Neuropsychologia 52 (2014) 102–116

Contents lists available at ScienceDirect

Neuropsychologia journal homepage: www.elsevier.com/locate/neuropsychologia

The after-effects of bilingual language production Francesca M. Branzi a,n, Clara D. Martin b,c, Jubin Abutalebi d,e, Albert Costa a,f a

Center for Brain and Cognition (CBC), Universitat Pompeu Fabra, Barcelona, Spain Basque Center on Cognition, Brain and Language, San Sebastian, Spain IKERBASQUE, Bilbao, Spain d Vita-Salute San Raffaele University & Scientific Institute San Raffaele, Milan, Italy e Division of Speech and Hearing Sciences, The University of Hong Kong, Hong Kong, China f Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain b c

art ic l e i nf o

a b s t r a c t

Article history: Received 11 June 2013 Received in revised form 5 September 2013 Accepted 12 September 2013 Available online 18 October 2013

We explored the temporal course of bilingual language control after-effects to shed light on the scope of language control (local vs. global) and on the way in which language control is implemented (L1 inhibition or L2 over-activation). High-proficient bilinguals named objects across three blocks, first in their L1, then in their L2, and then again in their L1 (and conversely) while event-related brain potentials (ERPs) were recorded. Behaviorally we found only the L1 as being hindered by previous naming in the L2. In the ERPs we did not observe inhibitory effects in the N2 component time-window. However, the P2 component showed more positive-going deflections when the previous language slowed down naming latencies of the successive language. The P2 mean amplitude predicted naming latencies whereas the N2 did not. We conclude that in high-proficient bilinguals the P2 component is the marker of language control mechanisms other than inhibition, which are applied globally. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Bilingualism Bilingual language control ERPs Local and global control

1. Introduction The question of how bilingual speakers are able to control their two languages during speech processing has generated a substantial body of research in the last ten years. One of the most used paradigms to investigate bilingual language control has been the language switching paradigm (or mixed naming task) (Hernandez, Martinez, & Kohnert, 2000; Jackson, Swainson, Cunnington, & Jackson, 2001; Costa & Santesteban, 2004; Christoffels, Firk, & Schiller, 2007; Abutalebi et al., 2008; Verhoef, Roelofs, & Chwilla, 2009; Abutalebi et al., 2013; Meuter & Allport, 1999). In a mixed naming task, bilinguals are asked to name some pictures in their first language (L1) and some others in their second language (L2), with the presentation of those pictures mixed. Thus, participants continuously have to switch from one language to the other. So far, this research has been fruitful in highlighting the role of executive functions implicated in bilingual language control. However, there are contexts in which bilingual language control is applied even though bilingual speakers do not switch from one language to the other so frequently. In these situations, that more closely resemble real life, it is likely that the way (the nature of the control mechanisms) and the extent (how broadly these mechanisms operate) to which bilingual language control is applied, is qualitatively different than in mixed naming contexts.

n

Corresponding author. Tel.: þ 1 34 93 5421381. E-mail address: [email protected] (F.M. Branzi).

0028-3932/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.neuropsychologia.2013.09.022

One way to study how and to what extent bilingual language control is achieved in these situations, is to assess the after-effects of naming in one language upon the other language without mixing them. Addressing this issue is the main scope of the present study. Several studies have already been conducted to investigate how performance in one language is affected by the previous use of a different language, without mixing them (e.g., Runnqvist & Costa, 2012; Levy, Mc Veigh, Marful, & Anderson, 2007; Lee & Williams, 2001). For instance, amongst the literature on memory, Levy et al. (2007) showed that naming pictures in L2 negatively affects the subsequent recall of the corresponding L1 translations (the socalled RIF effect across languages, retrieval-induced forgetting). The RIF effect was interpreted as reflecting an inhibitory mechanism that suppresses the strong interference of the L1 lexical entry when the L2 correspondent has to be retrieved from memory. As a consequence of this L1 inhibition during L2 retrieval, the subsequent recall of L1 is hindered. However, the existence of this effect was recently questioned by Runnqvist and Costa (2012). In their study, the authors tested three different groups of bilinguals (low, medium and high-proficient in the L2) in a RIF paradigm similar to the one used by Levy et al. (2007). They found an opposite result as compared to Levy et al. (2007): naming a picture in L2 facilitated the subsequent recall of the translation in the L1, so no RIF effect was present. Thus, the hypothetical inhibition of the L1 during L2 retrieval is still unresolved. The two experiments described above investigated the aftereffect of one language upon the other in a memory task. However,

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

more relevant for the present purpose is a recent study in which the after-effects of one language on the other were measured during a blocked picture naming task (Misra, Guo, Bobb, & Kroll, 2012). Two groups of participants took part in this study: the first one was required to name a set of pictures in the L1 and then the same set of pictures in the L2. The second group instead was required to name a set of pictures in the L2 and then the same pictures in the L1. Hence, given authors were interested in evaluating whether or not L1 production was affected by previous naming in the L2, they compared across groups the same language (L1 or L2) before and after having used the other language. Results revealed facilitation effects when naming in the L2 followed naming the same items in the L1. This facilitation effect may be considered a classical priming effect, occurring when naming a picture that was already presented earlier in the same experiment. Interestingly, an inhibition effect was observed when naming in the L1 was preceded by naming in the L2. This inhibition effect was reflected both by the absence of priming effects in response times (RTs) and by an enhancement of the N2 component observed in event-related potential (ERP) measures. These results have been taken as evidence of persistent inhibition of the L1 during naming in the L2: inhibition has a negative after-effect when naming the same pictures later on in the L1. In summary, the aforementioned study reveals that L1 production is hindered when the same items are previously named in the L2. Nevertheless, there are still two major unresolved issues: (1) whether the hindered L1 naming affects only those specific items that have been named previously in the L2, or whether naming in the L2 hinders subsequent production in L1 in a global way (i.e., whether language control operates just locally or globally), and (2) whether such control is implemented through inhibition of the non-target language or not. These two main questions are the focus of the present study. Before going into the details of the present study, we will briefly expose the current debate on each of these key questions on bilingual language control. 1.1. The scope: Local versus global control Control mechanisms applied during bilingual speech production could occur in at least two different ways. On one hand, language control might be restricted to the task-relevant lexical items (local control). On the other, control mechanisms could affect the entire non-target lexicon (global control; De Groot & Christoffels, 2006). Both processes might be required for efficient language selection and there is some evidence indicating that they are carried out differently. In a recent functional magnetic resonance imaging (fMRI) study, Guo, Liu, Misra, and Kroll (2011) observed the recruitment of different brain systems for global control1 (dorsal left frontal gyrus and parietal cortex) and local control (dorsal anterior cingulate cortex and supplementary motor area). Regardless of the merits of revealing a functional dissociation between the brain areas involved in these two types of control, it is important to notice that in this study the effects related to global control were tested just considering repeated items2. A way to further deepen our knowledge on how broad these global inhibition processes are, is to examine the control effects on items not previously encountered. In the present study, we investigated the scope of bilingual language control by asking high-proficient bilinguals to name repeated and new items in the 1 Note that in Guo et al. (2011) global and local control referred specifically to inhibitory processes. 2 Note that the scope of language control was not addressed either by Misra et al. (2012), who presented participants only with repeated items, thus testing exclusively what we call here local language control.

