• Aucun résultat trouvé

Disentangling sources of difficulty associated with the acquisition of accusative clitics in French

N/A
N/A
Protected

Academic year: 2022

Partager "Disentangling sources of difficulty associated with the acquisition of accusative clitics in French"

Copied!
25
0
0

Texte intégral

(1)

Article

Reference

Disentangling sources of difficulty associated with the acquisition of accusative clitics in French

DELAGE, Hélène, DURRLEMANN, Stéphanie, FRAUENFELDER, Ulrich Hans

DELAGE, Hélène, DURRLEMANN, Stéphanie, FRAUENFELDER, Ulrich Hans. Disentangling sources of difficulty associated with the acquisition of accusative clitics in French. Lingua , 2016, vol. 180, p. 1-24

DOI : 10.1016/j.lingua.2016.03.005

Available at:

http://archive-ouverte.unige.ch/unige:149712

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

Disentangling sources of difficulty associated with the acquisition of accusative clitics in French

Hélène Delage * , Stephanie Durrleman, Ulrich H. Frauenfelder

FacultyofPsychologyandEducationalSciences,UniversityofGeneva,Geneva,Switzerland Received4March2015;receivedinrevisedform31March2016;accepted31March2016

Availableonline4May2016

Abstract

AccusativecliticpronounsareacquiredlaterthanbothnominativeandreflexivepronounsintypicallydevelopingFrench-speaking children.Howeverrecentresearchsuggeststhatnotallthesecliticsareequallydifficult,3rdpersonaccusativeclitics(ACC3)beingmore problematicthan1stand2ndperson.Inthisstudy,weexploredthreepropertieswhichcouldmakeACC3complex:(1)gendermarking, (2)discourse-participantindependentreference,and(3)optionalityinspokenFrench.Wehavecreatedspecificexperimentsassessing theroleofeachofthesepropertiesincliticacquisitionin41French-speakingchildrenaged4--8.Resultsshowthatallthreeproperties playaroleinmaking ACC3complex, withthestrongestinfluencecomingfromgendermarking. Implicationsoftheseresultsare discussedinlightofpreviousrelatedstudies.

©2016ElsevierB.V.Allrightsreserved.

Keywords: Accusativeclitics;Acquisition;French

1. Introduction

Accusativecliticpronouns(ACC)havebeenwidelyinvestigatedinstudiesoftheacquisitionofRomancelanguages.

Theyhaveattractedparticularinterestbecauseoftheirlateacquisitionascomparedtobothnominativeandreflexive pronounsintypicallydevelopingFrench-speakingchildren(Delage,2008;Hamannetal.,1996;Jakubowiczetal.,1998;

JakubowiczandRigaut,2000;VanderVelde,2003;Zesigeretal.,2010).Theircomplexityhaspositionedthemasa clinical marker ofspecificlanguageimpairment or‘SLI’(Paradis et al.,2003; ParisseandMaillart, 2004) andmore generallyofatypicallanguagedevelopment(Delage,2008;Tulleretal.,2011).Morerecently,ithasbeensuggestedfor Greek,RomanianandCatalanthatnotallACCareequallydifficult(Avrametal.,2015;CoeneandAvram,2011;Gavarro´

andForto´n,2014;TsimpliandMastropavlou,2007).Insteadwhatappearstobeproblematicisspecifically3rdperson clitics,while1stand2ndpersonACCprovetoberelativelysimpler.Thisasymmetryhasnowalsobeenconfirmedfor Frenchbyarecentstudy(Tulleretal.,2011).Theaimofthepresentworkistofurtherexplorethe3rdversus1stperson distinction for ACC in French to elucidate why 3rd person clitics are associated with delayed development. More specifically,whileallACCsharecertainproperties(listedina--cinTable1),thoseofthe3rdpersonpresentadditional www.elsevier.com/locate/lingua Availableonlineatwww.sciencedirect.com

ScienceDirect

Lingua180(2016)1--24

*Correspondingauthorat:FacultyofPsychologyandEducationalSciences,UniversityofGeneva,40boulevardduPontdArve,1211Geneve 4,Switzerland.Tel.:+41223799170.

E-mailaddresses:Helene.Delage@unige.ch(H.Delage),Stephanie.Durrleman@unige.ch(S.Durrleman),Ulrich.Frauenfelder@unige.ch (U.H.Frauenfelder).

http://dx.doi.org/10.1016/j.lingua.2016.03.005 0024-3841/©2016ElsevierB.V.Allrightsreserved.

(3)

characteristicswhichwouldrenderthemparticularlydifficult(listedin d--fin Table1) andwhicharethefocusofthis research.

Thisstudyconsiderstheimpactofeachofthesepropertiesspecificto3rdpersonACC,namely(i)themorphological markingfor gender,(ii)the discourse-participant-independent reference(i.e.,wherereferenceisnot tiedtospeaker- addresseeroles)and(iii)thelegitimateomissionor‘optionality’inspecificcontextsofspokenFrench.Theseproperties havebeenhypothesizedtoplayaroleindelayingtheacquisitionof3rdpersonACCinpreviousresearch,whichwereview below.

1.1. 3rdpersonACC

ResearchontheacquisitionofFrenchACC(1)hasshownthattheseemergemoreslowlythanboth(2)reflexiveand (3)nominativecliticsintypicallanguagedevelopment:

(1) Jean le voit John him sees

‘Johnseeshim’

(2) Jean se voit John himself sees

‘Johnseeshimself’ (3) Il voit Jean

He sees John

‘HeseesJohn’

Hamannetal.(1996)analyzedthespontaneouslanguageofaFrench-speakingchildfromage2to2;10andshowed thatwhilethenominativecliticwaspresentfromthefirstrecording(i.e.,atage2;0)andwasconsistentlyusedinmorethan halfofthechildproductions,ACCwerevirtuallyabsentuntilage2;6.Moreoveratage2;9,theaccusativewasfoundonlyin 14%ofthecontextswhereitshouldhavebeenobligatorilyrealized,1contrarytonominativecliticswhichoccurredin63%

ofsuchcontexts.AsimilardelaywasobservedbyJakubowiczetal.(1998)inanelicitedproductionstudyof20French childrenaged5;6to5;11.Inaddition,theseauthorsnotedadelaybetweentheaccusativeascomparedtothereflexive clitic(bothencoding3rdperson).Thusahierarchyemerges,withthenominativebeingacquiredbeforethereflexive, whichinturnisacquiredbeforetheaccusative.JakubowiczandRigaut(2000)confirmedthishierarchywithananalysisof naturallanguagecorporaaswellaswithanelicitedproductiontaskconductedwith12youngchildrenaged2to2;7.

Subsequently,thehierarchywasreplicatedin astudy of99childrenaged3;5to 6;5tested specificallywithelicited production(Zesigeretal.,2010).Thisstudyclearlyillustratedthedevelopmentaltrendoftheseelements,namelythat whilenominativeswereatceilingatage3,reflexivesapproachedthislevelatage4(85%)whileACConlydidsoatage6 (90%).

Insum,ACCareproducedsubstantiallylessthanothercliticsinearlychildhood,yieldingeitherfrequentomissionsof theobjectaltogether,ortheuseoffulllexicalDPs(DeterminerPhrases)instead.Elicitedproductionexperimentsforclitics inFrenchshowthatfrequentomissionsareindeedcoupledwithasubstantialuseoflexicalDPs:Atmeanage3;2,children omittedACC55.9%ofthetimeandproducedlexicalDPsinsteadin20.5%oftheirutterances(VanderVelde,2003).This Table1

Propertiesofaccusativeclitics(adaptedfromTulleretal.,2011).

AllACC Specifically3rdpersonACC

(a)Appearanceinanon-canonicalposition (d)Morphologicalmarkingforgenderandnumber(whileotherACCagreefornumber only)

(b)Co-occurrencewithanominativeclitic (e)Discourse-participantindependencefortheirreference(andthereforeunmarkedfor animacy)

(c)Non-localbinding (f)OptionalityinspokenFrenchunderdiscourse/lexicalconditions

1Meaningthatthesecontextsdidnotmeetspecificrestrictions(asfurtherdescribed)whichcouldleadtoACC3optionalityinspokenFrench.

