Cerebellar climbing fibers signal flexible, rapidly adapting reward predictions

Classical models of cerebellar computation posit that climbing fibers (CFs) operate according to supervised learning rules, correcting movements by signaling the occurrence of motor errors. However, recent findings suggest that in some behaviors, CF activity can exhibit features that resemble the instructional signals necessary for reinforcement learning, namely reward prediction errors (rPEs). Despite these initial observations, many key properties of reward-related CF responses remain unclear, thus limiting our understanding of how they operate to guide cerebellar learning. Here, we have measured the postsynaptic responses of CFs onto cerebellar Purkinje cells using two-photon calcium imaging in mice of both sexes to test how CFs respond to learned stimuli that either do or do not predict reward. We find that CFs can develop generalized responses to similar cues of the same modality, regardless of whether they are reward predictive. However, the extent of this generalization depends on temporal context, and generalization does not extend across sensory modalities. Further, learned CF responses are flexible, and can be rapidly updated according to new reward contingencies. Together these results reveal two properties of learned CF responses: their generalization across similar cues is graded by the temporal and sensory context in which the cues are learned, and, when reward contingencies are reversed, they update rapidly to track new learning rules. Significance Statement The cerebellum is best known for coordinating movement, and the climbing fiber teaching signals thought to instruct cerebellar motor learning were long thought to report movement errors exclusively. Recent work has shown that these instructional signals can also convey information about reward. Here, by recording from these climbing fiber inputs in mice learning which cues predict reward, we find that they carry flexible, context-sensitive reward predictions. Notably, these signals do not simply track which cue an animal should respond to: they generalize across similar cues depending on context, and they rapidly reorganize when reward associations change. These properties resemble those of midbrain dopamine neurons, the brain's canonical reward system. These findings reframe how cerebellar teaching signals contribute to reward-guided learning.