.

Ziem,Alexander(2008).FramesundsprachlichesWissen.BerlinandNewYork:MoutondeGruyter.

Zipf,GeorgeKingsley(1935).ThePsycho-biologyofLanguage.Boston:HoughtonMifflin.

Zwaan,RolfA.,Stanfield,RobertA.,andYaxley,RichardH.(2002).“LanguageComprehendersRoutinelyRepresenttheShapesofObjects?”PsychologicalScience13:168–71.

Zwicky,Arnold(1971).“InaMannerofSpeaking.”LinguisticInquiry2:223–33.

———.(1974).“‘HeyWhatsyourname!’”.ChicagoLinguisticSociety10:787–801.

———.(1994).“DealingoutMeaning:FundamentalsofSyntacticConstructions.”BerkeleyLinguisticsSociety20:611–25.

Page 456: The Oxford Handbook of Construction Grammar

General index

Page 1 of 11

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

GeneralindexTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

Generalindex

adjunct77,84,87,119,210,226,228,238,242,321,499–502affix17,43,75–76,78,83,85,88,90,92,162,193,259–60,262,264–65,355,403,416affixsubstitution264affixoids259–60agreement19,25,29,31,40,79,114–16,124,129,163–5,193,219–20,225,227,242,266,307–8,310–2,314,327,354,380,403–4,408,430alienability232,505–6analogy,analogical33,46,433–4argumentstructure8,11–12,18–21,47,76–77,90,98,100,104,137,140,150–1,179–81,188,217–8,228–30,235–9,241,244,247–49,251,262,290–2,302,312,327,351,362,367,387,391,393,427,439,443–6,448,450,452,454–56,461,468–69aspect23,81,105,123,181,228,230,261,286,300–2,326,389,425,427,463,468as-predicative103,302assembliesofneurons398,415association4,7,19,39,49–50,55,63–5,67–8,94,99,103,105,107,150,156,165,185,192,194,282,288–9,291–2,294,296,298,303–4,306,322,371,373,422,430,480,483associativelearningfromusage368,405,410BerkeleyConstructionGrammar5,94,96,111–3,115,117,119,121,123,125,127,129,131,133,242,248,306,308,312,339bipolarstructure7,174,192boundmorpheme256,258,268–9,350categorization4,16,49,50–2,64–5,68,100,103,119,173,192,194–95,201,206,238,242,244,364,366–7,369,374,380,388–9,423,450,488,494coercion82–3,147–8,164,167,237,327cognition17,26,51,378coining/coinage4,28,33–5,38–41,43–4,46,48,84,87,131,264,428collexeme9,100,288,291–3,321,373,472–3,483co-varying9,100,288

Page 457: The Oxford Handbook of Construction Grammar

General index

Page 2 of 11

distinctive9,100,291,295–303,472–3,483simple9,291–2,294,296–9collocation(s)18,26,38–9,65,99,101,103,128,225,281,287–8,290–1,295,298,420,428,451,454,461,464,466,472collocationalchange461,464collostruction(al)9,62,99–100,103–5,288,290–6,298–306,316,373,477,483–4,489complement2,6,10,22,24,29,57,60,62,75–76,82,100,103–4,113,115,118,123–5,127–9,131–2,137,139,142,146–7,202–3,206–7,210,214,217,224,310,312–3,319,324,334–5,356,361–3,371,403,411,425–7,430,432,434,463,469,489completeinheritance308,314,325,327complexsystems163,378compositional,compositionality8–9,101,276,280–3,285–6,423,436,439computation98,170,289,366,385–6,393,401–2,406computationalmodelingxiiicomputationalimplementation7,169configuralfrequencyanalysis301connectionistmodel106constituency6,76,80,113,186,203,217,224,487construal16,30,51,80,117,185,199,207,217,237,241,248,375,495–7,508construct5–6,64,70,139,142,144–6,148–9,186–7,257,285,312,320,339–40,411,427constructicon1,3,8,186,193,253,256–73,275,277,279,281,283,285,289,291,295,297,299,301,303,305,310,312,314,318,320,322,324,326,328,330,332,334,336,338,340,342,344,278,280,286–8,293,307–8,315–6,321,327,412,439,452–3,456constructionfrequency11,87,234,368constructionlearning266,375,502constructionalanalyzer188constructionalmap428,433–4constructionalpolysemy246,251constructionalrole231constructionalschema205–6,261–3,270,465constructionalization421,424,428–9,433context4,7,11–12,27,39,49–50,52–6,61,63,65–8,83,85,87,92,95–6,98–9,101,105,115,117–8,125,128,132,139,148,166,169–70,172–3,175,177–8,183–8,190,192,194,209,223,230,234,239,247–8,260,266,269,271,275,277,280–2,310,319–21,324,330–1,340–1,363,370,374–6,384,402,405,413,419–20,422–9,432–3,435–6,447,460–1,466,476–9,483–8,496,499,501,507contingencyofform-functionmapping368,373control102,123,161,172,174–5,178,386,468conventionalunit193–4,209,218conventionality193–4conversationalimplicature37,56conversion269–70,390corpuslinguistics11,99,107–8,366,459corpus/corpora95,97–8,100,104,106,294,304,366,377,458–9,461–2,466,468–9,474–5,481,484,490