(4)

ratetaperedoffwithtime,fallingatage4;2(21.5%omissionand9.3%lexicalDPs)anddroppingevenmoresharplyby age6;7(8.3%omissionand7.3%lexicalDPs).

Whenchildrendoattempttoproducetheclitic,ithasbeenshownthattheyoftenmakemistakesingender.Zesiger etal.(2010)reportedthatchildrenaged3;5to6;5frequentlymadegendererrorsonobjectclitics.Themostfrequent patternwasthefemininelareplacedbythemasculinele(26.8%),thereversebeingproducedsignificantlylessoften (10.4%).Tulleretal.(2011)alsonotedthat,inanelicitedproductiontask,17%ofaccusativecliticsproducedbychildrenat age6containedanerrorofgender.Mosterrorsoccurredwhenthecliticsequences(nominative+accusative)involveda mismatchingender.Inthesecontexts,theauthorsproposethatthechild’sworkingmemoryresourcesareparticularly taxed.

Interestingly, while the global comprehension of accusativecliticsappears to fare better than their production (Grüter,2005,2006;TullerandJakubowicz,2004;VanderVelde,2003;Zesigeretal.,2010),difficultywiththegender featurereportedforproductionneverthelesscarriesovertocomprehension.Comprehensionofcliticswasassessed byZesigeretal.(2010)viaatruth-valuejudgmenttaskinvolvingapuppetwhoproducedasentencewhendescribinga picture.Thissentencecontainedanobjectcliticwhicheithercorrespondedtothesexofthereferentinthepictureor not.Inthisforcedtwo-choicetask,childrenatage5showedchance-levelperformancefordetectingwhenthegender oftheaccusativecliticintheteststimulimismatchedthatofthereferentinthepicture.Participantsonlyreached75%

error detection at age 6. Once again, a difference between masculine and feminine ACC surfaced, with worse performancefordetecting agendererror withfeminine ascomparedto masculineclitics.Grüter etal. (2012)also reportedthat4year-oldSpanish-speakingchildrenencounterdifficultiesinmakinguseofthegendercueonobject clitics when identifying their referent in an eye-tracking study. Pirvulescu and Strik (2014) further tested the comprehensionofbothgenderandnumberoncliticandstrongpronounsinchildrenaged3to5yearsold,bymeansof apicturechoicetask.Theirresultsrevealedthatchildrenhaddifficultieswiththesefeaturalcueswhendetermining antecedentsforbothcliticandstrongpronouns,althoughthesedifficultiesweremorepronouncedwithclitics.They alsoshowedthattheyoungest(3yearold)grouptendedtousethemostrecentlymentionednountodeterminethe clitic’sreference,suggestingthatthisgroupappliedanotherprocessingstrategythanthatofrelyingontheclitic’sphi- features.Theyproposedthattheyoungerchildrenattemptedtokeepanaphoricdependenciesasshortaspossible duetoworkingmemorylimitations.Furthermore,theseauthorsalsoshowedthat3--4yearoldchildrenperformbetter withpluralaccusativeclitics,i.e.,les,thanwithsingularones,i.e.,le/la.Notethattheseitemsdifferwithrespectto gendermarking,in thattheformerbutnotthelatter morphologicallyencodethisfeature.GrüterandCrago(2012) investigatedtherelationshipbetweentheproductionofcliticsandworkingmemoryinthecontextofL2French.These authorsshowedanegativecorrelationbetweenworkingmemory(i.e.,thedigitspantask)andtheproductionofclitics bynativeChinese-speakingchildren.TheworkbyGrüterandCrago(2012)ispartofalargebodyofworkhighlighting thatthedifficultywith3rdpersonACCextendsbeyondtypicalmonolingualstoL2learners(GrondinandWhite,1996;

Grüter,2005;Paradisetal.,2003;Prévost,2006),internationallyadoptedchildren(DelcenserieandGenesee,2013;

Delcenserieetal.,2013;Gauthieretal.,2012)andsimultaneousbilinguals(MüllerandHulk,2001;Pirvulescuetal., 2014). Protracted development of this clitic also clearly emerges in various situations of atypical language development,suchasSpecificLanguage Impairment(Hamannetal.,2003;JakubowiczandTuller,2008;Paradis etal.,2003;Tulleretal.,2011),hearingloss(Delage,2008;DelageandTuller,2007;Jakubowiczetal.,2000;Tuller and Delage, 2014), as well as epilepsy (Monjauze, 2007). The relationship between the development of clitic productionandworkingmemoryhasoftenbeensuggestedtoexistinavarietyofthesecontexts(seee.g.Pirvulescu andStrik,2014;Prévost,2006;Tulleretal.,2011),however,GrüterandCrago(2012)andMateu(2015)aretheonly empiricalinvestigationsofthislink.Mateuhasthusdemonstratedthatomissionofaccusativecliticscorrelatedwith resultsobtainedinanon-wordrepetitiontaskinSpanish-speakingchildrenaged2--3.Howcanweaccountforthis phenomenon?Producinganaccusativecliticarguablyimpliesthattheprocessingsystemretainsandlinksinformation abouttwopositions:thepreverbalpositionwherethecliticispronounced(or‘spelled-out’)aftersyntacticmovement, andthepost-verbal, canonical,‘pre-syntacticmovement’position(or the‘gap’).Thissortofcognitivemanipulation arguably poses problems for immature cognitive systems and only become accessible to systems with higher computationalresources(Boothetal.,2000).

CliticomissionhasalsobeenaccountedforintermsoftheUniqueCheckingConstraintorUCC(seee.g.Gavarro´

et al., 2010; Wexler et al., 2003/2004). The UCC predicts clitic omissionin languages where thepast participle morphologicallyagreeswiththeaccusativeclitic,asisthecaseinFrench(e.g.Ill’apeinte,lafille:Heherpainted +feminine,thegirl).Thisisclaimedtostemfromyoungchildren’ssystemsnotbeingmatureenoughtocarryoutsuch acomplexoperation.IndeedtheDP‘pro’whichco-occurswithcliticsinthecanonicalobjectposition(see(4))must raise to spec,ClP by passingalso throughSpec,AgrOP, because pastparticipleagreement isrequired (see also Sportiche, 1996):

(4) [ClPCLi[AgrOPVP...DPi

(5)

Intheseinstances,thechildsystemoptsinsteadtonotprojectClP,sothatthe‘pro’DPonlyhastocheckasingleD- featureagainstAgrO,thusgivingrisetocliticomission.Itisrelevanttonotethatthisreasoningpredictsthatomission shouldnotonlyoccurwith3rdpersonACCinFrench,butalsowithothercliticssuchas1st/2ndpersonACC,sincepast participleagreementoccurswiththeselatterelementsaswell(Ilt’a décrite,Maria:‘Hehasyoudescribed+feminine Maria’).

1.2. 3rdpersonvs1stpersonACC

Incontrast tothe bulkof studies discussedabove,fewinvestigationshave beenconductedof cliticsin typically developingFrenchotherthanthoseofthe3rdperson(exceptforHamannetal.,1996,whocountedallaccusativeclitics andTulleretal.,2011describedbelow).ForotherRomancelanguages,afewauthorshaveobservedanasymmetric developmentbetween3rdperson(5)and1st/2ndpersonACC(6),withtheformeremergingslowerthanthelatter(Avram and Coene, 2008; Baauw, 2000; Coene and Avram, 2011; Gavarro´ and Forto´n, 2014; Silva, 2010; Tsimpli and Mastropavlou,2007):

(5) Jean le/la regarde John him/her looks-at

‘Johnislookingathim/her’ (6) Jean me/te regarde

John me/you looks-at

‘Johnislookingatme/you’

CoeneandAvram(2011)analyzedlongitudinalcorporaofthespeechoftwomonolingualRomanianchildren(aged 1;05--3;05)andfoundthatinobligatorycliticcontexts,ACCofthethirdpersonwerefrequentlyomitteduntilage3years, whilethoseofthefirstandsecondpersonwerealwayspresentfromtheearliestutterances.Thisasymmetricomission patternwasalsoobservedforveryyoungCatalan-speakingchildren(Gavarro´ andForto´n,2014),forGreek-speaking childrenwithSLIaged3;5to7(Tsimpli,2001)andforadultL2learners(TsimpliandMastropavlou,2007).

