The misleading effects of composite taxa in supermatrices

With the amount of available sequence data rapidly increasing, supermatrices are at the forefront of systematic studies. As an alternative to supertrees, supermatrices utilize a total evidence approach where different genes and other lines of data are merged into a single data matrix, which is then analyzed in an attempt to obtain the phylogeny that best explains the data. However, questions may arise when combining data sets in which one or more taxa do not have sequences available for each individual gene. Two possible solutions to this situation are to either leave all taxa separate and code unavailable sequences as missing, or to combine taxa at a level for which monophyly is assumed a priori. By reanalyzing the previous work of Madsen et al. (2001), we show that combining taxa may yield misleading results, i.e., hypotheses of relationships that are not supported by the underlying data.