This study compared adolescents on a continuum of risk of school dropout: adolescents in the mainstream program (low risk); in an at-risk program (medium risk); and in an alternative evening program for dropouts (high risk). The researcher wanted to determine what factors predict school dropout, with particular attention given to family and social support variables. School dropout was measured by proxy, using group membership as an indirect indicator of risk. A discriminant function analysis was performed. Results indicated that family cohesion, adult support, and peer support were predictors of group membership. The model greatly increased correct prediction of membership in each group over that of chance.