Information
management
in
Mandarin
child
speech,
maternal
speech,
and
adult
speech
§
Chiung-chih
Huang
a,b,*
aGraduateInstituteofLinguistics,NationalChengchiUniversity,No.64,Sec.2,Zhi-nanRd.,WenshanDistrict,
TaipeiCity11605,Taiwan
b
ResearchCenterforMind,Brain,andLearning,NationalChengchiUniversity,No.64,Sec.2,Zhi-nanRd.,
WenshanDistrict,TaipeiCity11605,Taiwan
Received27October2015;receivedinrevisedform22May2016;accepted8June2016
Availableonline16July2016
Abstract
ThisstudyinvestigatedinformationmanagementinMandarinchildspeech,maternalspeech,andadultspeech.Thethreetypesof
speechwereanalyzedtotestthehypothesisofpreferredargumentstructure(DuBois,1987,2003),andtheywerealsoanalyzedinterms
oftwotypesofdiscourseinformationmeasures.Thedataconsistedofeighthoursofmother--child(aged2;2to3;1)conversationandone
hourofadult--adultconversation.Theanalysisofthedatashowedthatthethreetypesofspeechdemonstratedsimilarpatternsin
grammaticalalignmentandthatallofthemfollowedtheconstraintsofquantity.Thediscourseinformationmeasures,ontheotherhand,
showedthatthethreetypesofspeechdifferedsignificantlyinthemeasureofinformationpressureandinthemeasureoflexicalreferential
density.Theresultsrevealedhowinformationmanagementmaybeassociatedwithchildren’sdevelopinglinguisticability,mothers’
speechadjustments,andadults’maturecompetence.
©2016ElsevierB.V.Allrightsreserved.
Keywords: Informationmanagement;Argumentstructure;MandarinChinese;Mother--childconversation;Languageacquisition
1. Introduction
1.1. Preferredargumentstructure
Informationmanagementindiscoursehasbeenextensivelystudiedinthefunction--orientedresearchoflanguage. Oneoftheimportanttheories ofinformationmanagementisDuBois’preferredargumentstructure(henceforthPAS) (1987,2003).PASconcernstheflowofinformationindiscourseanditsinteractionwiththeprimarynounarguments associatedwithverbphrases:Thesubjectofatransitiveverb(A),theobjectofatransitiveverb(O),andthesubjectofan intransitiveverb(S).
ThecentralnotionsofPAScanbeexpressedintheformoffourconstraints,whicharehypothesizedtobeuniversal across languages. Thefour constraints include two quantityconstraints and two role constraints. The twoquantity
www.elsevier.com/locate/lingua
Availableonlineatwww.sciencedirect.com
ScienceDirect
Lingua184(2016)53--68
§ThehandlingEditorsforthisarticlewereIanthiMariaTsimpliandHarryWhitaker.
*Correspondenceto:GraduateInstituteofLinguistics,NationalChengchiUniversity,No.64,Sec.2,Zhi-nanRd.,WenshanDistrict,TaipeiCity
11605,Taiwan.
E-mailaddress:cchuang@nccu.edu.tw.
http://dx.doi.org/10.1016/j.lingua.2016.06.001
constraintsaretheOneNewArgumentConstraintandtheOneLexicalArgumentConstraint.Thetworoleconstraintsare theGivenAConstraintandtheNon-lexicalAConstraint.
TheOne New ArgumentConstraint statesthat eachclause containsno more than one argumentcarrying new information,and theOneLexicalArgumentConstraintindicatesthat eachclausecontainsnomore thanonelexical argument.Thesetwoquantityconstraintsarelocatedinparalleldimensions:TheOneNewArgumentConstraintisinthe pragmaticdimension,whiletheOneLexicalArgumentConstraintisinthegrammaticaldimension.TheGivenAConstraint claimsthattheAroletypicallycarriesoldinformation,andtheNon-lexicalAConstraintsuggeststhatlexicalargumentsdo notappearintheArole.Thesetworoleconstraintsarealsolocatedinparalleldimensions:TheGivenAConstraintisinthe pragmaticdimension,whiletheNon-lexicalAconstraintisinthegrammaticaldimension.Table1summarizestherelation betweentheconstraintsanddimensionsofpreferredargumentstructure.
Thefourconstraintsshowthecorrelationbetweengrammaticalroles,pragmaticinformation,andmorphologicalforms ofarguments.Theseconstraintsnotonlyhypothesizethatthereareupperlimitstothequantityofinformationwhichcan be contained in clauses, but alsosuggest that the distribution of the information shapes an ergative alignment of grammaticalroles.
PASsuggestsauniversalergativepatternofreferentdistribution;thatis,fulllexicalnounphrasesrarelyoccurintheA role,butoverwhelminglyoccurintheSroleortheOrole.Inapragmaticparalleltothis,newinformationfreelyappearsin theSroleortheOrole,butnotintheArole.Inotherwords,SisalignedwithO,asopposedtoA.AlthoughDuBoisfirst derivedPASfromnarrativesinSakapultekMaya,anergativelanguage(DuBois,1985,1987),subsequentresearchhas shownthatitisacross-linguisticphenomenon.ThepatternsofPAShavebeendocumentedinawidearrayoflanguages, bothoftheergative-absolutiveandofthenominative-accusativetypes.TheseincludelanguagessuchasKorean,Nepali, Inuktitut,Finnish,Mapudungun,andRoviana(Clancy,1993,2003;GenettiandCrain,2003;AllenandSchröder,2003; Helasvuo,2003;Arnold,2003;Corston-Oliver,2003).
However,anumberofstudieshavereported counter-evidenceforthe hypothesizedergative patterning(O’Dowd, 1990;Chui,1992;Karkkainen,1996;Kumagai,2006;Lin,2009;Huang,2012).Thedatathesestudiesexaminedwereof typologicallydifferentlanguages(e.g.,EnglishinO’Dowd,1990,andMandarininLin,2009),ofdifferentgenres(e.g., conversation in Karkkainen, 1996, and narrative in Kumagai, 2006), of different information pressures (e.g., high informationpressureinO’Dowd,1990,andlowinformationpressureinHuang,2012),andofdifferenttypesofspeakers (e.g., adult speakers in Chui, 1992, and child speakers in Huang, 2012). These studies showed that the ergative patterningwas not borneout by the languagedata theyexamined. Instead, the data ofthesestudies exhibited an accusativepatternofreferentdistribution;thatis, theSroleandthe Arolewerefoundtocontainconsistentlylower percentagesofnewmentionsandlexicalmentionsthantheOrole.
Inadditiontotheergativepatterning,PASalsohypothesizestheexistenceofuniversalupperlimitsofinformation quantity,assuggestedbytheOneNewArgumentConstraintandtheOneLexicalArgumentConstraint.Thatis,there arelimitsforthenumberofnewmentionsandthenumberoflexicalmentionsusedinaclauseindiscourse.Incontrast to the hypothesized ergative patterning, the hypothesized quantity limits appear to be more robust, supported consistentlybyevidenceacrosslanguages(Clancy,1993,2003;GenettiandCrain,2003;AllenandSchröder,2003; Helasvuo,2003;Arnold,2003;Corston-Oliver,2003).Infact,similarhypothesesconcerninginformationquantityhave beenproposedbyseveralearlierstudies.Forexample,PawleyandSyder(1983,2000)proposedthe‘oneclauseata time’constraint,whichsuggeststhatspeakerscanonlygivethecontentofoneclauseunitatatime.Chafe(1987)
proposedthe‘onerecallatatime’constraint,whichsuggeststhatonlyoneinactivatedconceptcanbeactivatedwithin asingleintonationunit.Similarly,Givo´n(1979)alsoobservedthatspeakersofBantulanguagesrestrictone‘bit’of informationperpropositionorVP.DuBois’OneNewArgumentConstraintappearstotakethematterastepfurther, suggestingthatitisnot simplythe amountof information,butalso thestatus oftheinformation thatissubject to constraint.Theconstraintappearstohaveacognitivebasis.AssuggestedbyDuBois,newinformationusuallyismore difficulttoprocessthangiveninformationsincenewinformationrequiresaspecialcognitiveefforttobringitintoan activatedstate;becauseofthis,themaximumamountofnewinformationtransferredperunitislimited.Inadditionto therestrictionofnewnessintheOne NewArgumentConstraint,theOne LexicalArgumentConstraintsuggestsa restrictiononlexicality.Inotherwords,thetwoquantityconstraintsindicatetwopresumablyuniversalupperlimitsof
Table1
Dimensionsandconstraintsofpreferredargumentstructure.
Grammar Pragmatics
Quantity OneLexicalArgumentConstraint OneNewArgumentConstraint
Role Non-lexicalAConstraint GivenAConstraint
informationquantityintwoparalleldimensions--thelimitofinformationnewnessinthepragmaticdimension,andthe limitofreferentiallexicalityinthegrammaticaldimension.