Tulleretal.(2011)studiedthisphenomenoninFrench-speaking adolescentssufferingfromdifferentpathologies.

Theseauthorsconductedanelicitedproductiontaskwith36typically-developing(TD)childrenaged6and11,aswellas 71adolescents(agerange11--20years)withatypicallanguagedevelopment.Theclinicalpopulationwasdividedinto threegroups:SLI,mild-to-moderatehearingloss(MMHL),andRolandicEpilepsy(RE).Both1stand3rdpersonclitics wereevaluated.Theauthorsreportthatallclinicalgroups,aswellastheyoungestTDgroup(aged6),showeddifficulty specificallywith3rdpersonACC.Moreprecisely,theTD11-year-oldgroupperformedatceilingforallcliticpronouns, whiletheTD6-year-oldswereatceilingperformanceonlyfor3rdpersonnominatives(97.4%)and3rdpersonreflexives (96.4%) butnot for 3rdpersonaccusative cliticswhichyielded anaccuracyof70.3%(versus90.6%for 1stperson accusativeclitics).Similarly,thegroupwithSLIproduced3rdpersonnominativecliticsand3rdpersonreflexivecliticsat meanratesof86.6%and93.6%respectively,while3rdpersonaccusativecliticswereproducedwithasignificantlylower meanrateof49.7%.Performancebythisgroupwith3rdpersonACCalsostrikesasignificantcontrastwiththatof1st personACC,whichrevealedameanrateof85%.ForMMHL,themeanrateofproductionof1stpersonaccusativeclitics washighaswell,attaining88.8%andinREitwasasmuchas95%,incontrasttothemeanproductionratesfor3rdperson ACC,80.9%forMMHLand85%forRE.Theseresultsattesttothestatusof3rdpersonACCasapersistentmarkerof atypicallanguagedevelopmentstemmingfromdifferentetiologies.Inaddition,mostcruciallyforthepresentstudy,3rd personACCweresignificantlymoresubjecttoavoidancethan1stpersonACCinbothTD-6-year-oldsandinadolescents withatypicallanguagedevelopment.

1.3. Sourcesofdifficultyfor3rdpersonACC

Whatmakesthesecliticsparticularlycomplex?Onepossibilityisthatthe3rdpersonfeatureitselfisthesourceoftheir difficulty,independentlyoftheaccusativecase.However,thepersonfeaturealonecannotaccountforthedifficultyof3rd personaccusativecliticssincenominativecliticsofthe3rdpersonareevenproducedearlierthanthoseofthe1stand2nd personinveryearlytypicaldevelopmentofFrench,betweentheagesof1;9and2;9(Pierce,1992;Hamannetal.,1996;

VanderVelde,1999).Nodifferencesarereportedbetweentheproductionratesof1stand3rdpersonnominativeclitics later,byage4(respectively85%and88%,Dupuy,2009).Byage6,childrenreachaproductionrateof98%forproduction ofboth1stand3rdpersonnominatives(Tulleretal.,2011),andcommitveryfewerrors:Inanelicitedproductiontask, typically-developingchildrenaged6to11(N=36,Mage=8;8)onlyproduced1.9%of3rdpersonnominativecliticswitha

(6)

gendererrorand0.6%withapersonerror(Delage,2008),contrastingwiththefrequentgendererrorson3rdperson accusativesreported in theliterature (as previouslydescribed).Giventhat ceilingperformanceisobserved with3rd personnominativesbyage6whiledifficultywith3rdpersonaccusativecliticspersists(Tulleretal.,2011),the3rdperson featurealonedoesnotseemtobethesourceofcomplexity,leadingtotheconclusionthatotherpropertiesspecifically associatedwith3rdpersonaccusativecliticsalsoaffectits acquisition,‘‘eachofthesepropertiesaddingcomplexity, separatelyandconjointly’’(Tulleretal.,2011:437).ItistothesespecificpropertiesproposedbyTulleretal.whichwenow turn.

Afirstpotentialsourceofdifficultyevokedforthe3rdpersonACCisthemorphologicalmarkingforgender.Indeed thesecliticsaremarkedforbothgenderandnumberinthesingular(le/la)whiletheircounterpartobjectcliticsofother personsonlyinflectfornumber.Itisworthrecallingthat1stand2ndpersonACCdonotmorphologicallyencodegenderin eithersingularorpluralforms(me/teforsingular,vous/nousforplural).Indeedallformsmaybeusedtoreferbothto masculineorfeminineantecedents.Givenpreviousevidenceshowingthehighprevalenceofgendererrorsinstudies targetingproductionof3rdpersonACC,thefactorofgendermaythusconstituteavaluablesourceofcomplexityofthis element.

Anotherpropertyof3rdpersonaccusativecliticsistheirreferencewhichischaracterizedbydiscourse-participant- independence,incontrasttoACCofthe1stand2ndpersonwhicharerestrictedtodiscourse-participantsforreference and,asaresult,areobligatorilyanimate(Tulleretal.,2011).Discourse-participantindependenceassuchwidensthe sourceofpossiblereferentsandmaythusalsobeasourceofdifficultyintheacquisitionof3rdpersonACC.Gibson(1998) considersthatbuildingastructureforanewdiscoursereferentincreasestheprocessingcost,whichmayleadto‘‘decays intheactivationsassociatedwithprecedinglexicalitems’’(Gibson,1998:12).Thiswouldexplainwhysentenceslike(7a), with1stor2ndpersonnominativeclitics,aresimplertoprocess(byadults)than(7b)wherethereferentofthe3rdperson cliticislacking(GibsonandWarren,1998).

(7a) ThestudentwhotheprofessorwhoIcollaboratedwithhadadvisedcopiedthearticle.

(7b) Thestudentwhotheprofessorwhotheycollaboratedwithhadadvisedcopiedthearticle

Athirdpropertyof3rdpersonACCwhichmayinfluencetheiracquisitionandwhichisspecificto3rdpersonACCas comparedtootherACCisthattheymaybefelicitouslydeletedinspokenFrenchwhenthereferenceissufficientlysalient inthediscourse(seeFo´nagy,1985;LambrechtandLemoine,1996;Lemoine,1997;Tuller,2000).Considertheexamples (8)belowwhichillustratethisoptionalityinspokenFrench(Fo´nagy,1985):

(8a) Voulez-vousquejevousdonnemonnumérodetéléphone?

want-youthatIyougivemynumberoftelephone

‘Doyouwantmetogiveyoumyphonenumber?’ Non,jeconnais__.

no,Iknow

‘No,Iknow__’

(8b) LesHauts-de-Seine,vousconnaissez__?

TheHauts-de-Seine,youknow

‘TheHauts-de-Seine,youknow__?’

(8c) Thegardenerwithanupwardheadmovementtowardatree:

J’abats__?

Ichop-down

‘Ichopdown__?’

Inadditiontodiscourse-saliencerestrictions,objectomissionisalsosubjecttolexicalrestrictions(seeFo´nagy,1985;

Tuller,2000),ascanbeobservedbelow(fromTuller,2000).Inotherwords,thirdpersonACCcanbeomittedonlyifthey areusedwithspecificverbs.

(9) LesacàdosdeLucpèseunetonnelevendredisoir.Çacontienttousseslivresettoussescahiers.

*T’asdéjàporté__?/T’asdéjàvu__?

ThebackpackofLukeweighsatontheFridayevening.Thatcontainsallhisbooksandallhisnotebooks.

*Youhaveevercarried__?/Youhaveeverseen__?

‘Luke’sbackpackweighsatononFridayevenings.It’sgotallofhisbooksandallofhisnotebooks.

Haveyoueverworn__?/Haveyoueverseen__?’

