THE AIMS AND FUNCTIONS OF IMPRISONMENT ARE REVIEWED, AND IT IS ARGUED THAT PROTECTION OF THE PUBLIC (AS OPPOSED TO RETRIBUTION, DETERRENCE, OR REFORMATION) SHOULD BE THE BASIS FOR DECISIONS REGARDING INCARCERATION. QUARANTINE--INCAPACITATING THE OFFENDER BY ISOLATING HIM OR HER FROM THE PUBLIC--IS PRESENTED AS AN INCARCERATIVE FUNCTION FOR WHICH IMPACT ON PUBLIC PROTECTION MAY BE ANALYZED RATIONALLY. AN EXAMINATION OF THE LIMITS OF THE POTENTIAL EFFECT OF QUARANTINE UNDER ACTUAL AND IDEAL CIRCUMSTANCES INDICATES THAT EXISTING PROPOSALS FOR INCREASING THE USE OF GENERAL INCAPACITATION WOULD REDUCE SERIOUS VIOLENT CRIME BY NO MORE THAN 10 PERCENT, AT A STAGGERING COST FOR PRISON CONSTRUCTION AND OPERATION. WITHIN THE CONTEXT OF MICHIGAN CRIMINAL LAW AND CORRECTIONS, TWO ALTERNATIVE PROPOSALS ARE PRESENTED: A SYSTEM OF SELECTIVE QUARANTINE BASED ON A PERSON'S POTENTIAL FOR FUTURE VIOLENCE WHICH COULD BE IMPLEMENTED WITHIN THE EXISTING AUTHORITY OF THE CORRECTIONAL SYSTEM AND A SYSTEM THAT EXPANDS ON THE CORRECTIONAL MODEL BY ADDING CHANGES WITHIN THE AUTHORITY OF THE JUDICIAL SYSTEM. THE SECOND JUDICIAL MODEL MAKES SELECTIVE QUARANTINE SENTENCING OBJECTIVE. HIGH RISK CASES PREVIOUSLY PLACED ON PROBATION WOULD FACE IMPRISONMENT, AND LOW RISK CASES PREVIOUSLY SENT TO PRISON WOULD BE DIVERTED THROUGH NONPRISON DISPOSITIONS. THE LENGTH OF INCAPACITATION FOR PRISONERS SENTENCED UNDER THE JUDICIAL MODEL WOULD BE REGULATED, AS IN THE CORRECTIONAL DISCRETION MODEL, TO A NET OF 5 YEARS FOR HIGH RISK AND 2 YEARS FOR MIDDLE RISK. LOW RISK CASES WOULD NOT BE SENTENCED TO PRISON. RISK OF FUTURE VIOLENCE WOULD BE DETERMINED BY USING ACTUARIAL METHODS SIMILAR TO THOSE USED BY INSURANCE COMPANIES. A BASE EXPECTANCY OF VIOLENCE HAS BEEN CONSTRUCTED AND VALIDATED FOR PAROLEES IN MICHIGAN. A COMPARISON OF THE CRIME REDUCTION POTENTIAL OF THE JUDICIAL AND CORRECTIONAL MODELS, BASED ON STATISTICS ON MICHIGAN OFFENDERS, REVEALS THE FORMER TO BE MUCH MORE EFFECTIVE (8.8 PERCENT REDUCTION, COMPARED TO 3.8 PERCENT FOR THE CORRECTIONAL MODEL). A COST-BENEFIT ANALYSIS, AGAIN WITH MICHIGAN DATA, SHOWS THAT THE JUDICIAL SELECTIVE QUARANTINE MODEL IS NEARLY TWICE AS COST-EFFICIENT AS MICHIGAN'S EXISTING SYSTEM, AND SEVERAL TIMES MORE COST-EFFICIENT THAN CERTAIN OTHER PROPOSALS FOR CHANGING CORRECTIONAL POLICIES. THE FEATURE THAT ENHANCES THE COST-EFFECTIVENESS OF SELECTIVE QUARANTINE IS THAT, UNLIKE GENERAL INCAPACITATION, IT MINIMIZES THE CONFINEMENT OF NONREPEATERS. IT IS CONCLUDED THAT SELECTIVE QUARANTINE IS A FEASIBLE APPROACH TO INCREASING PUBLIC PROTECTION AT TOLERABLE HUMAN AND FISCAL COSTS. FURTHER REFINEMENT OF THE STRATEGY IS URGED. SUPPORTING DATA ARE INCLUDED. (LKM)