This article estimates peer influences on the alcohol, tobacco and cannabis use of UK adolescents. We present evidence of large, positive and statistically significant peer effects in all three behaviours when classmates are taken as the reference group. We also find large, positive and statistically significant associations between own substance use and friends' substance use. When both reference groups are considered simultaneously, the influence of classmates either disappears or is much reduced, whereas the association between own and friends' behaviours does not change. The suggestion is that classmate behaviour is primarily relevant only inasmuch as it proxies for friends' behaviour.