Literature on the correlates and predictors of leisure-time physical activity among African-American population is sparse. This cohort study assessed correlates of leisure-time physical activity (specific large muscle activities during the past month at least three times a week) in a biracial population in 1987 and predictors for the adoption of this behavior 4 years later among those initially inactive. Random digit dialing methods were used to identify residents of two South Carolina communities in 1987. In 1991, 3,223 of these residents were resurveyed (62% response rate). In general, the correlates of leisure-time physical activity (education, > or = 12 years; nonsmoking; weight loss behaviors; and physician advice) were similar for each sex and race group. In 1987, the definition of leisure-time physical activity was not met by 831 (54% of 1,542) white women, 374 (76% of 489) African-American women, 586 (59% of 991) white men, and 126 (63% of 201) African-American men. Among those who were inactive in 1987, 22-24% of white adults and African-American men and 14% of African-American women adopted physical activity 4 years later. Twelve years or more of education was a predictor among white women (risk ratio = 1.7, 95% confidence interval 1.2-2.6) and African-American women (risk ratio = 3.1, 95% confidence interval 1.4-6.9), but not among men. Having a physician discuss physical activity was a predictor among white women (risk ratio = 1.9, 95% confidence interval 1.3-2.7), African-American women (risk ratio = 1.7, 95% confidence interval 0.9-3.2), white men (risk ratio = 2.0, 95% confidence interval 1.3-3.1), and African-American men (risk ratio = 2.7, 95% confidence interval 1.0-7.6). These results highlight the strong effect of educational attainment on adoption of healthy behaviors and support the involvement of physicians to promote physical activity among all race and sex groups.