The Determination of Parallel or Monophyletic Relationships: The Proteid Salamanders-A Test Case

Abstract

The basis of all comparative biology is the determination of phylogenetic relationships and lineages. The most difficult question in the determination of lineage is whether the similarity between two forms is the result of monophyly, parallel evolution, or convergent evolution. The erection of a new phylogeny or the testing of a previously proposed phylogenetic grouping involves several steps: (1) the relation of characters and their homologous character states-the morphocline; (2) determination within the morphocline of its primitive and derived states and their polarities; (3) evaluation of the character states by means of weighting analysis; and (4) construction of the new phylogeny. In this paper we have concentrated on the third step, and we have formalized a method for character state weighting. The criterion for character state importance is the information contained within the character and its states. Character states may be designated in increasing value according to the following groups: (I) loss character states which have the least amount of contained information because the complete absence of a character in two groups does not reveal whether the loss was monophyletic or polyphyletic; (II) characters demonstrating simplification by reduction or fusion; (III) characters whose degree of development is related to ontogenetic or allometric processes; (IV) characters belonging to functionally integrated character complexes; and (V) unique and innovative character states. Only the last two groups should be heavily weighted in phylogenetic estimation, since characters in the lower weight groups are more likely to have evolved in parallel. During the past 20 years the relationships of the salamander genera Necturus and Proteus have been interpreted as the result either of monophyly or of parallel evolution. Using the above weighting method, we have examined 18 previously studied characters which were used to define the family Proteidae. Of these 18 characters, we place none in weighting groups IV or V, 10 in group III, one in group II, and two in group I. Two characters are primitive for all salamanders, two require more data before they can be analyzed, and two are demonstrated as not being shared derived characters (for this family) under any criteria. The entire systematic debate on the reality of the family Proteidae rests on the evaluation of the karotype as a shared and derived character. We have evaluated the karotype as belonging to weighting group II, but others might consider it in group V. Other salamander families are characterized by shared and derived characters belonging to weighting groups IV and V. We have concluded that Proteus and Necturus are probably not derived from a common ancestor which was a perennibranchiate salamander. There are two remaining alternate hypotheses: (1) the ancestor was an unknown (probably extinct) metamorphosing terrestrial form which would have been in the same monophyletic family, and (2) the metamorphosing ancestors belong to two different families. Each of these hypotheses necessitates a different systematic treatment of the familial status of the two genera. However, at the present time, there is no morphological, paleontological, or biogeographical evidence to favor either of these hypotheses over the other. The analysis of the characters used to determine the relationships between these two genera also clearly demonstrates two important principles. First, similarities between two forms can easily be attained by similar ontogenetic or allometric patterns in which maximal morphological differences can be attained with minimal genetic change. Second, the search for parsimonious interpretations can be misleading if the phylogenetic process has not been economic or direct. The determination of relationships should not consider only the diagrammatic picture with the minimum number of convergences or parallelisms. Each morphocline must be analyzed independently and the concordances and discordances properly weighted and interpreted. The methodologies of cladism and pheneticism have been found lacking in the recognition of convergence and parallelism, but the so-called classical approaches also lack rigor and are too predisposed to apriorism