How the cerebellum could memorise movements

Abstract

THE neurones of the cerebellum are arranged in an extremely regular fashion and their synaptic connections are known in detail¹; this has led to considerable speculation about the functioning of these neuronal circuits. In particular, following a proposal that the cerebellum is involved in the learning of movements², there have been a number of theories about how the cerebellar circuitry could be used in this way^3–5. These theories are all based on the postulate that the signals relating to motor output which are to be stored in the cerebellum have been computed in the cerebral cortex, and transmitted from there to the cerebellum for storage. They assume that the initial learning process, during which the animal learns to produce motor outputs with favourable consequences for itself, takes place somewhere in the cerebral cortex and not in the cerebellum: only after the cerebral cortex has learned how to generate these motor outputs is the information for their production stored in the cerebellum. Recent results suggest, however, that this postulate is incorrect, and that the cerebellum is directly involved in the learning of motor actions which have satisfactory consequences for the animal. To account for these results I describe a new way by which the cerebellum could be used for storing information relating to movements.