Dopamine Contributions to Motivational Vigor and Reinforcement Driven Learning.

Abstract

Brain mechanisms for reinforcement learning and adaptive decision-making are widely accepted to critically involve the basal ganglia (BG) and the neurotransmitter dopamine (DA). DA is a key modulator of synaptic plasticity within the striatum, critically regulating neurophysiological adaptations for normal reinforcement driven learning, and maladaptive changes during disease conditions (e.g. drug addiction, Parkinson’s disease). Activity in midbrain DA cells are reported to encode errors in reward prediction, providing a learning signal to guide future behaviors. Yet, dopamine is also a key modulatory of motivation, invigorating current behavior. Prevailing theories of DA emphasize its role in either affecting current performance, or modulating reward-related learning. This thesis will present data aimed at resolving gaps in the literature for how DA makes simultaneous contributions to dissociable learning and motivational processes. Specifically, I argue that striatal DA fluctuations signal a single decision variable: a Value function (an ongoing estimate of discounted future rewards) that is used for motivational decision making ('Is It worth it?') and that abrupt deflections in this value function serve as temporal-difference reward prediction errors used for reinforcement/learning ("repeat action?”). These DA prediction errors may be causally involved in strengthening some, but not all, valuation mechanisms. Furthermore, DA activity on the midbrain-forebrain axis indicate a dissociation between DA cell bodies and their striatal terminals. I propose that this is an adaptive computational strategy, whereby DA targets tailor release to their own computational requirements, potentially converting an RPE-like spike signal into a motivational (value) message.