(7)

Insum,thelegitimateomissionof3rdpersonACCissubjecttodiscourserestrictions,butalsotolexicalones, whichmayrequiretimetobemastered,resultinginprolongedobjectomission.Childrenexposedtoavarietyofnull objectcontextssuchas seenabovemight thusassumethatsuchnullobjects have awider distributionthanthey actuallydointhetargetgrammar(Pérez-Lerouxetal.,2008).Indeedtheobjectomissionstageisattestedacross languages (Deen, 2006; Jakubowicz et al., 1997; Schaeffer, 1997, 2000; Wexler et al., 2003/2004,and others).

HoweverasPérez-Lerouxetal.(2008)haveshownempirically,childrenlearninglanguageswithacomplextypologyof nullobjects,suchasFrench,arelatertorestrictthedistributionofthesenullobjectsthanchildrenlearningalanguage withasimplerobjectdroptypology,suchasEnglish.Thisreasoningcanbeextendedtoexplaintheasymmetrywithin Frenchbetween3rdpersonACCand1stand2ndpersonACC,whicharetypologicallysimplerinthattheyarenever legitimatelydropped.

Theexplanationsbasedontheaboveproperties,initiallyreportedbyTulleretal.(2011),suggestthat3rdACC possessvariouspropertieswhichmayhaveanimpactontheiracquisition.Ithasalreadybeenshownthatamajorsource ofcomplexityofobjectcliticsistheirnon-canonicalpositionresultingfrominternalmerge(Belletti,1999).Otherauthors havealsoevokedprocessingdifficultiesinchaincrossingtoexplainthedifficultyofaccusativeclitics,asopposedto reflexiveones(SeeZesigeretal.,2010).Theseconsiderationshoweverdonotexplainthespecificdifficultyofthe3rd personACCascomparedtothe1st.Basedonthisobservation,thecomplexityof3rdpersonobjectcliticsisarguablythe resultofacombinationofseveralpropertiesasidefromtheirnon-canonicalposition.Aswehaveshown,theproperties associatedwith thirdpersonaccusativecliticsthatset themapartfrom1st and2nd,include theirmorphologically markingforgender,theirreferencenotbeingrestrictedtodiscourseparticipantsandtheirbeingsubjecttolegitimate omissionincertaincontexts.Thereasoningregardingthesepropertiesbeingcomplexisasfollows:(i)Ifmorphological agreementispartofthedifficultyinvolvedinproducingthirdpersonaccusativeclitics,becauseoftheneedtokeepin workingmemorythegenderofthereferent,weexpectthatACC3,whichdisplaygenderagreement,wouldbeharder than1stand2ndpersonACC(ACC1/2),whicharevoidofsuchagreement.(ii)Ifestablishingnon-discourse-participant referenceaddsdifficultybecausethelargersetofpossiblereferentsforACC3wouldtaxworkingmemoryresources, thenweexpectACC3tobehardertomasterthanACC1/2 whosesetofpossiblereferentsishighlyrestricted,i.e., respectivelythespeakerorthehearer.(iii)IfthelegitimateomissionofACC3incertaincontextsincreasesitsdifficulty becausechildrenbecomemorelikelytoovergeneralizethisomission,weexpectmoredifficultywiththemasteryofthose contextsthatdemandtheobligatorypresenceofACC3,adifficultywhichwouldnotexistforACC1/2,whichareinvariably obligatoryintheinput.

Inthisworkwethusproposetoassesstheempiricaleffectofgendermarking,discourse-participantindependenceand legitimateomission,byusingexperimentalmaterialswhichallowustodisentanglethesepropertiesof3rdpersonACC.

Wepredictthateachofthesepropertieswillhaveasignificanteffectonacquisition,andwespelloutthenatureofthe effectexpectedinourdescriptionofeachcorrespondingexperiment.

2. Method 2.1. Participants

Theparticipantswere41childrenaged4to8dividedintotwoagegroups:4--6and6--8.2Characteristicsofthesetwo groupsare presentedinTable2.Wechosethisagerangetoavoidbothceilingand flooreffects.3Allchildrenwere monolingualFrench-speakers(N=17,meanage=6;4)orhadacquiredFrenchbeforetheageofthree(N=24,mean age=6;6).Thelatter werechildrenwhowereborninSwitzerland(andthuswereexposedtoFrenchsincebirth)but whoseparents(eitheroneorboth)haveanothernativelanguage.4Notethatwewillpresenttheresultsofmono-and bilingualchildrentogethersincenosignificantdifferencewasfoundbetweenthesetwogroups(whateverthetaskused, asdetailedinAppendix1).Lastly,theparticipantsdidnotpresentanyparticularacademicorlanguagedifficultyandthus didnotreceivespeech-languagetherapy.

2Thefourexperimentshavethesameparticipants.However,notethatweadded10additional4-year-oldchildrentoexperiment3inorderto furtherassesstheimpactofdiscourse-participantindependenceinalargergroupofyoungchildren.

3IndeedDelage(2008)demonstratedthat8-year-oldtypically-developingchildrenalreadyattainceilingperformanceinelicitedproductionof accusativecliticswhile childrenyounger than4arelikelytohaveexcessivedifficultywithourgrammatical judgmenttask whichrequires metalinguisticknowledge(MaillartandSchelstraete,2007).

4AllthesechildrenwereschooledinmonolingualFrench-speakingclassesandhadoneFrench-speakingparent.Thisiswhywecanassume thattheyareFrench-dominantbilinguals.

(8)

2.2. Generalprocedure

AllchildrenweretestedindividuallybytwoexperimentersinasingleschoolinGeneva.Theiranswerswere(digitally) recordedforsubsequenttranscriptionandcoding.Alltranscriptionsweremeticulouslycheckedatleasttwice:afterthe initialcheck,asecondcheckwasdonebyadifferentcoderinordertoverifytheprecisionofalltranscriptionsandtocorrect anypotentialerrors.Incaseofdoubt,especiallywhentheelidedformoftheaccusativeclitic(l’)wasrequired,thethree coders(allnativeFrench-speakers)listenedtotheproblematicrecordingonceagaininordertoreachaconsensus.Inter- rateragreementthusattained100%foreachinstanceofthecliticwiththeelidedvowel(l’).Ifadoubtpersistedconcerning therealizationornotoftheaccusativecliticinsequencessuchas‘‘elle/ill’arrose’’,suchproductionswerecharacterizedas omissions.

Weusedatotaloffourexperimentaltasksfocusingonaccusativecliticsofwhichthreewerespecificallydesignedfor thisstudy.Aspreviouslyexplained,ourstudyaimstoexploretherespectiverolesofpropertieswhichshouldmake3rd personaccusative clitics(ACC3)more difficultto acquirethan1st(ACC1) and2nd person(ACC2):morphological marking for gender, discourse-participant independence and optionality in spoken French under specific (discourse and lexical) conditions.We havetherefore developedone experimental taskfor eachof thesepotential sourcesofdifficulty.Inaddition,wewantedtoreplicatetheresultsofTulleretal.(2011)whichclearlyshowedthegap betweenACC1andACC3(infavorofthefirst)in6-year-oldchildren.WealsowantedtoreplicateDelage(2008)who showedthattypically-developing8-year-oldsperformedatequallyhighratesforACC1andACC3.Withthisinmind,we focusonchildrenaroundtheageof6yearssoasto(i)checkforthepresenceoftheACC1/ACC3gapinchildrenyounger thanthosetestedinTulleretal.’sstudy,and(ii)documenttheevolutionfrompoorperformancetomasteryofbothACC1 andACC3.

Inthefollowingsections,wesystematicallypresentthepropertiesoftheACCthatareexamined,ourcorresponding prediction,experiment,resultsanddiscussion.Alltheparticipantscompletedthedifferenttasksinthesamesessionand inthesameorderasthatinwhichwepresentthesetasks.

Notethat,duetothenon-normaldistributionofthedata(confirmedbytheShapiro--Wilktest),ourmainanalyseswere conductedwithnon-parametrictests,withtheMann--Whitneytestforinter-groupcomparisonsandtheWilcoxontestfor intra-groupcomparisons,associatedwiththeFriedmanANOVAandSpearman’srankcorrelations.

