Computations in the cerebellum during control of a movement

Damage to the cerebellum typically causes dysmetria, affecting endpoint accuracy. What is the cerebellum computing that is so important for stopping a movement? In the marmoset, we have been studying saccadic eye movements, using silicon probes to quantify both the information that the cerebellar cortex receives via mossy fibers and climbing fibers, and the output it produces via Purkinje cells (P-cells). Our results suggest that the input encodes two different kinds of information: the goal of the movement in sensory coordinates, and the ongoing commands in motor coordinates. The output, as computed by populations of P-cells, produces a burst that rises analogous to predicting the real-time displacement of the eyes, and then suddenly and synchronously is inhibited when the eyes need to decelerate and stop. Thus, we speculate that the cerebellar cortex is computing in real-time the sensory consequences of the motor commands until it reaches a bound, as set by the desired goal of the movement, at which point the P-cells synchronously disinhibit the nucleus, thus stopping the movement.