Polymorphic Characters in Phylogenetic Systematics

?The use of discrete, intraspecifkally variable characters in parsimony analysis is re? viewed. Seven data sets (two from morphology and five from allozymes) were analyzed to (1) compare different methods for treating polymorphic characters, (2) test for phylogenetic infor? mation in polymorphic characters, and (3) determine if there is a relationship between homoplasy and intraspecific variability. The performance of eight methods was compared using five criteria (number of characters treated as informative, number of shortest trees, phylogenetic signal, num? ber of nodes supported by bootstrapping, and sensitivity to reduced sample size). Approaches that incorporate explicit frequency information perform best overall for all the criteria, although the "majority'' method ties for best for the bootstrapping criterion. Levels of phylogenetic infor? mation in the polymorphic characters differed greatly among data sets and methods. Polymorphic characters in most data sets contained significant phylogenetic structure using most methods, but only one, the frequency method, extracted significant signal from the polymorphic characters in all seven data sets. Fixed characters appear to contain more signal than polymorphic characters, and homoplasy is significantly and positively correlated with intraspecific variability. This study supports the traditional view that polymorphic characters are less reliable in inferring phylogeny but does not necessarily support their exclusion. Systematists working with morphological data often do not report intraspecific variation, the frequencies of different traits, or how polymorphic characters are screened and analyzed; this situation should change. [Polymorphic characters; char? acter coding; character selection; character weighting; homoplasy; allozymes; morphology] Polymorphism, used herein as discrete variation within species that is indepen? dent of ontogenetic and sexual variation, is a ubiquitous problem in phylogenetic anal? ysis. If characters evolve, they must vary within species, at least at some point in their history. Yet, there is a long tradition in systematics of excluding characters in which polymorphism is observed. This practice is evident from the scarcity of phylogenetic studies that report intraspe? cific variation but is difficult to document explicitly because systematists seldom mention the characters that are excluded, rarely give explicit criteria for discarding nonfixed characters, and justify this exclu? sion even less frequently. Presumably, the basis for excluding polymorphic characters is that they are less reliable in phylogenetic inference (e.g., Darwin, 1859; Simpson, 1961; Farris, 1966; Kluge and Farris, 1969; Mayr, 1969) and have traditionally been difficult to deal with analytically (e.g., Wi1 Present address: Section of Amphibians and Rep? tiles, Carnegie Museum of Natural History, Pitts? burgh, Pennsylvania 15213-4080, USA. E-mail: wiens@clpgh.org. ley, 1981). Despite the general stigma as? sociated with polymorphic characters, a number of methods have been developed for their treatment (mostly in the context of analysis of allozyme data), and the rel? ative merits of these methods have been deba ed vigorously (e.g., Farris, 1981; Mickevich and Mitter, 1981, 1983; Buth, 1984; Swofford and Berlocher, 1987; Crother, 1990; Campbell and Frost, 1993; Mabee and Humphries, 1993; Murphy, 1993). Few studies have used analyses of real data sets to examine the properties of polymorphic characters and to compare the different methods for using them (Mickevich and Mitter, 1981; Campbell and Frost, 1993). In this paper, I review the problem of using discrete, intraspecifically variable characters in parsimony analysis. Specifi? cally, I address three questions: (1) What is the best method for phylogenetic analysis of polymorphic characters? (2) Do poly? morphic characters contain useful phylo? genetic information? (3) Is there a relation? ship between levels of intraspecific variability and homoplasy in systematic characters? These questions are addressed