The challenging context of social work interventions require that most intervention studies will be derived from nonexperimental research designs. Two evaluation studies in this special issue employed nonrandomized designs to examine the efficacy of two programs--a police crisis intervention team designed to enhance officers' responses to mental health crisis and a program for pregnant incarcerated women. Each is a laudable effort to examine important intervention outcomes in insular settings. The design and execution problems explicated in both studies are familiar to researchers who venture into such practice arenas. Disappointing outcomes, as in these two studies, can sometimes be attributed to the problems of nonequivalent comparison designs or to unanticipated events. Intervention researchers must also recognize that programs often fail to demonstrate differences and can even result in adverse outcomes.