More useful criminal justice evaluation research would differentiate offenses and offenders on the basis of causal theory and would interrelate several levels of abstraction. This is illus trated in correctional practice evaluation by a linkage of be havior modification, symbolic interactionist, and sociocultural diffusion theory, from which three propositions on the effective ness of specific treatment methods for particular types of offenders are derived. Research thus far supports the validity of these propositions. Boards of autonomous criminologists and public representatives supervising criminal justice statistics and research agencies would foster more grounding of inquiries in policy- relevant theory.