Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 Available online at www.sciencedirect.com Journal homepage: www.elsevier.com/locate/cortex Note Language proficiency modulates the engagement of cognitive control areas in multilinguals Jubin Abutalebi a,b, Pasquale A. Della Rosa a, Guosheng Ding c, Brendan Weekes b, Albert Costa d and David W. Green e,* a Vita-Salute San Raffaele University and San Raffaele Scientific Institute, Milan, Italy Division of Speech and Hearing Sciences, University of Hong Kong, Hong Kong c State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China d Universitat de Pompeu Fabra & ICREA, Barcelona, Spain e Cognitive, Perceptual and Brain Sciences, University College London, United Kingdom b article info abstract Article history: Language proficiency should modulate the regions involved in language control in Received 29 March 2012 predictable ways during language switching. However, prior studies reveal inconsistent Reviewed 15 June 2012 effects on the regions involved in language monitoring [pre-Supplementary Motor Area/ Revised 3 July 2012 Anterior Cingulate Cortex (pre-SMA/ACC)] and language selection (left caudate) conceivably Accepted 22 August 2012 because variations in relative proficiency are confounded with other between-group differ- Action editor Roberto Cubelli ences. We circumvented this problem in an fMRI (functional Magnetic Resonance Imaging) Published online 1 September 2012 study of overt picture naming in trilingual participants. In this case, the difference between a high-proficient and a low-proficient further language can be assessed within subjects Keywords: with no between-group confound. We also used a monolingual group to assess the neural Bilingual correlates of switching between two categories of response within the same language. Cognitive control We report a novel result: relative language proficiency dissociates response of the pre- Language control SMA/ACC and left caudate during language switching. Switching between languages Multilingual increased pre-SMA/ACC response regardless of proficiency differences. By contrast, left Language switching caudate response did vary with proficiency differences. Switching from the most to the least proficient language increased the response. Within-language switching, as contrasted with between-language switching, elicited a comparable increase in pre-SMA/ACC response but a decrease in left caudate response. Taken together, our data support a wider role of pre-SMA/ACC in task monitoring and establish the critical role of the left caudate in the selection of the less proficient language in language switching. ª 2012 Elsevier Ltd. All rights reserved. 1. Introduction Language use and cognitive control are intimately related in bilingual language processing. For successful communication bilinguals have to control their two languages in order to select the correct language for use and to avoid unwanted interference from the language not in use. Bilinguals achieve this feat by engaging brain areas closely related to cognitive * Corresponding author. Cognitive, Perceptual and Brain Sciences, University College London, Gower Street, London WC1E 6BT, UK. E-mail address: [email protected] (D.W. Green). 0010-9452/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cortex.2012.08.018 906 c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 control such as the pre-Supplementary Motor Area/Anterior Cingulate Cortex (pre-SMA/ACC), prefrontal cortex and the left caudate (Abutalebi and Green, 2007). The relative demand on each region relates to its cognitive function. The pre-SMA/ ACC is important for monitoring the language context, for detecting conflict and for avoiding errors that may arise during language selection. Prefrontal regions are implicated in the top-down control required for selecting the correct language and eventual error correction (Hernandez, 2009) whereas the left caudate is more specifically implicated in selecting the intended language (Crinion et al., 2006). Indeed, lesions to the left caudate and the prefrontal regions may lead to errors in language selection such as pathological language switching in bilinguals (Abutalebi et al., 2000). The activation of the regions in the language control network might be expected to vary as a function of language proficiency. This issue can be addressed in the language switching paradigm in which participants name pictures in each of their two languages in an intermixed fashion contingent on a specific cue (see for review, Luk et al., in press). In this paradigm we would expect that switching into a less proficient language would increase demand on the left caudate and so increase its activation. If ACC activation reflects conflict or error avoidance then it too would show an increase on switching to the less proficient language. Alternatively