EACH OF 293 MINIMUM SECURITY INMATES WAS RATED BY STAFF MEMBERS ACCORDING TO THE LIKELIHOOD OF SERIOUS DISCIPLINARY INFRACTIONS AND ESCAPES FROM CUSTODY. FOLLOW-UP DATA WERE THEN OBTAINED, AND THE PREDICTIVE VALIDITY OF THESE RATINGS WAS COMPARED WITH THAT OF SELECTED INFORMATION EXTRACTED FROM INMATE FILES. IN PREDICTING OUTCOMES, ONLY ESCAPE RATINGS WERE STATISTICALLY SIGNIFICANT, AND NEITHER ESCAPE NOR DISCIPLINARY RATINGS CONTRIBUTED SIGNIFICANT CRITERION VARIANCE WHEN COMBINED WITH OBJECTIVE VARIABLES. FINDINGS THUS QUESTION THE ABILITY OF DECISIONMAKERS AND STATISTICAL COMPOSITES TO PREDICT THE OFFENDER BEHAVIOR, CHALLENGING THE ADVISABILITY OF SUCH PREDICTOR USE AS THE BASIS FOR INMATE DISPOSITIONS. GIVEN THE LIKELIHOOD, HOWEVER, THAT SUCH PREDICTIONS WILL NOT BE ABANDONED, EFFORTS SHOULD BE MADE TO INCREASE THEIR ACCURACY BY INCREASING THE QUALITY AND QUANTITY OF DATA UPON WHICH DECISIONS ARE MADE, ALONG WITH THE APPLICATION OF APPROPRIATE QUANTITATIVE METHODS. FURTHER, THERE IS THE NEED TO BALANCE THE RISK OF AN INMATE'S ENGAGING IN DISRUPTIVE AND DANGEROUS BEHAVIOR WITH THE DAMAGE OF EXCESSIVELY HARSH CONFINEMENT TO THE INMATE. TABULAR DATA AND REFERENCES ARE PROVIDED. (RCB)