Correlational learning rule for regulation of vestibular synapse strength by cerebellar inhibition: transfer of VOR gain from cerebellar cortex to brainstem.

<p>A: Schematic diagram of vestibular and cerebellar inputs to MVN neuron. The input from the vestibular periphery (V) arrives at an excitatory synapse with weight . The input from the cerebellum (C) arrives at an inhibitory synapse with nominal weight 1. The MVN output is treated as the linear combination of these two inputs. The input to the cerebellum denotes the retinal-slip training signal assumed to mediate learning in cerebellar cortex. B: Illustration of situation for gain down learning. The brainstem gain is too high so that the vestibular input on its own would produce an over-large vestibulo-ocular response. During training the cerebellum has learnt to produce an inhibitory input modulated in phase with the vestibular input, leading to cancellation which produces a 50% gain decrease (since and , a weight requires ). The same result could be produced without the need for cerebellar input if the vestibular afferent synaptic strength (represented by the weight ) was reduced during consolidation to half its value (since requires ). Hence in-phase (positively correlated) inputs to the MVN should drive a long-term depression (LTD) at vestibular afferent synapses during consolidation. C: Illustration of situation for gain up learning. In this case the out-of-phase cerebellar and vestibular afferent inputs to the MVN neurons should drive a long-term potentiation (LTP) of the vestibular afferent synapses during consolidation.