144 W. H. Ft. RIVERS systems I have considered possesses a single feature which is not compatible with social conditions arising out of this marriage. Apart from quantitative verification, I doubt whether it would be possible in the whole range of science to find a case where we can be more confident that one phenomenon has been conditioned by another. I feel almost guilty of wasting your time by going into it so fully, and should hardly have ventured to do so if this case of social causation had not been explicitly denied by one with so high a reputation as Professor Kroeber. I hope, however, that the argument will be useful as an example of the method I shall apply to other cases in which the evidence is less conclusive. The features of terminology which follow from the cross-cousin marriage were known to Morgan, being present in three of the systems he recorded from Southern India and in the Fijian system collected for him by Mr Fison. The earliest reference [Grant 1870: 276] to the cross-cousin marriage which I have been able to discover is among the Gond of Central India. This marriage was recorded in 1870, which, though earlier than the appearance of Morgan's book, was after it had been accepted for publication, so that I think we can be confident that Morgan was unacquainted with the form of marriage which would have explained the peculiar features of the Indian and Fijian systems. It is evident, however, that Morgan was so absorbed in his demonstration of the similarity of these systems to those of America that he paid but little, if any, attention to their peculiarities. He thus lost a great opportunity; if he had attended to these peculiarities and had seen their meaning, he might have predicted a form of marriage which would soon afterwards have been independently discovered. Such an example of successful prediction would have forced the social significance of the těr ology of relationship upon the attention oi <,tn" dents in such a way that we should have he spared much of the controversy which has long obstructed progress in this branch of Cx\ ology. It must at the very least have acted ^ " stimulus to the collection of systems of're[a tionship. It would hardly have been pos that now, more than forty years after thi pearance of Morgan's book, we are sti complete ignorance of the terminology of tela tionship of many peoples about whom volumes' have been written. It would seem impossible for instance, that our knowledge of Indian systems of relationship could have been what " it is today. India would have been the country in which the success of Morgan's prediction "' would first have shown itself, and such an ' event must have prevented the almost total -neglect which the subject of relationship h,n suffered at the hands of students of lndun sociology. REFERENCES Grant, C, 1870. Gazetteer of Central Provinces,' 2nd ed., Nagpur. Kohler, J., 1897. Zur Urgeschichte der Ehe. Z,\t-schrift für vergleichende Rechtswissenschaft, 12, pp. 187-353. Kroeber, A. L.> 1909. Classificatory Systems of Relationship. Journal of the Royal Anthropological Institute, 39, pp. 77-84. McLennan, J. F., 1876. Studies in Ancient His-' tory, London. Morgan, L. H., 1871. Systems of Consanguinity and Affinity of the Human Family. Smithsonian Contributions to Knowledge, 17, Washington. Rivers, W. H. R., 1914. The History of Melanes-ian Society, 2 vols. Cambridge. 8 Structural Analysis in Linguistics and in Anthropology Claude Lévi-Strauss Linguistics occupies a special place among the social sciences, to whose ranks it unquestionably belongs. It is not merely a social science like the others, but, rather, the one in which by far the greatest progress has been made. It is probably the only one which can truly claim to be a science and which has achieved both the formulation of an empirical method and an understanding of the nature of the data submitted to its analysis. This privileged position carries with it several obligations. The linguist will often find scientists from related but different disciplines drawing inspiration from his example and trying to follow his lead. Noblesse oblige. A linguistic journal like Word cannot confine itself to the illustration of strictly linguistic theories and points of view. It must also welcome psychologists, sociologists, and anthropologists eager to learn from modern linguistics the road which leads to the empirical ..knowledge of social phenomena. As Marcel Mauss wrote already forty years ago: "Sociology would certainly have progressed much further if it had everywhere followed the lead of the linguists____The close methodological analogy which exists between the two disciplines imposes a special obligation of collaboration upon them. F-ver since the work of Schräder2 it has been -unnecessary to demonstrate the assistance which linguistics can render to the anthropologist in the study of kinship. It was a linguist and a philologist (Schräder and Rose}3 who showed the improbability of the hypothesis of matrilin-eal survivals in the family in antiquity, to which so many anthropologists still clung at that time. The linguist provides the anthropologist with etymologies which permit him to establish between certain kinship terms relationships that were not immediately apparent. The anthropologist, on the other hand, can bring to the attention of the linguist customs, prescriptions, and prohibitions that help him to understand the persistence of certain features of language or the instability of terms or groups of terms. At a meeting of the Linguistic Circle of New York, Julien Bonfante once illustrated this point of view by reviewing the etymology of the word for uncle in several Romance languages. The Greek 0sioc, corresponds in Italian, Spanish, and Portuguese to zio and tio; and