Previous research on religiosity and substance use primarily used cross-sectional data or, at best, two waves of data separated by a year. In contrast, we use five waves of the National Youth Survey to determine whether religiosity predicts long-term trajectories of marijuana use and whether changes in religiosity predict changes in marijuana use over time. The results suggest that religious youths use marijuana less often initially and, in contrast to nonreligious youths, exhibit smaller increases in marijuana use over time. In fact, the results suggest that highly religious adolescents are unlikely to experience any increase in marijuana use over time. When religiosity changes over time, the initial level of religiosity does not predict changes in marijuana use. However, changes in religiosity are significantly related to changes in marijuana use. When adolescent religiosity increases, marijuana use tends to decrease, and vice versa. Adolescents who maintain their high levels of religiosity over time are less likely to use marijuana, while con-sistently nonreligious youths are less likely to decrease their marijuana use.