OBJECTIVE: The association between violent behavior and low serum total cholesterol levels was examined in a psychiatric inpatient population with diverse diagnoses. METHODS: The study used a case-control design to compare the cholesterol levels of patients in a long-term psychiatric hospital who had a history of seclusion or restraints (N=20) and those who did not (N=20). A low cholesterol level was defined as less than 180 mg/dL. RESULTS: A strong association was found between low cholesterol levels and violent behavior (odds ratio=15.49), an association that was not due to age, race, sex, or diagnosis. The finding was consistent whether mean levels or dichotomized levels of cholesterol were examined. Physical health, cholesterol-lowering medication, current alcohol use, or unusual diets could not explain the results. However, the raw frequency of episodes of seclusion or restraint as an indicator of the frequency of violent behavior was not associated with cholesterol level. Dichotomizing cholesterol levels at 180 mg/dL yielded high sensitivity (90 percent) for predicting violent behavior but at the cost of low specificity (65 percent). CONCLUSIONS: The results support the hypothesis that an association exists between low cholesterol and violent behavior among psychiatric patients but argue against using cholesterol level as a screening tool for predicting violent behavior.