Several State and local jurisdictions have adopted sentencing guidelines based on statistical models to reduce sentencing disparities. Compliance is voluntary, and provisions are made for judicial discretion. Primary sources for this evaluation were: data used to develop sentencing guidelines in the Denver District Court, the Cook County Circuit Court in Chicago, and the Essex County Court in Newark; information on the impact of sentencing guidelines in Denver and the Philadelphia Court of Common Pleas; and interviews in Denver, Chicago, and Philadelphia. A review of the process by which the Denver, Chicago, and Newark guidelines were created revealed that they did not have a sound empirical basis because researchers did not test the assumption implicit in their regression analysis that judges are homogeneous with respect to the weight they give various facts pertaining to the offense and the offender's background. Moreover, it is doubtful whether improved modeling techniques could ever capture the complexities and subtleties of sentencing in a way that could be translated into policy. A comparison of sentences imposed in Denver and Philadelphia before and after introduction of the guidelines suggested that they had little effect on the exercise of judicial sentencing discretion and failed to reduce sentencing disparity. Finally, the study found that the guidelines had no impact on plea bargaining in Denver, Chicago, and Philadelphia because of established patterns of discretionary decisionmaking.