Multiple neural systems contribute to human decision making and learning, each optimized for distinct constraints. For example, the dopamine-dependent habit learning system, implemented in cortico-basal-ganglia loops, provides slow but robust and long-lasting learning of associations between stimuli and actions that lead to favored outcomes. In contrast, prefrontal-cortex dependent working memory affords one-trial learning of a very limited amount of information, for a short period of time. In this talk, I will present a new experimental design and hybrid computational model I developed to extract the separate contributions of these two systems to human reinforcement learning. Behavioral and computational results show that we can disentangle their contributions, but also study their interactions. Furthermore, using results from genes, EEG and fMRI experiments, I will show that we can relate the separable behavioral components of learning to different neural mechanisms. We can thus more precisely isolate the cause of learning deficits in patient populations, such as schizophrenia. These studies highlight the fact that multiple brain mechanisms contribute to simple single behaviors, and that careful experimental design and computational mechanisms are required to disentangle their contributions to behavior.