Upload File

mapping prosody to reference in l2 - linguistics and...

  • Home
  • Documents
  • Mapping Prosody to Reference in L2 - Linguistics and ...hrohde/presentations/SchaferTakedaRohdeGruter... · Mapping Prosody to Reference in L2 Amy J. Schafer1), Aya Takeda1), Hannah
1
Mapping Prosody to Reference in L2 Amy J. Schafer 1) , Aya Takeda 1) , Hannah Rohde 2) and Theres Grüter 1) 1) University of Hawai‘i at Mānoa, 2) University of Edinburgh Presented at BUCLD 40, Nov 13-15, 2015; contact: [email protected] or [email protected] Experiment 1: Contrast on antecedents #) Pearson (2011; www.versanDest.com) Conclusions Par=cipants Akmajian, A., & Jackendoff, R. (1970). CoreferenLality and stress. LinguisMc Inquiry. Balogh, J. E. (2003). Pronouns, prosody, and the discourse anaphora weighLng approach. (Unpublished doctoral dissertaLon). UCSD. Chen, A., & Lai, V. T. (2011). Comb or coat: The role of intonaLon in online reference resoluLon in a second language. In Sound and Sounds. Studies presented to M.E.H. (Bert) Schouten on the occasion of his 65th birthday. UiL OTS. Grüter, T., Rohde, H., & Schafer, A. J. (in press). Coreference and discourse coherence in L2: The roles of grammaLcal aspect and referenLal form. LinguisMc Approaches to Bilingualism. Grüter, T., Rohde, H., & Schafer, A. J. (2014). The role of discourse-level expectaLons in non-naLve speakers’ referenLal choices. In Proceedings of BUCLD 38. Cascadilla Press. Huang, B. H. & Jun, S.-A. (2011). The effect of age on the acquisiLon of second language prosody. Language and Speech. Kehler, A., Kertz, L., Rohde, H., & Elman, J. L. (2008). Coherence and coreference revisited. Journal of SemanMcs. Kim, K., Grüter, T., & Schafer, A. J. (2013) Effects of event-structure and topic/focus-marking on pronoun reference in Korean. CUNY Poster. Pennington, M.C. & Ellis, N.C. (2000). Cantonese speakers’ memory for English sentences with prosodic cues. The Modern Language Journal. Roberts, L., Gullberg, M., & Indefrey, P. (2008). Online pronoun resoluLon in L2 discourse: L1 influence and general learner effects. Studies in Second Language AcquisiMon. Sorace, A. (2011). Pinning down the concept of ‘interface’ in bilingualism. LinguisMc Approaches to Bilingualism. Ueno, M., & Kehler, A. (2010). The interpretaLon of null and overt pronouns in Japanese: GrammaLcal and pragmaLc factors. Proceedings of the 32nd Annual MeeMng of the CogniMve Science Society. Zubizarreta, M. L. & Nava, E. (2011). Encoding discourse-based meaning: Prosody vs. syntax. ImplicaLons for second language acquisiLon. Lingua. • L2ers’ show significant use of contrasLve intonaLon on antecedents (Exp1), but not pronouns (Exp2) to determine pronoun reference. • L2ers’ success in using prosody may depend on the number, complexity and Lming of steps required to make the relevant mappings: Exp1: L+H* on referent à referent is salient à select as antecedent Exp2: L+H* on pronoun à establish set of potenLal antecedents à idenLfy most salient member of set à select other member of set as antecedent Age (years) Versant English Test #) (overall score, range 20-80) Self-rated English proficiency (out of 10) L1 English (n=48) 23.4 (18-40) --- 9.6 (8-10) L2 English (n=40) 25.6 (19-65) 51.2 (35-80) 6.1 (3-9) L1 Japanese (n=23) 25.2 (19-45) 47.2 (35-62) 5.9 (3-8) L1 Korean (n=17) 26.2 (21-65) 56.8 (37-80) 6.5 (3-9) Background & Mo=va=on How does CONTRASTIVE INTONATION/PROSODY affect naMve and non-naMve speakers’ interpretaMon of ambiguous pronouns? • ContrasLve intonaLon/prosody affects naLve-speaker processing, although its precise role in the interpretaLon of pronouns has received limited aDenLon. (1) …noLced a paramedic/PARAMEDIC calling the CAPTAIN/captain… Later Wendy bumped into him. (Balogh, 2003) (2) John hit Bill and then George hit HIM. (Akmajian & Jackendoff, 1970) • Prosody presents challenges in L2 acquisiLon, which vary with factors such as the L1-to-L2 mapping and prosodic funcLon. (Huang & Jun, 2011; Zubizarreta & Nava, 2011; Takeda, Schafer, & Schwartz, BUCLD40; Pennington & Ellis, 2000; Chen & Lai, 2011) • Pronoun interpretaLon is challenging in L2. This has been aDributed to the need to integrate informaLon from mulLple domains. (e.g., Roberts et al., 2008; Sorace, 2011) Previous work has shown a reduced effect of event structure, marked by grammaLcal aspect, on L2 speakers’ reference choices. (see below) The role of prosody in L2 pronoun resolu2on has not been inves2gated. Our goals: • Test how contrasLve intonaLon on potenLal antecedents (Exp1) and pronouns (Exp2) affects L1 