Basal Ganglia Control Of Reflexive Saccades: A Computational Model Integrating Physiology Anatomy And Behaviour

Abstract

It is hypothesised that the basal ganglia play a key role in solving the problem of action selection. Here we investigate this hypothesis through computational modelling of the primate saccadic oculomotor system. This system is an excellent target for computational modelling because it is supported by a reasonably well understood functional anatomy, has limited degrees of freedom, and there is a wealth of behavioural and electrophysiological data for model comparison. Here, we describe a computational model of the reflexive saccadic oculomotor system incorporating the basal ganglia, key structures in motor control and competition between possible actions. To restrict the likelihood of overfitting the model it is structured and parameterised by the known anatomy and neurophysiology along with data from a single experimental behavioural paradigm, then validated by testing against several additional behavioural experimental data without modification of the parameters. With this model we reproduce a range of fundamental reflexive saccadic results both qualitatively and quantitatively, comprising: the distribution of saccadic latencies; the effect of eccentricity, luminance and fixation-target interactions on saccadic latencies; and the effect of competing targets on saccadic endpoint. By investigating the model dynamics we are able to provide mechanistic explanations for the sources of these behaviours. Further, because of its accesibility, the oculomotor system has also been used to study general principle of sensorimotor control. We interpret the ability of the basal ganglia to successfully control saccade selection in our model as further evidence for the action selection hypothesis.