The experiment with community service in the canton of Vaud highlights the usefulness of controlled experiments. Many observers might have expected community work to be followed by lower rates of recidivism. Few, however, might have expected that recidivism is unrelated to how various sanctions affect employment record and social integration and that attitudes toward the offense, the punishment, and the criminal justice system may be the key variable through which sanctions affect future behavior. In the area of prescription of drugs, knowledge of long-term impacts of several drugs would be of particular interest. If heroin prescription has turned out to be highly efficient in terms of crime reduction among most seriously addicted participants, it is not certain how long such prescription could be continued without unfavorable side effects. Without controlled experiments, it might be difficult to obtain critical knowledge on this issue. In assessing ethical issues, researchers should not only consider what will likely be done to people during an experiment but also what they are likely to undergo without it. In this respect, heroin or any other form of maintenance program may be less intrusive than incarceration or forced long-term treatment, which many would view as morally superior. 15 notes and 29 references