Cursed complexity. Computational properties of subcortical neuronal microcircuitry in sensorimotor control

One of the big obstacles for understanding the nervous system is its inherent complexity. It poses problems when interpreting both experimental and theoretical studies since we are currently forced to consider only reduced variants of the actual circuitry of the brain. Since there exist problems that do not appear until a system is sufficiently complex, there are no guarantees that the results stemming from such reduced studies can be extrapolated to actually apply to the real brain. The initial part of the thesis investigates the properties of the spinocerebellar circuitry of the nervous system, and its role in motor control. Especially the cerebellum has been shown to play an important role in the coordination of fast movements, such as reaching and pointing. Paper I uses theoretical reasoning based on previously found experimental studies to show that the cerebellar circuitry should not be studied in isolation if the aim is to explore cerebellar function. The inputs provided by the pre-cerebellar circuits in the spinal cord and brain stem can significantly reduce the complexity of the problem that the cerebellar circuitry needs to solve. Papers II, IV and V investigate the properties of the mossy fiber pathways. Both the spinal border cell neurons that ascend the ventral spinocerebellar tract with sensorimotor information related to locomotion and the neurons of the cuneate nucleus that process tactile information are studied using behavioral stimulation, either in vivo (Paper V) or through modeling (Paper IV). The results indicate both that the overall activity of the circuitry provides the cerebellum with an easy to interpret encoding, but the individual neurons can at the same time segregate underlying features and details of the stimulus. This result can be seen as a parallel to the found statistics of spike generation in Paper III. Even though the neurons have complex electrodynamic properties, their average activity, described by their firing statistics is surprisingly similar between neurons with vastly different morphology. Paper VI reviews the theoretical grounds for sparse coding, and compares them to recent experimental findings, both in the cerebellum and the neocortex. While there are beneficial properties of certain sparse codes, the experimental results rather indicate that the circuitry both in the cerebellum and the neocortex do not actively maintain a sparse population code