20
This article was downloaded by: [Agnieszka Lijewska] On: 25 September 2014, At: 11:45 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK International Journal of Multilingualism Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rmjm20 Cognate facilitation in sentence context – translation production by interpreting trainees and non- interpreting trilinguals Agnieszka Lijewska a & Agnieszka Chmiel b a Department of Psycholinguistic Studies, Faculty of English, Adam Mickiewicz University, Poznan, Poland b Department of Translation Studies, Faculty of English, Adam Mickiewicz University, Poznan, Poland Published online: 23 Sep 2014. To cite this article: Agnieszka Lijewska & Agnieszka Chmiel (2014): Cognate facilitation in sentence context – translation production by interpreting trainees and non-interpreting trilinguals, International Journal of Multilingualism, DOI: 10.1080/14790718.2014.959961 To link to this article: http://dx.doi.org/10.1080/14790718.2014.959961 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &

Cognate facilitation in sentence context – translation production by interpreting trainees and non-interpreting trilinguals

  • Upload
    amu

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

This article was downloaded by [Agnieszka Lijewska]On 25 September 2014 At 1145Publisher RoutledgeInforma Ltd Registered in England and Wales Registered Number 1072954 Registeredoffice Mortimer House 37-41 Mortimer Street London W1T 3JH UK

International Journal of MultilingualismPublication details including instructions for authors andsubscription informationhttpwwwtandfonlinecomloirmjm20

Cognate facilitation in sentencecontext ndash translation productionby interpreting trainees and non-interpreting trilingualsAgnieszka Lijewskaa amp Agnieszka Chmielba Department of Psycholinguistic Studies Faculty of English AdamMickiewicz University Poznan Polandb Department of Translation Studies Faculty of English AdamMickiewicz University Poznan PolandPublished online 23 Sep 2014

To cite this article Agnieszka Lijewska amp Agnieszka Chmiel (2014) Cognate facilitation insentence context ndash translation production by interpreting trainees and non-interpreting trilingualsInternational Journal of Multilingualism DOI 101080147907182014959961

To link to this article httpdxdoiorg101080147907182014959961

PLEASE SCROLL DOWN FOR ARTICLE

Taylor amp Francis makes every effort to ensure the accuracy of all the information (theldquoContentrdquo) contained in the publications on our platform However Taylor amp Francisour agents and our licensors make no representations or warranties whatsoever as tothe accuracy completeness or suitability for any purpose of the Content Any opinionsand views expressed in this publication are the opinions and views of the authorsand are not the views of or endorsed by Taylor amp Francis The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information Taylor and Francis shall not be liable for any losses actions claimsproceedings demands costs expenses damages and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with in relation to or arisingout of the use of the Content

This article may be used for research teaching and private study purposes Anysubstantial or systematic reproduction redistribution reselling loan sub-licensingsystematic supply or distribution in any form to anyone is expressly forbidden Terms amp

Conditions of access and use can be found at httpwwwtandfonlinecompageterms-and-conditions

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Cognate facilitation in sentence context ndash translation productionby interpreting trainees and non-interpreting trilinguals

Agnieszka Lijewskaa and Agnieszka Chmielb

aDepartment of Psycholinguistic Studies Faculty of English Adam Mickiewicz University PoznanPoland bDepartment of Translation Studies Faculty of English Adam Mickiewicz UniversityPoznan Poland

(Received 16 July 2013 accepted 18 August 2014)

Conference interpreters form a special case of language users because the simultan-eous interpretation practice requires very specific lexical processing Word compre-hension and production in respective languages is performed under strict timeconstraints and requires constant activation of the involved languages The presentexperiment aimed at shedding more light on the effects of conference interpretingtraining on word production in the process of translation in sentence context In thestudy we tested trilingual interpreting trainees and matched non-interpreting trilin-guals Both groups were required to verbally produce L1 and L2 translation equiva-lents in response to L3 target words (L2ndashL3 cognates and non-cognates) presented intwo context constraints ndash high sentence context constraint and low sentence contextconstraint This enabled us to investigate whether expertise developed duringinterpreting training in translation between L1 and L2 (extensively practised) istransferable to translation from L3 to L2 (not practised at all) The study providedevidence for cognate facilitation and context effects on naming latencies However wefound no significant evidence of enhanced semantic processing of interpreting traineescompared to trilingual non-interpreters

Keywords cognate facilitation sentence context mental lexicon trilinguals inter-preting multilingualism

Introduction

One of the central questions in psycholinguistic experimentation with bilingual speakersand even more so with multilingual ones pertains to the nature of lexical processing intheir mental lexicons For years now researchers have been trying to resolve an ongoingdebate between two opposing hypotheses ie language selective versus language non-selective access to the bilingualmultilingual mental lexicon Until now there has beenabundant evidence in favour of the language non-selective access hypothesis comingfrom a number of empirical studies conducted within various paradigms eg in picturenaming (eg Costa Santesteban amp Cantildeo 2005 Hoshino amp Kroll 2008) in eye-tracking(eg Bartolotti amp Marian 2012 Ju amp Luce 2004) in word recognition and electro-physiological experimentation (eg Hoshino amp Thierry 2012 Peeters Dijkstra ampGrainger 2013) Studies on cognate processing and word processing in sentence contextcan offer further insight into the nature of multilingual mental lexicons

Corresponding author Email alijewskawaamuedupl

International Journal of Multilingualism 2014httpdxdoiorg101080147907182014959961

copy 2014 Taylor amp Francis

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The study presented in this article aims to shed more light on the nature of cognateprocessing in sentence context by trilinguals We have included conference interpretingtrainees in the study to reveal any effect of interpreting training on their lexicalprocessing Below we first present an overview of studies focusing on the processing ofcognates and sentence context effects on lexical processing both in bilinguals andmultilinguals Their results are reviewed in the context of two models of bilinguallexicon the Revised Hierarchical Model (Kroll amp Stewart 1994) and the BIA+ model(Dijkstra amp van Heuven 2002) Later we focus on the studies of lexical processing byinterpreters and group differences between professional interpreters trainees and non-interpreting controls The actual experiment involving translation of L3ndashL2 cognates andnon-cognates from L3 into L2 and L1 in manipulated context conditions brings interestingdata discussed in relation to the two above mentioned models

Processing of cognates

One of the key findings interpreted in favour of the language non-selective access is thecognate facilitation effect Cognates are words which share both form and meaning acrosslanguages (eg the word BUTTER is a cognate for English and German) The typicalfinding in psycholinguistic experiments investigating reaction times (RTs) to cognates isthat they take shorter time to respond to in comparison to language-specific controlwords This discrepancy between RTs obtained for cognates and non-cognates is taken toindicate that during lexical access lexicons of all languages known to an individual areactivated The cognate facilitation effect has been found in a number of studies carriedout with bilingual speakers (Antoacuten-Meacutendez amp Gollan 2010 Costa Caramazza ampSebastian-Galles 2000 Davis et al 2010 De Groot Dannenburg amp Van Hell 1994Dijkstra Grainger amp van Heuven 1999 Dijkstra Miwa Brummelhuis Sappelli ampBaayen 2010 Dijkstra Timmermans amp Schriefers 2000 Gollan Forster amp Frost 1997Lemhoefer Dijkstra amp Michel 2004 Lotto amp De Groot 1998 Saacutechez-Casas Garciacutea-Albea amp Davis 1992 Van Hell amp De Groot 1998 Van Hell amp Dijkstra 2002)

Experiments testing cognate facilitation in trilinguals are scarce (Dijkstra amp Van Hell2003 Lemhoefer et al 2004 Poarch amp Van Hell 2012 Szubko-Sitarek 2011Tymczyńska 2011 Van Hell amp Dijkstra 2002) A study by Van Hell and Dijkstra(2002) involved three experiments and two groups of participants The first groupcomprised Dutch-English-French trilinguals with a relatively low L3 proficiency and thesecond group consisted of Dutch-English-French trilinguals whose fluency in French washigher than in the first group and comparable to that in English Both groups performedlexical decision tasks (LDTs when participants need to decide if a given string of lettersis a word or not) on Dutch words and additionally the first group performed a wordassociation test (ie they were asked to produce single word associations in response toL1 words presented on the computer screen) Materials used for the experiments wereL1 words which were either Dutch-English cognates or Dutch-French cognates or non-cognates All tasks were run in an exclusively L1 context ie participants had not beeninformed of the multilingual nature of stimuli used in the experiments and no explicitreference to the participantsrsquo foreign languages was made in the course of theexperimentation Van Hell and Dijkstra observed a significant cognate facilitation effectfor Dutch-English cognates in both groups The cognate facilitation effect for Dutch-French cognates was recorded only in the second group which had a higher fluency inFrench According to Van Hell and Dijkstra (2002) these results lend further support tolanguage non-selective access hypothesis They claim that the fact that significant cognate

2 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

facilitation was observed in an exclusively monolingual context seems to contradict thestrict version of the Language Mode Hypothesis (Grosjean 1998 2001) which positsthat if bilingualsmultilinguals are in a monolingual environment they are able todeactivate their non-target languages (Van Hell amp Dijkstra 2002 p 787)

Similar conclusions were drawn from another study conducted with trilingualspeakers Lemhoefer et al (2004) tested a group of Dutch-English-German trilingualswho performed a LDT in their L3 The participants were asked to decide if the presentedstring of letters formed a German word or not The critical stimuli used in the study werecognates overlapping in Dutch and German with no cognate relation to English as wellas cognates overlapping across the three languages RTs recorded for Dutch-Germancognates were significantly shorter than for control German words and significantlylonger than RTs for Dutch-English-German ones This cumulative cognate facilitationeffect is even more interesting in view of the fact that the cognate facilitation effectobserved for Dutch-English-German cognates was not influenced by whether the partici-pants read an English text and performed an item recognition test before the LDT task ornot According to Lemhoefer et al the lack of context effect runs counter to thepredictions of the Language Mode Hypothesis put forward by Grosjean (1998 2001)which assumes that the context of experimental situation should affect activation levelsof languages in the participantsrsquo minds All in all Lemhoefer et al (2004) interpretedthe results of their experiments in favour of the language non-selective access intrilingual processing

The cognate facilitation effect has been accounted for by a number of models ofbilingual lexical processing the most influential being the Revised Hierarchical Model(RHM) (Kroll amp Stewart 1994) and BIA+ (Bilingual Interactive Activation Plus) model(Dijkstra amp van Heuven 2002) Even though these two models have been extended tomultilingualism (for details see Tymczyńska (2012) and Dijkstra (2003) respectively)accounts of cognate facilitation within the models will be presented within the bilingualcontext (ie in the form they were originally put forward)

The cognate facilitation effect has been accommodated by RHM for L2 processing inthe following manner Due to direct L1ndashL2 lexical links the meaning of an L2 word isaccessed more quickly when there is an overlap in lexical representations (word forms) ofL1ndashL2 translation equivalents (as in cognates) as compared to the lack of such an overlap(as in non-cognates) According to Dijkstra et al (2010) cognates differ from non-cognates within the BIA+ model in that the former share more orthographic semanticandor phonological features with their translation equivalents than the latter The cognatefacilitation effect is accommodated by the model in the following manner (Dijkstra et al2010 p 286)

the cognate facilitation effect in reading might in fact be an orthographicndashsemantic primingeffect overlapping orthographic and semantic representations of both languages becomeactive upon the presentation of one of the readings of the cognate leading to a facilitatedrecognition of cognates relative to non-cognates

Consequently the cognate facilitation effect is sensitive to frequency as well as cross-linguistic similarity (both orthographic and phonological) of translation equivalentsBIA+ was supported in cognate studies conducted by eg Dijkstra et al (2010) and Vogaand Grainger (2007) Even though Dijkstra et al (2010) explained the mechanics ofcognate facilitation within BIA+ for bilinguals there is every reason to assume that asimilar pattern of processing will also work for multilinguals within the Multilingual

International Journal of Multilingualism 3

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Interactive Activation model (MIA) which is the multilingual extension of BIA(Dijkstra 2003)

Sentence context effects

It needs to be stressed here that the vast majority of the bilingualmultilingual researchtested cognates and non-cognates by presenting them as single words Only in a relativelysmall number of studies did the researchers investigate cognates in a sentence contextGiven the fact that in everyday ie non-laboratory language use speakers extremelyrarely use isolated words investigating lexical processing in sentence context seems tohave more ecological validity than experimentation with single words

