Upload
others
View
14
Download
0
Embed Size (px)
Citation preview
Title Lexical Performance in L1: What is Native-like LexicalCompetence?
Author(s) Goya, Hideki
Citation 言語文化研究紀要 : Scripsimus(23): 1-26
Issue Date 2014-10-31
URL http://hdl.handle.net/20.500.12000/31134
Rights
言語文化研究紀要SCRIPSノM7SN0.23,2014
LexicalPerfbnnanceinLl:WhatisNative-likeLexicalCompetence?
HidekiGoya
Introduction
Vocabularyknowledgeisindispensibleinanylanguageto
communicatemeaningfullyandappropriately.However,various
difficultiesareinvolvedinachievingfull-developmentoflexical
competenceevenamongnativespeakers(e.g.,Stahl&Nagy,2006).
Previousstudiesinvestigatedwhatfactorsinfluencedprocessingand
learningvocabularyinthemothertongue.Thesestudiesfoundthatone
factoristhesimilarityofmeaningamongwords(Erten&Tekin,2008;
Finkbeiner&Nicol,2003;Nation,2001;Webb,2009).Morespecifically,
somefoundthatdictionarymeaningwasfoundtobeacauseofambiguity
influencingaccuratesynonym-usages(Azuma&VanOrden,1997;Millis
&Butto,1989)whileothersclaimedthatmoreinternalfactorsatamicro
levelsuchassemanticrepresentationmighthaveinfluencednative
speaker'ssuccessfulprocessing,whichaffectedsubsequentlearning
(Erten&Tekin,2008;Finkbeiner&Nicol,2003).Infact,inthe
word-senseresearchframework,themeaningsimilaritydeterminedby
thenumberofword-senses,asegmentofword-meaning,wasfoundtobe
influentialtoprocessingofnear-synonymsaswellassubsequentlearning
eveninthenativelanguage(Finkbeiner,2002;Finkbeiner&Nicol,2003;
Finkbenier,Forster,Nicol,&Nakamura,2004).
Thepresentstudylookedforanygeneralizablepatternsofnative
speakers'lexicalperformanceinrelationtotheword-senseeffectwhich
hasbeenscarcelydiscussedinthepsycho-semanticfield.Ifnative
-1-
speakers'lexicalcompetencecanbemeasuredbyobservingtheirlexical
performance,thiscanbeutilizedasabenchmarkforL2learners'lexical
developmentinthetargetlanguage,whichalsodevelopsmethodological
validityinmeasuringanddrawingadecisiveconclusion.Thiswillinturn
helpL2learnersandinstructorspredictpotentialdifficultiesinacquiring
newL2vocabulary.
TheoreticalBackground
Wordsinmind
Manyresearcheffortsweremadetoillustrateageneralpictureof
howwordsinanativelanguagewouldbesystematicallyorganizedina
speaker'smind.Oneoftheearlymodelsofwordorganizationinthefield
wascalledthehierarchicalnetworkmodel(Collins&Quillan,1969).The
modelpredictedthatwordswouldbeconceptuallyorganizedaccordingto
theirinterrelationshipofidiosyncraticfeatures.Morespecifically,itwas
hvoothesizedthatallwordsweresomehowrankedandinterconnected
basedonconceptualfeaturesinahierarchicalstructure(Figure1).
異|:難!。t*Swims
upstream
tolayeggs
CANARY
dangerous
F道ure1・HierarchicalNetworkModel(Collins&Quillan,1969)
ワム
Furthermore,themodelpostulatedthatthedistanceofnodelevels
(i.e.,CANARY,BIRD,orANIMALinFigure1)betweenwordswould
determinethespeedofresponsestowardsgivenstatementsinthe
experiment・Thatis,itwouldtakelongerforparticipantstoactivatethe
relationshipifcategoriesinthestatementwerefurtherapartthanifthey
wereclosertogetherinthemodel.Despiteplausibleaccountsofthe
latencyeffectamongvariousstatements,however,thismodelfailedto
adequatelyaccountforatypicalityissue(Anderson,2005).
Fromacognitivelinguisticviewpoint,linguisticknowledgecanbe
formedbycognitiveinteractionandexperiences(Bechtel&Graham,
1998).InBechtelandGraham'sview,thebasicstructureofword
knowledgeorganizationappeareddifferentlythanitdidintheprevious
model.Specifically,theprincipleunderlyingtheirnewperspectivewas
thatwordswereorganizedinaradicalorderbydegreeofmutual
closeness(Labov,1973).Theradicalmannerwellexpressesthatthe
closertothecenterawordislocated,themorelikelyitistorepresentthe
category(Bowerman,1978)orprototypicality(Aitchison,2003).
Aitchisonarguedfortheimportanceoftheprototypicalityofthe
categories,whichplausiblyaccountedforthedistanceofeachwordso
thatsomedifferencesinthecategorywouldneedmorelatentresponse
thanothers(Figure2).
- 3 -
Figi"・e2.PrototypicalityModel(adaptedfromAitchison,2003)
Inamorerecentapproach,furtherdevelopmentsinunderstanding
wordorganizationhasprovidedanovelview.CollinsandLoftus(1975)
attemptedtobringtogetherthesemanticinterrelationshipdescribed
aboveinamoreintegralmodel:thespreadingactivationmodel.This
modelincludednodeswhichwereinterconnectedwithwordsasseenin
thehierarchicalnetworkmodel・Whatwasdistinctivelydifferentfromthe
earliermodelswerethatthewordswerenolongerrankedina
hierarchicalorder.Rather,thewordswereinterconnectedbyaweb-like
networkwithlineswhoselengthhypotheticallyrepresentsthefamiliarity
andsimilarityofthewords.