and L2 speakers’ reference choices. • Compare effects of contrasLve intonaLon and grammaLcal aspect. Predic=ons: à IF prosody is generally challenging in L2, we expect L1-L2 differences in both Exp1 and Exp2. (NB: same L+H* L-H% contour in both experiments) à IF integraLng informaLon from mulLple domains is generally challenging in L2, we expect L1-L2 differences in both Exp1 and Exp2. à IF L2 speakers’ ability to use prosodic informaLon depends on the complexity of the mappings (L1 -> L2; prosody -> reference), we expect greater L1-L2 differences in Exp2 compared to Exp1. Experiment 2: Contrast on pronoun Method • Story conLnuaLon: aural context sentence + wriDen prompt (pronoun) + wriDen compleLon • 2 (contrast locaLon) x 2 (aspect) design Hear: DAVID served/was serving Paul a pint of beer. See: He _______________ Hear: David served/was serving PAUL a pint of beer. See: He _______________ Results Mixed-effect logis=c regression (LMER) isSource ~ Aspect * ContrLoc * Group + (1 + Aspect + ContrLoc | Subject) + (1 + Aspect + ContrLoc | Item) • Main effect of Contrast LocaLon (b = .89, p < .001), no interacLon with Group (b = .14, p = .63) • Main effect of Aspect (b = .42, p < .005), marginal Aspect x Group interacLon (b = .40, p = .09); effect of Aspect in L1 (b= .60, p < .001), but not in L2 (b= .25, p = .26). à More reference to accentuated antecedents in both groups: ContrasLve intonaLon on antecedents affects reference choices in L1 and L2. à Reduced effect of aspect in L2; replicates results from wriDen task. • AnnotaLon for coreference by two trained coders. He wanted Paul to get really drunk. (SOURCE) He thanked David. (GOAL) He insisted it was the best beer ever. (ambiguous: 6.8/8.8% of L1/L2 data) (missing: 0.7/2.3% of L1/L2 data) Par=cipants Age (years) Versant English Test #) (overall score, range 20-80) Self-rated English proficiency (out of 10) L1 English (n=48) 24.2 (18-49) --- 9.6 (6-10) L2 English (n=42) 31.1 (20-56) 51.0 (37-72) 5.9 (1-9) L1 Japanese (n=24) 35.5 (20-56) 50.0 (37-69) 5.5 (1-8) L1 Korean (n=18) 25.3 (20-48) 52.6 (40-72) 6.4 (3-9) Method • Story conLnuaLon: aural context sentence + 750ms silence + aural prompt (pronoun + adverb, to provide natural truncaLon point) + wriDen compleLon • 2 (pronoun stress) x 2 (aspect) design Hear: David served/was serving Paul a pint of beer. He/HE obviously See: ________ • AnnotaLon for coreference by two trained coders. (ambiguous: 7.3/8.9% of L1/L2 data; missing: 0.3/3.0% of L1/L2 data) Mixed-effect logis=c regression (LMER) isSource ~ Aspect * ProStress * Group + (1 + Aspect + ProStress | Subject) + (1 + Aspect + ProStress | Item) • Main effect of Pronoun Stress (b = -.49, p < .001), Pronoun Stress x Group interacLon (b = .72, p < .01); no effect of Pronoun Stress in L2 group alone (b = .10, p = .69) • Main effect of Aspect (b = .60, p < .001), no interacLon with Group (b = .27, p = .28); effect of Aspect in L2 group alone (b = .81, p < .001); Aspect x Pronoun Stress interacLon in L2 group (b = .81, p < .05) • Main effect of Group (b = 1.14, p < .001) à More reference to the Goal with stressed pronouns in L1 only. à Effect of Aspect emerges in L2 when conLnuaLon point is moved axer the adverb. Figure 2. Exp1: Mean proporLon of SOURCE-reference in conLnuaLons; 95% CIs Acknowledgments: Many thanks to Amanda Blake, Amber Camp, Bonnie Fox, Victoria Lee, Ivana Matson, Eric Stepans, and Alexis Toliva for assistance with data collecLon, annotaLon, and preparaLon of this poster. This research is supported by the NaLonal Science FoundaLon (BCS-1251450 to T. Grüter & A. Schafer). Event structure and pronoun interpreta=on in L1 & L2 (3) Emily Source brought/was bringing a drink to Melissa Goal . She _____ L1 speakers of English write more conLnuaLons with ‘she’ = SOURCE following imperfecLve vs perfecLve aspect (Kehler et al., 2008). This is not the case for L1- Japanese/Korean learners of English (Grüter et al., 2014, in press), even though (i) they reliably associate perfecLve/imperfecLve with completed/incomplete events in an independent task, (ii) aspect affects their choice of coherence relaLon in the conLnuaLon, and (iii) Japanese and Korean speakers show the same effect of aspect in their L1s (Ueno & Kehler, 2010; Kim et al., 2013). Figure 1. Context sentence with contrast on Source (A) and on Goal (B). A B A B1 B2 Figure 3. Broad-focus context (A) and conLnuaLon with unstressed (B1) and stressed (B2) pronoun. hDp://www2.hawaii.edu/ ~aschafer/snds.html#GRS Figure 4. Exp2: Mean proporLon of SOURCE-reference in conLnuaLons; 95% CIs

