Beruflich Dokumente
Kultur Dokumente
Editor
Harald Clahsen
University of Essex
Lydia White
McGill University
Editorial Board
Melissa F. Bowerman
Katherine Demuth
Brown University
Wolfgang U. Dressler
Universitt Wien
Nina Hyams
Jrgen M. Meisel
Universitt Hamburg
William OGrady
University of Hawaii
Luigi Rizzi
University of Siena
Bonnie D. Schwartz
Antonella Sorace
University of Edinburgh
Karin Stromswold
Rutgers University
Jrgen Weissenborn
Universitt Potsdam
Frank Wijnen
Utrecht University
Mabel Rice
University of Kansas
Volume 44
Developmental Psycholinguistics. On-line methods in childrens language
processing
Edited by Irina A. Sekerina, Eva M. Fernndez and Harald Clahsen
Developmental Psycholinguistics
On-line methods in childrens language processing
Edited by
Irina A. Sekerina
City University New York
Eva M. Fernndez
City University New York
Harald Clahsen
University of Essex
TM
2008
401'.930285--dc22
2007038990
Table of contents
Introduction
Irina A. Sekerina, Eva M. Fernndez and Harald Clahsen
Listofcontributors
Behavioralmethodsforinvestigatingmorphological
andsyntacticprocessinginchildren
Harald Clahsen
vii
xvii
Event-relatedbrainpotentialsasawindowtochildrens
languageprocessing:Fromsyllablestosentences
Claudia Mnnel and Angela D. Friederici
29
Usingeyemovementsasadevelopmentalmeasure
withinpsycholinguistics
John C. Trueswell
73
Lookingwhilelistening:Usingeyemovementstomonitor
spokenlanguagecomprehensionbyinfantsandyoungchildren
Anne Fernald, Renate Zangl, Ana Luz Portillo
and Virginia A. Marchman
97
Whatlurksbeneath:Syntacticprimingduringlanguage
comprehensioninpreschoolers(andadults)
Jesse Snedeker and Malathi Thothathiri
137
Languageacquisitionresearch.Apeekatthepast:
Aglimpseintothefuture
Helen Smith Cairns
169
Index
187
Introduction
IrinaA.Sekerina,EvaM.FernndezandHaraldClahsen
Thestudyofchildlanguageoccupiesauniqueplaceinresearchonchildrenscognitivedevelopment.Thiscomesasnosurprise,aslanguageisquiteclosetothe
coreofwhatitmeanstobehuman.Childrensuccessfullylearntheirnativelanguageinarelativelyshorttimeandwithouttheneedforformalinstruction.Languageisalsothemainvehiclebywhichwelearnaboutotherpeoplesthoughts;
therefore,cognitiveandlinguisticaspectsofhumandevelopmentmustbeintimatelyrelated.
Traditionalmethodsofinquiryinmodernlinguisticsandcognitivepsychologyhaveenabledustolearnagreatdealabouthowchildrenacquirelanguageand
thestagestheygothroughontheirwaytoadultcompetence(Pinker1995).But
empiricalstudiesonhowchildrenslanguagedevelopspublishedoverthelast30
or40yearshaveastrikingcharacteristicincommon:theytreatlanguageacquisitionasaprocessthatinvolvesbuildingastaticdatabasecalledthe grammar,tothe
exclusionofthemechanismsthatoperateinrealtimewhenthechildproducesor
comprehends language. The classic Competence/Performance distinction (e.g.,
Chomsky1964)providesausefulframeworkfordiscussingthisproblem:while
investigationsofchildlanguageacquisitionaregroundedontheassumptionthat
knowledgeoflanguageisputtoworkviaasetofprocessingmechanisms(performance),theprimaryconcerninacquisitionresearchhasbeenwithhowthatprincipledknowledge(competence)develops.McDaniel,McKeeandCairns(1996),
intheirseminalbookonassessingchildlanguage,describedhowtheknowledge
thatconstitutescompetencehadupuntilthenbeenextensivelystudied,andthey
documented the predominance of off-line experimental methods, that is, techniquesthatpromptedchildrentoactoutsentences,answerquestionsorprovide
grammaticalityjudgments,responsesthatcouldthenbecomparedtothoseprovidedbyadultsorbyolderoryoungerchildren.Armedwithempiricalevidence
ofthatsort,thefieldwasabletobegintoaddresssomeofthemostbasicquestions
aboutlanguagedevelopmentandtoformulateexplicitdescriptionsaboutthenatureofdevelopmentalsequences.
viii IrinaA.Sekerina,EvaM.FernndezandHaraldClahsen
Theeraoftraditionalresearchonlanguageacquisition,capturedsowellin
thevolumebyMcDanieletal.(1996),hasgrownintoamatureareaofinquiry
whoseinsightshaveledtoarichunderstandingaboutthedevelopmentoflinguisticcompetence.Buttimeshavechanged,asweenteranewerathattakesa
dynamicprocessingapproachtothestudyoflanguagedevelopment(Trueswell
thisvolume).Wearewitnessingagrowinginterestinthemechanismsthatunderlieproductionandcomprehensionabilitiesinchildren,ashiftfromafocus
oncompetencetoafocusonperformance.Thisenterprisehasbeensignificantly
facilitatedbyrecentadvancesintechnologiesthatpermittrackingbehaviorata
very fine temporal resolution, methods that have been successfully and extensivelyappliedtostudylanguageprocessinginadults.Suchnewtechniques,which
wewillcollectivelyrefertoas on-line,measurereactiontimes,trackeyegazes,
examinebrainactivity.Someofthesemethods,likeself-pacedreading,self-paced
listening,andcross-modalprimingbenefitfromhavingalong-standingtradition
in the study of adult language processing. Others, like eyetracking and neurophysiological techniques (Henderson & Ferreira 2004; Trueswell & Tanenhaus
2005;Carreiras&Clifton2004),arenewerbutquitepowerfuladditionstotheexperimentaltoolkit,particularlybecausetheyprovidethemeanstostudyingreat
detailveryearlyphasesofprocessing,andbecausetheyrelylittleonconscious
attentiontoormetalinguisticawarenessoflinguisticstimuli.
On-line methods have made their way into language acquisition research
withatrulyamazingspeed.Amere10yearsago,asdocumentedbyCecileMcKee
(1996) in her chapter on on-line methods in child language research, reaction
time methods (cross-modal priming in particular) dominated the scene, neuroimaginghardlyhavingapresence.EyetrackingwasfullyabsentfromMcKees
chapter.
When applied to the study of child language, on-line methods permit researchers to observe the interaction of grammar principles (competence) and
behaviorallimitationsand/orpreferences(performance),withagreaterlevelof
detailandagreaternumberofperspectivesthaneverbefore.Wecannowinvestigatehowchildrencoordinatemultiplesourcesofinformationinrealtimeand
arriveatsentencemeaningusinginformationextractednotonlyfromthewords
and structure of the sentence but also from the nonlinguistic context. The applicationofon-linemethodsalsomakesitpossibletotestchildrensperformance
limits,toseparateperformancefromcompetenceinassessingchildrensstaticand
developinglinguisticknowledge,anapproachthatpermitsbuildingandtesting
theoriesabouthowchildrenslanguageprocessingcontributestotheiracquisition
oflanguage(Fodor1998).
Thegrowingimportanceofon-linemethodsinchildlanguageresearchwas
evidentattheforumthatbroughtthisvolumeintobeing,theWorkshop on On-
Introduction
ix
x
IrinaA.Sekerina,EvaM.FernndezandHaraldClahsen
vestigatingchildrensgrammaticalprocessinginproductionandcomprehension.
Threeexperimentaltechniquesarepresentedindetail:speeded production,used
toexamineautomaticprocessesinvolvedinthespokenproductionofinflected
words;self-paced listening,employedinexaminingchildrensprocessingoftemporarilyambiguoussentences,specificallyrelative-clauseattachmentpreferences;
andcross-modal primingusedtodeterminewhetherdislocatedconstituents(e.g.,
frontedwh-phrases)arereactivatedatcorrespondinggappositionsduringprocessing.DespitetheavailabilityofERPsandeyetracking,behavioralmethodssuch
astheonesoutlinedbyClahsenstillhaveanimportantplaceindevelopmental
psycholinguistics,asthesetechniquescanbeappliedtostudyarangeofcomplex
andinterestinglanguagephenomena,providingtime-sensitivemeasuresthatrequireminimaltechnicalequipment.
Chapter2,Event-relatedbrainpotentialsasawindowtochildrenslanguage
processing:Fromsyllablestosentences(ClaudiaMnnelandAngelaFriederici),
provides an overview of the comprehensive research program to study neurocognitionoflanguage,corticalnetworksandcognitivefunctions,andlanguage
acquisitionusingERPsattheInstituteofNeuropsychologyoftheMax-PlanckInstituteforHumanCognitiveandBrainScienceinLeipzig.AngelaFriederici,head
oftheInstitute,isaleadingresearcherinapplyingmeasuresofbrainactivityin
adultsentenceprocessingandhasproposedaninfluentialneurocognitivemodel
oflanguageanditsextensiontotheareaoflanguageacquisition,whichsheherself
callsthe developmental cognitive neuroscience of language(Friederici2000,2002).
Intheirchapter,MnnelandFriedericidescribethesuccessfulapplicationof
ERPmethodstostudylanguageacquisitionininfantsfrombirthtothreeyears.
ThechapterdescribesfiveERPcomponentscloselylinkedtolanguage,andhow
theyreflecttheprocessingofphonological,semantic,andsyntacticinformation
inprogressivelyolderchildren,comparedtoadults.MnnelandFriedericioutlineERPresearchonanumberoflandmarksofchildlanguageacquisitionand
identifyneuralcorrelatesfordevelopmentalstagesinauditorylanguagecomprehension.TheresearchfindingssummarizedinChapter2include(a)workonsyllableandstressdiscriminationininfantsusingthepassiveoddballparadigm,(b)
investigationsoftheN400componentasareflexofphonotacticknowledge,early
word learning and knowledge of selectional restrictions for verbs, (c) research
measuringsensitivitytosentence-levelprosodiccueswiththeClosurePositive
Shift(CPS),and(d)studieselicitinganadult-likebiphasicELAN-P600componentinresponsetophrasestructureviolationsatthesentencelevel.Mnneland
FriedericidrawthechaptertoaclosebydemonstratinghowERPcomponentscan
beusedtoidentifyinfantsatriskforlaterdevelopinglanguageproblems,suchas
SpecificLanguageImpairment(SLI)anddyslexia.Thetechniquethereforecon-
Introduction
stitutesanewdiagnostictoolforveryearlyidentificationofchildrenwhowould
benefitfromintervention.
Chapter3,Usingeyemovementsasadevelopmentalmeasurewithinpsycholinguistics (John Trueswell), is the first of the three chapters dedicated to
eyetracking. Trueswell describes and evaluates free-viewing eyetracking also
knownasthevisualworldparadigmandtheworld-situatedeye-gazeparadigm
tostudysentence-level comprehension intoddlersandpreschool-agechildren.
JohnTrueswellandhisteamattheInstituteforResearchinCognitiveScienceat
theUniversityofPennsylvaniawerethefirsttoadaptthismethodtoinvestigate
how5-year-oldchildrencomprehendsyntacticallyambiguoussentencesandhow
sentence processing mechanisms develop (Trueswell, Sekerina, Hill & Logrip
1999).Trueswellbeginshischapterwithahistoryofeyetrackinginadultresearch
andcontinueswithatechnicaldescriptionofhead-mounted,remoteandpoor
manseyetrackers;anexplanationofcalibrationprocedures;andremarksabout
eyetracking data analysis. Trueswell then describes three linking assumptions
criticalformakingvalidinferencesaboutwhatchildrenseyegazepatternsreveal
about the development of sentence processing mechanisms, and, in particular,
referentialprocessing.
Inasecondpartofhischapter,Trueswelldiscussesrecentfindingsfromearly
oculomotordevelopment,visualsearch,andneurocomputationmodelsofvisual
attention,allofwhichareinformativewithrespecttounderstandingcharacteristicsofspatialattentionfrominfancyuntiltheageof3.Suchdiscussionprovidesa
solidbackdropforTrueswellsreviewoftheexperimentalevidenceaccumulated
overthepastdecadesaboutusingeyemovementstoinferhowchildrenresolve
prepositionalphraseattachmentambiguities,pronominalreference,andquantifierscope,aswellastheinfluenceofdiscoursefactorsinreferentialcommunicationtasks.
Chapter4,Howinfantslookastheylisten:Usingeyemovementstomonitor
on-linecomprehensionbyveryyounglanguagelearners(AnneFernald,Renate
Zangl,AnaLuzPortilloandVirginiaMarchman),takesonacomplementaryapproachtothefoundationalchapterbyTrueswell,walkingthereaderthrougha
detaileddescriptionofthelistening-while-lookingparadigmpioneeredbyAnne
Fernald,theleadingauthorofthischapter,andtheStanfordUniversityCenter
for Infant Studies. Listening-while-looking (LWL) is a version of free-viewing
eyetrackingadaptedforusewithinfants.Thetechniquewasdevelopedoutofa
desiretoovercomeanumberofshortcomingsofcommonlyusedoff-linemethodstoexaminelanguagecomprehensionbyinfants,suchasdiarystudies,parental-report checklists of vocabulary growth, experiments on word learning, and
early versions of the preferential-looking paradigm. These methods do not tap
intothereal-timepropertiesofspokenlanguageandreveallittleaboutthechilds
xi
xii IrinaA.Sekerina,EvaM.FernndezandHaraldClahsen
Introduction xiii
ducedbychildrenthroughstudiesemployingsophisticatedoff-linetechniquesto
probeunderlyinglinguisticknowledge,thoughstudiestakingadvantageofonlinetechniquestounderstandperformancemechanismsinchildren.Thechapter
addressesthecomplexityoftherelationshipbetweentheoryandpractice.Theorieshavedrivenexperimentalinnovation,whileatthesametimetheavailability
ofexperimentaltechniquespromotesthedevelopmentofnewtheories.
Cairns discusses the early interest on the problem of how an underlying
grammaticalsystemisacquired.Therealizationthatthespeechproducedbychildrenvastlyunder-representswhattheymightknowledtoashiftinintereston
howimplicitknowledgeoflanguagedevelops,anundertakingsustainedbythe
increasedsophisticationoftheoriesofgrammar.
Cairnsthenprovidesabriefhistoryofhowpsycholinguisticsdevelopedinto
aripeareaofinquiryexploringthemechanismsemployedintheproductionand
perceptionoflanguageinadults.Asforchildren,Cairnsnotesthatthepreoccupationwithcompetenceinlanguageacquisitionresearch,alongwithaconcern
tocontrolperformancefactors(suchaseffectsofmemoryortaskdemands),
resultedinapaucityofstudiesofchildrensperformance.
But,asthisvolumerepresents,andCairnsdiscussesatlengthinherchapter,
newquestionsareemergingthatdirectlyaddressthenatureofchildrenslanguage
processing.Cairnsreviewsarangeofstudiessomediscussedelsewhereinthis
volume, some presented at the Workshop, and others sampled from the literaturethatareaddressingquestionsaboutwhetherchildrenconstructrepresentationsthatresemblethoseconstructedbyadults,abouthowchildrenreviseinitial
parses,abouthowchildrensmemoryspanslimittheirperformance,abouthow
cross-linguisticresearchisidentifyinguniversaltendenciesinchildlanguageprocessing,andabouthowsomeofthesetechniquescanbeemployedfortheearly
detectionoflanguagedisorders.
Toconcludeherchapter,Cairnsofferssomespeculationsaboutwhatthefutureholdsforresearchinchildlanguagedevelopment,echoinganumberofthe
othercontributionstothisvolumewhenstressingtheneedformultipleandcomplementarymethodologicalapproaches.Progressiscalledforinparticularwith
respecttothequestionofhowchildrenoperateoninputtocreatenewgrammars,
onhowadult-likeprocessingskillsdevelop,andontheunderlyingneuralorganization.Wehopethatthisvolumewillleadtoinnovationintheseandrelated
questions.
Wecannotunderstatetheimportanceofthesourceforthisvolume,theWorkshop on On-Line Methods in Childrens Language Processing held at the GraduateCenteroftheCityUniversityofNewYork,March2223,2006.Theabstract
proceedingsforthepapersandposterspresentedattheWorkshoparepresently
available at http://www.qc.cuny.edu/~efernand/childlang/. The first joint scien-
xiv IrinaA.Sekerina,EvaM.FernndezandHaraldClahsen
tific gathering specifically dedicated to the emerging field of experimental developmental psycholinguistics, the Workshop gathered specialists in language
acquisition, psycholinguistics, neuroscience, and speech-language pathology.
Ostensibly, the objective was to provide a forum for discussing the advantages
andshortcomingsofusingon-linemethodstostudylanguageprocessinginchildren,rangingfrombehavioralmethodstoparadigmsinvolvingeyetracking,to
neurophysiologicaltechniques.Beyonddiscussingmethodology,theWorkshop
initiateddialogbetweenaninternationalandinterdisciplinarygroupofscholars,
whowereaffordedtheopportunitytoreflectonpastlandmarksofresearchinlanguageacquisition,summarizethecurrentstateofemergingresearchonlanguage
processinginchildren,andengageinlivelydebatesaboutfuturedirections.
Thecurrentvolumeofferssixpaperslooselybasedontalksdeliveredatthe
Workshop. We have asked the authors to concentrate on methodological matters,buttheyhavegonebeyondthatdirectiveandproducedchaptersthatserve
asmorethanintroductionstoexperimentalparadigms,sincetheyaddresssome
of the complex theoretical debates as well as provide a solid overview of child
languagedevelopment.
Wearehappytotakethisopportunitytoexpressourgratitudetothosewho
madetheWorkshopandthisvolumepossible.
First,wethankthe120presentersandattendeesfromtheUnitedStates,Europe,JapanandAustraliafortheirthought-provokingpapersandposters.
Second, we thank the co-directors of the Annual CUNY Human Sentence
ProcessingConference,JanetDeanFodorandDianeC.Bradley,forenthusiastically supporting this project andsharing the CUNY Conferenceinfrastructure
withus.
InorganizingtheWorkshopandeditingthevolume,wewereassistedbya
committeeofcommittedreviewersthatincludedJenniferArnold,HollyBranigan,
PatriciaBrooks,EdwardGibson,JeffreyLidz,EricPakulak,JulieSedivy,Valery
Shafer,MatthewTraxler,andVirginiaValian.Wewishtothankthem,alongwith
allthecontributingauthors,forsharingtheirexpertiseandkeepingupwithour
tightdeadlines.
WewouldalsoliketoexpressourgratitudetoLydiaWhite,Co-Editorofthe
Language Acquisition and Language Disorders series, Kees Vaes and the John
BenjaminsPublishingCompany,forencouragementandassistanceatallstages
ofproduction.
Finally,aredeeplygratefultotheNationalScienceFoundationforitsgenerousfinancialsupportofthisWorkshopthroughitsChildResearchInitiativegrant
#0518438.(TheopinionsexpressedintheWorkshoparethoseoftheorganizers,
presentersandattendeesanddonotnecessarilyrepresenttheviewsoftheNationalScienceFoundationoranyothergovernmentalorganization.)
Introduction xv
References
Carreiras, M. & Clifton, C., Jr. (Eds.). (2004). The on-line study of sentence comprehension:
Eyetracking, ERPs and beyond. Hove:PsychologyPress.
Chomsky,N.(1964).[Thedevelopmentofgrammarinchildlanguage]:Discussion.InU.Bellugi&R.Brown(Eds.),The acquisition of language,Vol.1:Monographs of the Society for
Research in Child Development, 29(1),3542.
Fodor,J.D.(1998).Parsingtolearn.Journal of Psycholinguistic Research, 27(3),339374.
Friederici,A.D.(2000).Thedevelopmentalcognitiveneuroscienceoflanguage:Anewresearch
domain.Brain and Language, 71,6568.
Friederici,A.D.(2002).Towardsaneuralbasisofauditorysentenceprocessing.Trends in Cognitive Sciences, 6(2),7884.
Henderson,J.M.&F.Ferreira(Eds.).(2004).The interface of language, vision, and action: Eye
movements and the visual world.Hove:PsychologyPress.
McDaniel,D.,McKee,C.,&Cairns,H.S.(Eds.).(1996).Methods for assessing childrens syntax.
CambridgeMA:TheMITPress.
McKee,C.(1996).On-linemethods.InD.McDaniel,C.McKee,&H.S.Cairns(Eds.),Methods
for assessing childrens syntax(pp.189212).Cambridge,MA:TheMITPress.
Pinker,S.(1995).Languageacquisition.InL.R.Gleitman&M.Liberman(Eds.),An invitation
to cognitive science, Vol.1.: Language(2nded.,pp.135182).CambridgeMA:TheMIT
Press.
Trueswell,J.C.&Tanenhaus,M.K.(Eds.).(2005).Approaches to studying world-situated language use: Bridging the language-as-product and language-as-action traditions.Cambridge
MA:TheMITPress.
Trueswell,J.C.,Sekerina,I.A.,Hill,N.,&Logrip.M.(1999).TheKindergarten-patheffect:
Studyingon-linesentenceprocessinginyoungchildren.Cognition, 73,89134.
List of contributors
Angela D. Friederici
MaxPlanckInstituteforHuman
CognitiveandBrainSciences
Stephanstr.1a
04103Leipzig,Germany
Tel.:+49/0341/99-40112
Fax:+49/0341/99-40113
angelafr@cbs.mpg.de
Harald Clahsen
UniversityofEssex
Colchester,CO43SQ,UK
Tel:+44/1206/87-2228
Fax:+44/1206/87-2198
harald@essex.ac.uk
Claudia Mnnel
MaxPlanckInstituteforHuman
CognitiveandBrainSciences
Stephanstr.1a
04103Leipzig,Germany
Tel.:+49/0341/99-40112
Fax:+49/0341/99-40113
maennel@cbs.mpg.de
Anne Fernald
StanfordUniversity
Stanford,CA94305
Tel:650-725-2400
Fax:650-725-5699
afernald@stanford.edu
Eva M. Fernndez
QueensCollegeandGraduateCenter
CityUniversityofNewYork
65-30KissenaBlvd.
Flushing,NY11367-1597
Tel:718-997-2867
Fax:718-997-2873
eva.fernandez@qc.cuny.edu
Virginia A. Marchman
StanfordUniversity
Stanford,CA94305
Tel:650-725-2400
Fax:650-725-5699
marchman@psych.stanford.edu
Ana Luz Portillo
StanfordUniversity
Stanford,CA94305
Tel:650-725-2400
Fax:650-725-5699
analuz@stanford.edu
xviiiListofcontributors
Irina A. Sekerina
CollegeofStatenIslandandGraduate
Center,CityUniversityofNewYork
2800VictoryBlvd.
StatenIsland,NY10314-6609
Tel:718-982-3760
Fax:718-982-4114
sekerina@mail.csi.cuny.edu
Jesse Snedeker
HarvardUniversity
WilliamJamesHall,33KirklandStreet
Cambridge,MA02138-2044
Tel:617-495-3873
Fax:617-384-7944
snedeker@wjh.harvard.edu
Marathi Thothathiri
HarvardUniversity
WilliamJamesHall,33KirklandStreet
Cambridge,MA02138-2044
Tel:617-384-8357
Fax:617-384-7944
malathi@wjh.harvard.edu
John C. Trueswell
UniversityofPennsylvania
3401WalnutStreet,Room302C
Philadelphia,PA19104-6228
Tel:215-898-0911
Fax:215-573-9247
trueswel@psych.upenn.edu
Renate Zangl
StanfordUniversity
Stanford,CA94305
Tel:650-725-2400
Fax:650-725-5699
ezangl@stanford.edu
chapter1
Whilemostfirstlanguageacquisitionresearchtodatehasfocusedonthedevelopmentofchildrenslinguisticcompetence,anumberofresearchteamshave
alsoinvestigatedthemechanismschildrenemploytoprocesssentence-leveland
word-levelinformationinrealtime,byapplyingexperimentaltechniquesfamiliarfromtheadultprocessingliteraturetochildren.Thischapterpresentsan
overviewofdifferentkindsofbehavioraltasksforinvestigatingbothmorphologicalandsyntacticprocessinginchildrenfocusingonthreetechniquesthat
wehaveexploredinourownresearchonchildrenson-linelanguageprocessing:
self-pacedlistening,cross-modalpriming,andspeededproduction.
1.
Introduction
In1996,CecileMcKeepresentedanoverviewoftheverysmallnumberofon-line
techniquessuitableforstudyingsyntacticprocessinginchildrenthatwereavailableatthetime.Thepurposeofthischapteristoprovideanupdatedoverview.My
focuswillbeonchildrensgrammaticalprocessingandondifferentkindsofbehavioral tasksforinvestigatingmorphologicalandsyntacticprocessinginchildren.
Languageprocessingcanbeconceivedofasasequenceofoperations,each
ofwhichtransformsalinguisticrepresentationofastimulusintoalinguisticrepresentationofadifferentform.Researchintolanguageprocessingexamineshow
linguisticrepresentationsareconstructedin real timeduringthecomprehension
and production of language and how different sources of information become
availableovertime.Tostudytheprocessesinvolvedinproductionandcomprehension as they occur, time-sensitive, so-called on-line, measures of language
processingarerequired.Theadvantagesofusingon-lineexperimentaltechniques
are that they allow us to tap into automatic unconscious processes involved in
2
HaraldClahsen
languagecomprehensionandproductionandthattheyminimizeparticipantsrelianceonexplicitormetalinguisticknowledge.Therearetwobasictypesoftimesensitivemeasuresavailabletoexaminelanguageprocessing:behavioralmeasures
(e.g.,comprehensionresponsetimesandproductionlatencies)andphysiological
measures(e.g.,event-relatedbrainpotentials(ERPs)andeyemovements).Asthe
latterwillbepresentedinotherchaptersofthisbook(seeMnnel&Friedericifor
ERPs,andTrueswellforeyemovementexperiments),Iwillonlybeconcerned
withbehavioralmeasuresoflanguageprocessinghere.
Beforepresentinganoverviewofbehavioralexperimentalmethods,itisnecessarytoestablishsomecriteriaagainstwhichthevariousmethodscanbeevaluated.Thefirstcriterionconcernsthetime-sensitivityofatechniqueandasksat
which point in time during language processing a particular measure is taken.
Clearly,ifatechniquemeasuresresponsesattheoffsetofastimulus,e.g.,atthe
endofasentence,itisnotparticularlyrevealingforunderstandingthemomentby-momentcharacteristicsofprocessesoccurringduringtheprocessingofthat
sentence.Thesecondcriterioniswhetherthestimuliarepresentedin a natural
wayallowingparticipantstoprocessthemusingnormallisteningorreading.As
we will see, this is particularly difficult to achieve for behavioral experimental
tasks.Thethirdquestionweaskiswhethertheexperimentaltaskassignedtoparticipantsischild-appropriate.Sometechniquesrequireadualtask,e.g.,monitoringforavisualtargetwhilelisteningtoasentence,whichmaybetoochallenging
foryoungchildren.Thefourthcriterioniswhetheratechniqueis linguistically
versatile,i.e.,applicabletoarangeofdifferentlinguisticphenomena.Finally,we
willaskwhetheratechniqueisfield-compatible.Thisreferstopracticalconsiderationsinrunningexperimentswithchildren.Insomecircumstances,forexample,itisimpossibletobringchildrenintothelab.Insuchcases,itwouldbe
advantageousifatechniquerequiredminimalequipmentsothatchildrencanbe
testedattheirschoolsortheirhomes.
Intheremainderofthischapter,Iwillconsiderbehavioralmethodsfirstfor
studyingon-linesentencecomprehensionandsecondforinvestigatinglanguage
production.Myfocuswillbeontheadvantagesanddisadvantagesofthreetechniquesthatwehaveexploredinourownresearchonchildrenson-linelanguage
processing: (i) the self-paced listening task to examine childrens comprehensionofambiguoussentences;(ii)thecross-modalpictureprimingtasktostudy
childrens comprehension of syntactic dependencies, specificallywh-dependencies, and (iii) the speeded production task to investigate processes involved in
childrensproductionofmorphologicallycomplexwords.
Behavioralmethods
2.
Theadultpsycholinguisticliteratureoffersarangeofbehavioralmethodsforinvestigatingon-linegrammaticalcomprehension,butonlyasmallnumberoftechniqueshavebeenusedwithchildren:wordmonitoringduringsentencecomprehension,proberecognition,speededgrammaticalityjudgment,self-pacedreading
andlistening,andcross-modalpriming.Whatiscommontothesetechniquesis
thattheycanbeusedwithchildrenfromabout4or5yearsofageonwardsto
studyrelativelycomplexsyntacticphenomena.Thestudyoflanguageprocessing
ininfantsrequiresdifferenttechniquesmeasuring,forexample,preferentiallooking and head-turning patterns (see Fernald, Zangl, Portillo, & Marchman this
volume). Here, I will first briefly present word monitoring, probe recognition,
andspeededgrammaticalityjudgment,andthendiscussinsomemoredetailselfpacedreading,self-pacedlisteningandcross-modalpriming.
2.1
Wordmonitoring
TylerandMarslen-Wilson(1981)wereamongthefirsttoinvestigateon-linesentencecomprehensioninchildren.Theyusedataskinwhichparticipantsmonitor
linguisticstimuli,e.g.,auditorilypresentedsentencesassuchasthosein(1),fora
particulartargetword,e.g.,thewordhand.Theparticipantsresponse,usuallyeitherabuttonpressoravocalresponse,indicatesthattheparticipanthasnotedthe
occurrenceofthetargetinthesentence.Word-monitoringtimesaremeasured
fromthetargetsoccurrenceinthesentencetotheparticipantsresponse.
(1) a. Johnhadtogobackhome.Hehadfallenoutoftheswingandhadhurt
hishand ontheground.
b. Johnhadtositontheshop.Hehadlivedoutofthekitchenandhad
enjoyedhishandinthemud
c. TheonsittoptohadJohn.Helivedhadandkitchentheouthisofhad
enjoyedhandmudinthe
TylerandMarslen-Wilson(1981)appliedthistechniqueto5-,7-,and10-yearoldchildrenandfoundthatthechildrensabilitytodetectawordtargetshowed
thesamepatternofdegradationasitdidinadultswiththeshortestmonitoring
timesforcontextuallyappropriatesentences(1a),followedbycontextuallyinappropriatesentences(1b),andsemanticallyandsyntacticallyanomaloussentences
(1c).Thisfindingwastakentoindicatethatchildrenandadultsanalyzesentences
inessentiallythesameway,i.e.,childrenlikeadultsusecontextinformationto
constructinterpretativerepresentationson-linewhichinturnfacilitatestherec-
HaraldClahsen
ognitionprocessofthetargetwords.Inthecategorymonitoringtask,however,in
whichchildrenhadtomonitorthesentencesforamemberofaparticularsemanticcategory(e.g.,Monitorforbodyparts),the5-year-oldsdifferedfromtheolderchildreninthatthefacilitatingeffectofcontextuallyappropriatesentences(1a)
wassmallerthanforthe7-and10-year-olds.TylerandMarslen-Wilson(1981)
attributedthisfindingtotheadditionalprocessingcostassociatedwithsemanticattributematchingpossiblyinconjunctionwithamoregeneralproblemwith
utilizingcertaintypesofpragmaticcuesduringsentencecomprehension.
TylerandMarslen-Wilsons(1981)word-monitoringtaskallowstheresearchertoexaminetheroleofdifferentkindsofcontextualinformationforwordrecognition.Thetaskprovidesatime-sensitivemeasureofwordrecognitionincontext
andallowslistenerstoprocesstheauditorilypresentedsentencesinanormalway.
However,asMcKee(1996)pointedout,adisadvantageofthistechniqueisthat
only a limited range of relations between target words and their host material
canbestudiedwiththistechnique.Itis,forexample,hardtoseehowcoreference
relationsandotherkindsofsyntacticdependenciescouldbeexaminedwiththis
technique.
2.2
Proberecognition
Intheproberecognitiontask,participantshearorreadasentence.Atsomepoint,
thepresentationofthesentenceisstoppedandparticipantsareaskedtodecide
whetheravisuallyorauditorilypresentedtargetword(probe)hadoccurredin
theprecedinglinguisticmaterial.Responsetimesaremeasuredfromtheonset
oftheprobeitemtothebeginningoftheparticipantsvocalresponseorbutton
press.Severalresearchershaveusedthistasktoexaminedifferentkindsofsyntacticdependenciesinadultsentencecomprehension(Bever&McElree1988;McElree&Bever1989;MacDonald1989;Bever&Sanz1997).Consider,forexample,
sentencessuchasthosein(2)fromMcElreeandBever(1989):
(2) a. Thedazedcabbiewhodrovethebeat-uptaxiwasresented(P1)
constantly(P2).
DAZED
b. Thedazedcabbiewhodrovethebeat-uptaxiwasresentful(P1)
constantly(P2).
DAZED
P1andP2indicatethepointsatwhichtheprobeitemsappeared.McElreeand
Beverdidnotfindanydifferencebetween(2a)and(2b)atP1,butattheendof
the sentence (P2), response times were significantly shorter for (2a) than (2b).
Behavioralmethods
Proberecognitiontimesareknowntoyieldfasterresponsetimesfortargetwords
thatwererecentlyperceivedthanforthosethatarefurtherawayfromtheendof
thesentence.Giventheassumptionthatpassivesentencessuchas(2a)contain
asyntacticgap of the dislocatedobject, the shorterproberecognitiontimes to
DAZEDin(2a)havebeeninterpretedasarecencyeffect,duetothereactivation
ofthedislocatedphrasethe dazed cabbieafterresented.
Mazuka (1998) applied this technique to groups of English and Japanesespeaking children as young as 4. Children had to listen to sentences involving
main and subordinate clauses and were probed on auditory word targets from
these sentences. Her results indicate differences in the way main and subordinateclausesareprocessedinthetwolanguages.Specifically,theEnglish-speakingchildrenshowedanadvantageforsubordinateclauses(asrevealedbyshorter
responsetimesinalexicalproberecognitiontask),whereastheJapanesechildren
hadshorterRTsformainclauses.
Fromamethodologicalperspective,onecrucialdisadvantageoftheproberecognitiontaskisthatitappearstobelesstime-sensitivethanotheron-linetechniquesandthatthetaskissensitivetoavarietyofstrategicprocesses(Gordon,
Hendrick,&Foster2000).Inmanystudies,probe-recognitiontimesaremeasured
attheendofthesentence.Thesedatadonottapon-linesyntacticprocessingas
itoccursbutaremorelikelytopickupsentence-finalwrap-upprocesses,which
may involve semantic rather than syntactic representations. A disadvantage of
studiesthatmeasuredprobe-recognitiontimesatwithin-sentencetestpointsis
thatthestimulussentenceshavetobeinterrupted,whichmakesthetaskrather
unnatural.
2.3
Speededgrammaticalityjudgment
Inthistask,participantsareaskedtojudgethegrammaticalityorungrammaticalityoflinguisticstimuliasquicklyaspossible.Timedorspeededgrammaticality
judgmenttaskshavebeenwidelyusedtoexamineadultssensitivitytovarious
typesofgrammaticalandsemanticinformation,orrelativeprocessingdifficulty.
Thegeneralassumptionisthatrelativeprocessingdifficultyshouldbereflected
inslowerresponsetimes,lowerresponseaccuracy,orboth(McElree&Griffith
1995, 1998). A variant of this task, the violation detection paradigm, has also
been applied to children (Wulfeck 1993; Kail & Diakogiorgi 1998; Kail 2004).
Consider,forexample,Kails(2004)studyinwhichthreeagegroupsofFrench
children(meanage:6;8,8;6and10;10years)andagroupofadultswereaskedto
detectagreementviolationsinsentencessuchas(3a)andwordorderviolations
insentencessuchas(3b):
6
HaraldClahsen
(3) a.
b.
The stimulus sentences were presented auditorily with normal intonation. Participantswereaskedtodecidewhetherasentencehadgoodgrammarandwere
specificallyinstructedtopressabuttonassoonastheydiscoveredanungrammaticality.Whileresponsetimestothegrammaticalsentenceswerenotanalyzed,
theresponselatenciesfortheungrammaticalsentencesweremeasuredfromthe
offsetoftheword(e.g.,*remplissentin(3a))thatmadeasentenceungrammatical.
Kail(2004)foundthatbothchildrenandadultswerefasterindetectingagreementviolationsthanwordorderviolationssuggestingdifferencesinsensitivityto
differenttypesofungrammaticality.
Fromamethodologicalperspective,itisnoteworthythatchildrensresponse
timesinthistaskweresubstantiallylongerthanthoseofadults.Thiswasthecase
notonlyforthe8-to-9-year-olds,whohadanoverallresponsetimeof2017ms,
butalsoforthe6-to-8-year-oldswhohadameanoverallresponsetimeof2573ms,
morethanthreetimesoftheadultgroup.Theseextremelylongresponsetimes
suggestthatthistaskisparticularlydifficultforchildrenand,moreimportantly,
thatthedataareunlikelytotapautomaticprocessesinvolvedinchildrenslanguage processing. Moreover, grammaticality judgment tasks have been subject
tomuchcriticismasthedegreetowhichsuchjudgmentsreflectimplicitgrammaticalcompetenceisunclear(Schtze1996).Itisalsonotobvioushowyoung
childreninterprettheinstructiontodecidebetweensentencesthathavegoodvs.
badgrammar.
2.
Self-pacedreadingandself-pacedlistening
In this task, sentences are presented segment-by-segment or word-by-word either visually or auditorily. Participants trigger the presentation of subsequent
segmentsbypressingapacingbutton.Inself-pacedreading,priorsegmentsor
wordsmayeitherstayonthescreenordisappearuponpressingthepacingbutton.Timesbetweenbuttonpressesarerecordedandprovideastep-by-steprecord
oftheparseasitunfolds.Thebasicrationaleunderlyingthistaskisthatincreased
readingorlisteningtimestoaparticularsegment(relativetothesamesegmentin
Behavioralmethods
acontrolcondition)indicaterelativelyhigherprocessingdifficultyatthispointin
thesentence(Just,Carpenter,&Wooley1982;Mitchell2004).
Theself-pacedreadingtaskhasbeenwidelyusedinadultsentenceprocessing
researchtoinvestigatearangeofphenomena,e.g.,theon-lineinterpretationof
temporarilyambiguoussentences(see,e.g.,Gibson,Pearlmutter,Canseco-Gonzalez,&Hickock1996)anddifferentkindsofsyntacticdependencies(e.g.,Clifton&Frazier1989),theprocessingofmulti-clausalstructures(Gibson&Warren
2004),etc.Resultsfromthesestudieshaveshownthatadultsareguidedbydifferenttypesofinformationduringparsingincludingphrase-structureinformation,
lexical-semanticinformation,andcontextualinformation(Mitchell2004).
Thereareafewstudiesthatusedself-pacedreadingorlisteningtoexamine
on-line sentence processing in children. Traxler (2002) studied subject-object
ambiguitiesinEnglish-speaking8-12-year-oldsusingtheself-pacedreadingtask.
Thematerialsincluded(i)sentencessuchas(4a)whichareknowntoproducea
cleargarden-patheffectinadults(becausethe girlisinitiallyinterpretedasthe
directobjectoftripped,ananalysisthathastoberevisedlaterintheclause),(ii)
sentences such as (4b) in which the postverbal NP is a semantically implausibleobjectoftheverb,and(iii)sentencessuchas(4c)thatcontainedintransitive
verbs.Thecontrolconditionsforallcaseswerecorrespondingsentencesinwhich
theembeddedverbandthepostverbalNPwereseparatedbyacomma,thereby
precludingthesubject-objectambiguity.
(4) a. WhenSuetrippedthe girl felloverandthevasewasbroken.
b. WhenSuetrippedthe tablefelloverandthevasewasbroken.
c. WhenSuefellthe policemanstoppedandhelpedherup.
Thechildrensreadingtimeswerefoundtobeshorterintheambiguousregion
(showninitalics)andlongerinthedisambiguatingregion(underlined)relative
tothecontrolconditionwithcommas.Likeadults(Traxler2005),8-to12-yearoldchildrentendedtomisanalyzethepostverbalNPinallthreeconditionsasa
directobjectindicatingthatchildren(andadults)preferthestructurallysimpler
analysisirrespectiveofsemanticplausibility.Theeffect,however,waslessstrong
intheintransitivecondition,suggestingthatsubcategorizationinformationwas
atleastpartiallyutilized.
Sentence processing in pre-literate children can be studied using the selfpacedlisteningtechnique,inwhichparticipantslistentosentencesbypressing
apacingbuttontoreceivesuccessivewordsorphrasalsegments.Thistechnique
hasbeenusedsuccessfullywithadults(Ferreira,Henderson,Anes,Weeks,&McFarlane1996;Ferreira,Anes,&Horine1996;amongothers)andhasbeenshown
tobesensitivetothesameeffectsthathavebeenobservedincorrespondingtasks
usingvisualstimuli.Booth,MacWhinneyandHarasaki(2000)investigated8-to
8
HaraldClahsen
12-year-oldchildrenson-linecomprehensionofrelativeclausesusingbothselfpacedreadingandself-pacedlisteningtasks.Theirmaterialsincludeddifferent
kindsofrelativeclausestructures:
(5) a. Themonkeythatfollowedthefrogleftthetreeinahurry.
b. Thedeerthatthetigerwatchedenteredthefieldfromtheside.
In (5a), both the antecedent NP and the relativized NP fulfill the grammatical
functionofsubjectwhereas(5b)containsanobjectrelative.Theresultsrevealeda
slow-downinbothreadingandlisteningtimesattherelativeclausemainclause
transitionforobjectrelatives(e.g.,(5b))comparedtosubjectrelatives(5a)indicatingincreasedprocessingdifficultyfortheformer.Boothetal.sfindingsconfirm
thatself-pacedlisteningandreadingyieldsimilarexperimentaleffects,notonly
inadults,butalsoinchildren.Felser,MarinisandClahsen(2003)andKiddand
Bavin(2007)usedtheself-pacedlisteningtasktoinvestigatehowchildrenprocess
ambiguoussentences.HerewewillconsidertheFelseretal.studyasanexample.
Behavioralmethods
possibletothemainpredicate,yieldingNP1attachment,whereasaccordingto
thelatter,ambiguousmodifiersareattachedtothemostrecentlyprocessedconstituent,yieldingNP2attachment.
Felseretal.(2003)tested6-to7-year-oldchildrenandadultcontrolsinaselfpacedlisteningtask.Theexperimenthada22designwiththefactorsPreposition
and Attachment yielding four conditions as illustrated in (6). All experimental
andfillersentencesweresplitupintofivesegmentsasshownin(7).Disambiguationusinggrammaticalnumberalwaysoccurredonthefourthsegment,i.e.,on
theauxiliary.Toensurethattheexperimentalsentencessoundedequallynaturalin
boththeofandthewithconditions,therelativeorderingofNP1andNP2wasreversedinthewith conditions.Additionaloff-lineandon-linecontrolexperiments
revealedthatNPorderbyitselfdidnotinfluenceattachmentdecisions.
(6) Thedoctorrecognized
a. Of-NP1: thenurseofthepupilswhowasfeelingverytired.
b. Of-NP2: thenurseofthepupilswhowerefeelingverytired.
c. With-NP1: thepupilswiththenursewhowerefeelingverytired.
d. With-NP2: thepupilswiththenursewhowasfeelingverytired.
(7) Thedoctorrecognized/thenurseofthepupils/who/was/feelingvery
tired.
Afterlisteningtoeachsegment,theparticipantswereaskedtopressabuttonona
dualpush-buttonboxasquicklyaspossibleinordertoreceivethenextsegment.
Theendofeachsentencewasindicatedbyatone.Toensurethattheparticipants
paidattentiontothetask,allexperimentalsentencesandhalfofthefillerswere
followedbyacomprehensionquestion,whichwasalsopresentedauditorily.
Table1providesanoverviewofthedifferentparticipantgroupsmeanreactiontimestothedisambiguatingauxiliary.
OnlytheadultgroupshowedasignificantinteractionbetweenPreposition
andAttachment,indicatingthattheirattachmentpreferenceswereinfluencedby
the type of preposition involved. The children differed from the adult controls
inthattheirdisambiguationpreferenceswerenotaffectedbythetypeofprepoTable 1. Meanreactiontimes(inms)forsegment4(adaptedfromFelseretal.2003).
of-NP1
of-NP2
with-NP1
with-NP2
Adults
High-Span
Children
Low-Span
Children
610
665
667
609
863
964
749
807
874
807
916
836
10
HaraldClahsen
sition (of vs. with) at all. Instead, the childrens on-line attachment preferences
were found to interact with their working memory span. While the high-span
childrenshowedapreferenceforNP1attachmentirrespectiveofthepreposition
involved,thelow-spanchildrenshowedanoverallpreferenceforNP2disambiguation.Theseresultsareincontrasttothefindingsfromarecentreadingstudywith
adults (Swets, Desmet, Hambrick, & Ferreira 2007) in which high-span adults
were found to favour local (NP2) attachment of RCs, whereas low-span adults
favourednon-local(NP1)attachment.Itisnotclearwhetherthesediscrepancies
areduetothedifferentmodalities(readingvs.listening)testedorduetodifferencesbetweenchildrenandadults.Inanycase,Felseretal.arguedthatduring
listening,thechildrenappliedoneoftwodifferentphrase-structurebasedlocality
principles,dependingontheirworking-memoryspan.Whereashigh-spanchildrenfollowaPredicateProximitystrategy,low-spanchildrentendtoassociate
therelativeclausewiththemostrecentlyprocessedNP.Thus,similarlytowhat
Traxler(2002)found,childrenseemtoapplythesamekindofphrase-structure
basedparsingheuristicsasadultsbutaremorelimitedintheirabilitytoexploit
lexical-semanticinformationduringon-lineambiguityresolution.
Behavioralmethods
niquecanbeappliedtoawiderangeoflinguisticphenomenaandthatitrequires
minimaltechnicalequipment(essentiallyaPCorLaptopandapush-buttonbox),
whichmakesitsuitableforuseoutsidetheexperimentallaboratory.FortheFelser
etal.(2003)study,forexample,itwasnotpossibletobringchildrenintothelaboratory(astheuniversitywasunwillingtocovertherequiredinsurance).Wethereforehadtoruntheexperimentsatthechildrensschools,whichcouldeasilybe
doneforaself-pacedlisteningexperiment.Thetaskassignedtoparticipantsisnot
particularlydemanding,eventhoughwesawsomeeffectsofworkingmemoryin
thechildren,whichmightreflecttaskdemandsthatdifferbetweenchildrenand
adults. A potential disadvantage of self-paced reading and listening is the segment-by-segmentorword-by-wordstimuluspresentation,whichyieldsrelatively
slow response timesin comparison to, for example, eye movement orERPexperimentsanddoesnotallowparticipantstoreadorlistentothesentencesinthe
usualway,eventhoughnewtechnologiessuchasinstantmessaging,onlinechats,
e-books,podcastsandwebcastsmakebothself-pacedreadingandself-pacedlisteningmorecommonplace.
2.
Cross-modalpriming
Inthistask,participantsarerequiredtonameor,morecommonly,makealexical
decisiontovisualtargetswhilelisteningtostimuluswordsorsentencesspoken
atnormalspeed.Therationaleisthattheprocessingofvisualtargetsisfacilitated
iftheyarepresentedimmediatelyaftertheauditorypresentationofanidentical
orsemanticallyrelatedword,orprime.Insentence-processingresearch,bothon
adultsandchildren,cross-modalpriminghasbeenusedtoexaminetheprocessingofsentence-internalreferentialdependencies,e.g.,bindingprinciples(Nicol
&Swinney1989;McKee,Nicol,&McDaniel1993),andoffiller-gapdependencies such as those in sentences involving wh-movement (e.g., Love & Swinney
1996,2007;Hestvik,Schwartz,Tornyova,&Datta2005;Roberts,Marinis,Felser,
&Clahsen2007)andobjectscrambling(e.g.,Clahsen&Featherston1999;Nakano,Felser,&Clahsen2002).
Withrespecttobindingprinciples,ithasbeenfoundthatinsentencessuch
as(8)bothadults(Nicol&Swinney1989)andpreschoolchildren(McKeeetal.
1993)respondedfastertovisualtargetssuchasLEOPARDinthereflexivethanin
thenon-reflexivecondition.
(8) Thealligatorknowsthattheleopardwithgreeneyesispatting
himself/himontheheadwithapillow.
[LEOPARD]
11
12
HaraldClahsen
This contrast suggests that a binding principle (according to which a reflexive
pronounmustbeboundbyalocalantecedentwithinthesameclause)affectsonlinesentenceprocessinginthatcoreferencebetweenthereflexiveanditsantecedentisimmediatelyestablished;seeMcKee(1996:195ff.)foradetaileddescription
ofthechildversionofthisexperiment.
Severalstudiesusingcross-modalpriminghaveexaminedtheprocessingof
filler-gapdependenciesinadults.LoveandSwinney(1996)studiedEnglishsentencescontainingobject-relativeclauses,suchasJimmy used the new pen that his
mother-in-law recently purchased,inwhichtheobject(the new pen)isdislocated
fromthesubcategorizingverb(purchased).LoveandSwinney(1996)foundthat
lexicaldecisiontimesontargetsappearingattheoffsetofpurchased,wherethe
gapis,weresignificantlyshorterfortargetsthatweresemanticallyrelatedtothe
objectoftheembeddedverbthanforunrelatedones,whereasatacontrolpositionprecedingtheverbpurchased,therewasnosuchdifference.Thesefindings
indicatethattheparserrecoversorreactivatesthegrammaticalandsemanticfeaturesofthedislocatedconstituent(the new pen)atapotentialgapsiteyielding
asemanticprimingeffectatthegappositionbutnotatthecontrolposition.An
alternativeinterpretationofthesefindingsistheso-calleddirectassociationaccountaccordingtowhichadisplacedargumentwillbelinkedtoitssubcategorizingverboncethisisencountered(Pickering1993;Traxler&Pickering1996)
yielding antecedent reactivation on or immediately after the main lexical verb
(purchasedintheexampleabove).
Antecedentprimingin(4-to6-year-old)childrenhasbeenstudiedbyLove
and Swinney (2007), Hestvik et al. (2005) and Roberts et al. (2007). Love and
SwinneyandRobertsetal.adoptedthecross-modalpictureprimingtaskfrom
McKeeetal.(1993)andHestviketal.(2005)usedanalternativepicture-naming
version.LetusconsidertheRobertsetal.studymoreclosely.
Foreachexperimentalsentence,thereweretwovisualtargets,arelatedtarget,
i.e.,apictureoftheindirectobjectnoun(e.g.,apictureofapeacockfor(9)),and
Behavioralmethods
anunrelatedtarget(e.g.,apictureofacarrotfor(9)).Visualtargetswereshown
attwopositions,(i)atthegapposition,i.e.,attheoffsetofthefinalwordofthedirectobjectNP,e.g.,afterpresentin(9),and(ii)atacontrolposition500msearlier.
Eachexperimentalsentencewaspresentedidenticallytofourgroupsofsubjects:the
firstonesawtherelatedtargetatthegapposition,thesecondgroupatthecontrol
position;thethirdgroupsawtheunrelatedtargetatthegapposition,andthefourth
groupatthecontrolposition.Duringthepresentationofthesentences,pictures
appearedonthecomputerscreen,andtheparticipantswererequiredtodecide
whethertheanimal/objectinthepicturewasaliveornot alive,bypushingbuttons
onapush-buttonbox.Responsetimesweremeasuredfromthepointatwhich
thepictureappearedonthescreentotheparticipantspressingoftheresponse
button.Toensurethattheparticipantspaidattentiontothetask,theywerealso
asked to respond to (yes-no) comprehension questions randomly interspersed
throughouttheexperimentaskingforoneofthemaincharacters.
The results from this study showed that the childrens reaction times were
sloweroverallthantheadultsandthatchildrensandadultsprocessingoffillergapdependencieswasaffectedbyworkingmemorydifferences.Forchildrenand
adultswithahighworkingmemoryspan,aPositionTargetTypeinteraction
was found indicating antecedent reactivation at the gap position in these participants.Allhigh-spanparticipantsrespondedmorequicklytoidenticalthanto
unrelatedpicturetargetsatthegapposition,andlexicaldecisiontimesforidenticalwereshorteratthegappositionthanattheearliercontrolposition;these
contrastsareillustratedinTable2.
Low-span children and adults, on the other hand, did not show any antecedent reactivation at the gap position. Interestingly, this did not compromise
theirabilitytounderstandtheexperimentalsentences,sincetheyansweredthe
comprehensionquestionsthatwereaskedaftertheauditorystimuliasaccurately
asthehigh-spanparticipants.Inanycase,thefindingthatworkingmemoryisa
relevantvariablefordiscoveringantecedentreactivationeffectsisconsistentwith
theresultsofearlierstudiesshowingthattheprocessingofcomplexsentencesin
Table 2. High-spanchildrenandadultsmeanreactiontimes(inms)topicturetargets
(adaptedfromRobertsetal.2007).
High-Span Children (N = 19)
Control Position Trace Position
Identical
Targets
Unrelated
Targets
1245
1158
694
678
1158
1211
692
709
13
1
HaraldClahsen
general,andoffiller-gapconstructionsinparticular,incursaworkingmemory
costinadults(Gibson&Warren2004;King&Kutas1995;Miyamoto&Takahashi2001;Nakanoetal.2002)andthatforchildren,memorycapacitymaybea
predictorofeffectivelanguageprocessing(e.g.,Boothetal.1999,2000;Gathercole&Baddeley1989).Oneconsequenceofthisisthatstudiesofsentenceprocessinginchildren(particularlyofcomplexsentences)shouldbeaccompaniedby
aworkingmemorytest(alongwithotherpretests).
Behavioralmethods
tualantecedentisthatparticipantsmightrealizethattargetswereprecedingwords
andstarttoanticipatethis,whichwouldchangethenatureofthetask,makingit
aconsciousrecalltaskratherthananunconsciousmeasureofon-lineprocessing.
Thiscanbeavoided,however,byusingalargerproportionofunrelatedtargetsthan
usual.Instudiesthatuseidentitytargets,theproportionofsentenceswithidenticaltargetstothosewithunrelatedtargetsisaboutoneintwenty,makingconscious
detectionofrepeatedwordsanunlikelypossibility.Moreover,anyamountofprimingduetotheformalorsemanticidentityoftheantecedentandthetargetcanbe
factoredoutbycomparingprimingeffectsatthegappositionwiththoseonthe
sametargetwordatcontrolpositions.
Detailedmethodologicaladviceforconstructingacross-modalprimingexperimentisgiveninMcKee(1996).Inadditiontothepointsmentionedthere,a
numberofpretestsarerequiredfortheconstructionofappropriatematerialsand
to rule out potentially confounding factors. For a picture-priming experiment,
a picture-classification task is necessary to ensure that the children are able to
correctlyclassifythetargetpicturesasaliveornotalive.Moreover,ifcomplex
sentencessuchasthosein(9)aretobeexaminedinthemainexperiment,the
childrensabilitytocomprehendthesekindsofsentencesneedstobepre-tested
alongwiththeirworkingmemoryspan.
Cross-modalprimingofferssomeadvantagesoverotherbehavioralmethods
for studying on-line sentence processing in children. It allows for the stimulus
materialstobepresenteduninterruptedandatanormalspeechrate,thusrenderingitmorenaturalthan,forexample,self-pacedreadingorlistening.Theuseof
picturetargetsinsteadofwrittenwordsmakesthetechniquesuitableforyoung
children who cannot yet read or write. Moreover, the possibility of presenting
visual targets at different positions during the auditory stimulus allows for the
discoveryofpotentialprimingeffectsoccurringon-lineatspecificpositionsin
thesentence.Anotheradvantageisthatparticipantsattentionisdirectedtowards
thedecisiontaskonthevisualtargetsandawayfromthesentencestimulithat
the researcher is interested in, thereby reducing the possibility of participants
responsesbeinginfluencedbyconsciousprocesses.
One disadvantage of this technique is that, unlike ERPs or eye movement
measures,cross-modalprimingdoesnotprovideacontinuousrecordofsentence
processing,butinsteadmoreofasnapshotviewofthestateofthelanguageprocessoratparticularpositionsinthesentence,namelyatthespecifictestpointsat
whichvisualtargetsarepresented.Anotherdisadvantageofcross-modalpriming
isthecomplexityofthetask,attentiontotheauditorysentencewhileperforming
adecisiontaskonapicturetarget.Thesedual-taskdemandsmighthavebeenthe
reasonwhyRobertsetal.(2007)failedtodetectantecedentreactivationinadults
andchildrenwithlowworkingmemory.Asanalternative,Hestviketal.(2005)
1
16
HaraldClahsen
haveproposedasupposedlymorechild-friendlypicture-namingversionofthe
taskinwhichpicturesshownduringlisteningtothesentenceshavetobenamed.
Hestviketal.foundantecedentreactivationeffectsin8-to11-year-oldchildren,
notonlyforthosewithhighbutalsoforthosewithlowworkingmemory.Note,
however, that their materials were simpler than those of Roberts et al. (2007),
involvingdirect(ratherthanindirect)objectgaps.Thus,Hestviketal.sfindings
replicateforolderchildrenwhatLoveandSwinney(2007)havealreadyfoundfor
4-to-6-year-oldsusingthealive/non-alivedecisiontask,buttheydonotnecessarilyshowthatthepicture-namingversionofthetaskistobepreferred.
3.
Whiletheadultpsycholinguisticliteratureoffersarangeofdifferentbehavioral
methodsforinvestigatinglanguagecomprehension,thereareonlyafewexperimentalparadigmsavailablethattapprocessesduringlanguageproduction,e.g.,
implicit priming (Roelofs 2002, among others), the picture-word interference
paradigm(e.g.,Schriefers,Meyer,&Levelt1990),syntacticpriming(e.g.,Pickering&Branigan1998),andspeededproduction(e.g.,Prasada,Pinker,&Snyder
1990).Ofthesethelattertwohavebeenadaptedtothestudyofchildrenslanguageproduction.Inthefollowing,Iwillfirstbrieflyexplainsyntacticpriming
andtheninsomemoredetailthespeededproductiontask.
3.1
Syntacticpriming
Whenpeopleproducesentencestheyarelikelytomaintainaspectsofsyntactic
structurefromonesentencetothenext,aphenomenonthatiscalledsyntactic
priming. The conditions under which syntactic priming occurs are thought to
reveal aspects of grammatical encoding during production. In syntactic primingstudies(see,e.g.,Bock,Loebell,&Morey1992;Pickering&Branigan1998),
subjectsprovidecontinuationsforpartialsentencesofbothprimefragmentsand
targetsentences.Primefragmentsaresuchthatthemostlikelycompletionisofa
particularform;forexample,forfragmentssuchas(10a)and(10c)acompletion
withaprepositionalobjectishighlylikely.Bycontrast,thetargetfragments(10b)
and(10d)endaftertheverbsothatsubjectshaveachoicebetweenaprepositional
andadouble-objectconstructionfortheircontinuation.