Page 458: The Oxford Handbook of Construction Grammar

General index

Page 3 of 11

BritishNationalCorpus(BNC)35,47,60,279,462,475,481,490CorpusofContemporaryAmericanEnglish(COCA)60,274,277IDScorpora467InternationalCorpusofEnglish(ICE-GB)292,315,483Lancaster-Oslo-BergenCorpus(LOB)xixPPCEME(ThePennParsedCorpusofEarlyModernEnglish)475,461PPCME2(ThePennParsedCorpusofMiddleEnglish)475SPRÅKBANKEN(variousSwedishcorpora)475TIME(TheTimeMagazine–fromthe1920stothe2000s)57,60,64,69YCOE(TheYork-Toronto-HelsinkiParsedCorpusofOldEnglishProse)475YORK(SaliTagliamonte'sCorpusofYorkEnglish)475YPC(TheYork-HelsinkiParsedCorpusofOldEnglishPoetry)475correspondenceprinciple237–8co-text433creativity28,35,396crosslinguisticgeneralization15,23,30,91,213,220cue98–9,181,304–5,315,368,371,372–3,487availability304–5reliability305validity98–9,304–5culture57,299,378,484,507dative,prepositional103,108,181,294,296–7,321,367,382,385–7,389,392–3,480–1declarativeclause308–10,313,317defaultinheritance8,18,21–3,84,138,243,257,328,479definitearticlereduction(DAR)503,505,507derivation3,4,6,43,71,73,77,133,139,144,162,240,269,428denominalverbs43,85,89deverbalnouns85determiner-nounagreement408dialect3,12,34,45,97,302–3,322,391,467–8,476–83,488–90,493,502–3,505–07discourse3–4,8,12,23,25,29,51,101–2,166,184–5,208,234,239,289,314,320,329–32,335,338,340–1,343–44,365,375,377–8,433,435,448,476–81,483–9,491,495,498,501discretecombinatorialneuronalassemblies(DCNAs)403–6,410–4,416distributionalanalysis7,213–4,216–17,219,224,230–1,368do-support75,79,313–4elaboration195,201–2,468–9embodiment374,378,400,410,416emergence,emergent170,415emergentgrammar194entrenchment4,21,51,55–6,193,247,258,278,304,306,314,319,328,348,360,368–9,377,459,479,483ergativelanguage25,220exemplartheory64,68expletiveinfixation84featurestructures6,43,112,118,132,135–6,138–40,142–4,150–2,154,188,311

