The effects of post-divergence gene flow on simultaneous divergence time (SDT) testing

Estimating simultaneous divergence across multiple taxonomic groups from genetic data is a common approach in phylogeographical studies, providing insight into the historical and evolutionary factors shaping species diversification. However, the robustness of divergence time estimates across multiple co-distributed species or population pairs in the face of post-divergence gene flow is often left unaddressed. Here, we use a simulation-based approach to test the robustness of estimating simultaneous divergence within a full-likelihood Bayesian framework when the model assumptions of no gene flow are violated. We generated simulated datasets of multiple population pairs with varying migration rates and estimate shared divergence times with the software package Ecoevolity, comparing them with estimates from population pairs experiencing no migration. Our goal was to identify the threshold at which migration rates begin to bias divergence time estimates, providing insight into the robustness of full-likelihood Bayesian methods under more complex demographic scenarios. We found that simultaneous divergence is incorrectly supported across a broad range of parameters space when assumptions are violated. We suggest that future empirical studies that use simultaneous divergence testing explicitly test for gene flow, or at the very least, explicitly consider the presence, absence, and implications of gene flow in their systems as part of their investigations. We assert that this result--biased support for simultaneous divergence--is just a specific example of a general behavior when the statistical assumptions of the model are not met or there is insufficient power given the data. We additionally recommend that future studies reorient their research framework to focus on trying to demonstrate non-simultaneous divergence in systems. As such, support for simultaneous divergence between groups should be interpreted as failure to resolve non-simultaneous divergence rather than evidence of true simultaneous divergence.