103

two languages. Through this manipulation, we were able to explore if language control is applied only on the critical items used in a naming block, or if it is exerted on the entire lexicon of the non-target language. 1.2. The mechanisms: Inhibition versus activation Levy et al. (2007) and Misra et al. (2012) interpreted their findings as evidence of L1 inhibition during L2 naming. Indeed, their results find a straightforward explanation in the Inhibitory Control model that explicitly predicts such inhibitory effects (IC model, Green, 1998). The main claim of this model is that, because both languages are active even when naming in only one language, a control mechanism is necessarily operating to suppress or reduce the interference of non-target lexical items (Green, 1998; Hermans, Bongaerts, de Bot, & Schreuder, 1998). The inhibition of lexical representations is supposed to be proportional to the amount of activation and potential interference of a given language. That is, the more interference from the non-target language, the greater the amount of inhibition that needs to be applied. Therefore, naming in the L2 requires a strong inhibition of the L1. Subsequently, naming in the L1 is hindered because it requires more resources to override a strong inhibition (see Meuter & Allport, 1999). However, the cost of naming in the L1 after naming in the L2 might also be interpreted under an alternative account, i.e., the persisting activation account. The persiting activation account (e.g., Yeung & Monsell, 2003; Philipp, Gade, & Koch, 2007) is based on the hypothesis that the stronger language (L1) is normally more active than the weaker one (L2). Thus, the L2 has to be over-activated (relative to the L1) during naming in the L2. When successively naming L1 items, this operation would cause a strong interference because of the carry-over effect of the previously over-activated language (L2; Philipp et al., 2007; for a review, see Koch, Gade, Schuch, & Philipp, 2010). This interference from previous L2 over-activation would increase the time necessary to retrieve the name of the picture in the target language: L1. The fact that two opposite accounts can explain the same phenomenon (slower L1 naming after L2 naming) may make it difficult to consider inhibition as a key feature of bilingual language control. This concern was actually acknowledged by Misra et al. (2012), who discussed that negative effects of a previous L2 naming over L1 naming could be explained by the persistent inhibition of L1 (Green, 1998; Meuter & Allport, 1999), but also by the persisting over-activation of L2 (e.g., Yeung & Monsell, 2003; Philipp et al., 2007). In the present study, we will test these two theoretical alternatives by evaluating not only the after-effects of naming in one language on the successive naming in the other language, but also the after-effects of returning to a previously abandoned language. 1.3. Present study In this study, we explored whether naming pictures in one language exerts after-effects on the successive other language with two purposes: to investigate to what extent bilingual language control mechanisms are applied (locally versus globally) and how these mechanisms are implemented (through inhibition versus activation). We conducted an ERP experiment in which participants were asked to name pictures in the L1 in the first block, and then to name pictures in the L2 in the second block (or conversely). In order to explore local after-effects of language control, we measured naming latencies and ERPs on items of the second block repeated from the first block. In order to explore global after-effects of language control, we measured naming latencies and ERPs on

104

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

new items introduced in the second block, which were not previously named in the first block. After naming in one language (in the first block) and in the other (in the second block), participants had to name a third block of pictures in the language they were using in the first block. This manipulation (third naming block) was added in the experiment in order to: (1) make a within-group comparison of the same language (L1 or L2) before (first block) and after (third block) having used the other language; (2) investigate if bilingual language control operated through inhibition or activation. To be more specific, we had various hypotheses regarding behavioral and ERP results, which were based on previous literature.

1.3.1. The scope: Local versus global control We hypothesized that language control effects would be indexed by modulations of the ERP P2 component, reflecting lexical access processes (Strijkers, Costa, & Thierry, 2010; Strijkers, Holcomb, & Costa, 2011; Strijkers, Baus, Fitzpatrick, Runnqvist, & Costa, 2013). Previous studies demonstrated that this component is a sensitive index of the ease of lexical access (e.g., Costa, Strijkers, Martin, & Thierry, 2009). For example, Strijkers et al. (2010) found more positive deflections for the less accessible lexical representations (L2) as compared to the more accessible lexical representations (L1). Hence, if naming in one language affects negatively the successive language, its lexical access might become harder, which should be reflected by larger P2 amplitudes. Regarding the effects related to item repetition, we based our predictions on the classical observation that repetition of an item during a naming task induces repetition priming effects. It results in faster RTs and more positive shifts (between 200 ms up to 600 ms) in the ERP waveforms, for the second presentation of a given item (e.g., Guillaume et al., 2009; Rugg & Nagy, 1989). Given that visual priming effects could occur in the P2 time-range, it is possible that they interact with those related to lexical access. This interaction could be problematic when interpreting ERP results, because the P2 effects related to “more effortful cognitive strategies” during lexical access are similar to those “facilitatory” effects induced by visual-conceptual priming: in both cases, more positive deflections in the P2 range are expected. However, as in Strijkers et al. (2011), we assumed that if we could find the P2 as related to language control after-effects without repeating items, namely with new items, this would suggest that P2 effects occur, at least in part, independently from those processes directly associated to visual priming. Hence, we propose the hypothesis that if language control is applied locally, only repeated items should be affected and consequently no repetition priming advantage should be observed. In fact, if naming in one language hinders the successive one, we expect that priming facilitation will be eliminated. Regarding new items, we expect them not to be affected by previous naming in the other language, if control is only local. On the other hand, if control mechanisms are global, we expect repeated and new items to be affected similarly. Precisely, if naming in one language affects globally the access to the other language, on one hand, we expect new items to be named more slowly and to elicit more positivegoing P2 amplitudes in the second naming block than in the first one. On the other hand, for repeated items we expect no repetition priming facilitation in naming latencies, along with more positivegoing P2 amplitudes. In summary, the P2 effects should be observed for both new and repeated items if control is applied globally, whereas just for repeated items if control is applied only locally. Importantly, L1 and L2 might show different after-effects of language control. In fact, if L1 only is hindered by previous naming

in the other language (e.g., Levy et al., 2007; Misra et al., 2012), we should observe hindered naming latencies and more positivegoing P2 amplitudes for L1 only, when comparing the second with the first block of naming. 1.3.2. The mechanisms: Inhibition versus activation Finally, we specifically investigated the after-effects of recovering a previously abandoned language, because this manipulation should generate qualitatively different effects depending on the nature of the control mechanisms involved (see Koch et al., 2010). In the switching literature the effects of recovering a previously used task have been evaluated widely through the so called "n-2 repetition cost"3. What is important for our purpose is that persistent inhibition (e.g., Meuter & Allport, 1999) and persistent activation accounts (e.g., Yeung & Monsell, 2003) make opposite predictions for language recovery after-effects. According to the persistent activation account (e.g., Yeung & Monsell, 2003), language recovery should be facilitated (repetition priming), because it implies recovering a ‘residual activation’. On the contrary, the persistent inhibition account predicts that recovering a previously inhibited language should induce an increase of RTs, due to the need of overcoming a ‘residual inhibition’ (e.g., Mayr & Keele, 2000; Gade & Koch, 2007). Based on these accounts, the following predictions can be made: if inhibition of the L1 is the control mechanism applied during bilingual language production (Misra et al., 2012), naming in the third block will imply overcoming inhibition. It should result in slower RTs as compared to naming in the first block. If language control does not involve inhibitory processes, naming in the third block will imply recovering a language-set which was previously activated (L2 over-activation). Thus, facilitation (or at least no cost) in L2 naming latencies should be observed in the third block as compared to the first one. Regarding the ERP effects, we expected the P2 to be larger if L1 recovery implies overcoming inhibition. On the contrary, we predicted no variation (or reduced positivity) in the P2 amplitude between the first and the third block of L2 naming, if language control requires recovering a previously activated language-set. For the sake of completeness, we included a general hypothesis on the modulation of another ERP component by language control. In fact, if inhibitory control is applied during bilingual language production, a modulation of the N2 ERP component should be observed. The N2 component has been described as an index of general control processes (Nieuwenhuis, Yeung & Cohen, 2004) and several authors focused on this component in order to study bilingual language control effects related to inhibition (Christoffels et al., 2007; Jackson et al., 2001; Verohef et al., 2009; Misra et al., 2012; Khateb et al., 2007). Hence, N2 component modulations should be observed in the present experiment if inhibitory processes are implicated in bilingual language control.

