The efficacy of bronchoscopy for the diagnosis of tuberculosis in patients infected with human immunodeficiency virus (HIV) has not been systematically evaluated. We therefore compared the diagnostic yield of bronchoscopy in 67 HIV-infected and 45 non-HIV-infected patients with culture-proven pulmonary tuberculosis. In all cases, acid-fast smears of sputum were negative or not obtained prior to bronchoscopy. Prebronchoscopic sputum culture yielded Mycobacterium tuberculosis in 34 (89 percent) of 38 HIV-infected patients and 26 (93 percent) of 28 non-HIV-infected patients from whom specimens were obtained. Bronchoscopy provided an early diagnosis of tuberculosis (positive acid-fast smear or granulomata on biopsy) in 23 (34 percent) of the HIV-infected patients and 20 (44 percent) of the patients without HIV infection. The sensitivities of the acid-fast smear and of mycobacterial culture of bronchoscopic specimens and postbronchoscopic sputum were similar in patients with or without HIV infection. In HIV-infected patients, granulomatous inflammation was noted on transbronchial biopsy in 11 (19 percent) of 59 patients with HIV infection, compared to 16 (43 percent) of 37 patients without HIV infection (p = 0.01). Nevertheless, transbronchial biopsy provided the exclusive means for an early diagnosis of tuberculosis in six (10 percent) of 59 HIV-infected patients. We conclude that the yield of bronchoscopy for the diagnosis of pulmonary tuberculosis in HIV-infected patients is similar to that in patients without HIV infection, and that transbronchial biopsy provides incremental diagnostic information not available from evaluation of sputum or bronchoalveolar lavage fluid.