When taken in an adequate dose, disulfiram usually deters the drinking of alcohol by the threat or experience of an unpleasant reaction. However, unless its consumption is carefully supervised by a third party as part of the formal or im-plied therapeutic contract, it is usually discontinued and the deterrent effect is therefore lost. In most studies, disulfiram administration has not been supervised and most reviews fail to stress the crucial importance of supervision. Unsupervised disulfiram has little or no specific effect. We have therefore reviewed all published clinical studies in which there was evidence that attempts had been made to ensure that disulfiram administration was directly supervised at least once a week. We found 13 controlled and 5 uncontrolled studies. All but one study reported positive findings, which were usually both statistically and clinically significant in controlled evaluations. In the sole exception, involving ‘skid-row alcoholics’, it seems that adequate supervision was not achieved. In general, the better the supervision, the better the outcome.Provided that attention is paid to the details of supervision and that supervisors are given appropriate training, supervised disulfiram is a simple and effective addition to psychosocial treatment programmes. Compared with unsupervised disulfiram or no disulfiram control groups, it reduces drinking, prolongs remissions, improves treatment retention and facilitates compliance with psychosocial interventions such as community reinforcement, marital and network therapies. The supervisor may be a health professional, workmate, probation officer or hostel worker but is usually a family member. Treatment should probably continue for a minimum of 12 months. Supervised disulfiram appears to be more effective than supervised naltrexone and may be more effective than unsupervised acamprosate. The crucial importance of supervising the consumption of disulfiram has been overlooked or minimised by many reviewers.