We present the theoretical background to a new method for measuring genetic variation for total fitness in Drosophila. The method allows heterozygous effects on total fitness of whole wild-type chromosomes to be measured under normal demography with overlapping generations. The wild-type chromosomes are competed against two balancer chromosomes (B1, B2, say), providing a standard genotype B1/B2 against which variation in the fitness effects of the wild-type chromosomes can be assessed. Fitness can be assessed in two ways: (i) at equilibrium of all three chromosomes under heterozygote advantage, and (ii) during displacement of one balancer by the other. Equilibrium with all three chromosomes present will be achieved only if the wild-type homozygote is not too fit, and if the fitnesses of the three heterozygotes are not too unequal. These conditions were not satisfied for any of a sample of 12 lethal-bearing chromosomes isolated from a random-bred laboratory population of Drosophila. At equilibrium, genotypic frequencies show low sensitivity to changes in genotypic fitness. Furthermore, where all four genotypes are viable and fertile, supplementary information from cages with only two chromosomes present and from direct measurements of pre-adult viability are required to estimate fitnesses from frequencies. The invasion method has the advantages of a greater sensitivity and of not requiring further data to estimate fitnesses if the wild-type homozygote is fertile. However, it requires that multiple samples be taken as the invasion progresses. In a discrete generation model, generation time influences fitness estimates from this method and is difficult to estimate accurately from the data. A full age-structured model can also be applied to the data from both types of experiment. For the invasion method, this gives fitness estimates close to those from the discrete generation model.