The dysfunction of reinforcement learning is emerging as a transdiagnostic dimension of mood and anxiety. Computational models of reinforcement learning may expedite our ability to identify predictors of response, thereby improving efficacy rates. We will will, first, examine the neural substrates of reinforcement learning in depression and anxiety, and, second, test a computational model of reinforcement learning as a predictor of response to an adapted version of behavioral activation psychotherapy. Subjects (N=10) will be enrolled in a two week evaluation, followed with a nine week weekly intervention program. Assessments will be conducted at baseline, and during the intervention as the 3-, 6-, 9-week follow-ups. Reinforcement learning will be measured using 3T magnetic resonance imaging during a computer task. All other measures include structured clinical interviews, questionnaires, and computer tasks.
