Nine polymorphic mer loci carried by 185 gram-negative fecal bacterial strains from humans and nonhuman primates are described. The loci were characterized with specific intragenic and intergenic PCR primers to amplify distinct regions covering approximately 80% of the typical gram-negative mer locus. These loci were grouped phylogenetically with respect to each other and with respect to seven previously sequenced mer operons from gram-negative bacteria (the latter designated loci 1, 2, 3, 6, 7, 8, and delta 8 by us here for the purpose of this analysis). Six of the mer loci recovered from primates are similar either to these previously sequenced mer loci or to another locus recently observed in environmental isolates (locus 4), and three are novel (loci 5, 9, and 10). We have observed merC, or a merC-like gene, or merF on the 5' side of merA in all of the loci except that of Tn501 (here designated mer locus 6). The merB gene was observed occasionally, always on the 3' side of merA. Unlike the initial example of a merB-containing mer locus carried by plasmid pDU1358 (locus 8), all the natural primate loci carrying merB also had large deletions of the central region of the operon (and were therefore designated locus delta 8). Four of the loci we describe (loci 2, 5, 9, and 10) have no region of homology to merB from pDU1358 and yet strains carrying them were phenylmercury resistant. Two of these loci (loci 5 and 10) also lacked merD, the putative secondary regulator of operon expression. Phylogenetic comparison of character states derived from PCR product data grouped those loci which have merC into one clade; these are locus 1 (including Tn21), locus 3, and locus 4. The mer loci which lack merC grouped into a second clade: locus 6 (including Tn501) and locus 2. Outlying groups lacked merD or possessed merB. While these mer operons are characterized by considerable polymorphism, our ability to discern coherent clades suggests that recombination is not entirely random and indeed may be focused on the immediate 5' and 3' proximal regions of merA. Our observations confirm and extend the idea that the mer operon is a genetic mosaic and has a predominance of insertions and/or deletions of functional genes immediately before and after the merA gene. chi sites are found in several of the sequenced operons and may be involved in the abundant reassortments we observe for mer genes.