2. Material and methods 2.1. Participants Forty early and high-proficient Catalan/Spanish bilinguals took part in the experiment and received monetary remuneration for their participation. All participants were right-handed and had normal or corrected-to-normal vision. Two of the participants in each group were excluded from the individual data sets due to movement artifacts during ERP recording. Therefore, the analyses reported below refer to 36 participants (25 females; mean age¼ 21.3 years 7 1.9). Language use history and self-assessed proficiency for all participants are reported in Table 1.

3 This cost is calculated in contexts where three tasks have to be alternated continuously and it is measured by the difference between n-2 repetition (e.g., ABA sequences) and n-2 switch trials (e.g., CBA sequences).

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

105

Table 1 Language use history and the self assessed proficiency. Language use scores represents a mean proportion (max. score¼1, min¼ 0) of languages' use in different periods of life: Preschool (from 0 to 5/6 years), Primary Education (from 5/6 to 12 years), Secondary Education and High school (from 12 to 18 years) and Adulthood (from 18 to the actual age). Language proficiency scores are on a 5 point scale, in which 5 represents a very high level and 1 a very low level of proficiency. The self-assessed index is the average of participants' responses for each domain (reading, writing, speaking and comprehension). L1

L2

Preschool Primary Education Secundary Education and High school Adulthood

Language use 0.7 (0.2) 0.7 (0.2) 0.7 (0.2) 0.6 (0.2)

Language use 0.4 (0.2) 0.4 (0.2) 0.4 (0.2) 0.4 (0.2)

Reading Writing Speaking Comprehension

Language proficiency 5 (0.2) 5 (0.4) 5 (0.3) 5 (0.2)

Language proficiency 5 (0.2) 5 (0.4) 5 (0.3) 5 (0.2)

We randomly assigned participants to two groups: “Start in L1” and “Start in L2” (18 participants in each group). The first group (Start in L1) performed the L1-L2-L1 version of the naming task while the second group (Start in L2) did the L2-L1-L2 version. 2.2. Materials One hundred and twenty-eight line-drawings of objects were selected from different databases (Snodgrass & Vanderwart, 1980; International Picture Naming Project, Szekely et al., 2004). The pictures represented objects belonging to a wide range of semantic categories (e.g., animals, body parts, buildings, furniture). Since the objects had to be named in Spanish and Catalan, the cognate status4 of the corresponding names was controlled, resulting in half of the items being Spanish/ Catalan cognates and the other half being non-cognates. The mean lexical frequency (LEXESP; Sebastián-Gallés, Martí, Carreiras & Cuetos, 2000) of the picture names was similar for cognates and non-cognates (non-cognates: 1.03 7 0.6; cognates: 1.147 0.6; t(1 2 6) ¼  1.075, p¼ 0.284). Pictures were grouped together in order to create four experimental lists, which were then randomized across participants. Each participant performed a picture-naming task in three consecutive blocks (start, switch, return). Within each single block, the language to be used for picture naming remained the same. The language changed from the first block (start block) to the next (i.e., switch block), and from the second block (switch block) to the third block (return block). In this manner, start and return blocks were named in the same language and switch block was named in the other language, resulting into two distinct series of blocks (i.e., L1-L2-L1 and L2-L1-L2). Importantly, in the start block, participants had to name a set of 64 pictures. The switch block included 32 repeated pictures (which were already presented in the start block) and 32 new pictures (randomly presented). The return block included 32 repeated pictures (which were the 32 new pictures of the switch block) and 32 new pictures (randomly presented). Trial order within each block was randomized. 2.3. Experimental procedure After having filled in the informed written consent and a language use/ proficiency questionnaire, each participant was tested individually in a soundproof room. Written instructions were presented in the language in which the participants had to start the naming task5. Participants were told they had to name pictures aloud as fast as possible in the language indicated by a language-cue sentence, presented at the beginning of each block (“Nombra estos dibujos en Español” or “Anomena aquests dibuixos en Catalan”—name these pictures in Spanish/ Catalan). Participants were told that the language cue would change across but not within blocks. As in previous studies (e.g., Misra et al., 2012; Guo et al., 2011), our participants were not familiarized with picture names in order to not create a bias in repetition effects across blocks. This was based on the assumption that the less

4 Here we consider cognates to be translation words that have similar phonology in the two languages (e.g.: “cocodrilo” in Spanish, “cocodril” in Catalan; meaning “crocodile” in English). Instead, non-cognate words have the same meaning but distinct phonology (e.g.: “calcetín” in Spanish, “mitjó” in Catalan; meaning “sock” in English). 5 Note that written instructions were provided in the language in which the task had to be initiated to introduce participants in the target language context. These instructions were detailed and extensive and the participants rarely asked for clarification. Hence, the interactions between participants and experimenter were minimal, and always in Spanish.

dominant language (L2) is likely to benefit more from the repetition than the more dominant one (L1; see Hernandez & Reyes, 2002). Stimuli were presented by means of Presentation software (Neurobehavioral systems: http://www.neurobs.com/) and vocal response latencies were recorded from the onset of the stimuli. Each trial began with a blank screen for 1000 ms, the picture to be named appeared for 1500 ms at the centre of the screen on a black background. Then, a fixation cross was presented for 500 ms. Participants had to name each picture as fast as possible during its presentation (the picture remained on the screen even after naming). Overall, the experimental session lasted approximately one hour and a half, including the electrodes' placement and removal.

2.4. Behavioral data analysis 2.4.1. Error analysis Any item belonging to the following categories was considered an “error”: (1) productions of incorrect names6; (2) verbal disfluencies (stuttering, utterance repairs, productions of non verbal sounds) (1%, SD ¼ 71); (3) recording failures (8,6%, SD ¼ 7 5); (4) naming inconsistencies (repeated items which were named incorrectly in the previous block) (in “others” see foot note n.6); (5) naming latencies which were three standard deviations above or below each individual's mean value (outliers) (0,8%, SD ¼ 7 1); (6) naming latencies faster than 500 ms (for motor-speech artifact free ERP analysis) (0,3%, SD ¼ 71). Overall, we identified 23% (SD¼ 7 5) of “errors” [(24% (SD ¼ 7 5) of “errors” in the group “Start in L1” and 22% (SD¼ 7 6) in the group “Start in L2” across all conditions (groups were not different in the percentage of errors; t(34) ¼.877, p ¼0.387)]. This high rate of “errors” is likely to be due to the highly restrictive criteria adopted to exclude trials. Behavioral and ERP analyses were conducted on the sub-set of error free data.

2.4.2. Naming latencies analysis Two series of analyses were performed on mean naming latencies. First, we explored the after-effects of bilingual language control by comparing the first two blocks of naming between groups. We performed a 2  2 ANOVA with Group (Start in L1 vs. Start in L2) and Language (L1 vs. L2) as variables. This analysis was performed on repeated items and on new items separately. Second, we explored the after-effects of language recovery within groups by comparing the first block (start) with the third one (return). With this comparison we aimed to test the effect of a previous language on the successive one, within groups. Hence, we compared the same language after (return block) and before (start block) the other language. Given in this case we were not interested in comparisons across groups, we run two separate (one for L1 recovery and one for L2 recovery) repeated measures ANOVAs (1  2) with Block (start vs. return) as variable. In this analysis, only new items from the return block were considered, because repeated items from this block were previously named in the switch block and not in the start block. All behavioral data were logarithm-transformed to achieve a normal distribution before statistical analysis. In the ANOVAs' Post-hoc analyses, we consistently applied the Bonferroni correction for multiple comparisons.

