Clinical laboratories are increasingly receiving requests to perform nucleic acid amplification tests for the detection of a wide variety of infectious agents. In this paper, the efficiency of nucleic acid amplification techniques for the diagnosis of respiratory tract infections is reviewed. In general, these techniques should be applied only for the detection of microorganisms for which available diagnostic techniques are markedly insensitive or nonexistent or when turnaround times for existing tests (e.g., viral culture) are much longer than those expected with amplification. This is the case for rhinoviruses, coronaviruses, and hantaviruses causing a pulmonary syndrome, Bordetella pertussis, Chlamydia pneumoniae, Mycoplasma pneumoniae, and Coxiella burnetii. For Legionella spp. and fungi, contamination originating from the environment is a limiting factor in interpretation of results, as is the difficulty in differentiating colonization and infection. Detection of these agents in urine or blood by amplification techniques remains to be evaluated. In the clinical setting, there is no need for molecular diagnostic tests for the diagnosis of Pneumocystis carinii. At present, amplification methods for Mycobacterium tuberculosis cannot replace the classical diagnostic techniques, due to their lack of sensitivity and the absence of specific internal controls for the detection of inhibitors of the reaction. Also, the results of interlaboratory comparisons are unsatisfactory. Furthermore, isolates are needed for susceptibility studies. Additional work remains to be done on sample preparation methods, comparison between different amplification methods, and analysis of results. The techniques can be useful for the rapid identification of M. tuberculosis in particular circumstances, as well as the rapid detection of most rifampin-resistant isolates. The introduction of diagnostic amplification techniques into a clinical laboratory implies a level of proficiency for excluding false-positive and false-negative results.