Does bilingualism hamper lexical access in speech production

This article was published in an Elsevier journal. The attached copy is furnished to the author for non-commercial research and education use, including for instruction at the author’s institution, sharing with colleagues and providing to institution administration. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright Author's personal copy Available online at www.sciencedirect.com Acta Psychologica 127 (2008) 277–288 www.elsevier.com/locate/actpsy Does bilingualism hamper lexical access in speech production? Iva Ivanova, Albert Costa * GRNC, Parc Cientı́fic, Hospital Sant Joan de Déu, Departament de Psicologia Bàsica, Universitat de Barcelona, Spain Received 21 September 2006; received in revised form 3 June 2007; accepted 8 June 2007 Available online 26 July 2007 Abstract In the present study, we tested the hypothesis that bilingualism may cause a linguistic disadvantage in lexical access even for bilinguals’ first and dominant language. To this purpose, we conducted a picture naming experiment comparing the performance of monolinguals and highly-proficient, L1-dominant bilinguals. The results revealed that monolinguals name pictures faster than bilinguals, both when bilinguals perform picture naming in their first and dominant language and when they do so in their weaker second language. This is the first time it has been demonstrated that bilinguals show a naming disadvantage in their L1 in comparison to monolingual speakers. Ó 2007 Elsevier B.V. All rights reserved. PsycINFO classification: 2720; 2340 Keywords: Bilingualism; Lexical access; Language production 1. Introduction In recent years, ample evidence has been brought to bear on the cognitive advantages associated with speaking two languages from an early age (Bialystok, 1999, 2001; Bialystok, Craik, & Ruocco, 2006; Bialystok & Martin, 2004; Costa, Hernández, & Sebastián-Gallés, in press; Ransdell, Arecco, & Levy, 2001; see Bialystok, 2005, for an overview). Researchers generally agree that these stem from the need to keep apart two representational systems and use each one appropriately. Could, however, the co-existence of the two linguistic systems also result in some sort of processing disadvantage for bilingual speakers? Specifically, does bilingualism affect the ease with which lexical representations are retrieved from the lexicon during speech production? The present article aims at answering this question. A review of the relevant literature seems to provide some evidence suggesting a bilingual disadvantage in lexical access in speech production. Kohnert, Hernandez, and Bates (1998) have shown that on the Boston Naming Test, * Corresponding author. E-mail address: [email protected] (A. Costa). 0001-6918/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.actpsy.2007.06.003 bilinguals scored below monolingual norms in both their languages (and this was later confirmed by Roberts, Garcia, Desrochers, & Hernandez (2002), who tested both bilinguals and monolinguals). In a verbal fluency task (generation of as many exemplars as possible of a given category), monolinguals outperformed bilinguals (at least clearly in the semantic categories), both when college-aged participants (Gollan, Montoya, & Werner, 2002) and healthy older adults (Rosselli et al., 2000) were tested. Also, bilinguals seem to experience more tip-of-the-tongue states (TOTs) than monolinguals (Gollan & Acenas, 2004, Gollan & Silverberg, 2001; but see Gollan & Brown, 2006, for a re-examination of the origin of TOTs). Perhaps the most relevant study for the present purposes is that conducted by Gollan, Montoya, Fennema-Notestine, and Morris (2005) (see also Mägiste, 1979), in which bilinguals were slower and less accurate in naming pictures in their dominant language than monolinguals. Interestingly, however, this difference between the two groups disappeared after several repetitions of the same stimuli. Several explanations for the observed bilingual disadvantage have been put forward (see Gollan et al., 2005). First, this disadvantage may stem from a disguised word frequency effect (see Mägiste, 1979; Ransdell & Fischler, Author's personal copy 278 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 1987). Given that presumably bilingual speakers use their dominant language less often than monolingual speakers, it is conceivable that the frequency values of the corresponding lexical representations are lower for the former group of speakers. Accordingly, the differential frequency with which words are used in the dominant language may affect the respective availability of lexical items in monolingual and bilingual speakers, and such a difference in availability can be enough to give monolinguals a head start in retrieving lexical representations. Another explanation of the bilingual disadvantage argues that bilinguals could suffer from cross-language interference. The basic idea here is that words from the language not used for production in a given communicative act become nevertheless activated and compete for selection with the lexical representations of the language being used (e.g. Green, 1998; Hermans, Bongaerts, de Bot, & Schreuder, 1998; Lee & Williams, 2001; but see Costa, 2005; Costa, Colomé, Gómez, & Sebastian-Galles, 2003; Costa, Miozzo, & Caramazza, 1999; Finkbeiner, Gollan, & Caramazza, 2006, for a discussion of this cross-language competition). Regardless of its specific origin, caution needs to be exercised when generalizing the observed bilingual disadvantage to other populations of bilingual speakers. This is because, in the studies reviewed above, bilingual speakers with switched dominance were tested. That is, for most of those individuals, their dominant language was not the first acquired language (L1). For instance, in Gollan et al.’s (2005) study, the dominant language of the participants did not correspond to their first acquired language (Spanish) but rather to a language that was acquired later (English acquired at the age of 3.5–4 years). As a consequence, the difference between the performance of English monolinguals and Spanish–English bilinguals in English (their dominant but second language) cannot be taken as a definitive proof of a bilingual disadvantage, and it is certainly difficult to attribute to a specific origin. In other words, one should put on hold the conclusion that lexical access in the first and dominant language of bilingual speakers is hampered by their bilingual status until the appropriate group of bilinguals is tested (see Ransdell & Fischler, 1987, for similar claims). The aim of the present study is to examine whether lexical access in the first and dominant language of bilingual speakers is less efficient than that of monolingual speakers. Importantly, there are reasons to suspect that bilinguals in their dominant and first language might perform differently from switched-dominance bilinguals in their dominant but second language. First, the age of acquisition (AOA) values of words in the dominant language for switched-dominance bilinguals and for monolinguals may be different (for the robustness of AOA effects, see Gerhand & Barry, 1998; Izura & Ellis, 2002; Morrison & Ellis, 1995, 2000). In this context, these bilinguals speaking in their second language (L2) would be slower than monolinguals not because of an intrinsic bilingual disadvantage but rather because the AOA values of the words tested in the experiment were different for the two groups (note that AOA effects hold even for groups of items acquired two years from one another; Izura & Ellis, 2002; see also Section 4). Second, switched-dominance bilinguals may use their two languages in different social and linguistic contexts. In other words, the use of some L1 and L2 items for these