he added that in certain regions of Italy the uncle is called barba. The "beard," the "divine" uncle-what a wealth of suggestions for the anthropologist! The investigations of the late A. M. Hocart into the religious character of the avuncular relationship and the "theft of the sacrifice" by the maternal kinsmen immediately come to mind.4 Whatever interpretation is given to the data collected by Hocart (and his own interpretation is not entirely satisfactory), there is no doubt that the linguist contributes to the solution of the problem by revealing the tenacious survival in contemporary vocabulary of relationships which have long since disappeared. At the same time, the anthropologist 146 CLAUDE LEVI-STRAUSS STRUCTURAL ANALYSIS IN LINGUISTICS AND IN ANTHROPOLOGY 147 explains to the linguist the bases of etymology and confirms its validity. Paul K. Benedict, in examining, as a linguist, the kinship systems of Southeast Asia, was able to make an important contribution to the anthropology of the family in that area.3 But linguists and anthropologists follow their own paths independently. They halt, no doubt, from time to time to communicate to one another certain of their findings; these findings, however, derive from different operations, and no effort is made to enable one group to benefit from the technical and methodological advances of the other. This attitude might have been justified in the era when linguistic research leaned most heavily on historical analysis. In relation to the anthropological research conducted during the same period, the difference was one of degree rather than of kind. The linguists employed a more rigorous method, and their findings were established on more solid grounds; the sociologists could follow their example in "renouncing consideration of the spatial distribution of contemporary types as a basis for their classifications."6 But, after all, anthropology and sociology were looking to linguistics only for insights; nothing foretold a revelation.' The advent of structural linguistics completely changed this situation. Not only did it renew linguistic perspectives; a transformation of this magnitude is not limited to a single discipline. Structural linguistics will certainly play the same renovating role with respect to the social sciences that nuclear physics, for example, has played for the physical sciences. In what does this revolution consist, as we try to assess its broadest implications? N. Troubetzkoy, the illustrious founder of structural linguistics, himself furnished the answer to this question. In one programmatic statement,8 he reduced the structural method to four basic operations. First, structural linguistics shifts from the study of conscious linguistic phenomena to study of their unconscious infrastructure; second, it does not treat terms as independent entities, taking instead as its basis of analysis the relations between terms; third, it introduces the concept of system - "Modern phonemics does not merely proclaim that phonemes are always part of a system; it shows concrete phonemic systems and elucidates their structure"9 -; finally, struc- tural linguistics aims at discovering getieraf laws, either by induction "or... by logical r in addition to,.the preceding ones, those attitudes . i; which are stylized, prescribed, and sanctioned: , tat>oos or privileges and expressed through a fec&d ritual. These attitudes, far from automat -"cally reflecting the nomenclature, often appear ■secondary elaborations, which serve to re-o|ve me contradictions and overcome the deficiencies inherent in the terminological system, ■fbis synthetic character is strikingly apparent .-■Bong the Wik Munkan of Australia. In this gtotip, joking privileges sanction a contradiction:, between the kinship relations which link two unmarried men and the theoretical relationship which must be assumed to exist between them in order to account for their later inarriages to two women who do not stand themselves in the corresponding relationship.22 'There is a contradiction between two possible svstems of nomenclature, and the emphasis ■placed on attitudes represents an attempt to integrate or transcend this contradiction. We can easily agree with Radcliffe-Brown and assert the existence of "real relations of interdependence between the terminology and the rest of the system."23 Some of his critics made the niistake of inferring, from the absence of a rigorous parallelism between attitudes and nomenclature, that the two systems were mutually independent. But this relationship of interdependence does not imply a one-to-one correlation. The system of attitudes constitutes, rather, a dynamic integration of the system of ■tcrfninology. Granted the hypothesis (to which we wholeheartedly subscribe) of a functional relationship between the two systems, we are nevertheless entitled, for methodological reasons, to treat independently the problems pertaining to each system. This is what we propose to do here for a problem which is rightly considered thepoint of departure for any theory of attitudes - that of the maternal uncle. We shall attempt to show how a formal transposition ofithe method of structural linguistics allows us: to shed new light upon this problem. Because the relationship between nephew and maternal uncle appears to have been the focus of significant elaboration in a great many primitive societies, anthropologists have devoted special attention to it. It is not enough t0 'K'tc the frequency of this theme; we must also account for it. tet us briefly review the principal stages in the development of this problem. During the entire nineteenth century and until the writings of Sydney Hartland,24 the importance of the mother's brother was interpreted as a survival of matrilineal