Zero cerebellum is actually amazingly rare, and probably leads to lots of compensatory changes, but my read is it causes far too many problems to isolate something this specific...
Shout-out to Flash & Hogan :) - here we're imagining the manifold as a temporal scaffold that you can map onto different outputs...
The cortex generates invariant dynamic primitives; the cerebellum reconfigures them to drive distinct policies.
Huge congrats to first author Martha Garcia-Garcia for leading this tour de force, and @somnirons.bsky.social & Michal Wojcik for a great collaboration!
www.biorxiv.org/content/10.6...
We simultaneously imaged premotor cortical output (L5PT) and cerebellar granule cells during VR & reaching tasks.
The cortex reused manifolds. But the cerebellum didn't shatter them—it reoriented them, rotating them apart while perfectly preserving their geometry! 🤯
The Problem: The cortex uses low-dimensional manifolds to generalize learning. But to prevent interference, circuits must separate overlapping contexts. Dogma says the cerebellum "shatters" inputs to orthogonalize...but that destroys the smooth geometry needed for continuous prediction and control.
How does the brain generalize past experiences without confusing memories? 🧠
Our lab's newest preprint reveals a fundamental division of labor between the cortex and cerebellum that solves this problem. (Check out the cortex-cerebellum video below! 👇) 🧵
