A 2470-bp section of the 23S ribosomal DNA from Serpulina hyodysenteriae and five biochemically different groups of weakly beta-haemolytic porcine intestinal Serpulina strains was sequenced. The similarity between the sequenced strains was high (96.85% to 99.84%). A phylogenetic tree was estimated by the maximum likelihood method. The sequenced strains formed three groups. Serpulina hyodysenteriae and biochemical group II ('S. intermedius') formed a cluster, but 20 nucleotide positions were different between the two, suggesting that biochemical group II is a separate species. Another cluster consisted of the closely related biochemical group IIIa ('S. murdochii') and IIIb/c (S. innocens) (99.84% similarity), while biochemical group IV (S. pilosicoli) constituted a separate group with a relatively low similarity (96.85% to 97.01%) to the other groups. Three primer pairs were designed for specific PCR detection of the clinically important S. hyodysenteriae and biochemical group II and IV. PCR amplification was accomplished with DNA extracted from bacterial colonies by a simple boiling procedure, and with DNA extracted directly from porcine stool samples using a bead beating extraction procedure. The level of detection for the direct extraction and amplification method was 5 x 10(5) cells added g-1 normal faeces.