A review of ethical concerns focuses on the legitimacy of prediction as a basis for sentences, the exact nature of such predictions and standards for their use, and predictor variables. The article cites empirical studies which suggest that potential crime reduction from incapacitation, without improvements in achieving arrests and convictions, is limited. In discussing the adequacy of predictions, the author identifies problems in the seven-point scale developed by the Rand Corporation to predict future high-rate offenders and validation concerns. The final section explores the feasibility of a hybrid policy that relies on observed variations in criminal career patterns to develop sentencing polices that take advantage of the differential reductions in crime associated with different offense types. Data from the histories of adults arrested in Washington, D.C., for robbery and burglary in 1973 demonstrate this approach. Graphs, tables, and 84 references.