Theirmodelwelldepictedrichconnectionsamongwordsin
speakers'minds・Figure3illustratestheword"fireengine"toshowhowit
wouldco-activatesurroundingwords.Forinstance,basedonthe"fire
-4-
engine"diagram,whentheword"fireengine"isprocessed,it
simultaneouslyactivates"house,""red,"and"truck"becausetheyare
directlyandcloselyconnected.Inotherwords,languageuserswill
immediatelythinkof"house,""red,"and"truck"whentheyseeorhear
theword"fireengine."Subsequently,theinitialactivationleadsto
spreadingactivationsofotherwordssuchas"street,""yellow,"or“
rosescontinuously.rosescontlnuously.
F漕L"で3.SpreadActivationModel(Collins&Loftus,1975)
Asseeninthediagram,thelexicalcategoriesarelinkednotbysemantic
relationshipsbutbythestrengthsanddistancesofthenodes.
Furthermore,sincethestrengthisrepresentedbythelengthofthe
connectionsbetweenwords,categoricalandprototypicalrelationships
arelogicallywellaccounted.
Psycholinguisticstudieshavewellsupportedthemodel'svalidityto
accountforthelearnersorganizationofwords.MeyerandSchvanveldt
(1971)askedparticipantsinastudytodecidewhethertargetwordswere
realwords(i.e.,lexicaljudgmenttest).Inthetask,thetargetword
-5-
appearedimmediatelyaftertheprimeworddisappeared.Iftherewere
contrastingdifferencesintheprocessingspeedbetweenthesemantically
relatedandnon-relatedpairs,itcouldbeconcludedthatthewordsmight
belinkedclosertogether.Indeed,theobtainedresultsindicatedthat
semanticallyrelatedpairssuchas"br・ead-butter"wereprocessed
quickerthannon-relatedpairssuchas"doctor-butter"(Meyer&
Schvanveldt,1971).MeyerandSchvanveldtattributedthelatencyeffect
tothefactthatsemantically-relatedwordswereorganizedcloselyand
strongly.
Morerecentmodel
Foramorespecificviewonthesegmentofword-meaning,
especiallyintermsofsimultaneousandspontaneousco-activation,
Finkbeiner(2002)proposedthesensemodel.Hehypothesizedthateach
wordincludesacertainnumberoflanguagespecificsegmentsofmeaning
or"word-sense"(shownasgrayishandwhitedotsinFigure4)and
sharedsenses(representedbyblackdotsinFigure4).Specifically,the
modelpredictedthesamepatternofprimingeffects,supportingtheidea
ofCO-activation(i、e.,whenweseeandhearaword,itwouldactivate
semanticallyrelatedwords).Additionally,thismodelpostulatedthata
translationasymmetrywasattributednottoaweakconnection(asseenin
therevisedhierarchicalmodel;Kroll&Steward,1994),butto
representationalasymmetrydeterminedbythenumberofword-sense.
Forinstance,theequivalentwordsblack(English)andkuroi(Japanese)
differinthenumberoflanguage-specificsenses.Theweakerlanguages
havefewersenses(i.e.,EnglishinthecaseofESLlearners).Theless
balancedtwolanguagesare,thelargerthedifferencebecomesbetween
thetwolanguagesintermsofnumberofsenses.Figure4illustratesthe
model.
戸○
LexicalFo面n
Representation
Lexicalscmannc
Representation
F遁["で聖.Finkbeiner'sSenseModel(AdaptedfromFinkbeiner,etal..
2004)
Inthesamevein,Rodd,Gaskell,andMarslen-Wilson(2002)
investigatedthesemanticpropertywithacontroversialclaimoflexical
ambiguityadvantage.Theyclaimedthatifwordsinapairwere
semanticallyambiguous,theywouldprocessthewordslower.Roddetal.
controlleddegreesofsemanticambiguityonabasisofthenumberof
word-senses"(i.e.,smallersegmentsofwordmeaning).Inaddition,they64
hypothesizedthatthewordmeaningadvantagewasactuallyasense
advantage(Roddetal.,2002).
Intheirseriesoflexicaljudgmenttestswithnativespeakersof
English,theparticipantswereaskedtojudgewhetherpresented
ambiguousandunambiguouswordswerealexicalitemornot(Roddetal.,
2002).Thelexicalvariables,includingconcreteness,familiarity,and
relatednessratings,werecontrolled.Theresultsshowedthatmultiple
word-sensesproducedaprocessingbenefit,whilemultiple
word-meaningsdelayedrecognition,andwordswithmanyword-senses
wererespondedtofasterwithfewererrors.Theyalsofoundthatlexical
decisiontimes(reactiontimeorRT)wereactuallyslowerforambiguous
wordssuchasbankthanforunambiguouswords,whereasdecisiontimes
forwordswithmanywordsensesweresignificantlyfasterthanforwords
withfewwordsenses(Roddetal.,2002).Basedonthefindings,they
ワー
suggestedthatitwasnotword-meaningbutword-sensethatwould
facilitatewordrecognition,andword-sensewasasemanticproperty
whichcouldbeascribedtosemanticambiguityinterferingwithlearning
semanticallyrelatedwords(Roddetal.,2002).
