Morgan's extravagance led Kroeber to reject sociological inter pretations almost in toto. In reply to criticism by Rivers, he has long since receded from this position, specifically admitting the correlation of certain terminological features with cross-cousin terminology. The issue was thus reduced to a difference in phi losophical attitude toward cultural phenomena. Rivers cham pioned a rigid determinism, contending that "not only has the Thurnwald, Die Gemeinde der Bdnaro, 1921, 6o sq. C. G. Selig man, The Melanesians of British New Guinea, 192o, 66, 481, 707.
general character of systems of relationship been strictly de termined by social conditions, but every detail of these systems has also been so determined." On the other hand, Kroeber insisted that "the infinitely variable play of the variable factors forbids any true determinations of causality of a sweeping character."' Kroeber was unquestionably right in rejecting Rivers' extreme claim, and in maintaining that since kinship terms were linguistic phenomena, they were subject to alterations of a linguistic char acter, e.g., extensions in meaning, independently of social con ditions. However, he did not sufficiently consider that language represents reality and that in so far as it related to social phe nomena it is likely to mirror them, even though imperfectly. The empirical facts of distribution indicate that the correlation between social custom and nomenclature is indeed far from perfect, but fairly high.
First of all, there is an undeniable tendency to designate by separate terms relatives with distinctive social functions. It is not sheer accident that Polynesians generally lack a term for the maternal uncle, but that a term appears with the avunculate. Furthermore, the same term applied to different individuals is an index of their sharing the same duties and claims, as Radcliffe Brown has pointed out for Australia.
Secondly, how are we to interpret the data of geographical dis tribution? Rivers found that neighboring Oceanians often differed more widely in their kinship systems than remote tribes. A survey of the world establishes this fact on a broader basis. Morgan's solution of resemblances in Dravidian and Iroquois nomenclature is impossible, but he put his finger on a real problem. When the Omaha of Nebraska resemble the Californian Miwok far more than their fellow-Siouans; when specific Omaha features crop up in some Assamese tribes and again on the upper Nile, the fact naturally arouses curiosity and calls for explanation. If the same
social factors uniformly accompany the terminological resem blance, a genuine correlation is established. Thus, the grouping of the mother's brother's son with the maternal uncle in the tribes cited is linked with paternal descent, while the classification of the father's sister's son with the father, and of the father's sis ter's daughter with the paternal aunt generally accompanies ma ternal descent. We are not dealing with a simple causal nexus, for there are matrilineal tribes like the Seneca and patrilineal tribes like the Ojibwa who do not override the generation principle in this fashion. But a functional relationship remains. It is con ceivable that the Omaha and Crow varieties of the bifurcate merg ing type depend on additional factors that may some time be discovered.
That, without prejudice to other functional relations, a high correlation obtains between a clan organization and the bifurcate merging system, seems certain. Clanless tribes like the Anda manese, Chukchee, Basin Shoshoneans, Hawaiians, have no termi nologies of this type ; characteristic clan organizations are almost uniformly coupled with them. Tylor's generalization, corroborated by Rivers and Lowie for Oceania and North America, respectively, seems to hold.' Yet it is possible to derive the terminology from the joint effect of the levirate and the sororate, institutions prob ably quite as widely distributed as the clan organization. True, they do not explain why in the common Iroquois variety of the type parallel cousins of the same sex are grouped together, while cross-cousins of the same sex are segregated under a common term of their own. Tylor's idea that the moiety represents the primeval clan does explain this; but large sections of the world with the bifurcate merging nomenclature—notably Africa—lack moieties.
In the present state of our knowledge, then, we may say that while social correlations of terminological features are undeniable, 'Rivers, Kinship and Social Organization, 1914; id., History of Mela nesian Society, II., II, passim,. Kroeber, Classificatory Systems of Relationship, Igo(); id., "California Kinship Systems," Univ. of Cal. Pub., vol. 12, no. 9, p.p. 385-396 (1917).