2.3. EXPERIMENT1:ACC1/ACC3gap

ThistaskwasusedtoverifythepresenceofthediscrepancybetweenACC1andACC3.Weexpectedtoreplicate resultsobtainedbyTulleretal.(2011)in6-year-olds,i.e.,toshowbetterperformanceforACC1comparedtothatofACC3.

Moreover,sincetheACC1/ACC3gapdisappearsatage8(Delage,2008),weexpectedtofindalesspronouncedgapin theoldergroup(6--8),comparedtotheyoungerone(4--6).

2.3.1. Taskandstimuli

Inordertotest productionofACC1and ACC3,weadministeredashortened versionoftheProductionProbefor PronounClitics(PPPC,Tulleretal.,2004)whichwasusedin2011bythesameauthorstohighlighttheACC1/ACC3gap.

Weadaptedthistaskbyremovingitemswhichrequiretheproductionofreflexivecliticsandbyaddingfillersinorderto avoidthatallitemsrequiredaccusativeclitics.Thefinaltaskcontained16test items(ACC1and ACC3)whichwere randomizedandprecededby2pre-testitemsandmixedwith4fillers(seeTable3).Appendix2providesthecompletelist ofalltargetitems.Theexperimenterelicitedcliticsbyaskingaquestionaboutadrawingappearingonacomputerscreen, asillustratedinTable4(forthetwopre-testitems).

Asdetailedintheinitialdescriptionofthistask(Tulleretal.,2011),theomissionoftheaccusativecliticselicitedinthese contextswouldleadtoanungrammaticalresponse,inthe sensethatdiscourseandlexicalrestrictionsallowingtheir potentiallegitimateomissionarenotpresent.Moreover,wheneverthechildansweredwithastructureotherthantheone expected(thatis,withaverbwhichdoesnotrequireanaccusativeclitic),theexperimenteraskedthechildtoreformulate his/heranswerbyspecifyingtheverbtouse.Belowisanillustrationofhowthiswasaccomplishedwithanexamplesuch Table2

Characteristicsofthetwoagegroups.

Group N Sex Agerange(year;month) Age:M(SD)(year;month)

4--6 21 10M11F 4;7--6;4 5;7(0;6)

6--8 20 10M10F 6;5--8;5 7;4(0;8)

(9)

asthatin(10),wheretheexpectedresponseisEllemelèche(‘‘It’slickingme’’)andthechildprovidedanutterancewhich doesnotrequireaclitic(10a)andispromptedtocorrecttheutterancewiththeappropriatetransitiveverb(10b)5: (10a) Lavacheestsympa

‘Thecowisnice’

(10b) Maiscommenttudiraisaveclécher?

‘Buthowcouldyousayitwith‘lick’?’

2.3.2. Results

Forthetotalof41childrenaged4--8,ACC1wereproducedat90%(SD=21),thusapproachingceilingperformance, whereasACC3wereproducedatonly76%(SD=25).Thedifferenceissignificant(T=30,p<.001,r=.41),whichclearly confirmsthe existence of the ACC1/ACC3gap. Notethat theseratesinclude,in additionto the correctlyproduced accusativeclitics,thoseaccusativecliticsthatwereproducedwithanerrorofgenderorperson.6Notethatproduction ratesofentirelycorrectACC1andACC3alsodiffered(T=103,p<.001,r=.37).TheACC1/ACC3gapisobservedboth inthegroupaged4--6(T=15.5,p<.01,r=.44)andintheoneaged6--8(T=0,p<.01,r=.42).Giventhatpreviouswork revealedanabsenceofthisgapin8-year-olds(ACC1vs.ACC3:p=0.3,Delage,2008)aswellasbeyondtheageof8,we cannowconfirmthatchildrendonotattainceilingperformanceforACC3beforetheageof8.Fig.1showsratesofACC1/

ACC3productionforourtwoagegroups,comparedtoratesobtainedin128-year-oldchildren(M=8;2,SD=0;3)tested inapreviousstudy(Delage,2008).

Finally,erroneouscliticsconsistofpersonerrors(3.7%,SD=8)forACC1andofgendererrors(12.2%,SD=11.7)for ACC3.Non-expectedresponsesconsistofcliticomission(9.1%,SD=19)anduseofDP(0.9%,SD=4.3)forACC1,and cliticomission(17.1%,SD=20.5),useofDP(5.5%,SD=10.1),non-response(0.9%,SD=3.3)andcaseerrors(0.1%, SD=0.3)forACC3.

Table3

CliticpronounselicitedbytheshortenedversionofthePPPC(EXPERIMENT1).

2pre-testitems 16testitems 4fillers

1xACC1 me

1xACC3 la

8xACC1 me

8xACC3 4masc.le 4fem.la

Allintransitiveverbs e.g.:ilnagehesswimming

Table4

Examplesofstimuli(EXPERIMENT1).

ACC3 Experimenter:

Quefaitlemonsieuravecsavoiture?

Whatsthemandoingwithhiscar? Expectedresponse:

Illalave.

Heherwashes

Heswashingit. ACC1 Experimenter:

Lui,ildit«Hé,Marie,quefaitlavache?»Toi,tuesMarie,quest-cequeturéponds?

Hesays«Hey,Mary,whatsthecowdoing?»YouareMary,whatdoyouanswer? Expectedresponse:

Ellemelèche.

Shemelicks

Itslickingme.

Note:IllustrationsaregiveninAppendix3.

5Thissamestrategywasappliedduringtheotherelicitationprobesofthepresentstudy.

6Notethatthetwooccurrencesofgrammaticaldativecliticswerecountedascorrectresponsessoastorenderourresultscomparabletothose reportedinTulleretal.(2011)whoproceededthisway.

(10)

2.3.3. Discussion

OneoftheaimsofthepresentresearchwastoconfirmthedelayofACC3ascomparedtoACC1,namelytheACC1/

ACC3 gap (Avram and Coene, 2008; Tsimpli and Mastropavlou, 2007; Tuller et al., 2011) and to investigate the developmentaltrajectoryofthisdelay.Ourresultshaveclearlyshownthisasymmetryandfurtherhighlightedforthefirst timethatthegapbetweenACC3andACC1persistsuntilwhenchildrenare8yearsold.

2.4. EXPERIMENT2:Morphologicalmarkingforgender

WedevelopedatasktotestspecificallyandindependentlygenderasapotentialfactorofdifficultyinproducingACC3.

Thistaskisdesignedsimilarlytotheprecedingtestsinceitalsoelicitsaccusativecliticsbyrequiringparticipantstogivean answertoa questionabout animage appearingonacomputer screen.Theoriginalityweadded isthe presenceof accusativecliticswhosegenderisneutralized.Thiswasattainedbyusingtheformwiththeelidedvowel(l’)whichdoes notshowovertagreementandcanthereforerefertoeitheramasculineorfeminineantecedent.Wealsovariedthegender ofbothnominativeandaccusativecliticstocreatedifferenttypesofcliticsequences(i.e.,anominativecliticimmediately followed byanaccusative clitic).Weexpected thattwoconsecutivecliticswithdifferent markingfor gender,(called mismatchsequences,e.g.,nominativemasculineil‘he’followedbyaccusativefemininela‘her’,asinillalave‘heher washes’) wouldbe more complexto processthan cliticssharing the samemorphological properties (calledmatch sequences).Indeed,the combinationoftwocliticsrequiresparticipantstoholdinworkingmemorytheirtwodistinct referentsincludingtheirrespectivemorphologicalproperties.Itthusseemslogicalthatsuchdiscordancebetweentwo morphological markings would require increased working memory demands and lead to confusion as well as subsequentlytogendererrorinsuchsequences.

Wepredictthatgenderagreementplays aroleintheacquisitionofACC3,andmorespecificallythatchildrenwill perform better on neutral sequences (without gender agreement on ACC3) than on other sequences (with overt agreement).Moreover,ifgendermismatchmakesindeedthetaskmorecomplex,inparticularduetotheadditionalcost involvedinstoringtwodifferentmorphologicalfeatures,weexpectperformancetobeworseonmismatchsequencesthan onmatchsequences.Thus,asTulleretal.(2011)underlinedintheirstudy,childrenarelikelytoconfusethegenderofthe nominativecliticandthegenderofthelastDPtheyhear,bothbeingofdiscordantgenderinthemismatchsequences.

