On the Nature of Sytactic Variation

7/29/2019 On the Nature of Sytactic Variation

1/20

Linguistic Society of America

On the Nature of Syntactic Variation: Evidence from Complex Predicates and Complex Word-FormationAuthor(s): William SnyderReviewed work(s):Source: Language, Vol. 77, No. 2 (Jun., 2001), pp. 324-342Published by: Linguistic Society of AmericaStable URL: http://www.jstor.org/stable/3086777 .

Accessed: 13/12/2011 07:17

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access toLanguage.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=lsahttp://www.jstor.org/stable/3086777?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/3086777?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=lsa


2/20

ON THENATURE OF SYNTACTICVARIATION:EVIDENCEFROM COMPLEXPREDICATESAND COMPLEXWORD-FORMATIONWILLIAMSNYDER

University of ConnecticutandHaskins Laboratories

The existence of substantiveparametric ariation n syntax,as characterizedn Chomsky1981,has been questioned n the more recentgenerative iterature,notablyin Borer 1984, Fukui1986,and Chomsky 1993. This articleprovides convergingevidence from child language acquisitionandcomparative yntax or the existence of a syntacticparameternthe classical sense of Chomsky1981, with simultaneous ffects on syntacticargument tructuree.g., verb-particle onstructions)andcomplexword-formationrootcompounding).The implicationsare firstthatsyntaxis indeedsubjectto pointsof substantiveparametric ariationas envisioned in Chomsky1981, and secondthat the time course of child language acquisition s a potentiallyrich source of evidence aboutthe innate constraintson languagevariation.*

1. THEORETICALACKGROUND. central question for syntactic theory is whethercrosslinguisticvariation s a 'deep' domain of inquiry-in otherwords, a domaininwhichgeneral,explanatoryprinciplesareoperative.The question s logically indepen-dent of the existence of substantiveuniversalsof human anguage.In principle,pointsof syntacticvariationcould be limited to superficial,listed idiosyncrasieswithin anotherwise invariantandrichly structuredanguage faculty.1.1. THE NATUREOF SYNTACTICVARIATION.The PRINCIPLES-AND-PARAMETERSRAME-

WORKntroduced n Chomsky 1981 permittedthe statement of highly abstractcon-straintsoncrosslinguisticvariation inthe form of parameterized rinciplesof UniversalGrammar, r UG), as well as "absolute" universals of humanlanguage (in the formof unparameterizedUG principles).Indeed,earlyresearch n the P&P framework edto the proposalof a numberof parameterized rinciples,each permitting wo or moredistinctparameter ettings with broadconsequences for the surface characteristics fthe resulting grammar.Travis 1984, for example, proposed parameterizedprinciplesof HEADGOVERNMENTnd THETA OVERNMENT,whose interaction with independentlymotivateduniversalsof government heoryaccounted or complex patternsof crosslin-guistic variation n word order.More recent researchwithin the generativetradition,however,has called into ques-tion theexistence of broad,parameterized rinciplesof thekindenvisioned n Chomsky1981. Notably, RichardKayne (1984) and Luigi Rizzi (1982) have emphasizedtheimportance f microparametersn accounting or thesyntacticvariationobservedacrossclosely relatedRomancedialects.Proposedmicroparametersretypically expressed nterms of lexically listed,morphosyntactic equirements f functionalheads. The recent* This article is a substantiallyrevised version of Snyder 1995b, ch. 2, ??1-3). The author s grateful,for numerouscomments and helpful suggestions,to the editors andanonymousreferees of Language,andto audiencesat BUCLD20, NELS 27, MIT,UMass-Amherst,McGill,CUNYGraduateCenter,SUNY StonyBrook, the Universityof Pennsylvania,and the Universityof Connecticut.This projecthas been supported

in part by an NSF Research andTrainingGrant o MIT for LanguageAcquisitionandComputation,by theMacDonnell-Pew Center for Cognitive Neuroscience at MIT, by Faculty Grantsfrom the University ofConnecticut ResearchFoundation,and by an NIH-NIDCDGrant to (PI) Diane Lillo-Martinand WilliamSnyder.

324


3/20

ON THE NATURE OF SYNTACTICVARIATION

emphasison microparameters,n combinationwith variousempiricalandconceptualproblems that were discovered in earliermacroparametric roposals,has led manygenerativists o doubt the existence of genuinemacroparametersf syntax(see amongothersRizzi 1989 on the subjacencyparameter).A competingview has beenchampionedbyMarkBaker nhis 1996work onpolysyn-thetic languages. Baker argues that both macroparameters nd microparameters reneeded to account for observedpatternsof crosslinguisticvariation.Baker attributesthe greaterexplanatoryrole of microparametersn the researchon Romancedialectsto a methodologicalartifact: he operationof macroparameterss evident only if onecomparesgeneticallyandtypologicallydiverselanguages,becauseclosely related an-guages tend to be extremelysimilarin theirmacroparametrichoices.Questions of methodology are critically important or the additional reason thatmacroparameters, y definition,are more abstract han the surfacecharacteristicsofgrammar hatthey help to determine.Thus,macroparametersrerelatively unlikelytobe discoveredby simple crosslinguistic comparisonsof surfaceproperties, n the ab-senceof a larger heoretical rameworkderivedfromfine-grainedanalysesof individuallanguages.1For example, the operationof a macroparameterould easily be obscuredby the existence of two distinctarraysof parameter ettings,each of which gives riseto similarsurfaceconstructions: givenlanguagecouldpresenta spurious ounterexam-ple to a validmacroparametriceneralization,by allowing(thesemblanceof) aparticu-lar surface constructionwithoutthe predictedgrammatical oncomitants.Hence, thepresentpaucityof convincingmacroparametricnalyses maywell reflectthe limited numberandvarietyof languagesfor which thereexist detailed,theoreticallysophisticatedgrammaticalanalyses;or indeed may reflect more general deficienciesin the grammaticalramework heoriesthat arecurrentlyavailable. To circumvent helimitationsof a purelycomparativeapproach,I add a novel source of evidence: thetime course of child languageacquisition.2

1.2. COMPLEXPREDICATES.hepresent nvestigation ocuses on argument tructure,and morespecificallyon structures hat aretypically analyzedas eitherCOMPLEX-PREDI-CATEor SMALL-CLAUSEonstructions. English, for example, permits a main verb tocombine with a secondarypredicateandforma new expression hatsemanticallyresem-bles a simpleverb.Examplesareprovided n 1. Theparadigmcases are the resultative(la), in which the main verb combineswith an adjective phrase(AP) (paint red), andthe verb-particle onstruction lb), in which the main verbcombines with a postverbalparticle(pick up).

(1) a. Johnpaintedthe house red. (resultative)b. Mary picked the book up/picked up the book. (verb-particle)c. Fred made Jeff leave. (make-causative)d. Fred saw Jeff leave. (perceptualreport)e. Bob put the book on the table. (Put-locative)f. Alice sent the letter to Sue. (to-dative)g. Alice sent Sue the letter. (double-objectdative)1In particular,Greenberg's(1966) program or the discovery of implicationaluniversals suffers fromthis difficulty. See Hoekstra & Kooij 1988 for discussion.2 In this way my studyfollows Croft'ssuggestionto supplement ypologicalevidence with 'othersourcesof data (e.g., direct or comparativehistorical evidence, child language development, and intralinguistic

variation)' (1995:91).

325


4/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)Certain syntactic analyses treat the main verb and secondary predicate of these construc-tions as forming a syntactic complex predicate. Analyses of this type can be found inLarson 1988a,b, 1990, Hale & Keyser 1993, Chomsky 1993, and Marantz 1993, al-though some of these authors do not extend the approach to the full range of construc-tions in 1. Alternative approaches include the small-clause analyses of Stowell 1983,Kayne 1985, Hoekstra 1988, and Den Dikken 1995, and the zero-syntax analysis ofPesetsky 1995. For expository convenience I will refer to the constructions in 1 simplyas complex predicates, but with the understanding that several different syntactic ap-proaches are possible.An illustration of the distinctive semantic properties of English complex predicatesis provided in 2. The simple transitive sentence in 2a describes a pure process or activity(cf. Vendler 1967), and is therefore more fully compatible with the simple durativemodifier for an hour, than it is with the telic (or endpoint-bounded) modifier in anhour.

