The subthalamic nucleus-external globus pallidus loop biases exploratory decisions towards known alternatives: a neuro-computational study.

Theories and models of the basal ganglia have mainly focused on the role of three different corticothalamic pathways: direct, indirect and hyperdirect. Although the indirect and the hyperdirect pathways are linked through the bidirectional connections between the subthalamic nucleus (STN) and the external globus pallidus (GPe), the role of their interactions has been mainly discussed in the context of a dysfunction (abnormal oscillations in Parkinson's disease) and not of its function. We here propose a novel role for the loop formed by the STN and the GPe. We show, through a neuro-computational model, that this loop can bias the selection of actions during the exploratory period after a change in the environmental conditions towards alternative responses. Testing well-known alternative solutions before completely random actions can reduce the time required for the search of a new response after a rule change. Our simulations further show that the knowledge acquired by the indirect pathway can be transferred into a stable memory via learning in the hyperdirect pathway to establish the blocking of unwanted responses. After a rule switch, first the indirect pathway learns to inhibit the previously correct actions. Once the new correct association is learned, the inhibition is transferred to the hyperdirect pathway through synaptic plasticity.