The performances of the BDProbeTec ET (Becton Dickinson) and COBAS AMPLICOR MTB (Roche) were retrospectively evaluated for detecting Mycobacterium tuberculosis complex in various respiratory specimens. The BACTEC and MGIT liquid culture system (Becton Dickinson) was used as a reference method. A total of 824 respiratory specimens, comprised of sputa, bronchoalveolar lavage fluid, and bronchial and tracheal aspirates from 580 patients, were evaluated. Out of 824 clinical specimens, 109 specimens from 43 patients were culture positive for M. tuberculosis. Of these 109 specimens, 67 were smear positive, 85 were positive by the COBAS AMPLICOR MTB test, and 94 were positive by the BDProbeTec ET. Of the 715 culture-negative specimens, 17 were positive by the auramine staining, 11 were positive by the COBAS AMPLICOR MTB test, and 12 were positive by the BDProbeTec ET. After discrepancy analysis and review of the patients' clinical data, 130 specimens from 50 patients were considered “true-positive” specimens. This resulted in the following sensitivities: microscopy, 61.5%; COBAS AMPLICOR MTB test, 78.0%; and BDProbeTec ET, 86.2%. The specificities of each system, based on the clinical diagnosis, were 99.7% for microscopy, 99.9% for the COBAS AMPLICOR MTB test, and 99.9% for the BDProbeTec ET. The data presented represent a considerable number of specimens evaluated with a considerable number of culture- and auramine-positive and culture-positive and auramine-negative results and therefore give a realistic view of how the data should be interpreted in a daily routine situation. Specifically, the data with regard to the culture-positive and auramine-negative specimens are useful, because in a routine situation, auramine-negative specimens are sometimes accepted, on clinical indications, to be analyzed by an amplification method.