Figure 4.

Contribution of cerebellar computation to navigation. Cerebellar cortex as an adaptive filter could transform an input signal consisting of self-motion sensory signals and a motor efference copy into a prediction of the sensory signal expected from voluntary movements and the previous sensory state. By comparing the sensory prediction with the actual sensory signal, the cerebellar output helps the system to detect novelty in the environment. Thanks to a parallel output sent to the inferior olive, this novelty signal is modified into a teaching signal which, if repeated in correlation with the sensory inputs, could contribute to modify the cerebellar model of prediction through LTD. Reciprocally, repeated inputs to the cerebellum may trigger LTP at the synapses gating those inputs and act as an unsupervised learning of the most relevant inputs. Depending on the type of signal (vestibular, tactile…) and possibly on the targeted lobules (IX-X vs. VII, crus I, and II), the output of the cerebellar computation could cancel the signal induced by voluntary self-motion and allow detection of novelty. This computation could have a double role: stabilizing perception during voluntary navigation and informing the navigator about the necessity to update his/her relative position in the context.