ONE OF THE DUTIES OF JAPANESE POLICE JUVENILE DIVISIONS IS TO SEPARATE THOSE WHO SHOULD BE REFERRED TO SOCIAL AGENCIES OR FAMILY COURT FROM THE MORE THAN HALF A MILLION YOUNG PERSONS WHO COME IN CONTACT WITH THE POLICE. SINCE JUVENILE DIVISIONS HANDLE ALL OFFENDERS UNDER AGE 19, INCLUDING THOSE COMMITTING TRAFFIC OFFENSES, SUCH SCREENING IS AN IMPORTANT POLICE FUNCTION. WHEN EXERCISING SUCH DISCRETION THE POLICE OFFICER MUST CONSIDER THE FOLLOWING: (1) THE LEGAL QUALIFICATION OF OF THE YOUTH FOR JUVENILE DELINQUENT STATUS; (2) MOTIVATION FOR THE ACT; (3) PRONENESS TO FUTURE DELINQUENCY; (4) HOME SITUATION; AND (5) RECOMMENDATIONS OF OTHER AGENTS OR VOLUNTEERS CONCERNED WITH THE CASE. THE FIRST PREDICTION TABLE WAS DEVELOPED IN 1960. THE PROCESS USED TO REVISE THE TABLE IN 1970 IS DESCRIBED. DATA GATHERED ON 2,944 OFFENDERS CONVICTED OF SERIOUS CRIMES WERE USED; OF THESE, 222 WERE REARRESTED WITHIN 6 MONTHS. CORRELATION COEFFICIENTS, CHI-SQUARE TESTS, AND FACTOR ANALYSES WERE USED TO ISOLATE FACTORS MOST CLOSELY ASSOCIATED WITH RECIDIVISM. THESE WERE DEVELOPED INTO A FIVE-CLASS PREDICTION TABLE WHICH WAS VALIDATED ON 7,389 NEW CASES IN 1972; PREDICTION SCORES PROVED RELIABLE FOR 72.2 PERCENT. THE TABLE CONSIDERS AGE, TYPE OF OFFENSE, MOTIVE, UNRECORDED OFFENSES, TRUANCY, RUNNING AWAY, STEALING, CARRYING A CONCEALED WEAPON, FAMILY SITUATION, AND CRIME AMONG FAMILY MEMBERS. SCORING IS EXPLAINED. (GLR)