A wide variety of drugs have been tested in experimental animals and several have been found that reduce voluntary alcohol drinking. The available evidence suggested that the same drugs also reduce alcohol drinking in alcoholics. Various factors limited or prevented the clinical use of the these drugs. Our working hypothesis has been that alcohol drinking is a learned response, reinforced primarily from alcohol in the brain, and that an alcoholic is a person in which this response and the related craving have become so strong that they dominate the behaviour and interfere with normal functioning. Learned responses are extinguished if they are made repeatedly while the reinforcement is blocked, and opiate antagonists appear to block the reinforcement from alcohol. A series of experiments support the hypothesis that drinking alcohol while an antagonist is present extinguishes the alcohol-drinking response in rats. The antagonists are non-addictive and at least naloxone appears to be safe. Clinical trials are now needed, but the present results suggest that this extinction procedure might be a useful adjunct to the treatment of alcoholism.