Figure 5: An illustration of various outcomes of disruptive selection and assortative mating, depending on parameter values. Each histogram represents the phenotype frequency distribution at a quasiequilibrium after 1,000 generations. Longer simulation runs do not qualitatively change the phenotype distribution for a given parameter combination. (a) With no disruptive selection and no assortative mating (, ), phenotypic variance follows a binomial (approximately normal) distribution whose variance depends on the level of polymorphism at the 40 additive loci, maintained by mutation-drift balance. (b) Using a combination of parameters close to the strongest disruptive selection and assortment observed in stickleback (, ), the population attains modestly higher phenotypic variance than in (a) but remains unimodal. (c) Larger parameter values (, ) rapidly lead to a weakly bimodal phenotype distribution that does not resolve itself into separate species even after 50,000 generations. (d) Speciation can occur when disruptive selection and assortment are near the upper end of plausible empirical values (, ).