Farthouat, J. 1 , Op de Beeck, M. 2 , Mary, A. 1 , Delpouve, J. 1, 2 , Leproult, R. 1 , Franco, A. 3 , de Tiège, X. 2 & Peigneux, P. 1
1 UR2NF, Neuropsychology and Functional Neuroimaging Research Unit at CRCN, Centre de Recherches en Cognition et Neurosciences and UNI - ULB Neurosciences Institute at the Université Libre de Bruxelles (ULB)
2 LCFC, Laboratoire de Cartographie fonctionnelle du Cerveau, UNI - ULB Neurosciences Institute, Université libre de Bruxelles (ULB)
3 UNESCOG, Unité de Recherche en Neurosciences Cognitives at CRCN, Centre de Recherches en Cognition et Neurosciences and UNI - ULB Neurosciences Institute at the Université Libre de Bruxelles (ULB)
In the present MEG study, we hypothesized that changes in neural oscillations synchronized with stimulus presentation during statistical learning (SL) might represent a robust alternative to standard behavioral tests. Indeed, cerebral electromagnetic activity can respond to stimulation by resonating at the frequency of stimulation. We predicted that during the audition of tritones embedded in a constant flow, the resonance frequency initially reflecting tones succession will gradually shift to a resonance at the level of the triplets, hence reflecting effective segmentation. Twelve pure sounds were randomly assigned to either a statistical stream, in which tones were always grouped as triplets, or to a random stream in which tones were randomly presented. In line with our predictions, we found increased triplet frequency power, and decreased tone frequency power, as compared to neighboring frequency bins, during the statistical as compared to the random streams. This effect was found robust in all participants (N=12) and took place over the first minutes of stimulation, reaching significance after 3 minutes of exposure. These results indicate that increased/decreased power in frequency-tagged responses is a marker of segmentation and a robust method to assess SL.