descent. This interpretation was based purely on speculation, and, indeed, it was highly improbable in the light of European examples. Furthermore, Rivers' attempt25 to explain the importance of the mother's brother in southern India as a residue of cross-cousin marriage led to particularly deplorable results. Rivers himself was forced to recognize that this interpretation could not account for all aspects of the problem. He resigned himself to the hypothesis that several heterogeneous customs which have since disappeared (cross-cousin marriage being only one of them} were needed to explain the existence of a single institution.26 Thus, atomism and mechanism triumphed. It was Lowie's crucial article on the matrilineal complex27 which opened what we should like to call the "modern phase" of the problem of the avunculate. Lowie showed that the correlation drawn or postulated between the prominent position of the maternal uncle and matrilineal descent cannot withstand rigorous analysis. In fact, the avunculate is found associated with patrilineal, as well as matrilineal, descent. The role of the maternal uncle cannot be explained as either a consequence or a survival of matrilineal kinship; it is only a specific application "of a very general tendency to associate definite social relations with definite forms of kinship regardless of maternal or paternal side." In accordance with this principle, introduced for the first time by Lowie in 1919, there exists a general tendency to qualify attitudes, which constitutes the only empirical foundation for a theory of kinship systems. But, at the same time, Lowie left certain questions unanswered. What exactly do we call an avunculate? Do we not merge different customs and attitudes under this single term?-And, if it is true that there is a tendency to qualify all attitudes, why are only certain attitudes associated with the avuncular relationship, rather than just any possible attitudes, depending upon the group considered? A few further remarks here may underline the striking analogy between the development of this problem and certain stages in the evolution of linguistic theory. The variety of possible 150 CLAUDE LEVI-STRAUSS attitudes in the area of interpersonal relationships is almost unlimited; the same holds true for the variety of sounds which can be articulated by the vocal apparatus - and which are actually produced during the first months of human life. Each language, however, retains only a very small number among all the possible sounds, and in this respect linguistics raises two questions: Why are certain sounds selected? What relationships exist between one or several of the sounds chosen and all the others?28 Our sketch of the historical development of the avuncular problem is at precisely the same stage. Like language, the social group has a great wealth of psycho-physiological material at its disposal. Like language too, it retains only certain elements, at least some of which remain the same throughout the most varied cultures and are combined into structures which are always diversified. Thus we may wonder about the reason for this choice and the laws of combination. For insight into the specific problem of the avunculate we should turn to Radcliffe-Brown. His well-known article on the maternal uncle in South Africa23 was the first attempt to grasp and analyze the modalities of what we might call the "general principle of attitude qualification." We shall briefly review the fundamental ideas of that now-classic study. According to Radcliffe-Brown, the term avunculate covers two antithetical systems of attitudes. In one case, the maternal uncle represents family audiority; he is feared and obeyed, and possesses certain rights over his nephew. In the other case, the nephew holds privileges of familiarity in relation to his uncle and can treat him more or less as his victim. Second, there is a correlation between the boy's attitude toward his maternal uncle and his attitude toward his father. We find the two systems of attitudes in both cases, but they are inversely correlated. In groups where familiarity characterizes the relationship between father and son, the relationship between maternal uncle and nephew is one of respect; and where the father stands as the austere representative of family authority, it is the uncle who is treated with familiarity. Thus the two sets of attitudes constitute {as the structural linguist would say) two pairs of oppositions. Radcliffe-Brown concluded his article by proposing the following interpretation: In the final analysis, it is desc that determines the choice of opposiTio^' In patrilineal societies, where the father anrf the father's descent group represent trad' tional authority, the maternal uncle is c, sidered a "male mother." He is gener?|l" treated in the same fashion, and sornetit- ' even called by the same name, as the mot] In matrilineal societies, the opposite occur' Here, authority is vested in the materija' uncle, while relationships of tenderness and familiarity revolve about the father and his descent group. It would indeed be difficult to exaggerate importance of Radcliffe-Brown's contributi m which was the first attempt at synthesis on ■ empirical basis following Lowie's authoritat and merciless criticism of evolutionist nie physics. To say that this effort did not entir succeed does not in any way diminish ■ homage due to this great British anthropology but we should certainly recognize that R ■ cliffe-Brown's article leaves unanswered si> fundamental questions. First, the avuncul does not occur in all matrilineal or all patril eal systems, and we find it present in so systems which are neither matrilineal nor pa lineal.30 Further, the avuncular relationship is not limited to two terms, but