Byexploringthesemanticpropertywithadifferentresearch
paradigm,Finkbeiner,Forster,Nicol,andNakamura(2004)investigated
word-senseinvolvementassemanticrepresentations.Theytestedboth
bilingualspeakersandnativespeakers(i.e.,Japanese-Englishbilinguals
andnativespeakersofEnglish).Theirmainfocuswastoshedlighton
primingasymmetrybetweenLIandL2・Theyespeciallyexaminedvia
taskdifferencesofhowtheycouldaccountforsuchasymmetricrelations
betweenLIandL2primingbyemployingword-sensesthroughdifferent
measurements.Asaresult,theyfoundtherobustmaskedL2-LI
translationpriminginsemanticcategorizationbutnotinlexicaldecisions,
TheyassumedthattherelationshipbetweenLIandL2lexicalsemantic
representationswasasymmetrical;LIwasrichbutL2wasnot.Theyalso
foundthatinlexicaldecisions,therobustprimingeffectwasobtainedin
themany-fewdirection,butnoprimingwasobtainedinthefew-many
directionusingthesamepairs.Theeffectinsemanticcategorization,on
theotherhand,wasobtainedinbothdirections.Theyconcludedthat
wordswithmanyword-sensesprimedotherswithfewword-sensesin
lexicaldecisionsandsemanticcategorization,butthesameword-pairsin
thefew-manydirectionproducedtheprimingeffectonlyinsemantic
categorization,notlexicaldecisions.
Whatisimportantisthatinaseriesofpsycholinguisticexperiments.
Finkbeineretal.(2004)verifiedtheasymmetricaleffectintranslation
tasks.Inshort,theresultconfirmedtheeffectobservedacrosslanguages
ifthestimulisharedword-senses.Moreimportantly,theyfoundtheeffect
withinalanguageamongnativespeakers・Thatis,inapairofsemantically
-8-
similarwordssharingsomesenses,thewordwithmoresenseswould
primetheotherwithfewersenses.
Tosum,thefindingsfromRoddetal.(2002)andFinkbeineretal.
(2004)collectivelysuggestedthatword-sensewasthesemanticproperty
thatwouldaffectsemanticprocessing.Specifically,wordswithmany
word-senseswouldfacilitaterecognitionofwordswithfewword-senses
withinalanguageaswellasacrosslanguages(Finkbeineretal,2004).
Amythinnative-speaker'slexicalcompetence
Asseenabove,manystudiesexaminedinwhatwaywordknowledge,
especiallywordmeaning,isorganizedinlearners'mindstobuildabasic
understandingofwhatconsistsofaccurate,appropriate,andnative-like
useofthetargetwords.Someresearchers,however,weredoubtfultothe
word-senseeffect・Forinstance,Gernsbacher(1984)arguedthatwords
withmorethanonemeaningweretypicallymorefamiliar;therefore,it
wasnaturalthatarobusteffectonlexicaljudgmentshouldbefound.
Althoughshefoundsupportingevidencetothiseffect,suchrobustness
wasfoundonlyinlexicaljudgmenttests(Borowsky&Masson,1996),
whichdroveotherresearcherstoinvestigatesuchtaskdependencyin
termsoftheeffect'srobustnessonword-sense(Finkbeineretal.,2004;
Roddetal.,2002).Inotherwords,theword-senseeffectfoundinthe
lexicaljudgmenttesthasbeenscarcelydiscussedinotherresearch
methodologies,whichwouldfailtogeneralizetheeffectinthe
psycholinguisticresearchparadigm.Suchtaskdependencyneedsmore
empiricalattentionforathroughdescriptionoftheword-senseeffect.
Furthermore,manystudiesinsecondlanguage(L2)vocabularywere
carriedoutbasedonapromisethatnativespeakersmusthave
full-fledgedlexicalcompetence.Thus,anyresultsgainedfromL2
participantsinexperimentsweredirectlycomparedtothatfromnative
-9-
speakersofthetargetlanguagebasedonthepremise.Whatisinteresting
isthatsuchcompetencehasbeentakenforgrantedasabenchmarkto
decidehowfarL2learnersofthetargetlanguagewerecomparedto
nativespeakers.Asaresult,someL2learnersatanadvancedlevelmight
beviewedasnotnative-likeduetoincongruentresultscomparedwith
misleadingresultsofnative-speakers'lexicalperformanceaswellaslack
ofmeasuringtheword-senseeffectintheexperiments.Inthissense,
whatitmeanstobenative-likeinlexicalcompetenceremainsunclearin
theframeworkofthesemanticjudgmenttest,whichdeservesmore
theoreticalattentionsinthefieldofpsycholinguistics.
Method
Researchquestions
Themajorgoalofthepresentstudywastodeterminelevelsof
native-likelexicalcompetence.Specifically,thestudyhypothesizedthat,
giventhattheexperimentthatinvolvedword-senseshowedthepriming
effectamongnativespeakers,theword-senseeffectcanbeobservablein
theframeworkofsemanticjudgmenttestaswell.Thisinturnhelpsus
determinewhatisnative-likeintermsofword-senseacquisition,which
willprovidemorevalidmethodologicalcluesforclarifyingtherelationship
betweenLItranslationandL2word-senseassemanticrepresentationin
L2lexicalcompetence・Inordertoexploretheaboveaspect,thefollowing
researchquestionswereaddressedintheframeworkofthesemantic
judgmenttest:
RQ1:Isnativespeakers'responsetowardssemanticallyrelated
wordsmoreaccurateandfasterthanunrelatedwords?
RQ2:Isnativespeakers'responsetowardssemanticallymore
similarwordslessaccuratethanlesssimilarwords?
- 1 0 -
RQ3:Isnativespeakers'responsetowardssemanticallymore
similarwordsslowerthanlesssimilarwords?
Researchdesign
Thestudyhadtwoindependentvariables:word-senseand
relatedness.Word-senseisusedtodeterminethedegreeofmeaning
similarity.Therelatednessdividedstimuliineithersemanticallyrelated
word-pairsorunrelatedword-pairs.Thedependentvariablesweremean
accuracyandreactiontime(RT)ofthesemanticjudgmenttest.
