For the isolation of mycobacteria from clinical specimens, we evaluated a method that used a thinly poured Middlebrook 7H11 agar plate (10 by 90 mm) that was examined microscopically. Inoculated plates were sealed, incubated, and examined at regular intervals for the appearance of microcolonies. Plates were examined microscopically, while still sealed, by focusing on the agar surface through the bottom of the plate and the agar. Plates were scanned at low power (x40 total magnification), and colony morphology was confirmed at intermediate power (x100 to x180 magnification). This method was compared with a traditional method that used macroscopic examination of standard mycobacterial media. By using all specimens submitted for mycobacterial culture over the duration of the study, the method was evaluated until 270 isolates of mycobacteria (Mycobacterium tuberculosis, n = 103; M. avium-M. intracellulare, n = 115; miscellaneous, n = 52) were detected. While the conventional method required an average of 23 days to the time of first detection of mycobacteria, the experimental method required an average of only 11 days. When limited to acid-fast stain-positive specimens that were culture positive for M. tuberculosis, the average interval to positivity was 7 days for the microcolony method compared with 17 days for the conventional method. With the experimental method, the microscopic colonial morphology allowed for the presumptive identification of M. tuberculosis colonies, which were distinguished by cording, and M. avium-M. intracellulare colonies, which were smooth and entire. Presumptive identification was complete for 83.5% of the M. tuberculosis isolates within 10 days and for 85% of the M. avium-M. intracellulare isolates within 11 days after inoculation. If the microcolony method was combined with a conventional tube medium, the composite would optimize for speed of recovery while providing the full sensitivity of the conventional method. In addition to reducing the interval to positivity, the microcolony method allows for the easy detection of mixed mycobacterial infections and yields a presumptive identification that facilitates the selection of a confirmatory gene probe test.