We examined the extent to which ethnicity, cannabis and alcohol use could predict prevalence of violence and self-harm in an inpatient psychiatric sample.We collected demographic and clinical data in a series of 141 adult psychiatric inpatients in Hamilton, New Zealand. The Alcohol Use Disorders Identification Test (AUDIT) and Cannabis Use Disorders Identification Test, Revised (CUDIT-R) were used to measure substance use. Clinical assessment and file review were used to verify histories of self-harm and violence.It was found that 66% had a history of violence, 54% of self-harm, and 40% of both; only 20% had neither. Cannabis use was found to significantly predict lifetime history of violence ( p = 0.02); other independent variables (gender, age, ethnicity, alcohol use, psychiatric diagnosis) did not. Self-harm was strikingly predicted by female gender ( p < 0.001), as well as by measures both of cannabis ( p = 0.025) and alcohol use ( p = 0.036); age, ethnicity and diagnosis did not reach significance. Less than 10% of patients were engaged with drug or alcohol services.Cannabis use is a significant predictor of lifetime violence among the severely mentally ill, while both alcohol and cannabis use predict self-harm. Few affected patients receive specific treatment for substance use comorbidity.