Présentation

Stinkeste, C., Vincent, M.A., Delrue, L., & Brunellière, A. (2021). Alpha and high-gamma oscillations support the interface between language prediction and speakers’ communicative intentions in spoken-language comprehension, Society for the Neurobiology of Language, 5-8 October, virtual conference.

The present study investigated whether predictions during spoken language comprehension are supported by alpha and high-gamma oscillations. High-gamma oscillations are usually assumed to reflect prediction errors (Lewis and Bastiaansen, 2015), whereas alpha-band oscillations are closely linked to the main fundamental function of attention (Klimesch, 2012). Attention is particularly critical for successful language comprehension and construction of sentence meaning (Boudewyn, & Carter, 2018) and speakers’ communicative intentions play an important role for the interpretation of sentence meaning. In line with previous studies showing that prediction shifts the attention to the expected event (Foxe et al., 1998; Mayer et al., 2016), we hypothesized a reduced power of alpha activity when the properties of incoming information were in line with those of predicted word in order to reinforce the initial interpretation of sentence meaning. Moreover, we expected to observe a higher decrease of alpha activity for the expected information when the communicative intentions of speakers provided a clear interpretation of the sentence. Differently, the power of high-gamma oscillations should be increased if an incoming information which was incongruent with the prior predictions and this effect should be enhanced by the communicative intentions of speakers. We conducted a first EEG experiment during which thirty-two French-speaking participants listened to semantically constraining sentences predicting a target word which was not presented. We focused on the power of alpha and high-gamma oscillations (respectively, 8-12 Hz, 60-100 Hz) associated with the article which could be either in agreement or in disagreement with the gender of the expected, yet not presented, word. The semantically constraining sentences were produced by the speaker with the intention of emphasizing the content of the sentence. We observed a reduced power of alpha-band oscillations when the gender of article was in agreement with that of the expected word and this pattern was found only when the speaker’s communicative intention was strong. A second EEG experiment was conducted during which thirty-two participants had to judge the speaker’s intention of communication in addition to listening and understanding the sentences. The stimuli were identical to those employed in the first experiment and the participants had the same characteristics as those included in the first experiment. The top-down attention focused on the speaker’s intention of communication led to the removal of predictive effects on alpha-band oscillations. Interestingly, the power of high-gamma oscillations was however stronger for the expected gender than for the unexpected gender when the speaker’s intention of communication was weak over a late time window. This suggested that prediction errors were elicited after the expected gender because of the expected, yet not presented and they were dependent on the speaker’s intention of communication. The top-down attention focused on the speaker’s intention of communication shaped frequency-specific brain dynamics supporting the interface between language prediction and speakers’ communicative intentions. It thus appears that alpha and high-gamma oscillations play different attentional roles related to prediction in spoken language comprehension.