Mood disorders represent a major public health and economic burden in the United States. Major depressive disorder (MDD) and bipolar disorder (BP) are chronic, debilitating diseases and existing strategies for relapse prediction and prevention could be improved. A clearer understanding of reward functioning, which is highly disrupted in active phases of MDD and BP, during the euthymic phase of illness may improve our model of how these disorders manifest and recur. Reward functioning deficits are documented extensively in the literature, but few studies have evaluated and directly compared self-report, task, and neural facets of reward functioning in a combined sample of remitted MDD and euthymic BP individuals over time. We hypothesized that, individuals with a history of mood disorders (HMD) would have lower trait positive affect, worse performance on reward processing tasks, increased connectivity from key regions in the reward circuit to the salience and emotion network (SEN) and default mode network (DMN), and more homogeneity in the SEN and DMN. We expected baseline self-reported reward responsiveness to predict fluctuations in mood symptoms over time in the HMD group and moderate the effect of diagnosis on mood symptoms over time. We explored whether reward task performance predicted depressive symptom change in a cross-lagged manner over time in the HMD group. Participants were 132 individuals with HMD and 42 healthy controls (HC). After undergoing symptom and diagnostic assessment by a clinician, participants completed resting-state and task-based fMRI, a neuropsychological battery including a reward task, and self-report measures of symptoms, reward and affective functioning. A subset of participants completed two additional study visits over a follow-up period of approximately three years, repeating the clinician assessments, self-report measures, and behavioral tasks. Self-reported affect and reward functioning differentiated the HMD and HC groups, whereas objective measurements of reward failed to differentiate groups or to predict naturalistic symptom fluctuation. Resting-state connectivity differences emerged that suggested increased connectivity between reward nodes and salience regions and decreased connectivity between reward nodes and default mode regions in HMD participants. Homogeneity, as assessed by variance between reward seeds and the SEN and DMN, did not differ between groups. Our hypotheses regarding moderators and mediators were not supported. Overall, rsFC may be more sensitive to trait biomarkers of disease, relative to behavioral performance and task-based fMRI findings.