Macalester College DigitalCommons@Macalester College
Psychology Honors Projects Psychology Department
May 2005 Do Bilinguals Access a Shared or Separate Conceptual Store? Creating False Memories in a Mixed-Language Paradigm Aaron Mitchel Macalester College
Follow this and additional works at: http://digitalcommons.macalester.edu/psychology_honors
Recommended Citation Mitchel, Aaron, "Do Bilinguals Access a Shared or Separate Conceptual Store? Creating False Memories in a Mixed-Language Paradigm" (2005). Psychology Honors Projects. Paper 1. http://digitalcommons.macalester.edu/psychology_honors/1
This Honors Project is brought to you for free and open access by the Psychology Department at DigitalCommons@Macalester College. It has been accepted for inclusion in Psychology Honors Projects by an authorized administrator of DigitalCommons@Macalester College. For more information, please contact [email protected]. Do BilingualsAccess
Abstract
A centralfocus of bilingualisrnresearch is therepresentation of two languagesystems
in memory.The Revised Hierarchical Model (Iftoll & Stewart,lgg4) predictsthat
bilingualsaccess conceptual information from both languages simultaneously. To
testthis hypothesis, the present study attempted to createfalse memories across languages.Twenty-two Spanish-English bilinguals participated in a mixedlanguage associativelist paradigm,False recognition of targetwords was significantlyhigher thanfalse recognition of controlwords, both within andbetween languages. These resultsprovide evidence for a sharedconceptual store and parallel activation of languages,supporting the Revised Hierarchical Model, Do BilingualsAccess
Do BilingualsAccess a Sharedor SeparateConceptual Store: Creating False
Memoriesin a Mixed-LanguageParadigm
The majorityof the world's populationspeaks two or morelanguages.
Despitethis, the majorityof psychologicalresearch into languagehas focused on
monolinguals.Recently, however, there has been a surgeof interestin bilingualism.
A centralissue in currentbilingual research concerns the representationoflanguages
in memory,Specifically, are two languagesstored together, or separately?
Additionally,how arethese two languagesactivated?
Thesequestions form the basis for two competingmodels of bilingual
languageprocessing The independencehypothesrs proposes that thereare distinct
andseparate memory stores for eachlanguage, such that processingin onelanguage
doesnot affectprocessing in the other, In contrast,the interdepmdence hypothesis
maintainsthat thereis a sirrgleintegrated flemory store(see Gerard & Scarborough,
1989).The goal of the presentstudy is to examineeach of thesemodels and address
thequestions raised earlier, the first of whichis how two languagesare accessed, or
activated.
Definitions
The terminologyused in the bilingualliteiature varies considerably, and so it is helpfulto haveprecise operational definitions. First, what constitutes bilingualism?Definitions of bilingualismcan range anywhere from informal experiencewith two or morelanguages to near-nativefluency in two languages.As
Francis(1999) suggests, a middle ground must be foundbetween these two extremes.
Thus,for the purposesof this paper,bilinguals are those who areable to Do BilingualsAccess
communicate,at somelevel, in morethan one language, This definitionis admittedly
very broadand simple, Howeverra narrowdefinition incorrectly excludes many
peoplewho wouldconsider themselves bilingual. For instance,some people can
speaka languageyet areunable to reador write in a language.At the sametime,
thereare many people who areable to reador write in a languagebut areunable to
speakthe language(this is particularlycommon if the languagewas learnedin a
schoolsetting).In addition,many people in theUnited States are native English speakerswho canneither read nor write andmay not speakwith "proper"grammar, yet arenative English speakers nonetheless. Because of thisvaried nature of languageproficiency and use, broad definitions of bilingualismare inevitable.
It is alsoimporlant to definethe termsler.ical and conceptual. A lexicon, definedherein, is a collectionof lexicalentries, or knowledgeabout a particularword that includesofihographic, phonologic, morphologic, syntactic, and semantic properties(Francis, 1999). Theuse of lexiconand lexical variesgreatly among studiesand models, and so the presentpaper uses a broaddefinition to incorporatea
Exeaterrange of literature.
Theconceptual level of representationconcerns the meaningsof words.
Thereis somedisagreement in the literaturebetween the termssemantic and conceptual,as some researchers will usethem interchangeably while otherstudies differentiatebetween the terms (Francis, 1999), Herein, the two termsare used interchangeablyunless otherwise noted. Both areused to referto meaning-leveL information. Do BilingualsAccess
Ac tivat ion of Languages
How is informationretrieved from the bilingual lexicon? When presented
with a word, do bilingualsactivate one lexicon at a time,both at the sametime, or do
theyuse cues to activateonly the appropriatelexicon? Beauvillain and Grainger
(1987)and Grainger and Dijkstra (1992) suggest that there is an initial activationof
bothlanguages, and that language selection occurs at a laterstage. Several studies
provideevidence for suchparallel activation of lexicons.
TheStroop effect is interferencethat occurswhen attempting to namean item
with incongruentmeaning and form(e.g., the word blue printedin yellow ink). While
thiseffect has been well establishedin monolinguals(c. f Macleod, I 991),do
bilingualsexperience this sameinterference when the printedwords are in one
language(language A) andcolor namingis in anotherlanguage (language B)? chen
andHo ( I 986)hypothesized that if only onelanguage is activatedat a time,then there
will be no Stroopeffect because there is no lexicalactivation of the printedwords in
languageB andtherefore no lexicalinformation to interferewith color naming,
However,interference in this conditionwas comparable to interferencein conditions
whereboth the printedwords and color namingwere in the samelanguage. Thus,
lexical inforrnationwas availableftom both languages,supporting the parallel
activationhypothesis (for a furtherreview of bilingualStroop experiments, see Smith, reeT).
Fufthersupport for parallellexical activation comes from Nas ( 1983).In a
lexicaldecision task, Nas instructed Dutch-English bilingual subjects to onlyrespond yesto Englishwords and no to anythingelse, including Dutch words. If only one Do BilingualsAccess
lexiconis availableat a time,then in a languagespecific task there should be no
interferencefrom the otherlanguage. However, participants were slower to respond
to non-Englishwords if thestimulus was eithEr a Dutchword or soundedlike a Dutch
word. Thisinterference is evidencethat both lexicons are activated simultaneously.
Finally,a seriesof eyetrackingexperiments reveal that bilinguals activate both languagesin parallel(Spivey & Marian,1999; Marian, Spivey, & Hirsch,2003;
Marian& Spivey,2003). Russian-Englishbilinguals were asked in onelanguage to pick up oneof four objectsplaced in frontof themand their gaze was followedwith afl eyetracker. Whenasked to pick up the targetobject, the participants'gaee briefly shiftedto an interlingualdistractor object whose name in the irrelevantlanguage was phoneticallysimilar to thetarget object (Spivey & Marian,1999; Marian, Spivey, &
Hirsch,2003). For example,when participants were told in Russianto pick up the stamp("marku"), their gazefixated upon the marker,the distractorobject. The
Englishword "maxker,"phonetically similar to "marku,"should not havebeen distractinghad only theRussian lexicon been available. However, because it was distracting(significantly moreso than the other two objects),these results suggest that bilingualsactivate both languagesin parallel. Additionally,it appearsthat the irrelevantlanguage cannot be deactivatedwhile in a completelymonolingual context
(Marianet, al, 2003). In anothereyetracking experiment, Marian and Spivey (2003) comparedRussian-English bilinguals with Englishspeaking monolinguals. While bothbilinguals and monolinguals experienced within-language competition (i.e.,
Englishdistractor items and English target objects), only bilingualsexperienced between-languagecompetition (i.e. Russian distractor items and English target items). Do BilingualsAccess
Thisdifference in performanceprovides further suppoft for theparallel activation of
languages.
Representationof Languagesin Memory
Thereis muchdebate over the natureof bilinguallanguage representation. Do
bilingualshave one lexicon for bothlanguages or hvo discreetlexicons? Likewise,
what is the natureof conceptualrepresentation in bilingualmemory? A reviewof the
literatureon the debatesuggests that the majorityof the evidencesupports the
existenceofseparate lexicons and a sharedconceptual store (Francis, 1999); thus, the
presentstudy focuses specifically on this evidence.
