Poster B54, Thursday, August 16, 3:05 – 4:50 pm, Room 2000AB
Bilingual language control: MEG evidence of inhibition in word production
Judy D. Zhu1,2, Robert A. Seymour1,2,3, Paul F. Sowman1,2;1ARC Centre of Excellence in Cognition and its Disorders, Sydney, Australia, 2Macquarie University, Sydney, Australia, 3Aston University, Birmingham, United Kingdom
Bilinguals have a remarkable ability to control which language to speak at any given time, and to switch between languages seamlessly. It has been suggested that bilinguals rely on inhibitory control to suppress the non-target language to ensure speech output occurs in the desired language. Previous electrophysiological studies report mixed findings as to when inhibition occurs in language switching. One possible reason for this discrepancy in findings is that previous experimental designs have not been well controlled in regards to confounding factors and trial-sequence effects. We addressed these issues in the current MEG study, employing the classical language-switching paradigm with an improved design. We tested sixteen Mandarin-English bilinguals in the MEG. We specifically recruited unbalanced bilinguals (all dominant in L1 Mandarin) so that we could examine how relative language proficiency modulates inhibition. Participants were instructed to name the digit they saw on each trial in either Mandarin or English, as indicated by the language cue (face of interlocutor, either Chinese or Caucasian). To eliminate the possible confound of cue-switching (i.e. cue changes whenever the language changes), which is present in most language-switching studies, we used two faces for each language and ensured that the cue changed on every trial. We generated well-controlled trial sequences by ensuring that there were no consecutive switch trials (previously shown to have a stacking effect on switch cost). A filler trial was inserted after every switch trial, so that no critical trial ever followed a switch trial. To examine language control in distinct stages of processing (preparation stage following cue onset, and naming stage following target onset), we separated cue onset and target onset by an interval of 750ms, which has been shown to be sufficient for optimal preparation. Behavioural naming latencies were submitted to linear mixed-effect modelling with language (Mandarin/English) and switch (stay/switch) as factors. An interaction was found between language and switch, replicating the well-known switch cost asymmetry in unbalanced bilinguals (i.e. it takes longer to switch back to L1 after naming in L2 than vice versa). Preliminary analysis of MEG data in the time domain using cluster-based permutation tests revealed a main effect of switch between 425-550ms following cue onset, and a main effect of language 200-300ms following target onset, which is typical of inhibitory control. These effects suggest that there are control processes involved at both stages. Upon seeing the face cues, bilinguals perform shifting and updating of task goals, biasing their language selection towards the required language. These processes are only required on switch trials, therefore an effect of switch is present in the cue window. Upon seeing the naming target, bilinguals inhibit the non-target language in order to produce speech in the target language. This happens on all trials, but speaking L2 requires stronger suppression of L1 than vice versa. Therefore, the effect of language, displaying signatures of inhibitory control, is present in the target window. Implications for models of bilingual language control will be discussed.
Topic Area: Multilingualism