- The Paleontological Society
We applaud Sumrall and Brochu's (2003) reanalysis of our data matrix and their success in finding trees shorter than those we reported. We also agree with some points in their comment, such as the importance of distinguishing between general properties of a method and operational attributes of any given application of the method. However, a number of points in their comment are misdirected. We shall try to clarify the most important areas of disagreement, keeping track of those that deal with stratocladistics in general, as opposed to our blastoid example.
PHYLOGENETIC AND CLADISTIC RESOLUTION
The distinction between precision and accuracy is not unique to phylogenetics; it is a concept native to all measurement and estimation. Yes, accuracy is hard to assess in phylogenetic studies of fossil organisms, and “we cannot know if a reduced number of optimal trees includes the true phylogeny” (Sumrall and Brochu, 2003, p. 189), but this sentence continues, “contained in a larger set.” Which set? Certainly not the cladistic solution-set, any more than that of stratocladistics. In contrast, Fox et al. (1999) compared cladistic and stratocladistic results to “true” (simulated) histories, and stratocladistics out-performed cladistics under a wide (and realistic) range of evolutionary models (Fisher et al., 2002). Given the importance of accuracy, we have always treated precision as secondary, and our discussion of the resolution of cladistic and stratocladistic results included explicit recognition of this (Bodenbender and Fisher, 2001, p. 363).
Sumrall and Brochu state, “Stratocladistics is said to recover fewer optimal trees, and to have better-resolved consensus trees, than standard parsimony analysis … ” (2003, p. 189; to simplify communication, we here mostly follow their usage of cladogram = tree, although this tends to conflate cladogram-trees with “evolutionary trees”). The first claim is accurate, but the second (attributed by them to Fox et al., 1999) has never been …