A new conceptual model of fronto-striatal interaction is described. It is suggested that frontal cortex encodes environmental cues that are used in the preparation of sequential motor programs as states of a temporally changing control signal. It is proposed that the basal ganglia use this information to facilitate desired motor patterns selectively and to inhibit competing representations. Similarity relations between states of the control signal are
assumed to cause interference amongst motor patterns. In the model, reduction of dopamine causes both a degradation in the quality of the control signal and a failure to resolve competition at output. In computer simulation studies the properties and behaviour of two contrasting connectionist architectures are explored as implementations of the ideas above. The results of damaging each model were investigated to model the reduction of dopamine (as occurs in Parkinson's disease) during the performance of learned movement sequences. In
one network the sequential behaviour is driven by recurrent connections from the output of a 'forward model'. The specific effects of damaging the forward model were investigated. A dynamical systems analysis of the patterns of motor interference both in the network and in Parkinson's disease is provided. In the other architecture that is explored, the control signal is composed of multiple endogenous oscillators. This model focuses on the interaction between the control signal and an explicit competitive action selection· process. It is shown that degrading either the quality of the control signal or disrupting the competitive processing can yield a variety of deficits that model parkinsonian impairments. An explicit computational account of how cortical and basal ganglia systems interact to subserve both sequencing and selection functions in normal behaviour is developed and it is shown how reduction of available dopamine gives rise to the particular pattern of deficits observed in Parkinson's
disease.