This study examines the predictive validity of the HCR-20 in a sample of 42 female patients admitted to a Dutch forensic psychiatric hospital. The findings are compared with those for a matched sample of 42 male forensic psychiatric patients. The inter-rater reliability of the HCR-20 was good for both female and male patients. There were significant differences between female and male patients in mean HCR-20 item scores, but the mean H, C and R-subscale scores and total score were comparable. The base rate for inpatient violence was similar for female (30%) and male patients (29%), but the base rate for violent recidivism after discharge was significantly higher for the male sample (43%) compared with the female sample (13%). For male patients, the HCR-20 demonstrated good to excellent predictive validity for violent outcome (violent recidivism and inpatient violence); however, predictive accuracy for female patients was much lower. In females, only the HCR-20 final risk judgment, and not the HCR-20 total score, demonstrated significant predictive validity for violent outcome. Copyright © 2005 John Wiley & Sons, Ltd.