Background: Depression has been proposed as a predisposing factor for cancer, but prospective studies have been inconclusive. We examined whether a high level of depressive symptoms, present for a long time, is associated with increased risk of cancer in the elderly. Methods: Data were obtained and analyzed from persons who lived in three communities (Massachusetts, Iowa, and Connecticut) of the Established Populations for Epidemiologic Studies of the Elderly, a prospective cohort study with a mean follow-up of 3.8 years that included 4825 persons (1708 men and 3117 women) aged 71 years and older. Chronically depressed mood was defined as present when the number of depressive symptoms exceeded specific cut points on the Center for Epidemiologic Studies-Depression scale at baseline (1988) and 3 and 6 years before baseline. New cases of cancer were identified from Medicare hospitalization records and death certificates. Results: Of the 4825 persons studied, 146 (3.0%) were chronically depressed. The incidence rate of cancer was 30.5 per 1000 person-years for the 146 persons with chronic depression and 21.9 per 1000 person-years for the 4679 nonchronically depressed persons. After adjustment for age, sex, race, disability, hospital admissions, alcohol intake, and smoking, the hazard ratio for cancer associated with chronically depressed mood was 1.88 (95% confidence interval = 1.13-3.14). The excess risk of cancer associated with chronic depression was consistent for most types of cancer and was not specific to cigarette smokers. Conclusion: When present for at least 6 years, depression was associated with a generally increased risk of cancer.