Three studies comparing 704 alcoholics' self-reports, observers' ratings, and breath-test determinations of alcohol intoxication found that (a) alcoholics' self-reports of their recent drinking were valid when they had not been drinking; (b) when they had been drinking, their self-reports were frequently invalid and underreported; (c) incidence of discrepant self-reports was not a function of level of intoxication; and (d) trained observers could identify only 50-67% of the Ss who were intoxicated but gave invalid self-reports. Breath analysis for blood alcohol concentration should be a mandatory component of alcohol treatment programs and alcohol treatment outcome evaluations. (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1979 American Psychological Association.