You are here

George Peter Murdock
Social structure

New York, Macmillan Co, 1949, xx+387 pages

A. L. Kroeber, R. Linton, R. H. Lowie, L. H. Morgan, A. R. Radcliffe-Brown, and W. H. R. Rivers
Whose creative efforts in the study of social organization are largely responsible for its present development

And to
J. Dollard, S. Freud, C. L. Hull, and A. G. Keller
Whose contributions to sociology and psychology
have immeasurably aided the present author
in furthering that development


This volume represents a synthesis of five distinct products of social science— one research technique and four systems of theory. It grows out of, depends upon, and reflects all five. It is the result of a conscious effort to focus several disciplines upon a single aspect of the social life of man— his family and kinship organization and their relation to the regulation of sex and marriage.
In intent, and hopefully in achievement, the work is not a contribution to anthropology alone, nor to sociology or psychology, but to an integrated science of human behavior.

Social structure. 1949
George P. Murdock
Social structure

The research technique upon which the volume depends, and without which it would not have been undertaken, is that of the Cross-Cultural Survey. Initiated in 1937 as part of the integrated program of research in the social sciences conducted by the Institute of Human Relations at Yale University, the Cross-Cultural Survey has built up a complete file of geographical, social, and cultural information, extracted in full from the sources and classified by subject, on some 150 human societies, historical and contemporary as well as primitive. From these files it is possible to secure practically all the existing information on particular topics in any of the societies covered in an insignificant fraction of the time required for comparable library research.

The author began the present study in 1941 by formulating a schedule of the data needed on the family, on kinship, on kin and local groups, and on marriage and sex behavior, and by abstracting such data from the files of the Cross-Cultural Survey on- all societies for which sufficient information had been reported. In a very few weeks he was able in this way to assemble the relevant materials for 85 societies, which are indicated by asterisks in the bibliography.

This number, though large, still fell far short of the cases required for reliable statistical treatment, and the author set out to secure further information by the usual methods of library research. Eventually he secured data on 165 additional societies, making a total of 250 in all. The labor required to secure these additional cases was immense, consuming well over a year of research effort or more than ten times that spent in obtaining the original 85 cases. Moreover, the results were both quantitatively and qualitatively inferior, since the author had to content himself in most instances with a single book or article in contrast to the complete source coverage for the cases derived from the Survey files. The informed reader who detects factual gaps or errors in our tabulated data will usually find, by referring to the bibliography, that they are due to the failure to use some recognized source. The author's only excuse for his incomplete coverage in the additional 165 societies is that he simply could not afford the extra years of research labor that would have been required to attain the degree of thoroughness achieved by the Cross-Cultural Survey. If the Survey ever reaches its goal of covering a representative ten per cent sample of all the cultures known to history, sociology, and ethnography, it should be possible to produce several studies like the present one, far more fully and accurately documented, in the time needed to compile and write this volume.

The use of statistics and of the postulational method of scientific inquiry has been contemplated from the beginning 1 as a major objective in the utilization of the accumulated materials of the Cross-Cultural Survey files. The present author has departed from plan mainly in abandoning the sampling technique in favor of using all available cases in areas such as South America and Eurasia ior which there are too few sufficiently documented cases to obtain an adequate sample. In other areas, too, he has occasionally chosen a society because a good source was readily accessible rather than because a sample was demanded. He has, however, sought consciously to avoid any appreciable over-representation of particular culture areas. In short, departures from, strict sampling, where they occur, reflect availability or non-availability of suitable sources and no other basis of selection. This explains why the sample includes 70 societies from native North America, 65 from Africa, 60 from Oceania, 34 from Eurasia, and 21 from South America rather than an approximately equal number from each of the five.

 To avoid any possible tendency to select societies which might support his hypotheses or to reject those which might contradict them, the author included all societies with sufficient information in the Cross-Cultural Survey files and adopted a standard policy for the additional cases. Having first determined that a particular society would meet the sampling criteria, he turned to the available sources and quickly flipped the pages. If there seemed at a glance to be data on kinship terminology, on sex and marriage, and on familial, kin, and local groups, he accepted the case before examining any of the information in detail, and resolved not to exclude it thereafter.

 This policy resulted in the inclusion of a number of societies for which the data are scanty and possibly unreliable. In nine instances— the Arawak, Fulani, Hiw, Huichol, Jivaro, Kamba, Mohave, Porno, and Sinhalese— the information proved so wholly inadequate in early tabulations that the resolution was abandoned and the cases excluded.

In at least nine other instances— the Getmatta; Hawaiians, Hupa, Mataco, Mikir, Nambikuara, Ruthenians, Twi, and Vai— similar inadequacies showed up later, but resolve stiffened and they were retained.

It would have been scientifically desirable to examine every negative or exceptional case to determine the countervailing factors apparently responsible for its failure to accord with theoretical expectations, since there can be no genuine exceptions to valid scientific principles. An attempt has been made to do so in Chapter 8 and occasionally elsewhere. To have followed this policy throughout, however, would have been impracticable in view of the fact that the 250 cases have been subjected to hundreds of different tabulations.

Although most sociologists and the functionalists among anthropologists fully recognize that the integrative tendency in the process of cultural change justifies the treatment of individual cultures as independent units for statistical purposes, many historical anthropologists and a few other social scientists still sus- pect that the fact of diffusion— the known dependence of most societies upon borrowing from others for a large proportion of their cultural elements— invalidates this statistical assumption: The author could doubtless argue his case at great length without convincing these skeptics. He therefore determined to face the issue squarely in an appendix, recalculating a series of tabulations using as units not individual tribes but culture areas and linguistic stocks, the two most widely recognized groupings of peoples with indisputable historical connections. Several trial calculations yielded results practically identical with those obtained with tribal units. This plan was abandoned, however, when a much more satisfactory method was discovered whereby our hypotheses could be validated by strictly historical means. This is done in Appendix A.

The fact that our historical test corroborates our statistical demonstration, coupled with the sampling method of selecting tribal units, with the confirmation from trial tests with other units, and with the specific disproof in Chapter 8 that historical connections significantly affect the forms of social organization, should shift the burden of proof squarely upon the skeptics. Unless they can specifically demonstrate, upon a scale at least comparable to that of the present work, that diffusion negates the tendency of cultures to undergo modification in the direction of the integration of their component elements, the assumption that historical contacts do not destroy the independent variability of cultural units stands immune to challenge.

As regards statistical operations, we have adopted two indices for each computation— Yule's Coefficient of Association ( Q ) for purposes of comparison and a Chi Square (x 2 ) index for showing the probability of the particular distribution occurring by chance. For advice on statistical methods, the author is greatly indebted to Dr. Irvin L. Child, Dr. Carl I. Hovland, Dr. Douglas H. Lawrence, Dr. Ben- net B. Murdock, Dr. Oystein Ore, Dr. Fred D. Sheffield, and Dr. John R. Wittenborn. It is scarcely necessary to state that there has been no rejection of tabulations yielding low or negative coefficients. In the very small number of cases in which a calculation has yielded a coefficient “with negative sign, analysis has revealed some defect in the hypothesis, which was then revised. The coefficients which test the corrected hypotheses in this volume are not infrequently too low to be statistically significant, but more important is the fact that out of hundreds of computations not a single one has turned out genuinely negative in sign. The chances of such an outcome, if -the theories tested were unsound in any substantial respect, are incredibly infinitesimal.

Without the Cross-Cultural Survey the present study could scarcely have beenmade, or its methods applied on so large a scale. The author consequently owes a deep debt of personal and professional gratitude to Dr. Mark A. May, Director of the Institute of Human Relations, for his unwavering support of the project. He is also greatly indebted to the Carnegie Corporation, and particularly to Charles Dollard, now its president, for financial aid to the Survey during a war emergency; to the Coordinator of Inter-American Affairs, and particularly to Dr. Willard Z. Park and his staff, for supporting an immense increment of materials from modern Latin American societies under a war project called the Strategic Index of the Americas; to the Navy Department, and particularly to Captain A. E. Hindmarsh, USN, and to Captain Harry L. Pence, USN, for making possible complete coverage during the war of the then Japanese-controlled islands of the Pacific. Appreciation for their assistance must also be acknowledged to Mrs. Aimee Alden, John M. Armstrong, Jr., Dr. Wendell C. Bennett, Ward H. Goodenough, Geoffrey Gorer, Mrs. Frances Campbell Harlow, Dr. Harry Hawthorn, Dr. Allan R. Holmberg, Dr. Donald Horton, Mrs. Lois Howard, Dr. Benjamin Keen, Dr. Raymond Kennedy, Dr. William Ewart Lawrence, Dr. Oscar Lewis, Professor Leonard Mason, Dr. Alfred Metraux, Dr. Alois M. Nagler, Dr. Benjamin Paul, Dr. Gitel Poznanski, Dr. John M. Roberts, Jr., Dr. Mary Rouse, Dr. Bernard Siegel, Dr. Leo W. Simmons, Mrs. Marion Lambert Vanderbilt, and the many other's who have been associated with the Cross-Cultural Survey in some of its aspects, but especially to Dr. Clellan S. Ford and Dr. John W. M. Whiting, who have been intimately connected with the project throughout its development.

Of the four systems of social science theory which have influenced this volume and are reflected in its results the first is that of sociology.
Here the author must repeat the acknowledgment 2 of his enormous intellectual indebtedness to Albert G. Keller, under whom he did his graduate work and with whom he was associated for many years as a junior colleague. From Professor Keller he acquired the conviction that the only road to genuine knowledge is the arduous path of science and the realization that social behavior in our own modern society can best be comprehended from a background of the comparative study of earlier and simpler peoples. Through Keller, too, he became aware of the major contributions of William Graham Sumner— the relativity of culture and the still incompletely appreciated fact of its affective basis and its permeation with sanction and moral values.

More important even than the above, however, was the discovery that culture is adaptive or “functional,” subserving the basic needs of its carriers and altering through time by a sort of mass trial-and-error in a process which is truly evolutionary, i.e., characterized by orderly adaptive change 3. This was Keller's own contribution—not inherited from Sumner— and of itself would justify for its author a place among the great social scientists of all time. So valid is this conception, and so vastly more sophisticated than the views of the nature of culture and cultural change held by anthropologists and other sociologists at the time, that it went almost completely unrecognized. Only in the last decade or two has this point of view gained general acceptance, largely through the “functionalism” of Malinowski and the more recent studies of “culture and personality,” yet even today Keller's priority is practically unknown. Most of what other anthropologists have subsequently learned from Malinowski had already become familiar to the present author through the influence of Keller. For these reasons he will be eternally thankful that he did his graduate work under Keller in sociology, though he might have received a technically more exacting training under Boas in anthropology.

Unfortunately, marked disadvantages accompanied these values, and made it necessary for the author to deviate from the Sumner- Keller tradition on important issues. A genuine respect for science seemed inconsistent with the intolerance of other approaches so often manifested by Professor Keller and his disciples, for work of obvious importance was being done by other sociologists and by psychologists and cultural anthropologists of several schools. Moreover, value judgments buttressed with cases too frequently masqueraded as science. Above all, The Science of Society ( New Haven, 1927) is so permeated with survivals of nineteenth century evolutionism, which historical anthropologists have long since disproved, that a very high degree of selectivity is required in accepting its conclusions, with the result that impatient, insufficiently skilled, or hypercritical scholars often prefer to reject it in toto. But however seriously discounted, the Sumner-Keller approach nevertheless remains perhaps the most influential single intellectual stimulus behind the present volume.

Among his other former sociological colleagues at Yale the author has profited particularly from professional and personal associations with E. Wight Bakke, Maurice R. Davie, Raymond Kennedy, James G. Leyburn, Stephen W. Reed, and Leo W. Simmons. Among sociologists in other institutions he has been especially influenced by the contributions to cultural theory of William F. Ogburn, by the rigorous scientific objectivity of George Lundberg, by the methodological versatility of Raymond V. Bowers, and by the contributions to social structure of Kingsley Davis, Robert Merton, and Talcott Parsons.

The second profound influence upon this work stems from the group of American historical anthropologists of whom Franz Boas was the pioneer and intellectual leader. The author must stress particularly his genuine indebtedness to this school and his sincere appreciation of its contributions, since he will find it necessary on occasion to take sharp issue with their conclusions. At the beginning of the twentieth century, thought and theory in anthropology and related social sciences were saturated with evolutionistic assumptions which barred the way to further scientific development. Boas and his disciples assumed as their principal function the riddance of this intellectual debris, and so skillfully and energetically did they pursue this task that by 1920 evolutionism in the social sciences was completely defunct. This feat was accomplished by concentration upon field research and by demonstration of the historical interrelatedness of the cultures in particular areas, such as the Plains and the Northwest Coast. In addition to revealing the fallacy of unilinear evolution, this activity established field research as the hallmark of professional anthropology— such early theorists as Bachofen, Durkheim, Frazer, Graebner, Lippert, Lubbock, Marett, McLennan, Schmidt, Sumner, Tylor, and Westermarck were not field workers— and thereby resulted in equipping anthropologists with that first-hand experience with alien cultures which is the best guarantee of realism in theoretical interpretations. The importance of these contributions can scarcely be overemphasized.

The school accomplished distressingly little, however, toward the advancement of cultural theory. Having exorcised the bogey of evolutionism, it could discover no promising new objectives. Boas himself, who has been extravagantly overrated by his disciples 4, was the most unsystematic of theorists, his numerous kernels of genuine insight being scattered amongst much pedantic chaff. He was not even a good field worker. 5 He nevertheless did convey to his students a genuine respect for ethnographic facts and for methodological rigor. In the hands of some of his followers, however, his approach degenerated into a sterile historicism consisting of rash inferences concerning prehistory from areal distributions. With others it became converted into an unreasoning opposition to all new trends in anthropology.

In Leslie Spier the virtues of the Boas approach find their purest distillation. Despite certain limitations in outlook, he is a sound and systematic ethnographer, he reaches historical conclusions with a caution and a respect for detail that evoke admiration, and he exhibits considerable capacity for theoretical formulations.
The author values his years of close association with Spier, and acknowledges a strong indebtedness to him for indoctrination into the tradition of American historical anthropology. Among other students of Boas, Robert H. Lowie has also done outstanding work, particularly in the field of social organization, and A. L. Kroeber appears to the author the leading anthropologist of his own and adjacent generations. Despite his defensiveness respecting the historical approach and his occasional rash use of its methods, Kroeber kept American anthropology alive through an ineffectual generation by his originality, his concern with vital issues, and his analytical insight. If a tree is to be judged by its fruits, Boas is justified by these three men, even though the whole weight of his personal influence after 1920 was directed toward stemming the natural development of a scientific anthropology.

From Clark Wissler, initially through his writings and subsequently through personal contact, the author first acquired an adequate appreciation of the relation of cultures to their geographic backgrounds and of the regional distribution of cultural elements, a contribution from American historical anthropology of fundamental importance. To Edward Sapir he is indebted for such linguistic knowledge as he possesses and for his initiation into fieldwork, and he readily acknowledges that the field of culture and personality owes its initial stimulus very largely to Sapir's extra- ordinary intuitive flair and verbal facility. It becomes increasingly apparent, however, that the permanent contributions of Sapir to cultural theory are relatively slight in comparison to those which he achieved in linguistic science. For Ralph Linton the author feels only the most profound respect. In him, the historical, functional, and psychological approaches are welded into a harmonious synthesis which typifies modern cultural anthropology at its best. In fields in which both he and the author possess professional competence, the latter has seldom been able to discover a significant difference of opinion on theoretical issues.

A decade ago the author might have been inclined to rank functional anthropology among the important influences upon his thinking. Not so today. Personal contact with Bronislaw Malinowski brought intellectual stimulation and some clarification respecting social institutions but no fundamental point of view not previously acquired from Keller. The work of Radcliffe-Brown on social organization appears exceedingly impressive upon superficial acquaintance and was, indeed, the factor which first induced the author to specialize in the field. On closer view, however, its virtues wane, and they fade into insubstantiality with intensive study. In the controversy between Kroeber and Radcliffe-Brown, for example, the expressed views of the latter seemed, and still seem, appreciably sounder, but in the actual analysis and interpretation of data the former has proved right and the latter wrong in nearly every instance.

The principal impact of functional anthropology appears to have been the revolution it has wrought in the younger generation of American anthropologists, who for the most part are no longer purely historical, or functional, or psychological, but who wield several instruments at the same time with an often extraordinary degree of skill. In the hands of these capable and catholic men lies the future of anthropology, and perhaps even that of social science in general. Among them the author acknowledges special intellectual indebtedness to H. G. Barnett, Fred Eggan, Clellan S. Ford, John P. Gillin, A. I. Hallowell, Allan R. Holmberg, Clyde Kluckhohn, W. E. Lawrence, Morris E. Opler, Alexander Spoehr, Julian H. Steward, W. Lloyd Warner, and John W. M. Whiting.

The third system of organized knowledge which has significantly influenced this volume is behavioristic psychology. Although the author had become acquainted with the work of Pavlov and Watson early in his career, and had reacted favorably, it was not until he met Clark L. Hull a decade ago and became familiar with the latter's work that he fully appreciated the soundness of the approach and recognized its extraordinary utility for cultural theory. Of all the systematic approaches to the study of human behavior known to the author, that of Hull exceeds all others in scientific rigor and objectivity, and it is the only one against which he can level no serious criticism. Without it, the main conclusions of the present work would have been virtually impossible. It appears capable of shedding more light upon cultural problems than any other product of psychological science. By contrast, the social psychologies, Gestalt, and even psychoanalysis seem comparatively slender. For making available to him this source of illumination the author is under the deepest debt to Professor Hull, as well as to John Dollard for introducing him to it. He is also greatly obligated to Carl I. Hovland, Donald G. Marquis, Mark A. May, Neal E. Miller, O. H. Mowrer, Robert Sears, and John Whiting for interpreting the principles of learning and behavior to him.

The fourth and final body of theory which has substantially influenced this volume is psychoanalysis. Recognition is fully and gladly accorded herewith to the genius of Freud, to the keenness of his insight into a previously obscure realm of phenomena, and to the extreme importance and essential soundness of his discoveries. Without disparagement of the above, and without denying the unquestionable value of Freudian therapeutic techniques, the author must nevertheless express his conviction that the theoretical system of psychoanalysis is in the highest degree obscure, that its hypotheses are frequently overlapping and even contradictory, and that it fulfills few of the requisites of a rigorous, testable, and progressive body of scientific knowledge. In his opinion, therefore, it is probably destined to disappear as a separate theoretical discipline as its findings are gradually incorporated into some more rigorous scientific system such as that of behavioristic psychology. Considerable progress has already been made in this direction, but the absorption is still very far from complete. The author has therefore been compelled to use unassimilated Freudian theory in two sections of this book, namely, in the interpretation of avoidance and joking relationships in Chapter 9 and in the analysis of incest prohibitions in Chapter 10.

For introducing him to this exceedingly significant though somewhat unsatisfactory scientific approach the author is deeply indebted to John Dollard. He is equally obligated to Earl F. Zinn for guiding him through fifteen months of analysis and for prior and subsequent exposition of Freudian principles. Neal Miller, Hobart Mowrer, and John Whiting have also participated in his education, and he has derived benefit from Erich Fromm, Karen Horney, and Abram Kardiner in the attempt to apply psychoanalytic principles to cultural materials.

In conclusion, a fitting tribute should be paid to the Institute of Human Relations and to its director, Dr. Mark A. May, for the enlightened policy of bringing these several disciplines and techniques together with the aim of developing a coordinated and interdisciplinary science of human behavior.

New Haven, Connecticut
September, 1948

George Peter Murdock

1. See G. P. Murdock, “The Cross-Cultural Survey,” American Sociological Review, V ( 1940), 369-70.
2. See G. P. Murdock, Studies in the Science of Society (New Haven, 1937), pp. vii-xx.
3. See A. G. Keller, Societal Evolution (New York, 1915).
4. Cf. R. H. Lowie, History of Ethnological Theory (New York, 1937), pp. 128-155.
5. Despite Boas' “five-foot shelf of monographs on the Kwakiutl, this tribe falls into the quartile of those whose social structure and related practices are least adequately described among the 250 covered in the present study.

The Nuclear Family

The family is a social group characterized by common residence, economic cooperation, and reproduction. It includes adults of both sexes, at least two of whom maintain a socially approved sexual relationship, and one or more children, own or adopted, of the sexually cohabiting adults. The family is to be distinguished from marriage, which is a complex of customs centering upon the relationship between a sexually associating pair of adults within the family. Marriage defines the manner of establishing and terminating such a relationship, the normative behavior and reciprocal obligations within it, and the locally accepted restrictions upon its personnel.
Used alone, the term “family” is ambiguous. The layman and even the social scientist often apply it undiscriminatingly to several social groups which, despite functional similarities, exhibit important points of difference. These must be laid bare by analysis before the term can be used in rigorous scientific discourse.
Three distinct types of family organization emerge from our survey of 250 representative human societies. The first and most basic, called herewith the nuclear family, consists typically of a married man and woman with their offspring, although in individual cases one or more additional persons may reside with them. The nuclear family will be familiar to the reader as the type of family recognized to the exclusion of all others by our own society. Among the majority of the peoples of the earth, however, nuclear families are combined, like atoms in a molecule, into larger aggregates. These composite forms of the family fall into two types, which differ in the principles by which the constituent nuclear families are affiliated, A polygamous family 1 consists of two or more nuclear families affiliated by plural marriages, i.e., by having one married parent in common 2. Under polygyny, for instance, one man plays the role of husband and father in several nuclear families and thereby unites them into a larger familial group. An extended family consists of two or more nuclear families affiliated through an extension of the parent-child relationship rather than of the husband-wife relationship, i.e., by joining the nuclear family of a married adult to that of his parents. The patrilocal extended family, often called the patriarchal family, furnishes an excellent example. It embraces, typically, an older man, his wife or wives, his unmarried children, his married sons, and the wives and children of the latter. Three generations, including the nuclear families of father and sons, live under a single roof or in a cluster of adjacent dwellings.
Of the 192 societies of our sample for which sufficient information is available, 47 have normally only the nuclear family, 53 have polygamous but not extended families, and 92 possess some form of the extended family. The present chapter will concern itself exclusively with the nuclear family. The composite forms of family organization will receive special attention in Chapter 2.
The nuclear family is a universal human social grouping. Either as the sole prevailing form of the family or as the basic unit from which more complex familial forms are compounded, it exists as a distinct and strongly functional group in every known society. No exception, at least, has come to light in the 250 representative cultures surveyed for the present study, which thus corroborates the conclusion of Lowie 3 :

“It does not matter whether marital relations are permanent or temporary; whether there is polygyny or polyandry or sexual license; whether conditions are complicated by the addition of members not included in our family circle: the one fact stands out beyond all others that everywhere the husband, wife, and immature children constitute a unit apart from the remainder of the community.”

The view of Linton 4 that the nuclear family plays “an insignificant role in the lives of many societies” receives no support from our data. In no case have we found a reliable ethnographer denying either the existence or the importance of this elemental social group. Linton mentions the Nayar of India as a society which excludes the husband and father from the family, but he cites no authorities, and the sources consulted by ourselves for this tribe do not substantiate his statement. Whatever larger familial forms may exist, and to whatever extent the greater unit may assume some of the burdens of the lesser, the nuclear family is always recognizable and always has its distinctive and vital functions—sexual, economic, reproductive, and educational—which will shortly be considered in detail. It is usually spatially as well as socially distinct. Even under polygyny a separate apartment or dwelling is commonly reserved for each wife and her children.
The reasons for its universality do not become fully apparent when the nuclear family is viewed merely as a social group. Only when it is analyzed into its constituent relationships, and these are examined individually as well as collectively, does one gain an adequate conception of the family's many-sided utility and thus of ita inevitability. A social group arises when a series of interpersonal relationships, which may be defined as sets of reciprocally adjusted habitual responses, binds a number of participant individuals collectively to one another. In the nuclear family, for example, the clustered relationships are eight in number: husband-wife, fatherson, father-daughter, mother-son, mother-daughter, brother-brother, sister-sister, and brother-sister 5. The members of each interacting pair are linked to one another both directly through reciprocally reinforcing behavior and indirectly through the relationships of each to every other member of the family. Any factor which strengthens the tie between one member and a second, also operates indirectly to bind the former to a third member with whom the second maintains a close relationship. An explanation of the social utility of the] nuclear family, and thus of its universality, must consequently be sought not alone in its functions as a collectivity but also in the services and satisfactions of the relationships between its constituent members.
The relationship between father and mother in the nuclear family is solidified by the sexual privilege which all societies accord to married spouses. As a powerful impulse, often pressing individuals to behavior disruptive of the cooperative relationships upon which human social life rests, sex cannot safely be left without restraints. All known societies, consequently, have sought to bring its expression under control by surrounding it with restrictions of various kinds 6. On the other hand, regulation must not be carried to excess or the society will suffer through resultant personality maladjustments or through insufficient reproduction to maintain its population. All peoples have faced the problem of reconciling the need of control with the opposing need of expression, and all have solved it by culturally defining a series of sexual taboos and permissions. These checks and balances differ widely from culture to culture, but without exception a large measure of sexual liberty is everywhere granted to the married parents in the nuclear family. Husband and wife must adhere to sexual etiquette and must, as a rule, observe certain periodic restrictions such as taboos upon intercourse during menstruation, pregnancy, and lactation, but normal sex gratification is never permanently denied to them.
This sexual privilege should not be taken for granted. On the contrary, in view of the almost limitless diversity of human cultures in so many respects, it should be considered genuinely astonishing that some society somewhere has not forbidden sexual access to married partners, confining them, for example, to economic cooperation and allowing each a sexual outlet in some other relationship. As a matter of fact, one of the societies of our sample, the Banaro of New Guinea, shows a remote approach to such an arrangement. In this tribe a groom is not permitted to approach his young wife until she has borne him a child by a special sib-friend of his father. Certain peasant communities in eastern Europe are reported to follow a somewhat analogous custom. A father arranges a marriage for his immature son with an adult woman, with whom he lives and raises children until the son is old enough to assume his marital rights 7. These exceptional cases are especially interesting since they associate sexual rights, not with the husband-wife relationship established by marriage, but with the father-mother relationship established by the foundation of a family.
As a means of expressing and reducing a powerful basic drive, as well as of gratifying various acquired or cultural appetites, sexual intercourse strongly reinforces the responses which precede it. These by their very nature are largely social, and include cooperative acts which must, like courtship, be regarded as instrumental responses. Sex thus tends to strengthen all the reciprocal habits which characterize the interaction of married parents, and indirectly to bind each into the mesh of family relationship in which the other is involved.
To regard sex as the sole factor, or even as the most important one, that brings a man and a woman together in marriage and binds them into the family structure would, however, be a serious error. If all cultures, like our own, prohibited and penalized sexual intercourse except in the marital relationship, such an assumption might seem reasonable. But this is emphatically not the case. Among those of our 250 societies for which information is available, 65 allow unmarried and unrelated persons complete freedom in sexual matters, and 20 others give qualified consent, while only 54 forbid or disapprove premarital liaisons between non-relatives, and many of these allow sex relations between specified relatives such as cross-cousins 8.
Where premarital license prevails, sex certainly cannot be alleged as the primary force driving people into matrimony.
Nor can it be maintained that, even after marriage, sex operates exclusively to reinforce the matrimonial relationship. To be sure, sexual intercourse between a married man and an unrelated woman married to another is forbidden in 126 of our sample societies, and is freely or conditionally allowed in only 24. These figures, however, give an exaggerated impression of the prevalence of cultural restraints against extramarital sexuality, for affairs are often permitted between particular relatives though forbidden with non-relatives. Thus in a majority of the societies in our sample for which information is available a married man may legitimately carry on an affair with one or more of his female relatives, including a sister-in-law , in 41 instances 9. Such evidence demonstrates conclusively that sexual gratification is by no means always confined to the marital relationship, even in theory. If it can reinforce other relationships as well, as it commonly does, it cannot be regarded as peculiarly conducive to marriage or as alone accountable for the stability of the most crucial relationship in the omnipresent family institution.
In the light of facts like the above, the attribution of marriage primarily to the factor of sex must be recognized as reflecting a bias derived from our own very aberrant sexual customs. The authors who have taken this position have frequently fallen into the further error of deriving human marriage from mating phenomena among the lower animals 10. These fallacies were first exposed by Lippert 11 and have been recognized by a number of subsequent authorities 12.
In view of the frequency with which sexual relations are permitted outside of marriage, it would seem the part of scientific caution to assume merely that sex is an important but not the exclusive factor in maintaining the marital relationship within the nuclear family, and to look elsewhere for auxiliary support. One such source is found in economic cooperation, based upon a division of labor by sex 13. Since cooperation, like sexual association, is most readily and satisfactorily achieved by persons who habitually reside together, the two activities, each deriving from a basic biological need, are quite compatible. Indeed, the gratifications from each serve admirably to reinforce the other.
By virtue of their primary sex differences, a man and a woman make an exceptionally efficient cooperating unit 14. Man, with his superior physical strength, can better undertake the more strenuous tasks, such as lumbering, mining, quarrying, land clearance, and housebuilding. Not handicapped, as is woman, by the physiological burdens of pregnancy and nursing, he can range farther afield to hunt, to fish, to herd, and to trade. Woman is at no disadvantage, however, in lighter tasks which can be performed in or near the home, e.g., the gathering of vegetable products, the fetching of water, the preparation of food, and the manufacture of clothing and utensils. All known human societies have developed specialization and cooperation .between the sexes roughly along this biologically determined line of cleavage 15. It is unnecessary to invoke innate psychological differences to account for the division of labor by sex; the indisputable differences in reproductive functions suffice to lay out the broad lines of cleavage. New tasks, as they arise, are assigned to one sphere of activities or to the other, in accordance with convenience and precedent. Habituation to different occupations in adulthood and early sex typing in childhood may well explain the observable differences in sex temperament, instead of vice versa 16.
The advantages inherent in a division of labor by sex presumably account for its universality. Through concentration and practice each partner acquires special skill at his particular tasks. Complementary parts can be learned for an activity requiring joint effort. If two tasks must be performed at the same time but in different places, both may be undertaken and the products shared. The labors of each partner provide insurance to the other. The man, perhaps, returns from a day of hunting, chilled, unsuccessful, and with his clothing soiled and torn, to :find warmth before a :fire which he could not have maintained, to eat food gathered and cooked by the woman instead of going hungry, and to receive fresh garments for the morrow, prepared, mended, or laundered by her hands. Or perhaps the woman has found no vegetable food, or lacks clay for pottery or skins for making clothes, obtainable only at a distance from the dwelling, which she cannot leave because her children require care; the man in his ramblings after game can readily supply her wants. Moreover, if either is injured or ill, the other can nurse him back to health. These and similar rewarding experiences, repeated daily, would suffice of themselves to cement the union. When the powerful reinforcement of sex is added, the partnership of man and woman becomes inevitable.
Sexual unions without economic cooperation are common, and there are relationships between men and women involving a division of labor without sexual gratification, e.g., between brother and sister, master and maidservant, or employer and secretary, but marriage exists only when the economic and the sexual are united into one relationship, and this combination occurs only in marriage. Marriage, thus defined, is found in every known human society. In all of them, moreover, it involves residential cohabitation, and in all of them it forms the basis of the nuclear family. Genuine cultural universals are exceedingly rare. It is all the more striking, therefore, that we here find several of them not only omnipresent but everywhere linked to one another in the same fashion.
Economic cooperation not only binds husband to wife; it also strengthens the various relationships between parents and children within the nuclear family. Here, of course, a division of labor according to age, rather than sex, comes into play. What the child receives in these relationships is obvious; nearly his every gratification depends upon his parents. But the gains are by no means one-sided. In most societies, children by the age of six or seven are able to perform chores which afford their parents considerable relief and help, and long before they attain adulthood and marriageability they become economic assets of definite importance. One need only think here of the utility of boys to their fathers and of girls to their mothers on the typical European or American farm. Moreover, children represent, as it were, a sort of investment or insurance policy; dividends, though deferred for a few years, are eventually paid generously in the form of economic aid, of support in old age, and even, sometimes, of cash returns, as where a bride-price is received for a daughter when she marries.
Siblings 17 are similarly bound to one another through the care and help given by an elder to a younger, through cooperation in childhood games which imitate the activities of adults, and through mutual economic assistance as they grow older. Thus through reciprocal material services sons and daughters are bound to fathers and mothers and to one another, and the entire family group is given firm economic support.
Sexual cohabitation leads inevitably to the birth of offspring. These must be nursed, tended, and reared to physical and social maturity if the parents are to reap the afore-mentioned advantages. Even if the burdens of reproduction and child care outweigh the selfish gains to the parents, the society as a whole has so heavy a stake in the maintenance of its numbers, as a source of strength and security, that it will insist that parents fulfill these obligations. Abortion, infanticide, and neglect, unless confined within safe limits, threaten the entire community and arouse its members to apply severe social sanctions to the recalcitrant parents. Fear is thus added to self-interest as a motive for the rearing of children. Parental love, based on various derivative satisfactions, cannot be ignored as a further motive; it is certainly no more mysterious than the affection lavished by many people on burdensome animal pets, which are able to give far less in return. Individual and social advantages thus operate in a variety of ways to strengthen the reproductive aspects of the parent-child relationships within the nuclear family.
The most basic of these relationships, of course, is that between mother and child, since this is grounded in the physiological facts of pregnancy and lactation and is apparently supported by a special innate reinforcing mechanism, the mother's pleasure or tension release in suckling her infant. The father becomes involved in the care of the child less directly, through the sharing of tasks with the mother. Older children, too, frequently assume partial charge of their younger siblings, as a chore suited to their age. The entire family thus comes to participate in child care, and is further unified through this cooperation.
No less important than the physical care of offspring, and probably more difficult, is their social rearing. The young human animal must acquire an immense amount of traditional knowledge and skill, and must learn to subject his inborn impulses to the many disciplines prescribed by his culture, before he can assume his place as an adult member of his society. The burden of education and socialization everywhere falls primarily upon the nuclear family, and the task is, in general, more equally distributed than is that of physical care. The father must participate as fully as the mother because, owing to the division of labor by sex, he alone is capable of training the sons in the activities and disciplines of adult males 18. Older siblings, too, play an important role, imparting knowledge and discipline through daily interaction in work and play. Perhaps more than any other single factor, collective responsibility for education and socialization welds the various relationships of the family firmly together.
In the nuclear family or its constituent relationships we thus see assembled four functions fundamental to human social life-the sexual, the economic, the reproductive, and the educational. Without provision for the first and third, society would become extinct; for the second, life itself would cease; for the fourth, culture would come to an end. The immense social utility of the nuclear family and the basic reason for its universality thus begin to emerge in strong relief.
Agencies or relationships outside of the family may, to be sure, share in the fulfillment of any of these functions, but they never supplant the family. There are, as we have seen, societies which permit sexual gratification in other relationships, but none which deny it to married spouses. There may be extraordinary expansion in economic specialization, as in modem industrial civilization, but the division of labor between man and wife still persists. There may, in exceptional cases, be little social disapproval of childbirth out of wedlock, and relatives, servants, nurses, or pediatricians may assist in child care, but the primary responsibility for bearing and rearing children ever remains with the family. Finally, grandparents, schools, or secret initiatory societies may assist in the educational process, but parents universally retain the principal role in teaching and discipline. No society, in short, has succeeded in finding an adequate substitute for the nuclear family, to which it might transfer these functions. It is highly doubtful whether any society ever will succeed in such an attempt, utopian proposals for the abolition of the family to the contrary notwithstanding.
The above-mentioned functions are by no means the only ones performed by the nuclear family. As a firm social constellation, it frequently, but not universally, draws to itself various other functions. Thus it is often the center of religious worship, with the father as family priest. It may be the primary unit in land holding, vengeance, or recreation. Social status may depend more upon family position than upon individual achievement. And so on. These additional functions, where they occur, bring increased strength to the family, though they do not explain it.
Like the community, the nuclear family is found in sub-human societies, although here the father is less typically a member and, where he is, is usually less firmly attached. But man's closest animal relatives possess, at best, only a rudimentary division of labor by sex. and they seem to lack culture altogether. The universal participation of the father in the human family would thus seem to depend mainly upon economic specialization and the development of a body of traditional lore to be transmitted from one generation to the next. Since both are products of cultural evolution—indeed, amongst the earliest of such—the human family cannot be explained on an instinctive or hereditary basis.
This universal social structure, produced through cultural evolution in every human society as presumably the only feasible adjustment to a series of basic needs, forms a crucial part of the environment in which every individual grows to maturity. The social conditions of learning during the early formative years of life, as well as the innate psychological mechanism of learning, are thus essentially the same for all mankind. For an understanding of the behavior acquired under such conditions the participation of the social scientist would seem as essential as that of the psychologist. It is highly probable, for instance, that many of the personality manifestations studied by depth psychology are rooted in a combination of psychological and social-cultural constants. Thus the “OEdipus complex” of Freud seems comprehensible only as a set of characteristic behavioral adjustments made during childhood in the face of a situation recurrently presented by the nuclear family 19.
Perhaps the most striking effect of family structure upon individual behavior is to be observed in the phenomenon of incest taboos. Since Chapter 10 will present in extenso a theory of the genesis and extension of incest taboos, we need not concern ourselves here with explanations. The essential facts, however, must be stated at this point, since an understanding of them is absolutely crucial to the further analysis of social structure. Despite an extraordinary variability and seeming arbitrariness in the incidence of incest taboos in different societies, they invariably apply to every cross-sex relationship within the nuclear family save that between married spouses. In no known society is it conventional or even permissible for father and daughter, mother and son, or brother and sister to have sexual intercourse or to marry. Despite the tendency of ethnographers to report marriage rules far more fully than regulations governing premarital and postmarital incest, the evidence from our 250 societies, presented in Table 1, is conclusive.
The few apparent exceptions, in each instance too partial to appear in the table, are nevertheless illuminating, and all those encountered will therefore be mentioned. Certain high Azande nobles are permitted to wed their own daughters, and brother-sister marriages were preferred in the old Hawaiian aristocracy and in the Inca royal family. In none of these instances, however, could the general population contract incestuous unions, for these were a symbol and prerogative of exalted status. Among the Dobuans, intercourse with the mother is not seriously regarded if the father is dead; it is considered a private sin rather than a public offense. The Balinese of Indonesia permit twin brothers and sisters to marry on the ground that they have already been unduly intimate in their mother's womb. Among the Thonga of Africa an important hunter, preparatory to a great hunt, may have sex relations with his daughter — a heinous act under other circumstances. By their special circumstances or exceptional character these cases serve rather to emphasize than to disprove the universality of intra-family incest taboos.

