Facial expressions in mice reveal latent cognitive variables and their neural correlates

Brain activity controls adaptive behavior but also drives unintentional incidental movements. Such movements could potentially be used to read out internal cognitive variables that are also neurally computed. Establishing this would require ruling out that incidental movements reflect cognition merely because they are coupled with task-related responses through the biomechanics of the body. Here we addressed this issue in a foraging task for mice, where multiple decision variables are simultaneously encoded even if, at any given time, only one of them is used. We found that characteristic features of the face simultaneously encode not only the currently used decision variables but also independent and unexpressed ones, and we show that these features partially originate from neural activity in the secondary motor cortex. Our results suggest that facial movements reflect ongoing computations above and beyond those related to task demands and demonstrate the ability of noninvasive monitoring to expose otherwise latent cognitive states.