This paper examines the links between individual adolescent smoking behavior and actual and perceived smoking behavior in the individual's school cohort. We hypothesized that students enrolled in schools with higher smoking prevalence among students in their grade are more likely to smoke subsequently. We also expected perceived school-level prevalence of smoking to have a greater impact than actual prevalence because the former is a more direct measure of perceived norms. Adjusting for demographics, actual school-level prevalence at baseline (grade 7) was strongly associated with smoking frequency one year later. However, the association disappeared after adjusting for individual smoking frequency at baseline. School-level prevalence did not moderate the association between individual's baseline and subsequent smoking frequency. Perceived prevalence of smoking among grade 8 students and two measures tapping the behavior of smaller peer groups--cigarette offers and exposure to friends and other peers who smoke--were associated with increased risk of smoking.