THIS THEORY ASSUMES THAT THE SENTENCING AUTHORITY ACTS IN A POSITIVE, RATIONAL MANNER. IF THE SENTENCING AUTHORITY BELIEVES THAT THE PURPOSE OF THE SENTENCE IS TO DETER CRIME, THE SENTENCE WILL REFLECT FACTORS WHICH WILL GIVE AN INDICATION OF THE INDIVIDUAL'S LIKELIHOOD TO REPEAT A CRIME. A SERIES OF GRAPHS ILLUSTRATE INCARCERATION AND NONINCARCERATION COSTS AS A FUNCTION OF TIME AND THE POINT AT WHICH THE COSTS TO SOCIETY OF REPEATED CRIME ARE LESS THAN THE COSTS OF INCARCERATION. VARIATIONS IN SENTENCING ARISE FROM VARIATIONS IN GOAL DIFFERENCES AND VARIATIONS IN FORECASTING PROBABILITY OF REPEAT CRIME. IN GENERAL, A PERSON WITH A PRIOR CRIMINAL RECORD RECEIVES A LONGER TERM THAN A PERSON WITH NO PRIOR RECORD. THE JUDGE WHO BELIEVES THAT PRISON REHABILITATES MAY HAND OUT A SENTENCE OF ONLY 2 YEARS WHILE THE JUDGE WHO BELIEVES THAT PRISON SHOULD BE PUNISHMENT MAY HAND OUT A SENTENCE OF 8 YEARS. MEANWHILE, THE JUDGE WHO BELIEVES THAT THE FELON CONVICTED OF STEALING A TELEVISION SET WORTH $100 IS LIKELY TO STEAL 10 SETS A YEAR MAY SET THE TERM AT 5-10 YEARS UNDER THE THEORY THAT THIS IS THE POINT AT WHICH THE DAMAGE FOR SOCIETY IS LESS THAN THE INCARCERATION COST. A GRAPH ILLUSTRATES SENTENCE VARIATIONS ACCORDING TO SENTENCING AGENTS' BELIEFS. THE CHANCES OF A GROUP OF LEGISLATORS PASSING UNIFORM SENTENCING LAW IS EXAMINED THROUGH A SERIES OF COST CURVES. THESE EXAMINE THE LEGISLATORS' POSITIONS, AREAS OF INDIFFERENCE, AND THE EFFECTS OF VARIOUS SIZES OF OVERLAP. LEGISLATIVE DETERMINATION OF SENTENCES MAY NOT SATISFY MOST REASONABLE CRITERIA OF WELFARE ECONOMICS UNLESS ONE CAN CONSTRUCT VERY ELABORATE SCHEMES OF INTERJURISDICTIONAL COMPENSATION OR IS 'LUCKY' ABOUT WHAT HAPPENS IN THE LEGISLATURE. JUDICIAL SENTENCING IS CALLED MORE RATIONAL. EMPIRICAL TESTING IS NEEDED. REFERENCES ARE APPENDED. (GLR)