if the region monitors language context, activation would increase on a switch trial but not as a function of language proficiency. However results to date using this paradigm present an inconsistent picture. One reason is that differences in relative language proficiency are confounded by other between-group differences. In a study with highly proficient SpanisheCatalan bilingual speakers, Garbin et al. (2011) reported that switching into the first language (L1) elicited greater activation only in the pre-SMA/ACC complex, while switching into the second language (L2) engaged the left caudate. In contrast, in a study with low proficiency ChineseeEnglish bilingual speakers, Wang et al. (2007) recorded no engagement of the caudate and ACC activation only when bilinguals switched into the low-proficient L2. In this study, we aimed to achieve a better characterization of the response of critical regions in the language control network during language switching. Instead of using bilinguals, we studied language switching in early multilingual speakers. With such participants, we can investigate within subjects how a difference in relative proficiency for the two further languages alters the response of these regions. We compared the neural response to switching to a highproficient L2 relative to L1 and the response to switching to a lower-proficient third language (L3) relative to L1. Moreover, we also compared neural response to language switching in our multilinguals with the neural response to a withinlanguage switching task in a group of monolinguals. This comparison allowed us to explore the selectivity of neural response to between-language switching. 2. Materials and methods Participants in the study comprised 14 healthy right-handed multilinguals (GermaneItalianeEnglish) and 14 healthy right- handed Italian speaking monolinguals matched for socioeconomic background, education and age (mean ¼ 23.35 years, SD ¼ 4.5). All were female with normal or corrected-to-normal vision. All multilinguals were from South Tyrol. They learned German from birth as their L1. They attended school from age six and were taught in German and Italian and so they learned Italian from age six. All participants also had a reasonable mastery of English (L3) which they learned at around the age of eight. We investigated language proficiency with translation tasks (see Abutalebi et al., 2007). For L1 and L2, subjects translated 81.1% of words correctly from L1 into L2 and 74.2% of words correctly from L2 to L1. Translation from L1 into L3 and L3 to L1 yielded lower accuracy rates of 64.8% and 69.2%, respectively. Hence, L2 was classified as a high-proficient and L3 as a medium-proficient language (see Supplemental material for demographic data). There was no significant correlation between AoA (Age of acquisition) and test performance for translation into L2 (R ¼ .223; p ¼ .444) or into L3 (R ¼ .328; p ¼ .252). The investigation was approved by the Ethics Committee of the University San Raffaele and informed consent was obtained from participants. 2.1. Task and procedures Participants (multilinguals and monolinguals) performed an overt picture-naming task (see Fig. 1A for details) with two runs for each naming context. For the multilinguals the order of presenting the two contexts (L1eL2 and L1eL3) was counterbalanced over participants. For all languages, 32 different pictures (8.5 8.5 cm) were selected from the Snodgrass and Vanderwart (1980) set. Each picture was repeated three times for each language across the conditions, totaling 96 stimuli in each of the two contexts for multilinguals and the single context for monolinguals. Four pre-randomized lists were created defining the order in which the stimuli appeared. All stimuli were checked for frequency and syllable length in each language, based on the norms for each of the languages (German: Genzel et al., 1995; Italian: Laudanna et al., 1995; English: Leech et al., 2001). Pictures with cognate names were excluded. Each picture was displayed for 2 sec, followed by an ISI (Inter-Stimulus Interval) of 1880, 3550, or 4950 msec for the purposes of optimizing statistical power. Trials could be switch trials or non-switch trials. For the multilinguals, these were defined by the language required on the prior trial. It was a switch trial if a picture-to-be-named was preceded by one named in a different language and a non-switch trial if a picture-to-be-named was preceded by one named in the same language. In total there were 48 switch trials (for each language) and 48 non-switch trials (for each language) in each experimental context. Switch trials could occur in an unpredictable manner. For monolinguals, switch and non-switch trials were defined by whether the same or a different category of naming response (noun or verb) was required on the current compared to the