Experimental designs are not appropriate for every evaluation study in criminal justice. Nonetheless, experiments are possible in many circumstances and can provide a powerful tool for developing criminal justice policy. There is no reason to exclude experimental designs at the outset, either for ethical, political, or practical reasons, although this is often the case in criminal justice study. The task is to identify under what conditions experiments can be successfully implemented. This article develops eight principles that may help researchers and practitioners assess when experimentation will be most feasible. First, there are generally fewer ethical barriers to experimentation when interventions involve the addition of resources; and second, there are generally fewer objections to experiments that test sanctions that are more lenient than existing criminal justice penalties. Third, experiments with lower public visibility will generally be easier to implement. Fourth, in cases in which treatment cannot be given to all eligible subjects, there is likely to be less resistance to random allocation. Fifth, randomized experiments are likely to be easier to develop if the subjects of intervention represent less serious threats to community safety. Sixth, experimentation will be more difficult to implement when experimenters try to limit the discretion of criminal justice agents who traditionally act with significant autonomy and authority. Seventh, it will be easier to develop randomized experiments in systems in which there is a high degree of hierarchical control. Finally, when treatments are relatively complex, involving multiple actions on the part of criminal justice agents, experiments can become prohibitively cumbersome and expensive. 26 references