Findings in previous research on the association of old age and depression are inconsistent due to a confounding of age changes and cohort differences. Using data from an accelerated longitudinal design from the National Institute of Aging Established Populations for Epidemiologic Studies of the Elderly, this study addresses three questions: (1) Does the age growth trajectory show an increase in depressive symptoms in late life? (2) Is there cohort heterogeneity in levels of depressive symptoms and age growth trajectories of depressive symptoms? (3) What social risk factors are associated with these effects? Results show evidence of substantial cohort variation in depression. There is also evidence for an age-by-cohort interaction effect. Specifically, depression declined with age more rapidly for earlier cohorts. The growth trajectories can be accounted for by factors associated with historical trends in education, life course stages, health decline, differential survival, stress, and coping resources. (Contains 10 notes, 3 tables, and 1 figure.)