Observed and simulated mismatch distributions. A) Mismatch distributions calculated for the total sample of IRS1 (black squares, n = 163) as well as for Model 1 (one spatial expansion) (open circles) and its 95% confidence interval (dotted lines). The Y axis stands for the average probability that two DNA sequences differ at a given number of sites represented on the X axis. B) Mismatch distribution obtained by simulation of Model 2 (two successive spatial expansions). The solid line shows the average mismatch distribution obtained from 1,000 simulations of the coalescent of 30 genes drawn in a single deme after two successive spatial expansions have occurred 10 and 30 mutational units of time ago. Dotted lines delimit an empirical 95% confidence interval for the mismatch distribution. C) Typical gene genealogies obtained for Model 2. Three main phases of coalescence (S, C1 and C2) are translated into the three modes of the mismatch distribution (see text).