Nevertheless studies testing cognates in sentence context are still very limited innumber and differ in terms of the chosen experimental paradigms eg word naming(Schwartz amp Kroll 2006) LDT (Duyck Van Assche Drieghe amp Hartsuiker 2007 VanHell amp De Groot 2008) translation (Van Hell amp De Groot 2008 Van Hell 2005) eye-tracking (Duyck et al 2007 Libben amp Titone 2009 Van Assche Drieghe Duyck Wel-vaert amp Hartsuiker 2011) The results recorded in these studies differ across paradigmsand do not always tally with findings from single word experiments

Results obtained from word naming suggest that the processing of cognates isaffected by context only if this context is not semantically constraining (ie low contexthenceforth LC) as in lsquoThe final word in this sentence ishelliprsquo where the sentence finaltarget word cannot be easily predicted As opposed to lsquoShe took a bite of the freshgreenhelliprsquo which is a semantically constraining context (ie high context henceforth HC)biased towards the final target word (APPLE) Schwartz and Kroll (2006) tested twogroups of Spanish-English bilinguals in word naming in L2 The materials includedidentical and non-identical cognates as well as Spanish-English interlingual homographsThe words were embedded in HC and LC sentences The sentences were presented usingthe Rapid Serial Visual Presentation (RSVP where words appear on the screen one-by-one) and the participants were asked to read out the target word RTs recorded in thestudy indicated a robust cognate facilitation effect in the LC condition however in theHC condition the effect was eliminated Schwartz and Kroll (2006) interpreted this findingin favour of the language non-selective access which is restricted when enough semanticinformation is available to suppress the non-target language representations

A similar pattern of results was obtained in an LDT In one of the experimentsreported on by Van Hell and De Groot (2008) a group of Dutch-English bilingualsperformed an L2 LDT where they were asked to decide if a letter string formed anEnglish word or not Target words were either (concrete or abstract) cognates or non-cognates embedded in HC LC sentences or presented in isolation (no context conditionNC) Significant cognate effect was observed only in the LC and NC conditions both forconcrete and abstract target words Van Hell and De Groot (2008) found no significantthree-way interaction between concreteness cognatersquos status and context This mightindicate that subtle meaning overlap variations between cognates and non-cognates donot influence the pattern of results

In the same paper Van Hell and De Groot (2008) reported on three experimentsinvolving L1ndashL2 (forward) and L2ndashL1 (backwards) translation Materials and participantswere similar to those in the LDT discussed above Unlike in word naming and LDTcognate facilitation effect was observed across all contexts and translation directions butin the HC condition the facilitation was heavily reduced

4 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Conditions of access and use can be found at httpwwwtandfonlinecompageterms-and-conditions

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Cognate facilitation in sentence context ndash translation productionby interpreting trainees and non-interpreting trilinguals

Agnieszka Lijewskaa and Agnieszka Chmielb

aDepartment of Psycholinguistic Studies Faculty of English Adam Mickiewicz University PoznanPoland bDepartment of Translation Studies Faculty of English Adam Mickiewicz UniversityPoznan Poland

(Received 16 July 2013 accepted 18 August 2014)

Conference interpreters form a special case of language users because the simultan-eous interpretation practice requires very specific lexical processing Word compre-hension and production in respective languages is performed under strict timeconstraints and requires constant activation of the involved languages The presentexperiment aimed at shedding more light on the effects of conference interpretingtraining on word production in the process of translation in sentence context In thestudy we tested trilingual interpreting trainees and matched non-interpreting trilin-guals Both groups were required to verbally produce L1 and L2 translation equiva-lents in response to L3 target words (L2ndashL3 cognates and non-cognates) presented intwo context constraints ndash high sentence context constraint and low sentence contextconstraint This enabled us to investigate whether expertise developed duringinterpreting training in translation between L1 and L2 (extensively practised) istransferable to translation from L3 to L2 (not practised at all) The study providedevidence for cognate facilitation and context effects on naming latencies However wefound no significant evidence of enhanced semantic processing of interpreting traineescompared to trilingual non-interpreters

Keywords cognate facilitation sentence context mental lexicon trilinguals inter-preting multilingualism

Introduction

One of the central questions in psycholinguistic experimentation with bilingual speakersand even more so with multilingual ones pertains to the nature of lexical processing intheir mental lexicons For years now researchers have been trying to resolve an ongoingdebate between two opposing hypotheses ie language selective versus language non-selective access to the bilingualmultilingual mental lexicon Until now there has beenabundant evidence in favour of the language non-selective access hypothesis comingfrom a number of empirical studies conducted within various paradigms eg in picturenaming (eg Costa Santesteban amp Cantildeo 2005 Hoshino amp Kroll 2008) in eye-tracking(eg Bartolotti amp Marian 2012 Ju amp Luce 2004) in word recognition and electro-physiological experimentation (eg Hoshino amp Thierry 2012 Peeters Dijkstra ampGrainger 2013) Studies on cognate processing and word processing in sentence contextcan offer further insight into the nature of multilingual mental lexicons

Corresponding author Email alijewskawaamuedupl

International Journal of Multilingualism 2014httpdxdoiorg101080147907182014959961

copy 2014 Taylor amp Francis

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The study presented in this article aims to shed more light on the nature of cognateprocessing in sentence context by trilinguals We have included conference interpretingtrainees in the study to reveal any effect of interpreting training on their lexicalprocessing Below we first present an overview of studies focusing on the processing ofcognates and sentence context effects on lexical processing both in bilinguals andmultilinguals Their results are reviewed in the context of two models of bilinguallexicon the Revised Hierarchical Model (Kroll amp Stewart 1994) and the BIA+ model(Dijkstra amp van Heuven 2002) Later we focus on the studies of lexical processing byinterpreters and group differences between professional interpreters trainees and non-interpreting controls The actual experiment involving translation of L3ndashL2 cognates andnon-cognates from L3 into L2 and L1 in manipulated context conditions brings interestingdata discussed in relation to the two above mentioned models

Processing of cognates

One of the key findings interpreted in favour of the language non-selective access is thecognate facilitation effect Cognates are words which share both form and meaning acrosslanguages (eg the word BUTTER is a cognate for English and German) The typicalfinding in psycholinguistic experiments investigating reaction times (RTs) to cognates isthat they take shorter time to respond to in comparison to language-specific controlwords This discrepancy between RTs obtained for cognates and non-cognates is taken toindicate that during lexical access lexicons of all languages known to an individual areactivated The cognate facilitation effect has been found in a number of studies carriedout with bilingual speakers (Antoacuten-Meacutendez amp Gollan 2010 Costa Caramazza ampSebastian-Galles 2000 Davis et al 2010 De Groot Dannenburg amp Van Hell 1994Dijkstra Grainger amp van Heuven 1999 Dijkstra Miwa Brummelhuis Sappelli ampBaayen 2010 Dijkstra Timmermans amp Schriefers 2000 Gollan Forster amp Frost 1997Lemhoefer Dijkstra amp Michel 2004 Lotto amp De Groot 1998 Saacutechez-Casas Garciacutea-Albea amp Davis 1992 Van Hell amp De Groot 1998 Van Hell amp Dijkstra 2002)

Experiments testing cognate facilitation in trilinguals are scarce (Dijkstra amp Van Hell2003 Lemhoefer et al 2004 Poarch amp Van Hell 2012 Szubko-Sitarek 2011Tymczyńska 2011 Van Hell amp Dijkstra 2002) A study by Van Hell and Dijkstra(2002) involved three experiments and two groups of participants The first groupcomprised Dutch-English-French trilinguals with a relatively low L3 proficiency and thesecond group consisted of Dutch-English-French trilinguals whose fluency in French washigher than in the first group and comparable to that in English Both groups performedlexical decision tasks (LDTs when participants need to decide if a given string of lettersis a word or not) on Dutch words and additionally the first group performed a wordassociation test (ie they were asked to produce single word associations in response toL1 words presented on the computer screen) Materials used for the experiments wereL1 words which were either Dutch-English cognates or Dutch-French cognates or non-cognates All tasks were run in an exclusively L1 context ie participants had not beeninformed of the multilingual nature of stimuli used in the experiments and no explicitreference to the participantsrsquo foreign languages was made in the course of theexperimentation Van Hell and Dijkstra observed a significant cognate facilitation effectfor Dutch-English cognates in both groups The cognate facilitation effect for Dutch-French cognates was recorded only in the second group which had a higher fluency inFrench According to Van Hell and Dijkstra (2002) these results lend further support tolanguage non-selective access hypothesis They claim that the fact that significant cognate

2 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

facilitation was observed in an exclusively monolingual context seems to contradict thestrict version of the Language Mode Hypothesis (Grosjean 1998 2001) which positsthat if bilingualsmultilinguals are in a monolingual environment they are able todeactivate their non-target languages (Van Hell amp Dijkstra 2002 p 787)

Similar conclusions were drawn from another study conducted with trilingualspeakers Lemhoefer et al (2004) tested a group of Dutch-English-German trilingualswho performed a LDT in their L3 The participants were asked to decide if the presentedstring of letters formed a German word or not The critical stimuli used in the study werecognates overlapping in Dutch and German with no cognate relation to English as wellas cognates overlapping across the three languages RTs recorded for Dutch-Germancognates were significantly shorter than for control German words and significantlylonger than RTs for Dutch-English-German ones This cumulative cognate facilitationeffect is even more interesting in view of the fact that the cognate facilitation effectobserved for Dutch-English-German cognates was not influenced by whether the partici-pants read an English text and performed an item recognition test before the LDT task ornot According to Lemhoefer et al the lack of context effect runs counter to thepredictions of the Language Mode Hypothesis put forward by Grosjean (1998 2001)which assumes that the context of experimental situation should affect activation levelsof languages in the participantsrsquo minds All in all Lemhoefer et al (2004) interpretedthe results of their experiments in favour of the language non-selective access intrilingual processing

The cognate facilitation effect has been accounted for by a number of models ofbilingual lexical processing the most influential being the Revised Hierarchical Model(RHM) (Kroll amp Stewart 1994) and BIA+ (Bilingual Interactive Activation Plus) model(Dijkstra amp van Heuven 2002) Even though these two models have been extended tomultilingualism (for details see Tymczyńska (2012) and Dijkstra (2003) respectively)accounts of cognate facilitation within the models will be presented within the bilingualcontext (ie in the form they were originally put forward)

The cognate facilitation effect has been accommodated by RHM for L2 processing inthe following manner Due to direct L1ndashL2 lexical links the meaning of an L2 word isaccessed more quickly when there is an overlap in lexical representations (word forms) ofL1ndashL2 translation equivalents (as in cognates) as compared to the lack of such an overlap(as in non-cognates) According to Dijkstra et al (2010) cognates differ from non-cognates within the BIA+ model in that the former share more orthographic semanticandor phonological features with their translation equivalents than the latter The cognatefacilitation effect is accommodated by the model in the following manner (Dijkstra et al2010 p 286)

the cognate facilitation effect in reading might in fact be an orthographicndashsemantic primingeffect overlapping orthographic and semantic representations of both languages becomeactive upon the presentation of one of the readings of the cognate leading to a facilitatedrecognition of cognates relative to non-cognates

Consequently the cognate facilitation effect is sensitive to frequency as well as cross-linguistic similarity (both orthographic and phonological) of translation equivalentsBIA+ was supported in cognate studies conducted by eg Dijkstra et al (2010) and Vogaand Grainger (2007) Even though Dijkstra et al (2010) explained the mechanics ofcognate facilitation within BIA+ for bilinguals there is every reason to assume that asimilar pattern of processing will also work for multilinguals within the Multilingual

International Journal of Multilingualism 3

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Interactive Activation model (MIA) which is the multilingual extension of BIA(Dijkstra 2003)

Sentence context effects

It needs to be stressed here that the vast majority of the bilingualmultilingual researchtested cognates and non-cognates by presenting them as single words Only in a relativelysmall number of studies did the researchers investigate cognates in a sentence contextGiven the fact that in everyday ie non-laboratory language use speakers extremelyrarely use isolated words investigating lexical processing in sentence context seems tohave more ecological validity than experimentation with single words

Nevertheless studies testing cognates in sentence context are still very limited innumber and differ in terms of the chosen experimental paradigms eg word naming(Schwartz amp Kroll 2006) LDT (Duyck Van Assche Drieghe amp Hartsuiker 2007 VanHell amp De Groot 2008) translation (Van Hell amp De Groot 2008 Van Hell 2005) eye-tracking (Duyck et al 2007 Libben amp Titone 2009 Van Assche Drieghe Duyck Wel-vaert amp Hartsuiker 2011) The results recorded in these studies differ across paradigmsand do not always tally with findings from single word experiments

