Microscopic examination of respiratory specimens for acid-fast bacilli (AFB) plays a key role in the initial diagnosis of tuberculosis, monitoring of treatment, and determination of eligibility for release from isolation. The objective of this study was to compare the sensitivity obtained with smears for detection of AFB (AFB smears) made directly from respiratory specimens (direct AFB smears) to that obtained with parallel smears made from concentrates of the specimens (concentrated AFB smears). A total of 2,693 specimens were evaluated; 1,806 were from the University of California Irvine Medical Center Medical Microbiology Laboratory (UCIMC), which serves a tertiary-care hospital with outpatient clinics, and 887 were from the Microbial Disease Laboratory at the California Department of Public Health (MDL), which receives specimens from outpatient facilities and clinics on Pacific islands. Of the 353 AFB culture-positive specimens at UCIMC, there was a statistically significant difference in the sensitivity of the direct AFB smear (34%) and that of the smear made from the concentrated specimen (58%) (P < 0.05). This was also true for the 208 specimens positive for Mycobacterium tuberculosis, for which the sensitivity of the direct smear was 42% (87 of 208) and that for the smear made from the concentrated specimen was 74% (154 of 208). At MDL, where all but 1 of the 45 culture-positive specimens grew M. tuberculosis, the sensitivity of the smear made from the concentrated specimen was 93% (42 of 45) and was not significantly higher than the sensitivity of the direct smear, which was 82% (37 of 45). By combining the results from both laboratories, 42 patients from whom at least three specimens were received were culture positive for M. tuberculosis. The cumulative results for the initial three specimens from these patients showed that the direct smear detected M. tuberculosis in 81% of these patients, whereas the smear made from the concentrate detected M. tuberculosis in 91% of these patients. In summary, when all culture-positive specimens are considered, the sensitivity of the direct smear compared to that of a smear made from the concentrated specimen was significantly different overall in the two different laboratory settings. However, this difference was reduced only if the cumulative results for the initial three specimens received from patients who were culture positive for M. tuberculosiswere evaluated.