Conference on Cognitive Computational Neuroscience

Dopaminergic projections originating in the ventral tegmental area (VTA) have long been associated with learning and adaptive behavior. The VTA is thought to mediate information encoding of its target neurons. This function is traditionally linked to the neuromodulatory property of dopamine. This work provides a network perspective on the function of the VTA. In silico experiments suggest that the function of the VTA is a product of its topological position between multimodal associative areas providing input signal indicating reward availability and prefrontal areas encoding behaviorally relevant information. A neural network model of the cortico-basal ganglia-thalamo-cortical loop is optimized to perform inference. Replacing elevated activity indicating reward in the VTA with baseline activity prevents encoding of an outcome in the prefrontal cortex. Furthermore, the same model fails to optimize without the reward signal even though all necessary information is available.