The Information Content of the Phylogenetic System

Farris, J. S. (Department of Ecology and Evolution, State University of New York, Stony Brook, New York 11794) 1979. The information content of the phylogenetic system. Syst. Zool. 28:483–519.—Both pheneticists and evolutionary taxonomists claim that their classifications contain more information on characters and degrees of similarity than does the phylogenetic system, but until recently no attempt had been made to measure the information content of phylogenetic classification. Extending previous results, I show that both distance and character information is better retained by phylogenetic than by phenetic classifications. The phylogenetic procedure of grouping according to synapomorphy corresponds to expressing distance information as a path-length distance, while phenetic clustering expresses distance information through clustering levels. It is impossible for a clustering-levels distance to fit distance data better than the best-fitting path-length distance. Classifications are most natural in Gilmour's sense, and have greatest information content per taxon, when they allow description of data in the fewest symbols. Most parsimonious trees have highest information content, are most natural, allow data to be completely summarized by the most succinct diagnoses, minimize exceptions to such complete diagnoses, and permit the greatest number of predictions between adjacent taxa. Values of Brothers' (1975) index of distinctness are perfectly described by a phylogenetic classification, but not by any other. There is no justification for either a phenetic or an evolutionary approach, since the phylogenetic system contains greater information than either on characters and similarities, while at the same time better reflecting phylogenetic kinship.