The study uses a school-based sample to test the social and familial risk and protective factors relating to cannabis use. Based on a self-completion survey of 2078 14-16-year-olds (mean age of 15 years) attending seven standard state-run secondary schools in south London, an assessment was made of rates and risk factors for cannabis use. Twenty-four per cent of the total sample had ever used cannabis, with 15% having done so in the month prior to assessment. In addition to greater likelihood of illicit drug use, lifetime cannabis users were less likely to spend time regularly with both their mothers and fathers, but more likely to spend free time with friends who smoked, drank alcohol and used illicit drugs, and with friends involved in criminal activities. Among those who had ever used cannabis, frequency of cannabis use was predicted (using linear regression) by two onset factors (earlier initiation of drinking and cannabis use were both linked to more frequent use) and two social factors (more time spent with drug-using friends and less time spent with the mother). Overall, the study showed that early onset, itself predicted by social networks, is linked to more frequent use of cannabis and that this appears to be sustained by less time spent with parents and more with drug-using peers.