Table 1
  Premarital Intercourse Postmarital Intercourse Marriage
Relative (of man) Forbidden Permitted Forbidden Permitted Forbidden Permitted
Mother 76 0 74 0 184 0
Sister 109 0 106 0 237 0
Daughter 81 0 198 0

The first consequence of these taboos is that they make the nuclear family discontinuous over time and confine it to two generations. If brother-sister marriages were usual, for example, a family would normally consist of married grandparents, their sons and daughters married to one another, the children of the latter, and even the progeny of incestuous unions among these. The family, like the community, the clan, and many other social groups, would be permanent, new births ever filling the gaps caused by deaths. Incest taboos completely alter this situation. They compel each child to seek in another family for a spouse with whom to establish a marital relationship. In consequence thereof, every normal adult in every human society belongs to at least two nuclear families—a family of orientation in which he was born and reared, and which includes his father, mother, brothers, and sisters, and a family of procreation 20 which he establishes by his marriage and which includes his husband or wife, his sons, and his daughters.
This situation has important repercussions upon kinship. In a hypothetical incestuous family, it would be necessary only to differentiate non-members from members and to classify each according to age and sex. An extremely simple kinship system would suffice for all practical needs. Incest taboos, however, create an overlapping of families and arrange their members into different degrees of nearness or remoteness of relationship. A person has his primary relatives 21—his parents and siblings in his family of orientation and his spouse and children in his family of procreation. Each of these persons has his own primary relatives, who, if they are not similarly related to Ego, rank as the latter's secondary relatives, e.g., his father's father, his mother's sister, his wife's mother, his brother's son, and his daughter's husband. The primary relatives of secondary kinsmen are Ego's tertiary relatives, such as his father's sister's husband, his wife's sister's daughter, and any of his first cousins. This stepwise gradation of kinsmen extends indefinitely, creating innumerable distinct categories of genealogical connection. To avoid an impossibly cumbersome system of nomenclature, every society has found it necessary to reduce the total number of kinship terms to manageable proportions by applying some of them to different categories of relatives. The principles governing this reduction, and the types of kinship structure which result under different social conditions, will be analyzed in Chapters 6 and 7.
Some of the intimacy characteristic of relationships within the nuclear family tends to flow outward along the ramifying channels of kinship ties. A man ordinarily feels closer, for example, to the brothers of his father, of his mother, and of his wife than to unrelated men in the tribe or the community. When he needs assistance or services beyond what his family of orientation or his family of procreation can provide, he is more likely to tum to his secondary, tertiary, or remoter relatives than to persons who are not his kinsmen. But to which of these relatives shall he turn? Owing to the rami£cation of kinship ties which results from incest taboos, a person may have 33 different types of secondary relatives and 151 different types of tertiary relatives, and a single type, such as father's brother, may include a number of individuals. All societies are faced with the problem of establishing priorities, as it were, i.e., of defining for individuals the particular group of kinsmen to whom they are privileged to turn fust for material aid, support, or ceremonial services. All cultures meet this problem by adopting a rule of descent.
A rule of descent affiliates an individual at birth with a particular group of relatives with whom he is especially intimate and from whom he can expect certain kinds of services that he cannot demand of non-relatives, or even of other kinsmen. The fundamental rules of descent are only three in number:

  • patrilineal descent, which affiliates a person with a group of kinsmen who are related to him through males only;
  • matrilineal descent, which assigns him to a group consisting exclusively of relatives through females
  • bilateral descent 22, which associates him with a group of very close relatives irrespective of their particular genealogical connection to him
  • double descent 23, combines patrilineal and matrilineal descent by assigning the individual to a group of each type.

An earlier generation of anthropologists completely misunderstood rules of descent, assuming that they meant a recognition of certain genealogical ties to the exclusion of others, e.g., that a matrilineal people is either ignorant of, or chooses to ignore, the biological relationship of a child to its father. Science owes a debt to Rivers 24 for pointing out that descent refers only to social allocation and has fundamentally nothing to do with genealogical relationships or the recognition thereof. It is now known that the Hopi and most other societies with matrilineal descent do not deny or ignore the relationship of a child to its father and his patrilineal kinsmen; frequently, indeed, they specifically recognize it by forbidding marriage with paternal as well as maternal relatives. A comparable situation prevails among patrilineal peoples. A number of Australian tribes, indeed, actually follow patrilineal descent while specifically denying the existence of any biological tie between father and child. In Africa and elsewhere, moreover, it is common for the illegitimate children of a married woman by another man to be unquestioningly affiliated by patrilineal descent with her husband, their “sociological father.”
Descent, in fine, does not necessarily involve any belief that certain genealogical ties are closer than others, much less a recognition of kinship with one parent to the exclusion of the other, although such notions have been reported in exceptional cases. It merely refers to a cultural rule which affiliates an individual with a particular selected group of kinsmen for certain social purposes such as mutual assistance or the regulation of marriage. The various types of kin groups which result from the several rules of descent will be analyzed in Chapter 3.
The incest taboos which regularly prevail within the nuclear family exert still another extremely important effect on social organization. In conjunction with the universal requirement of residential cohabitation in marriage, they result inevitably in a dislocation of residence whenever a marriage occurs. Husband and wife cannot both remain with their own families of orientation in founding a new family of procreation. One or the other, or both, must move. The possible alternatives are few in number, and all societies have come to adopt one or another of them, or some combination thereof, as the culturally preferred rule of residence.

  1. If custom requires the groom to leave his parental home and live with his bride, either in the house of her parents or in a dwelling nearby, the rule of residence is called matrilocal.
  2. If, on the other hand, the bride regularly removes to or near the parental home of the groom, residence is said to be patrilocal. It should be emphasized that this rule implies, not merely that a Wife goes to live with her husband, but that they establish a domicile in or near the home of his parents.
  3. Some societies permit a married couple to live with or near the parents of either spouse, in which case such factors as the relative wealth or status of the two families or the personal preferences of the parties to the union are likely to determine whether they will choose to reside matrilocally or patrilocally. The rule of residence in such cases is termed bilocal 25.
  4. When a newly wedded couple, as in our own society, establishes a domicile independent of the location of the parental home of either partner, and perhaps even at a considerable distance from both, residence may be called neolocal. In the ethnographic literature this rule is unfortunately often confused with patrilocal residence.
  5. A fifth alternative, which we shall term avunculocal 26 residence, prevails in a few societies which prescribe that a married couple shall reside with or near a maternal uncle of the groom rather than with the parents of either spouse or in a separate home of their own.

Though other rules of residence are theoretically possible, the five alternatives described above, either alone or in combination, cover all the cases actually encountered in our 250 sample societies. The Dobuans of Melanesia reveal a special combination of matrilocal and avunculocal residence whereby the two rules alternate with one another, periodically, throughout the married life of a couple. A more frequent compromise consists in requiring matrilocal residence for an initial period, usually for a year or until the birth of the first child, to be followed by permanent patrilocal residence. For this combination, which is really only a special variant of patrilocal residence, we propose the term matri-patrilocal as preferable to “intermediate” or “transitional” residence 27. The distribution of these various rules among our 250 societies is as follows: 146 patrilocal, 38 matrilocal, 22 matri-patrilocal, 19 bilocal, 17 neolocal, and 8 avunculocal. It is probable, however, that some of the tribes reported as patrilocal actually follow the neolocal rule.
Rules of residence reflect general economic, social, and cultural conditions. When underlying conditions change, rules of residence tend to be modified accordingly. The local alignment of kinsmen is thereby altered, with the result that a series of adaptive changes is initiated which may ultimately produce a reorganization of the entire social structure. The fundamental role of residence rules in the evolution of social organization is demonstrated in Chapter 8.
The primary effect of a rule of residence is to assemble in one locality a particular aggregation of kinsmen with their families of procreation. Patrilocal and matri-patrilocal residence bring together a number of patrilineally related males with their wives and children. Matrilocal and avunculocal residence aggregate matrilineal kinsmen and their families. Bilocal residence produces a local alignment of bilateral relatives. Out of such local clusters of kinsmen may arise two major types of social groups-extended families and clans. These will be considered, respectively, in Chapters 2 and 4. Neolocal residence is the only rule that definitely militates against the development of such larger aggregations of kinsmen.
The spouse who does not have to shift his residence in marriage enjoys certain advantages over the other. He or she remains at home in familiar physical and social surroundings, and his family of procreation can maintain a close connection with his family of orientation. The other spouse, however, must break with the past in some measure, and establish new social ties. This break is not serious where marriages are normally contracted within the same local community, for the spouse who has shifted residence still remains close enough to his family of orientation to continue daily face-toface associations. Where marriages are exogamous 28 with respect to the community, however, spouses of one sex find themselves living among comparative strangers, to whom they must make new personal adjustments and upon whom they must depend for the support, protection, and social satisfactions which they have previously received from relatives and old friends. They thus find themselves at a considerable psychological and social disadvantage in comparison with the sex which remains at home.
Although a change in community is theoretically possible under any rule of residence, the data from our sample reveal that it is rarely customary except when residence is patrilocal, matri-patrilocal, or avunculocal, i.e., when it is regularly the woman rather than the man who changes domicile. The evidence is presented in Table 2. While male dominance in consequence of physical superiority may be partially responsible for the vastly greater frequency with which women move to a new community in marriage, the author is inclined, with Linton 29, to seek the explanation mainly in economic factors, particularly those which derive from the division of labor by sex. These will be considered fully in Chapter 8 in connection with the exceedingly important bearing of the facts in question upon the evolution of social organization.

Table 2
Rule of Residence Community Exogamy or Tendency Thereto Local Community Neither Exogamous nor Endogamous Community Endogamy or Tendency Thereto
Patrilocal 54 40 7
Matri-patrilocal 5 6 2
Avunculocal 4 2 0
Bilocal 2 9 3
Neolocal 0 9 4
Matrilocal 5 2 17

Rules of residence, and especially the extent to which they involve a shift to a new community by the woman, are significantly related to the modes of contracting marriages. If a bride leaves her parental home when she marries, her nuclear family parts with a productive worker. Her parents, in particular, lose a potential source of help and support in their declining years. Moreover, they dispense with the potential assistance of a son-in-law, who would live with them and work for them if residence were matrilocal. Small wonder, then, if parents consent to allow a daughter to leave them in patrilocal or avunculocal marriage only if they receive some substantial compensation.
If a woman, even though she removes to her husband's home when she marries, still remains in the same community, the loss to her parents is less severe. The help and support which she and her husband can still give them may even amount to full compensation. If, however, she removes to another settlement, some form of remuneration becomes almost inevitable.
The modes of contracting marriage fall into two major classes: those with and those without consideration 30. Where a consideration is required, it may be rendered either in goods, or in kind, or in services, resulting respectively in the payment of a bride-price, in the exchange of a sister or other female relative for a wife, and in bride-service performed for the parents-in-law. Where no consideration is demanded, a marriage may be solemnized by an exchange of gifts of approximately equal value between the families of the contracting parties, or the bride may be released with a dowry of valuable goods, or the wedding may be devoid of any property transactions and be contracted through wife-capture, elopement, or a relatively informal initiation of cohabitation. Wife-capture is exceedingly rare as a normal mode of marriage, not appearing as such in any of the societies of our sample, and elopements are usually later legitimized by the performance of the customary ceremonies and property transactions.
The data from our 250 societies, as compiled in Table 3, reveal that some form of consideration ordinarily accompanies marriage when residence rules remove the bride from her home. The payment of a bride-price and the giving of a sister or other female relative in exchange for a wife are almost exclusively associated with patrilocal residence. Bride-service normally accompanies matri-patrilocal residence and is also common under matrilocal residence, from which it is not always clearly differentiated in the literature.

Table 3
  Marriages with Consideration  
Rule of Residence Bride Price Exchange of Women Bride Service Marriages
with Consideration
Patrilocal 103 10 2 25 140
Matri-patrilocal 6 0 13 2 21
Avunculocal 2 0 4 2 8
Bilocal 6 2 1 10 19
Neolocal 0 0 2 15 17
Matrilocal 4 0 8 24 36
Totals 121 12 30 78 241

Payment of a bride-price or other consideration is particularly common under patrilocal or matri-patrilocal residence when the bride is removed not only from her home but from the local community in which she 1 has grown up. The evidence, excluding other rules of residence, is shown in Table 4.

Table 4
  Marriages with Consideration  
Bride Is Removed from Her Local Community Bride Price Exchange of Women Bride Service Marriages without Consideration
Usually or always 44 6 6 4
Sometimes or often 27 2 4 8
Rarely or never 5 0 1 6

Individual cases are often illuminating. Among the Abelam, for example, the usual bride-price is omitted when the married couple come from the same hamlet. Among the Copper Eskimo, if a daughter remains in the community after marriage, she and her husband render assistance to her parents and no bride-price is required, but in the exceptional case of a patrilocal inter-village marriage the groom must compensate his bride's parents for their loss. Among the patrilocal Hupa and Yurok of northwestern California, a man who cannot afford the full bride-price pays half of it and resides with and works for his father-in-law in what is called “half-marriage.” Eight of our societies reveal an adjustment which is especially common in Indonesia and is there called ambil-anak 31. In these societies, when a family has daughters but no sons, an exception to the customary patrilocal rule is made in the case of one daughter, who is married without the usual bride-price and whose husband comes to reside matrilocally with her parents and takes the place of a son.
A bride-price is more than a compensation to parents for the loss of a daughter who leaves their home when she marries. It is commonly also a guarantee that the young wife will be well treated in her new home. If she is not, she can ordinarily return to her parents, with the result that her husband forfeits his financial investment in her. That these are the true functions of the bride-price has been pointed out by a generation of anthropologists, and students of ancient civilizations have reached the same conclusion 32. Seldom if ever is it regarded as a price paid for a chattel, or as comparable to the sum paid for a slave.
Because the conclusion of marriage is the regular means of establishing a nuclear family, because the marital relationship forms the very warp of the family fabric, and because the regulation of marriage through sex taboos produces far-reaching effects upon family structure itself, it has been necessary to consider these aspects of marriage. Others will be treated in later chapters, to illuminate other features of social structure. For wedding ceremonial, for divorce usages, and for other non-structural aspects of marriage, however, the reader must be referred to the works of Westermarck and other standard authorities.

.1. The terms “polygamy” and “polygamous” will be used throughout this work in their recognized technical sense as referring to any form of plural marriage; “polygyny” will be employed for the marriage of one man to two or more women, and “polyandry” for the marriage of one woman to two or more men.
2. Cf. M. K. Opler, “Woman's Social Status and the Forms of Marriage,” American Journal of Sociology, XLIX ( 1943), 144; A. R. Radcliffe-Brown, “The Study of Kinship Systems,” Journal of the Royal Anthropological Institute, LXXI (1941), 2.
3. R. H. Lowie, Primitive Society (New York, 1920), pp. 66-7. Cf. also F. Boas et al, General Anthropology (Boston, etc., 1938), p. 411; B. Malinowski, “Kinship,” Encyclopaedia Britannica (14th edit., London, 1929), XIII, 404.
4. R. Linton, The Study of Man (New York, 1936), pp. 153 (quoted), 154-5.
5. See Chapter 6 for a more detailed consideration of these relationships in connection with kinship structure.
6. These will be considered in detail in Chapter 9.
7. Cf. R. F. Kaindl, “Aus der Volksiiberlleferung der Bojken,” Globus, LXXIX (1901), 155.
8. A cross-cousin is the child of a father's sister or of a mother's brother. The children of a father's brother and of a mother's sister are technically known as “parallel cousins.”
9. For detailed information see Chapter 9.
10. See, for example, E. Westermarck, The History of Human Marriage (5th edit., New York, 1922), I, 72; A. M. Tozzer, Social Origins and Social Continuities (New York, 1925), p. 145.
11. J. Lippert, Kulturgeschichte der Menschheit in ihrem organischen Aufbau (Stuttgart, 1886-87), I, 70-4; II, 5.
12. See, for example, R. Briffault, The Mothers (New York, 1927), I, 608; W. G. Sumner and A. G. Keller, The Science of Society (New Haven, 1927), III, 1495--8, 1517; P. Vinogradoff, Outlines of Historical]urisprudence, I (New York, 1920), 203.
13. See W. G. Sumner and A. G. Keller, The Science of Society (New Haven, 1927), III, 1505-18.
14. Ibid., I, 111-40.
15. See G. P. Murdock, “Comparative Data on the Division of Labor by Sex,” Social Forces, XV ( 1937), 551-3, for an analysis of the distribution of economic activities by sex in 224 societies.
16. Cf. M. Mead, Sex and Temperament in Three Primitive Societies (New York, 1935).
17. The term “sibling” will be employed throughout this work in its technical sense as designating either a brother or a sister irrespective of sex.
18. Cf. R. Linton, The Study of Man (New York, 1936), p. 155.
19. Unlike other psychological systems, that of Freud thus rests on cultural as well as physiological assumptions. See G. P. Murdock, “The Common Denominator of Cultures,” The Science of Man in the World Crisis, ed. R. Linton (New York, 1945), p. 141.
20. For these very useful terms we are indebted to W.L. Warner.
21. The terms “primary, secondary, and tertiary relatives” are adopted from A. R. Radcliffe-Brown, “The Study of Kinship Systems,” Journal of the Royal Anthropological Institute, LXXI ( 1941), 2.
22. It has been suggested that this might more accurately be called “multilineal descent.” See T. Parsons, “The Kinship System of the Contemporary United States,” American Anthropologist, n.s., XLV (1943), 26.
23. See G. P. Murdock, “Double Descent.” American Anthropologist, n.s., XLII ( 1940), 555-61.
24. W. H. R Rivers, Social Organization (New York, 1924), p. 86. See also B.Z. Seligman, “Incest and Descent.” Journal of the Royal Anthropological Institute, LIX ( 1929), 248.
25. See E.A. Hoebel, “Comanche and Hekandika Shoshone Relationship Systems,” American Anthropologist, n.s., XLI (1939), 446. The term “ambilocal” has also been suggested for this rule.
26. This term is proposed in A. L. Kroeber, “Basic and Secondary Patterns of Social Structure,” Journal of the Royal Anthropological Institute, LXVII ( 1938). 301. For several years previously, however, it bad been used in classroom lectures by the present author—an interesting instance of parallel invention.
27. Cf. E.B. Tylor, “On a Method of Investigating the Development of Institutions,” Journal of the Royal Anthropological Institute, XVIll {1889), 2.45-69.
28. The term “exogamy” refers to a rule of marriage which forbids an individual to take a spouse from within the locaL kin, or status group to which he himself belongs. The complementary term “endogamy,” refers to a rule which requires him to marry within his own group.
29. R. Linton, The Study of Man (New York, 1936), p. 165.
30. Cf. E. Westermarck, The History of Human Marriage (5th edit., New York, 1922), n, 354-481.
31. For an extended discussion, with cases, see W. G. Sumner and A. G. Keller. The Science of Society (New Haven, 1927), ill, 1654-8.
32. See, for example, M. Burrows, “The Basis of Israelite Marriage,” American Oriental Series, XV (1938), 1-72.

Composite Forms of the Family

In a minority of societies, including our own, each nuclear family stands alone as a sort of independent atom in the community, as a unit separate from all others of its kind. In the great majority, however in 140 out of the 187 in our sample for which data are availablenuclear families are aggregated, as it were, into molecules. Clusters of two, three, or more are united into larger familial groups which commonly reside together and maintain closer ties with one another than with other families in the community. Physically, such composite family groups usually form a household, marked by joint occupancy either of a single large dwelling, or of a cluster of adjacent huts, or of a well-defined compound. Socially, the nuclear families thus associated are almost invariably linked to one another not only by the bond of common residence but also through close kinship ties. Unrelated or distantly related families, to be sure, occasionally form a common household, but this is not the usual practice in any of the societies studied.
The two principal types of composite families have already been briefly noted. The polygamous family, it will be recalled, consists of several nuclear families linked through a common spouse. The extended family includes two or more nuclear families united by consanguineal kinship bonds such as those between parent and child or between two siblings. In a hypothetical but typical case, for instance, such a group might embrace the families of procreation of a father and his two adult sons. Each son would be a member of two nuclear families—his own and his father's—and would have a primary relative, his brother, in the third. The father and mother would each, of course, have a primary relative, a son, in the two nuclear families of which they were not themselves members. While the norm naturally does not prevail in all individual cases, the evidence is nevertheless overwhelming that the characteristic and crucial bond of union in composite families of every type is the existence of close kinship ties linking members of the constituent nuclear families.
The polygamous family can appear, of course, only in societies which permit plural marriages. It cannot coexist with strict monogamy, which prevails in 43 societies in our sample as opposed to 195 which allow at least limited plural marriage. Theoretically, polygamy can assume any one of three possible forms:

  • polygyny or the marriage of one man to two or more wives at a time
  • polyandry or the coexistent union of one woman with two or more men
  • group marriage or a marital union embracing at once several men and several women.

Of these, only the first is common. Group marriage, though figuring prominently in the early theoretical literature of anthropology 1, appears never to exist as a cultural norm. It occurs in but a handful of the tribes in our sample, and then only in the form of highly exceptional individual instances. Its most frequent occurrence is among the Kaingang of Brazil, a tribe with exceedingly lax and fluid sexual associations. A statistical analysis 2 of Kaingang genealogies for a period of 100 years showed that 8 per cent of all recorded unions were group marriages, as compared with 14 per cent for polyandrous, 18 per cent for polygynous, and 60 per cent for monogamous unions. Even the Chukchee of Siberia and the Dieri of Australia, the societies to which the institution has been most frequently ascribed, cannot truly be said to practice group marriage. These tribes and others, to be sure, often extend sexual privileges to a group of males and females, but never the economic responsibilities upon which genuine marriage must likewise always rest. In fine, there is no evidence that group marriage anywhere exists, or ever has existed, as the prevailing type of marital union 3.
The polyandrous family occurs so rarely that it may be regarded as an ethnological curiosity. In only two of the tribes in our sample, the Marquesans of Polynesia and the Todas of India, is polyandry the normal and preferred form of marriage, although it appears in sporadic instances in a few other societies. Despite the paucity of cases there seems reason to assume that polyandry may sometimes be due to a scarcity of women resulting from the practice of female infanticide 4. Whatever its causes, however, polyandry is so infrequent a phenomenon that there is no justification for assigning to it, as was done by McLennan 5, an important place in the evolution of social organization.
Recent ethnographers 6 have shown an unfortunate tendency to apply the term “polyandry” to sporadic instances of the association of several men with one woman in contravention of cultural norms, or to cases where a woman enjoys sexual privileges with the brothers of her husband although she does not cooperate economically with them. The extension of the sexual rights of either partner in a marital union to his siblings-in-law of opposite sex is by no means a rare phenomenon. It was found by the present writer in all three of the tribes among which he has done personal field work, namely, the Haida, Tenino, and Trukese, and is attested for 41 societies in our sample, or considerably more than half of those for which pertinent information is available. Such sexual privileges, however, by no means constitute marriage. Cooper 7 is thus correct in his insistence that the term “polyandry” be reserved exclusively for a form of marriage which is socially sanctioned and culturally patterned and which involves economic cooperation and residential cohabitation as well as sexual rights. More recently the whole problem has been clarified in admirable fashion by Marvin Opler 8.
The polyandrous family may appear in two different forms, a non-fraternal and a fraternal form. Non-fraternal polyandry is exemplified by the Marquesans, among whom a number of unrelated men join the household of a woman of high status and participate jointly in economic responsibilities and sexual privileges. Fraternal polyandry is the usual form among the Toda, although they occasionally practice non-fraternal polyandry as well. Several brothers establish a common household with one woman. The exact paternity of her children is a matter of indifference and is established legally by a rite during her first pregnancy in which one brother presents her with a toy bow and arrow. In some subsequent pregnancy another brother may perform the ceremony. A child is always ascribed to the last husband to execute the rite, even though he may have been dead for years. It is perhaps significant that co-husbands among the Todas occupy one house when they are brothers, but in the occasional instances of non-fraternal polyandry they maintain separate dwellings where they are visited in rotation by their common wife 9.
A definition of polygyny is indispensable to an analysis of the polygynous family. Polygyny can be said to exist, of course, only when the plural marriages are contemporaneous, for if they are successive the second spouse is a stepparent or sociological parent to any children of the first who remain at home, and the structure is that of a nuclear family. Secondly, all the unions must be genuine marriages, involving residential cohabitation and economic cooperation as well as sexual association. Hence concubinage, even when culturally permissible, is to be distinguished from polygyny when it does not meet the economic criterion of marriage, and its presence is not necessarily inconsistent with true monogamy. Finally, the unions must have the support of culture and public opinion. The occurrence of occasional plural unions in defiance of law and custom, like bigamy and the maintenance of mistresses and their children in our own society, does not suffice to establish polygyny. Whenever these criteria are not met, a society must be classed as monogamous.
Granted that plural unions are not expressly forbidden by cultural taboos, how much social sanction do they require, and how numerous must they be, in order to justify speaking of the society as polygynous? With respect to the degree of social sanction, it is here suggested that a society be characterized as polygynous if plural unions enjoy superior prestige, as compared with monogamy, so that successful males in the society will seek to acquire secondary wives if they can. We have consequently classified as monogamous a few tribes in which plural unions occur occasionally but without indication that they are preferred. Most of these are instances in which a special exception is made for men whose first wives prove barren. Not only do such cases suggest no preference for polygyny but a secondary marriage under these conditions does not actually create a composite family; the first wife is an appendage to the nuclear family like an aged parent or an unmarried sibling.
A numerical criterion for polygyny is unsatisfactory, since monogamous unions nearly always outnumber polygynous ones at any given period of observation, even in societies where the preference for plural wives is extreme. This is assured by the natural sex ratio; except under extraordinary conditions and for short periods the number of females in a population cannot greatly exceed the number of males. Hence, roughly speaking, for every man who has two wives there must be another who lives in enforced celibacy. To be sure, this situation is rectified in part by the higher mortality of males and by the differential age of marriage, which is usually appreciably lower for females. Nevertheless, even in strongly polygynous societies it is principally the men who have reached a relatively ripe age that I are found to have several wives, the majority of males having to content themselves with one wife or none. Polygyny may be the ideal, (and it may even be attainable for most men, but they must ordinarily (wait for years before they can expect to purchase or inherit a second wife. An impartial observer employing the criterion of numerical preponderance, consequently, would be compelled to characterize nearly every known human society as monogamous, despite the preference for and frequency of polygyny in the overwhelming majority.
Rejecting the numerical criterion, we shall classify a society as polygynous whenever the culture permits, and public opinion encourages, a man to have more than one wife at a time, whether such unions are common or comparatively rare, confined to men of outstanding prestige or allowed to anyone who can afford them. On this basis, 193 societies in our sample are characterized by polygyny, and only 43 by monogamy and 2 by polyandry.
In a number of societies which are polygynous according to the above definition, monogamous unions may actually be considerably commoner for men of all ages because economic conditions strongly favor them. Only a few unusually energetic or capable men, perhaps, can successfully support two families. In other cases, polygyny is confined largely to chiefs or men of wealth and status, or is limited in other ways. So long, however, as it enjoys superior prestige and is not the exclusive prerogative of a very small status group, it is assumed to be the cultural norm.
The numerical problem shifts slightly in reference to the polygynous family. Here the issue is less what is culturally permitted or preferred than what is the normal social structure under which the majority of the population actually live. As will be seen in Chapter 8, monogamous and polygynous families exert a differential influence on kinship terminology and other aspects of social structure. Despite the superior prestige of plural marriages, the polygynous family cannot be expected to exercise a preponderant influence if it actually occurs only infrequently.In the case of polygynous societies where plural marriages are in actuality quite rare, consequently, we have chosen to designate the prevailing form of the family as monogamous rather than as polygynous. The dividing line has been set arbitrarily at 20 per cent. Any society in which plural unions, however strongly favored, actually occur in less than this percentage of all marriages, is considered monogamous with respect to the family though polygynous with respect to marriage. Of the 193 polygynous societies in our sample, 61 appear to fall below the 20 per cent limit.
In connection with polygyny, it becomes important to draw a distinction between primary and secondary marriages. A primary marriage is typically the first union which an individual contracts; a secondary marriage is any subsequent union. The distinction acquires importance from the fact that many cultures prescribe quite different rules in the two cases. The regulations governing primary marriages ordinarily exclude certain relatives as too closely akin and frequently define other relatives, e.g., cross-cousins, as preferred mates. In secondary marriages, though the choice of a spouse is sometimes wider, it is often much more limited. In particular, the relationship of the second to the first spouse may be a determining factor. This is true both of successive marriages following the death or divorce of previous spouses, and of polygamous unions.
The principal special regulations governing secondary marriages are the levirate and the sororate. The levirate is a cultural rule prescribing that a widow marry by preference the brother of her deceased husband, thus often becoming his secondary spouse as he is hers. When applied not to successive but to contemporary marriages, the levirate rule results in fraternal polyandry as it has been described above. The sororate, conversely, is a rule favoring the marriage of a widower with the sister of his deceased wife. When applied to plural rather than to successive marriages, it yields sororal polygyny or the preferred union of one man with two or more sisters 10. These terms should not be applied to societies where levirate or sororate unions occur only occasionally or incidentally, as they do for instance even amongst ourselves, but only when they are common and genuinely preferential.
Both the levirate and the sororate are exceedingly widespread phenomena. The former is reported to be present in a definitely preferential form in 127 of our sample societies, and to be absent or merely occasional in only 58, with no data on the remaining 65. The preferential sororate is reported present in 100 societies, absent in 59. Frequently levirate marriage is permitted only with a younger brother of the deceased husband, and the sororate only with a deceased wife's younger sister. The “junior levirate” and “junior sororate” are specified for 28 and 9 societies, respectively, among those enumerated above, and these numbers would doubtless be larger if information were complete. Other preferred secondary marriages, resembling the levirate and sororate but much less common than either, are those of a man with the widow of his paternal or maternal uncle, with his father's widow other than his own mother, and with the daughter of his deceased wife's sister or brother.
The polygynous family creates problems of personal adjustments which do not arise under monogamy, notably disputes arising from sexual jealousy and over the distribution of economic tasks in the feminine sphere of activity. A number of cultural solutions are apparent in the data. We have already noted the frequency with which co-wives are assigned separate dwellings. Another common solution is to give one wife, usually the one first married, a superior social status and to delegate to her a general supervisory authority over the feminine tasks of the household. Thirdly, sexual jealousy is commonly allayed by requiring the husband to cohabit with each wife in regular rotation. He may not actually have intercourse with an old or unattractive wife, or with one whom he has come to dislike, but he sleeps in their huts on the nights when their turns come around, and they are spared the mortification of public rejection.
These facts, coupled with information kindly furnished by Kimball Young, suggest some of the reasons for the failure of the Mormon experiment with polygyny in the United States. Wives of disparate origin were kept in one establishment, where nerves were easily frayed by too close contact. Status was insecure, since each successive wife was likely to be for a period the favorite of her husband, and with his support to exercise unwelcome authority over her elder co-wives. Sexually interested in the most recent addition to his menage, a husband tended to cohabit with her to the exclusion of the rest, who were thereby not only physically frustrated but socially humiliated. It was very probably their internal troubles in making the institution operate harmoniously, rather than external pressures, that induced the Mormons ultimately to abandon polygyny. That it can be made to work smoothly is perfectly clear from the evidence of ethnography.
One simple means of reducing the friction among co-wives is to institute sororal polygyny as the preferred form. Sisters, of course, have learned to adapt their behavior to one another in their family of procreation, and they carry their habits of cooperation and mutual tolerance with them when they move from the home of their father to that of their husband. Unlike other co-wives, they do not have to start from scratch, so to speak, in the learning of adaptive behavior. Our data show preferential sororal polygyny to be exceedingly widespread. It is reported for 14 of the 61 societies with permitted but infrequent polygyny. Among the 132 societies with general polygyny, it is the exclusive form in 18 and occurs in 38 others, not exclusively but preferentially, along with non-sororal polygyny. This total of 70 would doubtless be considerably increased if the data were more complete 11.
Since problems of personal adjustment among co-wives who are not sisters to each other become particularly acute when they must live in the same house, it is not surprising to find this fact either specifically or tacitly recognized in primitive societies. Among the Crow, Sinkaietk, and Soga, for example, the ethnographers report that co-wives regularly reside in the same dwelling if they are sisters but occupy separate habitations if they are unrelated. Gross statistics are even more convincing. In 18 of the 21 societies with exclusive sororal polygyny for which information is available, co-wives live together in the same house. In the majority of societies with nonsororal polygyny (28 out of 55) co-wives occupy separate dwellings, and in many of the cases where they share the same house it is a large dwelling in which they are provided with separate apartments. Tribes with both sororal and non-sororal polygyny, as might be expected, hold an intermediate position—16 with single and 8 with plural dwellings.
Sororal polygyny is peculiarly well adapted to matrilocal residence. Non-sororal polygyny, though possible under any other rule of residence, can occur only in exceptional cases when men regularly remove to the homes of their wives upon marriage. Under these circumstances the only women ordinarily available as secondary wives, without violating the residence rule, are the sister:; and other close female relatives of the first wife. Though uniquely suited to matrilocal residence, exclusive sororal polygyny can and does occur under other residence rules. Its distribution in our sample is as follows: present in 12 out of 38 societies with matrilocal residence, in 14 out of 176 with patrilocal, matri-patrilocal, or avunculocal residence, and in 6 out of 36 with bilocal or neolocal residence.
Extended families may be compounded from polygamous families, from monogamous nuclear families, or from both. The polygamous family, when it is absorbed in a larger familial aggregate, may be called the dependent polygamous family. When it stands alone, in the absence of extended families, it may be termed the independent polygamous family. The nuclear family, similarly, may be dependent or independent. In the presence of general polygamy or of extended families which subordinate it in a larger composite family, it may be called the dependent nuclear family. In the absence of either, when it stands alone as the only familial cluster in the society, it may be termed the independent nuclear family. The distribution of these types of family organization is compiled in Table 5 below.

