TO IDENTIFY VARIABLES THAT DISTINGUISH BETWEEN INMATES WHO ATTEMPT TO ESCAPE AND THOSE WHO DO NOT, THE RECORDS OF ALL ESCAPERS AND NONESCAPERS IN THE NORTH CAROLINA PRISON POPULATION WERE COMPARED, AND DATA ON ESCAPE-RELATED ATTITUDINAL AND ENVIRONMENTAL FACTORS WERE GATHERED IN A QUESTIONNAIRE SURVEY OF A MATCHED SAMPLE OF ESCAPERS AND NONESCAPERS. THE COMBINATION RECORD REVIEW/SURVEY APPROACH PRODUCED BETTER ESCAPE PREDICTIONS THAN DID EITHER APPROACH USED SEPARATELY. WORK RELEASE APPEARS TO BE THE MOST POWERFUL DETERRENT TO ESCAPE. ESCAPERS, PARTICULARLY MULTIPLE ESCAPERS, TEND TO BE CAREER CRIMINALS. THAT 40 PERCENT OF THE VARIANCE BETWEEN ESCAPERS AND NONESCAPERS REMAINS UNACCOUNTED FOR IN THE ANALYSIS SUGGESTS THE COMPLEXITY OF THE ESCAPE PHENOMENON. MULTIVARIATE ANALYSIS IS A USEFUL TOOL IN IDENTIFYING POTENTIAL ESCAPERS IN THAT IT SINGLES OUT VARIABLES WHICH MAY BE ASSOCIATED WITH ESCAPE WHEN CONSIDERED IN ISOLATION BUT NOT WHEN CONSIDERED TOGETHER WITH OTHER VARIABLES. THE ANALYSIS HAS THE FOLLOWING IMPLICATIONS: (1) ESCAPE OFTEN REPRESENTS A HEALTHY PROBLEM SOLVING RESPONSE, (2) MANY PROBLEMS ASSOCIATED WITH ESCAPES CAN BE SOLVED BY RESTRUCTURING ENVIRONMENTS RATHER THAN PERSONALITIES, (3) THE INMATE'S FAMILY SITUATION IS OF GREAT IMPORTANCE AND POSSIBLY COULD BE USED IN SETTING UP INCENTIVES NOT TO ESCAPE, AND (4) MANY INMATES REQUIRE COSTLY SURVEILLANCE ONLY FOR LIMITED, IDENTIFIABLE PERIODS. APPLICATIONS OF THE FINDINGS IN CORRECTIONAL DECISIONMAKING ARE DISCUSSED. A REVIEW OF OTHER ESCAPE PREDICTION STUDIES, A DISCUSSION OF THE THEORETICAL SYSTEMS MODEL UNDERLYING THE ANALYSIS, SUPPORTING DATA, AND A BIBLIOGRAPHY ARE PROVIDED. (LKM)