The original sample included 1,360 students; 680 received the intervention and 680 served as controls. The sample was 48.5 percent female and 51.5 percent male; 58 percent lived with married parents, 13 percent with remarried parents, and 29 percent with single parents. The eighth grade followup sample consisted of 1,075 students, 79 percent of the original sample. Level of substance use was assessed by having respondents rate the frequency with which they used each of nine substances during the past year (tobacco, alcohol, marijuana, LSD, amphetamines, barbiturates, heroin, inhalants, and cocaine). The ADM Program taught students decisionmaking, group process, and social network utilization skills. This intervention maintained a positive effect on mean tobacco use, but no differences were observed for mean alcohol, marijuana, or hard drug use. In a test of the intervention's differential effectiveness, students living with married parents reported lower mean tobacco use than control students living with married parents and program and control students living with single parents. Logistic regression analysis examining the proportion of users at followup revealed a negative program effect for alcohol and no differences for other substances. Subsequent attrition analyses strongly suggested that the positive effect for tobacco use at followup was most likely even stronger and that the negative effect for alcohol was spurious. The importance of examining both program and attrition effects when evaluating the impact of longitudinal preventive interventions is emphasized. The need to consider alternative models to guide the conceptualization and evaluation of adolescent substance use prevention programs is discussed. 29 references and 2 tables (Author abstract modified)