preceding trial. Again there were 48 switch and 48 non-switch trials. Our analysis concerned trials on which pictures were named with a noun (noun naming trials) to match the response of multilingual speakers in their L1. c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 907 Fig. 1 e In (A), the experimental paradigm employed in our study. Multilinguals had to name pictures respectively in an L1eL2 and L1eL3 switching condition. In each condition, the color of the image-to-be-named indicated the language. In the L1eL2 context, green pictures indicated naming in German and blue pictures in Italian. In the L1-L3 context, green indicated naming in German and red in English. In order to have a comparable condition, monolinguals were asked to generate a noun when the color of the picture was red and a verb for green pictures. In (B), the pattern of brain activity related to the differences between switching contrasts versus non-switch contrast images at the first level for both multilinguals and monolinguals and the respective box plot calculated in the left caudate cluster (M [ monolinguals). In (C), the pattern of brain activity related to the average positive effects of all switching versus non-switch contrast and relatively, the box plots as calculated at the pre-SMA/ACC cluster. Technical constraints precluded the recording of voice onset times in either group. Prior to scanning all participants were trained using a different set of pictures. 2.2. Scanning, image processing and preprocessing The fMRI-event-related technique was used (3T Intera Philips body scanner, Philips Medical Systems, Best, NL, eight channelssense head coil, sense reduction factor ¼ 2, TE ¼ 30 msec, TR ¼ 2400 msec, FOV ¼ 240 240, matrix size ¼ 128 128, 30 contiguous axial slices per volume, 210 volumes per each run, slice thickness ¼ 4 mm). Each run was preceded by 10 dummy scans that were discarded prior to data analysis of correct noun naming trials (i.e., trials on which pictures were named with a noun). Multilinguals only completed such trials whereas they were a subset of trials for the monolinguals. A high resolution structural MRI (Magnetic Resonance Imaging) was acquired for each participant (MPRAGE, 150 slice T1-weighted image, TR ¼ 8.03 msec, TE ¼ 4.1 msec; flip angle ¼ 8 , TA ¼ 4.8 min, resolution ¼ 1 1 1 mm) in the axial plane. SPM (Statistical Parametric Mapping) running on Matlab 6.5 was used for all preprocessing steps and statistical analysis. Slice-timing correction was carried out by interpolating the voxel time series using sinc interpolation and resampling with the middle (fifteenth) slice in time as a reference point. Functional volumes were realigned with the first one in the time series to correct for between-scan motion. The structural T1-weighted volume was segmented to extract a gray matter image for each subject, which was spatially normalized to a gray matter image of the MNI (Montreal Neurological Institute) template (MNI; http://www.bic.mni.mcgill.ca/ServicesAtlases/ ICBM152NLin2009). After normalization, all volumes were resampled in 2 2 4 mm voxels using sinc interpolation in space. Finally, the T2*-weighted volumes were smoothed using a Gaussian kernel with 8 mm full-width at halfmaximum (FWHM). 2.3. Statistical analysis A General Linear Model (GLM) analysis was performed at the first single subject level specifying eight regressors for multilinguals coding switch and non-switch trials, and four regressors for monolinguals coding for switch and nonswitch trials. Five contrasts of interest (four for multilinguals and one for monolinguals) were then assessed at the second level. 908 c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 For multilinguals (noun naming in different language contexts) - switching context; - switching context; - switching context; - switching context; into L1 versus L1 non-switch trials in L1eL2 into L2 versus L2 non-switch trials in L1eL2 into L1 versus L1 non-switch trials in L1eL3 into L3 versus L3 non-switch trials in L1eL3 For monolinguals (noun naming in a nouneverb context) - switching to noun naming versus. non-switch noun naming trials. Switching to noun naming matches the response required for L1 naming on switch trials in the multilingual participants. At the second level, a one-way ANOVA model was implemented in SPM5 and two contrast maps were computed: (1) An F-contrast (increases and decreases of activation) coding the differences among the five contrasts above (switching vs non-switch contrasts) with a threshold of p ¼ .019 FDR (False Discovery Rate) corrected at the voxel level. (2) A T-contrast calculating the average positive effect of the five contrasts above (resulting in areas in which an increase of activation was detected for all conditions) with a threshold of p ¼ .029 Family Wise Error (FWE) corrected at the cluster level. Effects