bilinguals is complementary rather than overlapping. Thus, switched-dominance bilinguals may not even be sure of the translation equivalents of the corresponding items in the other language, or, at least, they are likely to have more problems in retrieving items in language A that they use (almost) exclusively in language B. In fact, there is only one study assessing whether bilinguals speaking in their first and dominant language show a disadvantage in comparison to monolinguals (Ransdell & Fischler, 1987). These authors compared English monolinguals with bilinguals whose dominant and first-acquired language was English on four tasks (list recognition, free recall, lexical decision and object naming). They found that bilinguals were slower than monolinguals in list recognition and lexical decision, but – importantly – there were no significant differences between the groups on free recall or object naming. Although these results are suggestive, we believe that they do not provide a conclusive answer to the question of whether bilinguals suffer a disadvantage in lexical retrieval. This is because the manner in which the pictures were presented and latencies recorded in this experiment was far from optimal. The pictures were presented in sets of 10 on a paper sheet, and responses were recorded for each paper sheet with a stop-watch. That is, there was no independent measure for each target picture but rather for the whole set of 10 pictures. This procedure may have introduced considerable noise in the measurement, reducing the chances of detecting an effect. This is especially problematic if we consider that in this experiment there was actually a relatively large numerical difference between monolinguals and bilinguals in the expected direction (45 ms), indicating that the experiment might simply not have had sufficient sensitivity to lead to significant results. Furthermore, the latencies for the lexical decision task in the same study, where the authors did find that bilinguals were slower than monolinguals, were collected on a trial-by-trial basis by means of a computer. Thus, it is likely that the sensitivity of the different data collection methods used for the two tasks resulted in the difference between monolinguals and bilinguals being significant in one (lexical decision) but not in the other (picture naming). Given the considerations above, one may argue that the existing evidence is not conclusive regarding whether bilingualism exerts a negative effect on the retrieval of lexical items in the first and dominant language. That is, for all we know at present, efficiency of lexical retrieval in the first and dominant language of bilinguals may not be different to that of monolinguals. To asses this issue, we compared the performance of Spanish monolingual speakers (Group 1) to that of Spanish–Catalan bilinguals (Group 2) in a picture naming task. Crucially, the first and dominant lan- Author's personal copy I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 guage of the Spanish–Catalan bilingual group was Spanish. We made the following predictions. If bilingualism affected negatively the availability of lexical representations, naming latencies for the monolingual group should be faster than those of the bilingual group. If, on the contrary, there was no effect of bilingualism, naming latencies for the monolingual group (Group 1) and for the Spanish–Catalan bilingual group (Group 2) should be comparable. Furthermore, we included two sets of pictures: (a) pictures with low-frequency names and (b) pictures with high-frequency names. The purpose of this manipulation was twofold. First, we wanted to explore the behavior of the frequency effect over repetitions in order to be able to compare it to the potential bilingual disadvantage over repetitions. If the potential bilingual disadvantage is due to a frequency effect in disguise, then we would expect such an effect to be affected by repetition in a similar manner as the frequency effect is. The frequency manipulation also allowed testing the sensitivity of our design. That is, if our design was sensitive enough to detect reliable differences in naming latencies, we should expect a word-frequency effect for all groups of participants. In order to further control for the sensitivity of our design to detect reliable differences in naming latencies we included yet another manipulation. We assessed the naming performance in Spanish of a group of Catalan–Spanish bilinguals, whose first and dominant language was Catalan. This group served as a control, since regardless of the outcome of the comparison between the other two groups, we expected slower naming latencies when participants named the pictures in their non-dominant L2 than when they did so in their dominant L1. It is important to stress that the objective of the present study was not to adjudicate between the different explanations put forward to account for the bilingual disadvantage. Rather, we aimed at assessing whether such a bilingual naming disadvantage is actually present when bilinguals perform a picture naming task in their first and dominant language. That is, we aimed at exploring whether there is indeed a genuine bilingual disadvantage in lexical access in speech production. 2. Experiment: The effect of bilingualism on lexical access in speech production 2.1. Method 2.1.1. Participants Thirty-seven participants were included in each of the three Groups. Monolinguals used only Spanish for daily communication and had little knowledge of other language(s).1 Bilingual participants had acquired their 1 Monolinguals’ knowledge of another language was attained from the obligatory foreign language instruction hours included in the school curriculum. However, as can be seen on Table A, they rated their proficiency in the foreign language as rather low. 279 L1 (Spanish and Catalan, respectively) at home with their families, and did not have continuous exposure to their L2 until a (pre-) school age. Importantly, we took care to ensure that participants across all three groups did not differ on any factor other than linguistic status (monolingual vs. bilingual). That is, participants were matched in age and education: all of them were university students (most of them psychology students) at either the Autonomous University of Madrid (the monolingual group) or the University of Barcelona (the two bilingual groups). As a consequence, they were receiving education in the same educational system and had passed a common mandatory exam to be enrolled in university (see Table 1 for participants’ language history and proficiency ratings). All participants received a course credit in exchange for participation in the experiment. 