A comprehensive theoretical model of adolescent substance abuse, the Social Ecology Model, was developed. The model includes peer influence, school bonding, self-efficacy, and family and school climate variables. Several competing variations of the model were empirically tested using a confirmatory multivariate methodology (LISREL) on a high school sample (N = 1,373). Self-efficacy and school bonding collapsed into a single factor, which was significantly predicted by the family and school climate latent variables. These findings suggest that prevention approaches should include interventions effective in improving family and school climate for youths in addition to improving self-efficacy, school bonding, and peer relations. [Translations are provided in the International Abstracts section of this issue.]