6 By “incorrect names” we mean name disagreement with the “standard” label of the picture: “synonyms” (i.e., cup of tea for cup) (3.3%, SD ¼ 7 1), “coordinates/ different exemplar of the same category” (i.e., girl for woman) (3.7%, SD ¼ 7 2), “super-ordinates” (i.e., insect for beetle) (1%, SD ¼ 71), “cross-language errors” (0.7%, SD ¼ 7 1) and “others” ( incorrect names not attributable to the other categories of errors) (3.7%, SD ¼ 7 2).

106

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

2.5. Electrophysiological recording and analysis Electrophysiological data was recorded (Brain Vision Recorder 1.05; Brain Products) from 38 tin electrodes placed according to the 10–20 convention system. An electrode placed on the tip of the nose was used as a reference. Two bipolar electrodes were placed next to and above the right eye to register ocular movements. Two other electrodes were placed on the right and left mastoids for off-line re-referencing. Electrode Impedances were kept below 5 kΩ and EEG signal was recorded with a high cut-off filter of 200 Hz, with a sampling rate of 500 Hz. The EEG activity was filtered off-line using a 20 Hz high-pass filter and a 0.01 Hz low-pass filter (24 dB/octave). Eye blink artifacts were corrected using the procedure of Gratton and Coles (1989). The remaining artifacts (amplitude changes above or below 200 μV within 200 ms before and after stimulus event) were removed applying an automatic inspection on raw EEG data. Similar to other language production studies employing ERPs (for a review see Ganushchak, Christoffels, & Schiller, 2011), we followed the assumption that analyzing the ERP signal before the time of the actual response would lead to artifact-free ERP data. Therefore, EEG epochs were segmented from  100 to 500 ms with respect to picture onset, in order to avoid artifacts of speech contamination (e.g., Wohlert, 1993; Masaki, Tanaka, Takasawa, & Yamazaki, 2001). This segmentation procedure enables the avoidance of motor artifacts due to fast responses and latency jitter induced by muscular activity (e.g., Strjikers et al., 2010). EEG epochs corresponding to the same condition were averaged for each participant. Baseline correction was performed in reference to pre-stimulus activity [  100–0 ms] and individual averages were digitally re-referenced to the averaged mastoids. For each condition, the grand average was obtained by averaging individual averages, in order to obtain and analyze Event-Related Potentials (ERPs). In order to compare our results with previous literature on electrophysiological counterparts of language control (i.e., Strijkers et al., 2010, 2011, 2013; Misra et al., 2012; Christoffels et al., 2007; Jackson et al., 2001), we explored ERP component mean amplitudes. For ERP mean amplitude analyses, we selected a cluster of 12 electrodes centered on midline and lateral sites over three scalp regions: Midline (Fz, FCz, Cz, Cpz), Left Frontal-Central (F7, F3, FC3, C3), Right Frontal-Central (F8, F4, FC4, C4). Electrode's selection was based on the topographical distribution of the effects of interest. The mean amplitude of the ERP was calculated by considering two temporal windows: the first from 170 to 300 ms (P2 component) and the second from 300 to 500 ms (N2 component). We selected these two temporal windows by means of a careful visual inspection of the Grand Averages, and in agreement with previous evidence reporting language control effects in similar time-windows and over similar regions of the scalp (Misra et al., 2012; Strjikers et al., 2010; Christoffels et al., 2007) 7 . Importantly, these two different time-windows were centered on peaks visible in the global field power (GFP) (Lehmann & Skrandies, 1980; Picton et al., 2000) averaged over all subjects (Start in L1 and Start in L2), conditions (start, switch new, switch repeated, return new) and over the selected twelve electrodes. Similarly as in previous studies (e.g., Strijkers et al., 2013; Martin, Barceló, Hernández, & Costa, 2011), we run an automatic peak detection on the GFP that revealed two peaks: the first one, positive, peaking at 230 ms and the second one, negative, peaking at 430 ms. The two peaks were part of the two time-windows selected for statistical analyses. Moreover, 2-tailed paired t-tests were run separately for the two groups for each comparison of interest (Fig. 1a and b. start vs. switch repeated; Fig. 1c and d. start vs. switch new; Fig. 1e and f. start vs. return new) at each sampling point (2 ms) starting from target presentation (0 ms) until the end of the epoch (500 ms). This was done in order to establish the onset of the first temporal window of interest (P2) in which each condition started to diverge significantly (i.e., when at least a sequence of 10 consecutive t-test samples exceeded the 0.05 significance level 8). This analysis was done to have an overall view of the potential differences in latency between the groups and to allow the selection of the more appropriate time-windows (for the two groups) on which running the corrected repeated measures ANOVAs. Overall, this methodological procedure was adopted in order to have two objective measures (automatic peak detection on the GFP and paired-t-test) to determine the time-windows and ensure a limitation of the probability of type 1 errors (see Kilner, 2013). The same ANOVAs previously described in the behavioral analyses were performed to analyze ERP mean amplitudes in the

7 Note that we selected a late time-window to explore the N2 component (300–500 ms). This time-window was congruent with previous studies on language control (cf. Christoffels et al., 2007; Jackson et al., 2001; Misra et al., 2012), but other studies reported earlier time windows for the N2 component (e.g., Khateb et al., 2007). Importantly, the scalp distribution and topography of the component we observed were clearly the ones of the N2 component (see Jodo and Kayama, 1992; Pfefferbaum, Ford, Weller, & Kopell, 1985). We do not discuss the variability in the N2 peak latency since it is out of the scope of the study, but further research is needed to explore which factors modulate the latency of this component. 8 This was decided following Guthrie and Buchwald's (1991) proposed approach to avoid false positives.

two time-windows, centered on peaks visible in the GFP averaged over all subjects, conditions and electrodes of interest. In the ANOVAs' Post-hoc analyses we consistently applied the Bonferroni correction for multiple comparisons. 2.6. Multiple linear regression analysis between naming latencies and ERP components (P2, N2) A multiple linear regression analysis was run in order to evaluate whether the P2 and/or the N2 mean amplitudes could predict behavioral responses across all conditions. Naming latencies across all conditions were considered as the dependent factor and the P2 and N2 mean amplitudes as predictors. For each component the values of the mean amplitude was derived by averaging the values of the 12 electrodes on which we ran the statistical analysis.

3. Results 3.1. Behavioral data: Naming latencies analysis Mean naming latencies are reported in Fig. 2 and in Table 2. 3.1.1. START vs. SWITCH (repeated): L1 and L2 between groups In this analysis we aimed to assess whether naming pictures in one language affected the subsequent production of these same pictures in a different language. Hence, we considered naming latencies in the start block and those for the repeated items in the switch block, both when participants started naming in L1 and when participants started naming in L2. Hence, the comparison between naming latencies in each given language was between different participants. That is, the comparison of naming in the L1 in start block with naming in the L1 in switch block and naming in the L2 in start block with naming in the L2 in switch block was between subjects. The ANOVA performed on naming latencies revealed that the main effect of Language was significant [F(1, 34) ¼ 5.019, p ¼0.032], indicating that naming latencies were slower in the L2 than in the L1. The main effect of Group was not significant [F(1, 34) ¼1.933, p ¼0.173] but the Language  Group interaction was significant [F(1, 34) ¼13.120, p ¼0.001]. This interaction revealed that those participants that named the pictures in L2 in the start block were slower than those participants that named the pictures in L2 after having named them in the L1 (switch block) (p ¼0.011). Hence, naming in the L1 facilitated subsequent naming in the L2. Conversely, those participants that named pictures in the L1 in the start block were as fast as those that did so in the switch block (p ¼0.985). Hence, naming in the L2 did not seem to facilitate subsequent naming in L1. 3.1.2. START vs. SWITCH (new): L1 and L2 between groups In this analysis we aimed to assess whether naming pictures in one language affects the subsequent production of new pictures in a different language. Hence, we considered naming latencies for the new items in the start (indeed, by definition all pictures were new in this start condition) and switch blocks, both when participants started naming in the L1 and when participants start naming in the L2. Hence, as in the previous analysis, the comparison between naming latencies in each given language was between different participants. The ANOVA revealed a significant main effect of Language [F (1, 34) ¼ 5.621, p ¼0.024], indicating that naming in the L2 was slower than naming in the L1. The main effect of Group was not significant [F(1, 34) ¼1.719, p ¼0.199]. We also observed a significant Language  Group interaction [F(1, 34) ¼5.818, p ¼0.021]. This interaction reveals that naming latencies in the L1 were shorter for those participants that start naming pictures in the L1, than for those that start naming pictures in

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

107

Fig. 1. Paired t-test. (a) start L1 vs. switch L2 (repeated); (b) start L2 vs. switch L1 (repeated); (c) start L1 vs. switch L2 (new); (d) start L2 vs. switch L1 (new); (e) start L1 vs. return L1 (new) and (f) start L2 vs. return L2 (new).

the L2 and then switch into L1 (p ¼0.042), while naming latencies in the L2 were not affected by whether participants started with the L1 or not (p ¼ 0.825).