Page 459: The Oxford Handbook of Construction Grammar

General index

Page 4 of 11

Fisher-Yatestest291,296,325focus9,71,91,127,139,181,196–7,199,208,229,240,244,278,285,290,303,330,335,337–42,363,376,424–5,427,432,452,471,477,482,492,495form1–2,5,7–8,10–11,15,17–21,23,25,28,30–1,37,39–40,43,47,49–52,54–7,59–61,63–8,72,74–5,77–85,87,89–90,92,96–7,99,112–13,124,126,128,131,133–5,138–40,148–9,165–70,162,164,167,171,173–9,182–3,185–7,190,192–3,202,212–4,217–9,224–7,230,234–6,239–43,245–6,249,255–9,263,265–7,269–70,272–3,303,307–9,311,313–15,318,326,332,334,337,339,349,354,360,365–71,373–4,376–37,379,382,385,387,389–90,392–3,404,410,412,414–15,422,424–31,433–7,439–41,445,462,454–5,457,459–61,463–4,466–8,471,474,477,484–5,488–9,495,503,505–6formalism6,29,74–5,78,136,138,153,159,170–1,174–6,178,188–9,211–12,249–50,308,452,487,492formalization7,29,133,147,167,190,248–50,479frame6,45,131,140,147,149,196,206,227,235,237,242,250,298,305,335,343,353–5,370,388,390,412,420,430,445,448,450,452,486FrameSemantics6,232,235,250,452FrameNet6,12,249–50,486frequency/frequencies60,102,105,248,279,281,291,295–6,298,304–5,369,459,461,463–4,466–7fronting130,524fullentrytheory257function10–11,15–19,23–5,27–8,30,43,49,55,64,71,73–6,78,80,82,92,113,119–20,126,136–7,141,151,165,169–70,173,182,192,221,224,229–31,234,237,239,241,243,248,250,255–6,261,264,268–9,280,283,309,311,316,320,330,332–3,341–2,344,348–9,365–6,368–9,372–3,376–7,380–3,385–6,389–90,392,395–6,399,420,423,425,427–29,431,433–7,441,449,455–61,464,466,473–4,482,486,500fusion236–8,241,411,467future10,22,63,65–7,100,105,107,189,208,214,246,284,288,296,302–3,316,327,366,377–8,420,424–7,442,463,467,473,502goingto63,296,302,424–27,463,467shall100,303,473will66,296,302gardenpatheffects319generalization4,6–7,15–20,23–5,30,33,54,66,82,89,91,111,131–2,134,144,174,182,206,213,218–21,223–4,230,238–9,243–4,247–8,256,258,260,267,273,312,315,318,343,392–4,405,419,422–3,433,474,486,505gestalt175–6,182,199,217,226,228–9,494gradience423,489gradient50,328,413,419,436gradual11–12,46,50,54–5,57,61,68,163,171,213,348,353,356,358,361–4,398,407,410,416,419–21,424–6,428,432,436–7,469,474,482,489gradualness419,425,428,436grammaticalclasses7,193,197–8,201,385grammaticalrole7,206,212,214,363,380,386subject2,15,25–7,60–2,66–7,76–7,79,113–15,119–24,126–7,129,132,134,136–7,139,