Participants
AllparticipantswerenativespeakersofEnglish:fivemalesand15
females.Theiragerangedfrom18to21yearsold.Theyallvoluntarily
participatedintheexperimentin=20).Allofthemwerecollegestudents
whohadnotreceivedanyformalJapaneselanguagelessons.Anybilingual
speakerswereexcludedfromthegroupbythebackgroundquestionnaire.
Theircollegemajorswidelyvaried,butnoneofthemwasmajoring
English,Japanese,orlinguisticsexceptfortwoparticipants(secondary
EnglisheducationorEnglishstudies).
Taskandmaterials
Sema""で"吃'meattest.Thecomputer-basedtest,adaptedfrom
Jiang(2002,2004a),whereparticipantswereaskedtojudgewhetherthe
presentedEnglishwordpairswererelatedinmeaning.Thetestfeatured
79pairedEnglishwordsconsistingofsynonymouspairs(〃=39)and
non-semanticallyrelatedpairsin=40).Theorderoftheitem
presentationwasrandomizedforeachparticipant・Participantshadto
press"Y"onthekeyboardifapairseemedtoberelatedinmeaning;
otherwise,theyhadtopress"N"asunrelated.
-11-
TheListofS"功uli.Thestimuliusedinthisonlineexperimentwere
adaptedfromGoya(2014).Amongthesemanticallyrelatedpairs,two
sub-typeswerecreatedbycontrollingdegreesofmeaningsimilarity
betweenpairedstimuli:MoreSimilar(synonymssharingmultiple
word-senses)andLessSimilar(synonymssharingasingleword-sense).
AccordingtotheOne-wayANOVA,resultsshowedthatthemultiple
word-sensessharedpairshadsignificantlymoresensesthananyofthe
singleword-sensesharedpairs.Otherlexicalvariables(word-length.
semanticrelatedness,andwordfrequency)werealsocontrolled.
Accordingtoseveral̂ -tests,noneofthefactorsweresignificant.Inthis
sense,"[c]onsideringtheminimizedeffectsofothervariables(i.e.,
familiarity,relatedness,word-length,andword-frequency),anyresults
canbeattributedtothetestedvariable"(Goya,2014).Adescriptionof
stimuliispresentedinTable1.
Table1.
Desaぐわ"b〃ofstimuli
Examplesof"Numberof''Semantic
relatedpairssensesrelatedness
Moresimilarpairs(allow-permit)2.63
Lesssimilarpairs(admit-accept)1
5.14
5.21
F̂requency
ofpairs
''Lengthofwords
86.795.39
84.055.61
ハノote."Thenumberofword-sensesofeachwordwasdrawnfromthe
corpusWordNet(Fellbaum,1998).̂Semanticrelatednesswasmeasured
bytheadegreeofclosenessofthemeaningbetweenpairedwordsrated
byfivemonolingualspeakersofEnglish・"Wordfrequencywasmeasured
byaverageofpairedwordsdrawnfromBrowncorpus(Kucera&Fransis,
1967).''Wordlengthwasmeasuredbythemeannumberoflettersofthe
pairedwords.
-1 2 -
ApparatusandProcedures
Twotaskswereadministeredinthefollowingsequence.First,after
participantsfilledoutthebackgroundinformationquestionnaire,the
semanticjudgmenttestwasadministered・Theresearchermetallofthe
qualifiedparticipantsin=20)inthelaboratoryindividually.The
experimentwascarriedoutonaMaccomputer.Theparticipantswere
directedtorespondasquicklyaspossiblebypressinganykey.Aftera
cue,thestimulusword-pairremainedonthescreenfor500msoruntilany
buttonwaspressed.TheRTswererecordedfromtheonsetofthe
presentation.Thepresentationorderofthestimuliwasrandomizedfor
eachparticipantbySuperLab,apsycholinguisticexperimentation
software.
Result
Asmentionedearlier,thetotalnumberoftheparticipantswas20;
however,onewasexcludedfromfurtheranalysisduetosignificantlyslow
responses.Figure5and6presentthemeansandstandarddeviations
(SDs)ofaccuracyaswellasreactiontimes(RTs)accordingtothe
stimulustypes,suchasoverallitemsin=79),relatedpairsin=39),and
unrelatedpairsin=40).Accuraciesarepresentedinpercentage(%)
whileRTsarepresentedinmilliseconds(ms).
AccordingtoFigure5,themeanaccuracieswere89.67%iSD=7.79),
92.17%iSD=5.37),and85.79%iSD=15.70)fortheoverallitems,the
relateditems,andtheunrelateditems,specifically・Theratiobetweenthe
numberofthesemanticallyrelatedpairsandthatoftheunrelatedpairs
was39:40;therefore,thefinalizedchancelevelofbothgroupsaccuracy
was49.37%.Alloftheparticipants'judgmentwasabovethechancelevel,
whichindicatesthatthematerialsuccessfullyeliminatedanyguesswork
-1 3 -
onthesemanticjudgment.
94
92.17%
92旨③員団つ具昌旨関輯場単◎計。国。。。く
い密鎮駕《ふ.…。.1..:
霜ー90
88
85.79%86
84
82
Related Unrelated
Word-PairTypes
局ど["e5.Accuracyofjudgmentonrelatedandunrelatedword-pairs
AsforRTs,accordingtoFigure6,theirmeanswere1582.08msiSD
=329.39),1398.13msiSD=286.57),and1616.67msiSD=377.18)forthe
overallitems,therelateditems,andtheunrelateditems,specifically.
