Gaps as characters in sequence-based phylogenetic analyses.

In the analysis of sequence-based data matrices, the use of different methods of treating gaps has been demonstrated to influence the resulting phylogenetic hypotheses (e.g., Eernisse and Kluge, 1993; Vogler and DeSalle, 1994; Simons and May den, 1997). Despite this influence, a well-justified, uniformly applied method of treating gaps is lacking in sequence-based phylogenetic studies. Treatment of gaps varies widely from secondarily mapping gaps onto the tree inferred from base characters to treating all gaps as separate characters or character states (Gonzalez, 1996). This diversity of approaches demonstrates the need for a comprehensive discussion of indel (insertion or deletion) coding and a robust method with which to incorporate gap characters into tree searches. We use the term "indel coding" instead of "gap coding" because the term "gap coding" has already been applied to coding quantitative characters (Mickevich and Johnson, 1976; Archie, 1985). Although "indel coding" undesirably refers to processes that are not observed (insertions and deletions) instead of patterns that are observed (gaps), the term is unambiguous and does not co-opt established terminology. The purpose of this paper is to discuss the implications of each of the methods of treating gaps in phylogenetic analyses, to allow workers to make informed choices among them. We suggest that gaps should be coded as characters in phylogenetic analyses, and we propose two indel-coding methods. We discuss four main points: (1) the logical independence of alignment and tree search; (2) why gaps are properly coded as characters; (3) how gaps should be coded as characters; and (4) problems with a priori weighting of gap characters during tree search. LOGICAL INDEPENDENCE