Results obtained from word naming suggest that the processing of cognates isaffected by context only if this context is not semantically constraining (ie low contexthenceforth LC) as in lsquoThe final word in this sentence ishelliprsquo where the sentence finaltarget word cannot be easily predicted As opposed to lsquoShe took a bite of the freshgreenhelliprsquo which is a semantically constraining context (ie high context henceforth HC)biased towards the final target word (APPLE) Schwartz and Kroll (2006) tested twogroups of Spanish-English bilinguals in word naming in L2 The materials includedidentical and non-identical cognates as well as Spanish-English interlingual homographsThe words were embedded in HC and LC sentences The sentences were presented usingthe Rapid Serial Visual Presentation (RSVP where words appear on the screen one-by-one) and the participants were asked to read out the target word RTs recorded in thestudy indicated a robust cognate facilitation effect in the LC condition however in theHC condition the effect was eliminated Schwartz and Kroll (2006) interpreted this findingin favour of the language non-selective access which is restricted when enough semanticinformation is available to suppress the non-target language representations

A similar pattern of results was obtained in an LDT In one of the experimentsreported on by Van Hell and De Groot (2008) a group of Dutch-English bilingualsperformed an L2 LDT where they were asked to decide if a letter string formed anEnglish word or not Target words were either (concrete or abstract) cognates or non-cognates embedded in HC LC sentences or presented in isolation (no context conditionNC) Significant cognate effect was observed only in the LC and NC conditions both forconcrete and abstract target words Van Hell and De Groot (2008) found no significantthree-way interaction between concreteness cognatersquos status and context This mightindicate that subtle meaning overlap variations between cognates and non-cognates donot influence the pattern of results

In the same paper Van Hell and De Groot (2008) reported on three experimentsinvolving L1ndashL2 (forward) and L2ndashL1 (backwards) translation Materials and participantswere similar to those in the LDT discussed above Unlike in word naming and LDTcognate facilitation effect was observed across all contexts and translation directions butin the HC condition the facilitation was heavily reduced

4 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Cognate facilitation in sentence context ndash translation productionby interpreting trainees and non-interpreting trilinguals

Agnieszka Lijewskaa and Agnieszka Chmielb

aDepartment of Psycholinguistic Studies Faculty of English Adam Mickiewicz University PoznanPoland bDepartment of Translation Studies Faculty of English Adam Mickiewicz UniversityPoznan Poland

(Received 16 July 2013 accepted 18 August 2014)

Conference interpreters form a special case of language users because the simultan-eous interpretation practice requires very specific lexical processing Word compre-hension and production in respective languages is performed under strict timeconstraints and requires constant activation of the involved languages The presentexperiment aimed at shedding more light on the effects of conference interpretingtraining on word production in the process of translation in sentence context In thestudy we tested trilingual interpreting trainees and matched non-interpreting trilin-guals Both groups were required to verbally produce L1 and L2 translation equiva-lents in response to L3 target words (L2ndashL3 cognates and non-cognates) presented intwo context constraints ndash high sentence context constraint and low sentence contextconstraint This enabled us to investigate whether expertise developed duringinterpreting training in translation between L1 and L2 (extensively practised) istransferable to translation from L3 to L2 (not practised at all) The study providedevidence for cognate facilitation and context effects on naming latencies However wefound no significant evidence of enhanced semantic processing of interpreting traineescompared to trilingual non-interpreters

Keywords cognate facilitation sentence context mental lexicon trilinguals inter-preting multilingualism

Introduction

One of the central questions in psycholinguistic experimentation with bilingual speakersand even more so with multilingual ones pertains to the nature of lexical processing intheir mental lexicons For years now researchers have been trying to resolve an ongoingdebate between two opposing hypotheses ie language selective versus language non-selective access to the bilingualmultilingual mental lexicon Until now there has beenabundant evidence in favour of the language non-selective access hypothesis comingfrom a number of empirical studies conducted within various paradigms eg in picturenaming (eg Costa Santesteban amp Cantildeo 2005 Hoshino amp Kroll 2008) in eye-tracking(eg Bartolotti amp Marian 2012 Ju amp Luce 2004) in word recognition and electro-physiological experimentation (eg Hoshino amp Thierry 2012 Peeters Dijkstra ampGrainger 2013) Studies on cognate processing and word processing in sentence contextcan offer further insight into the nature of multilingual mental lexicons

Corresponding author Email alijewskawaamuedupl

International Journal of Multilingualism 2014httpdxdoiorg101080147907182014959961

copy 2014 Taylor amp Francis

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The study presented in this article aims to shed more light on the nature of cognateprocessing in sentence context by trilinguals We have included conference interpretingtrainees in the study to reveal any effect of interpreting training on their lexicalprocessing Below we first present an overview of studies focusing on the processing ofcognates and sentence context effects on lexical processing both in bilinguals andmultilinguals Their results are reviewed in the context of two models of bilinguallexicon the Revised Hierarchical Model (Kroll amp Stewart 1994) and the BIA+ model(Dijkstra amp van Heuven 2002) Later we focus on the studies of lexical processing byinterpreters and group differences between professional interpreters trainees and non-interpreting controls The actual experiment involving translation of L3ndashL2 cognates andnon-cognates from L3 into L2 and L1 in manipulated context conditions brings interestingdata discussed in relation to the two above mentioned models

Processing of cognates

One of the key findings interpreted in favour of the language non-selective access is thecognate facilitation effect Cognates are words which share both form and meaning acrosslanguages (eg the word BUTTER is a cognate for English and German) The typicalfinding in psycholinguistic experiments investigating reaction times (RTs) to cognates isthat they take shorter time to respond to in comparison to language-specific controlwords This discrepancy between RTs obtained for cognates and non-cognates is taken toindicate that during lexical access lexicons of all languages known to an individual areactivated The cognate facilitation effect has been found in a number of studies carriedout with bilingual speakers (Antoacuten-Meacutendez amp Gollan 2010 Costa Caramazza ampSebastian-Galles 2000 Davis et al 2010 De Groot Dannenburg amp Van Hell 1994Dijkstra Grainger amp van Heuven 1999 Dijkstra Miwa Brummelhuis Sappelli ampBaayen 2010 Dijkstra Timmermans amp Schriefers 2000 Gollan Forster amp Frost 1997Lemhoefer Dijkstra amp Michel 2004 Lotto amp De Groot 1998 Saacutechez-Casas Garciacutea-Albea amp Davis 1992 Van Hell amp De Groot 1998 Van Hell amp Dijkstra 2002)

Experiments testing cognate facilitation in trilinguals are scarce (Dijkstra amp Van Hell2003 Lemhoefer et al 2004 Poarch amp Van Hell 2012 Szubko-Sitarek 2011Tymczyńska 2011 Van Hell amp Dijkstra 2002) A study by Van Hell and Dijkstra(2002) involved three experiments and two groups of participants The first groupcomprised Dutch-English-French trilinguals with a relatively low L3 proficiency and thesecond group consisted of Dutch-English-French trilinguals whose fluency in French washigher than in the first group and comparable to that in English Both groups performedlexical decision tasks (LDTs when participants need to decide if a given string of lettersis a word or not) on Dutch words and additionally the first group performed a wordassociation test (ie they were asked to produce single word associations in response toL1 words presented on the computer screen) Materials used for the experiments wereL1 words which were either Dutch-English cognates or Dutch-French cognates or non-cognates All tasks were run in an exclusively L1 context ie participants had not beeninformed of the multilingual nature of stimuli used in the experiments and no explicitreference to the participantsrsquo foreign languages was made in the course of theexperimentation Van Hell and Dijkstra observed a significant cognate facilitation effectfor Dutch-English cognates in both groups The cognate facilitation effect for Dutch-French cognates was recorded only in the second group which had a higher fluency inFrench According to Van Hell and Dijkstra (2002) these results lend further support tolanguage non-selective access hypothesis They claim that the fact that significant cognate

2 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

facilitation was observed in an exclusively monolingual context seems to contradict thestrict version of the Language Mode Hypothesis (Grosjean 1998 2001) which positsthat if bilingualsmultilinguals are in a monolingual environment they are able todeactivate their non-target languages (Van Hell amp Dijkstra 2002 p 787)

Similar conclusions were drawn from another study conducted with trilingualspeakers Lemhoefer et al (2004) tested a group of Dutch-English-German trilingualswho performed a LDT in their L3 The participants were asked to decide if the presentedstring of letters formed a German word or not The critical stimuli used in the study werecognates overlapping in Dutch and German with no cognate relation to English as wellas cognates overlapping across the three languages RTs recorded for Dutch-Germancognates were significantly shorter than for control German words and significantlylonger than RTs for Dutch-English-German ones This cumulative cognate facilitationeffect is even more interesting in view of the fact that the cognate facilitation effectobserved for Dutch-English-German cognates was not influenced by whether the partici-pants read an English text and performed an item recognition test before the LDT task ornot According to Lemhoefer et al the lack of context effect runs counter to thepredictions of the Language Mode Hypothesis put forward by Grosjean (1998 2001)which assumes that the context of experimental situation should affect activation levelsof languages in the participantsrsquo minds All in all Lemhoefer et al (2004) interpretedthe results of their experiments in favour of the language non-selective access intrilingual processing

The cognate facilitation effect has been accounted for by a number of models ofbilingual lexical processing the most influential being the Revised Hierarchical Model(RHM) (Kroll amp Stewart 1994) and BIA+ (Bilingual Interactive Activation Plus) model(Dijkstra amp van Heuven 2002) Even though these two models have been extended tomultilingualism (for details see Tymczyńska (2012) and Dijkstra (2003) respectively)accounts of cognate facilitation within the models will be presented within the bilingualcontext (ie in the form they were originally put forward)

The cognate facilitation effect has been accommodated by RHM for L2 processing inthe following manner Due to direct L1ndashL2 lexical links the meaning of an L2 word isaccessed more quickly when there is an overlap in lexical representations (word forms) ofL1ndashL2 translation equivalents (as in cognates) as compared to the lack of such an overlap(as in non-cognates) According to Dijkstra et al (2010) cognates differ from non-cognates within the BIA+ model in that the former share more orthographic semanticandor phonological features with their translation equivalents than the latter The cognatefacilitation effect is accommodated by the model in the following manner (Dijkstra et al2010 p 286)

the cognate facilitation effect in reading might in fact be an orthographicndashsemantic primingeffect overlapping orthographic and semantic representations of both languages becomeactive upon the presentation of one of the readings of the cognate leading to a facilitatedrecognition of cognates relative to non-cognates

Consequently the cognate facilitation effect is sensitive to frequency as well as cross-linguistic similarity (both orthographic and phonological) of translation equivalentsBIA+ was supported in cognate studies conducted by eg Dijkstra et al (2010) and Vogaand Grainger (2007) Even though Dijkstra et al (2010) explained the mechanics ofcognate facilitation within BIA+ for bilinguals there is every reason to assume that asimilar pattern of processing will also work for multilinguals within the Multilingual

International Journal of Multilingualism 3

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Interactive Activation model (MIA) which is the multilingual extension of BIA(Dijkstra 2003)

Sentence context effects

It needs to be stressed here that the vast majority of the bilingualmultilingual researchtested cognates and non-cognates by presenting them as single words Only in a relativelysmall number of studies did the researchers investigate cognates in a sentence contextGiven the fact that in everyday ie non-laboratory language use speakers extremelyrarely use isolated words investigating lexical processing in sentence context seems tohave more ecological validity than experimentation with single words

Nevertheless studies testing cognates in sentence context are still very limited innumber and differ in terms of the chosen experimental paradigms eg word naming(Schwartz amp Kroll 2006) LDT (Duyck Van Assche Drieghe amp Hartsuiker 2007 VanHell amp De Groot 2008) translation (Van Hell amp De Groot 2008 Van Hell 2005) eye-tracking (Duyck et al 2007 Libben amp Titone 2009 Van Assche Drieghe Duyck Wel-vaert amp Hartsuiker 2011) The results recorded in these studies differ across paradigmsand do not always tally with findings from single word experiments

