Cerebellum and Habit

Project Summary Much of our behavior is habitual. Habits are defined as behaviors that are initiated automatically in response to stimuli, which means that neural representations of the likely outcome of the behavior (e.g., a reward) do not control these behaviors. The neural mechanisms underlying habit learning and performance are of clinical relevance because compulsive disorders, such as obsessive-compulsive disorder and drug addiction, are likely to involve imbalances between habit and goal-directed control over behavior. The most prominent theories posit that stimulus-action associations – neural representations of the stimulus that activate the habit action – are formed as a consequence of dopamine-mediated reinforcement learning and plasticity in corticostriatal circuits. Although lesion, inactivation and pharmacology studies of the dorsal striatum support this view, there is little neurophysiological evidence that dorsal striatal neurons represent stimulus-action associations. Here, we propose a novel hypothesis: that habit learning is a form of skill learning that, like other forms of skill learning, is dependent on sensory prediction error-based plasticity in the cerebellum. Skills are action sequences that are performed rapidly and accurately. Like habits, performance of skilled action sequences does not require a neural representation of the reward outcome (although skilled action sequences can be activated by such representations). Skilled action sequence learning depends on adjusting ongoing and future actions based on the difference between the action performed and the action predicted by the immediately-preceding action or stimulus. These differences are represented as sensory prediction errors in the cerebellum and, during skill learning, cerebellar output neurons come to represent action-action and stimulus-action associations. The activity of these neurons produces accurate movements at the correct time within the sequence. We propose that connecting a stimulus to an action sequence so that the action is performed automatically after stimulus presentation – i.e., learning a stimulus-action habit – requires the same cerebellar circuitry and supervised learning algorithm as learning the action sequence itself. The proposed experiments test the strong predictions of this hypothesis that disruption of cerebellar function will disrupt the formation and performance of stimulus-action habits, and that the firing of neurons in the deep cerebellar nuclei will reflect the formation of stimulus-action associations. Furthermore, based on anatomical studies that show reciprocal mono- and polysynaptic connections between the cerebellum and the infralimbic cortex, a structure required for habit behavior, we will specifically investigate whether these projections contribute to habit learning and performance. Establishing a role for the cerebellum in habit learning would be consistent with a growing literature indicating that the cerebellum participates in cognitive functions far beyond motor behavior. Most importantly, demonstrating a plausible candidate cerebellar mechanism for habit learning would constitute a paradigm shift in our understanding of how habits are learned and controlled.