← all systems · see it in The System
Cerebellum
The cerebellum (Latin for 'little brain') is a major feature of the hindbrain of all vertebrates. The human cerebellum does not initiate movement, but contributes to motor coordination, precision, and accurate timing: it receives input from sensory systems of the spinal cord and from other parts of the brain, and integrates these inputs to fine-tune motor activity.
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | >500 Mya (kind, hindbrain plan); gestational week 9 (one human cerebellum) |
| born precision | Myr (kind) |
| boundary | Inside: the cerebellar cortex (a thin sheet folded like an accordion into folia, with its molecular, Purkinje and granular layers), the white matter (arbor vitae) and the four deep cerebellar nuclei embedded in it (dentate, globose and emboliform = interposed, fastigial), in two hemispheres and the midline vermis, divided into ten lobules and three lobes. Outside: everything that sends it mossy fibres (pontine nuclei, spinal cord, vestibular nuclei and more), the inferior olive that sends its climbing fibres, the thalamus, red nucleus, vestibular nuclei and reticular formation its outputs reach, and the cerebral cortex its loops close on. All its connections run through the three paired cerebellar peduncles. |
| nested level | Part of the hindbrain (metencephalon) of the human brain, at the back of the skull below the cerebrum. Holds the prediction half of human-brain's L6 'selection and prediction' floor and the cerebellar lane of the brain's proprioception and balance channels; feeds the brain's movement outlet and speech (its damage gives dysarthria). |
| scale | organism |
Measures (30) published science
| neurons in the human cerebellum | 69 billion neurons |
| mass of the human cerebellum | 154 g |
| share of brain volume | 10 % |
| ratio of cerebellar to neocortical neurons | about 3.6, conserved across many mammals ratio |
| cerebellar cortical surface area | 1,590 (78% of neocortex) cm² |
| cerebellar cortical surface area (encyclopedia figure) | 730 cm² |
| cerebellar area relative to neocortex in macaques | at least 33 % |
| unfolded cerebellar cortex | 10 cm wide, almost 1 m long |
| mossy fibres | 200 million fibres |
| granule cells | approximately 60 billion cells |
| granule cells per mossy fibre | 200 to 1 ratio |
| Purkinje cells | 15 million cells |
| spines on one human Purkinje cell | up to 200,000 spines |
| climbing-fibre synapses on one Purkinje cell | up to 300 synapses |
| Purkinje cells per microzone | on the order of 1000 cells |
| Purkinje cells converging on one deep nuclear cell (cat) | about 860 cells |
| parallel fibre length | about 6 mm |
| Purkinje simple-spike rate in awake animals | around 40 Hz |
| complex-spike baseline rate | around 1 (never much above 10) Hz |
| pause after a complex spike | several hundred ms |
| Purkinje-cell storage capacity | up to 5 kilobytes, 40,000 input-output associations |
| silent synapses in an optimal perceptron | more than 50 % |
| time for a VOR memory to move beyond the cerebellar cortex | under 60 min still cortex-dependent; after 3 days barely affected by cortical blockade |
| cerebellar signalling energy on resting potentials | 54 % |
| cerebellar signalling energy on postsynaptic receptors | 22 % |
| cerebellar signalling energy on action potentials | 17 % |
| share of cerebellar cortical signalling energy used by Purkinje cells | 15 % |
| start of human cerebellar development | around gestational week 9, continuing beyond birth weeks |
| lateral hemispheres, humans and apes against monkeys | 2.7 times greater ratio |
| autism risk ratio after cerebellar injury | 36 x |
Open contentions (8) open contention
Where sources or readings disagree. Both stay on the map.
- C-413 Where do thoughts come from?
booked: No theory is proven, so the map lists the main ones side by side and doesn't pick one. It shows where thinking is handled in the brain, not where thoughts start.
the other reading: one more position for the registry, labelled as a proposal: the cerebellum generates optimized mental models whose updates are experienced in cortex as creative intuition ('a ha') (wp) - C-440 Does the body-position information sent to the cerebellum ever reach awareness?
booked: The map shows this information going into the cerebellum below awareness, and also shows the cerebellum's onward route to the cortex, which one 2023 review calls a main path for body sense.
the other reading: registered in C-440 - C-823 Does motor learning happen inside the cerebellum?
booked: The map follows the evidence that learning happens inside the cerebellum, and keeps the view that it only signals learning elsewhere alongside.
the other reading: the cerebellum only provides signals that promote learning elsewhere; Marr's sign (strengthening); the LTD evidence 'remains controversial' (wp) - C-824 How large is the surface of the cerebellum?
booked: The map uses 1,590 cm², 78% of the neocortex's area, from a 2020 detailed reconstruction, and keeps the earlier 730 cm² alongside.
