The point of introducing a classification scheme as part of an overall program of punitive sentencing as viewed from an ethical perspective is twofold: (1) to guarantee that only (though perhaps not all) the factors ethically relevant to a sentence determine that sentence and (2) to guarantee that offenders convicted of the 'same' offense get the 'same' punishment, while offenders convicted of 'different' offenses get 'different' punishments. For the purposes of this discussion, the relation of classification to sentencing is illustrated through an analysis of the 1980 proposals by the Pennsylvania Commission on Sentencing. Under the Pennsylvania system, an offender's classification after conviction and prior to sentencing is a function of two variables: 'offender score' and 'offense score.' Offense score is itself a function of 'offense rank' and four other factors of lesser importance. The classification scheme functions as a presumptive sentencing recommendation; it sets forth the sentence the court must mete out except as other considerations are taken into account. The guideline sentence, however, is to be fine-tuned by reference to a set of postclassification modifications. These considerations tranform the sentence as initially identified by classification into an actual sentence by means of a series of individualizing considerations, including the set of aggravators and mitigators as well as the catch-all provisions that authorize the sentencer to modify any guideline whenever all the facts show that such a sentence is 'clearly reasonable.' These postclassification factors raise doubts as to whether considerations of utility, cost/benefit, and efficiency may creep back into the sentencing system to such an extent that retributive concerns (presumably based on moral considerations) may be diluted or overpowered. Whether or not this occurs must be determined by empirical study. One table, an apendix, and 39 footnotes are provided.