AbstractTwo versions of a school‐based alcohol prevention programme for 7th grade students were evaluated against a non‐treated comparison condition. Both programme versions were based on a social‐cognitive theory approach to alcohol prevention, and differed in the level of role‐specifications for participating teachers and peer leaders. Twelve schools with 955 students m Bergen, Norway, participated in the study. Four schools were randomly assigned to each of the three conditions, and the programme was implemented during 10 class periods over 2 months in Spring, 1992. Pre‐ and post‐test surveys were conducted, assessing alcohol use, as welt as cognitive variables related to alcohol use in a specific context (intentions, attitudes, norms and self‐efficacy expectations). Results showed that the highly role‐specified (HRS) version had a higher degree of student involvement than the less role‐specified (LRS) version, indicating that the HRS programme was more successful in engaging students in alcohol prevention activities. A significant difference in programme effectiveness was found, as measured by an overall programme effect across all dependent variables. This effect was explained by significantly more positive outcomes in the HRS version compared to both the LRS and the non‐treated conditions, while the LRS version did not differ significantly from the no