1.2. Discourseinformationmeasures
Whiletheremaybeupperlimitsofinformationquantity,thereappearstobenogeneralconstraintspecifyingthelower limits.Previousstudieshavealsoinvestigatedinformationquantityintermsoftheentiretyofagivendiscourse,asinthe measureofinformationpressure(DuBois,1987)andthemeasureoflexicalreferentialdensity(Bickel,2003;Stolland Bickel,2009).Thesemeasuresreflectthe extentoftheinformativenessofagivendiscourse.
Informationpressurehastodowiththenumberofnewreferentsintroducedintoadiscourse.Whenthenumberofnew mentionsislarger,theinformationpressureishigher.Differentinformationpressuresalsoreflectthevaryingnumbersof referentstobetrackedintextsofcomparablesize.AsnotedbyDuBois(1987),informationpressureappearstoberelated to discourse genres. Insome genres, informationpressureis oftenhigh whilein others, pressureisoftenlow. For example, 3rd personstories about strangers maypresent a higher informationpressurethan intimate conversation betweenfamilymembers.
Whileinformationpressureisrelatedtonewness,themeasureoflexicalreferentialdensityconcernslexicality.Parallel toinformationpressure,lexicalreferentialdensityhastodowiththefrequencyofthelexicalNPsusedinagivendiscourse (Bickel,2003).SincetheintroductionofanewentityoftenmotivatestheuseofafulllexicalNP,lexicalreferentialdensity alsoappearstoberelatedtodiscoursegenres.Inaddition,previousstudieshaveshownthatlexicalreferentialdensityis alsorelatedtolanguagetypology.Thatis,languagesdifferinhowexplicittheyareinreferencing(Bickel,2003;Stolland Bickel,2009).1Inotherwords,languagesorgenresmaydiffergreatlyinlexicalreferentialdensity.
1.3. Informationmanagement:linguistic,pragmatic,andcognitiveabilities
Thedevelopmentoftheabilitytomanageinformationinvolveslinguistic,pragmatic,andcognitivedevelopmentin children.Asseenabove,importantaspectsrelatedtothelinguisticdomainincludethehypothesizeduniversalpreferred argumentstructure(DuBois,1987,2003),typologicallydifferentreferentialdensity(Bickel,2003;StollandBickel,2009), andlanguage-specificreferentialsystems.Childrenneedtoacquiretheseaspectsforappropriatereferencing.
In thepragmatic domain, previous studieshave shownhowyoung children, likeadults, choosereferentialforms accordingtoseveraldiscourse-pragmaticfactors(Allen,2000;Clancy,1993,1997;Guerrieroetal.,2006;Huang,2011; Narasimhanetal.,2005;Serratrice,2005).Thesefactorsarefeaturesofaccessibility,suchaswhetherthereferents mentionedarepresentinthephysicalcontextofthecurrentconversation,whetherthereferentshavebeenpreviously talkedaboutintheconversation,andwhetherthereferentsareanimate.Thesestudiesshowedthatchildrentendtouse moreinformativeforms(e.g.,lexicalnouns)forreferentsthatareinaccessibleforthefeatures,andlessinformativeforms (e.g.,pronouns,nullforms)forreferentsthatareaccessibleforthefeatures,demonstratingchildren’ssensitivitytothe pragmaticsofreference.
Asforthecognitivedomain,ithasbeensuggestedthatmanaginginformationindiscourserequiresanabilitybythe speakertotakeintoaccountthestateofmindoftheaddressee.Inotherwords,childrenneedtodevelopa‘theoryofmind’ forappropriatereferencing(Gundel,2009;GundelandJohnson,2013;HughesandAllen,2013;RozendaalandBaker, 2010;SalazarOrvigetal.,2010;Skarabela,2007).‘Theoryofmind’referstotheabilitytorealizethatanotherpersonisnot simplyaphysicalbeingbutacognitiveagent,havinghis/herownintentions,beliefs,desires,andperspectivesthatare differentfromone’sown.AssuggestedbyGundelandJohnson(2013),theappropriateuseofreferringformsinvolvesthe abilitytomakereasonableassumptionsabouttheaddressee’smemoryandstateofattentioninrelationtotheintended referent;i.e.,itinvolvestheattributionofmentalstatestoothers.
1.4. Thepresentstudy
Inchildlanguageresearch,itisimportanttoanalyzenotonlychildspeechbutalsomaternalspeechandadultspeech, inordertounderstandnotonlyacquisitiondata,butalsoinputdataandtargetdata.Ithasbeenwidelydocumentedthat maternalspeech(orchild-directedspeechingeneral)differsgreatlyfromadultspeechsincemothersoftenadjusttheir speechwhenaddressingchildren(e.g.,Nelson,1977;Snow,1989).Whilepreviousstudieshavecharacterizedmaternal speechintermsofsyntactic,phonological,orsemanticaspects,littlehasbeendonetoinvestigatethepragmaticaspects
1Mandarinallowsbothsubjectellipsisandobjectellipsis.SinceMandarindoesnothaveinflectionorcasemarkers,thepronominalsystemis
relativelysimple,consistingofpersonalpronounsanddemonstrativepronouns.MandarinNPtypesincludebarenouns,andnounsusedwith
ofmaternalspeech,letalonetheaspectofinformationmanagement.Asforchildspeech,littlehasbeendonetostudy children’sskillsofinformationmanagementattheearlystagesoflanguageacquisition.
ThisstudythusattemptstoinvestigateinformationmanagementinMandarinchildspeech,maternalspeech,andadult speech,focusingonthreeaspects.Thefirstaspectfocusesonthepatternsofthegrammaticalalignmentsinthethree typesofspeech.Asshownabove,oneimportanthypothesisofPASisthatthesubjectofatransitiveverb(A)isoftengiven andnon-lexical,andthatthesubjectofanintransitiveverb(S)isalignedwiththeobjectofatransitiveverb(O)forthe accommodationofnewmentionsand lexicalmentions.Inother words,PASsuggestsanergative S/Oalignmentof referentdistribution.However,myearlierstudyofMandarinchildlanguage(Huang,2012)showedthatMandarinchild speechdemonstratesanaccusativeA/Salignment,ratherthananS/Oalignment,castingdoubtontheuniversalityof PAS.InordertobetterunderstandthepatternobservedinMandarinchildspeech,thisstudyaimstoinvestigatewhether suchgrammaticalalignmentalsocharacterizesMandarinmaternalspeech(theinput)andadultspeech(thetarget).
Thesecondaspectfocusesonthe OneNewArgumentConstraintand theOne LexicalArgumentConstraint.As mentionedearlier,thetwoquantityconstraintsindicatethepresumablyuniversalupperlimitsofinformationquantityin clauses.Huang(2012)showedthatthetwoquantityconstraintswereobservedinMandarinchildspeech.Thisstudy attemptsto further examine whether the constraints alsohold for Mandarin maternalspeech and adultspeech.As mothersusuallyadjusttheirspeechwhenaddressingchildren,itisnotclearwhethertheconstraintscanbeobservedin bothadult-directedspeechandchild-directedspeech.IftheconstraintsholdforMandarinadultspeech,thefindingscan constituteanotherpieceofcross-linguisticevidencefortheuniversalityoftheconstraintsinadultspeech.Iftheconstraints holdforMandarinmaternalspeech,thismayindicatethattheconstraintsareratherrobust:Eveninadjustedchild-directed speech,mothers still followthe constraints of quantity. In other words,this studyattempts to examine whether the constraintsaresupportedby dataofdifferentspeechtypes.Therefore,thesecond focusofthisstudyistoexamine whetherthetwoconstraintscanbeobservedinMandarinchildspeech,maternalspeech,andadultspeech.
The third aspect focuses on comparing Mandarin child speech, maternal speech, and adult speech using two discourseinformationmeasures:themeasureofinformationpressureandthemeasureoflexicalreferentialdensity.Itis speculatedthatthesethreetypesofspeechwilldiffersignificantlyaccordingtothesetwotypesofmeasures,andthatthe differences will reflect children's developing linguistic ability, mothers’ speech adjustments, and adults’ mature competence. In other words, the analyses of information pressure and lexical referential density may reveal the characteristicsofthethreetypesofspeechininformationmanagement.
