There exists the practical possibility of randomly selected control groups for outcome assessments in mental health programs. Ethical considerations for randomization include distinguishing between innovation and "fooling around," and discriminatory implications of a first-come first-served selection procedure. Practical considerations include reviewing the differences between volunteers and non-volunteers and the utility of rigorous evaluation information. The strategies for implementing control groups in the research design, using administration and staff, include: (1) use of minimal program group; (2) separation of action and evaluation units; (3) inclusion of the lottery procedure on the informed consent form; and (4) allowance for particularly needy applicants. For dealing with participants the strategies include: use of a lottery for selection; informing potential participants about lottery; contacting and informing participants personally; when possible, providing subsequent services to the no-program group. (BN)