An automated cultivation system for mycobacteria, the MGIT 960 system (MGIT system), was compared in the clinical routine with two variants of Löwenstein-Jensen (L-J) medium. A total of 152 isolates were recovered from 2,015 specimens: 139 (91%) with the MGIT system and 127 (84%) with L-J media (P = 0.05). These included 68 isolates of Mycobacterium tuberculosis, of which 88% grew in the MGIT system and 93% grew in L-J media (P = 0.389), and 84 isolates of mycobacteria other than M. tuberculosis (MOTT), of which 94% grew in the MGIT system and 76% grew in L-J media (P = 0.003). More M. avium complex isolates were detected in the MGIT system (n = 65) than in L-J media (n = 50) (P = 0.001). Growth in the MGIT system was detected in 2 weeks for 78% of the isolates, whereas growth was detected in the two L-J media for 17 and 25% of the isolates, respectively. The mean times to detection of M. tuberculosis were 12 days in the MGIT system and 20 days in L-J media, and for M. avium complex the mean times to detection were 8 and 22 to 25 days, respectively. The contamination rates were similar (8.7 to 8.9%) in all media. A commercial amplification system (COBAS AMPLICOR) was evaluated for its ability to rapidly identify M. tuberculosis, M. avium, and M. intracellulare directly from 393 samples in MGIT system broth. A correct PCR result, as evaluated by culture or clinical data, was obtained for 96% of the samples, with inhibition being detected for 2% of the samples. Of the 89 results positive for M. tuberculosis, 91% were regarded as true positive, 8% were regarded as inconclusive, and 2% were considered false positive. For results positive for M. avium and M. intracellulare, 97 and 79%, respectively, were regarded as true positive. Increased rapidity and enhanced isolation of MOTT were obtained with the MGIT system. COBAS AMPLICOR was suitable for rapid identification of these three common pathogens from MGIT system broth.