Selection of Character Coding Method Is Not Phylogenetically Neutral: A Test Case Using Hominoids.

The early stages of phylogenetic inference from morphological data involve a sequence of choices about which analytical methods to employ. At each stage, the selection of one method over another can dramatically impact tree inference. Phylogenetic hypotheses are sensitive to decisions relating to which taxa and characters to select for analysis, whether and how to delimit character states, which taxa to use as outgroups, and how to account for character dependence. Using extant hominoids as a test case, I quantify the degree to which phylogenetic inferences are sensitive to the choice of method used to transform continuously scaled variables into categorical traits. I demonstrate that the character coding strategy significantly impacts hypotheses of character state identity and phylogenetic branching patterns. To avoid biasing evolutionary hypotheses, I recommend that continuously scaled characters be analyzed without prior discretization.