The association between level of cannabis consumption and mortality during a 15-year follow-up was studied in a cohort of 45,540 Swedish conscripts. The relative risk of death among high consumers of cannabis (use on more than 50 occasions) was 2.8 (95% confidence interval (1.9-4.1)) compared with non-users. However, after control for social background variables in a multivariate model, no excess mortality was found. A high level of consumption of other drugs was also associated with increased mortality; the relative risk of high consumption (greater than 50 times) was 4.6 (2.4-8.5) compared with non-users. After adjustment for social background a relative risk of 1.2 (0.8-1.9) remained; for those having used drugs intravenously more than once, the relative risk was 1.6 (0.9-2.7). Among causes of death a strong predominance was found for violent death, suicide or uncertain suicide being the single most important accounting for 34.4% of all deaths. The proportion of suicides increased sharply with the level of cannabis consumption.