The Cerebellum’s Role in Reward Processing

Theoretical models have been successfully developed to explain the role of the cerebellum in motor processing. These models emphasize the cerebellum’s role as a learning machine that changes and perfects ongoing processing through the quick comparison of what is going on in the real world, and what the brain predicts to happen via an “internal model.” Recent findings in humans and mice suggest that portions of the cerebellum are involved in reward processing; a process which has primarily been investigated in regards to the basal ganglia and its substructure, the ventral striatum. It is unclear what computational role the cerebellum plays in reward processing, as the cerebellum has been said to explicitly operate without reward information. To understand this, my dissertation used fMRI in adolescents and young adults to study the cerebellum’s contribution to social reward processing, and to differentiate it from the role of the ventral striatum in reward processing.

The Task

While in the fMRI scanner, participants were shown images of people from their age group. The participants were told to make one of two choices: predict that the person shown on the screen “liked” them or did not “like” them. Imagine this being similar to a dating app where you swipe left or right, but are trying to guess whether they would swipe left or right on you. After making their choice, the participant was shown a thumbs up image indcating that the person did like them, or a message saying “no response”. What we were interested in, is seeing how participants brains responsed when they were correct about someone liking them or not liking them.

The Results

The original results were published in the article titled []”I knew you weren’t going to like me.”](https://www.frontiersin.org/journals/behavioral-neuroscience/articles/10.3389/fnbeh.2019.00219/full) This title perfectly captures the phenomemon we wanted to study. Is being correct a rewarding sensation, even if it is about a negative situation? The original study showed this effect. The ventral striatum (a major reward region in the brain) went up in activity when participants were correct about someone liking them, and when they were correct about someone not liking them. In my project, I examined the cerebellum, which was not examined in the original study. My hypothesis was that the cerebellum is different from the ventral striatum, in that the ventral striatum cares more about reward (especially positive rewards), whereas the cerebellum cares more about being correct. My hypothesis was correct. I showed that neural resposnes to being correct about being liked and being correct about not being liked were more similar in the cerebellum, than they were in the ventral striatum. This indicates that the cerebellum is doing things differently from the ventral striatum and has a potentially unique contribution of being a general error processor, whereas the ventral striatum responds more strongly to positive rewards.

Why This Matters

These results show that although the cerebellum is involved in lots of different processes that other brain regions have been well mapped to, the cerebellum is contributing something unique. This unique contribution is similar to what the cerebellum does for motor processing; it helps process errors when things do go according to plan.

The Cerebellum’s Role in Social Reinforcement Learning

Currently, I am studying how the cerebellum contributes to social reinforcement learning. This is a process in which we can about other people or from other people. My current project involves adolescents who learn about two peers during an fMRI task. The participants learn whether two peers share similar interests or hobbies to themselves. I am examining the neural signals of when the participant shares a similar interest (e.g., both like to read), and when their interests are different (e.g., participant loves reading, but peer hates reading). In this study, I wish to examine how the cerebellum’s responses to these types of social feedback are different from the ventral striatum which is the more well known region involved in this process. To do this, I am using computational models of social reinforcement learning, in which we can calculate prediction errors based on models of how we believe participants are processing the peers and their feedback.