000000
m釦鋤和知加
11111l
]員のロ胃凶ロ。『○審目園【画の吻雫○門“
1616.67ms
.合一,堅臨鵠薄…VL_‐‐__‐‐_‐______、
1398.13,s瀞蕊織議蕊蕊蕊#
再禽!
8-…錐
UnrelatedRelated
Word-PairTypes
FIgui汐6.Reactiontime(RT)ofjudgmentonrelatedandunrelated
word-pairs
-14-
Toexaminewhetherthereisanysignificantdifferenceinthese
means,thedifferencesbetweenthemeansoftherelatedpairsandthatof
theunrelatedpairsweretestedusingtwo-tailedpairedt-tests.Asfor
accuracy,thedifferencewasnotsignificant(p=、12).Theresultsindicate
thatthegroupssemanticjudgmentwereequallyaccurateonbothrelated
andunrelatedword-pairs.AsforRTs,however,thedifferencewas
significant:̂ (18)=-4.59,p<.01,andd=.65.Theresultsindicatethat
thegroupssemanticjudgmentwassignificantlyfasterontherelated
word-pairsthantheunrelatedcounterparts.Table2indicatestheresult.
Table2.
ノセsuitsofthet-1~どSt0〃ノl絶ansofReactio〃刀切esbetweenRelateda"ぴ
Unrelateぴ姥"ぢ
A化α〃 SD r p
Related1398.13ms286.57-4.59.000
Unrelated1616.67ins377.18
‘ノ
、65
Acomparisonbetweensemanticallymoreandlesssimilar
word-pairswhosesimilaritywasdeterminedbythenumberof
word-sensesharedwithinapairwasalsoconducted.Themeanaccuracy
ofmoresimilarword-pairs(More)was92.37%iSD=6.74)whilethatof
lesssimilarword-pairs(Less)was91.97%iSD=7.31).AsforRTs,the
meanspeedofjudgmentonthemoresimilarword-pairswas1377.84ms
iSD=284.12)whilethatonthelesssimilarcounterpartswas1419.49ms
(、SD=314.98).Figure7and8indicatethosenumbers.Accuraciesare
presentedinpercentage(%)whileRTsarepresentedinmilliseconds
(ms).
- 1 5 -
94‐’。‐●P●1
92.37%
騨鍵蕊91.97%
208642
998888
]巨の鼻唇叩つ。一
畳目Eのど。否曽雪8く
鍔:.#需.…・‐
鱗,$騨・,輔
蕊
LessMore
Word-PairTypes
Figure7.Accuracyofjudgmentonmoreandlesssimilarword-pairs
1700
0000
0000
6543
1111
料宮⑫戸動己三『昌一唇貿唇の画雫○烏函
1419.49ms
11377雲34,s ‘j瞬・k、識冒・識懲識鼠__…-.‐
1200
LessMore
Word-PairTypes
Figure8.Reactiontimes(RTs)ofjudgmentonmoreandlesssimilar
word-pairs
Toexaminewhetherthereisanysignificantdifferenceinthese
means,thedifferencesofmeansbetweentherelatedpairsandthe
unrelatedpairswastestedusingtwo-tailedpairedt-tests.Theresults
indicatedthatbothdifferenceswerefoundtobenon-significant;t(18)
=、19andp=.85formeanaccuracyand(18)=一・96andp=.35formean
- 1 6 -
RTs・Inshort,nativespeakers'judgmentwasasaccurateandfastonboth
typesofstimuli,semanticallymoresimilaraswellaslesssimilar
word-pairsintheexperiment.
Discussion
Alloftheresearchquestionsaddressedinthepresentstudywereto
investigatenativespeakers'lexicalperformanceintermsofsubtle
meaningdifferences.Thefirstresearchquestionaskedwhethernative
speakers'responsestowardssemanticallyrelatedword-pairscouldbe
moreaccurateandfasterthantheirresponsestounrelatedcounterparts
inthesemanticjudgmenttest.Inthetest,theparticipantshadtodecide
whetherpresentedword-pairswererelatedinmeaning.Interestingly,the
studyfoundamixedresult.Specifically,thedifferenceinaccuraciesof
theparticipants"performancebetweenthesemanticallyrelatedand
unrelatedword-pairswasstatisticallynon-significant;however,howfast
theparticipantsrespondedtothestimulihadasignificantdifferenceip
=.00andd=.65).Thatis,theparticipantsrespondedmorequicklytothe
semanticallyrelatedword-pairsthantounrelatedword-pairsinthe
experiment.
Intuitivelyandlogically,itseemsnotsurprisingtoobservesuch
non-significantdifferenceinaccuracies.Theparticipantsinvolvedinthe
currentexperimentwereallcollegestudentsintheU.S.whomusthavea
widerangeofvocabularyknowledgeinordertoenrollinthefour-year
program.Inthissense,theirfullydevelopedlexicalknowledgeallowed
veryaccuratelexicalperformanceregardlessofthesemanticrelatedness
amongthestimulusword-pairs.Then,whydidittakerelativelylongerto
executetheirreactiontothesemanticallyunrelatedword-pairs?The
stimuliusedintheexperimentwereeithersemanticallyrelatedor
- 1 7 -
unrelated.Inorderfortheparticipantstodecidewhetherthepresented
word-pairswererelatedinmeaning,theparticipantshadtosearchforthe
sourceoftheirdecisionwhileprocessingthesetwowordsinmind.With
thisinmind,itcanbehypothesizedthattheparticipantswerequickeron
thesemanticallyrelatedword-pairsbecausethepairactuallyshared
commonsemanticinformationwhiletheyweresloweronthesemantically
unrelatedword-pairsbecausethepairdidnothaveanycommonality;the
participantshadtokeeponsearchingforacommonfeatureofmeaning
untiltheyterminatedtheirsearch・Therefore,itseemsnaturaltoobserve
suchmixedresults.
