A research perspective on the criminal justice system should have at least the following attributes: it should be structural and build on a combination of substantive theory and data. Further, it should be dynamic and incorporate information provided by experimental interventions. The working hypothesis of this paper is that a decision theoretic view of the performance of the criminal justice system provides such a perspective. Measurable outputs of the criminal justice system are conceptualized as the result of decisions made by agents in response to a variety of stimuli (including the policies of the system), and a reduced form framework follows, within which dynamic behavioral and technical relationships between components of the system can be specified. These in turn suggest the kinds of data that must be collected to operationalize the hypothetical model. There are substantial benefits derived from having a model explicitly linked to a unified perspective, with the greatest benefit being the obtaining of reliable prescriptions for policy. Using a decision theoretic framework expressed as a mathematical model, police officer functions are described by a system of discrete time, linear difference equations. The analysis concludes that if criminal justice processes are viewed in the context of discrete time, dynamic decisionmaking, the way the criminal justice system is analyzed and administered can be improved. The appendix lists possible variables for an optimal control model of police patrols, and 37 references are provided.