(10) Primingofprepositionalobjectconstruction
a. Therockstarsoldsomecocaine to an undercover agent
b. Thegirlhanded the paintbrush to the man
Behavioralmethods
Primingofdouble-objectconstruction
e. Therockstarsoldanundercoveragent......some cocaine
f. Thegirlhanded the man the paintbrush
Examiningthree-placepredicatessuchasthosein(10b),(10d),and(10f)astargets,PickeringandBranigan(1998)foundprimingeffectscausedbyprepositional object constructions (10a) and (10c), and double-object constructions (10e)
inadults.Thus,thecompletionof(10a)primedparticipantstocomplete(10b)
using a prepositional rather than a double-object construction and vice versa
foraprimesuchas(10e).Interestingly,itwasfoundthatwhileaprepositional
phrasewithfor(10c)primedtheprepositionalobjectconstruction(withto)for
give(10d),thesamelexicalitem(to)inadifferentsyntacticfunction(e.g.,inMary
brought a book to study),doesnotprimetheprepositionalobjectconstructionfor
give.Thisfindingsuggeststhatproductionprimingeffectsareabstractandsyntacticinnatureratherthanpurelybasedonlexicalinformation.Syntacticpriminginthesecaseshasbeenexplainedintermsofphrase-structurerules.Thus,the
constructionofaprimesentencesuchas(10a)involvesarulethatexpandsaVP
intoV+NP+PP.Onceemployedfortheconstructionoftheprimesentence,the
rulemayretainsomeresidualactivitywhenthetargetfragmentiscompletedthus
makinganNP+PPcompletionmorelikelythananNP+NPcompletion.
Productionpriminghasalsobeenusedwithchildren,withmodificationsin
the design. For example, Savage, Lieven, Theakston and Tomasello (2003) presented4-to-6-year-oldchildrenwithprimesentencesspokenbytheexperimenter
alongwithaprimepicture.Thechildwasthenaskedtorepeattheprimesentence, e.g., an active or a passive sentence. Then, the child was presented with
the target picture and asked Whats happening here? to examine whether the
childwaspromptedtoproduceanactiveorapassivesentencedependingonthe
primesentencepresentedbefore.Savageetal.(2003,2006)obtainedprimingeffects for 6-year-olds, whereas for 4-year-olds priming effects only occurred in
cases of high lexical overlap between primes and targets, i.e., when the prime
sentencecontainedthesamelexicalverbasrequiredforthetargetpicture.Savage
etal.tookthistomeanthat4-year-oldsrepresentationsofactivesandpassivesare
lexically-specificratherthanabstractorsyntacticinnature.Ontheotherhand,
Huttenlocher,VasilyevaandShimpi(2004)foundastructuralprimingeffectfor
4-year-oldsevenwhentheprimeandthetargetsentencesdidnotsharelexical
contentwords.Here,Iwillnotfurtherdiscusstheseconflictingresults.Itshould
benoted,however,thatSavageetal.sclaimsaboutthelackofabstractsyntactic
knowledgeinchildrenareonlybasedonproductionmeasures,whichmayunder-
17
18
HaraldClahsen
estimateachildslinguisticknowledge.Theirviewthat4-year-oldslackabstract
syntacticknowledgewouldbemoreconvincingiftheyhadconvergingevidence
fromothersources,e.g.,fromcomprehensionmeasures.
Forourpresentconcerns,itisimportanttonotethatthesyntacticpriming
technique as it stands is not time-sensitive as it does not provide any measure
ofthetime-courseofgrammaticalencoding.Wemayask,forexample,whether
priming effects in this task unfold predictively and how this precisely happens
overtime;seeSnedekerandThothathiri(thisvolume)forwaysofincorporating
time-sensitivemeasuresintothesyntacticprimingtask.
3.2
Speededproduction
Themeasurementofproductionlatenciesoffersawaytoexamineautomaticprocessesinvolvedinchildrensspokenproductions.Inaspeeded-productiontask,
participantsareaskedtoproduceasquicklyandaccuratelyaspossibleaparticularwordform,e.g.,aninflectedform(walked)foranauditorilypresentedverb
stem(walk).Accuracyratesandproductionlatenciesaremeasured,thelatterof
whichprovidethecrucialon-linemeasure.
Several research teams have used this technique to examine potential processing differences between regular and irregular inflection in adults (Prasada
etal.1990;Ullman1993;Beck1997;Lalleman,vanSanten,&vanHeuven1997;
Buck-Gengler,Menn,&Healy2004).Thepurposeofthesestudieswastodeterminetowhatextentthereal-timeproductionofaninflectedwordreliesonlexical
look-up,i.e.,uponretrievalofwholewordformsstoredinmemory,andtowhat
extentitdependsoncomputationalprocessesof,forexample,combiningstems
orrootswithaffixes(walk+-ed).Therationaleisthatifaninflectedwordformis
storedasawhole,thenretrievalshouldbefasterforhigh-frequencythanforlowfrequencyones,andthiscontrastshouldbemeasurableinproductionlatencies.
Thisisasensibleassumption,sincelexicalretrievalandstorageareknowntobe
affected by a words frequency. On the other hand, if regularly inflected forms
arecomputedfromtheirmorphologicalconstituentsduringproduction(rather
thanretrievedaswholewordformsfrommemory),thenthewordfrequencyofa
regularlyinflectedform(e.g.,thefrequencyofwalked)shouldnotaffectproductionlatencies.Hence,oftworegularlyinflectedformsthathavethesamestem
frequencybutdifferwithrespecttotheirpast-tensefrequency(e.g.,jumpandboil
whichbothhaveastemfrequencyof26permillionandpast-tensefrequencies
of32forjumpedand1forboiled, see Prasadaetal.1990),producingtheonewith
thelowerpast-tensefrequencyshouldnottakelongerthantheproductionofthe
high-frequencywordform.
Behavioralmethods
Thisparadigmhasproducedreasonablyclearandreplicableeffectsforadults.
Allthestudiesmentionedabovefoundafrequencyadvantageforirregulars,i.e.,
shorterresponsetimesforhigh-frequencythanforlow-frequencyirregulars,and
no corresponding advantage for high-frequency forms amongst regulars (see
Pinker 1999:129ff. for review). Let us consider more closely a study (Clahsen,
Hadler,&Weyerts2004)inwhichthespeededproductiontaskwasusedtoexaminemorphologicalprocessinginchildren.
19
20 HaraldClahsen
Table 3. Meanproductionlatencies(inms.)forhighandlow-frequency(regularand
irregular)participleforms(adaptedfromClahsenetal.2004).
Irregular-High
Irregular-Low
Regular-High
Regular-Low
Adults
(N = 35)
5- to 7-Year-Olds
(N = 20)
11- to 12-Year-Olds
(N = 20)
947
1002
958
947
1223
1283
1257
1188
1078
1130
1088
1049
The results from this study, presented in Table 3, can be summarized in three
points. Firstly, while adults produced hardly any morphological errors in participleformation,childrenwerefoundtooverregularizetheregular-tsuffixto
verbsthatrequiretheirregular-nsuffix,withhighererrorratesonlowthanon
high-frequencyirregulars.Bycontrast,over-applicationsoftheirregularsuffixto
regularverbswereextremelyrare(lessthan1%).Secondly,theoverallproductionlatencieswerefoundtodecreasewithage,5-to7-year-oldshavingamean
productionlatencyof1238ms,11-to12-year-oldsof1086ms,andadultsof963
ms.Thirdly,whereasallparticipantgroupshadshorterproductionlatenciesfor
high-frequencyirregularsthanforlow-frequencyones,bothagegroupsofchildrenshowedareversefrequencyeffectforregulars,longerproductionlatencies
forhighthanforlow-frequencyregulars.Thiscontrastwasmorepronouncedfor
the5-to7-year-oldchildrenthanforthe11-to12-year-olds.
Theseresultscanbetakentoindicatethattwomechanismsformorphologicalprocessing,lexicalstorageandmorphologicalcomputation,areemployedby
childrenaswellasbyadultsbutthatlexicalaccessislessefficientforchildren.
Overregularization errors arise when access to the lexical entry of an irregular
formfails.Consequently,childrenproducemoreofsucherrorsthanadults.Children took longer to produce participles than adults, another indication of less
efficientlexicalaccess.Finally,reverse-frequencyeffectsarisefromtheretrievalof
storedhigh-frequencyregularsthatinhibitmorphologicalcomputation(Pinker
1999).Hencetheproductionofhigh-frequencyregularsinvolvesmemoryaccess,
andthisinterfereswithmorphologicalcomputation(whichisavailableforboth
highandlow-frequencyregulars)inthatitslowsdowntheproductionofhighfrequencyregularsrelativetolow-frequencyonesforwhichmorphologicalcomputationisnotimpededbyanystoredforms.Slowlexicalretrievalincreasesthis
contrast,hencethedecreaseofthereversefrequencyeffectfromtheyoungerto
theolderchildgroups.
Behavioralmethods
21
22
HaraldClahsen
dren,frequencyinformationshouldbegatheredfromcorporaofspokenlanguage
andideallyfromcorporaofchild-directedspeech,becausefrequencydictionaries
orcorporaofwrittenlanguage(e.g.,newspapers)maycontainwordsunfamiliar
toyoungchildren.
Summarizing,thespeededproductiontaskprovidesanefficientmeasureof
processesinvolvedinlanguageproduction.Thetechniqueoffersatime-sensitive
measure,eventhoughtheresponselatenciesdonotonlyreflectproductionprocessesbutalsoincludethetimeneededforrecognizingtheverbstempresented.
Stimulus presentation is fairly natural requiring normal listening, but various
potentiallyconfoundingfactors(mentionedabove)needtoconsideredtoavoid
artifacts.Thetaskisnotdemandingand,inthemodifiedversion(Clahsenetal.
2004),appropriateforchildrenabovetheageof5.Indeed,noneofthechildren
testedfoundthetaskparticularlydifficult,andmostofthemenjoyedtheexperiment.Moreover,thetaskrequiresminimaltechnicalapparatus(aPCandamicrophone)andcanbeperformedinanyquietroom,evenoutsidetheresearch
laboratory.Asregardsitslinguisticversatility,thespeeded-productiontaskseems
tobewell-suitedtoexamineword-levelprocessing,e.g.,theproductionofinflectedwordformsofbarestems,pluralsfromsingulars,etc.,butlesssoforstudying
sentence-levelprocessing.
.
Summary
Thethreebehavioralmethodswefocusedon,self-pacedreadingandlistening,
cross-modal priming, and speededproduction, allprovide time-sensitive measures, an essential requirement for studying on-line processing. Unfortunately,
however,eachofthesemeasureshasitslimitations.Self-pacedreadingandlisteninghavearelativelylowtemporalresolution(compared,forexample,toERPsand
eyemovementmeasures)duetothewaythestimuliarepresented.Incross-modal
primingexperiments,responsetimesareonlymeasuredatspecifictestpointsin
asentencethusprovidingasnapshotviewofthestateofthelanguageprocessor
atthesepointsratherthanacontinuousmeasureofon-linesentenceprocessing.
Theresponselatenciesthataremeasuredinthespeeded-productiontaskinclude
therecognitiontimesrequiredfortheauditorystimuliandcannotbetakenasa
puremeasureoflanguageproductionprocesses.
Clearly,eyetrackingandERPsprovidebettermeasuresofthetime-courseof
processing,butthebehavioralmethodsdiscussedinthischapterwillnodoubt
haveaplaceinfutureresearchinthisfield,andthisisforanumberofreasons.
Firstly,anypsycholinguistictechnique(includingeyetrackingandERPs)hasits
limitationsandisindangerofproducingartifacts,e.g.,duetoanexperiments
Behavioralmethods
specifictaskdemands.Onewayaroundthisproblemistofindconvergingevidencefromothersources,e.g.,byreplicatinganeffectseenwithonetechnique
withadifferenttechnique.Behavioraltechniquescanbeusefulforthispurpose.
Secondly,behavioraltechniquesrequirerelativelylittletechnicalequipmentand
canbeadministeredwithoutbringingchildrenintotheresearchlaboratory.This
makes them ideally suited for piloting experimental designs, and for working
withpopulationsinout-of-reachplaces.Thirdly,comparedto,forexample,ERP
experiments which require many items per condition due to signal averaging,
the behavioral methods mentioned above require fewer critical items and can
typicallybeadministeredwithinasingleexperimentalsession,thusavoidingpotentialartifactssuchasthosecausedbytrainingeffectsandfatigue.Finally,unlikeERPoreyetrackingexperimentsforwhichanelectrodecaporahead-band
needstobeattachedtothechild,behavioraltechniquesdonotrequireanydirect
physicalcontactwithaparticipant.Itwillthereforebemucheasiertogetethical
approvalforbehavioralexperimentsthanforanytechniqueinvolvingphysiologicalmeasures,andtheremaybecircumstancesinwhichsuchconsiderationsare
adecisivefactor.
Acknowledgements
I am grateful to Sonja Eisenbeiss, Claudia Felser, Eva Fernndez, Ted Gibson,
TheodoreMarinis,LeahRoberts,andMattTraxlerforcommentsonanearlier
versionofthischapter.
References
Beck,M.-L.(1997).Regularverbs,pasttenseandfrequency:Trackingdownapotentialsource
ofNS/NNScompetencedifferences.Second Language Research,13,93115.
Bever,T.G.&McElree,B.(1988).Emptycategoriesaccesstheirantecedentsduringcomprehension.LinguisticInquiry,19,3543.
Bever,T.&Sanz,M.(1997).Emptycategoriesaccesstheirantecedentsduringcomprehension.
UnaccusativesinSpanish.LinguisticInquiry, 28,6891.
Bock,K.,Loebell,H.&Morey,R.(1992).Fromconceptualrolestostructuralrelations:Bridgingthesyntacticcleft.Psychological Review, 99,150171.
Booth,J.,Perfetti,C.,&MacWhinney,B.(1999).Quick,automaticandgeneralactivationof
orthographicandphonologicalrepresentationinyoungreaders.Developmental Psychology,35,319.
Booth, J., MacWhinney, B., & Harasaki, Y. (2000). Developmental differences in visual and
auditoryprocessingofcomplexsentences.Child Development,71,9811003.
23
2
HaraldClahsen
Buck-Gengler, C., Menn, L., & Healy, A. (2004). What mice trap tells us about the mental
lexicon.Brain and Language, 90,453464.
Carreiras,M.&Clifton,C.(1993).RelativeclauseinterpretationpreferencesinSpanishand
English. Language and Speech,36,353372.
Carreiras,M.&Clifton,C.(1999).Anotherwordonparsingrelativeclauses:EyetrackingevidencefromSpanishandEnglish.MemoryandCognition,27,826833.
Clahsen,H.&Featherston,S.(1999).Antecedentprimingattracepositions:Evidencefrom
Germanscrambling.Journal of Psycholinguistic Research,28,415437.
Clahsen,H.,Hadler,M.,&Weyerts,H.(2004).Speededproductionofinflectedwordsinchildrenandadults.Journal of Child Language,31,683712.
Carreiras,M.&Clifton,C.(1999).Anotherwordonparsingrelativeclauses:EyetrackingevidencefromSpanishandEnglish.Memory and Cognition,27,826833.
Clifton,C.&Frazier,L.(1989).Comprehendingsentenceswithlong-distancedependencies.
InG.M.Carlson&M.K.Tanenhaus(Eds.),Linguistic structure in language processing.
Dordrecht:Kluwer.
Cuetos,F.Mitchell,D.,&Corley,M.(1996).Parsingindifferentlanguages.InM.Carreiras,J.
Garca-Albea,&N.Sebastin-Galls(Eds.),Language processing in Spanish(pp.145187).
MahwahNJ:LawrenceErlbaumAssociates.
DeVincenzi,M.&Job,R.(1993).Someobservationsontheuniversalityofthelate-closure
strategy.Journal of Psycholinguistic Research,22,189206
Felser,C.,Marinis,T.,&Clahsen,H.(2003).Childrensprocessingofambiguoussentences:A
studyofrelativeclauseattachment.Language Acquisition,11,127163.
Fernndez,E.M.(2003).Bilingualsentenceprocessing, Amsterdam: JohnBenjamins.
Ferreira,F.,Henderson,J.Anes,M.Weeks,P,&McFarlane,D.(1996).Effectsoflexicalfrequencyandsyntacticcomplexityinspokenlanguagecomprehension:Evidencefromtheauditory moving window technique. Journal of Experimental Psychology: Learning, Memory
and Cognition, 22,324335.
Ferreira, F., Anes, M., & Horine, M. (1996). Exploring the use of prosody during language
comprehensionusingtheauditorymovingwindowtechnique.Journal of Psycholinguistic
Research,25,273290.
Frazier,L.&Clifton,C.(1996).Construal.CambridgeMA:TheMITPress.
Gathercole,S.,&Baddeley,A.(1989).EvaluationoftheroleofphonologicalSTMinthedevelopmentofvocabularyinchildren:Alongitudinalstudy.Journal of Memory and Language,
28,200213.
Gibson,E.,Pearlmutter,N.,Canseco-Gonzalez,E.,&Hickock,G.(1996).Recencypreferences
inthehumansentenceprocessingmechanism.Cognition,59,2359.
Gibson,E.&Warren,T.(2004).Reading-timeevidenceforintermediatelinguisticstructurein
long-distancedependencies.Syntax,7,5578.
Gilboy, E., Sopena, J., Clifton, C., & Frazier, L. (1995). Argument structure and association
preferencesinSpanishandEnglishcompoundNPs.Cognition, 54,131167.
Gordon,P.C.,Hendrick,R.,&Foster,K.L.(2000).Languagecomprehensionandprobelist
memory. Journal of Experimental Psychology: Learning, Memory, & Cognition, 26, 766
775.
Hestvik, A., Schwartz, R., Tornyova L., & Datta, H. (2005). Picture-naming shows children
reactivateantecedentsattracepositions.PosterpresentedattheGenerativeApproachesto
LanguageAcquisition(GALA)Conference,Siena,Italy.
Behavioralmethods
2
26 HaraldClahsen
Mitchell,D.C.(2004).On-linemethodsinlanguageprocessing:Introductionandhistorical
review.InM.Carreiras&C.Clifton(Eds.),The on-line study of sentence comprehension:
Eyetracking, ERP and beyond(pp.1532).Hove:PsychologyPress.
Miyamoto,E.&Takahashi,S.(2001).Antecedentreactivationintheprocessingofscrambling
inJapanese. MIT Working Papers in Linguistics,43,127142.
Nakano,Y.,Felser,C.,&Clahsen,H.(2002).AntecedentprimingattracepositionsinJapanese
long-distancescrambling.Journal of Psycholinguistic Research, 31,531571.
Nicol,J.&Swinney,D.(1989).Theroleofstructureincoreferenceassignmentduringsentence
comprehension.Journal of Psycholinguistic Research, 18,520.
Nicol,J.,Swinney,D.,Love,T.,&Hald,L.(2006).Theon-linestudyofsentencecomprehension:
Examinationofdual-taskparadigms.Journal of Psycholinguistic Research,35,215231.
Pickering, M. & Branigan, H. (1998). The representation of verbs: Evidence from syntactic
priminginlanguageproduction.Journal of Memory and Language,39,633651.
Pickering,M.(1993).Directassociationandsentenceprocessing:AreplytoGibson&Hickok.
Language and Cognitive Processes,8,163196.
Pinker,S.(1999).Words and rules. The ingredients of language.NewYorkNY:BasicBooks.
Prasada,S.,Pinker,S.,&Snyder,W.(1990).Someevidencethatirregularformsareretrieved
frommemorybutregularformsarerule-generated. Paperpresentedatthe31stAnnual
MeetingofthePsychonomicSociety,NewOrleans,LA.
Roberts,L.,Marinis,T.,Felser,C.,&Clahsen,H.(2007).Antecedentprimingattracepositions
inchildrenssentenceprocessing.Journal of Psycholinguistic Research,36(2),175188.
Roelofs,A.(2002).Storageandcomputationinspokenwordproduction.InS.Nooteboom,F.
Weerman,&F.Wijnen(Eds.),Storage and computation in the language faculty (pp.183
216).Dordrecht:Kluwer.
Savage,C.,Lieven,E.,Theakston,A.,&Tomasello,M.(2003).Testingtheabstractnessofchildrenslinguisticrepresentations:Lexicalandstructuralprimingofsyntacticconstructions
inyoungchildren.Developmental Science, 6,557567.
Savage,C.,Lieven,E.,Theakston,A.,&Tomasello,M.(2006).Structuralprimingasimplicit
learninginlanguageacquisition:Thepersistenceoflexicalandstructuralprimingin4year-olds.Language Learning and Development, 2,2749.
Schriefers,H.,Meyer,A.,&Levelt,W.(1990).Exploringthetimecourseoflexicalaccessin
language production: Picture-word interference. Journal of Memory and Language, 29,
86102.
Schtze,C.(1996).The empirical base of linguistics: Grammaticality judgments and linguistic
methodology.ChicagoIL:UniversityofChicagoPress.
Swets,B,Desmet,T.,Hambrick,D.Z.,&Ferreira,F.(2007).Theroleofworkingmemoryin
syntacticambiguityresolution:Apsychometricapproach.Journal of Experimental Psychology: General,136(1),6481.
Traxler, M. (2002). Plausibility and subcategorization preference in childrens processing of
temporarilyambiguoussentences:Evidencefromself-pacedreading.Quarterly Journal of
Experimental Psychology,55A,7596.
Traxler,M.(2005).Plausibilityandverbsubcategorizationintemporarilyambiguoussentences:
Evidencefromself-pacedreading.Journal of Psycholinguistic Research,34,130.
Traxler,M.&Pickering,M.(1996).Plausibilityandtheprocessingofunboundeddependencies:Aneye-trackingstudy.Journal of Memory and Language, 35,454475.
Traxler, M, Pickering, M., & Clifton, C. (1998). Adjunct attachment is not a form of lexical
ambiguityresolution.Journal of Memory and Language, 39,558592.
Behavioralmethods
Tyler,L.&Marslen-Wilson,W.(1981).Childrensprocessingofspokenlanguage.Journal of
Verbal Learning and Verbal Behavior, 20,400416.
Ullman,M.T.(1993).Thecomputationofinflectionalmorphology.PhDdissertation,MIT.
Wulfeck,B.(1993).Areactiontimestudyofgrammaticalityjudgmentsinchildren.Journal of
Speech and Hearing Research, 36, 12081215.
27
chapter2
Thepresentpapergivesanoverviewofourrecentresearchontheneurocognitionoflanguageacquisition.Ourresearchaimstogainamoredetailedunderstandingofthedevelopmentalstagesofthelanguageacquisitionprocessand
itsunderlyingbrainmechanisms.Here,weutilizethemethodofevent-related
brainpotentials,whichhasrevealedspecificelectrophysiologicalindicesfor
variousaspectsoflanguageprocessinginadults.Theseelectrophysiologicalparameterscanserveastemplatestodefinethehallmarksoflanguageacquisition.
Theresearchpresenteddemonstratesthatthemethodofevent-relatedbrain
potentialsisapowerfultooltoinvestigateandmonitorearlystagesoflanguage
acquisitionandprovidesfurtherinsightsintotheneuralcorrelatesoflanguage
processingininfantsandchildren.
1.
Introduction
The wonder of language acquisition, with its remarkable speed and high levels
of success, remains a mystery. At birth, infants are able to communicate their
basicneedsbydifferentwaysofcrying.Also,frombirthon,infantsshowapreferenceforthesoundoftheirnativelanguage.Followingthesefirstlanguage-related
stepsthereisafastprogressioninthedevelopmentofperceptiveandexpressive
language skills. At around 4 months, babies start to babble, the earliest stages
oflanguageproduction.Ameretwelvemonthsafterbirthmostbabiesstartto
speaktheirfirstwords,andabouthalfayearlaterevenshortsentences.Finally,
at the end of most childrens third year of life, they have acquired at least 500
wordsandknowhowtocombinethemintomeaningfulutterances.Thus,they
havemasteredtheentryintotheirnativelanguage:theyhaveacquiredacomplex
systemwiththetypicalsoundsofalanguage,thesesoundsarecombinedindif-
30
ClaudiaMnnelandAngelaD.Friederici
ferentwaystomakeupalargevocabulary,andthevocabularyentriesarerelated
togetherbymeansofsyntacticrules.Allofthisoccursatanamazingspeedandto
someextentindependentlyoftheenvironmentalconditions.
Althoughaquitedetailedoutlineofthelanguageacquisitionprocessexists
(seeBee&Boyd2007;Clark2003;Klann-Delius1999;Szagun2006),manyquestionsremain.Forexample,howdochildrenactuallylearntheirmothertongue
soeasilyandwhataretheexactdevelopmentalstages?Doeslanguageacquisition
happeninacontinuousmanneroradiscontinuousmanner?Here,thecontinuity
hypothesisoflanguageacquisitionholdsthatlanguageprocesseschangequantitativelyoverthecourseofthechildsdevelopmentbutthattheprocessesarein
principlepresentfromearlyonandinasimilarformtothewaytheyarepresent
inadults.Incontrast,thediscontinuity hypothesisproposesqualitativeratherthan
quantitativechangesduringthechildsdevelopmentuntilanadult-likestatusis
reached.Inthiscontext,then,whattypesofdevelopmentalchangesoccurinthe
brainduringlanguagedevelopment?Inotherwords,whatistheneurophysiologicalbasisofthevariousprocessesandstepsinlanguageacquisition?Studyinghow
childrensoreadilyacquirelanguageisnoteasilyaccomplished,becauseagood
dealoflearningtakesplacebeforethechildisabletospeakandshowovertresponsestowhatheorsheactuallyperceives.Forexample,itisnotaninsignificant
methodologicalchallengetodevelopwaysofaskingchildrenwhethertheyknow
awordbeforetheycandemonstrateaverbalreactiontothisquestion.Childrens
perceptivelanguageskillsdevelopmuchearlierthantheirexpressiveskillsbutare
forthemostpartbeyondthescopeofobservation.
Themethodofevent-relatedbrainpotentials(ERP)allowsustovirtuallylook
intothebrain,wheretheacquisitionoflanguageistakingplace,duringthecourse
ofacquisition.TheuseoftheERPmethodtoinvestigateon-linecognitiveprocessesinadultshasbeensuccessfullyproven.Inadults,thereareparticularERP
componentsthatappeartobespecifictovariousaspectsoflanguageprocessing.
Wealsousethismethodtostudylanguageacquisitionininfantsoverthecourse
ofthefirstthreeyearsoflifesincethismethodinheritsseveraladvantagesthat
becomeapparentspecificallyintheworkwithinfants.
Inthischapter,wedemonstratethattheERPmethodisapowerfultoolto
investigateandmonitorearlystagesoflanguageacquisition.Specifically,weshow
thattheERPmethoddeliversinformationabouttheneuralcorrelatesoflanguage
processesandthereforeprovidesabetterunderstandingofthehowandwhenof
thedevelopmentalstagesinthelanguageacquisitionprocess.First,weexplainthe
methodathandthoroughly(somedetailsaboutitssuccessfulapplicationinwork
withinfantsareprovidedintheAppendix).Then,webrieflyoutlinetheresearch
fieldaswellassketchthedevelopmentalstagesinlanguageacquisitionandtheir
associatedERPcomponents.Then,wegiveanoverviewofourownERPresearch
Event-relatedbrainpotentials
onthedifferentaspectsoflanguageacquisitionduringinfantsfirstthreeyearsof
life.Thus,wedescribe,bymeansofERP,theprocessingofphonological,semantic,andsyntacticinformationininfantsandchildrenanddiscusstheseresultsin
thelightofotherneurophysiologicalandbehavioralstudies.
2.
2.1
Whatismeasured?EEGandERP
Electroencephalography(EEG)isanon-invasivemethodusedtomeasurevoltage
fluctuationsonthescalpssurface.Thesevoltagefluctuationscomprisesummed
post-synapticelectricpotentialsgeneratedbysimilarlyalignedandsimultaneouslyfiringpyramidalcellsintheneocortex(formoredetailseeLopesdaSilva1991;
Speckman&Elger1993).Consequently,therecordedelectricactivityreflectsa
widerangeofbrainfunctions,includingvariousstatesofactivation,relaxation,
tiredness, engagement in cognitive tasks, etc. EEG data recorded for a certain
time period contain background activity as well as changes of electric activity
inresponsetosingleevents,suchaswordsandsentences.Inresponsetothose
events,therecordedbrainsignalcanbebrokenupintoitsfrequencybands,i.e.,
regularpatternsofelectricpotentialsinadefinedtimewindow(Fourieranalysis).
Therecordedsignalcanalsobeanalyzedforvoltagefluctuationsthataretimelockedtosensory,motor,andspecificallycognitiveevents,so-calledevent-related
brainpotentials(ERPs;forreviewsseeColes&Rugg1995;Fabiani,Gratton,&
Coles2000;Regan1989).Theseevent-relatedvoltagechangesarerelativelysmall
comparedtotheongoingbackgroundEEGactivity.Forthisreason,theinterpretationofEEGrawdatainrelationtosingleeventsis,ifnotimpossible,atleast
verydifficult(althoughtherehasbeensomeadvancementinsingle-trialanalysis;
seeforinstanceBansal,Sun,&Sclabassi2004;Holm,Ranta-Aho,Sallinen,Karjalainen,&Mller2006;Jungetal.2001).Toovercomethisproblem,repeated
presentationofthestimulusofinterest(e.g.,aspokensentence)andsubsequent
averagingarerequired.Inordertoanalyzetherecordedsignalonastimulus-or
event-relatedbasis,time-lockedepochsofinterest(forinstancetriggeredbythe
onset of a spoken sentence) are extracted from the EEG raw data. To obtain a
highsignal-to-noiseratioaminimumnumberoftrials(e.g.,30sentencesofthe
samesentencetype)isrequiredforaveraging.Priortoaveraging,optionalfilteringandeyemovementcorrectionadditionallyhelptoremoveartifactsandother
unrelatedbrainactivity.Dataprocessingandsubsequenttrialaveragingproduce
asmoothcurveofchangesinelectricactivityovertimesincetheyeliminateor
atleastreducetheeffectofrandomnoisedistributedbyeachofthesingletrials.
31
32
ClaudiaMnnelandAngelaD.Friederici
Theresultingwaveformrepresentstheaverageprocessingofastimulusovertime
andconsistsofasequenceofpositiveandnegativevoltagefluctuations,referred
toascomponents,waves,deflections,orpeaks.Theseevent-relatedcomponents
are associated with various sensory, motor, and cognitive processes and reflect
covertandovertinformationprocessing.TheN1component,anegativefluctuationaround100ms,forexample,reflectsearlysensoryresponsesandwasfound
tobemodulatedbyattention(e.g.,Hillyard,Vogel,&Luck1998;Mangun1995;
Woldorffetal.1993).
Insum,EEGmeasurementandsubsequentaveragingofstimulus-triggered
epochs to gain ERPs provide a direct, non-invasive measure of the temporal
courseofchangesinelectricactivitythatdirectlyrelatetoneuronalinformation
processing.
2.2
WhyshouldweuseERPs?Advantagesanddisadvantages
oftheERPmethod
Behavioralexperimentsmeasureovertresponsessuchasreactiontimeandnumberofcorrectanswers.Thesemeasuresprovideuswiththeendproductofthe
cognitive processes engaged during the perception and evaluation of a given
stimulus.However,thosetechniquesarenotcapableofmonitoringtheactualonlinecognitiveprocessesthatleadtotheobservedbehavior.Althougheyetracking
methodsmaydeliveron-lineparametersoftheongoinginformationprocessing,
these measures are nevertheless indirect indicators of the underlying neuronal
mechanisms.Incontrast,methodsinthefieldofCognitiveNeurosciencesuch
asneuroimagingtechniquesandtheERPmethodinformusabouttheon-line
stagesofinformationprocessinginthebrain.
TheERPmethodfeaturesexcellenttemporalresolution,asitdeliversinformationinmillisecondaccuracyaboutthetimecourseofbrainresponses.Inthis
wayERPsprovideamentalchronometry,i.e.,anexacttemporalsequencingof
informationprocessing.Incomparisontoneuroimagingtechniques(e.g.,fMRI
andPET),thespatialresolutionfortheidentificationoftheneuralgeneratorsof
theobtainedsignalisrelativelypoorinERPs,sincemaximalamplitudemeasures
atcertainelectrodesitesonlyprovideinformationaboutwhereneuralactivity,
evokedbycertainstimuli,arrivesatthescalpssurface.However,therearesome
sophisticatedsourcelocalizationtechniquesthatestimatetheneuralgenerators
basedonthemeasuredelectricscalppotential.Therearetwocategoriesofsource
localizationmethods:techniquesthatpostulatedistributedcurrentsources(e.g.,
the minimum norm-based technique LORETA; Pascual-Marqui 2002; PascualMarqui,Michel,&Lehmann1994)andtechniquesthatassumeequivalentcurrent
Event-relatedbrainpotentials
dipolesasneuralorigins(e.g.,theBESAtechnique;Scherg,Vajsar,&Picton,1989;
Scherg&vonCramon1986).Oneofthefewdevelopmentalstudiesthatapplied
sourcelocalizationinvestigatedage-relatedchangesintheauditorysystembased
ondipolesourcemodeling(Ponton,Eggermont,Khosla,Kwong,&Don2002).
Independentoftheappliedlocalizationtechnique,additionaldatafromfunctionalimagingandclinicalstudiesconsiderablyhelptoconstrainsourceanalysis.In
thefieldofdevelopmentalcognitiveneuroscience,however,thisinformationis
mainlyrestrictedtoadultstudies(foranoverviewonfMRIresearchonlanguage
processingseeforinstanceFriederici2004a),sincethereareonlyfewsystematic
childrenfMRIstudies(e.g.,Brauer&Friederici2007)andinvestigationsonaphasiainchildren(foranoverviewseeFriederici1994).
Inworkingwithinfantsandchildren,wheninvestigatingthequestionofhow
childrenactuallyacquiretheirmothertongueandhowtheirlanguageprocessing
abilitiesdevelopovertime,theadvantagesoftheERPmethodbecomereadilyapparent.ForEEGrecordings,noovertresponsesarenecessarysinceEEGdirectly
measures brain activity to stimuli, thus considerably facilitating work with infants.ThefactthatERPcomponentsaredirectindicatorsoftheunderlyingbrain
processesimpliesnotonlythattherearenotaskrequirementsformeasurement
butalsothatbrainprocessesevokedbycertainstimulimightbedetectablebefore
thereisabehavioralcorrespondenceobservableatacertainstageinthechilds
development. Although behavioral methods used in infant research, e.g., the
headturn paradigm,thepreferential looking paradigmandeyetracking techniques
(seechaptersinthisvolumebyClahsen;Fernald,Zangl,Portillo,&Marchman;
Trueswell;Snedeker&Thothathiri),requirealesscomplicatedset-upandcanbe
performedinamorenaturalsetting,thesemethodsareatthesametimemore
prone to external interferences. Regarding neuroimaging techniques, there are
stillsomelimitationsintheworkwithinfantsandyoungchildren(butseeRedcay,
Haist,&Courchesne2006;seealsoHebden2003;Meek2002onopticalimaging
ininfants).First,movementrestrictionsduringbrainscanningmakeitratherdifficulttoworkwithchildren.Second,thereisstillanongoingdiscussionwhether
theBOLDsignalinadultsiscomparabletotheoneinchildrenandwhetherthe
appliedadultmodelsareappropriateforinfantresearch(fordiscussionseeAndersonetal.2001;Marcar,Strassle,Loenneker,Schwarz,&Martin2004;Martin
etal.1999;Rivkinetal.2004;Schapiroetal.2004).
Inadditiontoconsideringhowamethoddiffersfromothersandhowpractical
itmightbe,itisalsoimportanttoconsiderhowdifferentmethodsdeliverdifferentkindsofinformation.Thedecisiontouseaspecificmethodhighlydependson
thekindofinformationsought:theneuronalcorrelatesofinformationprocessing
intheirspatialand/ortemporalresolutionorthebehavioralconsequencesthat
33
3
ClaudiaMnnelandAngelaD.Friederici
followfromtheseprocesses.Ideally,acombinationofvariousmeasuresusingthe
complementaryabilitiesofdifferentmethodsshouldbesought.
2.3
Whatdowegetintheend?ERPcomponentsandtheirinterpretation
Thecomponentsofevent-relatedbrainpotentialscanbedescribedintermsof
theiramplitude/polarity,theirlatency,andtheircharacteristicscalpdistribution/
topography.Amplitudeindicatestheextenttowhicharesponsetoanexperimentalstimulusiselicited,i.e.,theamountofneuralactivity.Dependentonthepole
orientationofthemeasuredelectricfield,neuralactivityisreflectedinpositive
andnegativedeflections,indicatingneuronaldischargingandcharging,respectively.Latencyreferstothepointintimeatwhichthisactivationoccursrelative
tostimulusonset.Giventhesetwoparameters,specificcomponentsarelabeledas
follows:waveswithanegative-goingdeflectionaredesignatedbyN,waveswith
apositive-goingdeflectionby P,whilethetime(inmilliseconds)fromstimulus
onsettocertainwavepeaksisindicatedbyanumber.TheN100component,for
example, refers to a negativity occurring at about 100 ms after stimulus onset.
Note,howeverthatcomponentsareoftenlabeledaccordingtotheorderoftheir
occurrence during stimulus processing (e.g., N1, P2, N2) and their functional
significanceratherthantheirpolarityandlatencyparametersperse.Forinstance,
theP300componentisknowntooccurinvariousoddballparadigmsinresponse
to deviant stimuli and reflects memory- and context-updating processes after
stimulusevaluation(Donchin&Coles1988).Dependentonstimulusdiscriminationdifficulty,stimuluscomplexity,andcategorizationdemandstheP300latency
variesbetween300msand700mspost-stimulus(e.g.,Katayama&Polich1998;
Daffneretal.2000).Scalp distributionortopographyprovidesinformationabout
acomponentsvoltagegradientoverthescalpatanypointintimeandtherefore
someinformationabouttheunderlyingneuroanatomicalactivity. Aspointedout,
conclusionsfromERPdataaboutthesourcesofneuralgeneratorscanonlybe
drawn in a restricted way when relying on the topographic information alone,
e.g.,lateralizationtoonehemisphereordistributionoverposteriorbrainregions.
Thelabelofacomponentcanalsoincludeinformationaboutitstopography,e.g.,
ELANforEarlyLeftAnteriorNegativity.Furthermore,somelabelsofERPcomponentsdepicttheparticularexperimentalparadigminwhichtheyareevoked,
e.g.,MMNforMismatchNegativity.
ERPcomponentsareconsideredtobeindicatorsoftheprogressionofinformationprocessingovertime.Earliercomponents(uptoabout100200msafter
theonsetofthestimulus)reflectessentiallyautomaticprocessesandaremodulatedbythephysicalpropertiesofastimulus,suchastheloudnessandpitchof
Event-relatedbrainpotentials
aspokenword.Latercomponentsarethoughttoreflecthigher-ordercognitive
processingthatisinfluencedbyapersonsintentionsandactions,suchasthose
presentduringadiscriminationtaskbetweenwordsandnon-words.Aspointed
out,ERPcomponentsdefinedbyspecificparametersarelikelytoreflectdifferent
brainmechanismsengagedinstimulusprocessing.Changesintheparametersof
aspecificcomponentindicatechangesintheunderlyingcognitivemechanisms.
Forinstance,aprolongedlatencymightpointtoaslowingdownofaspecificcognitiveprocessandareducedamplitudetoareductionintheprocessingdemands
orefficiency.Thus,ERPcomponentsareusuallyinterpretedwithrespecttoboth
theirunderlyingneuralmechanismsandtheirfunctionalsignificance.
TheinterpretationofERPcomponentsininfantsdemandssomeadditional
consideration.WhendealingwithinfantERPs,researchersshouldkeepinmind
theenormousphysiologicalchangesofthedevelopingbrain,concerningsynaptic
density,myelination,skullthickness,andfontanelstate(seeforexampleMrzljak,
Uylings,VanEden,&Judas1990;Pujoletal.2006;Uylings2006).Forinstance,
the reduced synaptic density results in greater slow wave activity, possibly explainingwhyinfantERPsdonotexhibitasmanywell-definedpeaksasadultERPs
(Nelson&Luciana1998).InfantERPsusuallyshowlargeramplitudesthanadult
data,possiblyduetodifferenceinskullthickness,andlongerlatenciesthanadult
ERPs, which, however, gradually decrease with increasing age (e.g., Jing & Benasich2006;Kushnerenkoetal.2002).Thus,whencomparingERPcomponents
acrossagegroupsoneshouldconsiderthesematurationalchanges.Here,paradigmsusedininfantERPexperimentsshouldbeconductedinadultsaswell,thus
achievingatargetadultERPpatternagainstwhichdevelopmentalcomparisons
canbemade.
2.
Whatdoesthistellusaboutlanguage?ERPcomponentsrelated
tolanguageprocessing
Inthedomainoflanguageprocessing,thereareatleastfivefunctionallydifferentcomponentsthatreflectphonological,semantic,andsyntacticprocessingin
adults. The following components have been observed in ERP studies on languageprocessing:(1)theMismatchnegativity(MMN),anegativitythatoccurs
ataround100250mspost-stimulusandindicatesthediscriminationofacoustically or phonetically different stimuli (e.g., Ntnen 1990; Opitz, Mecklinger,
Cramon,&Kruggel1999)andisalsomodulatedbylanguageexperience(Winkleretal.1999);(2)theN400,acentro-parietallydistributednegativityataround
400mspost-stimulusthatreflectslexical-semanticprocessesatbothwordlevel
(e.g.,Holcomb&Neville1990)andsentencelevel(e.g.,Kutas&Hillyard1980,
3
36
ClaudiaMnnelandAngelaD.Friederici
1983);(3)theE/LAN,aleftanteriornegativityataround150350mspost-stimulus,whichoccursforon-linesyntacticandmorphosyntacticprocesses(e.g.,Friederici,Pfeifer,&Hahne1993;Hahne&Friederici1999);(4)theP600,acentroparietallydistributedpositivityataround600mspost-stimulus,whichisrelated
toprocessesofsyntacticreanalysisandrepair(e.g.,Friederici&Mecklinger1996;
Kaan, Harris, Gibson, & Holcomb 2000; Osterhout & Holcomb 1992); and (5)
theClosurePositiveShift(CPS),acentrallydistributedpositivitythathasbeen
observedinresponsetotheprocessingofintonationalphrases(e.g.,Pannekamp,
Toepel,Alter,Hahne,&Friederici2005;Steinhauer,Alter,&Friederici1999).
Overthelastdecades,avastamountofstudiesinadultshasdemonstrated
that the processing of different aspects of linguistic information can be clearly
distinguished by means of these different ERP components. For a detailed description of the single components and the particular experimental conditions
inwhichtheyareevokedwereferthereadertorecentreviews(Friederici2002,
2004b;Kutas&Federmeier2000).
Ininfantsandyoungchildren,recentneurophysiologicalresearchhasdemonstratedthatmostERPcomponentsassociatedwithphonological,semantic,and
syntacticprocessesarequitesimilartotheonesobservedinadults.Thisfactindicatesthattherearequantitativeratherthanqualitativechangesinthelanguage
processes,reflectedbyparticularERPcomponents,duringinfantsandchildrens
developmentuntilanadult-likestatusisachieved.Asaforementioned,ERPcomponentsininfantsandchildrenoftenshowlongerlatenciesandlargeramplitudes
ascomparedtotheonesinadults,withagraduallatencyandpeakdecreaseasage
increases(forcomparisonsbetweenchildrenandadultERPdatainlanguageprocessingseeforinstanceHahne,Eckstein,&Friederici2004;Oberecker,Friedrich,
&Friederici2005).
3.
Inthefollowing,wegiveanoverviewofERPresearchonthedifferentlandmarks
oflanguageacquisitionduringachild'sfirstthreeyearsoflife,exemplifiedwith
experiments from our laboratory. We describe the processing of phonological,
semantic,andsyntacticinformationininfantsandchildrenanddiscusstheseresultsinthecontextofotherneurophysiologicalandbehavioralstudies.Figure1
showsthedevelopmentalstagesofauditorylanguageperceptionandtheirassociatedERPcomponents.Thedevelopmentalstagescanbeconsideredasinterrelated
phasesduringwhichnewinformationisderivedandprocessedbasedonpreviouslyacquiredknowledge.Fromearlyoninfantsareabletodiscriminatespeech
Event-relatedbrainpotentials
ERP correlates
Developmental stages
Identification of
intonational
boundaries
Identification of word
boundaries
Discrimination
of phoneme
parameters
Birth
Discrimination
of word stress
patterns
Sensitivity to
intonational
phrase boundaries
Lexical
processing
Sentence processing
Lexical
form
Lexical
semantics
Selection
restrictions
of verbs
12
14
19
Local
phrase
structure
building
32
Month of age
MMN
CPS
N400
ELAN - P600
Figure 1. DevelopmentalstagesoflanguageacquisitionandtheirassociatedERP
components(modifiedfromFriederici2005).
sounds and word stress patterns, which facilitates the identification of content
andfunctionwordsinthesententialcontext.Furthermore,infantsearlyabilityto
processprosodicinformationatthesentencelevel,presentinintonationalphrase
boundaries,aidsthedetectionofsyntacticboundaries.Theseprocesseseventually
allowchildrentoderivesyntacticrulesfromspeechinputandprovidethebasis
forthebuildingoflocalsyntacticstructuresandinterphrasalrelationships.The
timecourseoftheoutlineddevelopmentalstagesisbasedontheavailableERP
literatureininfantresearchandisthereforeonlyanapproximationoftheactual
timecourseinlanguagedevelopment.Thus,ouroverviewshouldbeunderstood
as an attempt to sketch the language acquisition process based on the current
knowledgewithintheframeworkofdevelopmentalcognitiveneuroscience.
3.1
Processingofphonological/prosodicinformation
Ingeneral,infantsfirststepsintolanguagearebasedonprosodicinformation.
Frombirthon,prosodiccuesfacilitatethesegmentationoftheincomingspeech
streamintostructuralelementsandthereforesupporttheacquisitionoflexical
andsyntacticunitsandeventuallythederivationofsyntacticorderingprinciples,
a process called prosodic bootstrapping. Infants first challenge is to extract the
phonological details from their mother tongue. This phonological information
comprises the actual speech sounds (phonemes), the rules according to which
thesesoundsarecombined(phonotacticknowledge),andtheprosodicpatterns
thathelptostructurethelanguageinputintounits,suchasinformationabouta
wordsstresspatternandasentencesintonationalcontour.
37
38
ClaudiaMnnelandAngelaD.Friederici
Event-relatedbrainpotentials
Amplitude
(normalized values)
(a)
Amplitude
(normalized values)
0.7633
/ba:/ long
-0.6605
0.7633
/ba/ short
-0.6605
Time (s)
0.202
Amplitude
(normalized values)
0 0.03
(b)
Acoustic
Parameters
0.7633
st
1 formant
2nd formant
3rd formant
4th formant
Pitch
Pitch min
Pitch max
0
-0.45
/ba:/ long
F3
C3
0.341
0 - 30 ms
776 Hz
1436 Hz
2735 Hz
4086 Hz
219 Hz
216 Hz
220 Hz
/ba/ short
F3 F4
C3 C4
F4
C4
deviant
standard
difference wave
F3
F4
C3
C4
-15 +V
0.4
s
0.8
Figure 2. Syllablediscrimination:(a)Thefiguredisplaystheacousticparametersofthe
shortsyllable/ba/(202ms)andthelongsyllable/ba:/(341ms).Inapassiveauditory
oddballparadigm,bothsyllablesarerepeatedlypresentedinthespecifiedfrequency
(standard5/6,deviant1/6).Thecriticalcomparisonconcernseachsyllableinitsroleas
standardinoneblockanddeviantintheother.(b)ERPdataanddifferencewaves(deviant-standard)of2-month-oldsinresponsetothelongsyllable/ba:/andtheshortsyllable
/ba/(modifiedfromFriedericietal.2002).
39
0 ClaudiaMnnelandAngelaD.Friederici
/ba:/ long
F3
F4
C3
C4
F3 F4
C3 C4
adults
2-month-olds
-15 +V
0.4
s
0.8
Figure 3. Syllablediscrimination:Differencewaves(deviant-standard)ofadultsand
2-month-oldsinresponsetothelongsyllable/ba:/inanauditoryoddballparadigm
(modifiedfromFriedericietal.2002).
able to discriminate long syllables from short syllables and that they display a
positivityintheERPasmismatchresponse(MMR).Here,theymoreeasilydiscriminatealongsyllableinastreamofshortsyllablesthanviceversa,whichcan
beexplainedbythegreaterperceptualsaliencyofalargerelementinthecontext
ofsmallerelementsthanviceversa(formoredetailseeFriedrich,Weber,&Friederici2004).
Sofar,severalERPstudieshavestudiedphonemediscriminationininfants
testingtheirabilitytodetectchangesinconsonantarticulation(Dehaene-Lambertz&Dehaene1994),consonantduration(Kushnerenkoetal.2001;Leppnen
etal.2002),vowelduration(Leppnen,Pikho,Eklund,&Lyytinen1999;Pihko
etal.1999),andvoweltype(Ceponiene,Lepist,Alku,&Ntnen2003;Cheour
etal.1998).Inthesestudies,theMMRappearedeitherasapositiveoranegative
deflectionintheERP.Forinstance,Kushnerenkoetal.(2001)presentedsleepingnewbornswithfricativesofdifferentdurationandobservednegativeMMRs.
Comparatively,Leppnenetal.(1999)andPihkoetal.(1999)investigatedsleepingnewbornsbypresentingphonemeswithdifferentvowellengthandreported
positiveMMRs.Thereisanongoingdiscussionaboutthenatureofthesedifferent
ERPresponsestoauditorychangedetection.Ourstudies,forinstance,showed
thattheoutcomeoftheseERPresponsesareaffectedbyatleasttwofactors:the
infantsstateofalertness(awakeorasleep)andtheparticulardatafilteringtechnique(seeWeber,Hahne,Friedrich,&Friederici2004).Furthermore,thechoice
of stimulus (discrimination difficulty or saliency) seems to have an impact on
thediscriminationresponse(Morr,Shafer,Kreuzer,&Kurtzberg2002).Also,the
transition from a positive to a negative MMR can be shown to be an effect of
Event-relatedbrainpotentials
advancingmaturation(seeKushnerenkoetal.2002;Morretal.2002;Trainoret
al.2003;foradiscussionofapossibleoverlapbetweenthetwocomponents,see
Morretal.2002).Despitethedifferencesintheappearanceofthedetectionof
phonemechangesintheERP,thecombineddatasuggestthattheinfantsability
toautomaticallydiscriminatebetweendifferentphonemesispresentfromearly
on.Here,ERPsmightbeusedtomarkdifferencesintheinfantsmaturationstate
thatarebeyondthescopeofobservation.
In addition to investigating syllable discrimination in infants with normal
languagedevelopment,wepursuedinourresearchthehypothesisthatamajor
underlyingcauseofSpecific Language Impairment (SLI)isadeficiencyintheprocessingofphonological/prosodicinformation(seeFriedrich,Weber,&Friederici
2004). SLI is defined as impairment in the expressive language domain in the
presence of otherwise normal development (see Leonard 1998). Consequently,
apart from their language deficits these children show normal intelligence and
donothaveanyneurological,sensory,ormotorproblems.Ifthehypothesisofan
impairedprocessingofphonological/prosodicinformationinSLIholds,infants
atriskforSLImightbedeficientindetectingdurationchangesinphonemesat
theageof2months.
Toaddressthisquestion,anotherstudycarriedoutbyourlaboratorytested
theabilityofchildrenwithariskforSLItodiscriminatelongfromshortsyllables
withthesameMMNprocedureasusedbefore(Friedrichetal.2004).Children
wereassignedtooneoftwogroups,beingatriskornotatriskforlaterdeveloping
SLI,basedonfamilyhistory.InFigure4,thedifferencewavesforbothgroups,
2-month-oldswithandwithoutriskforSLI,showapositivedeflection.However,
intheriskgroupthispositivewavereachesitsmaximumlaterthantheonein
theno-riskgroup.Thislatencydifferencepointstoaslowerspeedininformation
processingintheat-riskinfants.
Thus, infants at risk for SLI differ from those with no risk at the age of 2
monthsintheirperceptualERPparametersfordurationdiscrimination.Thisis
in line with recent ERP studies that investigated early differences in phoneme
processingbetweeninfantswithandwithoutafamilyriskofdyslexia.Specifically,
Leppnenetal.(2002)studiedERPresponsestoconsonantdurationchangesand
observedthat6-month-oldsat-riskdifferedfromcontrolsintheirinitialresponsivenesstosoundsaswellastheirabilitytodiscernsoundchanges.Guttormet
al. (2005) reported for at-risk newborns longer lasting positive ERP responses
toconsonant-vowelchangesthatwerecorrelatedwithpoorerreceptivelanguage
skillsmeasuredat2.5years.Althoughatthispointitisnotclearyetwhetherthe
childrenofourstudywilldevelopSLIatalaterage,thepresentdatasuggestthata
delayedauditorychangeprocessing,possiblycausedbyweakermemorytracesin
thesechildren,mightbeoneofthepotentialfactors(formoredetailseeFriedrich
1
2
ClaudiaMnnelandAngelaD.Friederici
/ba:/ long
F4
F3
F3 FZ F4
C3
C4
FZ
P3 PZ P4
C3
C4
P4
-10 +V
0.4
s
0.8
Figure 4. Syllablediscrimination:Differencewaves(deviant-standard)of2-month-olds
atriskandnotatriskforSLI(basedonfamilyrisk)inresponsetothelongsyllable/ba:/
inanauditoryoddballparadigm(modifiedfromFriedrich,Weber,&Friederici2004).
etal.2004).Consequently,ifthesechildrenalreadyhavedifficultiesatearlylanguagelearningstagesindetectingphonological/prosodiccues,theymightbedelayedorimpairedinutilizingthisinformationatlaterstagesoflexicalandsyntacticlearning.Tallaletal.(1996)suggestextensivetrainingwithartificiallyslowed
speechforchildrenwithSLItoovercomeabnormalperceptuallearningpresent
atearlydevelopmentalstages.Sincethementionedstudiesdemonstratethatthe
ERPmethodisabletodifferentiatebetweengroupsofchildrenatriskandnotat
riskforlaterlanguageproblems,specificelectrophysiologicalparameters,suchas
thelatencyofthemismatchresponse,mightbedevelopedasadiagnostictoolfor
veryearlyidentificationofchildrenwhowouldbenefitfromearlyintervention.
Event-relatedbrainpotentials
words,stressinformationcertainlyaidstheidentificationofwordonsets.Inorder
toinvestigatebymeansofERPatwhatageinfantsstarttoshowtheabilitytodiscriminatewordstress,weagainappliedtheMMNparadigmdescribedabove.
InthisERPstudy,4-and5-month-oldGermaninfantsweretestedontheir
abilitytodiscriminatebisyllabicpseudowordsstressedonthefirstsyllablefrom
bisyllabicpseudowordsstressedonthesecondsyllable(Weberetal.2004).Since
wordstressisdefinedbyanumberofacousticparameterswithsyllablelengthas
themostprominentone,weusedpseudowordsthatdifferedinthelengthoftheir
firstandsecondsyllables.Namely,weusedthetrochee/ba:ba/stressedonthefirst
syllable,andtheiamb/baba:/stressedonthesecond(seeFigure5(a)).Inapassiveauditoryoddballparadigm,infantswererepeatedlypresentedwithdeviant
/ba:ba/stimuliamongstandard/baba:/stimuliandviceversa.
Figure5(b)displaysthedifferencewavesforbothagegroups.Asignificant
negativedeflectionisonlypresentinthe5-month-oldgroup,forwhomonlythe
trocheeevokesanegativeMMR.Thisindicatesthatbytheageof5monthsinfants
are able to discriminate word stress in bisyllabic words, whereas 4-month-olds
arenot.Thediscriminationresponseisevidentinthenegative MMRtothetrochaicpattern,whichisthepredominantpatterninGerman.ThenegativeMMR,
incontrasttothepositiveMMRobservedinour2-month-olds(Friedericietal.
2002),mightbeattributabletoboththeinfantsadvancedmaturationstateaswell
as the lower processing demands of word stress discrimination (or the higher
saliencyofthestressdifferences)thandifferencesinvowelduration(seeKushnerenkoetal.2002;Morretal.2002).
Behavioralstudieshavedemonstratedthattheabilitytodiscriminatestress
patternsinbisyllabicwordsisnotpresentin6-month-oldinfants,buthasemerged
bytheageof9months(e.g.,Jusczyk,Cutler,&Redanz1993).Insupportofthe
prosodicbootstrappingapproach,Nazzi,Iakimova,Bertoncini,Frdonie,andAlcantara (2006) describe developmental effects for 8- to 16-month-olds for the
detection of syllables in fluent speech before two-syllable words are derived as
oneunit.Whilenosegmentationeffectwasfoundfor8-month-olds,12-montholdssegmentedindividualsyllablesfromthespeechstream,withmoreeasein
segmenting the second syllable, which is consistent with the rhythmic features
oftheirnativelanguageFrench.Interestingly,bytheageof16monthschildren
segmentedbisyllabicwordsaswholeunitsfromthespeechinput.Althoughthe
resultsofbehavioralandERPstudiesarenotdirectlycomparable,itseemsthat
before a discrimination reaction can be observed at a behavioral level, there is
evidence for a discrimination response in the measured brain activity. This is
supportedbythefactthatourstudyshowedastresspatterndiscriminationresponseintheERPevenin5-month-olds(Weberetal.2004).Similarly,Kooijman
andcolleagues(Kooijman,Hagoort,&Cutler2005,2006)foundwordrecogni-
3
ClaudiaMnnelandAngelaD.Friederici
0.7000
Amplitude
(normalized values)
(a)
trochee /ba:ba/
0
offset 1st CV onset 2nd CV
-0.6895
0
100
Time (ms)
Amplitude
(normalized values)
0
offset 1st CVonset 2nd CV
-0.6895
Amplitude
(normalized values)
100
183
278
1007
1018
958
1594
1573
1555
2858
2765
2800
4347
4236
4281
P3
5-month-olds
C4
P4
+V
0.4
s
0.8
/baba:/
F3 F4
C3 CZ C4
P3 P4
F4
F3
F4
CZ
0-183 ms
/ba:ba/
4-month-olds
F3
C3
750
0-100 ms
0
-0.6895
Time (ms)
Acoustic
Parameters
0.7000
-5
750
iamb /baba:/
(b)
355 405
0.7000
MMN
CZ
C3
P3
C4
P4
bandpass 1-15 Hz
trochee /ba:ba/
iamb /baba:/
Figure 5. Stresspatterndiscrimination:(a)Thefiguredisplaystheacousticparameters
ofthebisyllabicpseudowords/ba:ba/(trochee)and/baba:/(iamb).Inapassiveauditory
oddballparadigm,bothpseudowordsarerepeatedlypresentedinthespecifiedfrequency
(standard5/6,deviant1/6).Thecriticalcomparisonconcernseachpseudowordinitsrole
asstandardinoneblockanddeviantintheother.(b)Differencewaves(deviant-standard)of4-and5-month-oldsinresponsetothetrochaicandtheiambicstresspattern
(modifiedfromWeberetal.2004).
