Simplicity and Informativeness in Systematics and Phylogeny

Publisher Website
Google Scholar
Abstract
Advocates of syncretistic classification have generally held that the descriptive and explanatory roles of the biological reference system should be kept separate, and that description and explanation impose conflicting goals on classification. That view leads to contradictions; earlier demonstrations that the phylogenetic system satisfies both goals at once are discussed. Pheneticists have advocated ordinations as an alternative to heirarchic classification, claiming for them superior descriptive power. Ordinations are more general than had been thought, as hierarchies can be subsumed under such structures. This viewpoint allows a great many classificatory criteria to be analyzed under a factor or regression model. Present phenetic criteria for clustering amount to evaluating a multiple correlation as the sum of simple correlations. This fallacy can be avoided by regressing data multiply on classificatory factors (groups), but then perfect fit to data can always be achieved, so that the classificatory problem becomes one of the choosing among alternative sufficient bases. This problem is approached by analogy with simple structure criteria of factor analysis. Classificatory simple structure leads to the phylogenetic parsimony criterion. From this new viewpoint the phylogenetic system thus emerges again as a unified descriptive and explanatory structure. Objections to the descriptive power of the phylogenetic system are discussed with emphasis on recent claims by Rohlf and Sokal. Attempts to defend phenetic classification have comprised nothing other than abandoning or rendering untestable every claimed advantage of phenetic methods. There is thus no possible substantive basis for preference for phenetic grouping.