Actuarial violence risk assessments, many of which include the construct of psychopathy, have been shown to be superior to clinical judgment in the prediction of long-term risk of community violence and recidivism. While these instruments initially appeared to provide similarly accurate judgments of risk of institutional aggression, recent research has indicated that such assessments may be less robust in this setting. One explanation may lie in the types of aggression most frequently observed in each setting. Impulsive (or reactive/affective) is the type of physical aggression most commonly exhibited in psychiatric facilities. This research examines the relationship between risk assessments and aggression in an inpatient forensic setting, with such aggression categorized as impulsive, predatory or psychotic aggression. Consistent with previous research, impulsive aggression was the most frequent type observed (58%). Anger (as measured by the Novaco Anger Scale) and clinical issues (as measured by the HCR-20) were most associated with impulsive aggression, with AUC values of .73 and .71 respectively. In contrast, anger and psychopathy (as measured by the PCL-R) were more associated with predatory aggression, with AUC values of .95 and .84 respectively. Psychotic symptoms were highly associated with psychotically motivated aggression (AUC=.90). These results suggest that traditional violence risk assessments may have limited utility in predicting aggression in an institutional setting and that psychiatric symptoms and heightened affect are more relevant.