2. Methods
2.1. Participantsanddata
Thisstudyanalyzedthreetypesofspeech:childspeech,maternalspeech,andadultspeech.Thechildparticipants weretwoMandarin-speakinggirlswhilethemotherparticipantswerethechildren’smothers,whowerenativespeakersof Mandarin.ThetwofamilieslivedinthenorthernpartofTaiwan.Thedatacollectedconsistedofeighthoursofnatural mother--child conversation video-recorded in the participants’ homes, with four one-hour sessions with each dyad recordedwhenthechildrenwereattheagesof2;2,2;6/2;7,2;10,and3;1.Themotherswereattheageof38and39years, respectivelyatthetimeofdatacollection.Allofthedatawerecollectedinthelivingroomsofthetwohomes.
Thechilddataanalyzedinthisstudyincluded thechildren’sspeechproducedduringthe totaleighthoursofdata sessions.Asforthematernaldata,sincethemothersproducedmanymoreutterancesthanthechildren,20minofmaternal speechfromeachoftheeightsessions,i.e.,160minintotal,wereusedforanalysis.Thetotalnumbersofthecodedclauses inthechilddataandinthematernaldata(seethecodingschemebelow)were2419and3351,respectively.
The adult data were adopted from the NCCU Corpus of Spoken Chinese (Chui and Lai, 2008). The adult participantswere threedyads of Mandarin-speakingadultseach with a closerelationship: Oneof thedyads was composedofawife(55yearsold)andahusband(56yearsold);anotherdyadwascomposedofagirlfriend(25years old)andaboyfriend(25yearsold),andtheotherdyadwasconstitutedoftwofemalefriends(24and25yearsold). Eachdyadwasinvolvedinspontaneousconversationforabout20min.Inotherwords,theadultdataanalyzedinthe studyconsistedof60minofspeechproducedbythesixadults.Thetotalnumberofthecodedclausesintheadult datawas1688.
2.2. Codingscheme
Acodingschemewasadoptedtoanalyzethegrammaticalalignmentsandquantityconstraints.Eachclausewhich containedanovertfiniteverbinthedatawasidentifiedforanalysis.Thecoreargumentsofeachoftheseverbs,including nullarguments,werecodedforgrammaticalroles,informationstatuses,andreferentialforms.Thecodingschemeofthis studyisasfollows:
(1)Grammaticalroles
(a)TheArole:Itisthesubjectofagrammaticallytransitiveclause(e.g.,thepronounwo‘I’inwozaihuameiguihua‘I amdrawingarose’).
(b)TheOrole:Itisthe objectofagrammaticallytransitiveclause(e.g.,thenounmeiguihua ‘rose’in wozaihua meiguihua‘Iamdrawingarose’).
(c)TheSrole:Itisthesingleargumentofagrammaticallyintransitiveclause(e.g.,thenounmeimei‘sister’inmeimei zaiku‘Sisteriscrying’).
(2)Informationstatuses
(a)New:Anargumentisconsideredtobenewifthereferentitdenoteshasnotbeenmentionedinthepreceding 20utterances(Allen,2000;Chafe,1976,1987;DuBois,1987).
(b)Given:An argumentisconsidered tobegivenifthereferentitdenoteshasbeenmentionedin thepreceding 20utterances(Allen,2000;Chafe,1976,1987;DuBois,1987).
(3)Referentialforms
(a)Lexicalforms:Lexicalformsincludebarenouns(e.g.,mao‘cat’),nounphrases(e.g.,hongsedehua‘redflowers’), andpropernames(e.g.,YimingShushu‘UncleYiming’)
(b)Non-lexical forms: Non-lexical forms include null forms and pronominalforms (e.g., the pronoun wo ‘I’, the demonstrativezhe‘this’)
Thedatawerecodedbyatrainedresearchassistant,whowasanativespeakerofMandarinandagraduatestudentof linguistics.Inaddition,onefourthofthedatawereindependentlycodedbyanothertrainedresearchassistant,whowas alsoanativespeakerofMandarinandagraduatestudentoflinguistics.Cohen'sKappawasusedtodeterminethe inter-raterreliabilities.Thereliabilitiesforgrammaticalroles,informationstatuses,andreferentialformswere0.84,0.83,and 0.92,respectively.
2.3. Discourseinformationmeasures
Twotypesofdiscourseinformationmeasureswereincludedinthisstudy:(1)themeasureofinformationpressure,and (2)themeasureoflexicalreferentialdensity.Thedetailsofthesetwotypesofmeasuresarepresentedbelow. (1)Informationpressure(IP):Informationpressurehastodowiththenumberofnewreferentsintroducedintoadiscourse.
Thevalueoftheinformationpressureisobtainedbycalculatingtheratioofthenewcoreargumentstothetotalcore argumentsinthedata(DuBois,1987;Durie,2003),asshownbelow.
Informationpressure ðIPÞ¼# of newcorearguments # oftotalcorearguments
(2)Lexicalreferentialdensity(LRD):LexicalreferentialdensityhastodowiththefrequencyofthelexicalNPsusedina givendiscourse.Thevalueofthe lexicalreferentialdensityisobtainedbycalculatingthe ratioofthe lexicalcore argumentstothetotalcoreargumentsinthedata(Bickel,2003;StollandBickel,2009),asshownbelow.
Lexicalreferentialdensity ðLRDÞ¼# oflexicalcorearguments # oftotalcorearguments 3. Results
3.1. Grammaticalalignments
Thissectionpresentstheanalysisofthegrammaticalalignmentsinchildspeech,maternalspeech,andadultspeech. Morespecifically,weanalyzedthedistributionsofnew/giveninformationandlexical/non-lexicalreferentialformsinthe core arguments of the three types of speech. We attempted to examine whether similar patterns of grammatical alignmentscanbeobservedinthethreetypesofspeech.
3.1.1. Thedistributionsofnew/giveninformation
Table2apresentsthedistributionofthenewandgivenmentionswithineachgrammaticalroleinthechilddata.The tableshowsthatgivenmentionsoccurredmorefrequentlythannewmentionsinallofthegrammaticalroles;however, noticeabledifferenceswereobservedinthedistributionsofthenewandgivenmentionsinthethreegrammaticalroles.
Inordertoexaminewhetherthedifferenceswerestatisticallysignificant,aChi-squareanalysiswasconducted.Theresult showedthatthe distributionsofthenewand givenmentions weresignificantlydifferent acrosstheA,S,andO roles (x2
=308.18,d.f.=2, p<.001),suggesting thatthechildren’suseofnew andgivenmentionswasinfluencedby the grammaticalroles.
Aposthocresidualanalysiswasfurtherconducted2inordertoidentifythecontributionsmadebythedifferentcellsto
the resultingChi-square value.Asshownin Table 2b,the figuresin all of the cellsreachedstatistical significance, indicatingthatallofthecellsmadesignificantcontributionstotheresultoftheChi-squaretest.TheAroleandtheSrole weresignificantlylesslikelytoaccommodatenewmentions,whiletheOrolewassignificantlymorelikelytocontainnew mentions.Incontrast,theAroleandtheSroleweresignificantlymorelikelytoaccommodategivenmentions,whilethe Orolewassignificantlylesslikelytocontaingivenmentions.TheresultsrevealedthattheSrolealignsitselfmoreclosely withtheArolethanitdoeswiththeOroleinchildspeech.
Table3apresentsthedistributionofthenewandgivenmentionswithineachgrammaticalroleinthematernaldata.As seeninthetable,givenmentionsoccurredmorefrequentlythannewmentionsinallofthegrammaticalroles;however,the distributionsofthenewandgivenmentionsappeartopresentdifferentpatternsinthethreegrammaticalroles.AChi-square analysiswasfurtherconductedinordertoexaminewhetherthedifferenceswerestatisticallysignificant.Theresultshowed thatthedistributionsofthenewandgivenmentionsweresignificantlydifferentacrosstheA,S,andOroles(x2=197.05, d.f.=2,p<.001),suggestingthatthemothers’useofnewandgivenmentionswasinfluencedbythegrammaticalroles. Aposthocresidualanalysiswasconducted.Table3bindicatesthatthefiguresinallofthecellsreachedstatistical significance,indicatingthatallofthecellsmadesignificantcontributionstotheresultoftheChi-squaretest.TheAroleand theSroleweresignificantlylesslikelytobenew,whiletheOrolewassignificantlymorelikelytobenew.Incontrast,theA roleandtheSroleweresignificantlymorelikelytobegiven,whiletheOrolewassignificantlylesslikelytobegiven.The resultsrevealedthattheSrolealignsitselfmorecloselywiththeArolethanitdoeswiththeOroleinmaternalspeech,a patternsimilartothatobservedinchildspeech.
Asforadultspeech,Table4ashowsthatwhilegivenmentionsoccurredmorefrequentlythannewmentionsinallof thegrammaticalroles,thedistributionsofthenewandgivenmentionsappeartopresentdifferentpatternsinthethree grammaticalroles.AChi-squareanalysiswasalsoconducted,andtheresultshowedthatthedistributionsofthenewand givenmentionsweresignificantlydifferentacrossthethreegrammaticalroles(x2=218.51,d.f.=2,p<.001),suggesting thattheadults’useofnewandgivenmentionswasinfluencedbythegrammaticalroles.
