The rapid accumulation of multilocus data sets has led to dramatic advances in methodologies for estimating evolutionary relationships among closely related species, but relatively less advancement has been made in methods for discriminating between competing species delimitation hypotheses. Multilocus data sets provide an advantage in testing species delimitation scenarios because they offer a direct test of species monophyly and aid in the biological interpretation of such phenomena as allele-sharing and deep coalescent events. Most species tree estimation methods that are designed to analyze multilocus data sets require the a priori assignment of individuals to species categories and therefore do not provide a strategy to directly test competing species delimitation scenarios. An approach was recently proposed that utilizes a coalescent-based species tree estimation method to inform species delimitation decisions by comparing likelihood scores that measure the fit of gene trees within a given species tree. We use a multilocus nuclear and mitochondrial DNA sequence data set to both reexamine a recently proposed species delimitation scenario in the Etheostoma simoterum species complex and test the utility of species tree estimation methods in testing species delimitation hypotheses. Descriptions of species in the E. simoterum species complex of snubnose darters, a group of six teleost freshwater fish species, are based largely on male nuptial coloration. Most of the putative species are nonmonophyletic at every examined locus. Using a novel combination of Bayesian-estimated gene tree topologies, Bayesian phylogenetic species tree inferences, coalescent simulations, and examination of phenotypic variation, we assess the occurrence of shared alleles among species, and we propose that results from our analyses support a three-species rather than a six-species delimitation scenario in the E. simoterum complex. We found that comparing likelihood scores from the species tree estimation approach used across many potential delimitation scenarios resulted in a systematic bias toward over-splitting in the E. simoterum complex and failed to support a species delimitation scenario that was consistent with geography, phenotype, or any previous species delimitation hypothesis. Despite common expectations, we demonstrate that application of molecular approaches to species delimitation can result in the recognition of fewer, instead of a larger number of species. In addition, our analyses highlight the importance of phenotypic character information in providing an independent assessment of alternative species delimitation hypotheses in the E. simoterum species complex.