To evaluate the correspondence among measures of self-reported drinking, standard biological indicators and the reports of collateral informants, and to identify patient characteristics associated with observed discrepancies among these three sources of research data. Using data collected from a large-scale clinical trial of treatment matching with alcoholics (N = 1,726), these three alternative outcome measures were compared at the time of admission to treatment and at 12 months after the end of treatment. Patient self-reports and collateral reports agreed most (97.1%) at treatment admission when heavy drinking was unlikely to be denied. In contrast, liver function tests were relatively insensitive, with positive serum gamma-glutamyl transpeptidase (GGTP) values obtained from only 39.7% of those who admitted to heavy drinking. At 15-month follow-up the correspondence between client self-report and collateral report decreased to 84.7%, but agreement with blood chemistry values increased to 51.6%. When discrepancies occurred, they still indicated that the client' s self-report is more sensitive to the amount of drinking than the biochemical measures. Patients who presented discrepant results tended to have more severe drinking problems, more previous treatments, higher levels of pretreatment drinking and significantly greater levels of cognitive impairment, all of which could potentially interfere with accurate recall. In clinical trials using self-selected research volunteers, biochemical tests and collateral informant reports do not add sufficiently to self-report measurement accuracy to warrant their routine use. Resources devoted to collecting these alternative sources of outcome data might be better invested in interview procedures designed to increase the validity of self-report information.