Aposthocresidualanalysiswasfurtherconducted. Asshownin Table4b,the figuresinall ofthecellsreached statisticalsignificance,indicatingthatallofthecellsmadesignificantcontributionstotheresultoftheChi-squaretest.The AroleandtheSroleweresignificantlylesslikelytobenew,whiletheOrolewassignificantlymorelikelytobenew.In contrast,theAroleandtheSroleweresignificantlymorelikelytobegiven,whiletheOrolewassignificantlylesslikelyto
Table2a
Thedistributionofnew/giveninformationwithineachgrammaticalroleinthechilddata.
Child A S O x2 N % N % N % New 28 2.16 109 9.71 316 24.38 308.18*** Given 1268 97.84 1014 90.29 980 75.62 Total 1296 100 1123 100 1296 100 ***p<.001. Table2b
Posthocresidualanalysisforthechilddata.
Child A S O
New 13.75 3.15 16.6~
Given 13.7~ 3.1~ 16.65
Note: 5,significantlylow;~,significantly high: p<.05 ifjadjusted residualj>1.96;p<.01 if jadjusted
residualj>2.58;p<.001ifjadjustedresidualj>3.29.
2Aresidualisthedifferencebetweentheobservedandexpectedvaluesforacell.Aresidualanalysisidentifiesthosespecificcellsmakingthe
begiven.TheresultsrevealedthattheSrolealignsitselfmorecloselywiththeArolethanitdoeswiththeOroleinadult speech,apatternsimilartothoseobservedinchildspeechandinmaternalspeech.
Insum,anA/SalignmentwasobservedinthedistributionsofnewandgiveninformationinMandarinchildspeech, maternalspeech,andadultspeech.ThepatternwasinconsistentwiththeS/OalignmentsuggestedbyPAS. 3.1.2. Thedistributionsofreferentialforms
Thereferentialforms usedinthe child data,the maternaldata,and theadultdatawerealsoanalyzed.Table 5a
presentsthedistributionofthelexicalandnon-lexicalmentionswithineachgrammaticalroleinthechilddata.Thetable showsthatinthechilddata,non-lexicalformswereusedmorefrequentlythanlexicalformsinallofthegrammaticalroles; however,noticeabledifferenceswereobservedinthedistributionsofthelexicalandnon-lexicalmentionsinthethree grammaticalroles.Inordertoexaminewhetherthedifferenceswerestatisticallysignificant,aChi-squareanalysiswas conducted.Theresultshowedthatthedistributionsofthelexicalandnon-lexicalmentionsweresignificantlydifferent acrosstheA,S,andOroles(x2=514.56,d.f.=2,p<.001),suggestingthatthechildren’suseofreferentialformswas influencedbythegrammaticalroles.
Aposthocresidualanalysiswasfurtherconductedinordertoidentifythecontributionsmadebythedifferentcellsto the resulting Chi-square value.Asshownin Table 5b, the figuresin all ofthe cellsreached statisticalsignificance,
Table3a
Thedistributionofnew/giveninformationwithineachgrammaticalroleinthematernaldata.
Maternal A S O x2 N % N % N % New 20 1.31 91 4.99 206 13.47 197.05*** Given 1509 98.69 1731 95.01 1323 86.53 Total 1529 100 1822 100 1529 100 ***p<.001. Table3b
Posthocresidualanalysisforthematernaldata.
Maternal A S O
New 9.95 3.35 13.4~
Given 9.9~ 3.3~ 13.45
Note: 5,significantly low;~, significantly high:p<.05if jadjusted residualj>1.96;p<.01 if jadjusted
residualj>2.58;p<.001ifjadjustedresidualj>3.29.
Table4a
Thedistributionofnew/giveninformationwithineachgrammaticalroleintheadultdata.
Adult A S O x2 N % N % N % New 22 3.02 132 13.75 225 30.91 218.51*** Given 706 96.98 828 86.25 503 69.09 Total 728 100 960 100 728 100 ***p<.001. Table4b
Posthocresidualanalysisfortheadultdata.
Adult A S O
New 11.25 2.15 13.5~
Given 11.2~ 2.1~ 13.55
Note: 5,significantly low;~, significantly high:p<.05if jadjusted residualj>1.96;p<.01 if jadjusted
indicatingthatallofthecellsmadesignificantcontributionstotheresultoftheChi-squaretest.TheAroleandtheSrole weresignificantlylesslikelytoaccommodatelexicalmentions,whiletheOrolewassignificantlymorelikelytocontain lexicalmentions.Incontrast,theAroleandtheSroleweresignificantlymorelikelytoaccommodatenon-lexicalmentions, whiletheOrolewassignificantlylesslikelytocontainnon-lexicalmentions.TheresultsrevealedthattheSrolealigns itselfmorecloselywiththeArolethanitdoeswiththeOroleinchildspeech.
Table6apresentsthedistributionofthelexicalandnon-lexicalmentionswithineachgrammaticalroleinthematernal data.Asseeninthetable,non-lexicalmentionsoccurredmorefrequentlythanlexicalmentionsinallofthegrammatical roles;however,thedistributionsofthelexicalandnon-lexicalmentionsappeartopresentdifferentpatternsinthethree grammatical roles. A Chi-square analysis wasfurther conductedin orderto examine whether the differences were statisticallysignificant.Theresultshowedthatthedistributionsofthelexicalandnon-lexicalmentionsweresignificantly differentacrosstheA,S,andOroles(x2=412.42,d.f.=2,p<.001),suggestingthatthemothers’useoflexicaland non-lexicalmentionswasinfluencedbythegrammaticalroles.
Aposthocresidualanalysis wasconducted.Table 6bshowsthatthefiguresinallofthe cellsreachedstatistical significance,indicatingthatallofthecellsmadesignificantcontributionstotheresultoftheChi-squaretest.TheAroleand theSroleweresignificantlylesslikelytobelexical,whiletheOrolewassignificantlymorelikelytobelexical.Incontrast,
Table6a
Thedistributionoflexical/non-lexicalformswithineachgrammaticalroleinthematernaldata.
Maternal A S O x2 N % N % N % Lexical 238 15.57 326 17.89 673 44.02 412.42*** Non-lexical 1291 84.43 1496 82.11 856 55.98 Total 1529 100 1822 100 1529 100 ***p<.001. Table6b
Posthocresidualanalysisforthematernaldata.
Maternal A S O
Lexical 10.65 9.25 20.2~
Non-lexical 10.6~ 9.2~ 20.25
Note: 5,significantlylow;~,significantly high; p<.05 ifjadjusted residualj>1.96;p<.01 if jadjusted
residualj>2.58;p<.001ifjadjustedresidualj>3.29.
Table5a
Thedistributionoflexical/non-lexicalformswithineachgrammaticalroleinthechilddata.
Child A S O x2 N % N % N % Lexical 94 7.25 156 13.89 542 41.82 514.56*** Non-lexical 1202 92.75 967 86.11 754 58.18 Total 1296 100 1123 100 1296 100 ***p<.001. Table5b
Posthocresidualanalysisforthechilddata.
Child A S O
Lexical 15.35 7.35 22.3~
Non-lexical 15.3~ 7.3~ 22.35
Note: 5,significantlylow;~,significantly high; p<.05 ifjadjusted residualj>1.96;p<.01 if jadjusted
theAroleandtheSroleweresignificantlymorelikelytobenon-lexical,whiletheOrolewassignificantlylesslikelytobe non-lexical.TheresultsrevealedthattheSrolealignsitselfmorecloselywiththeArolethanitdoeswiththeOrolein maternalspeech,apatternsimilartothatobservedinchildspeech.
Asforadultspeech,Table7ashowsthatwhilenon-lexicalmentionsoccurredmorefrequentlythanlexicalmentions in allofthegrammaticalroles, thedistributionsof thelexicalandnon-lexicalmentionsappear topresent different patterns in the three grammatical roles. A Chi-square analysis was also conducted, and the result showed that the distributions of the lexical andnon-lexical mentions were significantly different across the three grammatical roles(x2=397.88,d.f.=2,p<.001),suggestingthattheadults’useoflexicalandnon-lexicalmentionswasinfluenced bythegrammaticalroles.
Aposthocresidualanalysiswasfurtherconducted.Table7bshowsthatitwasmainlytheAroleandtheSrolethat contributedtotheresultingvalueoftheChi-squaretest.Asshowninthetable,theArolewassignificantlylesslikelytobe lexical,whiletheOrolewassignificantlymorelikelytobelexical.Incontrast,theArolewassignificantlymorelikelytobe non-lexical,whiletheOrolewassignificantlylesslikelytobenon-lexical.ThedistributionintheSrole,however,didnot reachsignificance.
