&Cognition1998.26
(5).1002-1013 LexicalandconceptualprocessinginChinese-Englishbilinguals:Further evidenceforasymmetry HIMCHEUNGandHSUAN-CHIHCHENChineseUniversityofHongKong,HongKong ordingtotheasymmetrymodelofbilingualrepresentation(Kroll&Stewart,1994),thefirstlanguage(L1)lexiconiscloselytiedtoanunderlyingconceptualmemory,whereassecondlanguage(L2)itemsaremostlyassociatedwiththeirL1equivalents.AneofthisarchitectureisthatL1-to-L2,orforward,translationmustbemediatedbytheconceptualmemory,whereasL2-to-L1(backward)translationtakesadirectlexicalpath.Somepredictionsderivedfromthishypotheticalstructureweretestedinthepresentstudy,whichtookintoount,throughanalysisofcovariance,variationsinresponseproductiontime,conceptretrievaltime,andsomeothercharacteristicsassociatedwiththeindividualtestitems.ProficientChinese-Englishbilingualsweretestedondelayedproduction(Balota&Chumbley,1985),picturenaming,wordtranslation,andcategorymatching.Theexpectedasymmetricalpatternoftranslationlatencies(i.e.,forward>backward)wasdemonstrated,althoughitcouldbestatisticallyexplainedbytheitemcharacteristicoffamiliarity;matchinganL1itemtoacategorynamewasfasterthanmatchinganL2item,suggestingrelativelystrongL1conceptuallinks.ThepresentresultsarebestmodatedbyaformofasymmetrythatallowsfornondominantL2-conceptlinkage,theuseofwhichisconditionaluponthefamiliarityofthetestitemtothebilingual. KrollandStewart(1994)proposedatheoreticalframeworkforthestudyofthebilingualverbalmemorynamely,theasymmetrymodel,inmunicationsbetweenfirstlanguage(Ll)andsecondlanguage(L2)lexiconsareachievedviatwodifferentpathways.WhenanitemistranslatedfromanLltoanL2(i.e.,forwardtranslation),theunderlyingconceptisfirstactivatedbyitsLllabelandthenrecodedintoitsL2name.Thishypotheticalroutedoesnotassumeanydirectword-towordlinksbetweenthetwolexiconsandregardsconceptactivationasanecessarymediatingstepformunications.WhenanitemistranslatedfromanL2toanLl(i.e.,backwardtranslation),theL2labelisdirectlycodedintothecorrespondingL1nameatalexicallevelandconceptactivationisminimallyinvolved.Thisrouteexemplifiesdirectlexicalassociationsbetweenthetwolexiconsandrequiresnomediationthroughconceptsinbackwardtranslation.Themodelis,therefore,asymmetricalinthesensethatitprescribesqualitativelydif- ThisresearchwassupportedbyaDirectGrantforResearchfromtheSocialScienceandEducationPaneloftheChineseUniversityofHongKongandbyEarmarkedGrantsfromtheResearchGrantsCouncilofHongKongtoH.-
C.C.WethankChing-YeeChowandNga-YanChanfortheirassistanceinrunningtheexperiments.ThefirstauthoriscurrentlyaffiliatedwiththeDepartmentofPsychologyattheUniversityofOtago,NewZealand.CorrespondenceconcerningthisarticleshouldbeaddressedtoH.-
C.Chen,DepartmentofPsychology,TheChineseUniversityofHongKong,Shatin,
N.T.,HongKong(e-mail:ikechen@cuhk.edu.hk). eptedbypreviousassociateeditorKathrynT.Spoehr ferentroutesforthetwotypesoftranslation.ThemodelisdepictedinFigure1. InFigure1,thethicksolidarrowthatpointsbothwaysindicatesaverystronglinkbetweenconceptualmemoryandtheLllexicon.TheL2-to-Llthinsolidarrowrepresentstheprominentmunication,whichisoflesserstrength,responsibleforbackwardtranslation;forwardtranslationisplishedbyfirstgoingthroughthestrongLI-to-conceptrouteandthentheweakconcept-toL2pathrepresentedbythedashedarrow.Inprinciple,forwardtranslationcanalsobeachievedatalexicallevelthroughaweakLl-to-L2Iink.Themodel,therefore,identifieslexical/conceptualassociationsatthreedifferentlevelsofstrength,rangingfromthestrongest,representedbytheLl-conceptpath,totheweakest,realizedbytheL2-conceptlink.SincethethreeconstructsofanL1lexicon,anL2lexicon,andconceptualmemoryarefullyconnectedwithoneanother,themodel'semphasisonthelexicalandconceptualnatureofbackwardandforwardtranslation,respectively,isbynomeansabsolute.Alternativepathwayscouldbeemployed,whennecessary.Themodelsimplyformalizeswhatthebilingualtendstodoinnormalsituations,notwhatheorshemustdounderallpossibleconditions. Owingtoitsuniquearchitecture,theasymmetrymodelpredictsshorterlatenciesforbackwardthanforforwardwordtranslation,sincetheformerdoesnotnormallyrequireconceptess,whereasthelatterusuallynecessitatestheactivationofbothanLl-to-conceptandaconcept-to-L2link.Thisexpectedasymmetricalpatternofwordtranslationlatenciesisexactlywhathasbeen Copyright1998PsychonomicSociety,Inc. 1002 BILINGUALPROCESSING1003 Language1 .. -------. \ Concept Figure1.Aschematicrepresentationoftheasymmetrymodelofbilingualmemory.From"CategoryInterferenceinTranslationandPictureNaming:EvidenceforAsymmetricConnectionsBetweenBilingualMemoryRepresentation,"byJ.F.KroUandE.Stewart,1994,JournalofMemory&Language,33,pp,150.Copyright1994byAcademicPress.Adaptedwithpermission. found(see,e.g.,Kroll&Curley,1988;Kroll&Stewart,1994).Additionalsupportforthemodelhastodowiththesensitivityofthetranslationprocesstosemanticvariables.KrollandStewartreportedthatanizedwordlistshadanegativeeffectontranslationlatenciesinforwardbutnotinbackwardtranslation,indicatingthatonlytheformermighthaveinvolvedconceptactivationand,thus,wassensitivetoanization.Usingadifferentparadigm,Sholl,Sankaranarayanan,andKroll(1995)foundinarecentstudythat,ifconceptswererepeatedacrosspicturenamingandwordtranslation,positivetransferinresponselatenciesurredwithforwardbutnotwithbackwardtranslation.Thisfindingmustbetakenasevidencefordisparityinconceptualprocessingbetweenthetwotranslationconditions,forsuchtransferoftrainingcouldonlytakeplaceatasemanticlevel. Otherstudies,however,stressthesimilarityratherthanthedifferencebetweenthetwotranslationdirectionsanddemonstratepredominantconceptualprocessingforboth.LaHeij,Hooglander,Kerling,andvanderYelden(1996)reportedsignificantStroopfacilitatoryeffects,forbothforwardandbackwardwordtranslation,indicatingqualitativelysimilarconceptualprocessingforthetwolexicons.Indeed,ithasbeenshowninanumberofpreviousstudiesusingdifferentparadigmsthattheL2lexiconhasprominent,directassociationswiththeconceptualmemory,justasdoestheLllexicon(see,e.g.,Altarriba,1990;Chen,1990;Chen&Ho,1986;Chen&Ng,1989;Keatley,Spinks,&deGelder,1994).Thequestion,hence,isnotwhethercertaintranslationpathways exist;rather,itesunderwhatconditionstheactivationofaparticularsetofroutesisespeciallyencouragedandinwhatsituationsitisnot. Asaresult,morerecentdevelopmentintheconceptualizationoftheasymmetrymodelhasaddedontotheprototypicalversionsomeextradimensionsthatactuallyconditionbilinguallexical/conceptualprocessing.Theseconstraintsfurtherstrengthenthepredictivepowerofthemodel.TheworkofDufourandKroll(1995)providesanexampleofsuchendeavor.TheauthorsinvestigatedtheponentofstrengthofconceptualactivationassociatedwithbothLlandL2itemsandmanipulatedintheirstudythefactorofL2fluency.ForagroupofmorefluentandagroupoflessfluentEnglish-Frenchbilinguals,acategory-matchingtaskwasusedinwhichthesubjectwastodecideasquicklyaspossibleifatargetconceptwasamemberofacertaincategory.Boththetargetandthecategorynamevariedinpresentationlanguage(Englishvs.French).Resultsshowedacleardissociationofthepatternofmatchinglatenciesbetweenthetwosubjectgroups.ThemorefluentbilingualwasinsensitivetothecategorynamelanguageandgenerallyperformedbetterwithEnglishthanwithFrenchtargetnames.Thelessfluentbilingual,ontheotherhand,performedbetterwhenthelanguageofthecategorynamematchedthatofthetargetname.Itwasconcludedthat,forthemorefluentbilinguals,bothEnglishandFrenchwordsactivatedsimilarlyrichsemanticrepresentation,and,hence,languageofthecategorynamedidnotmakeadifferenceinmatchingperformance.Forthelessfluentbilinguals,however,FrenchwordsactivatedsignificantlylesssemanticinformationthandidEnglishwords,andsuchdiscrepancymanifesteditselfasaninhibitoryeffectwhentherewasamismatchoftargetandcategorynamelanguageinanexperimentaltrial:AnEnglishwordactivatedanumberofsemanticnodesthatwerenotsharedbythecorrespondingFrenchword,andtheactivationoftheseirrelevantnodesinterferedwithmatchingperformance.Suchinterferencesimplydidnoturwith(1)proficient,orbalanced,bilingualswhohadsimilaractivationlevelsforEnglishandFrenchwords,sothatveryfewirrelevantnodeswereactivatedevenwhentherewasalanguagemismatch;(2)lessfluentbilingualswhenthecategoryandthetargetnamewereinthesamelanguage,sothatnoirrelevantnodeswereactive.TheinterpretationofferedbyDufourandKrollqualifiestheasymmetrymodelbyputtingontheconstraintthatL2proficiencycorrelatespositivelywiththestrengthofL2-conceptassociations;itisinlinewiththefindingsofsomepreviousstudiesthatalsotakeadevelopmentalperspective(see,e.g.,Altarriba&Mathis,1997;Kroll,Dufour,Sholl,&Roufca,1993;Talamas&Kroll,1993;seealsoKroll&Sholl,1992). InadditiontotheindividualcharacteristicofL2proficiency,effectsofvariousstimulusattributesonthetranslationprocesshavealsobeenexplored.InherfirsttwoexperimentsusingthetasksofDutch-to-English(forward)wordtranslationandtranslationrecognition, 1004CHEUNGANDCHEN deGroot(1992a)reportedthatboththeurrencefrequencyandimageabilityoftheDutchstimuluswordsaffectedwordtranslationandtranslationrecognitionlatencies.Wordimageabilitywasoperationalizedastheeasinessofgeneratingconcretementalpictures/imageswhenacertaintestwordwasencountered.Itwasreportedthatbothhigh-frequencyandhigh-imageabilitywordsproducedshortertranslationlatenciesthandidlowfrequencyandlow-imageabilityones.Inanotherexperimentusingthemultipleregressionparadigm,theauthoridentifiedfourwordcharacteristicvariables,eachofwhichuniquelypredictedwordtranslationlatencies:frequencyofstimuluswords,frequencyofresponsewords,contextavailability(i.e.,theeasinessingupwithspecificcontextsinwhichwhatthetestworddenotedtendedtoappear),andcognatestatusoftranslations(i.e.,whetherthetestwordanditstargettranslationweresimilarinspellingandsound).Thesevariableswerenegativelycorrelatedwithtranslationlatencies.Tointerpretthefindings,deGroot(1992a)proposedahierarchicalarchitecturewithword-to-conceptandword-to-wordconnectionsofspecificstrengths.Theconceptualmemoryunderlyingthelexiconswasconceptualizedinadistributedmanner(seealsodeGroot,1992b).Testwordsthatmeasuredhighinthecriticalcharacteristicsactivatedespeciallynumeroussemanticnodes,andthisenhancedthespeedoflinkingupwithtargettranslations.Translationlatencieswerethereforeshortened. Inthesamevein,deGroot,Dannenburg,andvanHell(1994)examinedtheeffectsofvariouswordcharacteristicvariablesonbothforwardandbackwardtranslation.Meaning(semantic)variables,suchasimageability,contextavailability,anddefinitionuracy(i.e.,howeasilyanduratelyatestwordcouldbeverballydefined)associatedwithboththestimulusandtheresponsewords,werefoundtobecorrelatedwithbothforwardandbackwardtranslationlatencies,butsomewhatmorestronglyintheforwardthaninthebackwardcondition.Furthermore,thedifferentialeffectsofthesemanticvariablesonthetwotranslationtaskswerequalifiedbythecognatestatusoftranslations:Fornoncognates(testandtargetwordslookedandsoundeddifferent),thecorrelationbetweenlatencyandthesemanticvariableswashigherintheforwardthaninthebackwardcondition;forcognates(testandtargetwordslookedandsoundedsimilar),nodirectionalityeffectonthecorrelation,whichparabletothatfornoncognatesinbackwardtranslation,wasobserved.Itwasarguedthat,whenthetranslationswerecognates,theirsimilarorthographiesandphonologies(lexicallevelcharacteristics)encouragedtheuseofadirectlexicalpath,regardlessoftranslationdirection,whereas,fornoncognates,theuseofthelexicalpathwasnotparticularlyencouragedbyanylexicallevelsimilaritiesbetweenthetestandthetargetwords,and,therefore,wordtranslationdisplayedtheusualasymmetricalpattern.ThefindingsreportedbybothdeGroot(l992a)anddeGrootetal.(1994)arepatiblewiththeasymmetrymodelinthatforwardtranslationappears tobemoresensitivetosemanticvariablesthanisbackwardtranslation;theyalsoindicatethatstimulusattributesaffecttranslationlatencies,probablybyconstrainingtheactivationofthedifferentroutespostulatedbythemodelduringtranslation. Usingtheasymmetrymodelasaframework,thepresentstudyattemptstolooksystematicallyintothefunctioningofthevariouspathwaysinterconnectingthetwolexiconsandtheconceptualmemoryinproficientChineseEnglishbilinguals.Mostpreviousstudiesintheareahaveexaminedlanguagesbelongingtothesamefamilynamely,theIndo-Europeanlanguagefamily.Forinstance,AltarribaandMathis(1997)usednativeEnglishspeakerswhohadlearnedSpanish;LaHeijetal.(1996)recruitedDutch-Englishbilinguals;KrollandCurley(1988)lookedatEnglishandGerman,whereasDufourandKroll(1995)usedEnglish-Frenchbilinguals.Theselanguagesareallcloselyrelated,andtheydohaveanumberoflexicallevelsimilarities.Hence,oneisnotsurewhetherconclusionsgeneratedinthesestudieswouldapplytobilingualsspeakingtwovirtuallyunrelatedlanguages.IthasbeendocumentedthattheChineselanguagehasanumberofuniquecharacteristicsthatclearlydistinguishitfromEnglishintermsofcognitiveprocessing.Experimentally,forinstance,ithasbeendemonstratedthatlexicaldecisionisgenerallyfasterthannaminginChinese,whereasthereverseistrueinEnglish(seeChen,1992,1996,formoredetaileddiscussions).Orthographicfactorshavebeensuggestedtoountforsuchdiscrepancy(Chen,1996).TheneedtoreplicatebilingualfindingsbasedoncloselyrelatedEuropeantongueswiththedistinctivelanguagesofEnglishandChineseis,therefore,obviousforthegeneralizabilityoftheasymmetrymodel.Potter,So,vonEckardt,andFeldman(1984)diduseChinese-Englishbilinguals,buttheydidnotdirectlytestthemodel.ToevaluatetheuniversalapplicabilityoftheasymmetrymodelwithChineseandEnglishthereforeconstitutesamajormotivationofthepresentstudy. Anotherobjectiveofthisresearchistofurtherexploretheeffectsofcertainitemattributesonbilingualprocessingthathavenotbeensufficientlystudiedbyworkersinthearea.Inasense,thecurrentstudyattemptstoreplicatesomeofdeGroot's(1992a)anddeGrootetal.'s(1994)findingsinordertoobtainconvergingevidencefor,andthusreinforce,thenotionthatbilinguallexical!
