Phylogenetic analysis: where the parsimony criterion and the Linnean systematization meet

Cladists claim that phylogenetic analysis using the parsimony criterion can find the (single) true phylogeny given that a sufficient number of properties are included in the analysis, but what does this methodology lead to…?  The claim appears both indisputable and as a circularity at the same time. Is it true, an indisputible circularity or not true? 

The methodology is to first arrange the properties of existing biological species into characters and character states, and then to analyze the minimum number of assumed changes between character states that can explain the present distribution of properties among species on every possible phylogenetic hypothesis, in order to choose the hypothesis with the least number of assumed changes. The methodology is expected to “zoom in” on the true phylogeny with inclusion of an increasing number of properties.

Evaluation of this methodology first requires an answer to the question whether there are any particular kinds of phylogenetic hypothesis that are preferred by the methodology, because if there are, the methodology is not neutral among the different kinds of phylogenies.  The fact that the most parsimonious choice of each single character is one (of those) that extends over both sides of the basic dichotomy, since a change over this dichotomy is counted as zero changes, these choices will be preferred over those that extends over each side of the basic dichotomy. This skewness means that the methodology tends towards choices that are symmetrical over the basic dichotomy from those that aren’t. The influence of this tendency will also be enforced by an increasing objectivity in the coding of properties, since it minimizes the number of character states in each character, thus ending up with only two character states in each character, which will be optimized over the two sides of the basic dichotomy. Similarly, these character states will be optimized over the two sides of the succeding dichotomies till the end of the phylogeny (that is, till today’s species). It will thus ultimately prefer  a totally symmetrical phylogeny. The methodology can thus not find the (single) true phylogeny if it is asymmetrical.

Now, it does not mean that the methodology is either irrational or impossible, although highly unlikely to find the (single) true tree, since the (single) true tree is highly unlikely to be symmetrical. However, the next question is whether the methodology can find the (single) true phylogeny if it indeed is symmetrical. The answer to this question can be found by considering the situation when all characters have only two character states, and when all character states in each character are are optimized both sides of the basic dichotomy. In this situation, the two character states that are positioned closest to the basic dichotomy can be positioned on each side of the basic dichotomy, and similarly the succeeding character states can be positioned on each side of the succeeding dichotomies till the end of the tree. It means that the methodology will end up in the symmetrical tree that optimizes the relation between the characters (instead of between character states). Since characters are our abstract comprehensions of the properties of species, it means that the methodology optimizes our abstract comprehensions of properties on a symmetrical tree.

Now, this optimization appears strange, but it is neither irrational nor impossible. The (single) true phylogeny may well be symmetrical in a way that optimizes the relation between the abstract characters. However, a change of the character states in the character closest to the basic dichotomy does not change the topology of the tree, which a change in either of the succeeding character states does neither, meaning that the final tree of choice is predestined by the partitioning of properties into characters and character states: it will be the symmetrical tree that optimizes our partitioning of properties into characters and character states.

Interestingly, this end point of phylogenetic analysis using the parsomony criterion is the same as a Linnean organization of biological species using the same partitioning of properties into characters and characters states. The analysis is thus, in practice, only a method to find the Linnean systematization that optimizes one’s own comprehension of the origin of properties (although the systematization may require more ranks than the original Linnean systematization does).

Phylogenetic analysis using the parsimony criterion does thus not find the true phylogeny, but the best Linnean systematization of the true phylogeny. It finds what it denies.