Upload: truongnguyet

Post on 01-Apr-2018

214 views

Category:

Documents


2 download

Report

TRANSCRIPT

Page 1: Mapping Prosody to Reference in L2 - Linguistics and ...hrohde/presentations/SchaferTakedaRohdeGruter... · Mapping Prosody to Reference in L2 Amy J. Schafer1), Aya Takeda1), Hannah

MappingProsodytoReferenceinL2

AmyJ.Schafer1),AyaTakeda1),HannahRohde2)andTheresGrüter1)1)UniversityofHawai‘iatMānoa,2)UniversityofEdinburgh

PresentedatBUCLD40,Nov13-15,2015;contact:[email protected]@hawaii.edu

Experiment1:Contrastonantecedents

#)Pearson(2011;www.versanDest.com)

Conclusions

Par=cipants

Akmajian,A.,&Jackendoff,R.(1970).CoreferenLalityandstress.LinguisMcInquiry.Balogh,J.E.(2003).Pronouns,prosody,andthediscourseanaphoraweighLngapproach.(UnpublisheddoctoraldissertaLon).UCSD.Chen,A.,&Lai,V.T.(2011).Comborcoat:TheroleofintonaLoninonlinereferenceresoluLoninasecondlanguage.InSoundandSounds.StudiespresentedtoM.E.H.(Bert)Schoutenontheoccasionofhis65thbirthday.UiLOTS.

Grüter,T.,Rohde,H.,&Schafer,A.J.(inpress).CoreferenceanddiscoursecoherenceinL2:TherolesofgrammaLcalaspectandreferenLalform.LinguisMcApproachestoBilingualism.

Grüter,T.,Rohde,H.,&Schafer,A.J.(2014).Theroleofdiscourse-levelexpectaLonsinnon-naLvespeakers’referenLalchoices.InProceedingsofBUCLD38.CascadillaPress.

