In this study we estimated occurrence of the booster effect in a population infected with the human immunodeficiency virus (HIV) and assessed the relation between the booster effect, T-lymphocyte CD4 cell counts, tuberculosis risk categories, and HIV exposure categories. Patients were recruited from 13 participating sites of the Terry Beirn Community Programs for Clinical Research on AIDS (CPCRA). A two-stage tuberculin skin test was applied to 709 HIV-infected patients using the Mantoux method. An induration reading < 5 mm on the first test and > or = 5 on the second skin test defined the booster effect. Overall, 18 patients, or 2.7% (95% confidence interval, 1.6 to 4.2) experienced the booster effect. Boosted responses were seen in eight (2.1%) anergic patients, six (4.5%) nonanergic patients, and four (2.5%) with anergy status unknown. Boosting was noted in 1 of the 131 women enrolled. Age, race, CD4 cell count, injection drug use, anergy status, tuberculosis risk categories, and HIV exposure categories were not predictive of boosting. The booster effect occurs in a small percentage of HIV-infected patients tested, thus identifying small numbers of patients with latent tuberculosis infection. The two-stage procedure is probably of limited value in the diagnosis of latent tuberculosis in HIV-infected persons.