Combining trees and combining data

Abstract

The recent exchange between Rodrigo (1993) and Baum & Ragan (1993) highlights several current controversies. Rodrigo (1993) made several points concerning Baum's (1992) paper (and, by implication, Ragan, 1992a, b) on combining trees to reach solution concerning from differing datasets. I believe many of Rodrigo's comments are misplaced, both over how he perceives the problem and how he sees Baum & Ragan's (1993) proposed solution. The ideas expressed by both sets of authors impinge on the broader debate over the use of taxonomic congruence or character congruence in generating the most efficient summary of all available data to hand. The debate can be summarized as follows. Kluge (1989) made several assertions concerning what he believed to be conflicting approaches in phylogenetic inference when faced with data-sets from different sources (e.g. morphology, fossils, DNA, etc.). The apparent conflict is between character congruence (total e.g. Kluge, 1989) and taxonomic congruence (consensus, e.g. Nelson, 1979). In another paper, Jones & al. (1993: 93) defined total as the analysis of an unpartitioned body of ideally all the data available for group of terminal that is characterised as seek[ing] single, best-fitting hypothesis, which in cladistics involves maximising character congruence. The competing approach, taxonomic congruence, seeks consensus of obtained from different data sets .... This approach is characterized by three steps: (1) evidence is partitioned into 'different types of data' (Shaffer & al., 1991: 284), (2) hypotheses of relationships, i.e. fundamental cladograms, are obtained from each data set, and (3) a consensus of those topologies is constructed (Jones & al., 1993: 93). Following the characterization given above, it would appear that both approaches seek to maximize evidence, one derived directly from data derived from characters, the other from the relationships (cladograms) those data imply. The issue can be seen more precisely as identifying the nature of evidence. This could be answered (albeit simplistically) by stating that, in the total approach, data (direct observations into characters) are taken as while the taxonomic congruence approach uses as summaries of partitioned data (cladograms). The methods of Baum (1992) and Ragan (1992a, b; cf. Baum & Ragan, 1993) fall into the second category. Given two data-sets, each may suggest differing relationships among the taxa studied. Should the data-sets be combined into one analysis, or should the cladograms derived from each data-set be combined? Rodrigo suggested that, given two incompatible ... there are two ways of choosing between [them] ... (1): design critical experiment such that the experiment will