SOME RECENT PROGRESS WITH METHODS FOR EVOLUTIONARY TREES

Sequences of macromolecules have ''signals'' or patterns that arise fr om a number of sources, particularly from shared common history or phy logeny. We discuss methods for inferring evolutionary trees from these patterns or signals under five properties desired for an ideal method . These five desiderata are that the methods be efficient (fast), cons istent, powerful, robust, and falsifiable. Our conclusion is that corr ections for multiple changes in sequences are the most important facto r for any method to be consistent. Most optimality criteria, including compatibility and parsimony, become consistent when the sequences hav e appropriate corrections for multiple changes. Conversely, virtually no methods are consistent without adjustments for multiple changes. Ha damard conjugations are used to illustrate relationships between diffe rent methods and then illustrated by combining it with the closest tre e optimality criterion. The data used to illustrate these recent devel opments include DNA sequences used to study the origin of chloroplasts and also New Zealand skinks (Leiolopisma spp).