Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Rising Stars Session | Poster Sessions | Data Blitz
Poster F126
Neural correlates of reinforcement learning in the human cerebellum
Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom
Juliana E. Trach1 (juliana.trach@yale.edu), Samuel D. McDougle1,2; 1Yale University, 2Wu Tsai Institute
In the context of motor learning, the cerebellum is believed to encode a forward model, making predictions about expected outcomes of actions and using discrepancies between predicted and actual outcomes (i.e., prediction errors) to refine behavior. Increasing consensus that the cerebellum is involved in cognitive capacities beyond motor control has prompted questions about how the cerebellum might support prediction-error-based learning in non-motor tasks. Recent evidence in model organisms (mice, monkeys) implicates regions of the cerebellum (primarily Crus I and II) in reinforcement learning (RL; Sendhilnathan et al., 2020; Wagner et al., 2017). Here, we investigated cerebellar involvement in non-motor RL in humans. We used fMRI to scan participants (N = 32) during a RL task where they chose between two images on each trial and received probabilistic reward feedback on their choices (0.8s or 3s after response). Importantly, we randomized the button-presses associated with each image so that reward feedback was not associated with specific motor responses. We found robust neural correlates of RL in the human cerebellum. Specifically, we found activation in Crus I and II related to reward processing. Further, these regions exhibited activity correlated with trial-by-trial reward prediction errors, implicating the cerebellum in prediction error processing beyond motor learning. Finally, we found that the cerebellum preferentially responded to short latency (sub-second) feedback, and was less involved when feedback was delayed, echoing temporal constraints on cerebellar motor learning. Overall, this work implicates the human cerebellum in the processing of cognitive prediction errors during non-motor learning.
Topic Area: PERCEPTION & ACTION: Other