conceptualprocessingisconditionaluponcertaintestitemcharacteristics.Thepresentstudy,however,departsfromtheworkofdeGrootandhercolleaguesinthatadifferentapproach(i.e.,theanalysisofcovariance[ANCaVA]procedure)andadifferentpopulation(i.e.,proficientChinese-Englishbilinguals)areused.Picturenamingandwordtranslationlatencieswereusedtoestimatetheextenttowhichaparticularsetofbilingualprocessingrouteswasactive;extrainformationabouttheindividualtestitemswasalsoobtained.First,thetestitemcharacteristicsoffamiliarity,imageability,contextavailability,urrencefrequency,andnumberofphonemeswerecodedandusedascovariatesinaseries BILINGUALPROCESSING1005 ofANCOVAs.Detailsmunicationsamongthetwolexiconsandtheconceptualmemory(asrevealedbypicturenamingandwordtranslation)couldthusbeexaminedbothbeforeandaftertheitemcharacteristicvariableswereequatedacrossexperimentalconditions(i.e.,adjustedtozerobytheANCOVAprocedure).Second,thetasksofdelayedproductionandcategorymatchingwereusedtoestimateresponseproductionandconceptretrievallatencies,respectively,forindividualtestitems,andtheestimateswereusedascovariatesinanothersetofANCOVAs.Thiswastoexaminepatternsofinternalrouteactivationbothbeforeandafterresponseproduction,andconceptretrievallatenciesforthetestitemswereequatedacrossexperimentalconditions. Thementionedproceduresmayneedsomeelaboration.Inexaminingpatternsmunicationamongthelexiconsandtheconceptualmemory,thepresentstudycontrastedwordtranslationwithpicturenaming,ontheonehand,andforwardwithbackwardtranslation,ontheother(seeKroll&Stewart,1994;PotteretaI.,1984).AsSnodgrass(1984)hasargued,drawingconclusionsaboutinternalprocessesparisonsamongdifferenttasksmakestheassumptionthatthesetasksdonotdifferfromoneanotherinprocessesotherthanthoseinquestion.Forinstance,paringforwardtobackwardtranslationlatenciesforinformationabouttheinternaltranslationprocesswouldobviouslyrequiretheequationofresponseproductionacrossthetwotranslationconditions.Responselanguage,however,variessystematicallyacrossthetwoconditions(forwardandbackwardtranslationrequireL2andL1responses,respectively);itislikelythatLlandL2productionlatenciesareinherentlydifferent,independentofanyinternaltranslationprocesses.Similarly,paring,say,Llpicturenamingwithbackwardtranslationlatenciesforinformationaboutinternalprocessingpathswouldrequiretheequationofconceptess(retrieval)speedacrossthetwotasks.Nevertheless,Llpicturenamingalwayshaspicturestimuli,whereasbackwardtranslationusesL2stimulusitems,andthesetwotypesofstimulimaysimplyhaveverydifferentconceptretrievalspeeds.Thepresentstudy,hence,useddelayedproduction(Balota&Chumbley,1985)toestimateresponseproductionlatenciesandcategorymatchingtoestimateconceptretrievaltimesassociatedwithindividualtestitems.Theseestimateswereusedascovariatessothatinternaltranslationprocessescouldbeexaminedbothbeforeandaftertheireffectsweretakenintoount.Theuseofitemfamiliarity,imageability,contextavailability,urrencefrequency,andnumberofphonemesascovariatesinanothersetofANCOVAswasbasedonpreviousfindingsthattheywerecorrelatedwiththetranslationprocess(deGroot,1992a;deGrootetaI.,1994). Therewerefourhypothesesderivedfromtheasymmetrymodelandtestedinthepresentstudy.First,forwardtranslationwouldbeslowerthanbackwardtranslation,becausetheformerinvolvedbothconceptactivationand L2labeling,whereasthelatterrequiredonlyactivatingtheappropriatelexicalpath.Second,Llpicturenamingwouldbefasterthanbackwardtranslation,becauseLIpicturenamingisplishedviaaverystrongL1conceptlink,whereasthelexicalrouteusedinbackwardtranslationisrelativelyweak(seeFigure1).Third,L2picturenamingwouldbealmostasslowasforwardtranslation,becausetheformerisbasedonaweakL2conceptpathandthelatterrequiresbothconceptessandL2labeling.Fourth,matchinganLlitemtoacategorynamewouldbefasterthanmatchinganL2item,duetostrongLl-to-conceptlinks. Thepresentstudyconsistedofthreeexperiments.Experiment1testedthefirstthreehypotheseswithstimulimatchedinrecognitionthresholdacrossconditions.Experiment2hadindependentsubjectsrateallverbaltestitemsontheirfamiliarity,imageability,andcontextavailabilityandallpicturetestitemsontheirfamiliarityandcontextavailability.TheseratingswereusedtogetherwithurrencefrequenciesandnumberofphonemesforverbalitemsascovariatesinaseriesofANCOVAsthatretestedthefirstthreehypotheses.InExperiment3,delayedproductionandcategory-matchinglatencieswereobtainedforthetestitemsandusedascovariatestoreevaluatethesamethreehypotheses.Thefourthpredictionwasalsotestedwithcategory-matchingdata. EXPERIMENT1 Experiment1testedagroupofproficientChineseEnglishbilingualsonpicturenamingandwordtranslation.Comparisonsoflatenciesamongthesetaskspotentiallyrevealthebilingualmemorystructure(PotteretaI.,1984).Thepresenthypotheses,ordingtotheasymmetrymodel(Kroll&Stewart,1994),werethat(1)forwardtranslationwouldbeslowerthanbackwardtranslation;(2)namingpicturesinLlwouldbefasterthanbackwardtranslation;and(3)namingpicturesinL2wouldtakeaslongasforwardtranslation.Testitemsmatchedinrecognitionthresholdwereused. Method Subjects.Thirty-sixundergraduatesattheChineseUniversityofHongKong,whowereproficientChinese-Englishbilinguals,participatedinExperimentI.ThesesubjectshadbeenlearningEnglishformallyintheclassroomforatleast12yearsatthetimeoftesting.Allthesubjectsreportednormalorcorrected-to-normaleyesight. MaterialsandApparatus.FortypicturesandtheircorrespondingChineseandEnglishwrittennames,totaling120items,wereusedasstimuli.The40pictureshadbeenexperimentallyselectedfromSnodgrassandVanderwart's(1980)standardizedpicturesetinanearlierstudydoneinourlaboratory(Chen,Cheung,&Lau,1997).Theprincipleofselectionwasthatmeanrecognitionthresholdsacrossthethreeformsofpicture,Chineseword,andEnglishwordforanygivenpicture-conceptsshouldbeofleastdiscrepancies.Ineachselectiontrial.thesubjectwasinstructedtonameasquicklyaspossibleavisuallypresentedstimulus.whichcouldbeapicture.aChineseword.oranEnglishword.PicturesandChi- 1006CHEUNGANDCHEN nesewordsrequiredChineseresponses,whereasEnglishwordsrequiredEnglishresponses.Thestimuluswasinitiallypresentedfor20msec,followedbyaI.S-secmask.Ifthesubjectdidnotrespond,thestimuluswouldbepresentedagainwithalO-msecincreaseinpresentationduration.Thesequencewasrepeateduntilthestimuluswascorrectlynamedoruntilthepresentationdurationexceeded120msec.AGerbrandsT-4Afour-fieldtachistoscopewasusedforstimuluspresentation.Asaresult,80pictureconceptswereselected,withmeanrecognitionthresholds(inmilliseconds)of49.1,40.1,and47.0forthestimulustypesofpicture,Chineseword,andEnglishword,respectively;thecorrespondingerrorrateswere7.7,2.5,and3.0.Themaximumbetween-stimulustypedifferenceinthresholdwasonly9msec(cf.Potteretal.,1984,whoreportedthattherecognitionthresholdforpictureswas22msechigherthanthatforChinesewords).The40picture-conceptsusedinthepresentexperimentwererandomlyselectedfromthese80concepts. Stimuliweredisplayedonan80486puterinExperimentI.Areasofthedisplayswereapproximatelyequalforpictures,Chinesewords,andEnglishwords.puterwasconnectedtoavoicedetectionkeythatpedthetimerofputerupondetectingavocalresponse. Procedure.ExperimentIconsistedoftwoblocks(ChineseandEnglish)of10picture-namingtrialsandtwoblocks(forwardandbackward)of10wordtranslationtrials.Eachblockwasprecededby2practicetrials.Conceptswithineachblockwerepresentedinthesameorder.Theorderofblockpresentationwascounterbalancedacrosssubjects.Instructionsweregivenorallybytheexperimenterbeforeeachblock.Noconceptwasrepeatedinanyformforanysubject.Tobeginapicture-namingtrial,asquareofarraysofsmallcrosseswasdisplayedforIsectoserveasamaskandastartingsignal.Itwasfollowedbyapicturestimulusthatstayedonthemonitoruntilthesubjectresponded.Thesubjectwasinstructedtorespondbysayingthenameofthepictureoutloudasquicklyanduratelyaspossible.Nofeedbackwasgiven.Ifthesubjectdidnotknowtheanswer,heorshehadtogivethevocalresponse"skip."Theexperimentersatbesidethesubjectthroughouttheexperimentandrecordedthemissingtrials.Naminglatencywasdefinedasthetimeintervalbetweentheonsetofthepicturestimulusandtheonsetofthesubject'svocalresponse,whichwaspickedupbyavoicedetectionkey.Theexperimentalprocedureofwordtranslationwasthesameaspicturenaming,exceptthatwordsinsteadofpictureswerepresentedandtranslationsratherthannamesofpictureswererequired. ResultsandDiscussionPercenterror.Errorratesinthefourexperimental conditionsarepresentedinTable1.Drivenbythehypotheses,contrastsbetweenforward andbackwardtranslation,Llpicturenamingandbackwardtranslation,andL2picturenamingandforwardtranslationweretested.Theonlysignificantcontrastwasthatbetweenforwardandbackwardtranslation[t(78)=2.28,p=.026]indicatingagreaterpercenterrorinforwardthaninbackwardtranslation. Latencies:Analysesbysubject.Meanresponselatenciesinthevariousconditionswerecalculated,usingthesubjectastheunitofanalysis.TheyarepresentedinTable1.First,forwardtranslationwassignificantlyslowerthanbackwardtranslation[F(1,35)=6.35,MSe=34,052.6,p=.016].Thisresultisconsistentwiththeclaimofasymmetry,whichspecifiesmoreprocessingstagesforforwardthanforbackwardtranslation.Second,L1picture Table1MeanPictureNamingandWordTranslationLatencies(inMilliseconds;WithPercentError)forExperiment1: AnalysisbySubject ResponseLanguage Chinese English Task
M %