2.1.2. Materials A total of 50 pictures were chosen from the International Picture Naming Project database (http://crl. ucsd.edu/~aszekely/ipnp/). There were 25 pictures with high-frequency names (lemma frequency: 468, SD 377; word frequency: 339, SD 296) and 25 pictures with lowfrequency names (lemma frequency: 40, SD 21; word frequency: 24, SD 15). The difference between the frequency values of the two sets of pictures was significant [lemma frequency: t (48) = 5.65, p < .001; word frequency: t (48) = 5.30, p < .001]. The picture names across the high-frequency (HF) and the low-frequency (LF) sets were matched in length (syllables: HF – 2.52, SD 0.65; LF – 2.56, SD 0.71; phonemes: HF – 5.48, SD 1.39; LF – 5.72, SD 1.49; [both ts < 1]), and semantic category (see Appendix for a full list of the stimuli). The number of cognates (translation words with similar phonological properties) and non-cognates was the same in both sets (11 and 14, respectively). The HF cognates (lemma frequency: 626, SD 432; word frequency: 452, SD 320) were of a slightly higher frequency than the HF non-cognates (lemma frequency: 344, SD 286; word frequency: 250, SD 252) [lemma frequency: t (23) = 1.96, p = .06; word frequency: t (23) = 1.76, p = .09]. This was not true for the LF cognates (lemma frequency: 42, SD 22; word frequency: 26, SD 14) compared to the LF noncognates (lemma frequency: 39, SD 21; word frequency: 23, SD 16) [both ts < 1]. (Note, however, that cognate status was not a manipulation we initially planned to include in our experiment, and all analyses concerning cognate status were performed post-hoc). The difference between HF and LF cognates was significant, and so was the difference between HF and LF non-cognates [all ts > 3.3]. Regarding length, the HF cognates (syllables: 2.82, SD 0.75; phonemes: 6.09, SD 1.45) were longer than the HF non-cognates (syllables: 2.29, SD 0.47; phonemes: 5.00, SD 1.18) [syllables: t (23) = 2.17, p = .04; phonemes: t (23) = 2.08, p = .05], while the LF cognates (syllables: 2.27, SD 0.47; phonemes: 5.45, SD 0.93) contained less syllables than Author's personal copy 280 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 Table 1 Language history and self-assessed proficiency ratings for Groups 1–3 Age % Daily use of Spanish Now % Daily use of Spanish ages 2–10 Age exposed to Spanish Age speaking Spanish Age exposed to Catalan Age speaking Catalan Age exposed to other language Spanish proficiency Catalan proficiency Other language proficiency Difference L1–L2 Spanish monolinguals (Gr.1) Spanish–Catalan bilinguals (Gr. 2) Catalan–Spanish Bilinguals (Gr. 3) 23.2 97.5 97.6 0 (0) 1.5 – – 8.6 9.8 – 5.2 4.6 22.1 73.0 77.1 0 (0) 1.6 3.7 5.0 6.9 9.9 9.0 5.3 0.9 22.6 30.1 17.7 3.4 5.1 0 (0) 1.7 8.7 9.3 9.8 5.8 0.6 (3.2) (6.4) (7.6) (0.5) (4.6) (0.4) (2.0) (1.9) (4.3) (19.8) (16.8) (0.4) (2.2) (2.8) (2.5) (0.3) (0.9) (1.9) (0.9) (7.0) (20.4) (17.4) (2.2) (2.9) (0.6) (2.9) (0.8) (0.3) (1.8) (0.7) Standard deviations are given in brackets. ‘‘Age Exposed to Spanish/ Catalan’’ refers to the mean age at which participants received continuous exposure to these languages. ‘‘Age Speaking Spanish/ Catalan’’ refers to the mean age at which participants first started to speak these languages. ‘‘Age exposed to Other Language’’ refers to the mean age at which participants started formal instruction in a foreign language, in a classroom setting. Proficiency ratings are on a 1–10 scale, where 1 indicates ‘‘very little knowledge of the language’’, and 10 indicates ‘‘native proficiency’’. the LF non-cognates (syllables: 2.79, SD 0.80; phonemes: 5.93, SD 1.82) [syllables: t (23) = 1.88, p = .07; phonemes: t < 1]. The HF cognates contained more syllables than the LF cognates [syllables: t (20) = 2.05, p = .05; phonemes: t (20) = 1.23, p = .23]. Conversely, the HF non-cognates contained less syllables than the LF non-cognates [syllables: t (26) = 2.01, p = .05; phonemes: t (26) = 1.61, p = .12]. 2.1.3. Design and procedure The experiment consisted of 5 experimental blocks, amounting to 5 repetitions of the experimental pictures. There was an additional presentation of the pictures at the beginning of the experiment, which was considered a familiarization phase. Individual blocks were separated by a short pause. Each participant was assigned to one of 6 block orders created following a Latin square design. The order in which the pictures were presented in each block was randomized, with the restriction that no pictures whose names start with the same phoneme could immediately follow each other. An experimental trial had the following structure: (1) a fixation asterisk was presented for 700 ms; (2) an experimental picture was presented for 3000 ms or until the participant responded. The experiment was administered in a sound-attenuated, dimly-lit experimental room both at the University of Barcelona and the Autonomous University of Madrid. Participants were instructed to name the pictures as fast and as accurately as possible. The instructions were given in the following way to all participants: first the task was briefly explained to them orally and subsequently written instructions were presented on the computer screen. The instructions were given always in Spanish, and always by the same experimenter (first author). Participants were not informed about the goal of the experiment, and no reference was made to its relation to bilingualism. If participants produced an incorrect name for a picture during the familiarization phase, in the pause following it they were corrected. 2.1.4. Apparatus Due to the fact that we are comparing the performance of groups tested in different laboratories and with different equipment, we find it appropriate to describe the hardware and software we used in both cases in detail (see Table 2). Importantly, however, we are positive that any technical differences between the testing conditions for the three groups did not contribute in any way to the differential performance we observed between the monolingual group tested in Madrid, and the two bilingual groups (Spanish–Catalan and Catalan–Spanish) tested in Barcelona. More specifically, there is nothing in the parameters of both of the experimental computers used Table 2 Technical parameters of the computers used for testing the monolingual group in Madrid, and the two bilingual groups in Barcelona Madrid CPU RAM Video Card Intel Pentium II MMX 333 MHz Intel Pentium III 730 MHz 64MB 256MB ATI RAGE IIC AGP Display Nokia 447Zi Sound Board O. S. SoundBlaster 16 Microsoft Windows 95 (4.0, Build 950) DirectX 8.0 (4.08.00.0400) 2.9.05 13.39 ms DirectX version DMDX version Video refresh rate Barcelona Matrox Millenium G400 DualHead Samsung SyncMaster 17GLsi SB Audigy Audio Windows XP Prof (5.1 Build 2600 SP2) DirectX 9.0c (4.09.0000.0904) 3.1.21 16.59 ms Author's personal copy I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 which could produce a systematic pattern of faster or slower RTs. Both the response latencies and the actual verbal responses of participants were recorded by the same headset microphone with the help of the DMDX experimental software (Forster & Forster, 2003). The sensitivity of the voice-key was adjusted for each participant individually. 2.1.5. RT collection To make sure that the triggering of the voice key corresponded to the beginning of participants’ responses, all naming latencies were double-checked off-line. This was done by checking the sound files corresponding to each vocal response (these files were automatically recorded by DMDX in the course of the experiment and reflected a time window from picture onset until 3000 ms). In case the beginning of the participant’s response did not coincide with the triggering of the voice key, the corresponding latency was corrected to reflect the actual beginning of the response. For this purpose, the audio editor program Cool Edit (version 2000) was used. The off-line adjustments of the naming latencies were performed by the first author. Furthermore, in order to assess the reliability of these adjustments, a second rater, completely unaware of the purpose and predictions of the experiment, checked and adjusted the latencies for a subset of the responses (12%). This subset corresponded to the naming latencies of 22 participants (11 monolinguals and 11 Spanish–Catalan bilinguals) for the 1st and 2nd repetitions. When comparing the measurements of the two raters, we subtracted for each stimulus the adjusted values by the raters from one another, and averaged these differences for all stimuli. The average difference between the raters was .32 ms (SD 10.45 ms) for the monolingual group, and 1.39 ms (SD 11.38 ms) for the Spanish–Catalan bilingual group. When averaging the absolute differences between the raters, the inter-rater difference was 4.72 ms (SD 9.33 ms) for the monolingual group, and 5.78 ms (SD 9.90 ms) for the Spanish–Catalan bilingual group. Thus, it appears that the inter-rater reliability was quite high, and hence we can be rather confident about the general adjusted latencies obtained by the experimenter. Finally, errors and voice key failures were noted down by the experimenter online, and were subsequently double-checked offline. 