Abstract
Cerebellar computations are necessary for fine behavioural control and may rely on internal models for estimation of behaviourally relevant states. Here, we propose that the central cerebellar function is to estimate how states interact with each other, and to use these estimates to coordinate extra-cerebellar neuronal dynamics underpinning a range of interconnected behaviours. To support this claim, we describe a cerebellar model for state estimation that includes state interactions, and link this model with the neuronal architecture and dynamics observed empirically. This is formalised using the free energy principle, which provides a dual perspective on a system in terms of both the dynamics of its physical—in this case neuronal—states, and the inferential process they entail. As a demonstration of this proposal, we simulate cerebellar-dependent synchronisation of whisking and respiration, which are known to be tightly coupled in rodents, as well as limb and tail coordination during locomotion. In summary, we propose that the ubiquitous involvement of the cerebellum in behaviour arises from its central role in precisely coupling behavioural domains.
Introduction
The cerebellum is one of the most well-studied parts of the brain, yet its "universal" function remains a subject of intense debate. While traditionally viewed as a motor control center, it is now known to be involved in diverse functions ranging from social interaction to high-level decision-making. The challenge lies in explaining how a single, relatively simple neural architecture supports such a vast range of functions. This paper addresses that gap by proposing that the cerebellum doesn't just estimate individual states; it estimates the interactions between states, providing the necessary "glue" to keep complex behaviors synchronized and resilient to environmental noise.
Methods and Theoretical Framework
The authors utilize the Free Energy Principle (FEP) and Active Inference to formalize their theory. Under this framework, the brain is viewed as an inference engine that constantly minimizes "surprisal" (prediction error) by updating its internal model of the world. The cerebellum is modeled as a system that tracks hidden states ($x$) and hidden causes ($v$).
Anatomical Mapping
A major contribution of the study is mapping mathematical variables to specific cerebellar cell types. This provides a precise link between internal inferences and actual neuronal dynamics.
In this mapping, Mossy Fibers (mf) are associated with observations ($y$), Granule Cells (grc) with state estimates ($\mu_x$), and Purkinje Cells (Pjc) with hidden causes ($\mu_v$). This architecture allows the cerebellum to provide "top-down" constraints to extra-cerebellar regions, modeled here using Kuramoto oscillators to represent rhythmic brain activity.
Key Results
The model was tested through numerical simulations of two specific biological scenarios: orofacial coordination and locomotion.
Whisking-Respiration Synchronization
In rodents, whisking and respiration are tightly coupled. The simulations demonstrated that with "cerebellar contextualization," the model successfully maintained a tight phase-lock between these two systems, even when they had different intrinsic speeds or were subjected to external perturbations.
Locomotion and Ataxia
The model also successfully replicated "ataxic" behavior. When cerebellar output was removed, the simulated animal lost pairwise coordination between its front and hind limbs, and its tail movement became asymmetric—matching empirical observations in mice with cerebellar lesions. Quantitatively, the "With Cerebellum" models kept errors between behavioral states consistently low, while "Without Cerebellum" models saw errors double, representing a total loss of coordination.
Discussion and Insights
The authors argue that the cerebellum provides the context for behavior, not the content. It doesn't generate the rhythms of walking or breathing itself; instead, it ensures those rhythms are precisely coupled. This "universal computation" suggests that the cerebellum's repetitive architecture is optimized for estimating conditional dependencies between variables across the brain.
While the current model relies on linear approximations, future work will aim to incorporate more complex, nonlinear internal models and further integrate the role of Climbing Fibers in long-term learning. This framework places the cerebellum at a high level of the neural hierarchy, providing the empirical priors that refine and stabilize the activity of other brain regions.
Summary updated