Type of Family Organization Number of Societies
Independent nuclear families
With monogamy 24
With infrequent polygyny 23
Dependent nuclear families
With monogamy 16
With infrequent polygyny
(Also the 98 societies
with polygamous families tabulated below)
Independent polygamous families
With polyandry 2
With polygyny 51
Dependent polygamous families
With polyandry 0
With polygyny 45
Insufficient data for classification 63

Independent nuclear and polygamous families are alike differentiated from extended families by their transitory character.

Both dissolve and disappear as the parents die and the children establish new families of procreation. Normally, therefore, both consist of members of only two generations. Extended families, on the other hand, consist of three or more generations and are characterized by indefinite continuity over time. In societies with independent nuclear or polygamous families, marriage establishes a new family of procreation and separates both spouses, socially if not physically, from their respective families of orientation. Whenever a typical form of the extended family prevails, however, only one of the spouses breaks the tie with his family of orientation. The other remains at home, where he is joined by his spouse, and through his person his family of procreation is linked with his family of orientation in a composite familial aggregate. Since this process is normally repeated in each generation, the resulting extended family acquires temporal permanence. This continuity over time, which characterizes the extended family alone among the forms of family organization, classes it with unilinear kin groups and the community, which are also relatively permanent social groupings.
One comparatively rare form of the family occupies, in some respects, an intermediate position. This type, which we shall call the fraternal joint family, occurs when two or more brothers with their wives establish a common household. Whereas the basis of affiliation of the constituent nuclear families is the husband-wife relationship in the case of the polygamous family, and the parent-child relationship in the usual extended family, it is the brotherbrother relationship which unites the fraternal joint family. Like the independent nuclear and polygamous families, the fraternal joint family ordinarily includes only two generations and lacks temporal continuity. Like the extended family, however, the bond of union is a consanguineal rather than a marital tie. The fraternal joint family occurs as a common, but not exclusive, form of family organization among three of the tribes in our sample—the Iatmul of New Guinea, the Manus of Melanesia, and the Tenino of Oregon.
The several types of extended family depend primarily upon the prevailing rule of residence. To be sure—and this should be emphasized—no rule of residence will suffice of itself to create an extended family. Nevertheless, if economic or other factors are present which tend to unite contiguous families of adjacent generations into larger familial aggregates, the rule of residence which produces physical proximity among families of particular kinship connections will determine that these, and not others, will be the ones combined in the larger familial constellations.
Except for neolocal residence, which is inconsistent with extended family organization, any rule of residence tends to align families spatially in approximate accordance with a principle of descent. Oftentimes, however, this principle is not culturally recognized, or quite another rule of descent is followed. The rule of residence, by contrast, is always in harmony with the prevailing type of extended family organization. We shall consequently follow Kirchhoff 12 and name the types of extended families after their several residence rules, to wit, the patrilocal, matrilocal, bilocal, and avunculocal extended families.
Alternative names abound in the literature, e.g., patrilineal, matrilineal, and bilateral extended families; paternal, maternal, and avuncular families; patriarchal and matriarchal families. It is confusing, however, to speak of extended families as patrilineal or matrilineal when the rule of descent is bilateral rather than unilinear. Terms like “paternal” and “maternal” are too indefinite. Finally, “patriarchal” and “matriarchal” carry a misleading connotation of the attribution of domestic authority exclusively to one sex or the other. Residential terms have none of these disadvantages, and they correctly suggest the crucial differentiating factors. Occasionally, to avoid the confusing repetition of lengthy terms, we shall expand upon a useful suggestion by Lawrence 13 and speak of “patri-families,” “matri-families,” and “avuncu-families” instead of patri-local, matrilocal, and avunculocal extended families, respectively.
Where residence is patrilocal or matri-patrilocal, and factors favoring the development of an extended family are present, there appears a type of composite organization based upon the father-son relationship as the binding link. This is the patrilocal extended family, which includes the families of procreation of a man, his married sons, his sons' sons, etc. Under matrilocal residence, similarly, there results the matrilocal extended family, comprising the families of procreation of a woman, her daughters, and her daughters' daughters. In our sample of 250 societies, 52 are characterized by patrilocal and 23 by matrilocal extended families.
Under bilocal residence, either the son or the daughter, depending upon the circumstances of the particular case, may remain at home and thereby attach his family of procreation to his family of orientation. This results in a bilocal extended family, which unites the nuclear family of a married couple with those of some but not all of their sons, of some but not all of their daughters, and of some but not all of their grandchildren of either sex. Nuclear families of adjacent generations, in short, may be linked by any type of parent-child relationship. Extended families of this type are comparatively rare, numbering only 10 in our sample, since economic conditions usually exert a pressure in the direction either of patrilocal or of matrilocal residence and thus tend in time to upset the delicate balance of residential choice upon which the bilocal extended family rests.
Where avunculocal residence prevails, still another type of composite structure may emerge—the avunculocal extended family, of which our sample reveals seven instances. The Haida of British Columbia will serve as an illustration. In this tribe boys at about ten years of age take permanent leave of their parental homes and go to live with one of their maternal uncles in another village. The uncle's dwelling and his position as house chief are ultimately inherited by one of the nephews who reside with him, and this nephew often marries his uncle's daughter. Some of the sons of this nephew's sisters will in their turn come to live with him. Consequently a typical Haida household consists of the householder, his wife or wives, his young sons and unmarried daughters, several of his sisters' adolescent but unmarried sons, a sister's son who is married to his daughter, the young children of the latter couple, possibly other married nephews or daughters with their families, and occasionally even a grandnephew or two. In this instance the associated nuclear families are linked through two relationships, that between parent and daughter and that between maternal uncle and nephew. In some societies with this type of extended family, however, the nephew does not marry the daughter, so that the uncle-nephew link alone connects the associated nuclear families of adjacent generations.
Since independent nuclear and polygamous families are formed anew in each generation they are little influenced by historical vicissitudes. The temporal continuity of extended families, however, makes them subject to change as the conditions of life are modified. Repeatedly, for example, they have encountered the condition of an increase in population, outstripping the land or other resources available to them for economic exploitation. When this occurs, they commonly split, sending off branches which migrate and settle elsewhere. This fissive tendency has frequently been noted as a characteristic of societies with an extended family organization. Population growth and expansion may also convert an extended family into a clan. Indeed, as will be shown in Chapter 4, this is one of the commonest origins of clan organization.
A special investigation of the factors predisposing a society toward one or another type of family structure would doubtless yield illuminating results. Unfortunately the present study can shed little light directly upon this subject, since evidence was not gathered on exploitative techniques, division of labor, property, and other aspects of basic economy which presumably play a determining role. It may not be amiss, however, to advance a few tentative suggestions supported by a residual impression from the sources read.
The division of labor by sex in a particular economic setting may well determine in considerable measure the preferred form of marriage. Where women make an insignificant contribution to the economic life, as among the Todas 14, polyandry becomes a satisfactory adjustment. When the productive accomplishment of the two sexes is approximately equal, and a small unit is as efficient as a larger one, monogamy may be economically advantageous. When woman's economic contribution is large, and a man can produce enough in his sphere to satisfy the needs of several women, polygyny fits the circumstances. In suggesting the basic importance of economic factors we do not, of course, disclaim the auxiliary influence of others, e.g., the prestige value of plural wives or the sexual outlet offered by polygyny when continence is demanded during pregnancy and lactation.
The prevailing rule of residence is certainly an important element in the formation of extended families, but a consideration of the influences responsible for residence rules must be deferred until Chapter 8. Certain property factors probably operate to favor the continuity of the family and thus to promote the development of some form of extended family. Where possessions are few, or where they are readily movable or transferable, they offer no obstacle to family discontinuity. Where, however, property in agricultural land, large permanent houses, localized fishing or grazing sites, and other immovables constitutes a major form of wealth, especially if the most efficient unit for economic cooperation is larger than the nuclear family, each new family of procreation will tend to cleave to and assist the family of orientation of the spouse who can expect to share in the inheritance of such property. In this way, probably, the prevailing forms of property and the mode of its inheritance can predispose a society to a particular form of extended family.
The subject of the inheritance of property, having been introduced, deserves some parenthetical consideration. Evidence on inheritance rules was gathered incidentally during the course of the study, and some of the results may be presented here. In the literature 15, two primary modes of inheritance are commonly distinguished, namely, patrilineal and matrilineal. They are differentiated according to whether the preferred heir traces his relationship to the deceased through males or through females. Thus if a man's property passes to his son, inheritance is patrilineal; if to his sister's son, it is matrilineal; if to his brother, it may be either, with the rule becoming clear only when there are no surviving brothers. Table 6 summarizes our evidence on the mode of inheritance of masculine property in relation to rules of residence and descent.
It will be noted that matrilineal inheritance, though normally associated with matrilocal or avunculocal residence and with matrilineal descent, occurs in conjunction with patrilocal or matri-patrilocal residence in eight societies and with bilateral descent in three.

Table 6
Rules of Residence
and Descent
Patrilocal 87 6 5 98
Matri-patrilocal 10 2 0 12
Avunculocal 0 6 2 8
Matrilocal 5 13 2 20
Bilocal 8 0 3 11
Neolocal 9 0 1 10
Patrilineal 61 0 1 62
Double 6 1 3 10
Matrilineal 6 23 7 36
Bilateral 46 3 2 51
Total Cases 119 27 13 159

Moreover, patrilineal inheritance appears in five matrilocal and six matrilineal tribes. These results lend support to the contention of Lowie 16 that matrilineal and matrilocal customs are not inevitably linked in a consistent complex.
In actuality, the complexity of inheritance rules is such as to make the simple dichotomy of patrilineal and matrilineal highly inadequate for satisfactory analysis. In the first place, property may be destroyed, or given away, or disposed of by testament at the death of the owner, instead of descending to definite heirs. Secondly, the respective proprietary rights of the two sexes may exert a significant influence upon inheritance; there are usually articles such as clothing and occupational implements which are owned exclusively by the sex that uses them and inherited only by persons of the same sex, and there are objects such as money and valuables which may often be owned by either sex and transmitted to either sex. For this and other reasons, distinct rules of inheritance may prevail for different types of goods. Again, all the relatives of a given category may share alike, or a preference may be shown for the eldest or the youngest, as in primogeniture and ultimogeniture. Affinal relatives, such as the spouse or the children-in-law, may participate or be excluded; siblings may enjoy a preference over children, or vice versa; and so on. A really adequate study of property rights and inheritance in cross-cultural perspective still remains to be made.
Our digression into the field of the inheritance of property leads to the kindred subject of succession to positions of authority, which our study enables us to relate more closely to family structure. Final or supreme authority within the family, whatever the type of the latter, is almost universally vested in one person. In societies characterized by independent nuclear or polygynous families, this person is always, so far as our evidence goes, the husband and father. Where the extended family prevails, however, there are several alternatives. Authority within the household may be vested, for example, in the senior matron, as among the Iroquois, or in the senior maternal uncle, or—most commonly of all—in the paternal grandfather. Succession to this position may be either matrilineal or patrilineal, subject to most of the qualifications already noted in the case of inheritance. Similar roles, of course, prevail with respect to chiefship in the community and in the tribe, but these do not concern us here. Incomplete evidence as to the mode of succession in societies possessing extended families reveals that, with one exception, succession is always matrilineal in matrilocal and avunculocal extended families, and patrilineal in bilocal and patrilocal extended families. The single exception is the Ashanti of West Africa, who have the patrilocal extended family but matrilineal succession. In this society men exercise greater authority over their sororal nephews in other households than over their sons in their own.
Linton 17 distinguishes two fundamentally different types of family structure: the “conjugal family,” which capitalizes upon sexual attraction and consists of a nucleus of spouses and offspring surrounded by a fringe of comparatively unimportant relatives, and the “consanguine family,” which capitalizes upon the asexual associations formed during childhood and consists of a nucleus of blood relatives surrounded by a fringe of comparatively unimportant spouses. It is clear that the latter category corresponds closely to our extended family, whereas the “conjugal family” includes both the independent nuclear and independent polygamous families of our own classification. Linton regards the conjugal family as primordial, and he attributes the development of the consanguine family to a number of inherent advantages which it possesses. Among these he lists its permanency, the opportunity which this provides for the early development of adjustive habits, its superiority in numbers and hence in cooperative ability, its advantages in care for the aged and in selfprotection, and its greater suitability as a property-holding and status-giving tody. With this analysis the present author is in substantial accord.
Linton errs, however, in failing to recognize the universality of the nuclear family. He treats the extended (consanguine) family as an indivisible unit instead of viewing it as a cluster of nuclear families held together by interlocking primary relationships. He even goes to the extreme of stating that “spouses are of only incidental importance” in the extended or consanguine family. In composite families of this type, to be sure, certain of the economic and educational functions ordinarily fulfilled by the nuclear family are frequently shared in part by the larger group. Only in part, however, for, as bas already been pointed out, the constituent nuclear families always retain at least some distinctiveness as cooperative economic units and regularly bear the primary burden of child care, education, and socialization, in addition to which they maintain intact their unique sexual and reproductive functions. Despite such minor shortcomings, Linton's remains the most thoughtful analysis of composite forms of the family to be found in the existing literature, and is to be recommended as a supplement to the present chapter.

1. Cf. L. H. Morgan, Ancient Society (New York, 1877), p. 416; J. Lubbock, The Origin of Civilisation and the Primitive Condition of Man (5th ed., New York, 1892), pp. 86-98; J.G. Frazer, Totemism and Exogamy (London, 1910), IV, 151; W.H.R. Rivers, Social Organization (New York, 1924), p. 80; R. Briffault, The Mothers (New York, 1927), I, 614-781.
2. J. Henry, Jungle People (New York, 1941), p. 45n.
3. Cf. R.H. Lowie, Primitive Society (New York, 1920), pp. 49-62.
4. Among the Todas, for example, the numerical disproportion of the sexes is very marked. See W. H. R. Rivers, The Todas (London, 1906), pp. 477-80.
5. J. F. McLennan, Studies in Ancient History (London, 1876), p. 132.
6. See, for example, J. H. Steward, “Shoshoni Polyandry” American Anthropologist, n.s., XXXVIII (1936), 561-4; W.Z. Park, Paviotso Polyandry, American Anthropologist, n.s., XXXIX ( 1937), 366-8; 0. C. Stewart, “Northern Paiute Polyandry,” American Anthropologist, n.s., XXXIX ( 1937), 368-9; D. G. Mandelbaum, “Polyandry in Kota Society,” American Anthropologist, n.s., XL (1988), 574-83.
7. J.M. Cooper, “Temporal Sequence and the Marginal Cultures,” Catholic University of America Anthropological Series, X ( 1941 ), 52-3.
8. M. K. Opler, “Woman's Social Status and the Forms of Marriage,” American Journal of Sociology, XLIX (1943), 130-46.
9. New light on polyandry may be expected from Prince Peter of Greece, who is making a life study of the subject. Lecturing in the United States in 1948, he presented an illuminating functional interpretation of polyandry in western Tibet, where it is normally fraternal but becomes non-fraternal in exceptional matrilocal marriages where the family line is carried on by a daughter in default of sons.
10. In the literature, “sororate” is often applied to polygynous as well as to successive marriages with sisters, but this usage is here avoided as confusing.
11. Societies for which polygyny is reported without specifying its type have been regularly classed as non-sororal. These include 52 with general polygyny and 18 with infrequent polygyny.
12. P. Kirchhoff, “Kinship Organization,” Africa, V (1932), 190.
13. W.E. Lawrence, “Alternating Generations in Australia,” Studies in the Science of Society, ed. G. P. Murdock (New Haven, 1937), p. 319.
14. “The men, in addition to their political, religious, and ceremonial functions, do the herding, milking, churning, fuel gathering, building, trading, and cooking. In the absence of agriculture and important domestic arts, the female share in the division of labor by sex is confined to such comparatively minor activities as fetching water, mending and embroidering clothes, pounding and sifting grain, sweeping the floor, and cleaning the household utensils.” G. P. Murdock, Our Primitive Contemporaries (New York. 1934), p. 124.
15. See, for example, W.H.R. Rivers, Social Organization (New York, 1924), pp. 87-8.
16. R. H. Lowie, “The Matrilineal Complex,” University of California Publications in American Archaeology and Ethnology, XVI (1919), 29-45.
17. R. Linton, The Study of Man (New York. 1936), pp. 159-6S, et passim.

Consanguineal Kin Groups

Social groupings based upon kinship ties are called kin groups. The nuclear family is, of course, a kin group, since each of its members is a primary relative of each of the others. The polygamous family is also a kin group. In addition to primary relatives, however, it regularly includes a number of secondary relatives who are found only occasionally or incidentally in the household of a nuclear family, particularly stepparents, half siblings, and stepchildren. Extended families, whether of the patrilocal, matrilocal, avunculocal, or bilocal types, are likewise kin groups. Unlike nuclear and polygamous families, they regularly include tertiary relatives, such as a brother's son's wife or a father's brother's son in a patri-family, and often embrace remoter relatives as well.
It is important to note, however, that in every type of family organization the kinship bonds which link the members to one another are always in part affinal and never exclusively consanguineal. In the nuclear family, for example, the tie between father and mother, or husband and wife, is one of marriage; incest taboos universally prevent their being primary consanguineal relatives. in composite forms of the family a number of members are linked only by affinal bonds. But whether their ties are consanguineal or affinal, the members of a family, whatever its type, are also united by common residence. These characteristics of mixed consanguineal nod affinal ties and of common residence distinguish one of the major types of kin groups, that of the residential kin group, which includes all forms of the family.
A second major type of kin group is the consanguineal kin group. The members of such groups are bound together exclusively by consanguineal kinship ties. If an affinal connection can be traced between any two members, it is always indirect and incidental. Since persons whose relationship is primarily affinal cannot belong to the same consanguineal kin group, husband and wife can be co-members only under very special circumstances. Incest taboos and their extensions in the form of exogamous rules ordinarily require a person to marry outside of his own consanguineal kin group, with the result that spouses are normally members of different groups. Brother and sister, on the other hand, are always consanguineal kinsmen and members of the same group. Since marriage brings together in common residence both husband and wife, who cannot belong to the same consanguineal kin group, and since incest taboos ordinarily separate married brothers and sisters, who regularly belong to the same group, a consanguineal kin group cannot be characterized by common residence.
The basic characteristics of the two types of kin groups may be summed up as follows. The residential type is always characterized by common residence; the consanguineal type, never. The consanguineal type includes only consanguineal relatives; the residential type always excludes some consanguineal relatives and includes some affinal ones. The residential type regularly includes husband and wife but not brother and sister. The consanguineal type invariably includes both brother and sister but almost never both husband and wife. These basic characteristics apply to all kin groups of the two types, whatever their size or name and whatever the rule of residence or descent. A further generalization is that the form of a residential kin group is determined primarily by the prevailing rule of residence, that of a consanguineal kin group primarily by the rule of descent.
It has already been pointed out that descent is not synonymous with genealogical relationship. Even unilinear descent does not imply that the kinship tie with one parent is recognized to the exclusion of, or is closer than, that with the other, for there are instances of patrilineal descent in societies which do not recognize the biological connection of father and child. Descent refers solely to a cultural principle whereby an individual is socially allocated to a specific group of consanguineal kinsmen.
Kin groups represent, so to speak, the individual's second line of defense. When a person is in danger or in trouble, when he needs help in the performance of an economic task or a ceremonial obligation, whenever, in short, he requires a measure of assistance beyond what his own immediate family can provide, he can turn to the members of his larger kin group for aid or succor. Because they are bound to him by extended kinship ties, their obligation to help him is stronger than that of other members of the tribe or the community. He in turn, of course, is under reciprocal obligations toward them. Consanguineal kin groups are of particular importance, for a person ordinarily feels closer to his own “blood relatives” than to those who are related to him only through marriage.
To which of his consanguineal relatives can the individual turn for support, and to which of them does he owe reciprocal obligations? Consanguineal kinship ties ramify endlessly, and if carried far enough embrace all members of the local community or even of the tribe. Some discrimination must be made among them. There must be some selection of those with whom the individual is to be particularly closely affiliated. The decision must be reached at the time of his birth if confusion is to be avoided. The birth of a new child in any society inevitably affects the rights and privileges of other members. Some are compelled to assume new obligations. Others may gain new rights. The prospects of others are altered with respect to such matters as inheritance, succession, and marriage. In short, jural relations are modified in various ways, and everyone must know how his own are affected. Unless uncertainty and controversy are to reign, the social placement of the infant must be settled at the outset.
The date of one's birth may determine the age-grade to which one belongs. The place of birth may determine membership in a particular household, community, or residential kin group. But there are other kinds of social groups in which membership must also be decided at birth but where there are no obvious external guides to a decision among alternatives. Such a problem arises, for example, in the case of the offspring of a marriage between persons of different status groups, such as castes or social classes. To which caste shall a mulatto or Eurasian child belong; to which class the child of a freeman and a slave woman or of a noble father and a common mother? There are two alternatives of affiliation, and the decision, unless it is to become a bone of contention in every instance, must be prescribed in advance by a specific cultural provision.
The determination of the particular group of relatives with whom an individual is to be affiliated, and to whom he is to enter into reciprocal obligations, presents a precisely similar situation. In most societies these kinsmen begin to function shortly after his birthusually at a ceremony in which the newborn infant is given his name and is formally accepted by his relatives. But whether or not they assemble for such an occasion, they must know in advance , who they are, and how their jural rights are affected. This knowledge is supplied in all societies by a culturally formulated rule of descent. There is nothing inherently obvious or “natural” about any rule of descent. By virtue of incest taboos, which universally prevent marriage within the nuclear family, the father and mother cannot have the same kinsmen. H a child were affiliated with all the relatives of both parents, his consanguineal kin group would be double the size of theirs, and in a few generations the kin group would become coextensive with the community, or even the tribe, and would lose its signiflcance. It is imperative, therefore, that the kin group of the the child be restricted to approximately the size of that of either parent. To accomplish this, some of the members of the parents' kin groups must be eliminated from that of the child. There are three principal ways in which this can be done, and they constitute the three primary rules of descent.

  • Patrilineal descent accomplishes the selection by discarding the mother's kin group and affiliating the child exclusively with the consanguineal kin group of the father.
  • Matrilineal descent similarly discards all the father's relatives and assigns the child to his mother's kin group.
  • Bilateral descent accomplishes the same result by excluding some of the members of the father's kin group and some of those of the mother's and affiliating the child with a special group consisting of some of the members of both. In most cases these are his own nearest genealogical kinsmen, irrespective of through which parent they are related to him.

Each of these rules produces consanguineal kin groups of several characteristic types.
It is also possible for a particular society to combine two rules of descent. Combinations of matrilineal and patrilineal descent are particularly common. Thus in certain Indonesian societies with ambil-anak marriage the usual patrilineal rule is suspended for a generation in the case of a family without sons, and matrilineal descent through a matrilocally residing daughter supervenes to continue the family line. Among the Apinaye of Brazil, matrilineal descent prevails for females and patrilineal descent for males. Among the Mundugumor of New Guinea, on the other hand, a daughter is affiliated with her father, and a son with his mother, in a kin group which zigzags between the sexes from generation to generation 1 like sex-linked characteristics in heredity. Among the Buginese and Macassar of Celebes, the first, third, and other odd-numbered offspring are affiliated with the mother, and even-numbered progeny with their father 2. In all of these cases, it is important to note, one unilinear rule is applied in particular individual cases, the other in others. Both are not applied at the same time to one individual.
When patrilineal and matrilineal descent are applied together, and not alternatively in combinations like the above, their joint application is known as double descent 3. In this case the society possesses both patrilineal and matrilineal kin groups, and a person is affiliated at the same time with the patrilineal group of his father and the matrilineal group of his mother, the relatives of his father's matrilineal and of his mother's patrilineal group being discarded. Since under certain conditions double descent produces consanguineal kin groups of unique type, it might be ranked as a fourth primary rule of descent rather than as merely a combination of the patrilineal and matrilineal rules.
Our own society is characterized by bilateral descent and by the presence of kin groups of a typically bilateral type, technically called the kindred 4 but popularly known under such collective terms as “kinfolk” or “relatives.” Since bilateral kin groups represent an especially difficult problem, and one to which anthropologists have devoted relatively little attention, it will be preferable to give first consideration to the kin groups resulting from unilinear descent, i.e., patrilineal and matrilineal. These are reported in the literature under a welter of terms: clan, gens, lineage, moiety, phratry, ramage, sept, sib, etc. Several generations of anthropologists have made i notable progress in the analysis of these groups and in the assignment of appropriate terms to those of different type. We shall in general follow the classic work of Lowie 5, wherein is found the nearest approach to a complete clarification of the subject.
The consanguineal kin groups which result from patrilineal and matrilineal descent parallel one another almost exactly, being differentiated only in their modes of affiliation and in the particular relatives whom they include. Each constitutes a similar graded series, comprising types which reflect the varying degrees to which recognition of unilinear affiliation may be extended. These types bear identical names, the rule of descent being indicated by the adjectives “patrilineal” and “matrilineal” or by the prefixes “patri-” and “matri-”.
A consanguineal kin group produced by either rule of unilinear descent is technically known as a lineage when it includes only persons who can actually trace their common relationship through a specific series of remembered genealogical links in the prevailing line of descent. Oftentimes, though by no means always, a lineage consists of the unilinearly related persons of one sex who form the core of a patrilocal, matrilocal, or avunculocal extended family, together with their siblings of opposite sex who reside elsewhere and, of course, not including their spouses who live with them. Although our own society is bilateral, the patrilineal inheritance of surnames results in name-groups which resemble lineages. Thus all persons born with the surname Smith who can trace actual descent in the male line from a common ancestor constitute a sort of patrilineage. When the members of a consanguineal kin group acknowledge a traditional bond of common descent in the paternal or maternal line, but are unable always to trace the actual genealogical connections between individuals, the group is called a sib 6. If all persons born with the name Smith in our society regarded themselves as related, they would constitute a patri-sib. Some unilinear societies lack true sibs, possessing only lineages. The great majority, however, possess sibs, which are the most characteristic of all unilinear consanguineal kin groups. A sib normally includes several lineages, though these need not be culturally defined. Groups intermediate between sibs and lineages, which are found in some societies, may be called sub-sibs.

Occasionally two or more sibs recognize a purely conventional unilinear bond of kinship, more tenuous than that which unites a sib but nevertheless sufficient to distinguish the constellation of sibs, from others of its kind. A consanguineal kin group of this higher order is called a phratry.

When a society has only two sibs or phratries, so that every person is necessarily a member of one or the other, the dichotomy results in so many distinctive features in social structure that a special term, moiety, is applied to them. If our own society included only people named Smith and Jones, and if each group considered themselves patrilineally related, they would constitute patri-moieties.

