Two hundred and fifty-six sputum, bronchoalveolar lavage, and bronchial and tracheal aspirate specimens from 243 patients were tested for the presence of Mycobacterium tuberculosis complex by auramine fluorochrome staining, rRNA target amplification (Gen-Probe Amplified Mycobacterium Tuberculosis Direct Test [AMTD]), and PCR (Roche Amplicor Mycobacterium Tuberculosis Test [Amplicor PCR]). The results were compared with those of conventional Löwenstein-Jensen tube culture and BACTEC radiometric liquid culture. A total of 26 specimens from 18 patients were culture positive for M. tuberculosis. In addition, seven specimens were positive by staining and by culture for other Mycobacterium species but negative by nucleic acid amplification methods and were not included in the comparison. When compared with that for culture, the sensitivities of the techniques were as follows: for staining, 80.8%; for Gen-Probe AMTD, 84.6%; and for Roche Amplicor PCR, 84.6%. The specificities were 99.1, 98.7, and 99.1%, respectively. After resolution of discrepant results by review of the patients' clinical data, 29 specimens from 21 patients were considered positive, and the overall sensitivities, specificities, and positive and negative predictive values were 89.7, 100, 100, and 98.7% for culture; 75.9, 99.5, 95.7, and 96.9% for staining; 86.2, 100, 100, and 98.2% for Gen-Probe AMTD; and 82.8, 100, 100, and 97.9% for Roche Amplicor PCR, respectively. It is concluded that both nucleic acid amplification methods are rapid, sensitive, and specific methods for the detection of M. tuberculosis in respiratory specimens.