Description
The recent renaissance in the development of multilocus coalescent-based species tree inference methods has transformed the study of systematics; however, coalescent-based methods require a priori knowledge of species limits. A variety of methods of multilocus species delimitation are now available which provide potentially objective approaches to assign individuals to putative species; however, these methods may require knowledge of the species tree. This dichotomy illustrates the necessity of studying species delimitation and species tree inference in concert. Here I demonstrate a method of simultaneous multilocus coalescent-based species delimitation and species tree inference that does not require prior assumption of species limits or the species tree. This method uses the Bayes factor to compare the fit of competing hypotheses of species delimitation to the data, and can be used to compare non-nested hypotheses. The multilocus Bayesian species tree is inferred under each competing hypothesis while the fit of the hypothesis to the data is quantified using marginal likelihood estimation. Marginal likelihood scores (as estimated using path sampling, stepping stone, and the smoothed harmonic mean estimator) are then compared using the Bayes factor. Here I apply this method to the Western Rattlesnake (Crotalus viridis) species complex, a group for which the species limits are contentious and the species tree is unknown. I collected DNA sequence data for six loci (five nuclear introns and one mitochondrial coding gene) and 63 ingroup individuals. Hypotheses of species limits were generated using (1) historical subspecific designations and (2) clades on a guide mitochondrial gene tree that were iteratively clustered into increasingly inclusive groupings. For each hypothesis, the species tree and marginal likelihood were estimated (under three competing marginal likelihood estimators) using *BEAST. Resulting marginal likelihood scores were compared to one another using the Bayes factor. BPP was also used to delimit species within the C. viridis complex for comparison. Contrary to currently recognized taxonomy, I recovered very strong support using both the Bayes factor method and BPP that the C. viridis complex contains six independently evolving species, including cryptic species within the Northern Pacific Rattlesnake (currently C. o. oreganus). I applied this resulting taxonomy to infer the first dated multilocus species tree of the C. viridis complex, which is topologically discordant from the mitochondrial gene tree. This study successfully demonstrated a novel method of Bayesian multilocus species delimitation. The results presented here warrant revision of the taxonomy within the C. viridis complex and dramatically revise our understanding of the evolutionary history of this group.