From cerebellar texture to movement optimization

The cerebellum is a major site for supervised procedural learning and appears to be crucial for optimizing sensorimotor performance. However, the site and origin of the supervising signal are still elusive. Furthermore, its relationship with the prominent neuronal circuitry remains puzzling. In this paper, I will review the relevant information and seek to synthesize a working hypothesis that explains the unique cerebellar structure. The aim of this review was to link the distinctive functions of the cerebellum, as derived from cerebellar lesion studies, with potential elementary computations, as observed by a bottom-up approach from the cerebellar microcircuitry. The parallel fiber geometry is ideal for performing millisecond computations that extract instructive signals. In this scenario, the higher time derivatives of kinematics such as acceleration and/or jerk that occur during motor performance are detected via a tidal wave mechanism and are used (with appropriate gating) as the instructive signal to guide motor smoothing. The advantage of such a mechanism is that movements are optimized by reducing "jerkiness" which, in turn, lowers their energy requirements.