• Robots

A model of intrinsic processing in the basal ganglia

K. Gurney, P. Redgrave, and T. J. Prescott

REFERENCE: Gurney, K., Redgrave, P. and Prescott, T. J. (1998) A model of intrinsic processing in the basal ganglia. Society for Neuroscience Annual Meeting, 24, Los Angeles, 640.13.

NOTE: The poster panel numbers are enclosed in brackets - clicking on them will open a panel window.

Introduction

We have recently suggested (Redgrave et al 1997) that the basal ganglia (BG) play a crucial role in solving the action selection problem in vertebrates [1]. We now present a quantitative model of intrinsic BG processing which shows how this might occur. While much is known of the anatomy and physiology of the basal ganglia, comparatively little is known of the computational operations conducted within its functional architecture. The prevailing model (Albin et al 1989) - framed in terms of ‘direct’ and ‘indirect’ pathways - has several shortcomings [2]. First, it fails to emphasise the significance of several anatomically important pathways. Second, it remains to be developed as a full computational model so that its claims remain unsubstantiated. Third, It fails to accommodate recent clinical and animal data (Chesselet and Delfs 1996, Parent 1998). Given this state of affairs we believe the functional architecture of the BG is ready for review, an opinion shared by the authors of the current model (Albin et al 1995).

Aims

Methods

BG connectivity is very rich, and a principled approach is required to tease apart the strands of functionality within it. Starting from the computational premise of action selection, we examined the neural network architectures available for selection [3] and looked for correspondences with BG anatomy [4]. We then used the resulting architecture to build a computational model and subjected this to analysis of its equilibrium properties and general simulation.

Results

Conclusions

The functional decomposition of the BG architecture we have suggested here is supported by systemically elegant features (e.g. signal normalisation, synergistic DA modulation) and the model’s consistency with both neurophsyiological and behavioral data. We believe it will provide the basis for a programme to discover how the BG works to select actions in a robust, adaptive and hierarchical way.