Making inferences about speciation using sophisticated statistical genetics methods: look before you leap.

Understanding speciation is a fundamental aim of evolutionary biology and a very challenging one. Speciation can be viewed as the dynamics of genetic differentiation between populations resulting in substantial reproductive isolation (Gavrilets 2003). It was generally accepted that very small levels of migration prevent genetic differentiation among populations and, therefore, speciation. However, recent theoretical work showed that sympatric speciation is possible (Gavrilets 2003). Nevertheless, providing empirical evidence that gene flow occurred during speciation is challenging because several gene flow scenarios can explain observed patterns of genetic differentiation. Positive migration rate estimates alone do not prove ongoing gene flow during divergence. We also need to know whether migration took place before, during or after speciation. There is no statistical method specifically aimed at estimating gene flow timing, but several studies used the isolation with migration model (Hey & Nielsen 2004, 2007; Hey 2010b) to estimate this parameter and make inferences about speciation scenarios. It is tempting to use statistical methods to estimate important evolutionary parameters even if they do not appear explicitly in the inference model. Nevertheless, before doing so, we need to determine whether they can provide reliable results. In this issue of Molecular Ecology, Strasburg and Rieseberg (2011) present a simulation study that examines the degree to which gene flow timing estimates obtained from IMa2 (Hey 2010b) can be used to make inferences about speciation mode. Their results are sobering; gene flow timing estimates obtained in this way are not reliable and cannot be used to unequivocally establish if gene flow was ongoing during speciation.