A Comparison of Two Methods of Character Construction

Character construction is the partitioning of observed variation into discrete characters and character states.Although fundamental to hylogenetic inferences, there has been relatively little discussion of methods an B problems of character construction compared with discussion of the analysis of fully constructed data sets (Pogue and Mickevich, 1990).Some aspects of character construction may be relatively unproblematic because of the nature of the data.Thus, in aligned DNA sequences the character states can be clearly defined and, in morphology-based studies, simple presence-absence may delimit useful character states.However, morphological variation may involve character complexes or complex characters providing scope for alternative approaches to representing this variation as characters and their states.Wake (1993) recently presented some "non-traditional" neuroanatomical characters for caecilian amphibians, stressing that her study was preliminary and that she aimed to be both speculative and provocative.Subsequently, Wake (1994) discussed the general problems of "non-traditional" morphology, particularly how to sort discrete and independent characters from functionally complex units and how to best analyse them.Some of Wake's (1993) characters, particularly the first of her eye characters (El) and the first of her hypoglossal nerve characters (Hl), attracted my attention because, confronted with the same variation, I would have constructed very different characters.Here, I use these characters to help illustrate two distinct approaches to character construction and I compare the foundations and analytical consequences of each approach. Biological and Logical IndependenceCharacters used in phylogenetic inference may or may not be independent.Character independence is important in phylogenetic inference because characters are generally interpreted as providing independent evidence of relationships.If characters are not independent then the evidential significance of the underlying variation may be overweighted.In what follows I make use of a distinction between two kinds of independence, biological and logical, and the distinction as used here warrants some explanation.Two characters are not biologically independent if transformations in one are coupled to transformations in the other (e.g. because of pleiotropy or functional integration) so that they covary in their taxo-07483007/95/030297+12/