Theabovecontentioncancomplementarilyputforwardtheprevious
theoreticalmodelsofwordorganization.Giventhatthecontrolled
variablewasmeaningrelatednessandtheeffectwasfoundintermsof
reactiontime,thiscansuggestthatwordsinourmindmaybelinkedona
basisofthesemanticrepresentationassuggestedinthefield(Aitchison,
2003;Collins&Loftus,1975;Collins&Quillan,1969;Finkbeiner,2002;
Labov,1971;Meyer&Schvanveldt,1971).Inparticular,thePrototype
theory(Aitchison,2003;Labov,1971)postulatesthatwordsinthesame
semanticcategoryhypotheticallyareorganizedinaradicalmannerin
whichmorecentralwordscanbeplacedinthemiddleofwordswhileless
centralwordsmaybeplacedmoreperipherallydistance.Inthissense,the
semanticallyunrelatedwordsarehardlyco-activatedintheparticipants'
mindswhilesemanticallyrelatedwordsbecomehighlyco-activateddue
tocloserdistancesbetweenwords.Similarly,theSpreadActivationModel
(Collins&Loftus,1975;Meyer&Schvanveldt,1971)postulatesthat
wordsinourmindmayco-existinaweb-likenetwork.Furthermore,
distancesbetweenwordsrepresenthowfartheyareapartfromeach
otherinthenetwork.
Thesemodelscollectivelyandadequatelyaccountfortheslower
-1 8 -
reactiontimetothesemanticallyunrelatedwordsfoundinthepresent
investigation.Thatis,theparticipantsrespondedmorequicklytothe
semanticallyrelatedword-pairsbecausethewordswereco-activated
duetolessdistancebetweenthestimuliinourmindwhiletheyresponded
moreslowlytotheunrelatedword-pairsduetogreaterdistancebetween
thestimuliinthenetworkornolinkinthenetworkatall.Thus,the
participantsneededmoretimetosearchforthelink.Yet,onemayraise
thesequestions:whatconnectsthesewordsmthesamecategoryand
whatdeterminesthesemanticclosenessofwordsinthenetwork.
Thesecondandthirdresearchquestionsaddressedtheseissuesby
lookingintowhethernativespeakers'responsestowardsthesemantically
moresimilarword-pairscouldbelessaccurateandslowerthantheir
responsestowardsthesemanticallylesssimilarword-pairsduetomore
word-sensesinvolvedinlexicalprocessing.Lookingcloserathownative
speakersperformintheexperimentwouldpromoterelevantevidenceof
whatconnectswordsinourmindandwhatdeterminesthesemantic
closenessofwordsinthenetwork.Word-senseisatheoreticalconceptof
semanticrepresentationthatcanbecountedbyreferringtoanonline
corpus,WordNet(Fellbaum,1998).Giventhatword-senseispartofthe
semanticpropertyusedforlexicalprocessingamongnativespeakers
(Finkbeiner,2002;Roddetal.,2002),suchpolysemousaspectof
vocabularymightinterferewiththeparticipants'semanticjudgment
becausetheparticipantshavetoretrieveandprocessmanysemantic
representationsforthelexicalcomparisonofthepresentedword-pairs.
Surprisingly,however,thestudyfoundthattheparticipants
respondedtothesemanticallymoresimilarword-pairsasaccuratelyas
tothelesssimilarword-pairsaccordingtonon-significantdifferenceof
accuracyip=.85).AsforRT,althoughtheparticipantsresponded
relativelymorequicklytothesemanticallymoresimilarword-pairs(Af=
- 1 9 -
1377.84ms,SD=284.12)thantothesemanticallylesssimilarword-pairs
(M=1419.49ms,SD=314.98)accordingtothedescriptivestatistics,the
differencewasnotsignificantip=、35).Infact,thisisinlinewithGoya
(2014)indicatingthateventhoughtheEnglishnear-synonymswere
semanticallyverysimilaronthebasisofnativespeakers'ratingsthrough
apilottest,theparticipantswerecapableofdistinguishingthesubtle
differencesofmeaningdeterminedbythenumberofword-sense.
Likewise,otherstudiesinvestigatingL2vocabularyknowledge
amongESLlearners(Jiang,2002,2004a)discussednative-speakers'
full一fledgedlexicalknowledge.InJiang'sseriesofthesemanticjudgment
tests,thedatagainedfromthenativespeakersofEnglishwasusedto
determinehowwelltheESLlearnershadachievednative-likelexical
competence.Itfoundthattherewasnosignificantdifferenceinthelexical
performanceintermsofaccuracyandRT,whichindicatedthatnoneof
thelearnershadachievednative-likestatusduetoincongruentlexical
performancesbetweenL2learnersandnativespeakers(e.g.,Jiang,2002;
2004a).However,tothisdate,noneofthepreviousresearchhadfocused
oninwhatmannernativespeakerswouldspontaneouslyrespondtothe
semanticallyrelatedword-pairsdividedonabasisofthenumberof
word-sensesinanonlineexperiment.
Consideringnoeffectfoundinnativespeakers'accuracyandRTin
thesemanticjudgmenttest,thepresentfindingscontributed
methodologicaladvancementtothefieldofL2vocabularyacquisition.
Specifically,thepresentinvestigationprovidedabenchmarkof
native-likelexicalcompetencewhichwasdeterminedbynativespeakers'
lexicalperformanceinthesemanticjudgmenttest.Suchnovelviewof
native-likecompetencecansuggestwhatitmeanstohavefully
developedlexicalcompetenceinthenativelanguage.Thatis,thenative
speakersareconsiderablyaccurateandspontaneousindistinguishing
-2 0 -
near-synonymsregardlessofmeaningsimilarity.