Event-relatedbrainpotentials
MMN
Iamb /baba:/
F3
FZ
F4
F3 FZ F4
C3 CZ C4
P3 PZ P4
C3
CZ
C4
PZ
P4
-10 +V
0.4
s
0.8
Figure 6. Stresspatterndiscrimination:Differencewaves(deviant-standard)of
5-month-oldsatriskandnotatriskforSLI(basedonwordproductionattwoyears)in
responsetothetrochaicstresspattern/baba:/inanauditoryoddballparadigm(modified
fromWeberetal.2004).
tionresponsesintheERPtopreviouslyfamiliarizedwordsinastudytesting10month-oldsandeven7-month-olds.However,behavioralstudiesalsofindword
segmentationeffectsaroundtheageof78monthswhenthewordsstresspatternsfollowtherhythmicfeaturesoftheinfantsnativelanguage(e.g.,Houston,
Santelmann, & Jusczyk 2004; Jusczyk, Houston, & Newsome 1999). Together,
these findings emphasize the importance of stress cues for word segmentation
fromfluentspeech.
Tofollowuponthefindingthat5-month-oldinfantsareabletodiscriminate
differentlystressedpseudowords,weaimedtotestwhetherinfantsatriskforSLI
alreadyshowanimpairedstresspatterndiscriminationattheageof5months.
Another study carried out in our laboratory investigated the ERP responses of
infantsatriskforSLIbyusingthesameMMNparadigmwithbisyllabicpseudowords(seeWeberetal.2004).
Inthecurrentexperiment,infantswereretrospectivelygroupedintoinfants
beingatriskornotatriskforlaterSLIbasedontheirwordproductionperformance at the age of 24 months (Weber, Hahne, Friedrich, & Friederici 2005).
Childrenwhoat24monthshaveverylowwordproduction(asassessedbythe
ELFRA-2 measure; Grimm & Doil 2000) display at 5 months a reduced MMR
amplitudetothetrochaicpatterncomparedtotheirage-matchedcontrols,ascan
beobservedinthedifferencewavesinFigure6.Itfollowsthatinfantswithrisk
forSLI,asdeterminedbyadeficitinwordproductionattheageof24months,
6 ClaudiaMnnelandAngelaD.Friederici
seemtohaveanimpaireddiscriminationabilityforthelanguage-specifictrochaic
patternalreadyattheageof5months.Thisresultgivesrisetothenotionthatthe
processingandacquisitionofphonologicalinformationinearlyinfancymightbe
crucialforlaternormallanguagedevelopment.Inthiscase,theERPmismatch
responsecouldserveasanearlyidentificationofSLI.
Thecombinedresultsofthestudiesonphonemeandstresspatterndiscrimination provide evidence that speech segmentation in early infancy heavily dependsonphonological/prosodiccuesandthatthesecuesarelikelycontributors
to lexical acquisition. In their behavioral experiments, Nazzi, Dilley, Jusczyk,
Shattuck-Hufnagel,andJusczyk(2005)demonstratedthatindeedbothstresspatternandtypeofinitialphonemeinfluencewordsegmentationfromfluentspeech
withapreferenceforthepredominantpatternsoftheinfantsnativelanguage.
Event-relatedbrainpotentials
12-month-olds
-10 +V
10
CZ
19-month-olds
-10 +V
s
10
0.4 0.8 1.2 1.6
N400
CZ
Adults
-5 +V
s
5
0.4 0.8 1.2 1.6
N400
CZ
CZ
phonotactically
illegal nonwords
s
legal pseudowords
0.4 0.8 1.2 1.6
Figure 7. Phonotacticknowledge:ERPdataof12-month-olds,19-month-olds,and
adultsinresponsetopseudowords(phonotacticallylegal)andnonwords(phonotacticallyillegal)inapicture-wordparadigm(modifiedfromFriedrich&Friederici2005a).
thelexiconforpossibleentries,butthissearchfailsaspseudowordsarenotpart
ofthelexicon.Nonwords,however,donotinitiateasimilarsearchresponse.Apparently,nonwordsarenoteventreatedaslikelyentriesofthelexiconandthereby
possiblereferentsformeaning,sincetheyalreadyviolatethephonotacticrulesof
thelanguage.
We investigated childrens ERP responses to phonotactically legal pseudowords and phonotactically illegal nonwords to determine whether 12- and 19month-olds already have some phonotactic knowledge (Friedrich & Friederici
2005a).Inapicture-word paradigmchildrenwerepresentedwithsimplecolored
pictureswhilesimultaneouslylisteningtobasiclevelwordsspokenslowly.These
wordseithercorrectlylabeledthepicturecontentorwerepseudowordsornonwords.Assumingthatthepicturecontentinitiateslexical-semanticpriming,semanticintegrationdifficulties(reflectedinanenhancedN400amplitude)should
occur for pseudowords and nonwords, since these nonsense words do not semantically match the picture content. However, if infants at the age of 12 and
19monthspossessomephonotacticknowledge,theN400semanticviolationresponseshouldonlyappearforpseudowords,whichfollowthelanguage-specific
phonotacticrulesbutnotfornonwords,whichviolatethemandarethereforenot
treatedaslikelyreferentsforpicturelabels.
As can be seen in Figure 7, the ERP responses of 19-month-olds are quite
similar to the ones observed in adults. Both groups show a more negative responsetophonotacticallylegalpseudowordsthantophonotacticallyillegalnonwords.AdultsshowthetypicalN400response,startingataround400msafter
stimulus onset, most pronounced at central and parietal electrode sites. In 19month-olds,themorenegativedeflectionstopseudowordsalsostartataround
400 ms post-stimulus but are sustained longer than in adults, suggesting an
N400-like response in these children. In contrast, 12-month-olds do not show
differentialERPresponsestopseudowordsandnonwordscomparabletothose
ofadultsand19-month-olds.The12-month-oldsdoshowmorenegativitywith
7
8 ClaudiaMnnelandAngelaD.Friederici
pseudowords,mainlyatleftlateralfrontalsitesbetween800900msafterstimulusonset,whichsuggestssomekindoffacilitatedprocessingofphonotactically
legalpseudowords.However,thisfacilitationdoesnotseemtobebasedonlexical-semanticprocessing,asmarkedbyanN400response,butislikelytoreflect
favoredacoustic-phonologicalprocessingforthefamiliarphonemesequencesin
phonotacticallylegalpseudowords(foramoredetaileddiscussionoftheeffects
seeFriedrich&Friederici2005a).
TheresultsofthisERPstudyonprocessingofphonotacticinformationare
inlinewithbehavioralstudiesthatshowthat9-month-oldinfantsalreadyknow
somephonotacticrulesoftheirnativelanguage(Friederici&Wessels1993;for
areviewseeJusczyk1997).Here,infantswereabletousetheirknowledgeabout
phonotacticrulesregardingwordboundariestosegmentword-likeunitswhen
pseudowordswerepresentedeitherinisolationorincontext.
Insummary,wecanconcludethat19-month-oldspossesssomephonotactic
knowledge (indicated by an N400) and therefore treat as potential words phonotactically legal pseudowords but not phonotactically illegal nonwords. Thus,
phonotactically illegal nonwords are from very early on excluded from further
word learning.12-month-old infants seemnottohave establishedthiskindof
phonotacticknowledgeyetbutinsteadshowaphonologicalfamiliarityeffectto
phonotacticallylegalpseudowords.Takentogetherwiththeobservationfrombehavioralstudiesthat9-month-oldinfantshavesomeinitialphonotacticknowledgeonemayconcludethatthiskindofknowledgeadvancesandisappliedin
lexicalprocessing(andinthiscasemarkedbytheoccurrenceofanN400)onlya
fewmonthslaterinthechildsdevelopment.Thisassumptionissupportedbythe
findingthatthemechanismsthatunderlietheN400responsematurebetween12
and14months(seeFriedrich&Friederici2005a,2005b).
Event-relatedbrainpotentials
nationalphraseboundariesthuscouldprovideaneasyentryintothelexiconand
thesyntaxofagivenlanguage.
Intonationalphrase(IPh)boundariesaredefinedbyseveralcharacteristics.
First,thelastsyllableofanIPhcontainsachangeinpitch,markingalowora
highboundarytone.Second,thislastsyllableislengthenedascomparedtosyllableswithinthephrase.Third,thereisoftenapausefollowingtheIPhboundary.
Behaviorally,ithasbeenshownthatadultlistenersmakeuseofIPhboundaries
intheinterpretationofspokenutterances(seeCutler,Dahan,&vanDonselaar
1997).InERPstudieswithadults,theoffsetofintonationalphrasesisassociated
withapositivedeflectionwithacentro-parietaldistribution(Pannekamp,Toepel,
Alter,Hahne,&Friederici2005;Steinhauer,Alter,&Friederici1999).ThiscomponentisnamedClosure Positive Shift (CPS)sinceitisinterpretedasanindicator
fortheclosureofprosodicphrasesbyintonationalphraseboundaries.Inorder
togainfurtherinsightsintotheroleofprosodicinformationinlanguageacquisition,wehaveinvestigatedtheneurophysiologicalbasisoftheperceptionofsentence-levelprosodiccuesininfants.
InanERPstudyexaminingtheabilityof8-month-oldinfantstoidentifyIPh
boundaries,Pannekamp,Weber,andFriederici(2006)presentedinfantswithsentencesoftwoprosodicallycorrectsentencetypes,oneconsistingoftwoIPhswith
onesentence-internalIPhboundary(seeexample1)andtheotheronecontaining
threeIPhswithtwosentence-internalIPhboundaries(seeexample(2)).Thetwo
sentenceconditionsresultedfromthedifferentsyntacticstructureofthesentences.Theuseofintransitiveversustransitiveverbsleadstolateclosureversusearly
closure,respectively.ThedifferentsyntacticstructuresconsequentlyresultindifferentprosodicrealizationswithIPhboundariesatdifferentsentencepositions.
(1) Lena verspricht Mamazuflitzen und Getrnke zukaufen.
[Lena promises Mamatorun]IPh [and drinks tobuy]
(2) Lenaverspricht Mamazuhelfen und Getrnke zukaufen.
[Lenapromises]IPh [Mamatohelp]IPh[and drinks tobuy]
TheERPresultsinresponsetothetwodifferentsentencetypesforboththeinfantstudy(Pannekampetal.2006)andtheadultstudy(Pannekampetal.2005)
aregiveninFigure8.Inbothgroups,theoccurrenceofIPhboundariesoverthe
courseofthesentenceisfollowedbyapositiveshiftintheERP.Thus,forsentences with one IPh boundary one corresponding CPS was observed, and for
sentenceswithtwoIPhboundaries,twoCPS.Inadults,thepositiveshiftsstart
withanapproximatelatencyof500mstotheircorrespondingIPhboundaries,
whereas in infants the positive waves start about 1000 ms after the IPh offset.
ThisdevelopmentallatencyshifthasalreadybeenreportedinpriorinfantERP
9
0
ClaudiaMnnelandAngelaD.Friederici
8-month-olds
F3
C3
CZ
Adults
FZ
F4
F3
CPS
C4
C3
P4
P3
P3
2.0
s
4.0
C4
CZ
CPS1
CPS2
P4
PZ
PZ
sentences with
1 IPh boundary
2 IPh boundaries
F4
CPS
CPS1 CPS2
-10 +V
FZ
F3 FZ F4
C3 CZ C4
P3 PZ P4
-6 +V
0
2.0
s
4.0
Figure 8. Processingofintonationalphraseboundaries:ERPdataof8-month-oldsand
adultswithpositiveshifts(CPS)incorrelationtotheIPhboundariesineachsentence
condition(modifiedfromPannekampetal.2005,2006).
studies on semantic and syntactic processes and points to slower information
processingininfantsandchildren(seeforinstanceHahne,Eckstein,&Friederici
2004;Oberecker,Friedrich,&Friederici2005).TheCPSlatencytoeachofthe
IPhboundariesfurtherindicatesthatthisERPcomponentisnotamerereaction
totheonsetofacousticdifferences(lower-levelprocessing)butanindexforthe
underlyinglinguisticprocessoftheperceptionofintonationalphrases(higherlevelprocessing).SincetheinfantERPdataclearlyshowtheoccurrenceofthe
CPScomponentinresponsetoeachIPhboundary,thecurrentstudyindicatesin
electrophysiologicaltermsthatinfantsasyoungas8monthspossesstheability
toprocessprosodiccuesatthesentencelevel.Thisfindingsupportsthenotion
thattheneurophysiologicalbasisofprosodicprocessesthatarecrucialforspeech
segmentationisestablishedearlyduringlanguagedevelopment.
TheresultsofthisERPinvestigationareinlinewithbehavioralstudiesthat
showthatinfantsattheageofaround9monthsareabletodetectmajorsyntactic
phrasesinthespeechinputbyrelyingonprosodicinformation(Hirsh-Paseketal.
1987;Jusczyketal.1992;forareviewseeJusczyk1997).Recently,Soderstromand
colleagues(Soderstrom,Nelson,&Jusczyk2005;Soderstrom,Seidl,Nelson,&Jusczyk2003)haveshownthateven6-month-oldsaresensitivetophrase-levelprosodiccuesandthattheinfantsindeedutilizethisinformationtoidentifysyntactic
unitsinthespeechstream.Similarly,inlexicallearning,behavioralstudieshave
Event-relatedbrainpotentials
shownthat10-month-oldinfantsusephonologicalphraseboundaryinformation
toconstrainlexicalaccess(Christophe,Gout,Peperkamp,&Morgan2003;Gout,
Christophe,&Morgan2004).Takentogether,thesestudiesdemonstratethatinfantsarenotonlyabletoperceivephrase-levelprosodiccuesinearlyinfancybut
thattheyactuallyusethisinformationtosegmentfluentspeechintolexicaland
syntacticunits.FutureERPstudiesmayprovideadditionalinformationaboutthe
on-linebrainmechanismswheninfantsapplyprosodicinformationtostructure
thespeechinput.Whilebehavioralstudiesdelivertheinformationthatinfants
perceivesyntacticunitsonthebasisofprosodiccues,ERPstudiesmayshowhow
andwhenthesecuesareutilizedtodisentanglethespeechinput.
3.2
Processingoflexical-semanticinformation
Anotherchallengethatinfantshavetofaceinlanguagelearningistheacquisition
ofthemeaningsofwords.Infantssuccessivelybuilduptheirlexiconanddevelop
theabilitytomapwordsontoconceptualrepresentationsofobjectsoreventsand
viceversa.Fromearlyoninfantssegmentwordsfromfluentspeech,buthowand
whentheyactuallyknowwhataparticularwordmeansstillneedstobespecified.
Toanswerthisquestionandtounderstandtheneurophysiologicalmechanismsof
earlywordlearning,itisimportanttodeterminewhethertheinfantsbrainworks
similarly to the adult brain when processing meaningful words in meaningful
contexts,suchaspicturesorsentences.
IntheadultERPresearchonsemanticprocessing,theN400componenthas
beenestablishedasanindicatorfortheneuralmechanismsofsemanticintegrationofelementsintotheircontext(e.g.,Holcomb1993).TheN400isevokedin
response to both words and pictures that do not match the expectation build
upbypreviouslypresentedwords,sentences,pictures,andpicturestories(e.g.,
Friederici,Pfeifer,&Hahne1993;West&Holcomb2002).Theamplitudeofthe
N400 is inversely related to the expectation triggered by the semantic context
andthereforevarieswiththeeffortnecessarytointegrateastimulusintoagiven
situation.AnN400primingeffect,reflectedinareducedN400amplitude,thus
indicatestheeffectofsemanticprimingontheprocessofsemanticintegration.
Importantly, a reduced amplitude reflects ease in semantic integration due not
onlytopreviouslypresentedstimuli,i.e.,thecurrentsemanticcontext,butalso
tothesemanticknowledgeinlong-termmemory(forareviewseeKutas&Federmeier2000).SincetheN400hasbeensuccessfullyshowntobecorrelatedto
lexical-semanticprocessinginadultsandsincethiscomponentisalikelyindicatorforthedevelopmentofsemanticmemory,wehaveutilizedtheN400tostudy
1
2
ClaudiaMnnelandAngelaD.Friederici
theneurophysiologicalbasisandthedevelopmentalaspectsoflexical-semantic
processesinearlychildhood.
3.2.1 Processingoflexical-semanticinformationatwordlevel
Several studies in our laboratory have focused on the question of whether the
neuralmechanismsforsemanticprocessingobservedinadultsarealsopresent
duringearlylanguageacquisitionwheninfantshaveacquiredonlyafewwords.
Furthermore, these studies aim to determine at what point during the infants
development these neural mechanisms mature, by examining infants abilities
during word-level lexical-semantic processing as indexed by the N400 component(Friedrich&Friederici2004,2005a,2005b,2006).Usingacross-modalpicture-worddesign,theERPresponsesof12-,14-,and19-month-oldchildrenwere
recorded to slowly spoken, basic level words. While the children were looking
at sequentially presented pictures, they were acoustically presented with words
thatwereeithercongruousorincongruoustothepicturecontent.Ifinfantsare
alreadyabletointegratewordmeaningintosemanticcontext,providedhereby
thecontentsofthepictures,anN400primingeffectshouldoccur.Thesuccessful
comprehension of words should be reflected in a reduced N400 amplitude for
words that match their picture, while an enhanced N400 amplitude should be
presentfornon-matchingwords.
Inadults,thedescribedparadigmevokesanN400intheERPinresponseto
theincongruouswords(seeFigure9;Friedrich&Friederici2004).Thisnegative
deflectionismostpronouncedinthetimerangeof400to800mspost-stimulus
onset,predominantlyatcentro-parietalelectrodesites,butalsoextendstofrontal
sitesandlastsupto1200ms.AsshowninFigure9,in19-month-oldchildren
anN400-likeeffecttoincongruouswordswasobserved(Friedrich&Friederici
2004).Thisnegativity,withacentro-parietalmaximum,startsatabout400ms
post-stimulusandsustainsupto1400ms.Inchildren,theeffectstartslaterand
lasts longer than in adults, also showing a stronger involvement of the frontal
electrodesites.Fromthisitfollowsthateven19-month-oldchildrenshowalonglasting incongruity effect, which points to their ability to process the meaning
ofawordinitscontext.Thepicturecontentsuccessfullyactivatestheassociated
semanticknowledgethatfacilitatesorhamperssubsequentsemanticprocessing
whenawordmatchesordoesnotmatchthepicturecontent,respectively.
Anotherstudycarriedoutinourlaboratoryexaminedtheseearlywordlearningabilitiesincorrelationtothechildrenslaterexpressivelanguageskills(Friedrich & Friederici 2006). As described earlier, ERP data of 19-month-olds were
retrospectively grouped based on the childrens word and sentence production
Event-relatedbrainpotentials
19-month-olds
Adults
FC3
FZ
FC4
FC3
FZ
FC4
CP5
CZ
CP6
CP5
CZ
CP6
N400
N400
PZ
FC3 FZ FC4
PZ
CP5 CZ CP6
PZ
-4 +V
-10 +V
s
0 0.4 0.8 1.2 1.6
congruous words
incongruous words
s
0 0.4 0.8 1.2 1.6
Figure 9. Processingoflexical-semanticinformationatwordlevel:ERPdataof
19-month-oldsandadultsinresponsetosemanticallyincongruousandcongruous
wordsinapicture-wordparadigm(modifiedfromFriedrich&Friederici2004).
performanceattheageof30months(GermanlanguagedevelopmenttestSETK2; Grimm 2000). According to the childrens test results 19-month-olds were
classifiedbasedonwhethertheyhadlaterage-appropriatelanguageskillsoron
whethertheywereatriskforlaterlanguageproblems,thatis,basedonwhether
theyhadpoorlanguageproductionskillsatthewordorsentencelevel.Sinceimpaired language production is one of the major features of SLI, these children
seem to have an enhanced risk for later occurring SLI (see Leonard 1998; Rescorla,Roberts,&Dahlsgaard1997;Rescorla,Bascome,Lampard,&Feeny2001).
AsdisplayedinFigure10,childrenatriskforSLIdonotshowthesameN400
effectinresponsetotheincongruouswordsastheirpeers.Hence,childrenwho
showlanguageproductiondeficitsattheageof30monthsseemtobeimpaired
intheirwordprocessingabilitiesaboutoneyearbeforetheyexperienceproblems
intheirexpressivelanguageskills.Theresultssuggestthattheemergenceofthe
N400 during the development is related to childrens later expressive language
skillsandthatthemissingN400responseindicatesadelayedsemanticdevelopmentinthechildrenwithlaterlanguageproblems.Fromthisitfollowsthatthe
ERPmethodmaydeliverindicatorsforSLIriskatanearlierdevelopmentalstage
thanbehavioralmeasures.Whileitischallengingtotestchildrenwhoknowonly
afewwordsontheirexpressivelanguageskills,electrophysiologicalparameters
canbeobtainedfromearlyon.
3
ClaudiaMnnelandAngelaD.Friederici
At risk
Not at risk
FC3
FZ
FC4
FC3
FZ
FC4
CP5
CZ
CP6
CP5
CZ
CP6
N400
PZ
FC3 FZ FC4
PZ
CP5 CZ CP6
PZ
-12 +V
s
0
congruous words
incongruous words
Topinpointatwhatdevelopmentalstagetheneuralprocessesthatunderliethe
N400componentmature,anotherstudyofourlaboratoryfocusedonthedevelopmentalprogressionofwordprocessingabilitiesacrossdifferentagegroups(Friedrich&Friederici2005a,2005b).TheERPresponsesof12-,14-,and19-month-olds
toincongruousandcongruouspicture-wordpairswerecompared(seeFigure11).
Interestingly,theN400-likeeffecttoincongruouswordswasnotonlypresentin
19-month-oldsbutalsoin14-month-olds,althoughwithabroaderdistributionin
theyoungeragegroup.Incontrast,in12-month-oldinfantsthiseffectisnotyet
established.However,theseyoungestchildrenshowanearlynegativityinthetime
rangeof100to500mspost-wordonsetinresponsetocongruouswords(Figure
11).Thisearlynegativeeffectatlateralfrontalelectrodesiteswasalsopresentin
allotheragegroups.Thisnegativitycouldbeinterpretedasanearlyphonological-lexicalprimingeffectforthecongruentpicture-wordpairs.Morespecifically,
theauthorsproposeafacilitationoftheacoustic-phonologicalprocessingofthe
congruous words, which were expected due to lexical-semantic priming by the
pictures,whereastheincongruouswordswerenot(formoredetailseeFriedrich&
Friederici2005a,2005b).Thus,even12-month-oldsseemtohaveaninitiallexical
knowledgethatisactivatedbythepicturecontentandfavorstheacoustic-phonologicalprocessingofexpectedwords.TheabsenceofanN400inthesechildren,
Event-relatedbrainpotentials
14-month-olds
12-month-olds
-10 +V
10
PZ
0.8
-10 +V
1.2
1.6
10
PZ
s
0.4
0.8
1.2
1.6
N400
-8 +V
PZ
s
0.4
0.8
1.2
1.6
Adults
N400
-4 +V
PZ
s
0.4
0.8
1.2
1.6
F7
early negativity
10
N400
-10 +V
s
0.4
19-month-olds
s
0.4
0.8
1.2
congruous words
F7
PZ
incongruous words
1.6
however,supposesthattheirsemanticmemorystructuresarestilltoounspecificor
unstabletotriggerthemechanismunderlyingtheN400generation.
Similarlytothedescribedresults,earlynegativeeffectsintheERPhavebeen
observedinstudiesontheprocessingofknownandunknownwords.Negative
effectsinthetimerangeof200to400mspost-wordonsetweremorepronounced
inresponsetoknownwordsthantounknownwordsin20-andeven14-montholdchildren(Mills,Coffey-Corina,&Neville1993;1994;Mills,Prat,Zangl,Stager, Neville, & Werker 2004). Interestingly, in 20-month-olds but not yet in 14month-olds,ERPresponsesevendifferentiatedbetweenknownwordsandthose
words,whichdifferedintheirinitialphonemes,withmorenegativeresponsesto
theformer.Also,Thierry,Vihman,andRoberts(2003)foundearlynegativeERP
effectsinresponsetofamiliarwordsasopposedtounfamiliarwordsinthetime
range between 170 to 240 ms post-word onset. The combined results of these
studiessupportourviewthattheearlynegativeresponseatlateralsites,alsoobservedinourstudyforcongruouswords,reflectsafamiliarityeffect.Thisisinline
withERPeffectsin10-month-olds,whereafterseveralpresentationsoflowfrequencywordsanegativityemerges(Kooijmanetal.2005).Ingeneral,facilitated
phonologicalprocessingreflectedintheearlynegativitymightbeaffectedbyboth
familiarityonthebasisofrepeatedpresentationandexistingphonological-lexical
representationsinthelong-termmemory.
Insummary,inthecurrentERPstudiesonsemanticprocessingattheword
level,twoERPeffectsareobserved.First,anearlynegativityinresponsetocongruouswordsispresentinallagegroups,eveninthe12-month-olds.Weinterpretthisearlynegativeeffectasreflectingfacilitationofphonologicalprocessing
6
ClaudiaMnnelandAngelaD.Friederici
bylexicalpriming.Andsecond,thelatercentro-parietalnegativityintheERPfor
incongruouswordsispresentin14-and19-month-olds,aneffectreferredtoas
infantN400.Incontrasttotheirpeers,19-month-oldsatriskforlaterlanguage
impairmentdonotshowtheexpectedeffectofsemanticprocessingreflectedin
theN400.Sinceeven14-month-oldsshowanN400-likecomponent,thisimplies
thatat-riskchildrenaredelayedintheirsemanticdevelopmentforatleasthalfa
year.Theoccurrenceofaphonological-lexicalprimingeffectinallagegroupsindicatesthatnotonly14-and19-month-oldsbutalso12-month-oldsalreadycreatelexicalexpectationsfrompicturecontents,revealingthattheyalreadypossess
somelexical-semanticknowledge.However,infantsatthatagedonotyetdisplay
anN400semanticincongruityeffectthatispresentin14-month-olds.Fromthat
weassumethattheneuralmechanismsoftheN400maturebetween12and14
monthsofage.
ThedescribedERPstudiesshowthattheN400canbeutilizedasanindicatorforwordcomprehensioninagivensemanticcontexteveninearlychildhood.
TheresultthattheN400ispresentinthesecondyearoflifeimpliesthatthisERP
componentisausefultooltofurtherinvestigatehigher-levelsemanticprocessing
inveryyoungchildren.Inadditiontostudyingtheprocessingofsinglesemantic
units,thequestionariseswhenchildrenstarttoprocesssemanticrelationsbetweentheseunits,forinstanceinsententialcontext.
Event-relatedbrainpotentials
24-month-olds
19-month-olds
N400
-8 +V
0.8
N400
N400
-8 +V
PZ
0.4
Adults
1.2
1.6
PZ
0.4
0.8
1.2
1.6
PZ
-4 +V
PZ
s
0.4
0.8
1.2
1.6
semantically
correct
incorrect
expectationisviolatedwhentheverbattheendofthesentence(murdered) does
notsemanticallymeetthemeaningthatwassetupbythenouninthebeginning
of the sentence (honey). In the ERP responses to those incorrect sentences an
N400occursforthesemanticallyunexpectedsentenceendings,ashasbeendemonstratedinadultstudies(Friedericietal.1993;Hahne&Friederici2002).
Inaninfantstudyconductedinourlaboratory,theERPresponsestosemantically correct and incorrect sentences were analyzed in 19- and 24-month-old
children(Friedrich&Friederici2005c).Bothsentencetypesfollowedasimple
subject-verb-objectstructure.Semanticallyincorrectsentencescontainedobjects
thatviolatedtheselectionrestrictionsoftheprecedingverb,asinthesentence
The cat drinks the ballasopposedtoThe child rolls the ball.Ifchildrenareableto
processwordmeaninginsentencecontextandalreadypossesssomeknowledge
abouttheselection restrictions of particular verbs,anN400totheunexpected
sentenceendingsshouldbeobservableintheERPforincorrectsentences.
In both groups of children, the sentence endings of semantically incorrect
sentencesbutnotofcorrectsentencesevokedN400-likeeffectsintheERP,with
amaximumatcentro-parietalelectrodesites(Figure12).Incomparisontothe
adultdata,thenegativitiesinchildrenstartataboutthesametime,i.e.,ataround
400 ms post-word onset, but last longer. This suggests that semantically unexpectednounsthatviolatetheselectionrestrictionsoftheprecedingverbalsoinitiatesemanticintegrationprocessesinchildrenbutthattheseintegrationefforts
are maintained longer than in adults. Despite these processing differences, the
currentdataindicatethatchildrenattheageof24andeven19monthsareableto
processsemanticrelationsbetweenwordsinasentencequitesimilarlytoadults.
Otherelectrophysiologicalstudieshavefocusedontheprocessingoflexicalsemanticinformationatsentencelevelinolderchildren,namely5-to15-yearolds(Hahneetal.2004;Holcomb,Coffey,&Neville1992).Thesestudiesreported
7
8
ClaudiaMnnelandAngelaD.Friederici
N400-likeresponsestosemanticallyincorrectsentencesforallagegroups.Recently,Silva-Pereyraandcolleaguesinvestigatedsententialsemanticprocessingabilitiesinpreschoolersandin30-month-oldchildren(Silva-Pereyra,Klarman,Lin,
&Kuhl2005a;Silva-Pereyra,Rivera-Gaxiola,&Kuhl2005b).Inthepreschoolers
(3-and4-year-olds),sentenceendingsthatsemanticallyviolatedthepreceding
sentencephrasesevokedseveralanteriorlydistributednegativepeaks.Similarly,
in30-month-olds,ananteriornegativitybetween500800mspost-word-onset
wasobservedinresponsetosemanticallyanomaloussentences.Thedistribution
of these negativities did not match the usual centro-parietal maximum of the
N400.Therefore,itisaquestionwhetherthesenegativeERPresponsesindeedreflectsemanticintegrationprocesses.Here,itwouldhavebeenusefultohaveadult
datafromthesameexperimentalparadigmasabaselinetodeterminewhether
theusedparadigmevokesanN400component.Nevertheless,thesestudiesshow
differential responses to semantically incorrect and correct sentences in young
children.DespitethedifferenteffectsreportedintheERPstudiesonsentential
semanticprocessing,theresultsofourERPstudysuggestthatsemanticprocesses
atsentencelevel,asreflectedbyanN400-likeresponse,arepresentattheendof
childrenssecondyearoflife.
3.3
Processingofsyntacticinformation
Thesolecombinationofphonologicalfeaturesandsemanticunitsdoesnotyet
create meaningful language. In fact, a well-defined rule system is necessary to
relatetheelementsofasentencetogetherinanorganizedmanner,therebygiving
thesentenceitsstructure.Thecompositionofelementsisregulatedbythesyntacticrulesofaparticularlanguage.Theserulesareimportantinbothlanguage
comprehensionandproduction,sincetheyenablespeakersandlistenerstocommunicatewhodoeswhattowhom.Thus,syntacticrulesdefinethegrammatical
relationsofwordsandphrasesinasentence.Theanalysisofsyntacticrelations
betweenandwithinphrasesisacomplicatedandsophisticatedprocess,yetchildrenhaveacquiredthebasicsyntacticrulesoftheirnativelanguagebytheendof
theirthirdyearoflife.However,sofarnotmuchisknownabouttheunderlying
neurophysiologicalmechanismsoftheparticularstagesinsyntaxacquisition.
TheERPmethodpermitsinvestigatingtheneuralcorrelatesofsyntacticinformationprocessingbyapplyingthesyntactic violation paradigm.Here,syntacticallycorrectandsyntacticallyincorrectsentencesarepresented.Syntacticviolations are for example realized in morphosyntactic, phrase structure, or tense
violations.InadultERPstudies,twocomponentshavebeenobservedinresponse
tosyntacticallyincorrectsentencescontainingphrasestructureviolations.First,
Event-relatedbrainpotentials
theELAN,anearlyanteriornegativity,isinterpretedtoreflecthighlyautomatic
phrase-structure building processes (Friederici et al. 1993; Hahne & Friederici
1999).Second,theP600,alateroccurringcentro-parietalpositivity,istakentoreflectprocessesofsyntacticintegration(Kaanetal.2000)andcontrolledprocesses
ofsyntacticreanalysisandrepair(Osterhout&Holcomb1993;Friederici,Hahne,
&Mecklinger1996).ThisbiphasicERPpatterninresponsetophrasestructure
violationshasbeenfoundforbothpassiveaswellasactivesentenceconstructions
(Friedericietal.1993;Hahne&Friederici1999;Hahneetal.2004;Rossi,Gugler,
Hahne,&Friederici2005).
9
60 ClaudiaMnnelandAngelaD.Friederici
32-month-olds
24-month-olds
-10 +V
F7
-10 +V
ELAN
F7
Adults
-5 +V
ELAN
F7
F7
PZ
10
s
0.4 0.8 1.2 1.6
PZ
10
s
1.6
s
0.4 0.8 1.2 1.6
PZ
PZ
syntactically
correct
incorrect
P600
P600
P600
Otherneurophysiologicalstudieshaveaddressedtheprocessingofsyntactic
informationinolderchildren.Silva-Pereyraandcolleagues(2005a,2005b)examinedtheprocessingoftenseviolationsinactivesentencesinchildrenbetween30
and48months.Theauthorsreportedalatepositivityfortheolderchildrenand
a very late occurring positivity for the 30-month-olds. Furthermore, Hahne et
al.(2004)investigatedtheprocessingofphrasestructureviolationsinsyntacticallymorecomplicatedpassivesentences.TheELAN-P600patternwasseenin
7-to-13-year-oldchildren;however,6-year-oldsonlydisplayedalateP600butno
ELAN.Incomparisontotheseresults,theERPdataofthepresentstudiesshow
forthefirsttimethatthechildsbrainissensitivetophrasestructureviolationsin
activesentences,evenattheageofonly24months.TheERPsof32-month-olds
showbothinitialprocessesofstructurebuilding(ELAN)aswellaslateprocesses
ofsyntacticintegration(P600).Incontrast,theERPsof24-month-oldchildren
suggestadevelopmentalchangefrom2to2.5yearssinceinthesechildrenonly
aP600componentwithouttheexpectedELANoccurred.Thus,thedataindicate
that the two ERP components are somewhat independent and are likely to reflectdifferentprocessingmechanismsthatmatureatdifferentstagesduringthe
childrenslanguagedevelopment.
Insummary,thecombinedresultsindicatethatautomaticsyntacticprocesses
reflected by the ELAN are present earlier for active sentences than for passive
sentences during the childrens language development. Furthermore, the data
suggestthattheprocessesreflectedbytheP600areestablishedearlierduringthe
childsdevelopmentthanthosereflectedbytheELAN.Here,theERPmethod,in
contrasttobehavioralmethods,canhelptounderstandandsketchthedevelop-
Event-relatedbrainpotentials
mentalcourseofearlysyntaxacquisition,beforechildrencanactuallyproduce
comparablycomplexsyntacticconstructions.
.
Conclusion
Theresultsoftheneurophysiologicalresearchconductedinourandother laboratoriesdemonstratethatthemethodofevent-relatedbrainpotentialsisapowerful
tooltoinvestigateandmonitorearlystagesoflanguageacquisition.ERPsallowus
todescribetheunderlyingneurophysiologicalmechanismsofthelanguageacquisitionprocessaschildrendeveloptheirreceptivelanguageskills.Inthiscontext,
thedescribedexperimentsaimtoshownotonlythattherearespecificERPindicatorsofparticularlanguageprocessesininfantsandchildren,butalsothattheseindicatorscanbeusedastemplatestodefinethehallmarksoflanguageacquisition.
ThedescribedERPstudiesininfantsandchildrenbroadlycovertheprosodic,
semantic,andsyntacticaspectsoflanguageacquisitionduringthefirstthreeyears
oflife.WehavedemonstratedthattheERPmethoddeliversinformationabout
theneuralcorrelatesoflanguageprocessesandthereforeprovidesabetterunderstandingofthehowandwhenofthedevelopmentalstagesinthelanguageacquisitionprocess.Morespecifically,wehaveseenthataparticularERPcomponent
thatreflectsdiscriminationofphonologicalfeaturesispresenteveninnewborns
andcanthusbeusedtoexamineveryearlystagesoflanguageacquisition.Afurthercomponentthatindicateslexical-semanticprocessesinadults,theN400,is
registeredin14-month-olds,althoughnotyetin12-month-olds,andcanthereforebeusedtoinvestigatephonotacticknowledge,wordrecognitionandprocessingoflexical-semanticrelationsbetweenverbsandtheirargumentsinsentences.
Forthesyntacticdomain,anadult-likebiphasicERPpattern,theELAN-P600,is
presentin32-month-oldchildren,butnotyetin24-month-olds,fortheprocessingofstructuraldependencieswithinphrases,thuscharacterizingthedevelopmentalprogressionoftheacquisitionofsyntax.
ThecomparisonofadultandinfantERPdatashowsthatinfantERPcomponentsinresponsetospecificprocessesaresometimesdelayedascomparedtothe
adultcomponents,althoughthecomponentsrelativetimingmatchesthatofthe
adultdata.Specifically,localphrasestructurebuildingprecedeslexical-semantic
processesandthematicroleassignment.GiventhelatencydelayininfantERP
componentsontheonehand,buttheappearanceofthosecomponentsasinthe
adultdataontheotherhand,thecombinedresultsareinsupportofacontinuity
viewoflanguagedevelopment.
Insummary,theERPmethodisprovingtobeausefulresearchtoolinthe
workwithinfantsandchildren,especiallyduringearlystagesoflanguagelearning.
61
62
ClaudiaMnnelandAngelaD.Friederici
TheERPmethodplacesvirtuallynodemandsontheinfantsbehavioranddeliversanon-linemeasureofthebrainmechanismsunderlyingtheinfantslanguage
processingskills.Althoughwearestillfarfromadetailedoutlineoftheexactsteps
inthelanguageacquisitionprocess,usingtheERPmethodallowsresearchersto
studylanguagedevelopmentfromveryearlyonandthusprovidesthepossibility
tofurtherfillinthegapsinwhatweknowaboutlanguagedevelopmenttogaina
morefine-grainedpictureofacquisitionanditsneurophysiologicalbasis.
.
Future objectives
In the field of developmental neuroscience, researchers aim to achieve a more
detailedunderstandingoftheneuronalcorrelatesoflanguagelearning.Despite
theadvantagesandthefastadvancementsofneurophysiologicalandneuroimagingtechniques,thisresearchfieldconsiderablybenefitsfromthegroundbreaking
informationprovidedbybehavioralresearchers.Thus,effectiveinterdisciplinary
communicationisessentialtoscientificprogress,sincedifferentresearchareas,
usingvariousmethodsandfocusingondifferentaspectsoflanguagelearning,all
deliversinglepiecestothepuzzleoflanguageacquisition.
Furthermore, only an interdisciplinary communication between scientists
andearlychildcareprovidersaswellaseducatorsensuresaneffectiveuseofnew
findingsinthefieldofdevelopmentalcognitiveneuroscience.Informationmight
notonlyhelpabettergraspofthenormallanguageacquisitionprocessbutalso
delivernewinsightsintothenatureofimpairedlanguageacquisition.Importantly,oncetheERPmethodhasbeenfurtherdevelopedsothatconclusionscanbe
drawnnotonlyfromgroupdata,butalsofromindividualdata,certainERPcomponentsmightserveasearlyindicatorsofanimpairedlanguagedevelopment.If
theERPmethodcanbeutilizedasadiagnostictool,potentiallanguageproblems
canbeidentifiedearlyinthechildsdevelopmenttostartintervention.
References
Anderson,A.W.,Marois,R.,Colson,E.R.,Peterson,B.S.,Duncan,C.C.,Ehrenkranz,R.A.,et
al.(2001).Neonatalauditoryactivationdetectedbyfunctionalmagneticresonanceimaging.Magnetic Resonance Imaging, 19,15.
Bansal,P.,Sun,M.,&Sclabassi,R.J.(2004).Simulationandextractionofsingle-trialevoked
potentials.Proceedings of 26th Annual International Conference of the IEEE1.Engineering
in Medicine and Biology Society 2004. EMBC 2004, 200203.
Bee,H.&Boyd,D.(2007).The developing child.BostonMA:PearsonEducation.
Event-relatedbrainpotentials
Berg, P. & Scherg, M. (1994). A multiple source approach to the correction of eye artifacts.
Electroencephalography and Clinical Neurophysiology,90, 229241.
Bentin, S., Mouchetant-Rostaing, Y., Giard, M. H., Echallier, J. F., & Pernier, J. (1999). ERP
manifestations of processing printed words at different psycholinguistic levels: Time
courseandscalpdistribution.Journal of Cognitive Neuroscience,11(3),235260.
Brauer,J.&Friederici,A.D.(2007).Functionalneuralnetworksofsemanticandsyntacticprocessesinthedevelopingbrain.Journal of Cognitive Neuroscience,19(10), 115.
Brunia,C.H.M.,Mcks,J.,vandenBerg-Lenssen,M.M.C.,Coelho,M.,Coles,M.G.H.,
Elbert,T.,etal.(1989).CorrectingocularartifactsintheEEG:Acomparisonofseveral
methods.Journal of Psychophysiology,3,150.
Ceponiene,R.,Lepist,P.,Alku,H.A.,&Ntnen,R.(2003).Event-relatedpotentialindicesofauditoryvowelprocessingin3-year-oldchildren.Clinical Neurophysiology, 114(4),
652661.
Chatrian,G.E.,Lettich,E.,&Nelson,P.L.(1988).Modifiednomenclatureforthe10%electrodesystem. Journal of Clinical Neurophysiology, 5(2),183186.
Cheour,M.,Ceponiene,R.,Lehtokoski,A.,Luuk,A.,Allik,J.,Alho,K.,&Ntnen,R.(1998).
Development of language-specific phoneme representations in the infant brain. Nature
Neuroscience, 1,351353.
Christophe,A.,Gout,A.,Peperkamp,S.,&Morgan,J.(2003).Discoveringwordsinthecontinuousspeechstream:Theroleofprosody.Journal of Phonetics. Special Issue: Temporal
integration in the perception of speech,31(34),585598.
Clark,E.V.(2003).First language acquisition.Cambridge:CUP.
Coles,M.G.H.&Rugg,M.D.(1995).Event-relatedpotentials:Anintroduction.InM.D.Rugg
&M.G.H.Coles(Eds.),Electrophysiology of mind (pp.126).Oxford:OUP.
Cutler,A.,Dahan,D.,&vanDonselaar,W.(1997).Prosodyinthecomprehensionofspoken
language:Aliteraturereview.Language and Speech,40(2),141201.
Daffner,K.R.,Scinto,L.F.M.,Calvo,V.,Faust,R.,Mesulam,M.M.,West,W.C.,&Holcomb,P.
(2000).Theinfluenceofstimulusdevianceonelectrophysiologicandbehavioralresponses
tonovelevents.Journal of Cognitive Neuroscience,12(3),393406.
Dehaene-Lambertz, G., & Dehaene, S. (1994). Speed and cerebral correlates of syllable discriminationininfants.Nature, 370,292295.
Dien,J.(1998).Issuesintheapplicationoftheaveragereference:Review,critiques,andrecommendations.Behavior Research Methods, Instruments & Computers. Special Issue: Eventrelated brain potential methodology, 30(1),3443.
Donchin,E.&Coles,M.G.H.(1988).IstheP300amanifestationofcontextupdating?Behavioral and Brain Sciences,11,357374.
Edgar,J.C.,Stewart,J.L.,&Miller,G.A.(2005).DigitalfiltersinERPresearch.InT.C.Handy
(Ed.), Event-related potentials: A methods handbook (pp. 85113). Cambridge MA: The
MITPress.
Fabiani,M.,Gratton,G.,&Coles,M.G.H.(2000).Event-relatedpotentials:Methods,theory,
andapplications.InJ.Cacioppo,L.Tassinary&G.Berston(Eds.),Handbook of psychophysiology(2nded.,pp.5284).NewYorkNY:Plenum.
Friederici,A.D.(1994).FunktionaleOrganisationundReorganisationderSprachewhrend
derSprachentwicklung:EineHypothese.Neurolinguistik,8(1),4155.
Friederici,A.D.(2002).Towardsaneuralbasisofauditorysentenceprocessing.Trends in Cognitive Science, 6,7884.
63
6 ClaudiaMnnelandAngelaD.Friederici
Friederici,A.D.(2004a).Theneuralbasisofsyntacticprocesses.InM.S.Gazzaniga(Ed.),The
Cognitive Neuroscience III (3rded.,pp.789801).CambridgeMA:TheMITPress.
Friederici,A.D.(2004b).Event-relatedbrainpotentialstudiesinlanguage.Current Neurology
and Neuroscience Reports, 4,466470.
Friederici,A.D.(2005).Neurophysiologicalmarkersofearlylanguageacquisition:Fromsyllablestosentences.Trends in Cognitive Sciences,9,481488.
Friederici, A. D. & Mecklinger, A. (1996). Syntactic parsing as revealed by brain responses:
First-pass and second-pass parsing processes. Journal of Psycholinguistic Research, 25,
157176.
Friederici,A.D.&Wessels,J.M.I.(1993).Phonotacticknowledgeofwordboundariesanditsuse
ininfantspeechperception.Perception and Psychophysics, 54,287295.
Friederici,A.D.,Friedrich,M.,&Weber,C.(2002).Neuralmanifestationofcognitiveandprecognitivemismatchdetectioninearlyinfancy.NeuroReport, 13,12511254.
Friederici,A.D.,Hahne,A.,&Mecklinger,A.(1996).Temporalstructureofsyntacticparsing:earlyandlateevent-relatedbrainpotentialeffects.Journal of Experimental Psychology:
Learning Memory and Cognition,22,12191248.
Friederici,A.D.,Pfeifer,E.,&Hahne,A.(1993).Event-relatedbrainpotentialsduringnatural
speechprocessing:Effectsofsemantic,morphologicalandsyntacticviolations.Cognitive
Brain Research,1,183192.
Friedrich,M.&Friederici,A.D.(2004).N400-likesemanticincongruityeffectin19-montholds:Processingknownwordsinpicturecontexts.Journal of Cognitive Neuroscience, 16,
14651477.
Friedrich,M.&Friederici,A.D.(2005a).Phonotacticknowledgeandlexical-semanticprimingin
one-year-olds:Brainresponsestowordsandnonsensewordsinpicturecontexts.Journal of
Cognitive Neuroscience, 17(11),17851802.
Friedrich,M.,&Friederici,A.D.(2005b).Lexicalprimingandsemanticintegrationreflectedin
theERPof14-month-olds.NeuroReport,16(6),653656.
Friedrich,M.&Friederici,A.D.(2005c).Semanticsentenceprocessingreflectedintheeventrelatedpotentialsofone-andtwo-year-oldchildren.NeuroReport,16(6),18011804.
Friedrich,M.&Friederici,A.D.(2006).EarlyN400developmentandlaterlanguageacquisition. Psychophysiology,43,112.
Friedrich,M.,Weber,C.,&Friederici,A.D.(2004).Electrophysiologicalevidencefordelayed
mismatchresponseininfantsat-riskforspecificlanguageimpairment.Psychophysiology,
41,772782.
Gout,A.,Christophe,A.,&Morgan,J.L.(2004).Phonologicalphraseboundariesconstrain
lexicalaccessII.Infantdata.Journal of Memory and Language,51(4),548567.
Gratton,G.,Coles,M.G.H.,&Donchin,E.(1983).Anewmethodforoff-lineremovalofocularartifact.Electroencephalography and Clinical Neurophysiology,55,468484.
Grimm, H. (2000). Sprachentwicklungstest fr zweijhrige Kinder. Diagnose rezeptiver und
produktiver Sprachverarbeitungsfhigkeiten.Gttingen:Hogrefe.
Grimm, H. & Doil, H. (2000). Elternfragebgen fr die Frhererkennung von Risikokindern
(ELFRA1, ELFRA2).Gttingen:Hogrefe.
Guttorm,T.K.,Leppnen,P.H.T.,Poikkeus,A-M.,Eklund,K.M.,Lyytinen,P.,&Lyytinen,
H.(2005).Brainevent-relatedpotentialsmeasuredatbirthpredictlaterlanguagedevelopmentinchildrenwithandwithoutfamilialriskfordyslexia.Cortex. Special Issue: The
neurobiology of developmental disorders,41(3),291303.
Event-relatedbrainpotentials
Hahne,A.&Friederici,A.D.(1999).Electrophysiologicalevidencefortwostepsinsyntactic
analysis.Earlyautomaticandlatecontrolledprocesses.Journal of Cognitive Neuroscience,
11,194205.
Hahne,A.&Friederici,A.D.(2002).DifferentialtaskeffectsonsemanticandsyntacticprocessesasrevealedbyERPs.Cognitive Brain Research, 13,339356.
Hahne,A.,Eckstein,K.,&Friederici,A.D.(2004).Brainsignaturesofsyntacticandsemantic
processes during childrens language development. Journal of Cognitive Neuroscience, 16,
13021318.
Hebden,J.C.(2003).Advancesinopticalimagingofthenewborninfantbrain.Psychophysiology,40,501.
Hillyard,S.A.,Vogel,E.K.,&Luck,S.J.(1998).Sensorygaincontrol(amplification)asamechanismofselectiveattention:electrophysiologicalandneuroimagingevidence.Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 353(1373),
12571270.
Hirsh-Pasek,K.,Kemler-Nelson,D.G.,Jusczyk,P.W.,Cassidy,K.W.,Druss,B.,&Kennedy,L.
(1987).Clausesareperceptualunitsforyounginfants.Cognition,26(3),269286.
Holcomb,P.J.(1993).Semanticprimingandstimulusdegradation:Implicationsfortheroleof
theN400inlanguageprocessing.Psychophysiology,30,4761.
Holcomb,P.J.&Neville,H.J.(1990).Auditoryandvisualsemanticpriminginlexicaldecision:
Acomparisonusingevent-relatedbrainpotentials.Language and Cognitive Processes,5,
281312.
Holcomb,P.J.,Coffey,S.A.,&Neville,H.J.(1992).Visualandauditorysentenceprocessing:A
developmentalanalysisusingeventrelatedbrainpotentials.Developmental Neuropsychology,8,203241.
Holm,A.,Ranta-Aho,P.O.,Sallinen,M.,Karjalainen,P.A.,&Mller,K.(2006).Relationship
ofP300single-trialresponseswithreactiontimeandprecedingstimulussequence.International Journal of Psychophysiology,61,244252.
Houston,D.M.,Santelmann,L.M.,&Jusczyk,P.W.(2004).English-learninginfantssegmentation of trisyllabic words from fluent speech. Language and Cognitive Processes, 19(1),
97136.
Iacono,W.G.,&Lykken,D.T.(1981).Two-yearreteststabilityofeyetrackingperformanceand
acomparisonofelectro-oculographicandinfraredrecordingtechniques:EvidenceofEEG
intheelectro-oculogram.Psychophysiology,18(1),4955.
Ille,N.,Berg,P.,&Scherg,M.(1997).AspatialcomponentsmethodforcontinuousartifactcorrectioninEEGandMEG.Biomedical Techniques and Biomedical Engineering,42 (Suppl.
1),8083.
Jasper,H.H.(1958).Theten-twentyelectrodesystemoftheinternationalfederation.Electroencephalography and Clinical Neurophysiology,10,371375.
Jing,H.&Benasich,A.A.(2006).Brainresponsestotonalchangesinthefirsttwoyearsoflife.
Brain & Development,28(4),247256.
Jung,T.-P.,Makeig,S.,Westerfield,W.,Townsend,J.,Courchesne,E.,&Sejnowski,T.J.(2000).
Removalofeyeactivityartifactsfromvisualevent-relatedpotentialsinnormalandclinical
subjects.Clinical Neurophysiology,111(10),17451758.
Jung,T.-P.,Makeig,S.,Westerfield,M.,Townsend,J.,Courchesne,E.,&Sejnowski,T.J.(2001).
Analysisandvisualizationofsingle-trialevent-relatedpotentials. Human Brain Mapping,
14,166185.
Jusczyk,P.W.(1997).The discovery of spoken language.CambridgeMA:TheMITPress.
6
66 ClaudiaMnnelandAngelaD.Friederici
Jusczyk,P.W.,Cutler,A.,&Redanz,N.J.(1993).Infantspreferenceforthepredominantstress
patternsofEnglishwords.Child Development, 64,675687.
Jusczyk,P.W.,Hirsh-Pasek,K.,Nelson,D.G.,Kennedy,L.J.,Woodward,A.,&Piwoz,J.(1992).
Perceptionofacousticcorrelatesofmajorphrasalunitsbyyounginfants. Cognitive Psychology,24,252293.
Jusczyk,P.W.,Houston,D.M.,&Newsome,M.(1999).Thebeginningsofwordsegmentation
inEnglish-learninginfants.Cognitive Psychology,39(34),159207.
Kaan,E.,Harris,A.,Gibson,E.,&Holcomb,P.(2000).TheP600asanindexofsyntacticintegrationdifficulty.Language and Cognitive Processes,15,159201.
Katayama,J.&Polich,J.(1998).StimuluscontextdeterminesP3aandP3b.Psychophysiology,
35(1),2333.
Kenemans,J.L.,Molenaar,P.C.M.,Verbaten,M.N.,&Slangen,J.L.(1991).Removalofthe
ocularartifactfromtheEEG:Acomparisonoftimeandfrequencydomainmethodswith
simulatedandrealdata.Psychophysiology, 28(1),114121.
Klann-Delius,G.(1999).Spracherwerb.Stuttgart:Metzler.
Kooijman,V.,Hagoort,P.,&Cutler,A.(2005).Electrophysiologicalevidenceforprelinguistic
infantswordrecognitionincontinuousspeech.Cognitive Brain Research,24,109116.
Kooijman,V.,Hagoort,P.,&Cutler,A.(2006,March).Wordrecognitionincontinuousspeech
by7-month-oldinfants.PaperpresentationattheCUNYworkshopOn-lineMethodsin
ChildrensLanguageProcessing,NewYork,NY.
Kushnerenko, E., Cheour, M., Ceponiene, R., Fellman, V., Renlund, M., Soininen, K., et al.
(2001).Centralauditoryprocessingofdurationalchangesincomplexspeechpatternsby
newborns:Anevent-relatedbrainpotentialstudy.Developmental Neuropsycholgy,19(1),
8397.
Kushnerenko,E.,Ceponiene,R.,Balan,P.,Fellman,V.,Huotilainen,M.,&Ntnen,R. (2002).
Maturationoftheauditoryevent-relatedpotentialsduringthefirstyearoflife.NeuroReport, 13,4751.
Kutas,M.&Federmeier,K.D.(2000).Electrophysiologyrevealssemanticmemoryuseinlanguagecomprehension.Trends in Cognitive Science, 4,463470.
Kutas,M.&Hillyard,S.A.(1980).Readingsenselesssentences:Brainpotentialsreflectsemanticincongruity.Science,207,203205.
Kutas,M.&Hillyard,S.A.(1983).Event-relatedbrainpotentialstogrammaticalerrorsand
semanticanomalies.Memory & Cognition,11(5),539550.
Leonard,L.B.(1998).Children with specific language impairment.CambridgeMA:TheMIT
Press.
Leppnen,P.H.T.,Pikho,E.,Eklund,K.M.,&Lyytinen,H.(1999).Corticalresponsesofinfants
withandwithoutageneticriskfordyslexia:II.Groupeffects.NeuroReport, 10,969973.
Leppnen,P.H.T.,Richardson,U.,Pikho,E.,Eklund,K.M.,Guttorm,T.K.,Aro,M.,&Lyytinen,H.(2002).Brainresponsestochangesinspeechsounddurationsdifferbetweeninfants
withandwithoutfamilialriskfordyslexia.Developmental Neuropsychology,22,407422.
Lins,O.G.,Picton,T.W.,Berg,P.,&Scherg,M.(1993).OcularartifactsinEEGandevent-relatedpotential.II.Sourcedipolesandsourcecomponents.Brain Topography,6,6578.
LopesdaSilva,F.(1991).Neuralmechanismsunderlyingbrainwaves:Fromneuralmembranes
tonetworks.Electroencephalography and Clinical Neurophysiology,79,8193.
Mangun,G.R.(1995).Neuralmechanismsofvisualselectiveattention.Psychophysiology,32,
418.
Event-relatedbrainpotentials
Marcar,V.L.,Strassle,A.E.,Loenneker,T.,Schwarz,U.,&Martin,E.(2004).Theinfluenceof
corticalmaturationontheBOLDresponse:AnfMRIstudyofvisualcortexinchildren.
Pediatric Research,56(6),967974.
Martin,E.,Joeri,P.,Loenneker,T.,Ekatodramis,D.,Vitacco,D.,Henning,J.,&Marcar,V.L.
(1999).VisualprocessingininfantsandchildrenstudiedusingfunctionalMR.Pediatric
Research,46(2),135140.
Meek,J.(2002).Basicprinciplesofopticalimagingandapplicationtothestudyofinfantdevelopment.Developmental Science,5(3),371380.
Mills,D.L.,Coffey-Corina,S.A.,&Neville,H.J.(1993).Languageacquisitionandcerebral
specializationin20-month-oldinfants.Journal of Cognitive Neuroscience,5,317334.
Mills,D.L.,Coffey-Corina,S.A.,&Neville,H.J.(1994).Variabilityincerebralorganization
duringprimarylanguageacquisition.InG.Dawson&K.W.Fischer(Eds.), Human behavior and the developing brain(pp.427455).NewYorkNY:GuilfordPress.
Mills,D.L.,Prat,C.,Zangl,R.,Stager,C.L.,Neville,H.J.,&Werker,J.F.(2004).Languageexperienceandtheorganizationofbrainactivitytophoneticallysimilarwords:ERPevidence
from14-and20-month-olds.Journal of Cognitive Neuroscience, 16,14521464.
Morr,M.L.,Shafer,V.L.,Kreuzer,J.,&Kurtzberg,D.(2002).Maturationofmismatchnegativityininfantsandpre-schoolchildren. Ear and Hearing,23,118136.
Mrzljak,L.,Uylings,H.B.M.,VanEden,C.G.,&Judas,M.(1990).Neuronaldevelopmentin
humanprefrontalcortexinprenatalandpostnatalstages.Progress in Brain Research, 85,
185222.
Ntnen,R.(1990).Theroleofattentioninauditoryinformationprocessingasrevealedby
event-relatedpotentialsandotherbrainmeasuresofcognitivefunction.Behavioral and
Brain Sciences,13,201288.
Nazzi, T., Dilley, L. C., Jusczyk, A. M., Shattuck-Hufnagel, S., & Jusczyk, P. W. (2005). English-learninginfantssegmentationofverbsfromfluentspeech.Language and Speech,48,
279298.
Nazzi,T.,Iakimova,G.,Bertoncini,J.,Frdonie,S.,&Alcantara,C.(2006).Earlysegmentation
offluentspeechbyinfantsacquiringFrench:Emergingevidenceforcrosslinguisticdifferences.Journal of Memory and Language,54,283299.
Nelson, C. A. & Luciana, M. (1998). Electrophysiological studies II: Evoked potentials and
event-relatedpotentials.InC.E.Coffey&R.A.Brumback(Eds.),Textbook of pediatric
neuropsychiatry(pp.331356).WashingtonDC:AmericanPsychiatricPress.