Evidencefor separate le.ricons
Is lexicalinformation in bilingualsstored in oneshared lexicon or two
separatelexicons? The literature on thesubject suggests that the bilingual has two
distinctlexical representations in memory,one for eachlanguage (for a review,see
smith, 1997;Gollan & Kroll, 2001). In a fragmentcompletion task, Durgunofrlu and
Roediger(1987) found that the only variableto significantlyaffect completion rates was languageoverlap between study and test sections of the test. If the participants studiedthe list of wordsin the samelanguage as the completiontask, they hadhigher ratesof completion,suggesting that the participantswere accessing language-specific lexicons.Using a lexicaldecision task, Gerard and Scarborough (1989) found no cross-languagefacilitation from repetitionpriming among noncognate translations, thoughthey did find cross-languagefacilitation of cognatesand homographic noncognateswhere the wordsin eachlanguage.share an orthographicpattern. If the participantshad been accessing a shared lexicon,'then there would be facilitation Do BilingualsAccess
amongnoncognate translations as well ascognates and homographic noncognates.
Therefore,the resultsare consistent with thehlpothesis that lexicalinformation is
language-specific.
Evidencefor a shared conc eptual rep res ent ation
Thebody of literatureon conceptualrepresentationsin bilinguals points to a
common(or at leastpartially shared) conceptual-level episodic representation (for a
reviewof this literature,see Francis, 1999). Fromaneuropsychological standpoint, it
is generallybelieved that thetwo languagesof a bilingual arestored in shared,not
separate,anatomical areas (Paradis, 1997). Altarriba (1992) found greater priming
effectsfor translationequivalents than for unrelatedwords (e.g., table might havea
strongerpriming effect for silla [chair] thanventana[window]). Becausethe priming
occursacross languages on a semanticlevel, this suggests that the two languages
sharea conceptual-levelrepresentation, Likewise, Fox (1996)found significant
negativepriming (inhibition)of an attendedtarget word whenunattended
semanticallyrelated words or translationequivalents were presented in parafoveal
vision,or outsidethe "spotlight" of attention.In this experiment,Fox (199d)first
presenteda numberand required participants to respondas to whetheror not the numberwas even or odd. At the sametime, anunattended flanker prime was presentedin parafovealvision. Thenparticipants were asked to performa lexical decisiontask (i.e. "is this letterstring a word?"). If theunattended prime was semanticallyrelated to thetarget, then response times for thelexical decision task wereslower. For example,if cal is presentedin parafovealvision duringthe number task,then participants are slower at deciding perro (dog)is a word. Fox ( 1996)also Do BilingualsAccess
foundinhibition for translationequivalent target-prime pairs. This negativepriming
for translationequivalents and semantically related words is evidencethat conceptual
informationfrom both languagesis availablesimultaneously, suggesting that
bilingualsacces$ a sharedconceptual representation.
Modelsof Lexical/ConceptualRepres entation
A centralquestion in bilingualmemory research concerns the modelingof
bilingualcognitive structure. There are severalextensive reviews of modelsof
conceptuaylexicallevel episodic representation (e.g;, Dudsic, 1999; Kroll, 1993;
Kroll & De Groot, 1991).Therefore, each model will only be briefly reviewed.
Weinreich(1953) presented the first of thesemodels. He proposedthree types
of relationshipsbetween the two languagesof a bilingual coordinate,compound, and
subordinate.In coordinatebilingualism, each language has both a distinctlexical and
a distinctconceptual representation. In compoundbilingualism, each language
independentlyaccesses a single conceptual representation. In subordinate
bilingualism,the secondlanguage (LZ) is dependenton the first language(Ll) to
accessconcepts. Weinreich's outline of bilinguallanguage representations provided
muchof thebasis for subsequentmodels (Dudsic, 1999).
TheWortl Association model
TheWord Association model is similarto thesubordinate model. In theWord
Associationmodel, L2 accessesconcepts tttrough the lexicon of Ll (seeFigwe 1).
Whenpresented with a word in L2, a bilingualwill translatethe word into Ll and
accessthe conceptualrepresentation for thatword, Naminga pictruein L2 shouldbe a five-stepprocess: (1) recognizeimage; (2) retrieveconcept; (3) retrieveLl word; Do BilingualsAccess 10
(4) retrieveL2 word;(5) sayL2 word, Followingthis, picture naming should take
considerablylonger than translating a word fromLl to LZ becausetranslation has
fewersteps [(l) recognizeLl word;(2) retrieveL2 word;(3) sayL2 word]. potter,
So,Eckardt, and Feldman (1984) tested this hypothesis and found that latencies in
bothtasks were not significantlydifferent, Therefore,it takesapproximately the
sameamount of time to namea picturein L2 as it doesto translatea word from Ll to
L2. Thisresult implies that both Ll andL2 lexiconsaccess the conceptualstore
directly.This led Potteret aL ( I984)to proposethe Concept Mediation rnodel.
InserlFigure I abouthere
TheConcept Mediation model
TheConcept Mediation model is similarto Weinreich's(1953) compound
model. It proposesthat bilinguals mediate concepts directly between the separate
lexiconsand the conceptual store (see Figure 2). Becauseeach language has direct
accessto the conceptualstore, picture naming and translation require the same numberof processingsteps, and so Potteret al.'s( 1984)results are consistent with thismodel of representation.However, is therepresentational system static, or doesit changeover time?
InsertFigure 2 abouthere
The developmentalhypothesis Do BilingualsAccess I I
Thedevelopmental hypothesis states that, initially, bilingual cognitive
architectureis representedby theWord Association model, and as proficiency in the
L2 increases,there is a shifttowards a ConceptMediation model (Dudsic, l9g9;
Krolland Sholl, 1992; Chen, 1992). Dufour and Kroll (1995),using a sentence
verificationtask similar to thatof Smith,Shoben, and Rips (1974), observed that
morefluent bilinguals were able to successfullyuse conceptual links in within-and
cross-languageconditions while less fluent bilirrguals were more dependent upon
theirLl. As proficiencyincreases, there is a shift from translationstrategies towards
directconcept mediation. A numberof studieshave found that the amountof
interferencein a Strooptask is positivelycorrelated with proficiency,suggesting a
developmentalshift in processing(Chen & Ho, 1986;Mflgiste, 1984; Mflgiste, 1985).
Becausethe developmental shift is gradualrather than abrupt (Chen, 1992), there is a
periodwhere bilinguals use both translating and conceptual mediation strategies.
Therefore,evidence for the developmentalhypothesis also lends support for a mixed
representationalsystem, However, do bilingualswho arefluent in their L2 still use translatingstrategies? According to the developmentalhypothesis, once strong conceptuallinks are established bilinguals will only usedirect conceptual mediation andlexical links will deteriorate.However, this does not seemto bethe case. Cross* languageexperiments provide evidence that lexicallinks aremaintained by high proficiencybilinguals. Studies have found that translation from L2 to Ll is faster thantranslation ftom Ll to L2 for bilingualsof all proficiencylevels (Kroll, 1993;
Dudsic,1999). Thus, Iftoll andStewart (1994) proposed a mixedrepresentational modelof conceptuaVlexicallevel episodic representations, the Revised Hierarchical Do BilingualsAccess 12
Model(RHM).
The Rev{sedHterarchtcal Model
TheRHM is a hybridof theW'ord Association and Conceptual Mediation
models.It positsthat both lexical and conceptual-level links exist(see Figure 3). In
thismodel, there is a sharedconceptual store and language-specific lexicons, and both
translationand conceptual mediation strategies are used. It takesinto account
evidencefor a developmentalshift, since bilinguals rely heavilyon theirLl for retrievingconcepts while theyare acquiring their L2. However,as they become more prof,tcient,they rely lesson translationand are more able to accessconcepts directly.
A centralaspect of the RHM is its asymmetricalstructure. This modelassumes that asa resultof theway in whichsecond languages are learned, allL? wordsconnect to
Ll words,but all Ll wordsdo not necessarilyconnect to L2 words(Kroll & Stewaft,
1994).Thus, there is a strongerlexical link fromLZ to Ll (explainingwhy backwardstranslation is fasterthan forwards translation), and a strongerconceptual link betweenthe conceptualstore and the Ll lexicon(for a reviewof the RHM, see
Ifuoll & De Groot, 1997).The asymmetryof the modelis supportedby numerous cross-languagepriming studiesthat foundfaster information transfer from Ll to LZ thanfrom L2 to LI (e.g,,Dudsic, 1999; Kroll andStewart , 7994;Haftsuiker,
Pickering,& Veltkamp,2004).
