Abstract Representations of Sensorimotor Transformations in Human Premotor Cortex

Humans flexibly adapt their movements across contexts to implement a vast repertoire of skills. This means the motor system should not only encode movement-specific information, but also context-specific representations linking goals to actions. Here we asked if and how the human sensorimotor cortex encodes different latent mappings between visual goals and movements. Participants learned to adapt wrist movements under two distinct visuomotor transformations, where they controlled a visual cursor that was either rotated or reflected away from their hand movement. Representational similarity analyses on fMRI data collected during the task dissociated neural activity patterns related to transformation context, movement direction, and target location. We observed distinct representational profiles within sensorimotor cortex: Premotor cortex activity patterns tracked both movement direction and transformation context, whereas primary motor cortex activity patterns tracked movement direction but not transformation context. Stronger transformation-context encoding in dorsal premotor cortex was associated with better task performance, linking our neural effects to behavior. These findings suggest that premotor cortex can represent sensorimotor context abstracted from movement particulars, providing a candidate neural interface between abstract action policies and motor commands.