Nobre,A.C.&McCarthy,G.(1994).Language-relatedERPs:Scalpdistributionsandmodulationbywordtypeandsemanticpriming.Journal of Cognitive Neuroscience, 6(33),233
255.
Oberecker,R.&Friederici,A.D.(2006).Syntacticevent-relatedpotentialcomponentsin24month-oldssentencecomprehension.NeuroReport, 17(10),10171021.
Oberecker,R.,Friedrich,M.,&Friederici,A.D.(2005).Neuralcorrelatesofsyntacticprocessing
intwo-year-olds.Journal of Cognitive Neuroscience, 17,407421.
Opitz,B.,Mecklinger,A.,Cramon,D.Y.,&Kruggel,F.(1999).Combiningelectrophysiological
andhemodynamicmeasuresoftheauditoryoddball.Psychophysiology,36,142147.
Osterhout,L.&Holcomb,P.J.(1992).Event-relatedpotentialsandsyntacticanomaly.Journal
of Memory and Language,31,785804.
Osterhout,L.&Holcomb,P.J.(1993).Event-relatedbrainpotentialsandsyntacticanomaly:
Evidenceonanomalydetectionduringperceptionofcontinuousspeech. Language and
Cognitive Processes,8,413437.
67
68 ClaudiaMnnelandAngelaD.Friederici
Pannekamp, A., Toepel, U., Alter, K., Hahne, A., & Friederici, A. D. (2005). Prosody-driven
sentenceprocessing:Anevent-relatedbrainpotentialstudy.Journal of Cognitive Neuroscience, 17,407421.
Pannekamp,A.,Weber,C.,&Friederici,A.D.(2006).Prosodicprocessingatthesentencelevelin
infants.NeuroReport,17,675678.
Pascual-Marqui,R.D.(2002).Standardizedlowresolutionelectromagnetictomography(sLORETA): Technical details. Methods & Findings in Experimental & Clinical Pharmacology, 24,
512.
Pascual-Marqui,R.D.,Michel,C.M.,&Lehmann,D.(1994).Lowresolutionelectromagnetictomography:Anewmethodforlocalizingelectricalactivityinthebrain.International Journal
of Psychophysiology, 18,4965.
Pihko,E.,Leppnen,P.H.T.,Eklund,K.M.,Cheour,M.,Guttorm,T.K.,&Lyytinen,H. (1999).
Corticalresponsesofinfantswithandwithoutageneticriskfordyslexia:I.Ageeffects.
NeuroReport, 10,901905.
Ponton,C.W.,Eggermont,J.J.,Khosla,D.,Kwong,B.,&Don,M.(2002).Maturationofhuman
centralauditorysystemactivity:separatingauditoryevokedpotentialsbydipolesource
modeling.Clinical Neurophysiology,113,407420.
Pujol,J.,Soriano-Mas,C.,Ortiz,H.,Sebastin-Galls,N.,Losilla,J.M.,&Dues,J.(2006).Myelinationoflanguage-relatedareasinthedevelopingbrain.Neurology, 66,339343.
Redcay,E.,Haist,F.,&Courchesne,E.(2006,March).Speechperceptionrecruitsfrontalcortex
duringapivotalperiodoflanguageacquisition.PaperpresentationattheCUNYworkshopOn-lineMethodsinChildrensLanguageProcessing,NewYorkNY.
Regan,D.(1989).Human brain electrophysiology: Evoked potentials and evoked magnetic fields in
science and medicine.NewYorkNY:Elsevier.
Rescorla,L.,Bascome,A.,Lampard,J.,&Feeny,N.(2001).Conversationalpatternsinlatetalkersatage3.Applied Psycholinguistics,22,235251.
Rescorla,L.,Roberts,J.,&Dahlsgaard,K.(1997).Latetalkersat2outcomeatage3.Journal of
Speech, Language and Hearing Research,40,556566.
Rivkin,M.J.,Wolraich,D.,Als,H.,McAnulty,G.,Butler,S.,Conneman,N.,etal.(2004).ProlongedT*[2]valuesinnewbornversusadultbrain:ImplicationsforfMRIstudiesofnewborns.Magnetic Resonance in Medicine,51(6),12871291.
Rossi,S.,Gugler,M.F.,Hahne,A.,&Friederici,A.D.(2005).Whenwordcategoryinformation
encountersmorphosyntax:AnERPstudy.Neuroscience Letters,384,228233.
Schapiro,M.B.,Schmithorst,V.J.,Wilke,M.,ByarsWeber,A.,Strawsburg,R.H.,&Holland,
S.K.(2004).BOLDfMRIsignalincreaseswithageinselectedbrainregionsinchildren.
NeuroReport,15(17),25752578.
Scherg,M.&vonCramon,D.(1986).Evokeddipolesourcepotentialsofthehumanauditory
cortex.Electroencephalography and Clinical Neurophysiology,65,344360.
Scherg,M.,Vajsar,J.,&Picton,T.W.(1989).Asourceanalysisofthehumanauditoryevoked
potentials.Journal of Cognitive Neuroscience,1,336355.
Silva-Pereyra,J.,Klarman,L.,Lin,L.J.,&Kuhl,P.K.(2005a).Sentenceprocessingin30-montholdchildren:Anevent-relatedpotentialstudy.NeuroReport,16,645648.
Silva-Pereyra,J.,Rivera-Gaxiola,M.,&Kuhl,P.K.(2005b).Anevent-relatedbrainpotential
study of sentence comprehension in preschoolers: Semantic and morphosyntactic processing.Cognitive Brain Research,23,247258.
Event-relatedbrainpotentials
Soderstrom,M.,Nelson,D.G.K.,&Jusczyk,P.W.(2005).Six-month-oldsrecognizeclauses
embeddedindifferentpassagesoffluentspeech. Infant Behavior & Development,28(1),
8794.
Soderstrom,M.,Seidl,A.,Nelson,D.G.K.,&Jusczyk,P.W.(2003).Theprosodicbootstrapping
ofphrases:Evidencefromprelinguisticinfants.Journal ofMemory and Language,49(2),
249267.
Speckmann,E.-J.&Elger,C.E.(1993).NeurophysiologicalbasisoftheEEGandofDCpotentials. In E. Niedermeyer & F. Lopes Da Silva (Eds.), Electroencephalography. Basic principles, clinical applications and related fields(pp.1526).BaltimoreMD:Urban&Schwarzenberg.
Steinhauer,K.,Alter,K.,&Friederici,A.D.(1999).Brainpotentialsindicateimmediateuseof
prosodiccuesinnaturalspeechprocessing.Nature Neuroscience,2,191196.
Szagun,G.(2006).Sprachentwicklung beim Kind.Weinheim:Beltz.
Tallal,P.,Miller,S.L.,Bedi,G.,Wang,X.,Nagarajan,S.S.,Schreiner,etal.(1996).Language
comprehensioninlanguage-learning-impairedchildrenimprovedwithacousticallymodifiedspeech.Science,271(5245),8184.
Thierry, G., Vihman, M., & Roberts, M. (2003). Familiar words capture the attention of 11month-oldsinlessthan250ms.NeuroReport,14,23072310.
Trainor,L.,McFadden,M.,Hodgson,L.,Darragh,L.,Barlow,J.,Matsos,L.,&Sonnadara,R.
(2003).Changesinauditorycortexandthedevelopmentofmismatchnegativitybetween
2and6monthsofage.International Journal of Psychophysiology, 51,515.
Uylings, H. M. B. (2006). Development of the human cortex and the concept of critical or
sensitiveperiods.Language Learning, 56(s1),5960.
Weber,C.,Hahne,A.,Friedrich,M.,&Friederici,A.D.(2004).Discriminationofwordstress
in early infant perception: Electrophysiological evidence. Cognitive Brain Research, 18,
149161.
Weber,C.,Hahne,A.,Friedrich,M.,&Friederici,A.D.(2005).Reducedstresspatterndiscriminationin5-month-oldsasamarkerofriskforlaterlanguageimpairment:Neurophysiologialevidence.Cognitive Brain Research,25,180187.
West,W.C.&Holcomb,P.J.(2002).Event-relatedpotentialsduringdiscourse-levelsemantic
integrationofcomplexpictures.Cognitive Brain Research, 13,363375.
Winkler, I., Kujala, T., Tiitinen, H., Sivonen, P., Alku, P., Lehtokoski, A., et al. (1999). Brain
responses reveal the learning of foreign language phonemes. Psychophysiology, 36(5),
638642.
Woldorff,M.G.,Gallen,C.C.,Hampson,S.R.,Hillyard,S.A.,Pantev,C.,Sobel,D.,&Bloom,F.
E.(1993).Modulationofearlysensoryprocessinginhumanauditorycortexduringauditoryselectiveattention.Proceedings of the National Academy of Sciences,90,87228726.
Appendix to Chapter 2
1.
Inworkingwithinfants,wearefacedwithcertainlimitationsthatmaketheexperimentalprocedure much more challenging than with adults. Specifically, an abbreviated attention span,
69
70 ClaudiaMnnelandAngelaD.Friederici
limitedverbalandmotorskills,frequentlyoccurringhungerandtirednessnecessitateshortexperimentsthatworkwithoutinstructionsanddonotrequireverbalormotorresponses.Ideally,
passivelisteningparadigms(likethepassive oddball paradigmandsemantic/syntactic violation
paradigm), sometimes combined with passive viewing (like the picture-word paradigm), are
usedandadjustedtotheinfantsabilities.Inourlaboratory,wefindthatexperimentsshould
take no longer than 1015 minutes; unless the outcome is not affected by the infants being
asleep(inthiscasewehavefound2030minutesacceptable).Althoughthetrialnumberhasto
bedrasticallyreducedforthisreason,thepresentationofatleast3060trialsperconditionis
requiredforsubsequentaveraging.Thiscanbedonebyreducingthenumberofexperimental
conditionsandsimplifyingthestimulusmaterialascomparedtoadultexperiments.Toensure
ahighqualityoftherecordedEEGdata,wealsofinditusefultoprogramtheexperimentso
thatitcanbepausedatanypointintimeoraftercertainexperimentalblocks.Thisallowsthe
experimentertoimmediatelyreacttotheinfantsneeds,suchasrearrangingtheirseatingor
reachingforapacifier.
Passivelisteningandviewingparadigmsenablecomparisonsacrossagegroupsbecause
infants,children,andadultscanallbetestedunderthesameconditions.Foradults,thepassive
testingconditionissometimesdifficultbecauseitrequiresmaintainingattentionthroughout
theexperimentwithoutperformingatask.Therefore,inadultsitmightbeusefultomonitor
thesubjectsattentionviavideoandmotivatethemthroughgeneralinstructionsandtheannouncement of a post-experimental questionnaire on the experiment. Accordingly, possible
attention-related differences in testing conditions should be taken into consideration when
comparingdataacrossagegroupsordataofdifferentstudies.
2.
Duringtheset-upprocedure,theinfantsusuallysitontheirparentslap,watchingabookor
playing.WhilepreparingtheinfantsfortheEEGrecording,allstepsshouldbesimultaneously
explainedtotheparents.Theset-uptimeisindividuallyvariable(e.g.,dependentonhowshy
infantsareorwhethertheyarefamiliarwiththesituation)andisalsodependentontheinfants
age.Theyoungertheinfants(16months),thefastertheset-upshouldbecompletedtoavoid
early fatigue effects during the experiment. From about 4 months on it becomes more and
moreimportanttoofferinterestingdistractions,suchasjinglingtoys,handpuppets,ormovies.
Fromabout18monthson,theexperimentalprocedureshouldbeadditionallyexplainedtothe
childrenconvincingthemonhowfancyandcoolitistowearsuchahatorlettingthemdo
thesameset-uponadoll.Sincethisworkdemandsattentiontobothinfantsandparentsand
requiresveryflexiblearrangements,werecommendthattwoexperiencedexperimentersare
present.Also,theset-upshouldtakeplaceinapreparationroom(otherthanthetestingroom)
withlotsoftoysandbabyequipmenttomakeinfantsandparentsfeelcomfortable.
OurEEGrecordingequipmentconsistsofdifferentlysizedelasticcapsfittedwithalowdensitymontageofsilver-silverchloridelectrodes(EASY CAP GmbH).Wefindthatthislowdensitymontagesystemfitstherequirementsofourdevelopmentalresearchbecauseitiseasy
tosetupandwedonotaimtoperformsourcelocalization(forsourcelocalizationpurposes
seehigh-densityEEGrecordingsbyGeodesic Sensor Net, Inc.).Thenumberofelectrodesused
variesfrom9to32dependingmainlyontheinfantsage(headsize)andtheappliedparadigm.
AllelectrodesarearrangedaccordingtotheInternational1020SystemofElectrodePlace-
Event-relatedbrainpotentials
ment(Jasper1958)ortheextended1010System(Chatrian,Lettich,&Nelson1988),andare
attachedtothecapbyringadapters.Thismontagetechniqueallowsforsimplecleaningbyrinsingcapandelectrodesseparatelyinwarmwaterandforeasyexchangeofbrokenelectrodes.
To monitor eye movements (see data analysis), we place additional electrodes on the outer
cantiofbotheyesandontheinfra-andsupraorbitalridgesoftherighteye(electrooculogram;
EOG).Notethat,excepttheelectrodebelowtherighteye,allelectrodes(includingcommon
referenceandground)areimplementedinthecap,sothatonlyonesingleelectrodehastobe
placedattheinfantshead,whichallowsaneasyandquickset-up.Toensureproperelectrode
placementtheelasticcapisfixedbyacheststrapbecausechinstraps(oftenusedinadults)are
notcomfortableforinfants.Toapplytheconductancegelundertheringelectrodes,wesimply
useablunttipsyringeandcottonswabs.Notethattheconductancegelshouldbenon-abrasive
andnotcontainanypreservatives.
TheEEGrecordingtakesplaceinanoise-shieldedcabin,onlyequippedwithloudspeakers
andavideoscreentokeeptheinfantsfocusedontheexperiment.Duringtherecordings,the
infantssitontheirparentslaporinacarseatwiththeparentssittingnexttothem(preferably
thelatter,sinceitminimizesmovement).Ineithercase,itisveryimportanttocarefullyinstruct
theparents:notalking,nobouncingorswinging,andifnecessaryredirectingtheinfantsattentiontothescreenbypointing.Also,theparentsshouldwearearplugsorheadphonessothat
theycannotheartheacousticstimuli.Duringthewholetime,theinfantsaremonitoredviavideoandifnecessarytherecordingispaused.Toachieveeyefixationandreducemovementwe
usuallyprovideasilentvideo(e.g.,slowlymovingfish)andtrytoattracttheinfantsattention
byhandpuppets,soapbubbles,etc.whentheybecomeinattentiveandrestless.Ifnecessary,the
infantsalsomayhaveapacifier,bottle,etc.Giventhevariabilityintheinfantsstate,wehighly
recommendtocarefullydocumenttheexactcourseoftheexperimentandthespecifictesting
conditionstodeterminepossibleexternalinfluences,suchastimeoftheday,nursing,tiredness,
crying,useofbottleorpacifier,kindofdistraction.
3.
Theoff-lineprocessingoftherecordedEEGdatafocusesonthetime-lockedaveragingofthe
eventsofinteresttoderiveERPs.Insomecases,however,specificallyinnoisyinfantdata,prior
filteringandartifactreductionproceduresareunavoidabletoremoveunwantednoise,caused
byheadandeyemovement,perspiration,etc.
Dependentontheelectrodereferencingduringtherecording,thedatamayhavetoberereferenced.Inourexperiments,theEEGisrecordedtothecommonreferenceCz(anelectrode
thatisaboutequidistantfromallotherelectrodesandenablesstablerecording).Off-line,the
dataarethenre-referencedtoamathematicallylinkedreferencefromtwoseparatelyrecorded
electrodesitesthataresomewhatdistantfromtheothersitesandrelativelyinactive,e.g.,the
mastoids.Itisimportanttonotethatthechoiceofreferencehasaninfluenceontheamplitudes
aswellasthetopographyoftheERPsandshouldbeconsideredwhencomparingdataofdifferentstudies(fordiscussionseeDien1998).
Filteringreferstotheremovalofcertainfrequenciesfromthesignalthataresufficiently
differentfromthefrequenciesthatcontributetotheERPwaveform(amatterthatbecomes
morecomplicatedasthefrequencycontentofthenoiseandthesignalaremoresimilar).Accordingly,mostresearchersapplybandpassfiltersthatspare,forinstance,frequenciesbetween
71
72
ClaudiaMnnelandAngelaD.Friederici
0.01Hzand30Hz,knowntoreflectmostoftheportionoftheERPwaveform.Inthisway,
highfrequencies(e.g.,frommusclecontractions)aswellasveryslowfrequencies(e.g.,from
skinpotentials/perspiration)aresuppressed.Thelatterisespeciallyrelevantforinfantdatathat
areoftendrift-contaminatedandshouldnotenterERPaverageswithoutpriorfiltering.Here,
high-passfiltersmightbeevensetto0.5Hzor1.0Hz,iftheexperimentdoesnotfocusonthe
analysisofslowpotentials.Thus,filtersshouldbedesignedhavingboththefeaturesoftherecordedsignalandtheexpectedERPresultsinmind.Inotherwords,wegenerallyrecommend
theuseoffiltersbutalsoadvisetheircautiousapplicationbecausefilterscansubstantiallydistorttheERPdata(formoredetailonthefunctionanddesignoffilterssee,forexample,Edgar,
Stewart,&Miller2005).
Artifactrejectionproceduresareappliedtoincreasethesignal-to-noiseratiobyeliminating trials that contain artifacts. Most artifacts are substantially larger in amplitude than the
EEGsignalandoriginatefromvariousexternalsources,suchaseyemovement,headorbody
movement,andtechnicalequipment.Automaticartifactrejectionproceduresareforinstance
settoexcludealltrialsexceeding100V(inadults)or200V(ininfants).However,giventhe
high variability of infant data consideration, we recommend rejecting trials that exceed for
instanceastandarddeviationof80Vinaslidingtimewindowof200ms,ratherthansetting
anabsoluteamplitudecriterion.
Agreatdealofsignalcontaminationstemsfromocularartifacts,particularlyblinksthat
primarilycontributetotheEEGmeasuredatanteriorsites.Sinceinfants,incontrasttoadults,
cannotbegivenblinkinstructionsthecruderejectionofcontaminateddatamightleavetoo
fewtrials.Forthosecases,artifactcorrectionproceduresmightbemoreappropriatethanthe
completeexclusionofcontaminatedtrials.Therearevariouscorrectionproceduresthat,ina
firststep,mathematicallycalculatethepropagationbetweentheeyesandeachofthescalpelectrodesordeterminethespatialcomponentsoftheeyeactivity.Inthesecondstep,theestimated
EOG proportion (propagation factors) is subtracted from the EEG signal at each site (for a
comparisonofthedifferentmethodsseeBruniaetal.1989).Thecorrectiontechniquesdiffer
intheirmathematicalapproachandtheinvolvedtheoreticalassumptions,namelyregression
methods(e.g.,Gratton,Coles.,&Donchin1983;Kenemans,Molenaar,Verbaten,&Slangen
1991),dipolesourcemodeling(e.g.,Berg&Scherg1994;Lins,Picton,Berg,&Scherg1993),
principalcomponentanalysis(PCA,e.g.,Ille,Berg,&Scherg1997),andindependentcomponentanalysis(ICA,e.g.,Jungetal.2000).Intheanalysisofinfantdata,wehighlyrecommend
artifactcorrectionprocedurestoavoidlosingtoomanytrials.However,itisimportanttokeep
inmindthateachmethodhasitsownspecificflaws,suchasover-correctionofEEGsignalin
regressionmethods(sincetheEEGsignalalsopropagatestoEOGchannelsandcontributesto
theestimationofregressioncoefficients;seeIacono&Lykken1981).
Fortheaveragingprocedureaminimumnumberofartifact-freetrialsisrequired.Inour
laboratorywerequireeachinfanttodeliveratleast1020artifact-freetrialstotheindividual
averageandatleast2030subjectstocontributetotheoverallaverage.Thus,tocompletean
experimentweusuallytest3050infantswithasuccessrateof6585%,dependingonthetheir
age(e.g.,4-month-oldsareeasiertotestthan12-month-olds,whowanttogetupandwalk
around)andthedemandsoftheexperiment(e.g.,passivelisteningaloneoftenhasahigher
successratethanadditionalpassiveviewing).
chapter3
Thischapterdescribesandevaluatestheuseofeyetrackingmethodstostudy
thedevelopmentofspokenlanguageproductionandcomprehension.Theemphasiswillbeonunderstandingthechainofinferences,orlinkingassumptions,
researcherscommonlymakewhengoingfrommeasurementsofeyeposition
toconclusionsaboutattention,referenceandsentenceparsing.Itisarguedthat
theseassumptionsarevalid,thoughcareisneededwhendisentanglingdevelopmentalchangesinvisualattentionfromdevelopmentalchangesinlanguage
processingabilities.
1.
Introduction
Cognitivedevelopmentisoftenviewedastheacquisitionofknowledge:welearn
factsabouttheworldaroundus;welearnthewordsandthegrammarofalanguage,etc. Analternativewayofthinkingaboutcognitivedevelopment,which
has gained some traction recently in the developmental literature, is to treat it
astheacquisitionofdynamicskills:welearnhowtointeractwiththeworld;we
learnhowtoproduceandcomprehendalanguage, etc. Theworkdiscussedinthis
volumeisaboutthisdynamicprocessingapproachtodevelopment,particularly
asitpertainstolanguagedevelopment.Recentinterestinthisissuestemsinpart
from concurrent methodological advancements; it is now possible for instance
torecordchildrenseyemovementsastheycarryoutrelativelynaturaltasksinvolvinglanguage,suchasfollowingspokeninstructions,inspectingimagesthat
arebeingdescribed,andevenengaginginaspokenconversationwithinterlocutors.Theresultingeyemovements,whenlinkedwithlinguisticevents,provide
researchers with a record of each childs moment-by-moment consideration of
possiblereferentsintheworldandthustellusinsomedetailabouttheprocessthe
childisgoingthroughwhenderivingmeaningfromlinguisticforms.
7
JohnC.Trueswell
Thischapterdescribesandevaluatesthisvisualworldmethod(Tanenhaus,
Spivey-Knowlton,Eberhard,&Sedivy1995)andfocusesespeciallyonhowithas
beenappliedtosentenceprocessingresearchwithtoddlersandchildren.Theemphasisherewillbeonunderstandingthelinkingassumptionsnecessarytouseeye
movementstostudylanguagedevelopment.Thatis,thischapterwillexplorethe
chainofinferencesresearchersusuallymakewhengoingfrommeasurementsof
dartingeyestoconclusionsaboutattention,referenceandevensentenceparsing.
Theplanistostepthroughtheselinkingassumptionsandexploretheextentto
eachisvalidandhoweachmightinteractwithknowndevelopmentalchangesin
attention.
I hope to convince the reader that conclusions drawn from developmental
researchusingthisvisualworldparadigmrequirecarefulconsiderationofhow
certainattentionalskillsdevelop,inparticular,thedevelopingabilitiestoengage
inthecontrolofinformationcollection(acomponentofattentional control)and
informationre-characterization (acomponentofcognitive control). Iwilldiscuss
howthesetwokindsofattentionalabilitieschangeoverdevelopment,andhow
thesechangesmightbearupontheinterpretationofeyemovementresearchin
psycholinguistics.Withrespecttoinformationcollection,itiswellknownthat
the eye movements generated during the visual interrogation of the world are
drivenbybothexogenousandendogenousfactors(i.e.,bybothbottom-upvisual
factorsandexperience-relatedgoalssetbytheindividual).Withrespecttoinformationre-characterization,itiswellknownthathumansroutinelycharacterize
perceptualinputalongseveraldifferentdimensionsatseverallevelsofabstraction. Languageisperhapstheparadeexampleofthis;wecharacterizelinguistic
input acoustically, phonologically, syntactically, semantically and referentially,
witheachcharacterizationhavingitsownrepresentationaldimensions.Adultlistenersmustbeabletocontrolthecontentofthesecharacterizationsinreal-time
and override certain characterizations when conflicting evidence arises within
andacrosstheselevels.Indeed,theskillofdealingwithconflictturnsouttobe
importantinthedevelopmentofsentencecomprehensionabilities.
With this broader understanding of how attentional and cognitive control
abilitiesdevelop,researchersarelikelytomake(andarealreadymaking)significantadvancesinunderstandinghowthedynamicsoflanguagecomprehension
andproductionemergeintheyoungchild.Itismyhopethattouringthesefacts
herewillallowotherstotakeadvantageofthevisualworldmethod,andthatit
willfacilitatetheoreticaladvancementsinunderstandinglanguageacquisitionas
thedevelopmentofadynamicinformationprocessingskill.
2.
Usingeyemovementsasadevelopmentalmeasure
Cooper(1974)wasthefirsttouseeyemovementsasareal-timemeasureofadults
spoken language processing abilities. In a series of eyetracking experiments, it
wasobservedthatadultlistenersrapidlyfixatepicturesdepictingthereferentsof
heardspeech,oftenmid-word,priortothecompletionoftheutterance.Thiswork
receivedfairlylimiteddiscussioninthepsycholinguisticcommunityuntilthereintroductionofthismethodbyTanenhausandcolleagueswhoexploredtheeye
gazeoflistenersinthenaturalsettingoffollowingspokeninstructionstomove
aboutobjectsintheworld(Tanenhausetal.1995).Tanenhausetal.demonstrated
thatwhenadultparticipantsfollowspokeninstructionstomanipulateobjectsina
task-relevantvisualcontext,fixationstotheseobjectsarealsocloselytime-locked
totheelementspresentintheunfoldingutterancethatsignalabstractrepresentationalunits.Itwasthereforepossiblefromthisworktoinferagreatdealabout
thelexical(e.g.,Allopenna,Magnuson,&Tanenhaus1998)andsyntactic(e.g.,
Spivey,Tanenhaus,Eberhard,&Sedivy2002)hypothesesthatadultsconsideras
thespeechisperceived.Sincepublicationofthisseminalwork,agrowingbody
ofresearchhasdemonstratedthateyemovementscanbeusedtotracethetime
courseofadultlanguagecomprehension,productionandevendynamicconversation.(SeetheeditedvolumesofHenderson&Ferreira2004;andTrueswell&
Tanenhaus2005;forthoroughreviews.)
2.1
Eyetrackingtechniquesforusewithchildren
Thedevelopmentofaccuratehead-mountedandremote eyetrackingsystemshas
madeitpossibletoconductsimilarvisualworldstudieswithyoungchildren,toddlersandeveninfants.Head-mountedsystems(Figure1)usehighlyminiaturized
camerasandopticsmountedonavisor(twocameras,onetrainedontheeyeand
theotheronthesurroundingvisualworld).Inthesesystems,thevideooutputfrom
theeyecameraisanalyzedinrealtimetocalculatethecurrentlocationofthepupil
(i.e.,thecentralpositionofallthedarkestpixels)andthecenterofthecornealreflection(i.e.,thecentralpositionofthebrightestpixels).Duringaninitialcalibrationprocedure,thesecoordinatesaremappedontocoordinatesinthescenevideo.
Thisistypicallydonebyaskingtheparticipanttolookatlocationsintheworldthat
correspondtoparticularpixelcoordinatesinthescenevideo.Foreachlocation,
thepupilandcornealreflectioncoordinatesintheeyecameraaresampledand
pairedwithacoordinatepositioninthescenecamera.(Informally,thecomputeris
beingtoldthattheparticipantseyeballlookslikethiswhentheparticipantislookinghereanditlookslikethiswhentheparticipantislookingoverhere,etc.).The
7
76
JohnC.Trueswell
Figure 1. Head-mountedeyetracking.
resultingmatrixofcoordinates(tripletsofpupil,cornealreflectionandposition
coordinates)isthenanalyzed.Thisanalysiscreatesamulti-dimensionallinearor
nonlinearregressionequationthatreflectsthebestfitbetweentheeye-calibration
coordinatesandthescene-calibrationcoordinates.Thisequationcanthenbeappliedinrealtimethroughouttheexperiment,suchthatforanypupilandcorneal
coordinates,thecorrespondingscenecoordinateisgeneratedandplottedontopof
thescenevideo(usuallyasamovingdotorcrosshair).
Thiscalibrationprocedurecanbedifficulttousewithchildrenbecauseitrequiresthechildtoholdhis/herheadstillwhilefixatingatargetlocationinthe
world. However, some calibration procedures eliminate this problem; in the
point-of-lightcalibrationprocedure,theexperimenterholdsasmalllight(such
asasmallLED)whiletheparticipantfollowsthelightaroundwithhis/hereyes.
Theeyetrackingcalibrationsoftwarethensamplesthepositionofthisbrightlight
inthescenevideoandpairsitwiththepupilandcornealcoordinatesfromthe
eyevideo,therebycreatingacalibrationmatrix.Thisproceduredoesnotrequire
thechildtoholdstill,andsubstantiallydecreasescalibrationtimeandincreases
calibrationaccuracy.
Remoteeyetrackingsystems(Figure2)worklikehead-mountedsystemsexcepttheopticsarehousedoffthehead,requiringnovisor.Thesesystemsrequire
trackingoftheheadaswell,eitherviavideo-basedmethods(e.g.,theTobii1750)
orbymagneticheadtracking(e.g.,theASLandISCANsystems).Remotesystems
arebecomingincreasinglypopularbecausetheycanbeeasiertousewithtoddlers
andeveninfants(e.g.,Aslin&McMurray2004;S.Johnson,Slemmer,&Amso
2004).Mostremotesystemsmapdirectionofgazedirectlyontothecoordinates
ofacomputervideodisplay,ratherthanascenecamera,allowingforsimpleautomaticcodingofeyeposition.However,itisalsopossibletousesuchsystemsto
generateathree-dimensionalvectoroftheparticipantsgazeinthephysicalworld
ratherthanavirtualworld.
Usingeyemovementsasadevelopmentalmeasure
B . T obii R emote
A . A SL R emote
Figure 2. Remoteeyetracking.
Figure 3. Poormanseyetracker.
Finally,severallabs(includingmyown)alsousewhatweaffectionatelycall
thepoormanseyetracker(Figure3,seealsoSnedeker&Thothathirithisvolume).Inamodifiedpreferentiallookingprocedure,avideocameraislocatedin
thecenterofaplatformthathasbeenplacedinfrontofthechild.Thiscamerais
trainedonthechildsfaceandeyes.Objectsareplacedontheplatform,usually
in four different quadrants around the camera. Direction of gaze toward each
quadrantcanbecodedfromthevideoofthechildsface;atrainedcodercanuse
adigitalvideoeditingsystemtostepthroughthevideoframe-by-frame,recording shifts in gaze. Hand coding of this sort is quite time consuming; it takes
approximatelyanhourtocodetentofifteenexperimentaltrialswheneachtrial
consistsofoneortwoutterances.However,nocalibrationprocedureorexpensiveeyetrackingequipmentisrequired.Thishand-codingprocedurealsotoleratesconsiderableheadmovementswithoutsubstantiallossincodingaccuracy.
Wehavefoundthatinter-coderreliabilityisusually9095%onaframe-by-frame
basis(Snedeker&Trueswell2004).Similarhand-codingproceduresareusedin
77
78
JohnC.Trueswell
picture-viewingtaskswithinfantsandtoddlers(see,e.g.,Fernald,Zangl,Portillo,
&Marchmanthisvolume;Swingley,Pinto,&Fernald1999).
2.2
Dataanalysis
Regardless of the data collection technique used by the experimenter, similar
analysescanbeperformedontheresultinggazerecord.Foreachtrialofinterest, the childs direction of gaze is linked to the onset of critical speech events
(e.g.,theonsetofcriticalwordsinasentence)andthenaveragedacrosstrialsand
participants.Forexample,Trueswell,Sekerina,HillandLogrip(1999)evaluated
thetimecoursewithwhich5-year-oldchildrenvisuallyinspectasetoffourpossible referents, relative to critical word onsets in a sentence. The children were
instructedtolookatacentrallylocatedsmileyfacestickerandthentofollow
instructionstomovesomeoftheobjects.Forpurposesofillustrationconsidera
hypotheticaltrialinwhichparticipantsheard:Look at the smiley face. Now put the
frog thats on the napkin into the box.
AphotographofasamplesceneforthisitemispresentedinFigure4.1Objects
includethetarget(afrogonanapkin),thecompetitor(afrogonaplate),acorrect
goal(anemptybox)andanincorrectgoal(anemptynapkin).Theupperright
panelofFigure4showstheeyegazerecordsfromfivehypotheticaltrials.Thezero
mspoint(wherethex andy axesmeet)indicatestheonsetofthespokenword
put.Inaddition,theonsetsofthenounsaremarked(frog,napkinandbox).On
trialone,thehypotheticalparticipantinitiatedalooktothetargetabout400ms
aftertheonsetofthewordfrog andthenlaunchedalooktothecorrectgoallater
inthesentence.Ontrialtwo,thefixationonthetargetbeginsabitlater.Ontrial
three,thefirstfixationisonthecompetitor,followedbyafixationonthetarget
andthenthecorrectgoal.Ontrialfour,thefixationsequenceistarget,incorrect
goal,andcorrectgoal.Trialfiveshowsanothertrialwheretheinitialfixationison
thecompetitor.ThelowerrightpanelofFigure4providesaplotoftheproportion
oflooksovertimeforthefourregions,averagedacrosstrialsforthishypothetical
participant.Thesefixationproportionsareobtainedbydeterminingtheproportionoflookstothealternativeobjectsateachtimeslice(asderivedfromthetrial
samples)andshowhowthepatternoflookstoobjectschangesasthesentence
unfolds.2Theprobabilitiesdonotsumto1.0becausemostparticipantswereini1. ThisfigureismodeledafterasimilardiscussionofeyemovementsappearinginTanenhaus
andTrueswell(2005).
2. Like most psycholinguistic studies, several similar target trials are provided to the same
participant;e.g.,inadditiontothefrogitem,theremightbeaniteminvolvingcows:Put the
Usingeyemovementsasadevelopmentalmeasure
Figure 4. Calculatinggazeproportionsovertime(modifiedfromTanenhaus&
Trueswell2005).
tiallyfixatingonthesmileyface,whichisnotplottedhere.Ifitwereplotted,looks
tothesmileyfacewouldsteadilydropovertimewhilechildrenbegintoinspect
thetask-relevantobjects.
Researchersoftendefineatimewindowofinterest.Forexample,onemight
wanttofocusonthelookstothetargetandcompetitorinatimeregionstarting
200msaftertheonsetofthewordfrog andending200msaftertheonsetofthe
wordnapkin.This200msoffsetisdesignedtotakeintoaccountthatittakesabout
200 ms for a participant to program an eye movement to a target (e.g.,Matin,
Shao, & Boff 1993, though see below). The proportion of looks to objects, the
timespentlookingatthealternativeobjects(essentiallytheareaunderthecurve,
whichisasimpletransformationofproportionoflooks),andthenumberand/or
proportionoflooksgeneratedtoobjectsinthistimeregioncanthenbeanalyzed.
Thesedifferentmeasuresareallhighlycorrelatedbutinprincipleofferslightly
differentpicturesoftheeyemovementrecord.
TheactualdatafromthisconditionintheTrueswelletal.studyarereproducedinFigure5.Focusingonlyonthelookstothetargetandthecompetitor,
onecanseethattheselooksarefairlywelltime-lockedwiththeonsetofwords;
79
80 JohnC.Trueswell
tar get
competitor
incor r ect
goal
Figure 5. Childrens(5yearolds)proportionoflooksovertimetopotentialreferent
objectsinresponsetoPut the frog thats on the napkin into the box.FromTrueswell,
Sekerina,HillandLogrip(1999).Copyright1999,ElsevierPress.
first,lookstoboththetargetandcompetitor(thetwofrogs)risesharplyupon
hearingthefirstnoun,frog,andremainequallydistributedbetweenthesetwoobjectsuntilnapkin,atwhichtimeparticipantsbegintolookmoreatthetarget(the
frogonthenapkin).Similarly,lookstothecorrectgoalriseuponhearingbox.
Itisnotthecasethattheeyessimplydarttoobjectsthatbestmatchthenouns
mentionedintheinput.Forinstance,attheonsetofthenounnapkin,eyepositiondoesnotsplitbetweenthetwonapkinsinthescenelikeitdidforthetwo
frogswhenhearingfrog.Rather,lookstothetarget(thathasthenapkinunderit)
prevailoverlookstotheincorrectgoal(theemptynapkin).Whywouldthisbe?
Themostplausibleexplanationisthatthisisduetothesyntacticpositionofthe
nounnapkin inthesentence;thisnounispartofarelativeclausethatunambiguouslymodifiestheNPthe frog(i.e.,the frog thats on the napkin);assuch,theNP
the napkinmustrefertothenapkinunderthefrog,nottheemptynapkin.Similar
time-course data has been reported for adults (e.g., Spivey et al. 2002; Tanenhausetal.1995;Trueswelletal.1999)andreplicatedinotherchildren(Hurewitz,
Brown-Schmidt, Thorpe, Gleitman, & Trueswell 2001; Snedeker & Trueswell
2004),allofwhichsuggeststhatgazedirectionistightlyrelatedtothelinguistic
eventsincomplexsentencesandthatreferenceisbeingcomputedbythechild
andadultlistenerinrealtime.
Usingeyemovementsasadevelopmentalmeasure
3.
Linking assumptions
Itiscrucialtoconsiderthelinking assumptions,orchainofinferences,thatwejust
rapidlyranthroughwhenevaluatingdatalikethoseinFigure5.Ascompellingas
thisprobabilityplotis,howcanweconfidentlygofromeyegazepatternstothe
conclusionthatchildlistenerscomputereferentialhypothesesinreal-time?In
ordertoanswerthisquestion,thereareatleastthreecruciallinkingassumptions
worthevaluatingfurther.
(1) Eye position indicates the childs current attentional state, and attention is
drivenbypropertiesoftheworldandbythegoalsofthechild.
(2) Intasksrequiringthelinkingofspeechtoavisualreferentworld,visualattentioncanbeusedasanindicationofreferentialdecisions.
(3) Referentialdecisionscaninturnbeusedbytheresearchertoinferthechilds
syntacticparsingdecisions,insofarastheseparsingdecisionswerenecessary
todeterminethereferent.
Theremainingsectionsofthischapterunpackeachoftheselinkingassumptions
andexaminethecurrentexperimentalliteratureforvalidationoftheseassumptions. In particular, Section 3.1 (Eye position is a metric of spatial attention in
adults, children and infants)examinesthefirstofthesethreelinkingassumptions.
Section3.2(Eye movements can be used to infer referential and syntactic decisions)
examinesthelattertwoassumptions.Whereverrelevant,thesesectionsalsoexplorehowobserveddevelopmentalpatternsmightinteractwiththeselinkingassumptionsandthereforemodifytheconclusionsthatcanbedrawnwhenusing
thevisualworldmethodindevelopmentalpsycholinguistics.
3.1
Eyepositionisametricofspatialattentioninadults,
childrenandinfants
81
82
JohnC.Trueswell
(Forsourcesandreferencesoneyemovements,seeKowler1995;Liversedge&
Findlay2001;Rayner1998.)
Forthenormallydevelopingnewborn,mostoftheseanatomicalproperties
oftheretinaareinplaceatbirthordeveloprapidlyduringthefirstmonthsoflife.
Basicfundamentaloculomotorabilitiesarealsoinplacequiteearly;saccades,fixations,andeventheabilitytosmoothlypursueaslowlymovingobjectallemerge
quicklyduringthefirstsixmonthsoflifeandareknowntobewellinplaceby
the childs first birthday (for a review, see Colombo 2001). As discussed below
however,quantitativedevelopmentalchangesineyemovementabilitiesdooccur
wellafterthefirstbirthday.
Usingeyemovementsasadevelopmentalmeasure
controlledexogenously,i.e.,byvisualpropertiesoftheworldthatcaptureourattention.Regionsofspacethatarehighlydistinguishedfromotherareas(especially
sudden onsets of motion, a.k.a., motion transients) draw our gaze quite rapidly
andautomatically(e.g.,Franconeri&Simons2003;Jonides&Yantis1988;Yantis&
Jonides1984,1990).3Incontrast,experience-drivenexpectationsandnavigational
plansalsocontributetothevisualselectionofanobjectoraregionofspace,and
thuscontributeendogenouslytoattentionalcontrol. Undermanynaturalviewing
conditions,endogenousfactorsmustoverrideexogenousinfluencessoastoallow
fortheguidanceofattentiontoobjectsthataretask-relevantbutotherwisevisually
lesssalient(e.g.,Guitton,Buchtel,&Douglas1985;Hallett1978).
Thereisstillgreatdebateintheattentionliteratureregardingthedetailsof
howobjectsareselectedforattention(e.g.,invisualsearchtasks).However,many
currentneurocomputationalmodelsofattentionproposeaparallelselectionprocess (e.g., Findlay 2004; Rao, Zelinsky, Hayhoe, & Ballard 2001; Zelinsky, Rao,
Hayhoe,&Ballard1997).Here,theentirevisualfieldatanymomentintimeis
characterizedbymultiplesaliencymaps(colorsaliency,motionsaliency,texture
saliency,etc.).Goal-directedorientingisdrivenbyactiveintegrationofthesesaliencymaps;forinstance,searchingforatoyfrogwouldbehypothesizedtoincludeparallelconsultationofthosespatialregionsdistinguishedbyrelevantcolors,motions,textures,etc.Aviewersmemoryforthespatialpositionofobjects
isalsolikelytoplayarole(e.g.,arecentmemoryforthelocationofatoyfrog).
Aviewersworkingmemoryforthespatialpositionofobjectsappearstobelimitedtoasmallsetofitems(e.g.,Pylyshyn1994).Howeveritisalsobelievedthat
theextenttowhichaviewerholdsthevisualdetailsoftheseobjectsinmemory
dependsgreatlyonthetask.Forinstance,inmanysimplemanualtasks,theperceivermayrelyontheexternalworldasakindofvisualmemorybyassuming
thatthevisualfeaturesofobjectsintheworldremainunchangedovertime(e.g.,
Ballard,Hayhoe,&Pelz1995).
Quiteclearly,eyemovementmeasurementsfromthevisualworldparadigm
rely on the participants accurate implementation of visual selection processes,
forwhichtask-relevantendogenousfactorsareexpectedtooverrideexogenous
factors:Participantsareexpectedtofixateonwhatisrelevantforcarryingoutthe
instruction,notonwhatismostcolorfuloreye-catching. Infact,inanimportant
sense, the linguistic input can be viewed in this approach as a straightforward
characterizationoftheparticipantscurrentspatialgoals(Look at the smiley face,
Pick up the frog, etc.). Giventhis,itseemsurgenttounderstandthedevelopmental
timecourseofendogenousvisualselectionabilities,particularlyinnonlinguistic
3. SeealsoRuzandLupiez(2002),Yantis(1993),andYantisandJonides(1996),fordebates
overwhatexactlycapturesattentionandwhetherattentioncaptureistrulyautomatic.
83
8
JohnC.Trueswell
tasks.Withoutknowledgeofthesefacts,onewillnotbeabletoadequatelyinterpretdevelopmentalvisualworldfindings,particularlyastheyapplytotheoriesof
languageprocessingandlanguagedevelopment.
Asitturnsout,endogenouscontrolofattentionbyinfantsbeginsquiteearly
inlifebutappearstohaveaprotracteddevelopmentalprofile.Forinstance,infantswhoare3.5to6.0monthsofageoroldercanlearntorepositiontheireyes
soastoanticipatetheappearanceofanobjectinpredictablelocations(Haith,Hazan,&Goodman1988;McMurray&Aslin2004).Thesestudies,however,donot
involveproceduresthatrequireinfantstooverrideexogenousfactorsthatmight
influenceattention.Towhatextentcaninfantsandtoddlersdothis?Inthefirst
studyofthiskind,M.Johnson(1995)placed4-month-oldsinamodifiedantisaccadetask:Childrenhadtolearnthatthesuddenonsetofaspatialcueonthe
leftpredictedtheonsetofarewardingvisualstimulusontheright,andviceversa.
Adultsinsuchtasks(e.g.,Hallett1978;Guittonetal.1985)rapidlylearnsimultaneously to inhibit attention shifts to the briefly presented spatial cue (i.e., to
inhibitpro-saccadestothecue)andtogenerateanticipatorylookstothereward
location(i.e.,togenerateanti-saccades,despitethesuddenonsetontheotherside
of the screen). M. Johnson (1995) found that 4-month-olds did learn over the
courseoftheexperimenttoinhibitpro-saccades.Thatis,theylearnedtonotlook
attheflashedspatialcuepresumablybecauseitbecameclearthatthiseventwas
perfectlycorrelatedwiththepresenceofarewardontheothersideofthescreen.
However,thesesameinfantswereneverabletogenerateanti-saccades.Thatis,
theydidnotlearn to movetheir eyesto the rewardlocationpriortoits visual
onset.Importantly,aswejustmentioned,infantsinthisagerangecananticipate
thelocationofarewardobjectwhena(non-anti-saccade)spatialcueisprovided
(i.e.,theHaithetal.1998,result).Takentogetherthen,itappearsthatthebest
thata4-month-oldcandoiscounter-actbutnotcompletelyoverrideexogenous
contributionstoattention.
Recently, Scerif, Karmiloff-Smith, Campos, Elsabbagh, Driver and Cornish
(2005) examined the development of anti-saccade abilities over a much larger
age range (8 to 40 months). Scerif et al. found that the proportion of pro-saccadessteadilydecreaseswithinthisagerange(from100%downtoapproximately
20%)whereastheproportionofanti-saccadessteadilyincreases(from0%toapproximately40%).Thatis,itappearsthattheabilitytosimultaneouslycounter-act
exogenous factors while promoting endogenous factors has a fairly protracted
developmentalprofile;childrenundertheageofthreeyearsaremoresusceptible
toexogenousfactorsthanolderchildren.
Usingeyemovementsasadevelopmentalmeasure
Eyemovementscanbeusedtoinferreferentialandsyntacticdecisions
Wenowturntolinkingassumptions2and3,whichcanbecombinedandrestated
morepreciselyasfollows.
8
86 JohnC.Trueswell
Ifataskrequireslinkingspeechtoavisualreferentworld,eyemovementexperimentscanbedesignedtouncoverthelistenersongoingreferentialdecisions
and,byinference,theirongoingsyntacticparsingdecisions.Notethatthisdoes
notmeanthatatalltimeswherethechildislookingiswhatthechildisconsideringasthereferent.Eyemovementsinvisualselectiontasksreflectgoal-directed
behaviorandassuch,studiesinwhichreferenceisnecessarytoachievesomegoal
(suchasactingonspokeninstructions)permitsaresearchertoinferreferential
andsyntacticdecisions.
Usingeyemovementsasadevelopmentalmeasure
tar get
competitor
Figure 6. Childrens(5-year-olds)proportionoflooksovertimetopotentialreferent
objectsinresponsetoPut the frog on the napkin into the box.FromTrueswell,Sekerina,
HillandLogrip(1999).Copyright1999,ElsevierPress.
ity;linkingaPPtoaverbisclaimedtobecomputationallysimplerthanlinkingit
toanNP.Foreitheroftheseparsingreasons,ourlinkingassumptionsleadusto
predictthatchildren(iftheyparseinoneofthesemanners)shouldunderthese
conditionsstartconsideringtheemptynapkin(theincorrectgoal)asapossible
referentofnapkin.Thisisbecausethemostplausiblegoalforputtingafroginthis
caseistheemptynapkin(notthenapkinthatalreadyhasafrogonit).Ifasimple
conjunctionheuristicisatwork,theresultshouldbesimilartotheunambiguous
sentence(i.e.,weshouldagainseeincreasedlookstotargettheonlyquadrant
thathasbothafrogandanapkin).Figure6presentstheprobabilityplotforthis
temporarilyambiguoussentence.
Consistentwiththeparsing/referencelinkingassumptions,lookstotheincorrectgoaldoinfactincreasesoonafterhearingnapkin inthiscondition,apattern
thatisreliablydifferentfromthatinFigure5whenthephrasewasunambiguously
amodifier.Noticealsothatasaconsequenceofinterpretingon the napkinasa
goal rather than a modifier phrase, children are having trouble distinguishing
betweenthetwofrogs(theyarelookingequallyatboththetargetandcompetitorfrogsforanextendedperiodoftime;seeFigure5).Thisadditionalpatternis
alsoexpectedundertheparsingandreferenceassumptions;ifon the napkin isnt
parsedasamodifier(butratherasagoal),thenthisphraseisnolongerinformativefordistinguishingbetweenthetwofrogs.
SincethepublicationofTrueswelletal.(1999)numerousotherstudieshave
been conducted that also use childrens eye movement patterns during spoken
87
88
JohnC.Trueswell
Usingeyemovementsasadevelopmentalmeasure
representationsthatchildrenareformingduringdevelopment.Forinstance,the
presence of structural priming patterns that are independent of the particular
lexicalitemsusedinanutterancecaninprinciplebequiteinformativeforissues
pertainingtothelevelsofabstractionthatchildrenareabletooperateoverwhen
acquiringandprocessingalanguage.
Itisimportanttonotethatmostoftheserecentstudiesusedtheparticular
methoddescribedabove,inwhichparticipantsactuponobjectsbasedonspoken
instructions.However,somestudies(e.g.,Arnoldetal.inpress)askedchildrento
decideifaspokensentenceaccuratelydescribedavisuallyco-presentpicture.Here
thegoal-directedbehavioralsorequireslinkingspeechtothevisualreferentworld,
andasaresultcanprovideinformativepatternsrelatedtoreferentresolution.It
hasbeenourexperiencethatsimplyaskingchildrentopassivelylookatpictures
andlistentoastoryleadsthemtobecomemoreeasilydistractedandlesslikelyto
inspectthevisualscenes.Thisobservationisconsistentwiththelinkingassumptionsdiscussedabove:goal-directedbehaviorthatrequiresreferentiallinkageto
theworldismuchmorelikelytoyieldinterpretableeyemovementpatterns.
89
90 JohnC.Trueswell
what is known about the brain systems that support cognitive control of this
sort.Frontallobesystems(particularly,leftandrightprefrontalcortex)havebeen
implicatedinadultstosupportarangeofcognitivecontrolabilities(e.g.,Bunge
etal.2002;Thompson-Schilletal.1998).Theseverysamebrainregionsalsoshow
some of the most delayed neuroanatomical maturational profiles; for instance,
myelinationofneuronswithinthesefrontalsystemsisnotcompleteuntilquite
lateinnormalhumandevelopment,i.e.,aslateas510yearsofage,ifnotlater
(seeDiamond2002,foradiscussion).
Interestingly,thiscognitiveimpulsivitywasalsoobservedforthe5-year-olds
intheTrueswelletal.(1999)Put-study.Consideragaintheeyemovementpatterns
inFigure6,fortheambiguoussentencePut the frog on the napkin into the box.
Children never consistently converged on the intended target frog (looking just
asoftenatthecompetitor),suggestingthattheyneverrealizedthaton the napkin
couldbeaModifieroftheNPthe frog.Infact,childrensultimateactionssuggested
theyhadnotfullyrecoveredfromtheirgarden-path:Childrenwereatchanceselectingbetweenthetwofrogs,andfrequently(60%ofthetime)movedtheselected
frogtotheincorrectgoalplacingthefrogontheemptynapkin,orplacingthe
frogontheemptynapkinandthenintothebox.Thisdifficultywasclearlyrelated
toambiguity,sincethesesamechildrenmadeessentiallynoerrorsinresponseto
unambiguousmaterials:Put the frog thats on the napkin into the box.
Researchers who are not predisposed to thinking of child language use as
an emerging dynamic process might interpret such child failures as indicating
anagerangeatwhichchildrenlacksomeknowledge;perhapstheyhavenotyet
acquiredtherestrictive(NPmodifying)PPstructure.Twofindingsruleoutthis
possibilityandpointtotheaccountofchildrenfailingtorevise.First,Hurewitzet
al.(2001)showedthatchildrenwithintherelevantagerangecanproducerestrictiveNPmodifiers(e.g.,the frog on the napkin, the one on the napkin)whenasked
abouttwovisuallyco-presentreferents.Thesesamechildrenneverthelessgoon
tomisinterpretput-sentenceslike(1)inexactlythesamewayasthatfoundby
Trueswelletal.(1999).Thissuggeststhatinthetraditionalsense,thesechildren
haveacquiredthisstructure.
Second,similarparsingfailuresincomprehensionhaverecentlybeenseenina
specialpopulationofadultsspecifically,anindividualwithafocallesiontofrontalloberegionsknowntoberesponsibleforcognitivecontrol(Novick2005;see
alsoNovick,Trueswell,&Thompson-Schill2005,foradiscussion).Inthisstudy,
thisindividual(PatientNJ)wasgivenabatteryofneurocognitivetasksdesignedto
testfrontallobefunctioning,butalsothePut-taskdescribedabove.NJshowedthe
characteristicdeficitsincognitivecontrol,i.e.,notbeingabletoinhibitpre-potent
responsestovariousstimuli.Interestingly,NJshowedaparsingpatternquitesimilarto5-year-olds;hefailedtoreviseonambiguoustrials,moving(forinstance)
Usingeyemovementsasadevelopmentalmeasure
thefrogontotheemptynapkin(theincorrectgoal)andthenintothebox(thecorrectgoal).Like5-year-olds,NJmadenoerrorsonunambiguousversionsofthese
sentences(Put the frog thats on the napkin into the box).Alsoquiteinterestingly,
NJhasbeenfoundtohavedifficultyresolvinghighlybiasedambiguouswordsas
well (Novick, Bedny, Kan, Trueswell, &Thompson-Schill in preparation). Thus,
NJ, who has deficits in cognitive control, shows precisely the sorts of linguistic
processing deficits one might expect if cognitive control plays a role in parsing
andinterpretation,i.e.,aninabilitytorecoverthesubordinatemeaningofahighly
biasedambiguousstructureorahighlybiasedambiguousword.
Thissurprisingassociationbetweenspecificfrontallobedeficitsandgardenpathrecoverybodeswellfordynamicprocessingaccountsofchildlanguagedevelopment.Giventhatfrontal-lobeneuralsystemsaresomeofthelastregionsof
the brain to fully mature anatomically, it is completely plausible that childrens
dynamicprocessingsystemsarehinderedbydelayeddevelopmentofsystemsresponsible for engaging cognitive control, specifically the ability to recharacterizeotherwisesupportedinterpretationsoflinguisticinput.Khanna,Bolandand
Cortese(2006)explorethisandrelatedhypothesesastheypertaintochildrens
developingabilitytoresolvewordsenseambiguity.
Interestingly,childrensinabilitytousejoint-attentioncontextualconstraints
inreferentialcommunicationtasksmayberelatedtotheseissues.Forexample,
Epleyetal.(2004)foundthat5-year-oldsactegocentricallywhenselectingareferent,sometimespickingasareferentanobjectthatwasvisibleonlytothechild
andnottotheadultspeaker.However,pickingtheintendedcommon-ground
object(i.e.,theobjectthatwasvisibletoboththespeakerandthelistener)always
requiredthechildtoselectthesubordinatemeaning(orlessprototypicalmeaning)ofthereferentialexpression.Whenthisiscontrolledfor(aswasthecasein
Nadig&Sedivy2002),childrensuseofjointeyegazereturns.Takentogether,the
datasuggestthatchildrenweighmultiplelinguisticandnonlinguisticconstraints
whenmakingreferentialdecisions.However,ifthismultipleconstraintprocess
requires overriding potent linguistic tendencies, cognitive control is necessary
anddifficultymayensue.
.
Acommonwayofthinkingaboutcognitivedevelopmentisasthegradualacquisitionofknowledgeabouttheworld.Alternatively,wecanthinkofdevelopment
astheacquisitionofdynamicskills:Welearnhow to interactwiththeworld;we
learnhow to produce and comprehendalanguage,etc. Thischapterhasreviewed
somekeyexperimentalfindingswithinthedevelopmentalpsycholinguisticslit-
91
92
JohnC.Trueswell
eraturethatencouragesthisdynamic,functionalwayofthinkingaboutlanguage
learning.Studiesthathaverecordedtheeyemovementsofyoungchildrenasthey
hearspokeninstructionshavetodatebeenquitesuccessfulatuncoveringongoingreferentialandsyntacticprocessesastheyoccuroverthecourseofhearing
eachsentence.
Anevaluationofthelinkingassumptionsnecessarytointerpretfindingsfrom
thismethodologysuggeststhattheseassumptionsarevalid.However,cautionand
careisnecessarywhenperformingsuchresearchbecausedevelopmentalchanges
inattentionalcontrolandcognitivecontrolcaninprincipleinteractwithobservationsfromthismethod.Itisimportanttonotethatthisconcernistrueofany
experimental method when applied to the study of development; the onus falls
onthedevelopmentalresearchertounderstandandevenseekouttheseinteractionsintheirexperimentalfindings.Otherwise,developmentalobservationscan
beeasilymisattributedtotheresearcherstheoreticaltopicofinterest.Forinstance,
thepresentevaluationofthevisualworldmethodologysuggeststhatcaremust
betakeninunderstandinghowgeneralattentionalcontrolandcognitivecontrol
changewithage.Developmentalshiftswereidentifiedintherelativecontribution
ofexogenousandendogenousfactorswhenitcomestothedirectionofspatialattention,particularlyinyoungerchildren(3yearsofageandyounger).Inaddition,
developmentalshiftsexistingeneralcognitivecontrolabilitieswellintoachilds
10thyearoflife.Childrenshowadomain-generaldifficultyoverridinginitialcharacterizationsofstimuli.Thissamedifficultyisalsomanifestedinlanguageprocessing:Childrensometimeshavedifficultyoverridingtheirinitialcharacterizationof
asentenceandhencesometimesfailtorecoverfromgarden-paths.
No doubt as our understanding of visual attention in the infant and child
grows,significantadvanceswillalsosimultaneouslyoccurinourunderstanding
oflanguagelearningandlanguageprocessing,particularlyintherelativelynaturalsettingofdiscussingvisuallyco-presentreferents.Thevisualworldmethod
servesasanimportantnewwayofevaluatingthedynamicsoflanguageusein
theyoungchild.Significanttheoreticaladvanceshavebeenmadethroughtheapplicationofthisandotherreal-timemeasuresoflanguageuse.Indeed,thevisual
worldmethodinparticularhasshownitselftobeextremelyvaluableforunderstandinglanguagerepresentationanduseasitdevelopsfrominfancyintoadulthood.Themethodiswellsuitedforexperimentalinvestigationsatmultiplelevels
oflinguisticrepresentation(phonological,lexical,syntacticandreferential)and
offersimportantinsightintothefine-graintemporaldynamicsofthesesystems
astheygrowandmature.
Usingeyemovementsasadevelopmentalmeasure
Acknowledgments
ThisworkwaspartiallyfundedbyagrantfromtheNationalInstitutesofHealth
(1-R01-HD37507).
References
Allopenna,P.D.,Magnuson,J.S.,&Tanenhaus,M.K.(1998).Trackingthetimecourseofspokenwordrecognition:Evidenceforcontinuousmappingmodels.Journal of Memory and
Language, 38,419439.
Arnold, J. E., Brown-Schmidt, S., & Trueswell, J. C. (in press). Pronoun comprehension in
youngchildren.Language and Cognitive Processes.
