Fossil Caminalcules

Cladogram

Phenogram in a phylogeny, phenetic groupings will fail to link the genealogically closest branch points and place emphasis on the degree of phenetic divergence.

This criticism, which was Farris's fatal shot at phenetics, has come back to haunt phylogenetic approaches as well. Felsenstein (1982) pointed out that unequal rates of evolution, combined with short internal internode lengths, is an impediment to the success of phylogenetically based methods employing parsimony. This was shown in a simulation study that demonstrated a zone of chaos, where rates of change were highly unequal. Huelsenbeck and Hillis (1993) identified this type of tree topology as the "Felsenstein zone" (Figure 2.11), which hoisted Farris on his own petard. Although all phylogenetic techniques may fail under some extreme circumstances, it does turn out that the UPGMA clustering approach appears to under-perform relative to other groups in recovering the correct evolutionary relationships.

Problems with phenetic clustering can be seen in Sokal's (1983a, 1983b, 1983c) comprehensive study of the caminalcules, a group of synthetic creatures whose complete history, including phylogeny, fossils, and recent species are known by definition. The phenetic classification is superposed on the phylogeny in Figure 2.12. Groups A, B, C, and D, labeled by common shading patterns, are those clustered above one arbitrary level of phenetic similarity, whereas numbered subgroups (e.g., A1, A2) are clustered above another arbitrary but still higher level of similarity. B1 is the direct phyletic ancestor of B2, but the classification groups them as of equal rank. In another case, C1 gives rise to D2, which in turn gives rise to C4. The A group is of particular interest. It includes the ancestral group, A1, and a derived

Figure 2.11. Four tree topologies of four taxa with different branch lengths. In Region I, methods such as parsimony and neighbor joining do well at recovering the correct tree, although the Unweighted Pair Group Method (UPGM) and Lake's invariant method do relatively poorly. Performance falls off in region II, where branches are long. Region FZ is the Felsenstein zone, a set of topologies that cannot be readily analyzed successfully by any algorithms to recover the correct evolutionary tree. (After Huelsenbeck and Hillis 1993.)

Figure 2.11. Four tree topologies of four taxa with different branch lengths. In Region I, methods such as parsimony and neighbor joining do well at recovering the correct tree, although the Unweighted Pair Group Method (UPGM) and Lake's invariant method do relatively poorly. Performance falls off in region II, where branches are long. Region FZ is the Felsenstein zone, a set of topologies that cannot be readily analyzed successfully by any algorithms to recover the correct evolutionary tree. (After Huelsenbeck and Hillis 1993.)

Figure 2.12. Phylogeny of the caminalcules. Shaded pattern unites group at the 0.0 phenon (similarity) level. Numbered subgroups are united at the 0.5 level. Vertical axis indicates arbitrary time units. (After Sokal 1983c.)

"radiation," (A2, A3, A4, A5, A6) as groups of equal rank. But A1 on the phenogram appears to be strongly derived, whereas it is the ancestor group of the entire clade! The groupings, therefore, do not present a consistent picture as to genealogy, nor is there a way to connect the groups, given the absence of information or hypotheses on polarity of evolutionary change or derived states in common. Indeed, it is the precision of identification of defining character states in Hennigian trees that allows us to see precisely how a tree is defined. By contrast, phenetic indices amalgamate many characters into one index.

The controversy over phenetics has died because cladistic techniques have won the hearts and minds of systematists and evolutionary biologists. Overt phylogenetic approaches now dominate the pages of molecular and morphological evolution journals. Parsimony has struck most as a sensible hypothesis of evolutionary deriva-

tion, but even more importantly, most systematists have recognized the utility of identifying monophyletic groups with a system whose information can be mapped simply as character changes on a tree of evolutionary relationships. Indeed, although parsimony dominates the thinking of evolutionary biologists concerned with morphology, molecular data sets have lent themselves to other approaches that perform as well as parsimony in many cases (see the next section).

0 0

Post a comment