Objective: To further characterize cerebellar learning processes as well as learning impairment pattern in patients with cerebellar dysfunction.

Background: Supervised learning (i.e., learning from error) and reinforcement learning (i.e., learning from reward) are two of the most fundamental learning processes in the brain (Doya 1999). Recently, cerebellar granular cells were found to encode reward omission in mice (Wagner et al. 2017) and the cerebellar Purkinje cells were identified to track reward-related learning in monkeys (Sendhilnathan et al. 2020). These new findings challenge the traditional concept that the cerebellum relies on supervised learning to modify motor activities and behavior while the basal ganglia is dependent on reinforcement learning (Cagliori et al. 2019).

Method: Nineteen patients with cerebellar ataxia and 57 age and sex-matched controls (3 per patient) completed cognitive and neuropsychological tests as well as 2 learning experiments. Experiment 1 consisted of a reinforcement learning task. Experiment 2 provided an internal control by allowing decisions to be made based on either episodic memory or reinforcement learning.

Results: In experiment 1, patients with cerebellar ataxia were impaired in two ways compared to controls. First, patients sub-optimally adjusted their rate of learning throughout the task. Second, they made fewer decisions based on the values learned through reinforcement. These results were confirmed in experiment 2, where patients with cerebellar ataxia were again impaired at reinforcement learning but used episodic memory for decisions.

Conclusion: Patients with cerebellar dysfunction have deficits in reward-based learning. This study also provides further evidence that an intact cerebellum is necessary not just for supervised learning but also for reinforcement learning.

References: Doya, Kenji. “What are the computations of the cerebellum, the basal ganglia and the cerebral cortex?.” Neural networks 12.7-8 (1999): 961-974.
Caligiore, Daniele, et al. “The super-learning hypothesis: Integrating learning processes across cortex, cerebellum and basal ganglia.” Neuroscience & Biobehavioral Reviews 100 (2019): 19-34.
Wagner, Mark J., et al. “Cerebellar granule cells encode the expectation of reward.” Nature 544.7648 (2017): 96-100.
Sendhilnathan, Naveen, et al. “Neural correlates of reinforcement learning in mid-lateral cerebellum.” Neuron 106.1 (2020): 188-198.

« Back to 2022 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/impaired-reinforcement-learning-in-patients-with-cerebellar-ataxia/