MCMCTREE

To assess the timing of historical processes that may have led to diversification in <i>Ocymyrmex</i>, a dated phylogeny for the genus was inferred using MCMCTREE v4.9j in the PAML package (Phylogenetic Analysis by Maximum Likelihood; (Yang, 2007)). Divergence ages were reconstructed using the 80% taxon completeness matrix, trimmed with 0.97 cut-off in SPRUCEUP v2020.2.19 (Borowiec, 2019), and the tree resulting from partitioned analysis of this data set as input topology (Appendix C). This tree was pruned to include only one representative of each <i>Ocymyrmex</i> species, and two outgroup taxa (<i>Nesomyrmex</i> <i>madecassus</i> and <i>Cardiocondyla thoracica</i>) were also removed prior to this analysis. Given the unknown fossil record of <i>Ocymyrmex</i>, secondary ages estimated by Blaimer et al. (2018) were used to calibrate two nodes in our tree. The split between <i>Ocymyrmex</i> and <i>Cardiocondyla</i> was calibrated with 46.5–71.9 Ma and the split between <i>Tetramorium </i>and the <i>Cataulacus</i> genus-group of which both <i>Ocymyrmex</i> and <i>Cardiocondyla</i> are part of (= root node) was calibrated with 60.1–84.9 Ma. Both calibration ranges were derived from 95% highest posterior density (HPD) intervals across four analyses from Blaimer et al. (2018) to account for uncertainty.We first estimated the Hessian gradient in MCMCTREE (set usedata = 3), then the estimates from the Hessian gradient were used to perform approximate likelihood estimation (usedata = 2). We performed four independent runs, with the following parameters: burnin was set to 500,000; sampling was performed every 100 iterations until 1,000,000 samples were gathered; the model was set to 7 (GTR); an independent rates model was applied (clock=2); and we used a single data partition. Convergence and effective sample size (ESS) were visualized in TRACER v1.7 (Rambaut et al., 2018).