Null model selection, compositional bias, character state bias, and the limits of phylogenetic information

Evolutionary trends and processes can distort phylogenetic information in s equences such that they do not reliably reflect the evolutionary processes that generate them. This fact of molecular evolution has a ubiquitous influ ence on the ability of researchers to adequately reconstruct genealogical r elationships and histories of the processes of molecular evolution. This fe ature of phylogenetic inference can limit the capacity of researchers to ad equately specify a relevant null hypothesis for testing hypothesis of relat ionships, data informativeness, and processes of molecular evolution. We sh ow how this feature of historical inference also influences the exactness o f the relative apparent synapomorphy analysis (RASA) test for phylogenetic signal and demonstrate how a permutation modification of the null hypothesi s can improve the robustness of the underlying distributional assumption of the test. The RASA test (using either null model) was found not only to ap propriately reject the combinability of independent lines of evidence for t he relationships among the Physalaemus pustulosus frog species group, but a lso to be more appropriately sensitive to individual uninformative data set s than commonly used tree-based measures of signal, including the consisten cy index, the retention index, and the permutation tail probability test st atistic.