the other reading: 730 cm² (wp, high-resolution MRI) - C-825 Does the cerebellum do the same computation everywhere?
booked: The map records one repeated circuit used by nearly independent modules, and lists what it might compute without ruling.
the other reading: a universal cerebellar transform, one computation everywhere (Schmahmann 1996, as cited in Diedrichsen 2019; Doya per wp); multiple functionality, the same circuit implementing distinct computations (Diedrichsen 2019); forward model / state estimation (Ito 2008, Hull 2020); timing system (Ivry); sensory-to-motor coordinate transform (Pellionisz & Llinas); fundamentally sensory (Bower) - C-826 What do the cerebellum's climbing fibres signal?
booked: The map follows most researchers: they signal errors or surprises and drive learning. Other readings are kept alongside.
the other reading: the climbing fibres shape cerebellar output directly; or the message lies in ensemble synchrony and rhythm ('a drawing by Escher', wp) - C-827 Do lampreys and hagfish have a cerebellum?
booked: The map records that the brain plan the cerebellum grows from goes back more than 500 million years, and keeps the view that these fish have barely any distinct cerebellum alongside.
the other reading: in lampreys and hagfish the cerebellum is barely distinguishable from the brainstem, small paired nuclei like the vestibulocerebellum (wp) - C-830 How many granule cells are in the cerebellum?
booked: The map says tens of billions and keeps all the published figures (40, about 50, and about 60 billion).
the other reading: see booked: the sources give different numbers
Partners and parts published science
Built from / builds (14)
- Vertebrate brain
- Purkinje cells
- Cerebellar granule cells
- Golgi cells (and Lugaro and unipolar brush cells)
- Stellate and basket cells
- Bergmann glia
- Deep cerebellar nuclei
- Mossy fibres
- Climbing fibres
- Human brain
- Memory
- Classical conditioning
- Speech (human spoken language)
- Learning
Sits inside / holds (2)
Trades with (valence) (12)
- Spinal cord via sys:electrochemical-gradient · Cerebellum gain
- Human (Homo sapiens) via sys:sensory-environment · Cerebellum gain
- Thalamus via sys:electrochemical-gradient · Cerebellum gain
- Human brain via sys:electrochemical-gradient · Cerebellum gain
- Blood via sys:blood · Cerebellum gain
- Ventral tegmental area via sys:electrochemical-gradient · Cerebellum even
- Vestibular system via sys:gravitational-field · Cerebellum gain
- Inferior olive via sys:electrochemical-gradient · Cerebellum gain
- Basal ganglia via sys:electrochemical-gradient · Cerebellum even
- Memory via sys:electrochemical-gradient · Cerebellum gain
- Classical conditioning via sys:electrochemical-gradient · Cerebellum gain
- Learning via sys:electrochemical-gradient · Cerebellum gain
expressed by (12)
- Encapsulation & Modularity (Abstraction)
- Distributed Autonomy/Decentralized Execution (Subsidiarity)
- Internal Coordination/Regulation
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Multi-Scale Data Fusion / Triangulation
- Trail-Marking / Externalized Memory (Stigmergy)
- Surprise / Change-Point Detection
- Quick Harm Avoidance
- Proactive Defense
- Resilience & Repair
- Quality Control & Error Correction
floor filled by (19)
- Human (Homo sapiens)
- Thalamus
- Human brain
- Vestibular system
- Purkinje cells
- Cerebellar granule cells
- Golgi cells (and Lugaro and unipolar brush cells)
- Stellate and basket cells
- Bergmann glia
- Deep cerebellar nuclei
- Mossy fibres
- Inferior olive
- Pontine nuclei
- Memory
- African elephant (genus Loxodonta: bush and forest elephants)
- Classical conditioning
- Speech (human spoken language)
- Learning
- Human voice
Kinds (9)
- Vestibulocerebellum (flocculonodular lobe)
- Spinocerebellum (vermis and paravermis)
- Cerebrocerebellum (lateral hemispheres)
- Mammalian cerebellum
- Cerebellum of cartilaginous and bony fishes
- Mormyrid valvula cerebelli
- Lamprey and hagfish cerebellum
- Cerebellum-like structures
- Cephalopod cerebellum analogue
wired (7)
- Procedural memory (skills and habits)
- Cerebellar nuclei
- Sensorimotor cortex
- Amygdala
- Neocortex (six-layered isocortex)
- Purkinje cells
- Inferior olive
Notes published science
- Unfold the wrinkled cerebellum and you get a ribbon about 10 cm wide and almost a metre long, with 78% of the surface area of the whole neocortex packed into a fist-sized lump at the back of the head; in a macaque it is only about a third (src:sereno-2020).