Moreprecisely,thepredominantpatternthattheseauthorshavefoundinvolvedthechildusingthegenderofthelastDPof thequestionforthenominativecliticandthentheoppositegenderfortheaccusativeclitic,asshownbelow:

(11)Experimenter: Que fait Pierre à la dame?

what does Pierre to the lady

‘WhatisPierredoingtothelady?’ Expectedresponse: Il la coiffe.

he her doing-hair

‘He’sdoingherhair’ TD-6y-o: Elle le coiffe.

she him doing-hair

‘She’sdoinghishair’

Fig.1. Correctproductionrates(%)andstandarddeviationsforACC1andACC3in4--6,6--8and8-year-oldagegroups(EXPERIMENT1).

(11)

2.4.1. Taskandstimuli

Threetypesofsequences,allinvolvingsingularaccusativeclitics,wereelicitedinthistask(examples12--14):

(12) Neutralsequenceswithanominativecliticmarkedforgender(il/elle‘he/she’)followedbytheformofACC3with theelidedvowel(l’),suchasil/ellel’arrose‘he/sheissprinklinghim/her’

(13) Matchsequenceswithnominativeandaccusativecliticssharingthesamegendermarking(ille‘hehim’,ellela

‘sheher’),suchasille/ellelamontre‘he’spointingathim/she’spointingather’

(14) Mismatchsequenceswithnominativeandaccusativecliticssharingdifferentgendermarking(illa‘heher’,elle le‘shehim’),suchasilla/ellelemontre‘he’spointingather/she’spointingathim’

Therewereatotalof18testitemsaltogether(seeTable5),precededby4pre-testitemsandalternatedwith6fillers.

Beforethetesting,childrenwerefamiliarizedwiththepictogramsused(fromPicteliasoftware).

Ineachtrial,the experimenterstartedbyproducinga sentenceinanSVO ordertodescribethe transitiveaction illustratedonthescreen(seeexamplesinAppendix4).Thepicturescorrespondingtothesubjectandtotheverbthen disappearedinordertoavoidtheinfluenceofthecanonicalpositionofthethreepicturesonthescreen(thepictureonthe leftofthescreenrepresentingthesubject,theoneinthemiddletheverbandtheoneontherighttheobject),withonlythe pictureoftheobjectremaining.Atthispoint,theexperimenteraskedaquestiontoelicittheappropriatesequenceofclitics.

Table6illustratesthisprocedureforneutral,matchandmismatchsequences.Note,too,thatpictogramsrepresentingthe verbswere animated to betterillustrate the actions, which made this taskmore attractive for very young children.

Appendix5providesthecompletelistofalltargetitems.

2.4.2. Results

AFriedman ANOVA analysis (conducted onresults of the 41 children)shows a significant effect of the type of sequences(r=17.60,dl=2,p<.001),whichthusconfirmsourinitialhypothesis.Neutralsequences(consideredasthe lesscomplexsequences)wereproducedcorrectly(57%,SD=39)atsignificantlyhigherratesthanmatchsequences (47%,SD=39;T=51.5,p<.01,r=.29)andmismatchsequences(39%,SD=35;T=21,p<.001,r=.39).Ontheother hand,performanceonmatchsequenceswasbetterthanonmismatchsequences,eventhoughthedifferenceisnolonger significant if we adopt the Bonferroni correction (T=79.5, p<.05 but >.017, r=.22). Note that production rates correspondtoentirelycorrectsequences,withoutgendererrors(whichwedescribebelow).Thisisbecausewhatwe wantedtoobservewasspecificallythechildren’scapacitytousecorrectmorphologicalagreementontherelevantitems.

Regardingdifferencesbetweenagegroups,4--6year-oldsperformedlesswellthan6--8year-oldsonthethreetypesof clitic sequences(neutral: U=103, p<.01, r=.43; match: U=106, p<.01, r=-.42; mismatch:U=113,p<.017, r=.39),showingadevelopmentaleffect.Thetwogroupsalsoshowedaneffectofthetypeofsequences(for4--6:

r=7.18,dl=2,p<.05;for6--8:r=12.46,dl=2,p<.01).Fig.2illustratesthattheyoungergroupalwaysperformedless well than the older one, with the same pattern being attested in both groups, i.e., better performance for neutral sequences,followedbymatchsequencesandfinallybymismatchsequences.

Turningnowtothepatternoferrorsproduced(onlyonaccusativeclitics),Table7providesthedistributionofthese errorsinthetwoagegroups.WhenACC3werenotproduced,productionofalexicalDPwasthemostcommonanswer.7 Recallthatduringthistask,whenthechildhadtoreply,theonlypictureremainingonthescreenwastheonedepictingthe object;thiscanexplainwhylexicalDPisthemostfrequentinappropriateresponse.Fortheotheranswers,the4--6year old group made more omission errors than the 6--8 year old group (U=112.5, p<.01, r=.39), but did not differ significantlyongender errors(p=.073) whichthus persist andwere observedequally frequently in bothdirections (masculineintofeminineandviceversa).Moreover,weobservedmoregendererrorsonACC3inmismatchsequences (23.6%) than match sequences (10.6%; T=15, p<.01, r=.30). This confirms the specific difficulty due to the manipulationoftwodifferentmorphologicalfeaturesinsequencescontainingtwoclitics.

Lastly,ifweconsiderthegendererrorsproducedonlyinmismatchsequences,weobservethatthemajorityofthese errorsleadstoasinglegendermarking(insteadoftwo)ofthetwoclitics.Indeed,ofallthegendererrorsinmismatch sequences,69%consistofapplyingoneuniquegendertothetwoclitics,asin(15).Theother31%ofgendererrors consistofinvertingthemorphologicalmarkings(16).

7Thesamepatternisobservedforthevariousgenderconditions,withapredominanceoflexicalDPs(25%oftheresponsesforACC3 productioninneutralsequences,35%inmatchsequencesand40%inmismatchsequences),ascomparedtoomissions(13%inneutral sequences,8%inmatchsequencesand5%inmismatchsequences).TherelativelyhighrateofACC3omissioninneutralsequencescouldbe attributedtothefactthatwecharacterizeasomissionanycaseinwhichthethreenativecoderscouldnotreachaconsensusabouttherealization ornotoftheaccusativecliticinsequencessuchas‘‘elle/illarrose"(asmentionedinSection2.2).

(12)

(15) Experimenter:Quefaitlafilleaveclecroissant?‘Whatisthegirldoingwiththecroissant’ Expectedresponse: Ellelemange.

Shehimeats

‘She’seatingit’

*STA(5;5y.o.): ellelamange.

Shehereats (16) *PHI(5;5y.o.): illamange.

Hehereats

Table8indicates,foreachagegroup,theproportionofgendererrorsproducedinmismatchsequencesonnominative andonaccusativeclitics,aswellasthetypeoferrors.Twoerrortypeswerepossible:(1)eitherthechildinversedthe gendersofthetwoclitics(onbothnominativeandaccusative);(2)ors/hesubstitutedthegenderofonesingleclitic(the nominativeortheaccusative),thusleadingtooneuniquegenderforthesequence.

2.4.3. Discussion

Twoobservationsconcerningtheeffectofgendermarkingareofparticularinterest.Firstly,wehavedemonstratedthat itismoredifficultforchildrentoproducecliticsequencesinwhichgenderofACC3isovertlymarkedthantoproduce sequencescontainingtheformwiththeelidedvowell’wherethereisnoovertgendermarking.8Thiscouldbesurprising Table5

Sequenceselicitedbythetaskongendermarking(EXPERIMENT2).

18testitems 6fillers

6neutralsequences 3xill

3xellel

6matchsequences 3xille

3xellela

6mismatchsequences 3xilla

3xellele

Allintransitiveverbs e.g.:ilplonge

hesdiving

Table6

Examplesofstimuli(EXPERIMENT2).