The results of these two analyses showed a very interesting pattern of results. First, for repeated items, naming in the L2 was sensitive to whether such items have been already named in the

108

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

Fig. 2. Naming latencies. Mean naming latencies in each block (START, SWITCH, RETURN) are reported separately for the two groups (Start in L1, Start in L2). Error bars are referred to standard errors (SE).

Table 2 Naming latencies. Mean naming latencies in each block (START, SWITCH, RETURN). In parentheses are reported the standard errors (SE). Subscripts near to means indicate to which group of participants (Start in L1 or Start in L2) the data belong.

those in the start block. This result suggests that L2 recovery was not affected by previous naming in the L1 (switch block). 3.2. ERP data

L1 START SWITCH repeated SWITCH new RETURN repeated RETURN new

835StL1 836StL2 917StL2 824StL1 916StL1

L2 (25) (29) (29) (29) (34)

913StL2 819StL1 909StL1 852StL2 931StL2

(22) (29) (30) (27) (31)

L1 or not. However, naming in the L1 was not affected by this factor. Second, for new items, naming in the L2 was not sensitive to previous exposure to the L1, while naming in the L1 was indeed affected by previous use of the L2. However, these two analyses have one important shortcoming inherent to the blocked design, namely the fact that the comparisons for a given language (L1 or L2) are necessarily between subjects. We managed to reduce the potential influence of individual differences on the observed results by conducting the following analysis. 3.1.3. START vs. RETURN blocks, new items In this analysis we aimed to assess whether naming pictures in one language affects the subsequent production of new pictures in a different language that has been previously used. However, unlike previous analyses, here we focused on within subjects performance, by comparing the naming latencies in the start block, with those in the return block. It is of note that responses in these two blocks are given by any given participant in the same language. The difference between the two blocks then is that between the start and the return blocks, participants are asked to name pictures in a different language (the switch block). To avoid potential confusions, we considered here only new items. That is, the items that in the return blocks were not named neither in the start nor in the switch blocks. 3.1.3.1. L1 recovery. In this analysis the main effect of Block [F(1, 17) ¼9.492, p¼ 0.007] was highly significant, indicating that naming latencies in the return block were significantly slower than those in the start block. This result suggests that L1 recovery was negatively influenced by previous naming in the L2 (switch block). 3.1.3.2. L2 recovery. In this analysis the main effect of Block [F(1, 17) ¼0.469, p ¼0.503] was not significant, indicating that naming latencies in the return block were not different from

ERP Grand Averages relative to naming in the L1 start vs. L1 switch repeated/new are presented in Figs. 3 and 4. ERP Grand Averages of naming in L2 start vs. L2 switch repeated/new trials are reported in Figs. 5 and 6. Lastly, in Figs. 7 and 8 are reported the ERP Grand Averages for naming in start vs. return new blocks for the two groups (Start in L1 and Start in L2) separately. Visually, all conditions revealed similar patterns until the first negative peak (N100), that was clearly observable at frontal and central sites. After this first negative peak, a positive waveform was visible, peaking around 200 ms. Successively, the positivity turned into a negative-going shape until the end of the epoch (500 ms).

3.2.1. START vs. SWITCH (repeated): L1 and L2 between groups In the ERP analysis for the P2 time-window (170-300 ms), the main effect of Language showed a trend towards significance [F(1, 34)¼ 3.072, p¼ 0.089], suggesting an overall larger positivity for L2 naming. The main effect of Group was not significant [F(1, 34) ¼ 0.105, p ¼0.748]. Interestingly, the Language  Group interaction was highly significant [F(1, 34) ¼11.685, p ¼0.002]. Post-hoc analyses revealed differences that were almost significant: the P2 was larger for L1 in switch block as compared to L1 in start block (p ¼0.07). Conversely, the P2 amplitude for L2 in switch block was not different from that of L2 in start block (p¼ 0.18). In the ERP analysis for the N2 time-window (300 to 500 ms), the main effect of Language [F(1, 34)¼ 2.386, p ¼ 0.132] and the main effect of Group [F(1, 34) ¼1.46, p ¼0.235] were not significant. Interestingly, the Language  Group interaction was highly significant [F(1, 34) ¼ 27.778, p o0.001], indicating that the N2 component was less negative for L1 in switch block as compared with L1 in start block (p ¼0.004). Conversely, the N2 component was not reduced for L2 in switch block as compared to L2 in start block (p ¼0.427). 3.2.2. START vs. SWITCH (new): L1 and L2 between groups In the ERP analysis for the P2 time-window (170–300 ms), we did not observe any significant main effect of Language [F(1, 34) ¼ 0.719, p ¼ 0.402] or Group [F(1, 34) ¼0.808, p ¼0.375]. The Language  Group interaction was significant [F(1, 34) ¼4.41, p¼ 0.043], indicating that the P2 component was larger for L1 in switch block as compared to L1 in start block (p ¼0.08). Conversely, the P2 amplitude was not different for L2 in switch block as compared to L2 in start block (p ¼0.78).

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

109

Fig. 3. L1 start vs. L1 switch repeated. Grand Average ERPs for naming in the L1 in the first block (START, Group Start in L1) and in the second block (SWITCH repeated, Group Start in L2). Negative is plotted up.

In the ERP analysis for the N2 time-window (300–500 ms), the main effect of Language was not significant [F(1, 34) ¼1.167, p ¼0.288], as well as the main effect of Group [F(1, 34) ¼2.126, p ¼0.154]. The Language  Group interaction was marginally significant [F(1, 34) ¼ 3.84, p ¼0.058], indicating that the N2 component was less negative for naming in the L1 in switch block as compared to L1 in start (p ¼0.059). Conversely, the N2 was not reduced for L2 in switch block as compared to L2 in start block (p ¼0.48).

3.2.3.2. L2 recovery. In the ERP analysis for the early temporal window (170–300 ms), the main effect of Block was not significant [F(1, 17) ¼2.743, p ¼0.116], suggesting L2 recovery does not affect the P2 time-window. In the ERP analysis for the N2 time-window (300–500 ms), the main effect of Block was not significant [F(1, 17) ¼0.521, p ¼0.48]. This result reveals that the L2 recovery does not affect the N2 time-window.

3.2.3. START vs. RETURN blocks, new items 3.2.3.1. L1 recovery. In the ERP analysis for the early temporal window (170–300 ms), the main effect of Block came close to significance [F(1, 17) ¼3.707, p ¼ 0.071], suggesting that the effects of language recovering are already present in the P2 time-window. Specifically, we observed more positive P2 deflections during L1 recovery. In the ERP analysis for the N2 time-window (300–500 ms), we found a significant main effect for Block [F(1, 17) ¼4.915, p¼ 0.041]. This result confirms the visible attenuation of the N2 component for the L1, when it must be recovered after naming in the L2.

