Phenetics

Phenetic approaches group taxa by their overall similarity (Felsenstein 1982; Neff and Marcus 1980; Sneath and Sokal 1973; Sokal 1986; Sokal and Sneath 1963). A matrix of taxa by character states is used to calculate a correlation (or distance) matrix among the taxa. These correlations (distances) are then employed in a grouping algorithm to construct a tree, by successive pairings, according to successively decreasing correlations (e.g., the Unweighted Pair Group Method, or UPGMA, described in Sneath and Sokal 1973). The resulting tree has the advantage of a defined root and branching topology, characterized by given levels of overall similarity (correlation), which is amenable to a hierarchical organization. The groups are defined to a degree by all of the characters under consideration in producing the clustering. A set of subgroups included within a group need not have the set of character states that uniquely define the larger group. By contrast, cladistically defined groups have the important characteristic that all subgroups have the same character states as those that define the more inclusive group, as well as some unique states of their own, which define the lower taxonomic level of the subgroups.

Although trees can be readily constructed by the UPGMA, phenetic methods can also be employed to define clusters for the sake of defining distinct groupings, such as fossil "species" (Budd and Coates 1992; Gingerich 1979). Species are defined as phenetically clustered specimens, distinct from other clusters. Stratophenetics has been used to merge phenetic groups between strata to define temporally separated lower level taxa. (See Smith 1994 for an excellent discussion of these approaches.)

The theory behind the original phenetic approach supposed the premise of non-specificity (Sokal and Sneath 1963). Genes had sufficiently nonspecific (pleiotropic) effects across the phenotype that any large sampling of characters would reflect the genome and, therefore, would record genealogy. Different sets of characters (e.g., cephalic and pygidial in trilobites, larval and adult in moths) would therefore lead inevitably to the same classification. Incongruencies in classifications based on different suites of characters would therefore falsify the hypothesis.

Although a ferociously contentious literature exists on the relative abilities of phenetic and phylogenetic algorithms to produce more congruence among character sets (see discussion in Farris 1983; Mickevich 1978; Rohlf, Colless, and Hart 1983;

Rohlf and Sokal 1981; Schuh and Farris 1981; Sokal 1983c, 1986; Wiley 1981), it is not clear that the degree of congruence of either technique is especially good in any event (Mickevich 1978, Rohlf et al. 1983). This may be because the degree of nonspecificity is very limited. Poor congruence may derive from mosaic evolution. Differential rates of evolution of different character sets within the same mono-phyletic group would tend to produce different phenetic groupings using the different character sets (Farris 1971). But this would also weaken results using phylogenetic systematic methods (Huelsenbeck and Hillis 1993). Figure 2.10 demonstrates how mosaic evolution could cause differential phenetic groupings based on different character sets.

Grouping by overall similarity is more likely to lead to spurious conclusions from the genealogical point of view. Groups that have split off in the distant past but have diverged little phenetically will be grouped as close relatives. By contrast, groups that have split more recently but have diverged phenetically to a great degree will be grouped at a lower level of overall correlation. Farris (1971) gave the simple example of classifying birds, crocodiles, mammals, and snakes. A phenetic classification would group most closely the snakes and crocodiles, on the basis of similarity of both ancestral and derived characters. The lack of divergence between these two groups would obscure the genealogy relative to the other more phenetically divergent groups. To the degree that evolutionary rates are unequal in different branches

0 0

Post a comment