- You can't tickle yourself because your cerebellum predicts exactly what your own fingers will feel and cancels it: in the scanner, the touch areas fire less for self-made tickles than for someone else's (src:blakemore-1998).
- By the epigenetic clock the cerebellum is the slowest-ageing part of the body: in the oldest old it reads younger than its owner (src:horvath-2015).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
signals into and out of the cerebellum (reader's view)
- L1: What arrives: the body's own movement (muscle length, tension, joints), head motion and gravity, light and sound, pain, copies of the cerebral cortex's commands and plans, reward, and the mismatches when things go wrong
- L2: The two input systems: about 200 million mossy fibres (from the pontine nuclei, spinal cord, vestibular nuclei and elsewhere) and the climbing fibres of the inferior olive, plus modulatory fibres; all enter through the three paired peduncles
- L3: Granular layer: glomeruli where mossy-fibre rosettes meet granule-cell claws under Golgi-cell inhibition; expansion recoding into tens of billions of granule cells
- L4: Molecular and Purkinje layers: parallel fibres through the flattened Purkinje dendrites, one climbing fibre per Purkinje cell, stellate and basket inhibition; plasticity at parallel fibre-Purkinje synapses under the climbing-fibre teaching signal
- L5: Deep cerebellar nuclei (and vestibular nuclei for the flocculus): the only output; Purkinje inhibition shapes their firing
- L6: Modules and internal models: what each module does for the system it loops with, calibrating movement and, through loops with prefrontal and association cortex, thought, language and emotion; predictions compared with what follows
the boardroom (shared floors)
- L3: Granular layer: mossy-fibre rosettes in glomeruli, each contacted by up to 20 granule-cell claws and infiltrated by Golgi-cell inhibition; 200 million mossy fibres are recoded into tens of billions of granule cells (200 to 1), which Marr proposed encode combinations of inputs
- L4: Molecular and Purkinje layers: parallel fibres cross the flattened Purkinje dendrites at right angles (up to 200,000 spines per cell) while each Purkinje cell gets exactly one climbing fibre; stellate and basket cells inhibit; the parallel-fibre synapses change strength under the climbing-fibre teaching signal; Purkinje cells fire simple spikes (~40 Hz) and climbing-fibre complex spikes (~1 Hz) and send inhibition (GABA) to the nuclei. Organised in microzones of ~1000 Purkinje cells sharing one somatotopic receptive field
- L5: Deep cerebellar nuclei (dentate, interposed, fastigial) and, for the flocculus, the vestibular nuclei: the sole output, combining Purkinje inhibition with mossy-fibre and climbing-fibre collaterals; a second site where synaptic strength is modified
- L6: Modules and their internal models: hundreds to thousands of near-independent modules (multizonal microcomplexes: an olive cluster, its microzones and a nuclear cluster), each built the same way; read by the field as forward models that predict the sensory consequences of a command and are compared with what follows. Whether every module runs one computation (a universal cerebellar transform) or the same circuit runs several (multiple functionality) is contested
way in: Balance and head motion (vestibular mossy fibres)
- L1: Angular and linear head motion, and gravity
- L2: Semicircular canals and otolith organs; vestibular afferents and the vestibular nuclei send mossy fibres through the inferior peduncle, mainly to the flocculonodular lobe
way in: Blood supply (oxygen and glucose that pay for signalling)
- L1: Three paired arteries: superior cerebellar (SCA), anterior inferior (AICA) and posterior inferior cerebellar (PICA)
- L2: The energy budget of the cerebellar cortex: most signalling energy is spent on resting potentials and postsynaptic receptors; Purkinje cells use only 15%
way in: Climbing fibres from the inferior olive (errors and unexpected events)
- L1: Mismatch in the world and body: retinal slip during head turns, a reach that goes wrong, an air puff to the cornea, an unexpected event
- L2: Inputs from the spinal cord, brainstem and cerebral cortex converge on the inferior olive; for the eyelid response a hardwired pathway from the cornea
- L3: Inferior olive neurons, coupled by gap junctions within a microzone's cluster, fire at low rates and send one climbing fibre (about 10 branches) per Purkinje cell, with collaterals to the deep nuclei
way in: From the cerebral cortex (cortico-ponto-cerebellar mossy fibres): copies of motor commands, and the inner channels
- L1: Activity of the cerebral cortex: motor areas (a copy of the current motor command) and association areas such as prefrontal area 46
