More than 20 near full-length genome sequences have been reported for human polyomavirus JC (JCV). These have previously been classified into seven genotypes, and additional subtypes, which exhibit geographical associations. One of these genotypes, Type 4, has been suggested to be a recombinant of Types 1 and 3. We have investigated the pattern of diversity, and evolutionary relationships, among these sequences. In direct contradiction of a recent report, we found that different phylogenetic methods gave consistent results for the phylogenetic relationships among strains. The single known strain representing Type 5 was shown to be a mosaic of sequences from Types 2 and 6, although whether this recombination occurred in vivo or in vitro is not clear. In contrast, there was no substantial evidence that Type 4 strains are recombinant; rather they seem to be simply divergent examples of Type 1. On the assumption that the major genotypes of JCV diverged with human populations, the rate of synonymous nucleotide substitution was estimated to be around 4x10(-7) per site per year, about 10 times higher than a previous estimate for primate polyomaviruses.