Aslin, R. N. & McMurray, B. (2004). Automated corneal-reflection eye-tracking in infancy:
Methodologicaldevelopmentsandapplicationstocognition.Infancy,6,155163.
Ballard,D.,Hayhoe,M.,&Pelz,J.(1995).Memoryrepresentationsinnaturaltasks.Cognitive
Neuroscience,7,6680.
Bialystok,E.(2001).Bilingualism in development: Language, literacy, and cognition.NewYork
NY:CUP.
Bialystok,E.&Shapero,D.(2005).Ambiguousbenefits:Theeffectofbilingualismonreversing
ambiguousfigures.Developmental Science,8,595604.
Bunge,S.A.,Dudukovic,N.M.,Thomason,M.E.,Vaidya,C.J.,&Gabrieli,J.D.E.(2002).
Developmentoffrontallobecontributionstocognitivecontrolinchildren:Evidencefrom
fMRI.Neuron, 33,301311.
Choi,Y.&Trueswell,J.C.(inpreparation).Puttingfirstthingslast:TheKoreankindergartenpatheffect.Manuscriptinpreparation.
Cohen,M.E.&Ross,L.E.(1977).Saccadelatencyinchildrenandadults:Effectsofwarning
intervalandtargeteccentricity.Journal of Experimental Child Psychology, 23,539549.
Cohen,M.E.&Ross,L.E.(1978).Latencyandaccuracycharacteristicsofsaccadesandcorrectivesaccadesinchildrenandadults.Journal of Experimental Child Psychology, 26,517
525.
Colombo,J.(2001).Thedevelopmentofvisualattentionininfancy.Annual Review of Psychology,52,337367.
Cooper,R.M.(1974).Thecontrolofeyefixationbythemeaningofspokenlanguage.Cognitive
Psychology,6,84107.
Davidson,M.C.,Amso,D.,Anderson,L.C.,&Diamond,A.(2006).Developmentofcognitive control and executive functions from 413 years: Evidence from manipulations of
memory,inhibition,andtaskswitching.Neuropsychologia,44,20372078.
Diamond,A.(2002).Normaldevelopmentofprefrontalcortexfrombirthtoyoungadulthood:
Cognitivefunctions,anatomy,andbiochemistry.InD.Stuss&R.Knight(Eds.),Principles
of frontal lobe functioning(pp.466503).NewYorkNY:OUP.
Epley,N.,Morewedge,C.K.,&Keysar,B.(2004).Perspectivetakinginchildrenandadults:
Equivalentegocentrismbutdifferentialcorrection.Journal of Experimental Social Psychology, 40,760768.
93
9 JohnC.Trueswell
Fernald,A.&Hurtado,N.(2006).Namesinframes:Infantsinterpretwordsinsentenceframes
fasterthanwordsinisolation.Developmental Science, 9,F33F40.
Findlay,J.M.(2004).Eyescanningandvisualsearch.InJ.M.Henderson&F.Ferreira(Eds.),
The interface of language, vision, and action: Eye movements and the visual world(pp.135
160).NewYorkNY:PsychologyPress.
Franconeri,S.L.&Simons,D.J.(2003).Movingandloomingstimulicaptureattention.Perception & Psychophysics, 65,9991010.
Fukushima,J.,Hatta,T.,&Fukushima,K.(2000).Developmentofvoluntarycontrolofsaccadiceyemovements.I:Age-relatedchangesinnormalchildren.Brain Development, 22,
173180.
Golinkoff,R.M.,Hirsh-Pasek,K.,Cauley,K.M.,&Gordon,L.(1987).Theeyeshaveit:Lexical
andsyntacticcomprehensioninanewparadigm.Journal of Child Language,14, 2345.
Guitton,D.,Buchtel,H.A.,&Douglas,R.M.(1985).Frontallobelesionsinmancausedifficultiesinsuppressingreflexiveglancesandingeneratinggoal-directedsaccades.Experimental Brain Research,58,455472.
Haith,M.M.,Hazan,C.,&Goodman,G.S.(1988).Expectationandanticipationofdynamic
visualeventsby3.5-month-oldbabies.Child Development,59,467479.
Hallett,P.E.(1978).Primaryandsecondarysaccadestogoalsdefinedbyinstructions.Vision
Research,18,12791296.
Henderson,J.M.&Ferreira,F.(Eds.).(2004).The interface of language, vision, and action: Eye
movements and the visual world.NewYorkNY:PsychologyPress.
Huang,Y.&Snedeker,J.(2007).Frommeaningtoinference:Evidenceforthedistinctionbetween lexical semantics and scalar implicature in online processing and development.
Manuscriptsubmittedforpublication.
Hurewitz,F.,Brown-Schmidt,S.,Thorpe,K.,Gleitman,L.R.,&Trueswell,J.C.(2001).One
frog,twofrog,redfrog,bluefrog:Factorsaffectingchildrenssyntacticchoicesinproductionandcomprehension.Journal of Psycholinguistic Research, 29,597626.
Johnson,M.H.(1995).Theinhibitionofautomaticsaccadesinearlyinfancy.Developmental
Psychobiology, 28,281291.
Johnson,S.P.,Slemmer,J.A.,&Amso,D.(2004).Whereinfantslookdetermineshowthey
see:Eyemovementsandobjectperceptionperformancein3-month-olds.Infancy,6,185
201.
Jonides,J.&Yantis,S.(1988).Uniquenessofabruptonsetincapturingattention.Perception and
Psychophysics,43,346354.
Khanna,M.M.Boland,J.E.,&Cortese,M.J.(2006).Developmentofsentencecontextuse:
Whenandhowdochildrenknowthattagisalabelandnotagame?Paperpresentedat
theworkshoponOn-Line Methods in Childrens Language Processing. NewYorkNY.March
2122,2006.
Kidd,E.&Bavin,E.L.(2005).Lexicalandreferentialcuestosentenceinterpretation:Aninvestigationofchildrensinterpretationsofambiguoussentences.Journal of Child Language,
32,855876.
Kowler,E.(1995).Eyemovements.InS.M.Kosslyn&D.N.Osherson(Eds.),An Invitation to
Cognitive Science, Vol.2: Visual Cognition (2ndedn., pp.215265).CambridgeMA:The
MITPress.
Liversedge,S.&Findlay,J.(2001).Saccadiceyemovementsandcognition.Trends in Cognitive
Science, 4, 614.
Usingeyemovementsasadevelopmentalmeasure
Matin,E.,Shao,K.C.,&Boff,K.R.(1993)Saccadicoverhead:Informationprocessingtime
withandwithoutsaccades.Perception & Psychophysics,53,372380.
McMurray,B.&Aslin,R.N.(2004).Anticipatoryeyemovementsrevealinfantsauditoryand
visualcategories.Infancy,6,203229.
Nadig,A.&Sedivy,J.C.(2002).Evidenceofperspective-takingconstraintsinchildrenson
linereferenceresolution.Psychological Science,13,329336
Nappa, R., Trueswell, J. C., & Gleitman, L. R. (2006). Effects of attention and intention on
sentenceparsing.Paperpresentedatthe31st Annual Boston University Conference on Language Development,Boston,MA.
Novick,J.M.(2005).CognitivecontrolandtheroleofBrocasareainsentenceprocessing.PhD
dissertation,UniversityofPennsylvania.
Novick,J.M.,Trueswell,J.C.,&Thompson-Schill,S.L.(2005).Cognitivecontrolandparsing:
Re-examining the role of Brocas area in sentence comprehension. Journal of Cognitive,
Affective, and Behavioral Neuroscience, 5,263281.
Novick,J.M.,Bedny,M.,Kan,I.,Trueswell,J.C.,&Thompson-Schill,S.L.(inpreparation).
TheroleofLIFGincognitivecontrol,languageproduction,andlanguagecomprehension:
Asingle-casestudy.Manuscriptinpreparation.
Posner,M.I.(1980).Orientingofattention.Quarterly Journal of Experimental Psychology,32,
325.
Pylyshyn,Z.W.(1994).Someprimitivemechanismsofspatialattention.Cognition,50,363
384.
Rao,R.P.N.,Zelinsky,G.J.,Hayhoe,M.M.,&Ballard,D.H.(2001).Eyemovementsiniconic
visualsearch.Vision Research,42,14471463.
Rayner,K.(1998).Eyemovementsinreadingandinformationprocessing:20yearsofresearch.
Psychological Bulletin, 124,372422.
Ross,S.M.&Ross,L.E.(1983).Theeffectsofonsetandoffsetwarningandpost-targetstimuli
onthesaccadelatencyofchildrenandadults.Journal of Experimental Child Psychology,
36,340355.
Ruz,M.&Lupiez,J.(2002).Areviewofattentionalcapture:Onitsautomaticityandsensitivitytoendogenouscontrol.Psicolgica,23,283309.
Salman,M.S.,Sharpe,J.A.,Eizenman,M.,Lillakas,L.,Westall,C.,&To,T.etal.(2006).Saccadesinchildren.Vision Research, 46(89),14321439.
Scerif,G.,Karmiloff-Smith,A.,Campos,R.,Elsabbagh,M.,Driver,J.,&Cornish,K.(2005).To
lookornottolook?Typicalandatypicaldevelopmentofoculomotorcontrol.Journal of
Cognitive Neuroscience,17,591604.
Sekerina,I.A.&Brooks,P.J.(2007).Eyemovementsduringspoken-wordrecognitioninRussianchildren.Journal of Experimental Child Psychology, 98, 2045.
Sekerina,I.A.,Stromswold,K.,&Hestvik,A.(2004).Howdoadultsandchildrenprocessreferentiallyambiguouspronouns?Journal of Child Language,31,123152.
Snedeker,J.&Trueswell,J.C.(2004).Thedevelopingconstraintsonparsingdecisions:Therole
oflexical-biasesandreferentialscenesinchildandadultsentenceprocessing.Cognitive
Psychology,49,238299.
Song,H.,&Fisher,C.(2005).Whosshe?Discoursestructureinfluencespreschoolerspronouninterpretation.Journal of Memory & Language,52,2957.
Spivey,M.J.,Tanenhaus,M.K.,Eberhard,K.M.,&Sedivy,J.C.(2002).Eyemovementsand
spokenlanguagecomprehension:Effectsofvisualcontextonsyntacticambiguityresolution.Cognitive Psychology,45,447481.
9
96 JohnC.Trueswell
Swingley,D.,Pinto,J.P.,&Fernald,A.(1999).Continuousprocessinginwordrecognitionat24
months.Cognition,71,73108.
Tanenhaus,M.K.,Spivey-Knowlton,M.J.,Eberhard,K.M.,&Sedivy,J.C.(1995).Integrationofvisualandlinguisticinformationinspokenlanguagecomprehension.Science, 268,
16321634.
Tanenhaus,M.K.&Trueswell,J.C.(2005).Eyemovementsastoolforbridgingthelanguageas-productandlanguage-as-actiontraditions.InJ.C.Trueswell&M.K.Tanenhaus(Eds.),
Approaches to studying world-situated language use: Bridging the language-as-product and
language-as-action traditions.CambridgeMA:TheMITPress.
Thompson-Schill,S.L.,Swick,D.,Farah,M.J.,DEsposito,M.,Kan,I.P.,&Knight,R.T.(1998).
Verbgenerationinpatientswithfocalfrontallesions:Aneuropsychologicaltestofneuroimagingfindings.Proceedings of the National Academy of Science,95,1585515860.
Thothathiri,M.&Snedeker,J.(inpress).Syntacticprimingduringlanguagecomprehensionin
three-andfour-year-oldchildren.Journal of Memory and Language.
Trueswell,J.C.,Sekerina,I.,Hill,N.M.,&Logrip,M.L.(1999).Thekindergarten-patheffect:
Studyingon-linesentenceprocessinginyoungchildren.Cognition,73,89134.
Trueswell,J.C.&Tanenhaus,M.K.(Eds.).(2005).Approaches to studying world-situated language use: Bridging the language-as-product and language-as-action traditions.Cambridge
MA:TheMITPress.
Yang,Q.,Bucci,M.P.,&Kapoula,Z.(2002).Thelatencyofsaccades,vergence,andcombined
eyemovementsinchildrenandadults.Investigative Ophthalmology & Visual Science, 43,
29392949.
Yantis,S.(1993).Stimulus-drivenattentionalcaptureandattentionalcontrolsettings.Journal of
Experimental Psychology: Human Perception and Performance, 19,676681.
Yantis,S.&Jonides,J.(1984).Abruptvisualonsetsandselectiveattention:Evidencefromvisualsearch.Journal of Experimental Psychology: Human Perception and Performance,10,
601621.
Yantis,S.&Jonides,J.(1990).Abruptvisualonsetsandselectiveattention:Voluntaryverses
automaticallocation.Journal of Experimental Psychology: Human Perception and Performance,16,121134.
Yantis,S.&Jonides,J.(1996).Attentionalcapturebyabruptonsets:Newperceptualobjectsor
visualmasking?Journal of Experimental Psychology: Human Perception and Performance,
22,15051513.
Zelinsky,G.J.,Rao,R.P.N.,Hayhoe,M.M.,&Ballard,D.H.(1997).Eyemovementsrevealthe
spatiotemporaldynamicsofvisualsearch.Psychological Science,8,448453.
chapter4
Thelooking-while-listeningmethodologyusesreal-timemeasuresofthetime
courseofyoungchildrensgazepatternsinresponsetospeech.Thisprocedure
islowintaskdemandsanddoesnotrequireautomatedeyetrackingtechnology,similartopreferential-lookingprocedures.However,thelooking-whilelisteningmethodologydifferscriticallyfrompreferential-lookingprocedures
inthemethodsusedfordatareductionandanalysis,yieldinghigh-resolution
measuresofspeechprocessingfrommomenttomoment,ratherthanrelyingon
summarymeasuresoflookingpreference.Becausechildrensgazepatternsare
time-lockedtospeechandcodedframe-by-frame,each5-minexperimentresponselatenciescanbecodedwithmillisecondprecisiononmultipletrialsover
multipleitems,basedondatafromthousandsofframesineachexperiment.
Themeticulousproceduresrequiredinthecollection,reduction,andmultiple
levelsofanalysisofsuchdetaileddataaredemanding,butwellworththeeffort,
revealingadynamicandnuancedpictureofyoungchildrensdevelopingskillin
findingmeaninginspokenlanguage.
1.
Introduction
Developmental studies of comprehension in very young children have relied
traditionallyonoff-linemeasures,responsesmadeaftertheoffsetofthespeech
stimulusthatdonottapintothereal-timepropertiesofspokenlanguage.Studies
of incremental processing by adults rely on on-line measures that monitor the
timecourseofthelistenersresponseinrelationtokeypointsinthespeechsignal.
Becausecomprehensionoccursrapidlyandautomaticallywithouttimeforreflection,itisrevealingtostudythelistenersinterpretationduringspeechprocessing
98 AnneFernaldetal.
andnotjustafterward.Classicon-linebehavioraltechniquesusedtoinvestigate
incrementalspeechprocessingbyadultsincludephonememonitoring(Cutler&
Foss1977),gating(Grosjean1985),andcross-modalpriming(Marslen-Wilson
&Zwitserlood1989),amongothers.Althoughsomeofthesehavebeenadapted
forusewithschool-agedchildren(Cutler&Swinney1987;Walley1993;Clahsen
thisvolume),thetaskdemandsareoftenproblematicforyoungerchildren.Researchwithadultsusingautomatedeyetrackingtechniqueshasbeenextremely
productiveinrecentyears,providingsensitiveon-linemeasureshighinecological validity (Tanenhaus, Magnuson, Dahan & Chambers 2000). This technique
alsobeenusedwith45-year-oldchildren(Trueswell,Sekerina,Hill,&Logrip
1999;Snedeker&Trueswell2004).
Here we describe a simpler but equally powerful experimental method for
monitoringthetimecourseofcomprehensionbyinfantsandveryyoungchildren,
which we call the looking-while-listening procedure. In this procedure, children
lookatpairsofpictureswhilelisteningtospeechnamingoneofthepictures.Their
gazepatternsarevideotapedandeyemovementsaremeasuredwithhighprecision
inrelationtorelevantpointsinthespeechsignal.Usingthelooking-while-listening
paradigm,wehaveshownthatspeedandefficiencyininfants'on-lineresponses
tofamiliarwordsincontinuousspeechincreasedramaticallyoverthesecondyear
forbothEnglish-andSpanish-learningchildren(Fernald,Pinto,Swingley,Weinberg,&McRoberts1998;Hurtado,Marchman,&Fernald2007),that24-montholdsareabletoprocessphoneticinformationincrementally(Swingley,Pinto,&
Fernald 1999; Fernald, Swingley, & Pinto 2001), and that individual differences
inchildrensspeechprocessingefficiencyarerelatedtotheirleveloflexicaland
grammaticaldevelopment(Fernald,Perfors,&Marchman2006).Theseandother
recentfindingsvalidatethelooking-while-listeningparadigmasapowerfulnew
methodforexploringhowveryyoungchildrentakeadvantageofperceptualand
linguisticfeaturesofthespeechtheyhearinrelationtoinformationinthevisual
world,astheylearntofindmeaninginspokenlanguage.
1.1
Researchontheearlydevelopmentofreceptivelanguageskills
Longbeforetheycanspeaktheirfirstwords,infantsbegintorevealtheiremergingknowledgeoflanguagebyrespondingmeaningfullytothespeechtheyhear.
Yetbecausecomprehensionisamentaleventthatcanonlybeinferredindirectly
fromachildsbehaviorinaparticularcontext,theearlydevelopmentofreceptive
languagecompetencehasbeenlessaccessibletoscientificinquirythandevelopmentalgainsinspeechproduction.However,researchoninfantcognitionover
thepastthreedecadeshasledtovaluableexperimentalmethodsforreadingthe
Lookingwhilelistening
mindsofveryyounglanguagelearners,techniquesthathavemadeitpossible
toexplorethedevelopmentaloriginsofunderstandingingreaterdepthandwith
greater precision. Much of this research has examined how infants become attunedtosoundpatternsintheambientlanguageoverthefirstyearoflife(e.g.,
Werker&Tees1984;Kuhl2004),andattendtospeechpatternsrelevanttolanguagestructure(e.g.,Jusczyk1997;Saffran2002).Thesestudiesshowthatover
thefirstyearinfantsbecomeskilledlisteners,abletomakedistributionalanalyses
ofphoneticfeaturesofspokenlanguage,andthattheyformsomekindofacoustic-phoneticrepresentationforfrequentlyheardsoundpatterns(e.g.,Halle&de
Boysson-Bardies 1994). Such accomplishments are often cited as evidence for
earlywordrecognition.Butsincethisselectiveresponsetofamiliarwordscan
occurwithoutanyassociationbetweenparticularsoundpatternsandmeanings,
itisperhapsappropriatelyviewedasevidenceforpattern-detectionabilitiesprerequisiteforrecognizingwordsincontinuousspeech.Muchlessisknownabout
howlanguage-specificprocessingstrategiescontinuetodevelopbeyondthefirst
year,aschildrenintheirsecondandthirdyearsbegintoappreciateregularitiesat
higherlevelsoflinguisticorganization,usingtheiremerginglexicalandmorphosyntacticknowledgetomakesenseofspokenwordsandsequencesofwordsin
combination(seeFernald&Frankinpress;Naigles2002).
1.2
Observationalandoff-lineexperimentalmeasures
ofchildrensskillincomprehension
Betweentheagesof10and14monthschildrentypicallybegintoshowsignsof
associatingsoundpatternswithmeanings,speakingafewwordsandappearingto
understandmanymore.Bytheendofthesecondyear,theyrevealprogressinunderstandingthroughincreasinglydifferentiatedverbalandbehavioralresponses
tospeech.Butgrowthinreceptivelanguagecompetenceishardertoobservethan
growthinproductiveabilities,becausetheprocessesinvolvedincomprehension
areonlypartiallyandinconsistentlyapparentthroughthechildsspontaneousbehavior.Scientificstudiesofearlycomprehensionhavemadeuseofquitedifferent
methodologiesthatfallintofourmaincategories(seeFernald2002):Diary studiesprovidedthefirstsystematicobservationaldataonearlycomprehensionabilities,describinghowyoungchildrenappeartointerpretspeechintheireveryday
activitiesandinteractions(e.g.,Lewis1936;Bloom1973).Studies of vocabulary
growth use parental-report checklists to track changes in the estimated size of
thechildsreceptivelexicon;suchchangescorrelateininterestingwayswithlater
grammaticaldevelopment(e.g.,Fenson,Marchman,Thal,Dale,Reznick,&Bates
2007).Naturalistic experiments on comprehensionusebehavioralresponsestotest
99
100 AnneFernaldetal.
infants ability to identify familiar words (e.g., Benedict 1979) and understand
words in combination (e.g., Shipley, Smith, & Gleitman 1969). Experiments on
word learningconstitutethelargestareaofresearchrelatedtoearlyunderstanding:oneapproachfocusesonhowcognitivebiasesguideinferencesaboutword
meanings(e.g.,Markman1989);othersexplorehowchildrenuselinguistic(e.g.,
Katz,Baker,&McNamara1974)andpragmaticknowledge(e.g.,Tomasello2000)
toguidetheseinferences.
Alloftheseapproachesrelyonoff-linemeasuresofcomprehension,i.e.,assessmentsofunderstandingbasedonobservingthechildsbehaviorafterhearing
aparticularlinguisticstimulus.Inthecaseofdiaryobservationsandparentalreportchecklists,thejudgmentthatachilddoesordoesnotunderstandaword
suchasdogorcupismadeinformallybyadultswhointeractregularlywiththe
child in many different contexts. In the case of off-line measures used experimentally,thesejudgmentsarebasedonthechildsbehaviorinamorecontrolled
situation,withaclearlydefinedresponsemeasuresuchaschoosinganobjector
pointingtoapicturegiventwoormorealternatives.Whatthesemeasureshave
incommonisthattheyallarebasedonchildrensresponsestoaspokenwordor
sentenceafteritiscompleteratherthanasitisheardandprocessed.Whilesuch
off-lineproceduresenableresearcherstoassesswhetherornotachildresponds
systematically in a way that indicates understanding, they do not tap into the
real-timepropertiesofspokenlanguageandthusreveallessaboutthechildsdevelopingefficiencyinidentifyingandinterpretingfamiliarwordsincontinuous
speech.
1.3
Thedevelopmentofpreferential-lookingmeasures
forassessingcomprehension
In1963RobertFantzpublishedthefirststudyusingapreferential-lookingmethod with infants, showing that even newborns looked selectively at some visual
stimulioverothers.AlthoughHaith(1980)laterquestionedwhetherpreference
was the appropriate way to characterize infants selective looking behavior, the
findingsthatemergedfromdozensofpreferential-lookingstudiesinthisperiod
suggestedthatcertainearlyvisualbiasesappearedtobeindependentofprevious
experiencewithparticularstimuli.Adaptingthepreferential-lookingprocedure
toinvestigatecross-modalperceptionininfants,Spelke(1976)presentedinfants
with two visual stimuli, only one of which matched a simultaneously presented auditory stimulus, and found that infants looked significantly longer to the
matchingthantothenon-matchingvisualstimulus.Thisauditory-visualmatch-
Lookingwhilelistening 101
ingprocedurewaslatermodifiedindifferentwaysbyinvestigatorsinterestedin
thedevelopmentoflanguagecomprehensionintheearlyyearsoflife.
The first experimental procedures for testing infants knowledge of object
wordsinacontrolledsettingwerealsointroducedattheendofthe1970s.Benedict(1979)foundthat12-month-oldswouldorientreliablytoafamiliarobject
whenitwasnamed,evenwhennonverbalbehaviorssuchasgazeandpointingby
thespeakerwereeliminated.Inamorepreciselycontrolledprocedureusingeye
movementsasanindexofwordrecognition,Thomas,Campos,Shucard,Ramsey
and Shucard (1981) compared the ability of 11- and 13-month-old infants to
identifyafamiliarnamedobjectfromanarrayofcompetitorsmatchedforvisual
salience.Thefindingthatsome12-month-oldscouldidentifythecorrectreferents
ofafewfamiliarwordswashardlysurprising,sincethatmuchwasknownfrom
observationalstudiesandparentalreport.Whatthisnewmethodofferedwasa
way to assess word recognition more objectively. Unlike the informal observationsofcomprehensionusedearlier,thisprocedureenabledtheexperimenterto
standardizestimuluspresentations,todefinecarefullywhichbehaviorscounted
asacorrectresponse,andtoeliminategesturalandothernonverbalcuesfromthe
experimenterthatmightindicatethetargetobject.
TheinnovativestudiesbySpelke(1976)andThomasetal.(1981)provided
a foundation for later research by Golinkoff, Hirsh-Pasek, Cauley and Gordon
(1987)andbyReznick(1990)inwhichpreferential-lookingmeasureswerefurtheradaptedforuseinassessingearlylanguagecomprehension.Theversionof
themethoddevelopedbyGolinkoffetal.,nowknownastheintermodalpreferentiallookingparadigm(IPLP),hasbeenusedinnumerousstudies(e.g.,Hollich,Hirsh-Pasek,&Golinkoff2000;Shafer&Plunkett1998;Meints,Plunkett,
&Harris2002).Intheproceduresusedinthesestudies,infantsareshowntwo
picturesofobjectsastheyhearspeechnamingoneoftheobjects.Insomestudies
gazepatternsarecodedinreal-timeusingabuttonboxtorecordfixationsonthe
targetanddistracterobjectsandshiftsbetweenthetwopictures.Thedependent
measures typically used as an index of comprehension in such studies include
totallookingtimetothetargetpictureanddurationoflongestlooktothetarget
picture.Withthesemeasures,researchersusingtheIPLPhavebeenabletoinvestigateseveralinterestingquestionsaboutchildrensearlylexicalandsyntactic
knowledgeforexample,whetherornotinfantsataparticularagelookrelatively
longertothecorrecttargetpicturethantothedistracterpicturewhenaskedto
identifythetargetobjectinasentencesuchasFind the cat(e.g.,Golinkoffetal.
1987),ortointerpretaspatialpropositioninasentencesuchasLook at the cat on
the table(Meintsetal.2002).
102 AnneFernaldetal.
1.
Theevolutionofthelooking-while-listeningprocedure:
Movingtoreal-timemeasures
Informed by the studies of Thomas et al. (1981), Golinkoff et al. (1987), and
Reznick(1990),ourresearchgroupbegantouseamodifiedversionofthepreferential-lookingmethodtoinvestigatewhetherparticularfeaturesofinfant-directedspeech,suchasexaggeratedpitchandvowellengthening,mightmakeiteasier
foryounglanguagelearnerstoidentifyfamiliarwordsinfluentspeech.Ourinitial
goalinmodifyingthepreferential-lookingparadigmwastoincreasethesensitivity, reliability, and validity of the measures, by making minor modifications
totheprocedurethatservedtoeliminateconfoundingvariables.Earlierpreferential-lookingstudieshaduseddifferentstimuliastargetanddistracterobjects,
thus potentially confounding object salience with target status. And some had
alsofailedtocounterbalancesideoftargetobjectpresentation,anotherconfound
thatmadeitdifficulttointerpretinfantsselectivelookingbehaviorunambiguously.Toaddresstheseconcerns,wemadesurethatalltargetobjectswerealso
presentedasdistracterstoreducetheinfluenceofinitialobjectpreferences,and
that side of presentation was fully counterbalanced to control for possible side
bias.Theseandotherminorbutpotentiallyinfluentialchangeswereundertaken
toincreaseinternalvalidityinourexperimentaldesigns.Inaddition,wemadea
majorchangeinthemeasuresusedtocaptureinfantsgazepatternsinresponse
tospeech:Ratherthancodingeyemovementsinreal-timeusingabuttonbox,as
wasthepracticeinotherpreferential-lookingproceduresatthetime,webeganto
codeeyemovementsfromthevideotapes,frame-by-frameinslowmotion.This
changeenabledustoeliminatefromourmeasurementsthenoiseintroducedby
theca.300mslatencyoftheobservertopressthebutton,acriticalfirststepinthe
directionofachievinggreaterprecisioninourdependentmeasures.Thesemodificationsresultedinalabor-intensiveversionoftheoriginalprocedure,requiring
severalhourstocodetherecordofeach5-mintestsession,buttheenhancedreliabilityofthemeasurementsjustifiedtheeffort.
However,wewerestillnotyettakingfulladvantageofthepotentialforincreased temporal resolution in our measures of infants gaze patterns. In the
Golinkoffetal.(1987)paradigm,wordrecognitionwasoperationalizedasatendencytolooklongeratthenamedtargetpicturethanatthedistracterpicture,
withlooking-time-to-targetaveragedovera6-smeasurementwindowfollowing
offsetofthespeechstimulus.Asimilarcriterioniscommontodayinmostlabs
usingthisprocedure(Hollichetal.2000;Schafer&Plunkett1998),andweadopteditinitiallyaswell.The6-sresponsewindowreflectedapremisethatseemed
entirely reasonable at the time. Although psycholinguistic research with adults
hadrecentlyshownthatexperiencedlistenerscanprocessspeechincrementally,
Lookingwhilelistening 103
generatinghypothesesaboutwordidentitybasedonwhattheyhaveheardupto
thatmoment(Marslen-Wilson&Zwitserlood1989),wesimplyassumedthatinfantswouldbeconsiderablyslowerthanadultsinprocessingcontinuousspeech.
Andsincetherewerenodatayetavailableonthetimecourseofreal-timespeech
processingbyinfants,wehadtodiscoverthehardway,throughalongprocessof
trialanderror,howwrongwewereinthisassumption!
Inthreeearlystudieswetestedinfantsatdifferentagestoinvestigatetheinfluenceofprosodicfeaturesontheirabilitytorecognizefamiliarwords.While
theseresultswerepromising(Fernald,McRoberts,&Herrera1992;seeFernald,
McRoberts,&Swingley2001,forareviewofthesefindings),itbecameincreasinglyclearthatweneededtounderstandbetterhowthetimecourseofinfants
responseschangedwithdevelopmentoverthesecondyear.Forexample,when
usingapercent-correctmeasureaveragedovera6-smeasurementwindow,we
weresurprisedto findthat 24-month-olds apparently performedless well than
18-month-olds.Thiscounterintuitiveresultsuggestedthatourmeasuresoflookingtimewerefailingtocapturethegainsinaccuracyexpectedtowardtheendof
thesecondyear,anobservationthatledustomakeparametricreductionsinthe
timewindowoverwhichlooking-timewasaveraged.Whenweuseda4-smeasurementwindow,thedatabegantomakemoresense,andwitha2-swindow,the
predictedimprovementinwordrecognitionfinallybecameclear:for18-montholdsthemeanpercentageoflookingtimetothetargetpictureduringthismuch
shorterwindowwasca.60%,andfor24-month-oldsitroseto80%.Byadopting
the6-scodingwindowthatwasstandardatthetime,wehadgreatlyunderestimatedtheaccuracyof24-month-oldsinthiswordrecognitiontask;theseolder
infantshadinfactorientedquicklytothetargetpictureuponhearingitnamed
andhadlookedatitfor23seconds,butthentheytendedtolookbackatthe
otherpictureortolookaway.Sincetheselook-backsandlook-awaysbythe24month-oldsinmostcasesfollowedacorrectresponsewithinafewseconds,they
wereactuallyasignofrapidprocessing;however,whenthispost-responsenoise
wasaveragedintothepercent-looking-to-targetaveragedovera6-swindow,the
24-month-oldslookedlessoverallandthusappearedtobelessaccuratethan18month-olds.Thisdiscrepancywasobviouslynotbecausetheolderchildrenwere
slowerorlessreliableinrecognizingfamiliarwords,butratherbecausetheyrespondedmorequicklythantheyoungerchildrenandthentendedtoloseinterest
inthetargetpicturemidwaythroughthe6-swindow.
Itwasclearfromtheseanalysesthatinfantswereabletoidentifyfamiliarspokenwordsmuchmorerapidlythanwehadimaginedpossible.Thusatthispoint
wemadetwoimportantproceduralchanges:First,webegantomeasureinfants
eye movements from the onset of the target word, not the offset. And second,
ratherthanaveraginglookingtimeoveranarbitrary,fixedcodingwindow(which
10 AnneFernaldetal.
mightbesuitableatoneagebutnotatanother),webegantocodeeyemovements
atthefinestlevelofresolutionpossiblegiventhelimitsofourtechnology.Inthe
firstpublishedstudyusingtimecoursemeasuresofspokenwordrecognitionwith
infants(Fernaldetal.1998),thislimitwas100ms,theresolutionofourtimecodegeneratoratthattime.Insubsequentstudiesthelevelofresolutionincreased
to33ms,thedurationofasinglevideoframe.Thischangewascriticalinenabling
ustomovefromtheglobalmeasureoftotallookingtimetothetargetpictureto
amuchmoreprecisemeasureofreactiontime,capturingthechildslatencyto
shiftfromthedistractertothetargetpicture.Throughthisincrementalprocess
of refining our analysis techniques and improving our coding technology, the
looking-while-listeningprocedurehasbecomeanincreasinglypowerfulmethod
formonitoringthetimecourseofinfantscomprehensionofcontinuousspeech,
enablingustomeasurereactiontimeaswellasaccuracyinwordrecognitionin
veryyounglanguagelearners.Theresultisatestingprocedurewithminimaltask
demandsthatcanbeusedwithbothinfantsandadults,yieldingeyemovement
datacomparableinreliabilityandprecisiontodatafromadultstudiesthatrequire
technicallymoresophisticatedautomatedeyetrackingmethods(e.g.,Tanenhaus
etal.2000;Henderson&Ferreira2004).
2.
Inthissectionweprovideanoverviewofthelooking-while-listeningprocedure,
followingthetraditionalformatofaMethodssectioninaresearchreport(i.e.,
describingparticipants,procedure,etc.).However,consistentwiththefocuson
methodologyinthisvolume,theinformationineachsectionextendsbeyondthe
detailsrelevanttoanyparticularstudy,integratingourexperienceusingthisparadigm with many different experimental designs and with participants ranging
from12-month-oldinfantstoadults.Ourgoalhereistopresentanoverviewof
thelooking-while-listeningparadigmatafunctionallevel,discussingthelogicof
eachstepintheprocedure,frompreparingandrunninganexperimenttocoding
eyemovementsandanalyzingthedatausingseveraldifferentmeasuresofefficiencyinspokenlanguageprocessing.
Asmentionedintheintroduction,thelooking-while-listeningprocedureis
superficiallysimilartoapreferential-lookingprocedureinthatinfantsareshown
twopicturesoneachtrialandhearspeechnamingoneofthepicturesaswerecordtheirgazepatternsinresponsetothespeechsignal.Butthestaticnotionof
preferenceisirrelevantforourpurposes.Ratherthanconstruinginfantslookingbehaviorinresponsetospokenlanguageasmotivatedbypreference,weare
Lookingwhilelistening 10
interestedinhowchildrenestablishreferencebymakingsenseofspokenlanguage
frommomenttomoment,aprocessofincrementalinterpretationthatishighly
dynamicinadultcomprehension(e.g.,Knoeferle&Crocker2006).Ratherthan
relyingexclusivelyonasinglepreferencescorebasedontotallookingtothetarget
averagedoverafixedcodingwindowfollowingthespeechstimulus,wearealso
interestedinthetimecourseoflookingtothereferentasthesentenceunfolds.To
achievethisgoalwecodechildrensgazepatternsoff-linethroughcarefulframeby-frameinspection,enablingprecisemeasurementoftheirlatencytoinitiatean
eyemovementtowardtheappropriatereferent,time-lockedtocriticalmoments
in thespeechsignalon each trial. Thus the looking-while-listening procedure
incorporates the same sensitive temporal measures used in eyetracking studies
withadultsandolderchildren(seeTrueswell&Tanenhaus2005),differingfrom
those methods in only three noteworthy respects: first, we typically use visual
displays with only two alternatives rather than more complex scenes involving
fourormoredisplays;second,wedonottouseanautomatedeyetracker,because
comparableprecisionandreliabilitycanbeachievedusingourhigh-resolution
videocodingprocedures;andthird,aninherentconstraintinworkingwithinfantsandveryyoungchildrenisthattheyhavelimitedandfluctuatingattention;
thuswehavetodesignexperimentslastingjustafewminuteswithonly3040
trials,whichyieldmuchless(andpotentiallynoisier)datathanthelongerexperimentswithhundredsoftrialsthatarepossibleinstudieswithadults.
2.1
Participants
Thelooking-while-listeningparadigmhasbeenusedeffectivelyinourlaboratory
withinfantsasyoungas14monthsofageaswellasolderchildrenandadults.
Whenwetested12-month-oldsinseveralpilotstudiesusingthisprocedure,we
foundtheirperformancetobeclosetochance,asdidZangl,Klarman,Thal,FernaldandBates(2005)inastudyusingasimilarprocedure.While12-month-olds
arehappytolookatthepicturesdisplayed,theyarelikelytofixateononlyone
picture and ignore the other on any given trial, shifting less frequently overall
thaninfantsjusttwomonthsolder.Thispatternofresultswithyoungerinfants
reflectstheirlimitedlinguisticknowledge,butmayalsoresultfromattentional
limitationsatthisage.
2.2
Preparingvisualstimuli
In studies in which we assess recognition of familiar words, the visual stimuli
consistofpicturesofrealobjectswithwhichinfantsandtoddlersareveryfamil-
106 AnneFernaldetal.
iar.Thetargetobjectschosenforeachstudyhavenamesthatarehighlylikelyto
beunderstoodbychildrenintherelevantagerange,basedonvocabularynorms
fortheMacArthur-BatesCommunicativeDevelopmentInventory(Fensonetal.
2007). Realistic images of common objects judged to be prototypical for childrenateachagearedrawnfromvarioussources,includingimagebanksonthe
Internetandpicturestakenwithdigitalcameras.Forstudiesinwhichchildren
aretaughtandthentestedonnovelwords,weusepicturesofconstructedobjects
withwhichchildrenhavenopriorexperience.Insomecases,objectsareshown
againstauniformgraybackground;however,instudieswherethesametargetobjectsareusedrepeatedlyacrosstrials,objectsmaybepresentedonmorediverse
andcomplexbackgroundsthatvaryincolorandpattern,inordertomaintain
childrensinterest.
Allimagesusedasstimuliareeditedsotheobjectsareapproximatelythesame
sizeandareinformallymatchedinvisualcomplexityandbrightness.Whenchoosingandeditingpicturesforaparticularstudy,itisimportanttobalancethevisual
salienceofthetargetanddistracterpicturesoneachtrial,keepinginmindthatany
givenpictureisnotinherentlysalientbutonlyinrelationtootherpicturesinthe
stimulusset.Ifonepictureinapairismuchmoreengagingthananother,itismore
difficulttotellwhetherlookingtothetargetpictureonagiventrialwasinfluenced
bythespeechstimulusratherthanbybaselinevisualpreferences.Wehavefound
thatyoungerinfantsinparticulargenerallyfindimagesofanimateobjectsmore
interestingthaninanimateones,andthusmightinitiallyattendmoretoapicture
ofananimalifitwaspairedwithapictureofanartifact.Toreducetheeffectsof
thispotentialbias,wepresentyokedpairsofanimates(dogandbaby)onsometrials,andpairsofinanimates(shoeandcar)onothers.Althoughcounterbalancing
measures,suchaspresentingeachstimuluspictureasbothtargetanddistracteran
equalnumberoftimesineveryexperiment,aredesignedtomitigatetheeffectsof
differentialvisualsalienceamongstimuli,suchdifferencesshouldbereducedas
muchaspossiblebecausetheycontributenoisetothedata.
Insomeexperimentsthepotentialforvisualsaliencedifferencesisparticularlyacuteduetothenatureofthestimulusobjectsused,andinthiscasepilot
testingisadvisable.Toevaluatethesuitabilityofnewimagesaspotentialstimuli,
candidatepicturesarepairedwithotherpotentialstimuluspicturesandpresented
tochildrenintheappropriateagerangewithoutanyverbalinformation,todetermine whether participants orient more to one picture than the other based
ondifferencesinvisualsalience.Ifchildrendoindeedfixateonepicturesignificantlymorethantheother,thepicturesneedtobeeditedfurthertoadjustfactors
likehue,saturation,andvisualcomplexitythatmightcontributetodifferential
salience.Themodifiedstimulithenneedtobepilot-testedagainuntilchildren
areequallylikelytofixatebothpictures.Balancingtherelativevisualsalienceof
Lookingwhilelistening 107
stimulusobjectsisanimportantstepinmaximizingthesensitivityofthemeasuresinthelooking-while-listeningprocedure,especiallywhenworkingwithinfantsyoungerthantwoyearsofage.Thisextrastepisnotaslabor-intensiveasit
mayseem,ifsuchpicture-pilottrialsforpotentialstimuliinanewexperiment
canbeincludedasfillertrialsinanotherongoingexperimentwithchildreninthe
sameagerange.
2.3
Preparingauditorystimuli
108 AnneFernaldetal.
acousticonsetandoffsetsofkeywordsthatarecrucialformeasuresofresponse
latency.Forexample,wheninvestigatingchildrensefficiencyininterpretingfamiliarobjectnamesinsimpleEnglishsentencessuchasFind the car,determining
theonsetanddurationofthetargetwordissufficienttotime-lockthechildsresponsetotheunfoldingobjectname(e.g.,Fernaldetal.1998,2006).Inthiscase,
thenounonsetisthecriticalpointatwhichthechildbeginstoaccumulatephoneticinformationnecessarytointerpretthetargetwordandidentifythereferentobject.However,thecriticalpointmaycomeearlierincaseswherelinguistic
informationencounteredbeforethenounispotentiallyinformative.Forexample,
nounsinSpanishhaveobligatorygrammaticalgenderandareprecededbygender-markeddeterminers.InastudyinvestigatingwhetherSpanish-learningchildrencouldtakeadvantageofgender-markedarticlestoidentifyanupcomingobjectname,Lew-WilliamsandFernald(2007)measuredchildrenseyemovements
tothetargetobjectinrelationtothefirstpotentiallyinformativeelementinthe
sentence,inthiscasetheprenominalarticle.Thisstudyrequiredmeasurement
oftheonsetandoffsetofthedetermineraswellasthetargetnoun,andtokensof
eachstimulussentencewerechosentomaximizecomparabilityonthesedimensionsacrossthestimulusset.Thatis,thebest-matchingtokenswereselectedfrom
thelargersetofexemplarsrecordedforeachsentence.Thesewerethenfurther
editedsothatkeypointsinthespeechstreamcouldbecarefullymatchedwithin
oneframeacrosssentences.
By making the critical speech stimuli in each condition prosodically comparable,wecanreducetheeffectsofpotentiallyconfoundingfactorsthatmight
influencechildrensresponses.However,maximizingexperimentalcontrolinthis
wayalsohasalessdesirableconsequence:weendupwithastimulussetthatis
highly repetitive and potentially monotonous overall, which has drawbacks in
termsofmaintainingchildrensattention.Experiencehasshownusthatrepeatingshortsentenceswiththesameintonationcontourontrialaftertrialcauses
childrentoloseinterestsoonerthanwhentheyhaveamorevariedsequenceof
speechstimulitolistento.Onestepinreducingthisproblemistoaddatagquestion or statement to the speech stimulus on each trial, following the sentence
containingthetargetwordsentencebya1-sinterval.Theseshortsentencesareall
attention-getterssuchasCan you find it?orCheck that out! Thesamespeakerwho
producedthestimuliforthestudyrecordsthesetagsentenceswithlivelyintonationthatvariesfromtokentotoken,andtheyareappendedtothespeechstimuli
in a way that maximizes variability from trial to trial by alternating sentences
withrisingandfallingpitchcontours.Asecondstepinincreasingvariabilityisto
interspersetesttrialswithfillertrials,asdescribedbelow.
2.
Lookingwhilelistening 109
Constructingstimulusorders
Thevisualandauditorystimuliarecombinedtoconstructtwoormoredifferent
stimulusorders.Althoughinearlierstudiesusingthismethodweincludedonly
1020trialsineachexperimentalorder,inmorerecentstudieswehavelearned
fromexperiencehowtogethighqualitydataonalargernumberoftrials.Oneinsightwastheimportanceofincorporatingtherightamountofvisualandauditory
complexityinexperimentalandfillertrialsacrosstheentirestimulusset,finding
theappropriatebalancethatisoptimalforaparticularagegroup.Byfine-tuning
stimulus sets in this way, we can now test 18 month-olds on 3040 trials, and
olderchildrenon4050trials.Trialtypesarecounterbalancedsothatthetarget
objectappearsontheleftandrightsidesanequalnumberoftimes,anddoesnot
appearonthesamesidemorethanthreetrialsinarow.
Inadditiontousingtagsentencestoincreaseprosodicvariabilitywithinexperimentaltrials,itisalsoimportanttoincludefillertrialsthatintroducevisualas
wellasacousticnoveltyacrossthestimulusset.Ifastudyrequiresusingmultiple
tokensoftrialsofoneparticulartype,thenthetotalnumberoftrialsmayinclude
upto30%fillertrials,designedtoreducerepetitivenessandtomaintainchildrens
interest.Fillertrialsmightconsistofjazzypicturesthatarevisuallymorecomplex
andcolorfulthantheexperimentalstimuli,suchasamarketsceneorhotairballoons,accompaniedbyanexclamationsuchasGood job heres another picture!
Theymightalsoconsistoftrialtypeswearepilottestingforotherexperiments,
aslongasthetargetobjectsaresufficientlydifferentfromthosethechildrenare
being tested on. Thus one critical factor in maximizing childrens attentiveness
throughouttheexperimentistocarefullybalancevariationandrepetitionacross
thestimulusset.Anothercriticalfactoristoprovideastimulussetthatischallenging,butnottoochallenging,forchildrenateachparticularage.
2.
Apparatus
110 AnneFernaldetal.
Figure 1. (A)Configurationoftestboothwithrear-projectionscreenusedinthe
looking-while-listeningprocedure;(B)schematictimelineforatypicaltrial.
camera,andthechildsheadisca.1mfromthescreen.Children4yearsandolder
sitaloneonanon-swivelchair,alsoadjustedtothecorrectheightsotheireyesare
centeredonthevideorecording.Whenaparentisholdingthechildduringthe
testsession,theadultwearsdarkglassesmadeopaquebycoveringthelenseson
theinsidewithclothtape,sothevisualstimulicannotbeseen.Althoughpicture
sizemayvaryfromoneexperimenttoanother,thepicturesusedinourstandard
setupareprojectedatasizeofca.3650cm,alignedhorizontallyatadistanceof
60cmfromoneanother,withthecenterofeachimageslightlyabovethechilds
eyelevel.Speechstimuliaredeliveredthroughacentralloudspeakerlocatedon
thefloorbelowthescreen.Theblack-and-whiteanalogvideocamera,sensitiveto
lowlightlevels,isfittedwithazoomlensthatisfocusedonthechildsfaceonly.
Positioningeachchildrelativetothecameraandfocusingthecameraforeach
sessionarecrucialtocollectinguseabledata;onlywhenthechildseyesareclearly
visibleisitpossibletoreliablyidentifytheshiftsinfixationthatarecrucialfor
calculatingresponsespeed.
The experiment is run using PsyScope software controlled by a Macintosh
computerlocatedinanadjacentcontrolroom.Oneachtrialthiscomputerout-
Lookingwhilelistening
putsthevisualstimulitotherear-screenprojectorinthetestingroom,andalsoto
amonitorinthecontrolroom.Thevideorecordofthechildseyemovementson
eachtrialissenttoamixerinthetestingroomwhichintegratesthevideosignal
withgraphicinformationaboutparticipantandtrialnumber,stimulusorder,and
theonsetandoffsetofauditoryandvisualstimuli.Thiscombinedinformation
isthenfedtoatimecodegeneratorwhereitistime-stampedandsenttoaVCR
which records the complete video signal, including the childs eye movements,
theeventinformation,andthetimecode.Thiscompositevideorecordisthen
digitizedonasecondcomputerinthecontrolroom.
2.6
Testingprocedure
Eachtestsessionisprecededbya15-minperiodduringwhichtheparentcompletestheproceduresforinformedconsentandtheexperimenterinteractswith
thechild.Whenbothparentandchildarecomfortableandreadytobegin,theyare
escortedtothetestingroom.Thelightsinthetestingroomarealreadydimmed,
andimagesofpuppetsorcartooncharactersfamiliartoyoungchildrenareonthe
screen.Alampmountedontopofthescreenprovidesjustenoughlighttoavoid
completedarkness,ensuringthatthevisualdisplayisthemostinterestingthingin
sightforthechild.Inatypicaltestingsituation,twoexperimentersareinvolved:
thefirstexperimenterseatstheparentandchildinthetestingroomandchecks
thatthecameraisoptimallypositioned;thesecondexperimenter,locatedinthe
controlroom,speakstothechildoverthemicrophonetofamiliarizeherwiththe
soundsourceanddirectsattentiontothevisualdisplay.Ifnecessary,thesecond
experimenteralsocoordinateswiththeexperimenterinthetestingroomtofocus
thecameraonthechildsface.Oncethechildisateaseandtheparenthasputon
thedarkenedsunglasses,thefirstexperimenterleavestheroomandthetestsessionbegins.
On a typical test trial, the target and distracter pictures are shown for 2 s
priortotheonsetofthespeechstimulus,providingthechildwithenoughtimeto
checkoutbothpictures.Asmentionedearlier,usingeachobjectastargetinone
trialanddistracterinanothertrialconstitutesanimportantcontrol,offsettingthe
problemthatonepicturemaybefixatedfirstbecauseitwasnamedonaprevious
trialorsimplybecauseitisvisuallymoresalient.Figure1Bshowsthetimelinefor
atypicaltrial.Trialslastonaverage68s,includingthe2sbeforetheonsetofthe
speechstimuli,andareseparatedbyan800msbriefintervalwhenbothscreens
areblank.Theentireexperimentlastsabout56min.
Ineyetrackingstudieswithadults,participantsareinstructedtolookatacentralfixationpointbeforerespondingtotheverbalstimulus,andonmosttrials
111
112 AnneFernaldetal.
theywillinglycomply.Thisstephastheadvantagethattheadultstartseachtrialat
thesameneutrallocationequidistantfromallthevisualstimuli;thusinprinciple
alatencytoshiftfromthecenterpointtooneofthestimulusobjectscanbeassessedoneverytrial.Asimilarmethodhasbeenusedineyetrackingexperiments
withpreschoolchildren,whoarerequestedtolookatacentralsmileyfaceatthe
start of a new trial (see Trueswell this volume). In the looking-while-listening
procedurewedonotuseacentralfixationpoint,forthesimplereasonthatinfants
andveryyoungchildrenwillnotfollowsuchinstructions.Althoughsomepreferential-lookingstudieshaveusedacentralfixationpointattrial-onset,nonehave
reporteddataonhowlongchildrenactuallymaintainfixationatcenter.Weexploredthisissueinarecentstudybypresentingabrightgeometricimagebetween
thetwostimuluspicturesto26-and36-month-olds,inanefforttobringtheir
attentiontocenteratthestartofthespeechstimulus(Portillo,Mika,&Fernald
2006).However,withinmillisecondschildrenatbothageshadshiftedrandomly
toonesideortheotherbeforethetargetwordwasspoken.Sincemostchildrenon
mosttrialswerealreadylookingatthetargetordistracterobjectbeforehearing
thetargetword,thereisnoevidencethatacentralfixationpointisactuallyeffectiveinstudieswithveryyoungchildren.
2.7
Pre-screeningforcodeabletrials
The video record of each test session is integrated with information about the
participant(e.g.,subjectnumber,gender,age)aswellasinformationrelevantfor
dataanalysis(e.g.,stimulusorderandtrialtype)usingthecustomsoftwareEyecoder.Thisinformationisnotvisibletoobserversduringprescreeningorcoding.
AsshowninFigure2,thevideorecordofthetestsessionalsoincludesadditional
informationthatisvisuallyaccessibleandusefultocoders,suchasthedate,time
code,subjectnumber,trialnumber,andvisualmarkersthatappearatthepoints
whenthepicturesandspeechstimulusarepresentedtothechildoneachtrial.Becausecodersareblindtotrialtype,theyhavenoaccesstothevisualstimuliseen
bythechildandtheycodeeyemovementsinsilencewithnoaccesstotheverbal
stimuli. Thus thesemarkers are necessary for time-lockingvisual andauditory
stimulioneachtrialinrelationtothechildseyemovements.
Pre-screening precedes coding and serves the primary purpose of flagging
anynon-useabletrialsonwhichthechildsresponsewasaffectedbyfactorsunrelatedtotheauditoryandvisualstimuli.Twopre-screeners,blindtothehypothesis as well as to side of target presentation, independently watch each testing
sessioninrealtime,withthesoundoninordertoidentifyanytalkingbythechild
or parent during the trial. Because the pre-screeners have access to the verbal
Lookingwhilelistening 113
Figure 2. Screenshotofchildsfaceduringtesting,showinginformationaccessibleto
observersastheyconductframe-by-framecoding.ThePictureandSoundcodes
appearonthescreenastheimagesandspeechsoundsarepresentedtothechild,
enablingcoderstolinktherecordofeyemovementstotheonsetandoffsetofvisualand
auditorystimuliacrosseachtrial.
stimuli,theyareneverthesameindividualswhocodeeyemovementsforthat
study.Thetaskofthepre-screeneristomarkintheEyecoderrecordanytrialsthat
shouldbeeliminatedfromtheanalysisoftheexperimentaltrialsforanyofthe
followingreasons:thechildwasnotlookingatthepicturespriortosoundonset;
theparentorchildwastalkingduringthetrial;thechildwasfussyorinattentive;
thechildchangedpositionsothefacewasnotvisible,etc.Onlyafterbothprescreenershaveagreedonalluseableandnon-useabletrialsdoeseyemovement
codingbegin.
2.8
Codingeyemovements
Coding involves reducing the continuous record of the childs eye movements
oneachtrialtoaseriesofdiscreteevents.Giventhatourgoalistomeasurethe
latencyofthechildsshiftingazefromthedistractertothenamedtargetpicture
inrelationtocriticalpointsinthespeechstream,aswellasthedurationoffixationstothetargetanddistracterpicturesoneachtrial,high-resolutioncodingis
essential.Highlytrainedcoders,unawareoftrialtypeandtargetlocation,useEyecodertomoveframe-by-framethroughthedigitalrecordofeachtestsession.The
coderstartsbysynchronizingthetimecodewiththeonsetofthespeechstimulus,
markingtheframetowhichsubsequenteventsarereferenced.Figure3illustrates
thetimelineofeyemovementscapturedintheEyecoderrecord,fromtheonset
oftheutterancetotheoffsetofthepictureontwotrials.Thecoderstaskisto
11 AnneFernaldetal.
Figure 3. Samplecodingrecordfortwo4-strialsusingEyecoder.Eachlineindicatesthe
timeatwhichthecoderjudgedthatachangeoccurred,eitherinthestimuli,e.g.,from
sound ontosound off,and/orinthepositionofthechildsfixation,e.g.,fromrighttooff.
advancethevideorecordoneframeatatime,indicatingeachtimeachangeoccursinthechildsfixationwithoneoffoureventcodes:left(onleftpicture),right
(onrightpicture),off (betweenthetwopictures),andaway(off-task,lookingat
parent).Thecoderalsomarkstheframeswhenthevisualcuesforsoundand
picturesappearanddisappear,indicatingthattheauditoryandvisualstimuli
areonoroff.Closeattentionisrequiredtopreciselycodethefirstframeinwhich
theinfantinitiatesashiftfromoneobjecttotheotherandiscurrentlynotfixating
either.Sinceourlatencymeasuresarebasedonsuchshifts,codersvieweachof
thesetransitionpointsseveraltimesbytogglingbackandforthbetweenframes
tobesureoftheirjudgment.
2.9
Reliabilitycoding
Becauseerrorsinidentifyingtheonsetandoffsetofvisualfixationscanhavea
substantial impact on the accuracy of reaction time (RT) analyses, coders undergoseveralweeksoftrainingandpractice,withclosesupervision.Eachnovice
codermustcompleteastandardtrainingsetoffourtestsessionspre-selectedto
representparticularcodingchallengesindifferentagegroups.Theymustachieve
areliabilityscoreof96%orbetteronthetrainingprotocolbeforetheyareauthorizedtocodeactualdata.Aftercompletingtraining,allcoderscontinuetopar-
Lookingwhilelistening 11
ticipateinregularreliabilitycheckstoassesstheextenttowhichthatallcodersare
calibratedaccordingtolabstandards.Codersareneverinformedthattheyare
codingasessionchosenforreliabilitypurposes,toensurethattheyarenotbeing
extracautiousbecausetheyknowitisareliabilitytest.
116 AnneFernaldetal.
theotherofthetwopicturesonamajorityofthe90consecutiveframesoneach
3-strial.Sincethelikelihoodofagreementismuchhigherduringthesesustained
fixationsthanduringperiodsoftransition,itispossiblefortwocoderstoreceivea
veryhighentire-trialagreementscore,yetstilldisagreeabouttheexactframeson
whichshiftsingazebeginandend.Thisisproblematicbecausejudgmentsabout
theinitiationanddurationofshiftsarecrucialinmeasuresofreactiontimeand
accuracy.Thusthetraditionalentire-trialscorealoneistoolenientameasureof
reliability,althoughincombinationwithamorestringentreliabilitymeasure,it
canhelpidentifydiscrepanciesamongcoders.Shift-specific reliability scoresfocus
onlyonsequencesofframeswhereshiftsoccur.Theshift-specificcalculationisa
percentageofframesfromshiftstarttofinishonwhichcodersagree,differingby
oneframeatmost.Ifentire-trialandshift-specificscoresarebelow95%and90%,
respectively,anindependent,highly-experiencedcoderexaminesthediscrepanciesbetweencodersandmakesafinaldecisionaboutthegazepatternsforthe
sessionunder review.At that time,coders whomay haveshiftedstandardsfor
judgingeyegazearere-trainedasnecessary.
Lookingwhilelistening 117
Figure 4. Schematicrepresentationofdifferenttypesofresponsepatternsonhypotheticaltrialsinwhicheyemovementsaretime-lockedtotheonsetofthetargetnoun.The
dashedverticallineindicatesaveragetarget-nounoffset.Thissequenceoftrialsismeant
toillustratedifferentclassesofpossibleresponsepatternsandisnotrepresentativeofthe
actualdistributionoftrialtypesobservedintestsessionswithparticularchildren.
thecorrectpictureforatleast1400msaftertheonsetofthetargetnoun.Trial9is
alsoatarget-initialtrial;however,onthistrialthechildshiftedawayincorrectly
tothedistracterpictureafterhearingabout433msofthenoun.Trials3and4
aredistracter-initial trialsonwhichthechildshiftedquicklytothetargetpicture,
yieldingreactiontimesof500msand400msrespectively.Ontrial7,thechild
alsostartedonthedistracter,butthistimetheshiftwasinitiatedonly133msafter
nounonset.Suchveryshortshiftlatenciesarenotincludedinthecalculationof
mean reaction time, because they are likely to have been initiated prior to the
pointwherethechildhadenoughacousticinformationfromthenountomake
aninformedresponseandthentoinitiateaneyemovement.Ontrial8,thechild
againstartedonthedistracterpicture,maintainingfixationonthewrongpicture
forthedurationofthetrial,withoutevershiftingtothereferent.Onothertrials
inFigure4,thechildwaseitherbetweenpicturesoroff-taskattheonsetofthe
noun.Pleasenotethatthesequenceoftrialtypesshownheredoesnotrepresent
118 AnneFernaldetal.
the distribution of response patterns for a typical experimental session with a
childatanageforwhichtheverbalandvisualstimuliareappropriate.Trials14
showthemostcommontypesofcorrectresponsesondistracter-initialandtarget-initialtrials;trials59showpatternsofcorrectandincorrectrespondingthat
arecommon,butoccurlessfrequently,andtrials1012representawaytrials
thatarerelativelyinfrequentandwouldnotbeincludedinanalysesofreaction
time.ThusthegoalinFigure4isnottoshowarepresentativesequenceoftrials
foraparticularchild,butrathertoprovideexamplesofdifferentgazepatterns
thatillustratethevariabilitypossibleinchildrensresponseswithinthefirst1800
msfollowingtheonsetofthenoun.