M % Picturenaming 992 5.0 1,241 13.6 Wordtranslation 1,239 9.4 1,348 16.7 namingwasfasterthanbackwardtranslation[F(l,35)=68.04,MSe=16,078.5,p<.001].ThisresultconfirmsthehypotheticalcontrastbetweenaverystrongL1conceptlinkandaprominentbutweakerL2-to-L1directlexicallinkspecifiedbythemodel.Third,namingpicturesinL2tookaboutaslongastranslatingL1itemsintotheirL2equivalents[F(l,35)=3.02,MSe=44,952.9,P>.05].ThisfindingpatiblewiththeclaimthatforwardtranslationinvolvesbothdetailedconceptualprocessingandL2labelingandis,thus,asslowasL2picturenaming,whichisbasedonaweakconceptualpath. Latencies:Analysesbyitem.Meanresponselatencieswerealsocalculatedusingthetestitemastheunitofanalysis.Forwardtranslation,again,wasslowerthanback- wardtranslation[F(1,78)=4.47,MSe=80,854.1,p<.04]. Thisresultpatiblewiththeasymmetryclaim,whichspecifiesmoreprocessingstagesforforwardthanforbackwardtranslation.L1picturenamingwasfasterthanbackwardtranslation[F(l,78)=29.11,MSe=41,364.7,p<.01];namingpicturesinL2tookaboutaslongastranslatingLlitemsintotheirL2equivalents[F(l,78)=2.02,MSe=98,650.l,p>.10].Thepatternofresultsisidenticaltothatbasedontheindividualsubjects. Usingstimulusitemswithverysimilarrecognitionthresholdsacrossexperimentalconditions,Experiment1obtaineddatasupportiveoftheasymmetrymodel.Bothanalysesbysubjectandanalysesbytestitemrevealedthatforwardtranslationwasslowerthanbackwardtranslation,suggestingthatverbalitemsinthetwolexiconsweredifferentiallyprocessed.Forwardtranslationwasalsomoreerror-pronethanbackwardtranslation.Thefactthatbilingualsspeakingtwoverydifferentlanguages(oneIndo-EuropeanandoneEastAsian)wereexaminedinthepresentstudyfurtheraddstotheuniversalapplicabilityoftheasymmetrymodel,asevidenceforthemodelhasbyandlargebeenbasedoncloselyrelatedEuropeantongues.CommunicationsfromtheLltotheL2lexiconappearedtobeconceptuallybased(i.e.,requiringbothconceptactivationandL2labeling),forforwardtranslationwasasslowasL2picturenaming,whichreliedonaweakL2-conceptpath.L1picturenamingwasfasterthanbackwardtranslation,thusconfirmingthehypotheticalcontrastbetweenastrongLI-concepttieandarelativelyweakL2-to-Lllexicallink.Overall,thepresentresultsdosuggestaformofasymmetrythatspecifies BILINGUALPROCESSING1007 (1)differentinternalprocessingstagesforL1andL2itemsand(2)lexical/conceptualassociationsofdifferentstrengths. EXPERIMENT2 AsshownbydeGrootetal.(1994),certainitemlevelcharacteristicsarecorrelatedwiththeprocessingoflexicalitemstoasignificantextent.Althoughexactlyhowthesecharacteristicsaffectlexicalprocessingisnotwellunderstood,itisreasonablethattheireffectsshouldbetakenintoconsiderationwheninterpretingdataonvariouslexicalprocessingmechanisms.Experiment2obtainedmeasuresofsomeitemcharacteristicsassociatedwiththestimuliusedintheprecedingexperiment,usingthemascovariatesinaseriesofANCOVAsthatreanalyzedthepicture-namingandwordtranslationdataofExperiment1andretestedthethreehypotheses.SubjectiveratingsfromanindependentgroupofproficientChineseEnglishbilingualsonthefamiliarity,imageability,andcontextavailabilityoftheverbaltestitems,aswellastheirurrencefrequenciesandnumberofphonemes,wereobtained.Forthepicturestimuli,familiarityandcontextavailabilityratingsweregathered.ThisexperimentdepartedfromtheworkofdeGrootetal.andwouldthusaddtoitbecause(1)thisexperimentexaminedthedistinctivelanguagesofChineseandEnglish,theinteractionsbetweenwhichmightbeverydifferentfromthoseamongrelatedEuropeanlanguages,and(2)thisexperimentusedtheANCOVAmethod,whichalloweddirectevaluationoftheasymmetrymodelandstatisticalcontrolofitemattributesatthesametime. Method Subjects.TwelveundergraduateandgraduatestudentsattheChineseUniversityofHongKongwereaskedtoratethetestitemsusedintheprecedingexperimentoncertaincharacteristics.ThesesubjectswereallproficientChinese-EnglishbilingualswhohadbeenlearningEnglishasasecondlanguageformallyforatleast12yearsatthetimeoftesting.TheirgenerallanguagebackgroundandL2(English)proficiencywereveryparablewiththoseoftheExperimentIsubjectsbecause:(I)thetwosubjectgroupshadundergonethesamelanguagecurriculuminhighschool;(2)theyhad,byandlarge,thesamelanguageexperienceathomeintermsofthepredominantlymonolingualenvironmentofspokenlanguage(i.e.,Cantonese-Chinese);and(3)theyhadbeenrequiredtodemonstratethroughexaminationacertainlevelofEnglishproficiencybeforetheycouldenrollattheuniversity.Allthesubjectsinthepresentexperimentreportednormalorcorrected-to-normaleyesight.NoneofthemhadparticipatedinExperiment1. Materials.Thesame120stimulusitemsasthoseusedinExperimentI(i.e.,40picture-concepts,togetherwiththeirChineseandEnglishwrittennames)wereagainusedinthepresentexperiment. Procedure.Thesubjectswereaskedtoratethe120itemsonthefollowingcharacteristicsbyusinga5-pointscale. Familiarity(bothwordsandpictures).Thiswasdefinedasthenumberoftimesthesubjecthadexperiencedtheitem(deGrootetal.,1994;Noble,1953),withI=neverseen,heard.oruseditbeforeand5=he/shehadseen,heard,oruseditalmosteveryday. lmageabi/ity(wordsonly).Thiswasdefinedashoweasilyandquicklyitwaspossibleeupwithamentalimageinresponsetotheitem(deGrootetal.,1994;Paivio,Yuille,&Madigan,1968),with1=verydifficultandslowand5=veryeasyandquick. Contextavailability(bothwordsandpictures).Thiswasdefinedashoweasyitwastoproducecontextsorcircumstancesinwhichtheitemmightappear(deGrootetaI.,1994;Schwanenflugel,Harnishfeger,&Stowe,1988),withI=verydifficultand5=veryeasy. Instructionswerepresentedtothesubjectbothorallybytheexperimenterandinprintedform.ThelanguageofinstructionwasChinese,withwhichthepresentsubjectsweremorefamiliar.Itwasstressedthatwhatshouldberatedwerecharacteristicsofthewordsorthepicture-imagesthemselves,nottheconceptsthattheyrepresented. Inadditiontothementionedcharacteristics,thevariablesofurrencefrequencyandnumberofphonemeswerealsodefinedfortheChineseandEnglishitems.FrequenciesoftheEnglishwordswereobtainedfromFrancisandKucera(1982)andthoseoftheChineseitemsfromAFrequencyCountforWordsCommonlyUsedbyHongKongJuniorSecondarySchoolStudents(HongKongDepartmentofEducation,1986). ResultsandDiscussionBecausethecovariates(i.e.,theitemcharacteristic measures)obtainedinthecurrentexperimentweretiedtotheindividualtestitems,performanceofanANCOVAwaspossibleonlywiththetestitemastheunitofanalysis.Allsubsequentanalysesare,therefore,basedonindividualtestitems.MeansforthevariousitemcharacteristicvariablesarepresentedinTable2. One-wayANOVAswereperformedinordertoexaminedifferencesinthesemeasuresamongpicture,Chinese,andEnglishstimuli.Significantdifferenceswerefoundonlyforcontextavailability[F(2,117)=12.27,MSe=0.16,p<.001]andurrencefrequency[F(l,78)=4.81,MSe=8,972.6,p=.03].Forcontextavailability,boththecontrastbetweenpicturesandChinesewords[F(l,78)=23.87,MSe=0.16,P<.001]andthatbetweenpictures andEnglishwords[F(l,78)=11.57,MSe=0.13,p= .001]weresignificant.TheseanalysesshowedthatthepresentlyusedEnglishtestwordsweremorefrequentthantheirChinesetranslations;thepicturestimuliweremoreeasilyputintocontextsthanboththeChineseandtheEnglishwords. RetestingthehypothesesofExperiment1.MeasuresoftheitemcharacteristicswereusedascovariatesinasetofANCOVAsthatreanalyzedthepicture-namingandwordtranslationdataofExperiment1,usingthetest Table2MeansfortheItemCharacteristicVariables(WithStandard Deviations)inExperiment2 Pictures Chinese English Variable MSDMSDM SD Familiarity 4.5 Contextavailability4.4 Imageability Numberofphonemes Frequency 0.24.40.44.50.30.34.00.54.10.4 4.30.34.40.24.21.63.91.828.056.174.5121.7 Note-Frequency=urrencefrequencyperonemillionwords;allmeasuresotherthanfrequencyareona5-pointscale. 1008CHEUNGANDCHEN itemastheunitofanalysis(simplecorrelationsbetweenpicturenaming,wordtranslationlatencies[fromExperiment1],andtherelevantitemcharacteristicvariablesareshowninAppendixA).Thereanalysesweredrivenbythepreviouslytestedhypotheses.First,ithadbeenhypothesizedthatforwardtranslationwouldbeslowerthanbackwardtranslation.parisoninvolvedonlywordstimuli,and,therefore,allthevariableswereenteredascovariates.Thefivecovariatestogetherountedforasignificantportionofthetranslationlatencyvariance[F(5,72)=4.41,MSe=64,731.5,p=.001].Theonlysignificantcovariatewasfamiliarity[t(72)=-2.68,p=.009].Aftertheadjustment,forwardtranslationwasnofasterorslowerthanbackwardtranslation[F(l,72)=2.18,MSe=64,731.5,p=.14],withadjustedmeansof1,345.1and1,257.0,respectively.Thefirsthypothesisderivedfromtheasymmetrymodelwasnotsupported. Second,ithadbeenhypothesizedthatL1picturenamingwouldbefasterthanbackwardtranslation.parisoninvolvedbothwordandpicturestimuli,and,thus,onlythecovariatesofcontextavailabilityandfamiliaritywereused.Thetwocovariatestogetherountedforasignificantportionofthenaming/translationlatencyvariance[F(2,76)=10.34,MSe=33,370.l,p<.001].Nevertheless,onlyfamiliaritywasasignificantcovariate[t(76)=-3.42,p=.001].Aftertheadjustment,Llpicturenamingwasstillfasterthanbackwardtranslation[F(l,76)=26.79,MSe=33,370.1,P<.001],withadjustedmeansofl,000.9and1,233.3,respectively.Thesecondhypothesiswassupported. Third,ithadbeenhypothesizedthatL2picturenamingwouldbenofasterorslowerthanforwardtranslation.parisoninvolvedbothwordandpicturestimuli,and,therefore,onlythecovariatesofcontextavailabilityandfamiliaritywereused.Thetwocovariatestogetherountedforasignificantportionofthenaming/translationlatencyvariance[F(2,76)=8.0,MSe=83,840.5,p=.001].Nevertheless,onlyfamiliaritywasasignificantcovariate[t(76)=-2.66,p=.01].Aftertheadjustment,L2picturenamingwasnofasterorslowerthanforwardtranslation[F(l,76)=0.28,MSe=83,840.5,p=.60],withadjustedmeansof1,301.8and1,344.3,respectively.Thethirdhypothesiswassupported. Theonlychangeofresultsfromtheprecedingtothepresentexperiment,whichmeasuredcertainitemcharacteristicsandusedthemascovariatesinthereanalyses,wasthatthetwowordtranslationtasksnolongerdisplayeddifferentresponselatencies.ThestraightforwardinterpretationisthatthedifferenceintranslationtimefoundinExperiment1couldpossiblybeexplainedbyitemfamiliarity,independentofanyprocessingasymmetries.Nevertheless,thepresentANCOVAsonlyusethetestitemastheanalysisunit,andcautionmustthereforebetakennottooverinterprettheresults.Whatthepresentfindingssuggestisanalternativewaytointerpretthetranslationtimeasymmetrysofrequentlyobservedinpreviousstudies(whichdidnotcontrolforvariationsintestitemfamiliarity);theactivationpatterns ofthetranslationroutesspecifiedbytheasymmetrymodelmayactuallybeconditionaluponthefamiliarityofthetestitems.Futureresearchintheareashould,therefore,(1)attemptsomeformofcontroloveritemfamiliaritysoastorenderthedatamoreinterpretableand(2)exploreinamoreconcretewayhowitemfamiliarityaffectspatternsmunicationamongthelexiconsandtheconceptualmemory. EXPERIMENT3 TheaimofExperiment3istwofold.(l)ItevaluatesthepropositionthattheL1lexiconhasstrongerconceptuallinksthantheL2lexicon.
