TWENTY RESIDENTS OF A TRAINING SCHOOL AND 13 NONDELINQUENT HIGH SCHOOL BOYS OF THE SAME AGES WERE ASKED A SERIES OF QUESTIONS ABOUT THE LIKELIHOOD OF PUNISHMENT FOR A NUMBER OF OFFENSES RANGING FROM MURDER THROUGH NEGLIGENT MANSLAUGHTER, AGGRAVATED ASSAULT, ROBBERY, BURGLARY, AND AUTO THEFT. THEN THEY WERE GIVEN THREE SITUATIONS INVOLVING A FELONY AND WERE ASKED IF THEY COULD SEE THEMSELVES COMMITTING THE CRIME. THE TRAINING SCHOOL BOYS HAD SIGNIFICANTLY HIGHER PERCEPTIONS OF THE CHANCES OF ARREST AND PUNISHMENT, OVERESTIMATING ACTUAL ARREST RATES BY 25 PERCENT AND ACTUAL CONVICTION RATES BY 18 PERCENT. NONDELINQUENTS OVERESTIMATED ARRESTS RATES BY 14 PERCENT BUT UNDERESTIMATED CONVICTION RATES BY 16 PERCENT. HOWEVER, WHEN PRESENTED WITH THE HYPOTHETICAL SITUATIONS THE DELINQUENTS WERE SIGNIFICANTLY MORE LIKELY TO SEE THEMSELVES AS COMMITTING THE CRIME AND NOT BEING ARRESTED. THE NONDELINQUENTS WERE FAR LESS LIKELY TO PICTURE THEMSELVES COMMITTING THE CRIME AND WERE FAR MORE AFRAID THEY WOULD BE ARRESTED IF THEY DID COMMIT IT. IT IS SUGGESTED THAT ADOLESCENT BOYS ARE NOT AS RATIONAL, DISPASSIONATE, AND CALCULATING AS PROFESSIONAL CRIMINALS IN DECIDING WHETHER, WHEN, AND HOW TO ENGAGE IN CRIMINAL BEHAVIOR; THAT DELINQUENTS BELIEVE ENGAGING IN SUCH BEHAVIOR ENHANCES STATUS; AND THAT THE PROSPECT OF PUNISHMENT IS NOT SO FRIGHTENING SINCE THEY HAVE ALREADY SURVIVED A TRAINING SCHOOL EXPERIENCE. (GLR)