A common outcome of disruptive selection experiments between two differentiated populations which produce disadvantageous hybrids is an increase in homogamy. Experiments reported here result in another outcome when 'classical' selection experiments are redesigned. In these modified experiments, frequencies of genotypes in the mating population were not artificially maintained at parity but were instead determined from progeny proportions in the previous generation, in these selection lines another outcome, apart from an increase in homogamy, was demonstrated. Under a high selection coefficient against heterozygotes, elimination of a homozygote and the corresponding fixation of the other was observed. No selection line demonstrated the maintenance of two differentiated populations concurrently with the selection process of heterozygote disadvantage. A high number of generations of selection under this population genetical process is necessary to increase differences between two populations. However, the instability of gene frequencies which results in fixation or elimination of a homozygote is shown to be extremely rapid by comparison. Classical experiments were repeated and after 21 generations of selection there was no increase in divergence. For lower selection coefficients, high levels of introgression are apparent, and hence the genetical distinctness of the two populations decreases over time. This is in addition to the problem of an unstable equilibrium under selection against heterozygotes. Both aspects are important but not previously considered in experimental evidence for speciation models for which their implications are discussed.