Insum,anA/Salignmentwasobservedinthedistributionsofthelexicalandnon-lexicalinformationinMandarinchild speechandmaternalspeech.Adultspeech,however,didnotdemonstrateasignificantpatternofalignment.
Thissectionhasanalyzedthedistributionsofinformationstatusesandreferentialformsinchild,maternal,andadult speech.Theresultsshowedthatanergativepatternwasnotobservedinthethreetypesofspeech,castingdoubtonthe universalityofPAS.Instead,Mandarin-speakingchildren,mothers,andadultstendtoaligntheSrolewiththeArole, ratherthanwiththeOrole,demonstratinganaccusativealignmentintheirspeech.
Interestingly, although the three typesof speech demonstrated similar patternsin grammatical alignment, some noticeable differenceswere alsoobservedin the distributions.Forexample, itappears that the proportionof given referentsishigherformaternalspeechthanforeitherchildoradultspeech,ascanbeseeninTables2a,3aand4a.In addition,itappearsthattheproportionofnon-lexicalmentionsisconsiderablylowerforadultspeechthanforeitherchildor maternalspeech,ascanbeseeninTables5a,6aand7a.Furtheranalysisconcerningthedifferencesbetweenthethree typesofspeechispresentedinSection3.3.
3.2. Quantityconstraints
Thissectionpresentstheanalysisofinformationquantityinchildspeech,maternalspeech,andadultspeech.More specifically,wetestedtheOneNewArgumentConstraintandtheOneLexicalArgumentConstraintproposedbyDuBois (1987)byanalyzingthenumbersofnew/givenmentionsandthenumbersoflexical/non-lexicalmentionsintheclausesof thethreetypesofdata.Theseconstraintsindicatethepresumablyuniversalupperlimitsofinformationquantitycontained inclauses.WeattemptedtoexaminewhethersuchconstraintscanbeobservedinMandarinchildspeech,maternal speech,andadultspeech.
Table7a
Thedistributionoflexical/non-lexicalformswithineachgrammaticalroleintheadultdata.
Adult A S O x2 N % N % N % Lexical 70 9.62 325 33.85 431 59.20 397.88*** Non-lexical 658 90.38 635 66.15 297 40.80 Total 728 100 960 100 728 100 ***p<.001. Table7b
Posthocresidualanalysisfortheadultdata.
Adult A S O
Lexical 16.75 0.3n.s. 17~
Non-lexical 16.7~ 0.3n.s. 175
Note: 5,significantly low;~, significantly high;p<.05if jadjusted residualj>1.96;p<.01 if jadjusted
3.2.1. OneNewArgumentConstraint
TheOneNewArgumentConstraintindicatesthateachclausecontainsnomorethanoneargumentcarryingnew information.Table8displaysthefrequencyoftheclauseswithzero,one,ortwonewargumentsinthedata.Asseeninthe table,thepercentagesoftheclauseswithtwonewargumentsinthethreetypesofdatawereverylow.Theywere0.66%, 0.18%,and0.71%inthechilddata,thematernaldata,andtheadultdata,respectively.Themajorityoftheclausesinthe datacontained nonewarguments atall (81.93%, 90.72%, and78.26%). The results thusconfirmed the OneNew ArgumentConstraint.Inother words,the children,the mothers,and theadultsall tendedtoproduceclauses which containednomorethanonenewargument.
Sinceonlytransitiveclausescanhavemorethanonecoreargument,furtheranalysiswasconductedtoanalyzethe clausesinthedataintermsoftransitivity.AsseeninTable9,only1.23%,0.39%,and1.65%ofthetransitiveclausesinthe threetypesofspeechhadtwonewarguments.Mostofthetransitiveclausesinthedatahadeitheroneorzeronew argument,andthepercentagesoftheclauseswithzeronewargumentweremuchhigherthanthoseoftheclauseswith onenewargumentinthethreetypesofspeech.Asfortheintransitiveclauses,onlyafewintransitiveclauseshadonenew argument;theoverwhelmingmajorityoftheintransitiveclauseshadzeronewargument(90.29%,95.01%,and86.25%). TheresultsthusconfirmedtheOneNewArgumentConstraint.Inotherwords,thechildren,themothers,andtheadultsall tendedtoproduceclauseswhichcontainednomorethanonenewargument.
3.2.2. OneLexicalArgumentConstraint
TheOneLexicalArgumentConstraintclaimsthateachclausecontainsnomorethanonelexicalargument.Table10
presentsthefrequencyoftheclauseswithzero,one,ortwolexicalargumentsinthedata.Asseeninthe table,only 1.65%,3.04%,and2.61%oftheclausesinthethreetypesofspeechcontainedtwolexicalarguments.Themajorityofthe
Table8
Frequencyofclauseswithzero,one,ortwonewarguments.
New Child Maternal Adult
N % N % N % 0 1982 81.93 3040 90.72 1321 78.26 1 421 17.40 305 9.10 355 21.03 2 16 0.66 6 0.18 12 0.71 Total 2419 100 3351 100 1688 100 Table9
Transitivityandfrequencyofclauseswithzero,one,ortwonewarguments.
New Child Maternal Adult
Transitive Intransitive Transitive Intransitive Transitive Intransitive
N % N % N % N % N % N % 0 968 74.69 1014 90.29 1309 85.61 1731 95.01 493 67.72 828 86.25 1 312 24.07 109 9.71 214 14.00 91 4.99 223 30.63 132 13.75 2 16 1.23 0 0.00 6 0.39 0 0.00 12 1.65 0 0.00 Total 1296 100 1123 100 1529 100 1822 100 728 100 960 100 Table10
Frequencyofclauseswithzero,one,ortwolexicalarguments.
Lexical Child Maternal Adult
N % N % N %
0 1667 68.91 2216 66.13 906 53.67
1 712 29.43 1033 30.83 738 43.72
2 40 1.65 102 3.04 44 2.61
clausesinthedatahadnolexicalargumentsatall(68.91%,66.13%,and53.67%).TheresultsthussupportedtheOne LexicalArgumentConstraint.Inotherwords,thechildren,themothers,andtheadultsalltendedtoproduceclauseswhich containednomorethanonelexicalargument.
Furtheranalysiswasconductedtoanalyzetheclausesinthedataintermsoftransitivity.Theresultsareshownin
Table11.Asseeninthetable,only3.09%,6.67%,and6.04%ofthetransitiveclausesinthethreetypesofspeechhadtwo lexicalarguments.Mostofthetransitiveclausesinthedatahadeitheroneorzerolexicalargument.Asfortheintransitive clauses,whileafewoftheintransitiveclauseshadonelexicalargument,thevastmajorityoftheintransitiveclauseshad zerolexicalargument(86.11%,82.11%,and66.15%).TheresultsthussupportedtheOneLexicalArgumentConstraint. Inotherwords,thechildren,themothers,andtheadultsalltendedtoproduceclauseswhichcontainednomorethanone lexicalargument.
Inshort,thetwoconstraintsweresupportedbytheresultsofthepresentstudy.Itwasnotablyrarethatthe Mandarin-speakingchildren,mothers,oradultsproducedclauseswithmorethanonenewargumentorwithmorethanonelexical argument.Inotherwords,theseconstraintsapplynotonlytoadultspeech,assuggestedbypreviousstudies,butalsoto maternalspeechandtochildspeech.Theupperlimitsofinformationquantitycanbeobservedinallofthethreetypesof speech.
WhileallofthethreetypesofspeechfollowedtheOneNewArgumentConstraintandthe OneLexicalArgument Constraint,itisworthnotingthatthedistributionspresentedinthetablesabovealsorevealsomedifferencesbetweenthe threetypesofspeech.Forexample,Tables8and9showthatmaternalspeechappearstohaveahigherproportionof clauseswithzeronewargumentcomparedtotheothertwotypesofspeech.Inaddition,Tables10and11showthatadult speechappearstohavefewerclauseswithzerolexicalargumentsthantheothertwotypesofspeech.Furtheranalysis concerningthedifferencesbetweenthethreetypesofspeechispresentedinSection3.3.
3.3. Discourseinformationmeasures
Asseenabove,whilechildspeech,maternalspeech,andadultspeechdemonstratedsimilarpatternsintheanalyses ofgrammaticalalignmentsandquantityconstraints,acloserlookatthethreetypesofspeechrevealedsomenoticeable differencesintheproportionsofnewinformationandlexicalinformation.Inordertoobtainamorecompletepicture,this sectionpresentstheanalysisoftwotypesofdiscourseinformationmeasures.Morespecifically,wemeasuredthevalues ofinformationpressure(DuBois,1987;Durie,2003)andthevaluesoflexicalreferentialdensity(Bickel,2003;Stolland Bickel,2009)forthe threetypesofspeech.
