We examine the third phase of Wright's shifting-balance theory of evolution, the exportation by migration of favorable gene combinations from a fitter subgroup to the rest of the population. The equations are deterministic and are studied numerically. Most of the models studied involve 2-9 loci in which all intermediates between two extreme genotypes are equally unfit. If the favored combination consists of dominant alleles, it is usually fixed even if the migration rate is two orders of magnitude less than the selection coefficient, and if the combination is recessive, one order. Although Wright thought of migration as being essentially one-way, two-way migration does not significantly alter the results. We conclude that, whatever weaknesses the Wright theory may have, they are not in phase III.