Evaluation planning and design should begin before an LRE program is implemented and should incorporate process and a variety of outcome measures. During planning, the evaluation should be designed around the research questions and around the resources available, rather than around assumptions founded exclusively on basic research principles. Strict evaluation methodology is not required when programs are relatively inexpensive, innocuous for participants, and fairly standard; however, researchers should realize that evaluations that are designed during or after program implementation often do not allow research questions to be adequately answered, since issues central to answering these questions do not arise until too late. The best way to allow valid causal inferences about program impact is to use random assignment of clients to comparison groups. If this is not possible or is objectionable, a useful alternative is the biased-assignment design wherein clients are assigned to comparison groups according to a measured criterion that can be statistically controlled. A third and weaker alternative is the multiple groups pretest-posttest design. The one-group pretest-posttest design is the weakest of the four designs covered in this paper and should be avoided when seeking valid causal inferences. If LRE is to become a standard and widely used component of juvenile justice, it should first undergo the kind of empirical scrutiny required to make refinements and appropriate qualifications. 22 references