This paper reviews the published research literature on the use of physical restraint with mentally retarded adults and children. Research on three types of restraint is included. One type involves one or more person(s) holding another. A second method is where a mechanical device is fitted to limit movement or reduce injury. The third type is where the person voluntarily applies a personal or mechanical restraint. The following conclusions emerged: (a) there are numerous processes which contribute to the outcomes associated with restraint, and these are poorly understood; (b) different processes mediate the outcomes for contingent and noncontingent restraint; (c) both noncontingent and contingent restraint can result in long-term reductions in target behaviours, especially when fading procedures are employed (noncontingent restraint) and where staff or carers are involved in the treatment plan (contingent restraint); (d) self-restraint seems to be maintained by the reinforcing effects of the restraint procedure or by escape from the aversive consequences of self-injury; (e) there are (negative) reinforcing consequences for staff who use restraint procedures in service settings; (f) and both staff and clients risk injury, especially from emergency or unplanned restraint.