Neutralsequences Experimenter:

Regarde,legarçonessuieunegrossetachesurlatable.Alorstutesouviens?Quefaitlegarçonaveclatache?

Look,theboyiswipingabigspotoffthetable.Sodoyouremember?Whatistheboydoingtothespot? Expectedresponse:

Illessuie.

Heitwipes

Heswipingit. Matchsequences Experimenter:

Regarde,lafilleprendunebanane.Alorstutesouviens?Quefaitlafilleaveclabanane?

Look,thegirlistakingabanana.Sodoyouremember?Whatisthegirldoingwiththebanana? Expectedresponse:

Ellelaprend.

Shehertakes

Shestakingit. Mismatchsequences Experimenter:

Regarde,legarçonmontreunegirafe.Alorstutesouviens?Quefaitlegarçonaveclagirafe?

Look,theboyispointingatagiraffe.Sodoyouremember?Whatistheboydoingtothegiraffe? Expectedresponse:

Illamontre.

Heherpointsto

Hespointingatit. Note:IllustrationsaregiveninAppendix4.

8Note,however,thatsincegenderagreementisnotvisibleontheformswithanelidedvowel,wecannotknowifthesecliticsareaccurateornot withrespecttothisfeature.

(13)

sincethe so-called neutralcondition actuallyinvolves resyllabificationand the deletionofthe vowelwhich couldbe assumedtobeadditionalprocessingoperations(LeveltandWheeldon,1994).Ithaspreviouslybeenclaimedthatgender oncliticsisdifficultforyoungchildren(Zesigeretal.,2010,a.o.),althoughthismarkingisunproblematicwhenoccurringon articles(JakubowiczandNash,2003).Forthefirsttime,wehavebeenabletobringtolightthedifficultyassociatedwith genderoncliticsthroughacomparisonwithcliticswithoutthisovertagreement.Thisconfirmsthestrongimpactofgender markingonACC3.Moreoveritisnoteworthythattheformofthecliticwiththeelidedvowelisstillabstractlyspecifiedfor thegenderfeatureascanbeseenfromitstriggeringofparticipleagreementinthetargetgrammar(e.g.ill’aprise‘heit took[feminine]).However,theelidedformiscruciallynotovertlymarkedforthegenderfeature,whichaffectsthechildren’s performance.Thistestifiestothecomplexifyingroleofovertgendermarkinginproduction,asdoesthefactthat6to8- year-oldscontinuetomakeasignificantrateofgendererrorsonACC3.9NotealsothatresultsofbothPirvulescuand Belzil(2008)andAvrametal.(2015)providesupportfortheviewthatcomplexityisincreasedwithovertgendermarking in sequences of nominative and accusative clitics. Hence, Pirvulescuand Belzil (2008) provide experimental data showingthatFrench-speakingchildrenaged1;10to2;10produceACC3significantlymoreoftenwhenthesubjectis‘‘tu’’

‘you’whichisvoidofmorphologicalagreement.Avrametal.(2015)reportedresultsforataskwhichfocusedontheroleof gendermatch/mismatchinACC3productionin Romanian,whichisanullsubjectlanguage.Theyreportedveryfew Fig. 2. Correct productionrates(%) andstandard deviations forneutral, match andmismatch sequences in4--6and 6--8agegroups (EXPERIMENT2).

Table7

Distribution(%andnumber)ofunexpectedanswersproducedinsteadofACC3in4--6and6--8agegroups(EXPERIMENT2).

LexicalDPs Omissions Gendererrors

4--6y.o. 65%

(157/240)

25%

(59/240)

10%

(24/240)

6--8y.o. 69%

(87/126)

10%

(13/126)

21%

(26/126)

Table8

Characterizationofgendererrorsappearingonmismatchsequencescontainingbothnominativeandaccusativeclitics(EXPERIMENT2).

Locusofgendererrors Typeoferrors

Nominativeclitics Accusativeclitics Oneuniquegender Inversionofgenders

4--6y.o. 53%

(17/32)

47%

(15/32)

58%

(11/19)

42%

(8/19)

6--8y.o. 30%

(6/20)

70%

(14/20)

81%

(13/16)

19%

(3/16)

9Anotherargumentcomesfromthegendererrorsproducedonnominativesclitics(11%insequencescontaininganovertagreementonACC3, versus4%insequenceswithoutsuchagreement).

(14)

gendererrorsanddidnotfindanydifferencebetweencliticproductioninresponsestogendermatch/mismatchcontexts.

Thissuggeststhatthecomplexityofmatch/mismatchsequencesistheresultofthetwoclitics(nominative+accusative) beingovertlymarkedforgender.

Secondly,wehavealsodemonstratedtheeffectofcombiningtwodifferentmorphologicalgendermarkingsthroughthe inclusionofstructureswithbothamatchandamismatchingenderonthenominativeandaccusativeclitics.Whatour resultsillustrateisthatthemismatchsequences(illa/ellele)aremoredifficultthanmatchsequences(ille/ellela)in production.Thisisthefirsttimethatthishasbeenformallytestedforclitics(withovertagreements).Howcanweaccount forthis?Couldthisberelatedtointerferencecreatedbythemismatchinfeaturesbetweentheobjectcliticchainandthe interveningtraceleftbythemovedsubject(seealsoAvram,2011;Zesigeretal.,2010)?Thisseemsunlikelyinlightofa recentstudybyBellettietal.(2012)showingthatonlyfeaturesfunctioningasattractorsforsyntacticmovemententerinto thecomputationofintervention.Asgenderisnotpartofthephi-featurecomplexthatdeterminesmovementinFrench,itis notexpectedtoaffectlocalityconstraintsonchainformation.10Anotherpossibleexplanationisintermsofcognitiveload:

duringproduction,onenot onlyhastokeepinmindthemorphologicalmarkingsthatarelinkedtothecorresponding referents11butalsoneedstoapplythismarkingonanotherrelevantelement,thepronouninourcase.Onestrategytodeal with thisisto re-applya featural marking previously activated by anotherelement in the utterance. Recall thatour predominant errorpatternisconsistent withthis perspective,namelythata singlemorphologicalmarking wasoften erroneouslyappliedacrosselementsinthecliticsequenceforbothagegroupstested.Thatworkingmemoryishighly taxed by such constructions and canthus bring about gender mismarking was already suggested in Tuller et al.

(2011:436).12Moreprecisely,theseauthorsevokedtheideathatthemostrecentlyusedgenderwouldleadtofavoring thisgenderinthechild’sresponse.Similarly,PirvulescuandStrik(2014)showedthat3year-oldstendedtousethemost recently mentioned noun to determine the clitic’s reference and attributed this recency effect to working memory limitations.Wealsofoundthispatternfrequentlyamongstouryoungestgroup,aged4--6(seeTable8).However,thisis notthepredominantpattern,notablyintheoldergroup,aged6--8,sincetheywereabletoaccuratelyrealizegenderonthe nominativeclitic,plausiblybecausetheirworkingmemoryissignificantlybetteratthisage.Nevertheless,theycontinued tomakegendererrorsontheaccusativeclitic.Thiscouldbeduetothefactthatthesubjectinourtaskisalwaysaboyora girl,whosegenderisnatural,comparedtothegrammaticalgenderoftheobjectwhichiseitherananimaloraninanimate object.Thisspecificityofthistask(withonlynaturalgenders)rendersthegenderofthesubjectthemostsalientoneand thusitprimesthegender ofthe object.Recallthatinour taskchildrenperformedbetterin thematchcondition,and preciselyinthiscondition,thestimuliwhichthe childheardcontainedfourinstancesofeither feminineormasculine gender,suchas:‘‘Look,thegirl(feminine1)istakingthebanana(feminine2).Whatisthegirl(feminine3)doingtothe banana(feminine4)?’’Thismayarguablyhaveyieldedaprimingeffect,i.e.,whereparticipantsaremorelikelytousea particularformifithasrecentlybeenemployed(seee.g.Bock,1986;WeinerandLabov,1983fortheseprimingeffectsin syntax).Giventhis,themultipleinstancesoffeminineormasculinegenderinthematchconditionmaywellhaveincreased thechild’saccuracyascomparedtoinstanceswherenosuchpredominanceofaspecificgender waspresentinthe precedingstimuli.