Results obtained from word naming suggest that the processing of cognates isaffected by context only if this context is not semantically constraining (ie low contexthenceforth LC) as in lsquoThe final word in this sentence ishelliprsquo where the sentence finaltarget word cannot be easily predicted As opposed to lsquoShe took a bite of the freshgreenhelliprsquo which is a semantically constraining context (ie high context henceforth HC)biased towards the final target word (APPLE) Schwartz and Kroll (2006) tested twogroups of Spanish-English bilinguals in word naming in L2 The materials includedidentical and non-identical cognates as well as Spanish-English interlingual homographsThe words were embedded in HC and LC sentences The sentences were presented usingthe Rapid Serial Visual Presentation (RSVP where words appear on the screen one-by-one) and the participants were asked to read out the target word RTs recorded in thestudy indicated a robust cognate facilitation effect in the LC condition however in theHC condition the effect was eliminated Schwartz and Kroll (2006) interpreted this findingin favour of the language non-selective access which is restricted when enough semanticinformation is available to suppress the non-target language representations

A similar pattern of results was obtained in an LDT In one of the experimentsreported on by Van Hell and De Groot (2008) a group of Dutch-English bilingualsperformed an L2 LDT where they were asked to decide if a letter string formed anEnglish word or not Target words were either (concrete or abstract) cognates or non-cognates embedded in HC LC sentences or presented in isolation (no context conditionNC) Significant cognate effect was observed only in the LC and NC conditions both forconcrete and abstract target words Van Hell and De Groot (2008) found no significantthree-way interaction between concreteness cognatersquos status and context This mightindicate that subtle meaning overlap variations between cognates and non-cognates donot influence the pattern of results

In the same paper Van Hell and De Groot (2008) reported on three experimentsinvolving L1ndashL2 (forward) and L2ndashL1 (backwards) translation Materials and participantswere similar to those in the LDT discussed above Unlike in word naming and LDTcognate facilitation effect was observed across all contexts and translation directions butin the HC condition the facilitation was heavily reduced

4 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The study presented in this article aims to shed more light on the nature of cognateprocessing in sentence context by trilinguals We have included conference interpretingtrainees in the study to reveal any effect of interpreting training on their lexicalprocessing Below we first present an overview of studies focusing on the processing ofcognates and sentence context effects on lexical processing both in bilinguals andmultilinguals Their results are reviewed in the context of two models of bilinguallexicon the Revised Hierarchical Model (Kroll amp Stewart 1994) and the BIA+ model(Dijkstra amp van Heuven 2002) Later we focus on the studies of lexical processing byinterpreters and group differences between professional interpreters trainees and non-interpreting controls The actual experiment involving translation of L3ndashL2 cognates andnon-cognates from L3 into L2 and L1 in manipulated context conditions brings interestingdata discussed in relation to the two above mentioned models

Processing of cognates

One of the key findings interpreted in favour of the language non-selective access is thecognate facilitation effect Cognates are words which share both form and meaning acrosslanguages (eg the word BUTTER is a cognate for English and German) The typicalfinding in psycholinguistic experiments investigating reaction times (RTs) to cognates isthat they take shorter time to respond to in comparison to language-specific controlwords This discrepancy between RTs obtained for cognates and non-cognates is taken toindicate that during lexical access lexicons of all languages known to an individual areactivated The cognate facilitation effect has been found in a number of studies carriedout with bilingual speakers (Antoacuten-Meacutendez amp Gollan 2010 Costa Caramazza ampSebastian-Galles 2000 Davis et al 2010 De Groot Dannenburg amp Van Hell 1994Dijkstra Grainger amp van Heuven 1999 Dijkstra Miwa Brummelhuis Sappelli ampBaayen 2010 Dijkstra Timmermans amp Schriefers 2000 Gollan Forster amp Frost 1997Lemhoefer Dijkstra amp Michel 2004 Lotto amp De Groot 1998 Saacutechez-Casas Garciacutea-Albea amp Davis 1992 Van Hell amp De Groot 1998 Van Hell amp Dijkstra 2002)

Experiments testing cognate facilitation in trilinguals are scarce (Dijkstra amp Van Hell2003 Lemhoefer et al 2004 Poarch amp Van Hell 2012 Szubko-Sitarek 2011Tymczyńska 2011 Van Hell amp Dijkstra 2002) A study by Van Hell and Dijkstra(2002) involved three experiments and two groups of participants The first groupcomprised Dutch-English-French trilinguals with a relatively low L3 proficiency and thesecond group consisted of Dutch-English-French trilinguals whose fluency in French washigher than in the first group and comparable to that in English Both groups performedlexical decision tasks (LDTs when participants need to decide if a given string of lettersis a word or not) on Dutch words and additionally the first group performed a wordassociation test (ie they were asked to produce single word associations in response toL1 words presented on the computer screen) Materials used for the experiments wereL1 words which were either Dutch-English cognates or Dutch-French cognates or non-cognates All tasks were run in an exclusively L1 context ie participants had not beeninformed of the multilingual nature of stimuli used in the experiments and no explicitreference to the participantsrsquo foreign languages was made in the course of theexperimentation Van Hell and Dijkstra observed a significant cognate facilitation effectfor Dutch-English cognates in both groups The cognate facilitation effect for Dutch-French cognates was recorded only in the second group which had a higher fluency inFrench According to Van Hell and Dijkstra (2002) these results lend further support tolanguage non-selective access hypothesis They claim that the fact that significant cognate

2 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

facilitation was observed in an exclusively monolingual context seems to contradict thestrict version of the Language Mode Hypothesis (Grosjean 1998 2001) which positsthat if bilingualsmultilinguals are in a monolingual environment they are able todeactivate their non-target languages (Van Hell amp Dijkstra 2002 p 787)

Similar conclusions were drawn from another study conducted with trilingualspeakers Lemhoefer et al (2004) tested a group of Dutch-English-German trilingualswho performed a LDT in their L3 The participants were asked to decide if the presentedstring of letters formed a German word or not The critical stimuli used in the study werecognates overlapping in Dutch and German with no cognate relation to English as wellas cognates overlapping across the three languages RTs recorded for Dutch-Germancognates were significantly shorter than for control German words and significantlylonger than RTs for Dutch-English-German ones This cumulative cognate facilitationeffect is even more interesting in view of the fact that the cognate facilitation effectobserved for Dutch-English-German cognates was not influenced by whether the partici-pants read an English text and performed an item recognition test before the LDT task ornot According to Lemhoefer et al the lack of context effect runs counter to thepredictions of the Language Mode Hypothesis put forward by Grosjean (1998 2001)which assumes that the context of experimental situation should affect activation levelsof languages in the participantsrsquo minds All in all Lemhoefer et al (2004) interpretedthe results of their experiments in favour of the language non-selective access intrilingual processing

The cognate facilitation effect has been accounted for by a number of models ofbilingual lexical processing the most influential being the Revised Hierarchical Model(RHM) (Kroll amp Stewart 1994) and BIA+ (Bilingual Interactive Activation Plus) model(Dijkstra amp van Heuven 2002) Even though these two models have been extended tomultilingualism (for details see Tymczyńska (2012) and Dijkstra (2003) respectively)accounts of cognate facilitation within the models will be presented within the bilingualcontext (ie in the form they were originally put forward)

The cognate facilitation effect has been accommodated by RHM for L2 processing inthe following manner Due to direct L1ndashL2 lexical links the meaning of an L2 word isaccessed more quickly when there is an overlap in lexical representations (word forms) ofL1ndashL2 translation equivalents (as in cognates) as compared to the lack of such an overlap(as in non-cognates) According to Dijkstra et al (2010) cognates differ from non-cognates within the BIA+ model in that the former share more orthographic semanticandor phonological features with their translation equivalents than the latter The cognatefacilitation effect is accommodated by the model in the following manner (Dijkstra et al2010 p 286)

the cognate facilitation effect in reading might in fact be an orthographicndashsemantic primingeffect overlapping orthographic and semantic representations of both languages becomeactive upon the presentation of one of the readings of the cognate leading to a facilitatedrecognition of cognates relative to non-cognates

Consequently the cognate facilitation effect is sensitive to frequency as well as cross-linguistic similarity (both orthographic and phonological) of translation equivalentsBIA+ was supported in cognate studies conducted by eg Dijkstra et al (2010) and Vogaand Grainger (2007) Even though Dijkstra et al (2010) explained the mechanics ofcognate facilitation within BIA+ for bilinguals there is every reason to assume that asimilar pattern of processing will also work for multilinguals within the Multilingual

International Journal of Multilingualism 3

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Interactive Activation model (MIA) which is the multilingual extension of BIA(Dijkstra 2003)

Sentence context effects

It needs to be stressed here that the vast majority of the bilingualmultilingual researchtested cognates and non-cognates by presenting them as single words Only in a relativelysmall number of studies did the researchers investigate cognates in a sentence contextGiven the fact that in everyday ie non-laboratory language use speakers extremelyrarely use isolated words investigating lexical processing in sentence context seems tohave more ecological validity than experimentation with single words

Nevertheless studies testing cognates in sentence context are still very limited innumber and differ in terms of the chosen experimental paradigms eg word naming(Schwartz amp Kroll 2006) LDT (Duyck Van Assche Drieghe amp Hartsuiker 2007 VanHell amp De Groot 2008) translation (Van Hell amp De Groot 2008 Van Hell 2005) eye-tracking (Duyck et al 2007 Libben amp Titone 2009 Van Assche Drieghe Duyck Wel-vaert amp Hartsuiker 2011) The results recorded in these studies differ across paradigmsand do not always tally with findings from single word experiments

Results obtained from word naming suggest that the processing of cognates isaffected by context only if this context is not semantically constraining (ie low contexthenceforth LC) as in lsquoThe final word in this sentence ishelliprsquo where the sentence finaltarget word cannot be easily predicted As opposed to lsquoShe took a bite of the freshgreenhelliprsquo which is a semantically constraining context (ie high context henceforth HC)biased towards the final target word (APPLE) Schwartz and Kroll (2006) tested twogroups of Spanish-English bilinguals in word naming in L2 The materials includedidentical and non-identical cognates as well as Spanish-English interlingual homographsThe words were embedded in HC and LC sentences The sentences were presented usingthe Rapid Serial Visual Presentation (RSVP where words appear on the screen one-by-one) and the participants were asked to read out the target word RTs recorded in thestudy indicated a robust cognate facilitation effect in the LC condition however in theHC condition the effect was eliminated Schwartz and Kroll (2006) interpreted this findingin favour of the language non-selective access which is restricted when enough semanticinformation is available to suppress the non-target language representations

A similar pattern of results was obtained in an LDT In one of the experimentsreported on by Van Hell and De Groot (2008) a group of Dutch-English bilingualsperformed an L2 LDT where they were asked to decide if a letter string formed anEnglish word or not Target words were either (concrete or abstract) cognates or non-cognates embedded in HC LC sentences or presented in isolation (no context conditionNC) Significant cognate effect was observed only in the LC and NC conditions both forconcrete and abstract target words Van Hell and De Groot (2008) found no significantthree-way interaction between concreteness cognatersquos status and context This mightindicate that subtle meaning overlap variations between cognates and non-cognates donot influence the pattern of results

In the same paper Van Hell and De Groot (2008) reported on three experimentsinvolving L1ndashL2 (forward) and L2ndashL1 (backwards) translation Materials and participantswere similar to those in the LDT discussed above Unlike in word naming and LDTcognate facilitation effect was observed across all contexts and translation directions butin the HC condition the facilitation was heavily reduced

4 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

facilitation was observed in an exclusively monolingual context seems to contradict thestrict version of the Language Mode Hypothesis (Grosjean 1998 2001) which positsthat if bilingualsmultilinguals are in a monolingual environment they are able todeactivate their non-target languages (Van Hell amp Dijkstra 2002 p 787)

Similar conclusions were drawn from another study conducted with trilingualspeakers Lemhoefer et al (2004) tested a group of Dutch-English-German trilingualswho performed a LDT in their L3 The participants were asked to decide if the presentedstring of letters formed a German word or not The critical stimuli used in the study werecognates overlapping in Dutch and German with no cognate relation to English as wellas cognates overlapping across the three languages RTs recorded for Dutch-Germancognates were significantly shorter than for control German words and significantlylonger than RTs for Dutch-English-German ones This cumulative cognate facilitationeffect is even more interesting in view of the fact that the cognate facilitation effectobserved for Dutch-English-German cognates was not influenced by whether the partici-pants read an English text and performed an item recognition test before the LDT task ornot According to Lemhoefer et al the lack of context effect runs counter to thepredictions of the Language Mode Hypothesis put forward by Grosjean (1998 2001)which assumes that the context of experimental situation should affect activation levelsof languages in the participantsrsquo minds All in all Lemhoefer et al (2004) interpretedthe results of their experiments in favour of the language non-selective access intrilingual processing

