Background: Over the past decade, there has been a reduction in the incidence of tuberculosis in New York City and in the United States. However, the reduction has been confined mainly to U.S.-born persons. Understanding the reasons for the lack of reduction among non-U.S.-born persons may lead to new strategies for tuberculosis control. Methods: We performed DNA fingerprinting with the IS6110 insertion sequence of the organisms isolated from patients with culture-positive tuberculosis in northern Manhattan from 1990 to 1999. The goal was to identify the strains responsible for multiple infections, presumably through recent transmission (clusters of cases), as well as the strains found in only one patient, presumably representing reactivation of latent infection. Results: Of 546 available isolates of Mycobacterium tuberculosis, 261 (48 percent) belonged to a cluster and 285 (52 percent) did not. In multivariate analysis, significant predictors of noncluster status included birth outside the United States (odds ratio for a strain causing a cluster among non-Hispanic foreign-born patients, 0.31; 95 percent confidence interval, 0.14 to 0.66; odds ratio among Hispanic foreign-born patients, 0.51; 95 percent confidence interval, 0.30 to 0.88), age greater than 60 years (odds ratio, 0.37), and diagnosis after 1993 (odds ratio, 0.50). All these characteristics appeared to be associated with reactivation disease rather than with tuberculosis due to recent transmission. Homelessness was associated with clustering (odds ratio, 1.78; 95 percent confidence interval, 0.99 to 3.20) and therefore with recent transmission. Conclusions: These findings from northern Manhattan suggest that among foreign-born persons, tuberculosis is largely caused by reactivation of latent infection, whereas among U.S.-born persons, many cases result from recent transmission. Strategies for the control and elimination of tuberculosis among foreign-born persons at high risk should be directed toward the treatment of latent tuberculosis infection.