The most widespread characteristic of unilinear consanguineal kin groups is exogamy, i.e., the rule requiring all members to seek their spouses in some other group. The smaller the kin group, in general, the stronger the tendency toward exogamy. Lineages, for example, are sometimes completely exogamous when sibs are only partially so, and moieties are non-exogamous more often than any of the smaller unilinear groups. Social units that simulate kin groups but are not based upon a rule of descent, e.g., the pseudo-moieties of some tribes that live on opposite sides of a village square or oppose one another in games, and whose membership is determined on some non-kinship basis, must not be confused with genuine unilinear kin groups, even of non-exogamous type.
For purposes of structural analysis, such as the interpretation of kinship terminology, unilinear societies whose kin groups are completely non-exogamous must usually be treated as though they were bilateral in descent, for endogamous unions prevent the expected spatial and social alignments of kinsmen. Of the 178 unilinear societies in our sample, only 10 reveal a complete absence of exogamy. In half of these instances the kin groups appear to be emergent and not yet fully elaborated. Thus the Balinese, Tongans, and Tswana I have patri-lineages, apparently based upon patrilocal residence, without having yet developed a rule of exogamy, and non-exogamous matri-lineages are similarly found among the matrilocal Kallinago. The Ontong-Javanese possess non-exogamous groups of both types, the patri-lineages being land-owning units and the matri-lineages house-owning groups. In five other instances the kin groups appear decadent and on the point of disappearing. Among the Kababish of the Sudan and the Kurds of Iraq the introduction of Islam has destroyed sib exogamy by introducing preferential marriage with the father's brother's daughter. The patrilineal kin groups of the Fox, Pima, and Tewa of North America are likewise clearly decadent.
The above list of societies with unilinear but non-exogamous kin groups might have been expanded slightly by a somewhat broader definition. Thus the Ruthenians and Yankees might have been placed in the same category because of their patrilineally inherited name groups, which are possibly survivals of sibs. The Buin of Melanesia have matrilineally inherited totems, and the Edo of Nigeria have patrilineally inherited food taboos, either of which may represent decadent or incipient sibs. The Washo Indians have patrilineal descent groups whose only function is apparently to oppose one another in games. For various reasons, however, it has seemed preferable not to class these phenomena with kin groups except for such purposes as testing the tendency of exogamy to be associated with any unilinear grouping of kinsmen. On comparable grounds, the following have not been classed as moieties: the two divisions of the Sabei, since they are presumably merely local; the endogamous Tartharol and Teivaliol branches of the Toda, since they show marked affinity with castes; and the two patrilineal divisions of the Longuda, since they are primarily religious and ceremonial in character. With perhaps no better justification, however, the non-exogamous dual divisions of the Pukapukans and the Yuchi have been classed, respectively, as matrilineal and patrilineal moieties. Since borderline cases raise difficulties in any classification, all instances encountered have been enumerated above.
A classification of the 175 unilinear societies of our sample by type of kin group and the prevalence of exogamy is shown in Table 7 below. The inclusion of tribes with double descent in both columns accounts for the apparent numerical excess.

Type af Kin Group and Prevalence of Exogamy Patrilineal Descent MatrilinealDescent
With exogamous moieties 10 19
With non-exogamous moieties and exogamous sibs 4 5
With moieties and other kin groups, all nonexogamous 3 0
With exogamous phratries 9 5
With exogamous sibs 74 33
With non-exogamous sibs and exogamous lineages 4 0
With non-exogamous sibs and lineages 3 0
With exogamous lineages only 10 5
With non-exogamous lineages only 6 3
Totals 123 70

Another common characteristic of lineages, sibs, and moieties is totemism. For an analysis of this controversial phenomenon the reader must be referred to other sources 7, since its bearing on the formal structuring of social relations is comparatively slight. One of the most widespread features of the so-called totemic complex is the assigning of animal names to kin groups. The explanation may be quite simple. If a people who call their fraternal orders Eagles, Elks, and Moose, who dub their professional baseball teams Cubs, Orioles, and Tigers, who denote the student bodies and teams of their academic institutions Bulldogs, Panthers, Terrapins, and (Golden Bears, who use the Gopher and the Wolverine to symbolize I states and the Bull Moose, the Donkey, and the Elephant to represent political parties, and who see their American Eagle competing for world power or struggling for international peace with the British Lion and the Russian Bear—if such a people were to have sibs, would they be likely to arrive at names for them very different from the Bear, Beaver, Hawk, Turtle, Wolf, etc., chosen by the Iroquois Indians?
If social groups arE: to receive names, animal designations are as obvious as any. The important fact, however, is that consanguineal kin groups have names, not that these are so commonly those of animals. Oftentimes, of course, the names are derived from plants, natural objects, localities, chiefs, or ancestors rather than from birds and mammals. Naming is important because a common name can identify a member of a kin group who resides apart from his relatives, thus helping to keep alive the consciousness of group membership. It is probable, indeed, that the extension of a distinguishing name to all persons born in a locality, and its retention by those who have left their homes in marriage, is one of the principal means by which lineages and sibs have evolved 8.
Totemic food taboos can serve a similar purpose. Even in our own society, members of certain religious sects are commonly distinguished by their abstention from meat on certain days or from pork at all times. Though the elements of the totemic complex are doubtless diverse in kind and origin, many of them certainly fulfill the same function of supporting the social unity of the consanguineal kin group in the face of the dispersion of its members. In societies which possess both patrilineal and matrilineal lineages, sibs, or phratries, double descent introduces no novel structural features. Kin groups with both rules of descent exist, and the individual is affiliated with both the patrilineal group of his father and the matrilineal group of his mother. Among the Ashanti, for example, an individual inherits his “blood” through affiliation with his mother's matri-sib and his “spirit“ through membership in his father's patrisib, both of which are exogamous and totemic 9. The Herero, too, have exogamous, totemic matri-sibs and patri-sibs, the former primarily social in character, the latter religious 10. In no case do such groups diller in any significant respect from those found under a single unilinear rule of descent.
If, however, a society with double descent possesses moieties among its various component kin groups, there tends to emerge with exogamy an entirely new type of structure which may be called the bilinear kin group. In contrast to both unilinear and bilateral groups, a bilinear kin group is composed of persons who are affiliated with one another by both patrilineal and matrilineal ties, including those who stand to one another in such relationships as own siblings, parallel cousins, paternal grandfather and son's child, and maternal grandmother and daughter's child. Excluded from one's own group will be all persons who are related only patrilineally or only matrilineally to Ego, as well as those who are not connected in either line.
Bilinear kin groups, technically called sections, have long been known in aboriginal Australia, and from the time of Galton 11 their bilinear character has been dimly recognized. Knowledge of the subject was greatly advanced by Deacon 12, who discovered sections in Melanesia and interpreted them correctly. It remained for Lawrence 13, however, to clear up the entire matter in what is certainly one of the most original and significant contributions in the entire literature on social organization. Radcliffe-Brown 14, in other respects the outstanding authority on aboriginal Australian social structure, has seen fit to criticize Lawrence in an article which deals with ethnographic minutiae and evades the real issue. It therefore becomes necessary to emphasize that on this important matter Lawrence is right whereas Radcliffe-Brown has been consistently wrong.
The crucial issue concerns the determinants of social systems of the Australian type. The interpretation guessed by Galton, confirmed by D eacon, conclusively demonstrated by Lawrence, and accepted here is that such systems result from the interaction of patrilineal and matrilineal kin groups in the presence of moieties and rigorous exogamy. Radcliffe-Brown, who has disagreed with Deacon 15 as well as with Lawrence, ascribes them to the influence of kinship terminology 16. Depending upon the type of kinship system they happen to have, some Australian tribes will allegedly be characterized by preferential cross-cousin marriage, others by preferred marriage with a second cross-cousin, kin groups being entirely secondary and inconsequential 17. When forced by the facts to recognize that marriage is commonly also permitted with relatives to whom other kinship terms are applied, Radcliffe-Brown attempts to escape the dilemma by statements to the effect that these relatives “stand in an equivalent relation” to the preferred spouse. Analysis of the data shows that those who “stand in an equivalent relation” are regularly members of the same bilinear kin group. In other words, what really regulates marriage is not kinship terminology, as Radcliffe-Brown asserts, but the prevailing kin groups, as Lawrence demonstrates.
One of the most definite conclusions of the present work (see Chapters 7 and 10) is that kin groups are the primary determinants of both kinship terminology and marriage rules. Nowhere else in the world can we find evidence for the primacy of either of the latter, and Australia could hardly be expected to constitute the only exception, even if Lawrence had not adduced the proof. In addition to providing the most satisfactory explanation of the known facts, Lawrence's interpretation accounts for such peculiarities as the dispute between Mathews and Spencer as to whether the Arunta are matrilineal or patrilineal, whereas in reality they are both. Above all, however, it brings Australian social organization for the first time within the same frame of reference as social systems elsewhere in the world, composed of identical elements, accountable to the same influences, and differentiated only by the relative complexity of its configurations. Radcliffe-Brown's interpretation, on the other hand, would leave native Australian institutions in that morass of the seemingly bizarre, unique, and scientifically inexplicable whence Lawrence has rescued them.
When mysticism is swept away, the essential nature of the so-called “two-class,” “four-class,” and “eight-class” systems of Australian social organization is relatively easy to comprehend. All of them are based on a combination of matrilineal and patrilineal kin groups which always includes both exogamous matri-moieties and the exogamous patri-lineages or patri-sibs which constitute the core of the ubiquitous local groups or “hordes.” In the absence of any complicating factors, these two groups in conjunction produce a situation in which the native-born members of each local horde or patri-clan, i.e., the members of the localized patri-lineage or patri-sib, are divided in alternating generations between the two matri-moieties. A male Ego, his father's father, his son's son, and the siblings and parallel cousins of all of them fall into one moiety; Ego's father, his son, and the siblings and parallel cousins of both fall into the other. Since the matri-moieties extend throughout the tribe, alternating generations in all local groups throughout the society are affiliated. Exogamous rules permit a man to marry any woman of the opposite matri-moiety provided she is not a member of his own patri-sib; she must come from another local group and from one of the alternate generations there which are equated with the moiety to which his own father and son belong. This is the socalled “two-class system,” and differs from an ordinary system of exogamous matri-moieties only through the addition of patrilineal descent.
In a considerable part of the Australian area of double descent, patrilineal descent has been extended beyond the local group so that all the patri-sibs of the tribe are aggregated into two interspersed sets, which constitute exogamous patri-moieties. These patri-moieties intersect the two matri-moieties to form four sections. In a “four-class system” of this type, each section forms a bilinear kin group, united by double descent. For any given individual, all the members of his own section are both patrilineally and matrilineally related to him, i.e., they all belong both to his patri-moiety and to his matri-moiety. A second section includes all the persons who belong to his patri-moiety but to the opposite matri-moiety. A third section embraces all matrilineal kinsmen who do not belong to his patri-moiety. The fourth section includes everyone who is related to him in neither the male nor the female line, i.e., who belongs both to the opposite patri-moiety and to the opposite matrimoiety. Since both moieties are exogamous, it is from this fourth section, and this one only, that he can take a spouse. Double moiety exogamy explains the otherwise peculiar fact that in Australia marriage is permitted into only one other kin group, whereas in most societies with exogamous moieties and sibs a person may marry into any of the sibs of the opposite moiety.
Cleared of the obscurities which befog the descriptions in much of the literature, Australian “four-class systems” exhibit none of the complexities which readers have been led to expect. They are, on the contrary, exceedingly easy to comprehend. For the guidance of the layman, Table 8 lists the section membership of the more important primary, secondary, and tertiary relatives of a male Ego. Relatives are abbreviated according to a system proposed elsewhere, 18 which will be followed hereinafter.
In a still more limited area in native Australia there occurs the so-called “eight-class system,” wherein each section is subdivided into two subsections. This results from the imposition of an exogamous taboo upon one's mother's patrilineal kinsmen, which is extended throughout the society to form a third moiety dichotomy. This third exogamous dichotomy divides the members of each section into two groups, those belonging to one of the third pair of moieties and those belonging to the other. In the wife's section, for example, cross-cousins are grouped with MoFa in the prohibited subsection, and Ego is compelled to seek a wife in the other subsection, which includes his FaMo and certain second cross-cousins. It is in consequence of this third moiety dichotomy that a person is allowed to take a spouse from only one of the seven subsections other than his own, and that the nearest eligible mate is normally a second cross-cousin who belongs neither to his own matri-moiety, nor to his patri-moiety, nor to the third moiety which includes his mother's patrilineal kinsmen. For further details the reader must be referred to Lawrence's work and to the descriptive literature.

  Memberof Ego's Patri-Moeity Members of Opposite Patri-Moeity
Relative Ego's Matri-Moeity Opposite Matri-Moeity Ego's Matri-Moeity Opposite Matri-Moeity
FaFa X
FaMo X
MoFa X
MoMo X
Fa X
Mo X
FaBr, FaSi X
MoBr, MoSi X
FaBrWi, FaSiHu X
MoBrWi, MoSiHu X
WiFa X
WiMo X
Br, Si X
Wi X
FaBrSo, FaBrDa
FaSiSo, FaSiDa X
MoBrSo, MoBrDa X
MoSiSo, MoSiDa X
WiBr, WiSi X
BrWi, SiHu X
WiBrWi, WiSiHu X
So, Da X
BrSo, BrDa X
SiSo, SiDa X
SoWi, DaHu X
WiBrSo, WiBrDa X
WiSiSo, WiSiDa X
SoSo, SoD X
DaSo, DaDa X

The Ranon of the New Hebrides, and apparently also the Pentecost, reveal a “six-class system,” a variation apparently not reported for Australia. The six sections are produced by the intersection of three exogamous patri-sibs by two exogamous matri-moieties. A man can take a wife from only one of the three sections of the opposite moiety, namely, the one which belongs neither to his own nor to his mother's patri-sib. True bilinear kin groups or sections have never been reported outside of Australia and a limited area in Melanesia. Despite occasional allegations to the contrary, the complex social systems of eastern Indonesia, of the Naga tribes of Assam, of the Ge tribes of east central Brazil, and of such North American tribes as the Cherokee bear no relationship to those of Australia. Perhaps the most obvious criterion of a genuine bilinear kin group is that no individual can belong to the same group as either his father, his mother, his son, or his daughter. It is also likely to be diagnostic if he is permitted to marry into only one other kin group in the society. The crucial factor, however, is the presence of double descent with moieties and exogamy. The only society in our sample for which bilinear kin groups seem possible, though unreported, is the Wogeo of New Guinea, who have exogamous matri-moieties and local exogamy in connection with a group based on patrilocal residence. That they apparently lack some type of “marriage classes” is presumably due to the considerable measure of deviation from the normal residence rule.
Double descent should not be confused with bilateral descent, which, unlike the former, is not a simple combination of patrilineal and matrilineal rules. The distinction becomes clear in examining Ego's relation to his four grandparents. He is aligned in the same kin group as his father's father under patrilineal descent, in the same group as his mother's mother under matrilineal descent, and in different kin groups with each under double descent. In none of (these instances, however, does he find himself in the same kin group with either his father's mother or his mother's father. Under bilateral descent, however, he is affiliated equally with all four grandparents, and all four, being secondary relatives of his, will necessarily be members of any consanguineal kin group of bilateral type to which he himself belongs. Bilateral descent, instead of being a combination of patrilineal and matrilineal, reflects a complete absence of any unilinear emphasis.
The commonest type of bilateral kin group, as previously noted, is the kindred. In our own society, where its members are collectively called “kinfolk” or “relatives,” it includes that group of near kinsmen who may be expected to be present and participant on important ceremonial occasions, such as weddings, christenings, funerals, Thanksgiving and Christmas dinners, and “family reunions.” Members of a kindred visit and entertain one another freely, and between them marriage and pecuniary transactions for profit are ordinarily taboo. One turns first to them for aid when one finds oneself in difficulties. However much they may disagree or quarrel, they are expected to support one another against criticism or affronts from outsiders. The kindred in other societies has comparable characteristics and functions.
Bilateral kin groups have received little attention from anthropological theorists. Consequently ethnographers rarely notice their presence and. almost never report their absence. Specific descriptions or clear inferences attest the presence of kindreds in 33 societies in our sample, though further research would doubtless reveal them in others. Kindreds are occasionally reported for patrilineal societies, such as the Bena, Ojibwa, and Tikopia, and for matrilineal tribes, such as the Hopi, Iroquois, and Nayar, but the overwhelming majority are recorded for bilateral societies or for tribes with non-exogamous sibs or lineages, like the Fox and Tswana. They appear especially common with bilocal residence, though they also occur frequently with neolocal residence. In general, they are clearly associated with an absence of, or a minimal stress upon, unilinear descent. Probably they will ultimately appear to be characteristic of most bilateral societies. However, since kindreds normally exhibit a tendency toward exogamy comparable to that of lineages, the fact that 13 bilateral societies in our sample reveal no bilateral extension of sex prohibitions suggests that kindreds are probably completely lacking in at least some of these cases. If this is true, there are a few societies with no consanguineal kin groups intermediate between the nuclear family and the community.
Since the cultural rule of bilateral descent corresponds exactly with the facts of genealogical relationship, and since most peoples recognize the biological connection between a child and both of its parents, it might be expected that the majority of societies would choose this alternative as the cultural norm. In our sample of 250, however, only 75, representing but 30 per cent of the total, follow the rule of bilateral descent. The comparative rarity of bilateral descent and the widespread prevalence of other rules, all of which appear inconsistent with well-known biological facts, certainly demand explanation.
Suggested explanations have been numerous and diverse. The evolutionist anthropologists 19 of the nineteenth century assumed that social evolution must have begun with the matrilineate because primitive man could not have known the facts of physical paternity. In their opinion, patrilineal institutions evolved later as the male sex gradually achieved dominance, while bilateral descent became established only with the emergence of higher civilization and the accompanying realization of the equal roles of both parents. American anthropologists 20 of the early twentieth century, in criticizing the evolutionists for ignoring the nuclear family, ascribed priority to bilateral descent and regarded the matrilineate as a relatively late development, without, however, accounting for its origin. Historical anthropologists 21 of several persuasions-British, Austrian, and American-have regarded unilinear descent as so anomalous that they attribute it to a limited number of cultural inventions and to worldwide diffusion from its points of origin. All these hypotheses will be analyzed in Chapter 8, where their defects and their inconsistency with distributional facts will be demonstrated.
For enlightenment we must turn to theorists who consider the functional significance of the several types of consanguineal kin groups. Linton 22 advances several suggestions, among them the hypothesis that “an emphasis on unilinear descent is an almost unavoidable accompaniment of the establishment of family units on the consanguine basis.” If this theory were correct, unilinear descent should tend to be strongly associated with the presence of extended families, and bilateral descent with their absence. Our data, however, fail strikingly to confirm this expectation. To be sure, patrilineal descent occurs in 69 per cent (36 out of 52) of the societies in our sample with patrilocal extended families, and matrilineal descent in 73 per cent (22 out of 30) of those with matrilocal or avunculocal extended families, but the same unilinear rules also occur in 60 per cent (68 out of 113) of the societies which completely lack any kind of extended family.
Lowie 23 believes that “the transmission of property rights and the mode of residence after marriage have been the most effective means of establishing the principle of unilateral descent.” The alleged influence of the inheritance of property cannot be tested, since Lowie does not indicate what particular inheritance rules should coexist with bilateral descent. The hypothesis that residence rules can be instrumental in establishing unilinear descent is also advanced by Linton 24, who states: “Matrilineal descent is normally linked with matrilocal residence, patrilineal with patrilocal.” This theory is supported by the data compiled in Table 9, and will be fully validated in Chapter 8.

Rules of Residence Matrilineal Descent Patrilineal Descent Double Descent Bilateral Descent Totals
Matrilocal and avunculocal 33 0 0 13 46
Patrilocal and matri-patrilocal 15 97 17 39 168
Neolocal and bilocal 4 8 1 23 36
Totals 52 105 18 75 250

With such support from the evidence, we must agree with Lowie and Linton that a fixed rule of residence, whereby spouses of a particular sex regularly continue to reside with or near their linear relatives of the same sex after marriage, is conducive to unilinear rather than bilateral descent. The question still remains, however, of whence residence rules acquire the power to override the common knowledge of the genealogical bond between a child and both parents, and to produce rules of descent which affiliate a child with the relatives of only one parent. The fact that certain of the latter are congregated in the same locality by any unilocal rule of residence, thereby rendering their affiliation the more obvious, unquestionably provides an important part of the answer. It is clear, however, that unilocal residence alone does not produce unilinear descent, as is attested by the 52 societies in our sample which are bilateral in descent though characterized by matrilocal or patrilocal residence.
A supplementary hypothesis states that unilinear kin groups possess certain advantages not shared by the bilateral kindred, and that these add their weight to the factor of proximity in many cases t and help to tip the scales in favor of matrilineal or patrilineal descent. Radcliffe-Brown 25, for example, attributes unilinear forms of organization to “certain fundamental social necessities,” namely, a need for the precise formulation of jural rights so as to avoid conflicts, and a need for continuity in the social structure which defines such rights. Linton 26 mentions the sanie factors. These suggestions would seem to warrant a reexamination of the kindred. The most distinctive structural fact about the kindred is that, save through accident, it can never be the same for any two individuals with the exception of own siblings. For any given person, its membership ramifies out through diverse kinship connections until it is terminated at some degree of relationship—frequently with second cousins, although the limits are often drawn somewhat closer or farther away than this and may be rather indefinite. The kindreds of different persons overlap or intersect rather than coincide. Those, for example, of first cousins, the sons of two brothers, have part of their membership in common—the near relatives through their respective fathers—and the rest distinct; the kinsmen of either cousin through his mother do not belong to the kindred of the other.
Since kindreds interlace and overlap, they do not and cannot form r discrete or separate segments of the entire society. Neither a tribe nor a community can be subdivided into constituent kindreds. This intersecting or non-exclusive characteristic is found only with bilateral descent. Every other rule of descent produces only clearly differentiated, isolable, discrete kin groups, which never overlap with others of their kind.
One result of this peculiarity is that the kindred, though it serves adequately to define the jural rights of an individual, can rarely act as a collectivity. One kindred cannot, for example, take blood vengeance against another if the two happen to have members in common. Moreover, a kindred cannot hold land or other property, not only because it is not a group except from the point of view of a particular individual, but also because it has no continuity over time. Hence under circumstances favorable either to the communal ownership of property or to the collective responsibility of kinsmen, the kindred labors under decided handicaps in comparison to the lineage or sib.
A particular disadvantage of the kindred appears in the instances in which an individual belongs to the kindreds of two other persons and thereby becomes involved in conflicting or incompatible obligations. If they get into serious difficulties with one another, for example, he may be required to avenge the one and yet to defend the other. If they become estranged, both are likely to turn to him for support and to subject him to emotional conflict and strain. The reader can supply numerous examples from the rankling family quarrels in our own society. In a tribe segmented into lineages, sibs, or moieties, however, the individual knows exactly where he stands in such instances. If both disputants are members of his own kin group, he is expected to remain neutral and to use his good offices to compose their differences. If neither is a member, the affair is none of his business. If one is a member but the other is not, he is expected to support his sibmate, regardless of the rights in the matter. In short, most conflict situations are simply and automatically resolved.
The Tenino of Oregon illustrate how conflicts in ceremonial obligations can arise in a sibless society. In this tribe weddings are solemnized by an elaborate series of property exchanges between the kindreds of the bride and the groom. The relatives of the bride, of both sexes, bring clothing, baskets, bags, vegetal foods, and other articles produced in the feminine sphere of economic activity. Those of the groom bring horses, skins, meat, and other masculine products. Each participant then exchanges his .eifts with a particular member of the other kindred. It nearly always happens, of course, that a number of people are related to both the bride and the groom, and are forced to decide on which side they shall participate; they cannot play two contradictory roles at once. Moreover, the numbers in either party must be the same. These problems are settled only after protracted discussions among the parties concerned and persons in authority, and they not infrequently generate jealousy, friction, and injured feelings.
Under unilinear descent such conflicts could never arise. All kin groups produced by patrilineal, matrilineal, or double descent are discrete social units. The role of every participant in a ceremonial activity, of every bystander in a dispute of any sort, is automatically defined for him by his kin group membership. This advantage may well account in considerable measure for the marked preponderance of unilinear descent throughout the world.
The kindred is roughly comparable to the lineage, not only in approximate size but also in the fact that the genealogical connections of all of its members to Ego are known and traceable. Are there larger bilateral kin groups which are comparable to the sib both in size and in the traditional rather than demonstrable bonds of kinship which unite their members? No such groups have received extensive theoretical consideration in the literature. Nevertheless, the results of the present study indicate that there is one type of larger bilateral kin group which is fairly common, and which exerts an influence upon kinship terminology and sex behavior comparable to that exerted by sibs and other recognized consanguineal kin groups.
This group is most clearly observable in the endogamous local community which is not segmented by unilinear consanguineal groupings of kinsmen. By virtue of the rule or strong preference for local endogamy, the inhabitants are necessarily related to one another through intermarriage, although they cannot always trace their exact kinship connections. They are consequently bound to one another not only by common residence but also by consanguinity, as is, in fact, usually specifically recognized. Within such a group the only social structuring is commonly into families, which may be of either nuclear, polygamous, or extended type. Except for family ties, the strongest sense of identification is usually with the community as a whole, which is viewed as a consanguineal unit in relation to other communities in a manner quite comparable to the attitude towards one's own sib in a unilinear society. In our sample, endogamous localized kin groups of this type are attested with reasonable certainty for the Aymara, Chiricahua, Comanche, Cuna, Inca, Kiowa Apache, Mentaweians, Nuba, Pawnee, Ruthenians, Shoshone, Sinkaietk, Siriono, Taos, and Wichita, and they are probably present in at least some of the following: Carib, Cayapa, Copper Eskimo, Kaingang, Mataco, Nambikuara, Tupinamba, and Washo. In view of the fact that this type of kin group will later be found to be substantially associated with other social phenomena, such as kinship terminology, it seems desirable to give it specific recognition and a distinctive name. So far as the author's knowledge goes, it has not been christened, and none of the terms that are standard in social organization appears applicable. He has therefore sought for a new name which would be as short and distinctive as “sib” and which already has both a local and a genealogical connotation.
Search led eventually to the social organization of ancient Attica, where a local group called the deme (pronounced like “deem”), roughly comparable to an English parish, was substituted for a unilinear descent group in the political reforms of Cleisthenes, who made membership in it hereditary 27. While our source does not reveal whether the original deme was endogamous, it was certainly a local group and, at least in later times, was also consanguineal. It therefore seems to suit our purposes adequately, and, in addition to being short and easily pronounceable, has the advantage of being already established in English dictionaries. That the term has appeared briefly in the anthropological literature 28, only to be quickly forgotten, does not prevent our reintroducing it with our own definition. Henceforth, therefore, we shall regularly employ “deme” for an endogamous local group in the absence of unilinear descent, especially when we are regarding it as a kin group rather than as a community.
The widespread tendency to extend incest taboos throughout any kin group, which will be considered in Chapter 10, naturally does not leave the deme unaffected. On the basis of the tradition of common descent among its members, exogamy can and does extend, in many instances, from the kindred to the deme, as from the lineage to the sib. When this occurs, local exogamy replaces local endogamy, and the constitution of the deme is fundamentally changed. Members are required to marry outside of the community, and in doing so naturally conform to the prevailing rule of residence. Exogamous demes, where unilinear descent has not become established, may be distinguished according to residence rules as patri-demes and matri-demes. If necessary to differentiate the original endogamous deme from them, it may be called the endo-deme. As will be seen in the next chapter, exogamous demes constitute one of the two principal origins of clans.

1. See M. Mead, Sex and Temperament in Three Primitive Societies (New York, 1935), pp. 176-7.
2. See R. Kennedy, “A Survey of Indonesian Civilization,” Studies in the Science of Society, ed. G. P. Murdock (New Haven, 1937), p. 291.
3. For a detailed exposition see G. P. Murdock, “Double Descent,” American Anthropologist, n.s., XLII (1940), 555-81. Cf. also R. F. Fortune, “A Note on Some Forms of Kinship Structure,” Oceania, IV (1933 ), 1-9.
4. See W.H.R Rivers, Social Organization (New York, 1924), p. 16.
5. R.H. Lowie, Primitive Society (New York, 1920). A second edition of this work (New York, 1947) contains a few significant corrections and addenda.
6. See R.H. Lowie, Primitive Society (New York, 1920), p. 111. This exceedingly useful term has not yet achieved the universal acceptance which it deserves.
7. See especially A.A. Goldenweiser, History, Psychology, and Culture (New York., 1933), .PP 213-356; J.G. Frazer, Totemism and Exogamy (4 vols., London, 1910); R.H. Lowie, Primitive Society (New York, 1920), pp. 137-45.
8. Cf. R.H. Lowie, Primitive Society (New York, 1920), pp. 157-8.
9. See R.S. Rattray, Ashanti (Oxford, 1923), pp. 77-8.
10 See H.G. Luttig, The Religious System and Social Organization of the Herero (Utrecht. 1934), pp. 58-67.
11. F. Calton, “Note on the Australian Marriage Systems,” Journal of the Royal Anthropological Institute, XVIII (1889), 70-2.
12. A.B. Deacon, “The Regulation of Marriage in Ambrym,” Journal of the Royal Anthropological Institute, LVII (1927), 325-42.
13. W.E. Lawrence, “Alternating Generations in Australia,” Studies in the Science of Society, ed. G.P. Murdock (New Haven, 1937), pp. 319-54.
14. A.R. Radcliffe-Brown, “Australian Social Organization,” American Anthropologist, XLIV (1947), 151-4.
15. See A.R. Radcliffe-Brown, “Regulation of Marriage in Ambrym,” Journal af the Royal Anthropological Institute, LVII (1927), 347.
16. See A.R. Radcliffe-Brown, “The Social Organization of Australian Tribes,” Oceania, I (1930), 43-5 et passim.
17. See A.R. Radcliffe-Brown, “Three Tribes of Western Australia,” Journal of the Royal Anthropological Institute, XLIII (1913), 190-3.
18. G.P. Murdock, “Bifurcate Merging,” American Anthropologist, n.s., XLIX (1947), 56, n. 2. 19. See especially J.J. Bachofen, Das Mutterrecht (Stuttgart, 1861); J.F. McLennan, Studies in Ancient History (London, 1876); L.H. Morgan, Ancient Society (New York, 1877).
20. See especially J.R Swanton, “The Social Organization of American Tribes,” American Anthropologist, n.s., VII (1905), 663-73; R. H. Lowie, ) “Social Organization,” American journal of Sociology, XX (1914 ), 68-97.
21. See especially W.J. Perry, The Children of the Sun (New York, 1923); W. Schmidt and W. Koppers, Volker und Kulturen (Regensburg, 1924); R.L. Olson, “Clan and Moiety in Native America,” University of California Publications in American Archaeology and Ethnology, XXX (1933), 351-422.
22. R Linton, The Study of Man (New York, 1936), p. 166.
23. R.H. Lowie, Primitive Society (New York, 1920), p. 157.
24. R. Linton, The Study of Man (New York, 1936), p. 169.
25. A.R. Radcliffe-Brown, “Patrilineal and Matrilineal Succession,” Iowa Law Review, XX (1935), 301-3. I
26. R. Linton, The Study of Man (New York, 1936), pp. 160-2. 166-7.
27. See E.M. Walker, “Athens: the Reform of Cleisthenes,“ Cambridge Ancient History, IV (New York, 1926), 142-8.
28. Cf. A.W. Howitt and L. Fison, “On the Deme and the Horde,” Journal of the Royal Anthropological Institute, XIV (1885), 142.

The Clan

Two major types of kin groups have previously been distinguished.

One, the residential kin group, is based primarily upon awe of residence. It necessarily includes both husband and wife, since they always reside together. Almost as inevitably it cannot include married brothers and sisters, since they are separated by incest taboos and residence rules, and can consequently rarely live together. Hence the relatives aggregated in a residential kin group always include some whose kinship ties are primarily affinal, as in the case of husband and wife or of stepparent and stepchild, and others whose ties are purely consanguineal, as in the case of parent and child or of two brothers or parallel cousins. The most characteristic of residential kin groups are the several varieties of families described in Chapters 1 and 2.

The second major type is the consanguineal kin group, which is regularly based on a rule of descent rather than of residence. In consequence, it almost never includes both husband and wife, or any other pair of affinal relatives, but always includes brothers and sisters, after as well as before their marriage. It also embraces all other consanguineal relatives under the prevailing rule of descent as far as connections are traced for the purposes of group membership. In view of these characteristics it can practically never be a residential unit. Its principal forms—the lineage, sib, phratry, moiety, section, kindred, and deme—have been analyzed in Chapter 3.

