Of 638 reported assaultive and disruptive incidents in 1979 in a maximum security hospital, 221 (34.6%) were batteries. Whether an incident was a battery or not was a function of time and location. Preludes to battery were noted by staff in only 23.1% of cases and were high-frequency behaviors with low predictive value. Impact and injury to patients and officers occurred through both batteries themselves and efforts to subdue the batterer. With length of hospitalization controlled, the distributions of preadmission criminal charges were not significantly different between batterers and controls. Policies that alter patterns in the timing and location of interaction among patients and in the response of officers or other personnel to batteries in progress offer promise for reducing incident and injury rates.