This thesis is a theoretical study of the genotypic configurations of isolated populations after selection, when more than one locus is considered. Computer simulations are used to confirm predictions. Two cases, in which a pair of interacting mutations, without dominance, under constant natural selection pressure, were studied. When the mutations were complementary, genetic diversity was found across populations under negative selection when the interaction was sufficiently large. When the mutations were antagonistic, genetic diversity was found across small populations under positive selection when the interaction was sufficiently large. Complementary pairs of mutations, arising in environments which vary over time, were found to have greater frequencies and fixation probabilities, and smaller fixation times, than those without interaction, providing the interaction was sufficiently large. Fixation probabilities were smaller when the interaction was small, and larger when the interaction was large, than when there was no interaction. This was because the selection effect was stronger, therefore more variable, when interaction was large, and weaker when small, than when there was no interaction. This study concluded that interaction between loci is advantageous in traits for which selection direction varies, if the population consists of isolated subgroups. Therefore, interactions between loci affecting such traits are not unlikely. Artificial selection programmes with similar objectives, applied to lines at different initial allele frequencies, were shown to have different long-term results when there was interaction between loci. Artificial selection programmes with similar two-trait objectives applied to lines at different initial allele frequencies, or applied with inaccurate estimates of current genetic parameters, were also shown to have different long-term results when there is pleiotropy. It was concluded that genetic diversity across wild or domesticated breeding populations is likely, and will be maintained if these populations are kept relatively isolated and breed in different environments. Selection on crosses between such lines may be highly advantageous for current, or future selection goals, but results are highly unpredictable.