3.3.1. Informationpressure(IP)andlexicalreferentialdensity(LRD)
TheIPvaluesforchild,maternal,andadultspeechareshowninTable12.Thethreevaluesweresignificantlydifferent (x2=164.62,d.f.=2,p<.001).Aposthocmultiplecomparisontest(MarascuiloandMcSweeney,1977)showedthatthe
Table11
Transitivityandfrequencyofclauseswithzero,one,ortwolexicalarguments.
Lexical Child Maternal Adult
Transitive Intransitive Transitive Intransitive Transitive Intransitive
N % N % N % N % N % N % 0 700 54.01 967 86.11 720 47.09 1496 82.11 271 37.23 635 66.15 1 556 42.90 156 13.89 707 46.24 326 17.89 413 56.73 325 33.85 2 40 3.09 0 0.00 102 6.67 0 0.00 44 6.04 0 0.00 Total 1296 100 1123 100 1529 100 1822 100 728 100 960 100 Table12
Informationpressure(IP)inchildspeech,maternalspeechandadultspeech.
Child Maternal Adult
IP(new/total) 0.12(453/3715) 0.06(317/4880) 0.16(379/2416)
x2 164.62***
Posthoc M<C,C<A,M<A(M<C<A)
IPvalueformaternalspeechwassignificantlylowerthanthatforeitherchildoradultspeech,andthattheIPvalueforchild speechwassignificantlylowerthanthatforadultspeech.Inotherwords,adultspeechcontainedthelargestproportionof newmentions,whilematernalspeechcontainedthesmallestproportionofnewmentions.
TheLRDvaluesforchild,maternal,andadultspeechareshowninTable13.Thethreevalueswerealsosignificantly different(x2=127.79,d.f.=2,p<.001).Aposthocmultiplecomparisontest(MarascuiloandMcSweeney,1977)showed thattheLRDvalueforchildspeechwassignificantlylowerthanthatforeithermaternaloradultspeech,andthattheIP valueformaternalspeechwassignificantlylowerthanthatforadultspeech.Inotherwords,adultspeechcontainedthe largestproportionoflexicalmentions,whilechildspeechcontainedthesmallestproportionoflexicalmentions.
AsseeninTables12and13,adultspeechhadboththehighestIPandthehighestLRDwhencomparedwiththeother twotypesofspeech.Inotherwords,adultspeechprovidednewmentionsandlexicalformsmorefrequentlythanmaternal speechorchildspeech.Interestingly,theresultsshowedthatchildspeechhadahigherIPthanmaternalspeech,butthat maternalspeechhadahigherLRDthanchildspeech.Inotherwords,thechildrenweremorelikelythantheirmothersto providenewinformation,whilethemothersweremorelikelythanthechildrentosupplylexicalforms.
Tosummarize,theIPsandtheLRDsforthethreetypesofspeech,whichindicatetheproportionsofnewmentionsand lexicalmentionsinthedata,canbedemonstratedinTable14.Adultspeech,notsurprisingly,wascharacterizedbythe highestIPandthehighestLRD;maternalspeechwascharacterizedbythelowestIPandamiddleLRD,andchildspeech, ontheotherhand,wascharacterizedbyamiddleIPandthelowestLRD.
3.3.2. Newmentionsandlexicalmentions
Tofurtherinvestigatehowthedistributionsofnewmentionsandlexicalmentionsdifferedinthethreetypesofspeech, analysiswasconductedtoobtaintheratiosofnewmentionstolexicalmentionsinthethreetypesofspeech,asshownin
Table15.Theresultsshowedthatthereweremorelexicalmentionsthannewmentionsinallofthethreetypesofspeech. However,thedifferencebetweenthenumberoflexicalmentionsandthenumberofnewmentionswaslargerinmaternal speechthanintheothertwotypesofspeech.Themothersproducedaboutfourtimesasmanylexicalmentionsasthey didnewmentions;theadultsproducedmorethantwiceasmanylexicalmentionsastheydidnewmentions,andthe childrenproducedlessthantwiceasmanylexicalmentionsastheydidnewmentions.
While ithasoftenbeen suggestedthat the association between newinformationand lexical mentions isstrong (DuBois,1987;Clancy,2003),Table15showsthatthisassociationisonlypartial,andthatthestrengthoftheassociation mayvaryaccordingtothetypeofspeech.Sincethemothersproducedconsiderablymorelexicalmentionsthannew mentions,wesuspectthatitmaybemorelikelyforthemothersthanforthechildrenortheadultstouselexicalformsfor referentswhich weregiven.Furtheranalysis wasconductedto investigatethe proportions oflexical mentionsused
Table14
SummarytableofIPandLRD.
Child Maternal Adult
IP Middle Lowest Highest
LRD Lowest Middle Highest
Table13
Lexicalreferentialdensity(LRD)inchildspeech,maternalspeechandadultspeech.
Child Maternal Adult
LRD(lexical/total) 0.21(792/3715) 0.25(1237/4880) 0.34(826/2416)
x2 127.79***
Posthoc C<M,C<A,M<A(C<M<A)
***p<.001.
Table15
Ratiooflexicalmentionstonewmentionsinchildspeech,maternalspeech,andadultspeech.
Child Maternal Adult
Newmentions 453 317 379
Lexicalmentions 792 1237 826
forreferringtonewreferentsandforreferringtogivenreferentsinthethreetypesofspeechinordertoexaminewhether maternalspeechwascharacterizedbyahigherfrequencyofgivenlexicalmentions.TheresultsareshowninTable16. AsshowninTable16,thedistributionsinthethreetypesofspeechweresignificantlydifferent(x2=166.93,d.f.=2, p<.001).Aposthocmultiplecomparisontest(MarascuiloandMcSweeney,1977)showedthatthepercentageofgiven lexical mentions in adult speech was significantly lower than that in either child or maternal speech, and that the percentage in child speech wassignificantly lower than that in maternal speech. In other words, maternal speech containedthelargestproportionofgivenlexicalmentions,whileadultspeechcontainedthesmallestproportionofgiven lexicalmentions.Itappearsthatmaternalspeechwascharacterizedbyahigherfrequencyofgivenlexicalmentions comparedwiththeothertwotypesofspeech.
Excerpt1demonstrateshowalexicalformwasusedtorefertoagivenreferentinmaternalspeech.Asseeninthis example,themotherusedthelexicalform‘Barbie’severaltimestorefertoadoll,whichwasagivenreferentinthecontext. Excerpt13
*MOT: a nide Babi lei? PRT your Barbie PRT ‘(Where’s)yourBarbie?’’ *CHI: en.
‘Um.’
*MOT: Babi tang zai dishang shi-bu-shi? Barbie lie on floor right ‘Barbieislyingonthefloor,right?’ *CHI: en.
‘Um.’
*MOT: ba Babi jian qilai. BA Barbie pick up ‘PickBarbieup.’
In additiontothe analysisof givenlexicalmentions,a relatedanalysis istoexamine the useofnon-lexicalnew mentionsinthethreetypesofspeech.Analysiswasconductedtoexaminetheproportionsofnewreferentswhichwere mentionedwithlexicalformsandwithnon-lexicalformsinthethreetypesofspeech.TheresultsareshowninTable17.As seeninTable17,thedistributionsofthelexicalformsandnon-lexicalformsinthethreetypesofspeechweresignificantly
Table17
Theproportionsofnewreferentsmentionedwithlexicalformsandwithnon-lexicalforms.
New Child Maternal Adult x2 Posthoc
N % N % N % Lexical 288 63.58 231 72.87 366 96.57 131.20*** C<M,C<A,M<A(C<M<A) Non-lexical 165 36.42 86 27.13 13 3.43 M<C,A<C,A<M(A<M<C) Total 453 100 317 100 379 100 *** p<.001. Table16
Theproportionsoflexicalmentionsusedfornewreferentsandforgivenreferents.
Lexical Child Maternal Adult x2 Posthoc
N % N % N % New 288 36.36 231 18.67 366 44.31 166.93*** M<C,C<A,M<A(M<C<A) Given 504 63.64 1006 81.33 460 55.69 C<M,A<C,A<M(A<C<M) Total 792 100 1237 100 826 100 ***p<.001. 3
different(x2=131.20,d.f.=2,p<.001).Aposthocmultiplecomparisontest(MarascuiloandMcSweeney,1977)showed thatthepercentageofnon-lexicalnewmentionsinadultspeechwassignificantlylowerthanthatineithermaternalorchild speech,andthatthepercentageinmaternalspeechwassignificantlylowerthanthatinchildspeech.Inotherwords,child speech contained the largest proportion of non-lexical new mentions, while adult speech contained the smallest proportionofnon-lexicalnewmentions.Itappearsthatchildspeechwascharacterizedbyahigherfrequencyof non-lexicalnewmentionscomparedwiththeothertwotypesofspeech.
