THE STUDY WAS CONDUCTED IN AN ALL-MALE, MAXIMUM-SECURITY PSYCHIATRIC INSTITUTION IN ONTARIO, CANADA, AND UTILIZED 14 CLINICAL AND DEMOGRAPHIC VARIABLES. TWO GROUPS WERE STUDIED SEPARATELY: A CROSS-SECTIONAL GROUP OF 181 MEN WHO WERE POTENTIALLY ELIGIBLE FOR RELEASE TO A LESS-SECURE SETTING AND A GROUP DISCHARGED TO A LESS SECURE SETTING. IT WAS PREDICTED THAT AN ADMISSION OFFENSE OF HOMICIDE, A DIAGNOSIS OF PERSONALITY DISORDER, AND A YOUNGER AGE OF ADMISSION WOULD BE POSITIVELY CORRELATED WITH LENGTH OF STAY BECAUSE THESE VARIABLES ARE RELATED TO PERCEIVED DANGEROUSNESS. IN THE CROSS-SECTIONAL GROUP, EDUCATION WAS THE BEST PREDICTOR. THIS WAS FOLLOWED BY AGE ON ADMISSION AND AN OFFENSE OF HOMICIDE. IN THE DISCHARGE GROUP, AN OFFENSE OF MURDER WAS THE BEST PREDICTOR FOLLOWED BY A DIAGNOSIS OF RETARDATION. THE REASON FOR THE RELATIONSHIP BETWEEN THE LENGTH OF STAY AND BOTH YEARS OF EDUCATION AND A DIAGNOSIS OF RETARDATION ARE OBSCURE BUT ARE PROBABLY NOT RELATED TOO THE PATIENTS' PERCEIVED DANGEROUSNESS. THERE IS SUGGESTIVE EVIDENCE FROM OTHER RESEARCH THAT YEARS OF EDUCATION AND RETARDATION ARE NOT RELATED TO PERCEIVED PATIENT DANGEROUSNESS. THERE ARE ALSO SEVERAL POSSIBLE EXPLANATIONS FOR THE RELATION OF EDUCATION AND RETARDATION TO LENGTH OF STAY WHICH DO NOT INVOLVE PERCEIVED DANGEROUSNESS. PERHAPS THE MOST OBVIOUS INVOLVES THE BUREAUCRATIC STRUCTURE OF INSTITUTIONS; A FURTHER EXPLANATION IS THAT BRIGHTER PATIENTS ARE SIMPLY BETTER ABLE TO PRESENT ARGUMENTS SUPPORTING THEIR RELEASE AND ARE MORE LIKELY TO TRY AND SECURE THE SUPPORT OF LAWYERS FOR THEIR CASE. REFERENCES ARE NOTED. (AUTHOR ABSTRACT MODIFIED--MJW)