3.3. Multiple linear regression analysis between naming latencies and ERP components (P2, N2) In order to characterize the relationship between both the P2 and N2 components with respect to behavioral measures, we carried out a multiple regression analysis. Results indicated that overall the ERP amplitudes explain a significant proportion of naming latencies' variance [R2 ¼.048, F(2, 141)¼3.561, p¼ 0.031] (see Fig. 9). Moreover, the P2 mean amplitudes predicted significantly naming latencies [beta¼0.254, t(1 4 1)¼2.66, p¼ 0.009], while the N2 mean amplitudes did not [beta¼  .112, t(1 4 1)¼ 1.17, p¼ 0.244].

110

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

Fig. 4. L1 start vs. L1 switch new. Grand Average ERPs for naming in the L1 in the first block (START, Group Start in L1) and in the second block (SWITCH new, Group Start in L2). Negative is plotted up.

4. Discussion In the present study, we investigated the after-effects that one language may exert upon the successive use of the other language for two reasons. First, to explore whether the after-effects of language control occurs only for lexical items already activated in the previous language (local control), or rather if they are extended to the whole non-target lexicon (global control). Second, to shed light on the way in which language control is implemented (inhibition vs. activation). We investigated the time course of these local and global effects with a particular emphasis on the P2 component of the ERPs, that is considered a marker of lexical access in bilingual speech production (e.g., Strijkers et al., 2010, 2011, 2013) and the N2 component which, in turn, is related to inhibitory processes (e.g., Jodo and Kayama, 1992; Pfefferbaum et al., 1985; Jackson et al., 2001). We hypothesized that if control occurs only locally, the language control after-effects should influence only repeated items. On the other hand, if control occurs globally, we would expect to see repeated and new items being affected similarly by language control processes. We tested the after-effects of returning to a previously ‘abandoned’ language in order to explore whether putative behavioral and ERP differences for

naming in the L1 after the L2 may be explained by means of the L1 inhibition, or rather by an over-activation of the L2. 4.1. The scope: Local versus global control We observed similar behavioral and ERP effects for both repeated and new items, suggesting that language control is likely applied globally. Interestingly, the patterns of these effects were asymmetrical, for both repeated and new items. In other words, when bilinguals named repeated items in the L2, that were previously presented in the L1, the naming was facilitated. Indeed, verbal responses were faster as compared to the items named in the L2, without prior presentation in the L1 (i.e., when the L2 naming occurred in the starting block). The same facilitation did not occur for the L1. Indeed, when naming pictures in the L1, after having named them in the L2, latencies were not faster relative to latencies of the L1 naming in the starting block. Similarly, for new items we found that the L1 was negatively affected by the previous naming in the L2, with the L1 being slower in switch block than in start block. Conversely, we did not observe any increase of naming latencies for the L2 naming in switch block, as compared to the L2 naming in start block.

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

111

Fig. 5. L2 start vs. L2 switch repeated. Grand Average ERPs for naming in the L2 in the first block (START, Group Start in L2) and in the second block (SWITCH repeated, Group Start in L1). Negative is plotted up.

It is noteworthy that Misra et al. (2012) reported a similar asymmetry in naming latencies with repeated items. This result has been explained by suggesting that inhibition mechanisms applied on the L1 (during naming in the L2) may have erased priming facilitation for naming latencies. However, the authors recognized that the same result could also be explained by assuming that the less dominant language (L2) benefits more than the dominant one (L1) from an initial picture-language binding (L2 over-activation). Following this hypothesis, the L1 would be disadvantaged from the resulting competition during later naming blocks, and the repetition priming effect would be observed only in the L2. Consistent with this alternative interpretation is the observation we made that the N2 component did not show inhibitory effects (enhanced negativity) modulated by language order. This was the case for repeated but also for new items. Based on the common assumption that the N2 component reflects inhibitory mechanisms (e.g., Misra et al., 2012; Christoffels et al., 2007; Jackson et al., 2001; Khateb et al., 2007), we can assume that no inhibition was at play in our study. Rather, in the present study the effects of language control appeared earlier than the N2, that is, in the P2 time-window. Our prediction was that language order may modulate the P2 component, which reflects the ease of lexical access (e.g., Strijkers et al., 2010, 2011, 2013). The basic idea was

that if a previously used language could hinder the lexical access of the successive language, we would expect larger P2 effects for this latter as compared to the same language without prior use of the other. Our results are in line with this prediction, since we found a larger P2 for L1 (switch block) after the L2, relative to the L1 first (start block). Moreover, we found this P2 modulation for both repeated and new items, suggesting that P2 effects for the L1 reflect control mechanisms applied globally. This result is particularly important because it demonstrates that the effects related to the P2 for repeated items did not depend from those processes directly associated to visual priming. In fact, given the ERP effects related to item repetition are positive shifts that could start between 200 ms up to 600 ms (e.g., Guillaume et al., 2009; Rugg & Nagy, 1989), it is possible that there are confounds between priming effects and the effects related to lexical access in the P2 time-range. However, in line with a recent study (Strijkers et al., 2011), we found that the P2 component was sensitive to language control after-effects without the need of repeating items (with new items), hence confirming that the P2 effects observed for local control may occur, at least partly, independently from those elicited by visual priming. Furthermore, object recognition and conceptual processing during L1 and L2 naming should be similar (at least for concrete objects in early and high-proficient

112

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

Fig. 6. L2 start vs. L2 switch new. Grand Average ERPs for naming in the L2 in the first block (START, Group Start in L2) and in the second block (SWITCH new, Group Start in L1). Negative is plotted up.

bilinguals), hence it would be hard to explain the differences in the P2 effects between the L1 and the L2 in terms of visual-conceptual effects instead of lexical access effects. Finally, additional support for the fact that P2 is at the origin of the after-effects of language control comes from the regression analysis where we found P2 mean amplitudes positively correlated with naming latencies. In other words, the more positive the P2 deflections across all conditions, the slower the naming latencies. As to the results relative to the N2 component effects, in the present study, naming in the switch block produced less negative N2 deflections than naming in the start block, in the case of L1 naming. Again, this was observed for both repeated and new items. To the extent to which inhibitory effects should be seen in the N2 time-window (enhanced negativity), our ERP results might be seen at odds with those of Misra et al. (2012). In fact, on one side we found behavioral asymmetries for repeated and new items, suggesting that the L1 is globally hindered by previous naming in the L2. However, on the other side, we found reduced N2 amplitudes for the L1 in switch block as compared to the L1 in start block. Given these results and those from the regression analysis, it is very likely that the reduced negativity we observed in the N2 time-window for the L1 is an after-effect of early processes starting in the P2 time-window.

