Analysis of alternative homologies with parsimony and step-matrix recoding.

This paper describes how prior uncertainty of the homology of different parts of organisms, and therefore of the characters contained in those parts, can be analysed by recoding the set of characters into a complex character. In this complex character, states represent combinations of states and homologies among the original characters, and transformation costs between states are set to values that maximize (weighted) homology. This allows a proper treatment of both the alternative homologies and the inapplicability of the characters contained in parts that can be absent or present. All the methods described are implemented in the program TNT, with a relatively simple syntax. This facilitates the treatment of characters in body parts of uncertain homology, including parts that can appear or disappear making their characters inapplicable. The analyses can also be combined with definitions of other types of dependences (e.g. morphofunctional constraints, inapplicabilities determined from characters outside of the parts of ambiguous homology). Although recoding into combinations of states and homologies allows handling relatively low numbers of characters per part (because of computation time and RAM requirements), TNT also implements options that facilitate creating alternative static alignments, allowing approximate analysis of larger number of characters.