The use of the polymerase chain reaction (PCR) for assessing treatment response in tuberculosis was investigated. Serial sputum samples were analyzed from 10 Tanzanian patients treated for smear-positive pulmonary tuberculosis, including 4 who relapsed after initially successful treatment. A one-tube nested PCR with a colorimetric detection system was compared with microscopy and culture. Samples were found to be negative by microscopy before they were by PCR or culture, often remaining positive 1-2 months longer by PCR than by culture. For the 76 samples available for both culture and PCR, there was a 76% (58/76) agreement between the methods. Nine samples were negative by culture but positive by PCR; 7 were either negative (5) or equivocal (2) by PCR despite being positive by culture. Two of the 4 relapse cases were detected earlier by PCR than by culture. These results demonstrate that PCR is a promising method for assessing treatment response in pulmonary tuberculosis.