Dopaminergic computations for perceptual decisions

Abstract

Studies linking the brain’s dopamine signals with learning and decision making have enjoyed enormous progress using predominantly value-based decision-making tasks. However, recent studies have demonstrated pervasive dopamine signaling also during perceptual decision making. These signals have been shown to depend on both feedback and perceptual parameters, such as perceptual decision confidence and sensory statistics. Here, we review recent studies investigating dopamine signals in simple and complex forms of perceptual decision tasks across species and dopaminergic circuits. We discuss how reinforcement learning (RL) models can account for key aspects of learning during perceptual decision making and its dopaminergic underpinnings, thus bridging the gap with the literature on dopamine in value-based decisions. Finally, we propose that RL may provide a promising framework to address current challenges in the dopamine literature, such as explaining the function of its heterogeneous responses and its role in learning from naive to expert.