Although previous studies have identified a protective effect of marriage on risky health behaviors, gaps remain in our understanding of how marriage improves health, particularly among African Americans. This study uses longitudinal data to take selection into account and examines whether marital trajectories that incorporate timing, stability, and duration of marriage affect health risk behaviors among a community cohort of urban African Americans followed for 35 years (N = 1,049). For both men and women, we find six marital trajectories. Men and women in consistently married trajectories are less likely to smoke, drink heavily (women only), and use illegal drugs than those in unmarried or previously married trajectories. Late marrying men do not fare worse in midlife than men in earlier marrying trajectories, but late marrying women show increased risk of midlife drug use. Results suggest policies supporting marriage may have an impact on health but only if stable unions are achieved. (Contains 4 tables and 2 figures.)