Brain bases of language selection: MEG evidence from Arabic-English bilingual language production

Abstract

Much of the world's population is bilingual, hence, language selection is a core component of language processing in a significant proportion of individuals. Though language selection has been investigated using artificial cues to language choice such as color, little is known about more ecologically valid cues. We examined with MEG the neurophysiological and behavioral effects of two natural cues: script and cultural context, hypothesizing the former to trigger more automatic language selection. Twenty Arabic-English bilinguals performed a number-naming task with a Match condition, where the cue and target language of response matched, and a Mismatch condition, with opposite instruction. The latter addressed the mechanisms responsible for overriding natural cue-language associations. Early visual responses patterned according to predictions from prior object recognition literature, while at 150-300 ms, the anterior cingulate cortex showed robust sensitivity to cue-type, with enhanced amplitudes to culture trials. In contrast, a mismatch effect for both cue-types was observed at 300-400 ms in the left inferior prefrontal cortex. Our findings provide the first characterization of the spatio-temporal profile of naturally cued language selection and demonstrate that natural but less automatic language-choice, elicited by cultural cues, does not engage the same mechanisms as the clearly unnatural language-choice of our mismatch tasks.

Keywords: MEG; bilingualism; language selection; natural cue; speech production.

Figure 1

Graphic depiction of all types of stimuli utilized in the experiment and correct responses for each of the experimental conditions.

Figure 2

Trial structure for the presentation of the stimuli.

Figure 3

Mean reaction times and error rate as a function of task, cue type and language of the stimuli.

Figure 4

ROI results for a 2×2 ANOVA in the visual cortex and fusiform gyrus, activation averaged across subjects. On the waveform plots, the shaded regions indicate significant differences in activation as a factor of the utilized cue. In the left hemisphere, greater activity was observed when Script was utilized as a cue in comparison to when Culture was used for the same purpose. The opposite was true for the right hemisphere. Significance was determined using a non-parametric, permutation test (Maris and Oostenveld, 2007) performed from 50 to 200 ms (10,000 permutations). The whole brain comparisons show the subtraction of the activity elicited by Script cued conditions from the activity elicited by Culture cued conditions. In consequence, blue reflects increased activity for Script while yellow reflects increased activity for Culture.

Figure 5

ROI results for pairwise comparisons in the ACC, activation averaged across subjects. On the waveform plots, the shaded regions indicate significantly greater activity when culture was utilized as a cue in comparison to when Script was used in the Match task. Significance was determined using a non-parametric, permutation test (Maris and Oostenveld, 2007) performed from 150 to 300 ms (10,000 permutations).

Figure 6

ROI results for pairwise comparisons in the ACC, activation averaged across subjects. On the waveform plots, the shaded regions indicate significantly greater activity when culture was utilized as a cue in comparison to when Script was used in the Mismatch task. Significance was determined using a non-parametric, permutation test (Maris and Oostenveld, 2007) performed from 150 to 300 ms (10,000 permutations).

Figure 7

ROI results for pairwise comparisons in the LIPC, activation averaged across subjects. On the top, the locations of the ROIs are indicated in blue, green and red. On the waveform plots, the shaded regions indicate significantly greater activity when there was a mismatch in the association between cue and target language. Significance was determined using a non-parametric, permutation test (Maris and Oostenveld, 2007) performed from 300 to 400 ms (10,000 permutations).

Figure 8

Presentation of the stimuli for Experiment 2.

Figure 9

ROI results for pairwise comparisons in the ACC between Numerosity and Number naming conditions in Experiment 2, activation averaged across subjects. Significance was determined using a non-parametric, permutation test (Maris and Oostenveld, 2007) performed from 150 to 300 ms (10,000 permutations). No region showed reliable differences in activation between conditions in this analysis.