InsertFigure 3 abouthere
Problems with the RevisedHierarchical Model Do BilingualsAccess 13
A key concernwith the RHM is its asymmetricnature. TheRHM assumes
that forwardtranslation is largelyconceptually mediated while backwardtranslation
is not conceptuallymediated, However, there is evidencethat both forwardand
backwardtranslation are conceptually mediated (e.g., Bloem & La Heij, ?003;La
Heij, Hooglander,Kerling, van dervelden, 1996). Additionally, the RHM cannot
accountfor evidenceindicating that bilingualmemory is a dynamicrepresentational
system,and that effectsof translationdirection are not fixed characteristics(Heredia,
1997). Ll canfall in strengthand L2 canbecome the dominantlanguage. Likewise,
if Ll andLZ fluencyare comparable, then priming effectswill be symmetrical(Kotz,
2001). Thus,translation direction effects are depurdant upon proficiency, which is a
dynamicvariable.
In addition,the RHM doesnot takeinto accountthe full representationaland
processingvariation within thebilingual individual (Crosjean, 1998). Each bilingual
individualvaries greatly from the next,and any static representational model is going
to havetrouble accounting for thisvariation. Kim, Relkin,Lee, and Hirsch (1997)
foundthat in Broca'sarea, second languages are spatially separated when acquired in
adulthood,but not separatedwhen acquired during an earlylanguage acquisition
stageof development,Therefore, when the secondlanguage is acquiredcan have an
effecton the representationof languages.Moreover, a high motivationlevel can eliminatetranslation direction effects (Luna & Perachio,2002). Motivationlevel and ageof acquisitionare just two of the variablesthat caninfluence the language processingof bilinguals.Heredia (1997) suggests that rather than referring to Do BilingualsAccess 14
languagesin representationalmodels as Ll andL2, itmight bebetter to "describe
bilingualmemory as a functionof languagedominance" (p. 38).
Finally,there is evidencethat not all wordsare represented identically in the
conceptu4lstore and that translationequivalents do not alwaysactivate the same
concept.De Croot(1993) reviews several studies that indicate wordrtype effects in
processingtasks among bilinguals, While translation times for concretewords are
consistentwith the RHM, translationtimes for abstractwords are contrary to whatthe
RHM would predict(see also Heredia, lggT). Accordingto the RHM, all words
shouldbe represented similarly. However, representation is not independentof
concreteness.The RHM cannotexplain such word-type effects, Furthermore, Blot,
Zarate,and Paulus (2003) suggest that switching to Ll from L2 not onlypermits
strongactivation of concepts,but alsoactivation of newconcepts. This indicates that
the two languagesof a bilingualdo not alwaysaccess the sameconcepts, contrary to
whatthe RHM predicts.[nstead, Blot et al. (2003)propose a systemthat incorporates
differentiatedconcepts rather than simply a cofirmonconceptual store,
Thedis tributed conceptual feature model
The distributedconceptual feature model accounts for word type effectsand differentiatedconcepts. It proposesthat each lexicon accesses a set ofconceptual features,or meaningelements (De Groot, 1992),The meaningsof translationsdo not alwayscompletely overlap. For instance,the Englishwordp/ay is a very general term,which canenQompass playing a spoft,a boardgame, or an instrument. o'to However,the Spanishwordjugar canmean, play," but is usedto referto playing a spottor a boardgame, while tocar is usedto referto playingan instrument,Each of Do BilingualsAccess 15
thesemeaning elements would be represented by a differentnode (see Figure 4), and
so the conceptualrepresentations of play andjugar only partiallyoverlap.
Additionally,concrete translations, cognates, and nouns often share more meaning
elementsthan abstract words, noncognates, and verbs (Van Hell & De Groot,lgg8).
Thedistributed conceptual feature model accounts for word-typeeffects and the
differentiationof concepts.However, recent literature has pushed for the inclusionof
a third levelof representation,the lemmalevel, to makea distinctionbetween
semanticand conceptual representations (Grosjean, 1998; Pavlenko, 1999; Hartsuiker
et al.,2004).
InsertFigure 4 abouthere
Thefinal model of conceptual/lexicallevel episodic representation is the distributedconceptual/lexical feafure model (Kroll & De Groot, l9g7). In this model, lexical featuresare distributed in a similarmannor as conceptual features and it incorporatesa lemmalevel (seeFigure 5). The lemmalevel of representationin this contextincludes syntacfic and semantic information, and the lexiconincludes orthographicinformation and other physical properties of a word. It is meantto 'orepr€sent thepatterns of activationbetween word formsand meanings...that may allow the bilingual'stwo languagesto be influencedby oneanother and to share accessto a commonpool of lexicaland conceptual feafures but, at thesame time, enablefunctional autonomy when only one langiageis active"(Kroll & De Groot,
1997,p.l9l). Thus,ifthebilingualisinamonolingualcontext,thenthelemmawill Do BilingualsAccess 16
weightactivafion of the relevantlanguage accordingly. This model,like the distributedconceptual model,accounts for word-typeeffects and the differentiation of concepts' However,because it alsoincludes a distributedlexical representation, thelevel of facilitationor interferenceis not onlya functionof conceptualoverlap,
but alsoof consistencybetween word forrns,
InsertFigure 5 abouthere
Roedigerand McDermott (lggj)
To testthe predictions madeby the RHM anddistributed moders, mqny researchershave begun replicatingclassic cognitive psychology experiments (e.g., Strooptask) in a mixed-language paradigm.One such experimental task is the Deese-Roediger-McDermott task(DRM). Thistask, first employedby Deesein 1959'was used by Roediger& McDermott(19g5) to explorethe creation of false memories'The ta$k consists of presentingparticipants with an auditorylist of words that arehighly semantically or conceptuallyrelated to a criticalnon-presented target word (cNw). For example, table,sit, regs,secr,and coucharepresented (along with ten otherassociates), ail of whichare highry associated with thecNV/' chair. Participantsthen complete a recognitiontask, indicating whether the test item has beenpresented earlier in oneofthe studylists. Testitems fall into oneof three categories; presentedwords, non_presented words, and CNW'. In the DRM paradigr4cNWs havea substantialryhigher rate of false recognitionthan non-presented words. Fufthermore,the falserecognition rate for Do BilingualsAccess t7
cNWs was comparableto the hit ratefor presentedwords, suggesting that
pafiicipants"were unableto distinguishitems actually presented from the critical
lures[cNWs] thatwere not presented"(Roediger & McDennott,1995, p. 808).
qssdigerand McDermott (1995) explained these results by suggestingthat false
recognitionis producedthrough the activation of implicitassociative responses,
Falserecognition occurs at encoding,as items associated with the presentedobject
areactivated (e.9., when table and seat are presented, chair is activated).They also
askedparticipants to makea remembervs, lvtow judgmentfor itemspresented in the
recognitiontask. A rememberjudgment is when"the subjectcan mentally relive the
experience(perhaps by recallingits neighbors,what it madethem think of, whatthey
weredoing when they heardthe word, or physicalcharacteristics associated with its
presentation)"(Roediger & McDermott,1995, p. 807). In contrast,a know decision
is "madewhen subjects are confident that the item occurredon the list but areunable
to reexperience(i.e., remember) its occurrence"(p. 807). Contraryto whathad been
expected,they foundthat participants remembered, rather than knew, the presentation
of the CNWs,indicating that "conscious recollection" can create false memories.
Presentinglists of highlyassociated words can create false memories. In essence,these false memories of theCNWs are created through conceptual activation of associates.Roediger and McDermott (1995) were looking at monolingual activationof a singlelexicon. However, one would predict that if bilingualsactivate a singleconceptual store, then false recognition should occur across languages. That is, if the studylists arein onelanguage and the recognitiontask in other,CNWs Do BilingualsAccess l8
shouldstill be incorrectlyremembered more frequently than other non-presented
words,
Kawasaki-Miyaji,Inoue, & Yama(2003) tested this prediction, using a cross-
languageDRM paradigmto createfalse memories across two languages.They
presentedlists in eitherEnglish or Japanese,and then gave a mixed-language
recognitiontest. Recognitionof presentedwords was better when the languagewas
matched,CNWs were falsely recognized at approximatelythe samerate as correct
recognitionof presenteditems. Thus, even when the study and test languages are
different,there is still a highrate of falserecognition. This result appears to support
the interdependentstorage hypothesis, as conceptual information is beingactivated
andthen accessed in bothlanguages. However, Kawasaki-Miyaji et al. (2003)
interprettheir resultsiN supportingneither the independentor interdependentstorage
hypotheses.They maintain that if the interdependenthypothesis is accurate,then
languagecongruency between study lists andtest items should have no effect. In
otherwords, one would not expectto find a greaterperformance on test iternswhen
the studyword andtest word werepresented in the samelanguage, Since they found
a languagecongruency effect, they concluded that their datadid not supportthe
interdependencehypothes is.