Huang,B.H.&Jun,S.-A.(2011).TheeffectofageontheacquisiLonofsecondlanguageprosody.LanguageandSpeech.Kehler,A.,Kertz,L.,Rohde,H.,&Elman,J.L.(2008).Coherenceandcoreferencerevisited.JournalofSemanMcs.Kim,K.,Grüter,T.,&Schafer,A.J.(2013)Effectsofevent-structureandtopic/focus-markingonpronounreferenceinKorean.CUNYPoster.Pennington,M.C.&Ellis,N.C.(2000).Cantonesespeakers’memoryforEnglishsentenceswithprosodiccues.TheModernLanguageJournal.Roberts,L.,Gullberg,M.,&Indefrey,P.(2008).OnlinepronounresoluLoninL2discourse:L1influenceandgenerallearnereffects.StudiesinSecondLanguageAcquisiMon.

Sorace,A.(2011).Pinningdowntheconceptof‘interface’inbilingualism.LinguisMcApproachestoBilingualism.Ueno,M.,&Kehler,A.(2010).TheinterpretaLonofnullandovertpronounsinJapanese:GrammaLcalandpragmaLcfactors.Proceedingsofthe32ndAnnualMeeMngoftheCogniMveScienceSociety.

Zubizarreta,M.L.&Nava,E.(2011).Encodingdiscourse-basedmeaning:Prosodyvs.syntax.ImplicaLonsforsecondlanguageacquisiLon.Lingua.

•  L2ers’showsignificantuseofcontrasLveintonaLononantecedents(Exp1),butnotpronouns(Exp2)todeterminepronounreference.

•  L2ers’successinusingprosodymaydependonthenumber,complexityandLmingofstepsrequiredtomaketherelevantmappings:Exp1:L+H*onreferentàreferentissalientàselectasantecedent

Exp2:L+H*onpronounàestablishsetofpotenLalantecedentsàidenLfymostsalientmemberofsetàselectothermemberofsetasantecedent

Age(years)

VersantEnglishTest#)(overallscore,range20-80)

Self-ratedEnglishproficiency(outof10)

L1English(n=48) 23.4(18-40) --- 9.6(8-10)L2English(n=40) 25.6(19-65) 51.2(35-80) 6.1(3-9)L1Japanese(n=23) 25.2(19-45) 47.2(35-62) 5.9(3-8)L1Korean(n=17) 26.2(21-65) 56.8(37-80) 6.5(3-9)

Background&Mo=va=on

HowdoesCONTRASTIVEINTONATION/PROSODYaffectnaMveandnon-naMvespeakers’interpretaMonofambiguouspronouns?

•  ContrasLveintonaLon/prosodyaffectsnaLve-speakerprocessing,althoughitspreciseroleintheinterpretaLonofpronounshasreceivedlimitedaDenLon.

(1)…noLcedaparamedic/PARAMEDICcallingtheCAPTAIN/captain…LaterWendybumpedintohim.(Balogh,2003) (2)JohnhitBillandthenGeorgehitHIM.(Akmajian&Jackendoff,1970)

•  ProsodypresentschallengesinL2acquisiLon,whichvarywithfactorssuchastheL1-to-L2mappingandprosodicfuncLon.

(Huang&Jun,2011;Zubizarreta&Nava,2011;Takeda,Schafer,&Schwartz,BUCLD40;Pennington&Ellis,2000;Chen&Lai,2011)

•  PronouninterpretaLonischallenginginL2. ThishasbeenaDributedtotheneedtointegrateinformaLonfrommulLpledomains.(e.g.,Robertsetal.,2008;Sorace,2011)Previousworkhasshownareducedeffectofeventstructure,markedbygrammaLcalaspect,onL2speakers’referencechoices.(seebelow)

TheroleofprosodyinL2pronounresolu2onhasnotbeeninves2gated.

Ourgoals:

•  TesthowcontrasLveintonaLononpotenLalantecedents(Exp1)andpronouns(Exp2)affectsL1andL2speakers’referencechoices.

•  CompareeffectsofcontrasLveintonaLonandgrammaLcalaspect.