(2) a. John hammered the metal (for an hour)/(?? in an hour).b. John hammered the flat metal (for an hour)/(?? in an hour).c. John hammered the metal until flat (?for an hour)/(?? in an hour).d. John hammered the metal flat (?for an hour)/(in an hour).Addition of the attributive adjective flat in 2b, or even the adverbial modifier until flatin 2c, does not substantially alter the acceptability of the aspectual modifier in an hour.Yet, creation of the complex predicate (resultative) in 2d profoundly alters the aspectualproperties of the sentence, as indicated by the full acceptability of in an hour. Thecomplex predicate thus exhibits the aspectual character of an accomplishment predicate,in which the eventual flatness of the metal provides a natural endpoint (or telos) forthe hammering process.

The availability of the complex-predicate constructions in 1 varies across languages.Romance, for example, appears to be a strong candidate for a language group in whichcomplex predicates of the English type are systematically excluded. The Romancelanguages have long been noted to contrast with English and other Germanic languagesin that they exclude resultative constructions of the type in 2d (cf. Green 1973, Kayne1984, and especially Levin & Rapoport 1988). Furthermore, the Romance languagessystematically lack direct counterparts to the English verb-particle, make-causative, anddouble-object dative constructions.3 If we speculate that the availability of the complex-predicate family of constructions is indeed a point of parametric variation, then theresultative construction is perhaps the most appropriate diagnostic for the family'savailability, because it does not involve any idiosyncratic, closed-class lexical items(in contrast to the verb-particle construction), and because it displays, in an especially

3Two caveats are in order. First, it should be noted that Romance does provide at least superficialcounterpartso some of the otherEnglish constructions 1) that have received complex-predicateanalyses.This may simply indicate that some of the surfaceforms in 1 areambiguous n structure.Also, it shouldbenotedthatthe Germanic anguages,whichgenerallyresembleEnglishin permittingmostof the constructionsin 1, do not necessarilypermitALL of the constructions.For example, the English double-object(double-accusative)construction ig) lacks a directcounterpartn German,as illustrated n i, where morphologicaldative-markingnotaccusative-marking)s requiredon the definite articleof the indirectobject.Hence,evenlanguagesthat allow complex predicates n general,may disallow specific complex-predicate onstructionsfor independentreasons.

(i) *Hans hat den Mann das Geld gegeben.Hans has the-Acc man the-Acc money given'Hansgave the man (acc.) the money (acc.).'

326


5/20


clear-cutform, the characteristic emanticpropertiesof the complex-predicateclass(e.g., the creationof an accomplishmentpredicate, n 2d, froman activityverb andanadjective).Evidence from child languageacquisitionsupports hehypothesisthatEnglishcom-plex-predicateconstructionsconstitute a naturalclass, interrelatedby shareddepen-dence on a single, parametricpropertyof English.StromswoldandSnyder(1995 and1997) have employed longitudinal ranscriptdatafrom the CHILDESdatabase(Mac-Whinney& Snow 1985, 1990) in a studyof the spontaneous peechof twelve childrenlearningEnglish.Age of firstclearuse servedas a measureof acquisitionfor each ofthe sentence types in lb-g, all of which are used with high frequencyin the speechof adults and older children.4The majorresult, supportedby a varietyof statisticalmeasures,was thatevery child acquired he sentence-types n lb-g as a group.5Thus,evidence fromchild language acquisitionsupports he view that the complex-predicateconstructionsof English dependon a single, parametric ropertyof thegram-mar. As soon as a child acquiresany one of these constructions, he others quicklyfollow. Yet, a majorquestionremains:What,precisely, is the parametric roperty hatthe children are acquiring?In particular,can the propertybe representedwithin thelexical entryforsome single,abstract unctionalhead,or is it a moreglobalcharacteris-tic of the grammar hat cannot be reduced to the propertiesof any single lexical item?A possibleanswerto thesequestions s suggested by researchon the syntaxof Dutchand Afrikaans:complex-predicate onstructions n these languagescommonlyinvolveovert morphological compounding.This suggests that the morphological availabilityof productiveroot compounding,as a global propertyof the language,could perhapsbe the crucialprerequisite or syntacticavailabilityof complex predicates.In Dutch(Neeleman& Weerman1993,Neeleman1994), theword-order ossibilitiesfor resultativesandverb-particle ombinationsareunusuallyrestrictive(3a,b) (Neele-man & Weerman:436,exx. 6-7).

(3) a. dat Jan de deur (vaak) groen (*vaak) verfdethat John the door (often) green (*often) painted'thatJohn often paintedthe door green'b. dat Jan het meisje (vaak) op (*vaak) merktethat John the girl (often) up (*often) noticed'that John often noticed the girl'Despite the usual flexibility of word order in the Dutch Mittelfeld,an adverb cannotintervenebetween a verb and an associated resultpredicate(3a) or particle(3b).Similarly,LeRoux (1988) reports hatAfrikaansverb-particle ombinations e.g., af+ kyk'off + look') behaveas a unitin a varietyof syntacticcontexts, as forexamplewhen V-raisingappliesto an embedded clause in 4a, b (Le Roux:241,ex. 9a).

(4) a. Hy sal nie [die antwoordeby my e] kan af + kyk nie.he will not the answers from me can off look not'He will not be able to crib from me.'b. *Hy sal nie [die antwoordeby my af e] kan kyk nie.he will not the answers from me off can look not'He will not be able to crib from me.'4The resultativeconstruction a unfortunately adto be excluded from the spontaneous-speech nalysis,because of its extremely low frequency n the speech of both children and adults.5A considerablevariety of possible nongrammatical xplanationsfor this patternhave been tested andruled out. The details are reported n Snyder& Stromswold1997.

327


6/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)

Both Neeleman and Le RouxanalyzetheDutch/Afrikaansacts as follows: thecomplexpredicatesin these examples are morphologicalcompounds.In other words, certaincomplexpredicates n Dutch and Afrikaanshave not only the semanticproperties,butalso the morphologicalproperties,of a single, complex word.6

1.3. MORPHOLOGICALOMPOUNDS.he present investigation tests the following hy-potheses: (1) English complex predicates necessarilyinvolve a morphologicalcom-pound at some abstract evel of grammatical epresentation, ven though they do notexhibit the morphologicalcharacteristics f a compound n the surface form of a sen-tence, (2) the pointof grammarhat children areacquiringwhen they suddenly beginproducingEnglish complex-predicate onstructions, s the knowledgethat the type ofcompoundingrequired or complex predicates s available n English,and(3) the rele-vant type of compounding is PRODUCTIVE,NDOCENTRICOOT OMPOUNDING.Moreprecisely,the proposal s thatthe constructions n 1 all dependon the markedvalueof a parameterhat s fundamentally parameter f morphological ompounding:

