Neural representations of moving systems.

The cerebellum is necessary for moving smoothly and accurately, but this does not imply that the cerebellum generates or modifies movement control signals. Cerebellar function can be explained by assuming that it is involved in constructing neural representations of moving systems, including the body, its parts, and objects in the environment. To draw a technological analogy, the cerebellum could be a neural analogue of a dynamical state estimator, or a part of one. This explanation is able to account not only for cerebellar involvement in motor control, motor learning, and certain kinds of reflex conditioning, but also cerebellar involvement in certain kinds of perceptual and cognitive tasks unrelated to the production of movements. Evidence for the hypothesis that the cerebellum is involved in a neural analogue of state estimation is (1) across phylogeny, cerebellar morphology reflects animals' use of particular sensory systems for analyzing their own movements and the movements of objects in the environment; (2) cerebellar "oculomotor" neurons are active in relation to movements of salient objects in the environment, regardless of whether the animal moves its eyes to look at them; (3) compensatory eye movements have dynamic characteristics indicating that the control signals are constructed from an underlying optimal head state representation; and (4) the motor symptoms of cerebellar dysfunction resemble the effects of faulty state estimation in artificial control systems. The state estimator hypothesis explains the participation of the cerebellum in controlling, perceiving, and imagining systems that move.