Current theory in the molecular epidemiology of tuberculosis holds that tuberculosis cases harboring Mycobacterium tuberculosis strains with a common deoxyribonucleic acid (DNA) fingerprint are the result of recent M. tuberculosis transmission. Here we propose a mathematical approach independent of DNA fingerprinting to estimating the percentage of recent transmissions responsible for current tuberculosis incidence. The "short-term reproductive number" of tuberculosis is defined as the average number of tuberculosis cases developing within 1 year of infection. Multiplying the short-term reproductive number by the number of tuberculosis cases in each year and dividing by the subsequent year's tuberculosis case burden equals the proportion of tuberculosis cases in the subsequent year that are due to recent transmission. We carried out separate calculations for human immunodeficiency virus (HIV)-negative and HIV-positive tuberculosis cases. We applied the model to pulmonary (infectious) tuberculosis cases diagnosed in New York City during 1989-1993, using tuberculosis and AIDS surveillance data. Model-based estimates of the proportion of tuberculosis due to recent transmission were lower than estimates based on DNA fingerprints. Reconciliation of these divergent estimates may require the re-estimation of model parameters from data collected de novo, additional model development, and further advances in DNA fingerprinting methods.