Excerpt2demonstrateshowanewreferentwasmentionedinchildspeechwithanon-lexicalform,anullforminthis case.Asseeninthisexample,thechildusedanullformwhenintroducinganewreferent.Themotherrespondedwitha questionforclarification.
Excerpt2
*CHI: mama wo yao chi Ø. Mommy I want eat Ø ‘Mommy,IwanttoeatØ.’ *MOT: chi sheme?
eat what
‘Whatdoyouwanttoeat?’
Insum,theanalysesshowedthatmaternalspeechwascharacterizedbyafrequentuseofgivenlexicalmentions;child speechwascharacterizedbyafrequentuseofnon-lexicalnewmentions.Adultspeech,ontheotherhand,demonstrated astrongassociationbetweennewreferentsandlexicalforms.
4. Discussionandconclusion
ThisstudyhaspresentedananalysisofinformationmanagementinMandarinchildspeech,maternalspeech,and adultspeech.Similarities and differenceswere foundbetween the three typesof speechin relationto grammatical alignments,quantityconstraints,anddiscourseinformationmeasures.
TheresultsshowedthatratherthandemonstratinganS/Oalignment,aspredictedbyPAS,thethreetypesofspeech demonstratedanA/Salignment,revealingthatMandarinischaracterizedbyanaccusativepattern,andalsothatchildren aresensitivetothispatternearlyon.Theresultssuggestedthatanergativepatternofgrammaticalalignmentmaynotbe linguisticallyuniversal.
Asfor the quantityconstraints, this study showed that the One New ArgumentConstraint and the One Lexical ArgumentConstraintare supported by Mandarin child speech,maternal speech,and adultspeech. While previous studieshaveshownthatthequantityconstraintsaresupportedbyadultspeechcross-linguistically,thisstudyprovided evidencethattheseconstraintscanalsobeobservedinchildspeechatthe earlystages oflanguageacquisition.In addition,whilematernalspeechwascharacterizedbyalargerproportionoflexicalmentionsthanchildspeechandadult speech, the results demonstrated that the mothers alsofollow the upper limits of lexical quantity indicated by the constraints.
Whilethethreetypesofspeechshowedsimilarpatternsingrammaticalalignmentsandquantityconstraints,thisstudy furtherdemonstrated howmeasuresof discourseinformationcanreveal thedifferences between thethree typesof speechin informationmanagement, and howthese differencesmay reflectthe characteristics ofthe three typesof speech.Asseen inthe results,adultspeechwascharacterizedby thehighestIPandthe highestLRD; itwasalso characterizedbyastrongassociationbetweennewreferentsandlexicalforms.Theresultsappeartoreflectthenatureof adult--adultconversation:Boththespeakerandthelistenerinadult--adultconversationarecompetentlanguageusers. Conversationaltopicsinadult--adultconversation,unlikethoseinconversationinvolvingchildren,usuallygobeyondthe ‘here-and-now’.Newreferentsandlexicalexpressionsthustendtobeusedmoreofteninadultspeechinordertomeet thecommunicativeneedsofadultspeakers.
Maternalspeechisdirectedtolinguisticallyandcognitivelydevelopingchildren.Whilemothershavematurelinguistic ability,theytendtoadjusttheirspeechwheninteracting withchildren.Asseeninthe results,maternalspeechwas characterizedbythelowestIP.Itappearsthatthemotherstendedtorestricttheintroductionofnewreferentsinorderto easetheinformationprocessingloadforthechildren.Maternalspeechwasalsocharacterizedbyalargeproportionof givenlexicalmentions;thatis,themotherstendedtouseinformativeforms(i.e.,lexicalforms)torefertogivenreferents. Thismayresultfromthemothers’attempttoensurethechildren’scomprehensionofthereferentsmentioned.Theresults appeartoreflectthemothers’attempttofacilitatetheircommunicationwiththechildren.Thisfindingisconsistentwith thosereportedinHughesandAllen(2015)andinClarkandBernicot(2008).HughesandAllenfoundahighproportionof
lexicalNPsforaccessiblereferentsintheiradult-childdata,andtheyshowedthatthephenomenonwaslargelyduetothe repetitionofapartofthepreviousutterance.ClarkandBernicotalsofoundatendencyforadultstorepeatlexicalitemsin child-directed speech. They showed thatadults repeat lexicalitems to check and confirm children’s communicative intentions,tocorrecterrorsinwhatchildrenhavesaid,ortosignal thattheyareattendingtochildren’sutterances.In addition,itappearsthatthehighproportionsofgivenlexicalmentionsinmaternalspeecharealsorelatedtowhatKüntay andSlobin(1996,2002)havecalled‘‘variationsets’’.Variationsetsareacertainkindofdiscourseforminchild-directed speech, characterized by a sequence of utterances with a constantintention, but varyingforms (SeeExcerpt 1 for example).Theinteractivefunctionofvariationsets,assuggestedbyKüntayandSlobin,istomaximizethechanceof comprehensionand/orcomplianceonthepartofthechildthroughtherepetitionofthesamecontent.Variationsetsthus appeartomotivatethefrequentuseoflexicalmentionsforgiveninformation.Interestingly,thecharacteristicofalarge proportionofgivenlexicalmentionsinmaternalspeechwasalsofoundinteachertalk,asshownintheresultsofKumpf (2003).Itappearsthatthischaracteristicreflectsthesimilaritiesbetweenmaternalspeechandteachertalk:Bothmothers and teachers speak to addressees with less competence, and they both share the goal of trying to ensure the comprehensionoftheiraddressees.
Asforchildspeech,itwascharacterizedbythelowestLRD.Thatis,thechildrentendedtousepronominalformsoreven nullforms,whicharelessinformativethanlexicalforms.Childspeechwasalsocharacterizedbyalargeproportionof non-lexicalnewmentions.Inotherwords,whenthechildrenintroducednewreferents,thereferentialformstheyusedmaybe non-informative.Onepossibleexplanationforthisfindingisthatthechildrenwerepragmaticallyimmatureandnotsensitive totheperspectiveoftheirinterlocutors.However,Skarabelaetal.(2013)reportedthatthisphenomenonmaybeassociated withchildren'ssensitivitytothepresenceorabsenceofjointattention.Theyshowedthatchildrenusepronominalandnull formsfornewreferentslargelywhenthosereferentsarealreadyclarifiedthroughthepresenceofjointattention.Asalso suggestedbyAllenetal.(2015),childrenaresensitivetosubtleaspectsofinformationflow,andtheyattendtotheinterplay betweenaccessibilityfactors(e.g.,theinterplaybetween‘newness’and‘jointattention’)whenchoosingreferentialforms.In otherwords,thelargeproportionofnon-lexicalnewmentionsinchildspeechmaynotresultfromchildren'spragmatic deficiency;instead,mostofthesenon-lexicalnewmentionsmayshowchildren'sadjustmentoftheirmessagesforthe interlocutors,thusreflectingtheirdevelopingabilitytoacquireatheoryofmind.Furtherstudiesareneededinordertobetter understandhowtheinterplaybetweenaccessibilityfactorsinfluenceschildren'sreferentialchoice.
ThisstudyhasshownsomesimilaritiesanddifferencesininformationmanagementbetweenMandarinchildspeech, maternalspeech,andadultspeech.Thethreetypesofspeechweresimilarinthattheydemonstratedanaccusative pattern of grammatical alignment, and that they followed the quantity constraints. However, they were different in discourseinformationmeasures,i.e.,inthevalueofinformationpressureandinthevalueoflexicalreferentialdensity.By examiningthethreetypesofspeech,thisstudyhasshownhowinformationmanagementisassociatedwithchildren's developinglinguisticability,mothers’speechadjustments,andadults’maturelinguisticcompetence.
Acknowledgements
Iwouldliketoexpressmyappreciationtotheparticipantsofthisstudyfortheirkindsupportandtothethreeanonymous reviewersfortheirinsightfulcomments.TheresearchreportedinthispaperwasfundedbyagrantfromtheMinistryof ScienceandTechnology,Taiwan,R.O.C.(NSC101-2410-H-004-183).
References
Allen,S.E.M.,2000.Adiscourse-pragmaticexplanationforargumentrepresentationinchildInuktitut.Linguistics38,483--521.
Allen,S.E.M.,Schröder,H.,2003.PreferredargumentstructureinearlyInuktitutspontaneousspeechdata.In:DuBois,J.W.,etal.(Eds.),
PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.301--338.
