||<p/>In chapter 2 a number of the mechanisms are discussed through which genetic polymorphisms can be maintained in natural populations: overdominance, frequency dependent selection and neutral alleles with associative overdominance. The overdominance model is emphasized because overdominance is also the basic feature of the associative overdominance model. Different theoretical relationships between number of heterozygous loci and fitness are explored, including their implications with regard to mean population fitness and selection coefficients at individual loci in an ideal population. From these, King's threshold model for multiple gene action on fitness proved to be the most satisfactory in all respects: it accomodates fairly high selection coefficients at individual loci without implying too heavy a load; it further explains different inbreeding depressions for different organisms and for different environments, as well as genotype by environment interaction. The model of associative overdominance, with the incorporation of King's threshold model for multiple gene action, has been chosen as an operational hypothesis for explanation of my experimental results (chapter 3) and as a basis for the simulation study (chapter 4).<p/>Chapter 2 further discusses the implications of associative overdominance (which is a result of overall linkage disequilibrium in finite populations) when linkage disequilibrium is generated artificially by using a small sample to found a new population. In this situation pseudo-frequency dependent selection is expected to occur at selectively neutral loci. An experimental design is proposed which distinguishes between apparent and real frequency dependent selection.<p/>Chapter 3 presents the experiments: individuals from two laboratory stocks of <em>Tribolium castaneum</em> HERBST, together with their F <sub><font size="-1">1</font></sub> , were used to initiate a set of polymorphic populations (for the black locus) with different frequencies of the marker allele. These experiments, jointly taken, indicate that the black locus itself is selectively neutral under the current experimental conditions and rule out the possibility of real frequency dependent selection. There was however apparent selection against the mutant allele due to initial linkage disequilibrium. This linkage disequilibrium is described in terms of the different genotypic backgrounds of the components (wild type and mutant stock and F <sub><font size="-1">1</font></sub> ) of the founder population: in the mutant stock there is an excess of homozygosity which may be randomly distributed over the chromosomes or may be partially or wholly concentrated in a chromosome region near the marker locus. This confirms the expectation formulated in the Introduction (chapter 1). The initial linkage disequilibrium, in these experiments is not so much due to small samples from the founder stocks as to the different genotypic backgrounds of the founder stocks, and, with respect to neutral loci, implies associative dominance rather than associative overdominance. The apparent decrease in selection against the b allele is a result of the approach to linkage equilibrium.<p/>A comparison of the fitness differences among the original marker genotypes (wild type, mutant black and F <sub><font size="-1">1</font></sub> ) on the one hand and the marker genotypes of an F <sub><font size="-1">2</font></sub> population on the other band, showed that the fitness loci closely linked to the marker locus and the joint non-linked fitness loci made approximately equal contributions to the fitness contrast between the two founder stocks (i.c. a lower fitness of the mutant stock). It also showed that, under the current experimental conditions, differential viability only played a minor role, if any, in the gene frequency changes of the b allele in the pooled populations.<p/>Chapter 4 presents a computer model for (stochastic) simulation of the population experiments. This model is based on the hypothesis of overdominance at the chromosomal level and on the assumption that only the marker chromosome contributes to the fitness difference between the founder stocks. For this purpose FRASER'S technique of binary representation of genotypes was adopted.<p/>After correction for some discrepancies between the simulation model and the experiments (in the simulation only the marker chromosome is considered and selection acts through differential viability), the results of simulation proved to be in fairly good agreement with the experimentally obtained results. The simulation model can readily be adapted to other situations, e.g. both founder stocks being 'inbred', tracking the gene frequencies at more than one neutral locus, and any arbitrary function relating the number of heterozygous loci to fitness.<p/>The final conclusion from both the experiments and the simulation study is, that after introducing the relatively 'inbred' mutant stock into the wild type population, a great deal of the genetic material of the mutant stock is lost by natural selection. For practical breeding this implies a risk of losing part of the genetic material, as a result of natural selection, from small samples of (relatively) inbred populations which are introduced into a breeding stock. Of course, the breeder may artificially select in favour of a fitness-neutral character introduced by the 'fresh' genetic material; however, the effect of artificial selection may be greatly reduced when (in the initial generations) natural selection outweighs artificial selection, since natural selection 'acts' against the desired character through linkage disequilibrium with fitness loci.