Recently, there has been a growing recognition that the 'almost-nothing-works' credo fostered by Martinson's study of corrections programs (1974) is invalid. The evidence continues to mount that some programs are effective in producing long-term change in offenders and a significant reduction in recidivism. Most of the recent successful programs for this study were tested with methodologically impressive evaluation research. Thirty-three percent used true experimental designs with random assignment of subjects to experimental and control groups. Program effectiveness has been demonstrated in both institutional and community-based environments for both juveniles and adults encompassing offender types from predelinquents to hardcore offenders and recidivistic adult criminal heroin addicts. Many successful programs were based on a social learning rather than a medical (disease) model of criminal conduct. Most effective programs aimed at broadening clients' social perceptions or repertoire of adaptive behaviors rather than attempting to cure some underlying emotional disorder. All effective programs were multifaceted, including such techniques as family therapy, contingency contracting, behavioral counseling, role-playing and modeling, vocational and social skills training, interpersonal cognitive problem-solving training, and peer-oriented behavioral programs. Corrections policymakers must derive policy from empirical studies, policy positions based in the theorizing of radical criminology, radical non-intervention, or just desert models are doomed to failure. In fact, even empirically based policy has value only if it is implemented through programs whose structure, content of intervention, and staff performance conform with proven success principles, including the practice of matching offender characteristics to program components. Systems and operations research must be increased if successful programs are to be integrated into ongoing corrections systems. Seventy-nine references are listed.