Recent data on the cerebellum require new models and theories.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Motor Control

Background:

• The cerebellum's perceived simplicity historically made it a focus for theoretical studies.
• The Marr-Albus model established error-based supervised learning as a key theory for cerebellar function.
• This framework has guided decades of cerebellar research.

Purpose of the Study:

• To review theoretical progress in cerebellar research within the Marr-Albus framework.
• To discuss experimental findings that reveal greater complexity in cerebellar mechanisms.
• To highlight the need for an integrated model unifying cerebellar functions.

Main Methods:

• Literature review of theoretical and experimental cerebellar research.
• Synthesis of findings related to motor skill learning and error-based supervised learning.
• Analysis of new data on molecular, cellular, and network complexity.

Main Results:

• Theoretical progress has been made within the error-based supervised learning framework.
• Experimental findings reveal complex molecular/cellular mechanisms, new cell types, and recurrent connections.
• The cerebellum is involved in non-motor functions and other learning forms.

Conclusions:

• Existing models need to incorporate new experimental findings.
• An integrated cerebellar model is required to unify its diverse computational roles.
• Future research should focus on bridging theoretical and experimental gaps.