2.1.6. Data analysis In order to explore the effects of interest for the present study, naming latencies and error rates were submitted to two different analyses. First, to assess whether bilinguals are at a disadvantage when naming in their first and dominant language, we compared the performance of the Spanish–Catalan bilingual group to that of the Spanish monolingual group. Second, to assess whether there is any difference between naming in L1 and in L2, we compared the performance of the two bilingual groups. In each of these ANOVA analyses, we declare three factors: ‘‘Rep- 281 etition’’ (repetitions 1–5) and ‘‘Frequency’’ (high vs. low frequency) as within-subject factors, and ‘‘Group of Participants’’ (Group 1 vs. Group 2, or Group 2 vs. Group 3) as a between-subject factor.2 3. Results Four types of responses were scored as errors: stuttering, utterance repairs, missing responses or wrong language responses. Also, naming latencies greater than two standard deviations from each participant’s mean were eliminated. This led to the trimming of 3.62% of the data for Group 1, 3.15% for Group 2 and 4.25% for Group 3. 3.1. Comparison 1: Group 1 (Spanish Monolinguals) and Group 2 (Spanish–Catalan bilinguals) In the naming latency analysis, the main effect of ‘‘Group of Participants’’ was significant [F1 (1, 72) = 4.70, MSE = 43,625, p = .03; F2 (1, 48) = 152.86, MSE = 912, p < .001], revealing faster naming latencies for the monolinguals (573 ms) than for the Spanish–Catalan bilinguals (606 ms) (See Table 3). The main effect of ‘‘Repetition’’ was also significant [F1 (4, 288) = 35.00, MSE = 875, p < .001; F2 (4, 192) = 66.04, MSE = 343, p < .001], as was the main effect of ‘‘Frequency’’ [F1 (1, 72) = 369.42, MSE = 1059, p < .001; F2 (1, 48) = 26.64, MSE = 9919, p < .001]. The interaction between ‘‘Group of Participants’’ and ‘‘Repetition’’ was not significant [F1 < 1; F2 (4, 192) = 1.84, MSE = 192, p < .12]. However, ‘‘Group of Participants’’ interacted with ‘‘Frequency’’ [F1 (1, 72) = 11.52, MSE = 1059, p = .001; F2 (1, 48) = 7.19, MSE = 912, p = .01], such that the frequency effect was slightly larger for the Spanish–Catalan bilinguals (54 ms) than for the monolinguals (38 ms). The interaction between ‘‘Repetition’’ and ‘‘Frequency’’ [F1 (4, 288) = 16.39, MSE = 286, p < .001; F2 (4, 192) = 9.96, MSE = 343, p < .001] was also significant, indicating that the frequency effect was reduced with repetitions. However, the frequency effect was still present for the last repetition. The three-way interaction between ‘‘Frequency’’, ‘‘Repetition’’ and ‘‘Group of Participants’’ was significant by subjects [F1 (4, 288) = 2.63, MSE = 286, p = .03] but not by items [F2 (4, 192) = 1.23, MSE = 255, p = .30]. Perhaps the most relevant interaction for our present purposes is that between ‘‘Group of Participants’’ and ‘‘Frequency’’. A closer look at this interaction revealed that 2 Since there were more males in the monolingual (12) and the Catalan– Spanish bilingual groups (10) than in the Spanish–Catalan bilingual group (2), we performed a separate analysis with only female participants (25 from each group) and the results were virtually identical to the ones obtained with all participants. The mean difference between the monolinguals (Group 1) and the Spanish–Catalan bilinguals (Group 2) was 34 ms and the mean difference between the Spanish–Catalan bilinguals (Group 2) and the Catalan–Spanish bilinguals (Group 3) was 52 ms. Therefore, in the analyses reported below we included all 37 subjects for each group. Author's personal copy 282 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 Table 3 Mean response latencies in ms (RT) and error rate percentages (%E) for the monolinguals (Group 1), Spanish–Catalan bilinguals (Group 2) and Catalan– Spanish bilinguals (Group 3) Group Frequency Repetition 1 RT% 2 4 5 RT% E RT% E RT %E RT %E Monolinguals (Group 1) HF LF D LF – HF 568 626 58 3.78 7.78 4.00 550 596 46 2.81 5.19 2.38 553 582 29 1.84 4.22 2.38 557 582 25 0.97 4.22 3.25 543 574 31 1.73 3.68 1.95 Sp-Cat Bilinguals (Group 2) HF LF D LF – HF 597 662 65 1.51 8.11 6.60 580 641 61 1.51 5.95 4.44 572 631 59 0.65 4.32 3.67 575 621 46 1.84 3.14 1.30 572 612 40 1.71 2.81 1.08 D Mono – Sp-Cat Bilinguals Cat-Sp Bilinguals (Group 3) E 3 33 HF LF DLF–HF 649 718 69 38 3.57 10.16 6.59 623 683 60 34 2.49 8.32 5.84 617 675 58 29 1.95 4.32 2.37 608 662 54 33 1.62 4.97 3.35 611 653 42 1.41 3.68 2.27 HF – high frequency; LF – low frequency. the difference between monolinguals and bilinguals was smaller for high-frequency words (25 ms) [F1 (1, 72) = 2.86, MSE = 20,515, p = .09; F2 (1, 24) = 91.71, MSE = 466, p < .001] than for low-frequency words (41 ms) [F1 (1, 72) = 6.57, MSE = 24,170, p = .01; F2 (1, 24) = 76.02, MSE = 1358, p < .001]. In the error analysis, the main effect of ‘‘Group of Participants’’ was not significant [all ps > .1]. The main effect of ‘‘Repetition’’ [F1 (4, 288) = 16.93, MSE = 12.67, p < .001; F2 (4, 192) = 12.82, MSE = 11.31, p < .001] and ‘‘Frequency’’ [F1 (1, 72) = 77.02; MSE = 23.12; p < .001; F2 (1, 48) = 11.18; MSE = 107.59; p = .002] were significant. ‘‘Group of Participants’’ did not interact either with ‘‘Repetition’’ [both Fs < 1], or ‘‘Frequency’’ [F1 < 1; F2 (1, 48) = 1.43, MSE = 10.73, p = .29]. The interaction between ‘‘Repetition’’ and ‘‘Frequency’’ was significant [F1 (4, 288) = 6.82, MSE = 11.02; p < .001; F2 (4, 192) = 4.49, MSE = 11.31; p = .002]. The three-way interaction was also significant [F1 (4, 288) = 3.19; MSE = 11.02; p = .01; F2 (4, 192) = 3.58; MSE = 6.62; p = .008]. The main result observed in this comparison was the monolingual advantage over bilinguals in naming latencies across the five repetitions of the experimental pictures (see Figs. 1 and 2). This effect appeared to be larger for the lowfrequency items than for the high-frequency items. 3.2. Comparison 2: Group 2 (Spanish–Catalan bilinguals) and Group 3 (Catalan–Spanish bilinguals) In the naming latency analysis, there was a main effect of ‘‘Group of Participants’’ [F1 (1, 72) = 6.36, MSE = 55,611, p = .01; F2 (1, 48) = 181.20, MSE = 1429, p < .001], such that the Catalan–Spanish bilinguals (606 ms) were slower than the Spanish–Catalan bilinguals (650 ms). The main effects of ‘‘Repetition’’ [F1 (4, 288) = 42.91, MSE = 1060, p < .001; F2 (4, 192) = 68.46, MSE = 478, p < .001] and ‘‘Frequency’’ [F1 (1, 72) = 290.84, MSE = 1951, p < .001; F2 (1, 48) = 28.09, MSE = 13,929, p < .001] were also sig- nificant. The interaction between ‘‘Group of Participants’’ and ‘‘Repetition’’ was significant only in the items analysis [F1 (4, 288) = 1.34, MSE = 1060, p = .25; F2 (4, 192) = 3.80, MSE = 261, p = .005]. The interaction between ‘‘Group of Participants’’ and ‘‘Frequency’’ was not significant [both Fs < 1]. The interaction between ‘‘Repetition’’ and ‘‘Frequency’’ was significant [F1 (4, 288) = 10.76, MSE = 340, p < .001; F2 (4, 192) = 4.90, MSE = 478, p = .001], revealing that low-frequency words benefited more from repetition. Importantly, the difference between the high- and low-frequency words in the last repetition was still significant for both groups. The three-way interaction between ‘‘Group of Participants’’, ‘‘Repetition’’ and ‘‘Frequency’’ was not significant [both Fs < 1]. In the analysis of error rates, the three main effects were significant: ‘‘Group of Participants’’ [F1 (1, 72) = 6.41, MSE = 34.36, p < .01; F2 (1, 48) = 5.77, MSE = 25.82, p = .02]; ‘‘Repetition’’ [F1 (4, 288) = 22.51, MSE = 13.87, p < .001; F2 (4, 192) = 17.78, MSE = 11.87, p < .001]; and ‘‘Frequency’’ [F1 (1, 72) = 123.99, MSE = 20.99, p < .001; F2 (1, 48) = 12.03, MSE = 146.23, p = .001]. Only the interaction between ‘‘Repetition’’ and ‘‘Frequency’’ was significant [F1 (4, 288) = 13.25, MSE = 11.77, p < .001; F2 (4, 192) = 8.87, MSE = 11.87, p < .001]. The main result observed in this comparison is the faster naming latencies when the task is performed in the bilinguals’ first and dominant language than when it is performed in the bilinguals’ second and non-dominant language (see Figs. 1 and 2). 