Identification of individuals infected with Mycobacterium tuberculosis is essential for the control of tuberculosis (TB). The specificity of the currently used tuberculin skin test (TST) is poor because of the broad antigenic cross-reactivity of purified protein derivative (PPD) with BCG vaccine strains and environmental mycobacteria. Both ESAT-6 and CFP-10, two secretory proteins that are highly specific for M. tuberculosis complex, elicit strong T-cell responses in subjects with TB. Using an enzyme-linked immunospot (ELISPOT)-IFN-gamma assay and a restricted pool of peptides derived from ESAT-6 and CFP-10, we have previously demonstrated a high degree of specificity and sensitivity of the test for the diagnosis of TB. Here, 119 contacts of individuals with contagious TB who underwent TST and the ELISPOT-IFN-gamma assay were consecutively recruited. We compared the efficacy of the two tests in detecting latent TB infection and defined a more appropriate TST cutoff point. There was little agreement between the tests (k = 0.33, P < 0.0001): 53% of the contacts with a positive TST were ELISPOT negative, and 7% with a negative TST were ELISPOT positive. Furthermore, respectively 76 and 59% of the ELISPOT-negative contacts responded in vitro to BCG and PPD, suggesting that most of them were BCG vaccinated or infected with nontuberculous mycobacteria. The number of spot-forming cells significantly correlated with TST induration (P < 0.0001). Our in-house ELISPOT assay based on a restricted pool of highly selected peptides is more accurate than TST for identifying individuals with latent TB infection and could improve chemoprophylaxis for the control of TB.