Effective treatment of tuberculosis requires adherence to a minimum of 6 months treatment with multiple drugs. To improve adherence and cure rates, directly observed therapy is recommended for the treatment of pulmonary tuberculosis. We compared treatment outcomes among all culture-positive patients treated for active pulmonary tuberculosis (n = 372) in San Francisco County, California from 1998 through 2000. Patients treated by directly observed therapy at the start of therapy (n = 149) had a significantly higher cure rate compared with patients treated by self-administered therapy (n = 223) (the sum of bacteriologic cure and completion of treatment, 97.8% versus 88.6%, p < 0.002), and decreased tuberculosis-related mortality (0% vs. 5.5%, p = 0.002). Rates of treatment failure, relapse, and acquired drug resistance were similar between the two groups. Forty-four percent of patients who received self-administered therapy had risk factors for nonadherence and should have been assigned to directly observed therapy. We conclude that treatment plans that emphasize directly observed therapy from the start of therapy have the greatest success in improving tuberculosis treatment outcomes.