Although psychopathy is recognized as a relatively strong risk factor for violence among inmates and mentally disordered offenders, few studies have examined the extent to which its predictive power generalizes to civil psychiatric samples. Using data on 1,136 patients from the MacArthur Violence Risk Assessment project, this study examined whether the 2 scales that underlie the Psychopathy Checklist: Screening Version (PCL:SV) measure a unique personality construct that predicts violence among civil patients. The results indicate that the PCL:SV is a relatively strong predictor of violence. The PCL:SV's predictive power is substantially reduced, but remains significant, after controlling for a host of covariates that reflect antisocial behavior and personality disorders other than psychopathy. However, the predictive power of the PCL:SV is not based on its assessment of the core traits of psychopathy, as traditionally construed. Implications for the 2-factor model that underlies the PCL measures and for risk assessment practice are discussed.