Thirty-four schools (n=7426 consented sixth graders, 71% of the eligible population) were randomized to conditions to test the hypothesis that "Skills for Adolescence" (SFA) is more effective than standard care in deterring and delaying substance use through middle school. One-year posttest data were collected from 6239 seventh graders (84% of those eligible). Initiation of "ever" and "recent" use of five substances for baseline nonusers and changes in recent use for baseline users by experimental condition were compared using mixed model regression to control for school clustering. For pretest nonusers, recent cigarette smoking was lower for SFA than controls (P<.05), as was lifetime marijuana use (P<.06). There were also three Treatment x Ethnicity interactions around drinking behaviors. Hispanics in SFA were less likely to ever and recently drink, and to recently binge drink than Hispanic controls; there were no treatment differences among non-Hispanics. For baseline users, there were three significant SFA delays in transition to experimental or recent use of more "advanced" substances: drinking to smoking, drinking to marijuana use, and binge drinking to marijuana.