New Model Challenges How Brain Controls Movement, Rewrites Textbook Theory
Researchers propose the cerebellum controls walking through simple, hardwired neural architecture rather than learning—overturning decades of mainstream neuroscience. The finding could reshape how companies develop brain-computer interfaces, prosthetics, and neuromorphic AI systems, potentially unlocking cheaper, more efficient designs based on biological principles overlooked for years.
Originaltitel: Beyond synaptic plasticity: a summary of a linear model of the cerebellar locomotor computation
We present a summary of ideas that attempt to explain how the locomotor cerebellum may represent and process information. It includes the proposals that (i) the main network computation is a passive and unlearned effect of cell type morphologies and neural architecture; (ii) information has topographically defined spatial dimensions; (iii) it is coded at collective level, at any instant, in any random sample of functionally grouped signals; (iv) topographical organization extends outside the cerebellum and maps to the peripheral nervous system; and (v) learning and memory are at microzone level and in a supplementary role. The aim is to provide competition for traditional learning models and to challenge some common assumptions. We have found that the main resistance to the proposals is in these areas: loyalty to the traditional model; mathematically, the computation is unexpectedly unsophisticated; on the face of it, the mechanism is resource-heavy; we propose that neuroanatomy automates motor coordination and converts feedback into motor output in real time. Some of the ideas are contentious. We argue, nonetheless, that the proposals can explain the evidence.