Page 460: The Oxford Handbook of Construction Grammar

General index

Page 5 of 11

142–4,149,151–2,180–1,207,209–10,215–16,220,224,227,231,240,244,251,255,309,311–12,318,323,333,338–41,349,353,356,361,363,375,381–3,386,394–6,403,429–34,439,445,450,454,456,459–60,465–6,500–1object2,8,10,12–13,15,19–20,57,60,62,76–7,79–81,92,113–14,116,120–2,126–8,137,140–1,151,165,180–1,195,197,207–10,212,214–20,223–4,226,231,235,239,255,257,261–3,272,274,292,302,319,331,333,338,356,360,363–4,369,374–5,381–2,395–6,401,403,411,415,463,470–1,480,498directobject22–3,25,77,79,80–1,92,120,207,214–9,224,261,381–2,498indirectobject121,207,210,231,480grammaticaltheory34,290–91,300,303,306,326CognitiveGrammar7–8,71–2,75,92,191–210,248,306,314,491GeneralizedPhrase-StructureGrammar(GPSG)135Head-drivenPhraseStructureGrammar(HPSG)6,92,133,135,212LexicalFunctionalGrammar(LFG)71,212Mainstreamgenerativegrammar(MGG)318RoleandReferenceGrammar(RRG)212UnificationGrammar135,154WordGrammar224,250,477–9,492grammaticalization11,55–6,63–6,99,260,283,379,419–27,434–6,460–3,469,489,497,504headfeatures118,128head-directionparameter24hierarchicallexicon256–7,260,266,273host-classexpansion446,469,471hypoanalysis269iconicity242,505–6identity19,35,128–9,175,183–4,196,200,214,327,466,488,502,507–9idioms/idiomaticity100–1,279,285–7,289decodingidioms176encodingidioms276formalidioms277substantiveidioms308imperativeclause309–10indexation220,225–6,228inflection135,228,244,265,267,369,389,428informationstructure9,16,29,71–2,98,239,329–35,337,339–43,380–1,486inheritance6,8,15,18,21–3,43,70,79–82,84–5,87,90,134,136–8,144,146,176,180,219,239,243–6,251,257,308,312,314,325,327–8,442,456,479hierarchy18,21–2,43,79–81,84–5,87,136,243link21,246type-based134instancerelation144subpartrelation144inputfrequency314–15,320,368–9instantiation6,75,92,125,127,195,389

Page 461: The Oxford Handbook of Construction Grammar

General index

Page 6 of 11

interface5,7,16,71–3,90,150,171,173,175,182,249,290,333,492,509interfacerules5,71–73,90interrogativeclause128,309–10intersubjectivity494introspection5,100islandconstraints25,318item-specificinstances248Lancashire12,97,480–2,493,502–3,505–7language3–4,6–8,10–12,16,18,21,23–7,30–4,40,45,47,49–56,60–5,67–8,70–1,73–4,76,79,88–91,93–4,97,100,103,105–7,111–13,117,126,134,138–9,142–3,153–5,159–60,163,165–73,175–6,178,181–2,184–95,208–9,212,216,218,222–4,231,233–4,240,242,244,248,250,256–7,266,270–1,274,277–8,286,288,290–2,294,302–4,306,309,312,314–5,320–1,326,332–3,339,347–9,350–7,359–61,363–81,383–402,404,406–11,413–17,419–22,433–4,436,438–9,441–7,449–67,469,471,473–85,488–9,491–2,504,506–76languageacquisition3,30,50,56,97,106–7,153,188,223,233,291–2,303,306,309,347–9,351,353,355,357,359,361,363–73,375,377–9,392–3,396,489languagechange11,16,50,65,68,233,291–2,303,315,349,378,419–20,436,438,442,454,458,460,466,474,497,504languagevariation11,306,417,420–1,424,426,428,430,432,434,436,439,441,443,445,447,449,451,453,455,457,459,461,463,465,467,469,471,473,475,477,479,481,483,485,487,489,41–92,494,496,498–500,502,504,506,508LanguagesAvestan451,453–54,456Batsbi266English9–10,12,17–8,21–3,25,29,31–2,35,37–43,45–8,56–8,60–3,65–7,69,74–5,79–82,84–5,88–9,97–8,100,102–4,112–13,115,121,127,132,134,137,143,151,163,181,207,213–17,219,224,227–8,238,240,245,251,255–6,258–61,264,267–8,270–4,284,289,292,295,300,302–3,308,310–11,313–27,332,339,341–2,351–2,367,369–71,377,383,386,388,424–5,433,440–1,445,450–1,456,458–9,462–72,475,478–84,489–90,493,495,498–99,503,506–8Chinese25,259–62Dutch12,97,259–60,268–73,288,302–3,317,480,482–3,493,499–02,508German1,97,103–4,155,163–4,238,270–1,289,295,302,322,347,371,408,429,445,450–1,456,467–8Finnish97,238,340,344,352,442,479French27,66,97,118,238–9,322,331–2,335,344,353,454,464,467Gothic445,451,457Hungarian1,271Indonesian261,443,450Italian207,263,267Kabyle341,344Latin265–66,440–1,443–5,451,453–4,456,467Latvian451Japanese12,79,338,262–3,322,386,438,444–5

