Many common models of policy implementation assume that all the goals of any system are imposed on it from outside, by legislatures or other discretionary authorities. For example, bail is supposed to be used solely to require the presence of accused persons at their trials. However, bail systems are administered through ongoing criminal justice institutions which develop their own requirements, mainly involving the need to keep functioning efficiently. In addition, criminal justice officials are recruited through political channels and must be responsive to political demands. Thus, in the administration of bail, officials have developed a set of informal requisites which influence, modify, and perhaps negate behavior aimed at performing the system's formal functions. For example, accused prostitutes tend more than other offenders to jump bail and can also raise money easily. However, data from Peoria, Ill., for the period 1972-74 show that no attempt was made to increase the amount of bail required of accused prostitutes or to require more stringent prebail investigations during the study period. A 1967 study of bail in Manhattan, N.Y., Criminal courts produced similar results. The tendency of prostitutes to move geographically and the savings from avoiding trials may provide informal incentives for bail administrators to grant bail to accused prostitutes. Similarly, many court systems have been accused of operating informal preventive detention programs simply by setting the amount of bail beyond the reach of defendants proved dangerous. The article concludes that the informal pressures in the bail administration could be identified if research models contained variables representing what happens within the system itself. Thirteen references and three notes are provided.