presupposes four, namely, brother, sister, brother-in-law, ami nephew. An interpretation such as Radcliffe-Brown's arbitrarily isolates particular elements of a global structure which must be treated as a _ whole. A few simple examples will illustrate this twofold difficulty. The social organization of the Trobriand Islanders of Melanesia is characterized by matrilineal descent, free and familiar relations between father and son, and a marked antagonism between maternal uncle and nephew.3' On the other hand, the patrilineal Cherkess of the Caucasus place the hostility between father and son, while the maternal uncle assists: his nephew and gives him a horse when he marries.32 Up to this point we are still within the limits of Radcliffe-Brown's scheme. But: let " us consider the other family relationships involved. Malinowski showed that in the Trobriands husband and wife live in an atmosphere of tender intimacy and that their ieh tionship is characterized by reciprocity. The relations between brother 'and sister, on the.: STRUCTURAL ANALYSIS IN LINGUISTICS AND IN ANTHROPOLOGY 151 hand, are dominated by an extremely taboo. Let us now compare the situation 'lSjjje Caucasus. There, it is the brother-sister lfl| rionship which is tender - to such an extent 'hat among the Pschav an only daughter adopts" a "brother" who will play the cus-„ brother's role as her chaste bed com-° nion-33 But the relationship between spouses P entirely different. A Cherkess will not appear - nubile w't'1 ms w^e anc^ v*s'ts k£r on^ secret. According to Malinowski, there is no neater insult in the Trobriands than to tell a tniift-that he resembles his sister. In the Caucasus there is an analogous prohibition: It is forbidden to ask a man about his wife's health. When we consider societies of the Cherkess and Trobriand types it is not enough to study the correlation of attitudes between father/son and uncle/sister's son. This correlation is only one aspect of a global system containing four types of relationships which are organically linked, namely: brother/sister, husband/wife, father/son, and mother's brother/sister's son. The two groups in our example illustrate a law which can be formulated as follows: In both groups, the relation between maternal uncle arid nephew is to the relation between brother and sister as the relation between father and son is to that between husband and wife. Thus if we know one pair of relations, it is always possible to infer the other. Let us now examine some other cases. On Tonga, in Polynesia, descent is patrilineal, as among the Cherkess. Relations between husband and wife appear to be public and harmonious. Domestic quarrels are rare, and although the wife is often of superior rank, the husband "... is nevertheless of higher authority in all domestic matters, and no woman ^entertains the least idea of rebelling against that authority."34 At the same time there is great freedom between nephew and maternal ;. imcle. The nephew is fabu, or above the law, in relation to his uncle, toward whom extreme familiarity is permitted. This freedom strongly contrasts with the father-son relationship. The father is tapu; the son cannot touch his father's : head or hair; he cannot touch him while he eats, sleep in his bed or on his pillow, share "his food or drink, or play with his possessions. .However, the strongest tapu of all is the one between brother and sister, who must never be together under the same roof. Although they are also patrilineal and patri-local, the natives of Lake Kutubu in New Guinea offer an example of the opposite type of structure. F. E. Williams writes: "I have never seen such a close and apparently affectionate association between father and son____"3S Relations between husband and wife are characterized by the very low status ascribed to women and "the marked separation of masculine and feminine interests____5,36 The women, according to Williams, "are expected to work hard for their masters... they occasionally protest, and protest may be met with a beating."37 The wife can always call upon her brother for protection against her husband, and it is with him that she seeks refuge. As for the relationship between nephew and maternal uncle, it is "... best summed up in the word 'respect'... tinged with apprehensiveness,"38 for the maternal uncle has the power to curse his nephew and inflict serious illness upon him (just as among the Kipsigi of Africa). Although patrilineal, the society described by Williams is structurally of the same type as that of the Siuai of Bougainville, who have matrilineal descent. Between brother and sister there is "... friendly interaction and mutual generosity.... "39 As regards the father-son relationship, Oliver writes, "... I could discover little evidence that the word 'father' evokes images of hostility or stern authority or awed respect."40 But the relationship between the nephew and his mother's brother "appears to range between stern discipline and genial mutual dependence. ..." However, "... most of the informants agreed that all boys stand in some awe of their mother's brothers, and are more likely to obey them than their own fathers----"41 Between husband and wife harmonious understanding is rare: "... there are few young wives who remain altogether faithful... most young husbands are continually suspicious and often give vent to jealous anger. ..marriages involve a number of adjustments, some of them apparently difficult____"42 The same picture, but sharper still, characterizes the Dobuans, who are matrilineal and neighbors of the equally matrilineal Trobrian-ders, while their structure is very different. Dobuan marriages are unstable, adultery is 152 CLAUDE LEVI-STRAU5S STRUCTURAL ANALYSIS IN LINGUISTICS AND IN ANTHROPOLOGY 153 widespread, and