Can Partial Warp Scores Be Used as Cladistic Characters?

Abstract

Zelditch et al. (1995) have recently proposed that partial warp scores derived from shape differences among organisms can be used as cladistic characters for phylogenetic inference. The current study represents a test of Zelditch et al.’s proposal. A group of different—shaped imaginary fishes, interrelated by a known phylogeny, was artificially generated by subjecting an ancestral fish to evolutionary change along a predetermined tree. Care was taken to ensure that invoked shape changes occurred only once on the tree and that shape changes occurred independently of each other. These conditions were set in order to maximally predispose any emergent “shape characters,” which might be derived from shape differences among terminals, to cladistic analysis. Thirty-eight landmarks were recorded for each of the eight terminals resulting from the contrived evolutionary tree. These landmarks were transformed into shape coordinates. The shape coordinates for each of the terminals were then individually contrasted with those of the ancestor by means of the computer program TPSPLINE (Rohlf, 1994). The resulting partial warp scores were recoded as cladistic characters following the protocol of Fink et al.(1994) and subjected to cladistic analysis. Two most parsimonious trees (MPTs) resulted, one of which matched the “true” tree. Although this tree matched the true tree in topology, the reconstructed ancestral shape states implied by parsimony for this tree were quite different from those of the true tree. Whereas the true tree contained no homoplasy whatsoever, the two inferred MPTs implied considerable shape homoplasy (Retention Index = 0.48). An UPGMA of raw partial warp scores also retrieved the correct tree topology.