Experimental bias resulting from using volunteers in alcoholism research.