An evaluation of chemotherapy was made for 256 patients with pulmonary infection due to Mycobacterium kansasii. Studies of in vitro resistance of M. kansasii and Mycobacterium tuberculosis to various drugs revealed important differences between the two species that should be considered when planning chemotherapy. Partial or no in vitro resistance to 1 microgram of isoniazid/ml did not adversely affect the time required for conversion of sputum from positive to negative, whereas complete resistance to isoniazid had a statistically significant (P less than 0.001) adverse effect. All of the 115 patients who received treatment that included rifampin (initial and retreatment) had sputum conversion within four months. However, for regimens that did not include rifampin, 127 (90%) of 141 patients had conversion within four months (P less than 0.01). After completion of chemotherapy that included rifampin, there were no relapses among 64 patients, but after therapy without rifampin, four (7%) of 59 patients relapsed. The greater efficacy of the regimens that included rifampin may be due at least partially to the fact that more drugs were administered than in regimens that did not include rifampin. The use of p-aminosalicylic acid seems unwarranted in the treatment of disease due to M. kansasii. Since rifampin may be the most potent drug for treatment of this disease, the recommended combinations for initial treatment are (l) rifampin, streptomycin, isoniazid, and ethambutol, or (2) rifampin and two other drugs to which the organism is sensitive.