Key Takeaways
- The cerebellum activates during pure motor imagery (without movement, without intention to move) generating forward-model predictions of the rehearsed action.
- Forward models are internal predictors: a copy of the motor command (efference copy) is fed through the cerebellum, which estimates the sensory consequence of the action before it happens.
- When predicted and intended outcomes diverge, climbing-fibre error signals update the cerebellum’s internal model: even when the action is purely imagined.
- The olivo-cerebellar circuit drives the initial phase of skill acquisition; the motor cortex consolidates what the cerebellum has learned into durable retention.
- Timing-critical skills (surgical sequences, musical phrasing, conversational pacing, athletic rhythm) refine through the same prediction-error mechanism whether you rehearse in body or in mind.
The cerebellum runs forward models, internal predictions of movement timing, during pure mental rehearsal, with no muscle activation. When you imagine a sequence, the cerebellum compares its prediction against the rehearsal’s intended outcome, and any mismatch triggers a climbing-fibre error signal that rewrites the internal timing model. You rewire skill from imagination alone.
This article belongs to our hub on mental rehearsal and visualization, where imagined practice is examined in mechanistic detail.
Imagine a movement without lifting a finger and the brain still runs the prediction. The cerebellum rehearses the timing, refining the skill whether the action happens in body or in mind.
Is the Cerebellum Involved in Motor Imagery?
Yes: the cerebellum activates during pure motor imagery, with no overt movement at all. Functional MRI work shows cerebellar engagement during imagined hand action that overlaps substantially with the activation seen during actual execution. Imagined practice is not metaphorically motor; it is literally motor, run off-line.
The empirical case opens with Lotze and colleagues, who in 1999 used fMRI to compare cortical and cerebellar activation during executed and imagined hand movements. Their result was unambiguous: imagined movement produced cerebellar activation that overlapped substantially with executed-movement activation, with the difference being magnitude rather than precise location. The cerebellum was online during pure imagination, not waiting for the muscles.
Two decades later, the framework matured. Sokolov, Miall, and Ivry’s 2017 review in Trends in Cognitive Sciences synthesised more than thirty years of cerebellar imaging and lesion work into a single architectural claim: the cerebellum is not a movement-execution device; it is an adaptive prediction engine, and prediction runs whether or not movement follows. Their account integrates motor, cognitive, and sensory predictions under one circuit, with the cerebellum supplying the brain with continuously updated forecasts of what should come next.
The most recent reconceptualisation comes from Rieger and colleagues (2023), who argued that action imagery is fundamentally a predictive process. Internal prediction is not a side effect of imagery: it is the essential mechanism that makes imagined practice useful for error detection and skill refinement. Without prediction, you would have no way to know an imagined rehearsal was wrong.
A senior leader in his late fifties came to me to rehearse the opening of a board meeting: the exact pause, the inhale, the first syllable. He could not physically rehearse without an audience and a stake. The architecture is the same one Lotze documented in 1999 and Sokolov, Miall, and Ivry mapped in 2017: his cerebellum was running predictions of how each beat of the opening should land, comparing them against his intended effect, and refining the timing model in the silence of his study.
What Is a Forward Model in Neuroscience?
A forward model is the brain’s internal predictor: a circuit that takes a copy of an outgoing motor command and computes, in advance, the sensory consequences that command should produce. The cerebellum runs forward models continuously, comparing each prediction against the actual sensory return and updating itself when the two diverge.
It is part of the broader work on peak performance systems that frames how skill is built without physical reps.
The forward-model concept was formalised by Grush in a 2004 Behavioral and Brain Sciences treatment that has shaped two decades of motor neuroscience. His emulation theory described the cerebellum as a Kalman-filter-like emulator that runs an off-line copy of the motor system. When you move, an efference copy of the command is sent to this emulator, which simulates the action’s sensory consequences in millisecond-scale advance. Perception then arrives, and the brain compares the simulation against reality.
Efference copy (a duplicate of the outgoing motor command, sent internally rather than to the muscles) is the input that lets the cerebellum predict without waiting for movement to occur. It is the same mechanism that explains why you cannot tickle yourself: the cerebellum predicts the sensory consequence of your finger touching your skin, the prediction matches the actual sensation, and the brain cancels the response. Tickle requires unpredictability the cerebellum cannot generate from your own efference copy.
In motor imagery the loop runs without the muscles. The motor system issues an action plan, the efference copy reaches the cerebellum, the forward model produces a sensory prediction, and an inhibitory gate at the spinal level stops the actual command from reaching muscle fibres. You get the prediction without the movement. van Kemenade and colleagues (2018) demonstrated the comparative architecture directly: the cerebellum, angular gyrus, and middle temporal gyrus each play distinct roles in monitoring whether imagined and executed actions match their predicted consequences, with the cerebellum carrying the timing-prediction load.
This is the structural answer to a question that bothers many of my high-performing clients: why does mental rehearsal feel useful when nothing physical is happening? Something physical IS happening. At the level of internal model computation, the cerebellum cannot tell the difference between an imagined rehearsal and a withheld execution. The forward model runs in both cases. The error signal runs in both cases. The model update runs in both cases.
The wider science of imagined movement is detailed in the neuroscience of motor imagery.
