ABSTRACT
Multistate characters were ordered whenever possible, following the basic principles presented in Kitching et al. (1998). The same phylogenetic weight and hence the same phylogenetic importance was given to all characters used in the analysis, following Wiens (2001). In fact, as pointed out by him, most phylogenetic studies neglected the fact that by using ordered and unordered characters without corrective weighting, a different weight and thus different phylogenetic importance is, in reality, given to the different characters. For example, if species A and B share state 2 of a certain ordered character X and if species A and C share state 2 of a certain unordered character Y, this will favour a grouping of A and B, since they share two evolutionary steps on the cladogram (from state 0 to state 1 of character X, and then from state 1 to state 2 of this character, since this is an ordered character), while A and C share only one evolutionary step on the cladogram (directly from state 0 to
state 2 of character Y, since this is an unordered character). Thus, a different weight and hence a different phylogenetic importance is, in reality, attributed to the characters X and Y since, after all, A and C share, like A and B, state 2 of a phylogenetic character. Therefore, in this simple example, in order to render the phylogenetic importance of the character Y equal to that of the character X, the phylogenetic weight of the character X should be multiplied by y2, that is, by l/(n -l), where n is the total number of states of the character. In this way, the sharing of state 2 of the unordered character Y by A and C is ranked as equally important as the sharing of state 2 of the ordered character X: both characters have a total weight of 1. Consequently, in order to attribute the same phylogenetic importance to all characters used in the present analysis, i.e., in order to give an equal total weight of 1 to all these characters, the following procedure was undertaken: a) the weight of unordered characters, as well as of ordered characters with two states, was not changed, i.e., it remained 1; b) the weight of ordered characters with three, four and five states (there are no ordered characters with more than five states in the analysis) was multiplied, as described above, by a corrective factor of respectively, y2, and % • However, it is important to note that the phylogenetic outcomes resulting from the two other alternative methodologies concerning the ordering/unordering of characters, i.e., the exclusive use of unordered characters and the use of ordered characters without normalisation of their total weights, will be carefully examined and discussed in Chapter 5.
