The Human Cerebellum Encodes Temporally Sensitive Reinforcement Learning Signals.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The cerebellum, traditionally linked to motor control, also plays a role in non-motor functions like reinforcement learning (RL).
• Animal models show cerebellar involvement in reward processing and prediction errors, but human data is limited.

Purpose of the Study:

• To investigate cerebellar reinforcement learning (RL) signals in humans using fMRI.
• To understand the temporal constraints and network interactions of cerebellar RL signals.

Main Methods:

• Functional magnetic resonance imaging (fMRI) in human participants performing a reinforcement learning (RL) task.
• Manipulation of feedback delay to assess temporal sensitivity of cerebellar signals.
• Connectivity analyses to identify coactivating brain regions.

Main Results:

• Robust RL signals, including reward processing and prediction errors, were found in cognitive cerebellar regions.
• Cerebellar RL signals were temporally sensitive, detectable with short feedback delays but not with delays over one second.
• These temporal effects were specific to the cerebellum, unlike in the ventral striatum or hippocampus.
• Cerebellar reward prediction error activity correlated with behavior under immediate feedback conditions.
• Connectivity analyses showed coactivation between cognitive cerebellum and prefrontal cortex/caudate nucleus during RL feedback.

Conclusions:

• The human cerebellum contributes to cognitive reinforcement learning (RL) in a temporally constrained manner.
• These findings highlight functional parallels between human and animal models of cerebellar RL.
• The cerebellum's role in RL suggests it as a potential intervention target for reward processing disorders.