References
Sokolov, A. A., Miall, R. C., & Ivry, R. B. (2017). The Cerebellum: Adaptive Prediction for Movement and Cognition. Trends in Cognitive Sciences, 21(5), 313–332. https://doi.org/10.1016/j.tics.2017.02.005
Diedrichsen, J., Hashambhoy, Y. L., Rane, T. D., & Shadmehr, R. (2005). Neural Correlates of Reach Errors. Journal of Neuroscience, 25(43), 9919–9931. https://doi.org/10.1523/jneurosci.1874-05.2005
Galea, J. M., Vázquez, A., Pasricha, N., Orban de Xivry, J.-J., & Celnik, P. (2010). Dissociating the Roles of the Cerebellum and Motor Cortex during Adaptive Learning: The Motor Cortex Retains What the Cerebellum Learns. Cerebral Cortex, 21(8), 1761–1770. https://doi.org/10.1093/cercor/bhq246
Miall, R. C. (2024). Motor imagery, forward models and the cerebellum: a commentary on Rieger et al., 2023. Psychological Research. https://doi.org/10.1007/s00426-023-01916-7
Rieger, M., Boe, S. G., Ingram, T. G. J., Bart, V. K. E., & Dahm, S. F. (2023). A theoretical perspective on action consequences in action imagery: internal prediction as an essential mechanism to detect errors. Psychological Research. https://doi.org/10.1007/s00426-023-01812-0
Grush, R. (2004). The emulation theory of representation: Motor control, imagery, and perception. Behavioral and Brain Sciences, 27(3), 377–396. https://doi.org/10.1017/s0140525x04000093
Popa, L. S., & Ebner, T. J. (2019). Cerebellum, Predictions and Errors. Frontiers in Cellular Neuroscience, 12, 524. https://doi.org/10.3389/fncel.2018.00524
Lotze, M., Montoya, P., Erb, M., Hülsmann, E., Flor, H., Klose, U., Birbaumer, N., & Grodd, W. (1999). Activation of Cortical and Cerebellar Motor Areas during Executed and Imagined Hand Movements: An fMRI Study. Journal of Cognitive Neuroscience, 11(5), 491–501. https://doi.org/10.1162/089892999563553
van Kemenade, B. M., Arikan, B. E., Podranski, K., Steinsträter, O., Kircher, T., & Straube, B. (2018). Distinct Roles for the Cerebellum, Angular Gyrus, and Middle Temporal Gyrus in Action–Feedback Monitoring. Cerebral Cortex, 29(4), 1520–1531. https://doi.org/10.1093/cercor/bhy048
The cortical counterpart to cerebellar timing is unpacked in how mental rehearsal activates the motor cortex.
What the First Conversation Looks Like
When a high-performing client comes to me asking why their preparation isn’t translating into the moment, the work begins with timing: not content. I want to know what they rehearse, when they rehearse it, and whether what they rehearse is what should happen or when each part of it should land. The distinction is everything. Most accomplished people have rehearsed the words a hundred times and never once rehearsed the cadence. We start by separating those two layers, mapping the timing-critical decisions in their actual context, and building rehearsal that the cerebellum can use. The architecture I described in this article is not abstract; it is the substrate of the work.
Frequently Asked Questions
Does mental rehearsal of timing-critical skills actually rewire the cerebellum?
Yes: the cerebellum updates its forward model during imagined practice, driven by climbing-fibre prediction-error signals that fire whether the action is executed or simulated. Functional MRI work since the late 1990s has documented overlapping cerebellar activation during imagined and executed movement, and the contemporary framework treats action imagery as fundamentally predictive. Imagined timing rehearsal is not metaphorically rewiring the cerebellum; it is doing it through the same mechanism physical practice uses.
What is the difference between visual imagery and motor imagery for skill learning?
Visual imagery rehearses what an action looks like from the outside; motor imagery rehearses what producing the action feels like from inside the body. Only motor imagery, first-person kinesthetic, engages the cerebellum’s forward-model circuitry and produces the prediction-error updates that drive skill change. Watching yourself succeed in third person updates the visual system; feeling yourself produce the action in first person updates the motor system, including the cerebellar timing layer.
How does the cerebellum’s prediction error differ from dopamine prediction error?
They are biologically distinct signals in different circuits serving different functions. The cerebellum’s prediction-error pulse travels through climbing fibres from the inferior olive and updates motor and timing models. The dopamine prediction-error signal travels from midbrain dopamine neurons and updates reward-based action selection in the basal ganglia. Both are called prediction error, but conflating them obscures that one calibrates when an action should land, the other calibrates whether the outcome was worth pursuing.
Can mental rehearsal replace physical practice for timing-dependent skills?
Not entirely: but it can carry the rate-limiting step. The cerebellum updates its forward model from imagined practice, but the motor cortex requires consolidation through sleep and a smaller amount of physical practice to lock the adaptation into long-term retention. Mental rehearsal is most useful when physical rehearsal in the actual context is impossible (a board-meeting opening, a difficult family conversation, a high-stakes recital) and in those cases the cerebellar timing update is often the layer that closes the gap.
How long should a mental rehearsal session run for cerebellar adaptation?
Roughly 15 to 20 minutes of focused first-person kinesthetic imagery, repeated three to four times per week, layered onto whatever physical practice is available. Sessions longer than 20 minutes show diminishing returns because cognitive fatigue degrades imagery vividness and the prediction-error updates lose precision. The dose-response pattern across more than 100 documented motor-imagery interventions points to short, frequent, vivid sessions rather than marathon blocks, and crucially, sessions that rehearse timing, not just content.
