In this regard the exogamy hypothesis in the broader sense enjoys an obvious superiority. Exogamous kin groups occur both in southern Africa and in many sections of America from which exogamous moieties have never been reported. Doubtless here, too, we must reckon to a considerable extent with the effect of diffusion, which repeatedly carried the Dakota principle to non-exogamous tribes. Yet when we apply the method of variation to the best-studied regions of the globe, our confidence in the essential correctness of the exogamy hypothesis is considerably strengthened. In Oceania it is the non-exogamous Polynesians who fail to distinguish the maternal and paternal sides, while the generally exogamous Melanesians recognize the principle of bifurcation. In North America, the non-exogamous tribes are either bifurcating but fail to merge the collateral and lineal lines or neither bifurcate nor merge.[42] Certain instances are especially illuminating because they permit a refinement of the method of variation by the practical or total elimination of other factors to account for the phenomena. Thus on the northwest coast of America we find certain tribes like the Kwakiutl and Nootka who are not organized in strictly exogamous groups, and here neither merging nor bifurcation occurs. “The terms for ‘uncle’ and ‘aunt’ refer equally to the father’s and mother’s fraternity;” and specific terms distinguish father and mother from more remote kindred. When we compare such systems with those of the more northern and exogamous tribes, _viz._, the Tsimshian, Haida and Tlingit, we discover at once a striking difference. In all these terminologies men of the father’s are distinguished from those of the mother’s moiety or clan; and the collateral lines are wholly, or almost entirely, merged in the lineal lines.[43] Here we are not dealing simply with a contact phenomenon, for no good reason can be given why the Tlingit system should not have extended southward or the Kwakiutl system to the north. Nor are we simply confronted by a difference of tribal affiliation: while the Kwakiutl and Nootka belong to the same stock, and affinity has recently been claimed for the Tlingit and Haida languages, the Tsimshian stand apart. It is the difference in social organization that runs parallel with the difference in nomenclature.