Commercial tests measuring IFN-gamma responses to ESAT-6 and CFP-10 are available for diagnosing Mycobacterium tuberculosis infection. Measures that minimize cost and complexity will facilitate their application in less-developed countries. We investigated whether overlapping peptides representing both ESAT-6 and CFP-10 are required to detect M. tuberculosis infection in a high TB-burden country, and whether they can be combined in a single pool. ESAT-6 and CFP-10 peptides were compared in IFN-gamma enzyme-linked immunospot (ELISPOT) in 183 HIV-negative smear-positive TB cases and 1673 HIV-negative household contacts. Separate peptide pools for each antigen were compared with a combined pool in 498 contacts. Forty per cent of responsive contacts recognized both antigens, 51% only ESAT-6 and 10% only CFP-10, whereas 56% of responsive cases recognized both antigens, 30% only ESAT-6 and 13% only CFP-10. Accordingly, CFP-10 response rates were higher for TB cases (odds ratio 2.409, P<0.001). Low purified protein derivative response rates indicated that responses to CFP-10 only were non-specific in contacts. Agreement between peptides in separate versus combined pools was good (kappa=0.758, r=0.840). Therefore a combined ESAT-6/CFP-10 peptide pool provided maximum sensitivity and efficiency, but CFP-10 was mainly required to detect active disease.