(5) COMPOUNDINGARAMETER:he grammar {disallows*, allows} formation ofendocentriccompoundsduringthe syntacticderivation.[*unmarked alue]The idea behind the formulation n 5 is thatmorphologicalcompoundscan be createdin at least two ways: as deliberatecoinages (independentlyof the settingof 5), and asautomaticproductsof syntacticderivation(when 5 assumes the markedvalue). Thelatterprocess accounts for the extremeproductivityof endocentriccompoundinginEnglish (takingthe markedsetting of 5), where a compound such as frog man, forexample,can be called into service to designatea man with almostanytype of connec-tion to frogs: a man who resembles a frog, behaves like a frog, or collects frogs, forexample.As detailedin Bauer 1978, the situation s quite different in French(takenhere tohave the unmarked ettingof 5), where thecorresponding ompoundhommegrenouille(lit. 'manfrog') is restricted o its original, exical sense of 'underwater iver'. Deliber-atecoinagesof the French ypehave aninterpretationixed at the time of coinage,whilesyntacticallyderivedcompoundsof the type permittedby English can be interpretedcompositionally, n muchthe same way as syntacticphrases.For this reason,Englishroot compoundscan be createdspontaneously,unconsciously,to fit the needs of themoment.Thepresentview of productivecompoundings an extensionof ideas in Baker1988.Baker abandons he strong exicalisthypothesisof Chomsky 1970, andarguesinsteadthatcertainprocesses of word-formation ccurby means of the syntacticcombinationof heads (especially head-to-headmovement). On his approach,morphologysimplyimposes well-formedness conditions on heads, and applies equally to heads formedduring,and heads formed outsideof, the syntacticderivation.As in the presentdiscus-sion,operationsof word-formationhatoccur n thesyntaxare associatedwithparticularproductivity,while operationsof word-formation hat take place 'in the lexicon' (i.e.outsidethe syntacticderivation)are less productive.Yet actually allowing English root-compounding o take place in the syntax is adepartureromBaker'ssystem.Bakerassumes hatthe formationof Englishcompoundsoccurs in the lexicon, because of the generic,nonreferentialnterpretation f the N in

6 Notice, however,thatneitherDutchnor Afrikaans ONSISTENTLY requiresanovertcompound n complex-predicateconstructions.Forexample,as discussedin Neeleman 1994, DutchV2-movementroutinely sepa-rates the main verb from the remainderof a complex predicate.

328


7/20


an English (gerundive) N-V compound such as man-watching (Baker 1988:78-81).Following DiSciullo & Williams 1987, he assumes that words are islands with respectto referential properties. At the same time, he takes the process of N-incorporation inNahautl, for example, to involve syntactic head-to-head movement, with the result thatthe N kocillo 'knife' in a structure such as ki-kocillo-tete'ki '3sS/3sO-knife-cut' bindsa trace outside the complex word, and is potentially referential ('he cut it with theknife'; Baker 1988:79, based on Merlan 1976).An alternative explanation for this difference in referentiality, however, would bethat N-incorporation in Nahuatl involves head-to-head movement of the N out of thehead position of a full NP, while root compounding in English involves the directsyntactic merger of two heads (cf. Chomsky 1995, 1998 on Merge as a generalizedtransformation). If referential interpretation of the N depends on having a full NP (andperhaps a DP) in the tree, but only N-incorporation is compatible with this additionalstructure, then we can capture the observed difference in referentiality even if Englishroot-compounding occurs in the syntax.7

2. METHODANDRESULTS. wo empirical predictions follow immediately from theidea that the formation of complex predicates depends on syntactic compounding. First,across languages, the availability of complex predicates (as found in English) shouldpattern closely with availability of productive root compounding (e.g., N-N compound-ing). Second, in children acquiring English, the age at which complex predicates arefirst used productively should correspond very closely to the age at which novel rootcompounds are first produced.

2.1. CROSSLINGUISTICURVEY. he first prediction was evaluated by a crosslinguisticsurvey, the major results of which are summarized in Table 1. The survey was limited tolanguages for which native informants were readily available, but nonetheless included asubstantial range of language groups: Afroasiatic, Austroasiatic, Austronesian, Finno-Ugric, Indo-European (Germanic, Romance, Slavic), Japanese-Korean, Niger-Kordo-fanian (Bantu), and Sino-Tibetan, as well as American Sign Language and the language-isolate Basque.8

RESULTATIVESAmericanSign Language yesAustroasiatic Khmer) yesFinno-Ugric(Hungarian) yesGermanic(English,German) yesJapanese-Korean Japanese,Korean) yesSino-Tibetan Mandarin) yesTai (Thai) yesBasque noAfroasiatic(EgyptianArabic,Hebrew) noAustronesian Javanese) noBantu(Lingala) noRomance(French,Spanish) noSlavic (Russian,Serbo-Croatian) no

TABLE. Results of crosslinguistic survey.

PRODUCTIVEN-N COMPOUNDINGyesyesyesyesyesyesyesyesno (?)nononono

7 The present approach o root compounding,based on syntactic mergerof heads, is furtherdevelopedin Roeperet al. 2001.8Notice that healternative pproach f relyingon referencegrammars, ather hannative-speaker onsult-ants,would have permitteda largersample,but with an associated risk that the terminologyanddiagnosticscould be inconsistent across sources. See Newmeyer 1998, ?3.4.1 for discussion.

329


8/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)A language was judged to have productive N-N compounding only if it permitted

truly novel (nonlexical) N-N compounds, and did not require any overt morphologicalor syntactic connective to combine the nouns.9 As can be seen in Table 1, complexpredicates (as diagnosed by resultatives of the type in 2d) patterned quite closely withproductive root-compounding (as diagnosed by grammaticality of novel N-N com-pounds). Examples are provided in Appendix A.10'11Observe that Basque provides a clear example of a language in which nominal com-pounding is fully productive, yet resultatives are unavailable. Hence, despite the strongtendency illustrated in Table 1 for nominal compounding and resultatives to patterntogether, the relationship must be unidirectional: root compounding is a necessary, butnot sufficient condition for the availability of resultatives.12Notice also that Arabic and Hebrew receive here a tentative classification as non-compounding languages, despite the availability of construct-state expressions. This isbecause Semitic construct states overlap, in their morphological and syntactic proper-ties, both with nominal compounds (Borer 1988) and with possessive phrases (Ritter1991). Yet, given that both Arabic and Hebrew lack the resultative construction, eitherdecision on the productivity of compounding will be consistent with the more generalpattern observed in the crosslinguistic survey: Resultatives are available in a givenlanguage ONLYif nominal compounding is productive.13While the evidence reported here, from my own informant work, is fully consistent

9For example, French constructions nvolving the prepositionde or a were excluded (e.g., sortie desecours 'emergencyexit', lit. 'exit OF escue'). Likewise, constructionsnvolving obliquedeclensional mor-phology closely corresponding o Frenchde or a would not be classified as N-N compounds;cf. Russianprazdnikpjesni 'song festival', lit. 'festival OFsongs'. Notice that all the languagesin the presentsamplepermitN to serve as a nominalmodifierwith the help of an adpositionand/oroblique case-marking,butthe availabilityof bare-Ncompoundingdistinguishesa propersubset of the languages.10The resultspresentedhere diverge in some cases from the resultsreported n Snyder 1995b, becauseof moreflexible diagnosticcriteria n the presentstudy.In Snyder1995ba potentialresultativeconstructionwas excludedif it containedanymaterialabsentfrom the Englishresultative,such as the ASL wordglossedas BECOME n AppendixA, la. In the present study,the element BECOME n ASL, and haj in Thai,areregardedas possible overt counterparts o a null morpheme n the English resultative(cf. among othersSnyder1995a).Use of the predicatepaint the housered, as a paradigm ase of the resultative,has also beenreconsidered n light of the findingthatcertain anguages(e.g., Javanese)allow this as the SOLExampleofan (apparent) esultative.Hence, unavailabilityof paint the house red is taken as evidence for unavailabilityof resultativesmoregenerally,but a languagehas been counted as genuinely permittingresultativesonly ifadditionalexamplesare attested(e.g., beat the metal lat, wipe the table clean). Finally,a broaderrangeofexampleshas been considered n the presentstudy, to assess the productivityof nominalcompounding na given language. See Miyoshi 1999 for discussion of problems with the Snyder 1995b diagnostic forcompoundingwhen appliedto Japanese.1l If we count Slavic, Romance,andGermanicas distinctgroupings,the crosslinguistic sampleincludeda total of thirteen anguage groups. While this is relatively small, the observed contingency nonethelessreachesstatisticalsignificanceby Fisher ExactTest; two-tailedp = .00466. In otherwords,the probabilityof the observed associationoccurringby chance, if resultativesandcompounds n fact variedindependentlyacross language groupings,would be aboutfive chances in a thousand.12If onechose to excludefromthe class of trueresultatives hose constructionsnvolvingextramorphology(e.g., translativecase-marking n Hungarian), hen languages in the category of Basque, with productivecompoundingbut no resultatives,would become more numerous n this survey.13Clark(1993) providesseveral reasonsto doubt thatHebrew constructstates areequivalentto Englishcompounds.First,theirproductivity s relativelylow in spokenHebrew,as evidencedby the fact thatmanylexical borrowings nitially broughtinto the language as compoundshave since been replaced with non-compound orms(1993:173). Second, in contrast o English,where nominalcompounding s a majorsourceof children's novel words by the age of two to three years, childrenacquiringHebrewmake virtuallynoproductiveuse of compounding hrough he age of six years (1993:175).