Page 462: The Oxford Handbook of Construction Grammar

General index

Page 7 of 11

Persian22,24Polish263Russian97,265,275,302,317,438,445,457OldChurchSlavic440,451OldEnglishxx,440,445,450–51,456,470,475OldIrish440,451OldSaxon451OldHighGerman445,451,467OldNorse-Icelandic443–5,450–1,456–7Saami266learnedattention366lexeme26,33,113–14,117–20,122,125,131,135–36,139–41,148–52,191,260,263–64,266,291–92,301,336,382,384,389,437,441lexical4–6,8–11,15,18,26,28,30,33,43,45–7,58–9,62–4,67,71,73–9,81,97–8,106,113–14,118–9,123–4,126,128–31,136–9,141–2,150–1,154,156–8,164–7,177–9,182–3,185,188–9,193–6,204,212,217,224,234–8,241,246,249,251,256–8,260,262–4,266–7,270–1,276,278,283–4,287,290–2,296,298,301,306,308,311,314,319,349–50,352,354–6,358,367–71,377,380,382–4,387–93,395,398–99,401–5,408,410–16,428,437–43,445–9,451–2,455–6,459–60,463,466,468–9,471–2,492,507lexicalconstruction11,157–8,167,183,291,296,301,466lexicalentry75,142,241,257,387,452lexicalitem33,58,62,74–6,78–9,81,97,123,130,194,196,237,298,319,369,402–3,441–2,449,451lexicalsemanticproperties249link21–2,76,125,145,156,159,183,225–6,234,245–6,251,328,333,360,364,392,394,403–4,407–8,411–13,415inheritance21,246instance245–46polysemy246subpart145,245linking/valencecompletion119,122,132locality6,136,138,146locativealternation20,241maximalityfeature118,128,131,147mentalsimulation7,105metaphor39,51,188–189,242,280,293,461,464,470metaphor(ical)19,22,39,77,207–8,246,251,283,293–4,297,302,470–1metaphoricalextension246,251metaphysicalcommitments211–12metonymy242,461,470mini-construction238–39mismatch150,216–7,231,324,354,406,408–9,412–14,416,423,436,496modality84,222,290,302,381–2,384–5,392,394,463–4modifier77,82,202–3,210,259,432,460,482–3morphology8,31,44,48,74–5,84–5,88–9,138,149,156,193–4,247,255–7,259,261,263–

Page 463: The Oxford Handbook of Construction Grammar

General index

Page 8 of 11

9,271–3,315,339,380,429,443,448,450,460,477,486verb-to-nounderivation43morphophonemicchange11,461,466morphosyntacticchange461motivation20,55,242–3,249,251,281,377,381,384,390,428,434–5multivariateanalysis464,499,505multiword-expressions281network3–4,8,10,15,18,21,27,106,132,138,161,194,219–22,224,239,244,310,313–15,320,322–3,327,349–51,357,363–4,370,302,305,415,421,427,433,435,459–60,471,478,482multiple-inheritance138neurobiologicalembodiment410neuronalcircuitry400objectraising80paradigmaticrelationship256participantrole236–8particle12,81,271–3,283,289,292,321,389,403,493,498–99,506placement12,321,389,493,489–99,506verb271–2,389pasttense66,75,88,183,245,265,302,344,369,503,505patternofcoining34–5,37,40–2,46,48,131Performance-GrammarCorrespondenceHypothesis320,328phonologicalinformation151,240,380phrasestructurerules76,78,83phraseology367–8pied-piping9,86,308,319–21,323–4pied-pipedpreposition86possessive26,222,270,302,322postureverb(s)302,468pragmatics16,24,52,55,64,131,240,334,378,477,485–6,494prefabricatedphrases(‘prefabs’)18,26,65prepositionstranding308,319–21,323PrinciplePrincipleofMaximizedEconomy243PrincipleofMaximizedExpressivePower243PrincipleofMaximizedMotivation243processingAvoidCompetingSubcategorizorprinciple319constraints328,387ValencyCompletenessprinciple319–20productivity8,11,28,42–47,62–3,83–5,89–90,92,235,242,247–8,258,284,294,315,356–8,369,425,461,464–5,471profile60,197–202,210,284–5,370propositionalstructure71prototypes195,197,245,299,372,422,436,494