Combinedwiththefindingsfromthefirstresearchquestion,the
descriptivestatisticsofaccuracyandspeedoftheirperformanceinthe
semanticjudgmenttestmayalsosuggestthatthesemanticallyrelated
wordscanbelinkedtogetherbyword-sensesiftheyshareany;therefore.
theparticipantscanactivatethestimuliquicker,whichcansupportthe
sensemodelproposedbyFinkbeinerandhiscolleague(2004).Thiscanin
turnsuggestthatword-senseisasemanticpropertynativespeakersuse
forlexicalprocessing.Despitethestatisticallynon-significanceofthe
word-senseeffectfoundinthepresentinvestigation,theparticipants
respondedrelativelymorequicklytothesemanticallymoresimilar
word-pairsthantolesssimilarwords.Becausethenumberword-sense
wascontrolledamongthestimuli,thesemanticallyrelatedword-pairs
shareatleastasingleword-sense,andsuchcommonalitymighthelpthe
participantsdrawquickerconclusions.Inotherwords,thiscommon
word-sensecanbeasourceforparticipants'semanticjudgment,whichin
turnsuggestsword-sensecanbeapartoflexicalrepresentationthat
consistsofword-organization.
However,wecannotdenythatmeaningsimilarityonthebasisof
someotheroverlookedlexicalfactorsmighthaveinfluencedthe
participants'lexicalprocessing,whichmightconnectthesemantically
relatedwordsinanetwork.Therefore,thepresentfindingpressesaneed
formoretheoreticalclarificationonwhetherL2speakerscanrespondto
words-pairsthatshareword-senseasaccuratelyandquicklyastothe
word-pairsthatnativespeakersconceiveassemanticallysimilarbutdo
notactuallyshareanyword-sensesuchassil"-s“pid,example-sample.
ordoubt-suspectii.e.,noneofthemshareanyword-sensewithinapair).
Nonetheless,thepresentfindingscanshedmorelightonthe
methodologicaladvancementofword-senseintwoways:(1)theyare
- 2 1 -
methodologicallyusefulininvestigatingtheword-organizationmodeland
(2)theycanbeusedtoassesslexicalcompetenceintermsofsemantic
judgmenttestinfuturestudiesinthenativeandthesecondlanguage.
Conclusion
Vocabularyknowledgeisessentialinanyaspectoflanguageuse.
Themyththatthelexicalcompetenceofnativespeakersisfully
developedhasbeentakenforgrantedinL2vocabularyresearch,
especiallyinrelationtothesemanticjudgmenttests・Inthisstudy,such
competencewasdescribedintermsofnativespeakers'lexical
performanceinrelationtotheword-senseeffect.Whatwasfoundinthe
presentinvestigationwasthatnativespeakersrespondedsignificantly
morequicklytothesemanticallyrelatedword-pairsthantounrelated
word-pairs.Thisempiricalfindingservesasevidenceofthe
psycholinguisticmodelsofword-organization.Moreover,thepresent
findingfurtherstheoreticaldevelopmentintheareaofwhatwaywordsin
themindbecomeco-activatedinrelationtootherlexicalitems.
Inadditiontotheabovecontribution,thisnovelfindingmayalsoadd
amethodologicaladvancementtothefieldofL2vocabularyacquisition.
Thatis,thepresentempiricalfindingcanserveasabenchmarkof
native-likelexicalcompetenceinordertomakeadirectcomparisonwith
L2learners.Specifically,thiscanbereferredtowheninvestigating
whetherL2learnershavealsoacquiredtheword-senseforlexical
retrievalandsubsequentprocessingintheirlanguageuse.
Asforalimitationofthepresentstudy,theresearchdesignonly
allowedthestudytolookatdecontextualizedempiricalevidence,which
restrictsgeneralizabilityofthefindings.Thus,providingthecontextsfor
particularnear-synonymsintheexperimentwilladdthesupplemental
-2 2 -
informationofhownativespeakerswillprocessthetargetwordsin
relationtotheembeddedcontext.Thepresentstudylookedathownative
speakersrespondedtothesemanticallyrelatedwordswhichshared
commonword-sensesinthepair.Inordertoverifythatwordsare
interconnectedwiththeword-senseintheword-organizationinourmind.
theresearchneedstofocusonword-pairsthatdonotactuallyshareany
commonword-senseitemsdespitetheirmeaningclosenessratedby
nativespeakers.Investigatingthisbyhowquicklynativespeakers
respondtothemwilldeterminewhetherwordsareconnectedintermsof
word-sense.
Reference
Aitchson,J.(2003).WordsintheMind:Anintroductiontothemental
lexicon・London:Blackwell.
Anderson,J.R.(2005).Cog刀itivePs}'“oノogya"ぴだs〃刀p"でa"0775.New
York;Worth.
Azuma,T.,&VanOrden,G.C.(1997).Whysafeisbetterthanfast:The
relatednessofaword'smeaningsaffectslexicaldecisiontimes.
JournalorMemorya"〃Langiノage,36,484-504.
Bechtel,W.,&Graham,G.(1998).ACompa"わ〃toCog刀itiveSoだ"Ce・
Maiden,MA:BlackwellPublishersLtd.