Thissectionexaminesdifferentapproachestoanalyzingdataonthetimecourse
ofchildrenseyemovementsastheylookatpicturesandlistentospeechthatreferstooneofthepictures.Firstwepresentplotsofchildrensshiftinggazepatterns
inrelationtoparticularwordsinthestimulussentenceasitunfoldsovertime.
Nextwedescribehowtoderivediscretemeasuresofchildrensspeechprocessing
efficiencyfromthetimecourseinformationforuseinstatisticalanalyses,focusingonreactiontimeandaccuracy.Andfinally,weaddressthequestionsofhow
stablesuchtimecoursemeasuresofprocessingefficiencyarewithinindividual
Lookingwhilelistening 119
children,andwhethertheyaremeaningfullyrelatedtootherdimensionsofearly
languagecompetence.
3.1
Plottinggraphsofthetimecourseofchildrenseyemovements
inrelationtospeech
Ausefulfirststepinexaminingthedataistoprepareanonset-contingent(OC)
plot,whichdividestrialsaccordingtowherethechildislookingattheonsetof
thecriticalwordinthestimulussentence.AnOC-plottracksseparatelythetime
courseofparticipantsresponsesfortarget-initialtrialsanddistracter-initialtrials
asthestimulussentenceunfolds.Atthebeginningofatrial,thechildhasnoway
ofknowingwhichobjectwillbenamed,andsoisequallylikelytobelookingat
thetargetorthedistracterpictureattheonsetofthetargetword.Thusthebehaviorthatconstitutesacorrectresponsevarieswiththepositionofthechildseyes
attheonsetofthetargetword:ondistracter-initialtrials,thechildshouldquickly
shift awayfromthedistractertothenamedtargetpicture;however,ontarget-initialtrials,thecorrectresponseisnottoshiftbuttostayput.TheOC-plotinFigure
5providesagraphicoverviewofthesetwodifferentresponsepatterns,usingdata
fromoneconditioninacross-sectionalstudywithchildreninthreeagegroups:
18-,24,and36-month-olds(Zangl&Fernald2007).Foreachparticipant,trials
weregroupedcontingentonwhichpicturethechildwasfixatingattheonsetofthe
targetnoun.Plottedonthey-axis,ateach33-msintervalfromtargetwordonset,
isthemeanproportionoftrialsonwhichchildrenatthatpointarelookingata
picturethatisdifferentfromwheretheystartedattarget-wordonset.BecauseOCplotscapturechildrenstendencytoshiftawayfromtheoriginalstartingpoint,
theyshowbothcorrectandincorrectresponses:ondistracter-initialtrials,ashift
awaytothetargetpictureisacorrect response,whileontarget-initialtrials,ashift
awaytothedistracterpictureisanincorrectresponse.Thusachildwithperfect
accuracywouldshiftquicklytothetargetpictureon100%ofthedistracter-initial
trials,andwouldnevershiftawayontarget-initialtrials.
ThetopthreelinesinFigure5trackthemeanproportionofdistracter-initial
trials on which children at each age have correctly shifted from the distracter
and are now looking at the target at each 33-ms interval, plotted over participantsfromtheonsetofthetargetnoun.Thethreelowerlinestrackresponseson
target-initial trials foreachagegroup,plottingthemeanproportionoftrialson
whichchildrenhaveshiftedawayfromthetargetateachtimepointandarenow
lookingattheincorrectobject.Itisclearfromtheverydifferenttrajectorieson
distracter-andtarget-initialtrialsthatthe36-month-oldsweremorelikelythan
the younger children to respond correctly in both ways. On those trials when
120 AnneFernaldetal.
theystartedoutonthedistracterpicture,the36-month-oldsbeganshiftingtothe
correctreferenthalfwaythroughthetargetnoun.However,whentheyhappened
tostartoutonthetargetpicture,theytendedtomaintainfixationonthecorrect
referentratherthanshiftingaway.Incontrast,the18-month-oldsrespondedless
efficientlyonbothcounts:ondistracter-initialtrials,theirgazepatternssuggested
theyweremorelikelytoshiftmoreslowlyandlessreliablytothetargetpicture,
andontarget-initialtrials,theyweremuchmorelikelytofalse-alarmbyshifting
awayfromthecorrectpicture.Ifchildrenwereequallylikelytomakedistracterto-target shifts and target-to-distracter shifts in response to a particular object
name,thisresponsepatternwouldsuggesttheywereunabletoidentifythetarget
wordand/ortomatchitwiththecorrectreferent.Buttotheextentthatchildren
arequickandreliabletoshifttothecorrectreferentondistracter-initialtrials,and
alsotendtomaintainfixationontarget-initialtrialswithoutshiftingaway,wecan
inferthattheyareabletointerpretthespokenwordefficientlyandhaveassociatedthenamewiththerightobject.Byprovidinginformationonbothtypesof
correctresponse,theOC-plotinFigure5offersaglobalviewofdevelopmental
Lookingwhilelistening 121
changesinchildrensefficiencyinidentifyingtheappropriatereferentasthetarget
nounisheard.
Asimplerwaytographthesamedataistouseaprofile plot,asshowninFigure6.Thisplottracksthemean proportion of looking to the target pictureateach
timeinterval,measuredfromtarget-nounonset,averagedoverparticipantsforthe
samethreegroupsofchildrenasinthepreviousanalysis.TheoverallresultsinFigure6are,ofcourse,thesameasthoseshowninFigure5,withthe36-month-olds
respondingmorereliablyandreachingahigherasymptotethantheyoungerchildren.Notethatthecurvesarerelativelyflatatthebeginningofthetargetnoun,beforechildrenhadaccumulatedenoughphoneticinformationtoidentifytheword.
Butvisualinspectionoftheslopesofthethreecurves,andoftherelativepoints
atwhichtheybegintorise,suggeststhatthe36-month-oldsasagroupresponded
morequicklytothetargetwordthandidtheyoungerchildren.IncontrasttoOCplots,profileplotscombineresponsesfrombothtarget-anddistracter-initialtrials
toshowtheoverallmeanproportionoftrialsonwhichthechildislookingtothe
correct referent at each time point, regardless of whether the child was already
lookingatthetargetpictureorhadjustshiftedtherefromthedistracter.
Figures7and8provideanotherexampleofhowOC-plotsandprofileplots
providedifferentvantagepointsonthesamedata.Thedatarepresentedinthese
figurescomefromapilotstudyofchildrensuseofsemanticcuestoidentifythe
122 AnneFernaldetal.
referentofanupcomingnoun(Fernald2004).Hereweaskedwhether26-montholds,likeadults(Altmann&Kamide1999),wouldrespondmorequicklytofamiliarnounspresentedinsentenceframeswithasemanticallyrelatedverb(e.g.,
Eat the cookie) than in frames with a semantically unrelated verb (e.g., See the
cookie). Six object words were presented in both related and unrelated frames,
carefullysplicedtocontrolfordurationsofcarrierphrasesandtargetwords.Note
thatthegraphsofchildrensresponsesinthesetwoconditionsareplottedfrom
sentenceonsetratherthanfromtargetnounonset,becausewepredictedthatthe
veryfirstwordinthesentencewouldhaveaninfluenceontheirspeedoforienting
tothetargetobject.Indeed,onrelated-frametrials,26-month-oldsbeganorientingtothereferentmuchsoonerthanonunrelated-frametrials;theOC-plotin
Figure7showsthatresponsesontarget-anddistracter-initialtrialsbegantodivergeasthechildrenheardtheverb,beforethetargetwordhadbeenspoken.In
contrast,onunrelated-frametrialschildrenhadtowaitforthetargetnoun,since
noearliercueswereavailabletofacilitateidentificationofthereferent.InFigure
8,thesamedataarepresentedinaprofileplot,showingthatchildrenlookedmore
Figure 7. OC-plotshowingateachtimepointthemeanproportionstrialsonwhich
26-month-oldshaveshiftedawayfromthepicturetheystartedonatsentenceeither
correctlyfromthedistractertothetargetpicture(ondistracter-initialtrials,toptwo
curves),orincorrectlyfromthetargettothedistracter(ontarget-initialtrials,lowertwo
curves).Verbalstimuliweresentenceswithrelatedandunrelatedcarrierframesprecedingfamiliartargetnouns.Dashedlinesdemarcatemeasurementwindowsforresponses
duringtheverb,determiner,andnoun.Errorbarsrepresentstandarderrorsoverparticipants(adaptedfromFernald2004).
Lookingwhilelistening 123
Figure 8. Aprofileplotshowingthemeanproportionoftrialsonwhich26-month-olds
fixatethetargetpictureonrelated-frameandunrelated-frametrials.Dashedlinesdemarcatemeasurementwindowsforresponsesduringtheverb,determiner,andnoun.Error
barsrepresentstandarderrorsoverparticipants(adaptedfromFernald2004).
overalltothetargetpictureonrelated-thanonunrelated-frametrials.Unlikethe
dataonage-relatedchangesinwordrecognitionshowninFigure6,therewasno
differenceinasymptotebetweenthetwoconditionshere,because26-month-olds
eventuallyconvergedonthetargetobjecttothesameextentonbothkindsoftrials.WhattheprofileplotinFigure8showsinsteadisasubstantialdifferencein
thetimingofchildrensresponsebetweenconditions.Likeadults,26-month-olds
wereabletotakeadvantageofthesemanticinformationintheverbtoestablish
referencemorequickly,butonlyonrelated-frametrials.
3.2
Derivingmeasuresofprocessingefficiencyfromtimecoursedata
Althoughcontinuousplotsofchildrenseyemovementdataprovideadynamic
pictureofthetimecourseoftheirresponsestoparticularwordsintheunfolding
sentence,theydonotdirectlyrepresentthediscretemeasuresthataremostconvenientforpurposesofstatisticalcomparison.Thetwodependentmeasuresused
mostfrequentlyinourresearcharereaction time,orlatencytoorienttothetarget
word,andaccuracy,basedonlookingtimetothecorrectreferentcalculatedover
particularregionsofthesentence.
12 AnneFernaldetal.
Lookingwhilelistening 12
Figure 9. Reactiontimeanalysis:meanlatenciestoinitiateashiftingazefromthe
distractertothetargetpicture,measuredfromthebeginningofthespokentargetword,
for15-,18-,and24-montholdinfants.Thisanalysisincludedonlyshiftsoccurringon
distracter-initialtrials,i.e.,thosetrialsonwhichtheinfantwasinitiallylookingatthe
incorrectpictureandthenshiftedtothecorrectpicturewithin1800msoftarget-word
onset(adaptedfromFernaldetal.1998).
revealed that over the second year of life, during the same period when most
infantsshowavocabularyspurtinspeechproduction,theyalsomakedramatic
gainsinreceptivelanguagecompetencebyincreasingthespeedwithwhichthey
canidentifyfamiliarwordsandmatchthemwiththeappropriatereferent.These
cross-sectional findings have now been replicated in a longitudinal study with
English-learninginfants(Fernaldetal.2006),andwehavealsofoundcomparable
resultswithinfantsfromLatinofamilieslivingintheUS,learningSpanishastheir
firstlanguage(Hurtado,Marchman,&Fernald2007).
IntheFernaldetal.(1998)study,alldistracter-to-targetshiftsthatoccurred
betweentheonsetofthetargetwordandtheendofthetrialwereincludedinthe
RTanalysis.However,insubsequentexperimentswehavebeenmoreselective,
excludingveryshortresponselatenciesthatpresumablyreflecteyemovements
programmedbeforethestartofthetargetword.Everydistracter-to-targetshift
thatisinterpretableasacorrectresponsehasmultiplecomponents,eachrequiring processing time that can only be estimated. These components include the
timerequiredtoaccumulateenoughphoneticinformationtoidentifythespoken
wordandthentoassessitsrelevancetothecurrentlyfixatedpicture,andthetime
requiredtoinitiateaneyemovement,ifashiftisrequired.Basedonthefindings
ofHaithetal.(1993),itseemedreasonabletoassumethatinfantsinthelooking-
126 AnneFernaldetal.
while-listeningprocedureneedaround200mstoprogramaneyemovement,or
perhapslongergiventheneedtodisengagefromaninterestingdistracterpicture
beforeshiftingtothetarget(Hood&Atkinson1993).Allowingtimefor100ms
orsoofphoneticinformationtoaccumulate,ourfirstestimatewasthatitwas
reasonabletoexcludefromtheRTanalysisanyshiftsthatoccurredinthefirst333
msfromtheonsetofthetargetword(Fernaldetal.2001);inmorerecentstudies
witholderinfants,wehaveusedacutoffof300ms(Zangl&Fernald2007).
InadditiontoexcludingveryshortlatenciesfromtheRTanalysis,ontheassumptionthattheywereprogrammedbeforethetargetwordwasheard,itisalso
importanttoexcludeverylonglatenciesthatarealsounlikelytobeinresponseto
thetargetword.Todevelopguidelinesforsettingtheboundaryforsuchdelayed
shifts,weexamineddistributionsoffirstshiftsatdifferentages.Thehistogramsin
Figure10showthedistributionsofRTsonbothdistracter-andtarget-initialtrialsfromastudyoffamiliarwordrecognitionby18-and21-month-olds(Fernald
etal.2001).Infantsshiftedtothetargetbeforetheendofthetrialon88%ofthe
distracter-initialtrials;on12%ofthesetrialstheynevershiftedatall.Notethatthe
firstshiftsonsixtrialsfellbelowthelowercutoffpointof333ms,andthuswere
excludedfromtheRTanalysis.Inthisdatasetwedecidedon1800msastheupper
boundaryforshiftstobeincludedintheRTanalysis.Thiscutoffwaschosento
eliminateresponsesinfluencedbythesecondrepetitionofthenoun(sinceinthis
particularstimulusseteachtargetwordwasspokentwicepertrial),anditexcludedoutliersmorethan2SDgreaterthanthemeanofthedistribution.Forpurposes
ofcomparison,thedistributionofresponselatenciesonthe200target-initialtrials
inthisstudyisalsoshowninFigure10B.Althoughinfantsshiftedrandomlytothe
distracteronsometrialsasthetargetwordwasspoken,morethanhalftheinfants
didnotshiftatall,maintainingfixationonthecorrectpicture.
Thedeterminationofappropriatecutoffpointsforexcludingtrialsfroman
RTanalysisisanimportantdecisionthatmayvarysomewhatfromstudytostudy,
dependingontheexperimentalquestionandtheageofthechildreninthestudy.
Thelowercutoffisrelativelyconstrained;ineyetrackingstudieswithparticipants
ofdifferentages,thiscutoffhasvariedfrom200400ms(Bailey&Plunkett2002;
Ballem&Plunkett2005),withshorterintervalstypicallyusedwithadults(e.g.,
Tanenhaus,Magnuson,Dahan,&Chambers2000)andwithchildrenolderthan
24months(Fernaldetal.2006;Zangl&Fernald2007).Establishingtheupper
cutoffisnotalwaysasstraightforward,althoughonereasonableapproachisto
identifyoutliersbyexaminingthedistributionofshiftsforoutliers,asshownin
Figure10.
Oneadditionalissueofconcernincalculatingresponselatenciesistheproblemofsparsedata,alwaysariskininfantstudiesthathaveveryfewtrials.Itis
importanttokeepinmindthatRTscanonlybecalculatedonthosetrialswhere
Lookingwhilelistening 127
Distracter-to-Target Shifts
A.
82%
No Shift
12%
6%
8
Number
of
Trials
4
333
667
B.
2333
2667
Target-to-Distracter Shifts
Shifts < 1800 ms
36%
No Shift
55%
333
667
2667
thechildhappenstostartoutonthedistracterandthenshiftstothetargetwithin
theRT-window(e.g.,3001800msfromtarget-wordonset),asubsetofthedata
thatoftenincludesfewerthanhalfthetotalnumberoftrials.Forthisreason,in
experimentaldesignswithtwoormorewithin-subjectconditions,itcaneasily
happenthatnotallchildrencontributeRTdataandthusmustbeexcludedfrom
someanalyses.Forexample,inahypotheticalexperimentwith20criticaltrials,
10ineachcondition,eachchildwillonaveragehaveeightuseabletrials,withtwo
trialscodedasoff orawayattarget-wordonset.Oftheeightuseabletrialsin
each condition, some children may have six distracter-initial trials in each, all
withshiftsthatoccurwithintheappropriateRT-window.Inthiscasethemean
RTineachconditionwouldbebasedonsixtrialsnotanimpressivenumberby
thestandardsofstudieswithadults,butsubstantialforRTstudieswithinfants.
However,themorelikelyscenarioisthatchildrenbychancewillhaveonlyfour
128 AnneFernaldetal.
distracter-initialtrialsineachcondition,andsomeofthesewillbetoofastortoo
slowtobeincludedintheRTanalysis.Andofcoursesomechildrenbychance
willhaveevenfewerdistracter-to-targetshifts,andthusmayendupwithnoRTs
tocontributetotheanalysisatall.Inearlierstudieswehavesometimeshadtouse
meanRTsthatwerebasedononlytwotrialspercondition,butthislowcriterion
canresultinverynoisydataandworksagainstfindingpositiveresults.Toavoid
this disappointing outcome, when designing an experiment in which latency
measuresarecritical,itisimportanttomakeeveryefforttomaximizethenumberofpotentialRTtrials.Thismeansincludingnomorethantwowithin-subjects
factors,sinceeachadditionalfactorreducesthenumberofdistracter-initialtrials
ineachconditionthatwillpotentiallyyieldRTs.Anotherapproachistodouble
the overall number of trials by observing each participant in two separate sessions,scheduledadayorsoapart.
3.2.2 Accuracy
Accuracyreflectshowreliablychildrenlookatthecorrectreferent,operationalizedasthemeantimespentlookingatthetargetpictureasaproportionoftotal
timespentoneitherthetargetorthedistracterpicture,averagedoveraparticular
regionofinterest.WhiletheRTanalysisisbasedonlyondistracter-initialtrials,
accuracyincludesbothtargetanddistracter-initialtrials,assessinglookingtimeto
thereferentregardlessofwhetherthechildstartedoutonthetargetpictureorhad
toshifttothetargetfromthedistracter.Dependingontheexperimentalquestion,
accuracymaybemeasuredacrossasinglebroadtimewindow,orovermultiple
smallertimewindows.Forexample,whenassessingdevelopmentalchangesininfantsaccuracyinrecognizingfamiliarwordsinsimplesentenceframes(Wheres
the doggy?), accuracy was calculated as the mean proportion of looking to the
targetoverthebroadwindowextendingfrom300to1800msfromtheonsetof
thetarget-word.Thiswouldbeequivalenttotheareaunderthecurvesoverthat
windowintheprofileplotshowninFigure6.
However,amultiple-windowanalysiswasmoreappropriateforthestudydescribedearlieronchildrensuseofsemanticinformationfromtheverbtoidentify
thereferent(Fernald2004).AsshownintheprofileplotinFigure8,wedefined
fourregionsofinterest:1:verb,2:determiner,3:noun,4:post-noun.Notethat
themeasurementwindowscorrespondingtothesefourregionsincorporatethe
estimated300msassumedtobenecessaryforprocessingtheinitialspeechsegments and mobilizing an eye movement; thus the verb window begins 300 ms
intotheverbandextends300msbeyondtheoffsetofthisword.Ourprediction
wasthattheaccuracycurvesonrelated-frameandunrelated-frametrialswould
begintodivergeattheendoftheverbwindow,withsignificantdifferencesemergingwithinthenextfewhundredmillisecondsaschildrenmadeuseofinforma-
Lookingwhilelistening 129
tionintheverbtofindthetargetpicture,evenbeforetheyheardthenoun.The
accuracyanalysisofthesedataispresentedinFigure11,showingthemeanproportionsoflookingtimetothetargetpicturecalculatedovereachofthefourtime
windows.Aspredicted,thedifferenceinaccuracybetweenrelatedandunrelated
trialswassignificantinboththedeterminerandthenounwindows,indicating
thatchildrencouldidentifythecorrectreferentusingsemanticinformationfrom
therelatedverb.Onrelated-frametrials,childrenwerealreadyfixatingtheappropriatereferent75%ofthetime,onaverage,bythebeginningofthetargetnoun;
onunrelated-frametrials,incontrast,themeanproportionoflookingtimetothe
correctpicturereached75%onlyaftertheendthenoun.
130 AnneFernaldetal.
Fernald2007)andmorecomplexsentencestructure(Thorpe&Fernaldunderreview)alsoimprovesdramaticallyoverthisperiod.Theseinvestigationsofon-line
processingefficiencybyveryyounglanguagelearnersprovidenewinsightsinto
theearlydevelopmentofreceptivelanguagecompetence,complementingresults
fromstudiesoflexicalandgrammaticalgrowththatarebasedonmoretraditional
measuresofspeechproductionoverthefirstthreeyears.
Butcharacterizingtypicalpatternsoflanguagegrowthovertimeisjustone
perspective in developmental research; another central goal is to characterize
variationamongchildren.Forexample,youngchildrenvarywidelyinthesizeof
theirproductivevocabulary,andone15-month-oldmayproducemorethan50
wordswhileanotherhasnotyetstartedtospeakatall,withconsiderablevariabilityapparentingrammaticalaswellaslexicalgrowthovertheearlyyears(Bates,
Dale,&Thal1994).Researchusingon-lineprocessingmeasuresoflanguageunderstandingcanaddressimportantquestionsaboutdifferencesamongchildren
withinanagegroup,aswellasbetween-groupdifferences.Isspeedofprocessing
at any given age a stable measure for an individual child? That is, are children
whorespondmorequicklyonaverageinidentifyingfamiliarwordsat18months,
relativetothemeanRTforchildrenatthatage,thesamechildrenwhorespond
relativelymorequicklyatlaterages?Andhowdoindividualdifferencesinefficiencyofspokenlanguageprocessingrelatetoindividualdifferencesinlanguage
growth,asassessedbystandardmeasuresoflexicalandgrammaticalknowledge?
Inparticular,doesprocessingefficiencyininfancypredictlanguageandcognitive
outcomesatlaterages?
To begin to address these questions, we conducted a longitudinal study of
59 English-learning infants, testing them in the looking-while-listening procedureat15,18,21,and25monthsofage(Fernaldetal.2006).Childrensspeed
andaccuracyinspokenwordrecognitionincreasedsignificantlyoverthisperiod,
consistentwithearliercross-sectionalresearch.Toexploretherelationofon-line
measuresofspeechprocessingskilltomoretraditionalmeasuresoflinguisticdevelopment,parentalreportsofvocabularyandgrammaticalusageweregathered
atfivetimepointsacrossthesecondyear,alongwithastandardizedtestoflexical
knowledgeat25months.Speedandaccuracyinspeechprocessingat25months
wererobustlyrelatedtolexicalandgrammaticaldevelopmentacrossarangeof
measuresfrom12to25months.Analysesofgrowthcurvesrevealedthatchildren
whowererelativelyfasterandmoreaccurateinspokenwordrecognitionat25
months were also those who had experienced faster and more accelerated vocabularygrowthacrossthesecondyear.
Thesefindingsledtotheobviousnextquestion:towhatextentdoindividual
differencesinprocessingefficiencyininfancypredictlaterlanguageandcognitiveoutcomes?Inarecentfollow-upstudy(Marchman&Fernaldunderreview),
Lookingwhilelistening 131
30ofthechildrenfromtheFernaldetal.(2006)longitudinalstudyweretestedat
theageof8yearsontheKaufman Assessment Battery for Children, Second Edition
(KABC-II)andtheClinical Evaluation of Language Fundamentals, Fourth Edition
(CELF-4) standardizedassessmentsofcognitiveandlanguageskills.Multipleregressionanalyseswereusedtoevaluatethelong-termpredictivevalidityoftwo
measuresininfancyexpressivevocabularyandmeanRTat25monthsinrelation
toschool-ageoutcomes.Inlightofthepresumedlinksbetweenefficiencyofspokenlanguagecomprehensionandworkingmemoryinolderchildrenandadults,
wealsoexaminedrelationsbetweenprocessingspeedininfancyandperformance
workingmemory.MeanRTinthelooking-while-listeningtaskat25monthswas
significantlycorrelatedwithscoresonCELF-4(r=.52to.55)andKABC-II
(r=.43to.70)at8years,takingvocabularysizeintoaccount,and relations
were strongest to performance on the working memory subscale. Moreover, it
was childrens speed in identifying the target word on challenging trial types,
requiring them to integrate semantic or morphosyntactic cues, that accounted
forthemostvariance;meanRTtofamiliarwordsinsimplerframesdidnotadd
predictive power. This prospective longitudinal study is the first to reveal the
long-termpredictivevalidityofon-linemeasuresofprocessingefficiencybyvery
younglanguagelearners,showingthatindividualdifferencesintheefficiencyof
spokenlanguageinterpretationattheageoftwoyearspredictchildrenssuccess
incognitiveandlanguagetasksinlaterchildhood.
.
Conclusions
Itisfascinatingandrevealingtowatchinfantslookattheworldastheylistento
speechinacarefullycontrolledexperimentalcontext.Theirgazepatternsprovide
awindowontheirreferentialdecisions,astheyseekmeaninginthewordsthey
arehearinginrelationtotheobjectstheyarelookingat,allwithinfractionsofa
second.Agreatdealofresearchonearlylanguagedevelopmentaimstocharacterizewhatwordschildrenknowataparticularage,asifwordswereacquiredin
anall-or-nonefashion;however,themethodsandresultswehavedescribedhere
takeadifferentperspective,focusingonthegradualdevelopmentofchildrensefficiencyinusingtheiremerginglexicalknowledgetointerpretspokenlanguage.
Accordingtothisview,iftheratherstaticnotionofacquisitionisappropriateto
lexicaldevelopmentatall,thenlearningtomakesenseofaspokenwordislikeacquiringaskillratherthanacquiringathing,withanemphasisongradualmastery
ratherthanonpossession.Infantsmayrespondtomoreandmorewordsoverthe
secondyear,buttheyalsolearntorespondwithincreasingspeedandefficiencyto
132 AnneFernaldetal.
eachofthewordstheyarelearning,andtorecognizethesewordsinmorediverse
andchallengingcontexts.
The looking-while-listening methodology described here uses fine-grained
measures of the time course of childrens gaze patterns in response to speech
toexploretheearlydevelopmentoflanguageunderstanding.Ontheonehand,
thisprocedureistechnicallysimpleintermsofstimuluspresentation,similar
topreferential-lookingproceduresinthatitislowintaskdemandsanddoes
notrequireautomatedeyetrackingtechnology.However,thelooking-while-listening methodology differs critically from preferential-looking procedures in
termsofthequantitativemethodsusedfordatareductionandanalysis,yielding
high-resolutionreal-timemeasuresofspeechprocessingratherthanrelyingon
summarymeasuresoflookingpreference.Becausegazepatternsaretime-locked
tothespeechsignalandcodedframe-by-frameinthison-lineparadigm,each5minexperimentyieldsdatafromthousandsofframesaboutthechildsdynamic
responsetotheunfoldingsentence.Asdescribedinthischapter,themeticulous
proceduresinvolvedinthecollection,reduction,andmultiplelevelsofanalysis
ofsuchdetaileddataarecertainlynotsimple.Buttheyarewellworththeeffort,
revealingadynamicandnuancedpictureofyoungchildrensdevelopingskillin
findingmeaninginspeech.
Acknowledgements
ThankstoJonathanBerger,NereydaHurtado,GeraldMcRoberts,AmyPerfors,
JohnPinto,DanSwingley,andmanyotherstudentsandco-workerswhohaveall
contributed to the gradual refinement of the looking-while-listening paradigm
describedinthischapter.WearegratefultotheNationalInstituteofChildHealth
andHumanDevelopment(HD42235)fortheirgeneroussupportofourresearch.
Correspondenceshouldbeaddressedto:AnneFernald,DepartmentofPsychology,StanfordUniversity,Stanford,CA94305.
References
Altmann,G.T.M.&Kamide,Y.(1999).Incrementalinterpretationatverbs:Restrictingthe
domainofsubsequentreference. Cognition, 73(3),247264.
Aslin,R.N.(1981).Developmentofsmoothpursuitinhumaninfants.InD.F.Fisher,R.A.
Monty,&J.W.Senders(Eds.),Eyemovements: Cognition and visual perception.Hillsdale
NJ:LawrenceErlbaumAssociates.
Bailey,T.M.&Plunkett,K.(2002).Phonologicalspecificityinearlywords. Cognitive Development, 17,12651282.
Lookingwhilelistening 133
13 AnneFernaldetal.
Golinkoff,R.M.,Hirsh-Pasek,K.,Cauley,K.M.,&Gordon,L.(1987).Theeyeshaveit:Lexical
andsyntacticcomprehensioninanewparadigm.Journal of Child Language, 14,2345.
Grosjean,F.(1985).Therecognitionofwordsaftertheiracousticoffset:Evidenceandimplications.Perception and Psychophysics, 38(4),299310.
Haith,M.M.,Wentworth,N.,&Canfield,R.(1993).Theformationofexpectationsinearly
infancy.Advances in Infancy Research, 8,251297.
Haith,M.M.(1980).Rules that babies look by: The organization of newborn visual activity.PotomacMD:LawrenceErlbaumAssociates.
Halle,P.A.&deBoysson-Bardies,B.(1994).Emergenceofanearlyreceptivelexicon:Infants
recognitionofwords.Infant Behavior and Development, 17, 119129.
Henderson,J.M.&Ferreira,F.(Eds.).(2004).The interface of language, vision, and action: Eye
movements and the visual world.NewYorkNY:PsychologyPress.
Hollich, G., Hirsh-Pasek, K., & Golinkoff, R. M. (2000). Breaking the language barrier: An
emergentistcoalitionmodelofwordlearning.Monographs of the Society for Research in
Child Development,65(3),SerialNo.123.
Hood,B.M.&Atkinson,J.(1993).Disengagingvisualattentionintheinfantandadult.Infant
Behavior & Development, 16,405422.
Hurtado,N.,Marchman,V.A.,&Fernald,A.(2007).SpokenwordrecognitionbyLatinochildrenlearningSpanishastheirfirstlanguage.Journal of Child Language, 34(2),227249.
Juszcyk,P.W.(1997).The discovery of spoken language. CambridgeMA:TheMITPress.
Katz,N.,Baker,E.,&McNamara,J.(1974).Whatsinaname?Astudyofhowchildrenlearn
commonandpropernames.Child Development, 45,469473.
Knoeferle,P.&Crocker,M.W.(2006).Thecoordinatedinterplayofscene,utterance,andworld
knowledge:Evidencefromeyetracking.Cognitive Science,30,481529.
Kuhl,P.K.(2004).Earlylanguageacquisition:Crackingthespeechcode.Nature Reviews Neuroscience, 5(11),831841.
Lew-Williams,C.&Fernald,A.(2007).YoungchildrenlearningSpanishmaderapiduseofthe
grammaticalgenderinspokenwordrecognition.Psychological Science,18, 193198.
Lewis,M.M.(1936).Infant speech: A study of the beginnings of language.London:Routledge
&KeganPaul.
Marchman,V.A.&Fernald,A.(2007).Speedoflanguageprocessingandvocabularyknowledgeininfancypredictschool-agecognitiveandlanguageoutcomes.Manuscriptsubmittedforpublication.
Markman, E. M. (1989). Categorization and naming in children. Cambridge MA: The MIT
Press.
Marslen-Wilson, W. & Zwitserlood, P. (1989). Accessing spoken words: The importance of
wordonsets.Journal of Experimental Psychology: Human Perception and Performance,15,
576585.
Meints, K., Plunkett, K., & Harris, P. (2002). What is on and under for 15, 18- and 24month-olds?Typicalityeffectsinearlycomprehensionofspatialprepositions.British Journal of Developmental Psychology, 20,113130.
Naigles,L.R.(2002).Formiseasy,meaningishard:Resolvingaparadoxinearlychildlanguage.Cognition,86,157199.
Portillo,A.L.,Mika,S.,&Fernald,A.(2007).Interpretingbehaviorsinpreferentiallooking
procedures:Dochildrenseekaword/objectmatch,orfleeamismatch,orboth?Presented
attheBiennialMeetingoftheSocietyforResearchinChildDevelopment,Boston,March
29April12007.
Lookingwhilelistening 13
Reznick,J.S.(1990).Visualpreferenceasatestofinfantwordcomprehension.Applied Psycholinguistics,11,145166.
Saffran, J. (2002). Constraints on statistical language learning. Journal of Memory and Language,47, 172196.
Shafer, G. & Plunkett, K. (1998). Rapid word learning by 15-month-olds under tightly controlledconditions.Child Development, 69,309320.
Shipley,E.F.,Smith,C.S.,&Gleitman,L.(1969).Astudyoftheacquisitionoflanguage:Free
responsestocommands.Language, 45, 322342.
Snedeker,J.&Trueswell,J.C.(2004).Thedevelopingconstraintsonparsingdecisions:Therole
oflexical-biasesandreferentialscenesinchildandadultsentenceprocessing.Cognitive
Psychology, 49(3),238299.
Spelke, E. S. (1976). Infants intermodal perception of events. Cognitive Psychology, 8, 553
560.
Swingley,D.,Pinto,J.,&Fernald,A.(1999).Continuousprocessinginwordrecognitionat24
months.Cognition, 71(2),73108.
Tanenhaus,M.K.,Magnuson,J.S.,Dahan,D.,&Chambers,C.G.(2000).Eyemovements&
lexicalaccessinspoken-languagecomprehension:Evaluatingalinkinghypothesisbetween
fixationsandlinguisticprocessing.Journal of Psycholinguistic Research, 29(6),557580.
Tanenhaus,M.K.,Spivey-Knowlton,M.J.,Eberhard,K.M.,&Sedivy,J.C.(1995).Integrationofvisualandlinguisticinformationinspokenlanguagecomprehension.Science, 268,
16321634.
Thomas,D.G.,Campos,J.J.,Shucard,D.W.,Ramsay,D.S.,&Shucard,J.(1981).Semantic
comprehensionininfancy:Asignaldetectionanalysis.Child Development, 52, 798803.
Thorpe, K. & Fernald, A. (2006). Knowing what a novel word is not: Two-year-olds listen
throughambiguousadjectivesinfluentspeech.Cognition, 100, 89433.
Thorpe,K.&Fernald,A.(2007). Developingefficiencyinon-lineinterpretationofadjectivenounphrases.Manuscriptsubmittedforpublication.
Tomasello,M.(2000).Doyoungchildrenhavesyntacticcompetence?Cognition, 74, 209253.
Trueswell,J.C.&Tanenhaus,M.K.(Eds.).(2005).Approaches to studying world-situated language use: Bridging the language-as-product and language-as-action traditions.Cambridge
MA:TheMITPress.
Trueswell,J.C.,Sekerina,I.,Hill,N.,&Logrip,M.(1999).Thekindergarten-patheffect:Studyingon-linesentenceprocessinginyoungchildren.Cognition,73,89134.
Walley,A.C.(1993).Theroleofvocabularydevelopmentinchildrensspokenwordrecognition
andsegmentationability.Developmental Review,13,286350.
Werker,J.F.&Tees,R.C.(1984).Cross-languagespeechperception:Evidenceforperceptual
reorganizationduringthefirstyearoflife.Infant Behavior and Development,7,4963.
Zangl,R.&Fernald,A.(2007).Increasingflexibilityinchildrenson-lineprocessingofgrammaticalandnoncedeterminersinfluentspeech.Language Learning and Development, 3,
199231.
Zangl,R.,Klarman,L.,Thal,D.,Fernald,A.,&Bates,E.(2005).Dynamicsofwordcomprehensionininfancy:Developmentsintiming,accuracy,andacousticdegradation.Journal of
Cognition and Development, 6,179208.
chapter5
Howdoyoungchildrenrepresentthestructureofanutterance?Dotheyemploy
abstractsyntacticcategories?Oraretheirrepresentationsmoreconcreteand
lexicallylimited?Ourrecentworkbringstogethertheworld-situatedeye-gaze
paradigmandsyntacticprimingtoexplorethesequestions.Webeginbyreviewingtheoriesofsyntacticdevelopmentanddescribingpreviousstudiesof
syntacticprimingduringchildrenslanguageproduction.Thenweintroduce
ourmethodforexploringprimingduringcomprehension.Nextwepresenta
seriesofexperimentsonpriminginadults,4-year-oldsand3-year-olds.Ineach
casetheparticipantsinterpretationisinfluencedbythestructureofpriorutterances,evenintheabsenceoflexicaloverlap.Weconcludethatyoungchildren
(andadults)employabstractsyntacticrepresentationsduringon-linesentence
comprehension.
1.
Introduction
In the past ten years researchers have made extensive use of on-line methods
toexplorewhatchildrenunderstandatdifferentages,howrapidlytheyunderstand it, and what kinds of information they use to reach this understanding.
Our recent work brings together two existing methods (syntactic priming and
theworld-situatedeye-gazeparadigm)toexploreadifferentquestion:whatisthe
nature and scope of young childrens grammatical representations and how do
theychangeoverdevelopment?Webeginbybrieflyreviewingalternatetheories
aboutthedevelopmentofsyntaxanditsmappingtosemantics.Nextwedescribe
thephenomenonofsyntacticprimingduringproduction,beforeintroducingour
methodforexploringprimingduringspokenlanguagecomprehension.Wethen
describe a series of experiments on syntactic priming from comprehension to
138 JesseSnedekerandMalathiThothathiri
comprehensioninadults,4-year-oldsand3-year-olds.Ineachcasewefindrobust
evidencethatparticipantsinterpretationofutterancesisinfluencedbythestructureofpriorutterancesevenintheabsenceoflexicaloverlap.Theseresultsindicatethatyoungchildren(andadults)employabstractstructuralrepresentations
duringon-linesentencecomprehension.Finally,wediscusshowthistechnique
canbeusedtoexplorethenatureoftheseabstractrepresentationsandtheirrelationtoindividuallexicalitems.
1.1
Theacquisitionofargumentstructure:
Earlyabstractionoritem-basedframes?
Aperennialquestioninlinguisticsishowtocharacterizetherelationsbetween
syntacticroles,semanticroles,andourknowledgeofspecificpredicates(seee.g.,
Baker 1988; Goldberg 1995; Grimshaw 1990; Jackendoff 2002). Any theory of
theserelationsmustaccountfortwofacts.First,therearesystematiclinkagesbetweenmeaningandsyntacticstructurethatarerobustacrossverbsandlanguages
(Fillmore1968;Baker1988;Dowty1991;Levin1993).Forexample,agentsofactionsgenerallysurfaceassubjects,andthemesasdirectobjects.Second,although
theselinkagesaresystematic,theydonotfullypredictthesyntacticpositionofan
argument.Similarpropositionscanbeexpressedusingdifferentsurfacesyntactic
forms,dependingontheverb,itsmorphologicalformandotherfactorssuchas
discourse structure and the phonological weight of constituents (1)(2). Since
muchofthisvariationdependsupontheverbintheutterance,allviabletheories
makeuseoflexically-specificinformation(thoughtheyvaryinwhetheritissyntacticorsemanticandhowitisused).
(1) a. Thepossiblehousingcollapsefrightenedtheyoungcouple.
b. Theyoungcouplefearedthepossiblehousingcollapse.
(2) a. Arielsentahalf-eatenpomegranatetoChris.
b. ArielsentChrisahalf-eatenpomegranate.
While every theory of language acquisition must acknowledge these two facts,
theoristsdifferinwhichtheyseeasprimary.Foronegroupoftheorists,therobustnessofthesyntactic-semanticcorrespondencesisseenasevidencethatthe
linkagesbetweenmeaningandstructureareinnatepropertiesofuniversalgrammarthatplayaroleinlanguageacquisition(Grimshaw1981;Pinker1984;Gleitman1990).Forexample,Pinkerssemanticbootstrappinghypothesis(1984)proposesthatchildrencometothetaskoflanguageacquisitionwithasetofthematic
roles(e.g.,agentandpatient),asetofsyntacticfunctions(e.g.,subjectanddirect
object),andsomedefaultrulesforlinkingonetotheother.
Whatlurksbeneath 139
10 JesseSnedekerandMalathiThothathiri
1.2
Novel-verbgeneralizationasawindow
ontogrammaticalrepresentations
Recentworkonthenatureofchildrensstructuralrepresentationshasfocusedon
theirabilitytocomprehendandproducesentenceswithnovelverbs.Animpressive number of production experiments has demonstrated that children under
3.5primarilyusenewverbsinwaysthatmimictheinput,failingtogeneralize
argumentstructurealternationsfromoneverbtoanother(seeTomasello2000
forareview).Forexample,Tomasello&Brooks(1998)presented2-year-oldchildrenwithanovelverbineitheratransitiveoranintransitiveconstruction(e.g.,
The puppy is meeking the ball orThe ball is meeking)andthenattemptedtoelicit
theunmodeledconstruction.Forexample,intransitiveswereelicitedbyasking
questionswhichplacedthethemeinsubjectposition,ensuringthatitwouldbe
giveninformationinanyresponse(What did the ball do?).Despitethisdiscourse
pressure,thechildrenusedtheconstructionthathadbeenmodeledbytheadult
almost90percentofthetime.Theauthorsconcludedthatchildrensearlyconstructionsareverb-specific,andthatabstractverb-generalconstructionsdevelop
graduallyduringthepreschoolyears.
Incontrast,severalnovel-verbcomprehensionstudieshavefoundevidencefor
abstractconstructionsinchildrenbetween20and36monthsofage.Thesestudiesexplorethisgeneralizationbytestingwhetheryoungchildrencanusebroad
semantic-syntacticmappingstointerpretthethematicrolesassignedbyanovel
verb.Forexample,Fisherandcolleagueshavefoundthatchildrenasyoungas20
monthssystematicallyprefertomaptransitivesentencestocausedmotionevents
(ratherthanself-generatedmotionevents)butshownosuchpreferenceforintransitivesentences(Fisher2002a;Yuan,Fisher,&Snedeker2007).Furtherevidence
comesfromchildrenscomprehensionofreversibletransitives.By21monthsof
age, children systematically interpret the subject of a transitive sentence with a
novelverbastheagentoftheaction(Gertner,Fisher,&Eisengart2006).
Howcanwereconciletheproductivitypresentinthesecomprehensionstudieswiththelackofgeneralizationobservedintheproductionstudies?Mostauthorssuggestthatonesetoffindingsrevealsthechildsworkadaygrammar,while
theotherreflectstask-specificstrategiesorlimitations.Forexample,Tomasello
and colleagues have suggested that the preferential-looking studies may reflect
emergingandincompleterepresentationsthatinitiallyplaylittleroleineveryday
comprehensionandproduction(see,e.g.,Savage,Lieven,Theakston,&Tomasello
2003).IncontrastFisher(2002b)hasarguedthatlowproductivityduringnovel
verbproductioncannotbetakenasevidencefortheabsenceofabstractrepresentations.Whetheraverbcanappearinaparticularargumentstructurealternationdependsonacomplexsetofsemanticconstraints(Levin1993).Forexample,
Whatlurksbeneath 11
thecausativealternation,usedbyTomaselloandBrooks(1998),isrestrictedto
verbsthatencodeanexternally-causedmannerofmotion.Consequently,evena
learnerwithabstractrepresentationsofargumentstructuremightbeunwillingto
extendnovelverbstounattestedconstructionsbecauseshelacksfullknowledge
ofthesemanticconstraintsonthealternationorisuncertainaboutthemeaning
oftheverb.Whilethechildrenintheseproductionstudiesclearlymaptheverbs
toappropriateevents,itisunclearhowpreciselytheirinterpretationoftheverb
matchestheonethattheexperimentershadinmind.Extractingthemeaningof
anovelverbfromavisualsceneisdifficultevenforadults(Gillette,Gleitman,
Gleitman,&Lederer1999;Snedeker&Gleitman2004).
Whilethenovelverbparadigmshavebeenextremelyinformative,theyhave
twolimitationswhichhaveledresearcherstoseekoutothermethods.First,these
paradigms necessarily explore childrens structural representations by probing
theirknowledgeofthematiclinkingrules.Togeneralizeanargumentstructure
alternationorinterpretasentencewithanovelverb,achildmustknowhowsemanticrolesaremappedontosyntacticpositions.Butaswenotedearlier,while
structuralabstractionsarenecessaryforadult-likelinkingrules,linkingrulesare
notnecessaryfor abstractions. Thus a paradigmwhichallowsustoinvestigate
abstractionswithoutrequiringknowledgeoflinkingruleswouldbevaluable.
Second,somehavequestionedtheconclusionsthatcanbedrawnfromnovelverbparadigms.Forexample,Ninio(2005)arguesthatchildrensabilitytouse
or interpret novel verbs in unattested constructions does not necessarily demonstratethattheyhavelinguisticrepresentationswhichareabstract.Insteadshe
suggeststhatchildrensgrammaticalknowledgeisstoredinaformatthatislexicallyspecificandconcrete.However,underunusualcircumstances,likethosein
the generalization studies, knowledge of one verb can be extended to another
throughaprocessofstructuralanalogy.
Theimpactofthisargumentdependsonourconceptionofanalogy.Consider
achildwhohearsThebunnyisgorpingtheduckwhilewatchingtwovideosin
whichoneactorpullsanotherbythefeet.Ifshetransfersstructuralknowledge
fromknownverbstothenovelverbbyvirtueofthestructureoftheutteranceor
thefactthattheyareallverbs,thenthisprocesswouldpositpreciselythekind
ofstructuralgeneralizationsthattheverbislandhypothesisdenies.However,it
would also be possible to form such an analogy without invoking higher-level
linguisticcategories.Childrencouldsimplytranslateorsubstituteanovelverb
withaknownformthathasthesameapparentmeaning(e.g.,gorpmeanspull).
Thiswouldallowthemtoapplyitem-specificknowledgewithoutinvokinglarger
generalizations.Thusknowingthatknowledgeistransferredfromaknownverb
toanovelverbdoesnottellusaboutthenatureoftherepresentationsthatunderliethistransferortheknowledgethatchildrendrawuponintheircomprehension
12 JesseSnedekerandMalathiThothathiri
andproductionofknownwords.Below,wediscusshowsyntacticprimingparadigmscanhelpshedlightontheseworkadaystructuralrepresentations.
2.
Inanelegantandextensiveseriesofexperiments,Bockhasdemonstratedthat,in
adultspeakers,prioruseofasyntacticconstructionalterstheprobabilitythatit
willbeusedagain(Bock1986;Bock1989;Bock&Loebell1990;Bock,Loebell,&
Morley1992).Forexample,adultswhohavejustproducedaprepositional-object
dative(POe.g.,The girl handed a paintbrush to the man)arerelativelymorelikely
todescribeanotherpictureusingasecondPOdative,whilethosewhohavejust
producedadouble-objectdative(DOe.g.,The girl handed the man a paintbrush)
arerelativelymorelikelytodescribethepicturewithaDOdative.This priming
isstructural inthatitoccursevenwhenthemeaningsoftheprimesentencesare
controlled and the prime and target sentences have no content words in common.Inadditiontopicturedescription,priminghasbeendemonstratedduring
writtensentencecompletion,spokensentencecompletion,anddialoguewitha
confederate(Branigan,Pickering,Stewart,&McLean2000;Branigan,Pickering,
&Cleland2000;Pickering&Branigan1998).Thisprimingisnotduetolexicalor
prosodicsimilarityalone(Bock&Loebell1990).Theprimingpersistsregardless
ofwhetherthespeakerproducestheprimeormerelyreadsorhearsit,demonstratingthattheseeffectsarisefromtheactivationofstructuralrepresentationsor
proceduresthatarecommontobothproductionandcomprehension.
This technique has several advantages for studying the nature of childrens
structuralrepresentations.First,itallowsustoexplorehowchildrenuseknown
verbsundercontrolledconditions.Whilenovel-verbtasksareaneffectivewayto
controlforpriorverb-specificlearning,interpretingtheirrelevanceforlanguage
processingmaynotbestraightforward.Asmentionedabove,successatanovelverbgeneralizationtaskiscompatiblewithrelianceonlexically-specificrepresentationsforcomprehensionandproduction,supplementedbyanalogicalproblemsolvingstrategies.Bylookingforabstractstructuralprimingwithknownverbs
(for which children potentially could possess adequate lexically-specific representations),wecanbetterevaluatetheimportanceofabstractrepresentationsin
childrenseverydaylanguageuse.
Second,thistechniqueallowsustocomparetherelativestrengthofabstract
andlexically-specificpriming.Wecanassessthisbycomparingtheprimingeffectsundertwoconditions:(1)whenthereisnolexicaloverlapbetweentheprime
andtarget(e.g.,prime:The boy threw his dog a ball,target:The teacher gave the
student new books);and(2)whentheprimeandtargetsentencessharethesame
Whatlurksbeneath 13
verb(e.g.,prime:The boy gave his dog a ball,target:The teacher gave the student
new books).Inadultsprimingisstrongerwhenthereislexicaloverlap(Pickering
&Branigan1998butseeKonopka&Bock2005).Thissuggestseitherthatadults
are using lexically-specific representations in addition to abstract ones, or that
lexicalitemsareconnectedtoabstractrepresentationsbylinkswhichcanthemselvesbeprimed(seeSection7.2).Bycomparingthestrengthofstructuralprimingbetweenverbsandstructuralprimingwithinverbs,wecanelucidatetherelationbetweenlexicalandabstractprocessesinchildrenandwecanexplorehow
bothprocesseschangeoverdevelopment.Thismayhelpreconcilethecompeting
findingsfromthenovel-verbproductionandcomprehensionstudies.
Structuralpriminghasonlyrecentlybeenusedtostudythenatureofyoung
childrens representations. We are aware of four published studies that explore
productionpriminginchildren.Oneexaminedtheproductionofnounphrases
in3-and4-year-oldsusingaconfederatedialogparadigm(Branigan,McLean,&
Jones2005).Thechildrenshowedstrongabstractprimingwhichwasfurtherenhancedwhentheprimeandtargetsharedthesameheadnoun.Whilethisstudy
suggestsdevelopmentalcontinuityinpriming,itdoesnotaddressthequestions
thatmotivatethecurrentwork.Thecontroversyinlanguageacquisitioncenters
onthestatusofverbsinthegrammarsofyoungchildren.Thepsychologicalreality of nouns is uncontentious (for example, the verb island hypothesis posits a
categoryofnounsthatfillintheslotsoflexically-specificverbalframes).1
Theremainingthreestudiesusedapicturedescriptionparadigm.Inthisparadigm,participantsareshownpicturesofsimplescenes.Duringtheprimetrials
the scene is described for the participant, who is typically asked to repeat this
description.Theprimetrialisimmediatelyfollowedbyatesttrialduringwhich
theparticipantissimplyshownapictureandaskedtodescribeit.Critically,both
theprimeandtestpicturesdepicteventswhichcouldbedescribedusingtwosyntacticallydistinctforms(e.g.,DOandPOdativesorthepassiveandactiveforms
ofthetransitive).
1. The confederate dialog task used by Branigan and colleagues (2005) may tap different
mechanismsthanthepicturedescriptiontasksdescribedbelow.Inadults,primingeffectsin
dialogparadigmsappeartoreflecthigher-levelprocessesfacilitatingcoordinationduringdialog, in addition to lower-level structural priming (Branigan et al. 2000; Pickering & Garrod
2004).ThetaskusedintheBraniganchildstudymayhavepromoteddirectcomparisonofthe
primeandtargetutterances,furtherencouragingparallelism.Oneverytrialtheexperimenter
andchildeachputdownacardwithacoloredobjectonit,theexperimenterdescribedhercard
(e.g.,the red catorthe cat that is red),thechilddescribedhis,andthenbothparticipantsraced
topickupthecardsiftheymatched.Theprimingeffectswerelargerthanthoseobservedin
parallelstudieswithadultsandpriorstudieswithchildren(82%matchestotheprimeinthe
absenceoflexicaloverlap).
1 JesseSnedekerandMalathiThothathiri
Thepicturedescriptionstudieswithchildrendifferfromthosewithadultsin
severalrespects.First,studieswithchildrentypicallyemployasmallnumberof
testitems.Whilethishelpstoensurethatthestudyisshortenoughtobecompletedbeforethechildsattentionwanders,itpreventstheexperimentersfromascertainingwhethertheobservedeffectsarerobustacrossitems.Second,forthesame
reason,fewifanyfilleritemsareused.Finally,thecriticalindependentvariables
areoftenmanipulatedbetweensubjects(ratherthanwithinsubject)orblocked
and presented in separate testing sessions. In adult studies, fillers and within
subjectmanipulationsareusedtodecreasetheprobabilitythatparticipantswill
becomeawareofthecriticalmanipulationanddevelopstrategiesspecifictothe
experimentalsituation.Giventhelimitedmetalinguisticabilitiesofpreschoolers
(Gombert1992),developmentalistsaretypicallynotconcernedaboutthispossibility.However,notethatbothofthesechangescouldincreasetheamountof
priming in the child studies. If all the primes are of the same type for a given
participantandthesetrialsareinterruptedbyfewornofilleritems,thenpriming
fromonetrialmaylingerandsummatewithprimingfromthenext.
Thefirststudytoexplicitlyexplorepriminginyoungchildrencomparedpassive and active transitive constructions (Savage et al. 2003). Children were assigned to either a high or a low lexical overlap condition. In the high overlap
condition,primesentencesusedpronounsthatcouldpotentiallyberepeatedin
the target descriptions (e.g., It got pushed by it). In the low overlap condition,
primesentencesusednounsthatcouldnotberepeatedinthetargetdescriptions
(e.g., The bricks got pushed by the digger). 6-year-olds showed priming in both
overlapconditions,but3-and4-year-oldsshowedpriminginthehighoverlap
conditiononly.Thustheauthorsconcludedthatwhile6-year-oldshaveabstract
representations,3-and4-year-oldsprimarilyrelyonlexically-specificrepresentationsinvolvingpronounsandsomegrammaticalmorphemes.Incontrast,Huttenlocherandcolleaguesfoundabstractstructuralprimingin4-and5-year-olds
forbothtransitiveanddativeconstructions(Huttenlocher,Vasilyeva,&Shimpi
2004).While,thesestudiesdifferedinseveralrespects,twofactorsseemparticularlyrelevant.Thefirstisthenumberoftimesthattheprimewasrepeated.The
childrenintheHuttenlocherstudyheardeachprimeutterancejustonce,while
those in the Savage study heard it four times. Repetition of a single utterance
mayprimarilyengagethemechanismsresponsibleforlexically-specificpriming
(e.g.,verbalmemoryseeKonopka&Bock2005),resultinginlittleornoabstract
priming. A subsequent study by Savage and colleagues provides some preliminarysupportforthispossibility(Savage,Lieven,Theakston,&Tomasello2006).
Theyfoundthatolder4-year-olds(meanage4;11)showedweakprimingeffects
whenasingleprimewasusedrepeatedly,butshowedrobustprimingwhenavarietyofprimeswereemployed.Thesecondcriticaldifferenceinthepriorstudies
Whatlurksbeneath 1
onabstractverbalprimingistheexactageoftheparticipants.The4-year-oldsin
theHuttenlocherstudywereolderthanthoseintheSavagestudy(meanageof4;8
ascomparedto4;2).Thusthediscrepancycouldbeexplainedifweassumethat
abstractproductionprimingemergessometimearoundfourandahalf.However,
recentunpublishedreportsofproductionpriminginyoungerchildrencomplicatethispicture.Gamez,Shimpi,andHuttenlocher(2005)foundnostructural
primingofdativesinapicturedescriptiontaskwith3.5-to4.5-year-olds,while
SongandFisher(2004)foundrobuststructuralprimingin3-year-oldchildren
usingasentenceimitationtask.
3.
Recentworkinourlabexploressyntacticprimingduringon-linecomprehension.
Sinceproductiontasksareoftenmoredifficultforchildrenthancomprehension
tasks(Hirsh-Pasek&Golinkoff1996),thismayprovideamoresensitivemeasureof
childrenslinguisticknowledgeandallowustotestyoungerchildren.Thesestudies
useaworld-situatedeye-gazeparadigmthattapson-linesentenceprocessing(the
visualworldparadigm).Wemeasureparticipantseyemovementswhiletheylistentoinstructionsandmanipulateobjects.Undersuchcircumstances,eyemovementstotheobjectsaretightlylinkedtotheunfoldingutterancesandaresensitivetolexicalandstructuralprocessinginbothadults(e.g.,Allopenna,Magnuson,
&Tanenhaus1998;Tanenhaus,Spivey-Knowlton,Eberhard,&Sedivy1995)and
children (Trueswell, Sekerina, Hill, & Logrip 1999; Snedeker & Trueswell 2004;
Snedeker&Yuaninpress;andTrueswellthisvolume).Byusingatechniquewith
goodtemporalresolution,wecanexplorethelocusoftheprimingeffectandrule
outalternateexplanationsthatmightapplytoprimingduringproduction.
3.1
Thepoormanseyetracker
Most researchers employing the visual world paradigm use head-mounted or
table-mountedeyetrackerstomeasurefixationpatterns(seeTrueswellthisvolume).Inourlabweuseamethodwecallthepoormanseyetracker,inwhicha
hiddencameraisusedtovideotapetheparticipantsdirectionofgaze.Theset-up
issimple.Theparticipantsitsinfrontofaninclinedpodiumwithfourshelves,
oneineachquadrant.Acameraisplacedbeneaththepodiumwithitslensaligned
withaholeinthecenterofthedisplay.Thecameraisfocusedontheparticipants
faceandisusedtorecordhisorhereyemovements,whicharelatercodedusing
frame-by-frameviewingonadigitalVCR.
16 JesseSnedekerandMalathiThothathiri
Thereareseveraladvantagestothismethod.First,wehavefoundthatmore
preschools,parents,andchildrenarewillingtoparticipateinresearchthatuses
familiarandnoninvasivetechnology.Second,thehiddencameratechniqueisfar
lessexpensivethaneyetracking,makingitaccessibletomoreresearchers.Allthat
is needed is a laptop computer with speakers, a small camera, a stage for displayingprops,andsomemethodforcodingvideotapeframe-by-frame(e.g.,an
editing deck with a jog-shuttle knob or a computerized coding system). Since
mostofthisequipmentisalreadypresentinthetypicallanguageacquisitionlab,
theparadigmcanbeeasilyadoptedbyexperimenterswhoarejustbeginningto
exploreon-linemethods.Third,theequipmentislight,compact,andveryeasy
tosetup,thustheparadigmdoesnotrequireadedicatedtestingroomandcan
bereadilytransportedtoschoolsordistantpopulations.Finallythepoormans
eyetrackeravoidssomeofthetechnicallimitationsofothereyetrackingmethods.
