Oscillatory characterization of sensory wordform pre-activation in the visual and auditory domains

Résumé

Prediction has been proposed to be a fundamental cognitive mechanism. However, empirical data regarding the neural correlates of the predictions themselves are still scarce. One outstanding question, especially within the language processing literature, is what degree of perceptual detail is carried by predictive representations. In the present thesis, we investigated sensory pre-activation by comparing activity before two highly predictable words differing only along a sensory dimension. This paradigm was conducted with written or spoken words to examine the role of sensory modality, and with either fixed or variable delays, using time-frequency analysis of MEG signals. Our results showed that pre-word activity was modulated by expected word-form features both in the auditory and visual domains, whilst timing only interacted with predictive effects in the auditory domain. Congruent with prior literature, phonological pre-activation localized at superior temporal areas in the theta band, and visual prediction at ventro-occipital cortex in the beta band. Differences across modalities may reflect preferential oscillatory signatures of different brain areas as well as specific characteristics of the speech and written texts. These results contribute to the understanding of the role of predictive processing in general and in the linguistic domain specifically, showing that word-form pre-activations maybe generated when context clearly biases towards a specific word. By identifying the oscillatory bands that characterize such pre-activation, we hope these results will help interpret pre-target word brain activity in other more naturalistic set-ups.