Thereare two chiefconcerns with theKawasaki-Miyaji et al. (2003)study.
The first is their interpretationof theresults. The higher correct recall rate for matched-languagewords is not necessarilyevidence against the interdependence hypothesisThe recognition of thepresented words is not entirelyon a conceptual level,but insteadis largelyon a lexicallevel. In matched-languagepresented Do BilingualsAccess l9
words,there are not onlyconceptual memory cues, but alsolexical cues. The
presenceof lexicalcues in onecondition and not in theother led to asymmetrical
activationof lexicalinformation. One would only expectrecognition rat€s between
matched-languageand non matched-languagewords to be the sameif lexical
informationwas available in bothconditions. Hewever, because bilinguals to not
accessa singlelexical store for both languages,lexical information is not goingto be
availablein the non matched-languageconditions, Therefore, any difference in
recognitionrates is morelikely evidencethat thereare two separatelexicons than
evidenceagainst a sharedconceptual store. Kawasaki-Miyaji's (2003) results are in
fact congruentwith mostcurrent models of bilingualmemory representation that
proposeseparate lexicons and a sharedconceptual store, such as the Revised
HierarchicalModel (Kroll & Stewart,1994).
Thesecond concern is with theirmethodology. They presented each study list
in onelanguage, and the recognitionlist alsoin a singlelanguage. They failed to take
into accountwhat Grosjean (1998) termed "language mode." Grosjean(1998)
suggeststhat there are differences in performancedepending on what languagemode
a bilingualperson is in at that moment. If the participantsare in a bilingualmode, theninformation from both languagesmay be morereadily available than if theyare in a monolingualm.ode. Because Kawasaki-Miyaji et al.'s(2003) study lists were in onelanguage (e.9., the list for chair would eitherbe entirelyin Englishor entirelyin
Japanese),it is plausiblethat their participantswere in a monolingualmode during encodingand this mighthave had an effecton their results,Instead, what the present studywill do is presentlists in Englishonly, Spanish only, and in bothEnglish and Do BilingualsAccess 20
Spanishcombined (e.g., the list for chair tnclvdesitems in Englishand items in
Spanish;see Appendix A). Thegoal of presentingthe wordsin a mixedlanguage list
is to exploreany languagemode effects. If both languagesare present during
encoding,then the participants should be in a bilinguallanguage mode. Likewise,if
the studylist is Englishonly or Spanishonly thenpafticiparts should be in a monolingualmode. Thus,if falserecognition is greaterfor mixedlists, then this wouldbe evidence for thepresence of a languagemode effect.
Thegoal of thepresent study was to examinethe nature of conceptual-level episodicrepresentation in Spanish-Englishbilinguals, In additionto exploring languagemode effects, I alsohope to providefurther illumination of the independent- interdependentdebate. To testwhether conceptual information is activatedin both languagessimultaneously, I employed a DRM falsememory task in a cross-language paradigm.If falserecognition of the CNWswere significantly more frequent than falserecognition of non-presentedwords, then this would support the interdependencehypothesis. Furthermore, the Revised Hierarchical Model (I{roll &
Stewaft,1994) predicts asymmetrical rates of falserecognition, More specifically, wordsstudied in the Ll shouldhave higher rates of falserecognition. This asymmetryis theresult of greaterconceptual mediation irr the Ll. Similady,I predict thatthere will be a positivecorrelation between rate of falserecognition and language proficiency.As L2 proficiencyincreases, conceptual mediation should increase (as is congruentwith the developmentalhypothesis); thus, conceptual activation of the criticallure is greaterand should result in higherrates of falserecognition.
Method Do BilingualsAccess 2I
Participants
Twenty-two undergraduatesfrom MacalesterCollege participated in a half-
hour experimentfor five dollars, Fifteenwomen and sevenmen participatedin the
study. Participantswere Spanishand Englishbilinguals with varying proficiency
levels. As mentionedearlier, bilinguals are definedas thosewho are ableto
communicate,at somelevel, in more than one language,in this caseEnglish and
Spanish.
Eachparticipant's proficiency was measuredin both Englishand Spanish
using a Curriculum Based Measurement(CBM). Originally used to assessreading
skill, it has beenadapted to measurelanguage proficiency (for a review of CBMs, see
Marston, 1989;Hasbrouck & Tindal, 1992). I alsogave parlicipants a languageuse
questionnairethat gatheredinformation about the languageprofile of eachparlicipant
(e,g,,when they learnedeach language and how proficient they feel in each
language).There were 16 nativeEnglish speakers,7 native Spanishspeakers, and 1
native Albanian speaker(this participant felt more proficient in English than in
Spanish).
Materials
I usedE-prime software (Psychology Software Tools, 1995-2000)on
computersrunning Pentium [V processorsto presentstimuli and gather data. The
presentstudy used 18 lists from the appendixof Roedigerand McDermott (1995).
Six lists were dropped from the original 24 due to translation difficulties (this matter will be addressedlater). Roedigerand McDermott createdthese lists from Russell and Jenkin's word associationnorm study in 1954. The 18 studv lists were Do BilingualsAccess 22
arbitrarilydivided into threegroups: Spanishonly, Englishoniy, anda mixed
Englishand Spanish group, Therewere three versions of the experimentwith
differentlist orderingto providecounterbalancing.
Eachlist consistedof the 15most coflrmon associates of thetarget word
(CNIV) For example,the list for the targetword chair is table, stt,legs, seat, couch, desk,recliner, sofa, wood, cushion, swivel, stool, sitting, rocking,and bench. The only exceptionis thelist for Spider,which included l4 wordsafterfeelers was removedbecause there was no translationequivalent in Spanish.The words were presentedvisually on a computerscreen in sequentialorder, and were on thescreen for a durationof two seconds.In betweeneach list participantswere asked to completemath questions. The math questions were difficult algebraic questions wherethe par-ticipant needed to solvefor'{
Becausethe stimuli were presented visually, the present study's design deviatesfrom Roedigerand McDermott's (1995) original design that presented auditorystimuli.
Gallo,McDerrnott, Percer, and Roediger (2001) found that auditory presentation of the studylists led to greaterfalse recognition than visual presentation, However, visualmodalities still createhigh rates of falserecognition (Gallo et al.,2001;
Kawasaki-Miyajiet al.,2003); thus, I wouldnot predictany modality effects to confoundthe results.
Thelists for theSpanish and mixed language conditions were translated into
Spanishby four differentnative speakers of Spanish,The first codertranslated
Englishlists into Spanish.The second coder then translated this back into English. Do BilingualsAccess 23
Thethird codertranslated the originalEnglish lists into Spanishagain in orderto
providea secondset of Spanishlists. Thefinal coderresolved any discrepancies
betweenthe first threecoders, The finalized lists were compiled by comparingthe
translationsof the first andsecond coders. If the forwardand backward translations
matched,then I acceptedthe translation.If the forwardand backward translations did
not match,then I sentthe first andthird coder'stranslations to the fourthcoder ro
resolvethe discrepancy,I first askedthe fourth coderto translateeach word from
Spanishto English.I thenasked which word would be a mereaccurate translation of
the Englishword. In total,there were 9 wordswhich weresent to the fourthcoder,
andall exceptfeelers(as mentioned earlier, this wasremoved from the list) were
successfullyresolved.
Therecognition task list consistedof 144words, 72 studiedand 72 non-
studied. Thestudied words were drawn from the l"t, 6*, lOth,and l3th serial
positionsfrom eachlist (thisis consistentwith theselection method used by Roediger
& McDermott,1995). Half of the "studied"test items were translation equivalents of
theoriginal word in thestudy list (i,e.,the "studied" words were tested in eitherthe
sameor differentlanguage as they had been studied). Of the 72 studiedtest items, 36
werein thesame language and 36 werein theother language.