Becauseoftheirweightandfragility,head-mountedeyetrackersarenotsuitable
forchildrenunderaboutfourandahalf.Table-mountedeyetrackerscanbeused
withchildrenofanyage.However,mostcurrentmodelshavedifficultytracking
gazeifthechildsheadisinmotion.Consequently,theyareill-suitedforuseduringactouttasksandwithchildrenwhohavedifficultysittingstill.
Althoughnewtosentenceprocessing,thepoormanseyetrackerissimply
avariantofthepreferential-lookingparadigmswhicharewidelyappliedindevelopmentalpsychology(Fantz1961;Fagan1970;Spelke1979).Intermodalpreferential-lookingstudiestypicallyshowveryhighinter-coderreliabilityespecially
whenframe-by-framecodingisemployed(Hirsh-Pasek&Golinkoff1996).Many
oftheseintermodalstudieshavelookedatchildrenscomprehensionofspoken
language(Hirsh-Pasek&Golinkoff1996).Whenframe-by-framecodingissynchronizedwithaspeechstimulus,theparadigmisquitesimilartotheeyetracking
paradigmsusedinsentenceprocessing(seeFernald,Zangl,Portillo,&Marchman
thisvolume).Thesetechniqueshaveproventobesensitiveenoughtoexplorethe
resolutionofpronounsinpreschoolers(Song&Fisher2002)andimprovements
inthespeedofwordidentificationbetween15and24months(Swingley,Pinto,
&Fernald1999).
InourlabwerecordandcodeeyemovementsusingDVCAMequipmentand
tape stock. Unlike most formats DVCAM has audio-lock recording which ensuresthattheaudioandvideotracksremainsynchronized.Codingiscompleted
intwosteps.Thefirstcoderlistenstothetapewiththeaudioonandnotesthe
timeoftheonsetandoffsetofthesentence.Thesetimepointsareusedtodefine
theperiodduringwhichtheeyemovementswillbecodedandtosynchronizethe
eyemovementdatatothespeechstreamduringanalysis.Asecondcoderviews
thetapewiththeaudiooffandnotestheonsetofeachchangeingazeandthedirectionofthesubsequentfixation.Thedirectionofafixationiscodedasbeingin
Whatlurksbeneath 17
oneofthequadrants,atcenter,orawayfromthedisplay.Iftheparticipantseyes
areclosedornotvisible,theframeiscodedasmissingandthedataareexcluded
fromtheanalysis.Asubsetofthetapesisindependentlycodedbyanadditional
observerandinter-coderreliabilityisgenerallyhigh(>90%inthestudiespresentedbelow).
Tovalidatethismethod,SnedekerandTrueswell(2004)performedadirect
comparisonofdatacollectedwiththehiddencameraanddatacollectedwitha
head-mountedeyetracker.Theyfoundthatthetwomethodswerequitecomparable:theyconvergedonthesamefixationlocationfor93%ofthevideoframes
andproducedsimilaramountsoflostdata(23%offrames).Severalaspectsof
ourproceduremaybecriticaltoachievingthislevelofaccuracy.First,theroom
is well lit and the camera is tightly focused on the participants face, allowing
thecoderstoseetheirisandthusdetermineeyeposition.Second,participants
areplacedclosetothedisplayandtheirchairispositionedsothattheirgazeis
centeredatthelocationofthecamera.Thisensuresthatgazestoeachofthefour
quadrantscantypicallybedistinguishedbythedirectioninwhichtheeyesrotate
andnotmerelybytheextenttowhichtheydoso.Theimageoftheparticipants
faceonthehiddencameraismonitoredthroughouttheexperimenttoensurethat
theparticipantremainsproperlypositioned.Finally,toensurethatcodersreceive
frequentfeedbackabouttherelationbetweeneyepositionandgazedirection,we
elicitapredictablesequenceofgazesfromtheparticipantsatthebeginningof
eachtrialbylayingoutthepropsinaconsistentorder(clockwisefromtheupper
left)anddrawingtheirattentiontoeachone.
18 JesseSnedekerandMalathiThothathiri
3.2
Stimuli
Dativesarewell-suitedtoourpurposesforthreereasons.First,theyappeartobe
acquiredquiteearly;childrencomprehendandproducebothformsbyagethree
(Campbell & Tomasello 2001; Gropen, Pinker, Hollander, Goldberg, & Wilson
1989). Second, the two dative constructions differ primarily in their syntactic
structure and in the mappings between thematic roles and syntactic positions,
andonlyslightlyinmeaning(ifatall;seeBaker1997).Thus,primingusingdativesoffersareasonablyclearcaseofstructuralprimingindependentofsemantics.
Finally,dativesarecommonlyusedinstudiesofproductionpriminginadults,
facilitatingcomparisonsacrossexperiments.
Eachexperimentalblockconsistedof:(1)somefillersentences(whichwere
notdatives);(2)twoprimesentencesthatwereeitherDOorPOdatives(e.g.,DO:
Give the lion the ball;PO:Give the ball to the lion);and(3)afinaltargetsentence,
whichwasalsoaDOorPOdative(e.g.,DO:Bring the monkey the hat;PO:Bring
the money to the bear).Inallofourstudiesprimetypeandtargettypewerefully
crossedandmanipulatedbetweensubjects,resultinginfourpossibleconditions
(DOprime-DOtarget;DOprime-POtarget;POprime-DOtarget;POprime-PO
target).Eachparticipantinagivenstudywasrandomlyassignedtooneofthese
conditions.
Our goal was to determine whether DO and PO datives would prime the
interpretation of subsequent utterances that used a different verb and had no
common content words. To link this priming to eye movements we made use
ofawell-studiedphenomenoninwordrecognition,thecohorteffect(MarslenWilson&Welsh1978).Asaspokenwordunfolds,listenersactivatethelexical
itemsthatsharephonologicalfeatureswiththeportionofthewordthattheyhave
heard.Inthevisualworldparadigm,thisprocessresultsinfixationstothereferentsofwordsthatsharefeatureswiththetargetword(Allopennaetal.1998).
Theseeffectsareparticularlystrongatthebeginningofaword,whenallofthe
Whatlurksbeneath 19
phonological information is consistent with multiple words (the members of a
givencohort).Inourstudiesweusedprimingasatop-downconstraintwhich
mightmodulatetheactivationofdifferentmembersofaphonologicalcohort.
Ontargettrials,thesetoftoysthataccompaniedtheutterancecontainedtwo
itemsthatwerephonologicalmatchestotheinitialpartofthedirectobjectnoun
(seeFigure1).Onewasanimateandhenceapotentialrecipient(e.g.,amonkey)
whiletheotherwasinanimateandhenceamorelikelytheme(e.g.,money).Thus
theoverlapinwordonsets(e.g.,mon)createdalexicalambiguitywhichwas
tightlylinkedtoashort-livedambiguityintheargumentstructureoftheverb.We
expectedthatprimingoftheDOdativewouldleadtheparticipantstointerpret
thefirstnounasarecipient,resultinginmorelookstotheanimatematch,while
primingofthePOdativewouldleadparticipantstointerpretitasathemeresultinginmorelookstotheinanimatematch.Theinstructionswereprerecordedby
aspeakerwhousedanenthusiastictoneandslowdelivery.
3.3
Selectingadependentvariable
Dependentmeasuresthatarecommonlyusedineyetrackingstudiesincludefirst
gazeduration, latency,andtotal fixation time. Young childrensfirstlooksmay
notbereliablyguidedbymemoryforaparticularobjectinaparticularlocation
(Fernald,Thorpe,Hurtado,&Williams2006).Therefore,wedidnotcalculatefirst
gazedurationorlatency,andanalyzedtotalfixationtimeonly.Wewillreferto
totalfixationtimesimplyaslooks.Inmostofthesestudiesouranalysesfocused
ontheintervalduringthetargettrialsinwhichtheidentityofthedirectobject,
andhencetheargumentstructureoftheverb,wastemporarilyambiguous(e.g.,
mon).Withinthisinterval,wewereinterestedinlookstothepotentialanimate
recipient(e.g.,monkey)andthepotentialinanimatetheme(e.g.,money).Wewill
refertothesetwoitemsasanimal andobject respectively.Specifically,wewereinterestedinwhetherthetypeofprimesentenceinfluencedhowmuchthechildren
lookedtoeithertheanimalortheobject.Weexploredthreedifferentdependent
variables:(1)lookstotheanimalasaproportionofalllooks;(2)lookstotheobjectasproportionofalllooks;and(3)thedifferencebetweentheproportionof
lookstotheanimalandtheobject.
Becauseeyemovementsareinfluencedbyfactorsotherthantheonesmanipulated(e.g.,visualsalienceandnamefrequency;seeHenderson&Ferreira2004
fordiscussion),lookstooneofthetwoitems(animalorobject)maybehigher
thanlookstotheotherirrespectiveoftheexperimentalcondition.Thus,ceiling
orflooreffectsmightleadustofindsignificanteffectsforoneofthemeasures
butnottheothers.Inourpilotstudieswefoundthatanalysesofthedispreferred
10 JesseSnedekerandMalathiThothathiri
itemappearedtobemoresensitivetoprimingeffects.Forexample,ifparticipants
preferredtolookattheanimalirrespectiveoftheexperimentalcondition,then
wewouldfindreliableeffectsofprimingintheanalysisoflookstotheobjectbut
notintheanalysisoflookstotheanimal.Therefore,ourprimarymeasureinthe
studiesthatfollowwillbelookstothedispreferreditem.
.
Therewas,unfortunately,oneroadblocktousingcomprehensionprimingtoexplorethedevelopmentofstructuralgeneralizations.Theexistenceofthisformof
primingiscontroversialinadults,raisingthepossibilitythatprimingparadigms
are insensitive to structural representations in comprehension, or that abstract
syntaxplaysaweakerroleincomprehensionthanproduction(Townsend&Bever
2001).Critically,whileseveralpriorstudieshaveexploredstructuralprimingduringcomprehension,noneofthemprovideunambiguousevidenceofabstractsyntacticprimingforthekindsofsentencesthatweintendedtoexplore(post-verbal
ambiguitiesinargumentstructure).Twoofthestudiesthatfoundrobustpriming
effects(Luka&Barsalou2005;Noppeney&Price2004)usedstructureswhichwere
notsemanticallyequivalent(e.g.,relative-clauseattachmentambiguities),leaving
openthepossibilitythattheprimingeffectsweresemanticratherthansyntactic.
In addition, these studies used measures with a coarse temporal grain, creating
uncertaintyaboutwhethertheeffectswereduetoinitialstructuralanalysisorlater
reanalysis.Tworecenteyetrackingstudieshavesearchedforevidenceofpriming
duringon-linecomprehension.ScheepersandCrocker(2004)studiedtheprocessingofGermantransitivesentenceswithcasemarking.Theyfoundthattheonline interpretation of ambiguously marked preverbal arguments was influenced
bypriorunambiguouslymarkedprimesentences.Incontrast,usingtheEnglish
dativealternation,Arai,VanGompel,andScheepers(2007)lookedforprimingof
theinterpretationofpost-verbalarguments.Theyfoundprimingwhenprimeand
targetsentencescontainedthesameverb,butnotwhentheycontaineddifferent
verbs(seealsoBranigan,Pickering,&McLean2005).
Thus,whilethereisrobustevidenceforverb-specificprimingduringcomprehensionthescopeandstatusofabstractprimingisuncertain.Ourfirstexperimentrevisitedtheroleofabstractstructuralinformationduringadultlanguage
comprehension. We tested 28 undergraduates in a between-verb priming task.
Inthisstudy,weusedaprimingparadigminwhicheachsentencewasactedout
(act-outprimingtask,hereafter).Eachparticipantheardfourblocksofinstructions.Ineachblock,thefirsttwosentenceswerefillers(non-datives),thenexttwo
wereDOorPOprimedativesentences,andthelastwasatargetDOorPOda-
DO Prime
PO Prime
0.4
0.3
0.2
0.1
0
DO Target
PO Target
DO Prime
PO Prime
0.4
0.3
0.2
0.1
0
DO Target
PO Target
12 JesseSnedekerandMalathiThothathiri
persistswhenprimesentencesarenotenactedormappedontoavisualarray.If
primingpersistsundertheseconditionswecanruleoutthepossibilitythatactionsarecriticaltoproducingprimingduringcomprehension.
Participantspassivelylistenedtoastorythatcontainedtheprimesentences
andthenactedoutatargetsentenceusingasetofprops(storyprimingtask,hereafter).Specifically,theparticipantsweretoldthattheyweregoingtolistentotwo
voices(BobandSusan)whichwouldbepresentedonacomputer.Bobwouldtell
stories,whileSusanwouldgivetheminstructionstoactout.Atthebeginningof
eachtrial,Bobsvoicetalkedabouteventsthathappenedinachildrensstorethe
daybefore.SusansvoicetheninterruptedwithIts my turn. Are you ready?followed
bytheactualinstruction.Thefirstthreetrialswerepracticeitemsthatdidnotinvolvedatives,whilethesubsequenttrialsalternatedbetweencriticaltrialsandfiller
trials. On critical trials, Bobs last two sentences were DO or PO dative primes,
whileSusanstargetinstructionwasalsoaDOorPOdative.Therewereatotalof6
criticaltrialswithread, teach, sing, show, sell, andfeed astheprimeverbsand bring,
pass, throw, send, toss, andhand asthetargetverbs.AsFigure2illustrates,theeffectsofprimingpersistedinthestoryprimingtask,demonstratingthatpriming
occursevenwhentheprimeutteranceisnotmappedtoanactionoravisualdisplay.Againtherewasaweakpreferenceforlookstotheanimalandsoouranalyses
focusedonlookstotheobject.ParticipantswhohadheardPOprimesweremore
likelytolookattheobjectthanthosewhohadheardDOprimes.
The two experiments reported here demonstrate comprehension-to-comprehensionpriminginadultswhendifferentverbsareusedinprimeandtarget
sentences.Thesestudiesextendthefindingsofthepriorcomprehensionstudies
inseveralways.First,theydemonstratethatprimingoccursevenwhensemantically equivalent dative sentences are used, thus minimizing the possibility that
the effects are semantic rather than syntactic. Second, they show that priming
unfolds soon after the onset of the first noun, which was on average less than
550 ms after verb onset. This suggests that priming influences initial syntactic
analyses.Inaddition,theseresultscomplementScheepersandCrocker(2004)by
showing priming duringthe interpretation ofpost-verbal arguments.Abstract,
non-verb-specificinformationappearstoinfluencecomprehensionevenaftera
specificverbhasbeenencountered.
We attribute the divergence between our findings and those of Arai, Van
Gompel,andScheepers(2007)totwodifferencesbetweenthestudies.First,Arai
and colleagues used a single prime before each target trial while we used two.
Previousevidencesuggeststhatencounteringmultipleverbsinastructureleads
tostrongerstructuralpriming(Pickering&Branigan1998;Savageetal.2006).
Second,Araiandcolleaguespresentedthetargetsentenceimmediatelyafterthe
primewithnointerveningverbalmaterials,resultinginalagofapproximately
Whatlurksbeneath 13
1200 ms between prime and target trials. The lag in our studies was considerablylonger,intermsofbothinterveningutterancesandelapsedtime(approximatelyfivesentencesand3060secondsfortheact-outtaskandtwosentences
and 45 seconds for the story priming task). A recent study by Konopka and
Bock(2005)suggeststhatthedistancebetweentheprimeandthetargetaffects
therelativemagnitudeoflexically-specificprimingandabstractpriming.They
foundlexically-specificprimingonlywhenthetargetimmediatelyfollowedthe
prime.Incontrast,abstractprimingwasnumericallygreaterwhenasinglesentenceintervenedbetweenthetargetandprime(primingatlag1>primingatlag
0)andremainedrobustacrossasmanyasthreeinterveningsentences.Konopka
andBockattributelexically-specificprimingtoanexplicitmemoryfortheprime
sentencewhichdecaysrapidly.Abstractpriming,theyargue,involvesaformof
implicitlearning,paralleltothesettingofconnectionweightsinaneuralnetwork
(Chang,Dell,&Bock2006).Extendingthisproposaltocomprehensionpriming
generates the prediction that abstract priming would be greater in the present
experimentswhilelexically-specificprimingwouldbegreaterinthestudybyArai
andcolleagues.Systematicinvestigationoftheeffectoftheprime-to-targetlagon
comprehensionprimingwillberequiredtovalidatethisspeculativeaccount.
.
Armedwithaparadigmthatwassensitivetoabstractstructuralpriminginadults,
wesetouttodiscoverwhetherthesameparadigmcouldbeappliedtochildren.
Wefocusedourworkontwoagegroups:young4-year-oldsandyoung3-yearolds.Young4-year-oldswereofinterestbecausetheyhavefailedtoshowabstract
primingeffectsintwoproductionprimingstudies(Savageetal.2003;Gamezet
al.2005)despiteshowingfairlyrobustgeneralizationinnovel-verbproduction
tasks(Tomasello2000).Thisraisesthepossibilitythatnovel-verbgeneralization
paradigms may not reflect the structure that underlies everyday language use
(Ninio 2005). Young 3-year-olds were of interest because they typically fail to
generalizeinnovel-verbproductionstudies(Tomasello2000).Thusevidenceof
abstractpriminginthisagegroupwouldchallengetheempiricalbasisoftheverb
islandhypothesis.
Ofcourse,onecannotsimultaneouslytestthesensitivityofamethodandthe
existenceofthephenomenonthatitissupposedtobesensitiveto.Failuretofind
parallel effects of abstract structural priming in young children could indicate
eitherthatthetaskisnotappropriatetoexplorepriminginthisagegroup,orthat
childrenfailtoemployabstractrepresentationsinon-linecomprehension.Thus
we began by testing our task on an uncontroversial phenomenon: within-verb
priming in 4-year-olds. Within-verb priming can be mediated by abstract structure or lexically-specific representations. Given the results of the prior production
priming studies and the novel-verb generalization studies we would expect to find
robust priming within verbs in any task that is sensitive to priming of the relevant
representations.
5.1
Within-Verb Priming
0.5
0.4
DO Prime
PO Prime
0.3
0.2
0.1
0
DO Target
PO Target
Twenty young 4-year-olds (M = 4;1) participated in an act-out priming task, similar to the one we conducted with adults. All prime and all target sentences used
the verb give, which is the most frequent dative verb in the input to children and
in speech between adults. The temporary ambiguity in each target sentence was
created by using an animal as the recipient in the DO sentences and a compound
noun beginning with the same word as the theme in the PO sentences (e.g., DO:
Give the bird the dog bone; PO: Give the birdhouse to the sheep), resulting in a long
ambiguous region (400 ms).
The children performed the right action on 89% of the target trials. Twothirds of the errors were role reversals (e.g., giving the bird to the dog bone in response to Give the bird the dog bone). Most of these were in the mixed conditions,
where the prime type did not match the target type, suggesting that children were
sometimes led down the wrong path by the prime sentences. Across conditions,
children looked at the animal more than the object, thus our analysis focused on
looks to the object (Figure 3, left panel). Those children primed with PO sentences (where the first noun is the inanimate theme) looked more at the object than
those primed with DO sentences (where the first noun is the animate recipient).
Between-Verb Priming
0.5
DO Prime
0.4
PO Prime
0.3
0.2
0.1
0
DO Target
PO Target
Whatlurksbeneath 1
Thuswefoundthat4-year-oldchildrensinterpretationoftemporarilyambiguous
dative sentenceswasrapidlyinfluencedbytheprevioussentencesthattheyhad
heard.Becauseweusedthesameverbinbothprimeandtargetsentences,this
effectcouldreflecteitherverb-specificorabstractpriming.
To explore whether 4-year-olds have structural representations that are
broaderthanindividualverbs,weconductedaparallelstudyofacross-verbpriming.Theprimesentencesusedshowandbring, whilethetargetswerethesameas
thoseusedinthewithin-verbprimingstudy,andthususedtheverb give.Thirtyeightyoung4-year-oldsparticipated(M=4;0).Thechildrenperformedthecorrectactionon90%ofthetargettrials.Ouranalysisoflookstothedispreferred
item(theobject)revealedasignificanteffectofprime(Figure3,rightpanel).As
predicted,thoseprimedwithPOsentenceslookedmoreattheobjectthanthose
primedwithDOsentences.Thuswefoundthat4-year-oldchildrensinterpretationoftemporarilyambiguousgive sentenceswasinfluencedbythepreviousshow
orbring sentencesthattheyhadheard.Thisprimingacross verbsdemonstrates
that4-year-oldshavestructuralrepresentationsofdativeutterancesthatarenot
boundtoindividualverbs.
Therearehowever,twolimitationstothisfinding.First,whiletheseresults
clearlydemonstratebetween-verbprimingforthetargetverbgive,theycannot
telluswhetherthisprimingoccursacrossawiderrangeofdativeverbs.Theverb
give isunique:itisthemostfrequentdativeverbintheinputandundersome
theoriesithasaprivilegedroleintheacquisitionofdativeconstructions(Goldberg1995;Ninio1999).Second,althoughthefindingsoftheproductionpriming
taskshavebeenmixed,mosttheoristswouldagreethat4-year-oldsdemonstrate
somedegreeofabstractstructuralgeneralizationinnovel-verbproductiontasks.
Muchofthedebateaboutthenatureofchildrensrepresentationshascenteredon
3-year-olds.Thusitwasimperativetoextendthesefindingstoyoungerchildren
andtoawidervarietyofdativeverbs.
.2
Experimentswith3-year-olds
Tovalidateourtechniquewithayoungeragegroupwebeganbyexploringwithin-verbprimingusingtheact-outprimingtask.Thirtyyoung3-year-olds(M=
3;1)participated.Eachchildheardoneoftwolistseachcontainingfourdifferent
dativeverbs,butwithinagivenblock,theprimeandthetargetverbswerethe
same. The verbs were pass, send, throw, and bring for group 1 and hand, show,
toss, andtakeforgroup2.Thephonologicalambiguitiesintargetsentencesdid
notdependuponcompoundnouns,becausewewereunsurewhether3-year-olds
16 JesseSnedekerandMalathiThothathiri
wouldknowthem.Insteadweusedanimal/objectnamepairsthatoverlappedin
theironsets(e.g.,DO:Show the horse the book;PO:Show the horn to the dog).
Because the 3-year-olds were slower and more variable in their eye movementsthanthe4-year-olds,weaveragedlookingtimeoveralongertimewindow
inouranalyses.Whileminimumsaccadelatenciescanbeassmallas133msfor
adultsinasimplevisualtask(Matin,Shao,&Boff1993),thelatenciesforyoung
childreninataskwhereallstimulistayvisiblethroughoutthetrialandthereis
phonological overlap amongst the visible items are likely to be higher. For example,Swingley,Pinto,andFernald(1999)foundmeanlatenciesof558msand
785msforadultsand24-month-oldsrespectively.Inourexperiments,average
latenciestolookatthefirstmentioneditemontenrandomlyselected,unambiguousprimetrialswere983msforthe3-year-olds(SD=292ms)and437msfor
the4-year-olds(SD=188ms).Therefore,ouranalysesfor3-year-oldsbegan200
msaftertheonsetofthefirstnounbut(conservatively)extendedupto2seconds
afternounonset.Becausethiswindowislikelytoincludelooksthatwereprogrammedafterthefirstnounwasdisambiguated,wemightexpecttoseeaneffect
oftargettypeinadditiontoanyprimeeffects.
Childrenperformedthecorrectactionon79%ofthetargettrials.Mosterrors
(72%)wereduetochildrennotactingoutDOsentences(pickingupthetoysbut
notcarryingouttheaction).Acrossconditions,childrenshowedaslightpreferencefortheobjectovertheanimal.Ouranalysisoflookstothedispreferreditem
(theanimal)foundasignificanteffectofprime(Figure4,leftpanel).Aspredicted,thoseprimedwithDOsentenceslookedmoreattheanimal.Unsurprisingly,
therewasalsoasignificanteffectoftarget,reflectingthedisambiguationofthe
directobjectduringthistimewindow.Inadditiontherewasamarginalinteractionbetweenprimeandtarget,suggestingthattheprimingeffectwasstrongerfor
DOtargetsentences.
The error rate in 3-year-olds was considerably higher than in 4-year-olds.
Becauseeyemovementsonerrortrialsarehardtointerpret,weperformedasecondaryanalysisexcludingthosetrialswheretherewasanerrorintheaction.The
effectofprimetypeonlookstothedispreferreditem(theanimal)persistedinthis
analysis.Thuswefoundwithin-verbprimingin3-year-oldchildrenusingeight
differentdativeverbs.ThoseprimedwithDOsentenceslookedrelativelymoreat
theanimalthanthoseprimedwithPOsentences.Theseresultsdemonstratethat
primingisnotrestrictedtofrequent,prototypicaldativeverbssuchasgive.
Because this within-verb priming could arise from either verb-specific or
more abstract representations, our next experiment examined between-verb
priminginthispopulation.Thirty-twoyoung3-year-olds(M=3;1)participated.
Eachchildwasassignedtooneoftwostimuluslists.Inonelistpass, send, throw,
andbringappearedastargetswhilehand, show, toss, andtake appearedasprimes.
Within-Verb Priming
0.5
DO Prime
PO Prime
0.4
0.3
0.2
0.1
0
DO Target
PO Target
Between-Verb Priming
0.5
DO Prime
PO Prime
0.4
0.3
0.2
0.1
0
DO Target
PO Target
In the second list, the role of the verbs was reversed. Thus the target and prime
sentences were the same as those in the within-verb priming study but they were
simply paired differently across participants. The two prime sentences in each
block used two different verbs because varied primes have been found to lead to
greater production priming in older children (Savage et al. 2006).
Children performed the right action on 75% of the target trials. Our analysis
of looks to the dispreferred item (the animal) found a significant effect of prime
only. As predicted, those primed with DO sentences looked more at the animal
(Figure 4, right panel). In a secondary analysis we excluded all trials where children committed errors. The effect of prime on looks to the animal persisted. Thus
we found that 3-year-olds interpretation of target dative sentences was influenced
by the previous dative sentences that they had heard, even when the prime and
target sentences used different verbs and give did not appear as a prime or target.
The results reported here demonstrate within- and across-verb priming in
both 3- and 4-year-old children. The across-verb priming results can only be
explained by representations that are not verb-specific. Therefore, these results
suggest that both 3- and 4-year-old children use abstract representations during comprehension. This priming appears across a variety of verbs and in an age
group that shows limited productivity in many novel-verb generalization tasks
(see Tomasello 2000).
6.
What is the source of this priming effect? Our methodology rules out an alternate
explanation for previous production priming results found in children. Because
18 JesseSnedekerandMalathiThothathiri
thealternateconstructionsusedinthechildprimingstudiesaredistinguishedby
thepresenceorabsenceofclosed-classwordsormorphemes(to fordativesorby
and the participle for passives), production effects could reflect the priming of
thesewordsratherthangrammaticalstructures.Weavoidedthispossibilityby
measuringprimingoftheroleassignedtothedirect-objectnoun,whichprecedes
thiscriticalmorpheme.Thisisclearestforthe4-year-oldswhereourentiretime
windowofanalysisprecededtheonsetofto(witha200msoffset).However,even
forthe3-year-olds,differencesbetweentheDO-andPO-primeconditionsbegin
toemergepriortotheonsetofthismorpheme.
Nevertheless,severalalternateexplanationsoftheseprimingeffectsremain.
Thefirst,andtheleastinteresting,invokesnorepresentationofthephrasestructureoftheutteranceortheargumentstructureoftheverb.Perhapschildrenin
theDOprimeconditionssimplyformedtheexpectationthattheanimalwouldbe
mentionedfirstwhilechildreninthePOprimeconditionslearnedtoexpectthat
theobjectwouldbementionedfirst.Theremaininghypothesesallinvokesyntax
inonewayoranother.
Thesimplestofthesehypothesesisthatourmanipulationdirectlyprimedthe
syntacticstructuresusedindouble-object(VNPNP)andprepositional(VNP
PP)datives(Pickering&Branigan1998).Thesestructureswouldactivatethethematicrolesassociatedwiththem,whichinturnwouldactivateanimacyfeatures
associatedwiththoseroles,resultingintheobservedeyemovements.Alternately,
our priming manipulation could have targeted the mapping between thematic
rolesoranimacyfeaturesontheonehandandsyntacticpositionsontheother.For
example,ifthelocusoftheeffectwasthemappingofthematicroles,DOprimes
wouldpotentiatetherecipientdirectobjectmapping,whilePOprimeswould
potentiateathemedirectobjectmapping.Sincetherecipientandthemeroles
are in turn correlated with animacy (the recipient is usually animate, theme is
usuallyinanimate),thiswouldgiverisetothepatternofeyemovementsseenin
our experiments. Alternatively, direct mappings between animacy features and
syntacticpositions(e.g.,animatedirectobjectorinanimatedirectobject)
mayhavebeenprimed.Allthreeoftheseformsofpriminghavebeenfoundin
adults during sentence production (syntactic structures: Bock & Loebell 1990;
animacy mappings: Bock et al. 1992; thematic role mappings: Chang, Bock, &
Goldberg2003).
Todisentanglethesedifferentpossibilitieswewillhavetoexamineabroader
rangeofprimeandtargettypestodeterminewhichofthesefeaturesmustoverlap
forrobuststructuralpriming.Wehavebegunexploringthisissuewiththestory
primingtask.Thistaskhastheadvantageofallowingustouseprimesentences
thatarenotcommandsandcannotbeactedout.Ourcurrentworkexaminesthe
effectsoffourkindsofprimes(4a-d)ontheinterpretationofDOandPOdatives.
0.5
0.4
0.3
0.2
0.1
DO Datives
PO Datives
Intransitive
Controls
Locative
Prepositions
Prime Type
Figure 5. Story priming task with 4-year-olds. Effects of four prime types on the interpretation of dative sentences (PO and DO targets collapsed) as measured by the difference between the proportion of looks to the animate match and the proportion of looks
to the inanimate match (higher values are consistent with DO priming).
To date, thirty-two 4-year-olds (eight per prime type) have participated in this
study of the nature of between-verb priming.
(4)
a.
b.
c.
d.
The preliminary results are promising (Figure 5). The observed pattern for the
DO and PO primes replicates and extends the results of the act-out priming task.
Participants who hear DO-primes show a strong preference for the animal during
the region of ambiguity, while those who hear PO-primes have no strong preference for either the animal or the object. Thus dative priming in children persists
even when prime sentences are not mapped to an array or an action, and when
the prime and target sentences differ along several syntactic and semantic dimensions (e.g., primes in the present study have subjects, include indefinite NPs and
in many cases abstract themes).
The locative preposition primes offer a preliminary answer to our questions
about the nature of these priming effects. These utterances have the same mapping between animacy and position as the DO datives (animatedirect object
and inanimatesecond NP). However, they have the same syntactic structure
160 JesseSnedekerandMalathiThothathiri
as the PO datives (NP PP) and a similar pattern of thematic role assignments
(theme,goal/recipient).Inthisconditionthelooksduringtheambiguousregion
clearlypatternwiththePOprimes:thelocativeprimeincreaseslookstotheobjectrelativetotheintransitivecontrolresultinginadifferencescorethatisreliablysmallerthattheDOprimesbutindistinguishablefromthePOprimes(see
Bock&Loebell1990forparallelfindingsinadultproductionpriming).Thuswe
tentativelyconcludethatcomprehensionprimingofdativesinpreschoolersisnot
linked to animacy features but may be attributable to the priming of syntactic
framesorpatternsofthematicroleassignment.
7.
Conclusions
7.1
Comparisontopreviousstudies
Whatlurksbeneath 161
andabstractrepresentations.Aswediscussbelow,ourparadigmallowsustoestimatetherelativesizeofabstractprimingandlexically-specificpriming.Inboth
agegroupswefindrobustevidenceforabstractprimingintheformofreliable
between-verbprimingbutnoreliableevidenceoflexically-specificpriming(i.e.,
nointeractionbetweenprimetypeandwithin-/between-verbprimingineither
agegroup).Infact,theevidenceforlexically-specificprimingisparticularlyweak
in the 3-year-olds, where the difference in effect size between the within-verb
andbetween-verbprimingisnegligible.Second,thegradedstrengthhypothesisis
weakenedbyarecentstudydemonstratingthat3-year-oldscangeneralizethedativealternationinanovel-verbproduction task(Conwell&Demuth2007).Thus,
by3yearsofage,someabstractrepresentationsareclearlystrongenoughtoinfluencebothcomprehensionandproduction,suggestingtheneedforanalternate
explanationoftheco-existenceofitem-specificuseandabstractstructuralrepresentations.Below,wedescribehowstructuralprimingcanbeusedtoinvestigate
onesuchalternateexplanationwhichhasbeenwidelyacceptedbythosestudying
adultsentenceprocessing.
7.2
Usingstructuralprimingtoinvestigatechildrensrepresentations
Thestructuralprimingtechniqueofferspromiseforexploringthetheoreticaland
developmentalissuesraisedintheIntroduction.Theoreticalworkonargument
structurehasconsistentlyacknowledgedbothbroadsyntax-semanticscorrespondencesandtherolethatlexicalinformationplaysinthesyntacticrealizationof
event structure (Dowty 1991; Levin 1993; Jackendoff 2002). Developmentally,
thereisatensionbetweenevidenceforearlyabstractrepresentations(e.g.,Fisher
2002b) and item-specific use (e.g., Tomasello 2000). Studying the relation betweenchildrenslexicalrepresentationsandtheirabstractrepresentationsmaybe
afruitfulavenueforresolvingthesequestions.
Lexical-specificity and abstract syntax have long been accepted and reconciledintheoriesofadultsentencecomprehension.Thedatahaveleftuswithlittle
choice.Forexample,TrueswellandKim(1998)foundthatreadingtimesfortemporarilyambiguoussentencecomplementslike(5)wereaffectedbybriefexposures(39ms)toone-wordprimes.
(5) Thephotographeracceptedthefirecouldnotbeputout.
Exposuretoaverbthattypicallytakesasentencecomplement(e.g.,realize)facilitatedambiguityresolution,whileexposuretoaverbthattypicallytakesadirect
object(e.g.,obtain)hinderedit.Notethatthiseffectcanonlytakeplaceinarepresentationalsystemwhichisbothlexicallyspecific(differentverbshaddifferent
162 JesseSnedekerandMalathiThothathiri
impacts) and abstract (the structural biases of one verb affected processing of
another).Likemanyinthefield,theauthorsaccountedforthesefindingsbypositingthatindividualverbsareassociatedwithabstractstructuralrepresentations
whichcanbeprimed(seee.g.,MacDonald,Pearlmutter,&Seidenberg1994).The
strengthofthelinkbetweentheverbandastructuralnodedependsonlearners
priorexposuretothatparticularverbinthatparticularsyntacticcontext.
Modelsofthiskindprovideanexplanationfortheco-existenceofgeneralization and item-specificity in young children (see Fisher 2002b). Perhaps young
childrenhavethesamelinguisticarchitectureasadults,butsimplylackexperience. Perhaps like adults, they have abstract syntactic representations, abstract
semanticrepresentations,andmappingsbetweenthetwo.However,becausetheir
experiencewithindividualverbsislimitedtheconnectionsbetweensomeindividualverbsandsomestructuresmaybeweakerorevenabsent.Werethistrue,
wewouldexpectchildrentosucceedwhenknowledgeoftheconstructionalone
issufficienttosolvetheproblem.Thisisgenerallythecaseinnovel-verbpreferential-lookingstudies.Thestructureisprovidedandtheuseofgenerallinkingrules
issufficienttointerprettheutterancewithoutintegratingverb-specificinformation.However,whenthetaskrequireschildrentousetheconnectionbetweenthe
verbandthestructure,wewouldexpectthatperformancewoulddependupon
(1)thechildspriorexperiencewiththeverbinthatstructureand(2)theirexperienceoftheverbinalternatestructures.Novel-verbproductionstudiesputthe
childinpreciselythesituationwheresheisleastlikelytobeabletolinktheverb
tothenewstructure:thereisnopriorassociationbetweenthetwoandthereis
a strong association between the verb and an alternate structure. Known-verb
primingstudiesallowthechildtomakeuseofpreviouslyacquiredassociations
betweenspecificverbsandabstractstructures.
Similarmodelshavebeeninvokedtoexplainthedifferencesbetweenwithin- and across-verb priming in production studies. As we noted above, while
structuralprimingoccursevenwhenutterancessharenocontentwords,some
researchers have found that priming is greater when the same verb is used in
bothtargetandprime(Pickering&Branigan1998).PickeringandBraniganexplainthispatternwithatheoryinwhichindividualverbsarelinkedtoabstract
combinatorial representations such as [NP, NP] and [NP, PP]. These abstract
combinatorialnodesaresharedbetweenverbs,leadingtoacross-verbpriming.
Inaddition,thelinkbetweenanindividualverbandacombinatorialnodecanbe
potentiated,leadingtoanadvantageforwithin-verbpriming.
Toexplorewhetherourresultsaccordwiththispattern,wecomparedwithin-
andacross-verbprimingin3-and4-year-oldchildren.For3-year-olds,theeffect
sizesforwithin-andacross-verbprimingwerepartial2=.21andpartial2=.17,
respectively.For4-year-olds,thewithin-verbandacross-verbprimingeffectsizes
Whatlurksbeneath 163
werepartial2=.45andpartial2=.16respectively.Thus,forbothagegroups
within-verbprimingappearstobestrongerthanacross-verbpriming.However,
theinteractionbetweenthetwotypesofprimingwasnotsignificantforeither
group.
7.3
Finalwords
Thestudiespresentedinthischapterdemonstratethattheon-lineinterpretation
ofdativeutterancescanbestructurallyprimedbypriorcomprehensionofother
dativesentences.Thisprimingeffectisrobustlypresentinadults,4-year-olds,and
3-year-oldsandappearsregardlessofwhethertheprimesentenceisactedoutor
mappedontoavisualdisplay.Theseeffectsarenot,ornotsolely,lexicallyspecific.
They persist when different verbs and nouns are used in the prime and target
sentences.Thebetween-verbprimingeffectdemonstratesthatchildrenasyoung
as3yearsemployabstractrepresentationsduringthecomprehensionofsentences
withknownverbs.
Future studies can shed light on important questions that remain. These
include the precise nature of the representations that can be primed (semantic, syntactic,ormappings betweensyntax and semantics),andtheconstraints
onprimingbetweenverbs(isprimingrestrictedtoverbswithsimilardistribution,similarmeaning,orboth?).Critically,futureprimingstudiescanelucidate
whetheryoungchildrenlikeadults,havealanguageprocessingsysteminwhich
lexical and abstract representations interact to produce both item-specific and
generalizedpatternsofuse.
Acknowledgments
We thank Cynthia Fisher and Steven Pinker for helpful suggestions and comments.WearegratefultoSnehaRao,JaneSung,AdriannaSaada,MeganPowell
andAlisonYoungfortheirassistanceincollectingandcodingdata.Thiswork
wasfundedbyagrantfromtheNationalScienceFoundation(NSF-BCS0623945)
andwethankthemfortheirsupport.TheadultexperimentssummarizedinSection4ofthispaperappearasExperiments1and2inThothathiriandSnedeker
(inpress,a).Theexperimentswith3-and4-year-oldsfromSection5arereported
inThothathiriandSnedeker(inpress,b).ThedatareportedinSection6come
fromasubsetoftheparticipantstestedinThothathiriandSnedeker(2007),Experiment1.
16 JesseSnedekerandMalathiThothathiri
References
Allopenna,P.D.,Magnuson,J.S.,&Tanenhaus,M.K.(1998).Trackingthetimecourseofspokenwordrecognitionusingeyemovements:Evidenceforcontinuousmappingmodels.
Journal of Memory & Language,38,419439.
Arai,M.,VanGompel,R.P.G.,&Scheepers,C.(2007).Primingditransitivestructuresincomprehension.Cognitive Psychology, 54(3), 218250.
Baker,M.(1988).Incorporation: A theory of grammatical function changing.ChicagoIL:UniversityofChicagoPress.
Baker,M.(1997).Thematicrolesandsyntacticstructure.InL.Haegeman(Ed.),Elements of
grammar(pp.73137).Dordrecht:Kluwer.
Bock,K.(1986).Syntacticpersistenceinlanguageproduction.Cognitive Psychology,18,355
387.
Bock,K.(1989).Closedclassimmanenceinsentenceproduction.Cognition, 31,163186.
Bock,K.&Loebell,H.(1990).Framingsentences.Cognition,35,139.
Bock,K.,Loebell,H.,&Morey,R.(1992).Fromconceptualrolestostructuralrelations:Bridgingthesyntacticcleft.Psychological Review,99(1),150171.
Bowerman,M.(1973).Early syntactic development: A cross-linguistic study with special reference
to Finnish. Cambridge:CUP.
Braine,M.D.S.(1976).Childrensfirstwordcombinations.Monographs of the Society for Research in Child Development, 41(1,Serialno.164).
Branigan,H.P.,McLean,J.F.,&Jones,M.W.(2005).Abluecatoracatthatisblue?Evidence
forabstractsyntaxinyoungchildrensnounphrases.InA.Brugos,M.R.Clark-Cotton,&
S.Ha(Eds.),The proceedings of the 29th annual Boston University conference on language
development(pp.109121).SomervilleMA:Cascadilla.
Branigan,H.P.,Pickering,M.J.,Stewart,A.J.,&McLean,J.F.(2000).Syntacticprimingin
spoken production: Linguistic and temporal interference. Memory & Cognition, 28(8),
12971302.
Branigan,H.P.,Pickering,M.J.,&Cleland,A.A.(2000).Syntacticco-ordinationindialogue.
Cognition, 75, B1325.
Branigan,H.P.,Pickering,M.J.,&McLean,J.F.(2005).Primingprepositional-phraseattachment during comprehension. Journal of Experimental Psychology: Learning, Memory &
Cognition, 31(3),468481.
Campbell,A.L.&Tomasello,M.(2001).TheacquisitionofEnglishdativeconstructions.Applied Psycholinguistics, 22,253267.
Chang,F.,Bock,K.,&Goldberg,A.E.(2003).Canthematicrolesleavetracesoftheirplaces?
Cognition, 90, 2949.
Chang, F., Dell, G. S., & Bock, K. (2006). Becoming syntactic. Psychological Review, 113(2),
234272.
Conwell, E. & Demuth, K. (2007). Early syntactic productivity: Evidence from dative shift.
Cognition, 103(2),163179.
Dowty,D.(1991).Thematicproto-rolesandargumentselection.Language, 67(3),547619.
Fagan,J.F.(1970).Memoryintheinfant.Journal of Experimental Child Psychology, 9(2),217
226.
Fantz,R.L.(1961).Theoriginofformperception.Scientific American, 204, 6672.
Whatlurksbeneath 16
Fernald,A.,Zangl,R.,Thorpe,K.,Hurtado,N.,&Williams,C.(2006).Learningtolistenahead
inEnglishandSpanish:Infantsusemultiplelinguisticandnonlinguisticcuesinonline
sentenceinterpretation. PaperpresentedattheWorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
Fillmore,C.J.(1968).Thecaseforcase.InE.Bach&R.T.Harms(Eds.),Universals in linguistic
theory (pp.188).NewYorkNY:Holt,RinehartandWinston.
Fisher,C.(2002a).Structurallimitsonverbmapping:Theroleofabstractstructurein2.5-yearoldsinterpretationsofnovelverbs.Developmental Science, 5(1),5564.
Fisher,C.(2002b).Theroleofabstractsyntacticknowledgeinlanguageacquisition:Areplyto
Tomasello.Cognition, 82,259278.
Gamez,P.B.,Shimpi,P.M.,&Huttenlocher,J.(2005).Emergingsyntacticrepresentationsas
evidentthroughpriming.Posterpresentedatthe2005SRCDBiennialMeeting,Atlanta,
GA.
Gertner,Y.,Fisher,C.,&Eisengart,J.(2006).Learningwordsandrules:Abstractknowledgeof
wordorderinearlysentencecomprehension.Psychological Science, 17,684691.
Gillette,J.,Gleitman,H.,Gleitman,L.,&Lederer,A.(1999).Humansimulationsofvocabulary
learning.Cognition, 73,135176.
Gleitman,L.R.(1990).Thestructuralsourcesofwordmeaning.Language Acquisition,1,3
55.
Goldberg,A.(1995).Constructions: A construction grammar approach to argument structure.
ChicagoIL:UniversityofChicagoPress.
Goldberg,A.(2006).Constructions at work: The nature of generalization in language.Oxford:
OUP.
Gombert,J.E.(1992).Metalinguistic development.ChicagoIL:UniversityofChicagoPress.
Grimshaw,J.(1990).Argument structure.CambridgeMA:TheMITPress.
Grimshaw,J.(1981).Form,function,andthelanguageacquisitiondevice.InC.L.Baker&J.
McCarthy(Eds.), The logical problem of language acquisition(pp.165182). Cambridge
MA:TheMITPress.
Gropen,J.,Pinker,S.,Hollander,M.,Goldberg,R.,&Wilson,R.(1989).Thelearnabilityand
acquisitionofthedativealternationinEnglish.Language, 65,203257.
Henderson,J.M.&Ferreira,F.(Eds.).(2004).The interface of language, vision, and action: Eye
movements and the visual world. NewYorkNY:PsychologyPress.
Hirsh-Pasek,K.&Golinkoff,R.M.(1996).The origins of grammar: Evidence from early language comprehension.CambridgeMA:TheMITPress.
Huttenlocher,J.,Vasilyeva,M.,&Shimpi,P.(2004).Syntacticpriminginyoungchildren.Journal of Memory and Language, 50,182195.
Jackendoff,R.(2002).Foundations of language brain, meaning, grammar, evolution.Oxford:
OUP.
Konopka,A.&Bock,K.(2005).Helpingsyntaxout:Howmuchdowordsdo?Paperpresented
atthe18thAnnualCUNYSentenceProcessingConference,Tucson,AZ.
Levin,B.(1993).English verb classes and alternations.ChicagoIL:UniversityofChicagoPress.
Luka,B.J.&Barsalou,L.W.(2005).Structuralfacilitation:Mereexposureeffectsforgrammaticalacceptabilityasevidenceforsyntacticprimingincomprehension.Journal of Memory
and Language, 52,436459.
MacDonald,M.C.,Pearlmutter,N.J.,&Seidenberg,M.S.(1994).Thelexicalnatureofsyntacticambiguityresolution.Psychological Review, 101(4), 676703.
166 JesseSnedekerandMalathiThothathiri
Marslen-Wilson,W.D.&Welsh,A.(1978).Processinginteractionsandlexicalaccessduring
wordrecognitionincontinuousspeech.Cognitive Psychology, 10(1),2963.
Matin,E.,Shao,K.C.,&Boff,K.R.(1993).Saccadicoverhead:Information-processingtime
withandwithoutsaccades.Perception & Psychophysics, 53,372380.
Ninio,A.(2005).Testingtheroleofsemanticsimilarityinsyntacticdevelopment.Journal of
Child Language, 32,3561.
Ninio,A.(1999).Pathbreakingverbsinsyntacticdevelopmentandthequestionofprototypical
transitivity.Journal of Child Language, 26, 619653.
Noppeney,U.&Price,C.J.(2004).AnfMRIstudyofsyntacticadaptation.Journal of Cognitive
Neuroscience, 16,702713.
Pickering,M.J.&Branigan,H.P.(1998).Therepresentationofverbs:Evidencefromsyntactic
priminginlanguageproduction.Journal of Memory & Language,39,633651.
Pickering,M.J.&Garrod,S.(2004).Towardamechanisticpsychologyofdialogue.Behavioral
and Brain Sciences, 27(2),169226.
Pinker,S.(1984).Language learnability and language development.CambridgeMA:Harvard
UniversityPress.
Savage,C.,Lieven,E.,Theakston,A.,&Tomasello,M.(2003).Testingtheabstractnessofchildrenslinguisticrepresentations:Lexicalandstructuralprimingofsyntacticconstructions
inyoungchildren.Developmental Science, 6,557567.
Savage,C.,Lieven,E.,Theakston,A.,&Tomasello,M.(2006).Structuralprimingasimplicit
learninginlanguageacquisition.Language Learning & Development, 2(1),2749.
Scheepers,C.&Crocker,M.W.(2004).Constituentorderprimingfromreadingtolistening:
Avisual-worldstudy.InM.Carreiras&C.J.Clifton(Eds.),The on-line study of sentence
comprehension: Eyetracking, ERP and beyond(pp.167185).Hove:PsychologyPress.
Snedeker,J.&Gleitman,L.R.(2004).Whyitishardtolabelourconcepts.InD.G.Hall&S.R.
Waxman(Eds.),Weaving a lexicon(pp.255293).CambridgeMA:TheMITPress.
Snedeker,J.&Trueswell,J.C.(2004).Thedevelopingconstraintsonparsingdecisions:Therole
oflexical-biasesandreferentialscenesinchildandadultsentenceprocessing.Cognitive
Psychology, 49,238299.
Snedeker,J.&Yuan,S.(inpress).Effectsofprosodicandlexicalconstraintsonparsinginyoung
children(andadults).Journal of Memory and Language.
Song,H.&Fisher,C.(2002).Youngchildrenssensitivitytodiscoursecuesinon-linepronoun
interpretation.Proceedings of the 26th Annual Boston University Conference on Language
Development.SomervilleMA:Cascadilla.
Song,H.-J.&Fisher,C.(2004).Syntacticprimingin3-year-oldchildren.Paperpresentedatthe
29thAnnualBostonUniversityConferenceonLanguageDevelopment,Boston,MA.
Spelke,E.S.(1979).Perceivingbimodallyspecifiedeventsininfancy.Developmental Psychology, 15(6),626636.
Swingley,D.,Pinto,J.P.,&Fernald,A.(1999).Continuousprocessinginwordrecognitionat24
months.Cognition, 71, 73108.
Tanenhaus,M.K.,Spivey-Knowlton,M.J.,Eberhard,K.M.,&Sedivy,J.C.(1995).Integrationofvisualandlinguisticinformationinspokenlanguagecomprehension.Science, 268,
16321634.
Thothathiri,M.&Snedeker,J.(inpress,a).Giveandtake:Syntacticprimingduringspoken
languagecomprehension.Cognition.
Thothathiri,M.&Snedeker,J.(inpress,b).Syntacticprimingduringlanguagecomprehension
inthree-andfour-year-oldchildren.Journal of Memory and Language.
Whatlurksbeneath 167
Thothathiri,M.&Snedeker,J.(2007).Grammaticalpriminginpreschoolers:Anewmethodfor
tacklingoldquestionsinlanguageacquisition.Manuscript.
Tomasello,M.(1992).First verbs: A case study in early grammatical development.Cambridge:
CUP.
Tomasello,M.(2000).Theitem-basednatureofchildrensearlysyntacticdevelopment.Trends
in Cognitive Sciences, 4,156163.
Tomasello,M.&Abbot-Smith,K.(2002).Ataleoftwotheories:ResponsetoFisher.Cognition,
83,207214.
Tomasello, M. & Brooks, P. (1998). Young childrens earliest transitive and intransitive constructions.Cognitive Linguistics,9,379395.
Townsend, D. J. & Bever, T. G. (2001). Sentence comprehension. Cambridge MA: The MIT
Press.
Trueswell,J.C.&Kim,A.E.(1998).Howtopruneagarden-pathbynippingitinthebud:Fastprimingofverbargumentstructures.Journal of Memory and Language, 39,102123.
Trueswell, J., Sekerina, I., Hill, N. M., & Logrip, M. L. (1999). The kindergarten-path effect:
Studyingon-linesentenceprocessinginyoungchildren.Cognition,73,89134.
Yuan,S.,Fisher,C.,&Snedeker,J.(2007).Countingthenouns:Simplestructuralcuestoverb
meaning.Manuscriptinpreparation.
chapter6
170 HelenSmithCairns
1963).Knowledgethatwasnon-existentin1959waswell-developedin1969,yet
stillprimitivecomparedto2007.Wehaveexperiencedanexplosionofknowledge
about child language in four decades. My second point is to illustrate that our
beliefsshapeourexperiences.Lestthissounddangerouslypost-modern,Iassure
youthatIbelieveweweremuchclosertotruthin1969thanin1959,andin2007
than1969.Asbeliefshapesexperience,theorydrivesmethodology.Inempirical
researchwedecidewhatweneedtoexperienceinordertotestourhypotheses,
andwedevelopmethodologiesinordertodoso.
In Aspects of the theory of syntax, published in 1965, Chomsky introduced
the Competence/Performance distinction. Competence was the speaker-hearersknowledgeofhislanguage;performance,theactualuseoflanguageinreal
time.ToquoteChomsky:Linguistictheoryisconcernedprimarilywithanideal
speaker-hearerwhoisunaffectedbysuchgrammaticallyirrelevantconditions
asmemorylimitations,distractions,shiftsofattentionanderrorsinapplyinghis
knowledgeofthelanguageinactualperformance(p.3).HegoesontosayTo
studyactuallinguisticperformance,wemustconsidertheinteractionofavariety
offactors,ofwhichtheunderlyingcompetenceofthespeaker-hearerisonlyone
(p.4).Thus,fromthebeginningthereweretwoconceptionsoflinguisticperformance:idealizedperformancethatwouldperfectlyreflectcompetenceandactual
performancetakingplaceinrealtime.
Intheearlydaysourgoalwastodescribechildrensgrammars(competence),
andourprimarymethodwastoobservenaturalisticspeech.Diarystudiesgave
waytomoresystematicmodesofdatacollection.Inhis1970bookThe acquisition
of languageDavidMcNeillhadachapteronmethodologyinwhichheaddressed
production,comprehension,andelicitedimitation,butproductionwasprimary.
Theearlypioneers,Brown,McNeill,Bloom,Braine,Menyuk,Bowermanand
others, learned an enormous amount from analyzing naturalistic speech. They
describedtheamazingphenomenonofChildEnglish.Thatsthelanguagespoken
byeveryonewhoislearningEnglish,whentheyarebetweentheagesof(roughly)
1and3,yetfewofthosepeopleeverhearitspoken.ChildEnglishisalanguage
considerablydifferentfromAdultEnglish,yetwithhighlypredictableandwellunderstoodcharacteristics.
Manyvaluableanalysesoftheacquisitionofbasicsentencestructure,negation,movement,functionwords,laterthepro-dropphenomenon,andevidence
forparametersettingwerebasedprimarilyonproductiondata.Fieldworkhas
contributedvaluableinformationontheacquisitionoflanguagesotherthanEnglish,andCHILDESremainsanimportantresourceforacquisitionists.However,
inthoseearlydaysourmethodologyflowedfromourdesiretodescribetheregularities in childrens speech, believing (correctly) that what was systematically
produced could give us information about the childs underlying grammatical
Languageacquisitionresearch 171
knowledge.Furthermore,wesoughtcommonfeaturesofspeechacrosschildren,
believing(alsocorrectly)thatifmanydifferentchildrenproducedsimilarnonadultsyntacticandmorphologicalforms,thiswouldpointtosimilarmechanisms
ofacquisition.
We learned a lot, but we made a lot of mistakes, as well. For instance, we
thoughtthatyoungchildrendonotrepresentfunctionwordsintheirearlylexicons(orfunctionalcategoriesintheirsyntax)norboundmorphemesintheirmorphologies.Whencross-linguisticworkbegan,wediscoveredthatthisabsenceof
boundmorphemesispeculiartolightlyinflectedlanguageslikeEnglish.Children
inhighlyinflectedlanguages,suchasItalian,produceboundmorphemesintheir
earliestutterances.Cleverbehavioralstudies,suchasthosebyShipley,Smith,and
Gleitman(1969)andGerkenandMcIntosh(1993)havesincedemonstratedthat
English-speakingchildrendo,infact,representfunctionwordsandboundmorphemeslongbeforetheyusetheminspeech.Sowelearnedthatchildrenknow
morethantheysay,andwerealizedthatweneededtoconcentratemoreondevelopingviablehypothesesaboutchildrensimplicitknowledgeoflanguage,rather
thanbeingsatisfiedwithanalysesofspeechbehavior.Thisisanexampleofour
beliefsshapingwhatwechoosetoexperienceempirically.Weinitiallybelieved
thatallweneededwastodescribespeechpatterns,sowelistenedcarefully.Later
wecametobelievethatunderlyinglinguisticknowledgeisimperfectlyreflected
inspeechandwedevelopedincreasinglysophisticated(mostlyoff-line)methods
fortestinghypothesesaboutchildrensunderlyinggrammars.
Simultaneously, linguistic theory became more sophisticated. In the early
daysoftransformationalgrammarwefocusedontheapplicationofphrasestructurerulesandtransformationalrules.Inchildrensspeechwesoughttodiscover
theirincreasinglysophisticatedapplicationoftherulesoflanguage;agreatdeal
oftheearlyworkonchildlanguage(speech)attemptedtoaccountforbasicsentence structure with hypothesized phrase structure and transformational rules
(e.g.,Brown&Hanlon1970;Menyuk1969).Aslinguistictheorybecamemore
technical,itpostulatedtheacquisitionofknowledgestructuresfarremovedfrom
anything that could be discovered by an analysis of actual speech. Since then,
ourgoalhasbeentocharacterizechildrensgrammarsandtofigureouthowthey
coulddevelopovertimebasedoninputconsistingofexclusivelypositiveevidence
constrainedbygeneralprinciplesofacquisition,cognition,anddevelopment.
Acquisitionresearchturnedfromananalysisofindividualconstructionsto
acquisitionofsystemsofthegrammar.Wewereabletodistinguishbetweenuniversalprinciples,whichwecouldassumetobeavailabletothechildatbirthand
language-particular aspects of individual languages, which had to be acquired
fromlinguisticinput.Researchturnedtoattemptstodemonstratethatfromthe
beginningchildrenobeyuniversalconstraints,suchasapplytomovementopera-
172 HelenSmithCairns
tions(Crain1987)andtoacquisitionofgrammaticalmodulessuchasbinding
(Wexler&Chien1985;McDaniel,Cairns,&Hsu1990)andcontrol(McDaniel,
Cairns, & Hsu 1990/1991). Acquisitionists theory construction became much
moreconstrained.Ifchildrenwerefoundtoappeartoviolateuniversalprinciples,
thenitwasessentialtodevelopanaccountofwhythiswasthecase.Forinstance,
pragmaticfactorswereinvokedtoaccountfortheapparentviolationofPrincipleBofthebindingtheory(Montalbetti&Wexler1985;McDaniel&Maxfield
1992).Limitedstructure-buildingabilitieswerehypothesizedtoaccountforapparentviolationofcontrolprinciples(Cairns,McDaniel,Hsu,&Rapp1994).
