Embed Size (px)
Citation preview
7.Discussion&Conclusion
5.Exp 1:English.LinearWordOrdervsSubjecthood
• Firstlookatreal-timeproductionofquestions• Planningisstructurallyincremental• Speakersstartwithsyntacticrolesevenwhenitconflictswithlinearwordorder
• Noevidencecovertdependenciesformulatedinthesamewayasovertdependencies
• Noevidenceinformationfocusaffectseye-movementsduringmessageformulation;Exp 1resultsnotconfoundedbyfocus• Inlinewithpriorworkshowinglateemergenceofdiscourse-pragmaticeffectsinproduction[8]
2.CurrentStudy• ResearchQuestion:Howdolinearwordorderandsubjecthood interacttoinformthestartingpointofmessageformulation?
• Englishobjectwh-questionscanteaseapartlinearity&subjecthood
didthenursestickle?
3.Hypotheses&Predictions
REFERENCES:[1] Levelt,1989;BockandLevelt,1994[2] Gleitman etal.,2007;Brown-Schmidt&Konopka,2008[3] Griffin &Bock,2000[4] Kuchinsky,2011;Konopka ,2012[5] Griffin &Bock,2000[6]Myachykov etal.,2011[7] Norcliffe etal.,2015;Sauppe etal.,2013[8] Ganuschak etal.,2014,2017ACKNOWLEDGEMENTS:[1] StudentOpportunitiesforAcademicResearch(USC),PatriceZhao[2] RussellEndowedFellowship(USC)[3] CEDLTravelGrant(LSA)
Whatcanwh-questionstellusaboutreal-timelanguageproduction?EvidencefromEnglishandMandarinMonicaDo {[email protected]}😺 Elsi Kaiser {[email protected]}😺 Pengchen Zhao {[email protected]}DepartmentofLinguistics,UniversityofSouthernCalifornia,LosAngeles,USALSA,SaltLakeCity,Utah,January4-7,2018
4.Experiment Design
• Languageproduction,likecomprehension,isincremental[1]• Whendescribinganimage,speakers:(1)ApprehendSceneà (2)FormulateMessageà (3)GrammaticallyAssembleMessage/SelectLexicalItemsà(4)PhonologicallyEncodeMessageà (5)BeginArticulation
• Whatfactorsdeterminewherewestartincrementallyformulatingmessages?• LinearAccounts: Startwithmostaccessiblelexicalconcept;mentionthatfirst(e.g.asthesubjectinEnglish).[2]
• StructuralAccounts:Startwithsubjectofthesentence;insertrelevantlexicalconceptintothe‘subjectslot’[3]
• Multi-factorialAccounts:Productionvariesduetoaccessibilityandstructure[4]
• Howdoweteaseaparttheseaccountsifsubjectsareoftenthefirstargumentsinasentence?• ActivevsPassives:Grammatical(notthematic)rolesdrivemessageformulation,butstillsubject-initial[5]
• Freewordorder:Russian,Finnish[6]||Verb-initial:Tzeltal,Tagalog[7]• But,resultscomplicatedbydiscourseand/ormorphologicalfactors
1.Introduction
• Participantsfirst sawsentencetypecue,then sawimage;producedthecuedsentencetype
Statement(S) ObjectWh-Question(Q)
• Verbsindicatedbyinstruments(e.g.feather),instrumentlocationindicatedsubjectcharacter
• 33targets;30fillers.Familiarizationsessionbeforeexperiment• MeasuredProportionoffixationstosubject,objectand verb,&Sub-Obj DifferenceScores
DeclarativesThenursestickledthechefs.
ObjectWh-QuestionsWhichchefsdidthenurses tickle?
LinearAccount:Linearlyfirstword Subject ObjectStructuralAccount:
Subject Subject SubjectMulti-Factorial:Bothinteract Subject ?????
