Abstract
Brain networks support learning across the multiple time scales over which intelligent behaviour unfolds. One of the key challenges in learning adaptive behaviour is the problem of temporal credit assignment: the process of identifying which set of past actions and observations, and their underlying neural representations, lead to the behavioural outcome observed in the present. In this thesis, I explore how interactions between the cerebellum and the cerebral cortex contribute to temporal credit assignment. The thesis is divided into two parts: in the first part a computational model of cerebro-cerebellar interactions for temporal credit assignment is developed. In this model a cerebellar, feedforward, network communicates with a cerebral, recurrent, network for efficient temporal credit assignment. The cerebellar signal, which contains information about future feedback that the cerebral cortex receives, influences the cerebral network such that appropriate activity patterns can be acquired for precise behaviour. The cerebellum learns to predict this feedback based on the neural representation in the cerebral cortex, thereby decoupling learning in cerebral networks from future feedback. When trained in a simple sensorimotor task the model shows faster learning and reduced dysmetria-like behaviours, in line with normal cerebellar function. These results indicate that cerebellar feedback predictions enable the cerebral cortex to acquire adaptive representations effectively by increasing the amount of temporal information available to each cerebral network.The cerebro-cerebellar model suggests that the cerebellum mediates behaviour by predicting feedback across a range of time scales. In the second part of the thesis I tested this hypothesis using an animal model. In particular I studied how the cerebellum contributes to interval timing, which is our ability to process temporal information in the seconds-to-minute range. The cerebellum is thought to be in volved in the generation and updating of internal models for control of movements with sub-second timing. Here I hypothesise that the cerebellum may also be involved in learning an internal model of supra-second stimulus time intervals. In order to test the predictive function of the cerebellum in the supra-second range, I trained rats to associate a sound duration with reward delivery. The effects of chemogenetic inactivation of cerebellar output from the lateral nucleus was investigated in expert rats. Analysis indicates that when internal time estimation is required animals show premature temporal judgements when cerebellar outflow is disrupted. This suggests that the cerebellar contributions to time processing are not restricted to sub-second intervals, but also include longer time intervals associated with cognitive processes such as decision making. Overall, this work provides a better understanding of how cerebro-cerebellar interactions support efficient temporal information processing.
| Date of Award | 9 May 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Richard Apps (Supervisor), Rui Ponte Costa (Supervisor), Jasmine Pickford (Supervisor) & Nadia L Cerminara (Supervisor) |