In a 12-year longitudinal study of a birth cohort 1577 men were questioned at the age of 19 and again at 31 years. Personality, social background and substance use were connected with subsequent consumption of alcohol, tobacco and cannabis, using correlation and discriminant analysis on the total and on two split samples, respectively. The preceding level of substance use was the best predictor for later use. Antisocial attitudes, depressiveness and psychosomatic complaints were associated as non-specific indicators. The following predictors were specific for: Alcohol: parental consumption, self-description as aggressive-extraverted; Tobacco: broken home, low level of vocational training, extraversion; Cannabis: alienation from society. The results are related to various theoretical frameworks and discussed with regard to possible efforts for prevention.