The effects of social punishment, positive practice, and timeout on the noncompliant behavior of four mentally retarded children were assessed in a multitreatment withdrawal design. When programmed, the experimental procedure occurred contigent on non-compliance to experimenter-issued commands. Commands were given at 55-sec intervals throughout each experimental session. The results showed (1) lower levels of noncompliance with social punishment than with the positive-practice or timeout conditions, and (2) that relatively few applications of social punishment were required to obtain this effect. The advantages of social punishment over other punishment procedures, considerations to be made before using it, and the various aspects of the procedure that contribute to its effectiveness were discussed.