The basal ganglia and frontal cortex interact intimately to facilitate adaptive action plans while suppressing those that are less adaptive. The dynamics of this circuitry in reinforcement learning and decision making have been explored via a series of inter-related computational models. The models suggest distinct neurobiological mechanisms associated with (a) action selection; (b) learning the probability of an action leading to reward; (c) holding in mind graded values of reinforcement magnitude in working memory; and (d) dynamic modulation of decision thresholds. I will present novel predictions arising from these models that have been confirmed in experiments with multiple patient populations, pharmacological manipulation, neuroimaging and genetics.