Evaluations have been conducted on a range of interventions for domestic violence, but this study focused on evaluations of arrest and programs for violent abusers. The article contrasts randomized designs used primarily in North American studies of arrest with the extant evaluations of abuser programs; it argues for the use of more theoretically informed contextual evaluations of criminal justice interventions. Using their own 3-year evaluation study of two Scottish abuser programs, the authors show how the contextual approach is attuned to both outcome and process and results in more empirically informed assessments of how change is achieved in the behavior and orientations of violent men. The authors argue that evaluations of criminal justice-based interventions should be designed to fit the phenomenon under consideration as well as the intervention itself. Further, evaluations of criminal justice programs must move beyond straightforward assessments of rates of recidivism as indicators of the effects of various interventions to a consideration of what achieves sustainable change in the orientations, beliefs, and actions of offenders. 35 references