Basal ganglia: structure and computations

Abstract

Until recently the basal ganglia of the mammalian brain have attracted little attention from theoretical neurobiologists. Traditional views of the functioning of the basal ganglia are based on their biomedical importance in disorders such as Parkinson's disease. Their contribution to normal brain functions has remained poorly understood. Experimental investigations over the past few decades have produced a wealth of detailed information about the structure of the basal ganglia and the physiological properties of their component neurones. It has become evident that the basal ganglia play a role in the selection and performance of learnt behaviours, and also in the effects of reinforcement on acquisition and maintenance of new behaviours. At present it is difficult to link the symptoms of basal ganglia disorders to these basic facts, in part because very few theoretical models attempt to incorporate the information that is now available. Computational modelling can help to advance theoretical understanding in this area by establishing explicit links between different levels of organization: from the effects of neurotransmitters such as dopamine on synaptic plasticity, through the dynamic interactions within subpopulations of neurons, to system-level interactions between the basal ganglia and cerebral cortex. The aim of this review is to outline existing knowledge of the basal ganglia in relation to previous computer modelling work, and to suggest ways of making use of the new experimental findings in the next generation of models.