The cognate facilitation effect has been accounted for by a number of models ofbilingual lexical processing the most influential being the Revised Hierarchical Model(RHM) (Kroll amp Stewart 1994) and BIA+ (Bilingual Interactive Activation Plus) model(Dijkstra amp van Heuven 2002) Even though these two models have been extended tomultilingualism (for details see Tymczyńska (2012) and Dijkstra (2003) respectively)accounts of cognate facilitation within the models will be presented within the bilingualcontext (ie in the form they were originally put forward)

The cognate facilitation effect has been accommodated by RHM for L2 processing inthe following manner Due to direct L1ndashL2 lexical links the meaning of an L2 word isaccessed more quickly when there is an overlap in lexical representations (word forms) ofL1ndashL2 translation equivalents (as in cognates) as compared to the lack of such an overlap(as in non-cognates) According to Dijkstra et al (2010) cognates differ from non-cognates within the BIA+ model in that the former share more orthographic semanticandor phonological features with their translation equivalents than the latter The cognatefacilitation effect is accommodated by the model in the following manner (Dijkstra et al2010 p 286)

the cognate facilitation effect in reading might in fact be an orthographicndashsemantic primingeffect overlapping orthographic and semantic representations of both languages becomeactive upon the presentation of one of the readings of the cognate leading to a facilitatedrecognition of cognates relative to non-cognates

Consequently the cognate facilitation effect is sensitive to frequency as well as cross-linguistic similarity (both orthographic and phonological) of translation equivalentsBIA+ was supported in cognate studies conducted by eg Dijkstra et al (2010) and Vogaand Grainger (2007) Even though Dijkstra et al (2010) explained the mechanics ofcognate facilitation within BIA+ for bilinguals there is every reason to assume that asimilar pattern of processing will also work for multilinguals within the Multilingual

International Journal of Multilingualism 3

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Interactive Activation model (MIA) which is the multilingual extension of BIA(Dijkstra 2003)

Sentence context effects

It needs to be stressed here that the vast majority of the bilingualmultilingual researchtested cognates and non-cognates by presenting them as single words Only in a relativelysmall number of studies did the researchers investigate cognates in a sentence contextGiven the fact that in everyday ie non-laboratory language use speakers extremelyrarely use isolated words investigating lexical processing in sentence context seems tohave more ecological validity than experimentation with single words

Nevertheless studies testing cognates in sentence context are still very limited innumber and differ in terms of the chosen experimental paradigms eg word naming(Schwartz amp Kroll 2006) LDT (Duyck Van Assche Drieghe amp Hartsuiker 2007 VanHell amp De Groot 2008) translation (Van Hell amp De Groot 2008 Van Hell 2005) eye-tracking (Duyck et al 2007 Libben amp Titone 2009 Van Assche Drieghe Duyck Wel-vaert amp Hartsuiker 2011) The results recorded in these studies differ across paradigmsand do not always tally with findings from single word experiments

Results obtained from word naming suggest that the processing of cognates isaffected by context only if this context is not semantically constraining (ie low contexthenceforth LC) as in lsquoThe final word in this sentence ishelliprsquo where the sentence finaltarget word cannot be easily predicted As opposed to lsquoShe took a bite of the freshgreenhelliprsquo which is a semantically constraining context (ie high context henceforth HC)biased towards the final target word (APPLE) Schwartz and Kroll (2006) tested twogroups of Spanish-English bilinguals in word naming in L2 The materials includedidentical and non-identical cognates as well as Spanish-English interlingual homographsThe words were embedded in HC and LC sentences The sentences were presented usingthe Rapid Serial Visual Presentation (RSVP where words appear on the screen one-by-one) and the participants were asked to read out the target word RTs recorded in thestudy indicated a robust cognate facilitation effect in the LC condition however in theHC condition the effect was eliminated Schwartz and Kroll (2006) interpreted this findingin favour of the language non-selective access which is restricted when enough semanticinformation is available to suppress the non-target language representations

A similar pattern of results was obtained in an LDT In one of the experimentsreported on by Van Hell and De Groot (2008) a group of Dutch-English bilingualsperformed an L2 LDT where they were asked to decide if a letter string formed anEnglish word or not Target words were either (concrete or abstract) cognates or non-cognates embedded in HC LC sentences or presented in isolation (no context conditionNC) Significant cognate effect was observed only in the LC and NC conditions both forconcrete and abstract target words Van Hell and De Groot (2008) found no significantthree-way interaction between concreteness cognatersquos status and context This mightindicate that subtle meaning overlap variations between cognates and non-cognates donot influence the pattern of results

In the same paper Van Hell and De Groot (2008) reported on three experimentsinvolving L1ndashL2 (forward) and L2ndashL1 (backwards) translation Materials and participantswere similar to those in the LDT discussed above Unlike in word naming and LDTcognate facilitation effect was observed across all contexts and translation directions butin the HC condition the facilitation was heavily reduced

4 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Interactive Activation model (MIA) which is the multilingual extension of BIA(Dijkstra 2003)

Sentence context effects

It needs to be stressed here that the vast majority of the bilingualmultilingual researchtested cognates and non-cognates by presenting them as single words Only in a relativelysmall number of studies did the researchers investigate cognates in a sentence contextGiven the fact that in everyday ie non-laboratory language use speakers extremelyrarely use isolated words investigating lexical processing in sentence context seems tohave more ecological validity than experimentation with single words

Nevertheless studies testing cognates in sentence context are still very limited innumber and differ in terms of the chosen experimental paradigms eg word naming(Schwartz amp Kroll 2006) LDT (Duyck Van Assche Drieghe amp Hartsuiker 2007 VanHell amp De Groot 2008) translation (Van Hell amp De Groot 2008 Van Hell 2005) eye-tracking (Duyck et al 2007 Libben amp Titone 2009 Van Assche Drieghe Duyck Wel-vaert amp Hartsuiker 2011) The results recorded in these studies differ across paradigmsand do not always tally with findings from single word experiments

Results obtained from word naming suggest that the processing of cognates isaffected by context only if this context is not semantically constraining (ie low contexthenceforth LC) as in lsquoThe final word in this sentence ishelliprsquo where the sentence finaltarget word cannot be easily predicted As opposed to lsquoShe took a bite of the freshgreenhelliprsquo which is a semantically constraining context (ie high context henceforth HC)biased towards the final target word (APPLE) Schwartz and Kroll (2006) tested twogroups of Spanish-English bilinguals in word naming in L2 The materials includedidentical and non-identical cognates as well as Spanish-English interlingual homographsThe words were embedded in HC and LC sentences The sentences were presented usingthe Rapid Serial Visual Presentation (RSVP where words appear on the screen one-by-one) and the participants were asked to read out the target word RTs recorded in thestudy indicated a robust cognate facilitation effect in the LC condition however in theHC condition the effect was eliminated Schwartz and Kroll (2006) interpreted this findingin favour of the language non-selective access which is restricted when enough semanticinformation is available to suppress the non-target language representations

A similar pattern of results was obtained in an LDT In one of the experimentsreported on by Van Hell and De Groot (2008) a group of Dutch-English bilingualsperformed an L2 LDT where they were asked to decide if a letter string formed anEnglish word or not Target words were either (concrete or abstract) cognates or non-cognates embedded in HC LC sentences or presented in isolation (no context conditionNC) Significant cognate effect was observed only in the LC and NC conditions both forconcrete and abstract target words Van Hell and De Groot (2008) found no significantthree-way interaction between concreteness cognatersquos status and context This mightindicate that subtle meaning overlap variations between cognates and non-cognates donot influence the pattern of results

In the same paper Van Hell and De Groot (2008) reported on three experimentsinvolving L1ndashL2 (forward) and L2ndashL1 (backwards) translation Materials and participantswere similar to those in the LDT discussed above Unlike in word naming and LDTcognate facilitation effect was observed across all contexts and translation directions butin the HC condition the facilitation was heavily reduced

4 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Similarly to other methodologies studies employing eye-tracking methods have alsoyielded discrepant results as far as cognate processing in context is concerned Libbenand Titone (2009) and Titone Libben Mercier Whitford and Pivneva (2011) recordedeye movements of French-English bilinguals when they read L2 sentences with cognatesinterlingual homographs and matched control words Early-stage comprehensionmeasures (eg first fixation duration gaze duration and skipping) as well as late-stagecomprehension measures (eg go-past time and total reading time) showed significantcognate facilitation for LC sentences In the HC condition only the early stage compre-hension measures indicated cognate facilitation In contrast Van Assche et al (2011)who also investigated cognate processing with the use of eye-tracking found cognatefacilitation effect in both contexts for both early- and late-stage comprehension measuresin their study The most general observation from the aforementioned studies seems to bethat only semantically rich context exerts influence on cognate facilitation

Unfortunately to the best of our knowledge no studies concerning the influence ofcontext on cognate processing with trilinguals have been conducted yet Hence thepresent study aims at filling this gap However before a detailed account of ourexperimentation is provided the paragraphs to follow will briefly introduce interpreters asa special case of multilinguals since they constituted one of the groups tested in thepresent study

Lexical processing by interpreters

Interpreters and interpreting trainees are deemed a special case of bilingualsmultilingualsdue to their specific use of languages known to them (Chmiel 2010) According toParadis (1994) interpreters are required to complete tasks lsquothat bilinguals in order tominimise interference try to avoidrsquo (p 319) This specific use of two or more languageswhere the language systems for the source language and target language have to bedeliberately kept active makes conference interpreters and trainees an interesting researchgroup that can offer valuable insights into language processing by multilinguals

A number of studies have compared professional interpreters with trainees and non-interpreters by looking at the performance of all groups in interpreting tasks (Dillinger1994 Fabbro Gran amp Gran 1991 Joumlrg 1997 Padilla Bajo Canas amp Padilla 1995Riccardi 1998 Sunnari 1995) and found consistent superiority of interpreters overtrainees as well as trainees over non-interpreters reflected in smoother delivery prioritisinginformation effective cognitive capacity management deeper semantic processing easierrestructuring of syntax and better verb anticipation Interpreting practice thus seems toaccelerate various linguistic tasks including semantic processing

Studies involving professional interpreters and non-interpreting bilinguals performingsingle word translation tasks bring inconclusive results Christoffels De Groot and Kroll(2006) found shorter translation latencies for interpreters performing English-Dutch(L2ndashL1) and Dutch-English (L1ndashL2) word translation as compared to bilingual non-interpreting students but no difference between interpreters and English teachers Thissuggests that single word translation is a skill that can be boosted not only throughextensive interpreting practice but also through concurrent use of both languages inanother context ie foreign language teaching Moreover Christoffels et al (2006) foundasymmetries according to translation direction for the (non-interpreting) student grouponly Translation into L2 was faster than into L1 offering further support for the RHM-based suggestion of asymmetrical lexical links

International Journal of Multilingualism 5

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

One of the variables manipulated in the study by Christoffels et al (2006) wascognate status A general cognate effect was found In both translation directions theeffect was smaller for high frequency words than for low frequency words This pertainedto all experimental groups ie interpreters and non-interpreting bilinguals (both studentsand English teachers) The authors speculated that the cognate effect might be attenuatedfor interpreters if they are able to better control the non-target language However nointeraction between group and cognate status was found thus confirming that the cognateeffect size does not depend on group characteristics

The issue of language control language switching activating and inhibiting theappropriate systems is very crucial when more than one language comes into play inproduction According to De Groot and Christoffels (2006 p 194)

bilingual language production requires the suppression of the non-target language and theactivation of the target language and (hellip) the degree of suppression of the non-targetlanguage depends on language strength the stronger the non-target language the more itneeds to be suppressed in order not to interfere with the current target language

The same authors suggest that translation equivalents (both words and multiword units)are linked via direct memory connections since they tend to co-occur which formsassociations between them In fact lsquothe more often the same two terms (words or longerphrases) co-occur in a translation act the stronger the memory connection between themwill bersquo (De Groot amp Christoffels 2006 p 198)

The above suggestions are largely supported by Paradis (1984) who also claims thatduring production both languages are activated and the unspoken language undergoesinhibition Paradis (1994) proposed the Activation Threshold Hypothesis according towhich any trace (in our case understood as the word) will have its activation thresholddepending on its frequency of activation and time since its last activation (p 320)In other words the more often and the more recent the activation of a given word thelower will be its activation threshold However for the successful production of a wordits competitors have to be inhibited According to Paradis (1994) if a given item isactivated its competitors are inhibited and as the wordrsquos activation threshold decreasesthe threshold for the competing items increases The same pertains to the whole languagesystem The non-selected language system is inhibited lsquoto prevent interference duringproduction but not sufficiently to preclude borrowing (hellip) or comprehension in the otherlanguagersquo (Paradis 1994 p 321) Interpreters are thus different from non-interpretingbilinguals because while the latter inhibit the non-selected language to facilitateproduction and comprehension in the selected one the former keep both language systemsactivated with varying activation thresholds Paradis (1994) even suggests that lsquothe formof the SL [source language] in the short term memory (STM) seems to inhibit to someextent access to its translation equivalentrsquo (p 322)