330


9/20


with this generalization,certainpotentialcounterexampleshave been reported n theRomance syntax literature.In particularItalian, like the other majorRomance lan-guages, lacks productive,endocentriccompounding.Yet Napoli (1992) argues thatItaliandoes permitcertainresultativeconstructions see also DiSciullo 1996).AlthoughItalian disallows 6, some speakersreportedlypermit7.

(6) *Gianni ha martellato l metallopiatto. (Napoli 1992: ex. 73)'Gianni hammered he metal flat.'(7) Ha dipintola macchinarossa. (ex. 74)'He paintedthe carred.'

Napoli arguesthat sentences with AP result predicatesare possible in Italian,butonly if the main verb is interepretedas 'focusing on the endpoint' of the event itdescribes(Napoli 1992:53).14 n otherwords,therelevantdifference s that in English,butnotItalian,one can adda resultAP to a simple activity predicateandtherebycreatean accomplishmentpredicate.Hence, I conclude that the type of resultativerequiringproductiverootcompoundings this 'English'type thatpotentiallyconvertsanactivityverb into an accomplishmentpredicate.Insummary, esultativesof theEnglishtypearefoundonly in languageswithproduc-tive endocentriccompounding.Nonetheless, confident identificationof surface con-structions romdifferent anguagesas grammatically quivalentor disparates clearlya delicate matter.Confidencein the general picture presentedby the crosslinguisticsurveywill be greatlyincreased f supporting vidence canbe providedfroma seconddomainof investigation. turnnowto convergingevidencefrom childlanguageacquisi-tion.

2.2. CHILDREN'SACQUISITION OF ENGLISH. he second prediction in ?1.3, that anygiven child learning English shouldacquirecomplex predicatesandproductiveendo-centriccompoundingat approximatelyhe same age, was tested in a studyof sponta-neous productiondatafor ten childrenfrom the CHILDESdatabase MacWhinney&Snow 1985, 1990).The tenchildrenwere a subsetof thosestudied n Snyder& Stroms-wold 1997. The age of acquisitionfor a given grammatical onstructionwas taken asage of first clearuse; latertranscriptswere checkedin all cases to confirmthatthe firstclear use was followed soon afterwardby regularuse (see Stromswold1996).Thediagnosticforproductivecompoundingwas novel N-N compounding;N-N com-poundingis the most frequentlyused form of compounding n English. To count asnovel, a child's N-N compoundcouldnot be a lexicalized form(e.g. toothbrush,applejuice), and the context of the child's utterancehad to supportthe interpretationhatthe compoundwas invented on the spot. Indeed,the lattercriterionwas surprisinglyeasy to satisfy, as childrenwere often found 'teaching'new compoundsto the adultsin the transcripts.14 According to Napoli, the distinguishingcharacteristicsof grammaticaland marginally grammaticalresultatives n Italianall serveto make the mainverb's natural ndpointmore salient. One of themoreexotic

focusing devices that she describes is emphaticdoublingof the resultpredicate,as in i.(i) Ho stirato a camicia piatta *(piatta).(exx. 109 and 112, pp. 74-75)'I ironed the shirt flat (flat).'

Interestingly,a similareffect of emphasison the resultpredicatehas been reportedby Demonte (1991) forat least one variety of Spanish,as shown in ii; this was broughtto my attentionby Liliana Sanchez andMarcelaDepiante.(ii) Pedro edific6 la casa *(muy) amplia. (Demonte 1991, ex. 2c)'Pedro built the house (very) wide.'

331


10/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)

2.6 -2.5 -2.4-2.3 -

0. 2.20X 2.1 -zi 2.0- -

1.9 -1.8-1.7 ,, , ,

1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6N-NCompound

FIGURE1. First N-N compoundversus first verb-particle ombination ages in years).The age of acquisitionof a varietyof complex predicateconstructionshad alreadybeen determined or each child in Snyder& Stromswold 1997. In additionto the ageof acquisitionof productiveN-N compounds,a numberof new controlmeasureswereobtainedfor each child: the age at which the child's mean lengthof utterance MLU)first reachedorexceeded 2.5 words; heageof firstclearuse of a lexical N-N compound,such as toothbrush;and the age of first clear use of an adjective-nouncombination,such as big dog. The MLU measure was a control for the possibility that complexpredicatesand productive compoundingmight be acquired together simply becauseboth form a part of the 'grammarexplosion' that occurs at the transitionbetweenBrown's (1973) Stages II and III.More generally,MLU = 2.5 serves as a proximatedevelopmentalmilestone,allowing one to assess the contributionof generaldevelop-mentalfactorsto the time course of acquisitionfor compoundingandcomplexpredi-cates. Lexical N-N compounds and adjective-nouncombinations serve as closelymatchedcontrols for the conceptualcomplexityandlengthof utteranceof novel N-Ncompounds.Theresults, nbrief,wereas follows. Ages of first clear use of a novel N-N compoundwere EXCEPTIONALLYell correlated with the ages of acquisition reported in Snyder &Stromswold 1997 for verb-particleconstructions(lb) (r = .98, t(8) = 12.9, p


11/20


were generally acquired somewhat later than the other complex predicates, as discussedin Snyder & Stromswold 1997.When the contribution of each of the control measures is subtracted out, through apartial regression procedure, all of the above correlations remain statistically significant,except for the correlation between compounding and double-object datives. The double-object construction thus appears to be something of an outlier among the complex-predicate constructions, when viewed in relation to morphological compounding. Afterpartialing out the contribution of the ages at which MLU first reaches or exceeds 2.5words, a statistically significant portion of the remaining variance in the ages of acquisi-tion for novel N-N compounding can still be accounted for by the ages of acquisitionfor verb-particle constructions (r = .94, t(7) = 7.41, p = .0001), causative/perceptualconstructions (r = .77, t(7) = 3.14, p = .0164), put-locatives (r = .88, t(7) = 4.87,p = .0018), or to-datives (r = .80, t(7) = 3.41, p = .0133), but not double objectdatives (r = .59, t(7) = 1.95, p = .0919, marginally significant).

Similarly, when ages of first clear use of a lexical N-N compound are partialed out,a significant portion of the remaining variance in ages of acquisition for novel N-Ncompounding can still be accounted for by verb-particle combinations (r = .95, t(7)= 7.72, p = .0001), causative/perceptual constructions (r = .79, t(7) = 3.34, p =.0124), put-locatives (r = .90, t(7) = 5.54, p = .0009), or to-datives (r = .86, t(7)= 4.55, p = .0026), but not double object datives (r = .37, t(7) = 1.06, p = .3259,NS). Finally, when ages of first clear use of an adjective-noun combination are partialedout, a significant portion of the remaining variance in ages of acquisition for novel N-N compounding can once again be accounted for by verb-particle combinations (r =.95, t(7) = 8.45, p = .0001), causative/perceptual constructions (r = .82, t(7) = 3.77,p = .0070), put-locatives (r = .91, t(7) = 5.87, p = .0006), or to-datives (r = .88,t(7) = 4.99, p = .0016), but not double object datives (r = .48, t(7) = 1.43, p= .1954, NS). (First clear uses of novel compounds, lexical compounds, and A-Ncombinations are provided in Appendix B.)