Borowsky,R.,&Masson,M.E.J.(1996).Semanticambiguityeffectsin
wordidentification..ノひurnalofExperime"麹IPs}ノ℃方ology:Learning,
A化"7o〃ノト&Cog刀ノ."",22,63-85,
Bowerman,M.(1978).Theacquisitionofwordmeanings:Aninvestigation
intosomecurrentcongluicts.InN.Waterson&C.E.Snow(Eds.),
刀'ledevelopmentofcommunノ℃a"'on(pp.263-287).Chichester,UK:
Wilery.
- 2 3 -
Collins,A.M.,&Loftus,E.F.(1975).Aspreadingactivationofsemantic
processing.PsychologiでaIReVI・ew.82407-428.
Collins,A.M、,&Quillan,M.R.(1969).Retrievaltimefromsemantic
memory.Journalofためa/Leamj"ga"dI/e功a/BehawDr卜8,
240-247.
Erten,I.H.,&Tekin,M.(2008).Effectsonvocabularyacquisitionof
presentingnewwordsinsemanticsetsversussemantically
unrelatedsets.System,36,407-422.
Fellbaum,C.(1998).WordNet.'anelec"℃"/でle幻でaldatabase.Cambridge,
MA:MITPress.
Finkbeiner,M.(2002).TowardsapsycholinguisticmodelofadultL2
lexicalacquisition,representation,andprocessing.
Finkbeiner,M.,&Nicol,J.(2003).Semanticcategoryeffectsinsecond
languagewordlearning・AppliedPs}/℃加"刀g【"苫"℃s,24.369-383.
Finkbeiner,M.,Forster,K.,Nicol,J.、&Nakamura,K.(2004).Theroleof
polysemyinmaskedsemanticandtranslationpriming.Jou"7alof
〃e/刀oノー ya〃ぴZ,anguage.51,1-22.
Gernsbacher,M.A.(1984).Resolving20yearsofinconsistentinteractions
betweenlexicalfamiliarityandorthography,concreteness,and
polysemy.""rnalofExpe""]entalPsychology.Genem4皿a
256-281.
Goya(forthcoming).Relationshipbetweenword-senseinL2andLI
translationinword-knowledgedevelopment.1心ASELE.ノα"刀al.42
Jiang,N.(2002).Form-meaningmappinginvocabularyacquisitionina
secondlanguage.StuめどsinSecondLanguageAcquisノ"'on,18.
148-169.
Jiang,N.(2004a).Semantictransferanditsimplicationsforvocabulary
teachinginasecondlanguage.7乃e〃b咋rnLanguage.ノburnal,88,
416-432.
- 2 4 -
Kucera,H.,&Francis,W.N.(1967).CompU垣"vnalAnalysisof
Present一吃.yAmericanEンフglish・Providence:BrownUniversity
press・
Labov,W・(1973).TheBoundariesofWordsandtheirMeaning.InBailey,
C一J.N.andR.W.Shuy(Eds.),NewI1灼ysofAnalyzing吃"匂"b〃ノカ
English,(pp.340-373).Washington,DC:GeorgetownUniversity
Press.
Meyer,D.E.,&Schvanveldt,R.W.(1971).Facilitationinrecognizing
pairsofwords:Evidenceofadependencebetweenretrieval
operations.、ノburnalofExpe"海entalPsychology90,227-235.
Millis,M.L.,&Button,S・B.(1989).Theeffectofpolysemyonlexical
decisiontime:Nowyouseeit,nowyoudon't.Memory&Cog刀/加刀,
17,141-147.
Nation,I.S.P.(2001).LearningI/bcabα曲ryinAnotherLanguage.
Cambridge:CambridgeUniversityPress.
Rodd,J.,Gaskell,G.,&Marslen-Wilson,W.(2002).Makingsenseof
semanticambiguity:Semanticcompetitioninlexicalaccess.、ノ'ournal
ofMemoryandLanguage,芋6245-266.
Stahl,S.A.,&Nagy,W.E.(2006).TeachingWorぴ〃'eanings.Mahwah,NJ:
LawrenceErlbaumAssociates.
Webb,S.(2008).Theeffectsofcontextonincidentalvocabularylearning.
ReadinginaFore噌刀Language,20,232-245.
-25-
Abstract
LexicalPerformanceinLI:、M1atisNative-likeLexicalCompetence?
HidekiGoya
心的辞書に関する研究において、我々の語棄知識は意味的に結びついてい
ると言われている。またプライミングに関する研究では、word-sense
(Finkbeiner,2002)は母語話者の語桑使用に影響を与える事が分かっている。
このことより、心的辞書における語貧は、word-senseが共有される事によっ
て結びついていると考えられるが、これまでの意味判断実験において、
word-senseは十分に研究されているとは言えない。その上SLAに関する研
究では、母語話者の語糞知識は十分に熟達しているので、意味判断実験におい
て統計学的な差異を示さないと考えられたまま、word-senseを含まない実験
結果において第二言語話者の語童知識の熟達度の基準となっていた。
本実験ではword-senseを統制した語棄のベアを用い、母語話者in=20)
の意味判断実験での意味処理を観察した。被験者はコンピューターによる反応
速度を測る実験(オンライン実験)で、同義語の組み合わせ(〃=39)が意味
的に似ているかどうかの判断を行った。結果として被験者の意味判断の正確さ
に統計学的な差は観察されなかったが、意味的に似通ったペアに対しては意味
判断が統計学的に遅かった。以上の結果はword-senseに関する実験の方法論
的発展を示唆している。それは、(1)心的辞書内の語棄は意味的に組織され、
(2)共有されるword-senseの数の差は、母語話者の意味判断の語童処理に
影響を及ぼさない、という事であった。以上の結果は意味判断実験を用いる母
語研究と第二言語習得研究において、被験者の語蕊能力を評価する際の基準と
なるであろう。
-2 6 -