A lot of research in this area both in psycholinguistics and Interpreting Studies hasbeen done with reference to two languages One of the few studies testing the influence ofinterpreting experience in the trilingual context was the one reported by Tymczyńska(2012) She tested professional interpreters conference interpreting students and non-interpreting trilinguals in a single word translation task from L1 to L3 She manipulatedthe cognate status of target translations ie some L3 target translations were cognateswith their L2 equivalents but not L1 stimulus words A moderate L2ndashL3 cognate effectwas found in all experimental groups in the L1ndashL3 translation task

6 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The present study

The present study was designed to address research questions concerning the effects ofinterpreting training and sentence context on translation as well as the influence ofsentence context on cognate facilitation Our participants were asked to translate L3words into L1 and L2 even though this kind of translation is not a part of the interpretingtraining half of our participants (ie interpreting trainees) went through By choosingL3ndashL2 and L3ndashL1 translation for the purposes of our experimentation we wanted to testwhether skills developed during interpreting training (in L1ndashL2 and L2ndashL1 translation)such as eg accelerated semantic processing are potent enough to boost performance ofinterpreting trainees in the unpractised translation direction (from L3 to L1 and fromL3 to L2)

Admittedly the task applied does not fully reflect what actually happens wheninterpreting However we had to simplify the experimental task in order to avoidnumerous otherwise uncontrolled and confounding variables The main differencebetween the experimental task (translation of the final word in a sentence) and conferenceinterpreting is that in real life interpreters interpret whole texts and not single words Thesentences used in the study do not form a coherent text and interpreters are only asked toprovide translations of the final words to measure their production latencies with satis-factory precision Such a design guarantees in our opinion the control of the variablesimportant for our study and can offer insights about lexical production by interpretersunder various context constraints and about how this production is influenced by theinterpreting experience De Bot (2000 p 80) supports this approach and claims thatstudying word translation can help us understand the nature of interpreting since languageproduction and perception are lexically-driven processes lsquohighly dependent on the speedof accessing of words in the lexiconrsquo

It has to be noted that we use the term translation in line with the psycholinguistictradition ndash in Translation Studies or in Interpreting Studies the task applied in the presentstudy would be called interpretation as it pertains to an oral rendition of the source textinto the target text To be even more precise since the stimuli are visual and not aural thetask is actually sight translation ie oral rendition of the source text presented visuallySight translation actually happens when interpreters are asked to interpret excerpts ofwritten documents at a meeting

Additionally it is worth mentioning that interpreting is frequently viewed in theTranslation Studies tradition as a three-stage activity eg involving comprehensiondeverbalisation (perception of sense without the verbal form) and production (cf Seles-kovitch amp Lederer 1995) However the latter theory has been extensively criticised(Poumlchhacker 2004 Setton 2003) Also the present study builds upon abundantpsycholinguistic research confirming language non-selective access and cognate effectin linguistic processing Thus in line with the psycholinguistic approach it is assumedhere that translation (as used in the form of an experimental task in the present study)involves comprehension and production stages only

Research questions

In view of the experiments and their findings discussed above the present study aimed toaddress the following questions

(1) Will L3ndashL2 cognates be translated faster than non-cognates in translation intoboth target languages (L2 and L1)

International Journal of Multilingualism 7

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

(2) Does sentence context constraint affect cognate processing in translation(3) Do interpreting trainees use context more efficiently than non-interpreters in

translation from their L3 to L1 and L2

Method

A 2 (group interpreting trainees non-interpreting trilinguals) by 2 (context constrainthigh context constraint low context constraint) by 2 (target language L2 L1) by 2(cognate status cognate non-cognate) mixed factorial design was used with group as abetween-group variable and context target language and cognate status as within-subjectvariables

Participants

Two groups of participants took part in the present study The group of interpretingtrainees (henceforth INT) included 23 interpreting students (6 male 17 female) with themean age of 221 All of them were students of the Institute of Applied Linguistics atAdam Mickiewicz University They were enrolled in the final year of the BA programmewith Applied Linguistics as their major and conference interpreting as an additionalspecialisation Their exposure to conference interpreter training was at least 150 contacthours (note-taking German-Polish and Polish-German consecutive interpreting German-Polish and Polish-German simultaneous interpreting) over the course of one year Thegroup of non-interpreting trilinguals (henceforth NON-INT) comprised 20 studentsenrolled in the penultimate and the final years of the BA programme (3 male 17 femalemean age 21) at the Institute of German Studies at Adam Mickiewicz University

All participants from both groups had Polish as L1 German as L2 and English as L3Their German competence was at the C2 level within the Common European Frameworkof Reference for Languages (CEFR) Such a high level of German competence was aprerequisite for university programmes the participants were enrolled in Their Englishwas not lower than the CEFR level B2 since this was the level of the English classes theyall attended Furthermore all students scored at least 80 at the English test administeredbefore the experiment (suggesting that their language proficiency in English was at leastat the upper-intermediate level) Other languages known to the participants includedSpanish Russian French Swedish Italian Finnish but the participants themselvesconsidered them their L4s (acquired last and at the lowest level of proficiency of alllanguages known to them at the time of the experiment)

Materials

Forty-four English concrete nouns matched for frequency and length were selected for theexperiment (these were taken partially from Kujałowicz Chmiel Rataj amp Bartłomiej‐czyk 2008) Word frequencies were taken from the CELEX lexical database (BaayenPiepenbrock amp Gulikers 1995) Half of these words were L3ndashL2 cognates (eg UNCLE-ONKEL) and half of them were non-cognates (eg PEAR-BIRNE) The cognate status ofthe words was assessed by a group of 20 trilinguals taken from a population similar to theparticipants of the present study They used a seven-point Likert scale to decide onthe similarity of translation equivalents (1-very low similarity 7-very high similarity)The trilinguals were asked to use both spelling and pronunciation as similarity criteriaThe mean score was 63 out of 7

8 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

The experimental stimuli appeared as final words in English sentences (taken partiallyfrom Kujałowicz et al 2008) with two types of context constraint high ndash HC (exampleShe loves pepperoni pizza with double CHEESE) and low ndash LC (example Stir in somebroccoli tomatoes and CHEESE) Each word appeared in HC and LC but no participantsaw the same word twice Both the sentences and the target words were presented inblack font against a white background

The type of context constraint in the sentence was established through a probabilitycloze test performed by 83 Polish-English bilinguals They were presented with sentenceswithout the final words and were asked to complete the final words so that the sentencewould make sense If at least 70 of the respondents completed a given sentence with thesame word (also happening to be one of the experimental stimuli) that sentence wasconsidered a high context constraint If less than 30 of the respondents completed agiven sentence with the same word that sentence was considered a low context constraintThe sentences were also judged by a native speaker of English for their accuracy andnaturalness Examples of HC and LC sentences with cognates and non-cognates used inthe study are given in Table 1

Apparatus and procedure

The experiment was run on a Fujitsu-Siemens laptop in a normally lit room or asimultaneous interpreting booth in a classroom Each participant was tested individuallyand asked to produce a verbal response (translation equivalent) into a microphoneReaction times were recorded by means of a voicekey (Serial Response Box) The stimuliwere presented in black on a white screen in E-Prime 20 (Schneider Eschman ampZuccolotto 2002)

The procedure within trial was as follows First the participants saw a fixation pointfor 500 ms and then they saw a sentence without the final target word The presentationof the sentence was self-paced ie the participants were asked to read the sentencesilently and carefully and press a spacebar immediately after having finished reading itThen another fixation point appeared for 500 ms followed by the target word to betranslated orally into the respective language as quickly and as accurately as possible Theword was displayed until the participantrsquos verbal response was registered by the voicekeyIf no response was given for 10000 ms the target word disappeared and the nexttrial began

The presentation of sentences was blocked by target language and counter-balancedacross participants Each block contained 22 sentences (11 HC and 11 LC pseudo-randomised sentences) where half of the target words were cognates and half were

Table 1 Example of HC and LC sentences with cognates and non-cognates used in the study

High context Low context

Cognate On sunny days they picnic sitting onthe green GRASSNice flowers and vegetables grow inher GARDEN

In the background she painted theGRASSThe children love playing in theGARDEN

Non-cognate Only few survivors reached theislandrsquos SHOREShe loves pepperoni pizza withdouble CHEESE

There were several trucks near theSHOREStir in some broccoli tomatoes andCHEESE

International Journal of Multilingualism 9

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

non-cognates Each experimental block was preceded by a practice block with 10sentences similar to those used in the experiment proper

The participants were asked to complete a language history questionnaire Then taskinstructions were provided in L3 at the beginning of the experiment The participantswere asked to provide an accurate translation of the target word as quickly and accuratelyas possible All in all each participant performed two trial blocks (translation into L2 andtranslation into L1) preceded by practice blocks Each trial block was followed by arecognition task to check if the participants read the sentences carefully enough to processthem semantically The whole experimental procedure took no more than 25 minutes

Results and discussion

For each participant mean RTs were calculated for each of the six conditions ndash context(HC vs LC) by target language (L2 vs L1) by cognate status (cognate vs non-cognate)The data were trimmed as follows observations following failed voicekey response orwrong verbal response (wrong translation gap fillers that activated the voicekey beforethe actual response etc) were eliminated Observations longer than 20 SD above theparticipantrsquos mean and longer than 2000 ms were also eliminated as outliers It turned outthat accuracy for L3ndashL2 translation was below 75 so it was decided to remove thiscondition from the present analysis Hence for further analysis RTs recorded for each ofthe four conditions ndash context (HC vs LC) by cognate status (cognate vs non-cognate)were considered

We performed a mixed design ANOVA just for the L3ndashL1 translation direction on themean subject RTs with context and cognate status as within-subject variables and groupas a between-subjects variable The mean RTs as well as error rates are presented inTable 2

Even though numerically patterns of cognate facilitation seem different we found nosignificant main group effect F(141) = 20 p gt 05 which means that the groups did notdiffer above the chance level The analysis also yielded a significant context effectF(141) = 790 p lt 05 but the interaction between group and context failed to reachsignificance F(141) = 037 p gt 05 This means that translations were produced atdifferent speeds in HC and in LC but the pattern of these mean RTs did not differ betweengroups The main cognate effect turned out to be significant so generally cognates weretranslated faster than non-cognates (F(141) = 4477 p lt 000) However the interactionbetween cognate status and context turned out to be insignificant (F(141) = 056p gt 05) showing that cognate facilitation was not affected by context A priori planned

Table 2 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in thetwo groups

Groups Conditions Mean (SD) ER

INT HC Cognates 860 (212) 28Non-cognates 902 (211) 84

LC Cognates 1017 (212) 56Non-cognates 1091 (202) 196

Non-INT HC Cognates 810 (184) 14Non-cognates 900 (142) 98

LC Cognates 1034 (201) 77Non-cognates 1044 (159) 189

10 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

simple effect analyses in the form of paired-samples t-tests (with Bonferroni correctionapplied resulting in a significance level set at p lt 025) revealed no significant cognatefacilitation effects in neither of the context conditions (t(42) = minus226 p gt 25 for HCcondition and t(42) = minus209 p gt 25 for LC conditions respectively) The significantcontext effects corroborate the findings by Van Hell and De Groot (2008) Van Hell(2005) and Schwartz and Kroll (2006) suggesting that a highly predictable sentencecontext facilitates translation production The main cognate effect was also significantoffering further support to Dijkstra and Van Hell (2003) Lemhoefer et al (2004) and VanHell and Dijkstra (2002) Detailed explanation of the fact that L3ndashL2 cognate facilitationwas recorded in L3ndashL1 translation is given in the discussion of results of the redefinedstudy below

As mentioned above low accuracy for L3ndashL2 translation resulted in the eliminationof the data and the target language variable from the analysis As far as interpretingtrainees are concerned such a low accuracy rate may suggest that these participants hadproblems with this translation direction since they had rarely interpreted from L3 to L2since interpreting from L3 is generally practised only into L1 and not L2 The interpretertraining that our trainees were exposed to only included L1ndashL2 and L2ndashL1 directions

