A reanalysis of Drosophila viability data was undertaken to determine the role of genotype-environment interactions in the maintenance of polymorphism. Between-replicate variances of viabilities in chromosomal homozygotes and heterozygotes with the same mean fitnesses were compared, with the expectation that if the heterozygote variance were on the average greater, conditional overdominance would be prevalent; if it were less, partial dominance would be prevalent; and if it were the same, marginal overdominance of the type considered by Wallace (1968) would be the prevalent type of variation. In fact, heterozygote variance was slightly less. The work of Dempster (1955) and of Gillespie and Langley (1974) is cited to show that this situation can still lead to balanced polymorphisms. Their general model for genetic variation in populations, consistent with the viability data, is reinforced.