In this study of 102 patients with culture-positive chronic treatment failure or repeatedly relapsed pulmonary tuberculosis receiving chemotherapy, 56 received an injection of killed Mycobacterium vaccae as immunotherapy after 1 month of treatment. At the start of treatment, there was little difference between those receiving immunotherapy and the 46 patients in the control group receiving chemotherapy alone. Thereafter, the two groups diverged so that 1 yr later, 43 of 56 (77%) patients receiving M. vaccae had a successful outcome, in comparison with 24 of 46 (52%) patients receiving chemotherapy alone (P < 0.02). Successful results were obtained from patients infected with drug-resistant bacilli, 20 of 32 (63%) patients compared with 11 of 25 (44%) patients, respectively, as well as from fully drug-sensitive cases (23 of 24 compared with 12 of 21 patients; P = 0.004). At the final follow-up after 22 months, 13 of 56 patients receiving immunotherapy had an unfavourable outcome compared with 26 of 46 members of the control group (P = 0.0006). During the study, 16 patients died of tuberculosis (six after immunotherapy), and 12 were lost to follow-up. Not only was bacteriological success improved by immunotherapy, chest X-ray showed markedly better resolution of cavities and other radiological lesions, recovery of body weight was improved, and the mean erythrocyte sedimentation rate returned almost to normal (P < 0.001) in comparison with those receiving chemotherapy alone. These changes were seen even in those failing bacteriological cure, suggesting that the immunotherapy had been effective, but that bacilli were replicating in an extracellular situation, protecting them from its effects.