OBJECTIVE To examine prevalence and correlates of cannabis dependence among long-term cannabis users. METHODS A cross-sectional survey of patterns and experiences of cannabis use and dependence. METHODS A snowball sample of 243 long-term cannabis smokers, who were currently smoking 3-4 times a week, were recruited from the New South Wales North Coast, an area long associated with cannabis cultivation and use. METHODS A structured interview was administered, incorporating the following dependence measures: an approximation to a life-time DSM-III-R diagnosis, an approximation to a 12-month ICD-10 diagnosis, and the Severity of Dependence Scale (SDS). RESULTS Prevalence of a life-time DSM-III-R diagnosis of cannabis dependence was 57%, while 57% received an ICD-10 dependence diagnosis for the last year. Substantially fewer (15%) of the sample were diagnosed as dependent according to the Severity of Dependence Scale (SDS). Only 26% believed they had a problem with cannabis at least sometimes. There was general concordance between DSM-III-R and ICD-10 measures, but not between these and the SDS. ICD-10 and DSM-III-R dependence diagnoses were modestly correlated with age, life-time illicit drug use and quantity of cannabis use. Principal components analyses of the dependence measures provided little evidence for a unidimensional dependence syndrome for ICD-10 and DSM-III-R criteria. There was strong support for unidimensionality of the SDS. CONCLUSIONS Symptoms of cannabis dependence were common among these long-term users but only one-quarter perceived that they had a cannabis problem. There was no strong evidence for a unidimensional cannabis dependence syndrome.