THE RESEARCH STRATEGY CALLED FOR TWO ANALYSES. FIRST REGULARITIES IN TRANSITIONS BETWEEN COMMITMENT AND NEW OFFENSES OF THE CONSTRUCTION SAMPLE WERE USED TO DERIVE A CLASSIFICATORY SCHEME. THE RESULTS OF THIS FIRST STEP WERE THEN USED TO ORGANIZE THE DATA FOR THE VALIDATION SAMPLE SO THAT THE PREDICTIVE EFFICIENCY OF THE CLASSIFICATORY SCHEME COULD BE TESTED IN A SAMPLE OTHER THAN THE ONE USED TO DERIVE IT. RESULTS INDICATE THAT THE CATEGORY OF PRIOR OFFENSE IS A MODERATELY EFFICIENT PREDICTOR OF THE CATEGORY OF SUBSEQUENT OFFENSE. CATEGORIES RESEMBLING PERSON AND PROPERTY GROUPINGS FOUND IN EARLIER CLUSTER ANALYSES ARE ALSO FOUND HERE. (AUTHOR ABSTRACT MODIFIED)