The72 non-studiedwords consisted of l8 criticallures and 54 non-studied. non-targetwords, Half of thecritical lures were in Spanishand half werein English.
Therewere three critical luresin eachlanguage for all threeconditions (Spanish only,
Englishonly, andmixed). The 54 non-studiedtest items were never identical to or Do BilingualsAccess 24
semanticallyrelated to thewords in thestudy lists, Half of thenon-studied test items
werein Englishand half werein Spanish,
Design
Theexperiment used a 3 (studylist languagein English,Spanish, or mixed)
by 2 (testlanguage of Englishor Spanish)within-subjects design. The dependent
variableswere the recognition rate of testitems and the proficiency of theparticipant.
Procedure
Participantswere told the experimentwas exploring memory. Participants
wereinstructed that they would see a numberof lists,each followed by a math
question,and thenwould be askedto decideif wordshad been presented in oneof the
lists. Theexperimenter gave verbal instructions to supplementthe on*screen
instructions,and did not remainpresent during the experiment, There was a brief
practicesession to familiarizeparticipants to thepresentation of thelists and math
questions.The study items appeared on thescreen one at a timefor two seconds
each. After eachstudy list participantswere required to answera difficult math
question.The purpose of thesemath questions was to removeany memoryof list
wordsfrom short-termmemory. Participantshad fifteen seconds to answerthe math
questions.However, the experiment did not proceeduntil the endof thefifteen
seconds,even if the participantanswered the mathquestion before the endof the allottedtime. The experimenterinformed the participant that the mathquestions were designedto be difficult andso instructedparticipants to "do their best,"but to not becomeanxious or wolry if theycould not answerthe questionin theallotted time. Do BilingualsAccess 25
During the recognitiontest, a singleword appearedon the screenand
participants used a respoflsebox to indicate whether or not the word had been
presentedin one of the study lists. The participant'srssponse, the correctresponse,
and the accuracyof the parlicipant'sresponse were recorded, Feedbackabout the
accuracyof the participant's responsewas not availableto participants during the
recognitiontask.
Additionally, if the participant recognizedthe test item, a follow up question
askedthe participant to indicate whether they rememberedor knew that the item had
appeared. Participantsreceived verbal and written instrucfions for how to makethe
remember/knowjudgment. Theseinstructions defined a rememberexperience as
"one in which you can mentally relive the experience(perhaps by recalling its neighbors,what it made you think of, what you were doing when you saw the word, or physicalcharacteristics associated with its presentation)."In contrast,aknow judgment "is made when you are confident that the item occurred on the list but are
'remember') unableto reexperience(i.e. its occurrence"(taken from Roediger&
McDermott,1995, p. 807).
Results
LanguageProfi.ciency
Therewere two testsof L2 languageprof,rciency: an objectiveCBM measure anda subjectiveself report score on a scaleout of 10. As a measureof convergent validity,the scores from thesetwo testswere compared. Figure 6 illustratesthe positivecorrelation between CBM scoresand self-reported L2 proficiency. Do BilingualsAccess ?,6
InsertFigure 6 abouthere
This correlationwas significantat a .05 alphalevel , t = ,467, p = .02g,with an r2 of
'218. Thus, CBM scoresagreed with parlicipants'subjective opinion of their own
proficiency.
Recognition
During the recognitiontest, participants indicated whether or not a word had
beenpresented during the study session.Test items fell into three categories:words
that had been presentedin the study session(studied), words that were new and had
not beenpresented in the study session(unstudied), and CNWs that were new but
were the semanticallyrelated targetsof the study words (critical). Recognitionrates
were calculatedt for eachcategory, such that the recognition rate for the studiea
categoryrepresents coffect recognition,and the recognitionrates for unstudiedand
critical words representsfalse recognition. Studied items were coffectly recognized
57% of the time, unstudieditems falselyrecognized 13% of the time, and critical
items falsely recognized4T%ofthe time. Beforecomputing inferential statistics, the
propofiional datawas noffnalized using an arc sin transformation.
A one-way repeatedmeasures analysis of variance (ANOVA) revealedthat
there was a significant difference in recognition rate betweenstudied, unstudied,and
criticaltest items, E(?,42): 85.51,p < .001(see Figure 7), An LSD painvise comparisonindicated that both studied words and critical lures were recognizedmore frequently than unstudiedwords (p<.001), Studiedwords were recognizedmore
I Dueto a recurringmechanical error, there were instances in whichresponding to onetest wor6 resultedfu skippingover items in therecognition test. Thus,any individual word thathad a response time at or below 100ms wasremoved from the dataset. Therewere lg totatinstances of skipping Do BilingualsAccess ?7
oftenthan CNWs (p:.007); however, the effect size, as measured by Cohen'sd, of
thiscomparison was relatively small (.64) when compared with thedifference
betweencritical or studiedwords and unstudied w ords(2 . I 6 and3 . 0 I respectively).
This suggeststhat falserecognition rates for CNWsand correct recognition rates for
studiedwords were roughly similar. Overall,effect sizes and the estimated observed
power(L00) werelarge, suggesting that there wa.s a goodchance of correctly
rejectingthe null hypothesis.
InserlFigure 7 abouthere
Studylists fell into oneof threeconditions: English only, Spanish only, and
mixedEnglish and Spanish. Grosjean (1998) suggests that it is necessaryfor
bilingualresearch to considerthe language mode of theparticipants, such that one
might expectdifferent results if the participantswere in a bilingualmode as opposed
to a monolingualmode. Thus,the presentstudy predicted that the mixedlanguage
conditionshould result in higherfalse recognition of the criticallures. This
predictiorrwas not supportedby theresults. For bothstudied and critical lists Spanish
only listshad the highest rate of recognition(.64 and .489 respectively), followed by mixedlists (.596 arrd.474) and then English only lists (.aa3 and.436; see Table l). A
2 (studiedvs. criticalword type) X 3 ( Englishonly vs.mixed vs. Spanishonly study lists)repeated measures ANOVA confnmeda maineffect for wordtype, F (1, 2l) :
6.582,p:.016. In addition,there was a maineffect for studylanguage, F (2,42) =
6.058,p = .005. Therewas also a significantinteraction between word tlpe andstudy Do BilingualsAccess 28
language,F (2,42) = 4.37Q,p: .019.Thus, the main effect of studylanguage
appearsto be differentfor studiedand critical words. Two plannedpost-hoc
comparisonsrevealed that there was a languageeffect for studiedwords, F (2, 42)=
19.218,p < .001,but was not significant for criticalwords, F (2,4l) < l.
Specifically,for studieditems Spanish only andmixed conditions resulted in higher
recognitionrates than English only conditions (p< 001). Contraryto predictions,
therewere no significantlanguage mode effects on falserecognition of criticallures.