2.5. EXPERIMENT3:Discourse-participantindependence

Wenowopposediscourse-participantdependentaccusativeclitics(i.e.,1stor2ndpersoncliticswhosereferenceis directlytiedtothespeechact)anddiscourse-participantindependentclitics,namely3rdpersoncliticswhosereferenceis necessarilyotherthanthespeechactparticipants.Wepredictdiscourseindependencetoplayaroleintheacquisitionof ACC3,particularlyduetotheresourcesnecessaryinworkingmemorytokeepinmindthesyntacticreferent.Thuswe expecttoobservebetterperformanceinsequenceswithACC2comparedtothosewithACC3.

2.5.1. Taskandstimuli

InordertoelicitACC2andACC3naturallyandparticularlythosewithdiscourse-participantdependence(ACC2),we createdataskinwhichthechildhastodealwithtangibleactionsandrealcharacters(inspiredbyPirvulescuandHill’s study,2011).Thisnaturalcontextfacilitatesthe productionofdiscourse-dependentACCandmoreover seemstobe highlymotivatingforyoungchildren.Aswiththeprevioustask,thisonealsoelicitssequencesofclitics,thenominative alwaysbeing3rdpersonsingular(il‘he’)andthesecond(accusative)beingeither2ndpersonsingular(ACC2:te‘you’,

10Note,however,thatotherfeatureswhichdonotclearlyfunctionasattractorsforsyntacticmovementhavebeenrecentlyarguedtomodulate thecomprehensionofobjectrelativeclausesinFrench(seeDurrlemanandBentea,2015).

11Inourtask,thismarkingalwayscorrespondstogrammaticalgenderandnottoconceptualgender,i.e.,linkedtothesexofthereferent.

12NotethatDelageandFrauenfelder(2012)empiricallyverifiedtherelationshipbetweencomplexsyntax(i.e.,embedding)andworkingmemory inchildrenwithtypicallanguagedevelopment.

(15)

whosephonologicalformisthesameforafeminineormasculinereferent)or3rdpersonneutral13(ACC3:l’‘it’).Wechose toevaluateonlyneutral3rdpersonaccusativecliticsinordertoavoideffectsduetodifferencesinmorphologicalgender markingbetweensequencesbeingcompared.Indeed,thismarkingisneutralizedforboth2ndorneutral3rdperson accusativeclitics;sequencesofcliticswerethusdistinguishedonlybytheeffectsofdiscourse-participant(in)dependence andnotbymorphologicalgenderfeatures.Thetaskcontains16testitemsaltogether(seeTable9),precededby5pre-test itemsandmixedwith6fillers.Beforetesting,childrenwerefamiliarizedwiththeactions(andthecorrespondingverbs)to bepresentedduringthetask.Appendix6providesthecompletelistofalltargetitems.

Asalreadyexplained,theprocedureinvolvedaninteractivesituation.Oneexperimenter,aman,performedanaction eitheronthesecondexperimenter14(awomanwitheyescovered)oronanobjectsuchasaBarbiedoll,aplasticflower,a flashlight,anappleorevenasmallbox.Thewomanthenaskedthechildwhatthemanhaddone.Theinitialinstructions aregivenin(17).

(17)Femme:Regarde,jemebandelesyeux,jenevaisplusrienvoir!(Nomdel’expérimentateurhomme)vafairedes chosesmaiscommejenevoisrien,jenesauraipasexactementcequ’ilfait!Toi,tupeuxtoutvoiralorstuvasmedirece qu’ilfait,soitàmoi,soitàlaBarbieouàunobjet,soitilfaitdeschosestoutseul!Tuascompris?

Woman:‘Look,myeyesarecovered,Ican’tseeanything!(Nameofthemaleexperimenter)willdosomethingsbutasI cannotsee,Iwon’tknowexactlywhat!YOUcanseeeverything,soyouwilltellmewhatheisdoing,eithertome,ortoa Barbieortoanobject,orevenwhatheisdoingallbyhimself!Doyouunderstand?’

Whenquestioning,thewomanalwaysmentionedtheobjectthatthemanwasmanipulating.Ifsheherselfwasthis

‘‘object’’,sheusedthestrongpronounmoi‘me’inordertoavoidproducingacliticpronoun.ExamplesinTable10illustrate thisprocedureforsequenceswithACC2andwithACC3.

2.5.2. Results

Resultsobtainedbythe41childrenshowedadifferencebetweenthetwoitems,ACC2(M=75%,SD=30)beingmore frequentlyproducedthanACC3(M=66%,SD=32).Thisdifferencedoesnotattainstatisticalsignificance(p=.068).

This effect thus appears to beweaker than those observed for the properties studiedin the previous experiment.

Furthermore,thedifferencewasnolongerdetectablewheneachagegroupwasconsideredseparately,confirmingthe weaknessoftheACC2/ACC3gapinourpopulation.Turningnowtodifferencesbetweentheseagegroups,itisinteresting tonote that4--6 year-old childrenperformed significantlylesswell that6--8 year-oldson ACC3(U=107.5,p<.01, r=.35),whereasinter-groupdifferenceonlyrevealedamarginaltrendtowardsignificanceforACC2(p=.057).Fig.3 illustratesthesedifferencesforACC2andACC3.15WhenACCwerenotproduced,omissionswerethemostfrequent non-expectedresponses.Thus,outofallthenon-expectedanswers(N=179),83%wereomissions,followedbythe productionoflexicalDPs(15%).

2.5.3. Discussion

Ourresultssuggestatendencytowardbetterperformancewithpronounswhosereferenceislinkedtothediscourse(Il te). Thiscanbeexplained bythe factthatidentifyingthe referentfortheseelementsisfacilitatedbecausepotential referentsarerestrictedtothoseinvolvedinthespeechact:thespeaker(me)ortheinterlocutor(te).Howevertheweak effectofthispropertythatweobservedmaybeduetothefactthatitismasteredearlierinacquisition.Withthisinmind,we Table9

Sequenceselicitedbythetaskondiscourse-participantindependence(EXPERIMENT3).

16itemtests 6fillers

8sequenceswithdiscourse-participant dependence(ACC2):Ilte

8sequenceswithdiscourse-participant independence(ACC3):Ill

Allintransitiveverbs e.g.:ilbaille

hesyawning

13Amongsttheseitems,halfincludeitemswith[nonanimate]features(e.g.,apple)andhalftheBarbiepuppetwhichcouldbeconsideredby childrenasanimate.Wecomebacktothisinfootnote15.

14Suchasputtingonmake-up,tapping,tickling,pointingto,sprinkling,etc.

15Recallinfootnote13wepointedatthathalfofourACC3withdiscourse-independencereferredtoinanimateentitieswheretheotherhalf refereedtoinanimates.Nostatisticalrelevantdifferenceswereidentifiedbetweenthesetwoconditions(p=.45).

Références

Documents relatifs

It is, however, not very probable that similar citizen programmes will be set up for flora and insects in the near future, even though national structures with the means to

The aim of this presentation is the syntax- semantic interface for clitics analyses in which we will stress on clitic climbing over verb and raising verb.. Keywords Minimalist

Since error analysis in grammar serves as an effective way to improve writing proficiency, in order to make English teaching more efficient in our department,

languages which have such plural clitics (including modifications like -ān &gt; -ūn etc.) can have adopted them from Middle or New Persian in the way assumed

Chronic osteomyelitis treatment is complex, due to the difficulty to achieve therapeutic drug levels at the site of infection by systemic administration.. Our objective was to

Se pòt pensar que, puslèu que d’extasi (concebuda coma un artifici poetic que destòrç la fonologia normala en impausant una reculada de l’accent), s’agís

The syntactic analysis could still be saved, for example by introducing further fea- tures to distnguish clitics on the verb and preverbal clitics, but we prefer to treat

Although recent studies of the acquisition of Romance clitics by second language (L2) learners (see, amongst others Bruhn-Garavito and Montrul, 1996; White, 1996; Duffield and