(2)UsinganANCOVA,itreteststhehypothesesofthetwoprecedingexperimentsbyintroducingresponseproductionandconceptretrievallatenciesascovariates.DerivedfromthefirstobjectiveisthenewhypothesisthatmatchinganL1itemtoacategorynamewouldbefasterthanmatchinganL2item,assumingthatstrongconceptuallinksfacilitateconceptessand,therefore,category-matchingperformance.Thesecondobjectivewasmotivatedbytheanalysisofpicture-namingandwordtranslationlatenciesintoponents.Someofponents,suchasthetimeneededforproducingovertresponsesandthetimeneededforessingconceptsinresponsetostimuli,areirrelevanttothetranslationpathwaysspecifiedbytheasymmetrymodel.Forinstance,thecontrastbetweenforwardandbackwardtranslationtimesparingalatencycontainingL2responseproductiontime(i.e.,forwardtranslationlatency)withalatencycontainingLlresponseproductiontime(i.e.,backwardtranslationlatency).Sinceresponseproductiontimeisirrelevanttotheissueunderexamination,systematicvariationsinitshouldbecontrolledinsomeform.Asimilarproblemariseswhenpicture-naminglatencyparedagainstwordtranslationtime,fortheformercontainsconceptretrievaltimeinresponsetoapicture,whereasthelattercontainsconceptretrievaltimeinresponsetoaword,anditisconceivablethattherearesystematicbutextraneousvariationsinthetimeneededforconceptessduetodifferentstimulustypes(picturesvs.words).Possiblevariationsinresponseproductionandconceptesstimeshavenotbeencontrolledbypreviousstudies.OnthebasisofthesametestitemsasthoseusedinExperimentIbutadifferentgroupofproficientChineseEnglishbilinguals,thepresentexperimentemployedthetasksofdelayedproductionandcategorymatchingtoestimateeachtestitem'sresponseproductionandconceptretrievaltime,respectively. Therationaleforusingdelayedproductiontoindicatesheervocalresponselatencywasthatprocessingstagesprecedingresponseproduction,suchasstimulusrecognitionandphonologicalcodeess,wouldhavepletedbytheendofthe650-msecdelay(i.e.,theintervalbetweenstimulusonsetandresponsesignal)usedinthepresentexperiment.BalotaandChumbley(1985)usedthesametaskwiththesamedelayandobtainedsta- BILINGUALPROCESSING1009 bleresults.Categorymatchingwasusedinthepresentstudytoindicateconceptretrievaltime.Categoriesofdifferentbroadnesswereused,becausematchingacertaintargettoacategorynamemightnotrequirefullactivationandpreciseknowledgeoftheconceptunderlyingthattarget.Tojudgewhetheranitembelongedtoacertaincategorymightonlyneedinformationaboutsomeparticularaspectsoftheitem,suchasanimacyandmobility,dependingonthebroadnessofthecategoryname.Itwas,therefore,importanttousecategoriesofdifferentbroadness,whichsupposedlydemandeddifferentlevelsofconceptactivation,toexaminetheconceptretrievalprocess.Empirically,JaredandSeidenberg(1991)diddemonstratethatcategoriesofdifferentbroadnessbehaveddifferentlyinasemanticdecisiontask. Inthepresentexperiment,estimatesobtainedfrombothdelayedproductionandcategorymatchingwereusedascovariatesinaseriesofANCOVAsthatreanalyzedthepicture-namingandwordtranslationdataobtainedinExperimentIandretestedthethreehypotheses. Method Subjects.Thirty-sixproficientChinese-Englishbilingualswererecruitedinthepresentexperiment.Noneofthemhadparticipatedinthetwopreviousexperiments.TheirgenerallanguagebackgroundandL2(English)proficiencywereveryparabletothoseofExperiments1and2subjectsbecause(1)theyhadundergonethesamelanguagecurriculuminhighschool;(2)theyhadgenerallythesamelanguageexperienceathomeintermsofthepredominantlymonolingualenvironmentofspokenlanguage(i.e.,Cantonese-Chinese);and(3)theyhadbeenrequiredtodemonstratethroughexaminationacertainlevelofEnglishproficiencybeforetheycouldenrollattheuniversity. MaterialsandApparatus.MaterialsandapparatuswerethesameasthoseinExperiment1.Newmaterialsweretwotypesofcategorynamesforuseinthecategory-matchingtask:Therewere2remotecategories(livingandnonlivingthings)and21immediatecategories(furniture,electricalappliances,meansoftransport,mammal,fruit,weapon,metallicinstrument,stationery,an,clothing,thingstoread,silverware,timer,plant,mechanicaldevice,container,,musicalinstrument,building,signal,andvegetables).Duringcategorymatching,immediatecategorieswereassumedtodemandfullerconceptactivationofthetargetthandidmoreremoteones,duetotheirrelativelyhighsemanticspecificity.Theuseofthesetwotypesofcategorynamessupposedlyhaving differentconceptualdistancesfromanytargetitems,therefore,providedanopportunitytoassesstheeffectsproducedbydifferentdegreesofconceptactivation.The21immediatecategorynamesweregeneratedby10independentsubjects,whowereaskedtoproduceamosttypical,e-to-mindsuperordinatecategorynameforeachofthe40conceptsunderlyingthe120testitems.Interjudgeagreementonproducingthese21categorynamesforthe40conceptswashigherthan84%. Procedure.TheprocedureofdelayedproductionfollowedthatusedbyBalotaandChumbley(1985),anditwasidenticaltothatofpicturenaminginExperiment1,exceptthat(l)wordsinsteadofpictureswerepresentedand(2)thesubjectwasrequirednottonamethestimuluswordasquicklyaspossiblebuttodelaynaminguntilapairofparenthesescameonbracketingthestimulusword.Theparenthesesalwaysappeared650msecaftertheonsetofthetestword(Balota&Chumbley,1985).Responselatencywasdefinedastheintervalbetweentheonsetoftheparenthesesandtheonsetofthevocalresponse.Thisprocedurewasadoptedinordertoensurearelativelypuremeasureofmereresponseproductiontime,forprocessing,suchasessingtheappropriatephonologicalcode,andpreparationformotorresponseshouldhavepletedbytheendofthe650-msecdelay. Forcategorymatching,thesubjectwasrequiredtodecideasquicklyanduratelyaspossiblewhetherthepresentedstimulus(picture,Chineseword,orEnglishword)belongedtoasuperordinatecategory.Thefourfactors(i.e.,stimulustype,positivevs.negativematch,languageofthecategoryname,andimmediatevs.remotecategories)werecounterbalancedacrossconceptsandsubjects.Halfofthestimuliineachblockmatchedthecategoryname.Eachcategory-matchingtrialbeganwithasquareofarraysofsmallcrosses.Itwasfollowedbyasuperordinatecategoryname(EnglishorChinese),whichwasdisplayedfor1sec.Thestimulus(target)thenfollowedimmediately(i.e.,SOA=1,000msec),anditstayedonthemonitoruntilthesubjectrespondedbypressingeitherthedesignatedyesorthedesignatednoresponsekeyonthekeyboard.Nofeedbackwasgiven.Categorymatchinglatencywasdefinedastheintervalbetweentheonsetofthetargetandthekeypress. ResultsandDiscussionPercenterror.Errorratesforbothdelayedproduction andcategorymatchingarepresentedinTable3.Theonlyreliabledifferencewasthatbetweenimmediateandremotecategoriesforthecategory-matchingtask[F(1,117)=13.79,MSe=0.03,p<.001].Matchingitemstoremotecategorieswasingeneralmoreerrorpronethanmatchingitemstoimmediatecategories. Table3MeanDelayedProductionandCategoryMatchingLatencies(inMilliseconds; WithPercentError)forExperiment3:AnalysisbySubject Target Chinese English Pictures Task
M SD
M SD
M Delayedproduction 595 1.9 601 4.9 Immediatecategorymatching 971 7.1 1,093 7.9 956 9.2 (Chinesecategorynames) Immediatecategorymatching 1,068 7.8 1,229 8.1 1,129 8.8 (Englishcategorynames) Remotecategorymatching 1,154 13.8 1,370 21.7 1,099 12.5 (Chinesecategorynames) Remotecategorymatching 1,10612.5 1,207 18.1 L179 8.8 (Englishcategorynames) Note-Onlylatenciesofcorrect(andalsopositiveforcategorymatching)responsesarereported. 1010CHEUNGANDCHEN Latencies:Analysesbysubject.Meandelayedproductionandcategory-matchinglatencieswerecalculated,usingthesubjectastheunitofanalysis.TheyareshowninTable3.Fordelayedproduction,noeffectoftestlanguagewasfound.Forcategorymatching,onlythelatenciesofpositiveandcorrectresponseswereanalyzedandusedtoindicateconceptretrievaltimes.Thejustificationwasthatarelativelypureconceptessprocesscouldbeassumedforpositiveandcorrectresponses,whereasnegativeorincorrectresponsesmighthaveinvolvedother,additionalprocesses.Category-matchinglatenciesfrompositiveandcorrecttrialswereanalyzedbya3(Chinese,English,andpicturestimuli)X2(Chinesevs.Englishcategorynames)X2(immediatevs.remotecategories)mixeddesignANOVA.Categorynamewasaninsignificantfactor,nordiditinteractwithanyothereffects.Boththeimmediateversusremotecategoriescontrast[F(1,34)=8.76,MSe=76,590.8,p=.006]andtheeffectofstimulustype[F(2,68)=5.76,MSe=84,747.5,p=.005]weresignificant,thoughtheirinteractionwasnot.Theresultsindicatedthatthereexistedoveralldifferencesamongcategory-matchingtimesinresponsetopicture,Chinese,andEnglishitems.Morespecifically,forimmediatematching,EnglishitemsweremoreslowlymatchedthanbothChinese[t(35)=-2.34,p=.02]andpictureitems[t(35)=3.61,p=.001].RemotematchingwasslowerforEnglishthanforChineseitems[t(35)=-2.28,p=.03];thecontrastbetweenEnglishandpictureitemswasnotsignificant.TheshortermatchinglatenciesassociatedwiththeChinesethanwiththeEnglishitemsareindicativeofstrongerconceptuallinksfortheformer.ThelackofacategorynamelanguageeffectisconsistentwiththeresultsreportedbyDufourandKroll(1995),whoalsofailedtofindsucheffectinrelativelyfluentEnglish-Frenchbilinguals,asthecurrentsubjectswerealsoconsideredproficientintheirL2.Matchinganitemtoaremotecategorywas,inthepresentexperiment,generallyslower
(5).1002-1013 LexicalandconceptualprocessinginChinese-Englishbilinguals:Further evidenceforasymmetry HIMCHEUNGandHSUAN-CHIHCHENChineseUniversityofHongKong,HongKong ordingtotheasymmetrymodelofbilingualrepresentation(Kroll&Stewart,1994),thefirstlanguage(L1)lexiconiscloselytiedtoanunderlyingconceptualmemory,whereassecondlanguage(L2)itemsaremostlyassociatedwiththeirL1equivalents.AneofthisarchitectureisthatL1-to-L2,orforward,translationmustbemediatedbytheconceptualmemory,whereasL2-to-L1(backward)translationtakesadirectlexicalpath.Somepredictionsderivedfromthishypotheticalstructureweretestedinthepresentstudy,whichtookintoount,throughanalysisofcovariance,variationsinresponseproductiontime,conceptretrievaltime,andsomeothercharacteristicsassociatedwiththeindividualtestitems.ProficientChinese-Englishbilingualsweretestedondelayedproduction(Balota&Chumbley,1985),picturenaming,wordtranslation,andcategorymatching.Theexpectedasymmetricalpatternoftranslationlatencies(i.e.,forward>backward)wasdemonstrated,althoughitcouldbestatisticallyexplainedbytheitemcharacteristicoffamiliarity;matchinganL1itemtoacategorynamewasfasterthanmatchinganL2item,suggestingrelativelystrongL1conceptuallinks.ThepresentresultsarebestmodatedbyaformofasymmetrythatallowsfornondominantL2-conceptlinkage,theuseofwhichisconditionaluponthefamiliarityofthetestitemtothebilingual. KrollandStewart(1994)proposedatheoreticalframeworkforthestudyofthebilingualverbalmemorynamely,theasymmetrymodel,inmunicationsbetweenfirstlanguage(Ll)andsecondlanguage(L2)lexiconsareachievedviatwodifferentpathways.WhenanitemistranslatedfromanLltoanL2(i.e.,forwardtranslation),theunderlyingconceptisfirstactivatedbyitsLllabelandthenrecodedintoitsL2name.Thishypotheticalroutedoesnotassumeanydirectword-towordlinksbetweenthetwolexiconsandregardsconceptactivationasanecessarymediatingstepformunications.WhenanitemistranslatedfromanL2toanLl(i.e.,backwardtranslation),theL2labelisdirectlycodedintothecorrespondingL1nameatalexicallevelandconceptactivationisminimallyinvolved.Thisrouteexemplifiesdirectlexicalassociationsbetweenthetwolexiconsandrequiresnomediationthroughconceptsinbackwardtranslation.Themodelis,therefore,asymmetricalinthesensethatitprescribesqualitativelydif- ThisresearchwassupportedbyaDirectGrantforResearchfromtheSocialScienceandEducationPaneloftheChineseUniversityofHongKongandbyEarmarkedGrantsfromtheResearchGrantsCouncilofHongKongtoH.-