lnsert Table 1 and Figure 8 about here
A 2 (studiedvs. criticalword type)X 2 (samelanguage vs. differentlanguage)
repeatedmeasures ANOVA againconfirmed the predicted word tpe effect,F ( I , 2 1)
: 5.975,p: .023.There was a moderatelysignificant congruency effect, F (1,2l) =
2.975,p = .099. However,there was a significantinteraction between word typeand
congruencyeffect, F (1,21)= 11.198,p: 003. To examinethis interaction, two
plannedpost-hoc tests were conducted. Consistent with Kawasaki-Miyajiet al.'s
(2003)findings, studied words were coffectly recognized more frequently if theword
wastested in thesame language as at thetime of study. Whenthe study and test
languageswere the same, the mean recognition rate was .65; when they were
differentthe mean recognition rate was 49. Thiseffect was significant at a .05level,
! (21): 5.52Q,p<.001, However, this result does not suggestthat the two languages arenot conceptuallyinterdependent, as will be discussedlater. Additionally,there wasno significantstudy-test language congruency effect for criticalwords, t (21) Therewas no differencein falserecognition rate between critical luresthat were testedin thesame language as their target study list (e.g.,the list for/aaf was presentedentirely in English,and the CNW/ool wastested in English)and those that weretested in a differentlanguage. Thedevelopmental hypothesis predicts that as L2 proficiencyincreases so too shouldfalse recognition of criticallures. However, this was not supportedby thedata as falserecognition of thecritical lure was slightly negatively correlated with L2 proficiency,I: -.048,p = .833. Contraryto thedevelopmental hypothesis, an increasein L2 proficiencydid not cortespondwith an increasein conceptual mediationand false recognition of CNWs. Finally,the Revised Hierarchical Model (Kroll & Stewart,1994) incorporates Ll andL2 asymmetriesthat would predict a differencein falserecognition of CNWs betweenlists that werepresented in the participants'Ll andlists that were presented in the L2, Theresults do notsupport this prediction, as there was no significant differencebetween these two conditions,I(Zl) Discussion Implicationsfor theoriesof bilingual languagerepresentation Theprimary goal of thepresent study was to testtheories of conceptual-level episodicrepresentation of two languagesin memory.The interdependence hypothesismaintains that thereis a singleintegrated memory store (Gerard & Scarborough,1989), and as such would predict parallel activation of bothlanguages. If, in the DRM falsememory paradigm, participants were more likely to falsely recognizethe critical lure than non-studied, non-target words, then this would provide Do BilingualsAccess 30 evidencefor theinterdependence hypothesis. Indeed, false recognition rate of the criticallures was significantly higher than unstudied words. Moreover,the false recognitionrate for CNWsapproached the correctrecognition, or hit, ratefor studied words. Thesehigh.rates of falserecognition occurred between languages, suggesting thatthe lists of associateswere activating the target word in bothlanguages. This patternwas found regardless of thestudy and test languages, Because of thisparallel activationof languages,these results provide support for the interdependence hypothesis. Thedevelopmental hypothesis states that as L2 proficiencyincreases, so too shouldconceptual mediation (Kroll & De Groot,1997). If conceptualmediation increases,false recognition of theCNW shouldincrease. Contrary to thisprediction, thepresent study found no correlationbetween L2 proficiencyand false recognition of the CNW. This seemsto provideevidence against the developmentalhypothesis. It is possible,though, that the CBM usedin thepresent study may not havebeen an appropriatemeasure of proficiencyfor theDRM falsememory task. TheCBM was originallyused to assessreading ability, andso whenit is adaptedto measure languageproficiency, it is measuringreading proficiency in that language.For instance,one participant reported that shedid not know manyof the wordsused in the study,yet shedid relativelywell on the CBM. As mentionedearlier, there are many differenta.spects to languageprof,rciency. A personmight be good at readingin a languageand at the very sametime not havea largevocabulary. The CBM was assessingreading proficiency, not vocabularysize, and so may not havebeen an Do BilingualsAccess 31 appropriatemeasure of proficiency.Thus, fuither research is neededto examinethe effectof languageproficiency on falserecognition of criticallures in theDRM. G'osjean(1998) argues that bilingualresearch needs to takeinto accountthe languagemode of the participant.To examinelanguage mode effects, the present experimentdivided study lists into Englishonly, mixed, and Spanish only. I predictedthat the mixed language condition would result in a higherfalse recognition ratefor CNWsbecause it would establishthe padicipantin a bilingualmode as opposedto a monolingualmode. The results did not supportthis prediction, as there wereno significantdifferences between the three conditions. An alternativeexplanation of theseresults is that languagemode effects may not havebeen observed since it is not clearthat the participantwas actually in one modeor the other. It is perhapsuueasonable to assumethat viewinga fifteenword mixed-languagelist of associatesmight effectively establish a bilingualmode. Additionally,the listswere presented in a randomorder such that theparticipant wouldnot haveknown what type of list it was,suggesting that a bilingualmode might havebeen maintained throughout the experiment,regardless of study-list condition. Again, fi.utherresearch is neededto explorelanguage mode effects. Kawasaki-Miyaji,et al, (2003)found that rates of correctrecognition of studieditems were higher when the languageat the time of studyand at the time of testingwas congruent.They interpreted such language congruency effects as conflictingwith the interdependencehypothesis. The present study found similar languagecongruency effects, as the hit ratefor studieditems was significantly higher whenthe testand study languages were the same,However, as mentioned earlier, Do BilingualsAccess 32 this is not necessarilyinconsistent with theinterdependence hypothesis. Both the RHM (Kroll andStewart, 1994) and distributed models of languagerepresentation predictthat recognition is not merelya resultof conceptualoverlap, but alsoof lexicaloverlap. According to thesemodels, the greater the lexicaloverlap between studyand test, the higherthe hit ratefor studieditems shsuld be. Furthermore,the presentstudy found no languagecongruency effects for criticalwords, That is, CNWswere not morelikely to be falselyrecognized if thetest language was the same asthe languageof the studylist. Therewould be no lexicaloverlap for same languageCNWs because the rvord was never presented in a studylist. Becausethere is no lexicaloverlap, there should be no increasedrecognition rate. Thus,the results of thepresent experiment are consistent with theRHM (Kroll & Stewart,1994) and distributedmodels of languagerepres entation. Finally,the RHM (Kroll & Stewart,1994) predicts that there should be languageasymmetries in falserecognition rates for CNWs. No suchasymmetry was found in the presentexperiment. However, this might be explainedby the participants'high L2 proficiency. Languageasymmetries should decrease as the L2 proficiencyincreases and the participant gets closer to themythical "balanced bilingual"state. I recruitedparticipants from upperlevel Spanish classes and who were oftenSpanish majors. Theseven native Spanish speakers were all very proficientin Englishas they had beenliving in the UnitedStates for severalyears and hadbeen taking classes in English.Because all of theparticipants in thestudy were highly proficientin the L2, the RHM (I{roll & Stewart,1994) would not have predictedlarge LI-LZ asymmetriesin falserecognition of CNWs. Do BilineualsAccess 33 Limitations of thepresent study The first lirnitations of the present study are the translation difficulties that arosewhile putting togetherthe Spanishonly and mixed word lists. The word lists consistof associates,and so many times this resultedin a number of synonymsin each list. For example,the list for anger includes ire, wrath, rage, andenrage, When translating the lists into Spanish,it was difficult to find a separateand distinct Spanishword for eachof the Englishwords. Therewas not alwaysa direct translation equivalent, and this was problematic when attempting to createlists of associates.Because there might not be a separateword for wrath andanger, participants might incorrectly translatethe word for wrath as anger. If, due to translation errors, participantsare studying the CNWs, then false recognition of the critical lure would be spwiously high. This situationoccurred frequently for abstract words, as