Another explanation might be offered by the Activation Threshold Model by Paradis(1994) According to the model translation equivalents in the interpreterrsquos workinglanguages are generally activated in interpreting (for comprehension or productionpurposes depending on the source and target languages) Inhibition is the mechanism thatshould successfully prevent interference from the source language in the production in thetarget language In the present study it might have been the case that L3 stimuli inhibitedotherwise activated L2 equivalents to avoid potential interference which resulted inoffsetting cognate facilitation

The low accuracy of the non-interpreting trilinguals in interpreting into L2 could beexplained by their insufficient L3 competence (B2 level) but a more plausibleexplanation is (which may also be true for the interpreting group) that the experimentalstimuli in L3 inhibited L2 translation equivalents

We found no group effect and no influence of interpreting training on wordproduction by trilinguals in any condition which suggests that the training period or itsintensity was insufficient to influence the processes involved in lexical processing of thistype The lack of differences between interpreters and non-interpreters might stem fromthe nature of interpreter training to which our participants were exposed before the studyIt involved only L2ndashL1 and L1ndashL2 translation and skills thus developed might not betransferable to L3ndashL2 or L3ndashL1 translation This issue calls for further research on thenature language processing by interpreters and non-interpreters

The study redefined

As mentioned the effect of target language could not be analysed as it had been planneddue to low accuracy (lt75) in the L3ndashL2 translation part of the study However weredefined the experiment in order to be able to analyse the translation direction from aslightly different perspective Namely those participants of the original experiment whoseaccuracy in both translation directions was above 75 were selected and their naminglatencies were re-analysed Responses of other participants were excluded from the newanalysis In the paragraphs to follow a detailed report of the redefined experiment willbe provided

International Journal of Multilingualism 11

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Method

A 2 (cognate status cognate non-cognate) by 2 (context high constraint low constraint)by 2 (target language L1 L2) factorial design was used

Participants

Twenty participants (12 interpreting students and 8 non-interpreting trilinguals) whoseaccuracy in translation into both target languages was above 75 were selected out of thegroup described above The selected participants differed from the eliminated ones onlyin the accuracy score (at least 75 for the former and less than 75 for the latter)Interpreting students and non-interpreting trilinguals could be treated as a single groupsince no group effect was found in the analyses described in previous sections Howeverin order to double check for any group effects in the new group a non-parametric Mann-Whitney test was performed The fact that the test yielded no significant result (p gt 05)allowed for the selected participants to be treated as a homogenous sample

Results and discussion

Data included in this analysis and other analyses reported on in the paragraphs to followwere trimmed as described above In order to analyse data obtained from the redefinedgroup of participants a Repeated Measures ANOVA on participantsrsquo mean RTs wasperformed Table 3 presents participantsrsquo mean RTs (in ms) SD (in parentheses) and errorrates (ER in )

The Repeated Measures ANOVA yielded a significant main effect of the targetlanguage (F(119) = 63171 p lt 05) and a significant interaction between context andtarget language (F(119) = 13841 p lt 05) L3 nouns were translated 273 ms faster intoL1 than into L2 Moreover L3ndashL1 translation was 424 ms faster in the HC condition thanin the LC condition and the L3ndashL2 translation was 175 ms faster in HC than in LCcondition

There were also main effects of context (F(119) = 11848 p lt 05) and of the cognatestatus (F(119) = 1911 p lt 05) However the interaction between cognate status andcontext failed to reach the significance level (F(119) = 0356 p gt 05) This means thatgenerally the speed of translation was affected by context and the cognate status In orderto investigate the influence of context on the processing of cognates depending on thetarget language a set of a priori planned simple effects analyses within the two levels ofthe factor context were performed It needs to be noted that even though in the materialswe used only L3ndashL2 cognates (no L3ndashL1 cognates were included) when analysing dataa t-test with two levels of cognate status was performed not only for the L3ndashL2

Table 3 Mean RTs (in ms) SDs (in parentheses) and error rates (in ) in each condition in the twotranslation directions

L3ndashL1 translation L3ndashL2 translation

Conditions Mean (SD) ER Mean (SD) ER

HC Cognates 801 (172) 08 1002 (322) 33Non-cognates 896 (202) 35 1007 (279) 38

LC Cognates 996 (175) 00 1430 (224) 08Non-cognates 1051 (170) 94 1424 (232) 51

12 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

translation but also for the L3ndashL1 translation in order to check for any possible influenceof L3ndashL2 cognates on the production of L1 translation equivalents

The a priori planned simple effect analyses revealed a significant cognate facilitationeffect in both context constraints but only in the L3ndashL1 translation These analysesinvolved paired samples t-tests with Bonferroni correction applied resulting in asignificance level set at p lt 025 In HC cognates were translated 96 ms faster thannon-cognates (t(19) = minus280 p lt 025) and in LC the cognate facilitation effect was 55ms (t(19) = minus244 p lt 025) However no such effect was found for the L3ndashL2translation (in HC t(19) = minus07 p gt 025 in LC t(19) = 13 p gt 025) These results areparticularly interesting in view of the fact that only L3ndashL2 cognates were employed inthe study

As it transpires from the data translation from the third language (English) wassignificantly faster into the first language (Polish) than into the second language(German) This effect according to Francis and Gallard (2005) may be ascribed toproficiency The process of translation involves two subsequent processes ndash comprehen-sion and production Since production is more sensitive to proficiency effects thancomprehension the net effect favours production into the dominant language (L1) ratherthan into the weaker L2 This finding is also in line with Kujałowicz (2007)

The fact that context effects were found in translation lends further support to VanHell (2005) and Van Hell and De Groot (2008) However it needs to be noted that ourdata are not entirely in accord with the results from the two aforementioned studies Ashas been described above Van Hell (2005) as well as Van Hell and De Groot (2008)found significant cognate facilitation effects only in the LC condition which theyinterpreted as an indication of the fact that the semantically constraining context languagelimits non-selectivity in lexical processing In the present study however cognatefacilitation was recorded in both context constraints but only in L3ndashL1 translation Thisobservation is intriguing for three reasons

Firstly in our research we employed L3ndashL2 cognates (not L3ndashL1 ones) so theyshould have significantly affected processing in the L3ndashL2 translation more than in theL3ndashL1 one (see Tymczyńska 2012) However counter-intuitive our data are they stillindicate language non-selectivity in lexical processing exhibited by the trilingualparticipants If L3ndashL2 cognates influenced translation from L3 to L1 it means that L2translation equivalents must have been pre-activated to some degree when orthograph-ically similar L3 stimuli were presented

Secondly the absence of cognate facilitation effect in the L3ndashL2 translation can be aresult of inhibition in the production of L3ndashL2 translations especially that many of ourparticipants reported difficulties with the production of L2 equivalents in response to L3items in general The fact that accuracy in L3ndashL2 translation fell below 75 for morethan a half of the participants also suggests that translation to another foreign languagewas a rather difficult task As has already been mentioned Francis and Gallard (2005)suggested that translation is composed of two processes ndash comprehension and productionhence it might have been the case that cognate facilitation effect in comprehension wasattenuated by inhibition in production

Finally the results of our study may be interpreted as a reflection of the effects oflearning experiencelearning strategies applied by our participants which in turninfluenced the quality of lexical connections in the participantsrsquo mental lexicons Thefact that we observed cognate facilitation induced by L3ndashL2 cognates in L3ndashL1translation but not in L3ndashL2 translation has important implications for RHM (Kroll ampStewart 1994) and for BIA+ or its multilingual extension ndash MIA (Dijkstra 2003)

International Journal of Multilingualism 13

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

However since RHM was designed to model translation in bilinguals a number ofassumptions have to be made before RHM can be extended to our population oftrilinguals

Since with all probability most of our participants have acquired L3 in the context ofL1 (Polish) which is typical in the Polish educational system the majority of theparticipants are likely to have used L3ndashL1 translation and L1ndashL3 translation as avocabulary learning strategy By the same token the participants are very unlikely to haveused L3ndashL2 translation in their learning experience Consequently this might haveinduced strong L3ndashL1 lexical links and significantly weaker (or almost nonexistent) L3ndashL2 lexical links Furthermore our participants being quite proficient in L3 must havealready developed conceptual links between L3 and the conceptual store (see Francis ampGallard 2005) Hence as a result of learning experience L3 words are stronglyconnected (via lexical links) to their L1 equivalents and weakly connected to their L2equivalents Simultaneously L3 items are already linked to the conceptual store via theconceptual links Such a makeup of connections in the mental lexicons of our participantscan explain why we observed L3ndashL2 cognate facilitation in L3ndashL1 translation but not inL3ndashL2 translation The mechanics of cognate facilitation (or its lack) will be presented indetail with the use of an example of an L3ndashL2 cognate word BUTTER

When BUTTER is presented as a cue word in the L3ndashL1 translation task it first needsto be recognised by the visual system Following MIA (see Dijkstra 2003 Dijkstra et al2010) we assume that lexical representations of BUTTER get activated in L3 as well asin L2 due to the overlap in orthography and semantics of the representations Thenfollowing RHM it can be assumed that by way of lexical links between translationequivalents this activation travels to the L1 equivalent (MASŁO) ConsequentlyMASŁO receives activation from two sources (from the L3 and the L2 representationsof BUTTER) and these joint jolts of activation give MASŁO a head start in translation Incontrast when the to-be-translated L3 word is a non-cognate the activation received byits L1 equivalent via lexical links comes only from one source ie the L3 representationHence L1 equivalents are produced faster in response to L2ndashL3 cognates than inresponse to L3 non-cognates

In turn the lack of cognate facilitation in L3ndashL2 translation can be explained as aconsequence of the quality of connections between L3 and L2 equivalents When the L3ndashL2 cognate (eg BUTTER) is presented as a cue word in the translation task activationcannot travel between L2 and L3 lexical representations (as the links are weak) hencethe activation travels from L3 lexical representation to its conceptual representation andonly then (via conceptual links) to the lexical representation of BUTTER in L2 The sametakes place when L3 non-cognate words are translated into L2 Consequently as a resultof the weakness (or even lack) of lexical links between L2 and L3 items the L3ndashL2translation of cognates and non-cognates lasts equally long as in either case thetranslation is executed via the conceptual links which attenuate the facilitative (lexical)effects of cognates

Another interesting issue is the fact that cognate facilitation was not only present inHC but it was also almost twice the size of the cognate facilitation observed in the LCcondition This observation runs counter to other studies employing context which havebeen reported to this date This discrepancy might either be an artefact a result of noise inthe data due to a small sample or it can result from the qualitatively different processingin trilingual speakers

14 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

General conclusions

In this study we investigated the influence of sentence context on cognate facilitation inL3ndashL2 and in L3ndashL1 translation production We also aimed to compare interpretingtrainees with non-interpreters to test whether interpreting training leads to boostedperformance in the tested translation directions The data obtained suggest that translationfrom L3 to L2 was much more difficult for both groups of participants (very low accuracyrate) than from L3 to L1 Hence the former translation direction had to be initiallyexcluded from analysis and it could only be analysed once the design of the studywas redefined

In general our results provided evidence for the effect of sentence context ontranslation production leading to shorter latencies in a semantically constraining contextthan in a non-constraining one Also we found significant cognate facilitation effect inL3ndashL1 translation in that cognates were generally translated faster than non-cognatesThis finding is especially interesting if we take into account the fact that the studyemployed only L3ndashL2 cognates and no L3ndashL1 ones This pattern of results hasimplications for the representation of translation equivalents within RHM and it seemsto indicate that as a result of learning experience at the lexical level L3 words arestrongly connected to their L1 equivalents but only weakly (or not at all) to the L2 onesThe observation of possible influence of participantsrsquo educational experience on thestructure of their mental lexicons and lexical processing may prove useful infurther research

In turn cognate facilitation effect was found to be modulated by sentence contextonly once the study design was redefined However the observed pattern of themodulation is at odds with similar studies with bilinguals This might be a result of noisein the data or a peculiarity of lexical processing in L3 In order to address this problem ina more systematic manner further research with trilingual speakers seems to be in placeFuture experimentation should in particular address the question as to why cognatefacilitation is not observed in L3ndashL2 translation as well as to how sentence contextinfluences language processing in trilinguals