Predic=ons:

à  IFprosodyisgenerallychallenginginL2,weexpectL1-L2differencesinbothExp1andExp2.(NB:sameL+H*L-H%contourinbothexperiments)

à  IFintegraLnginformaLonfrommulLpledomainsisgenerallychallenginginL2,weexpectL1-L2differencesinbothExp1andExp2.

à  IFL2speakers’abilitytouseprosodicinformaLondependsonthecomplexityofthemappings(L1->L2;prosody->reference),weexpectgreaterL1-L2differencesinExp2comparedtoExp1.

Experiment2:Contrastonpronoun

Method•  StoryconLnuaLon:auralcontextsentence+wriDenprompt(pronoun)

+wriDencompleLon•  2(contrastlocaLon)x2(aspect)design

Hear:DAVIDserved/wasservingPaulapintofbeer.See:He_______________Hear:Davidserved/wasservingPAULapintofbeer.See:He_______________

Results

Mixed-effectlogis=cregression(LMER)isSource ~ Aspect * ContrLoc * Group + (1 + Aspect + ContrLoc | Subject) + (1 + Aspect + ContrLoc | Item)

•  MaineffectofContrastLocaLon(b=.89,p<.001),nointeracLonwithGroup(b=.14,p=.63)•  MaineffectofAspect(b=.42,p<.005),marginalAspectxGroupinteracLon(b=.40,p=.09);

effectofAspectinL1(b=.60,p<.001),butnotinL2(b=.25,p=.26).

à Morereferencetoaccentuatedantecedentsinbothgroups:ContrasLveintonaLononantecedentsaffectsreferencechoicesinL1andL2.

à ReducedeffectofaspectinL2;replicatesresultsfromwriDentask.

•  AnnotaLonforcoreferencebytwotrainedcoders.He wanted Paul to get really drunk. (SOURCE)He thanked David. (GOAL)He insisted it was the best beer ever. (ambiguous:6.8/8.8%ofL1/L2data)

(missing:0.7/2.3%ofL1/L2data)

Par=cipantsAge

(years)VersantEnglishTest#)

(overallscore,range20-80)Self-ratedEnglish

proficiency(outof10)L1English(n=48) 24.2(18-49) --- 9.6(6-10)L2English(n=42) 31.1(20-56) 51.0(37-72) 5.9(1-9)L1Japanese(n=24) 35.5(20-56) 50.0(37-69) 5.5(1-8)L1Korean(n=18) 25.3(20-48) 52.6(40-72) 6.4(3-9)

Method•  StoryconLnuaLon:auralcontextsentence+750mssilence+auralprompt

(pronoun+adverb,toprovidenaturaltruncaLonpoint)+wriDencompleLon•  2(pronounstress)x2(aspect)design

Hear:Davidserved/wasservingPaulapintofbeer.He/HEobviouslySee:________

•  AnnotaLonforcoreferencebytwotrainedcoders.(ambiguous:7.3/8.9%ofL1/L2data;missing:0.3/3.0%ofL1/L2data)

Mixed-effectlogis=cregression(LMER)isSource ~ Aspect * ProStress * Group + (1 + Aspect + ProStress | Subject) + (1 + Aspect + ProStress | Item)

•  MaineffectofPronounStress(b=-.49,p<.001),PronounStressxGroupinteracLon(b=.72,p<.01);noeffectofPronounStressinL2groupalone(b=.10,p=.69)

•  MaineffectofAspect(b=.60,p<.001),nointeracLonwithGroup(b=.27,p=.28);effectofAspectinL2groupalone(b=.81,p<.001);AspectxPronounStressinteracLoninL2group(b=.81,p<.05)

•  MaineffectofGroup(b=1.14,p<.001)

à  MorereferencetotheGoalwithstressedpronounsinL1only.à  EffectofAspectemergesinL2whenconLnuaLonpointismoved

axertheadverb.

