OBJECTIVE: The authors examined associations between patient-related characteristics and assaultiveness on six different psychiatric wards to determine 1) the relative contributions of demographic, disorder- related, and diagnostic variables to prediction of assaultiveness and 2) how ward composition and type of victim affect prediction of assaultiveness. METHOD: Hospital records of 1,025 inpatients residing in psychiatric wards within a 6-month time frame were reviewed for evidence of assaultiveness. Data on all 260 assaultive patients and a sample of 136 of the nonassaultive patients were analyzed with multiple regression to predict assaultiveness scores for each ward and each victim category. RESULTS: The findings indicated high rates of assaultive patients and assaults on fellow patients. Multiple regression results produced different predictors of assaultiveness for different wards but not for different categories of assault victim within each ward. Overall, age and sex consistently failed to predict assaultiveness, whereas greater assaultiveness was significantly associated with a greater proportion of time hospitalized since first admission. The most powerful unique predictors of assaultiveness scores were diagnostic distinctions derived from data on coexistent diagnoses. The most assaults were by acute patients whose diagnoses excluded organic mental disorder but included either bipolar disorder or personality disorder and longer-stay patients whose exclusive diagnosis was organic mental disorder. CONCLUSIONS: This identification, albeit modest, of risk factors for assaultiveness on different wards nevertheless provides information fundamental to the management policies of psychiatric institutions. The findings caution against aggregating different ward populations for research on assaultiveness and endorse the usefulness of coexistent diagnoses for predicting assaultiveness.