Page 464: The Oxford Handbook of Construction Grammar

General index

Page 9 of 11

prototypicalityofmeaning368,372pseudo-coordination468psycholinguistic(s)3,7,9–10,96,100,108,186,188,208,248,280,283,289,303–6,309,350,368,373,379,381,383,385–8,391,393–6,399pumping6,116,122,131–132seeunarybranchingconstruction138,147–8,152quantifier36,92,149,261,352RadicalConstructionGrammar7,93,197,203,209,211–19,221,223–7,229–31,248,306,308,311,314,422,425recency54,368,370reduction56–7,65,99,170,240,402,467,503,505–8redundancy248,368,372reduplication34,91,260–1,273referencetransfer5,82–3referent18,83,126,162,183,207,226,277,283,331,335–9,343,363,382,386,401,498relativeclause57,227,229,231,286,308–10,321–3,344,363,369relativefrequency89,459,461,463,469,506representationalcommitments212,215,224,230rules(verbandnoun)2,5,46,48,52,71–4,76,78,83,87–8,90,112–13,120,122,125,132,134,138,144,151,157,193–4,203,217–8,226,234,238,246,264,275,348,352,375,399–400,402,405–8,413–16,441salience368,371–3,375–76,384,471,504–5scalarreading36scanning198–9,208schema,schematic2–3,10,53–4,57–9,61–2,68,174,178,180,193–5,197–8,200–01,207–9,223,244–5,247,249–50,256,265,277–8,286–8,299,305,307–8,310,312,315,321,325–8,350,356–8,360–1,364,372,377,381,405,408,412,414–5,420,425,428–31,436–7,449–50,455,459,471,482schematicity62,68,193–4,223,249,256,307–8,327,425,449,471,482schematization7,192,194,206,278,287–8,425–6,431secondlanguageacquisition3,233,291,303,309,365–7,369–71,373,375,377,489semantic4–9,16,19,23–5,29,32,35–6,52,54–9,62,64–8,71–3,75–83,85,87,93,98,101–2,116,118–22,126,128–32,136–8,140–1,143–4,150–1,155–60,164–5,171,174,183,185,188–9,192–4,196,198–200,206–7,210,213,218,220–8,230–1,234,236–7,239–43,247–9,251,256,258–59,261–2,264,266,268–9,273,275–6,278–86,288,293–4,296,298–9,303,309,311,313–16,332–3,338,342–3,351,356,361,363,365,367,370,374,376–7,380,382–5,387–8,392,398–9,401,406,408,411,415–16,420,422,424–7,429–30,432–3,436–9,441–2,452,459,470–3,480,486,493,505–6semanticcoherenceprinciple237semanticextension206,470semanticrole70,119–20,132,225,236,338,480semiologicalfunction192sequencedetectors402,404–5,412,414,416sign1,5,74–6,79,83,92,134–6,138–40,142,145–6,150–1,349,422,487SignPrinciple142,151