- L2: Pontine nuclei relay it through the middle cerebellar peduncle, the largest of the three
way in: Global modulation (dopaminergic, serotonergic, noradrenergic and cholinergic fibres)
- L1: The brain's state (arousal and the like) carried by monoamines and acetylcholine
way in: Pain (cortico-cerebellar and spino-cerebellar)
- L1: Painful stimuli on and in the body
- L2: Descending cortico-cerebellar and ascending spino-cerebellar pathways, arriving through the pontine nuclei (mossy fibres) and the inferior olives (climbing fibres)
way in: Proprioception, spinocerebellar lane (mossy fibres)
- L1: Muscle length and tension and joint dynamics: the body's own mechanical state as it moves
- L2: Muscle spindles, Golgi tendon organs, joint and cutaneous receptors; dorsal and ventral spinocerebellar and cuneocerebellar tracts, and the trigeminal nerve for the face, carry it up through the inferior peduncle
way in: Reward and its expectation (granule-cell and climbing-fibre signals)
- L1: Reward delivered, expected or omitted (sugar water in the mouse studies), the outcome of the animal's own actions
- L2: Mossy-fibre and climbing-fibre pathways whose sources are not yet known: Hull lists how reward signals reach granule cells and climbing fibres as an open question
way in: Sight and hearing (visual and auditory mossy-fibre input)
- L1: Light and sound reaching the eyes and ears
- L2: Visual and auditory systems feed the spinocerebellum and the vestibulocerebellum; in humans the vermis is functionally coupled to primary visual cortex with a representation of the central visual field
way out: Back to the inferior olive (nucleo-olivary inhibition)
- L1: Inferior olive neurons, the climbing channel's L3
- L5: Small GABAergic nuclear cells project only to the inferior olive, to the same cluster that sends them climbing fibres
way out: To the cerebral cortex (dentate → ventrolateral thalamus → motor and prefrontal cortex)
- L1: Motor cortex (arm area of M1) and dorsolateral prefrontal cortex (area 46), each fed by its own region of cerebellar cortex
- L2: Ventrolateral thalamus relays to premotor and primary motor cortex, and to prefrontal area 46
- L5: Dentate nucleus (and interposed), glutamatergic projection cells
- L6: The module's prediction or correction, for the cortical area that sent the input
way out: To the descending motor systems (interposed → red nucleus; fastigial → vestibular nuclei and reticular formation)
- L1: Coordinated, accurately timed movement of body, limbs and the speech articulators; when it fails, ataxia: dysmetria, intention tremor, dysarthria, broad-based gait
- L2: Red nucleus, vestibular nuclei and reticular formation: descending motor systems to the spinal cord
- L5: Fastigial and interposed nuclei (spinocerebellum) project to the brainstem
- L6: Calibration of the detailed form of a movement (force, direction, speed, amplitude, timing), not its initiation
way out: To the eyes (flocculus → vestibular nuclei → oculomotor neurons)
- L1: Eye movements that hold the image steady on the retina when the head turns (the vestibulo-ocular reflex)
- L4: Floccular Purkinje cells, the only cortex that projects outside the deep nuclei, straight to the vestibular nuclei
- L6: The vestibulocerebellum's calibration of gaze against head motion
way out: To the reward circuit (lateral cerebellum → ventral tegmental area)
- L1: The ventral tegmental area's dopamine neurons, the reward-learning system of the striatum
- L5: Lateral cerebellar nuclei with a direct, monosynaptic connection to the VTA (mice)
Sources (24) published science
Blakemore S.-J., Wolpert D.M., Frith C.D. (1998) Central cancellation of self-produced tickle sensation. Nat Neurosci 1:635. doi:10.1038/2870 (behind its links)
Hull C. (2020) Prediction signals in the cerebellum: beyond supervised motor learning. eLife 9:e54073. doi:10.7554/elife.54073 (PMC7105376, CC BY) (behind its links)
Herculano-Houzel S. (2009) The human brain in numbers: a linearly scaled-up primate brain. Front Hum Neurosci 3:31. doi:10.3389/neuro.09.031.2009 (PMC2776484) (behind its links)
Kassardjian C.D., et al. (2005) The site of a motor memory shifts with consolidation. J Neurosci 25:7979. doi:10.1523/jneurosci.2215-05.2005 (PMC6725450) (behind its links)
Leto K., et al. (2016) Consensus paper: cerebellar development. Cerebellum 15:789. doi:10.1007/s12311-015-0724-2 (PMC4846577) (behind its links)
Wang S.S.-H., Kloth A.D., Badura A. (2014) The cerebellum, sensitive periods, and autism. Neuron 83:518. doi:10.1016/j.neuron.2014.07.016 (PMC4135479) (behind its links)
Memory (Wikipedia, rev 1380372913, 2026-10-15, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Christian KM, Thompson RF 2003. Neural substrates of eyeblink conditioning: acquisition and retention. Learn Mem. doi:10.1101/lm.59603 (behind its links)