Properties of matrix representation with parsimony analyses.

Baum (1992) and Ragan (1992b) independently devised a method that uses additive binary coding and parsimony to combine trees derived from di€ erent data sets, a procedure Ragan termed matrix representation with parsimony analysis (MRP). Because the method utilizes the topology of source trees rather than the orig inal data, (1) trees derived from di€ erent data types (e.g., molecular sequences, morphological characters, pairwise distances) and analy zed by di€ erent clustering techniques (e.g., maximum parsimony, maximum likelihood, neighbor joining) can be combined, and (2) the source patterns are evaluated on a moreor-less equal basis, so that the phylogenetic signal from datamatrices with a smaller number of characters is not swamped by those with a larger number (see M iyamoto, 1985; Hillis, 1987). The method is also unusual in that (1) trees with di€ erent terminal taxa can be combined, a feature that among consensus methods characterizes only the supertree method (Gordon, 1986; Steel, 1992) and the modi® ed semistrict algorithm of Lanyon (1993), and (2) it is less sensitive to con ̄ ict among source trees than are most conventional consensus techniques so that resolution is not necessarily lost as increasing numbers of con ̄ icting trees are analyzed (see also Purvis, 1995b). Although the appropriateness of MRP to phylogenetic inference has been discussed (Baum and Ragan, 1993; Rodrigo , 1993, 1996; Bruneau et al., 1995) and modi® cations to the method have been proposed (e.g., Purvis, 1995b; Ronquist, 1996), its properties, mechanics, and biases have not been considered in su cient depth. We discuss some of the properties of MRP, show how MRP di€ ers from standard consensus techniques, and explore some modi® cations to the method. Although other clustering methods, such as compatibility (Ragan, 1992a; Purvis, 1995b; Rodrigo, 1996), have also been suggested as methods for generating composite trees by using matrix representation, we will not discuss them.