Controlled drinking has recently become a controversial alternative to abstinence as an appropriate treatment goal for alcoholics. In this study we reexamine the evidence underlying a widely cited report by Sobell and Sobell of successful controlled drinking by a substantial proportion of gamma (physically dependent) alcoholic subjects in a behavior therapy experiment. A review of the evidence, including official records and new interviews, reveals that most subjects trained to do controlled drinking failed from the outset to drink safely. The majority were rehospitalized for alcoholism treatment within a year after their discharge from the research project. A 10-year follow-up (extended through 1981) of the original 20 experimental subjects shows that only one, who apparently had not experienced physical withdrawal symptoms, maintained a pattern of controlled drinking; eight continued to drink excessively--regularly or intermittently--despite repeated damaging consequences; six abandoned their efforts to engage in controlled drinking and became abstinent; four died from alcohol-related causes; and one, certified about a year after discharge from the research project as gravely disabled because of drinking, was missing.