Figure2.Exp1:MeanproporLonofSOURCE-referenceinconLnuaLons;95%CIs

Acknowledgments:ManythankstoAmandaBlake,AmberCamp,BonnieFox,VictoriaLee, IvanaMatson,EricStepans, and Alexis Toliva for assistance with data collecLon, annotaLon, and preparaLon of this poster.ThisresearchissupportedbytheNaLonalScienceFoundaLon(BCS-1251450toT.Grüter&A.Schafer).

Eventstructureandpronouninterpreta=oninL1&L2

(3)EmilySourcebrought/wasbringingadrinktoMelissaGoal.She_____

L1speakersofEnglishwritemoreconLnuaLonswith‘she’=SOURCEfollowingimperfecLvevsperfecLveaspect(Kehleretal.,2008).ThisisnotthecaseforL1-Japanese/KoreanlearnersofEnglish(Grüteretal.,2014,inpress),eventhough(i)theyreliablyassociateperfecLve/imperfecLvewithcompleted/incompleteeventsinanindependenttask,(ii)aspectaffectstheirchoiceofcoherencerelaLonintheconLnuaLon,and(iii)JapaneseandKoreanspeakersshowthesameeffectofaspectintheirL1s(Ueno&Kehler,2010;Kimetal.,2013).

Figure1.ContextsentencewithcontrastonSource(A)andonGoal(B).

A BA B1 B2

Figure3.Broad-focuscontext(A)andconLnuaLonwithunstressed(B1)andstressed(B2)pronoun.hDp://www2.hawaii.edu/~aschafer/snds.html#GRS

Figure4.Exp2:MeanproporLonofSOURCE-referenceinconLnuaLons;95%CIs

Prosody 2014

L2 Alignments - PaVE: Papilloma virus genome database · II-L2-1 OCT 95 L2 Alignments L2 Alignments L2 Alignments L2 Alignments L2 Alignments L2 Alignments L2 Alignments L2 Alignments

Pali Prosody

Music and Speech Prosody

Prosody & Information Structure

Prosody Introduction

Automatic Feedback for L2 Prosody Learning - core.ac.uk · Automatic Feedback for L2 Prosody Learning Anne Bonneau, ... was one of the first scientists to use melodic curve visualisation

PROSODY - School of Englishseas3.elte.hu/analysis/analysis7.pdf · English Phonological Analysis – Chapter 7 – PROSODY (Autumn 2013) 2 7.1. WHAT IS PROSODY? Prosody in linguistics

Terms for Discussing Prosody

Prosody and Fluent Reading

Prosody Modeling

Mapping novel L2 words into the mental lexicon: Evidence ...ˆneburg_talk_Greg.pdfstrengthen L2 word to L1 word form connections) and more stress on the integration of L2 words into

Prosody 2012

Pitch Tracking + Prosody

05. Concept mapping decision matrix for L2 teaching

Terms for Discussing Prosody. Basic Terms Prosody: The metrical pronunciation of a song or poem. Prosody: The metrical pronunciation of a song or poem

Suprasegmental features and Prosody

Prosody Training

Prosody X cPCI

The interface between syntax, prosody, and information structure: What we can learn from L2 speech Maria Luisa Zubizarreta University of Southern California

Finnish OT Prosody

Prosody s telephony software

The Effect of Concept Mapping on Iranian Pre- intermediate L2

Preschoolers use phrasal prosody online to constrain ... use phrasal prosody online to ... tional phrase ‘with the ... Preschoolers use phrasal prosody online to constrain syntactic

Sanskrit Prosody

The Use of Prosody in Syntactic Disambiguationwjh.harvard.edu/~pal/pdfs/prosody/price91.pdf · The use of prosody in syntactic disambiguation P.J. Price, M. Ostendorf, S. Shattuck-Hufnagel,

Prosody in Spoken Language Understanding

Terms for Discussing Prosody. What is prosody? The term “prosody” refers to discussions of the kinds of stressed and unstressed syllables in poems. The

Psycholinguistics Cannot Escape Prosody

Elements of Greek Prosody

Milton's Prosody (1901)

High-Quality Prosody Generation in Mandarin Text-to … · High-Quality Prosody Generation in Mandarin Text-to ... binary prosodic tree method in terms of ... Quality Prosody Generation

Functional Timing of Prosody

Prosody and Suprasegmentals

Prosody and NLP