C.C.WethankChing-YeeChowandNga-YanChanfortheirassistanceinrunningtheexperiments.ThefirstauthoriscurrentlyaffiliatedwiththeDepartmentofPsychologyattheUniversityofOtago,NewZealand.CorrespondenceconcerningthisarticleshouldbeaddressedtoH.-
C.Chen,DepartmentofPsychology,TheChineseUniversityofHongKong,Shatin,
N.T.,HongKong(e-mail:ikechen@cuhk.edu.hk). eptedbypreviousassociateeditorKathrynT.Spoehr ferentroutesforthetwotypesoftranslation.ThemodelisdepictedinFigure1. InFigure1,thethicksolidarrowthatpointsbothwaysindicatesaverystronglinkbetweenconceptualmemoryandtheLllexicon.TheL2-to-Llthinsolidarrowrepresentstheprominentmunication,whichisoflesserstrength,responsibleforbackwardtranslation;forwardtranslationisplishedbyfirstgoingthroughthestrongLI-to-conceptrouteandthentheweakconcept-toL2pathrepresentedbythedashedarrow.Inprinciple,forwardtranslationcanalsobeachievedatalexicallevelthroughaweakLl-to-L2Iink.Themodel,therefore,identifieslexical/conceptualassociationsatthreedifferentlevelsofstrength,rangingfromthestrongest,representedbytheLl-conceptpath,totheweakest,realizedbytheL2-conceptlink.SincethethreeconstructsofanL1lexicon,anL2lexicon,andconceptualmemoryarefullyconnectedwithoneanother,themodel'semphasisonthelexicalandconceptualnatureofbackwardandforwardtranslation,respectively,isbynomeansabsolute.Alternativepathwayscouldbeemployed,whennecessary.Themodelsimplyformalizeswhatthebilingualtendstodoinnormalsituations,notwhatheorshemustdounderallpossibleconditions. Owingtoitsuniquearchitecture,theasymmetrymodelpredictsshorterlatenciesforbackwardthanforforwardwordtranslation,sincetheformerdoesnotnormallyrequireconceptess,whereasthelatterusuallynecessitatestheactivationofbothanLl-to-conceptandaconcept-to-L2link.Thisexpectedasymmetricalpatternofwordtranslationlatenciesisexactlywhathasbeen Copyright1998PsychonomicSociety,Inc. 1002 BILINGUALPROCESSING1003 Language1 .. -------. \ Concept Figure1.Aschematicrepresentationoftheasymmetrymodelofbilingualmemory.From"CategoryInterferenceinTranslationandPictureNaming:EvidenceforAsymmetricConnectionsBetweenBilingualMemoryRepresentation,"byJ.F.KroUandE.Stewart,1994,JournalofMemory&Language,33,pp,150.Copyright1994byAcademicPress.Adaptedwithpermission. found(see,e.g.,Kroll&Curley,1988;Kroll&Stewart,1994).Additionalsupportforthemodelhastodowiththesensitivityofthetranslationprocesstosemanticvariables.KrollandStewartreportedthatanizedwordlistshadanegativeeffectontranslationlatenciesinforwardbutnotinbackwardtranslation,indicatingthatonlytheformermighthaveinvolvedconceptactivationand,thus,wassensitivetoanization.Usingadifferentparadigm,Sholl,Sankaranarayanan,andKroll(1995)foundinarecentstudythat,ifconceptswererepeatedacrosspicturenamingandwordtranslation,positivetransferinresponselatenciesurredwithforwardbutnotwithbackwardtranslation.Thisfindingmustbetakenasevidencefordisparityinconceptualprocessingbetweenthetwotranslationconditions,forsuchtransferoftrainingcouldonlytakeplaceatasemanticlevel. Otherstudies,however,stressthesimilarityratherthanthedifferencebetweenthetwotranslationdirectionsanddemonstratepredominantconceptualprocessingforboth.LaHeij,Hooglander,Kerling,andvanderYelden(1996)reportedsignificantStroopfacilitatoryeffects,forbothforwardandbackwardwordtranslation,indicatingqualitativelysimilarconceptualprocessingforthetwolexicons.Indeed,ithasbeenshowninanumberofpreviousstudiesusingdifferentparadigmsthattheL2lexiconhasprominent,directassociationswiththeconceptualmemory,justasdoestheLllexicon(see,e.g.,Altarriba,1990;Chen,1990;Chen&Ho,1986;Chen&Ng,1989;Keatley,Spinks,&deGelder,1994).Thequestion,hence,isnotwhethercertaintranslationpathways exist;rather,itesunderwhatconditionstheactivationofaparticularsetofroutesisespeciallyencouragedandinwhatsituationsitisnot. Asaresult,morerecentdevelopmentintheconceptualizationoftheasymmetrymodelhasaddedontotheprototypicalversionsomeextradimensionsthatactuallyconditionbilinguallexical/conceptualprocessing.Theseconstraintsfurtherstrengthenthepredictivepowerofthemodel.TheworkofDufourandKroll(1995)providesanexampleofsuchendeavor.TheauthorsinvestigatedtheponentofstrengthofconceptualactivationassociatedwithbothLlandL2itemsandmanipulatedintheirstudythefactorofL2fluency.ForagroupofmorefluentandagroupoflessfluentEnglish-Frenchbilinguals,acategory-matchingtaskwasusedinwhichthesubjectwastodecideasquicklyaspossibleifatargetconceptwasamemberofacertaincategory.Boththetargetandthecategorynamevariedinpresentationlanguage(Englishvs.French).Resultsshowedacleardissociationofthepatternofmatchinglatenciesbetweenthetwosubjectgroups.ThemorefluentbilingualwasinsensitivetothecategorynamelanguageandgenerallyperformedbetterwithEnglishthanwithFrenchtargetnames.Thelessfluentbilingual,ontheotherhand,performedbetterwhenthelanguageofthecategorynamematchedthatofthetargetname.Itwasconcludedthat,forthemorefluentbilinguals,bothEnglishandFrenchwordsactivatedsimilarlyrichsemanticrepresentation,and,hence,languageofthecategorynamedidnotmakeadifferenceinmatchingperformance.Forthelessfluentbilinguals,however,FrenchwordsactivatedsignificantlylesssemanticinformationthandidEnglishwords,andsuchdiscrepancymanifesteditselfasaninhibitoryeffectwhentherewasamismatchoftargetandcategorynamelanguageinanexperimentaltrial:AnEnglishwordactivatedanumberofsemanticnodesthatwerenotsharedbythecorrespondingFrenchword,andtheactivationoftheseirrelevantnodesinterferedwithmatchingperformance.Suchinterferencesimplydidnoturwith(1)proficient,orbalanced,bilingualswhohadsimilaractivationlevelsforEnglishandFrenchwords,sothatveryfewirrelevantnodeswereactivatedevenwhentherewasalanguagemismatch;(2)lessfluentbilingualswhenthecategoryandthetargetnamewereinthesamelanguage,sothatnoirrelevantnodeswereactive.TheinterpretationofferedbyDufourandKrollqualifiestheasymmetrymodelbyputtingontheconstraintthatL2proficiencycorrelatespositivelywiththestrengthofL2-conceptassociations;itisinlinewiththefindingsofsomepreviousstudiesthatalsotakeadevelopmentalperspective(see,e.g.,Altarriba&Mathis,1997;Kroll,Dufour,Sholl,&Roufca,1993;Talamas&Kroll,1993;seealsoKroll&Sholl,1992). InadditiontotheindividualcharacteristicofL2proficiency,effectsofvariousstimulusattributesonthetranslationprocesshavealsobeenexplored.InherfirsttwoexperimentsusingthetasksofDutch-to-English(forward)wordtranslationandtranslationrecognition, 1004CHEUNGANDCHEN deGroot(1992a)reportedthatboththeurrencefrequencyandimageabilityoftheDutchstimuluswordsaffectedwordtranslationandtranslationrecognitionlatencies.Wordimageabilitywasoperationalizedastheeasinessofgeneratingconcretementalpictures/imageswhenacertaintestwordwasencountered.Itwasreportedthatbothhigh-frequencyandhigh-imageabilitywordsproducedshortertranslationlatenciesthandidlowfrequencyandlow-imageabilityones.Inanotherexperimentusingthemultipleregressionparadigm,theauthoridentifiedfourwordcharacteristicvariables,eachofwhichuniquelypredictedwordtranslationlatencies:frequencyofstimuluswords,frequencyofresponsewords,contextavailability(i.e.,theeasinessingupwithspecificcontextsinwhichwhatthetestworddenotedtendedtoappear),andcognatestatusoftranslations(i.e.,whetherthetestwordanditstargettranslationweresimilarinspellingandsound).Thesevariableswerenegativelycorrelatedwithtranslationlatencies.Tointerpretthefindings,deGroot(1992a)proposedahierarchicalarchitecturewithword-to-conceptandword-to-wordconnectionsofspecificstrengths.Theconceptualmemoryunderlyingthelexiconswasconceptualizedinadistributedmanner(seealsodeGroot,1992b).Testwordsthatmeasuredhighinthecriticalcharacteristicsactivatedespeciallynumeroussemanticnodes,andthisenhancedthespeedoflinkingupwithtargettranslations.Translationlatencieswerethereforeshortened. Inthesamevein,deGroot,Dannenburg,andvanHell(1994)examinedtheeffectsofvariouswordcharacteristicvariablesonbothforwardandbackwardtranslation.Meaning(semantic)variables,suchasimageability,contextavailability,anddefinitionuracy(i.e.,howeasilyanduratelyatestwordcouldbeverballydefined)associatedwithboththestimulusandtheresponsewords,werefoundtobecorrelatedwithbothforwardandbackwardtranslationlatencies,butsomewhatmorestronglyintheforwardthaninthebackwardcondition.Furthermore,thedifferentialeffectsofthesemanticvariablesonthetwotranslationtaskswerequalifiedbythecognatestatusoftranslations:Fornoncognates(testandtargetwordslookedandsoundeddifferent),thecorrelationbetweenlatencyandthesemanticvariableswashigherintheforwardthaninthebackwardcondition;forcognates(testandtargetwordslookedandsoundedsimilar),nodirectionalityeffectonthecorrelation,whichparabletothatfornoncognatesinbackwardtranslation,wasobserved.Itwasarguedthat,whenthetranslationswerecognates,theirsimilarorthographiesandphonologies(lexicallevelcharacteristics)encouragedtheuseofadirectlexicalpath,regardlessoftranslationdirection,whereas,fornoncognates,theuseofthelexicalpathwasnotparticularlyencouragedbyanylexicallevelsimilaritiesbetweenthetestandthetargetwords,and,therefore,wordtranslationdisplayedtheusualasymmetricalpattern.ThefindingsreportedbybothdeGroot(l992a)anddeGrootetal.(1994)arepatiblewiththeasymmetrymodelinthatforwardtranslationappears tobemoresensitivetosemanticvariablesthanisbackwardtranslation;theyalsoindicatethatstimulusattributesaffecttranslationlatencies,probablybyconstrainingtheactivationofthedifferentroutespostulatedbythemodelduringtranslation. Usingtheasymmetrymodelasaframework,thepresentstudyattemptstolooksystematicallyintothefunctioningofthevariouspathwaysinterconnectingthetwolexiconsandtheconceptualmemoryinproficientChineseEnglishbilinguals.Mostpreviousstudiesintheareahaveexaminedlanguagesbelongingtothesamefamilynamely,theIndo-Europeanlanguagefamily.Forinstance,AltarribaandMathis(1997)usednativeEnglishspeakerswhohadlearnedSpanish;LaHeijetal.(1996)recruitedDutch-Englishbilinguals;KrollandCurley(1988)lookedatEnglishandGerman,whereasDufourandKroll(1995)usedEnglish-Frenchbilinguals.Theselanguagesareallcloselyrelated,andtheydohaveanumberoflexicallevelsimilarities.Hence,oneisnotsurewhetherconclusionsgeneratedinthesestudieswouldapplytobilingualsspeakingtwovirtuallyunrelatedlanguages.IthasbeendocumentedthattheChineselanguagehasanumberofuniquecharacteristicsthatclearlydistinguishitfromEnglishintermsofcognitiveprocessing.Experimentally,forinstance,ithasbeendemonstratedthatlexicaldecisionisgenerallyfasterthannaminginChinese,whereasthereverseistrueinEnglish(seeChen,1992,1996,formoredetaileddiscussions).Orthographicfactorshavebeensuggestedtoountforsuchdiscrepancy(Chen,1996).TheneedtoreplicatebilingualfindingsbasedoncloselyrelatedEuropeantongueswiththedistinctivelanguagesofEnglishandChineseis,therefore,obviousforthegeneralizabilityoftheasymmetrymodel.Potter,So,vonEckardt,andFeldman(1984)diduseChinese-Englishbilinguals,buttheydidnotdirectlytestthemodel.ToevaluatetheuniversalapplicabilityoftheasymmetrymodelwithChineseandEnglishthereforeconstitutesamajormotivationofthepresentstudy. Anotherobjectiveofthisresearchistofurtherexploretheeffectsofcertainitemattributesonbilingualprocessingthathavenotbeensufficientlystudiedbyworkersinthearea.Inasense,thecurrentstudyattemptstoreplicatesomeofdeGroot's(1992a)anddeGrootetal.'s(1994)findingsinordertoobtainconvergingevidencefor,andthusreinforce,thenotionthatbilinguallexical!
