If correctional evaluation research is to contribute to the accumulation of practical knowledge, it should be designed to test abstract behavioral science principles that explain why a particular type of program should change the future conduct of a specific category of clientele. Focus on such theory leads to nontraditional types of sampling, designed to apply principles to only those cases for which they are relevant rather than to whatever mix one happens to encounter at a particular setting at one moment in its history. In addition, organizational theory must be applied and tested to see whether—in the social, psychological, and political context where a program is studied—the clients have the experience that the behavioral science theory assumes the program is giving them.