Rampant Parallelism: An Appraisal of the Use of Nonuniversal Derived Character States in Phylogenetic Reconstruction

Abstract

There is an essential difference between convergence and parallelism, the former being due to analogous similarity, while the latter is based on a homologous relationship between characters. Careful application of the recognition criteria of homology and analogy may enable us to distinguish between convergence and true parallelism. Under the assumptions that one works in a genealogical domain wherein parallelism is rampant, and that parallel characters are homologues, derived parallelisms may successfully be used as indicators of propinquity of descent. Therefore, the principle of parsimony is considered a poor selection criterion in situations where parallel evolution is rampant. Character distributions concerning repeatedly evolved homologous structures have been described with a varying terminology: apomorphic tendencies, unique inside parallelism, underlying synapomorphy, nonuniversal derived character states. The paper illustrates the practical use and potential significance of such character state distributions for phylogenetic analysis by detailing several examples from marine triclads. These examples are grouped into cases with outside parallelism and with inside parallelism. In a case of outside parallelism the character state is also present in a more distantly related outgroup, while in cases of inside parallelism the parallelisms are restricted to the ingroup. A theoretical evaluation is given of the use of underlying synapomorphies in phylogenetic reconstruction, and in particular attention is paid to alternative explanations of the character state patterns, viz. convergence, secondary change, and retention of the primitive character state. The use of underlying synapomorphies is “Hennigian” in the most literal sense because Hennig himself used repeated derivations as indicators of genealogical relationship. Some of the practical rules for the application of underlying synapomorphy in phylogenetic inference are still fuzzy. Worthy of future theoretical discussions are questions on the frequency of the character states, cases of outside parallelism, and on falsification of presumed underlying synapomorphies. Repeated evolutionary derivation of structures may result from polyallelism, or from the retention of structural gene complexes and subsequent re-expression due to a change in regulatory mechanisms. For practical purposes, a fruitful synthesis may be achieved by combining a computerized parsimony analysis with the qualitative phylogenetic technique.