This study examined cross-sectional and prospective associations of exercise with depressed mood in a community-based sample of older men and women (aged 50--89 years in 1984--1987) in southern California. Regular strenuous exercise and exercise > or =3 times per week were reported; depressed mood was assessed by using the Beck Depression Inventory (BDI). After exclusion of persons with categorical depression and those rating themselves largely or extremely physically limited during the previous month, data on 932 men and 1,097 women were available for cross-sectional analysis. Exercise and depressed mood were reassessed for 404 men and 540 women in 1992--1995; these data were the focus of prospective analyses. In 1984--1987, exercise rates were high (>80%), and average BDI scores were low. Cross-sectional analyses indicated that before and after adjustment for covariates, exercise was significantly associated with less depressed mood. However, prospective analyses of the 944 persons who attended both clinic visits indicated no association between baseline exercise and either follow-up BDI score (p > 0.10) or change in BDI score between baseline and follow-up (p > 0.10). Results confirm that exercisers have less depressed mood. However, exercise does not protect against future depressed mood for those not clinically depressed at baseline.