The New York City tuberculosis (TB) case rate declined from 1991 to 1994 following more than a decade of increases. The present study investigated TB trends in New York City neighborhoods and their association with neighborhood-specific rates of application of directly observed therapy (DOT). Using Poisson regression models, TB trends in each of New York City's 30 health districts were classified as increasing, decreasing, or stable, as indicated respectively by significant positive, negative, or nonsignificant regression coefficient. Case counts increased in four health districts, decreased in 10, and were stable in 16. Decreasing TB was associated with a higher rate of application of DOT. TB cases among foreign-born persons increased in 12 health districts and were stable in 18, whereas cases among persons born in the United States decreased in 19 and were stable in 11 districts. Among the foreign-born, increasing TB was not associated with a lesser rate of application of DOT. These data provide some support for the role of DOT in containing TB, but also suggest that the application of DOT among foreign-born residents is less effective than among United States-born residents. This may be due to a greater proportion of TB cases among the foreign-born being due to reactivation of TB rather than new infection.