Thetheorythatlanguagesvaryparametricallyprovidedthebasisforanexplosionofresearchinvestigatinggrammaticalacquisitioninavarietyoflanguages.A
largeliteraturepointstothechildsveryearly(beforetheageofthree)acquisition
ofanumberoflanguagespecificproperties.Inmanylanguages,suchasSpanish
andItalian,thesubjectofsentencescanbephoneticallynull;inothers,suchas
English,itmustbepronounced.Thisisknownasthepro-dropparameter,andit
isapparentlysetveryearly(Hyams1986).TherearelanguagessuchasGerman
andDutchinwhichfiniteverbsappearinthesecondpositionofasentence,while
nonfiniteverbsappearattheend.Earlylearning(Poeppel&Wexler1993)ofthis
parameterdemonstratesnotonlythatchildrenhavelearnedasyntacticrequirementoftheirlanguage,butalsothattheycandiscriminatefinitefromnon-finite
verbs.ChildrenacquiringalllanguageswithastandardorderingofSubject,Verb,
andObject(e.g.,SVOinlanguagessuchasEnglishandSwedish,SOVinGerman)
learnlanguage-particularwordorderpossiblybeforetheyarecombiningwords
(Wexler 1999). Since word order follows from a parameter regulating phrasal
structure,thisdemonstratestheacquisition,notofalinearconstraint,butofa
structuralone.InFrenchtheplacementofthemorphemepas, indicatingnegation,variesaccordingtothefinitenessoftheverb,appearingbeforeafiniteverb
andafteranon-finiteone.Thefactthatchildrenacquirethisdistinctionaround
the age of two demonstrates the ability to discriminate between the two types
ofverbandalsoknowledgeofamovementoperationthatraisesfinite,butnot
non-finite verbs (Pierce 1992). Thus, cross-linguistic investigations of language
acquisitiondemonstrateresponseatveryearlyagestotheinformationrequired
by children to set parameters and acquire non-universal, language-specific aspectsoftheirlanguages.
Amajordiscoveryofthecross-linguisticstudyofacquisitionhasbeenthat
languagesspokenbymembersofnon-industrialculturesarenotprimitive,as
wasoncebelieved,buttohaveallthecomplexityofthelanguagesofadvanced
industrialcultures.Arguablythemostimportantcontributionincross-linguistic
workhasbeenthedemonstrationthatsignedlanguagesarefullyhumanlanguages, with all the universal and language-specific properties of spoken languages
Languageacquisitionresearch 173
(Klima&Bellugi1979).Inthenot-so-distantpastsignedlanguageswerethought
ofassimplegesturalsystems,notfullyformedlanguages.Nowweknowthatchildrenacquiringsignedlanguagesgothroughthesamestagesasdospeakingchildren:frombabblingtotelegraphicutterances,tofullyformedsentences(Petitto
1994).
Alloftheseinvestigationsofgrammaticaldevelopmentfocusedontheacquisitionofanunderlyingsystemofgrammar,orlinguisticcompetence.Otherlines
ofresearchinvestigatedlinguisticperformance.
Intheearlydayswethoughtofperformancefactors,memorylimitations,
speech errors, false starts, and disfluencies, as masking ideal speaker-hearer
speech. However, very early in adult psycholinguistics, under the influence of
peoplelikeJerryFodor,TomBever,MerrillGarrett,andGeorgeMiller,theother
meaningofperformance,astheproductionandcomprehensionofspeechbased
ontheunderlyinggrammar,wastakenseriouslyasafocusofexperimentalinvestigation.Earlyworktooktooliterallyelementsoflinguistictheoryaselements
of psycholinguistic theory, and enterprises such as the Derivational Theory of
Complexitygotthingsofftoarockystart.Butwesoonlearnedthattheoriesof
performance needed to be independent of linguistic theory, although in some
important sense compatible with it. Systematic behavioral responses to newly
discovered phenomena like garden-path sentences called for explanations that
lookedbeyondatheoryofcompetence.Thegrammarisagnosticaboutwhether
atemporarilyambiguousstructure,suchasEveryone believed JohnshouldcontainaclauseboundarybetweenbelievedandJohn,makingJohnthesubjectofan
embeddedclause,ornot,makingJohntheobjectoftheverbbelieved.Thecomprehensionsystem,however,hasanoverwhelmingpreferenceforthelatteranalysis.Thus,therearelanguageprocessingmechanismsthat,whiledependentupon
thegrammarforarangeofpossiblestructures,clearlyhaveparsingpreferences
thatareindependentofthegrammar.Experimentalpsycholinguistswereinterestedfromthebeginningindevelopingtheoriestoaccountforadultsprocessing
ofsentencesinrealtimebecausemanyofthepreferencesofthecomprehension
system are theoretically temporary and therefore not measurable with off-line
techniquesthatpredominantlymeasuretheproductofcomprehension.Sothey
developedon-linetaskssuchasphonemeandwordmonitoring,probetasks,and
otherreactiontimemeasures,whichreflectedthepropertiesofthemechanisms
engagedinlanguageprocessingasitunfoldsovertime.
Interest in childrens linguistic performance as an object of study (rather
thanasavehiclefortestinghypothesesabouttheircompetence)laggedfarbehind adult psycholinguistics. I think this was primarily because acquisitionists
werefocusedongrammaticaldevelopmentinchildrenandwantedtofigureout
howchildrensgrammarsprogressedfromtheinitialtotheadultstate.(Andwe
17 HelenSmithCairns
wantedourexperimentstobeasfreeaspossiblefromperformancefactors,such
aseffectsofmemoryandtaskdemands.)Adultgrammarsweretheprovinceof
linguists,whileadultprocessingwastheresponsibilityofexperimentalpsycholinguists.Thus,attentiontochildrensgrammaticaldevelopmenteclipsedinterestin
theirprocessingoflanguage.
In1996MITPresspublishedMethods of assessing childrens syntax (McDaniel,
McKee,&Cairns1996),whichreviewedalargevarietyof(mostly)off-linemethodsusedtoinvestigategrammaticaldevelopment.Therewereonlytwochapters
dealingwithon-linemethods,onebyKathyHirsh-PasekandRobertaGolinkoff
onintermodalpreferentiallookingandonethatwasageneralreviewofon-line
methodsbyCecileMcKee(McKee1996).Shepointedouttherarityofon-line
studiesrelativetooff-lineones,citingastudybyTylerandMarslen-Wilson(1981)
asoneoftheearliestattemptstoidentifyoperationsinthereal-timeprocessing
ofsentencesbychildren.Itwasawordmonitoringexperimentinvestigatingvariouscontextualeffectsonlanguageprocessingin5,7,and10-year-oldchildren.
(Ahistoricalnote:ThiswaswhentheJournal of Memory and Language wasthe
Journal of Verbal Learning and Verbal Behavior.)IntheirintroductionTylerand
Marslen-Wilsonwritethefollowing:Inmostoftheresearchonthedevelopment
oflanguagecomprehensionismeasuredattheendofanutterance,ratherthan
asitisbeingheard.Suchstudieshavelittletosayabouttheinternalstructureofprocessingevents.Toaddressthesekindsofissueswehavetousetasks
whichtaptheoperationsinvolvedinreal-timeoron-linesentenceprocessing
(p.400).Suchmethodshavebeenonlygraduallydevelopedovertheelevenyears
afterMcKeewroteherchapter.
This volume and the Workshop that inspired it is extremely important for
ourfieldbecauseitfocusesonthereal-timeoperationsalludedtobyTylerand
Marslen-Wilsonin1981,adomainofinquirythatisdestinedtobecomeanincreasinglyprominentaspectofchildlanguageresearch.Itisparticularlytimely
becauseon-linemethodologyhasmaturedsufficientlyinthelastdecadetoconstitutearobustfieldinitsownright.
Not surprisingly, as research in childrens language processing has come
intoitsown,thetheoreticalissuesandempiricalquestionsthatitaddressesare
primarily interested ingrammatical development.An interesting aspectofthis
Workshopandoftheworkdiscussedinthisvolumeisthatthenewquestionsare
directlyreflectedinthenewon-linemethodsthatarebeingdeveloped.Theory
identifies the hypotheses and questions that need to be addressed empirically,
andcleverpeoplefigureoutthemethodstodothat.Whileoff-linetaskscantest
hypothesesaboutthestatusofchildrensgrammars,itiscrucialtoknowwhether
those grammatical principles are applied as sentences are being processed. An
early and important study addressing this question was one by McKee, Nicol,
Languageacquisitionresearch 17
andMcDaniel(1993).Usingacrossmodalprimingparadigm,theydemonstrated
thatchildrenapplyknowledgeofPrincipleAon-line.Further,theyshowedthat
the mysterious tendency of some children to (apparently) disobey Principle B
manifests itself during sentence processing. Just those children who disobeyed
PrincipleBinanoff-linetruth-valuejudgmenttaskshowedactivationofillegal
antecedentsintheon-linetask,whilethosewhoobeyedPrincipleBinanoff-line
taskdidnot.ThestudybyHirsh-PasekandGolinkoff(1996)usingtheintermodal
preferential-lookingparadigm(whichtheyadaptedfromSpelke,whodeveloped
itin1979tostudyintermodalperception)demonstratedthatinfantshaveinternalizedthestandardwordorderofEnglishandcandistinguishtransitivefrom
intransitivesentenceframes.Thisresultwasparticularlyimportantbecausethe
infantstheystudiedwerenotyetusingcombinatorialspeech.
A closely related question is: What are the representations constructed by
childrenastheyprocesssentences?Obviously,theconstructionofarepresentationisbasedongrammaticalknowledge,soanyparticularrepresentationisprima facieevidenceoftheexistenceofthatknowledge.Forexample,experiments
reportedbyClahsen(thisvolume)demonstratingthatchildrenfillgapson-line
demonstratenotonlytheexistenceofemptycategoriesinchildrensrepresentations,butalsotheexistenceintheirgrammarsofmovementoperations.Snedeker
andThothathiri(thisvolume)arguethateffectivecross-verbalprimingintheir
structuralprimingstudiesdemonstratethatchildrenasyoungas3and4years
oldconstructabstractrepresentationsofverbandsentencestructure.Thenow
well-knownpreferenceofchildrenforagoalinterpretationofthefirstprepositionalphraseinsentenceslikePut the frog on the napkin into the box(Trueswell,
thisvolume)reflectaninitialstructuralanalysisinwhichthefirstPPisVPrather
thanNPattached.Italsoindicatesthatchildrenrepresenttheargumentstructure
ofverbsintheirlexicons.
Anumberofoff-linestudies(e.g.,Gerken&McIntosh1993)havedemonstratedthatchildrenwhoomitfunctionwordsinspeechincludetheminsentence
representations.WorkreportedatthisWorkshopbyKedar(2006)reinforcesthis
finding,butshowsthatevenyoungerchildrenthanthoseidentifiedbyoff-line
studiescanbeshowntoconstructinternalrepresentationsoffunctionwordsif
sufficientlysensitivemeasuresareused(forKedarlookinglatenciesintheintermodalpreferential-lookingparadigm).Thisworkiscriticaltoourunderstanding
ofwhenchildrenbegintorepresentfunctionalcategoriesintheirgrammars.The
lateuseoffunctionwordshadledearliertothebeliefthatearlygrammarsmay
lackfunctionalcategories(Radford1990).
Justasworkingrammaticaldevelopmentaskedquestionsaboutwhenchildrens grammars become adult-like, much processing work is driven by asking
howchildrensprocessingoperationsaresimilartoanddifferentfromthoseof
176 HelenSmithCairns
adults.Anexcellentexampleofthislineofresearchcomesfromtheon-lineprocessingoflexicalambiguities.DavidSwinney(1979),usingacross-modalprimingtask,demonstratedtheinitiallycounter-intuitivefindingthatuponencountering an ambiguous word in a sentence, adults retrieve all possible meanings,
independentofcontext,thenselectthecontextuallyappropriatemeaningbefore
theendofthesentence.Thisworkwasofgreatimportancebecauseforyearswe
had known about the role of context in sentence processing in general and in
disambiguationinparticular.Swinneyidentifiedthelocusofthecontexteffect,
when it does and does not influence the hearer, and revealed the existence of
unconsciousmentaloperationsthatcomeintoplayduringsentenceprocessing.
TenyearslaterSwinneyandPrather(1989),and,later,Love,Swinney,Bagdasaryan,andPrather(1999)showedthatthissameeffectobtainsforchildren.Atthis
WorkshopKhanna,BolandandCortese(2006)reportedstudiesoftheresolution
oflexicalambiguityinresponsetobiasingcontext.Usingacross-modalnaming
paradigm, they showed that second and third grade children (approximately 8
and9yearsold)arenotabletousebiasinginformationinthewayadultsdo,but
fourthgradechildren(10yearsold)are.Further,successfuluseofcontextwas
correlatedwithreadingability.
Other on-line work on lexical access and organization demonstrates that
childrenslexiconsareorganizedsimilarlytothoseofadults.Inaneyemovement
studyofcohorteffects,reportedatthisWorkshop,Sekerina(2006)showedthat
preschool children demonstrate the cohort effect, but that developmental progressionisinthespeedandefficiencywithwhichlexicalinformationisaccessed
andused.
Arelatedissueiswhether,duringon-linesentenceprocessing,childrenuse
the same kinds of information as adults do. Gibson, Breen, Rozen, and Rohde
(2006) demonstrate that relative clauses with object extraction are processed
more slowly than are those with subject extraction for both adolescents (age
1215)andadults.Thetwogroupsshowsimilargarden-patheffectsinsentences
withreducedsubjectrelativeclauses,effectswhichareattenuatedbythepresence
of verbs with a high frequency of use as a past participle. The tasks were selfpacedreadingandlistening.Thestrikingdifferencebetweenthetwogroupswas
revealedinthereadingtask,inwhichtheadults,butnottheadolescents,wereassistedbyplausibilityinformation.Note,however,thatintheGibsonetal.(2006)
studyparticipantsweremucholderthanchildreninmoston-linestudies.Itis
unclearwhetherself-pacedreadingcouldbedonewithmuchyoungerchildren,
giventhecomplexityandlengthofthematerials.Still,thelackofadult-likeuseof
plausibilityisevenmoreinteresting,giventhefactthatthepeoplestudiedwere
adolescentsandmighthavebeenexpectedtoexhibitfullyadult-likeprocessing
strategies.Kidd,Stewart,andSerratrice(2006),inaneyemovementstudy,showed
Languageacquisitionresearch 177
asimilartendencyof5-year-oldstorelymoreonstructuralinformationthanon
eitherplausibilityorinformationgainedfromthevisualscene.Inasentencelike
Chop the tree with the leaveschildrenignoretheimplausibilityofleavesasaninstrumentandrelyontheverbbiasforVP-attachmenttoproducetheimplausible
situation of using leaves to chop the tree. Note that this also demonstrates the
preferenceofthechildren,aswellastheadults,toconstructrepresentationswith
theprepositionalphraseattachedtotheVPratherthantotheNP.
Acrucialprocessingdifferencebetweenchildrenandadultsischildrensinabilitytoreviseinitialparsesbyshiftingfromaninitialstructuralrepresentation
toamoreappropriateone.Fabrizio,Guasti,andAdani(2006)usedaself-paced
listening experiment with Italian-speaking 9-year-olds to see whether number
agreementontheauxiliaryverbwouldcausethemtorepairaninitialsubjectrelativeanalysis.Manyofthechildrenwereunabletodothisand,asaconsequence,
misunderstoodthesentences.Fabrizioetal.(2006)suggestthatforchildrenstructuralinformationismoresalientthanagreementinformation,similartoGibson
etal.s(2006)andKiddetal.s(2006)findingsaboutstructuralinformationrelative
toplausibility.Inhiseyemovementexperiments,Trueswell(thisvolume)shows
thatchildrenabovetheageof8can,asdoadults,revisetheirinitialVPattached
representationofthefirstPPinPut the frog on the napkin into the boxwhenthey
hearthesecondPP,restructuringtheparsesothatthefirstPPisNP-attachedand
thesecondfulfillsthesub-categoricalrequirementsoftheverb.Childrenyounger
thaneightcannotperformthisreanalysis,eveninthepresenceofavisualcontext
containingtwofrogs,oneofwhichisonanapkin.
Explanationsforthedifferencesbetweentheprocessingabilitiesofchildren
andadultsleadon-lineresearcherstoexaminethedevelopmentofnon-linguistic cognitive capacities. Interestingly, these are often the performance factors
wehavesoughttocontrolinoff-lineexperiments.Khanna,Boland,andCortese
(2006)accountfortheabilityoffourthgraders,asdistinctfromthatofsecondand
thirdgraders,tousecontextforlexicaldisambiguationbypostulatingthematurationofinhibitionandselectioncapabilities.Theysuggestthatthesecapacitiesare
underdevelopedbeforetheageof8,whichisalsowhatTrueswellidentifiesasthe
agewhensuccessfulparsingrevisionstakeplace.Trueswellarguesthatincreased
revisionabilityisattributabletothedevelopmentofcognitivecontrolandexecutivefunction.Heclaimsthatchildrenmustovercomecognitiveimpulsivityinordertoreviseinitialhypothesesaboutstructureandmeaning.Gibsonetal.(2006)
believethattheirfindingthatplausibilityaffectslisteningbutnotreadingisaresultnotofmodalityper se,butofresourceallocation.Severalstudiespresentedat
thisWorkshopanddiscussedinthisvolumeidentifytheeffectsofmemoryspan.
Clahsen(thisvolume)reportsstudiesshowingthatchildrenwithlowmemory
spandonotshowthereactivationofantecedentsatgapsdemonstratedbychil-
178 HelenSmithCairns
drenwithhighermemoryspans.Theyalsotendtoattachstructurallyambiguous
relativeclausestothesecondNP,whilehighspanchildrenpreferthefirst.Fabrizioetal.(2006)foundthatchildrenwithhighermemoryspansweremorelikely
thanchildrenwithshortermemoryspanstouseagreementinformationtorevise
initialstructuralhypotheses.
Questionsabout the information usedinon-lineprocessinghavebeen extendedtocross-linguisticresearch.Insentencessuchas Put the frog on the napkin
English-speakingchildrenpreferVPattachmentforthePPpresumablybecause
theinitialverbissub-categorizedforalocativeargument.InKoreansentences
of this type, verb information is last, while initial information is case marking
on napkin, which, though ambiguous, is biased toward the locative. Choi and
Trueswell(2006)showthatbothKoreanandEnglishspeakingchildrenemploy
a(byhypothesis)universalstrategyofusingthefirstreliableinformationavailableintheirlanguage,verbalfortheEnglish-speakingchildren,casemarkingfor
theKoreanchildren.Furtherdemonstrationthatchildrenexploittheinformation
availableintheirlanguageisaseriesofstudiesbyFernaldandcolleagues(Fernald,Zangl,Thorpe,Hurtado,&Williams2006)showingthatSpanish-speaking
childrenasyoungas3usegendermarkingoftheadjectivetoidentifythereferent
of nouns on timed trials. Post-nominal adjectives in Spanish also facilitate the
processingofnounphrases,relativetoEnglishpre-nominalforms.
Amajorcontributionofon-lineresearchhasbeenanenhancedunderstanding of the nature of language disorders, and the promise of early detection of
childrenwhoareatrisk.InaseriesofERPstudiesFriederici(Mnnel&Friederici
thisvolume)identifiedanumberofmeasuresonwhichinfantsatriskforspecific
languageimpairmentdifferedfromthosewhowerenot.At-risk2-month-oldinfantsfailedtodiscriminatelongfromshortsyllablesasrapidlyasdidinfantswho
hadnofamilyhistoryofSLI.Theyalsoreportstudiesshowingthatinfantswith
afamilyhistoryofdyslexiarespondtodurationchangesandconsonantchanges
differently than do infants without such a family history. Retrospective studies
demonstrateimpairedstressperceptionandresponsestoincongruouswordsin
theERPresponsesofinfantswholaterhavelanguageproductiondeficits.Taken
together,thesestudiessuggestthatchildrenwhoareatriskforlanguagedisordersprocessspeechinputdifferentlythandochildrenwhodonotdevelopdisorders.On-linemethodsalsopromiseearlyidentificationofchildrenatrisk.Both
MarchmanandFernald(2006)andPakulakandNeville(2006)identifylowsocio-economicstatusasamajorpredictorofcognitiveandlinguisticdeficits.The
latterevenidentifydifferencesinbrainstructureassociatedwithSESdifferences.
PakulakandNevillereportalargestudyofavarietyofinterventiontechniques
withlowperformingSESchildren,withpromisingresults.Thesefindingshave
majorpublicpolicyimplicationsforearlychildhoodeducationandintervention.
Languageacquisitionresearch 179
VanderLelyandFonteneaus(2006)neurolinguisticworkonchildrenwithspecificlanguageimpairment(SLI)whohaveaspecificgrammaticalimpairmenthas
demonstratedaneuralsubstratespecializedforsyntacticprocessing.Herwork
bringsusfullcircle,fromanunderstandingoflanguageimpairmenttoinsight
aboutlinguisticfunctioningintheunimpairedbrain.
Several studies presented at this Workshop identified predictors in infancy
for language disorders later in life. One, however, demonstrates continuity betweenlanguageskillsintypicallydevelopingchildrenfromtheageof25-months
to8years.MarchmanandFernald(2006)conductedalongitudinalstudyofeye
movementsinresponsetopicturenaming.Thosechildrenwhorespondedmore
quicklyandaccuratelyat25monthsdemonstratedfastervocabularygrowthin
subsequentyears.Furthermore,inafollow-upstudywhenthesechildrenwere
8yearsold,theyweretestedonstandardizedlanguagemeasures.Theresponse
timesofthechildrenwhentheywereinfantscorrelatedsignificantlywiththeir
performanceonlanguageandcognitivetestssixyearslater.Thisremarkablestudy
demonstratesthecontinuityofverydifferentlanguageskillsoverdevelopmental
time.Theimportanceofsuchafindingforourunderstandingoftypicallanguage
acquisitioncannotbeover-estimated.
Myownwork(Cairns,Waltzman,&Schlisselberg2004)investigateschildrens
metalinguisticabilitytodetecttheambiguityoflexicallyandstructurallyambiguous sentences.Weare interested notin ambiguityresolution,butintheability
toreportthatasentencehastwopossiblemeanings.Wearguethatthisability
restsonthelexicalandstructuralprocessingoperationsstudiedon-line:accessof
multiplemeaningsofambiguouswordsinsentencesandtheabilitytoconstruct
structuralrepresentationsofsentences.Inparticular,wethinkthatambiguitydetectionreliesontheabilitytoreprocesslexicalrepresentationsandrevisestructuralrepresentations.Itisnocoincidencethattheageof8,whichiscrucialinthe
abilitytorevisestructuralanalyseson-line,istheageatwhichchildrenbeginto
beabletodetectstructuralambiguity.Inordertoconstructtwostructuralrepresentations(necessaryfortheperceptionoftheambiguityofstructurallyambiguoussentences)childrenmustescapefromwhatTrueswellcallscognitiveimpulsivity.Wefindthatambiguitydetectionisamassivepredictorofreadingability
inpre-readersthroughthirdgraders,andwearguethatisbecauseitisjustthose
psycholinguisticprocessesthatarerecruitedinskilledreading.Itwouldbeinterestingtoinvestigatewhethergoodearlyreadershavelesscognitiveimpulsivity
thandopoorreaders.Itisanidealoutcomeforon-lineinvestigationstoproduce
resultsthatelucidatenotonlyon-lineprocessingbutalsotheacquisitionofhigher
leveloperations,suchasmetalinguisticskillandreading.
Theimportanceofavarietyofmethodologiestoourunderstandingofchild
languagecannotbeover-estimated.Thisisbecausewemusthavetheoriesofat
180 HelenSmithCairns
leastthreelinguisticlevelsandtheiracquisitiontoaccountforlanguageuse.First,
we must have a theory of linguistic form and organization (the grammar and
lexicon)todefinethenatureoflexicalinformationandthestructuresthatcan
becomputedduringsentenceproductionandcomprehension.Suchtheoriesare
typicallytestedbyavarietyofoff-linemethods.Second,wemusthaveatheory
oftheprocessingoperationsinvolvedinaccessingthegrammarandlexiconin
productionandcomprehension.Suchatheorywillincludeparsingpreferences
andprinciplesoflexicalorganization(e.g.,frequency)aboutwhichthegrammar
isagnostic.On-linemethods,suchasthoseaddressedatthisWorkshopanddiscussedinthisvolume,arecrucialfortestingthesetheories.Finally,wemusthave
atheoryofneuralorganizationandoperationtoaccountforhowtheprocessing
operations are implemented and how knowledge of language is developed and
represented.Methodsprobingbrainfunctionandorganization,someofwhich
werepresentedatthisWorkshopanddiscussedinthisvolume,arecriticaltotest
thesetheories.Inthespiritofpredictingthepossibledirectionoffutureresearch,
Iwillsuggestwhereprogressneedstobemadeineachoftheseareas.
Current theories of the development of linguistic competence assume that
theinfantbeginswithinnateaccesstotheprinciplesandoperationsmadeavailablebyUniversalGrammarand,thus,doesnotneedtoacquirethemthroughinteractionwiththeenvironment.Language-particularaspectsofmorphologyand
syntax,aswellaslexicalrepresentations,mustbeacquiredthroughinformation
availableinthespeechofthechildscommunity.Advancesinlearnabilitytheory
demonstratethattheinformationthechildreceivesmustbeexclusivelypositive,
as opposed to negative, information. That is, the child does not have access to
informationaboutwhichlinguisticformsarenotavailableinhislanguage.Note
thatmostofthisconceptionofacquisitiontheoryderivesdirectlyfromlinguistictheoryitself.UniversalGrammarspecifiestheinnateaspectsoflanguagethat
childrendonothavetolearn,aswellastheparametersthatmustbesetthrough
experience.Iwouldliketosuggest,however,thatwedonotreallyhaveatheoryof
languageacquisition.AsMnnelandFriedericistateinthefirstsentenceoftheir
paperinthisvolume:Thewonderoflanguageacquisitionwithitsremarkable
speedanditshighsuccessremainsamystery.Decadesofresearchhaverevealed
an enormous amount of valuable descriptive information about what children
knowandwhentheyknowit,butwedonotyethaveatrulyexplanatorytheory.
Justasadultpsycholinguisticshassucceededindevelopingprocessingtheories
consonantwithbutindependentofthegrammar,weneedatheoryoflanguage
acquisition that accounts for how children operate on their linguistic input to
creategrammars.Similartoadultsinlanguageprocessing,childrenmustengage
cognitiveprocessesthatarenon-linguisticinordertooperateonthespeechinput
available in their environment. Slobin (1985) in a series of papers, books, and
Languageacquisitionresearch 181
chapters,describedthesetofoperatingprinciplesthatthechildbringstobear
onthelanguage learning process. Indeed,weneed atheoryoflanguage learninguniversalsthatallhumanchildrenemployinordertocreateaninternalized
grammar,shapedbybiologically-basedaccesstoUniversalGrammarinteracting
withthespeechofthechildscommunity.Valian(1990)hasaddressedthequestionofwhetherparameterscanbesetbyprecisetriggersintheinputlanguage,
orwhetherthechildusesparametricvariationtoconstructhypothesesaboutthe
language-particularaspectsofhislanguage.Adiscussionoftheadvantagesand
difficultiesassociatedwithbothconceptionswouldtakeustoofarafield;however,
Ibelievethatthisisthekindofquestionweshouldbeaddressingaswemovetowardthegoalofatrulyexplanatorytheoryoftheacquisitionofcompetence.
Similarly,weneedatheoryofhowprocessingskillsdevelop.Havingestablishedthatadultsusenon-linguisticprocessingoperations,wemustaccountfor
howinfantsgrowuptobeadultsinthisrealm,aswellasintherealmoflinguistic
competence. Studies presented at this Workshop suggest that properties of the
language,suchaswordorderconstraints,movementoperations,andlexicalinformationdriveearlyparsingpreferences.Buthowdochildrenacquireprocessing
operationsandpreferences,suchassubject-objectasymmetriesandtheexpectationofsubjectgaps,whicharenotdrivenbythelanguageandwhichthechild
cannot,inprinciple,observe?Lexicalorganizationandreceptiveaccessappears
tobesimilarforadultsandchildren,butwhatprinciplesoflexicaldevelopment
canaccountforthis?Canweidentifyuniversalprinciplesofstructuralandlexical
processing?Howcanweaccountforthefactthatsomekindsofinformationseem
tobelesssalientforchildrenthanforadults,e.g.,informationaboutmorphology and plausibility? Non-linguistic cognitive characteristics seem to drive the
acquisitionofprocessingtoamuchgreaterdegreethantheydotheacquisition
ofgrammar.Selection,inhibition,efficiencyoflexicalaccess,cognitivecontrol,
executivefunction,resourceallocation,andmemoryspanallseemtoplayrolesin
themovementofchildrentoadult-likeprocessingcapabilities.Weneedtheories
ofhowthesecognitiveabilitiesdevelopandhowtheyinteractwiththeapplication
ofprocessingoperationsderivedfromgrammaticalandlexicalknowledge.
Atheoryofthedevelopmentoflinguisticperformancemustaddresslanguage
productionaswellasreception.Webeganbyacknowledgingthatchildrenknow
morethantheysay:thequestioniswhyshouldthatbeso.Therehasbeensurprisinglylittleresearchintothelanguageproductionofyoungchildren,yetthereare
importantcross-linguisticsimilaritiespointingtotheroleofgenerallinguistic,
cognitive,ormotoricprinciplesinearlyspeechproduction.Ithasbeensuggested
thatthewell-knownvocabularyspurtthatoccursinthesecondyearoflifemay
bemorearesultofenhancedlexicalaccessthanofwordlearning(Dapretto&
Bjork2000).Atheoryofearlyproductionwill,liketheoriesofadultproduction,
182 HelenSmithCairns
addressthemechanismsbywhichmorphologicalandsyntacticformsarerealized
bytheproductionsystem,aswellastheprocessesoflexicalrepresentationand
retrieval(Garrett1988;Bock&Levelt1994).
Atheoryofthedevelopmentoftheneurologicalrepresentationoflanguage
and the brain mechanisms that underlie language learning and processing will
undoubtedlybeeventuallysubsumedasacomponentofdevelopmentalneurology.Itwillbecriticaltoidentifythoseareasofthebrainthatdevelopaslanguage
learning takes place, just as wenow knowquite abitaboutthetime-courseof
linguisticoperationsinthedevelopingbrain.
As we speculate about the future applications of research in all three areas
of language competence, processing, and the neurological substrate we are
struckbytheimportanceofworkinallthreetoidentifychildrenatriskforvarioustypesoflanguagedisorders.SeveralpresentationsattheWorkshopandinthe
wider literature suggest subtle precursors to specific language impairment and
other forms of language disorders. Early detection is extremely important, but
ofequalvalueistheapplicationofpsycholinguisticprinciplestointerventionin
disordersofspeech,language,andreading.Theconversationbetweenpsycholinguisticresearchersandpeopleonthefrontlinesofhelpingat-riskandlanguage
disorderedchildrenmustbegreatlyimproved.Thiswillrequireeffortsonboth
sidesofthatconversation.Psycholinguistsandotherlanguageresearchersneed
to reach out to make their findings accessible and relevant to speech-language
pathologistsandeducationalspecialists.Bythesametoken,practitionersmustbe
willingtolistentopeopleinvestigatingbasiclanguageprocessesandbeopento
theimplicationsofexperimentalworkforclinicalintervention.Enhancedcommunication across the disciplinary divides could result in improvement in the
livesofthousandsofchildren.
Whateverthefutureholdsforthefieldoflanguageacquisition,wecanpredictthatadvancesinourknowledgewillbeasgreatastheyhavebeeninthepast.
Alongwiththepresentvolume,theWorkshop on On-line Methods in Childrens
Language Processing servednotonlytodemonstratesophisticatednewmethodologiesbeingdeveloped.Itshowcasedthetalentedanddedicatedscholarswho
willleadthefieldforward.
Acknowledgments
IamgratefultoEvaFernndezandDanaMcDanielforinsightfulcommentson
previousdraftsofthischapter.
Languageacquisitionresearch 183
References
Bock,K.&Levelt,W.(1994).Languageproduction:Grammaticalencoding.InM.A.Gernsbacher (Ed.), Handbook of psycholinguistics (pp. 945978). New York NY: Academic
Press.
Braine,M.D.S.(1963).TheontogenyofEnglishphrasestructure:Thefirstphase.Language,
39, 114.
Brown,R.(1973).A first language: The early stages.CambridgeMA:HarvardUniversityPress.
Brown,R.&Fraser,C.(1963).Theacquisitionofsyntax.InC.N.Cofer&B.Musgrave(Eds.),
Verbal behavior and learning: Problems and processes(pp.158201). NewYorkNY:McGraw-Hill.
Brown, R. & Hanlon, C. (1970). Derivational complexity and order of acquisition in child
speech.InJ.R.Hayes(Ed.),Cognition and the development of language(pp.1154).New
YorkNY:Wiley.
Cairns,H.S.,McDaniel,D.,Hsu,J.R.,&Rapp,M.(1994).Alongitudinalstudyofprinciplesof
controlandpronominalreferenceinchildEnglish.Language, 70,260288.
Cairns,H.S.,Waltzman,D.,&Schlisselberg,G.(2004).Detectingtheambiguityofsentences:
Relationshiptoearlyreadingskill.Communication Disorders Quarterly, 25, 6878.
Choi,Y.&Trueswell,J.C.(2006).Puttingfirstthingslast:Across-linguisticinvestigationofthe
developingsentenceprocessingmechanism.PaperpresentedattheWorkshoponOn-Line
MethodsinChildrensLanguageProcessing,CUNY.
Chomsky,N.(1965).Aspects of the theory of syntax. CambridgeMA:TheMITPress.
Crain,S.(1987).Languageacquisitionintheabsenceofexperience.Behavioral and Brain Sciences, 14, 597650.
Dapretto,M.&Bjork,E.L.(2000).Thedevelopmentofwordretrievalabilitiesinthesecond
yearanditsrelationtoearlyvocabularygrowth.Child Development, 71, 635648.
Fabrizio,A.,Guasti,M.T.,&Adani,F.(2006).RelativeclauseprocessingbyItalianchildren:
A self-paced listening study. Paper presented at the Workshop on On-Line Methods in
ChildrensLanguageProcessing,CUNY.
Fernald,A.,Zangl,R.,Thorpe,K.,Hurtado,N.,&Williams,C.(2006).Learningtolistenahead
inEnglishandSpanish:Infantsusemultiplelinguisticandnon-linguisticcuesinonline
sentenceinterpretation.PaperpresentedattheWorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
Garrett,M.F.(1988).Processesinlanguageproduction.InF.J.Newmeyer(Ed.),Linguistics:
The Cambridge survey: Vol. III. Language: Psychological and biological aspects (pp.996).
Cambridge:CUP.
Gerken,L.A.&McIntosh,B.J.(1993).Thenterplayoffunctionmorphemesandprosodyin
earlylanguage.Developmental Psychology, 29, 448457.
Gibson,E.,Breen,M.,Rozen,S.,&Rohde,D.(2006).Languageprocessinginchildrenasmeasuredusingself-pacedreadingandlistening.PaperpresentedattheWorkshoponOn-Line
MethodsinChildrensLanguageProcessing,CUNY.
Hirsh-Pasek, K., & Golinkoff, R. M. (1996). The intermodal preferential looking paradigm:
Awindowintoemerginglanguagecomprehension.InD.McDaniel,C.McKee,&H.S.
Cairns(Eds.),Methods of assessing childrens syntax. CambridgeMA:TheMITPress.
Hyams,N.(1986).Language acquisition of the theory of parameters.Boston:Reidel.
18 HelenSmithCairns
Kedar,Y.(2006).Uncoveringearlygrammaticalknowledge:Differentmethodsandmeasures
capturespecificaspectsofinfantslinguisticprocessingcompetence.Paperpresentedat
theWorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
Khanna,M.M.,Boland,J.E.,&Cortese,M.J.(2006).Developmentofsentencecontextuse:
Whenandhowdochildrenknowthattagisalabelandnotagame?Paperpresentedatthe
WorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
Kidd,E.,Stewart,A.,&Serratrice,L.(2006).Canchildrenovercomelexicalbiases?Theroleof
thereferentialscene.PaperpresentedattheWorkshoponOn-LineMethodsinChildrens
LanguageProcessing,CUNY.
Klima, E. & Bellugi, U. (1979). The signs of language. Cambridge MA: Harvard University
Press.
Love, T., Swinney, D., Bagdasaryan, S., & Prather, P. (1999). Real-time processing of lexical
ambiguitiesbychildren.Proceedings of the 12th Annual CUNY Conference on Human Sentence Processing. NewYork:CityUniversityofNewYork.
Marchman,V.A.&Fernald,A.(2006).Onlinespeechprocessingefficiencyininfancyisrelated
bothtovocabularygrowthandtoschool-agelanguageaccomplishments.Paperpresented
attheWorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
McDaniel,D.&Maxfield,T.L.(1992).PrincipleBandcontrastivestress.Language Acquisition:
A Journal of Developmental Linguistics, 2, 337358.
McDaniel,D.,Cairns,H.S.,&Hsu,J.R.(1990).Bindingprinciplesinthegrammarsofyoung
children.Language Acquisition: A Journal of Developmental Linguistics, 1, 121138.
McDaniel,D.,Cairns,H.S.,&Hsu,J.R.(1990/1991).Controlprinciplesinthegrammarsof
youngchildren.Language Acquisition: A Journal of Developmental Linguistics, 1, 297336.
McDaniel,D.,McKee,C.,&Cairns,H.S.(1996).Methods for assessing childrens syntax. CambridgeMA:TheMITPress
McKee,C.(1996).On-linemethods.InD.McDaniel,C.McKee,&H.S.Cairns(Eds.),Methods
of assessing childrens syntax. CambridgeMA:TheMITPress.
McKee,C.,Nicol,J.,&McDaniel,D.(1993).Childrensapplicationofbindingduringsentence
processing.Language and Cognitive Processes, 8, 265290.
McNeil, D. (1970). The acquisition of language: The study of developmental psycholinguistics.
NewYorkNY:HarperandRow.
Menyuk,P.(1963a).Apreliminaryevaluationofgrammaticalcapacityinchildren.Journal of
Verbal Learning and Verbal Behavior, 2,429439.
Menyuk,P.(1963b).Syntacticstructuresinthelanguageofchildren.Child Development, 34,
407422.
Menyuk,P.(1969).Sentences children use. CambridgeMA:TheMITPress.
Montalbetti,M.M.&Wexler,K.(1985).Bindingislinking.Proceedings of the West Coast Conference on Formal Linguistics, 4, 228245.
Pakulak,E.&Neville,H.(2006).Exploringtherelationshipbetweenenvironment,proficiency,
andbrainorganizationinchildrenfromdifferentsocioeconomicbackgrounds.PaperpresentedattheWorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
Pierce,A.(1992).Language acquisition and syntactic theory: A comparative analysis of French
and English child grammars. Dordrecht:Kluwer.
Petitto, L. A. (1994). Modularity and constraints in early lexical acquisition: Evidence from
childrensearlylanguageandgesture.InP.Bloom(Ed.),Language acquisition: Core readings(pp.95126).CambridgeMA:TheMITPress.
Poeppel,D.&Wexler,K.(1993).Thefullcompetencehypothesis.Language, 69(1),133.
Languageacquisitionresearch 18
Radford,A.(1990).Syntactic theory and the acquisition of English syntax: The nature of early
child grammars of English. Oxford:Blackwell.
Sekerina,I.A.(2006).Spoken-wordrecognitioninRussianpreschoolers.Paperpresentedat
theWorkshoponOn-LineMethodsinChildrensLanguageProcessing,CUNY.
Shipley,E.F.,Smith,C.S.,&Gleitman,L.R.(1969).Astudyintheacquisitionoflanguage:Free
responsestocommands.Language, 45, 322342.
Slobin,D.I.(1985).Cross-linguisticevidenceforthelanguage-makingcapacity.The crosslinguistic study of language acquisition, Vol.2.: Theoretical issues (pp.11571256).HillsdaleNJ:
LawrenceErlbaumAssociates.
Spelke,E.(1979).Perceivingbimodallyspecifiedeventsininfancy.Developmental Psychology,
15, 626636.
Swinney,D.(1979).Lexicalaccessduringsentencecomprehension:(Re)considerationofcontexteffects.Journal of Verbal Learning and Verbal Behavior, 18, 645659.
Swinney,D.&Prather,P.(1990).Onthecomprehensionoflexicalambiguitybyyoungchildren:
Investigationsintothedevelopmentofmentalmodularity.InD.S.Gorfein(Ed.),Ambiguity processing. NewYorkNY:Springer.
Tyler,L.K.&Marslen-Wilson,W.D.(1981).Childrensprocessingofspokenlanguage.Journal
of Verbal Learning and Verbal Behavior, 20, 400416.
Valian,V.(1990).Nullsubjects:Aproblemforparameter-settingmodelsoflanguageacquisition.Cognition, 35,105122.
VanderLely,H.&Fonteneau,E.(2006).ERPinvestigationsintypicallydevelopingandlanguage-impaired children reveal a domain-specific neural correlate for syntactic dependencies.PaperpresentedattheWorkshoponOn-LineMethodsinChildrensLanguage
Processing,CUNY.
Wexler,K.(1999).Maturationandgrowthofgrammar.InW.C.Ritchie&T.K.Bhatia(Eds.),
Handbook of language acquisition. NewYorkNY:AcademicPress.
Wexler,K.&Chien,Y.-C.(1985).Thedevelopmentoflexicalanaphorsandpronouns.Papers
and Reports on Child Language Development, 24, 138149.
Index
A
accuracy 5,18,32,48,76,77,93,
103,104,114,116,118,119,123,
128130,147
acoustic-phonological 48,54
agreement 5,6,115,116,177,178
ambiguity 7,10,90,91,149,151,
154,159,161,176,179
ambiguousfigures 89
ambiguousregion 7,154,160
amplitude 32,34,35,36,4547,
51,52,56,72
analysis 7,25,31,33,58,64,65,
68,71,72,76,78,97,104,112,
113,115,118,121,125,126,128,
129,132,135,146,147,150,151,
154158,160,171,173,175,
177,184
antecedentpriming 12
anti-saccades 84
argumentstructure 138,140,
141,149,150,158,161
artifactcorrection 72
artifactrejection 72
attachmentpreference 8,9,10
attention ix,32,6971,73,74,
8185,9192,105,108,111,112,
114,144,147,170,174
attentionalcontrol 74,83,85,
92,96
auditorystimuli 13,22,107,
109,112,113
automaticprocesses x,6,18,34
auxiliaries 9,18,169,176
averaging 23,31,32,7072,103
B
basicsentencestructure 170,
171
binding 11,12,172,175,184
boundmorphemes 169,171
C
calibration xi,7577
CHILDES 170
ClosurePositiveShift(CPS) x,
36,49,50
coding xii,77,102105,112
116,118,146,163
cognitiveabilities 89,177,
180,181
cognitivecontrol 74,9092,
177,181
cognitiveimpulsivity 90,177,
179
cohorteffect 148,149,176
commonground 88,91
commonreference(ERP) 71
competence 1,6,23,98,99,119,
125,130,133,135,169,170,173,
180182,184
comprehension 14,8,9,13,
16,18,2326,52,56,58,59,63,
6669,7375,88,90,93101,
104,105,124,131,133135,137,
138,140,142,143,145,146,
150153,157,160,161,163167,
170,173,174,180,183,185
conjunctionheuristic 8687
context 3,4,7,14,19,25,34,37,
46,48,51,56,57,75,88,91,98,
100,131,132,162,174,176,185
continuityhypothesis 30
control 7,9,1015,21,65,74,
84,85,9095,102,107,108,
110,111,122,142,160,172,
177,183
conversation 73,75,182
coordinates 75,76
coreference 4,12
cornealreflection 75,76
counterbalancing 102,106,109
cross-linguistic xiii,8,88,171,
172,178,181,183
cross-modalpriming viii,x,
13,11,14,15,22,98,176
D
dative xii,142144,148150,
152,154159,161,163165
deficits 41,44,53,90,91,178
DerivationalTheoryof
Complexity 173
design ix,9,10,17,21,23,52,
72,86,102,105,107,127,129
developmentalpsycholinguistics
ix,x,xiv,75,81,91
discontinuityhypothesis 30
discourse xi,88,138,140
distracter-initialtrials 117,
119122,124128
double-objectconstruction 16,
17,142,148
Dutch 172
dyslexia x,41,178
E
EarlyLeftAnteriorNegativity
(ELAN) x,34,5961
electrooculogram(EOG) 71,
72
emptycategories 175
endogenous 74,8285,92
Event-RelatedPotentials(ERP)
11,23,26,3043,4564,
6769,71,72,166,178,185
executivefunction 177,181
exogenous 74,8285,92
eyemovements 2,7375,78,
81,82,85,88,89,92,9498,
101104,107,108,111113,
116119,124,125,145,146,148,
149,151,156,158,164
eyeposition 73,76,80,81,82,
85,147
188 Index
eyetracker xi,xii,77,105,145,
146
eyetracking viiixii,33,73,
7577,97,98,104,105,111,112,
126,132,149,150
F
fillertrials 59,79,107,109,152
filler-gapdependencies 11,12,
13,14
filtering 31,40,71,72
finiteverbs 172
firstlanguageacquisition 1,
125,169
fixation 71,78,81,82,93,
110112,114117,120,124,126,
145147,149
fovea 81
French 5,42,43,172
frequency 1823,31,39,44,55,
71,149,176,180
frontallobe 90,91
functionwords 37,170,171,175
functionalcategories 171,175
inflection 18
inhibition 177,181
input xiii,37,42,43,50,51,59,
74,8083,85,91,140,154,155,
171,178,180,181
inter-coderreliability 77,146,
147
intermodalpreferentiallooking
paradigm 101,146,175
intervention xi,42,62,178,182
intonationalphrase 36,37,
4850
intransitive 7,49,140,159,
160,175
K
Korean 88,178
H
headmovements 77,115
head-final 88
head-mountedeyetracker xi,
75,76,145147
L
languagedisorders xiii,178,
179,182
language learning universals
181
languageproduction 2,16,22,
26,29,53,54,95,137,164,166,
178,181,183
latency 3436,41,42,49,50,
61,82,93,96,102,104,105,
112114,123,124,128,149
learnabilitytheory 180
lesion 90
lexicalacquisition 46
lexicalambiguities 149,176
lexicaldecision 1114,124
lexicalpriming 54,56
lexical-semanticprocessing 46,
48,51,52
lexicon 46,47,49,51,99,171,
175,176,180
linguistictheory 170,171,173,
180
linkedreference(ERP) 71
linkingassumptions xi,73,74,
81,85,87,89,92
listeningspan 10
looking-while-listening ix,97,
98,102,104,105,107,109,110,
112,118,124,129,130,132
I
iamb 4345
idealspeaker-hearer 170,173
M
manualtasks 83
maturation 35,41,43,90,177
G
garden-path 7,9092,173
gaze 7581,83,85,86,88,91,
97,98,101,102,104,105,113,
115,116,118,120,124,125,131,
132,145147,149
German 19,42,43,53,150,172
goal-directed 83,86,89
gradedstrengthhypothesis 161
grammar vii,viii,xiii,6,73,
138,139,140,143,170,171,
173175,180,181
grammaticalgender 88,108,
178
grammaticalrepresentations
xii,137,140,160
measure viii,x,2,4,10,15,17,
18,22,31,3234,45,46,53,62,
7375,79,83,89,92,97108,
113116,123,124,128132,145,
149,150,173,175,178,179
memoryspan xiii,10,11,13,
14,15,177,178,181
metalinguistic viii,2,124,
144,179
mismatchnegativity(MMN)
34,35,37,38,41,4345
mismatchresponse(MMR) 38,
40,42,46
morphologicalprocessing ix,
1,19,20
morphology 88,180,181
motiontransients 83
movement 33,71,72,77
movementoperations 172,
175,181
myelination 35,90
N
N400 x,35,37,4648,5158,61
naturalisticspeech 170
navigationalplans 83
negation 170,172
neurocognitivetasks 90
nonfiniteverbs 172
novelverbs 88,106,140143,
152155,157,160162
O
oddballparadigm x,34,3840,
4245,69
ocularartifacts 72
oculardevelopment 81
oculomotor xi,82
off-line vii,xixiii,9,10,71,
97,99,100,105,171,173175,
177,180
on-line 13,5,710,12,14,15,
18,22,25,26,30,32,36,51,62,
9698,129,130,132,135,137,
138,145,146,150,151,153,160,
163,166,167,169,173179;see
real-timemeasures
onset-contingentplot 119123
operatingprinciples 181
outliers 126
overgeneralize 169
P
P600 x,36,37,5961
paradigm ixxii,xiv,5,14,16,
19,3335,3840,4245,47,
5260,69,70,74,75,83,98,
101,102,104,105,118,124,132,
137,139,141,143,145,146,148,
150,153,161,175,176
parallelselection 83
parameter 170,172,181
parsing 7,10,14,73,74,81,
8691,173,177,180,181
passivelisteningparadigm 69
pasttense 18,169
performance vii,viii,xii,xiii,
45,53,105,160,162,169,170,
173,174,177,179,181
phoneme 37,38,40,41,46,48,
55,98,173
phonememonitoring 98
phonological x,31,3538,41,
42,46,48,51,5456,58,61,
92,138,148,149,155,156
phonologicalfamiliarityeffect
48
phonological-lexicalpriming
effect 54,56
phonotactic x,37,4648,61
photoreceptors 81
phrasestructure x,7,8,10,17,
5861,158,171
phrasestructureviolation 58
60
picture-classificationtask 15
picturenaming 12,16,179
picture-viewingtask 78
picture-wordparadigm 47,
5355,70
plausibility 7,176,177,181
polarity 34
poormanseyetracker xi,xii,
77,145,146
positiveevidence 171
preferentiallookingparadigm
xi,3,33,77,85,97,100,102,
104,112,115,132,140,146,160,
162,174,175
prefrontalcortex 90
prepositionalobject
construction 16,17,142,
148,158
Index 189
prepositionalphrase xi,17,59,
86,148,175,177
preschol xi,xii,11,58,112,137,
140,144,146,153,160,176
prescreening 112
priming viii,x,xii,13,11,12,
1418,22,47,51,52,54,56,
88,89,98,137,139,142145,
148163,175,176
PrincipleA 175
PrincipleB 172,175,184
probabilityplot 81,87
proberecognition 35
probetasks 173
pro-dropparameter 172
production 1,2,1622,24,26,
45,5254,58,7375,94,98,
125,130,133,137,140143,
145,148,150,153155,157,158,
160162,164,170,173,180183
profileplot 121123,128
proportionoflooks 7880,87,
149,151,154,157,159
pro-saccade 4
prosody x,1,19,37,4143,
46,4851,61,103,107,108,
109,142
prosodicbootstrapping 37,43
psycholinguistictheory 173
pupil 75,76
push-buttonbox 9,11,13
Q
quantification
88
R
reactiontime viii,ix,9,13,14,
32,104,114,116118,123,124,
125,129,173
reactivation 5,1216,177
reading viii,2,3,68,10,11,15,
22,98,161,176,177,179,182
real-timemeasures xi,18,
75,92,97,102,132,174;see
on-line
reanalysis 36,59,150,177
recency 5,8
recovery 12,9092
reference 73,74,80,8688,
104,105,123
referent 47,73,75,7882,86
92,101,105,108,117,120123,
125,128,129,148,178
referentialcommunicationtask
88,91
regressionequation 76
relativeclause x,8,10,12,80,
86,150,176,178
reliabilitycoding 114
remoteeyetrackingsystem xi,
7577
repetition 14,109,126,144
representations ix,xii,xiii,
1,3,5,14,17,51,55,89,92,
99,116,117,137144,150,151,
153158,160163,175,177,179,
180,182
resourceallocation 177,181
retina 81,82
Russian 88
S
saccades 81,82,84,85,156
saliency 40,43,83,85
salient 83,106,111,177,181
scalpdistribution 34,63
scope xi,88,139,150
scrambling 11
self-pacedlistening viii,x,
13,611,177
self-pacedreading viii,3,68,
10,11,15,22,176
semanticbootstrapping 138,
139
semanticinformation x,5,7,
10,32,5157,123,128,129,138
semanticintegration 47,51,
5658
semanticpriming 12,47,51,54
semanticprocessing 35,36,46,
48,5052,5558,61
semanticviolation 47,5658,
69
sentenceprocessing x,xi,7,
1012,14,15,22,37,74,86,88,
145,146,161,174176
signedlanguages 172,173
simplicityparsingheuristic 88
socio-economicstatus 178
sourcelocalization 32,33,70
190 Index
Spanish 98,107,108,125,172,
178
spatialattention xi,81,82,92
spatialresolution 32
specificlanguageimpairment
(SLI) x,41,42,45,46,53,
54,178,179,182
speechstimuli 97,102,105108,
111113,116,146
speech-languagepathologists
182
speededproduction x,1,2,16,
18,19,21,22
splicing 10,107,122
spokenlanguage xi,xii,22,73,
75,85,97,100,104,105,124,
130,131,137,146,172
stress x,19,37,4246,107,178
structuralpriming xii,17,88,
89,139,142145,148,150154,
157,158,160,161,175
subjectrelative 8,176,177
syntacticambiguity xi,88
syntacticdependencies 2,4,7
syntacticinformation x,31,35,
36,58,60,138
syntacticintegration 59,60,66
syntacticphraseboundary 37,
48
syntacticpriming xii,1618,
137,142,145,150,153
syntacticprocessing xi,1,5,35,
50,60,92,179
syntacticreanalysis 36,59
syntacticviolation 5860,69
syntax 48,49,58,59,61,137,
139,150,158,161,163,170,171,
174,180
syntax-semanticsmappings
137,139,140,148,162,163
T
target-initialtrials 116120,
122,126,127
temporalresolution viii,22,
32,33,102,105,124,145,150
temporaryambiguity 151,154
thematicroles 61,138,140,148,
158,160
timecourse 18,22,32,37,75,
78,80,83,97,98,103105,115,
116,118,119,123,132
timeregion 79
time-sensitivemeasures x,1,2,
4,5,10,18,22
topography 34,71
trackingofthehead 76
transformationalrules 171
transitive xii,7,49,140,142
144,150,159,175
trochee 43,44
truth-valuejudgmenttask 175
U
universalgrammar 138
universalprinciples xiii,171,
172,181
usage-basedtheories 139
V
validity 98,102,129,131,
vector 76
verbbias 88,162,177,178
visualsalience 101,106,149
visualsearch xi,83,9496,124
visualselection 8386
visualstimuli 7,1115,82,100,
105,109114,118,124
visualworldmethod xi,74,75,
8184,88,89,92,98,145,148
vocabularyspurt 125,181
VPattachment 177,178
W
wh-movement 11
WisconsinCardSortingTask
89
wordmonitoring 3,4,173,174
wordorder 5,6,172,175,181
wordrecognition xii,4,61,
88,99,101105,123,124,126,
130,148
wordsegmentation 37,43,45,
46,50
wordstress 37,42,43,69
workingmemory 10,11,1316,
56,83,131
world-situatedeye-gaze
paradigm 137,145
wrap-up 5
In the series Language Acquisition and Language Disorders the following titles have been
published thus far or are scheduled for publication:
A complete list of titles in this series can be found on the publishers website, www.benjamins.com
45 Guijarro-Fuentes, Pedro, Maria Pilar LarraaGa and john CLibbens (eds.): First Language
Acquisition of Morphology and Syntax. Perspectives across languages and learners. Expected April 2008
44 sekerina, irina a., eva M. Fernndez and Harald CLaHsen (eds.): Developmental
Psycholinguistics. On-line methods in childrens language processing. 2008. xviii, 190 pp.
43 saviCkien, ineta and Wolfgang u. dressLer (eds.): The Acquisition of Diminutives. A crosslinguistic perspective. 2007. vi, 352 pp.
42 LeFebvre, Claire, Lydia WHite and Christine jourdan (eds.): L2 Acquisition and Creole Genesis.
Dialogues. 2006. viii, 433 pp.
41 torrens, vincent and Linda esCobar (eds.): The Acquisition of Syntax in Romance Languages. 2006.
viii, 422 pp.
40 deen, kamil ud: The Acquisition of Swahili. 2005. xiv, 241 pp.
39 unsWortH, sharon, teresa Parodi, antonella soraCe and Martha YounG-sCHoLten (eds.):
Paths of Development in L1 and L2 acquisition. In honor of Bonnie D. Schwartz. 2006. viii, 222 pp.
38 FranCesCHina, Florencia: Fossilized Second Language Grammars. The acquisition of grammatical
gender. 2005. xxiv, 288 pp.
37 MontruL, silvina a.: The Acquisition of Spanish. Morphosyntactic development in monolingual and
bilingual L1 acquisition and adult L2 acquisition. 2004. xvi, 413 pp.
36 bartke, susanne and julia sieGMLLer (eds.): Williams Syndrome across Languages. 2004.
xvi, 385 pp.
35 snCHez, Liliana: Quechua-Spanish Bilingualism. Interference and convergence in functional categories.
2003. x, 189 pp.
34 ota, Mitsuhiko: The Development of Prosodic Structure in Early Words. Continuity, divergence and
change. 2003. xii, 224 pp.
33 joseFsson, Gunlg, Christer PLatzaCk and Gisela Hkansson (eds.): The Acquisition of Swedish
Grammar. 2004. vi, 315 pp.
32 Prvost, Philippe and johanne Paradis (eds.): The Acquisition of French in Different Contexts.
Focus on functional categories. 2004. viii, 384 pp.
31 Marinis, Theodoros: The Acquisition of the DP in Modern Greek. 2003. xiv, 261 pp.
30 Hout, roeland van, aafke HuLk, Folkert kuiken and richard j. toWeLL (eds.): The Lexicon
Syntax Interface in Second Language Acquisition. 2003. viii, 234 pp.
29 Fernndez, eva M.: Bilingual Sentence Processing. Relative clause attachment in English and Spanish.
2003. xx, 294 pp.
28 sHiMron, joseph (ed.): Language Processing and Acquisition in Languages of Semitic, Root-Based,
Morphology. 2003. vi, 394 pp.
27 saLaberrY, M. rafael and Yasuhiro sHirai (eds.): The L2 Acquisition of TenseAspect Morphology.
2002. x, 489 pp.
26 sLabakova, roumyana: Telicity in the Second Language. 2001. xii, 236 pp.
25 CarroLL, susanne e.: Input and Evidence. The raw material of second language acquisition. 2001.
xviii, 461 pp.
24 Weissenborn, jrgen and barbara HHLe (eds.): Approaches to Bootstrapping. Phonological, lexical,
syntactic and neurophysiological aspects of early language acquisition. Volume 2. 2001. viii, 337 pp.
23 Weissenborn, jrgen and barbara HHLe (eds.): Approaches to Bootstrapping. Phonological, lexical,
syntactic and neurophysiological aspects of early language acquisition. Volume 1. 2001. xviii, 299 pp.
22 sCHaeFFer, jeannette C.: The Acquisition of Direct Object Scrambling and Clitic Placement. Syntax and
pragmatics. 2000. xii, 187 pp.
21 HersCHensoHn, julia: The Second Time Around Minimalism and L2 Acquisition. 2000. xiv, 287 pp.
20 kanno, kazue (ed.): The Acquisition of Japanese as a Second Language. 1999. xii, 180 pp.
19 beCk, Maria-Luise (ed.): Morphology and its Interfaces in Second Language Knowledge. 1998. x, 387 pp.
18 kLein, elaine C. and Gita MartoHardjono (eds.): The Development of Second Language
Grammars. A generative approach. 1999. vi, 412 pp.