A data based parsimony method of cophylogenetic analysis

Phylogenies of closely interacting groups, such as hosts and parasites, are seldom completely congruent. Incongruence can arise from biologically mean ingful differences in the histories of the two groups, or can be generated by artifactual differences that are merely the result of incorrect phylogen ies with weakly supported nodes. We present a method that distinguishes bet ween these sources of incongruence and identifies lineages that are respons ible for significant differences between phylogenies. We use the logic of c onditional combination in that we first test for statistically significant incongruence using the partition homogeneity test. Then we remove all possi ble combinations of taxa until a non-significant result of this test is ach ieved. Finally, we construct a 'combined evidence' phylogeny and then repos ition the incongruent taxa. This method produces trees for final comparison using reconciliation methods, but it includes only as many incongruence ev ents as can be statistically justified from the data sets. We apply this me thod to a host-parasite (gopher-louse) data set and identify many fewer inc ongruence events than do topology based analyses alone. Our method is broad ly applicable to comparisons of phylogenies of interacting taxa, such as ho sts and parasites, or mutualists. The method should also be useful for othe r problems involving comparisons of phylogenies, such as multiple gene tree s or cladistic biogeography.