Résumé : The purpose of the study was to evaluate the contribution of polymerase chain reaction (PCR) to the diagnosis of tuberculous infection in children. Two different PCR techniques were compared to the standard bacteriological methods for the detection of Mycobacterium tuberculosis in 157 specimens obtained from the respiratory system of 51 children. Patients were classified in three groups: 12 patients with active disease (57 specimens), 12 patients with silent tuberculous infection (23 tuberculosis (77 specimens). One PCR method (PCR/Ag85) used amplification of a fragment of the genes coding for the mycobacterial antigen 85 followed by hybridization of a probe specific for M. tuberculosis on the Southern blot of amplified DNA. The other PCR technique was a nested PCR (NPCR) using double amplification of a fragment of the insertion element IS6110 only present in the M. tuber culosis genome. The sensitivities of the different techniques, compared to the clinical diagnosis, were 7.0% for acid fast staining, 22.8% for culture, 24.6% for PCR/Ag85 and 44.9% for NPCR in active disease, 4.3% for culture, 8.7% for PCR/Ag85 and 28.6% for NPCR in silent tuberculous infection. The specificities were 100% for culture, 94.8% for PCR/Ag85 and 87.9% for NPCR. Among the 12 children clinically considered as having active tuberculosis, 1 had smear positive samples, 4 had at least one positive culture, 7 at least one positive PCR/Ag85 and 9 at least one NPCR positive sample. Among the 12 children having silent tuberculous infection, none had positive smears, 1 had one positive culture, 2 had at least one positive PCR/Ag85 and 5 at least one NPCR positive sample. Conclusion: Our study suggests that both PCR techniques, and especially NPCR, are able to detect M. tuberculosis DNA in specimens containing few micro-organisms. PCR methods are more sensitive than culture and the results are available more quickly. Testing multiple samples from the same individual increased the sensitivity. In view of occasional false-positive results, cultures remain the gold standard to establish definitive diagnosis of primary tuberculous infection in children.