Allen,S.E.M.,etal.,2015.Theroleofcognitiveaccessibilityinchildren’sreferentialchoice.In:Serratrice,L.,Allen,S.E.M.(Eds.),TheAcquisition
ofReference.Benjamins,Amsterdam,pp.123--153.
Arnold,J.E.,2003.Multipleconstraintsonreferenceforms:null,pronominal,andfullreferenceinMapudungun.In:DuBois,J.W.,etal.(Eds.),
PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.225--246.
Bickel,B.,2003.Referentialdensityindiscourseandsyntactictypology.Language79,708--736.
Chafe,W.,1976.Givenness,contrastiveness,definiteness,subjects,topics,andpointofview.In:Li,C.(Ed.),SubjectandTopic.AcademicPress,
London,pp.25--55.
Chafe, W.,1987.Cognitive constraintson informationflow. In:Tomlin, R.S.(Ed.), Coherence andGroundinginDiscourse. Benjamins,
Amsterdam,pp.21--51.
Chui,K.,1992.Preferredargumentstructurefordiscourseunderstanding.In:Proceedingsofthe14thconferenceonComputationalLinguistics
(COLING92),vol.4,AssociationforComputationalLinguistics,pp.1142--1146.
Chui,K.,Lai,H.-l.,2008.TheNCCUcorpusofspokenChinese:Mandarin,Hakka,andSouthernMin.TaiwanJ.Linguist.6(2),119--144.
Clancy,P.,1993.PreferredargumentstructureinKoreanacquisition.In:Clark,E.V.(Ed.),TheProceedingsoftheTwenty-FifthChildLanguage
Clancy,P.,1997.DiscoursemotivationsforreferentialchoiceinKoreanacquisition.In:Sohn,H.,Haig,J.(Eds.), Japanese/KoreanLinguistics, vol.6.CSLI,Stanford,CA,pp.639--657.
Clancy,P.,2003.Thelexiconininteraction:developmentaloriginsofpreferredargumentstructureinKorean.In:DuBois,J.W.,etal.(Eds.),
PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.81--108.
Clark,E.V.,Bernicot,J.,2008.Repetitionasratification:howparentsandchildrenplaceinformationincommonground.J.ChildLang.35,349--371.
Corston-Oliver,S.H.,2003.CoreargumentsandtheinversionofthenominalhierarchyinRoviana.In:DuBois,J.W.,etal.(Eds.),Preferred
ArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.273--300.
DuBois,J.W.,1985.Competingmotivations.In:Haiman,J.(Ed.),IconicityinSyntax.JohnBenjamins,Amsterdam,pp.343--365.
DuBois,J.W.,1987.Thediscoursebasisofergativity.Language63,805--855.
DuBois,J.W.,2003.Argumentstructure:grammarinuse.In:DuBois,J.W.,etal.(Eds.),PreferredArgumentStructure:GrammarasArchitecture
forFunction.JohnBenjamins,Amsterdam,pp.11--60.
DuBois,J.W., etal.,2003.PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam.
Durie,M.,2003.Newlightoninformationpressure:informationconduits,‘‘escapevalves’’,androlealignmentstretching.In:DuBois,J.W.,etal.
(Eds.),PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.159--196.
Genetti,C.,Crain,L.D.,2003.Beyondpreferredargumentstructure:sentences,pronouns,andgivenreferentsinNepali.In:DuBois,J.W.,etal.
(Eds.),PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.197--224.
Givo´n,T.,1979.OnUnderstandingGrammar.AcademicPress,NewYork.
Guerriero,S.,etal.,2006.ThedevelopmentofreferentialchoiceinEnglishandJapanese:adiscourse-pragmaticperspective.J.ChildLang.33,
823--857.
Gundel,J.K.,2009.Children’suseofreferringexpressions:whatcanittellusabouttheoryofmind?Cognit.Crit.1,73--100.
Gundel,J.K.,Johnson,K.,2013.Children’suseofreferringexpressionsinspontaneousdiscourse:implicationsfortheoryofminddevelopment.J.
Pragmat.56,43--57.
Helasvuo,M.-L.,2003.ArgumentsplitsinFinnishgrammaranddiscourse.In:DuBois,J.W.,etal.(Eds.),PreferredArgumentStructure:Grammar
asArchitectureforFunction.JohnBenjamins,Amsterdam,pp.247--272.
Huang,C.-c.,2011.ReferentialchoiceinMandarinchildlanguage:adiscourse-pragmaticperspective.J.Pragmat.43,2057--2080.
Huang,C.-c.,2012.PreferredargumentstructureinMandarinchildlanguage.TaiwanJ.Linguist.10(2),119--168.
Hughes,M.E.,Allen,S.E.M.,2013.Theeffectofindividualdiscourse-pragmaticfeaturesonreferentialchoiceinchildEnglish.J.Pragmat.56,15--30.
Hughes,M.E.,Allen,S.E.M.,2015.Theincrementaleffectofdiscourse-pragmaticsensitivityonreferentialchoiceintheacquisitionofafirst
language.Lingua155,43--61.
Karkkainen,E.,1996.PreferredargumentstructureandsubjectroleinAmericanEnglishconversationaldiscourse.J.Pragmat.25,675--701.
Kumagai,Y.,2006.Informationmanagementinintransitivesubjects:someimplicationsforthepreferredargumentstructuretheory.J.Pragmat.
38,670--694.
Kumpf,L.E.,2003.Genreandpreferredargumentstructure:sourcesofargumentstructureinclassroomdiscourse.In:DuBois,J.W.,etal.(Eds.),
PreferredArgumentStructure:GrammarasArchitectureforFunction.JohnBenjamins,Amsterdam,pp.109--130.
Küntay,A.,Slobin,D.I.,1996.ListeningtoaTurkishmother:somepuzzlesforacquisition.In:Slobin,D.I.,etal.(Eds.),SocialInteraction,Social
Context,andLanguage:EssaysinHonorofSusanErvin-Tripp.LawrenceErlbaumAssociates,Hillsdale,NJ,pp.265--286.
Küntay,A.,Slobin,D.I.,2002.Puttinginteractionbackintochildlanguage:examplesfromTurkish.Psychol.Lang.Commun.6,5--14.
Lin,W.-h.,2009.PreferredargumentstructureinChinese:acomparisonamongconversations,narrativesandwrittentexts.In:Xiao,Y.(Ed.),The
Proceedingsofthe21stNorthAmericanConferenceonChineseLinguistics(NACCL-21),vol.2. pp.341--357.
Marascuilo,L.A.,McSweeney,M.,1977.NonparametricandDistribution-FreeMethodsfortheSocialScience.Brooks/Cole,Monterey,CA.
Narasimhan,B., etal.,2005.ArgumentrealizationinHindicaregiver-childdiscourse.J.Pragmat.37,461--495.
Nelson,K.,1977.Facilitatingchildren’sacquisitionofsyntax.Dev.Psychol.13,101--107.
O’Dowd,E.,1990.Discoursepressure,genreandgrammaticalalignment--afterDuBois.Stud.Lang.14(2),365--403.
Pawley,A.,Syder,F.H.,1983.Naturalselectioninsyntax:notesonadaptivevariationandchangeinvernacularandliterarygrammar.J.Pragmat.
7,551--579.
Pawley,A.,Syder,F.H.,2000.Theone-clause-at-a-timehypothesis.In:Riggenbach,H.(Ed.),PerspectivesonFluency.UniversityofMichigan
Press,AnnArbor,MI,pp.163--199.
Rozendaal,M.I.,Baker,A.E.,2010.Theacquisitionofreference:pragmaticaspectsandtheinfluenceoflanguageinput.J.Pragmat.42,1866--1879.
SalazarOrvig,A.,etal.,2010.Dialogicalbeginningsofanaphora:theuseofthirdpersonpronounsbeforetheageof3.J.Pragmat.42,1842--1865.
Serratrice,L.,2005.TheroleofdiscoursepragmaticsintheacquisitionofsubjectsinItalian.Appl.Psycholinguist.26,437--462.
Skarabela,B.,2007.Signsofearlysocialcognitioninchildren’ssyntax:thecaseofjointattentioninargumentrealizationinchildInuktitut.Lingua
117,1837--1857.
Skarabela,B., etal.,2013.Jointattentionhelpsexplainwhychildrenomitnewarguments.J.Pragmat.56,5--14.
Snow,C.E.,1989.Understandingsocialinteractionandlanguageacquisition:sentencesarenotenough.In:Bornstein,M.H.,Bruner,J.S.(Eds.),
InteractioninHumanDevelopment.LawrenceErlbaum,Hillsdale,NJ,pp.83--104.
Stoll,S.,Bickel,B.,2009.Howdeeparedifferencesinreferentialdensity?In:Lieven,E.,etal.(Eds.),CrosslinguisticApproachestothe