Page 465: The Oxford Handbook of Construction Grammar

General index

Page 10 of 11

Sign-BasedConstructionGrammar6,43,45,94,116,133,135,137,139,141,143,145,147,149,151,154,248,306,308,311,318,387,456significancetest291,368sluice-stranding44,86,91sociolinguistics467,479,491–3,495,497,499,501–3,505,507–9specificity10,16,18,27,65,161,139,405,412,500–2speechacts239,351,363statisticallearning366,368statisticalpreemption28statistics,statistical9,108,304–5subject-auxiliaryinversion(SAI)243subjectivity572subject-verbagreement124,307–8,310–12subregularities17,244surfacestructure15,20,389symboliccomplexity193–4synonymy159,243,575taxonomichierarchy219,244–5taxonomicorganization150tense26,66,75,79,88,123,143,183,229,245,265,267,275,284,286,290,302,310,326,344,369,371–3,389,425,427,429–30,432,434,466,503,505textfrequency11,461trajector/landmarkalignment197,202type4,28,30,39,44–45,48,52,59–64,67,68,75,77,88,92,96,99,106,113,117,119,121,124–5,128,134–6,138–42,144–6,148–9,152,162,164,174,178–80,183,195,207,223,226–7,229,235,241–2,246–9,256–63,266,269,287,297,311,314–6,319–21,323–4,331–2,338,343,353,355,369–70,375,405,407,413,416,449–50,453,459,461,464,469,471,478,480,482,485–8,501–2,507typeandtokenfrequency4,63,319,369typeconstraints6,135,138,183typefrequency28,52,59–64,68,92,223,247–8,258,311,315,319–20,369,459,464underspecification152,424,436unarybranchingconstruction138,147–8,152unification6,76,112–15,129,135,137,148,154,159,188,241–2,339UniversalGrammar3,348,389universals23,213,222,231usage-based4–5,7–10,12,16,21,26–7,30,45,47,49–51,53,55,57,59,61,63,65,67–9,105,153,189,195,209,223,247–8,278–9,284–5,288,290,306,308,311,314–15,319–22,325,327,347–51,355,365–6,369,391,398,405,410,412,476,478–9,484,487,506valence6,41,47,112,115,118–128,130–2,137,141,145,149–50,242,311,411,429verbatimmemory28voice103,105,119,121,132,137,141,212–15,217–18,222,227–8,285–6,302,331,359,429–30,432,434,462WeightedR9,281,283,288wh-word29,35,86,124,320

Page 466: The Oxford Handbook of Construction Grammar

General index

Word-and-Paradigmmodel265Zipfiandistribution368–9

Page 467: The Oxford Handbook of Construction Grammar

Index of Constructions

Page 1 of 4

PrintPublicationDate: Feb2013 Subject: LinguisticsOnlinePublicationDate: Dec2013

IndexofConstructionsTheOxfordHandbookofConstructionGrammarEditedbyThomasHoffmannandGraemeTrousdale

OxfordHandbooksOnline

IndexofConstructions

(Mono)transitiveconstruction483[DATiswoe]construction(Proto-IndoEuropean)452–6AasNPpattern4,38–42,44,46–7Abstractconstruction3,8,10–11,28,36,45,76,78,235,307,310–12,315,317,328,368,412,414AbstractClausalconstruction9,309–11,326DeclarativeClauseconstruction310ImperativeClauseconstruction310InterrogativeClauseconstruction144,310RelativeClauseconstruction231,310,320–4,565,574All-cleftconstruction4,35–7AntitopicExclamativeconstruction151Applicativeconstruction141ArgumentStructureconstruction8,11,18,20–1,47,90,98,100,104,137,140,179,180–1,188,217–8,229–30,235–9,241,247–9,251,291–2,312,327,351,367,391,439,443–6,448,450,454–6AuxiliaryInitialconstruction146Begoingto/gonnafuture(English)296,302,426CausedMotionconstruction20,22,40–2,98–9,104,149–52,167,195,204,206–8,236–7,241,245–6,250,351,356–7,411ComplexPredicate(CP)construction22ComparativeCorrelativeconstruction9,308,318,325,355,543Conditionalsentences129Conditional-andConstruction527Definite-sConstruction(Dutch)268DeprofiledObjectconstruction8,239Ditransitiveconstruction17–8,20,28,30,92,98–9,108,137,181,193,195,204,207,237–8,240–1,246,251,292–5,299–300,302,307–8,351,358,369,389,459,470,480–4

