Richard T. O'Grady

Coding Multistate Characters, with Special Reference to the Use of Parasites as Characters of their Hosts

-This paper examines four methods of representing multistate transformation series with a numerical code. The three that have the broadest applications are Additive Binary Coding, Redundant Linear Coding, and Nonredundant Linear Coding. The fourth method, internal rooting, has specialized applications for coding basally dichotomous series. The first three methods are capable of unambiguously representing multistate trees with or without specified nodal values. The Nonredundant Linear Coding method is recommended because it produces the most efficient code. The Redundant Linear Coding method can produce unjustified weighting of areas of a multistate tree. We examine coding procedures for working with multistate transformation series of intrinsic organic characters, and with parasite phylogenies as multistate characters of their hosts. Use of the inclusive ORing technique to code for the occurrence of more than one parasite taxon per host taxon can result in the distortion of phylogenetic information and the drawing of incorrect inferences of relationships. [Multistate coding methods; host-parasite coevolution; inclusive ORing.] This paper is a technical note with three parts. The first demonstrates four multistate character coding methods, the second compares their properties, and the third examines some of their uses in phylogenetic studies. This last part is primarily concerned with the use of parasite data to infer host relationships. For this paper, we consider a multistate character to be any set of more than two organic or inorganic states that have, through some process, transformed from one into another. The order of transformation of these states will describe a multistate tree. This definition is broad enough to include phenomena such as organic character evolution, changes in ecological properties, host-parasite coevolution, and biogeographic

A measure of the information content of phylogenetic trees, and its use as an optimality criterion

The D measure is presented as a quantification of historical constraints in phylo- genetic data. It is derived from information-theoretic considerations, and offers increased res- olution of the criteria used in parsimony analyses. Thus, it can be used as an optimality measure in phylogenetic studies. The D measure is more sensitive to the evolutionary nature (symple- siomorphy, synapomorphy, autapomorphy) of characters on trees than two commonly used phylogenetic optimality measures: the consistency index; and the F-ratio. The consistency index does not distinguish between shared and unique characters on a tree. The F-ratio is affected by factors relevant to undirected, rather than directed, trees. Neither the D measure nor the F-ratio are biased a priori towards choosing postulates of reversal or parallelism in equal-length trees. Neither the D measure nor the F-ratio always indicate the shortest tree when used alone. The D measure can be used to choose among equal-length shortest trees that have been produced by standard parsimony techniques. (Systematics; phylogenetic trees; information theory; opti- mality criteria.) This paper has two purposes. First, we describe a measure, termed the D measure, that quantifies the amount of historical constraint on character evolution postu- lated by a phylogenetic hypothesis. Sec- ond, we suggest that the maximal value of this measure can be used as an optimality criterion to choose among trees when ho- moplasy in the data produces multiple