3.3. Post-hoc analyses involving cognate status To further explore whether the bilingual disadvantage was modulated in any way by the cognate status of the picture names, we conducted a post-hoc analysis in which we assessed the effect of the variable ‘‘Cognate Status’’ and its interaction with the variables ‘‘Frequency’’, ‘‘Repetition’’, Author's personal copy 283 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 Mono (Gr. 1) 700 Bil L1 (Gr. 2) Bil L2 (Gr. 3) 680 Naming Latencies (ms) 660 640 620 600 580 560 540 1 2 3 4 5 Repetition 25 Mono Bil L1 Bil L2 % Data Points 20 15 10 5 0 350 550 750 950 1150 Naming Latencies (ms) HF LF 680 630 580 530 Naming Latencies (ms) Naming Latencies (ms) Monolinguals (Gr. 1) 730 730 Spanish-Catalan Bilinguals (Gr. 2) HF LF 680 630 580 530 1 2 3 4 5 1 2 Repetition 3 4 Repetition 5 Naming Latencies (ms) Fig. 1. Panel A: Overall mean picture naming latencies for the Spanish Monolinguals (Group 1), the Spanish–Catalan Bilinguals (Group 2) and the Catalan–Spanish Bilinguals (Group 3), averaged across high-frequency and low-frequency picture names. Error bars represent the standard error. Panel B: Distribution of all naming latencies (in percentage) for the three groups of participants. The size of the interval is 50 ms. Mono – monolinguals (Group 1); Bil L1 – Spanish–Catalan bilinguals (Group 2); Bil L2 – Catalan–Spanish bilinguals (Group 3). Catalan-Spanish Bilinguals (Gr. 3) HF LF 730 680 630 580 530 1 2 3 4 5 Repetition Fig. 2. Mean picture naming latencies for the Spanish Monolinguals (Group 1), the Spanish–Catalan Bilinguals (Group 2) and the Catalan–Spanish Bilinguals (Group 3) for high-frequency and low-frequency picture names (HF – high frequency; LF – low frequency). and, especially, ‘‘Group of Participants’’. Note, however, that caution needs to be exercised when interpreting the results of this post-hoc analysis (especially those of the item analysis) because of the different samples of cognate (22) and non-cognate words (28). As above, we carried out two comparisons: between the monolinguals (Group 1) and Spanish–Catalan bilinguals (Group 2), and between the Spanish–Catalan bilinguals (Group 2) and the Catalan–Spanish bilinguals speaking L2 (Group 3). In the first comparison, that between the monolinguals (Group 1) and the Spanish–Catalan bilinguals (Group 2), the difference between cognates and non-cognates was virtually identical for the two groups, revealing the absence of a cognate effect (monolinguals: 2 ms; Spanish–Catalan bilinguals: 5 ms) (see Fig. 3). Furthermore, the bilingual disadvantage was very similar for both types of words (cognates: 32 ms; non-cognates: 35 ms). Thus, it appears that the bilingual disadvantage is not modulated by Author's personal copy 284 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 Mono Cognates Mono Non-Cognates Bil L1 Cognates Bil L1 Non-Cognates Bil L2 Cognates Bil L2 Non-Cognates 700 Naming Latencies (ms) 680 660 640 620 600 580 560 540 1 2 3 4 5 Repetition Fig. 3. Mean picture naming latencies for the Spanish Monolinguals (Mono), the Spanish–Catalan Bilinguals (Bil L1) and the Catalan–Spanish Bilinguals (Bil L2), represented separately for cognate and non-cognate words. cognate status. Furthermore, the variable ‘‘Cognate Status’’ did not interact with ‘‘Repetition’’ ([F1 (4, 288) = 1.32, MSE = 510, p < .26; F2 < 1]). In the second comparison, that between the Spanish– Catalan bilinguals (Group 2) and the Catalan–Spanish bilinguals (Group 3), the difference between cognates and non-cognates was significantly larger for L2 than for L1 (L2: 18 ms; L1: 5 ms; interaction Cognate Status X Group of Participants: [F1 (1, 72) = 8.59, MSE = 1802, p = .005; F2 (1, 46) = 4.40, MSE = 1266, p = .04]). In other words, latencies for cognates were faster than those for non-cognates but only when the task was performed in L2 ([F1 (1, 36) = 35.64, MSE = 1678, p < .001; F2 (1, 46) = 2.70, MSE = 8073, p = .11]). The difference between groups was present for both cognates (38 ms) and non-cognates (50 ms) (see Fig. 3). A further interesting result comes from the observation that the magnitude of the cognate effect was only present for high-frequency words (high-frequency words: 31 ms; low-frequency words: 4 ms). That is, cognate words show a benefit in processing times only provided that they are of high frequency. 4. General discussion The question assessed in this article was the following: is lexical access in the first and dominant language of bilingual speakers less efficient than in monolinguals? We aimed at answering this question by comparing the picture-naming performance of a group of Spanish monolinguals, a group of Spanish–Catalan bilinguals, whose dominant language was also their first-acquired language, and a group of Catalan–Spanish bilinguals who performed the naming task in their L2 (Spanish). Several interesting results were observed when comparing the naming performance of these individuals. First and most importantly, monolinguals named the set of pictures faster than bilinguals – both when the latter did so in their second language, and, more importantly, when they did so in their first and dominant language.3 Secondly, the bilingual disadvantage did not disappear with repetitions, was larger for low- than for high-frequency words and was present for both cognate and non-cognate words. Thirdly, a cognate effect was present only in L2 and for high-frequency words. In the following, we discuss the theoretical implications of these findings. 4.1. Bilingual disadvantage, repetitions and frequency effects The presence of a bilingual disadvantage in picture naming partially replicates Gollan et al.’s (2005) observations, extending them in the direction that even bilinguals speaking in their first and dominant language show a delay in lexical access as indexed by picture naming. However, our results also diverge in some ways from those of Gollan et al. (2005), where the difference in latencies between monolingual and bilingual speakers disappeared in the fourth and fifth repetitions of the stimuli. The question arises of why Spanish–English bilinguals would catch up with English monolinguals with repetitions, while Spanish–Catalan bilinguals do not with Spanish monolinguals. Before seeking possible explanations of this difference, it is useful to assess the interpretation given by Gollan et al. (2005) to the bilingual disadvantage. Gollan et al. (2005) argued that the lack of difference between monolinguals and bilinguals in the last repetition reveals that the bilingual disadvantage was due to a frequency effect in disguise, and that such a frequency effect gets 3 These differences cannot be attributed to differences in the name agreement values between the monolingual and the Spanish–Catalan bilingual groups. This is because, during the familiarization phase (the first time participants saw the pictures), these two groups performed very similarly: the total number of different-than-expected names was 82 for the monolingual group and 90 for the Spanish–Catalan bilingual group (t < 1). Author's personal copy I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 reduced over time. This argument was based on the notion that frequency effects