husband and wife constantly fear death induced by their spouse's witchcraft. Actually, Fortune's remark, "It is a most serious insult to refer to a woman's witchcraft so that her husband will hear of it"43 appears to be a variant of the Trobriand and Caucasian taboos cited above. In Dobu, the mother's brother is held to be the harshest of all the relatives. "The mother's brother may beat children long after their parents have ceased to do so," and they are forbidden to utter his name. There is a tender relationship with the "navel," the mother's sister's husband, who is the father's double, rather than with the father himself. Nevertheless, the father is considered "less harsh" than the mother's brother and will always seek, contrary to the laws of inheritance, to favor his son at the expense of his uterine nephew. And, finally, "the strongest of all social bond the one between brother and sister.44 What can we conclude from these exarnp]esi The correlation between types of descent forms of avunculate does not exhaust the r. lem. Different forms of avunculate can coexist with the same type of descent, whether ]■ . lineal or matrilineal. But we constantly fin, ■ same fundamental relationship betwciri four pairs of oppositions required to cons -the system. This will emerge more clearly from the diagrams which illustrate our examples The sign + indicates free and familiar relation;, and the sign - stands for relations < acterized by hostility, antagonism, or res ■.. (Figure 8.1). This is an oversimplification, but we can tentatively make use of it. We describe some of the indispensable refinerr farther on. Cherkess-patrilineal A = Trobriand-matrilineal O A Tonga-patrilineal O A Siuai-matrilineal Lake Kubutu-patrilineal Figure 8.1 The synchronic law of correlation thus sug-: ted may be validated diachronically. If we ''uftimar'ze' a^er toward, the evolution of faiiiily relationships during the Middle Ages, ' e find approximately this pattern: The brother's authority over his sister wanes, and that of t!ie: prospective husband increases. Simultaneously the bond between father and son is "iateried and that between maternal uncle atiti nephew is reinforced. This evolution seems to be confirmed by the documents gathered by Leon Gautier, for in the "conservative" texts (Raoul de Cambrai, Geste des Loherains, etc.},46 the positive relationship is established chiefly between father and son and is only gradually displaced toward the maternal uncle and nephew.47 Thus we see48 that in order to understand the avunculate we must treat it as one relationship within a system, while the system itself must be considered as a whole in order to grasp its structure. This structure rests upon four terms ■(brother, sister, father, and son), which are linked by two pairs of correlative oppositions in such a way that in each of the two generations there is always a positive relationship ■and a negative one. Now, what is the nature ■ of this structure, and what is its function? The -answer is as follows: This structure is the most elementary form of kinship that can exist. It is, properly speaking, the unit of kinship. One may give a logical argument to support ■this statement. In order for a kinship structure .to exist, three types of family relations must .always be present: a relation of consanguinity, a relation of affinity, and a relation of descent -in other words, a relation between siblings, a ^relation between spouses, and a relation between parent and child. It is evident that the ; structure given here satisfies this threefold requirement, in accordance with the scientific principle of parsimony. But these considerations are abstract, and we can present a . more direct proof for our thesis. The primitive and irreducible character of the basic unit of kinship, as we have defined it, is actually a direct result of the universal presence of an incest taboo. This is really saying that in human society a man must obtain a woman from another man who gives him a daughter or a sister. Thus we do not need to explain how the maternal uncle emerged in the kinship structure: He does not emerge - he is present initially. Indeed, the presence of the maternal uncle is a necessary precondition for the structure to exist. The error of traditional anthropology, like that of traditional linguistics, was to consider the terms, and not the relations between the terms. Before proceeding further, let us briefly answer some objections which might be raised. First, if the relationship between "brothers-in-law" is the necessary axis around which the kinship structure is built, why need we bring in the child of the marriage when considering the elementary structure? Of course the child here may be either born or yet unborn. But, granting this, we must understand that the child is indispensable in validating the dynamic and teleological character of the initial step, which establishes kinship on the basis of and through marriage. Kinship is not a static phenomenon; it exists only in self-perpetuation. Here we are not thinking of the desire to perpetuate the race, but rather of the fact that in most kinship systems the initial disequilibrium produced in one generation between the group that gives the woman and the group that receives her can be stabilized only by counter-prestations in following generations. Thus, even the most elementary kinship structure exists both synchronically and diachronically. Second, could we not conceive of a symmetrical structure, equally simple, where the sexes would be reversed? Such a structure would involve a sister, her brother, brother's wife, and brother's daughter. This is certainly a theoretical possibility. But it is immediately eliminated on empirical grounds. In human society, it is the men who exchange