The number of pathogens involved in community-acquired pneumonia, with varying susceptibilities to antimicrobials, is numerous constituting an enormous challenge for diagnostic microbiology. Differentiation of infections due to Streptococcus pneumoniae and those due to Mycoplasma pneumoniae, Chlamydophila pneumoniae, or L. pneumophila as well as those due to viruses is essential to allow correct decisions concerning the antibiotics to be administered. The sensitivity and specificity of real-time simplex and multiplex nucleic acid sequence-based amplification (NASBA), and simplex PCR were compared for the detection of M. pneumoniae, C. pneumoniae and Legionella spp. in respiratory specimens from hospitalized and outpatients with community-acquired pneumonia (CAP). Two hundred fifty one respiratory specimens were collected from 147 patients with CAP. NASBA was done using the NucliSens Basic Kit (bioMérieux). PCR for M. pneumoniae and C. pneumoniae was done as described earlier [Ieven, M., Ursi, D., Van Bever, H., Quint, W., Niesters, H. G. M., and Goossens, H. 1996. Detection of Mycoplasma pneumoniae by two polymerase chain reactions and role of M. pneumoniae in acute respiratory tract infections in pediatric patients. J. Infect. Dis. 173, 1445-14452.; Ursi, D., Ieven, M., Van Bever, H. P., and Goossens, H. 1998. Construction of an internal control for the detection of Chlamydia pneumoniae by PCR. Mol. Cellul. Probes. 12, 235-238.]. A real-time PCR was developed to detect L. pneumophila whereas a real-time NASBA was designed to detect Legionella spp. All samples with discordant results were re-analysed. Compared to an expanded gold standard the sensitivities of the different techniques, were 77.8%, 100%, and 100% for detection of M. pneumoniae; and 50%, 100%, and 50% for detection of L. pneumophila by PCR, real-time simplex NASBA, and real-time multiplex NASBA, respectively. C. pneumoniae was detected in two samples only. Simplex real-time NASBA proved to be more sensitive than simplex PCR and was also more sensitive than real-time multiplex NASBA, as previously found with spiked clinical specimens. It's practical attractiveness pleads for further optimalisation of the multiplex approach.