Cerebellar learning: adapting the vestibulo-ocular reflex (VOR)

A trial-by-trial rate model of the flocculus, vestibular nuclei and inferior olive. Learning happens at two sites: quickly in the flocculus (Ito's hypothesis) and slowly in the vestibular nuclei under Purkinje cell guidance (Miles and Lisberger's hypothesis). It is schematic: rates and time constants are in physiological ranges but are not fitted to any data set.

Circuit

Swipe sideways to see the whole diagram.

CEREBELLUM (FLOCCULUS) BRAINSTEM AND EYE mossy fibres vestibular input (plastic: brainstem site) parallel fibres (plastic: flocculus site) climbing fibre direct VOR pathway eye rotation seen through ×2 lenses Granule cells → parallel fibres head-velocity signal at successive moments of the turn Interneurons basket / stellate Purkinje cell flocculus simple spikes Semicircular canals → vestibular afferents Target neurons vestibular nucleus Relay neurons vestibular nucleus · fixed Eye motor neurons abducens / oculomotor Retina image slip Pretectum / AOS direction-selective motion detectors Inferior olive dorsal cap · retinal slip

Click any cell group or connection to read about it below. Moving dots are spikes, slowed down while a trial plays (thinned on the busiest pathways); each climbing-fibre spike is a single large dot.

Current trial: one head turn

Swipe sideways to see the whole trial.

Experiments

Paused
Spectacles (image magnification)
Vision
Manipulations
Trial
0
naive
VOR gain
1.00
eye ÷ head velocity
Climbing-fibre response
0%
chance per turn
Purkinje at peak turn
60 Hz
resting 60 Hz
Brainstem weight
1.00
vestibular → target

Learning across trials

Flocculus synaptic weights for each parallel-fibre group