WHILE THERE IS CONSIDERABLE EVIDENCE THAT IMPRISONMENT REDUCES CRIME THROUGH INCAPACITATION AND DETERRENCE, IT IS NOT KNOWN PRECISELY HOW MANY CRIMES CAN BE AVOIDED BY IMPRISONING CONVICTED CRIMINALS. A SECOND UNKNOWN IS THE EFFECT ON CRIME AND PRISON POPULATIONS OF IMPRISONING ONLY SERIOUS, HABITUAL OFFENDERS WITH EXTENSIVE PRIOR RECORDS, AS OPPOSED TO ALL PERSONS CONVICTED OF SERIOUS CRIMES. EARLY STUDIES OF CRIMINAL CAREERS SEEMED TO SUPPORT THE VIEW THAT ONLY A FEW REPEAT OFFENDERS NEED BE IMPRISONED IN ORDER TO REDUCE CRIME. BUT MORE RECENT STUDIES, INCLUDING AN ANALYSIS OF ANNUAL OFFENSE RATES AMONG ADULT ARRESTEES IN THE DISTRICT OF COLUMBIA, INDICATE THAT REPEAT OFFENDERS EXIST IN GREATER NUMBERS AND COMMIT FEWER CRIMES PER YEAR THAN HAD ONCE BEEN THOUGHT. A SECOND, MORE SERIOUS PROBLEM WITH SENDING ONLY OLDER, CONFIRMED FELONS TO PRISON IS THAT MANY OFFENDERS ARE IMPRISONED TOO LATE IN THEIR CRIMINAL CAREERS. SEVERAL STUDIES HAVE FOUND THAT, AS OFFENDERS GROW OLDER, THEY ACCOUNT FOR A DIMINISHING SHARE OF CRIME BUT AN INCREASING SHARE OF IMPRISONMENT. FROM AN INCAPACITATION POINT OF VIEW, IT APPEARS THAT THE PRIOR-RECORD APPROACH TO SENTENCING LOCKS UP OFFENDERS TOO LATE IN THEIR CAREERS AND TO A LARGE EXTENT INCAPACITATES THE WRONG GROUP OF OFFENDERS. NEITHER DOES THE POLICY OF TREATING LIGHTLY THE CRIMES OF JUVENILES AND ADULT FIRST-OFFENDERS BEAR UP FROM A DETERRENCE POINT OF VIEW. GIVEN THAT IMPOSING PROGRESSIVELY STIFFER SENTENCES WITH EACH CONSECUTIVE CONVICTION DOES DETER A LARGE NUMBER OF CRIMES, IT STILL IS NOT LOGICAL TO MINIMIZE THE PROBABILITY OF PUNISHMENT FOR THE MOST CRIMINALLY ACTIVE GROUP OF OFFENDERS. SUPPORTING DATA AND A LIST OF REFERENCES ARE INCLUDED. (LKM)