CONJECTURES AND REFUTATIONS

Phylogeneticists familiar with Popper's ideas have long regarded mostparsimonious trees as bestcorroborated hypotheses of relationship, but Faith (1992) feels that those trees are leastcorroborated instead. Not that he recommends least-parsimonious trees, rather (p. 268) "Popperian corroboration is properly applied to a higher-level cladistic hypothesis." That hypothesis being that analyzing the data has revealed something of phylogeny. And, conveniently, his permutation tail probability (PTP) establishes both the content (boldness) and the degree of corroboration of that hypothesis. But in fact no one quantity can measure both content and corroboration; and content, unlike PTP, is not calculated from presently available evidence. In trying to connect PTP to corroboration, Faith confuses pointwith cumulative probabilities, evidence with the hypothesis tested, and null models with accepted theories. In arriving at his view of most-parsimonious trees he repeats most of the same mistakes, as will be seen here.