Background: Violence risk prediction is a priority issue for clinicians working with mentally disordered offenders. Aims: To review the current status of violence risk prediction research. Method: Literature search (Medline). Key words: Violence, risk prediction, mental disorder. Results: Systematic/structured risk assessment approaches may enhance the accuracy of clinical prediction of violent outcomes. Data on the predictive validity of available clinical risk assessment tools are based largely on American and North American studies and further validation is required in British samples. The Psychopathy Checklist appears to be a key predictor of violent recidivism in a variety of settings. Conclusions: Violence risk prediction is an inexact science and as such will continue to provoke debate. Clinicians clearly need to be able to demonstrate the rationale behind their decisions on violence risk and much can be learned from recent developments in research on violence risk prediction. Declaration or interest: None.