An earlier study (Stephens and Phillips 1972) showed that primitive cultivars of Gossypium barbadense L., native to the Americas, could be subdivided into two geographical groups: one west of the Andes which is polymorphic for the corky alleles, cky and ck0, and the other, east of the Andes, which approaches monomorphism for cky, All races of G. hirsutum L., except race maarie-galante, are monomorphic for ck0. In the present study an almost identical grouping is found to hold for the presence or absence of fringe hairs on the floral nectary. The western group of G. barbadense is polymorphic for fringe hairs and the eastern group monomorphic for absence of fringe hair. Virtually all forms of G. hirsutum are monomorphic for presence of fringe hairs. The trivial and inconspicuo,us nature of the fringe hairs makes it difficult to imagine that they have any strong selective value. The corky alleles have no apparent phenotypic expression except in the rare cases where mixed plantings of G. barbadense and G. hirsutum race marie-galante have been established and followed by interspecific hybridization. Both characters, therefore, would seem to provide useful indicators of racial phylogenies. Applying this concept to primitive forms of G. barbadaense now established in Africa, India, and Polynesia, it is found that most of them belong to the eastern group with respect to fringe hairs and corky alleles. This finding is consistent with the commonly held but unsubstantiated opinion that they were introduced over the old Portuguese trade routes from Brazil and the West Indies and not via the old Spanish trade route (Mexico, Guam, Philippines). The origin of modern annual forms of G. barbadense is obscure. The origins of the Sea Island cottons can be traced back only to mid-18th century in the seaboard colonies of North America, and the Egyptian cottons to the early 19th century in the Nile Delta. The Sea Island cottons have the eastern pattern of fringe hairs and corky alleles, contrasting sharply with the Egyptian cottons which have the western pattern. The possible implications of these findings are discussed.