Publication Summary and Abstract
Humphries, M. D. (2002), The basal ganglia and action selection: A computational study at multiple levels of description, PhD Thesis, University of Sheffield.
An action selection problem arises whenever stimuli (external or internal) requiring behavioural responses co-occur or overlap in time. It has been proposed that the basal ganglia, a collection of nuclei predominantly located in the forebrain, are a crucial part of the neural substrate of the action selection process. The work presented here builds on a previous computational model of intrinsic basal ganglia processing (Gurney et al. 2001a, 2001b) which demonstrated that its function was consistent with a selection mechanism. Two groups of extensions to this original model are described. First, structural extensions, comprising the addition of new internal connections or of structures extrinsic to the basal ganglia. Second, the instantiation of the extended model at a lower level of abstraction through the use of spiking model neurons. Results from the former group showed that the selection and switching capabilites of the basal ganglia were maintained following all biologically-constrained additions and demonstrated the existence of other action selection mechanisms in the brain instantiated outside of, but in contact with, the basal ganglia. These results support the hypothesis that the basal ganglia play a crucial role in action selection. Modelling at a lower level of detail provided four main results. First, selection and switching were maintained. Second, inputs which represented equally urgent actions caused oscillations in basal ganglia output which were consistent with rapid alternations of selected actions. Third, neural noise was required for selection of low-level inputs and, paradoxically, for ensuring that a model neuron's output was a coherent representation of its input. Fourth, the successful simulation of the results from an in vitro study of a basal ganglia sub-circuit provided evidence for a channel-based architecture within the basal ganglia. These results emphasise the utility of modelling the same neural circuit at different levels of abstraction.
Copy of the article (PDF)