Reconstructing the demographic history during the early stages of population divergence is central to understand fundamental processes in evolutionary biology, such as how populations adapt to novel environments and how reproductive isolation between incipient species arises. One popular approach to test evolutionary scenarios and infer demographic parameters is to obtain summary statistics from a set of populations, and compare them to simulated data in order to choose the most likely model and infer its parameters. As models are always extreme simplifications of a complex reality, it is unclear what the effects of ignoring such complexities are on model choice and parameter estimations. Using simulations, we demonstrate that under realistic scenarios of population divergence, failure to account for population size changes in a daughter population or the ancestral population leads to severe biases in both model choice and parameters estimations. Similarly, ignoring metapopulation structure can also severely affect demographic inference. The inference of divergence time and changes in patterns of gene flow through time, some of the parameters of most interest in the study of speciation, can be strongly biased when using oversimplified models. We illustrate these issues reconstructing the demographic history of North Sea and Baltic Sea turbots (Schopthalmus maximus) by testing 16 Isolation with Migration (IM) and 16 secondary contact (SC) scenarios, modelling changes in Ne as well as the effects of linked selection and heterogeneous migration rates across the genome. As in the simulated data, failure to account for changes in Ne resulted in selecting SC models with a long period of strict isolation and divergence times preceding the formation of the Baltic Sea. In contrast, models accounting for Ne changes suggest the Baltic Sea turbot population originated from a very recent (<6 kya) invasion and diverged with constant gene flow from the North Sea. The results have implications for the study of speciation, high-lighting the potential effects of ancestral size changes, bottlenecks, and metapopulation structure on choices between competing scenarios. In general, extreme caution should be exercised when interpreting results of demographic model comparisons.
Biases in demographic modelling affect our understanding of recent divergence