Water samples collected from 28 dental facilities in six U.S. states were examined for the presence of Legionella pneumophila and other Legionella spp. by the PCR-gene probe, fluorescent-antibody microscopic, and viable-plate-count detection methods. The PCR and fluorescent-antibody detection methods, which detect both viable and viable nonculturable Legionella spp., gave higher counts and rates of detection than the plate count method. By the PCR-gene probe detection method, Legionella spp. were detected in 68% of the dental-unit water samples and L. pneumophila was detected in 8%. Concentrations of Legionella spp. in dental-unit water reached 1,000 organisms per ml or more in 36% of the samples, and 19% of the samples were in the category of 10,000/ml or above. L. pneumophila, when present in dental-unit water, never reached concentrations of 1,000/ml or more. Microscopic examination with fluorescent-antibody staining indicated that the contamination was in the dental-unit water lines rather than in the handpieces. Legionella spp. were present in 61% of potable water samples collected for comparative analysis from domestic and institutional faucets and drinking fountains; this percentage was not significantly different from the rate of detection of Legionella spp. in dental-unit water. However, in only 4% of the potable water samples did Legionella spp. reach concentrations of 1,000 organisms per ml, and none was in the 10,000 organisms-per-ml category, and so health-threatening levels of Legionella spp. in potable water were significantly lower than in dental-unit water. L. pneumophila was found in 2% of the potable water samples, but only at the lowest detectable level.(ABSTRACT TRUNCATED AT 250 WORDS)