TESTING PHYLOGENETIC HYPOTHESES WITH STRATIGRAPHY AND MORPHOLOGY—A COMMENT ON SMITH (2000)

Abstract

Smith (2000) critiqued several recently proposed phylogenetic methods that accommodate stratigraphic data as well as several simulation studies that examine the efficacy of such methods. Here, I will evaluate Smith's claims concerning: 1) the putative ad hoc nature of some tests using stratigraphic data; and 2) the relevancy of simulation studies to empirical studies. #### Ad hoc versus hypothetico-deductive tests Wagner (1995) used confidence intervals on observed stratigraphic ranges (Strauss and Sadler, 1989; Marshall, 1990) to test taxon durations hypothesized by a possible phylogeny. Smith (p. 764) dismissed this approach as ad hoc “weighting” of character and stratigraphic data. However, Wagner's use of confidence intervals is a hypothetico-deductive approach that assigns different roles to the two datasets. Morphology is used abductively (inferring a condition based on assumed properties of that condition; see Sober, 1988) to hypothesize that species X and Y are closest relatives among sampled taxa. If so, then: A) if one is ancestral, then the other's lineage diverged within the ancestor's duration; or B) if they are true sister taxa, then their lineages share a common divergence time. A and B both limit expected observable patterns in the fossil record. If stratigraphic data deviate from those expectations, then we reject the hypothesized durations and phylogeny. Smith (2001b) employed similar logic in a “stratigraphically restricted” phylogenetic analysis of echinoids, albeit without formalized rejection of possible relationships between old and (excluded) young taxa. This represents modus tollens deduction (i.e., dismissing an explanation because observation deviates from expectation). Why we even considered a hypothesis in the first place is irrelevant, except that we now cannot accept the premises that led to that hypothesis (see below). The germane point is that analyses formalizing expectations in a deductive statement are not ad hoc. #### The complexities of stratigraphy and the soundness of tests Although not ad hoc, hypothetico-deductive tests can be unsound (i.e., based on false premises). Smith …