The relationship between homoplasy and confidence in a phylogenetic tree

Abstract

This chapter provides a statistical assessment of confidence in a phylogenetic tree. The analysis is done with the help of bootstrap procedure. A set of 101 phylogenetic studies was extracted from the TreeBASE data base. The data sets included 50 morphological, 29 restriction fragment length polymorphisms (RFLP), and 22 DNA sequence data sets, ranging in size from 5 to 68 taxa and from 10 to 2226 characters. All studies were on green plants. Each data set was subjected to parsimony analysis using PAUP 3.1. Each data set was then bootstrapped (100 replicates) using heuristic search options. Results reveal several interesting differences among the three different kinds of data. RFLP data sets have the highest average consistency indices (CIs) and retention indices (RIs), the largest number of most parsimonious trees, and the lowest average level of resolution in any given minimal tree. Morphological data sets have the lowest bootstrap support, but their RIs are higher than those for sequence data sets. It can be concluded that the conventional view that confidence is directly related to the level of homoplasy in a data set is not supported by available data from phylogenetic studies. The consistency index is a measure of homoplasy, not robustness. Homoplasy and confidence are two separate issues—not completely independent perhaps, especially in carefully controlled simulation experiments—but effectively so in real data sets.