This study hypothesized that the presumed causal effects of illicit drug use on dropping out are mediated by three intervening variables. Drug use is expected to decrease motivation, to increase negative social sanctions, and to increase age-inappropriate adoption of social roles. In turn, decreased motivation, increased negative social sanctions, and increased adoption of age-inappropriate roles increase dropping out. Subjects were tested in the seventh grade (Time 1), in the eighth grade (Time 2), and during young adulthood (Time 3). The study involved 2,805 subjects. Models were estimated by using logistic regression. Time 1 drug use had a significant effect on not graduating from high school (measured at Time 3), controlling for gender, father's education, race/ethnicity, and Time 1 measures of deviance, distress, self-control, and grades. This effect was decomposed by the addition of three hypothesized mediating variables in the relationships: Time 2 measures of low motivation, negative social sanctions, and premature performance of competing social roles. The addition of Time 1 measures of these variables did not obviate the relationship, but the addition of Time 2 measures of the three hypothesized mediating variables to the equation reduced to insignificance the effect of drug use on not graduating from high school. The authors interpret the analyses as supporting the hypotheses. First, the effect of earlier drug use on later dropping out of school persisted after controlling for a large number of variables that are interpretable, on theoretical or empirical grounds, as common antecedents of drug use and dropping out. 1 table and 55 references