We now come to a third major type of kin group, which is based upon both a rule of residence and a rule of descent, insofar as the two can be reconciled, and which may therefore be called a compromise kin group. The problem of reconciliation arises only with unilocal residence and unilinear descent. Of the two principal varieties of bilateral consanguineal kin groups, the kindred is not a group at all, except from the point of view of a particular individual, and hence cannot be localized, whereas the deme is in the fullest sense both a residential and a consanguineal group without any necessity for compromise. We have already seen, however, that incest taboos and the residential cohabitation of husband and wife prevent the localization of a unilinear consanguineal kin group in its entirety under any rule of residence. This can be approximated only by combining a unilocal rule of residence with a consistent unilinear rule of descent and effecting a compromise whereby some affinal relatives are included and some consanguineal kinsmen excluded. Of the several possibilities, as we shall see, all involve the exclusion of the adult consanguineal kinsmen of one sex and the inclusion of the spouses of those of the opposite sex.
A compromise kin group is commonly larger than an extended family, but the alignment of kinsmen is identical. The principal distinction is the addition of a unilinear rule of descent as an integral factor in the structure of the group. The core of a unilocal extended family always consists of persons of one sex who are in fact unilinearly related, but this relationship is purely incidental, need not be formulated, and is frequently not even recognized. The bond of union is primarily and often exclusively residential. In a compromise kin group, on the other hand, the unilinear relationship of the core of the group is at least as crucial an integrating fact as the residential alignment.
The existence and essential nature of compromise kin groups has been dimly recognized for some time, but they have commonly been confused with the extended family on the one hand or with the sib on the other, depending upon whether the particular author has chosen to stress the residential or the consanguineal bond. Even Lowie 1, who has cleared up satisfactorily so much of the confusion about unilinear kin groups inherited from the past, has failed to make explicit the crucial distinction here involved, although he appears to understand it. The clarification of this issue, which is fundamental to an adequate comprehension of unilinear institutions, is the primary objective of the present chapter.
The selection of an appropriate term for compromise kin groups presents the most serious problem of nomenclature encountered in the present work. In the literature they are rarely distinguished from sibs, and the same term is usually applied indiscriminately to both groups. The term most widely used is clan, and it is this which we have chosen to adopt after considerable hesitation rather than clutter an already overburdened nomenclature with a new term. There are, it must be admitted, significant disadvantages in this choice. Foremost among them is that fact that “clan” has been used in two other distinct senses in the technical literature on social organization. From the time of Morgan until very recently most American anthropologists have used it to designate a matrilineal sib in contradistinction to a patri-sib, for which the term “gens” has been employed. This usage is now obsolescent, since increasing recognition of the essential similarity of sibs under the two rules of descent renders their terminological distinction no longer necessary. Far more serious is the use of “clan” by British anthropologists for any unilinear consanguineal kin group of the type which we, following Lowie, have termed the “sib.” Among recent American anthropologists, “clan” and “sib” have run a close race for acceptance in this sense. We have given preference to “sib” over “clan” primarily because the former has never been applied to kin groups other than consanguineal ones with unilinear descent, so that its use by us in this sense could lead to no confusion.
Despite the disadvantages of employing “clan” for compromise kin groups, and particularly the initial confusion which it may create in the minds of some readers, our choice has been dictated by a number of positive advantages. In the first place, the term itself is too apt and too widely known to be discarded completely, as we should have to do if we chose another term after having preferred “sib” as a designation for the most typical of unilinear consanguineal kin groups. In the second place, “clan” has in fact been frequently used by anthropologists for compromise kin groups, and is indeed almost the only term that has ever been applied to them. Even when specifically differentiated from sibs in the literature, they are most commonly called “localized clans.” Finally, our use and proposed redefinition of the word accord very closely with its popular connotations. Webster 2, for example, gives the following as the primary definition of clan: “A social group comprising a number of households the heads of which claim descent from a common ancestor…. ” Any group which comprises households related through their heads necessarily includes wives with their husbands but not married sisters with their brothers, and is consequently not a consanguineal kin group or sib but a kin group of the compromise type. Our proposal thus returns the word to its original non-technical meaning.
For a group to constitute a genuine clan it must conform to three major specifications. If any one of the three is lacking, the group is not a clan, however greatly it may 'resemqle one in composition and external appearance. In the first place, it must be based explicitly on a unilinear rule of descent which unites its central core of members. Unilocal extended families and exogamous demes reveal a composition identical with that of clans, from which they differ only in lacking the unifying principle of descent; unilinear descent is absent in the case of demes, absent or incidental in the case of extended families. In the second place, to constitute a clan a group must have residential unity. This cannot exist if the residence rule is inconsistent with that of descent, e.g., patrilocal or neolocal when the latter is matrilocal. Nor can it exist if any appreciable degree of individual deviation is permitted from the normal rule of residence. In the third place, the group must exhibit actual social integration. It cannot be a mere unorganized aggregation of independent families like those residing in a block in an American residential suburb. There must be a positive group sentiment, and in particular the in-marrying spouses must be recognized as an integral part of the membership.
Above all, it should be emphasized that a clan does not result automatically from the coexistence of compatible rules of residence and descent. Even if residence is strictly patrilocal and descent patrilineal, for example, and the households of the patrilineally related males are actually aggregated in a specific part of the community, the group does not necessarily constitute a clan. The residents may be merely related neighbors. The ethnographer must observe evidences of organization, collective activities, or group functions before he can characterize them as a clan.
The Dobuans of Melanesia provide an example of groups with the membership but not the organization of clans. In consequence of alternating residence and matrilineal descent, the members of a local community consist of unilinearly related males and females with the spouses who are residing temporarily with them. The noteworthy lack of any social integration between these in-marrying spouses and the native-born members of the community, however, prevents recognition of the local group as a genuine clan.
It has already been noted that the normal method of effecting a compromise between a rule of residence and a rule of descent in order to form a clan is by including one sex among the adult members of the consanguineal kin group, by excluding their siblings of opposite sex, and by including the spouses of the former. Logically, there are four ways in which this can be accomplished, since there are two unilinear rules of descent-patrilineal and matrilineal -and two sexes, either of which might theoretically be selected for inclusion as the core of the clan. Actually, only three of the four possibilities occur in the societies of our sample.
The first possibility in clan formation is to localize a patri-lineage or patri-sib around its male members through a rule of patrilocal residence. The group which results includes all the males and the unmarried females of the lineage or sib, together with the wives of the married males. The married females of the lineage or sib, the sisters of the males, are excluded because they have gone to live with their husbands in patrilocal residence in other clans. It is appropriate to designate such a group either as a patrilocal clan or as a patrilineal clan, since it depends equally upon rules of residence and descent. Following a useful suggestion by Lawrence 3, we shall ordinarily call it a patri-clan.
The distinction between the sib or lineage which is localized and the clan which results from its localization can perhaps be clarified by reference to our own society. Though we lack both sibs and clans, the reader can readily visualize their existence, assuming that our rule of residence were patrilocal rather than neolocal and that our rule of descent were patrilineal rather than bilateral. In such a case our surnames, which are patrilineally inherited, would correspond to both clans and sibs. All males with the surname Wolf, for example, would be members of the Wolf sib, and they would reside together in one or more localized groups, which could be called Wolf clans. All women, however, would change both their local groups and their surnames upon marriage. They would belong to the sibs of their brothers, to the clans of their husbands. In the case of males, the surname Wolf would indicate both sib and clar. membership. In the case of women, however, sib membership would be indicated by the maiden name, clan membership by the married name. Miss Mary Wolf would be a member of the Wolf sib, even after she married John Heron and joined the Heron clan. Mrs. James Wolf, nee Fox, would belong to a Wolf clan but to the Fox sib.
The second possibility in clan formation is to localize a matrilineage or matri-sib around its female members through a rule of matrilocal residence. The resulting matri-clan (matrilocal or matrilineal clan) embraces all the females and the unmarried males of the sib, together with the husbands of the married females. It does not include the adult males of the sib, since these have married and taken up residence with their wives in other matri-clans.
The third possibility consists in localizing a matri-sib around its male rather than its female members. This is accomplished through avunculocal residence, whereby an unmarried male leaves his paternal home to live with his maternal uncle and brings his wife there when he marries. The resulting group, which may be called an avunculocal clan or avuncu-clan, includes a group of matrilineally related males who stand to one another in the relation of uncle and nephew or of real or classificatory brothers, together with their wives and unmarried children, who of course belong to different sibs. The adult women of the clan are the wives and sisters-in-law of the men of their own generation. The adult and married women of the sib reside in other clans with their husbands.
Occasionally in an avuncu-clan the nephews may marry the daughters of their maternal uncles, with the result that at least some of the married women of the clan are the daughters of the older men. To the extent that this is true, the group represents a combination of matri-clan and avuncu-clan, with its females belonging to one sib and its males to another and with matrilocal residence characterizing the former and avunculocal residence the latter. An avuncuclan, in fine, resembles a patri-cian in the fact that its core consists of a group of unilinearly related males, but it resembles a matriclan in its acceptance of matrilineal rather than patrilineal descent as the basis of affiliation.
The fourth possibility in clan formation is as yet purely hypothetical. It would consist in localizing a patri-sib around its adult females. This could be accomplished by a cultural provision that unmarried females take up their residence with a paternal aunt and bring their husbands to the aunt's home when they marry. Such a rule of residence, it if should be discovered in future field research, might be called amitalocal on the basis of a suggestion by Lowie 4 that the term “amitate” be adopted for a special relationship with the father's sister paralleling “avunculate,” which is generally used for the comparable relationship with the mother's brother. The type of clan resulting from such a residence rule in conjunction with patrilineal descent would then be called an amitalocal clan or amita-clan.
Among the 250 societies of our sample, the available information, which is often fragmentary, indicates the presence of patri-clans in 72, matri-clans in 11, and avuncu-clans in 4. Clans are definitely absent in 131 tribes, while for 32 unilinear societies the data are too scanty to determine whether clans exist or not. Amita-clans are not reported in our sample, nor has the author ever encountered a description in his general ethnographic reading. That avuncu-clans are not simply a local peculiarity, dependent upon a unique concatenation of circumstances, is shown by the distribution of the four cases in our sample. Two instances (the Haida and Tlingit) are reported from the Northwest Coast of North America, one (the Longuda) from Northern Nigeria in Africa, and one (the Trobrianders) from Melanesia in Oceania.
The distinction between compromise and consanguineal kin groups, or specifically between clan and sib, is no idle exercise in classificatory ingenuity. On the contrary, it is a matter of profound functional importance. The two kin groups not only differ in constitution, but they appear to be characteristically associated with different functions. These can be illustrated from a tribe known to the author through personal field work.
The Haida of British Columbia, in addition to the nuclear family, are characterized by one type of residential kin group, the avunculocal extended family; by two types of consanguineal kin groups, the matri-sib and the matri-moiety; and by one type of compromise kin group, the avuncu-clan. Each extended family occupies a large plank dwelling. Each avuncu-clan comprises the inhabitants of a particular village-a group of matrilineally related adult males together with their wives, their unmarried or recently married daughters, and their young sons who have not yet left to join the household of a maternal uncle. The wives and children of the adult males of the clan belong, of course, to the opposite matri-moiety. A sib consists of the adult males of an avuncu-clan, and of their sisters and the daughters and young sons of the latter who all reside in other villages. A clan and the corresponding sib thus coincide in somewhat less than half of their membership-their male members over ten years of age, the approximate time at which the shift to avunculocal residence occurs.
The functions of the several Haida kin groups are quite distinctive. The moiety regulates marriage, which is strictly exogamous. It also channelizes rivahies and regulates ceremonial property exchanges. Potlatches, for example, are invariably given to members of the opposite moiety. The extended family is the unit of ordinary domestic life, of primary economic cooperation, of trade, and of property accumulation. The nuclear family, in addition to its usual functions, is the group which gives a potlatch; the wife is technically the donor in major potlatches, but she is assisted by her husband, and their children are the beneficiaries of the resulting enhancement in status. The clan is the community, i.e., the face-to-face group of daily social intercourse. It is also the basic political unit, each clan being independent of all others. All property rights in land are held by the clan, under the trusteeship of its chief.
Movable goods are owned by the extended family or by individuals. Intangible property rights, on the other hand, are vested in the sib. Tills group owns a fund of personal names, of ceremonial titles for houses and canoes, of totemic crests, and of exclusive rights to songs and ceremonies. Mythology, too, is largely associated with the sib, and it is this group which regulates inheritance and sucession. The sib, moreover, is the ceremonial unit; its members are invited collectively to feasts and potlatches, and they assist one another in preparing for and conducting these ceremonies when any one of them is the host. Finally, the duty of blood vengeance for the murder or injury of one of its members falls upon the sib. Warfare, however, is the function of the clan, whether motivated by vengeance, self-defense, or the desire for booty and slaves.
The role of Haida women in disputes between the men of two villages is especially illuminating. Those who have married from one community into the other find themselves tom by divided loyalties; they are wives and clansmen of one group of disputant men, sisters and sibmates of the other. So long as relations are merely strained, these women act as mediators and strive to settle the differences; they can, of course, move from one village to the other unmolested. If, however, matters reach the breaking point, and war ensues, the women cleave to their husbands rather than to their brothers 5. In the ultimate crisis, in short, clan ties override the bonds of sib membership. This is particularly enlightening because it contradicts the statement by Linton 6 that in societies organized on a consanguine basis, of which the Haida are certainly an outstanding example, “spouses are of only incidental importance.”
The ethnographic literature, despite the prevalent confusion, contains numerous intimations of a similar distribution of functions between clan and sib. In general, the clan seems to function primarily in the economic, recreational, political, and military spheres, while the sib is associated with totemism and ceremonial, acts as a unit in life crisis situations, and regulates marriage and inheritance.
Future field research and a critical reexamination of the existing literature will doubtless clarify the entire situation.
With respect to relative size, clans fall into two major categories. The larger is coextensive with the local community, as in the case of the Haida, and may therefore be designated as a clan-community. The smaller type of clan forms merely a segment of the community, such as a ward in a village, or a hamlet when the community consists of a cluster of segregated clans. Since the Spanish word barrio can connote divisions of either type, the group will be called a clan-barrio. Of the societies in our sample with patri-clans, 45 are characterized by clan-communities and 27 by clan-barrios. Only clan-communities are found in the four avunculocal tribes. The preponderance of clan-communities disappears in matrilocal societies for reasons which will be made clear in Chapter 8. Nine of those in our sample possess clan-barrios exclusively, whereas only two are organized into clan-communities, and both of these, the Vedda of Ceylon and the Yaruro of Venezuela, live in migratory bands rather than settled villages.
For purposes of exposition, the chapter has been written up to this point as though clans originated in societies already possessing unilinear consanguineal kin groups, through a process of localization resulting from the application of a unilocal rule of residence consistent with the preexisting rule of descent. To have complicated the synchronic analysis with diachronic considerations would have resulted only in confusing the lay reader. Now that the analysis is complete, however, the problem of origins may be faced. As this is done, the reader should be specifically warned to disregard the implicit assumption of the temporal priority of the sib to the clan. In actuality, the clan is almost universally the first to make its appearance, and gives rise to the lineage or sib, rather than vice versa. While much of the proof must be deferred until Chapter 8, the essential facts regarding sequences must be summarized here, in order to forestall misunderstanding.
It may be stated categorically, disregarding only the most exceptional circumstances, that lineages and sibs invariably arise out of clans by the extension of the recognition of unilinear affiliation from the sex which forms the core of the clan to their siblings of opposite sex who have moved away in marriage. Until this is done, by some mechanism such as sib names, totemism, or inherited taboos, the society possesses clans but not sibs. There are presumably such cases in our sample, but so widespread is the confusion between clan and sib in the literature that wherever we have found indisputable evidence of the former we have felt compelled to assume the existence of the latter. Once sibs or lineages have come into existence, provided they have acquired the characteristic of exogamy, they lead an independent life and are capable of surviving long after the clans from which they sprang have disappeared. No fewer than 56 of the societies in our sample, for instance, lack clans though they possess unilinear consanguineal kin groups-matrilineal in 30 cases, patrilineal in 23, and groups of both types in 3. In 32 other tribes with sibs or lineages, clans are not reported and may very possibly be absent. In other words, on statistical grounds a unilinear society is approximately as likely to lack clans as to possess them.
Clans doubtless originate in some instances merely from a unilocal rule of residence, the incidental result of which is to assemble in one locality a number of unilinearly related adults of one sex together with their spouses and children. In such instances, however, both a principle of organization and a rule of descent must be developed. It is probable, therefore, that clans arise more often from kin groups which, in addition to a similar aggregation of relatives, already possess a principle of organization and consequently require only the development of a rule of descent to be converted into clans. There are two types of kin groups which meet these specifications-unilocal extended families and exogamous demes. All that is necessary to convert an extended family into a clan-barrio, or an exogamous deme into a clan-community, is to give cultural recognition to the unilinear affiliation of the core of the group. When this is extended to their out-marrying siblings, lineages are produced in the one case, sibs in the other. Both theoretical considerations and the evidence of ethnography, therefore, make it probable that clans, and ultimately lineages and sibs as well, arise in most instances either from extended families or from demes which have become exogamous. The initial phase in the evolution from extended family to clanbarrio is exemplified by the Havasupai of Arizona. On the basis of a rule of matri-patrilocal residence this tribe has developed groups called “camps,” which are actually patrilocal extended families. Spier 7 has shown very convincingly how a relatively slight accentuation of factors already in existence, such as an increased emphasis upon patrilineal inheritance and an extension of kinship recognition in the paternal line, would give rise to genuine patricians. The Tongans of Polynesia are one of several tribes in our sample which have achieved this step without, however, having yet developed a rule of exogamy.
The Mentaweians of Indonesia illustrate the initial stages of a parallel development from matrilocal extended families to clanbarrios. In this society the village, which has the characteristics of an endogamous deme, is divided into sections, each consisting of a communal house, a number of surrounding dwellings, and outlying field-houses. These local divisions, which are the primary social, economic, and religious units of Mentaweian society, bear many of the earmarks of extended families comprising a number of nuclear families occupying separate houses. They have not yet become exogamous, however, nor has recognition been accorded to matrilineal descent, which alone is needed to convert them into matri-clans.
A clan-community may arise from a clan-barrio through a process of population growth and fission, which ultimately leaves only a single clan in a community which previously had several. At least as commonly, however, it can develop directly out of an exogamous deme. As incest taboos spread throughout a deme, an inherent (tendency in all consanguineal kin groups (see Chapter 10), the rule of endogamy becomes gradually converted into a preference for local exogamy. This produces matri-demes where the rule of residence is matrilocal, patri-demes under patrilocal residence. For reasons to be explained in Chapter 8, matri-demes are rare and are confined almost exclusively to societies with a migratory band organization. In our sample only the Arapaho and Cheyenne reveal matri-demes, incipient in both cases. If either tribe were to adopt strict local exogamy with matrilocal residence, the band would consist of matrilineally related women with their husbands and children. If then this unilinear connection were culturally recognized through a rule of matrilineal descent, and if the out-marrying men retained their kin affiliations with the band members, the tribe would be characterized by matrilineal clan-communities and matrisibs like the Vedda and Yaruro. Far more common, however, is the transition from patri-demes to patri-cians and patri-sibs. Our sample includes 13 societies with bilateral descent and a marked tendency toward local exogamy, i.e., with patri-demes. These are the Blackfoot, Eromangans, Hupa, Ona, Quinault, Semang, Shasta, Takelma, Teton, Walapai, Wapisiana, Yaghan, and Yurok. The normal process by which patri-demes may be converted into patrilocal clan-communities, with the eventual emergence of patri-sibs, may be illustrated by reference to the Hupa of northwestern California.
In consequence of the rules of patrilocal residence and local exogamy, the inhabitants of a Hupa village or settlement consist mainly of patrilineally related males and their children with wives whom they have secured from other communities. There is, however, one important exception to the patrilocal rule. Certain poor men, who constitute a not insignificant minority, pay only half of the ordinary bride-price and reside matrilocally with their parents-in-law, working off the balance through bride-service. The children of these “half-married” men continue to live in their mothers' village and are recognized as members thereof. Since a household may include daughters of the owner with their half-married husbands, as well as sons with their wives and children, it is as much a bilocal as a patrilocal extended family. Lowie 8 has clearly pointed out two of the three modifications which would be necessary to produce patri-clans, then patri-sibs.
First, the patrilocal rule of residence would have to be made invariable; this would clarify the patrideme structure. Second, patrilineal descent would have to be recognized as a bond of union among the resident males; this would produce genuine patrilocal clan-communities.
Third, “there must be a means of fixing the affiliation of the female no less than of the male members,” e.g., by giving a common name to all persons born in the village whereby the original affiliation of out-marrying women could be permanently indicated. Once this were accomplished, full-fledged patri-sibs would have come into existence.
The nearly universal derivation of lineages and sibs from clans, and of clans from unilocal extended families and exogamous demes accords with the views of all American anthropologists who have given thoughtful consideration to the problem. More recently than Spier and Lowie, Titiev 9 has come to the same conclusion. The evidence in favor of this hypothesis is overwhelming, as will be seen in Chapter 8. The facts of diachronic sequence render misleading the common designation of clans as “localized” consanguineal kin groups, and demand the acceptance of a distinctive term. If “clan” strikes the reader as unsatisfactory, he is invited to suggest a more appropriate name.

1. R.H. Lowie, Primitive Society (New York, 1920), pp. 111-85.
2. Webster's New International Dictionary of the English Language (rev. edit., Springfield, 1923), p. 409.
3. W.E. Lawrence, “Alternating Generations in Australia,” Studies ln the Science of Society, ed. G. P. Murdock (New Haven, 1937), p. 319.
4. R.H. Lowie, in American Anthropologist, n.s., XXXIV (1932), 534.
5. The common confusion between sib and clan led Swanton to make one of his rare misstatements of fact in regard to this point, for he states that in such cases women support their sibmates against their husbands. See J. R. Swanton, “Contributions to the Ethnology of the Haida,” Memoirs of the American Museum of Natural History, VIII (1909), 62.
6. R. Linton, The Study of Man (New York, 1986), p. 159.
7. L. Spier, “A Suggested Origin for Gentile Organization,” American Anthropologist, n.s., XXIV (1922), 489.
8. R.H. Lowie. Primitive Society (New York, 1920), pp. 157-8.
9. M. Titiev, “The Influence of Common Residence on the Unilateral Classification of Kindred,” American Anthropologist, n.s., XLV (1943), 511-30.

The Community

Anthropologists from Morgan to Lowie have shown far more interest in the forms of the family, the sib, and the clan than in the organization of social groups upon a strictly local basis. Sociologists, on the other hand, have for some time manifested a strong interest in community organization, and a parallel concern has recently been developing in anthropology, with especially noteworthy contributions from Steward 1 and Linton 2.
The sociological term community is here chosen in preference to less definite or less descriptive alternatives, such as “ocal group” and “band,” as the generic designation for groups organized on a predominantly local basis. It has been defined as “the maximal group of persons who normally reside together in face-to-face association.” 3 The community and the nuclear family are the only social groups that are genuinely universal. They occur in every known human society, and both are also found in germinal form on a sub-human level.
Nowhere on earth do people live regularly in isolated families. Everywhere territorial propinquity, supported by divers other bonds, unites at least a few neighboring families into a larger social group all of whose members maintain face-to-face relationships with one another. Weyer 4, in demonstrating this fact for the Eskimo, has pointed out that community organization provides individuals with increased opportunities for gratification through social intercourse, with more abundant sustenance through cooperative food-getting techniques, and with insurance against temporary incapacity or adversity through mutual aid and sharing. To these advantages may be added protection through numbers and the economies possible with specialization and a division of labor. The chances of survival thus seem to be materially enhanced through community organization, and this, together with the directly perceived gains, doubtless accounts for its universality.
Communities differ in type with their mode of life. Where subsistence depends largely upon gathering, hunting, or herding, which usually require migration from place to place at dillerent seasons of the year, the local group consists typically of a number of families who habitually camp together. This type of community is called a band. Agriculture, on the other hand, favors more permanent residence in a single settlement, though exhaustion of the land may compel the community to move to a new site every few years. Fixed residence is also consistent with a fishing economy and even with a bunting economy under exceptional conditions where game is plentiful and non-migratory. With more or less settled residence, the community may assume the form either of a village, occupying a concentrated cluster of dwellings near the center of the exploited territory, or of a neighborhood, with its families scattered in semi-isolated homesteads, or of some compromise between the two, like the rural American town with its dispersed farm homesteads and its local center with church, school, post office and general store. It is also possible for people to live in settled villages at one season of the year and in migratory bands at another. Of the 241 societies in our sample for which information is available, 39 are organized in bands, 13 in neighborhoods lacking prominent nuclei, and 189 in villages or towns.
In size, the community at its lower limit, approached for example by the Reindeer Chukcbee, consists of two or three families. The upper limit is seemingly set by “the practical impossibility of establishing close contacts with developing habitual attitudes toward any great number of people.” 5 For this reason, presumably, large urban aggregations of population tend to become segmented, when geographical mobility is not excessive, into local districts or wards which possess the outstanding characteristics of communities. A study by Goodenough 6 reveals a maximum range of from 13 to 1,000 in average community population, with 50 as the mean for tribes with migratory bands 7, 250 for those with neighborhood organization, and 300 for those with settled villages. The normal size of the community was shown by the same study to depend largely upon the prevalent type of food quest. Under a primarily hunting, gathering, or fishing economy, for example, the community averages somewhat fewer than 50 persons, whereas under an agricultural economy with animal husbandry it attains a mean population of about 450.
The community appears always to be associated with a definite territory, whose natural resources its members exploit in accordance with the technological attainments of the culture. Under a hunting or gathering economy, the lands of the community are ordinarily owned and exploited collectively 8, although in some instances, as Speck 9 bas shown for many of the Algonquian tribes of northeastern North America, they are divided into individual family tracts. The situation tends to be similar in herding societies. Under agriculture, the tillable land is sometimes collectively owned and periodically redistributed among families. Much more frequently, however, it becomes allocated as feudal or private property, although the nonagricultural portions of the community's territory may continue to be collectively owned and utilized. The territorial basis of the community survives even under a mercantile or industrial economy, despite the decline in the relative importance of land as a source of livelihood.
In consequence of its common territory and of the interdependence of its constituent families, the community becomes the principal focus of associative life. Every member is ordinarily acquainted more or less intimately with every other member, and has learned through association to adapt his behavior to that of 1 each of his fellows, so that the group is bound together by a complex network of interpersonal relationships. Many of these become culturally patterned, yielding standardized relationships like those of kinship and those based on age and sex status, which facilitate social intercourse, and many are aggregated into clusters around common interests, forming groups such as clans and associations which help to bind the families of the community to one another.
Since it is mainly through face-to-face relations that a person's behavior is influenced by his fellows-motivated, cued, rewarded, and punished-the community is the primary seat of social control. Here it is that deviation is penalized and conformity rewarded. It is noteworthy that ostracism from the community is widely regarded as the direst of punishments and that its threat serves as the ultimate inducement to cultural conformity. Through the operation of social sanctions, ideas and behavior tend to become telatively stereotyped within a community, and a local culture develops. Indeed, the community seems to be the most typical social group to support a total culture. This, incidentally, provides the theoretical justification for “community studies,” a field in which anthropologists, sociologists, and social psychologists alike have shown a marked interest in recent decades.
Under conditions of relative isolation, each community has a culture of its own. The degree to which this is shared by neighboring local groups depends largely upon the means and extent of intercommunication. Ease of communication and geographical mobility may produce considerable cultural similarity over wide areas, as, for example, in the United States today, and may even generate important social cleavages which cut across local groupings, as in the case of social classes. For most of the peoples of the earth, however, the community has been both the primary unit of social participation and the distinctive culture-bearing group.
United by reciprocal relationships and bound by a common cult ture, the members of a community form an “in-group,” 10 characterized by internal peace, law, order, and cooperative effort. Since they assist one another in the activities which gratify basic drives, and provide one another with certain derivative satisfactions obtainable only in social life, there develops among them a collective sentiment of group solidarity and loyalty, which has been variously termed syngenism, we-feeling, esprit de corps, and consciousness of kind.
Social life, despite the manifold advantages and rewards which reinforce it, also involves incidental frustrations. The individual , must curb certain of his impulses if he is to secure the cooperation of his fellows, and when he fails to do so he experiences the application of painful social sanctions. These frustrations, as always, generate aggressive tendencies 11. The latter cannot, however, be fully expressed within the in-group, lest mutual aid be withdrawn and further sanctions imposed. Consequently they are displaced toward the outside and drained off in the form of antagonistic sentiments and hostile behavior toward other groups. Intergroup antagonism is thus the inevitable concomitant and counterpart of in-group solidarity.
The tendency to exalt the in-group and to depreciate other groups, a phenomenon technically known as “ethnocentrism,” 12 though perhaps originally associated primarily with the community, has, with broadening social horizons, become characteristic of all human social groups. Today, for example, it runs the gamut from “local pride,” “college spirit,” and the esprit de corps of a business organization to religious intolerance, race prejudice, the “class struggle,” and international conflict. However deplorable from an ethical point of view, it is as inevitable as social life itself; at best it is capable only of being directed into channels that are socially less seriously disruptive.
Since its members are experienced in face-to-face cooperation, a community is ordinarily able to achieve concerted action, at least in emergencies, whether it does so under informal leaders or under chiefs and deliberative bodies with culturally d.eflned authority and functions. Moreover, as the fundamental locus of social control, it ' maintains internal order and conformity to traditional norms of behavior, if not through formal judicial organs and procedures, at least through the collective application of sanctions when public opinion is aroused by serious deviations. Basically, then, the community is a political group, as well as a localized, face-to-face, culture-bearing in-group. In it is to be sought the germ of government, however simple and informal the organs and procedures. It should be pointed out incidentally that government, as seen in cross-cultural perspective, has a second primary function. Besides serving as a means of channelizing collective action and social control, which justifies it to the governed, it offers to those in authority an opportunity to use their power for selfish aggrandizement. To the barbaric chieftain, the feudal lord, and the municipal boss, alike, accrue special privilege and pelf. So long as rulers preserve law and order, and their exploitative activities are not disproportionate to the social services they render, people do not ordinarily begrudge them their personal emoluments. Excesses in exploitation, however, tend to precipitate changes in the governing personnel.
Social relationships, even face-to-face ones, are probably never confined exclusively to the community, unless it is completely isolated, like the Polar Eskimos who, when first visited by Ross 13, were surprised to discover that they were not the only human inhabitants of the earth. Trade, intermarriage, and other forces create personal ties between members of different communities, on the basis of which peace and order may be widely extended. The war-likeness and atomism of simple societies have been grossly exaggerated. Primitive man is as capable as ourselves of perceiving the advantages to be obtained through peaceful intercourse with his neighbors, and of controlling his ethnocentric prejudices in order to reap these advantages. Even in regions where war is endemic, it is not waged all of the time nor with all surrounding groups; at the worst, armed truces or temporary alliances occur. Far more commonly, however, peaceful intercommunication prevails as the norm over wide areas.
The extension of personal relationships beyond the community may be facilitated by various cultural devices, e.g., local exogamy, blood brotherhood, safe-conduct, and market peace. It may be regularized by the development of social groups which cut across community lines, e.g., sibs, religious sects, and social classes. Finally, it may be consolidated by political unification, by the organization of a number of local groups under a single district, tribal, or state government. While many societies have followed this last course, an approximately equal number have developed no genuine political integration transcending the community. Evidence on pre-European governmental organization is available for 212 of our sample societies. In 108, each community is politically independent; in 104, definite governmental institutions unite several or many communities into larger organized groups of varying magnitude.
Among the factors favoring wider political organization, settled life appears to be peculiarly important. Table 10 shows that the bands of migratory tribes are usually politically independent, whereas the villages and settlements of sedentary populations are more commonly organized into larger aggregates.

Community Organization Bands Neighborhoods Villages Totals
Politically independent 28 8 68 104
Politically dependent 5 4 93 102
Totals 33 12 161 206

The problem of achieving concerted action and maintaining law and order becomes far more complex in a larger political society than in a single community. Informal modes of consensus, reciprocity, and social control do not operate where face-to-face association is lacking, and must be supplemented by formal mechanisms and procedures. The interpersonal relationships which bind the members of the larger society together are, of necessity, relatively abstract or conventional rather than concrete or face-to-face. To be sure, they are ordinarily patterned after the intimate relationships developed within the community, but these become formalized and stereotyped as they are extended. The habits of personal interaction which largely govern the relationship of a villager and his local headman, for example, are conventionalized in terms of formal etiquette and of explicitly defined rights and duties when they are extended to apply to the impersonal relationship of a subject to his tribal chief or king. Similarly, rules of judicial procedure tend to supplant informal discussion, systems of taxation and tribute to replace gift-giving, and specialized officials to take over the several functions of the unspecialized local headman.
Even with complex governmental organization, the community normally survives as a political unit, albeit a subordinate one, and a relative simplicity and face-to-face quality still characterize, as a rule, its regulative forms 14. For this reason, comparative studies of community organization are not vitiated by differences in political complexity. One may, however, seriously question the validity of those comparative studies of government which deal with the largest political aggregates in diverse societies, whether they be communities, organized tribes, or complex states. The Arunta band and the Inca empire, for example, are not comparable units, although it might well be profitable to compare the former with the local Peruvian ayllu, or the governmental institutions of the Incas with the Dahomean monarchy. No special analysis of political structures was made for the present study, and none will be attempted here. The community, , however, is one of the social groupings which we shall find to operate significantly in the channeling of kinship nomenclature and sexual behavior, and for this reason it has been necessary to analyze it in its relations to the larger political society as well as to its constituent kin groups.
One type of social structure which often transcends the community is the organization into social classes. Information on class stratification was assembled in the hope that the material might prove significant in the interpretation of sexual and kinship behavior. Although this hope has not on the whole been realized, the data are summarized in Table 11 as possibly of general interest.