disappear with stimuli repetitions, and consequently, if the difference between monolinguals and bilinguals were due to frequency, it should also disappear with repetitions. However, the empirical evidence regarding word frequency on which such an argument is based is rather weak. This is because, in most experimental studies (e.g., Caramazza, Costa, Miozzo, & Bi, 2001; Jescheniak & Levelt, 1994; Jescheniak & Levelt, 1994; Levelt, Praamstra, Meyer, Helenius, & Salmelin, 1998; Navarrete, Basagni, Alario, & Costa, 2006), including the present one, frequency effects were still present after extensive repetitions of the stimuli.4 At present, it is difficult to know the origin of the contrastive results observed in our study and in Gollan et al.’s, given that differences in terms of the participants, items and procedure used discourage any meaningful comparison. Note, however, that even in Gollan et al’s study there was a non-significant trend in the expected direction in the 4th and 5th repetitions (30 ms. approximately). In sum, we consider that, without further support, at this stage it appears more problematic than otherwise to embrace the assumption that the bilingualism effect disappears with repetition. On the contrary, one could make the argument in the following way: that, in fact, the pervasive presence over repetitions of frequency effects and bilingualism effects seem to be more the rule than the exception, and – consequently – may call for a common explanation. In this respect, another result reported in our study is instructive: the difference between monolinguals and bilinguals was more pronounced for low-frequency words than for high-frequency words (see also Gollan, Montoya, Cera, & Sandoval, in press). In other words, the larger frequency effect for the bilingual group in comparison to the monolingual group is mostly carried by the low-frequency words. This finding is consistent with an interpretation of the bilingual disadvantage in terms of a difference in the frequency values of the words for the two populations as proposed by Gollan and colleagues (e.g., Gollan & Acenas, 2004; Gollan et al., 2005; Gollan & Silverberg, 2001; see also Mägiste, 1979; Ransdell & Fischler, 1987, for the original claim). Indeed, even our L1-dominant bilinguals claimed to use their L1 73% of the time, as compared to the 97.5% of monolinguals (this difference was significant [t (72) = 7.16, p < .001]). That is, the proposal that the bilingual disadvantage reveals a word-frequency effect in disguise in fact predicts a modulation of this disadvantage 4 There is in fact one study (Griffin & Bock, 1998) in which the frequency effect did disappear with the 3rd repetition of the stimuli. Thus, one could object that the persistence of the frequency effect over multiple repetitions is caused by variables such as the phonological properties of the stimuli, to the effect that the low-frequency words result inherently more difficult than high-frequency words, independently of frequency. However, we deem such a possibility unlikely: note that Navarrete et al. (2006) found a robust frequency effect with 4 repetitions, while their participants did not have to produce the target picture names at all (they were substituted with a pronoun). 285 depending on the frequency values of the words included in the experiment, and that its magnitude would be larger for low-frequency words. This is because a reduced frequency of use of high-frequency words by the bilingual participants would still render these words of rather high frequency for them, whereas a reduction of the frequency of use of medium to low-frequency words would render them of very low frequency for these speakers. In other words, lowering word frequency values (in the case of bilinguals) would affect much more low-frequency words than high-frequency words, and, as a consequence, the bilingual disadvantage would be more evident for the former set of words (see Gollan et al., in press, for a more detailed account). A possible concern with this interpretation is that it predicts that the frequency effect should be still larger for the Catalan–Spanish bilingual group (speaking L2) than for the Spanish–Catalan bilingual group (speaking L1). However, this prediction was not borne out by the data. At first sight, this fact is problematic for the argument developed above insofar as the Catalan–Spanish bilinguals speak Spanish (their L2), much less than the Spanish–Catalan bilinguals speak Spanish (their L1). Therefore, one should expect, everything else being equal, that the difference between the two groups would be more pronounced for low-frequency words (which is exactly what Gollan et al., in press, found for L2 speakers – they suffered larger frequency effects when speaking in L2 than when speaking in L1). However, the lack of differences between the two groups in relation to the frequency effect may stem from other factors, especially from the different AOA values of the words.5 The AOA variable may have played a role in our observations in the following way. The AOA of lexical items is highly correlated with their frequency values. Hence, unless the two variables are deliberately decoupled (which is not the case in our experiment), both of them are likely to contribute to the observed differences between the two sets of words. This would lead to an overestimation of the frequency effect for all groups of participants. (That is, part of the difference attributed to frequency may in fact stem from the correlated variable AOA). Note, however, that the relative contribution of these two variables to the observed latency differences may be different for each group. For the Spanish monolinguals and the Spanish– Catalan bilinguals, the contribution of AOA should be similar, given that their first-acquired language is the same (Spanish). However, the contribution of the AOA variable 5 An alternative explanation of this pattern of results appeals to floor effects: i.e., monolinguals have reached an asymptote level for highfrequency words and cannot speed up their naming latencies any further. However, note that naming latencies for high-frequency words for Spanish–Catalan bilinguals also stay virtually the same after the 3rd repetition. Given this observation, if one were to interpret the difference between monolinguals and bilinguals as revealing a floor effect, the question would still remain of why bilinguals reached such an asymptote level at higher latencies than monolinguals. Author's personal copy 286 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 in the case of Catalan–Spanish bilinguals may be smaller. This is because they are typically exposed to their L2 at about age 4, and it is to be expected that they would have acquired some high- and some low-frequency items almost simultaneously. In other words, those Spanish words that are acquired with a difference of, say, two years (from 2 to 4) for the monolingual and Spanish–Catalan bilingual groups, are likely to have been acquired at the same age (or within a smaller time interval) by the Catalan–Spanish group, therefore removing (or reducing) any effect of AOA between these words for this group. This scenario may explain why there is no difference between the two bilingual groups regarding the word frequency effect. Even if the frequency effect is actually larger when speaking in L2 (as argued above) than when doing so in L1, such an effect would be masked by the larger contribution of AOA to naming in L1 than to naming in L2. This is certainly a tentative explanation, but it can also capture some recent results reported by Gollan et al. (in press), in which the frequency effect was much larger in the weak than in the dominant language. Crucially, in this study, the first acquired language of participants was the weak one. As a consequence, naming latencies for the weak language may have shown both a frequency effect and an AOA effect, while latencies for the dominant language may have shown mostly a frequency effect. Thus, in this case, there may have been an overestimation of the frequency effect (because of the contribution of AOA) for the weak language. 