Predictive coding during action observation revealed by human electrocorticographic activity

Predictive coding is a theoretical framework that has received much attention for its ability to generate testable hypotheses on how multiple brain regions integrate information during cognitive functions. Given relatively large sensorimotor delays, during social interactions, predicting the behavior of others is crucial to enable joint actions or provide competitive advantages. The action observation network (AON) has been extensively studied, but how information is integrated across its main nodes remains poorly understood. Here we leverage the high spatial and temporal resolution of intracranial Electrocorticography (ECoG), to characterize how the key nodes of the AON - including precentral, supramarginal and visual areas - exchange information. We found more top-down beta oscillation from precentral to supramarginal contacts during the observation of predictable actions while more bottom-up gamma oscillation from visual to supramarginal contacts were measured for unpredictable actions. These results, in line with predictive coding, provide critical evidence towards an understanding of how nodes of the AON integrate information to process the actions of others.