One hundred water samples (32 from clinical units and 68 from private households) were examined for Legionella by culture, fluorescence in situ hybridization (FISH) and polymerase chain reaction (PCR). Twenty-four samples were positive by culture (22 L. pneumophila; 2 non-pneumophila species), 36 by FISH (32 L. pneumophila; 4 non-pneumophila species) and 75 by PCR (41 positive for L. pneumophila; 26 positive for L. pneumophila and a non-pneumophila species; 8 positive for non-pneumophila species). PCR and FISH results were compared to bacterial culture as the "gold standard" method by calculating sensitivities and specificities, respectively: PCR assays, 96% and 47%; FISH assays, 67% and 72%, respectively. In comparison with FISH the lower specificity of PCR is probably caused by dead Legionella bacteria and/or free Legionella DNA in potable water, and the higher sensitivity of PCR may be explained by the detection limit of fluorescence microscopy. In conclusion, the relatively high specificity, sensitivity and quickness of the FISH assay offer significant advantages over conventional PCR and culture-based techniques.