THE RESEARCH WAS BASED ON THE CASE OF 586 MENTALLY ILL OFFENDERS WHO WERE TRANSFERRED BETWEEN JULY 1969 AND APRIL 1971 FROM FARVIEW STATE HOSPITAL, PENNSYLVANIA'S ONLY MAXIMUM SECURITY MENTAL HOSPITAL, TO CIVIL MENTAL HOSPITALS AS A RESULT OF A SUCCESSFUL CLASS ACTION SUIT (DIXON VS. ATTORNEY GENERAL OF THE COMMONWEALTH OF PENNSYLVANIA, 1971). THE 'DIXON PATIENTS' CLAIMED THAT THEY WERE BEING HELD UNCONSTITUTIONALLY AFTER THE EXPIRATION OF THEIR ORIGINAL CRIMINAL COMMITMENTS BECAUSE THEY WERE CONSIDERED DANGEROUS. THE VALIDITY OF PREDICTION OF DANGEROUSNESS WAS ANALYZED FROM DATA ON THE DIXON PATIENTS OBTAINED FROM PENNSYLVANIA STATE MENTAL HOSPITALS AND POLICE RECORDS. IN TERMS OF WHETHER IT IS POSSIBLE TO PREDICT THE LEVEL OF VIOLENT BEHAVIOR OF INDIVIDUALS IN ONE ENVIRONMENTAL SETTING FROM THEIR LEVEL OF VIOLENCE IN PRECEDING SETTINGS, THE RESULTS SHOW THAT THERE WAS LITTLE PREDICTIVE VALIDITY FOR BEHAVIORAL VARIABLES WHETHER THE SETTING OF INTEREST WAS THE SECURITY HOSPITAL, THE CIVIL MENTAL HOSPITAL, OR THE COMMUNITY. THE VARIABLE OF AGE REMAINS THE SINGLE MOST EFFICIENT PREDICTOR OF VIOLENCE BOTH INSIDE AND OUTSIDE INSTITUTIONS; YOUNG MEN ARE MORE VIOLENT THAN OLDER MEN. THE PREDICTION DEVICE DEVELOPED IN THE RESEARCH PRODUCED NEGATIVE RESULTS IN PREDICTING VIOLENCE, AS DO OTHER METHODS OF PREDICTION CITED IN THE LITERATURE REVIEW; THE AUTHOR STRESSES THAT EXTREME CAUTION SHOULD BE EXERCISED BEFORE SUPPORTING ANY POLICY WHICH WOULD JUSTIFY CONTINUED APPLICATION OF THE LABEL 'DANGEROUS,' AS THE CONSEQUENCES -PROLONGED MAXIMUM SECURITY CONFINEMENT -- ARE SEVERE. IT IS NOT ENOUGH TO RESEARCH MORE ACCURATE METHODS OF PREDICTING VIOLENCE, AS THE PERFECTION OF SUCH TECHNIQUES WOULD NEVER BY ITSELF JUSTIFY A SOCIAL POLICY OF PREVENTIVE CONFINEMENT. AN INDEX, BIBLIOGRAPHY, SUPPORTING STATISTICS, AND RELATED FORMS ARE INCLUDED.