6.Exp 2:Mandarin.LinearWordOrdervsInformationFocus
• Speakers(n=30)looktoverb firsttodetermineSubj/Obj• Differencesbetweendecl &quesemerge~400ms• Differencesbecomesignificant~600ms• Subj-Obj differencescoresindeclarativeslarger thaninobjectwh-questions(|z|=2.67)
Decla
rativ
esQ
uest
ions
0 200 400 600 800 1000
0.00.10.20.30.40.50.60.70.8
0.00.10.20.30.40.50.60.70.8
Time(ms) from Image Onset
Prop
ortio
n of
Fixa
tions
Subj
Obj
Shaded areas indicate 95% confidence interval
Decla
rativ
esQ
uest
ions
0 200 400 600 800 1000
0.00.10.20.30.40.50.60.70.8
0.00.10.20.30.40.50.60.70.8
Time(ms) from Image Onset
Prop
ortio
n of
Fixa
tions
Region Subj Obj Verb
• Speakerslooktoverb firsttodetermineSubj/Obj characters• Decl andquesdonotdiffer200-1000msafterimageonset• Subj-Obj differencescoresdonotdiffer(|z|s <1.4)• KeyPattern:Speakersfixatesubjectinbothdecl &ques;donotconsiderobject ineither
• ResearchQuestions:(1)TowhatextentdidinformationfocusdrivecompetitivelookstotheobjectinExp 1(English)?(2) Isplanningdifferentforovertvscovertdependencies?
• MandarinChinese(Subject-Verb-Object)• Wh-questionsanddeclarativeshavethesamelinearwordorderDeclarative:护士们 枪毙了 厨师。Thenurses shotthechefs.ObjectWh-Question:护士们 枪毙了 哪个厨师?Thenurses shotwhichchefs?
• Eye-movementsdifferencescannot beduetosurfacewordorder
• NativeMandarinspeakers(n=35)• Exp 2conductedinMandarin;itemsdifferedtoaccountforlexicaldifferences
• KeyPattern:Speakerslooktothesubject beforeobject indecl &ques(~400ms),butconsidertheobject moreinquesthanindecl
• Messageformulationmodulatedbysyntacticstructure• Decl: Rapidriseinlookstosubjectonly~400ms• Ques: Riseinlookstosubject&object~400ms
• Howdolinearwordorderandsubjecthoodinteract?• Theyareseparable,competitiveeffects• Subjecthood isprivileged overlinearwordorderduringmessageformulation• But,linearwordorderisnotruledout:Itcompeteswithsubjecthood
• But,LinearWordOrderorInformationFocus?• wh-wordsareinformationallyfocused elements• PossibleAlternativeAccount:Informationfocusdrovecompetitivelookstoobjectwh-phraseinEnglishquestions.
Fig2:Eye-MovementsAfter MessageFormulation
• After windowofinterest,fixationpatternsreflectlinearwordorder,asexpected• Tightgaze-to-speechcoordination:Speakerslooktotheto-be-mentionedimagebeforenamingit• Decl: Subjplannedbeforespeechonset• Ques: Obj plannedbeforespeechonset
Decla
rativ
esQ
uest
ions
−1600−1400−1200−1000 −800 −600 −400 −200 0 200 400 600 800 1000
0.10.20.30.40.50.60.70.8
0.10.20.30.40.50.60.70.8
Time(ms) Aligned To Speech Onset
Prop
ortio
n of
Fixa
tions
Dosp
eakersgen
erallybeh
avelikeweexpe
ctth
emto
?
Shaded areas indicate 95% confidence interval
Decla
rativ
esQ
uest
ions
0 200 400 600 800 1000
0.00.10.20.30.40.50.60.70.8
0.00.10.20.30.40.50.60.70.8
Time(ms) from Image Onset
Prop
ortio
n of
Fixa
tions
Region Subj Obj Verb
Decla
rativ
esQ
uest
ions
0 200 400 600 800 1000
0.00.10.20.30.40.50.60.70.8
0.00.10.20.30.40.50.60.70.8
Time(ms) from Image Onset
Prop
ortio
n of
Fixa
tions De
clara
tives
Que
stio
ns
−1600−1400−1200−1000 −800 −600 −400 −200 0 200 400 600 800 1000
0.00.10.20.30.40.50.60.70.8
0.00.10.20.30.40.50.60.70.8
Time(ms) Aligned to Speech Onset
Prop
ortio
n of
Loo
ks
Fig4:Eye-MovementsAfter MessageFormulation
• After windowofinterest,fixationpatternsreflectlinearwordorder,asexpected• Tightgaze-to-speechcoordination:Speakerslooktotheto-be-mentionedimagebeforenamingit• Decl &Quesshowsamepattern: Subjplannedbeforespeechonset
MessageFormulationisMulti-Factorial
Agent Linearwordorder
Subjecthood isprivileged
Someotherfactorscanstillplayarole
Butnotallinthesameway
Subject
Informationfocus?
Fig1:Eye-MovementsImmediatelyAfter ImageOnset
Fig3:Eye-MovementsImmediatelyAfterImageOnset
Subj
Obj
Dosp
eakersgen
erallybeh
avelikeweexpe
ctth
emto
?