the women, and not vice versa. It remains for further research to determine whether certain cultures have not tended to create a kind of fictitious image of this symmetrical structure. Such cases would surely be uncommon. We come now to a more serious objection. Possibly we have only inverted the problem. Traditional anthropologists painstakingly endeavored to explain the origin of the avunculate, and we have brushed aside that research by treating the mother's brother not as an extrinsic element, but as an immediate given of the simplest family structure. How is it then 154 CLAUDE LEVI-STRAUSS that we do not find the avunculate at all times and in all places? For although the avunculate has a wide distribution, it is by no means universal. It would be futile to explain the instances where it is present and then fail to explain its absence in other instances. Let us point out, first, that the kinship system does not have the same importance in all cultures. For some cultures it provides the active principle regulating al! or most of the social relationships. In other groups, as in our own society, this function is either absent altogether or greatly reduced. In still others, as in the societies of the Plains Indians, it is only partially fulfilled. The kinship system is a language; but it is not a universal language, and a society may prefer other modes of expression and action. From the viewpoint of the anthropologist this means that in dealing with a specific culture we must always ask a preliminary question: Is the system systematic? Such a question, which seems absurd at first, is absurd only in relation to language; for language is the semantic system par excellence; it cannot but signify, and exists only through signification. On the contrary, this question must be rigorously examined as we move from the study of language to the consideration of other systems which also claim to have semantic functions, but whose fulfillment remains partial, fragmentary, or subjective, like, for example, social organization, art, and so forth. Furthermore, we have interpreted the avunculate as a characteristic trait of elementary structure. This elementary structure, which is the product of defined relations involving four terms, is, in our view, the true atom of kinship.49 Nothing can be conceived or given beyond the fundamental requirements of its structure, and, in addition, it is the sole building block of more complex systems. For there are more complex systems; or, more accurately speaking, all kinship systems are constructed on the basis of this elementary structure, expanded or developed through the integration of new elements. Thus we must entertain two hypotheses: first, one in which the kinship system under consideration operates through the simple juxtaposition of elementary structures, and where the avuncular relationship therefore remains constantly apparent; second, a hypothesis in which the building blocks of the system are already of a more complex order I the latter case, the avuncular relation shi while present, may be submerged within a dif* ferentiated context. For instance, we can con" ceive of a system whose point of depai iure j|e in the elementary structure but which adds ^ the right of the maternal uncle, his wife, and at the left of the father, first the father's sister j1Kj then her husband. We could easily demonstrate that a development of this order leads to a parallel splitting in the following generation The child must then be distinguished according to sex - a boy or a girl, linked by a relation which is symmetrical and inverse to the turns occupying the other peripheral positions in the structure (for example, the dominant position of the father's sister in Polynesia, the South African nhlampsa, and inheritance by the mother's brother's wife). In this type of structure the avuncular relationship continues to prevail, but it is no longer the predominant one. In structures of still greater complexity, the avunculate may be obliterated or may merge with other relationships. But precisely because it is part of the elementary structure, the avuncular relationship re-emerges unmistakably and tends to become reinforced each time the system under consideration reaches a crisis - either because it is undergoing rapid transformation (as on the Northwest Coast), or because it is a focus of contact and conflict between radically different cultures (as in Fiji and southern India), or, finally, because it is in the throes of a mortal crisis (as was Europe in the Middle Ages). We must also add that the positive and negative symbols which we have employed in the above diagrams represent an oversimplification, useful only as a part of the demonstration. Actually, the system of basic attitudes comprises at least four terms: an attitude of affection, tenderness, and spontaneity; an attitude which results from the reciprocal exchange of prestations and counter-prestations; and, m~ addition to these bilateral relationships, two unilateral relationships, one which corresponds to the attitude of the creditor, the other: to that of the debtor. In other words there are:.: mutuality ( = ), reciprocity ( ± ), rights (+h and obligations (—). These four fundamental attitudes are represented in their reciprocal relationships in Figure 8.2. STRUCTURAL ANALYSiS IN LINGUISTICS AND IN ANTHROPOLOGY 155 Figure 8.2 In many systems the relationship between two individuals is often expressed not by a single attitude, but by several attitudes which together form, as it were, a "bundle" of attitudes (as in the Trobriands, where we find both mutuality and reciprocity between husband njid wife). This is an additional reason behind the difficulty in uncovering the basic structure. We have tried to show the extent to