Samples of the polymorphic butterflyDanaus chrysippusare analysed from six well separated sites in East Africa. Morph-ratio clines are described for four diallelic genes A, B, C and L, each of which influences the visual phenotype. Each of the four clines has a different orientation, consistent with an hypothesis that the polymorphism originated from hybridization between a number of polytypic demes which have at various times undergone range expansion. Allopatric subspeciation in isolated Pleistocene refugia is postulated. The phenotype of each geographical race is shared with one of the morphs within the hybrid zone; other sympatrically maintained polymorphic forms are normally confined to the hybrid zone. Wright's isolation-by-distance model best explains the present distribution of gene frequencies. Morph-ratios differ significantly between the sexes and are sometimes associated with heterozygote excess; gametic and genotypic disequilibria are general throughout the region and suggest the clines are maintained by strong natural selection. Seasonal cycling of phenotype frequency is believed to result from extensive migratory movements rather than natural selection. Female-biased sex-ratio, which is also seasonal, and Haldane rule effects, result from hybrid breakdown when genetically distinct demes meet and interbreed. Oscillating sex-ratios and frequency of colour genes are functionally linked by negative feedback. The polymorphism owes its origin to allopatric evolution but is now maintained sympatrically.