the distinctions betweenabstract words are less definite than between concretewords. Becauseof this translation difficulty, I removed two study lists that were particulady difficult to translatefrom each condition (the two lists from the English only conditionwere chosenat random), Another translation difficulty occurred as a result of dialect differencesin Spanish. There are many dialect differencesbetween the Spanishspoken in Spain and Latin America. There is even considerabledifference betweendialects of Spanishwithin Latin America. There were severalwords that would be translated one way in a particular country but differently in another country (e.g., the words for cake andjam vary acrossdialects). When dialectical discrepanciesoccurred, the fourth coder was askedto determinewhich translationwas the most cornmonand Do BilingualsAccess 34 mostappropriate. However, it is quitepossible that there were words that the participantdid not know dueto dialecticaldifferences. Thesecond limitation of thepresent study is its useof outdatedand culturally specificnorming data to createthe associate lists. Theassociate lists were taken from Russelland Jenkin's norming study from 1954.Fifty-one years later, some of the associatesare out of date.For instance,one of theitems for mountainis molehill, comingfrom the idiomaticphrase, you are makinga mountainout of a molehill. This phraseis not nearlyas popular as it oncewas, and unless one knows the phrase, molehillis not goingto activatemauntain. It is alsolikely that in the pasthalf- centurythe f,rfteenmost corilnon associatesof a word havechanged. In additionto beingoutdated, Russell and Jenkin's norming data from 1954is culturallyspecific. A largenumber of the associateswere taken from language-specific idioms (e.9., molehill (mountain),molasses (slow), andhaystack (needle)). If a participantis not familiarwith the idiom,then it is lesslikely to activatethe CNW andcreate false memories. While there was strong false recognition of critical lures, the norms were problematic when attempting to createthe study lists. Not only were some lists dropped from the study, but also when creatingthe mixed-languagelists, I was occasionally forced to abandonthe random assignmentof study-list words to one languageor another. That is, words hke molehll/ were left in English becauseit was impossibleto translatethem into Spanish.Thus, there may havebeen some bias when compiling the study lists, and so it is important to correct this by creating new lists from more curTentnorming studies. Do BilingualsAccess 35 FutureDirections In futurestudies, it is necessaryto establisha new,updated set of associate lists thatwill resolvemany of the limitationsof the presentstudy. Ratherthan attemptingto translate50 yearold lists of associativenorTlrs into anotherlanguage, it is importantto createthe lists through norming studies in bothEnglish and Spanish. This avoidsmany of the pitfallsof attemptingto trzurslatea long list of very similar synonyrnsinto anotherlanguage, and also ensures that the list will be culturally relevant.The present study used lists filled with culturallyspecific idioms and severallists were removed because of translationdifficulties. Simply creating new listsfrom Englishand Spanish associative norrns will alleviatemuch of these difficultiesand should result in greaterfalse recognition of CNWs. In additionto new lists,a moreappropriate and valid rneasureof proficiency shouldbe explored.While the CBM wasa validmeasure of proficiency,it was assessingreading proficiency, and so wasnot appropriatefor the DRM falsememory task,which was heavily influenced by vocabularysize, In otherwords, a highscore on theCBM did not necessarilytranslate into betterperformance on theDRM becausethe two tasksrequired different t;pes of knowledge.Thus, in orderto accuratelyexamine the effect of languageproficiency on conceptualmediation and the creationof falsememories, it is importantto developa moreappropriate measure of proficiency. Finally,the paradigm should be amendedin someway suchthat language modeeffects can be examined.The present study attempted to exploresuch effects by havingmixed and single language study lists. However,this may not an effective Do BilingualsAccess 36 way to put participantsin eithera bilingualor monolinguallanguage mode. Perhaps eachparticipant could receive only onestudy-list cortdition and read several mixed or singlelanguage passages before the beginning of theexperiment. This might be a moreeffective way to artificiallycreate a languagemode and so allow for the explorationof languagemode effects on falserecognition rates. Overall,the findingsof thepresent experiment provide evidence for the interdependencehypothesis. False recognition of CNWswas much higher than the falserecognition of unstudied,non-target words across two languages,suggesting that therewas parallel activation of bothlanguages during encoding, These results suppofimodels of languagerepresentation such as the RHM (Kroll & Stewart,1994) anddistributed models. However, before drawing any furtherconclusions, it is importantto establishnew lists of associatesas well asa newproficiency measure. Do BilingualsAccess )t References Altarriba,J. (1992).The representation of translation equivalents in bilingual memory. In R. Harris(ed.), Cognitive Processing in Bilinguals,(pp. 157- 174).Amsterdam; North-Holland. Beauvillain,C. & Grainger,J. (1987),Accessing interlexical homographs: Some limitationsof a language-selectiveaccess. Journal of Memoryand Language, 26,658-672. Bloem,L, & La Heij, W. (2003) Semanticfacilitation and semantic interference in wordtranslation: Implications for modelsof lexicalaccess in language production.Journal of Memory& Language,48(3),468-488. Blot,K, I.,Zatate,M. A., & Paulus,P. B. (2003).Code-switching across brainstormingsessions: Implications for therevised hierarchical model of bilinguallanguage processing . Experimental Psycholo g,,, 5 0(3), I 7 I - I 83. Chen,H-C., (1992), Lexicalprocessing in bilingualor multilingualspeakers. In R. Harris(ed.), Cognitive Processing in Bilinguals,(pp. 253-264).Amsterdam: North-Holland. Chen,H-C., & Ho, C. (1986).Developmental study of thereversed Stroop effect in Chinese-Englishbilingtals, Journal of GeneralPsychologt, I I3(2), l2l-125. De Groot,A. (1992).Bilingual lexical representation. a closer look at conceptual representations.In R. Frost& L. Katz (eds.),Orthography, phonologt, morpholog,t,and meaning, (pp. 389-412). Amsterdam: Elsevier. De Groot,A. (1993),Word-Type effects in bilingualprocessing tasks: Support for a mixed-representationalsystem. In K. De Bot & T. Huebner(eds.), The Do Access 38 bilinguallexicon, (pp 27-51).Amsterdam: John Benjamins. De Groot,A., & Kroll, J, (eds.)(1997). Tutorialsin bilingualism:Psycholinguistic perspectives.Mahwah, NJ: LawrenceErlbaum. Dudsic,J. A. (1999),Priming asymmetries in Chinese-Englishbilinguals: A seriesof single-subjectstudies, Dissertation Abstracts International Section A: Humanit[es& SocialSciences, d1(l-A), 152. Dufour,R., & Kroll, J F. (1995).Matching words to conceptsin two languages:A testof the conceptmediation model of bilingualrepresentation. Memory & cognition,23(Z), 166-1 80. Durguno$Iu,A. Y. (1997).Bilingual reading: Its components,development, and otherissues. In A. De Groot& J. Kroll (eds), Ttttorialsin btlingualism, Psycholinguisticperspectives, (pp 255-276).Mahwah, NJ: Lawrence Erlbaum. Durgunofrlu,A. Y., & Roediger,H. L. (1987) Testdifferences in accessingbilingual memory.Journal of Memory&Language, 26(4),377-391. Fox,E. (1996),Cross-language priming from ignoredwords: Evidence for a cornmon representationalsystem in bilinguals.Journal of Memory& Language,J5(3), 353-370. Francis,W. S. (1999).Cognitive integration of languageand memory in bilinguals: Semanticrepresentation. Psychological bulletin, I 25 (2), 193-222. Gallo,D., McDermott,K., Percer,J,, & Roediger,H. (2001),Modality effects in alse recalland false recognition. Journal of ExperimentalPsychology: Learning, Memory,and Cognition,27(2), 339-353. Do BilingualsAccess Gerard,L. D., & Scarborough,D, L. (1989).Language specific lexical accessof homographsby bilinguals. Journal of Experimentnl Psycholog,,:Learning, Memory, and Cognitian, I 5(2), 305-3I 5. Gollan, T. H., & Kroll, J. F. (2001),Bilingual lexical access,In B. Rapp (Ed.), The handbookof cagnitive neuropsychalogt: ll/hat deficits reveal about the uman mind. (pp.32l-345). Philadelphia,PA: PsychologyPress Grainger,J., & Dijkstra, T. (1992). On the representationand use of language informationin bilinguals. In R.J.Hanis (Ed ), Cognitive processingin hilinguals,(pp. 207-252), Amsterdam:Elsevier. Grosjean,F. (1998). Studyingbilinguals: Methodological and conceptualissues. Bilingualism: Languageand Cognition,1, 13l-149. Hartsuiker,Pickering, & Veltkamp (2004). Is synta:rseparate or