which the preceding analysis is indebted to outstanding contemporary exponents of the sociology of primitive peoples. We must stress, however, that in its most fundamental principle this an-: alysis departs from their teachings. Let us cite as an example Radcliffe-Brown: The unit of structure from which a kinship system is built up is the group which I call an "elementary family," consisting of a man and his wife and their child or children----The existence of the elementary family creates three special kinds of social relationship, that between parent and child, that between children of the same parents (siblings), and that between husband and wife as parents of the same child or children____The three relationships that exist within the elementary family constitute what I call the first order. Relation-ships of the second order are those which depend on the connection of two elementary families through a common member, and are such as father's father, mother's brother, wife's sister, and so on. In the third order are such as father's brother's son and mother's brother's wife. Thus we can trace, if we have genealogical information, relationships of the fourth, fifth or nA order.50 ..." The idea expressed in the above passage, that -the biological family constitutes the point of .departure from which all societies elaborate Their kinship systems, has not been voiced solely by Radcliffe-Brown. There is scarcely an idea which would today elicit greater consensus. Nor is there one more dangerous, in our opinion. Of course, the biological family is ubiquitous in human society. But what confers upon kinship its socio-cultural character is not what it retains from nature, but, rather, the essential way in which it diverges from nature. A kinship system does not consist in the objective ties of descent or consanguinity between individuals. It exists only in human consciousness; it is an arbitrary system of representations, not the spontaneous development of a real situation. This certainly does not mean that the real situation is automatically contradicted, or that it is to be simply ignored. Radcliffe-Brown has shown, in studies that are now classic, that even systems which are apparently extremely rigid and artificial, such as the Australian systems of marriage-classes, take biological parenthood carefully into account. But while this observation is irrefutable, still the fact (in our view decisive) remains that, in human society, kinship is allowed to establish and perpetuate itself only through specific forms of marriage. In other words, the relationships which Radcliffe-Brown calls "relationships of the first order" are a function of, and depend upon, those which he considers secondary and derived. The essence of human kinship is to require the establishment of relations among what Radcliffe-Brown calls "elementary families." Thus, it is not the families (isolated terms) which are truly "elementary," but, rather, the relations between those terms. No other interpretation can account for the universality of the incest taboo; and the avuncular relationship, in its most general form, is nothing but a corollary, now covert, now explicit, of this taboo. Because they are symbolic systems, kinship systems offer the anthropologist a rich field, where his efforts can almost (and we emphasize the "almost") converge with those of the most highly developed of the social sciences, namely, linguistics. But to achieve this convergence, from which it is hoped a better understanding of man will result, we must never lose sight of the fact that, in both anthropological and linguistic research, we are dealing strictly with symbolism. And although it may be legitimate or even inevitable to fall back upon a 156 CLAUDE LEVI-STRAUSS naturalistic interpretation in order to understand the emergence of symbolic thinking, once the latter is given, the nature of the explanation must change as radically as the newly appeared phenomenon differs from those which have preceded and prepared it. Hence, any concession to naturalism might jeopardize the immense progress already made in linguistics, which is also beginning to characterize the study of family structure, and might drive the sociology of the family toward a sterile empiricism, devoid of inspiration. NOTES 1 Marcel Mauss, "Rapports reels et pratiques de la psychologie et de la sociologie," Journal de Psychologie Normale et Pathologique (1924); reprinted in Sociologie et Anthropologie (Paris: 1951), p. 299. 2 O. Schräder, Prehistoric Antiquities of the Aryan Peoples, trans. F. B. Jevons {London: 1890), Chapter XII, Part 4. 3 Ibid. See also H. J. Rose, "On the Alleged Evidence for Mother-Right in Early Greece," Folklore, XXII (1911), and the more recent studies by George Thomson, which support the hypothesis of matrilineal survivals. 4 A. M. Hocart, "Chieftainship and the Sister's Son in the Pacific," American Anthropologist, n.s., XVII (1915); "The Uterine Nephew," Man, XXIII, No. 4 (1923); "The Cousin in Vedic Ritual," Indian Antiquary, LIV (1925); etc. 5 Paul K. Benedict, "Tibetan and Chinese Kinship Terms," Harvard Journal of Asiatic Studies, VI (1942); "Studies in Thai Kinship Terminology," Journal of the American Oriental Society, LXIII (1943). 6 L. Brunschvicg, Le Progres de la conscience dans la Philosophie occidentale (Paris: 1927), II, p. 562. 7 Between 1900 and 1920 Ferdinand de Saussure and Antoine Meillet, the founders of modern linguistics, placed themselves determinedly under the wing of the anthropologists. Not until the 1920's did Marcel Mauss begin - to borrow a phrase from economics -to reverse this tendency. 8 N. Troubetzkoy, "La Phonologie actuelle," in Psychologie du langage (Paris: 1933). 9 Ibid., p. 243. 