Social Class Stratification Slavery Present Slavery Absent No Data on Slavery Totals
Complex structure of social classes 16 14 2 32
Hereditary aristocracy and commoners 15 18 6 39
Social classes based directly on wealth 10 5 0 15
Wealth distinctions without formal classes 7 16 3 26
Social classes absent 0 72 2 74
No data on class stratification 14 8 42 64
Totals 62 133 55 250

Slavery is distinguished in the table from other types of class structure, and the societies possessing and lacking slaves as a definite status group are enumerated in separate columns according to the class typing of the rest of the population. When war captives receive little differential treatment and are speedily adopted into the tribe, a society is considered as lacking a true slave class. A class structure is ranked as complex if it includes three or more definitely stratified groups other than slaves, or if it is complicated by the presence of hereditary and endogamous castes. A distinction is made between types of class structure based predominantly upon wealth and those in which priviliged status is reported as primarily hereditary. For a number of societies, differences in wealth are reported to exist but to be associated with no important differences in behavior, thus resembling individual distinctions in skill, valor, and piety rather than status gradations in the stricter sense. These have been distinguished in the tabulation from other classless societies on the one hand and from wealth-stratified societies on the other.
As might be expected, social stratification is especially characteristic of sedentary populations. Slavery, for example, is reported present in 55 societies with settled villages or neighborhoods and absent in 94, whereas it occurs in only 3 of the tribes organized in migratory bands and is specified to be absent in 33. Genuine social classes appear in none of the societies of our sample that are organized in bands, but occur in a majority of those with settled communities, as Table 12 reveals.

Class Stratification Bands Settled Communities Totals
Complex structure of social classes 0 31 31
Hereditary aristocracy and commoners 0 38 38
Social classes based directly on wealth 0 14 14
Wealth distinctions without formal classes 7 19 26
Social classes absent 27 44 71
Totals 34 146 180

Social classes operate not only to unite members of different local groups but also to segment the community itself and to complicate its social structure. Thus a village may be divided into nobles and commoners or into a number of castes. Participation tends to be greater within such groups than between them, and significant cultural differences may emerge. It has been shown by Warner 15 for example, that a typical New England city is segmented horizontally into six social classes, each with its distinctive cultural characteristics, that intimate social participation is confined primarily to members of the same “clique” within a social class and secondarily to persons belonging to cliques in the same stratum, and that intercourse between classes takes place largely through more formal associations which override class boundaries.
Probably the most significant differences in the internal organization of the community result from the varying ways in which its structure is integrated with that of the varying types of kin groups. In many instances, as has been seen in Chapters 3 and 4, the community itself may be a kin group. Local groups of this type may be collectively designated as kin-communities. Among the 222 societies in our sample for which sufficient information is available on community organization, there are 81 with kin-communities. They include 15 with endogamous bilateral demes, 13 with exogamous patri-demes, 2 with matri-demes, 45 with patrilocal clancommunities, 2 with matrilocal clan-communities, and 4 with avunculocal clan-communities. In some other societies the community is normally divided into a number of clan-barrios. Local groups of this type may be called segmented communities. In our sample, 36 societies are characterized by segmented communities—27 with patrilocal clan-barrios, and 9 with matrilocal clan-barrios. Local groups which are neither segmented into clans, nor themselves organized as clans or demes, may be called unsegmented communities. In our sample, 105 societies possess unsegmented communities. In 48 of them neither clans nor extended families are present, and in 17 others clans are absent and extended families unreported. If clans are absent but extended families are present, the community can be regarded as only partially rather than completely unsegmented. This is the case in 40 of our societies—7 with bilocal extended families, 19 with patri-families, 10 with matri-families, and 4 with avuncu-families.
Communities of any of the above types may be further classified, on the basis of the presence or absence of social classes, as stratified communities or unstratified communities. In our own society, for example, communities are normally stratified but unsegmented.
The classification proposed by Steward 16 for band organization may be compared with that presented above. Steward's “patrilineal band,” which is said to be characterized by “land ownership, political autonomy, patrilocal residence, band or local exogamy, and patrilineal land inheritance,” 17 includes both our patri-deme and our patrilocal clan-community. His “matrilineal band” embraces our matri-deme and our matrilocal clan-community. His “composite band,” which is stated to differ from the “patrilineal band” in not having “band exogamy, patrilocal residence, or land inheritance by patrilineal relatives,” 18 would include both the endogamous deme and the unsegmented community in our classification, and probably the segmented community as well.
A recurrent feature of community organization, noted by Linton 19, is an internal division into factions, usually two in number. We need instance here only the famous Tartharol and Teivaliol divisions of the Todas, the rivalrous districts of Faea and Ravenga on the tiny isle of Tikopia, the “hostile” and “friendly” factions among the Hopi, and the moiety cleavages of the Apinaye and many other tribes. Miner 20 has described a striking dual alignment in a rural French-Canadian parish, based ostensibly on affiliation with different political parties.
So widespread are such factional divisions, so frequently is their number precisely two, so commonly do they oppose one another in games and other activities, and so often are their reciprocal relations marked by rivalry, boasting, and covert forms of aggression that the phenomenon seems scarcely accidental. Ethnocentrism suggests a possible common function. A dual organization of a community, or of a larger social group, may provide a sort of safety valve whereby agression generated by in-group disciplines may be drained off internally in socially regulated and harmless ways instead of being translated into out-group hostility and warfare. If this highly tentative hypothesis is valid, opposing factions should be more characteristic of peaceful than of warlike communities. Perhaps herein lies the fundamental social justification of a two-party political system in a modern democratic state.
The analysis of familial, kin, and local groups in the foregoing chapters by no means represents a complete survey of human social organization. Economic, recreational, religious, and ceremonial associations, for example, have been mentioned but not discussed, and the same is true of age, sex, and status groupings. Only a fraction, indeed, of the interpersonal and group relationships which constitute the social situations within which man's behavior is learned and enacted have been presented, and the environmental and technological factors, which are likewise of great importance in influencing behavior, have been omitted altogether. The nature of our primary scientific task has limited us to a consideration of the social groups which appear peculiarly effective in channeling kinship and sexual behavior.

1. J.H. Steward, “The Economic and Social Basis of Primitive Bands,” Essays in Anthropowgy Presented to A.L. Kroeber (Berkeley, 1936), pp. 331-50.
2. R. Linton, The Study of Man (New York, 1936), pp. 209-30.
3. G.P. Murdock, C.S. Ford, A.E. Hudson, R. Kennedy, L.W. Simmons, and J.W.M. Whiting, “Outline of Cultural Materials,” Yale Anthropological Studies, n (1945), 29.
4. E.M. Weyer, The Eskimos (New Haven, 1932), pp. 141-4. Cf. also, J.H. Steward, “The Economic and Social Basis of Primitive Bands,” Essays in Anthropology Presented to A.L. Kroeber (Berkeley, 1936), pp. 332-3.
5. R. Linton, The Study of Man (New York, 1936), p. 218.
6. W. H. Goodenough, “Basic Economy and the Community” (unpublished article, 1941). This study, undertaken in the files of the Cross-Cultural Survey at the suggestion of Professor W.F. Ogburn, covered 40 tribes for which reliable population data were available.
7. An independent estimate by Steward also arrives at 50 persons as the average population of a band and finds, in addition, that the area exploited by a band averages approximately 100 square miles. See J.H. Steward, “The Economic and Social Basis of Primitive Bands,” Essays in Anthropology Presented to A.L. Kroeber (Berkeley, 1936), p. 333.
8. Cf. J.H. Steward, “The Economic and Social Basis of Primitive Bands,” Essays in Anthropology Presented to A.L. Kroeber (Berkeley, 1936), pp. 332-S.
9. F.G. Speck, “The Family Hunting Band as the Basis of Algonkian Social Organization,” American Anthropologist, n.s., XVII (1914), 289-305; “Family Hunting Territories and Social Life of Various Algonkian Bands of the Ottawa Valley”; Memoirs of the Canada Department of Mines Geological Survey, LXX (1915), 1-10; “Kinship Terms and the Family Band among the Northeastern Algonkian,” American Anthropologist, n.s., XX (1918), 143-61; “Mistassini Hunting Territories in the Labrador Peninsula,” American Anthropologist, n.s., XXV (1923), 452-71; “Family Hunting Territories of the Lake St. John Montagnais,” Anthropos, XXII (1927), 387-403; Penobscot Man (Philadelphia, 1940), pp. 203-12.
10. Cf. W.G. Sumner, Folkways (Boston, 1906), p. 12.
11. Cf. J.Dollard, L. W. Doob, N. E. Miller, O.H. Mowrer, and R.R. Sears, Frustration and Aggression (New Haven, 1939).
12. See W. G. Sumner, Folkways (Boston, 1906), p. 13; G.P. Murdock, “Ethnocentrism,” Encyclopaedia of the Social Sciences, V (1931), 613-14.
13. J. Ross, A Voyage of Discovery (London, 1819), p. 110.
14. Cf. the New England “town meeting” in our own society.
15. W.L. Warner and P.S. Lunt, The Social Life of a Modem Community (New Haven, 1941).
16. J.H. Steward, “The Economic and Social Basis of Primitive Bands,” Essays in Anthropology Presented to A.L. Kroeber (Berkeley, 1936), p. 331.
17. Ibid., p. 334.
18. Ibid., p. 338.
19. R. Linton, The Study of Man (New York, 1936), p. 229.
20. H.M. Miner, St. Denis (Chicago, 1939), pp. 58-00, 68-9.

Analysis of Kinship

The scientific significance of kinship systems was first appreciated by Morgan 1 in what is perhaps the most original and brilliant single achievement in the history of anthropology. That many of Morgan's particular interpretations are no longer acceptable does not diminish the luster of his work. Since his day, major contributions to the theory and analysis of kinship have been made by Rivers, Kroeber, Lowie, and Radcliffe-Brown, and important additional light has been shed on the subject by Aginsky, Eggan, Evans-Pritchard, Gifford, Kirchhoff, Lawrence, Lesser, LeviStrauss, Malinowski, Opler, Sapir, Brenda Seligman, Spier, Spoehr, Tax, Thurnwald, Warner, White, and others. No other anthropological topic, in all probability, has been the beneficiary of so much creative effort. Hence the present author, in sketching the background for his own contribution, must perforce draw heavily on the work of his predecessors.
A kinship system differs in one important respect from the types of social organization previously considered. In the various forms of the family, sib, clan, and community, interpersonal relationships are structured in such a manner as to aggregate individuals into social groups. A kinship system, however, is not a social group, nor does it ever correspond to an organized aggregation of individuals. It is merely, as the name implies, a structured system of relationships, in which individuals are bound one to another by complex interlocking and ramifying ties. Particular kinship bonds, isolated from others, may and often do serve to unite individuals into social groups, such as a nuclear family or a lineage, but kinship systems as wholes are not, and do not produce, social aggregates.
The point of departure for the analysis of kinship is the nuclear family 2. Universally it is in this social group that the developing child establishes his first habits of reciprocal behavior, his first interpersonal relationships. He learns to respond in particular ways toward his father, his mother, his brothers, and his sisters, and to expect certain kinds of behavior in return. His responses, however individualized at first, are gradually modified, as learning and socialization progress, to conform in general to the prevailing cultural norms. Once learned, his behavior in these primary intra-family relationships tends to be extended or “generalized” to persons outside of the family as his circle of personal contacts broadens 3. Such generalized behavior is. rewarded, and reinforced when consistent with cultural norms. Uherwise it is not rewarded, or is even punished, and thus becomes extinguished. Discrimination arises, and a situation is created in which distinctive modes of behavior toward other persons can occur and become established, either by trial and error or through imitative learning. In either case, it is the parents, elder siblings, other relatives, and neighbors who set the standards and exert the pressures which ultimately produce conformity with social expectations.
Intra-family relationships are not only the first to be learned in infancy and childhood; they continue to be an individual's most intimate relationships in adulthood. The child, now grown up and married, tends to recreate with his own children and spouse the behavior which his parents exhibited toward him, his siblings, and one another. Family relationships are of necessity highly functional, since they are universally involved in many of life's most important activities-economic cooperation, household routines, sex, reproduction, child care, and education. It is scarcely surprising, therefore, that they set the standard for all other kin relationships-standards to which the latter must conform or from which they must be differentiated.
Within the nuclear family are found eight characteristic relationships. Though functionally differentiated, all tend to be characterized, as compared with extra-family relationships, by a high degree of reciprocal cooperation, loyalty, solidarity, and affection. Despite cultural differences, each of the eight primary relationships reveals a markedly similar fundamental character in all societies, in consequence of the universality of the family's basic functions. These relationships, with their most typical features, are as follows:

  • Husband and wife
    economic specialization and cooperation; sexual cohabitation; joint responsibility for support, care, and upbringing of children; well defined reciprocal rights with respect to property, divorce, spheres of authority, etc.
  • Father and son
    economic cooperation in masculine activities under leadership of the father; obligation of material support, vested in father during childhood of son, in son during old age of father; responsibility of father for instruction and discipline of son; duty of obedience and respect on part of son, tempered by some measure of comradeship.
  • Mother and daughter
    relationship parallel to that between father and son, but with more emphasis on child care and economic cooperation and less on authority and material support.
  • Mother and son
    dependence of son during infancy; imposition of early disciplines by the mother; moderate economic cooperation during childhood of son; early development of a lifelong incest taboo; material support by son during old age of mother.
  • Father and daughter
    responsibility of father for protection and material support prior to marriage of daughter; economic cooperation, instruction, and discipline appreciably less prominent than in father-son relationship; playfulness common in infancy of daughter, but normally yields to a measure of reserve with the development of a strong incest taboo.
  • Elder and younger brother
    relationship of playmates, developing into that of comrades; economic cooperation under leadership of elder; moderate responsibility of elder for instruction and discipline of younger.
  • Elder and younger sister
    relationship parallel to that between elder and younger brother but with more emphasis upon physical care of the younger sister.
  • Brother and sister
    early relationship of playmates, varying with relative age; gradual development of an incest taboo, commonly coupled with some measure of reserve; moderate economic cooperation; partial assumption of parental role, especially by the elder.

All of the above relationships, naturally with local elaborations, are found in any complete family with at least two children of each sex. A typical male in every society, at some time in his life, plays the roles of husband, father, son, and brother in some nuclear family, and a female, those of wife, mother, daughter, and sister. Incest taboos, however, prevent a man from being husband and father in the same family in which he is son and brother, and a woman from being wife and mother in the family where she is daughter and sister. Both, on marrying, become members of a nuclear family other than that into which they were born. Hence, as we have seen, every normal adult individual in any society belongs to two nuclear families, the family of orientation in which he was born and reared and the family of procreation which he establishes by marriage. He is a son or daughter and a brother or sister in the former, a husband or wife and a father or mother in the latter. It is this universal fact of individual membership in two nuclear families that gives rise to kinship systems. If marriages normally took place within the nuclear family, there would be only family organization; kinship would be confined to the limits of the family. But by virtue of the fact that individuals regularly belong to two families, every person forms a link between the members of his family of orientation and those of his family of procreation, and ramifying series of such links bind numbers of individuals to one another through kinship ties.
The term primary relatives is applied to those who belong to the same nuclear family as a particular person-his father, mother, sisters, and brothers in his family of orientation, and his husband or wife, his sons, and his daughters in his family of procreation. Each of these relatives will have his own primary relatives, most of whom will not be included among the primary relatives of Ego. From the point of view of the latter these may be called secondary relatives. Potentially, a person can have 33 distinct kinds of secondary relatives, namely:

  • FaFa (paternal grandfather)
  • FaMo (paternal grandmother)
  • FaBr (paternal uncle)
  • FaSi (paternal aunt)
  • FaWi (stepmother)
  • FaSo (half brother)
  • FaDa (half sister)
  • MoFa (paternel grandmother)
  • MoMo (maternel grandmother)
  • MoBr (maternel uncle)
  • MoSi (maternel aunt)
  • MoHu (stepfather)
  • MoSo (step brother)
  • MoDa (step sister)
  • BrWi (sister-in-law)
  • BrSo (nephew)
  • BrDa (niece)
  • SiHu (brother-in-law)
  • SiSo (nephew)
  • SiDa (niece)
  • WiFa (or HuFa)
  • WiMo (or HuMo)
  • WiBr (or HuBr)
  • WiSi (or HuSi)
  • WiHu (or HuWi, i.e., co-spouse)
  • WiSo (or HuSo)
  • Willa (or HuDa)
  • SoWi (stepson)
  • SoSo (grandson)
  • SoDa (granddaughter)
  • DaHu (son-in-law)
  • DaSo (grandson)
  • DaDa (granddaughter)

Each secondary relative, in turn, has primary relatives who are neither primary nor secondary relatives of Ego, and who may thus be termed tertiary relatives. Among these there are 151 possibilities, including eight great-grandparents, eight first cousins, the spouses of all uncles, aunts, nephews, and nieces, and many others. It would be possible in similar fashion to distinguish quaternary relatives (like first cousins once removed), quinary relatives (like second cousins), etc., but for our purposes it will be sufficient to class all who are more remote than tertiary relatives as distant relatives.
Primary relatives are linked by bonds of blood or biological kinship, with one exception, namely, husband and wife, who, because of incest taboos, are linked only by a marital bond. This gives rise to a fundamental dichotomy in relatives at all levels. Whenever the connection between two relatives, whether primary, secondary, tertiary, or distant, includes one or more marital links, the two have no necessary biological relationship and are classed as affinal relatives. WiMo, DaHu, and MoBrWi are examples. Relatives between whom every connecting link is one of blood or common ancestry, on the other hand, are known as consanguineal relatives.
The science of genetics tells us the exact probabilities of common heredity between consanguineal relatives of each degree. A person can be expected to share, on the average, exactly half of the hereditary factors or genes of any primary consanguineal relative. He inherits half of those possessed by his father and half of those carried by his mother. He transmits approximately half of his own to each son and daughter. With each brother and sister (except identical twins) he is likely to share 50 per cent of the half inherited from the father plus 50 per cent of the half inherited from the mother, or half of the total heredity of the sibling. The common heredity of secondary consanguineal relatives approximates one quarter, and of tertiary relatives one eighth. If a pair of kinsmen can trace consanguineal relationship in two lines, the common heredity is the sum of those in both lines. For example, a man and his FaSiDa, being tertiary relatives, ordinarily have one eighth of their heredity in common, but where cross-cousin marriage prevails and the FaSiDa is at the same time a MoBrDa, the couple will share one quarter of the same biological potentialities. In homogeneous and closely inbred populations, of course, the actual bio logical similarity between related individuals can considerably exceed the proportions indicated above, but even in the most heterogeneous society it cannot be less 4.
Kinship systems constitute one of the universals of human culture. The author is not aware of any society, however primitive or decadent, that does not recognize a system of culturally patterned relationships between kinsmen. Remembrance of kinship ties naturally tends to disappear with time and with remoteness of actual relationship, but social groupings, like those based on common residence or descent, often help to preserve the memory or tradition of certain kinship bonds for surprising periods. Indeed, the author knows of no society which does not reckon kinship well beyond tertiary relatives, at least in some directions. In many small tribes every member acknowledges some specific kinship tie with every other member. In aboriginal Australia, where preoccupation with kinship was carried to unusual lengths, it is said that a native could, at least theoretically, traverse the entire continent, stopping at each tribal boundary to compare notes on relatives, and at the end of his journey know precisely whom in the local group he should address as grandmother, father-in-law, sister, etc., whom he might associate freely with, whom he must avoid, whom he might or might not have sex relations with, and so on.
Even if we ignore for the moment some of the finer distinctions between relatives made by some societies, any individual in any society has potentially seven different kinds of primary relatives, 33 of secondary relatives, 151 of tertiary relatives, and geometrically increasing numbers of distant relatives of various degrees. To associate a distinctive pattern of behavior with each potentially distinguishable category of relationship would be impracticable and intolerably burdensome, and no society attempts to do so. The problem is solved in all societies by reducing the number of culturally distinguished categories to a manageable number through grouping or coalescence. The varying methods by which such coalescence is accomplished give rise to many of the principal differences in kinship structure. Before they are considered, however, it is necessary to introduce the subject of kinship terminology.
Part of the reciprocal behavior characterizing every relationship between kinsmen consists of a verbal element, the terms by which each addresses the other. Although some peoples commonly employ personal names even among relatives, all societies make at least some use of special kinship terms, and the great majority use them predominantly or exclusively in intercourse between relatives. An interesting and fairly common usage intermediate between personal names and kinship terms is called teknonymy 5. In its most typical form it consists in calling a person who has had a child “father (or mother) of So-and-so,” combining the parental term with the child's name, instead of using a personal name or a kinship term.
Kinship terms are technically classified in three different ways—by their mode of use, by their linguistic structure, and by their range of application 6. As regards their use, kinship terms may be employed either in direct address or in indirect reference. A term of address is one used in speaking to a relative; it is part of the linguistic behavior characteristic of the particular interpersonal relationship. A term of reference is one used to designate a relative in speaking about him to a third person; it is thus not part of the relationship itself but a word denoting a person who occupies a particular kinship status. In English, most terms for consanguineal relatives are employed in both ways, though “nephew” and “niece” are seldom used in direct address. Terms for affinal relatives are rarely used in address by English speakers, consanguineal terms or personal names being substituted. Thus a man normally addresses his mother-in-law as “mother,” his stepfather as “father,” and his brother-in-law by the latter's given name or a nickname. The special terms of address in English are mainly diminutive or colloquial, e.g., “grandpa,” “granny,” “auntie,” “dad,” “papa,” “ma,” “mummy,” “hubby,” “sis,” and “sonny.” Some peoples have completely distinct sets of terms for address and reference, others make only grammatical distinctions or none at all, and still others have varying combinations.
Terms of reference are normally more specific in their application than terms of address. Thus, in English, “mother” as a term of reference ordinarily denotes only the actual mother, but as a term of address it is commonly applied also to a stepmother, a mother-in-law, or even an unrelated elderly woman. Moreover, terms of reference are usually more complete than terms of address. It may be customary to use only personal names in addressing certain relatives, or a taboo may prevent all conversation with them, with the result that terms of address for such kinsmen may be completely lacking. Furthermore, terms of address tend to reveal more duplication and overlapping than do terms of reference. For these reasons, terms of reference are much more useful in kinship analysis, and are consequently used exclusively in the present work.
When classified according to linguistic structure, kinship terms are distinguished as elementary, derivative, and descriptive 7. An elementary term is an irreducible word, like English “father” or “nephew,” which cannot be analyzed into component lexical elements with kinship meanings. A derivative term is one which, like English “grandfather,” “sister-in-law,” or “stepson,” is compounded from an elementary term and some other lexical element which does not have primarily a kinship meaning. A descriptive term is one which, like Swedish farbror (father's brother), combines two or more elementary terms to denote a specific relative. In actual use, the qualifying lexical element in derivative terms is quite commonly dropped unless there is need to be precise. Thus, in English, a man is as likely to say “my son” as “my stepson” in referring to his wife's son by a previous marriage, and in many societies a mother's sister is optionally referred to as “mother” or designated by a derivative term translatable as “little mother.” In all languages it is apparently possible to resort to a descriptive term if the reference of any other term is ambiguous. Thus in English, if I mention “my sister-in-law” and am asked to specify which one, I can refer to either “my brother's wife” or “my wife's sister,” or even to “my elder brother's second wife,” etc. Except for such supplementary clarifying use, descriptive terms appear only sporadically in kinship terminologies, except, as our data show 8, in a band extending across central Africa from west to east, where a number of Sudanese, Nilotic, and Bantu tribes use descriptive terminology very freely.
As regards range of application, kinship terms are differentiated as denotative and classificatory. A denotative term is one which applies only to relatives in a single kinship category as defined by generation, sex, and genealogical connection. Sometimes such a term, for a particular speaker, can denote only one person, as in the case of English “father,” “mother,” “husband,” “wife,” “father-in-law,” and “mother-in-law.” 9. Often, however, a denotative term applies to several persons of identical kinship connection, as do the English words “brother,” “sister,” “son,” “daughter,” “son-in-law,” and “daughter-in-law.” A classificatory term is one that applies to persons of one or more kinship categories, as these are defined by generation, sex, and genealogical connection. Thus, in English, “grandfather” includes both the father's father and the mother's father; “aunt” denotes a sister of either parent or a wife of either a maternal or a paternal uncle; “brother-in-law” applies equally to a wife's or a husband's brother or a sister's husband; and “cousin” embraces all collateral relatives of one's own generation, and some of adjacent generations, irrespective of their sex, their line of genealogical connection, or even their degree of remoteness. It is primarily through the liberal use of classificatory terms that all societies reduce the number of kinship categories from the thousands that are theoretically distinguishable to the very modest number, perhaps 25 as an approximate average, 10 which it bas everywhere been found practicable to recognize in actual usage.
From Morgan's day until quite recently it was customary to speak of “classificatory” and “descriptive” kinship systems, the former being regarded as characteristic of primitive tribes and the latter of civilized mankind. These distinctions are now known to be wholly erroneous. The words “classificatory” and “descriptive” refer to particular terms, not to whole systems of terminology. As a matter of fact, with the exception of a few African societies which resort to descriptive terminology on a wholesale scale, every known kinship system is classificatory in the sense of making extensive use of classificatory terms. The systems of western Europe, including our own, employ classificatory terms at least as freely as do those of the average primitive tribe. Indeed, the English system is identical in type to the systems of the Andamanese pygmies, the Ona of Tierra del Fuego, and the Eskimos, and is even technically classified as an “Eskimo system.” 11
The several categories of primary relatives (Fa, Mo, Br, Si, Hu, Wi, So, Da) are denoted by as many different terms in the great majority of societies. These terms are nearly always elementary, though in seven societies in our sample siblings are called by descriptive terms, e.g., “father's daughter” or “mother's daughter” instead of a special term for sister. Terms for primary relatives may be either denotative or classificatory, but if they are the latter they usually include one primary and one or more secondary relatives rather than two categories of primary relatives. Exceptions occur, though they are rare. Occasionally, for example, a term meaning “spouse” is employed for both husband and wife, or a term meaning “child” for both son and daughter, or a sibling term is used by both sexes but denotes “brother” to one and “sister” to the other. In general, however, all primary relatives are terminologically differentiated from each other. In addition, the majority of societies distinguish elder from younger brothers and sisters by separate terms, thus fully reflecting all functional differences in relationships within the nuclear family.
In the designation of secondary, tertiary, and distant relatives, though new elementary terms are applied to distinctive relatives, derivative and descriptive terms appear with increasing frequency.
Denotative terms become rare with secondary relatives and practically disappear with tertiary kinsmen, giving way to classificatory terminology. This results in large measure, of course, from the increasing number of potentially distinguishable categories—33 for secondary and 151 for tertiary relatives—and from the correspondingly increased practical necessity of reducing the recognized number by grouping or coalescence. This can be achieved either by extending a term originally denoting some primary relative to one or more categories of secondary or remoter kinsmen, or by applying a distinctive term to several categories of secondary, tertiary, or distant relatives. Our own kinship system follows the latter method exclusively, reflecting the isolated character of our nuclear family, but in cross-cultural perspective the former method is rather more common.
A classificatory term can arise only by ignoring one or more fundamental distinctions between relatives which, if given full linguistic recognition, would result in designating them by different denotative terms. The pioneer researches of Kroeber and Lowie 12 have led to the recognition of six major criteria which, when linguistically recognized as a basis of terminological differentiation, yield denotative terms but the ignoring of any one of which produces classificatory terms. These criteria are generation, sex, affinity, collaterality, bifurcation, and polarity. They are the criteria employed above in calculating the number of potential categories of primary, secondary, and tertiary relatives. In addition, the same authors have isolated three subsidiary criteria—relative age, speaker's y sex, and decedence—the linguistic recognition of which makes a classificatory term less inclusive or a denotative term more specific. These nine criteria have an empirical as well as a logical basis; severally and in combination they appear to include all the principles actually employed by human societies in the linguistic classification and differentiation of kinsmen. Each will now be considered individually.
The criterion of generation rests on a biological foundation. The facts of reproduction automatically align people in different generations: Ego's own generation, which includes brothers, sisters, and cousins; a first ascending generation, which embraces parents and their siblings and cousins; a first descending generation, which includes sons, daughters, nephews, and nieces; a second ascending or grandparental generation; a second descending or grandchildren's generation; and so on. Since marriages in most societies normally occur between persons of the same generation, affinal relatives tend to be aligned by generation in the same manner as consanguineal relatives. Most kinship systems give extensive recognition to generation differences. Our own, for example, ignores them in only a single unimportant instance, namely, when the term “cousin” is applied to a “cousin once (or twice) removed,” i.e., one or two generations above or below Ego. Perhaps the most striking examples of classificatory terms resulting from the ignoring of generation differences are found in the so-called “Crow” and “Omaha” types of crosscousin terminology. In the former, the children of the father's sister are called by the same terms as paternal uncle and aunt, while the mother's brothers' children are classed with fraternal nephews and nieces. In an Omaha system, the situation is roughly reversed; the father's sisters' children are classed with sororal nephews and nieces, and the mother's brothers' children with maternal uncles and aunts 13.
The criterion of sex derives from another biological difference, that between males and females, and is also widely taken into account in kinship terminology. Our own system, for example, ignores sex in respect to only one basic term, namely, “cousin.” Some societies employ a single classificatory term for both a son and a daughter, or for a parent-in-law of either sex. The commonest instances of the ignoring of sex in kinship terminology are found, however, in the second descending and second ascending generations, where many societies have terms approximately equivalent to “grandchild” or “grandparent.” It is, of course, in precisely these generations that an individual is most likely to find relatives who are mainly too young to be significantly differentiated sexually or too old to be sexually active.
The criterion of affinity arises from the universal social phenomena of marriage and incest taboos. In consequence of the latter, marital partners cannot normally be close consanguineal relatives. Among relatives of like degree, therefore, whether they be primary, secondary, tertiary, or distant, there will always be one group of consanguineal kinsmen, all equally related biologically to Ego, and a second group of affinal relatives whose connection to Ego is traced through at least one marital link and who are biologically unrelated or only remotely related to him. This difference is widely recognized in kinship terminology. In our own system, for example, it is completely ignored only in the term “uncle,” which includes the husbands of aunts as well as the brothers of parents, and in the word “aunt,” which similarly includes the wives of uncles as well as the sisters of parents. Elsewhere we recognize affinity only partially, through the use of derivative terms with the prefix “step-” or the suffix “-in-law,” in which respect we differ from most societies, who ordinarily apply elementary terms to affinal relatives. Classificatory terms resulting from the ignoring of this criterion are particularly common in societies with preferential rules of marriage. For example, under a rule of preferential cross-cousin marriage with the FaSiDa, the latter may be called by the same term as wife, and a single term may suffice for FaSi and WiMo.
The criterion of collaterality rests on the biological fact that among consanguineal relatives of the same generation and sex, some will be more closely akin to Ego than others. A direct ancestor, for example, will be more nearly related than his sibling or cousin, and a lineal descendant than the descendant of a sibling or cousin. Our own kinship system consistently recognizes the criterion of collaterality and, with the sole exception of “cousin,” never employs the same term for consanguineal kinsmen related to Ego in different degrees. The majority of societies, however, ignore collaterality with greater frequency, and in this way arrive at various classificatory terms. The phenomenon of grouping lineal and collateral kinsmen, or relatives of different degrees, under a single classificatory term is technically known as merging 14. Among the relatives most commonly merged are a parent and his sibling of the same sex, a sibling and a parallel cousin (child of a FaBr or MoSi), a wife and her sister, and a son or daughter and a nephew or niece.
The criterion of bifurcation (forking) applies only to secondary and more remote relatives, and rests on the biological fact that they may be linked to Ego through either a male or a female connecting relative. Recognition of this criterion involves applying one term to a kinsman if the relative linking him to Ego is male and quite another term if the connecting relative is female. Our own kinship system ignores the criterion of bifurcation throughout, and derives many of its classificatory terms from this fact. Thus we call a person “grandfather” or “grandmother” irrespective of whether he is the father's or the mother's parent; “uncle” or “aunt” regardless of the sex of the parent through whom the relationship is traced; “brotherin-law” or “sister-in-law,” “nephew” or “niece,” “grandson” or “granddaughter,” without considering the sex of the connecting relative. The majority of societies, however, make terminological distinctions between some or most of these relatives.
The criterion of polarity 15, the last of the six major criteria for differentiating kinship terminology, arises from the sociological fact that it requires two persons to constitute a social relationship. Linguistic recognition of this criterion produces two terms for each kin relationship, one by which each participant can denote the other. When polarity is ignored, the relationship is treated as a unit and both participants apply the same classificatory term to each other. In our own kinship system polarity is recognized throughout, with the sole exception of the term “cousin.” The fact that two brothers, two sisters, two brothers-in-law, or two sisters-in-law also apply the same term to one another is really an incidental result of the recognition of other criteria, as becomes clear when we observe that the same term can be used for the same relative by a relative of opposite sex, in which case the reciprocal term is different. Polarity is occasionally ignored in sibling relationships, as where a brother calls his sister by the same term that she uses for him, and in avuncular relationships, as where a maternal uncle and his sisters' children refer to one another by the same term. It is most commonly ignored, however, in the terms used by relatives two generations removed; grandparents and grandchildren in many societies refer to one another by identical terms.
The criterion of relative age reflects the biological fact that relatives of the same generation are rarely identical in age. Of any pair, one must almost inevitably be older than the other. While ignored completely in our own kinship system, and not treated as one of the six basic criteria in our theoretical analysis, relative age is widely taken into account in kinship terminologies. A significant majority of all systems differentiate terminologically between elder and younger siblings of the same sex, and 100 out of 245 in our sample do likewise for siblings of opposite sex. Some societies, e.g., the Yuman tribes of the American Southwest 16, make extensive age distinctions in terminology, differentiating, for example, the elder and younger siblings of a parent and the spouses and children of an elder and a younger sibling.
The criterion of speakers sex rests on the biological fact that the user of a kinship term, as well as the relative denoted by it, is necessarily either a male or a female. Kinship systems which recognize this criterion will have two terms for the same relative, one used by a male speaker and the other by a female. Among the Haida 17, for example, there are two denotative terms for father, one employed by sons and one by daughters. To the Haida, the English word “father” would appear classificatory. The criterion of speaker's sex often operates in conjunction with the criterion of sex, with the result that the sameness or oppositeness of sex of the speaker and relative may appear more important than the actual sex of either. This is especially common in sibling terminology, where one term (or a pair distinguishing relative age) may be used by a man for his brother and by a woman for her sister, while a different term is employed by a man for his sister and by a woman for her brother.
The criterion of decedence, the last and least important of the nine, is based on the biological fact of death. Like the criterion of bifurcation, it applies particularly to secondary relatives and depends upon the person through whom kinship is traced. But whereas the crucial fact in bifurcation is whether the connecting relative is male or female, in the criterion of decedence it is whether that relative is dead or alive. A very few societies, especially in California and adjacent areas 18, have two kinship terms for certain relatives, one used during the lifetime of the connecting relative, the other after his death. The distinction occurs almost exclusively in terms for relatives who are potential spouses under preferential levirate or sororate marriage. With the death of an elder brother, for example, the status of his wife relative to his younger brother may undergo a sharp change; under the levirate she is now destined to marry the younger brother, whereas previously she had not been available to him as a wife. While not itself of great consequence, decedence completes the roster of criteria which, through linguistic recognition or non-recognition, yield most if not all of the known variations in kinship nomenclature.
Though of fundamental importance for analysis, the foregoing criteria do not of themselves explain differences in kinship terminology. The crucial scientific problem is that of discovering the factors which have led different peoples to select or reject particular V criteria as a basis for differentiating kinsmen of some categories, and equating others, in arriving at a practicable number of culturally recognized categories out of the hundreds or thousands of potentially distinguishable ones. Before a solution to this problem is sought, consideration must be given to the question of the inter-relations between kinship terminology and kinship behavior.
As has already been indicated, terms of address form an integral part of the culturally patterned relationships between kinsmen, even though they are an aspect of habitual verbal rather than gross muscular behavior. Terms of reference, on the other hand, are linguistic symbols denoting one of the two statuses involved in each such relationship (or both statuses where their polarity is ignored). However, since any status is defined in terms of the culturally expected behavior in the relationship in which it is embedded, there are a priori reasons for assuming a close functional congruity between terms of reference and the relationships in which the denoted kinsmen interact. The data analyzed for the present study provide abundant empirical support for this assumption, and most students of kinship have arrived at the same conclusion. Radcliffe-Brown 19 sums up existing knowledge on the subject very adequately in the statement:

“we can expect to find, in the majority of human societies, a fairly close correlation between the terminological classification of kindred or relatives and the social classification. The former is revealed in kinship terminology, the latter . . . specifically in the attitudes and behavior of relatives to one another.”