Page 468: The Oxford Handbook of Construction Grammar

Index of Constructions

Page 2 of 4

DirectedMotionconstruction179,180–3DoubleObjectconstruction10,181seeDitransitveConstructionEllipsisconstruction91ExtraposedExclamativeconstruction144Filler-Gapconstruction9,308,317–9,322,325–6,328Futureconstruction100,302–3WillFutureconstruction296Going-toFutureconstruction296Head-Complementconstruction6,124,146–7,312–13Idiomconstruction2,8IndefiniteDeterminerconstruction28,148Intransitiveconstruction96,219,311,351,356–8,360,484Intransitivemotionconstruction245,351InterrogativeExclamativeconstruction146Inversionconstruction,aux-first6,28,124,180,240,243InvertedExclamativeconstruction145InvertedNegativeAdverbPreposingconstruction146–7ISISconstruction142–3It-cleftconstruction(English)469,560Letaloneconstruction239,275–6LightVerbconstruction80Long-distanceDependencyconstruction21,25Long-participle(Slavic)429Morphologicalconstructions8,11,84,161,164,255–6,258,260,263,273,311,429,433English[V-able]464English[V-ment]465Englishbe-construction470–1compoundANcompound270exocentriccompound263NNcompound82,258–9NVcompound259resultativeV V compounds261suru-compounds273MorphologicalVerbconstruction219NominalExtrapositionconstruction8,239,339–41Nounconstruction176NofNPconstruction18,28,82,240,503,507NCI-construction465NounPhraseconstruction/construction183Noun-Preposition-Noun(NPN)pattern/construction44–5,86–7,91,277Passiveconstruction23–4,137,215–6,219,224,287,302,356,357–9get-passive(English)462Passive-Ditransitiveconstruction31

1 2

Page 469: The Oxford Handbook of Construction Grammar

Index of Constructions

Page 3 of 4

Phrasalconstruction15,26,70,85,91,118,120,127–9,137,150,164,178,256,287Pied-Pipingconstruction320Pied-Pipedwh-RelativeClauseconstruction320–1Possessiveconstruction302,322PrenominalModificationconstruction431,433PrepositionNounconstruction/PNconstruction17–8,21PrepositionalDativeconstruction297,386,480–1PrepositionalPhraseconstruction/PNPconstruction18,21Presentativeconstructions508Preteritconstruction148–9Pseudo-coordinationconstruction(Swedish)468,542Questionconstruction28–9,240,320Non-subjectQuestionconstruction28Rateconstructions129Referencepointconstructions493,500–1Resultativeconstructionxii,2,21–2,58,60,104,217,237,246–7,251,261,358IntransitiveResultativeconstruction21TransitiveResultativeconstruction21–2SentenceComplementconstruction29Sluicingconstruction44Sound+Motionconstruction85–6,91StrandedQuestionconstruction320Strandedwh-RelativeClauseconstruction321Subject-AuxiliaryInversionconstruction28,180,240,243Subjectlesstagconstruction134Subject-Predicateconstruction6,21,125,142,146–7,312–13substantiveVNPconstructions279Tense-Aspect-Mood(TAM)constructionsxx,228–9TheXertheYerconstruction17,235seeComparative-CorrelativeconstructionThere-construction95,96,246Time-awayconstruction81Transitiveconstruction19,67,140–1,214,219,311–12,351,356–60,483Transitivizingconstruction40TransitiveVerbconstruction79–80,219,360Unary(non-branching)constructions,pumping6,116,130,137–38,147–8,151–2VerbalPredicateconstruction219VerbPhraseconstruction/VPconstruction15,19,21,28,77,137,240,244,312VNPconstruction79,279–85,289Verb-Locativeconstruction369Verb-Object-Locativeconstruction369Verb-Particleconstructionxii,12,80–1,91,96,280,296,302,499Voiceconstructions218,227–8ActiveVoiceconstruction213–15PassiveVoiceconstruction215,217,227

Page 470: The Oxford Handbook of Construction Grammar

Index of Constructions

Wayconstruction(English)46,58,76–8,80,83,85–6,90,235,237,247,302,471What'sXdoingYconstruction(English)xx,56,63,66,128,277,446Wh-cleftconstruction35–7Wordconstructionscat73,85,114,116–8,121,123–5,139–40,142–3,146–8,156,173–84,193,200–4,311,320,430,432,434flower192,194,196,200,295,358,380–1,394–5jumped36,177,179,180,182–3