The present study used a multilevel approach with multiple informants to determine whether, at individual level, association with substance-using peer groups, and, at class level, incidences of intoxicated students in school premises, are related to students' own substance use. Additionally, it tested the hypothesis that such school incidences affect the closeness of the relation between association with substance-using peers and students' own substance use. Multilevel regression models were estimated separately for drunkenness and cannabis use on the basis of cross-sectional data from 3,925 students of eighth and ninth grades in Switzerland (mean age 15.3, S.D. 0.9) and their teachers (N=220). For both drunkenness and cannabis use, the results confirmed that association with substance-using peers is strongly related to individual substance-use. A higher level of students' own cannabis use and a closer relation between association with cannabis-using peers and the students' own cannabis use were found in classes where students saw others coming cannabis-intoxicated to school or taking cannabis in school premises. Such relations were not found for alcohol. It appears that cannabis use at school or shortly before arriving at school creates an atmosphere that favors cannabis use whether or not students are associated with cannabis-using peers. Establishing an overarching environment of disapproval appears to be an effective means of preventing cannabis use by adolescents.