Introduction: Speech is inherently audio-visual (AV). It is well established that gain in performance afforded by AV speech is greater the smaller performance is in the unimodal conditions (inverse effectiveness). Coherently, it has been suggested that behaviourally AV speech is of even greater importance to non-native speakers (Wang 2008). Candidate multisensory integration (MSI) sites underpinning AV speech perception have been proposed, but little is known about speech-related MSI sites during non-native language perception. Previous results highlight showed that the left superior temporal gyrus (STG) plays a critical role in integrating AV speech and that the congruency (i.e. the temporal relationship) between the sensory modalities modulates activation in this area (Calvert 2001).

Aims: We hypothesized that the effects of AV congruency during speech comprehension in non-native language (L2) would involve the STG differentially as compared to native the language (L1).

Methods: Twenty-one healthy native speakers of Spanish (L1) who had learned English (L2) before adulthood participated in this fMRI study. Mean age of acquisition of the L2 was 11 years. Participants were scanned in two separated runs, one in each of two languages (English or Spanish). Five conditions were presented during each run in a blocked design: auditory alone (A-), visual alone (-V), congruent audiovisual (AVc), incongruent audiovisual (AVi) and a rest condition (blank screen). Runs and conditions were counterbalanced across subjects. Thus, participants could listen (A-), see (-V) or simultaneously hear and see (AVc and AVi) a male Spanish/English bilingual speaker at each respective condition. FMRI data sets were acquired applying a sparse sequence (TR=8-s. TA= 2-s.).

Results: The post-scanning recognition test returned above chance performance for AVi, AVc and A stimuli. The imaging results showed that the posterior portion of the STG was bilaterally more active in the AVc condition than in the AVi condition when presented with L2 stimuli, but not L1. The opposite contrast did not reveal any significant activation. Audio-visual congruency effects between languages showed the involvement of the left posterior STG when presented with L2 more than L1. On the other hand, L1 showed a higher activation than L2 in the anterior portion of the STG. Left posterior STG showed an interaction effect across languages and congruency conditions. Thus, its activation was higher during AVc than AVi for L2, and just the opposite for L1.

Conclusions: In accord with Calvert (2001), we confirm that AV speech congruency modulates activity in the posterior STG. We expand this result showing that this same effect is larger for L2 than for L1. We speculate that the left STG may play a key role in integrating AV speech under a impoverished informational conditions (e.g during L2 comprehension) and that this same area is less responsive in a less demanding task when the visual channel benefit is of lesser importance.