INTRODUCTION AND AIMS: To examine predictors of cannabis use at 5 year follow up in an Australian Aboriginal cohort. DESIGN AND METHODS: A longitudinal study consisting of two waves of data collection 5 years apart was conducted. Of the 100 Aboriginal residents (aged 13-36 years) interviewed about cannabis use in 2001, 83 were re-interviewed in 2005-2006 from three remote communities in Arnhem Land, Northern Territory, Australia. Self-reported cannabis use was categorised at each time point (none; former use, quit > or = 3 months; lighter use, < 6 cones, 2-3 times weekly; daily use, > or = 6 cones, daily) and summarised as any current use, heavy use, dependence or cessation. Other substance use, employment and involvement in school or training were also compiled. RESULTS: Most respondents who reported cannabis use at baseline again reported use at follow up. A history of petrol sniffing predicted later heavy cannabis use (P < 0.05). Trends were evident for men to have persisting cannabis use, and for employment and/or engagement in school or training to be associated with cannabis cessation. DISCUSSION AND CONCLUSIONS: Ongoing heavy cannabis use is commonplace in this Aboriginal cohort and raises concerns for the physical, social and psychiatric burden on these already vulnerable communities. Prevention, treatment and intervention programs developed with communities are badly needed.