Perceptuo-motor speech units in the brain with COSMO, a Bayesian model of communication

Nowadays, three kinds of theories still discuss the nature of speech units: auditory theories (Diehl, Lotto, and Holt 2004), motor theories (Galantucci, Fowler, and Turvey 2006) and perceptuo-motor theories (Schwartz et al. 2012). Briefly, supporters of the auditory (resp. motor, perceptuo-motor) theories claim that speech units are represented in the brain by auditory (resp. motor, perceptuo-motor) information. Several neuroimaging studies have revealed the role of the motor system during speech perception (Skipper, Devlin, and Lametti 2017). However, a recent study by Cheung et al. (2016) (noted C-2016 in the rest of the paper) questions the presumed motor nature of the activations observed in the motor cortex. By analysing high-density direct human cortical recordings, they observed a similar structure between neural patterns in the motor and auditory cortex during a perception task but different from neural patterns in the motor cortex during a production task. They conclude that the motor cortex does not contain motor but rather auditory representations in speech perception. Intrigued by this conclusion, we assess COSMO, our Bayesian model of communication (Moulin-Frier et al. 2012), on a comparable study. Indeed, we consider that representations analyzed in the C-2016 study are not clearly identified. Specifically, the distinction between stored and computed information appears unclear. Then, by simulating an analogous perception task, we also observe a similar structure in the two cortical areas. However, this similarity disappears by using noisy stimuli, revealing the differences between representations in the auditory and motor pathways.