This prospective study was undertaken to determine the interpretation of "scanty-positive" acid-fast bacilli on microscopy and to reevaluate simultaneous microscopy and culture of sputum for the accurate diagnosis of pulmonary tuberculosis (PTB). A total of 2,560 specimens were processed from 727 patients. There were 435 positive specimens (17.0 percent), originating from 139 patients, 10 by microscopy only, 176 by culture only, and 249 on both microscopy and culture. Review of the hospital records showed that 107 patients had PTB, 1 had Mycobacterium kansasii colonization, and 31 were thought not to have PTB. Sensitivity and specificity were 53.1 and 99.8 percent for microscopy, 81.5 and 98.4 percent for culture, and 77.6 and 100 percent for microscopy and culture, respectively. Seventy-five microscopy specimens (46 patients) were reported as scanty-positive, of which five (four patients) were deemed false positives, yielding a positive predictive value of 93.3 percent. In those patients with positive sputum microscopy, acid-fast bacilli were detected in one of the first four specimens. Seven isolates (three patients) were mycobacteria other than tubercle (0.27 percent of specimens and 1.6 percent of mycobacteria cultured). Despite the ready availability of laboratory evidence of disease, only 73 percent of cases were diagnosed by ward staff and 36 percent notified by the primary physician. Eleven patients (10.3 percent) died, six of whom had not received diagnoses of PTB before death. Sputum microscopy and culture remains reliable despite Bayesian predictions when applied to a population with a decreasing incidence of tuberculosis.