Accuracy of phylogenetic analysis including and excluding polymorphic characters

Intraspecific variation is ubiquitous in systematic characters, yet systematists often do not deal with polymorphism explicity. For example, morphological systematists typically exclude characters in which any or "too much" polymorphism is observed, and molecular systematists often avoid intraspecific variation by sampling a single individual per species. Recent empirical studies have suggested that polymorphic characters contain significant phylogenetic information but are more homoplastic than fixed characters. Given these two observations, should including polymorphic characters increase or decrease accuracy? We addressed this question using simulated data sets that also show a strong relationship between homoplasy and intraspecific variability. Data sets were generated with eight species, two alleles per locus, and a variety of branch lengths, number of loci, and sample sizes (individuals sampled per species). The data sets were analyzed using eight parsimony coding methods (with and without a priori and successive weighting) and different variability thresholds for excluding polymorphic characters. Excluding polymorphic characters decreased accuracy under almost all conditions examined, even when only the more variable characters were excluded. Sampling a single individual per species also consistently decreased accuracy. Thus, two common approaches for dealing with intraspecific variation in morphological and molecular systematics can give relatively poor estimates of phylogeny. In contrast, the unweighted frequency method, including polymorphic characters and sampling a reasonable number of individuals per species (n ≥ 5), can give accurate results under a variety of conditions.