One possible way to explain the absence of inhibitory effects in our ERP data is to assume that the nature of language control mechanisms could be influenced by L2 proficiency. In Misra et al. (2012) participants were low-proficient bilinguals. As in the case of switch costs in mixed naming tasks (e.g., Meuter & Allport, 1999) the authors found a similar asymmetry interpreted as stemming from inhibitory control on the L1. Conversely, in our study participants were high-proficient bilinguals. If the magnitude of the N2 negativity reflects the activity required for response inhibition, the absence of these effects in our data is consistent with the evidences suggesting that high-proficient bilinguals control their languages through mechanisms other than inhibition (Costa et al., 2004; Costa, Santesteban, & Ivanova, 2006). In line with this idea, no P2 effect was found in low-proficient bilinguals (Misra et al., 2012) and we could hypothesize that, in the present study, the P2 effects reflect these language-specific selection mechanisms applied during the early stage of lexical access. These mechanisms would be applied during lexical access to cope proactively with the persisting activation of the memory trace of the previous language, which is stronger for the L2 (for a review, see Koch et al., 2010). Thus, persistent L2 activation would carry over and interfere with the subsequent naming in the L1 more than the other way round. Later on, the control of this interference

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

113

Fig. 7. L1 recovery. Grand Average ERPs for naming in the L1 in the first block (START, Group Start in L1) and in the third block (RETURN new, Group Start in L1). Negative is plotted up.

would affect the P2 component and the time necessary to retrieve the name of the picture, affecting the L1 more than the L2. Testing the present experimental design in a group of late and lowproficient bilinguals would provide further evidence to validate these assumptions. All in all, the assumption that language control differs qualitatively between high and low proficient bilinguals is in line with recent functional magnetic resonance imaging (fMRI) evidence (e.g., Garbin et al., 2011) showing that language switching in early and high-proficient bilinguals involves different brain networks from those involved for low-proficient bilinguals (e.g., Wang, Xue, Chen, Xue, & Dong, 2007). However, in order to understand better the nature of the mechanisms of bilingual language control, we examined language recovery effects. 4.2. The mechanisms: Inhibition versus activation As aforementioned, by adding a third block of language we were able to compare, within-groups, naming in one language before and after having used the other language. Testing the effects of recovering a recently abandoned language is informative on the presence of inhibitory control processes, given that persisting activation and persisting inhibition accounts predict opposite

outcomes (see Koch et al., 2010). Interestingly, the behavioral patterns obtained from the comparison between start and return blocks were different for the L1 and the L2. Precisely, L1 recovery induced a behavioral cost, whereas L2 recovery did not. One possible interpretation could be that the behavioral cost of L1 recovery would derive from inhibitory control. In other words, because the L1 was inhibited during L2 naming in the switch block, overcoming this inhibition would have induced an increase in naming latencies. However, we did not observe any such inhibitory effect reflected by the N2 component. Instead, the P2 showed more positive deflections during L1 recovery. Conversely, L2 recovery did not induce any significant effect on the P2 component. The enhanced P2 effect for L1 recovery had an after-effect on the N2 component (reduced negativity). These results may confirm that the P2 component may be at the origin of language control after-effects in high-proficient bilinguals. The fact that we observed the P2 and not the N2 as being sensitive to language recovery manipulations is particularly relevant, since language recovery effects make explicit predictions regarding the mechanisms involved in language control (inhibition vs. activation). One possibility is that we did not find inhibitory effects in the N2 time-window because this component may not

114

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

Fig. 8. L2 recovery. Grand Average ERPs for naming in the L2 in the first block (START, Group Start in L2) and in the third block (RETURN new, Group Start in L2). Negative is plotted up.

be a reliable index of inhibitory control, at least in bilingual language production studies9. Granted this as a possibility, we may not exclude that the P2 effects observed here reflect inhibitory control on L1. However, it is important to note that while the N2 has been related to inhibition across a large number of scientific evidences, the P2 component has never been associated to inhibitory processes. Hence, our findings may actually indicate that the mechanisms of language control in high-proficient bilinguals occur in a qualitatively different way from those of lowproficient bilinguals, even when languages are not mixed. These control mechanisms related to the effects occurring at the P2 timewindow may operate in a proactive manner at the early stage of lexical access, in order to allow bilingual speakers to control the languages efficiently. Therefore, in order to explain the behavioral asymmetries in language recovery we propose an alternative

9

Despite the N2 component received a lot of attention in the domain of language control investigations (e.g.,Christoffels et al., 2007; Jackson et al., 2001; Verohef et al., 2009) the results derived from these studies are far from being consistent. In fact, an N2 enhancement has been reported for the L2 and not for the L1 by Jackson et al. (2001), the opposite pattern was reported by Christoffels et al. (2007), while Verhoef et al. (2009) did not report any N2 modulation related to language control effects.

account. We observed that RTs for naming in the L1 and L2 in switch and return blocks were almost identical. A possible reason could be that, in the second block and any other following, the two languages have been activated and used once. Thus, this bilingual mode context (see also 10Grosjean, 2001), where both the languages are in some way already active, could have induced an adjustment of the L1's selection threshold, such as the L1 becomes slower in the return block. Hence, the L1 would be slower in the return block not because of inhibition during the naming in the L2 (switch block), but because after the switch block the languages are adjusted to a bilingual mode context. In line with this interpretation, Costa et al. (2004, 2006) proposed a similar account to explain the paradoxical speed disadvantage for L1 in mixed naming tasks. According to the authors, the L1 is slowed down because its

10 Grosjean (2001) extended the view of global control (inhibition and activation processes) by assuming that bilinguals are able to balance (unconsciously) the activation levels of the two linguistic systems. This could vary depending on the specific communicative setting. Therefore, the activation of the two languages (context-dependent) is reflected by a position on a “LANGUAGE MODE CONTINUUM”. On one end of the continuum just one language is strongly activated (“MONOLINGUAL mode”). On the other end of the continuum the two languages are about equally activated (“BILINGUAL mode”).

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

115

Fig. 9. Scatter plots. Scatter plots and fitted lines describe the relationship between mean naming latencies and mean amplitudes (P2 and N2).

selection threshold is heightened to favor L2 selection. Hence, it could be that also in this case L1's selection threshold is raised after naming in the L2 (switch block) determining an increase of naming latencies and an enhancement of the P2 positivity in the return block. The presence of the P2 effects and the absence of inhibitory ones in the N2 time-window suggests that the L1 threshold's adjustment, to adapt to the bilingual mode context, could explain the asymmetries of language recovery in highproficient bilinguals.

5. Conclusion Overall, the present study provides evidence of the effects of speaking in one language upon the successive use of the other language in single naming contexts. We suggest that these effects reflect the involvement of language control processes which likely operate on the whole language set (global control), given that we found new and repeated items as affected similarly. Moreover, our study provides strong support that language control is influenced by L2 proficiency. Contrary to a similar experiment in which inhibition effects were reported in lowproficient bilinguals (Misra et al., 2012), our findings suggest that the fashion in which high-proficient bilinguals achieve language control could be qualitatively different and might not be explained by inhibitory control mechanisms.

Acknowledgments We are grateful to Cristina Baus for her constructive comments on this manuscript. This work was supported by grants from the Spanish government (PSI2008-01191, PSI2011-23033, Consolider Ingenio 2010 CSD2007-00012) and the Catalan government (Consolidat SGR 2009-1521). Francesca M. Branzi was supported by a predoctoral fellowship from the Spanish Government (FPU-2009– 2013). Clara D. Martin was supported by the Spanish Government

(Grant Juan de la Cierva) and is now supported by the Basque Foundation for Science (IKERBASQUE) and the BCBL institution.