of interests for each of the five contrasts were then plotted for the activation cluster of the left caudate (x ¼ 6, y ¼ 16, z ¼ 4) highlighted in contrast one; and for the pre-SMA/ ACC cluster of activation (x ¼ 4, y ¼ 2, z ¼ 56) resulting from contrast two. In order to compute the significance of the effects plotted at the second level, beta values were extracted for each condition included in the second level one-way ANOVA (Analysis of Variance) at the left caudate (x ¼ 6, y ¼ 16, z ¼ 4) and at the pre-SMA/ACC (x ¼ 4, y ¼ 2, z ¼ 56) at the single subject level. Two one-way ANOVA analyses, one for the left caudate and one for the pre-SMA/ACC, were computed on the extracted beta values for each of the five switching conditions. Post-hoc comparisons using Tukey (HSD, honestly significant difference) post-hoc tests were then computed to assess significant differences between the five switching conditions. 3. Results The mean total error percentage for naming in multilinguals was 1.04% (SD ¼ 1.03) for L1, 4.86% (SD ¼ 2.62) for L2 and 27.08% (SD ¼ 3.65) for L3, confirming that L3 was the less proficient language. The mean total error percentage for noun naming (i.e., naming a picture with a noun rather than a verb) in the monolingual group was 5.21% (SD ¼ 1.38). There was no significant relation between AoA and in-scanner naming accuracy (R ¼ .158; p ¼ .590 between L2 AoA and L2 accuracy; and R ¼ .072; p ¼ .806 between L3 AoA and L3 accuracy). However, picture naming in L1 elicited significantly more errors in the monolinguals than in the trilinguals (p ¼ .003, T-test) and we consider the implications in the discussion. Turning to the functional neuroimaging results, the four switching contrasts in multilinguals are reported in Fig. 2, in order to illustrate the areas involved in language control such as the caudate and the pre-SMA/ACC complex. The pre-SMA/ ACC complex was engaged by all four switching conditions in multilinguals whereas the left caudate was more engaged specifically when multilinguals had to switch into L3 (i.e., red color in Fig. 2). The main effect of all differences among the switching contrasts resulted in extensive bilateral caudate activity and two smaller foci in the thalamus. No difference was found in the left prefrontal cortex and the ACC (see Fig. 1B). The average positive effects of all switch contrasts engaged only the pre-SMA/ACC area (see Fig. 1C). As to the plots of the left caudate cluster (i.e., significant differences between the various switching conditions, see Table 1 for beta values), it is important to underline that switching to noun naming in monolinguals differed significantly in the left caudate when compared to the switching conditions in multilinguals (see Table 1 for p-values). It showed decreased response. In multilinguals, caudate activity increased with switching into L2 compared to switching into L1 in the L1eL2 context. Caudate activity increased further when switching into L3 compared to switching into L1 in the L1eL3 context. As reported in Table 1, our HSD Tukey tests reported for this latter comparison a trend of significance (p ¼ .056). The plots of the pre-SMA/ACC cluster (calculated as the area of average positive effect of all switching contrasts), revealed a different pattern (see Table 1 for the beta values). Switching between languages in both contexts in multilinguals and switching into noun naming in monolinguals activated the pre-SMA/ACC to the same degree. 4. Discussion Our study aimed to resolve inconsistent results in the literature on the neural response to language switching. A key determinant of response is likely to be differences in language proficiency but to establish this requires control of other individual differences. Such control is difficult when differences in relative language proficiency are tested in betweengroup studies. We circumvented the problem by studying language switching in trilingual speakers for whom their second and third languages differed in proficiency. We establish that relative language proficiency is a key factor affecting regional neural response during language switching. More specifically, we were able to dissociate responses in two dominant regions of the language control network: the left caudate and the pre-SMA/ACC. Left caudate response varied with proficiency differences but the pre-SMA/ACC did not. The left caudate showed the greatest increase for switching from the most (L1) to the least proficient language (L3). Such an outcome is consistent with the left caudate’s role in selecting the name in the required language in the face of interference from the alternative language. Its relative deactivation when 909 c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 Fig. 2 e Brain activity patterns for the four different switching conditions in multilinguals (p < .001 uncorrected at