In implicit plea bargaining, the defendant pleads guilty under the expectation of receiving a more lenient sentence. In charge-reduction plea bargaining, the prosecutor downgrades or eliminates charges in exchange for a guilty plea to the reduced charges. In judicial plea bargaining, the judge, after consulting with the prosecutor and defense counsel, offers the defendant a specific guilty plea sentence. In sentence-recommendation plea bargaining, the prosecutor, in exchange for a guilty plea, recommends a particular disposition to the judge. Research has generally focused on plea bargaining in general rather than on the varieties of plea bargaining, explaining plea bargaining on the basis of caseload pressures, the difficulty of proving factual guilt for many defendants, and the desire to have flexible sentencing standards. These three themes can be interrelated through the concept of sentence discount schedules, which reflect the concern of judges and prosecutors for substantive rather than formal justice. Stochastic models of the four plea bargaining forms and mixtures of them are designed to predict both aggregate guilty rates and equilibrium sentence discounts. Using these with empirical data from different U. S. cities indicates that the average strength of the case, the caseload pressure, and, to a lesser extent, the local sentencing norms determine the likelihood of particular plea bargaining systems being used. However, testing predictions for specific cities presents difficulties due to problems in official statistics. Nevertheless, informal roles which emerge locally are at least partially influenced by the formal control structure they subvert. Figures and a list of 49 references are supplied.