Double-object datives are thus a POSSIBLE exception to the compounding/complex-predicate generalization. While the ages of acquisition for double-object datives aresignificantly correlated with the ages for novel N-N compounding, the correlation be-comes nonsignificant after one subtracts out the contribution of a control measure (MLU= 2.5, A-N, or lexical N-N), through partial regression. At least two explanations arepossible: First, the English double-object construction may not in fact depend on theavailability of productive root compounding. Second, the double-object dative maydepend on both productive compounding and some other, late-acquired prerequisite;hence, compounding alone would be a relatively weak predictor of when the doubleobject datives become available to the child.

The first possibility predicts that double-object datives and novel compounds can

andIsobe (2001) find a close associationbetweensuccessfulproductionof novel compoundsand successfulcomprehensionof resultatives n a laboratory tudy of Japanesechildren.Snyderand Chen (1997) reportthat childrenacquiringFrench,a languagewith the negative settingof the compoundingparameter, cquirethe N-de-N paraphrase f English N-N compounds significantly ater than the paraphrasesor English put-locatives, make-causatives,and verb-particleconstructions; hus, as expected, the ability to form N-de-Nexpressions is not a prerequisite or these Frenchargument tructures.Slabakova(1999) reportsthat adultEnglish-speakers earningSpanishexhibit similarperformanceacross tasks testing theirunderstandinghatN-N compoundings unproductiven Spanish,and taskstestingtheirunderstandinghatEnglish-typedouble-object datives, verb-particle onstructions,andresultativesareunavailable n Spanish.(See also Slabakova1997 for relatedfindings from adultspeakersof Slavic languageslearningEnglish.)

333


12/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)enter the child's speech in either order. The second possibility, however, makes adistinctiveprediction:no child will beginto producedouble-objectdativessignificantlyearlierthannovel compounds.This predictionwas checkedagainstthe data from theten childrenin the study, and appears o be correct.Of the ten childrenstudied,fiveproducedtheir first double-objectdative later than their first novel N-N compound,threeproduced heir firstdouble-objectdativein the sametranscript s their first novelN-N compound,andtwo (Eve andAllison)firstproduceda double-objectdative earlierthan their first novel compound.Yet, for neither Eve nor Allison is the delay statistically significant.Eve producedexactly one double-objectdative in her corpusbefore the first novel N-N compound.The relativefrequencyof double-objectdatives andnovel N-N compounds n her laterspeech (basedon the last two transcriptsn hercorpus)was 4:15 (datives:compounds).Hence, samplingone double-objectdative before the first novel compound is fullyconsistent with the two constructionsbecomingavailableconcurrently,and the gap isnonsignificantby modifiedsign test (p = .211, NS). Similarly,Allison producedonlytwo double-objectdatives before her first novel N-N compound.Basedon the relativefrequencyof 4:3 (datives:compounds)n herspeech in the last transcript f hercorpus,the gap is againnonsignificant,andfully consistentwith concurrentacquisitionof thetwo constructions p = .327, NS). Hence, despitethe relative 'noisiness' of the datafordoubleobjectdatives(and eaving opentheidentityof theproposedsecondprerequi-site), the availableevidence still supports he conclusionthat the double-objectdativehas productivecompoundingas one of its prerequisites.16

3. DiscusSION. havepresented onvergingevidence,fromcrosslinguisticvariationand child language acquisition,for a strongassociationbetween complex predicatesandmorphologicalcompounds.Thesefindingsareproblematicor theview thatpointsof parametric ariationn syntaxarestrictlyconfined to the lexical entriesof functionalheads, such as ToandDO see in particularChomsky1993). This view would requirethatsome single, independentlymotivated unctionalhead finds a naturalrole both incomplexpredicatesand in a morphologicalcompoundsuch as coffeecup.Yet, Englishcompounds are well known to resist the inclusion of overt functional morphology(Kiparsky1982), renderingdoubtfulany proposalof a null functionalhead in such acompound.The inclusion of covert syntacticmaterial n root compoundswould perhapsgainplausibility f root compoundscould in fact be syntacticallyderivedfrom whole sen-

16 A refereesuggests thatthe second factor could relate in some way to lexical learning,for example ifthe earlydouble-objectdativesalwaysinvolved the same one or two verbs. Examinationof thedata revealedthat the first verb used in a double-objectdative was give (for 5 children),get (3), send (1), or read (1).Otherearly double-objectdatives (producedbefore the first to-dative) involved bring, build, buy, make,show, tell, or write (in the sense of 'draw').The first verb used in a to-dative was read (4), give (3), show(1), or get (1).While the range of early verbs in double-objectdatives was reasonably large (not simply one or twoverbs), the notionthatlexical learningcould have been exertingan excessive influence on the ages of firstuse for double-objectdatives does receive some support.First, for six of the children,the frequencyofdouble-objectdatives was initially rather ow, and then gradually ncreased.Hence, the ages of first clearuse would have been a relatively 'noisy' measureof grammaticalknowledgefor these children.Second, asnoted in Snyder& Stromswold 1997, the childrenmanagednot to overextend the double-objectdative toverbswith which it is impossiblefor adults(cf. *SuewhisperedChrisa secret). This fact suggests thatevenwhen childrenrecognize that the double-objectdative is syntacticallypossible in English, they are stillconservative n deciding, for a verb, whetherit is compatiblewith the structure.Thus, lexical learning,inthis sense, could perhapsbe the 'second prerequisite' or double-objectdatives.

334


13/20


tences, as proposedin Lees 1960, 1970 and Levi 1973, 1974, 1975. Such proposalshave been sharplycriticizedby Downing (1977), who argues that there is no fixed,finite set of possible semantic relations between the modifier and the head in Englishnominalcompounds see also Gleitman& Gleitman1970).Rather, hepossibleseman-tic relationsvary as a function of pragmatic actors,such as whetherthe compound sintended 'as a categorylabel or merely a demonstrativedevice' (Downing 1977:839),and also as a function of the semantic class of the head N (e.g., naturalobject vs.synthetic object). If Downing is correct that the possible interpretations f an EnglishN-N compoundcannot be deduced fromthe detailsof its syntacticderivation, hen theargumentor a richsyntacticstructuren suchcompounds s correspondinglyweakened.Thus, somehow reducingthe compoundingparameterof 5 (repeatedbelow) to theinformationcontained n the lexical entryof a single functionalhead,would seem tobe a distortionof a qualitativelydifferenttype of parameter.

(5) COMPOUNDINGARAMETER:he grammar {disallows*, allows} formation ofendocentriccompounds duringthe syntacticderivation.[*unmarked alue]The setting of the parameter n 5 is presumablya general propertyof morphology,potentially affectinga vast rangeof open-class morphemes.Thisparameters 'lexical'in the very generalsense that it governs principlesof wordformation,but it is not bynaturetied to any single lexical entry.Hence, 5 is perhapscompatiblewith the viewthat the child's acquisitionof syntax reduces to acquisitionof the lexicon, but onlyunder a very broadsense of 'lexicon' that would include quite general propertiesofword formation, ndependentof any single lexical entry.An interestingcharacteristicof the lexical hypothesis,as developedin Borer 1984,Fukui1986,andChomsky1993,is thatacquisitionof syntaxmight proceed n amannerparallelto word learning:The child learningsyntax would simply be acquiringthelexical entriesforclosed-class,nullandovert,word-level tems.Thetypeof connectionbetween syntactic knowledge and the lexicon that is most compatible with thecompounding/complex-predicate arameter,however, does not permitan exclusively'word-learning'approach o the acquisitionof syntax, althoughit is of acquisitionalinterest n the somewhatweakersense that it adds a new domain of morphology(viz.complex word formation)as a possible source of evidence aboutlanguage-particularpropertiesof syntax.17The centralrole of productivity n characterizing herelevanttype of morphologicalcompounding s also noteworthy.As observedby Spencer(1991:323-24), the genera-tive literatureon morphologicalcompoundinghas (surprisingly)ended to neglect theissue of productivity n general,and theissue of crosslinguisticvariation n productivityin particular. f one adoptsthe view thatproductivity s an essentialpropertyof Ger-manic compounding, then the lack of PRODUCTIVE-N compounding in Romance mightin itself be takenas a clearexampleof a pointof grammaticalvariation hat cannot be

