and Woese 2007). Microbial taxonomists have traditionally been much
more concerned with phenotypic characters than with phylogenetic relationships (Goodfellow et al. 1997, 26). This is due in part to con- jugation, a process in which bacteria are able to pass genetic material (and thus ultimately phenotypic traits) to other individuals that are not their own offspring. Such “cross-borrowing” tends to decrease the ex- planatory utility of cladistic analysis to the extent that these represent lineages of evolving populations of organisms. There are many cladistic analyses of prokaryotic life; indeed, the idea that the tree of life is organized into three ancient domains, the archaea, the eubacteria, and the eukaryotes, is based on the discovery of an an- cient split among the prokaryotes. But there is an important sense in which these trees are histories of gene lineages rather than organisms.9 For one thing, these phylogenies are based on genetic data: the compari- son of homologous genes. But more importantly, if cross-borrowing is a regular feature of bacterial life, we cannot assume that closely related genes—genes dating back to a recent common ancestor—are parts of the genome of closely related organisms. Among the eukaryotes, gene histories and organism histories are typically (though not universally) concordant. When two versions of a gene complex begin accumulating changes independently of each other, it will be because those genes’ lin- eages are contributing to two organism lineages evolving independently of each other. If genes are often transferred laterally, that is an assump- tion we can no longer make. We cannot treat the branching pattern of gene evolution as a proxy for the branching pattern of the organism lineages of which they are a part. Moreover, there is a set of technical challenges to the tractability of their approach. Parsimony analysis is computationally intensive. Find- ing the most likely phylogeny is made difficult by the fact that the num- ber of possible “trees” increases exponentially as more taxa are added to the analysis. Three taxa can be arranged in only three different rooted tree topologies (ones that show both phylogenetic relationships be- tween taxa and also pick out a particular taxon as being the ancestor of all the others). Five can be arranged in 105. Ten can be arranged in 3.4 × 107, while twenty can be arranged in 8.2 × 1021 (Quicke 1993, 59). It is true that there are algorithms to find the most parsimonious tree that do not involve exhaustively searching all the possible tree topologies. Even so, in practice computation tractability will always be an issue for cladistic analysis. In short, cladistics is explanatorily powerful, with a well-defined ra- tionale. These are great virtues in a taxonomic system. But it purchases these virtues at the cost of abandoning an explicit representation of