Simulation scenarios and models used for demographic inference. Times are given in units of $4NREF$ generations. (A) We simulated data under an IM scenario with divergence times from 0.05 to 1.6. In IM$B$ scenarios one population experiences a contraction to 1–64% of its previous size at time T_B followed by exponential growth. The IM$AB$ scenario reflects an ancestral population size contraction followed by expansion. In the IM$AE$ scenario, the ancestral population undergoes an expansion. For both AB and AE scenarios, we tested a range of times (0.25, 0.5, 1) and strengths of the ancestral contraction (AB = 1/4, 1/16, 1/64 of previous size) and expansion (AE = 4×,16×,64× the ancestral N_e). (B) Simulations of recent divergence scenarios, with a T_S of 0.05 for IM scenarios and 0.1 for SC scenarios. Data were simulated under the IM and SC scenarios with 64 possible combinations of ancestral expansions and bottlenecks. Simulations were ran both with symmetric and asymmetric migration, for a total of 256 demographic scenarios. (C) demographic models used for inference with moments and dadi, that is, the basic IM and SC models as well as modifications that include an ancestral population size change as well as a bottleneck followed by growth in one of the daughter populations. Migration rate is asymmetric (i.e., different m₁₂ and m₂₁ parameters) in all inference models. For the analyses of the empirical data, we used the models graphically represented in panel C, and modifications of these models accounting for heterogeneity in migration rates (2M) and N_e (2N models) across the genome, as well as modifications accounting for both (2N2M models).