17A refereemadetheinteresting uggestionthat unctionalheadsmightstillplaya centralrole in determin-ing whetherroot compounding s productive n a given language. Specifically, if the presenceof functionalheads INTERFERESith compounding, hen compoundingmightbe productiveprecisely in those languageswithouta layerof functionalstructurenterveningbetween the head andthemodifierof would-becompounds.This idea, in my opinion,warrants urther nvestigation.Yet, it shouldbe noted that t falls outside the realmof the lexical hypothesis. One would need a way to force the presence of a layer of functional structurebetween the head and the modifier of every would-be compound n a noncompounding anguage,and it isunclearhow to accomplishthis, except by meansof a global parameter f the sortdisallowedby the lexicalhypothesis.

335


14/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)tied to any single lexical item. With the evidence from syntactic complex-predicateconstructions,however, the argument hatproductivityof compounding s a genuinepoint of parametricvariationbecomes considerablystronger.Moreover,the evidencefrom complex predicatesmakes it clear that the point of parametricvariationcannotbe restricted, n its consequences,to morphologyproper.The natureof the connectionbetweenproductivecompoundingandcomplexpredi-cates is an importantssue, but will only be touchedon here, as it remains a topic ofongoingresearch. Foradditionaldiscussion see Snyder1995a,1995b,?2.4,andBeck &Snyder2001). Inbrief,the connectionappearso be semantic ncharacter:he distinctivesemantic characteristics hat unify the complex-predicateconstructionsderive froma mode of semanticcompositionavailableonly within endocentriccompounds.Therestrictioncan be statedmorepreciselyas in 8.

(8) COMPLEXREDICATEONSTRAINT:wo syntactically independent expressionscan jointly characterize he event-type of a single event-argument,only iftheyconstitutea singleword(endocentric ompound)at thepointof semanticinterpretation.To see how 8 works,recall thatthe resultativepredicaten 2d describesa (telic)accom-plishmentevent, while the nonresultativepredicates n 2a-c instead describe (atelic)activities.(2) a. John hammered he metal (for an hour)/(?? n an hour).b. Johnhammered he flat metal (for an hour)/(?? n an hour).c. John hammered he metal until flat (?for an hour)/(?? n an hour).d. Johnhammered he metal flat (?foran hour)/(inan hour).

Moreover, he accomplishment vent describedby 2d comprisestwo subparts:Parsons(1990) calls these a development subpart the activityof hammering he metal) and aculminationsubpart the achievement event in which the metal finally becomes flat).Crucially,both the verb hammered nd theadjectivephraseflat participaten character-izingtheevent-typedescribedbythe verbphrase; nthiscase, hammered ontributeshedevelopment,and lat contributes heculmination,of anaccomplishment-typevent.18Accordingto 8, this state of affairs s possible only if hammeredand lat aresubpartsof an endocentriccompoundat the point of semanticinterpretation LF). Yet, theseexpressions clearlyfunctionindependentlyn the syntax,as evidencedby the fact thatthey are discontinuousin the sentence's surface structure.Hence, formation of therelevant endocentriccompoundmust take place duringthe syntacticderivation,andsuchcompound ormation n the syntaxis possible preciselybecauseEnglishtakes themarkedsetting of the compoundparametern 5.19'20

18As discussedin ?2.1, Italiandiffers fromEnglishin permitting esultativeAP predicatesonly whenthemain verbalreadyhas a salientend-point or the resultativepredicate o modify. In thepresent erms,Italiancannotperformrootcompoundingduring he syntacticderivation,and thereforecannot combinean activitypredicatewith a stativepredicate o createa descriptionof an accomplishment vent.19Recall that overt combinationof the resultphrasewith the main verb is foundin Dutchexamplessuchas 3, above. A referee adds that Swedish and Norwegian similarly permit optional,overt incorporation fa resultadjectiveto the left of the main verb.

20 Extensionof this general approach o the remaining ypes of English complex-predicate onstructionsin 1 requires hatthe aspectualproperties event structure)of those constructionsbe composedin a similarway, fromsyntactically ndependent xpressions. Analyses of this type have been proposed n Snyder1995afor verb-particle onstructionsanddativeconstructions.Moregeneralsupport or the approach omes fromthe observation hat most or all of the constructions n 1 are associated with distinctiveaspectualproperties;for discussion, see in particularTenny 1994.

336


15/20

ON THE NATURE OF SYNTACTICVARIATION4. CONCLUSION.anguage acquisition and comparative syntax provide convergingevidence for a parameter that determines both the availability of productive, endo-centric compounding, and the availability of a range of syntactic complex-predicate

constructions. This compounding parameter resists reduction to the lexical entry fora functional head or other closed-class lexical item, however, because no suchclosed-class lexical item has yet been independently motivated in root compounds.Thus the compounding parameter appears to be a substantive parameter, in theclassical sense of Chomsky 1981. My study demonstrates the considerable potentialof child language acquisition as a testing ground for hypotheses about the natureof Universal Grammar.

APPENDIX A: CROSSLINGUISTICDATA(Transliteration f non-Romanwritingsystems is only approximate.)

(Al) Examplesof resultativesa. HE HAMMER METAL BECOME FLAT. (ASL, word-by-wordgloss)[The word glossed as BECOME s obligatory.]b. John hammered he metal flat. (English)c. Hans hammertdas Metall flach. (German)'John hammers he metal flat'd. A munkas apos-ra kalapacsoltaa f6met. (Hungarian)the worker flat-TRANSammer-PsThe metal

'The worker hammered he metal flat.'(The resultpredicate s marked or translativecase; cf. also Maracz1989:223)e. John-ga teeburu-okiree-ni hui-ta. (Japanese)John-NoMable-Aceclean wipe-PST(The predicatekireeniappears n a tenseless form used both for resultpredicatesand adverbials)f. Kira wai daik kpaet. (Khmer)Kira hit metal flat

'Kirabeat the metal flat.'g. John-i teibl-ul kekuti tak-at-ta. (Korean)John-NOMable-Acc clean polish-PST-COMPLEMENTIZER'Johnwiped/polishedthe table clean.'

[The predicatekekutiappears n a tenseless form used both for resultpredicatesand adverbials]h. Ta ba tie guan da ping. [Tones omitted] (Mandarin)(s)he BA iron pipe hit flat'(S)he beat the iron pipe flat.'

[The directobject tie guan necessarily appears n a preverbalBA phrase.]i. Ja: t'up lo:ha?haj bae:n. [Tones omitted] (Thai)NEG-IMPERammer metal HAJ (be-)flat'Don't hammerthe metal flat.'

[The particleHAJis obligatory.](A2) Paraphrases equired n place of resultativesa. Lokoda taroktuelhadide haete?osbohoh mosoto?han. (Arabic,Egyptian)Lokodabeat metal/ironuntil-it-became lat

'Lokoda beat the metal until it became flat.'b. Gorri-z atz-azal-ak pintatzenari naiz. (Basque)red-with/in inger.covering-PL ainting AUX'I am paintingmy finger nails with/inred.'c. Jean a martel6 le metaljusqu'a ce qu'il 6tait plat. (French)Johnhas hammered he metal until-to that that-itbe.PsT lat'John hammered he metal until it was flat.'d. Dani tzavaA ?it ha-bayet bi-?adoom. (Hebrew,Modem)Dani paintedP(ACC)he-house in-red'Dani paintedthe house in (the color) red.'