10 hoc. ext. 11 Ibid., p. 245; Roman Jakobson, "prj der historischen Phonologie," Tra ^ Cercle linguistique de Prague, \\r iy, ''' and also Jakobson, "Remarques sur ]>£Vo[ ) tion phonologique du russe," ibid \\ i ] < 12 W. H. R. Rivers, The History ofMelanJa < Society (London: 1914), passim; Social ol ' ganization, ed. W. J. Perry I London; 1.9 Chapter IV. ' 13 In the same vein, see Sol Tax, "Some! lems of Social Organization," in FredE (ed.), Social Anthropology of North Ainen-can Tribes (Chicago: 1937), 14 Roman jakobson, "Observations sur leMn sement phonologique des consonnes/" , ceedings of the Third International Congress * of Phonetic Sciences (Ghent: 1938). ::; 15 K. Davis and W. L. Warne^ "Structural An- : alysis of Kinship," American Anthropoid gist, n.s., XXXVII (1935). ^ 16 Thus at the end of the analysis carried out by these authors, the term husband is replaced by the formula: C272d/0 SUla s/Ego (Ibid.) There are now available two works which employ a much more refined logical apparatus and offer greater interest in terms both of method and of results. See E G. Lotms-bury, "A Semantic Analysis of the Pawnee Kinship Usage," Language, XXXII, No. 1 (1956), and W. II. Goodcnough. "The Componential Analysis of Kinship," ibid. 17 As will be seen in Chapter V, I have, now'i* refined this formulation. 18 A. I . Kroeber, "Classificatory Systems of . Relationship," Journal of the Royal Anthro-.'. pological Institute, XXXIX (1909). 19 We must except the remarkable work of W. L. Warner, "Morphology and Functions of the Australian Murngin Type of Kinship," American Anthropologist, n.s., XXXII-XXXIII (1930-1931), in which his analysis;! of the system of attitudes, although funda- . mentally debatable, nevertheless initiates a new phase in the study of problems of kinship. 20 A. R. Radcliffe-Brown, "Kinship Termin-ology in California," American Anthropoid-gist, n.s., XXXVII (1935); "The Study of Kinship Systems," Journal of the Royal An-.- : thropological Institute, LXXI (1941). 21 M. E. Opler, "Apache Data Concerning the Relationship of Kinship Terminology to STRUCTURAL ANALYSIS IN LINGUISTICS AND IN ANTHROPOLOGY 157 ibciai Classification," American Anthropologic n.s., XXXIX (1937); A. M. Hal pern. Yuma Kinship Terms," American ^Anthropologist, n.s., XLIV (1942). '.'J, p. Thomson, "The Joking Relationship ' aJlCj Organized Obscenity in North Queensland," American Anthropologist, n.s., XXXVU (1935). 2ji:: Radcliffe-Brown, "The Study of Kinship >■ Systems," op. cit., p. 8. This later formuláři, tion seems to us more satisfactory than his jt>55 statement that attitudes present "a ^■fairly high degree of correlation with the ■terminological classification" (American An-%éwpologist, n.s., XXXVII [1935], p. 53). 24;S Sydney Hartland, "Matrilineal Kinship and " the Question of its Priority," Memoirs of the ^■■American Anthropological Association, No. 4 (1917). 1 í: i W. H. R. Rivers, "The Marriage of Cousins -.in India," Journal of the Royal Asiatic Soci-........ etyi\u\y, 1907). 26 Ibid., P- 624. 27 R. H. Lowie, "The Matrilineal Complex," University of California Publications in American Archaeology and Ethnology, XVI, No. 2 (1919). 28 Roman Jakobson, Kindersprache, Aphasie and allgemeine Lautgesetze (Uppsala: 1941). 29 A. R. Radcliffe-Brown, "The Mother's '-. Brother in South Africa," South African journal of Science, XXI (1924). 30 As among the Mundugomor of New 'J.':'.,-, Guinea, where the relationship between ma->;"-;':■*■ ternal uncle and nephew is always familiar, although descent is alternately patrilineal or matrilineal. See Margaret Mead, Sex and J;'- Temperament in Three Primitive Societies ■'"J- (New York: 1935), pp. 176-185. š:31ř B. Malinowski, The Sexual Life of Savages in Northwestern Melanesia (London: 1929), . ..ý. 2 vols. 32 Dubois de Monpereux (1839), cited in M. ;;vj-' Kovalevski, "La Familie matriarcale au Cau- case," L'Anthropologie, IV (1893). 33 Ibid. 34 E. W. Gifford, "Tonga Society," Bernice P. Bishop Museum Bulletin, No. 61 {Honolulu: ■ ■ 1929), pp. 16-22. 35 F. E. Williams, "Group Sentiment and Primitive Justice," American Anthropologist, n.s., XLIII, No. 4, Part 1 (1941), p. 523. 36 F. E. Williams, "Natives of Lake Kutubu, Papua," Oceania, XI (1940-1941), p. 266. 37 Ibid., p. 268. 38 Ibid., p. 280. See also Oceania, XII (1941-1942). 39 Douglas L. Oliver, A Solomon Island Society: Kinship and Leadership among the Siuai of Bougainville (Cambridge, Mass.: 1955), p. 255. 40 Ibid., p. 251. 41 Ibid., p. 257. 42 Ibid., pp. 168-9. 43 R, F. Fortune, The Sorcerers of Dobu (New York: 1932), p. 45. 44 Ibid., pp. 8, 10, 62-4. 45 G. E. Howard, A History of Matrimonial Institutions, 3 vols. (Chicago: 1904). [46 Translator's note: The "Chansons de Geste," which survive in manuscript versions of the twelfth to the fifteenth century, are considered to be remodelings of much earlier originals, dating back to the age of Charlemagne. These poems of heroic and often legendary exploits also constitute a source of information on the family life of that period.] 47 Leon Gautier, La Chevalerie (Paris: 1890). See also: F. B. Gummere, "The Sister's Son," in An English Miscellany Presented to Dr. Purnivall (London: 1901); W. O. Farns-worth, Uncle and Nephew in the Old French Chanson de Geste (New York: 1913). 48 The preceding paragraphs were written in 1957 and substituted for the original text, in response to the judicious remark by my colleague Luc de Heusch of the Universite Libre of Brussels that one of my examples was incorrect. I take this opportunity to thank him. 49 It is no doubt superfluous to emphasize that the atomism which we have criticized in Rivers refers to classical philosophy and has nothing to do with the structural conception of the atom developed in modern physics. 50 A. R. Radcliffe-Brown, "The Study of Kinship Systems," op. cit., p. 2.