sharedbetween languages:Cross-linguistic syntactic priming in Spanish-Englishbilinguals. PsychologicalScience, I 5(6), 409-414. Hasbrouck,J. E, & Tindal, G. (Spring, 1992).Curriculum-based oral readingfluency nornu for studentsin grades2-5. TeachingExcepttonal Children, 24 (3),41- 44. Heredia,R. R. (1997).Bilingual memory and hierarchicalmodels: A casefor languagedominance . CurrentDirections in PsychologicalScience, 6,34-39. Kawasaki-Miyaji,Y., Inoue,T., & Yama, H. (2003). Cross-linguisticfalse recognition:How do Japanese-dominantbilinguals process two languages: Japaneseand English? Psychologia, 46, 255-267, Kim, K, H. S., Relkin, N. R,, Lee, K., & Hirsch,J. (1997).Distinct corticalareas Do BilingualsAccess 40 associatedwith nativeand second languages. Nature, -f88(6638), 17l-174. Kotz, S.A. (?001).Neurolinguistic evidence for bilinguallanguage representation: A comparisonof reactiontimes and event-related brain potentials. Bilingualism: Language& Cognition.Special Issue: The cognitive neuroscience af bilingualism,4(2), | 43- | 54. Kroll, J. F,, (1993),Accessing conceptual representations for wordsin a second language.In K. De Bot & T, Huebner(eds.), The bilingual lexicon, (pp 53- 8l). Amsterdam:John Benjamins. Kroll, J. F., & De Groot,A. (1997).Lexical and conceptual memory in thebilingual: Mappingform to meaningin two languages.In A. De Groot& J. Kroll (eds.), Tutorialsin hilingualism.Psycholinguistic perspectives, (pp 169-199). Mahwah,NJ: LawrenceErlbaum. Kroll, J.F.& Sholl,A, (1992),Lexical and conceptual memory in fluentand nonfluentbilinguals. In R. Harris(ed.), Cognitive Processing in Bilinguals, (pp. 191-204).Amsterdam: North-Holland. Kroll, J. F., & Stewaft,E. (1994).Category interference in translationand picture naming:Evidence for asymmetricconnection between bilingual memory representations.Journal of Memory& Language,33(2), 149-174. La Heij, W., Hooglander,A., Kerling,R,, & vander Velden, E. (1996).Nonverbal contexteffects in forwardand backward word translation:Evidence for conceptmediation. Journal of Memory& Language,-t-t(5), 648-665. Luna,D., & Peracchio,L. A. (2002)."Where there is awill,..":Motivatiou as a moderatorof languageprocessing by bilingualconsumers. Psycholog,, & Do BilingualsAccess 41 Marketing. special Issue: Psychology,marketing, &psycholinguistics, I9(7- 8), 573-593. Macleod, C.M. (1991). Half a centuryof researchon the Stroopeffect: An integrative review. PsychologiculBulletin, I 09, 163-203. Maegiste,E. (1985).Development of intra- and interlingualinterference in bilinguals. Journal of psycho l inguistic research, I 4(2), 137-1 5 4, Maegiste,E. (1984). Strooptasks and dichotictrmslation: The developmentof interferencepatterns in bilinguals. Journal of Experimental Psychologt; Learning, Memory, & Cognition, I 0(2), 304-3I 5. Marian, V., & Spivey,M. (2003).Bilingual and monolingualprocessing of competinglexical items.Applied Psycholin guis tics, 2 4(Z), 173 - lg3. Marian, V., Spivey,M., & Hirsch, J. (2003).Shared and separatesystems in bilingual languageprocessing: Converging evidenie from eyetracking and brain imaging,Brain &Language, 8d(l), 70-82. Marston,D. (1989),A curriculum-basedmeasurement approach to assessing academicperformance: What it is and why do it. In M. R. Shinn (Ed ), Curriculum-basedmeasurement: Assessing special chiklren(pp. f 8-78).New York: Guilford. Nas, G. (1983). Visual word recognitionin bilinguals:Evidence for cooperation betweenvisual and soundbased codes during accessto a commonlexical store.Journal ofVerbal Learntng and VerbalBehavior, ZZ, 5ZE-534. Paradis,M. (1997). The cognitiveneuropsychology of bilingualism. In A. De Groot & J. Ifuoll (eds.), Tutorlals in bilingualism. Psycholinguist[cperspectives, (pp Do BilingualsAccess 42 331-354). Mahwah,NJ: LawrenceBrlbaum. Pavlenko,A. (1999). New approachesto conceptsin bilingual memory. Bilingualism:Language and Cognition,2 (3),209-230. Potter,M. C., So, K., von Eckardt,B.,& Feldman,L, B, (1984).Lexical and conceptualrepresentation in beginningand proficient bilinguals.Journal of VerbalLearning & VerbalBehavior, ?-l(l), 23-38. Roediger,H., & McDermott,K (1995). Creatingfalse memories: Remembering words not presentedin lists. Journal of Experimental Psychologt: Learning, Memory,and Cognition,2 I (4), 803-814. Smith, M, (1997). How do bilingualsaccess lexical information? In A. De Groot & J. Kroll (eds,), Tutorials in bilingualism. Psycholinguisticperspectives, (pp 145-168).Mahwah, NJ: LawrenceErlbaum. Smith, E. E., Shoben,E. J., & Rips, L. I. (1g74).Sfucture and processin semantic memory:A featuralmodel for semanticdecisions. Psychological review, 8/(3),214-24t. Spivey,M. J., & Marian, V. (1999).Cross talk betweennative and secondlanguages: Partialactivation of an irrelevantlexicon. Psychological Science, 10,281- 284. Van Hell, J. G,, & De Groot, A. M. B. (1998).Conceptual representation in bilingual memory: Effects of concretenessand cognatestatus in word association. B ilingualtsm: Language & Cognition, I (3), 193-211 . Weinreich,U. (1953). Languagesin contact:Findings and problems, New York: Linguistic Circle of New York. Reprintedin 1964by Mouton, The Hague. Do BilingualsAccess 43 Acknowledgements I wouldlike to thankthe Macalester College Psychology Depzutment. without theirsuppoft, financial and otherwise, I would not havebeen able to completethis project. I wouldalso like to thankmy thesisadvisors, Janet Oh and BrookeLea, for their continuedsupport throughout the formationand completion of this project I thankSarah Dart for her role in my honorscommittee. I thankMarissa Weyerfor her commentsand suggestions on an earlierdraft of this manuscript. Finally,I thankthe numerous translators, Alejandro Carrion Mendndez, Maria Cielo Parodi,and Jacquelyn Durin, aswell asthe MacalesterCollege Spanish Department for their help creatingthe stimuli, Do BilingualsAccess 45 FigureCaptions FieureL TheWord Association model (adapted from Potteret al,, 1984), Fieure2. TheConcept Mediation model (adapted from Potteret al., 1984). Fieure3, TheRevised Hierarchical Model (adapted from Kroll & Stewafi,1994). Fieure4. Thedistributed conceptual feature model (adapted from De Groot,1992). Fisure 5. The distributedconceptuaVlexical feature model (adapted from Kroll & De Groot, 1991). Fieure6. CBM scoresby self-reportedproficiency. Fieure7. Meanrecognition rates for unstudiedwords, studied words, and critical lures. Fieure8. Meanrecognition rate by wordtype and study list condition. Fieure9. Languagecongruency effects for studiedworfu. 0.600 @ @ q o H o.loo tr tr o '4 6-0.300 o o tr E r! o E o.zoo 0.100 0.000 WordType I studied f,l critical o fit E C o :|i 0.40 tr Et o (J q E o.so tr ct o E $panish only StudyList Conditlon E o.so trfit E .E5 0.40 tr) o oC) E o.so g fit o E $ame Language DifferentLanguage AppendixA. EnglishOnly Tareet word King Black Foot Needle River Soft queen white shoe thread water hard England dark hand pin stfeam light crown cat toe eye lake pillow prince charred kick sewing Mississippiplush George night sandals shary boat loud dictator funeral soccer point tide cotton palace color yard prick swim fur Studied throne grief walk thimble flow touch worfu chess blue ankle haystack run fluffy rule death arm thorn barge feather subjects ink boot hurl creek furry monarch bottom inch injection brook downy royal coal sock syringe fish kitten Ieader brown smell cloth bridge skin reign gray mouth knitting winding tender AppendixA (continued) SpanishOnly (rranslation) Cold Sweet Spider caliente hot 6cido 8our telarafla web nieve snow caramelo candy insecto insect tibio wdrrn azfcar sugar bicho bug invierno winter amargo bitter susto fright hielo ice bueno good mosco flY mojado wFt sabor taste ar6cnido arachnid frigido frigid diente tooth arrastrarse crawl Studied fresco chilly agradable nice tarantula tarantula worfu calor heat miel honey veneno poison clima weather soda soda morder bite congelar freeze chocolate chocolate escalofriante creepy aire air corazon heart animal animal temblor shiver pastel cake feo uglY Artico Arctic agrio tart pequeflo small *t"feelers" escarcha frost tafta pie removed Appendix A (continued) SpanishOnly (translat ion) Target word Music Fruit Doctor nota flote manzana apple enfermera nurse sonido sound vegetal vegetable enfermo sick piano piano naranja orange abogado Iattlter cantar sing kiwi kiwl medicina medicine radio radio citrico citrus salud health banda band maduro ripe hospital hospital melodia melody pera pear dentista dentist Studied trompeta horn platano banana remedio remedy words concierto concert baya berry indispuesto ill instrumento f.nstrument cereza cherry paciente patient sinfonia symphony cesta basket officina ,ffirc jm= jazz jugo juice estetoscopio stethoscope orquesta orchestra ensalada salad cirujano surgeon arte drt bol bowl clinica clinic ritmo rhythm c6ctel cocktail curaci6n cure