4.2. Bilingual disadvantage and cognate effects Two other instructive results were obtained in our experiment. First, the bilingual disadvantage was present both for cognate and non-cognate words. Second, a cognate facilitation effect was present only in L2 and for high-frequency words. Although these results stem from a post-hoc analysis, and consequently, they need to be interpreted cautiously, we consider it pertinent to discuss them a little further. The first result reveals that the bilingual disadvantage is quite robust and does not depend on the phonological similarity between translation words. This was true both when the naming task was conducted in L1 (where no cognate effect was present) and when the task was performed in L2. The fact that a cognate facilitation effect was present only for those bilinguals speaking in L2 is not a surprising result, given that previous studies have shown that cognate effects are much more likely to be observed when participants name pictures in their L2 than in their L1.6 Thus, 6 Indeed, Costa, Caramazza, and Sebastián-Gallés (2000) did find cognate effects for the L1 of Spanish–Catalan bilinguals. While we are not able to give a precise account of this difference, one should keep in mind that (1) the cognate effect in Costa et al. (2000) was smaller in magnitude for L1 than for L2; and (2) our experiment was not designed to explore the cognate effect, and therefore our stimuli may not have been ideal to obtain a significant cognate effect in L1. if anything, these results show the reliability of the asymmetrical cognate effect for L2 and L1 (see Costa et al., 2000; Gollan, Fennema-Notestine, Montoya, & Jernigan, 2007; Hernández, Costa, Sebastián-Gallés, Juncadella, & Reñé, 2007). Second, the cognate effect for the Spanish–Catalan bilinguals was present only for high-frequency words. What would be the origin of the modulation of the cognate effect by word frequency? This question may find an explanation in the following framework. According to one explanation of the cognate facilitation effect, this effect arises as a consequence of the extra activation sent by the target’s translation to its phonological properties that, in the case of cognate words, are shared (to some extent) with those of the target word (see Costa, 2005; Costa et al., 2000; Costa, La Heij, & Navarrete, 2006; Costa, Santesteban, & Caño, 2005). That is, in the course of naming a picture (e.g. gato/ gat [cat]), the semantic system activates both the phonological properties of the target lexical node (Spanish ‘‘gato’’) and those of its translation in the non-response language (Catalan ‘‘gat’’) via the activation of the corresponding lexical nodes. Given that cognate words (‘‘gato’’ – ‘‘gat’’) share more phonological properties than non-cognates words (Spanish ‘‘mesa’’ – Catalan ‘‘taula’’ [table]), phonological encoding would be faster for cognates than for noncognates, leading to the observed facilitatory effect. On this view, an important factor predicting the magnitude of the cognate effect is the amount of activation that is sent by the target’s translation lexical node to the shared phonological properties – the stronger the activation sent, the larger the cognate effect. In principle, lexical items with a higher resting activation level would tend to send more activation to their corresponding phonological properties. The activation of lexical items may depend on several properties, such as, e.g., the language to which they belong, their frequency, etc. In this scenario, high-frequency translation words would send more activation to the corresponding phonological properties than low-frequency translation words, increasing the chances of observing a cognate effect. Note also that this framework gives a ready explanation for the larger cognate effects observed in L2 than in L1 (L1 lexical representations, when L1 is not the target language, would propagate more activation than L2 lexical representations) (see also Gollan & Acenas, 2004).7 7 It is possible that the presence of the cognate effect only for highfrequency words is simply due to the fact that the high-frequency cognates had higher frequency values than the high-frequency non-cognates (see Section 2.1.2). This argument would work because we found that the frequency effect was larger for bilinguals than for monolinguals. However, it leads to two predictions which are not supported by the data. First, we should have observed a significant cognate effect for the Spanish–Catalan bilinguals (speaking in L1) – which we did not. Second, if the cognate effect we observed were purely a frequency effect, it should be similarlysized in both bilingual groups (for whom we found similarly-sized frequency effects), rather than being larger for the group speaking in L2. In all, we deem unlikely such an explanation of the cognate effect pattern we find. Author's personal copy 287 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 5. Conclusion In the present study, we showed that bilingual speakers exhibit a lexical access disadvantage in language production in comparison to monolinguals even when the task is performed in their first and dominant language. A likely explanation of this effect can be found if we interpret it as a frequency effect in disguise, according to which bilingual speakers would need more time to retrieve lexical items from their L1 lexicon given that they use these words less often than monolingual speakers. nova was supported by a Pre-doctoral scholarship (FPU: AP2005-4496) from the Spanish Government. Requests for reprints should be addressed to Albert Costa. The authors thank Jose Manuel Igoa from the Autonomous University of Madrid for his assistance in testing the monolingual participants, to Kenneth Forster and Xavier Mayoral for their technical assistance, to Jasmin Sadat for performing the latency adjustments for a portion of the data, and to Tamar Gollan, Mikel Santesteban, Eduardo Navarrete and two anonymous reviewers for their useful comments on previous versions of this manuscript. Acknowledgements This research was supported by a grant from the Spanish Government (SEJ05 62542CV00568007). Iva Iva- Appendix. Materials used in the experiment High-frequency items Spanish name a + Estrella (Estela) + Libro (Llibre) + Caballo (Cavall) + Reloj (Relotge) + Zapato (Sabata) + Coche (Cotxe) + Corazón (Cor) + Camarero (Cambrer) + Vaso (Vas) + Teléfono (Telèfon) + Iglesia (Esglesia) Huevo (Ou) Lluvia (Pluja) Manzana (Poma) Cuchillo (Ganivet) Queso (Formatge) Mancha (Taca) Espejo (Mirall) Cerdo (Porc) Perro (Gos) Vela (Espelma) Llave (Clau) Ventana (Finestra) Silla (Cadira) Mesa (Taula) Low-frequency items English Tr. Frequency Star Book Horse Watch Shoe Car Heart Waiter Glass Phone Church Egg Rain Apple Knife Cheese Blot Mirror Pig Dog Candle Key Window Chair Table 227 1088 354 284 73 688 848 148 209 449 604 114 333 62 86 62 91 368 78 339 83 128 526 270 964 b Spanish name English Tr. Frequency + Martillo (Martell) + Cubo (Cubell) + Payaso (Pallaso) + Volcán (Volcà) + Flecha (Fletxa) + Globo (Globus) + Jarra (Gerra) + Violı́n (Violı́) + Cañon (Canyó) + Pincel (Pinzell) + Escoba (Escombra) Calcetı́n (Mitjó) Grifo (Aixeta) Muela (Queixal) Gusano (Cuc) Zanahoria (Pastanaga) Lavadora (Rentadora) Fresa (Maduixa) Seta (Bolet) Buho (Mussol) Cepillo (Raspall) Bandeja (Safata) Tenedor (Forquilla) Hacha (Destral) Mariposa (Papallona) Hammer Bucket Clown Volcano Arrow Baloon Jug Violin Cannon Paintbrush Broom Sock Tap Molar Worm Carrot Washing machine Strawberry Mushroom Owl Hair brush Tray Fork Axe Butterfly 30 34 23 31 25 57 22 29 1 20 16 8 39 8 24 13 6 16 2 25 29 60 21 35 35 + Cognate words. a The Catalan translations are given in brackets. b 1 per 1,000,000 word frequency ratings were taken out of the LEXESP-Corco database for Spanish. Author's personal copy 288 I. Ivanova, A. Costa / Acta Psychologica 127 (2008) 277–288 References Bialystok, E. (1999). Cognitive complexity and attentional control in the bilingual mind. Child Development, 70, 636–644. Bialystok, E. (2001). Bilingualism in development: Language, literacy, and cognition. New York: Cambridge University Press. Bialystok, E. (2005). Consequences of bilingualism for cognitive development. In J. F. Kroll & A. M. B. de Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. 