Predictive coding of visual feedback of actions in human parieto-frontal circuits

Abstract

Executing visually guided movements requires integrating motor commands with visual feedback of the moving hand. Yet, how the motor and visual systems interact during movement remains unquantified. Using real-time motion tracking in immersive virtual reality during fMRI (MOTUM), we manipulated motor execution and hand visual feedback orthogonally in 24 human participants performing reach-to-grasp actions. Execution and hand visual feedback converged in a bilateral parieto-frontal network, including the anterior intraparietal sulcus (aIPS), anterior superior parietal lobule (aSPL), and premotor cortex. In the left aIPS extending into aSPL, combining movement with visual feedback elicited a subadditive response—less activity than predicted by the sum of movement and visual feedback alone—revealing circuitry tuned to the predicted visual consequences of action. Conversely, the posterior cerebellum showed a superadditive response, consistent with internal model updating. Dynamic causal modeling demonstrated a dynamic reversal in information flow along the dorsal parieto-frontal pathway (aSPL–PMd): effective connectivity was excitatory from parietal to frontal cortex when the moving hand was visible, but reversed from frontal to parietal cortex when visual feedback was removed. These findings show that the human parieto-frontal network integrates action and visual feedback predictively, suppressing activity when feedback matches internal predictions and dynamically reconfiguring information flow based on visual availability.