conceptualprocessingisconditionaluponcertaintestitemcharacteristics.Thepresentstudy,however,departsfromtheworkofdeGrootandhercolleaguesinthatadifferentapproach(i.e.,theanalysisofcovariance[ANCaVA]procedure)andadifferentpopulation(i.e.,proficientChinese-Englishbilinguals)areused.Picturenamingandwordtranslationlatencieswereusedtoestimatetheextenttowhichaparticularsetofbilingualprocessingrouteswasactive;extrainformationabouttheindividualtestitemswasalsoobtained.First,thetestitemcharacteristicsoffamiliarity,imageability,contextavailability,urrencefrequency,andnumberofphonemeswerecodedandusedascovariatesinaseries BILINGUALPROCESSING1005 ofANCOVAs.Detailsmunicationsamongthetwolexiconsandtheconceptualmemory(asrevealedbypicturenamingandwordtranslation)couldthusbeexaminedbothbeforeandaftertheitemcharacteristicvariableswereequatedacrossexperimentalconditions(i.e.,adjustedtozerobytheANCOVAprocedure).Second,thetasksofdelayedproductionandcategorymatchingwereusedtoestimateresponseproductionandconceptretrievallatencies,respectively,forindividualtestitems,andtheestimateswereusedascovariatesinanothersetofANCOVAs.Thiswastoexaminepatternsofinternalrouteactivationbothbeforeandafterresponseproduction,andconceptretrievallatenciesforthetestitemswereequatedacrossexperimentalconditions. Thementionedproceduresmayneedsomeelaboration.Inexaminingpatternsmunicationamongthelexiconsandtheconceptualmemory,thepresentstudycontrastedwordtranslationwithpicturenaming,ontheonehand,andforwardwithbackwardtranslation,ontheother(seeKroll&Stewart,1994;PotteretaI.,1984).AsSnodgrass(1984)hasargued,drawingconclusionsaboutinternalprocessesparisonsamongdifferenttasksmakestheassumptionthatthesetasksdonotdifferfromoneanotherinprocessesotherthanthoseinquestion.Forinstance,paringforwardtobackwardtranslationlatenciesforinformationabouttheinternaltranslationprocesswouldobviouslyrequiretheequationofresponseproductionacrossthetwotranslationconditions.Responselanguage,however,variessystematicallyacrossthetwoconditions(forwardandbackwardtranslationrequireL2andL1responses,respectively);itislikelythatLlandL2productionlatenciesareinherentlydifferent,independentofanyinternaltranslationprocesses.Similarly,paring,say,Llpicturenamingwithbackwardtranslationlatenciesforinformationaboutinternalprocessingpathswouldrequiretheequationofconceptess(retrieval)speedacrossthetwotasks.Nevertheless,Llpicturenamingalwayshaspicturestimuli,whereasbackwardtranslationusesL2stimulusitems,andthesetwotypesofstimulimaysimplyhaveverydifferentconceptretrievalspeeds.Thepresentstudy,hence,useddelayedproduction(Balota&Chumbley,1985)toestimateresponseproductionlatenciesandcategorymatchingtoestimateconceptretrievaltimesassociatedwithindividualtestitems.Theseestimateswereusedascovariatessothatinternaltranslationprocessescouldbeexaminedbothbeforeandaftertheireffectsweretakenintoount.Theuseofitemfamiliarity,imageability,contextavailability,urrencefrequency,andnumberofphonemesascovariatesinanothersetofANCOVAswasbasedonpreviousfindingsthattheywerecorrelatedwiththetranslationprocess(deGroot,1992a;deGrootetaI.,1994). Therewerefourhypothesesderivedfromtheasymmetrymodelandtestedinthepresentstudy.First,forwardtranslationwouldbeslowerthanbackwardtranslation,becausetheformerinvolvedbothconceptactivationand L2labeling,whereasthelatterrequiredonlyactivatingtheappropriatelexicalpath.Second,Llpicturenamingwouldbefasterthanbackwardtranslation,becauseLIpicturenamingisplishedviaaverystrongL1conceptlink,whereasthelexicalrouteusedinbackwardtranslationisrelativelyweak(seeFigure1).Third,L2picturenamingwouldbealmostasslowasforwardtranslation,becausetheformerisbasedonaweakL2conceptpathandthelatterrequiresbothconceptessandL2labeling.Fourth,matchinganLlitemtoacategorynamewouldbefasterthanmatchinganL2item,duetostrongLl-to-conceptlinks. Thepresentstudyconsistedofthreeexperiments.Experiment1testedthefirstthreehypotheseswithstimulimatchedinrecognitionthresholdacrossconditions.Experiment2hadindependentsubjectsrateallverbaltestitemsontheirfamiliarity,imageability,andcontextavailabilityandallpicturetestitemsontheirfamiliarityandcontextavailability.TheseratingswereusedtogetherwithurrencefrequenciesandnumberofphonemesforverbalitemsascovariatesinaseriesofANCOVAsthatretestedthefirstthreehypotheses.InExperiment3,delayedproductionandcategory-matchinglatencieswereobtainedforthetestitemsandusedascovariatestoreevaluatethesamethreehypotheses.Thefourthpredictionwasalsotestedwithcategory-matchingdata. EXPERIMENT1 Experiment1testedagroupofproficientChineseEnglishbilingualsonpicturenamingandwordtranslation.Comparisonsoflatenciesamongthesetaskspotentiallyrevealthebilingualmemorystructure(PotteretaI.,1984).Thepresenthypotheses,ordingtotheasymmetrymodel(Kroll&Stewart,1994),werethat(1)forwardtranslationwouldbeslowerthanbackwardtranslation;(2)namingpicturesinLlwouldbefasterthanbackwardtranslation;and(3)namingpicturesinL2wouldtakeaslongasforwardtranslation.Testitemsmatchedinrecognitionthresholdwereused. Method Subjects.Thirty-sixundergraduatesattheChineseUniversityofHongKong,whowereproficientChinese-Englishbilinguals,participatedinExperimentI.ThesesubjectshadbeenlearningEnglishformallyintheclassroomforatleast12yearsatthetimeoftesting.Allthesubjectsreportednormalorcorrected-to-normaleyesight. MaterialsandApparatus.FortypicturesandtheircorrespondingChineseandEnglishwrittennames,totaling120items,wereusedasstimuli.The40pictureshadbeenexperimentallyselectedfromSnodgrassandVanderwart's(1980)standardizedpicturesetinanearlierstudydoneinourlaboratory(Chen,Cheung,&Lau,1997).Theprincipleofselectionwasthatmeanrecognitionthresholdsacrossthethreeformsofpicture,Chineseword,andEnglishwordforanygivenpicture-conceptsshouldbeofleastdiscrepancies.Ineachselectiontrial.thesubjectwasinstructedtonameasquicklyaspossibleavisuallypresentedstimulus.whichcouldbeapicture.aChineseword.oranEnglishword.PicturesandChi- 1006CHEUNGANDCHEN nesewordsrequiredChineseresponses,whereasEnglishwordsrequiredEnglishresponses.Thestimuluswasinitiallypresentedfor20msec,followedbyaI.S-secmask.Ifthesubjectdidnotrespond,thestimuluswouldbepresentedagainwithalO-msecincreaseinpresentationduration.Thesequencewasrepeateduntilthestimuluswascorrectlynamedoruntilthepresentationdurationexceeded120msec.AGerbrandsT-4Afour-fieldtachistoscopewasusedforstimuluspresentation.Asaresult,80pictureconceptswereselected,withmeanrecognitionthresholds(inmilliseconds)of49.1,40.1,and47.0forthestimulustypesofpicture,Chineseword,andEnglishword,respectively;thecorrespondingerrorrateswere7.7,2.5,and3.0.Themaximumbetween-stimulustypedifferenceinthresholdwasonly9msec(cf.Potteretal.,1984,whoreportedthattherecognitionthresholdforpictureswas22msechigherthanthatforChinesewords).The40picture-conceptsusedinthepresentexperimentwererandomlyselectedfromthese80concepts. Stimuliweredisplayedonan80486puterinExperimentI.Areasofthedisplayswereapproximatelyequalforpictures,Chinesewords,andEnglishwords.puterwasconnectedtoavoicedetectionkeythatpedthetimerofputerupondetectingavocalresponse. Procedure.ExperimentIconsistedoftwoblocks(ChineseandEnglish)of10picture-namingtrialsandtwoblocks(forwardandbackward)of10wordtranslationtrials.Eachblockwasprecededby2practicetrials.Conceptswithineachblockwerepresentedinthesameorder.Theorderofblockpresentationwascounterbalancedacrosssubjects.Instructionsweregivenorallybytheexperimenterbeforeeachblock.Noconceptwasrepeatedinanyformforanysubject.Tobeginapicture-namingtrial,asquareofarraysofsmallcrosseswasdisplayedforIsectoserveasamaskandastartingsignal.Itwasfollowedbyapicturestimulusthatstayedonthemonitoruntilthesubjectresponded.Thesubjectwasinstructedtorespondbysayingthenameofthepictureoutloudasquicklyanduratelyaspossible.Nofeedbackwasgiven.Ifthesubjectdidnotknowtheanswer,heorshehadtogivethevocalresponse"skip."Theexperimentersatbesidethesubjectthroughouttheexperimentandrecordedthemissingtrials.Naminglatencywasdefinedasthetimeintervalbetweentheonsetofthepicturestimulusandtheonsetofthesubject'svocalresponse,whichwaspickedupbyavoicedetectionkey.Theexperimentalprocedureofwordtranslationwasthesameaspicturenaming,exceptthatwordsinsteadofpictureswerepresentedandtranslationsratherthannamesofpictureswererequired. ResultsandDiscussionPercenterror.Errorratesinthefourexperimental conditionsarepresentedinTable1.Drivenbythehypotheses,contrastsbetweenforward andbackwardtranslation,Llpicturenamingandbackwardtranslation,andL2picturenamingandforwardtranslationweretested.Theonlysignificantcontrastwasthatbetweenforwardandbackwardtranslation[t(78)=2.28,p=.026]indicatingagreaterpercenterrorinforwardthaninbackwardtranslation. Latencies:Analysesbysubject.Meanresponselatenciesinthevariousconditionswerecalculated,usingthesubjectastheunitofanalysis.TheyarepresentedinTable1.First,forwardtranslationwassignificantlyslowerthanbackwardtranslation[F(1,35)=6.35,MSe=34,052.6,p=.016].Thisresultisconsistentwiththeclaimofasymmetry,whichspecifiesmoreprocessingstagesforforwardthanforbackwardtranslation.Second,L1picture Table1MeanPictureNamingandWordTranslationLatencies(inMilliseconds;WithPercentError)forExperiment1: AnalysisbySubject ResponseLanguage Chinese English Task
M %
