DNA hybridization in animal taxonomy: a critique from first principles.

DNA hybridization is a "distance method" for phylogenetic reconstruction and, as such, shares a set of assumptions, advantages, and problems with other techniques that do not directly employ character data. The technique purports to measure the average percent mismatch of homologous nucleotide sequences between the single-copy genomes of species. This measurement, as any other, is subject to considerations of accuracy and precision. While replicate measurements and technical modifications can improve precision, the accuracy of such measurements is limited by the equivalence of genomes under comparison. Such routine events in genome evolution as gene duplication and deletion may complicate the interpretation of DNA hybridization distances. Beyond measurement limitations, the most serious potential distortions of distances are due to biased sequence sampling and homoplasy. These problems, however, do not necessarily preclude phylogenetic reconstruction, and their effects may be mitigated by numerical corrections. Homoplasy, in particular, is a difficulty faced by all methods of phylogenetic inference. If such distortions can be eliminated, mitigated by correction, or shown to be trivial, pairwise tree-construction strategies should provide reliable estimates of phylogeny.