The recognition that incomplete lineage sorting is a common and expected pattern in comparative DNA sequence data has revolutionized the theory and practice of phylogenetics. This recognition was fueled in part by the realization that replicated sampling within lineages is necessary to fully describe evolutionary relationships and in part by the further realization that different genomic regions are expected to yield different gene trees despite the same underlying divergence history. As such, most empirical studies, especially those interested in the relationships among closely related species, use multiple individuals per lineage in combination with multiple loci to make inferences about evolutionary histories. As these multilocus data sets increase coverage of the focal genomes, however, they are likely to include non-neutral regions. I examined the extent of this problem for inference of evolutionary relationships among 11 species of soft pines (Pinus subgenus Strobus) using a 167 nuclear gene data set. First, patterns of non-neutral evolution were examined using standard tests of neutrality based on the site-frequency spectrum and polymorphism-to-divergence ratios. Second, species-level phylogenies were inferred using standard approaches based on the reconciliation of gene and species trees. Finally, effects of non-neutral evolution on the inference of species trees were examined empirically and through simulation. The results to date are consistent with the idea that non-neutral evolution impacts inference of species-level phylogenies and that the recalcitrancy of this clade to phylogenetic resolution is due partially to signals of natural selection. Future phylogenetic studies, therefore, need to consider carefully the effect of increased genomic sampling.