M % Picturenaming 992 5.0 1,241 13.6 Wordtranslation 1,239 9.4 1,348 16.7 namingwasfasterthanbackwardtranslation[F(l,35)=68.04,MSe=16,078.5,p<.001].ThisresultconfirmsthehypotheticalcontrastbetweenaverystrongL1conceptlinkandaprominentbutweakerL2-to-L1directlexicallinkspecifiedbythemodel.Third,namingpicturesinL2tookaboutaslongastranslatingL1itemsintotheirL2equivalents[F(l,35)=3.02,MSe=44,952.9,P>.05].ThisfindingpatiblewiththeclaimthatforwardtranslationinvolvesbothdetailedconceptualprocessingandL2labelingandis,thus,asslowasL2picturenaming,whichisbasedonaweakconceptualpath. Latencies:Analysesbyitem.Meanresponselatencieswerealsocalculatedusingthetestitemastheunitofanalysis.Forwardtranslation,again,wasslowerthanback- wardtranslation[F(1,78)=4.47,MSe=80,854.1,p<.04]. Thisresultpatiblewiththeasymmetryclaim,whichspecifiesmoreprocessingstagesforforwardthanforbackwardtranslation.L1picturenamingwasfasterthanbackwardtranslation[F(l,78)=29.11,MSe=41,364.7,p<.01];namingpicturesinL2tookaboutaslongastranslatingLlitemsintotheirL2equivalents[F(l,78)=2.02,MSe=98,650.l,p>.10].Thepatternofresultsisidenticaltothatbasedontheindividualsubjects. Usingstimulusitemswithverysimilarrecognitionthresholdsacrossexperimentalconditions,Experiment1obtaineddatasupportiveoftheasymmetrymodel.Bothanalysesbysubjectandanalysesbytestitemrevealedthatforwardtranslationwasslowerthanbackwardtranslation,suggestingthatverbalitemsinthetwolexiconsweredifferentiallyprocessed.Forwardtranslationwasalsomoreerror-pronethanbackwardtranslation.Thefactthatbilingualsspeakingtwoverydifferentlanguages(oneIndo-EuropeanandoneEastAsian)wereexaminedinthepresentstudyfurtheraddstotheuniversalapplicabilityoftheasymmetrymodel,asevidenceforthemodelhasbyandlargebeenbasedoncloselyrelatedEuropeantongues.CommunicationsfromtheLltotheL2lexiconappearedtobeconceptuallybased(i.e.,requiringbothconceptactivationandL2labeling),forforwardtranslationwasasslowasL2picturenaming,whichreliedonaweakL2-conceptpath.L1picturenamingwasfasterthanbackwardtranslation,thusconfirmingthehypotheticalcontrastbetweenastrongLI-concepttieandarelativelyweakL2-to-Lllexicallink.Overall,thepresentresultsdosuggestaformofasymmetrythatspecifies BILINGUALPROCESSING1007 (1)differentinternalprocessingstagesforL1andL2itemsand(2)lexical/conceptualassociationsofdifferentstrengths. EXPERIMENT2 AsshownbydeGrootetal.(1994),certainitemlevelcharacteristicsarecorrelatedwiththeprocessingoflexicalitemstoasignificantextent.Althoughexactlyhowthesecharacteristicsaffectlexicalprocessingisnotwellunderstood,itisreasonablethattheireffectsshouldbetakenintoconsiderationwheninterpretingdataonvariouslexicalprocessingmechanisms.Experiment2obtainedmeasuresofsomeitemcharacteristicsassociatedwiththestimuliusedintheprecedingexperiment,usingthemascovariatesinaseriesofANCOVAsthatreanalyzedthepicture-namingandwordtranslationdataofExperiment1andretestedthethreehypotheses.SubjectiveratingsfromanindependentgroupofproficientChineseEnglishbilingualsonthefamiliarity,imageability,andcontextavailabilityoftheverbaltestitems,aswellastheirurrencefrequenciesandnumberofphonemes,wereobtained.Forthepicturestimuli,familiarityandcontextavailabilityratingsweregathered.ThisexperimentdepartedfromtheworkofdeGrootetal.andwouldthusaddtoitbecause(1)thisexperimentexaminedthedistinctivelanguagesofChineseandEnglish,theinteractionsbetweenwhichmightbeverydifferentfromthoseamongrelatedEuropeanlanguages,and(2)thisexperimentusedtheANCOVAmethod,whichalloweddirectevaluationoftheasymmetrymodelandstatisticalcontrolofitemattributesatthesametime. Method Subjects.TwelveundergraduateandgraduatestudentsattheChineseUniversityofHongKongwereaskedtoratethetestitemsusedintheprecedingexperimentoncertaincharacteristics.ThesesubjectswereallproficientChinese-EnglishbilingualswhohadbeenlearningEnglishasasecondlanguageformallyforatleast12yearsatthetimeoftesting.TheirgenerallanguagebackgroundandL2(English)proficiencywereveryparablewiththoseoftheExperimentIsubjectsbecause:(I)thetwosubjectgroupshadundergonethesamelanguagecurriculuminhighschool;(2)theyhad,byandlarge,thesamelanguageexperienceathomeintermsofthepredominantlymonolingualenvironmentofspokenlanguage(i.e.,Cantonese-Chinese);and(3)theyhadbeenrequiredtodemonstratethroughexaminationacertainlevelofEnglishproficiencybeforetheycouldenrollattheuniversity.Allthesubjectsinthepresentexperimentreportednormalorcorrected-to-normaleyesight.NoneofthemhadparticipatedinExperiment1. Materials.Thesame120stimulusitemsasthoseusedinExperimentI(i.e.,40picture-concepts,togetherwiththeirChineseandEnglishwrittennames)wereagainusedinthepresentexperiment. Procedure.Thesubjectswereaskedtoratethe120itemsonthefollowingcharacteristicsbyusinga5-pointscale. Familiarity(bothwordsandpictures).Thiswasdefinedasthenumberoftimesthesubjecthadexperiencedtheitem(deGrootetal.,1994;Noble,1953),withI=neverseen,heard.oruseditbeforeand5=he/shehadseen,heard,oruseditalmosteveryday. lmageabi/ity(wordsonly).Thiswasdefinedashoweasilyandquicklyitwaspossibleeupwithamentalimageinresponsetotheitem(deGrootetal.,1994;Paivio,Yuille,&Madigan,1968),with1=verydifficultandslowand5=veryeasyandquick. Contextavailability(bothwordsandpictures).Thiswasdefinedashoweasyitwastoproducecontextsorcircumstancesinwhichtheitemmightappear(deGrootetaI.,1994;Schwanenflugel,Harnishfeger,&Stowe,1988),withI=verydifficultand5=veryeasy. Instructionswerepresentedtothesubjectbothorallybytheexperimenterandinprintedform.ThelanguageofinstructionwasChinese,withwhichthepresentsubjectsweremorefamiliar.Itwasstressedthatwhatshouldberatedwerecharacteristicsofthewordsorthepicture-imagesthemselves,nottheconceptsthattheyrepresented. Inadditiontothementionedcharacteristics,thevariablesofurrencefrequencyandnumberofphonemeswerealsodefinedfortheChineseandEnglishitems.FrequenciesoftheEnglishwordswereobtainedfromFrancisandKucera(1982)andthoseoftheChineseitemsfromAFrequencyCountforWordsCommonlyUsedbyHongKongJuniorSecondarySchoolStudents(HongKongDepartmentofEducation,1986). ResultsandDiscussionBecausethecovariates(i.e.,theitemcharacteristic measures)obtainedinthecurrentexperimentweretiedtotheindividualtestitems,performanceofanANCOVAwaspossibleonlywiththetestitemastheunitofanalysis.Allsubsequentanalysesare,therefore,basedonindividualtestitems.MeansforthevariousitemcharacteristicvariablesarepresentedinTable2. One-wayANOVAswereperformedinordertoexaminedifferencesinthesemeasuresamongpicture,Chinese,andEnglishstimuli.Significantdifferenceswerefoundonlyforcontextavailability[F(2,117)=12.27,MSe=0.16,p<.001]andurrencefrequency[F(l,78)=4.81,MSe=8,972.6,p=.03].Forcontextavailability,boththecontrastbetweenpicturesandChinesewords[F(l,78)=23.87,MSe=0.16,P<.001]andthatbetweenpictures andEnglishwords[F(l,78)=11.57,MSe=0.13,p= .001]weresignificant.TheseanalysesshowedthatthepresentlyusedEnglishtestwordsweremorefrequentthantheirChinesetranslations;thepicturestimuliweremoreeasilyputintocontextsthanboththeChineseandtheEnglishwords. RetestingthehypothesesofExperiment1.MeasuresoftheitemcharacteristicswereusedascovariatesinasetofANCOVAsthatreanalyzedthepicture-namingandwordtranslationdataofExperiment1,usingthetest Table2MeansfortheItemCharacteristicVariables(WithStandard Deviations)inExperiment2 Pictures Chinese English Variable MSDMSDM SD Familiarity 4.5 Contextavailability4.4 Imageability Numberofphonemes Frequency 0.24.40.44.50.30.34.00.54.10.4 4.30.34.40.24.21.63.91.828.056.174.5121.7 Note-Frequency=urrencefrequencyperonemillionwords;allmeasuresotherthanfrequencyareona5-pointscale. 1008CHEUNGANDCHEN itemastheunitofanalysis(simplecorrelationsbetweenpicturenaming,wordtranslationlatencies[fromExperiment1],andtherelevantitemcharacteristicvariablesareshowninAppendixA).Thereanalysesweredrivenbythepreviouslytestedhypotheses.First,ithadbeenhypothesizedthatforwardtranslationwouldbeslowerthanbackwardtranslation.parisoninvolvedonlywordstimuli,and,therefore,allthevariableswereenteredascovariates.Thefivecovariatestogetherountedforasignificantportionofthetranslationlatencyvariance[F(5,72)=4.41,MSe=64,731.5,p=.001].Theonlysignificantcovariatewasfamiliarity[t(72)=-2.68,p=.009].Aftertheadjustment,forwardtranslationwasnofasterorslowerthanbackwardtranslation[F(l,72)=2.18,MSe=64,731.5,p=.14],withadjustedmeansof1,345.1and1,257.0,respectively.Thefirsthypothesisderivedfromtheasymmetrymodelwasnotsupported. Second,ithadbeenhypothesizedthatL1picturenamingwouldbefasterthanbackwardtranslation.parisoninvolvedbothwordandpicturestimuli,and,thus,onlythecovariatesofcontextavailabilityandfamiliaritywereused.Thetwocovariatestogetherountedforasignificantportionofthenaming/translationlatencyvariance[F(2,76)=10.34,MSe=33,370.l,p<.001].Nevertheless,onlyfamiliaritywasasignificantcovariate[t(76)=-3.42,p=.001].Aftertheadjustment,Llpicturenamingwasstillfasterthanbackwardtranslation[F(l,76)=26.79,MSe=33,370.1,P<.001],withadjustedmeansofl,000.9and1,233.3,respectively.Thesecondhypothesiswassupported. Third,ithadbeenhypothesizedthatL2picturenamingwouldbenofasterorslowerthanforwardtranslation.parisoninvolvedbothwordandpicturestimuli,and,therefore,onlythecovariatesofcontextavailabilityandfamiliaritywereused.Thetwocovariatestogetherountedforasignificantportionofthenaming/translationlatencyvariance[F(2,76)=8.0,MSe=83,840.5,p=.001].Nevertheless,onlyfamiliaritywasasignificantcovariate[t(76)=-2.66,p=.01].Aftertheadjustment,L2picturenamingwasnofasterorslowerthanforwardtranslation[F(l,76)=0.28,MSe=83,840.5,p=.60],withadjustedmeansof1,301.8and1,344.3,respectively.Thethirdhypothesiswassupported. Theonlychangeofresultsfromtheprecedingtothepresentexperiment,whichmeasuredcertainitemcharacteristicsandusedthemascovariatesinthereanalyses,wasthatthetwowordtranslationtasksnolongerdisplayeddifferentresponselatencies.ThestraightforwardinterpretationisthatthedifferenceintranslationtimefoundinExperiment1couldpossiblybeexplainedbyitemfamiliarity,independentofanyprocessingasymmetries.Nevertheless,thepresentANCOVAsonlyusethetestitemastheanalysisunit,andcautionmustthereforebetakennottooverinterprettheresults.Whatthepresentfindingssuggestisanalternativewaytointerpretthetranslationtimeasymmetrysofrequentlyobservedinpreviousstudies(whichdidnotcontrolforvariationsintestitemfamiliarity);theactivationpatterns ofthetranslationroutesspecifiedbytheasymmetrymodelmayactuallybeconditionaluponthefamiliarityofthetestitems.Futureresearchintheareashould,therefore,(1)attemptsomeformofcontroloveritemfamiliaritysoastorenderthedatamoreinterpretableand(2)exploreinamoreconcretewayhowitemfamiliarityaffectspatternsmunicationamongthelexiconsandtheconceptualmemory. EXPERIMENT3 TheaimofExperiment3istwofold.(l)ItevaluatesthepropositionthattheL1lexiconhasstrongerconceptuallinksthantheL2lexicon.