References Abutalebi, J., Annoni, J. M., Seghier, M., Zimine, I., Lee-Jahnke, H., Lazeyras, F., et al. (2008). Language control and lexical competition in bilinguals: An eventrelated fMRI study. Cerebral Cortex, 18, 1496–1505. Abutalebi, J., Della Rosa, P. A., Ding, G., Weekes, B.S, Costa, A., & Green, D. W. (2013). Language proficiency modulates the engagement of cognitive control areas in multilinguals. Cortex, 49, 905–911. Christoffels, I. K., Firk, C., & Schiller, N. O. (2007). Bilingual language control: An event-related potential study. Brain Research, 1147, 192–208. Costa, A., & Santesteban, M. (2004). Lexical access in bilingual speech production: Evidence from language switching in highly proficient bilinguals and L2 learners. Journal of Memory and Language, 50, 491–511. Costa, A., Santesteban, M., & Ivanova, I. (2006). How do highly proficient bilinguals control their lexicalization process? Inhibitory and Language-Specific Selection mechanisms are both functional. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32, 1057–1074. Costa, A., Strijkers, K., Martin, C. D., & Thierry, G. (2009). The time course of word retrieval revealed by event-related brain potentials during overt speech. PNAS, 106, 21442–21446. De Groot, A. M. B., & Christoffels, I. K. (2006). Language control in bilinguals: Monolingual tasks and simultaneous interpreting. Bilingualism: Language and Cognition, 9, 189–201. Gade, M., & Koch, I. (2007). The influence of overlapping response sets on task inhibition. Memory & Cognition, 35, 603–609. Ganushchak, L. Y., Christoffels, I. K., & Schiller, N. O. (2011). The use of electroencephalography in language production research: A review. Frontiers in Psychology, 2, 1–6. Garbin, G., Costa, A., Sanjuan, A., Forn, C., Rodriguez-Pujadas, A., Ventura, N., et al. (2011). Neural bases of language switching in high and early proficient bilinguals. Brain and Language, 119, 129–135. Gratton, G., & Coles, M. G. H. (1989). Generalization and evaluation of eyemovement correction procedures. Journal of Psychophysiology, 3, 14–16. Green, D. W. (1998). Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1, 67–81. Grosjean, F. (2001). The bilingual's language modes. In: J. L. Nicol (Ed.), One mind, two languages: Bilingual language processing (pp. 1–22). Cambridge, MA: Blackwell Publishers. Guillaume, C., Guillery-Girard, B., Chaby, L., Lebreton, K., Hugueville, L., Eustache, F., et al. (2009). The time course of repetition effects for familiar faces and objects: An ERP study. Brain Research, 1248, 149–161.

116

F.M. Branzi et al. / Neuropsychologia 52 (2014) 102–116

Guo, T., Liu, H., Misra, M., & Kroll, J. F. (2011). Local and global inhibition in bilingual word production: fMRI evidence from Chinese–English bilinguals. NeuroImage, 56, 2300–2309. Guthrie, D., & Buchwald, J. S. (1991). Significance testing of difference potentials. Psychophysiology, 28, 240–244. 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. Hernandez, A. E., & Reyes, I. (2002). Within- and between-language priming differ: Evidence from repetition of pictures in Spanish–English bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 726–734. Hernandez, A. E., Martinez, A., & Kohnert, K. (2000). In search of the language switch: An fMRI study of picture naming in Spanish–English bilinguals. Brain and Language, 73, 421–431. Jackson, G. M., Swainson, R., Cunnington, R., & Jackson, S. R. (2001). ERP correlates of executive control during repeated language switching. Bilingualism: Language and Cognition, 4, 169–178. Jodo, E., & Kayama, Y. (1992). Relation of a negative ERP component to response inhibition in a Go/No-go task. Electroencephalography and Clinical Neurophysiology, 82, 477–482. Khateb, A., Abutalebi, J., Michel, C., Pegna, A. J., Lee-Jahnke, H., & Annoni, J.-M. (2007). Language selection in bilinguals: A spatio-temporal analysis of electric brain activity. International Journal of Psychophysiology, 65, 201–213. Kilner, J. M. (2013). Bias in a common EEG and MEG statistical analysis and how to avoid it. Clinical neurophysiology: Official journal of the International Federation of Clinical Neurophysiology, 124, 2060–2068. Koch, I., Gade, M., Schuch, S., & Philipp, A. M. (2010). The role of inhibition in task switching: A review. Psychonomic Bulletin & Review, 17, 1–14. 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. Lehmann, D., & Skrandies, W. (1980). Reference-free identification of components of checkerboard-evoked multichannel potential fields. Electroencephalography and Clinical Neurophysiology, 48, 609–621. Levy, B. J., Mc Veigh, N., Marful, A., & Anderson, M. C. (2007). Inhibiting your native language: The role of retrieval-induced forgetting during second-language acquisition. Psychological Science, 18, 29–34. Martin, C. D., Barceló, F., Hernández, M., & Costa, A. (2011). The time course of the asymmetrical “local” switch cost: Evidence from event-related potentials. Biological Psychology, 86, 210–218. Masaki, H., Tanaka, H., Takasawa, N., & Yamazaki, K. (2001). Error-related brain potentials elicited by vocal errors. Neuroreport, 12, 1851–1855. Mayr, U., & Keele, S. W. (2000). Changing internal constraints on action: The role of backward inhibition. Journal of Experimental Psychology: General, 129, 4–26. Meuter, R. F. I., & Allport, A. (1999). Bilingual language switching in naming: Asymmetrical costs of language selection. Journal of Memory and Language, 40, 25–40.

Misra, M., Guo, T., Bobb, S., & Kroll, J. F. (2012). When bilinguals choose a single word to speak: Electrophysiological evidence for inhibition of the native language. Journal of Memory and Language, 67, 224–237. Nieuwenhuis, S., Yeung, N., & Cohen, J. D. (2004). Stimulus modality, perceptual overlap, and the go/nogo N2. Psychophysiology, 41, 157–160. Pfefferbaum, A., Ford, J. M., Weller, B. J., & Kopell, B. S. (1985). ERPs to response production and inhibition. Electroencephalography and Clinical Neurophysiology, 60, 423–434. Philipp, A. M., Gade, M., & Koch, I. (2007). Inhibitory processes in language switching: Evidence from switching language-defined response sets.. European Journal of Cognitive Psychology, 19(3), 395–416. Picton, T. W., Bentin, S., et al. (2000). Guidelines for using human event-related potentials to study cognition: Recording standards and publication criteria. Psychophysiology, 37, 127–152. Rugg, M. D., & Nagy, M. E. (1989). Event-related potentials and recognition memory for words. Electroencephalography & Clinical Neurophysiology, 72, 395–406. Runnqvist, E., & Costa, A. (2012). Is retrieval-induced forgetting behind the bilingual disadvantage in word production? Bilingualism: Language and Cognition, 15, 365–377. Sebastián-Gallés, N., Martí, M., Carreiras, M., & Cuetos, F. (2000). LEXESP: Lexíco informatizado del español. Barcelona: Ediciones Universitat de Barcelona ([LEXESP: Spanish informatized lexic]). Snodgrass, J. G., & Vanderwart, M. (1980). A standardized set of 260 pictures: Norms for name agreement, familiarity and visual complexity. Journal of Experimental Psychology: Human Learning and Memory, 6, 174–215. Strijkers, K., Costa, A., & Thierry, G. (2010). Tracking lexical access in speech production: Electrophysiological correlates of word frequency and cognate effects. Cerebral Cortex, 20, 912–928. Strijkers, K., Holcomb, P., & Costa, A. (2011). Conscious intention to speak facilitates lexical access during overt object naming. Journal of Memory and Language, 65, 345–362. Strijkers, K., Baus, C., Fitzpatrick, I., Runnqvist, E., & Costa, A. (2013). The temporal dynamics of first versus second language speech production. Brain and Language, 127, 6–11. Szekely, A., Jacobsen, T., D'Amico, S., Devescovi, A., Andonova, E., Herron, D., et al. (2004). A new on-line resource for psycholinguistic studies. Journal of Memory and Language, 51, 247–250. Verhoef, K. M. W., Roelofs, A., & Chwilla, D. J. (2009). Role of inhibition in language switching: Evidence from event-related brain potentials in overt picture naming. Cognition, 110, 84–99. Wang, Y., Xue, G., Chen, C., Xue, F., & Dong, Q. (2007). Neural bases of asymmetric language switching in second-language learners: An ER- fMRI study. NeuroImage, 35, 862–870. Wohlert, A. B. (1993). Event-related brain potentials preceding speech and non speech oral movements of varying complexity. Journal of Speech, Language and Hearing Research, 36, 905–987. Yeung, N., & Monsell, S. (2003). Switching between tasks of unequal familiarity: The role of stimulus-attribute and response-set selection. Journal of Experimental Psychology: Human Perception & Performance, 29, 455–469.