The Hierarchical Ordering of Characters as a Solution to the Dependent Character Problem in Numerical Taxonomy

Summary Where there are many taxonomically useful characters occurring in some but not all OTUs (“inapplicable characters”), the resultant “distortion” of character‐space may be troublesome in interpreting the similarity matrix. In numerical taxonomic work at and above the generic level considerable inapplicability of characters due to “missing organs” cannot be avoided. A solution may be to provide a hierarchical ordering of characters such that similarity is calculated over never‐missing “primary” characters whose scores are modified according to the similarities of their dependent characters (secondary, tertiary etc.). This, however, has the effect of “weighting” primary characters such that their presence/absence must always override the contribution of the secondary character scores. It is not evident that this is necessarily desirable, but an empirical examination of the method on three sets of data has produced reasonably acceptable results.