The validity of self-report of drug use has been found to vary widely. Moreover, previous research has focused on samples of adults. In 1996-1998, human immunodeficiency virus (HIV)-infected adolescents and high-risk, noninfected adolescents (n=182) were recruited at 16 locations in 13 US cities into the Reaching for Excellence in Adolescent Care and Health (REACH) project, to the authors' knowledge, the first national study of disease progression among HIV-positive adolescents who were infected through sexual behavior or injection drug use. Self-report of marijuana use was assessed through audio computer-assisted self-administered interviewing (ACASI). Urines were tested for marijuana at a certified laboratory by using the enzyme-multiplied immunoassay technique. Conditional kappas for 2-, 5-, and 7-day self-reports were 0.57, 0.71, and 0.69, respectively . Maximum sensitivity was obtained from a combination of ACASI and urine drug testing. Contrary to previous studies, the data suggest that if a single evaluative instrument is to be used for prevalence, ACASI is more sensitive than urine drug testing for marijuana overall, but particularly for HIV-infected adolescents.