Background and Purpose-The longer-term natural history of depression after stroke is poorly understood. We estimate frequency, predictors, and associations of depression up to 5 years after stroke in a population-based study. Methods-Data from 3689 patients registered in the South London Stroke Register 1995 to 2006 were used. Baseline data included age, sex, ethnicity, socioeconomic status, and stroke severity. At 3 months and at 1, 3, and 5 years, survivors were assessed for depression (Hospital Anxiety and Depression; depression subscale score Ͼ7 indicates depression), cognition, disability, activity, accommodation, employment, and social networks. Associations with depression were investigated with logistic regression. Data are reported with OR and 95% CI. Results-Depression frequencies were 33% (30%-36%), 28% (25%-30%), 32% (30%-35%), and 31% (27%-34%) at 3 months and at 1, 3, and 5 years after stroke, respectively. Forty-eight percent of patients were not depressed at any time point; 49% to 55% of depressed patients at 1 assessment remained depressed at follow-up; and 15% to 20% of patients at each assessment were new cases. Predictors of depression included stroke severity, inability to work, and impaired cognition. Associations with depression at follow-up included impaired cognition, lack of family support, institutionalization, inability to work, functional dependence, and low activity level. Conclusions-Frequency of depression up to 5 years after stroke is 30%; however, it is a dynamic situation with recovery and new cases diagnosed over time. These findings support the need for regular assessment of depression and its associated factors and for the development of effective interventions to reduce depression after stroke.