The ability to select actions based on internalized goals is a significant domain of animal fitness, and particularly crucial in humans. These behaviors are guided by an ability to weigh the positive and negative effects of an action and to learn from experience. Individuals with neuropsychiatric disorders share common defects in this cognitive domain, yet a circuit understanding of this computational dysfunction is unclear. Further progress requires a closer association between the genes that cause neuropsychiatric disorders and the circuits that underlie observed abnormalities. In this thesis, I begin by with an overview of current nosological and etiological understanding of neuropsychiatric disease as well as current challenges in developing circuit hypotheses of dysfunction. I move on to characterize a quantitative multidimensional behavioral assay in mice that gives key insight into value-based action in this model system. Because of its role in regulating motor output and reinforcement learning, the striatum was identified as a potential circuit junction mediating critical cognitive computations. In vivo imaging of the direct and indirect pathway of the dorsomedial striatum revealed broad overlap in encoding reward costs and benefits in these cell populations, with the indirect pathway acting as a circuit substrate for cost-benefit interactions. Finally, we leveraged these techniques to characterize goal-directed dysfunction in the Nrxn1α model of neuropsychiatric dysfunction. We isolated this deficit to excitatory projections from forebrain regions using conditional region-specific ablations of Nrxn1α. In these mice, we observed abnormalities in encoding features of reward that serve as the circuit correlate to observed choice abnormalities. In sum then, this thesis attempts to synthesize quantitative behavioral, genetic and in vivo physiological techniques to characterize a circuit intermediary between genetic mutations and neuropsychiatric cognitive symptoms.