the voxel level with a cluster extent of k [ 10) superimposed on the brain template. The 2nd level T-maps are overlaid and rendered on the mean structural image of the study sample with MRIcron (http://www.sph.sc.edu/comd/rorden/mricron/). The yellow color indicates switching into L1 (L1eL2 context), dark blue indicates switching into L2 (L1eL2 context), green indicates switching into L1 (L1eL3 context), and red indicates switching into L3 (L1eL3 context). Color mixtures indicate that brain regions were engaged by more than one switching condition. monolingual speakers switched to naming the picture e a more familiar type of response e rather than generating a verb in response to it, indicates a broader role in the resolution of response conflict (Ali et al., 2010). By contrast, switching between languages increased pre-SMA/ACC response regardless of differences in relative proficiency. Such an outcome is consistent with the view that the preSMA/ACC monitors the language context for bilingual or Table 1 e Beta values for each contrast extracted at left caudate [L6 16 4] and pre-SMA/ACC [L4 2 56] at the single subject level. Reported p values are from HSD Tukey tests. The numbers (1e5) in the second column characterize the switching contrast to which the contrast of the first column was compared [1 [ switching into L1 (L1eL2 context); 2 [ switching into L2 (L1eL2 context); 3 [ switching into L1 (L1eL3 context); 4 [ switching into L3 (L1eL3 context); 5 [ switching into nouns (monolinguals)]. (*) [ Significant difference, (^) [ trend of significance. Left caudate e cluster at x ¼ 6, y ¼ 16, z ¼ 4 Mean differences 1. Switching into L1 (L1-L2 context) 2. Switching into L2 (L1-L2 context) 3. Switching into L1 (L1-L3 context) 4. Switching into L3 (L1-L3 context) 5. Monolinguals (switching into nouns) 2 3 4 5 1 3 4 5 1 2 4 5 1 2 3 5 1 2 3 4 .1892088 .2415788 .5109324 .9957076 .1892088 .4307876 .3217235 1.1849164 .2415788 .4307876 .7525112 .7541288 .5109324 .3217235 .7525112 1.5066399 .9957076 1.1849164 .7541288 1.5066399 (*) (*) (^) (^) (^) (*) (*) (*) (^) (*) pre-SMA/ACC e cluster at x ¼ 4, y ¼ 2, z ¼ 56 p-Value Mean differences p-Value .958 .903 .344 .008 .958 .519 .765 .001 .903 .519 .056 .077 .344 .765 .056 .000 .008 .001 .077 .000 .1790382 .0888992 .1819147 .1688550 .1790382 .0901390 .3609529 .0101832 .0888992 .0901390 .2708139 .0799558 . 1819147 .3609529 .2708139 .3507697 .1688550 .0101832 .0799558 .3507697 .954 .997 .951 .969 .954 .996 .610 1 .997 .996 .820 .998 .951 .610 .820 .679 .969 1 .998 .679 910 c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 multilingual speakers as part of a more general role in task monitoring. Conceivably, response of the pre-SMA/ACC and the caudate reflect general task difficulty or, more specifically, differences in the ease of switching between languages in the same or in a different language family. We can reject both alternative accounts of the data. If task difficulty was explanatory, response in pre-SMA/ACC in monolinguals (who showed increased error in picture naming) would be greater than that shown by trilingual speakers in their L1 and less than that revealed by these speakers naming in their L3 (English). In fact as indicated in Table 1 there was no difference. Similarly we should predict increased response in the left caudate in comparison with naming in L1 in trilingual speakers. But the data showed the opposite effect. Increased caudate response when switching into English (L3) from German (L1) relative to switching into Italian (L2) from German (L1) might alternatively reflect a greater difficulty of switching within a language family (German and English are both non-Romance languages) than between language families (Italian is a Romance language). If so, the converse should hold: switching into German (L1) should be more difficult when switching from English (L3) than from Italian (L2). Inspection of Table 1 indicates that switching into L1 in an L3 context does not increase activation more than switching into L1 in an L2 context either for the left caudate or for the preSMA/ACC. On the bases of these data then we have no reason to believe that language family makes a difference to the neural response to switching but it is possible that more extreme differences (e.g., between European and nonEuropean languages) may do so. Our results underline the importance of cognitive control to language use in bilingual and trilingual speakers but the brain regions activated are not special to bilinguals. We presumed their use in monolingual speakers as they switched between noun and verb naming and would expect their recruitment when multilingual speakers perform the same task within each of their languages. The importance of our data is that they show for the first time that differences in language proficiency, defined within participants, differentially