Tax 20 states that as a general rule, subject to some exceptions, “Persons toward whom ego behaves in the same manner he will call by the same term; … persons to whom ego behaves in a different manner he will call by different terms.”
The congruity between kinship terms and behavior patterns, though firmly established as an empirical generalization, is nevertheless not absolute. Thus as Opler 21 points out, among the Apache tribes of the Southwest differences in behavior between kinsmen sometimes exist without any corresponding differentiation in nomenclature, and terminological distinctions occur in the absence of important differences in social function. Such exceptions are specifically recognized by both Radcliffe-Brown and Tax, and they account for the more cautious statement by Lowie 22 of the same conclusion:

“Where relatives whom other people distinguish are grouped together there is some likelihood that the natives regard them as representing the same relationship because they actually enjoy the same privileges or exercise the same functions in tribal life. When relatives whom other peoples group together are distinguished, there is some probability that the distinction goes hand in hand with a difference in social function.”

The fact that a people applies a single classificatory term to a variety of different relatives embraced in one conventional kinship category does not imply, of course, that even the standardized behavior exhibited toward all of them is identical. In a society, for example, which extends the kinship term for mother to the father, other polygynous wives, to maternal aunts, and to the wives of the father's brothers, people do not confuse these women with one another. Everyone knows his own mother, reacts toward her in an especially intimate manner, and when necessary can readily distinguish his “classificatory mothers” from her by the use of descriptive terms. Toward these other “mothers” he acts in a generically similar manner, e.g., with respect, helpfulness, and tokens of fondness, but with an attenuation in warmth and responsiveness proportionate to their social remoteness 23. This point is so obvious that it would require no notice if it had not been ignored by some earlier anthropologists and overemphasized by some later ones 24.
The use of a single term for several categories of relatives usually means that the behavior toward the functionally less significant ones (the stepmothers and aunts in the above instance), though not identical with that toward the closest or most important relative (the mother), is in general more like the latter than like that exhibited toward other comparable relatives (e.g., a grandmother, paternal aunt, or mother-in-law) who are denoted by different terms. This follows a normal process of linguistic classification. A strictly parallel illustration may be cited from the field of economic relationships. The word “tenant” is applied to a person who holds tenure in landed property either for life, or from year to year, or at will, or at sufferance. In each case his relationship to his landlord is different, but all have more in common with one another than with other subordinate economic statuses, such as those of a bailee, a debtor, an employee, or a servant. This relative generic similarity suffices to account for the common term.
Although there are substantial grounds for assuming an essential congruity between kinship terms and the culturally patterned behavior toward the relatives they denote, this by no means implies either:

  1. that the behavior patterns in particular societies are as sharply differentiated from one another as the associated terms, or
  2. that the associated behavior patterns in different societies show an approximately equal degree of differentiation.

With the exception of derivative and descriptive terms, which constitute a distinct minority, all kinship terms are independent words, and as such are completely and thus equally differentiated from one another. Patterns of kinship behavior, on the other hand, run the gamut between practical identity and extreme dissimilarity, with countless intermediate gradations. The application of completely differentiated terms to incompletely and variably differentiated phenomena results inevitably in a lack of strict comparability.
In different societies the conditions leading to the differentiation of behavior patterns in kin relationships may vary widely. In societies with complex forms of unilinear social organization, for example, the presence of such groupings as moieties, sibs, clans, sections, and extended families may, when added to the universal differentials of behavior within the nuclear family and between persons of different sex and age, readily complicate social interaction to such a degree that every linguistically recognized kinship category may be associated with a highly distinctive set of culturally patterned responses. On the other hand, bilateral societies with simple or amorphous forms of social organization, though known empirically to recognize on the average an approximately equal number of elementary kinship terms, may have a very much smaller number of distinctive behavior patterns to differentiate. Even with special terms assigned to each primary relative, and with terminological distinctions made between other relatives on the basis of such universal differences as those between the sexes, between different generations, and between affinal and consanguineal kinsmen, there may still be leeway for other terminological distinctions before the maximum practicable number of terms is reached. In this twilight zone, so to speak, distinctions in terminology may arise or be perpetuated which have little relation to patterned kinship behavior.
If this interpretation is correct, distinctions in kinship terminology should appear in association with relatively insignificant functional differences between kinsmen more often in bilateral than in unilinear societies. This can be statistically tested. From the point of view of a male Ego, the relative age of a non-marriageable female relative is inherently of little functional significance-much less, for example, than the differences between primary and other relatives, between persons of different generations, between males and females, between consanguineal and affinal relatives, or between marriageable and non-marriageable females. Age-differentiating terms used by a brother for his sisters constitute an example. The extent to which they are associated with bilateral descent, taken as a rough indication of relative simplicity of social organization, is shown in Table 13. The results show an appreciable tendency toward the anticipated association, which can be expressed statistically by a coefficient of association of + .33. A parallel test of male-speaking terms for paternal aunts yields a coefficient of association of + .31 for the adhesion between bilateral descent and the occurrence of separate terms for FaElSi and FaYoSi.

Table 13
Male-speaking Kinship Terms for Elder and Younger Sister Bilateral Descent Unilinear Descent
Separate terms 41 59
The same term 39 106

While societies with relatively few and simple forms of social structure tend to show less differentiation in the kinship behavior associated with different kinship terms, it is dangerous to assume the absence of such differentiation. It may merely be difficult to discover, eluding all but the most meticulous field investigation. The author can cite an illuminating instance from his own field experience.
When he first went into the fleld in 1932, among the Haida of the Northwest Coast, the author was much impressed with the rich material on kinship behavior that had been gathered in many places by students of Radcliffe-Brown and Malinowski, and he was inclined to be somewhat critical of his American colleagues for the comparative paucity of like material reported in their monographs. This attitude was confirmed when he returned from the Haida with a wealth of data on patterned kinship behavior 25 that even exceeded his own anticipations. Some of his colleagues nevertheless continued to assure him that they had not overlooked the possibility of unearthing such material in their own field work, but that in certain areas, particularly among the bilateral tribes of the Plateau, Great Basin, and California, the most exhaustive research had failed to reveal any significant patterning of kinship behavior beyond the most obvious family relationships. The author kept his peace, but resolved to take advantage of the first opportunity to put the issue to the test.
The opportunity came in 1934, when a month was spent among the Tenino, a typical Plateau tribe in central Oregon. Persistent inquiries about kinship revealed the inevitable differentiation of relationships within the nuclear family and a few behavioral norms obviously correlated with age and sex, but beyond this no significant patterning of behavior. The evidence concerning such matters as mother-in-law avoidance and the avunculate was completely negative, and the field notes bulged with specific denials of relationships characterized by respect, joking, special privileges, and the like. The negative evidence seemed convincing, and on his return the author, both in conversation with his university colleagues and in classroom lectures, retreated from his previous position and admitted the likelihood of the absence of all but an irreducible minimum of patterned kinship behavior in simple sibless tribes.
A second field trip to the Tenino in the summer of 1935 brought out new material. Working not directly with kinship but with other aspects of culture, such as property, house-building, sex, and childhood training, the author found bits of kinship material unexpectedly coming to light in a variety of other connections. When this material was analyzed, it was found to reveal a surprising quantity of kinship patterning, e.g., an obligation on the part of the paternal grandfather to instil hardihood in his grandson by the imposition of physical ordeals, a fairly typical joking relationship between a father's sister's husband and his wife's brother's child, a patterned privilege of property appropriation between brothers-in-law, and permitted sexual license between siblings-in-law of opposite sex. With such material in hand the author had no alternative save to retract his retraction.
That kinship nomenclature is closely correlated with culturally patterned norms of behavior toward relatives must be assumed. This assumption accords with a priori reasoning, with the overwhelming testimony of the data surveyed for the present study, and with the experience and the declared or admitted views of nearly all competent anthropological authorities. Further exploration of the subject would become primarily an exercise in semantics, a study of the relation between words and the things they denote. Moreover, it would be irrelevant, for the real scientific problem is not to derive terminology from patterned behavior, or vice versa, but to explain both phenomena on the basis of causal factors lying outside of the kinship complex. The next chapter will seek such an explanation for variations in kinship terminology, and Chapters 9 and 10 will analyze the causes of those aspects of kinship behavior which relate to the regulation of sex.
In both cases the determinants must be independent variables, i.e., causal factors arising outside the realm of kinship phenomena. Such factors can be expected to exert an influence on both behavior patterns and nomenclature. In some cases they may affect both at the same time and in like degree. In others they may change initially only the patterns of kinship behavior, setting in motion an adaptive process which with the passage of time produces congruent modifications in terminology. Sometimes, perhaps, they may even alter first the kinship terms, with behavior undergoing subsequent adjustment, but this is probably relatively rare since new words and new meanings of old words do not ordinarily precede the things they designate. In any event, the ultimate effect of an outside causative factor is to alter both relationships and terminology, which always maintain their essential integration.

1. L.H. Morgan, “Systems of Consanguinity and Affinity of the Human Family,” Smithsonian Contributions to Knowledge, XVII (1870), 1-590.
2. Cf. B. Malinowski, “Kinship,” Man, XXX ( 1930), 23-5.
3. Cf. E. E. Evans-Pritchard, “The Nature of Kinship Extensions,” Man, XXXII (1982), 13.
4. The geneticist will realize, of course, that this discussion deals with probabilities and ignores chance permutations of genes as well as certain sexchromosome phenomena, mutations, identical twinning, etc.
5. See E. B. Tylor, “On a Method of Investigating the Development of Institutions,” Journal of the Royal Anthropological Institute, XVIII ( 1889), 248.
6. Cf. R. H. Lowie, “A Note on Relationship Terminologies,” American Anthropologist, n.s., XXX ( 1928), 264.
7. Cf. R. H. Lowie, “Kinship,” Encyclopaedia of the Social Sciences, VIII ( 1932), 568; K. Davis and W. L. Warner, “Structural Analysis of Kinship,” American Anthropologist, n.s., XXXIX (1937), 303.
8. Cf. also R.H. Lowie, Culture and Ethnology (New York, 1917), pp. 105-7.
9. These have been called “isolating terms.” See K. Davis and W. L. Warner, “Structural Analysis of Kinship,” American Anthropologist, n.s., XXXIX (1937), 300-1.
10. Cf. A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX ( 1909), 79.
11. Cf. L. Spier, “The Distribution of Kinship Systems in North America,” University of Washington Publications in Anthropology, I (1925), 79.
12. See, in particular, A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 77-85; R. H. Lowifc, “Relationship Terms,” Encyclopaedia Britannica ( 14th edit., London, 1929), XIX, 84-9. Cf. also K. Davis and W. L. Warner, “Structural Analysis of Kinship,” American Anthropologist, n.s., XXXIX ( 1937), 291-313.
13. In consequence of merging (see below) it often happens that the term for FaSiSo is “father,” that for MoBrSo is “son,” and that for MoBrDa is “daughter” in a Crow system, and that the term for MoBrDa is “mother” in an Omaha system. It has often been overlooked that this is only an incidental and not a fundamental feature of Crow and Omaha terminology.
14. See R.H. Lowie, Culture and Ethnology (New York, 1917), p. 109.
15. In the literature this criterion is commonly called “reciprocity.” See, for example, A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX ( 1909), 80-1; R. H. Lowie, Culture and Ethnology (New York, 1917), pp. 165-6. Since the names for all the other criteria suggest a basis of differentiation, whereas “reciprocity” suggests equivalence, the name “polarity” is adopted for the sake of consistency. Moreover, “reciprocity” has another recognized technical meaning in anthropology; see B. Malinowski, Crime and Custom in Savage Society (London, 1926), pp. 24-7
16. See L. Spier, “The Distribution of Kinship Systems in North America,” University of Washington Publications in Anthropology, I (1925), 75-6.
17. See G. P. Murdock, “Kinship and Social Behavior among the Haida,” American Anthropologist, n.s., XXXVI ( 1934), 360-2.
18. See A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 79. The term “decedence” is proposed by the present author.
19. A.R. Radcliffe-Brown, “Kinship Terminologies in California,” American Anthropologist, n.s., XXXVII ( 1935), 531. Cf. also W. H. R. Rivers, Kinship and Social Organisation (London, 1914), pp. 11-12.
20. S. Tax, “Some Problems of Social Organization,” Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), pp. 20-1.
21. M.E. Opler, “Apache Data concerning the Relation of Kinship Terminology to Social Classification,” American Antliropologist, n.s., XXXIX (1937), 202-5.
22. R.H. Lowie, Culture and Ethnology (New York, 1917), p. 100. 23. Cf. B.Z. Seligman, “Incest and Descent,” Journal of the Royal Anthropological Institute, LIX (1929), 271.
24. See, for example, B. Malinowski, “Kinship,” Man, XXX (1930), 29.
25. See G.P. Murdock, “Kinship and Social Behavior among the Haida,” American Anthropologist, n.s., XXXVI (1934), 355-85.

Determinants of Kinship Terminology

Analysis of the theoretical literature reveals that six groups of external factors have been proposed as determinants of kinship terminology. These are:

  1. multiple historical influences
  2. morphological differences in language
  3. elementary psychological processes
  4. universal sociological principles
  5. customs of preferential marriage
  6. the constitution of kin and local groups

The theories of particular authors, of course, frequently take several factors into account. A discussion of each of the six types of theories will serve to highlight the views of the present writer and to relate them to those of his predecessors.
The attribution of kinship terminology to multiple historical influences has Kroeber as its leading exponent. The kinship publications of this author, in addition to valuable descriptive and invaluable analytical contributions, include a series of controversial papers. He began with sharp attacks on the use of kinship terms for evolutionist reconstruction 1, and continued with criticism of later functional and scientific interpretations 2. Yet he has never denied that social institutions can exert an influence on social structure 3, he has himself proposed at least one significant correlation between kinship terminology and social structure 4, and he concludes his latest controversial article 5 with the suggestion that all students of kinship can “meet on the common ground that the determinants are multiple and variable.” The present author finds no difficulty in agreeing with Kroeber as to the multiplicity of determining influences. Indeed, he finds himself in substantial agreement with practically all of Kroeber's theoretical statements, save for a few obiter dicta.
An example of such obiter dicta occurs in the following statement by Kroeber 6 of his historical position:

“Kin-term systems … are subject to modification from within and without. There is always a sufficient number of such ‘accidents’ to disguise the basic patterns more or less… the essential features of the pattern are … likely to be the ones which have the greatest historical depth. The search for them therefore implies a willingness and ability to view data historically. Without such willingness, it is as good as impossible to separate the significant from the trivial … and the work becomes merely sociological, an affair of schemes.”

The wholly acceptable statement to the effect that kinship terminology can change only in consequence of historical events, including internal modifications and borrowings from without, is followed by incidental assumptions leading to the unwarranted characterization of all except historical interpretations as “schemes.” Lowie 7, after making his bow to historical interpretations with the statement that “features of kinship terminology are distributed like other ethnographical phenomena and must be approached in the same spirit,” exposes the fallacy in Kroeber's remarks when he points out that sociological and historical interpretations are reconcilable when similar causal factors are operative in historically unrelated areas.
The crucial criterion as to the relative appropriateness of comparative or “sociological,” as contrasted with purely historical, methods in anthropology is the limitation of possibilities 8. Where there are no practical limits to the variety of responses which people can make in particular situations, cultural forms can vary endlessly with little comparability between those of unrelated societies, with the result that satisfactory interpretation must depend very heavily upon historical investigation of local and regional influences. Examples include language, since the possibilities of phonetic and morphological variation are immense; ceremonial, since there are no limits to the variety of rites which can be devised and assembled; folktales, since themes and subjects are endless; and much of art, technology, and other aspects of culture. In all such cases, the overwhelming majority of cross-cultural similarities must necessarily be attributed to diffusion. Moreover, since such phenomena are in large measure unique and regionally distributed, problems arising from their influence on other aspects of culture must be attacked locally and by historical methods. The situation is quite different where there are practical limitations to the variety of responses which people can make. Under such conditions cultural similarities will appear in many different places, irrespective of historical contacts, and the influences they exert on other aspects of culture can be treated as comparable. As a result, comparative analysis becomes a highly useful supplement to historical research, and may lead to valid generalizations which are not only valuable in themselves but often capable of being applied with profit by historians.
Does the criterion of limited possibilities apply to kinship terms and their alleged social determinants? The answer to this question is an unambiguous affirmative. While kinship terms themselves show unlimited variability, the methods of classifying them do not. With regard to each of the nine criteria of classification, for example, there are only two alternatives; a people can choose only to recognize or to ignore generation, sex, affinity, etc., in assigning a kinship term to a particular relative. There are also limited numbers of possible or practicable marriage forms (monogamy, polyandry, polygyny), preferential marriages (levirate, sororate, cross-cousin marriage, and a few others), rules of descent (bilateral, matrilineal, patrilineal), rules of residence patrilocal, matrilocal, avunculocal, (neolocal, bilocal), family forms (nuclear, polygamous, extended), unilinear kin groups (moieties, phratries, sibs, lineages), and so on. All of these alternatives are widely distributed in historically independent regions, and insofar as they interact with one another or with other aspects of culture it is reasonable to assume that they exert a parallel influence and to seek the character and degree of such influence by comparative studies.
The data in Chapter 8 will demonstrate the extraordinary extent of parallelism, both in kinship terminology and in types of kin and local groups. There are few forms which are not represented in all five of the world's main continental and insular areas. Even more striking, if possible, are the wide divergences in the forms found within each linguistic stock of which our sample includes several tribal representatives. Since language relationships are the most incontrovertible of all evidences of historical contact, we arrive at the curious paradox that kinship terminology and the forms of social organization often differ precisely where historical connections are indisputable and show resemblances where historical relationships are inconceivable. The distributions, in short, are quite unlike those which anthropologists regularly encounter with respect to such phenomena as languages, folktales, cultivated plants, or types of traps. They not only suggest but demand the intervention of determinants that are not historically limited. In seeking these, of course, the author by no means denies that in each individual case the phenomenon has developed through a specific historical process 9.
When historical causes are exclusively unique or local, the weighing of their respective influence depends wholly upon the judgment of the historian; there is no independent basis of comparison. Historians notoriously differ widely in their judgments, as witness the varied theories of the causation of the American Revolution or the Civil War. When, however, some of the causes of an historical phenomenon are also known to have operated in other situations, comparison of these situations provides some independent basis for estimating the influence to be ascribed to them in particular cases. The wider the comparisons, and the larger the proportion of the causative factors that are susceptible to comparative analysis, the more reliable will be the conclusions as to the weights to be assigned to particular causes. Under favorable circumstances, therefore, the inductions from comparative studies become appreciably more dependable than inferences from historical analysis.
Kinship data, in the opinion of the author, provide a nearly ideal case, for not some but many or most of their determinants, including those of the greatest apparent significance, are found in historically diverse societies, and the number of independent cases is sufficient to give statistical reliability to the inductions drawn from their comparison.
A second group of theories attributes variations in kinship terminology to morphological differences in language. According to Gifford 10, for example, “kinship systems are first of all linguistic phenomena … and only secondarily social phenomena. As such they … constitute an archaic and highly refractory nucleus, which yields unevenly and only here and there to influences from … social structure.” Kinship terms as words must of course conform to the morphological principles of the particular language 11, but the way in which relatives are terminologically classified bears no necessary relation to the nature of the language. As Tax 12 points out, methods of classification often differ considerably, not only in tribes of the same linguistic stock but even in those which are so closely related that the actual terms are practically identical. Moreover, as Lowie 13 rightly states, “language represents reality and … in so far as it is related to social phenomena it is likely to mirror them.”
Kirchhoff 14 attributes the bifurcate merging and generation types of kinship terminology to differential principles of word formation. Since all features of linguistic morphology are regionally distributed in consequence of historical contacts and migrations, linguistic interpretations of kinship terminology constitute only a special class of historical interpretations and are answerable in the same way. The only significant example of the influence of language on nomenclature which has come to light during the present study is the previously mentioned tendency to use descriptive terms very freely among tribes within a band across central Africa. This phenomenon undoubtedly has an historical origin, and is very possibly due to some morphological feature common to the languages of the area.
A discussion of linguistic factors seems a logical place into which to inject the observation that kinship terms frequently exemplify the phenomenon of “cultural lag.” 15 “One factor that must always be considered,” as Lowie 16 correctly states, “is the time element. A recently acquired custom may not yet have developed an appropriate nomenclature, while … the nomenclature may survive after the custom has become obsolete.” A large proportion of the discrepancies between the kinship terms reported in our sources and the forms theoretically anticipated under the prevailing social conditions are precisely those adapted to conditions which actual or inferential historical evidence indicates existed somewhat prior to the period of observation. It should be emphasized, however, that such “survivals” appear far more frequently in association with functionally insignificant relationships than with functionally important ones. Thus, though they indicate a tendency for kinship terminology to adjust somewhat tardily to changes in social determinants, they do not provide a basis for reconstructing the forms of social institutions in the distant past, as Morgan and other early theorists attempted to do 17. Authorities ranging from Kroeber 18 to Radcliffe-Brown 19 now agree that long-range historical inferences cannot legitimately be drawn from kinship terminologies. It will be shown in Appendix A, however, that short-range historical reconstruction is possible with the aid of survivals in kinship terminology which have not been integrated with the rest of the social structure.
A third group of theories attributes kinship terminology to certain elementary psychological or logical processes. “Terms of relationship,” maintains Kroeber 20, “reflect psychology, not sociology,” and subsequently explains that by psychological factors he means those that are “directly expressive of a manner of thought.” These seem to include the underlying logic in linguistic morphology. RadcliffeBrown and his students make extensive use of logical assumptions. For example, the “rules” of “uniform descent,” “uniform siblings,” “uniform mates,” and “uniform ascent,” as defined by Tax 21 —meaning, in combination, that if a kinsman whom Ego denotes by term A has any primary relative whom Ego denotes by term B, then Ego will also tend to apply the term B to a comparable relative of any other kinsman whom he also calls A—really rest on certain assumptions about the nature of the fundamental psychological processes underlying reasoning and association. As a matter of fact, they are in substantial accord with the important psychological mechanism known as “generalization.” 22
Properly used, elementary psychological processes can be of great assistance in the interpretation of kinship phenomena, and the present author will draw freely upon them. One fundamental caution, however, must always be observed. Psychology can never be used alone to explain any phenomenon of culture. It merely provides a mechanism by which historical and other influences are translated into patterns of behavior under particular social conditions. Without a knowledge of the conditions of behavior, which cultural anthropology alone can furnish, no comprehension of psychological principles can yield an explanation of cultural forms. But when the conditions and circumstances affecting a change in culture are known, psychology can greatly assist us in comprehending the kind and range of cultural modifications that are likely to ensue. The above-mentioned “rules” propounded by Tax, for example, provide a substantial basis for predicting the future terminological classification of the primary relatives of a person to whom the kinship term A has recently been extended (see Theorem 1 below ).
Incidentally, they also furnish the first satisfactory explanation of a phenomenon which has long puzzled theorists, namely, why in a society which applies sibling terms to all members of Ego's sib and generation, the parallel cousins who are not sib members, as well as those who are, nearly always are likewise called by sibling terms 23.
The derivation of kinship terminology from assumed universal sociological principles finds its principal exponent in RadcliffeBrown. For example, in discussing the hypothesis advanced by Sapir 24 that bifurcate merging terminology may result from the operation of the levirate, Radcliffe-Brown 25, though accepting the asserted correlation between the two phenomena, refuses to regard one as the effect and the other as the cause but maintains that both “are the result of a single sociological principle, … the social equivalence of brothers.” According to his student, Tax 26, Radcliffe-Brown “believes that the necessity for social integration is the fundamental cause of all social institutions—that they have the function of keeping the society integrated.”
With hypotheses of the above type the present writer has neither sympathy nor patience. In the first place, the alleged principles are mere verbalizations reified into causal forces. In the second, such concepts as “equivalence of brothers” and “necessity for social integration” contain no statements of the relationships between phenomena under varying conditions, and thus lie at the opposite extreme from genuine scientific laws. Thirdly, being unitary in their nature, they provide no basis for interpreting cultural differences; they should produce the same effects everywhere. To be sure, numerous “principles” are alleged, and it is tacitly assumed that when the anticipated effects of one do not occur some other countervailing principle is operative, but nowhere are the conditions set forth under which particular principles give way to others 27. Fortunately, Radcliffe-Brown's claim to eminence as a student of kinship rests on other and more substantial grounds. The term “sociological” as applied to determinants of kinship ordinarily refers, not to the invoking of universal principles, but to the attribution of classificatory terminology to the influence of particular social institutions 28. These fall into two groups—customs of preferential marriage and the constitution of kin and local groups such as family, clan, sib, and moiety. Rivers 29 expresses one extreme of opinion on their influence when he asserts that “the details which distinguish different forms of the classificatory system from one another have been directly determined by the social institutions of those who use the systems.” Sapir 30 expresses the opposite extreme in alleging “that the factors governing kinship nomenclature are very complex and only in part capable of explanation on purely sociological grounds.” No authority on the subject denies any influence to sociological factors. The position of the present author lies between those expressed by Rivers and Sapir—closer to the latter in admitting that kinship determinants are complex and multiple, to the former in assigning great weight to sociological determinants.
Rules of marriage may affect kinship terminology when they produce a situation in which Ego can trace a kinship connection to some relative in two different ways. When this happens, the criterion which might otherwise be recognized in distinguishing the two types of bonds tends to be ignored, with the result that relatives of both types tend to be called by the same term. Thus Lowie 31 maintains that wherever, as in native Australia, the preferred mate is always a consanguineal relative, special affinal terms tend to be absent. The criterion of affinity is ignored because the spouse and his or her relatives are also consanguineal kinsmen and can continue to be designated by consanguineal terms. A regular rule of marriage by sister exchange, i.e., when two men obtain wives by exchanging their sisters, leads, says Rivers 32, to the ignoring of affinity and to the terminological equating of several pairs of relatives, namely, FaSiHu and MoBr, MoBrWi and FaSi, SoWiBr and DaHu, DaHuSi and SoWi. A rule of preferential marriage with a cross-cousin (a child of either FaSi or MoBr) will allegedly lead, in similar fashion, to the use of the same terms for WiFa, MoBr, and FaSiHu, for WiMo, FaSi, and MoBrWi, and for cross-cousins, spouses, and siblings-in-law 33.
Sapir 34 has suggested and others 35 have agreed that preferential levirate and sororate marriages (and the same holds true for fraternal polyandry and sororal polygyny respectively) may operate to minimize the criterion of collaterality and thus produce kinship terminology of the so-called “bifurcate merging” 36 type. The argument runs thus: if Ego's mother normally marries the father's brother when the father dies, there will be a tendency to employ the same terms for Fa and FaBr, for Br and FaBrSo, for Si and FaBrDa, for So and BrSo, and for Da and BrDa since to Ego such persons are likely to play similar family or kinship roles. In the same way the sororate would equate MoSi with Mo, MoSiSo with Br, MoSiDa with Si, and a woman's sisters' children with her own.
Preferential secondary marriages with a kinsman of an older or younger generation are similarly alleged to result in ignoring the criterion of generation and thus in classifying relatives of different generations under a single term. Thus Aginsky 37, Gifford 38, Lesser 39, Lowie 40, and Rivers 41 have accepted a suggestion originally made by Kohler 42 that rules of preferential marriage with a wife's brother's daughter or a mother's brother's widow will produce, or help to produce, respectively, the Omaha and Crow types of terminology for cross-cousins.
Customs of preferential marriage, in the opinion of the present writer, are likely to influence kinship terminology when they apply to all or most marriages within a society, but not when they apply only to occasional unions or to a distinct minority of all that take place. For this reason he professes skepticism concerning the suggested determination of kinship terminology by secondary marriages. Every marriage with a BrWi or MoBrWi is necessarily a second marriage for the woman; every one with a WiSi or WiBrDa, a second (or subsequent) marriage for the man. By the very nature of things such marriages can constitute only a minority of all unions. There must always be a first marriage before there can be any secondary marriage at all. Moreover, if a brother or a maternal uncle dies, his widow is available as a wife to only one of the surviving brothers or nephews, and where there are none she must marry someone else. Hence levirate marriages, even where they are possible and preferred, can actually occur in only a fraction of all cases. Similarly, when a man's wife dies she may have no unmarried sister or brother's daughter whom he can marry in her stead 43. Even when polygyny prevails, as many observant ethnographers have noted, the majority of men at any given time, or of all men during their lifetimes, normally have only a single wife; only those who survive well into middle age tend to have additional wives. The assumption that the kinship usages of an entire society are determined by the secondary marriages of a relatively small minority of the total population seems clearly deficient in realism.
This criticism is supported by the ethnographic evidence. In a quantitative test of various theories of the causation of bifurcate merging, made by the present writer 44, the levirate-sororate hypothesis of Sapir failed to receive reliable statistical confirmation, while alternative explanations propounded by Rivers, Lowie, and Kroeber were substantially corroborated. Further factual disproof will be advanced later in the present chapter under Propositions 28, 29, and 30.
The sixth and last of the factors which have been proposed as determinants of kinship terminology is the constitution of kin and local groups. Moieties, as pointed out by Tylor 45 and Rivers 46, by aligning all members of a society in two unilinear groups, tend to group together various relatives who are ordinarily distinguished and thus to result in assigning them a common kinship designation. s Thus moieties are very widely associated with bifurcate merging 47. Some of the same results can be brought about by smaller exogamous kin groups or sibs, as Lowie 48 and Kroeber 49 have pointed out. A rule of patrilocal or matrilocal residence can bring about a similar alignment of kinsmen 50.
It has been observed 51 that cross-cousin terms of the Crow type tend to be found in tribes with matrilineal sibs, and terms of the Omaha type in those with patrilineal sibs. Iroquois terminology 52 is also common in tribes with both types of descent, and White 53 has suggested a possible basis of differentiation: “When the clan system is young and weak the kinship system will be of the Dakota-Iroquois type, regardless of the sex in which descent is reckoned. As the clan system develops, however, and comes to exert its influence more and more upon the social life of the tribe, the Dakota-Iroquois terminology will be transferred into the Crow type in a matrilineal society and into the Omaha type in a patrilineal society.”
Determinants of this last general type seem to the present author perhaps the most significant of all. On a priori grounds, the alignment of relatives in extended families, clans, sibs, and moieties, as these are created by rules of residence and descent, would seem to provide precisely the kinds of social situations in which classificatory groupings of kinsmen would be most likely to arise. Abundant evidence for this assumption will be presented below. The contrary hyothesis, namely, that kinship classifications arise first and then give rise to sibs and other comparable groups 54, does not appear well founded.
The foregoing critical survey of the various theories of the determination of kinship nomenclature should reinforce an important conclusion, to wit, that the causal factors actually operating in any particular situation are always multiple. No single factor or simple hypothesis can account for all observable effects. From this it follows that different determinants must often exert their pressure in opposite directions. What operates is therefore a sort of parallelogram of forces, and the phenomena which ensue represent, not the effects of particular forces but the resultant of them all. Often, indeed, the influences exerted by opposing factors may be so evenly balanced that a relatively insignificant supplementary factor may suffice to tip the scales. Opler 55 records a particularly apt illustration. Speaking of the fact that some Apache tribes classify the maternal aunt with the mother, while others with similar social institutions call her by a distinct term, he notes that either practice may represent a functional adaptation to the situation. Such factors as matrilocal residence are consistent with the former usage, such factors as separate nuclear family dwellings with the latter. Since the Apache are thus faced with a choice between alternatives of approximately equivalent functional utility, minor or even irrelevant factors might decide the issue, here in one way, there in another.
Since multiple factors are nearly always operative, perfect statistical correlations between any particular kinship determinant and the terminological features that it tends to produce should never be expected, even if the hypothesis is entirely sound. The factor of the time lag which commonly intervenes between the appearance of a causal influence and the resulting change in nomenclature also operates to reduce the magnitude of statistical coefficients. Bearing these facts in mind, the reader will realize that positive coefficients of only moderate magnitude may often reflect genuine causal relations of considerable significance.
The presentation and testing of the present writer's theory of the determination of kinship terminology will take as its model the so-called postulational method of scientific inquiry 56. In this procedure, the most rigorous of scientific methods, all logical or rational operations are performed prior to the final empirical test, so that no fallible mental processes can intervene between the survey of the evidence and the formulation of explanatory hypotheses, to bias or distort the latter—a basic defect in much of social science theory.
In its essence, the postulational method requires the formulation of a set of hypotheses of a general character, called “postulates,” and of a series of derivative propositions of a more specific character, called “theorems.” Postulates are commonly too broad or general to be capable of direct validation. They are verified through the theorems deduced from them, the latter being formulated in such a manner that they can be projected against a body of facts and tested by merely enumerating agreements and disagreements or by making some other equally simple operation. Each theorem must be subjected to careful logical analysis in order to bring to light any further postulates or axioms which may be necessary to derive it. The theorems for testing any postulate should be as numerous, as diverse, and as representative as possible. Only after the whole framework of postulates and theorems has been systematically worked out, with all terms carefully defined and all implicit assumptions made explicit, is it put to the test by projecting the theorems against the facts. If even a single theorem fails to check with the facts, the postulate from which it is derived must be considered invalid, and the logical structure must be revised and tested again.
We shall fall somewhat short of this scientific ideal. Unfortunately the postulational method, however economical and precise as a tool of research, does not lend itself to simplicity of exposition. In order not to repel the reader, therefore, we shall dispense with as much of the usual elaborate apparatus as we can, and will strive for simplicity. Moreover, a completely rigorous use of the method would require us to reduce much of anthropological, sociological, and psychological theory—we draw heavily upon all three—to a series of logically precise and explicitly defined propositions. While this is a consummation devoutly to be wished, it is a task for the general social scientist and not for the author of a specialized contribution. To attempt it here would be presumptuous and probably premature. We shall therefore attempt a compromise and confine ourselves to our limited field. As a consequence, the propositions which we shall call “postulates” will really have the character of what are technically known as “first-order theorems,” and what we call “theorems” will actually be “theorems of the second order.” Our real postulates lie, only partially formulated, in our fundamental assumptions, which will shortly be made explicit.
In conjunction with the theorems derived from our own postulates, we shall occasionally test hypotheses advanced by other authors and also some which are not derivable from our postulates but are tentatively proposed as meriting consideration. These will be labeled as “propositions” in order to distinguish them from genuine theorems, although they will be numbered consecutively with the latter.
The testing of theorems and propositions will be accomplished by a simple enumeration of the societies in our sample of 250 which agree and disagree with each. Agreement will mean in some instances that the same kinship term, in others that different terms, are applied to two relatives, disagreement being the opposite in either case. When one of a pair of relatives may be called by either of two terms, one the same as and the other different from that applied to the other, the case is counted as half in agreement and half in disagreement. Cases of derivative terms are similarly split. Thus if MoSi is being compared with Mo, and the term for the former is a derivative one such as “little mother,” the terms for the two relatives are considered half the same and half different.
The same device is followed when two equally competent authorities differ, the one reporting agreement and the other a disagreement. Genuinely doubtful or inferential cases, moreover, are given only half weight. In all totals, however, fractions are carried to the next whole number.
The results of each enumeration are summarized by a statistical coefficient, accompanied by an index of reliability. The coefficient used is Q, the coefficient of association proposed by Yule 57. Since the sampling distribution of Q has not been established, the reliability of the associations has been determined by using the chi square (x2 ) test of independence, corrected for continuity, based on the formula given by Snedecor 58. The statistical tables include, along with the numerical data and the Q values, not the x2 values but an indication of the probability of obtaining a x2 value as large as or larger than that actually obtained, on the basis of random sampling alone. Thus 1000 indicates a probability of less than 1 in 1,000; 100, of less than 1 in 100; 20, of less than 1 in 20; 10, of less than 1 in 10; 5, of less than 1 in 5; 2, of less than 1 in 2—or, phrased in another way, reliability at or better than the one-tenth-of-one-per-cent level, the one per cent level, the five per cent level, the ten per cent level, the twenty per cent level, and the fifty per cent level of confidence, respectively. The x2 column is left blank when the reliability is very low, i.e., at less than the fifty per cent level or below 2. When Q is +1-00 or —1.00, the y2 value becomes inaccurate or inappropriate because of small cell frequencies, and an asterisk is shown in its place.
Coefficients of association range from +1.00, indicating perfect positive association, to — 1.00, indicating perfect negative association. A coefficient of .00, or close enough thereto in either direction to be presumed to diverge only through the chance of the particular sample, indicates complete independence, or an absence of association. In the testing of a theorem, such a coefficient constitutes lack of substantiation, unless other samples consistently show positive coefficients with good reliability. Any negative coefficient, of course, constitutes definite invalidation of a theorem, unless the index of reliability is very low. Positive coefficients provide confirmation, though this must be considered merely tentative if the indices of reliability are consistently low. In general, a theorem may be considered validated only when every enumerative test yields a positive coefficient of association, the decisiveness of the validation depending upon the consistency of the coefficients and the level of reliability. When the data were gathered, the author did not intend to undertake an analysis of the determinants of kinship, but expected to use kinship terminology only as one of the aspects of social structure which may operate to channelize sexual behavior. Consequently he recorded only such kinship data as promised to be useful for the purpose in mind. Thus material was gathered on kin relationships between males and females, but not on patterned behavior between pairs of male relatives or pairs of female kinsmen. Similarly, kinship terms were noted only when employed by males for female relatives; no terms for male relatives were collected, and no female-speaking terms for females. Only male-speaking terms for female relatives, therefore, are available for the testing of theorems. While they are probably sufficiently conclusive, a complete demonstration must await comparable analysis of other kinship terms.
The hypothesis to be advanced for the determination of kinship terminology will be formulated in a single very general postulate, from which 26 distinct theorems will be derived and individually tested. The postulate is not designed to cover the entire field of kinship nomenclature but only the area within which the great majority of recognized kinship problems lie, namely, the classification and differentiation of terms for secondary and tertiary relatives of Ego's generation and of the first ascending and first descending generations. Here uniform treatment is possible because the pertinent criteria are the same—generation, sex, affinity, collaterality, bifurcation, and polarity. For primary relatives the subsidiary criteria of relative age 59 and speaker's sex assume equal importance, and provision should be made for them in the formulation of a postulate. On the other hand, in regard to the second and higher ascending and descending generations, and to distant relatives, extension and classification become so general that most and sometimes all of the major criteria are ignored.
The postulational method requires that all assumptions upon which a postulate is based be made explicit. Although this is almost never done in the theoretical writings of social science, we shall make a serious effort to conform to the scientific ideal. The terms to be employed in the postulate will be precisely defined—another requirement of the method—during the exposition of the underlying assumptions.
Our first assumption is that all human behavior, including that which is called cultural, conforms to the fundamental principles of behavior as these are being laid bare by psychologists. Culture change is the product of individual behavior in the mass, operating' over time and adapting to the changing conditions of existence through such mechanisms as trial-and-error learning and imitation 60.
All cultural responses in particular situations are similarly understandable in terms of the established habits of the reacting individuals, their motivations, and the environmental and social conditions under which they must behave. We assume, in short, that there is no conflict whatsoever between valid psychological and cultural interpretations of behavior.
It is assumed, secondly, that all cultural phenomena are historical. We specifically disclaim any evolutionary, cyclical, or other process of change according to which cultural forms are interpreted otherwise than as products of the prior events and existing conditions in the particular local context. Since behavior, whether individual or collective, depends upon the reciprocal interaction of the same factors—external stimuli and conditions, the habitual response tendencies and motivations of the behaving organisms, and an innate mechanism of behavior—we recognize no conflict between history and psychology. We must insist, however, that the psychological mechanism operates only with the materials which history provides, including under history the life histories of individuals. In default of such materials, psychology can offer no valid interpretation of any cultural phenomenon. The social scientist may—indeed, he must resort to the psychologist for answers to the question “how?” But for a solution of problems concerned with “what?” “when?” “where?” or even “why?” he must look to history for the independent variables.
It is assumed, in the third place, that the terminological classification of kinsmen is but a special case of linguistic classification, and / that its function is to designate types of relatives in accordance with their socially relevant common characteristics, such as the patterned behavior expected from them. Whenever there is occasion to individualize a relative, personal names provide a universal means of so doing.
It relatives under a single term can occur only on the basis of regular and perceptible similarities, or the absence of regular and significant dissimilarities, between them. This assumption is derived from the psychological process of “stimulus generalization,” as identified and described by Hull 61. Generalization is the mechanism by which any response, learned in connection with a particular stimulus or pattern of stimuli, tends also to be evoked by other stimulus situations in proportion to their similarity to the original one. It follows from our first assumption that any cultural phenomenon in which —responses (such as kinship terms) associated with particular stimulus objects (such as relatives) are transferred to other stimulus contexts will reflect in the behavior of numbers of people the same mechanism that is observable in any single individual. Since, however, the cultural phenomenon in question involves symbolic mental processes, i.e., those of language, as well as pluralistic behavior on the part of many individuals in a society, it seems advisable not to speak of the “generalization” of kinship terms but to use the word extension (and the verb “extend”) to denote the social counterpart and derivative of the psychological process.
A fifth assumption is that different kinship terms will be applied to two or more relatives to the extent that they exhibit regular and significant dissimilarities or the absence of regular and perceptible similarities. This assumption is derived from the psychological process of “discrimination” as defined by Hull 62 by a procedure paralleling that for “generalization.” Discrimination is the fundamental mechanism by which the generalization of responses along a continuum of decreasing similarity in the stimuli evoking them is checked at the point where the behavior ceases to be adaptive and a situation is created favorable to the appearance of other responses. In reference to kinship terminology we shall employ the term differentiation (and the verb “differentiate”) to denote the symbolic and social counterpart of the psychological process of discrimination.
Our sixth assumption is that the extension or differentiation of kinship terms depends in every individual case upon the total net effect of all similarities and dissimilarities exhibited by the relatives in question. Similarities and dissimilarities can be classified into three groups:

