The cerebellum is usually pictured as the brain’s movement specialist. But evidence suggests it may also help us learn events that unfold over seconds or longer—by predicting what feedback will come next. Timing is crucial for wide-ranging behaviours: from our ability to catch a ball, cross a road, perceive music or plan our commute to work.

The brain must handle both quick bodily movements and slowly changing situations. The ability to encode temporal information across a wide range of time scales is essential for generating adaptive behaviour. Yet how temporal information processing in the brain enables adaptive, goal-directed behaviour remains unclear.

The cerebellum has traditionally been linked to brief timing, such as coordinating movement. More recent studies, however, have provided evidence for cerebellar involvement in the supra-second timing range. In people with damage to parts of the cerebellum, supra-second time intervals could be overestimated or underestimated.

Other studies found the cerebellum active with the prefrontal cortex during longer timing tasks, and stimulation of cerebellar connections could rescue timing deficits in schizophrenia. Together, these reports question the idea that the cerebellum exclusively influences sub-second timing.

The review proposes that its contributions to longer timing arise from interactions with the cerebral cortex. Here is the shift in perspective: timing used to be attributed to two distinct circuits, one for brief intervals and one for longer intervals.

Recent evidence suggests that this division might not be as clear cut. The newer view is a set of brain-wide feedback loops supporting different scales of temporal information processing, rather than two completely separate timing systems. The proposed model works like a coach who watches a learner’s current position and predicts the feedback that will arrive later.

A cerebellar network communicates with a cerebral network, forming a two-learner system. The cerebellar network learns to predict future feedback from the current state of the cerebral network. The main network then integrates these internal feedback signals.

By making feedback available beyond the cerebral network’s own timescale, the cerebellar signal helps the cerebral network acquire activity patterns for precise behaviour. The model suggests that the cerebellum mediates behaviour by predicting feedback across a range of time scales.

In movement-related tasks, the model learned faster and showed fewer coordination-like problems. The feedback predictions helped especially when little feedback came from the environment or the body, which is highly relevant for reward-based learning.

The same model also generated richer captions, particularly longer captions, suggesting that cerebellar feedback can help the cerebral network learn richer visual and language context. Timing once seemed divided between one brain system for split-second events and another for longer intervals.

The newer view instead links both to interacting loops across the brain, suggesting that our sense of time is less neatly separated than previously thought. But here is the catch: another experiment found that shutting down part of the cerebellum did not impair rats’ ability to estimate a longer target interval or learn a new one.

The review suggests this task may not have required a well-timed movement. The rats could begin the nose poke early, and other repeated movements might have helped timing accuracy. So the absence of impairment may reflect the kind of behaviour being tested: the repeated movements could support timing, or the nose poke itself may not depend strongly on the cerebellum.

How the cerebellum interacts with the cerebral cortex during skilled, goal-directed behaviour remains an open question. The review does not claim that the mechanism is settled. Future studies are needed to understand how the cerebellum supports longer intervals, and longer behaviour may be built from many brief steps.

The proposed picture is a brain working across multiple time horizons: the cerebellum receives a copy of ongoing activity and returns predictions or feedback signals needed for learning. For someone outside neuroscience, the practical promise is that the cerebellum appears critical for learning temporal tasks on timescales longer than a second.

Its predictions and feedback may also support behavior across multiple horizons, including motor control, decision-making, and long-term motor planning. The review’s central idea is that the cerebellum may help the rest of the brain learn difficult, extended sequences by supplying predictions across multiple timescales.

That could matter for movement, decisions, rewards, and language.