The model is used to estimate the average sentences imposed on bank robbers and high-level cocaine dealers. While the first analysis focuses on offense characteristics, the offender's background, and particularly on whether there was a plea bargain or court conviction, a more detailed analysis focuses on the potential effect of recently introduced Federal sentencing guidelines in reducing disparities. The results indicate that plea bargaining is statistically significant in its relationship to sentence concessions, but there was only scant evidence that the frequency of guilty pleas was related to the size of these concessions. The author tentatively concludes that preguidelines disparity was probably a major problem in Federal courts, which may be addressed by the new guidelines. 6 tables, 2 figures, 1 appendix, and 54 notes (Author abstract modified)