Another issue which calls for further research is the fact that we observed nosignificant differences in the use of sentence context by interpreting trainees and by non-interpreting trilinguals This might be interpreted as evidence for the lack of acceleratedsemantic processing on the part of interpreting trainees which would be in line with theresults obtained by Christoffels et al (2006) who found no difference in single wordtranslation tasks between interpreters and non-interpreting English teachers andconcluded that efficient word retrieval was not uniquely relevant for interpreting Sincethe present study involved translation directions which had not been practised duringinterpreting training before the study we could not conclude whether the lack of thedifference between groups stemmed from the selected translation directions or fromthe insufficient amount of interpreting training which had not yet led to observabledifferences in the speed of semantic processing in participants Hence further researchshould involve the comparison of translation performance in the practiced translationdirections and also at different stages of interpreting training

ReferencesAntoacuten-Meacutendez I amp Gollan T H (2010) Not just semantics Strong frequency and weak cognate

effects on semantic association in bilinguals Memory amp Cognition 38 723ndash739 doi103758MC386723

International Journal of Multilingualism 15

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Baayen R H Piepenbrock R amp Gulikers L (1995) The CELEX lexical database (Release 2)[CD-ROM] Philadelphia PA Linguistic Data Consortium University of Pennsylvania

Bartolotti J amp Marian V (2012) Language learning and control in monolinguals and bilingualsCognitive Science 36 1129ndash1147 doi101111j1551-6709201201243x

Chmiel A (2010) Interpreting studies and psycholinguistics A possible synergy effect In D GileH Gyde amp N K Pokorn (Eds) Why translation studies matters (pp 223ndash236) AmsterdamJohn Benjamins

Christoffels I K De Groot A M B amp Kroll J F (2006) Memory and language skills insimultaneous interpreting Expertise and language proficiency Journal of Memory andLanguage 54 324ndash345 doi101016jjml200512004

Costa A Caramazza A amp Sebastian-Galles N (2000) The cognate facilitation effectImplications for models of lexical access Journal of Experimental Psychology LearningMemory and Cognition 26 1283ndash1296 doi1010370278-73932651283

Costa A Santesteban M amp Cantildeo A (2005) On the facilitatory effects of cognate words inbilingual speech production Brain and Language 94(1) 94ndash103 doi101016jbandl200412002

Davis C Snchez-Casas R Garca-Albea J E Guasch M Molero M amp Ferr P (2010)Masked translation priming Varying language experience and word type with SpanishndashEnglishbilinguals Bilingualism Language and Cognition 13 137ndash155 doi101017S1366728909990393

De Bot K (2000) Simultaneous interpreting as language production In B Englund-Dimitrova ampK Hyltenstam (Eds) Language processing and simultaneous interpreting (pp 65ndash88)Amsterdam John Benjamins

De Groot A M B amp Christoffels I K (2006) Language control in bilinguals Monolingual tasksand simultaneous interpreting Bilingualism Language and Cognition 9 189ndash201 doi101017S1366728906002537

De Groot A M B Dannenburg L amp Van Hell J (1994) Forward and backward wordtranslation by bilinguals Journal of Memory and Language 33 600ndash629 doi101006jmla19941029

Dijkstra T (2003) Lexical processing in bilinguals and multilinguals The word selection problemIn J Cenoz B Hufeisen amp U Jessner (Eds) The multilingual lexicon (pp 11ndash26) DordrechtSpringer Netherlands Retrieved from httpwwwspringerlinkcomcontentvu385p748l632845

Dijkstra T Grainger J amp van Heuven W J B (1999) Recognition of cognates and interlingualhomographs The neglected role of phonology Journal of Memory and Language 41496ndash518 doi101006jmla19992654

Dijkstra T Miwa K Brummelhuis B Sappelli M amp Baayen H (2010) How cross-languagesimilarity and task demands affect cognate recognition Journal of Memory and Language62 284ndash301 doi101016jjml200912003

Dijkstra T Timmermans M amp Schriefers H (2000) On being blinded by your other languageEffects of task demands on interlingual homograph recognition Journal of Memory andLanguage 42 445ndash464 doi101006jmla19992697

Dijkstra T amp Van Hell J G (2003) Testing the language mode hypothesis using trilingualsInternational Journal of Bilingual Education and Bilingualism 6(1) 2ndash16 doi10108013670050308667769

Dijkstra T amp van Heuven W J B (2002) The architecture of the bilingual word recognitionsystem From identification to decision Bilingualism Language and Cognition 5 175ndash197doi101017S1366728902003012

Dillinger M (1994) Comprehension during interpreting What do interpreters know that bilingualsdonrsquot In S Lambert amp B Moser-Mercer (Eds) Bridging the gap Empirical research insimultaneous interpretation (pp 155ndash188) Amsterdam John Benjamins

Duyck W Van Assche E Drieghe D amp Hartsuiker R J (2007) Visual word recognition bybilinguals in a sentence context Evidence for nonselective lexical access Journal ofExperimental Psychology Learning Memory and Cognition 33 663ndash679 doi1010370278-7393334663

Fabbro F Gran B amp Gran L (1991) Hemispheric specialization for semantic and syntacticcomponents of language in simultaneous interpreters Brain and Language 41(1) 1ndash42doi1010160093-934X(91)90108-D

16 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Francis W S amp Gallard S L K (2005) Concept mediation in trilingual translation Evidencefrom response time and repetition priming patterns Psychonomic Bulletin amp Review 12 1082ndash1088 doi103758BF03206447

Gollan T H Forster K I amp Frost R (1997) Translation priming with different scripts Maskedpriming with cognates and non-cognates in Hebrew-English bilinguals Journal of ExperimentalPsychology Learning Memory and Cognition 23 1122ndash1139 doi1010370278-73932351122

Grosjean F (1998) Studying bilinguals Methodological and conceptual issues BilingualismLanguage and Cognition 1(2) 131ndash149 doi101017S136672899800025X

Grosjean F (2001) The bilingualrsquos language modes In J L Nicol (Ed) One mind twolanguages Bilingual language processing (pp 1ndash22) Oxford Blackwell

Hoshino N amp Kroll J F (2008) Cognate effects in picture naming Does cross-languageactivation survive a change of script Cognition 106 501ndash511 doi101016jcognition200702001

Hoshino N amp Thierry G (2012) Do SpanishndashEnglish bilinguals have their fingers in two piesndashoris it their toes An electrophysiological investigation of semantic access in bilinguals Frontiersin Psychology 3 1ndash6 doi103389fpsyg201200009

Joumlrg U (1997) Bridging the gap Verb anticipation in German-English simultaneous interpretingIn M Snell-Hornby Z Jettmarovaacute amp K Kaindl (Eds) Translation as interculturalcommunication (pp 217ndash228) Amsterdam and Philadelphia PA John Benjamins

Ju M amp Luce P A (2004) Falling on sensitive ears Constraints on bilingual lexical activationPsychological Science 15(5) 314ndash318 doi101111j0956-7976200400675x

Kroll J F amp Stewart E (1994) Category interference in translation and picture naming Evidencefor asymmetric connections between bilingual memory Representations Journal of Memoryand Language 33(2) 149ndash174 doi101006jmla19941008

Kujałowicz A (2007) Cross-linguistic factors in multilingual lexical processing a case oftrilingual speakers with English or German as L2 (Unpublished doctoral dissertation) AdamMickiewicz University in Poznan

Kujałowicz A Chmiel A Rataj K amp Bartłomiejczyk M (2008) The effect of conferenceinterpreting training on bilingual word production Paper presented at the 39th PoznańLinguistic Meeting Gniezno

Lemhoefer K Dijkstra T amp Michel M (2004) Three languages one ECHO Cognate effects intrilingual word recognition Language and Cognitive Processes 19 585ndash611 doi10108001690960444000007

Libben M R amp Titone D A (2009) Bilingual lexical access in context Evidence from eyemovements during reading Journal of Experimental Psychology Learning Memory andCognition 35 381 doi101037a0014875

Lotto L amp De Groot A M B (1998) Effects of learning method and word type on acquiringvocabulary in an unfamiliar language Language Learning 48(1) 31ndash69 doi1011111467-992200032

Padilla P Bajo M T Canas J J amp Padilla F (1995) Cognitive processes of memory insimultaneous interpreting In J Tommola (Ed) Topics in interpreting research (pp 61ndash72)Turku University of Turku

Paradis M (1984) Aphasie et traduction Meta 29(1) 57ndash67 doi107202003781arParadis M (1994) Towards a neurolinguistic theory of simultaneous translation The framework

International Journal of Psycholinguistics 10 319ndash335Peeters D Dijkstra T amp Grainger J (2013) The representation and processing of identical

cognates by late bilinguals RT and ERP effects Journal of Memory and Language 68315ndash332 doi101016jjml201212003

Poarch G J amp van Hell J G (2012) Cross-language activation in childrenrsquos speech productionEvidence from second language learners bilinguals and trilinguals Journal of ExperimentalChild Psychology 111 419ndash438 doi101016jjecp201109008

Poumlchhacker F (2004) Introducing interpreting studies London and New York NY RoutledgeRiccardi A (1998) Interpreting strategies and creativity In A Beylard-Ozeroff J Kraacutelovaacute amp

B Moser-Mercer (Eds) Translatorsrsquo strategies and creativity (pp 171ndash179) Amsterdam JohnBenjamins

International Journal of Multilingualism 17

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14

Saacutechez-Casas R M Garciacutea-Albea J E amp Davis C W (1992) Bilingual lexical processingExploring the cognatenon-cognate distinction European Journal of Cognitive Psychology4 293ndash310 doi10108009541449208406189

Schneider W Eschman A amp Zuccolotto A (2002) Prime userrsquos guide Pittsburgh PAPsychology Software Tools

Schwartz A I amp Kroll J F (2006) Bilingual lexical activation in sentence context Journal ofMemory and Language 55(2) 197ndash212 doi101016jjml200603004

Seleskovitch D amp Lederer M (1995) A Systematic approach to teaching Interpretation SilverSpring MD The Registry of Interpreters for the Deaf

Setton R (2003) Words and sense Revisiting lexical processes in interpreting Forum 1 139ndash168Sunnari M (1995) Processing strategies in simultaneous interpreting lsquoSaying it allrsquo vs synthesis

In J Tommola (Ed) Topics in interpreting research (pp 109ndash119) Turku Centre forTranslation and Interpreting University of Turku

Szubko-Sitarek W (2011) Cognate facilitation effects in trilingual word recognition Studies inSecond Language 1 189ndash208

Titone D Libben M Mercier J Whitford V amp Pivneva I (2011) Bilingual lexical accessduring L1 sentence reading The effects of L2 knowledge semantic constraint and L1ndashL2intermixing Journal of Experimental Psychology Learning Memory and Cognition 371412ndash1431 doi101037a0024492

Tymczyńska M (2011) Lexical processing in online translation tasks The case of Polish-English-German professional and trainee conference interpreters (Unpublished doctoral dissertation)Adam Mickiewicz University in Poznan Retrieved from httpsrepozytoriumamuedupljspuihandle105933287indexjsp

Tymczyńska M (2012) Trilingual lexical processing in online translation recognition Theinfluence of conference interpreting experience In D Gabryś-Barker (Ed) Cross-linguisticinfluences in multilingual language acquisition (pp 151ndash167) Heidelberg and New York NYSpringer

Van Assche E Drieghe D Duyck W Welvaert M amp Hartsuiker R J (2011) The influence ofsemantic constraints on bilingual word recognition during sentence reading Journal of Memoryand Language 64(1) 88ndash107doi101016jjml201008006

Van Hell J (2005) The influence of sentence context constraint on cognate effects in lexicaldecision and translation In J Cohen K T McAlister K Rolstad amp J MacSwan (Eds)Proceedings of the 4th international symposium on bilingualism (pp 2297ndash2309) SomervilleMA Cascadilla Press

Van Hell J G amp De Groot A M B (1998) Conceptual representation in bilingual memoryEffects of concreteness and cognate status in word association Bilingualism Language andCognition 1 193ndash211 doi101017S1366728998000352

Van Hell J G amp De Groot A M B (2008) Sentence context modulates visual word recognitionand translation in bilinguals Acta Psychologica 128 431ndash451 doi101016jactpsy200803010

Van Hell J amp Dijkstra T (2002) Foreign language knowledge can influence native languageperformance in exclusively native contexts Psychonomic Bulletin amp Review 9 780ndash789doi103758BF03196335

Voga M amp Grainger J (2007) Cognate status and cross-script translation priming Memory ampCognition 35 938ndash952 doi103758BF03193467

18 A Lijewska and A Chmiel

Dow

nloa

ded

by [

Agn

iesz

ka L

ijew

ska]

at 1

145

25

Sept

embe

r 20

14