A Phylogenetic Analysis of the Caminalcules. II. Estimating the True Cladogram

Abstract

-The ability of numerical methods to estimate the true cladogram is examined, using the Caminalcules as an example. This group of was generated artificially according to principles believed to resemble those operating in real organisms. Estimated cladograms obtained by numerical methods are only moderately good estimates of the true cladogeny. Of the various models applied in numerical cladistics, the closeness of approximation to the true cladogeny is in the following order: best-Wagner parsimony and Camin-Sokal parsimony; second-polymorphism parsimony; third-character compatibility and Fitch's nonsequential method; fourth-UPGMA phenograms; worst-Dollo parsimony. Since the data matrix contains NC (no comparison) states, numerical cladistic algorithms that take account of these states give better results than those that do not. For distance Wagner algorithms, there is little difference in the outcome between midpoint rooting and rooting with a zero vector or with the true ancestor. By separate permutations of the order of the OTUs that are input to the numerical algorithm, trees of varying length were obtained from which the shortest could be chosen. Shortest trees, computed globally over all OTUs or separately for each Caminalcule genus, are not necessarily those that best estimate the true cladogeny. Numerical cladistic estimates separately computed for genera are poorer than those based on the entire taxon. The effects of homoplasy and divergence on the results of phenetic and cladistic algorithms are investigated in general and employed to explain in detail the differences among the true cladogram, the phenogram, and the estimated cladogram. The agreements observed are all in cases where given taxa diverge greatly from ancestral stems. Discrepancies are due to (A) parallelisms in the cladogeny affecting phenetic similarities and (B) divergence of cladistically close relatives increasing the relative phenetic similarity of cladistically more distant relatives. These two cases are about equally frequent. [Phenetic classifications; cladistic classifications; estimated cladograms; homoplasy; Wagner trees; Caminalcules; numerical taxonomy.] The availability of the true cladogeny of the Caminalcules, a group of artificially created organisms, has presented an opportunity for examining various principles and practices of current interest to systematists. In a previous paper (Sokal, 1983a), I presented for the first time the images of all 77 Recent and fossil Caminalcules, definitions of their characters, and the data matrix employed in this study, as well as a cladogram of their true phylogeny. A new standard phenogram was also featured. The current study investigates the accuracy of numerical estimates of the true cladogram from an assemblage of Recent species and examines the reasons for the departure of numerical phenetic and cladistic dendrograms from the correct genealogy. The following two sets of related questions are examined in this paper. (1) How reliable are various numerical approaches for estimating the true cladogeny? Can reasons be found for their varying degrees of success? How does the nonnumerical, traditional Hennigian method compare with numerical approaches? (2) How does Hennigian classification based on the true cladogeny compare with a classification based on numerical phenetic procedures applied to the images of the Caminalcules? What evolutionary phenomena are responsible for the observed discrepancies between the phenetic classification and that based on the true cladogeny? The question of the relevance of the Caminalcules to the study of real taxonomic organisms has been discussed in Sokal (1983a) and need not be repeated here. For relevant parameters, the Recent and fossil Caminalcules provide values that resemble those resulting from data sets of real organisms. Consequently, questions raised by the Cam-