Encoding and control of motor prediction and feedback in the cerebellar cortex

Abstract

Extensive research implicates the cerebellum as a forward internal model that predicts the sensory consequences of motor commands and compares them to their actual feedback, generating prediction errors that guide motor learning. However, lacking is a characterization of how information relevant to motor control and sensory prediction error is processed by cerebellar neurons. Of major interest is the contribution of Purkinje cells, the primary output neurons of the cerebellar cortex, and their two activity modalities: simple and complex spike discharges. The dominant hypothesis is that complex spikes serve as the sole error signal in the cerebellar cortex. However, no current hypotheses fully explain or are completely consistent with the spectrum of previous experimental observations. To address these major issues, Purkinje cell activity was recorded during a pseudo-random manual tracking task requiring the continuous monitoring and correction for errors. The first hypothesis tested by this thesis was whether climbing fiber discharge controls the information present in the simple spike firing. During tracking, complex spikes trigger robust and rapid changes in the simple spike modulation with limb kinematics and performance errors. Moreover, control of performance error information by climbing fiber discharge is followed by improved tracking performance, suggesting that it is highly important for optimizing behavior. A second hypothesis tested was whether climbing fiber discharge is evoked by errors in movement. Instead, complex spikes are modulated predictively with behavior. Additionally, complex spikes are not evoked as a result of a specific ‘event’ as has been previously suggested. Together, this suggests a novel function of complex spikes, in which climbing fibers continuously optimize the information in the simple spike firing in advance of changes in behavior. A third hypothesis tested is whether the simple spike discharge is responsible for encoding the sensory prediction errors crucial for online motor control. To address this, two novel manipulations of visual feedback during pseudo-random tracking were implemented to assess whether disrupting sensory information pertinent to motor error prediction and feedback modulates simple spike activity. During these manipulations, the simple spike modulation with behavior is consistent with the predictive and feedback components of sensory prediction error. Together, this thesis addresses a major outstanding question in the field of cerebellar physiology and develops a novel hypothesis about the interaction between the two activity modalities of Purkinje cells.