Design of control populations to be used in the simulation studies of assortative mating is discussed. In order to be useful for 25 generations a control population should satisfy the following requirements: (i) phenotypic correlation of mates,rP, close to zero; (ii) negligible amount of random drift; and (iii) large number of loci controlling the trait under consideration. It was concluded that under the assumptions of the design used in the present study the number of animals to produce a negligible amount of random drift (200 mating pairs) is much smaller than the number of animals necessary forrpto be sufficiently close to zero to exclude unconscious assortative mating (400 mating pairs). The minimum number of loci required allowing the trait to show a sufficiently large amount of genetic variation even in long‐term selection studies depends, among other things, on the selection intensity and the number of generations involved. In more intensely selected populations more loci are required. In long‐term selection experiments with 25 generations and 25% selected, the number of loci required, each with a very small equal effect, was 2500. An unselected control population fulfilling these requirements behaves similarly to the predictions of both the Hardy‐Weinberg model and the infinitesimal model.