337


16/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)e. [Tukangpande-nipun]mande wesi ngantosgepeng.workerforge-poss beat iron until flat'The blacksmithbeat the iron until (it was) flat.'f. Joe abisi ndakona modobo motani.

Joe he-painthouse with paintred'Joe paintedthe house with redpaint.'g. Ivan pokrasildom v krasnyjtsvet.Johnpaint-PsT ouse in red color'Johnpaintedthe house in the color red.'h. Johnje ofarbao kucu u crveno.Johnis paintedhouse in red'Johnpaintedthe house in (the color) red.'i. Juangolpe6 el hierro hastaque estabaplano.John beat-PsT he iron until that be-PsT lat

'Johnbeat the iron until it was flat.'(A3) Examplesof novel compoundsa. BANANA BOX

(for 'a box in which bananasare stored')b. liburu-kutxa

(Javanese)

(Lingala)

(Russian)

(Serbo-Croatian)

(Spanish)

(ASL, word-by-wordgloss)(Basque)book box

(for 'a box in which books arestored')(See Saltarelli 1988:262 on the productivityof nominalcompounding n Basque.)c. worm can (English)(for 'a can in which fishing bait is stored')d. Wurmkanne (German)worm + cane. giliszta ved6r (Hungarian)wormbucketf. bananabako (Japanese)banana+ box(Rendakuconverts hako to bako.)

g. kapongjole:n (Khmer)can wormh. pelley-thong (Korean)worm + cani. you ji (Li & Thompson p.50, ex. 53, tones omitted) (Mandarin)oil stain (for 'a stain caused by spilling oil on clothing')(See Li & Thompson1981:48-54 on the productivityof nominalcompounding n Mandarin.)j. na:mmap'raw (Thai)liquidcoconut'coconutjuice'(See Warotamasikkhadit 963:59-69 and Fasold 1968 for early generativetreatmentsof nominal

compounding n Thai.)(A4) Paraphrases equired n place of novel compoundsa. sandu?el moz (Arabic,Egyptian)box-of banana[construct tate construction]b. boite aux vers (French)can for-the wormsc. kufsattulaAimcan-of worm [construct tateconstruction]d. bok ngangge wadahpisangbox for containbanana(s)e. linzanzampo-na mpambocan for wormse. bankadlja chervej

can for wormsf. konzervaza crvecan for wormsg. bote de/con gusanoscan of/with worms

(Hebrew,Modem)(Javanese)(Lingala)(Russian)(Serbo-Croatian)(Spanish)

338


17/20

ON THE NATURE OF SYNTACTIC VARIATIONAPPENDIX B: EXAMPLES FROM CHILDREN'S SPEECH (AGES IN YEARS)

(B1) First novel N-N compoundsCHILD AGE NOVEL N-N COMPOUNDAdam 2.26 tattoomanAllison 2.33 animalcupApril 2.08 pig bookEve 1.83 pig ( = peg) toyNaomi 1.92 bunny girlNathaniel 2.47 Big + Bird bookNina 1.99 zoo bookPeter 1.87 tape + recorderbuttonSarah 2.59 ribbon hatShem 2.25 bunny + rabbit record(B2) First lexical N-N compounds:CHILD AGE LEXICAL N-N COMPOUNDAdam 2.26 apple juiceAllison 1.72 baby dollApril 1.83 applejuiceEve 1.50 tomato soupNaomi 1.75 tape (re)corderNathaniel 2.47 snow ballNina 1.98 peanutbutterPeter 1.77 suit caseSarah 2.27 ice creamShem 2.21 orangejuice(B3) FirstA + N combinations:CHILD AGE A + N COMBINATIONAdam 2.26 big hornAllison 1.62 big babyApril 1.83 brown bearEve 1.50 good girlNaomi 1.68 bad girlNathaniel 2.47 little boyNina 1.96 little rabbitPeter 1.93 big tunnelSarah 2.30 bad lionShem 2.21 good juice

REFERENCESBAKER,MARK.1988. Incorporation:A theory of grammatical function changing. Chicago:University of Chicago Press.BAKER,MARK. 1996. The polysynthesis parameter.Oxford: Oxford University Press.BAUER, LAURIE.1978. The grammar of nominal compounding, with special reference toDanish, English, and French. Odense: Odense University Press.BECK,SIGRID,and WILLIAMNYDER.001. The resultative parameterand restitutive again.Proceedings of the 25th Boston University Conference on Language Development.Somerville, MA: Cascadilla Press, to appear.BORER,HAGIT. 984. Parametricsyntax: Case studies in Semitic and Romance languages.Dordrecht: Foris.BORER, AGIT. 1988. On the parallelism between compounds and constructs. Yearbook ofMorphology 1.45-66.BROWN,ROGER.1973. A first language: The early stages. Cambridge,MA: HarvardUniver-sity Press.CHOMSKY,NOAM. 1970. Remarks on nominalization. Readings in English transformationalgrammar, ed. by Roderick A. Jacobs and Peter S. Rosenbaum, 184-221. Waltham,MA: Blaisdell.CHOMSKY,NOAM. 1981. Lectures on government and binding. Dordrecht: Foris.CHOMSKY,NOAM.1993. A minimalistprogramfor linguistic theory. The view from Building

339


18/20

LANGUAGE,VOLUME77, NUMBER 2 (2001)20: Essays in honor of Sylvain Bromberger, ed. by Kenneth Hale and Samuel JayKeyser, 1-52. Cambridge, MA: MIT Press.

CHOMSKY, NOAM. 1995. The minimalist program. Cambridge, MA: MIT Press.CHOMSKY, NOAM. 1998. Minimalist inquiries: The framework. Cambridge, MA: MIT, MS.CLARK,EVE V. 1993. The lexicon in acquisition. Cambridge: Cambridge University Press.CROFT, WILIAM.1995. Modem syntactic typology. Approaches to language typology, ed.by Masayoshi Shibatani and Theodora Bynon, 85-144. Oxford: Clarendon Press.DEMONTE, VIOLETA. 1991. Temporal and aspectual constraints on predicative adjectivephrases. Current studies in Spanish linguistics, ed. by Hector Campos and FerandoMartinez-Gil, 165-200. Washington, DC: Georgetown University Press.DEN DIKKEN, MARCEL. 1995. Particles: On the syntax of verb-particle, triadic, and causativeconstructions. New York: Oxford University Press.DISCIULLO, ANNA-MARIA. 1996. Romance verbs and variation. Paper presented at the 26thLinguistic Symposium on Romance Languages. Universidad Aut6noma Metropolitana,Mexico City.DISCIULLO, ANNA-MARIA, and EDWIN WILLIAMS. 1987. On the definition of word. Cam-

bridge, MA: MIT Press.DOWNING, PAMELA. 1977. On the creation and use of English compound nouns. Language53.810-42.FASOLD,RALPHW. 1968. Noun compounding in Thai. Chicago: University of Chicagodissertation.FUKUI, AOKI. 986. A theory of category projections and its applications. Cambridge,MA:MIT dissertation.GLEITMAN,LILAR., and HENRYGLEITMAN. 970. Phrase and paraphrase: Some innovativeuses of language. New York: Norton.GREEN,GEORGIA.973. A syntactic syncretism in English and French. Issues in linguistics:Papers in honor of Henry and Renee Kahane, ed. by Braj B. Kachru,257-78. Urbana,IL: University of Illinois Press.GREENBERG, OSEPH.1966. Some universals of language with particular reference to the