modulate activity in core regions of the language control network during language switching. Switching language increases demand on a region associated with monitoring the language in use (pre-SMA/ACC) and switching to the least proficient language increases demand in the region implicated in selecting the language in use (left caudate). Such demands change with proficiency and are likely to underlie the adaptive response to gray matter in these regions (Abutalebi et al., 2012; Zou et al., 2012). The prefrontal cortex revealed no differential effect of language switching conceivably because it plays a role in the sustained inhibition of a language not in use. It would then not be engaged in a context where one or more languages are required as in the present study but would be engaged where a single language was required consistent with recent findings (Parker-Jones et al., 2012; Guo et al., 2011). In summary, by examining differences in relative language proficiency within trilingual speakers we were able to dissociate responses of two core regions of the language control network during language switching. The pre-SMA/ACC region responded comparably to both between-and within-language switches and, within our trilingual speakers, was less sensitive to the effects of proficiency. By contrast, response of the left caudate increased markedly with a switch from the most to the least proficient language. Further work is needed to characterize neural responses to language switching in trilingual speakers for whom the less proficient language is a member of a distinct language family. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.cortex.2012.08.018. references Abutalebi J, Della Rosa PA, Green DW, Hernandez M, Scifo P, Keim R, et al. Bilingualism tunes the anterior cingulate cortex for conflict monitoring. Cerebral Cortex, 22(9): 2076e2086, 2012. Abutalebi J and Green DW. Bilingual language production: The neurocognition of language representation and control. Journal of Neurolinguistics, 20(3): 242e275, 2007. Abutalebi J, Brambati SM, Annoni JM, Moro A, Cappa S, and Perani D. The neural cost of the auditory perception of language switches: An event-related fMRI study in bilinguals. Journal of Neuroscience, 27(50): 13762e13769, 2007. Abutalebi J, Miozzo A, and Cappa SF. Do subcortical structures control ‘language selection’ in polyglots? Evidence from pathological language mixing. Neurocase, 6(1): 51e56, 2000. Ali N, Green DW, Kherif F, Devlin JT, and Price CJ. The role of the left head of caudate in suppressing irrelevant words. Journal of Cognitive Neuroscience, 22(10): 2369e2386, 2010. Crinion J, Turner R, Grogan A, Hanakawa T, Noppeney U, Devlin JT, et al. Language control in the bilingual brain. Science, 312(5779): 1537e1540, 2006. Garbin G, Costa A, Sanjuan A, Forn C, Rodriguez-Pujadas A, Ventura N, et al. Neural bases of language switching in high and early proficient bilinguals. Brain and Language, 119(3): 129e135, 2011. Genzel G, Kerkhoff G, and Scheffter S. PC-gestuetzte Standardisierung des Bildmaterials von Snodgrass & Vanderwart (1980). Neurolinguistic, 9(1): 41e53, 1995. Guo T, Liu H, Misra M, and Kroll JF. Local and global inhibition in bilingual word production: fMRI evidence from ChineseeEnglish bilinguals. NeuroImage, 56(4): 2300e2309, 2011. Hernandez AE. Language switching in the bilingual brain: What’s next? Brain & Language, 109(2e3): 133e140, 2009. Laudanna A, Thornton AM, Brown G, Burani C, and Marconi L. Un corpus dell’italiano scritto contemporaneo dalla parte del ricevente. In Bolasco S, Lebart L, and Salem A (Eds), III Giornate Internazionali Analisi Statistica Dati Testuali. Rome, Italy: CISU, 1995: 103e109. Leech G, Rayson P, and Wilson A. Word Frequencies in Written and Spoken English: Based on the British National Corpus. London: Longman, 2001. Luk G, Green DW, Abutalebi J, and Grady C. Cognitive control for language switching in bilinguals: A quantitative meta-analysis on functional neuroimaging studies. Language and Cognitive Processes, in press. Parker-Jones O, Green DW, Grogan A, Pliatsikas C, Filippopolitis K, Ali N, et al. Where, when and why brain activation differs for c o r t e x 4 9 ( 2 0 1 3 ) 9 0 5 e9 1 1 bilinguals and monolinguals during picture naming and reading aloud. Cerebral Cortex, 22(4): 892e902, 2012. Snodgrass JG and Vanderwart M. A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity and visual complexity. Journal of Experimental Psychology, 6(2): 174e215, 1980. 911 Wang Y, Xue G, Chen C, Xue F, and Dong Q. Neural bases of asymmetric language switching in second-language learners: An ER-fMRI study. NeuroImage, 35(2): 862e870, 2007. Zou L, Guosheng D, Abutalebi J, Shu H, and Peng D. Structural plasticity of left caudate in bimodal bilinguals. Cortex, 48(9): 1197e1206, 2012. All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.