(2)UsinganANCOVA,itreteststhehypothesesofthetwoprecedingexperimentsbyintroducingresponseproductionandconceptretrievallatenciesascovariates.DerivedfromthefirstobjectiveisthenewhypothesisthatmatchinganL1itemtoacategorynamewouldbefasterthanmatchinganL2item,assumingthatstrongconceptuallinksfacilitateconceptessand,therefore,category-matchingperformance.Thesecondobjectivewasmotivatedbytheanalysisofpicture-namingandwordtranslationlatenciesintoponents.Someofponents,suchasthetimeneededforproducingovertresponsesandthetimeneededforessingconceptsinresponsetostimuli,areirrelevanttothetranslationpathwaysspecifiedbytheasymmetrymodel.Forinstance,thecontrastbetweenforwardandbackwardtranslationtimesparingalatencycontainingL2responseproductiontime(i.e.,forwardtranslationlatency)withalatencycontainingLlresponseproductiontime(i.e.,backwardtranslationlatency).Sinceresponseproductiontimeisirrelevanttotheissueunderexamination,systematicvariationsinitshouldbecontrolledinsomeform.Asimilarproblemariseswhenpicture-naminglatencyparedagainstwordtranslationtime,fortheformercontainsconceptretrievaltimeinresponsetoapicture,whereasthelattercontainsconceptretrievaltimeinresponsetoaword,anditisconceivablethattherearesystematicbutextraneousvariationsinthetimeneededforconceptessduetodifferentstimulustypes(picturesvs.words).Possiblevariationsinresponseproductionandconceptesstimeshavenotbeencontrolledbypreviousstudies.OnthebasisofthesametestitemsasthoseusedinExperimentIbutadifferentgroupofproficientChineseEnglishbilinguals,thepresentexperimentemployedthetasksofdelayedproductionandcategorymatchingtoestimateeachtestitem'sresponseproductionandconceptretrievaltime,respectively. Therationaleforusingdelayedproductiontoindicatesheervocalresponselatencywasthatprocessingstagesprecedingresponseproduction,suchasstimulusrecognitionandphonologicalcodeess,wouldhavepletedbytheendofthe650-msecdelay(i.e.,theintervalbetweenstimulusonsetandresponsesignal)usedinthepresentexperiment.BalotaandChumbley(1985)usedthesametaskwiththesamedelayandobtainedsta- BILINGUALPROCESSING1009 bleresults.Categorymatchingwasusedinthepresentstudytoindicateconceptretrievaltime.Categoriesofdifferentbroadnesswereused,becausematchingacertaintargettoacategorynamemightnotrequirefullactivationandpreciseknowledgeoftheconceptunderlyingthattarget.Tojudgewhetheranitembelongedtoacertaincategorymightonlyneedinformationaboutsomeparticularaspectsoftheitem,suchasanimacyandmobility,dependingonthebroadnessofthecategoryname.Itwas,therefore,importanttousecategoriesofdifferentbroadness,whichsupposedlydemandeddifferentlevelsofconceptactivation,toexaminetheconceptretrievalprocess.Empirically,JaredandSeidenberg(1991)diddemonstratethatcategoriesofdifferentbroadnessbehaveddifferentlyinasemanticdecisiontask. Inthepresentexperiment,estimatesobtainedfrombothdelayedproductionandcategorymatchingwereusedascovariatesinaseriesofANCOVAsthatreanalyzedthepicture-namingandwordtranslationdataobtainedinExperimentIandretestedthethreehypotheses. Method Subjects.Thirty-sixproficientChinese-Englishbilingualswererecruitedinthepresentexperiment.Noneofthemhadparticipatedinthetwopreviousexperiments.TheirgenerallanguagebackgroundandL2(English)proficiencywereveryparabletothoseofExperiments1and2subjectsbecause(1)theyhadundergonethesamelanguagecurriculuminhighschool;(2)theyhadgenerallythesamelanguageexperienceathomeintermsofthepredominantlymonolingualenvironmentofspokenlanguage(i.e.,Cantonese-Chinese);and(3)theyhadbeenrequiredtodemonstratethroughexaminationacertainlevelofEnglishproficiencybeforetheycouldenrollattheuniversity. MaterialsandApparatus.MaterialsandapparatuswerethesameasthoseinExperiment1.Newmaterialsweretwotypesofcategorynamesforuseinthecategory-matchingtask:Therewere2remotecategories(livingandnonlivingthings)and21immediatecategories(furniture,electricalappliances,meansoftransport,mammal,fruit,weapon,metallicinstrument,stationery,an,clothing,thingstoread,silverware,timer,plant,mechanicaldevice,container,,musicalinstrument,building,signal,andvegetables).Duringcategorymatching,immediatecategorieswereassumedtodemandfullerconceptactivationofthetargetthandidmoreremoteones,duetotheirrelativelyhighsemanticspecificity.Theuseofthesetwotypesofcategorynamessupposedlyhaving differentconceptualdistancesfromanytargetitems,therefore,providedanopportunitytoassesstheeffectsproducedbydifferentdegreesofconceptactivation.The21immediatecategorynamesweregeneratedby10independentsubjects,whowereaskedtoproduceamosttypical,e-to-mindsuperordinatecategorynameforeachofthe40conceptsunderlyingthe120testitems.Interjudgeagreementonproducingthese21categorynamesforthe40conceptswashigherthan84%. Procedure.TheprocedureofdelayedproductionfollowedthatusedbyBalotaandChumbley(1985),anditwasidenticaltothatofpicturenaminginExperiment1,exceptthat(l)wordsinsteadofpictureswerepresentedand(2)thesubjectwasrequirednottonamethestimuluswordasquicklyaspossiblebuttodelaynaminguntilapairofparenthesescameonbracketingthestimulusword.Theparenthesesalwaysappeared650msecaftertheonsetofthetestword(Balota&Chumbley,1985).Responselatencywasdefinedastheintervalbetweentheonsetoftheparenthesesandtheonsetofthevocalresponse.Thisprocedurewasadoptedinordertoensurearelativelypuremeasureofmereresponseproductiontime,forprocessing,suchasessingtheappropriatephonologicalcode,andpreparationformotorresponseshouldhavepletedbytheendofthe650-msecdelay. Forcategorymatching,thesubjectwasrequiredtodecideasquicklyanduratelyaspossiblewhetherthepresentedstimulus(picture,Chineseword,orEnglishword)belongedtoasuperordinatecategory.Thefourfactors(i.e.,stimulustype,positivevs.negativematch,languageofthecategoryname,andimmediatevs.remotecategories)werecounterbalancedacrossconceptsandsubjects.Halfofthestimuliineachblockmatchedthecategoryname.Eachcategory-matchingtrialbeganwithasquareofarraysofsmallcrosses.Itwasfollowedbyasuperordinatecategoryname(EnglishorChinese),whichwasdisplayedfor1sec.Thestimulus(target)thenfollowedimmediately(i.e.,SOA=1,000msec),anditstayedonthemonitoruntilthesubjectrespondedbypressingeitherthedesignatedyesorthedesignatednoresponsekeyonthekeyboard.Nofeedbackwasgiven.Categorymatchinglatencywasdefinedastheintervalbetweentheonsetofthetargetandthekeypress. ResultsandDiscussionPercenterror.Errorratesforbothdelayedproduction andcategorymatchingarepresentedinTable3.Theonlyreliabledifferencewasthatbetweenimmediateandremotecategoriesforthecategory-matchingtask[F(1,117)=13.79,MSe=0.03,p<.001].Matchingitemstoremotecategorieswasingeneralmoreerrorpronethanmatchingitemstoimmediatecategories. Table3MeanDelayedProductionandCategoryMatchingLatencies(inMilliseconds; WithPercentError)forExperiment3:AnalysisbySubject Target Chinese English Pictures Task
M SD
M SD
M Delayedproduction 595 1.9 601 4.9 Immediatecategorymatching 971 7.1 1,093 7.9 956 9.2 (Chinesecategorynames) Immediatecategorymatching 1,068 7.8 1,229 8.1 1,129 8.8 (Englishcategorynames) Remotecategorymatching 1,154 13.8 1,370 21.7 1,099 12.5 (Chinesecategorynames) Remotecategorymatching 1,10612.5 1,207 18.1 L179 8.8 (Englishcategorynames) Note-Onlylatenciesofcorrect(andalsopositiveforcategorymatching)responsesarereported. 1010CHEUNGANDCHEN Latencies:Analysesbysubject.Meandelayedproductionandcategory-matchinglatencieswerecalculated,usingthesubjectastheunitofanalysis.TheyareshowninTable3.Fordelayedproduction,noeffectoftestlanguagewasfound.Forcategorymatching,onlythelatenciesofpositiveandcorrectresponseswereanalyzedandusedtoindicateconceptretrievaltimes.Thejustificationwasthatarelativelypureconceptessprocesscouldbeassumedforpositiveandcorrectresponses,whereasnegativeorincorrectresponsesmighthaveinvolvedother,additionalprocesses.Category-matchinglatenciesfrompositiveandcorrecttrialswereanalyzedbya3(Chinese,English,andpicturestimuli)X2(Chinesevs.Englishcategorynames)X2(immediatevs.remotecategories)mixeddesignANOVA.Categorynamewasaninsignificantfactor,nordiditinteractwithanyothereffects.Boththeimmediateversusremotecategoriescontrast[F(1,34)=8.76,MSe=76,590.8,p=.006]andtheeffectofstimulustype[F(2,68)=5.76,MSe=84,747.5,p=.005]weresignificant,thoughtheirinteractionwasnot.Theresultsindicatedthatthereexistedoveralldifferencesamongcategory-matchingtimesinresponsetopicture,Chinese,andEnglishitems.Morespecifically,forimmediatematching,EnglishitemsweremoreslowlymatchedthanbothChinese[t(35)=-2.34,p=.02]andpictureitems[t(35)=3.61,p=.001].RemotematchingwasslowerforEnglishthanforChineseitems[t(35)=-2.28,p=.03];thecontrastbetweenEnglishandpictureitemswasnotsignificant.TheshortermatchinglatenciesassociatedwiththeChinesethanwiththeEnglishitemsareindicativeofstrongerconceptuallinksfortheformer.ThelackofacategorynamelanguageeffectisconsistentwiththeresultsreportedbyDufourandKroll(1995),whoalsofailedtofindsucheffectinrelativelyfluentEnglish-Frenchbilinguals,asthecurrentsubjectswerealsoconsideredproficientintheirL2.Matchinganitemtoaremotecategorywas,inthepresentexperiment,generallyslower
声明：
该资讯来自于互联网网友发布，如有侵犯您的权益请联系我们。