order of meaningful elements. Universals of language, 2nd edn., ed. by Joseph Green-berg, 73-113. Cambridge, MA: MIT Press.HALE, KENNETH, and SAMUELJAY KEYSER. 1993. On argument structure and the lexicalexpression of syntacticrelations.The view fromBuilding 20: Essays in honorof SylvainBromberger, ed. by Kenneth Hale and Samuel Jay Keyser, 53-109. Cambridge, MA:MIT Press.HOEKSTRA,TEUN. 1988. Small clause results. Lingua 74.101-39.HOEKSTRA, EUN,and JAN G. KooIJ. 1988. The innateness hypothesis. Explaining languageuniversals, ed. by John A. Hawkins, 31-55. Oxford: Blackwell.KAYNE, RICHARD.1984. Connectedness and binary branching. Dordrecht: Foris.KAYNE,RICHARD.985. Principles of particle constructions. Grammatical representation,ed. by Jacqueline Gueron, Hans-Georg Obenauer, and Jean-Yves Pollock, 101-40.Dordrecht: Foris.KIPARSKY,PAUL. 1982. Explanation in phonology. Dordrecht: Foris.LARSON,RICHARD. 1988a. Light predicate raising. Lexicon Project Working Papers 27.Cambridge, MA: MIT Center for Cognitive Science.LARSON,ICHARD.988b. On the double object construction.Linguistic Inquiry19.335-91.LARSON, ICHARD.990. Double objects revisited: Reply to Jackendoff. Linguistic Inquiry21.589-632.LEES,ROBERT . 1960. The grammarof English nominalizations. International JournalofAmerican Linguistics 26.3, part 2.LEES,ROBERT. 1970. Problems in the grammatical analysis of English nominal com-pounds. Progress in linguistics, ed. by Manfred Bierwisch and Karl E. Heidolph,174-86. The Hague: Mouton.LEROUX,CECILE. 988. On the interface of morphology and syntax. Stellenbosch Papers

in Linguistics 18. Stellenbosch: University of Stellenbosch.LEVI, UDITH . 1973. Where do all those other adjectives come from? Chicago LinguisticSociety 9.332-45.LEVI, UDITH. 1974. Onthe alleged idiosyncrasy of non-predicateNP's. Chicago LinguisticSociety 10.402-15.

340


19/20

ON THENATUREOF SYNTACTICARIATIONLEVI,JUDITH . 1975. The syntax and semantics of non-predicating adjectives in English.Chicago: University of Chicago dissertation.LEVIN, BETH, and TOVA R. RAPOPORT. 1988. Lexical subordination. Chicago LinguisticSociety 24.275-89.LI,CHARLES., and SANDRA. THOMPSON.981. MandarinChinese: A functional referencegrammar. Berkeley: University of California Press.MACWHINNEY,RIAN, nd CATHERINENOW. 985. The child language data exchange sys-tem. Journal of Child Language 12.271-96.MACWHINNEY,RIAN, nd CATHERINENOW. 990. The child language data exchange sys-tem: An update. Journal of Child Language 17.457-72.MARACZ, ASZL6.1989. Asymmetries in Hungarian. Groningen: Rijksuniversiteit Gro-ningen.MARANTZ,LEC.1993. Implications of asymmetries in double object constructions. Theo-retical aspects of Bantu grammar,ed. by Sam A. Mchombo, 113-50. Stanford, CA:Stanford University Press.MERLAN,FRANCESCA. 976. Noun incorporation and discourse reference in Moder Nahuatl.International Journal of American Linguistics 42.177-91.MIYOSHI, OBOHIRO.999. Compounds and complex predicates: Evidence from Japanesefor a 'global' parameter. Proceedings of the 23rd Boston University Conference onChild Language Development, vol. 2, ed. by Annabel Greenhill, Heather Littlefield,and Cheryl Tano, 453-61. Somerville, MA: Cascadilla Press.NAPOLI, ONNA o. 1992. Secondary resultative predicates in Italian. Journal of Linguistics28.53-90.NEELEMAN,D. 1994. Complex predicates. Utrecht: Onderzoekinstituutvoor Taal en Sprak(OTS).NEELEMAN,AD, and FREDWEERMAN.1993. The balance between syntax and morphology:Dutch particles and resultatives. Natural Language & Linguistic Theory 12.433-75.NEWMEYER,REDERICK. 1998. Language form and language function. Cambridge, MA:MIT Press.PARSONS,TERRENCE. 990. Events in the semantics of English. Cambridge, MA: MIT Press.PESETSKY,DAVID. 1995. Zero syntax: Experiencers and cascades. Cambridge, MA: MITPress.RITTER,LIZABETH.991. Two functionalcategories in nounphrases:Evidence fromModemHebrew. Syntax and semantics, vol. 25: Perspectives on phrase structure,ed. by SusanRothstein, 37-62. San Diego, CA: Academic Press.RIZZI,LUIGI. 982. Issues in Italian syntax. Dordrecht: Foris.RIzzI,LuIGI.1989. On the format of parameters.Commentaryon Lightfoot. Behavioral andBrain Sciences 12.355-56.ROEPER, THOMAS; WILLIAM NYDER; and KAZUKOHIRAMATSU.001. Learnabilityin a mini-malist framework: Root compounds, merger,and the syntax-morphology interface. Theprocess of language acquisition, ed. by Ingeborg Lasser. Frankfurt: Peter Lang, toappear.SALTARELLI,MARIO. 1988. Basque. London: Croom Helm.SLABAKOVA, ROUMYANA.1997. Zero acquisition: Second language acquisition of the param-eter of aspect. Montreal: McGill University dissertation.SLABAKOVA,ROUMYANA.1999. The complex-predicate / N-N compounding connection inL2 acquisition. Proceedings of the 23rd Boston University Conference on Child Lan-guage Development, vol. 2, ed. by Annabel Greenhill, HeatherLittlefield, and CherylTano, 642-53. Somerville, MA: Cascadilla Press.SNYDER, WILLIAM.995a. A neo-Davidsonian approach to resultatives, particles, and da-tives. North Eastern Linguistic Society 25.1.457-71.SNYDER,WILLIAM.995b. Language acquisition and language variation:The role of mor-phology. Cambridge, MA: MIT dissertation.SNYDER, WILLIAM, andDEBORAHCHEN.1997. The syntax-morphology interface in the acqui-

sition of French and English. North Eastern Linguistic Society 27.413-23.SNYDER, WILLIAM, and KARINSTROMSWOLD.997. The structure and acquisition of Englishdative constructions. Linguistic Inquiry 28.281-317.SPENCER,ANDREW.1991. Morphological theory: An introduction to word structure in genera-tive grammar.Oxford: Blackwell.

341


20/20

LANGUAGE,OLUME7, NUMBER (2001)STOWELL,TIM.1983. Subjects across categories. Linguistic Review 2-3.285-312.STROMSWOLD,ARIN. 996. Analyzing children's spontaneous speech. Methods for assess-ing children's syntax, ed. by Dana McDaniel, Cecile McKee, and Helen Smith Cairns,23-53. Cambridge, MA: MIT Press.STROMSWOLD,ARIN,and WILLIAM NYDER.1995. The acquisition of datives, particles, andrelated constructions: Evidence for delayed parametric learning. Proceedings of the19th Boston University Conference on Language Development, vol. 2, ed. by DawnMacLaughlin and Susan McEwen, 621-28. Somerville, MA: Cascadilla Press.SUGISAKI,OJI, ndMIWA SOBE. 001. Resultatives resultfrom the compounding parameter:On the acquisitional correlation between resultatives and N-N compounds in Japanese.West Coast Conference on Formal Linguistics 19, to appear.TENNY, CAROL. 1994. Aspectual roles and the syntax-semantics interface. Dordrecht:Kluwer.TRAVIS, ISA. 1984. Parametersand effects of word order variation. Cambridge, MA: MITdissertation.VENDLER,ZENO. 1967. Linguistics in philosophy. Ithaca, NY: Cornell University Press.WAROTAMASIKKHADIT,DOM. 1963. Thai syntax: An outline. Austin, TX: University ofTexas dissertation.Departmentof Linguistics [Received 12 May 1999;Universityof Connecticut revision received 22 June2000;341 Mansfield Road accepted 15 August 2000]Storrs,CT 06269-1145[[email protected]]

342

Documents

On the Nature of Sytactic Variation