Computational anatomy: the cerebellar microzone computation

Abstract

The cerebellum is a large brain structure. Most of the mass and volume of the cerebellum is made up by the cerebellar cortex. The outer layer of the cerebellar cortex is divided functionally into long, thin strips called microzones. We argue that the cerebellar microzone computation is the aggregate of simple unit computations and a passive effect of anatomy, unaided and unlearned, which we recreate in silico. This is likely to polarise opinion. In the traditional view, data processing by the cerebellum (stated very briefly) is the effect of learned synaptic changes. However, this has become difficult to reconcile with evidence that rate information is linearly conserved in cerebellar signalling. We present an alternative interpretation of cell morphologies and network architecture in the light of linear communication. Parallel fibre synaptic memory has a supporting role in the network computation. Lay summaryIt is not fully understood how the cerebellum represents information and uses it to generate motor outputs, and how outputs are coordinated to control movement. We model the cerebellar network computation, with a focus on the microzone computation, the middle layer of the network. The main proposal is the contentious idea that the core computation is a passive and unlearned effect of neuroanatomy, unit functions are linear, and the computation is mathematically unsophisticated. To provide functional context, we propose that the topography of input and output of cerebellar motor circuits could be a functional extension of modular cerebellar wiring which automates motor sequences. To illustrate how that could work, we present a simple model of anguilliform (eel-like) swimming.