Despite the importance of competition as an evolutionary determinant in natural populations there have been few studies of the genetical control of competitive ability. Here, we report the results of a biometrical analysis of four continuously varying traits which, between them, describe the competitive interactions in mixed cultures of Drosophila melanogas fee. The analysis involved the parental, F1, F2 and backcross generations (including all reciprocals) derived from crosses between two highly inbred lines isolated from the Texas population of D. melanogasfer. The competitive performance of each genotype in monoculture and in duoculture with a phenotypically distinct tester were assessed using a yield-density regression analysis. Appropriate genetic models were fitted using a variance weighted least squares procedure and the resulting genetic components of the generation means used to define the genetical architecture of competition. Of the four competitive parameters investigated here the e-value, which describes the competitive performance of the indicator genotype at a fixed reference density, was found to be determined by simple additive genetic effects with no evidence of significant dominance. Conversely, competitive performance in monoculture (intra-genotypic competition) did display a significant net dominance component and the observed values in the F1 and parental generations indicated some degree of heterosis. Of the two competitive parameters determining performance in duoculture (inter-genotypic sensitivity and inter-genotypic pressure) the former was found to have a complex genetic determination involving not only additive and dominance components of the progeny's own genotype but also dominance components of the F1 maternal genotypes. There were also additive-dominance and dominance-dominance non-ailelic interactions. Heterosis was evident, determined both by the progeny's own genotype and by one of the F1 maternal genotypes. All dominance and heterosis was directed towards reduced inter-genotypic sensitivity or, iii other words, superior competitive ability. The analysis of maternal effect components for inter-genotypic competitive pressure could not be accommodated for reasons described in the text, although the data provided evidence for their involvement. The fitting of a simplified model revealed significant additive and dominance components of similar magnitude together with heterosis determined by the progeny's own genotype. There was no evidence of non-allelic interaction. As before all dominance and heterosis was directed towards superior competitive ability (i.e., increased inter-genotypic pressure). Throughout the experiment, there was no evidence for sex-linkage in the determination of competitive parameters. This is thought to be a prerequisite for stability of the sex ratio in the intense competitive environment of natural populations. Possible interpretations of the genetical architecture of competition are discussed in the light of these results.