The police agencies involved in the study were the Los Angeles Police Department and the suburban police departments of Inglewood, which faces a serious crime problem, and Redondo Beach, which serves a complacent community with a moderately low crime rate. Three kinds of data are used throughout the study: data obtained through participant observation of police work, data obtained from the departments, and survey data based on an interview schedule administered to patrol officers and field supervisors. The findings discount explanations of police discretion derived from an all-pervasive police mentality or the unique characteristics of a police-citizen encounter. Officers clearly have well-defined views about how they should act on the street, views shaped by occupational socialization and departmental controls. Differences in the exercise of discretion between the departments studied can be explained by the differential impact of administrative controls in large and small departments. The greater uncertainty generated by administrative controls in small police departments leads to a pattern of leniency in police discretion, while the administrative characteristics of large departments stimulate strict enforcement. The predominant reform models for policing are limited in their failure to acknowledge and address the conflict between institutional controls and occupational socialization. What is needed is a system of institutions which will permit continuing reflection on the ends of police work while forcing the police officer to regularly reexamine the contradiction between the ends sought and the means used. Decentralization appears to be a step in this direction. The method used to construct the scale used in the study is described in the appendix, and about 165 bibliographic listings are provided. Tabular data are provided throughout the presentation.