NEURAL EVIDENCE FOR CEREBELLAR INTERNAL MODELS BASED ON DIRECTION-DEPENDENT PURKINJE CELL CODING OF SACCADE KINEMATICS AND ERROR PROCESSING

The cerebellum is hypothesized to implement internal models for motor control, yet how Purkinje cell populations encode these models remains unclear. We investigated Purkinje cell dynamics in macaques to determine how they shape saccade kinematics and error processing. By aligning activity to each cell’s complex spike preferred direction (CS-on), we uncovered a striking functional asymmetry.

For CS-on opposed saccades, simple spike activity peaks at saccade onset. Given the inhibitory nature of Purkinje cells, this early activity is consistent with a contribution to deceleration following movement onset. Correspondingly, complex spikes in this direction are primarily sensitive to post-saccadic errors, serving to update the system based on movement outcomes. Furthermore, simple spike activity in this population is dynamically modulated by peak velocity and shifts during gain adaptation, suggesting that CS-on opposed saccades receive online feedback from a forward model.

In contrast, for CS-on aligned saccades, simple spike activity peaks at saccade offset. This timing implies a release of inhibition that facilitates acceleration prior to offset. In this direction, complex spikes are more sensitive to pre-saccadic fixation errors, providing input for planning the initial motor drive. Unlike the opposed direction, these responses remain stable during adaptation, indicating that CS-on aligned saccades rely on a pre-programmed motor command characteristic of an inverse model.

Together, these results provide neural evidence that the cerebellum utilizes direction-dependent population codes to implement dual internal models, independently calibrating motor drive and feedback control to maintain saccadic accuracy.