417–432). New York: Oxford University Press. Bialystok, E., Craik, F. I. M., & Ruocco, A. C. (2006). Dual-modality monitoring in a classification task: The effects of bilingualism and ageing. The Quarterly Journal of Experimental Psychology, 59(11), 1968–1983. Bialystok, E., & Martin, M. (2004). Attention and inhibition in bilingual children: Evidence from the dimensional change card sorting task. Developmental Science, 7, 325–339. Caramazza, A., Costa, A., Miozzo, M., & Bi, Y. (2001). The specific-word frequency effect: Implications for the representation of homophones. Journal of Experimental Psychology: Learning, Memory and Cognition, 27, 1430–1450. Costa, A. (2005). Lexical access in bilingual production. In J. F. Kroll & A. M. B. De Groot (Eds.), Handbook of bilingualism: Psycholinguistic approaches. New York: Oxford University Press. Costa, A., Caramazza, A., & Sebastián-Gallés, N. (2000). The cognate facilitation effect: Implications for models of lexical access. Journal of Experimental Psychology: Learning, Memory and Cognition, 26(5), 1283–1296. Costa, A., Colomé, Á., Gómez, O., & Sebastian-Galles, N. (2003). Another look at cross-language competition in bilingual speech production: Lexical and phonological factors. Bilingualism: Language and Cognition, 6, 167–179. Costa, A., Hernández, M. & Sebastián-Gallés (in press). Bilingualism aids conflict resolution: Evidence from the ANT task. Cognition. Costa, A., La Heij, W., & Navarrete, E. (2006). The dynamics of bilingual lexical access. Bilingualism: Language and Cognition, 9, 137–151. Costa, A., Miozzo, M., & Caramazza, A. (1999). Lexical selection in bilinguals: Do words in the bilingual’s two lexicons compete for selection? Journal of Memory and Language, 41, 365–397. Costa, A., Santesteban, M., & Caño, A. (2005). On the facilitatory effects of cognate words in bilingual speech production. Brain and Language, 94, 94–103. Finkbeiner, M., Gollan, T., & Caramazza, A. (2006). Lexical access in bilingual speakers: What’s the (hard) problem? Bilingualism: Language and Cognition, 9(2), 153–166. Forster, K. I., & Forster, J. C. (2003). DMDX: A Windows display program with millisecond accuracy. Behavior Research Methods, Instruments, & Computers, 35(1), 116–124. Gerhand, C., & Barry, C. (1998). Word frequency effects in oral reading are not merely age-of-acquisition effects in disguise. Journal of Experimental Psychology: Learning, Memory and Cognition, 24, 267–283. Gollan, T. H., & Acenas, L. A. (2004). What is a TOT? Cognate and translation effects on tip-of-the-tongue states in Spanish–English and Tagalog–English bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 246–269. Gollan, T. H., & Brown, A. S. (2006). From tip-of-the-tongue (TOT) data to theoretical implications in two steps: When more TOTs means better retrieval. Journal of Experimental Psychology: General, 135(3), 462–483. Gollan, T. H., Fennema-Notestine, C., Montoya, R. I., & Jernigan, T. L. (2007). The Bilingual Effect on Boston Naming Test performance. The Journal of the International Neuropsychological Society, 13, 197–208. Gollan, T. H., Montoya, R., Cera, C., & Sandoval, T. (in press). When more use means smaller frequency effects: Aging, bilingualism, and the weaker links hypothesis. Journal of Memory and Language. Gollan, T. H., Montoya, R., Fennema-Notestine, C., & Morris, S. (2005). Bilingualism affects picture naming but not picture classification. Memory and Cognition, 33(7), 1220–1234. Gollan, T. H., Montoya, R., & Werner, G. A. (2002). Semantic and letter fluency in Spanish-English bilinguals. Neuropsychology, 16(4), 562–576. Gollan, T. H., & Silverberg, N. B. (2001). Tip-of-the-tongue states in Hebrew–English bilinguals. Bilingualism: Language and Cognition, 4, 63–84. Green, D. W. (1998). Mental control of the bilingual lexicosemantic system. Bilingualism: Language and Cognition, 1, 67–81. Griffin, Z. M., & Bock, K. (1998). Constraint, word frequency, and the relationship between lexical processing levels in spoken word production. Journal of Memory and Language, 38, 313–338. Hermans, D., Bongaerts, T., de Bot, K., & Schreuder, R. (1998). Producing words in a foreign language: Can speakers prevent interference from their first language? Bilingualism: Language and Cognition, 1, 213–229. Hernández, M., Costa, A., Sebastián-Gallés, N., Juncadella, M., & Reñé, R. (2007). The organization of nouns and verbs in bilingual speakers: A case of bilingual grammatical category-specific deficit. Journal of Neurolinguistics, 20, 285–305. Izura, C., & Ellis, A. W. (2002). Age of acquisition effects in word recognition and production in first and second languages. Psicológica, 23, 245–281. Jescheniak, J. D., & Levelt, W. J. M. (1994). Word frequency effects in speech production: Retrieval of syntactic information and of phonological form. Journal of Experimental Psychology: Learning, Memory and Cognition, 20(4), 824–843. Kohnert, K. J., Hernandez, A. E., & Bates, E. (1998). Bilingual performance on the Boston Naming Test: Preliminary norms in Spanish and English. Brain and Language, 65, 422–440. Lee, M.-W., & Williams, J. N. (2001). Lexical access in spoken word production by bilinguals: Evidence from the semantic competitor priming paradigm. Bilingualism: Language and Cognition, 4, 233–248. Levelt, W. J. M., Praamstra, P., Meyer, A. S., Helenius, P. I., & Salmelin, R. (1998). A MEG study of picture naming. Journal of Cognitive Neuroscience, 10(5), 553–567. Mägiste, E. (1979). The competing language systems of the multilingual: A developmental study of decoding and encoding processes. Journal of Verbal Learning and Verbal Behavior, 18, 79–89. Morrison, C. M., & Ellis, A. W. (1995). The roles of word frequency and age of acquisition in word naming and lexical decision. Journal of Experimental Psychology: Learning, Memory and Cognition, 21, 116–133. Morrison, C. M., & Ellis, A. W. (2000). Real age of acquisition effects in word naming and lexical decision. British Journal of Psychology, 91, 167–180. Navarrete, E., Basagni, B., Alario, F.-X., & Costa, A. (2006). Does word frequency affect lexical selection in speech production? Quarterly Journal of Experimental Psychology, 59(10), 1681–1690. Ransdell, S., Arecco, M. R., & Levy, C. M. (2001). Bilingual long-term working memory: The effects of working memory loads on writing quality and fluency. Applied Psycholinguistics, 22, 113–128. Ransdell, S. E., & Fischler, I. (1987). Memory in a monolingual mode: When are bilinguals at a disadvantage? Journal of Memory and Language, 26(44), 392–405. Roberts, P. M., Garcia, L. J., Desrochers, A., & Hernandez, D. (2002). English performance of proficient bilingual adults on the Boston Naming Test. Aphasiology, 16, 635–645. Rosselli, M., Ardila, A., Araujo, K., Weekes, V. A., Caracciolo, V., Padilla, M., et al. (2000). Verbal fluency and repetition skills in healthy older Spanish–English bilinguals. Applied Neuropsychology, 7, 17–24.