To determine the incremental cost of directly observed therapy (DOT) for patients with tuberculosis at low risk for treatment default, we applied a model of DOT effectiveness to 1,377 low-risk patients in California during 1995. The default rate for this cohort, which consisted of those with no recent history of substance abuse, homelessness, or incarceration, was 1.7%. The model predicted that DOT and self-administered therapy (SAT) cured 93.1 and 90.8% of these patients, respectively. DOT would initially cost $1.83 million more than SAT, but avert $569,191 in treatment cost for relapse cases and their contacts, for a net incremental cost of $1.27 million ($919 per patient treated), or $40,620 per additional case cured. The cost-effectiveness of DOT was sensitive to the default rate and relapse rate after completing SAT. DOT would generate cost savings only when the default and relapse rates were more than 32.2 and 9.2%, respectively. Given the low default rate and resulting high incremental cost of DOT, provision of DOT to low-risk patients in California should be evaluated in the context of resource availability, competing program priorities, and program success in completing self-administered therapy with a low relapse rate.