  1. the absence or presence of differences inherent in the very nature of kinship structure in consequence of the biology of heredity and of the universal cultural fact of family exogamy
  2. the absence or presence of recurrent but not universal features of social organization and of associated cultural rules of residence, descent, and marriage which have the effect of increasing or decreasing the degree of similarity or dissimilarity prevailing between relatives of particular kinship categories
  3. the absence or presence of other cultural or environmental factors which can affect the degree of similarity between categories of relatives, including in the main local and non-repetitive historical influences

It is assumed, in the seventh place, that the only inherent differences which are of fundamental significance in the classification of secondary and tertiary relatives are the six major criteria of generation, sex, affinity, collaterality, bifurcation, and polarity. In the postulate and theorems these six criteria will be designated as inherent distinctions. The subsidiary criteria of relative age, speaker's sex, and decedence are either rare in themselves or are recognized only infrequently and sporadically in the terminology for secondary and tertiary relatives, and will consequently not be classed for present purposes as inherent distinctions. This seventh assumption is derived from the classic analysis by Kroeber 63 of the factors underlying terminological differentiation.
The term kin-type, which will appear in the postulate and theorems, may now be defined. A kin-type is a class of relatives defined by all six major criteria, i.e., consisting exclusively of relatives between whom no inherent distinction exists. A kin-type can include only siblings of the same sex, but not all groups of brothers or of sisters constitute kin-types, since the criterion of collaterality often divides such a group into two kin-types, e.g., Mo and MoSi or SoWi and SoWiSi. The “equivalence of brothers” or of siblings, as used by Radcliffe-Brown 64, refers to the classificatory similarity of the relatives of a kin-type.
Our eighth assumption is that the six inherent distinctions vary in their relative efficacy in producing differentiation in kinship terminology. By relative efficacy we mean the magnitude of the influence v of any one, relative to that of the others, in causing terminological differentiation. Relative efficacy can be determined only inductively. Our data do not cover the criteria of sex and polarity, but they make possible an estimate of the relative efficacy of the other four inherent distinctions. This is assumed to be approximately of the order of 25 for generation, 5 for affinity, 1 for bifurcation, and 1 for collaterality on the basis of a wholesale tabulation of data from 221 of our sample societies. For comparability with bifurcation and collaterality, affinal and consanguineal kin-types in Table 14 were tabulated only within the same generation 65.

Table 14
Inherent Distinctions Pairs of Kin-types Compared Different Terms Same Terms
Number Percentage Number Percentage
Generation 276 33,071 97.7 796 2.3
Affinity 33 4,518 87.2 662 12.8
Bifurcation 16 1,179 51.6 1,125 48.4
Collaterality 11 678 40.3 1,004 59.7

It is assumed, in the ninth place, that practical considerations everywhere exert pressure in the direction of reducing the total number of kinship terms from the number necessary if all six inherent distinctions were fully and simultaneously recognized (33 secondary plus 151 tertiary kin-types, in addition to the several categories of primary relatives) to a relatively modest optimal number. This assumption is derived from observations by Kroeber66 and others to the effect that the total number of distinct kinship terms in different societies varies within a rather limited range around a mean of about 25. It follows from this assumption that in every kinship system at least some of the six inherent distinctions must be inoperative in at least some secondary and tertiary kin-types, and it follows from the eighth assumption that this will be most likely to happen in the case of those distinctions with the lowest relative efficacy, namely, collaterality and bifurcation. Gross statistics indicate that both of these are ignored or “overridden” approximately as often as they are observed.
It is assumed, in the tenth place, that inherent distinctions are wholly incapable of accounting by themselves for cross-cultural differences in kinship terminology, and are effective only in conjunction with other types of similarities and dissimilarities, which are therefore the real determining factors. This assumption is self-evident; factors which are universally the same can never account for differences 67.
An eleventh assumption is that certain recurrent but non-universal features of social structure can significantly increase or decrease the similarities or dissimilarities between particular categories of relatives in the societies where they prevail, thus accentuating or minimizing the effect of particular inherent distinctions and thereby operating as determinants of kinship terminology. This assumption is derived principally from the kinship theories of Rivers and Radcliffe-Brown. The number and variety of such features—forms and preferential rules of marriage, types of family and clan organization, rules of residence and descent, and varieties of unilinear and bilateral kin groups—provide a wide range of causal factors. That these are of greater significance as determinants of kinship terminology than local and non-repetitive historical influences is not an assumption of this study, though it is an overwhelming and not wholly expected conclusion from it.
An aspect of social structure or of associated cultural behavior which significantly increases the perceptible dissimilarities between relatives of different kin-types in particular societies will be called a social differential. An aspect of social structure or of associated cultural behavior which creates significant similarities between relatives of different kin-types will be called a social equalizer. Like inherent distinctions, social differentials and equalizers differ in relative efficacy, which can be determined only inductively. A rough estimate will be attempted at the end of the present chapter.
Social equalizers and differentials appear to exert then minimizing and accentuating effects through the creation of several different kinds of similarity and dissimilarity, of which the following may be distinguished:

  • Coincidence—a similarity between two kin-types owing to the probability that the members of both are the same persons, as can happen as a result of certain social equalizers. Sororal polygyny, for example, tends to equate WiSi with Wi.
  • Proximity—a similarity or dissimilarity in spatial relations. For example, matrilocal residence, by bringing Mo and MoSi together as close neighbors or even actual housemates, operates as a social equalizer to make more likely their designation by a single classificatory term, and by separating FaSi spatially from both of them operates as a social differential to favor the application of a special kinship term to her.
  • Participation—a, similarity or dissimilarity in group membership. Thus patrilineal descent places BrDa and Da in the same lineage, sib, or moiety and SiDa in a separate group, thereby promoting extension of the term for Da to BrDa and differentiation of the term for SiDa.
  • Analogy—a similarity relative to a parallel relationship. Extension of the term “mother” to MoSi, for example, acts as a social equalizer in the case of MoSiDa, increasing her likelihood of being called “sister” even under patrilineal descent when Si and MoSiDa are usually neither sibmates nor neighbors.
  • Immateriality—a negative similarity resulting from the functional unimportance of the relatives of two kin-types, whereby a sufficient basis for differentiating them is lacking. Immateriality appears chiefly with respect to distant relatives. In the English kinship system, for example, it operates as an equalizer to favor the extension of the term “cousin” to various distant relatives of little importance, without reference to distinctions of sex, generation, or collaterality.

Our twelfth assumption is that the forms of social structure are not determined by kinship patterns or terminology, or influenced in any major degree by them, but are created by forces external to social organization, especially by economic factors. It is assumed herewith, for example, that the available sources of food and the techniques of procuring it affect the sex division of labor and the relative statuses of the sexes, predisposing peoples to particular rules of residence, which can eventuate in the formation of extended families, clans, and sibs. It is further assumed that the prevailing types and distribution of property favor particular rules of inheritance, that wealth or its lack affects marriage (e.g., encouraging or inhibiting polygyny), and that these and other factors external to social structure can strongly influence rules of residence and marriage and through them the forms of social organization and kinship structure. This assumption is derived from the analysis by Lowie 68 of the origin of sibs, from our own supportive evidence as presented in Chapter 8, and from the various theorists from Marx to Keller who have stressed the importance of economic factors in cultural change 69.
Our thirteenth and final assumption is that a change in social structure which significantly alters the social equalizers and differentials affecting particular kin-types will be followed by adaptive changes in the pertinent kinship terms only after a lapse of time. This assumption is derived from the suggestion by Lowie 70 of the importance of “the time element” and especially from the sociological hypotheses of a “strain toward consistency” and a “cultural lag” between the constituent elements of a culture, as propounded respectively by Sumner 71 and Ogburn 72. Herein lies a probable explanation of many of the instances in which the data from particular societies do not correspond with the theoretical expectations expressed in our postulate. The underlying assumptions having been made explicit, and all relevant terms having been defined, the basic postulate may now be formulated.

Postulate 1: The relatives of any two kin-types tend to be called by the same kinship terms, rather than by different terms, in inverse proportion to the number and relative efficacy of (a) the inherent distinctions between them and (b) the social differentials affecting them, and in direct proportion to the number and relative efficacy of the social equalizers affecting them.

Rephrased in looser language, the postulate states that the extension and differentiation of kinship terminology is the product of the joint interplay of all inherent and cultural factors which significantly affect the degree of similarity or dissimilarity between particular categories of relatives.
The serious reader who feels that there is something peculiarly incomprehensible or baffling about kinship terminology—an illusion shared even by some anthropologists—is nevertheless urged to make an attempt to understand the theorems and their validation, since they are basic to a comprehension of the general theory of the evolution of social organization in Chapter 8 and of the principles governing the extension of sexual privileges and taboos as presented in Chapters 9 and 10. To paraphrase Lawrence 73, if any primitive tribe can invent a kinship system, any civilized reader should be capable of comprehending it.
The first theorem to be tested is of a rather special nature, and is given priority in order to provide a basis for the inclusion of certain relatives in the testing of subsequent theorems. It deals with the structural similarity between nuclear families in any society as a social equalizer facilitating the extension of kinship terms by analog). Whenever, for the various reasons to be disclosed in subsequent theorems, the term for “mother” is extended to MoSi, FaSi, FaBrWi, or MoBrWi, the daughters of such “classificatory mothers” tend to be called by the same term as sister, on the analogy of the relation between own mother and her daughter. Similarly when WiSi or BrWi is called by the term for “wife,” WiSiDa or BrDa, respectively, will tend to be called “daughter.” In all instances the social squalizer overrides the inherent distinction of collaterally, and in the case of WiSiDa that of affinity as well. While sample tabulations have shown that the same principle operates elsewhere, it is not necessary for our purposes to establish it for other relatives, and the theorem and its validation will consequently be concerned only with the relatives mentioned above.

Theorem 1: When secondary or tertiary relatives of any kin-type are called by a kinship term used to denote a primary relative, the daughters of such secondary or tertiary relatives tend to be called by the same kinship term as the daughter of the primary relative.
In this and all subsequent theorems it is naturally assumed that all factors other than the social equalizers and differentials under consideration are held constant.

Theorem 1 is subjected to test by the data compiled in Table 15. It is decisively validated by high and consistent positive coefficients of association and, in all but one instance, by very high indices of reliability.

Table 15
  Parent Called "Mother" Parent CalledOtherwise
Pairs of Relatives Child Called ''Sister" Child Called Otherwise Child Called ''Sister Child Called Otherwise Statistical Indices
  Q X2
FaSi-FaSiDa 18 22 42 156 +.50 1000
MoSi-MoSiDa 110 16 62 34 +.58 1000
FaBrWi-FaBrDa 85 9 50 24 +.64 1000
MoBrWi-MoBrDa 17 10 29 113 +.74 1000
  Parent Called "Wife" Parent Called Otherwise
  Child Called "Daughter Child Called "Daughter Child Called "Daughter Child Called "Daughter
BrWi-BrDa 25 4 96 68 +.67 100
WiSi-WiSiDa 10 4 44 33 +.30 2

The “rule of uniform descent” as defined by Tax 74 corresponds closely to Theorem 1, and is confirmed by the validation of the latter. It should be expressly pointed out that neither the rule nor the theorem invokes a psychological constant as the prime determinant of a cultural variable. On the contrary, they assume only that when a term has been extended to one relative (FaSi, MoSi, FaBrWi, MoBrWi, BrWi, or WiSi) for whatsoever reasons, family structure provides the necessary similarity to facilitate another extension with respect to the daughter of that relative.
Forms of marriage constitute a second principal group of social equalizers and differentials. The major forms are polygyny, polyandry, and monogamy. Polyandry is the general and preferred form in only two of the sample societies, the Marquesans and the Todas, and is consequently too rare for statistically reliable testing. Polygyny may be either common or infrequent. If it is exceptional in actual practice, even though permitted, it cannot be expected to exert any considerable influence upon the alignment or terminological classification of relatives, which should be essentially the same as under monogamy. We have consequently classed as monogamous for present purposes all societies in which less than 20 per cent of all marriages are plural unions, considering as polygynous only those with a higher percentage. Polygyny, moreover, may be either sororal or non-sororal. The polygynous societies have therefore been divided into two groups—one in which secondary wives are exclusively the sisters of the first wife and one in which non-sororal polygyny is common, even though sororal polygyny also occurs or is actually somewhat preferred. For purposes of computation, all polygynous societies for which exclusive sororal polygyny is not specifically reported have been classed as non-sororal.
Sororal polygyny operates as a social equalizer through both coincidence and participation. By making it probable that such relatives as MoSi and Mo, WiSi and Wi, MoSiDa and Si, and WiSiDa and Da will actually be the same persons or at least members of the same family as Ego and one another, it should facilitate the extension of kinship terms, within each pair, from the latter to the former.
Theorem 2: In the presence of sororal polygyny, terms for primary relatives tend to be extended, within the same sex and generation, to their collateral relatives through females.
The data compiled in Table 16 provide only tentative confirmation of the theorem since, though the coefficients of association are exclusively positive, the indices of reliability are very low throughout. There are indications that the results might be improved by a somewhat less rigorous definition of sororal polygyny.

Table 16
  Sororal polygyny Other Forms of Marriage
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
MoSi-Mo 11 7 117 94 +.12
WiSi-Wi 4 17 20 171 +.34 2
MoSiDa-Si 16 2 147 54 +.49 2
WiSiDa-Da 6 2 45 35 +.40

In the testing of many of the theorems which follow, it will be useful to employ a fourfold classification of relatives proposed independently by Lowie 75 and Kirchhoff 76. This classification is based upon the four possible permutations in the application of the inherent distinctions of collaterality and bifurcation. Either or both may be recognized, or both may be ignored, in the terminological designation of kinsmen. The four possibilities can be illustrated by the terms for parents and their siblings. Recognition of both bifurcation and collaterality produces so-called bifurcate collateral terminology, in which paternal and maternal uncles and aunts are terminologically differentiated both from parents and from one another. This yields separate terms for Fa, FaBr, MoBr, Mo, MoSi, and FaSi. The recognition of bifurcation combined with the ignoring of collaterality produces bifurcate merging terminology, in which the Fa and FaBr are called by one classificatory term, and. the Mo and MoSi by a second, while the MoBr and FaSi are denoted by distinct terms.
The recognition of collaterally but not of bifurcation yields so-called lineal terminology; as in our own system, FaBr and MoBr are grouped under one classificatory term, and FaSi and MoSi under a second, with separate denotative terms for each parent. Finally, the ignoring of both inherent distinctions leads to so-called generation terminology, in which the term for father is extended to both uncles, and often to remoter relatives as well, while a second classificatory term serves alike for mother and both aunts.
One or another of the above four types can be applied to any trio of kinsmen in the same generation and of the same sex, provided the second is a collateral relative of the first in the same line of descent and the third is related to Ego in the same manner as the second except for a difference in the sex of the connecting relative. The demonstrations of our theorems will, whenever appropriate, use six such trios, in two of which, the second and sixth in the list below, an additional inherent distinction, that of affinity, is involved. These trios, with the four alternative methods of distributing kinship terms among them, are tabulated below.

Kinship terms distribution methods
Bifurcate Collateral Lineal Terminology
Mo, MoSi, FaSi Mo, MoSi = FaSi
Mo, FaBrWi, MoBrWi Mo, FaBrWi = MoBrWi
Si, FaBrDa, FaSiDa Si, FaBrDa = FaSiDa
Si, MoSiDa, MoBrDa Si, MoSiDa = MoBrDa
Da, BrDa, SiDa Da, BrDa = SiDa
Da, WiSiDa, WiBrDa Da, WiSiDa = WiBrDa
Bifurcate Merging Generation Terminology
Mo = MoSi, FaSi Mo = MoSi = FaSi
Mo = FaBrWi, MoBrWi Mo = FaBrWi = MoBrWi
Si = FaBrDa = FaSiDa Si = FaBrDa, FaSiDa
Si = MoSiDa = MoBrDa Si = MoSiDa, MoBrDa
Da = BrDa, SiDa Da = BrDa = SiDa
Da = WiSiDa, WiBrDa Da = WiSiDa = WiBrDa

The fifth theoretically possible combination, that of equating FaSi with Mo and differentiating MoSi from both, or an equivalent grouping in other trios, almost never occurs in actual fact; there are only four sporadic instances for the six trios in our entire sample.
Sororal polygyny, by bringing collateral relatives through females into proximity with lineal relatives and thus separating them spatially from collateral relatives through males, tends to emphasize the inherent distinction of bifurcation as well as to minimize that of collaterality. It should thus operate to produce bifurcate merging terminology. A theorem is formulated to cover the two trios of relatives for which its influence is most directly relevant.

Theorem 3: Sororal polygyny tends to be associated with kinship terminology of the bifurcate merging type for aunts and for nieces by marriage. This theorem is subjected to test in Table 17. Despite the fact that one computation yields a perfect positive coefficient of association, Table 17 can be regarded as providing only tentative validation of Theorem 3 since the x2 index of reliability is low in the only instance where it can appropriately be applied.

Table 17
  Sororal polygyny Other Forms of marriage
Trio of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 9 9 81 129 +.23 2
WiBrDa-WiSiDa-Da 5 0 27 42 +1.00

Polygyny of the non-sororal type operates as a social differential rather than as a social equalizer. Whereas under sororal polygyny MoSi lives in close proximity to Mo, MoSiDa to Si, and WiSiDa to Da, these and all other collateral relatives are spatially separated from lineal relatives under non-sororal polygyny. The mother is now immediately surrounded by her co-wives, not her sisters; the sister and daughter by their half sisters, not their parallel cousins. In other words, the members uf the polygynous family intervene to separate the members of tlie nuclear family from aunts, nieces, and first cousins who might otherwise live in their immediate vicinity. This happens irrespective of rules of residence and descent, and the resulting spatial separation of collateral relatives outside of the polygynous family serves to inhibit the extension to them of the kinship terms for primary relatives. In short, non-sororal polygyny tends to prevent the occurrence of merging.
Theorem 4: In the presence of non-sororal polygyny, collateral relatives outside of the polygynous family tend to be terminologically differentiated from primary relatives of the same sex and generation.
Table 18, which assembles the data to test Theorem 4, yields five low positive coefficients and the only negative coefficient that we shall encounter in testing any of our theorems. This, coupled with the fact that all indices of reliability are low, suggests that the validation of the theorem is inconclusive.

Table 18
  Non-sororal polygyny Other Forms of Marriage
Pairs of Relatives Different Terms Same Terms Different Terms Same Term Statistical Indices
  Q X2
MoSi-Mo 53 58 48 70 +.14 2
FaBrWi-Mo 36 46 37 45 -.02
FaBrDa-Si 28 84 23 95 +.16 2
MoSiDa-Si 30 78 23 90 +.21 2
BrDa-Da 38 61 39 73 +.08
WiSiDa-Da 22 28 16 24 +.08

Careful analysis, however, counteracts this impression. Non-sororal polygyny, as it happens, is very highly associated with patrilocal residence, patrilineal descent, and the types of kin groups which depend upon these rules, all of which factors, as will presently be seen, exert an influence directly counter to that of polygyny, i.e., in favor of merging. The evidence from our 250 societies shows that non-sororal polygyny tends to occur in the presence of patrilocal residence and patrilineal descent by coefficients of association of +.66 and +.58, respectively, with a x2 index of reliability of 1000 in each case. Hence the tendency toward positive association in Table 18 and the appearance of five positive signs out of six, which could occur only about once in ten times by random sampling if the true relationship were zero, are attained in the teeth of the strongest of countervailing factors. Under these circumstances, the results must be regarded as very much more significant than they appear on the surface.
The influence of non-sororal polygyny is shown more clearly when Table 18 is corrected for one of the major groups of countervailing factors. This is done in Table 19 for the factor of patrilineal descent by eliminating from the compilation all societies with exogamous patrilineal lineages, sibs, and moieties. In the new tabulation all the coefficients are positive in sign, their average magnitude is appreciably raised, and in one instance reliability at the five per cent level of confidence is attained, despite the fact that the opposing factor of patrilocal residence is still not eliminated. Random sampling could produce six positive signs only once in 64 times. Theorem 4 can therefore be regarded as substantially validated.

Table 19
  Non-sororal polygyny Other Forms of Marriage
Pairs of Relatives Different Terms Same Term Different Terms Same Term Statistical Indices
  Q X2
MoSi-Mo 25 19 34 51 +.33 2
FaBrWi-Mo 16 11 28 29 -.30
FaBrDa-Si 12 33 18 69 +.16
MoSiDa-Si 11 22 20 68 +.06
BrDa-Da 23 17 28 50 +.41 20
WiSiDa-Da 7 9 12 16 +.02

It is already apparent, and will become increasingly clear below, that kinship terminology in patrilocal and patrilineal societies, in contrast to societies with other rules of residence and descent, regularly reflects two sets of opposing influences. The rules of residence and descent promote merging, whereas the associated phenomenon of general polygyny inhibits the tendency to extend terms from lineal to collateral relatives. Whether merging does or does not occur is therefore likely to depend upon such facts as which of the opposing influences has been longer established or possesses the greater relative efficacy.
By operating as a social differential to oppose merging, non-sororal polygyny militates against kinship terminology of either the generation or the bifurcate merging types. By virtue of its association with patrilocal residence and patrilineal descent, which favor bifurcation. it is more conducive to bifurcate collateral than to lineal terminology. In the usual polygynous society, collateral relatives on the maternal side are ordinarily differentiated from those on the paternal side by the fact that they reside in another clan or community and belong to different consanguineal kin groups. Both types of collateral relatives, moreover, are differentiated from lineal relatives; the half siblings, stepparents and stepchildren of the polygynous family intervene between them. Bifurcate collateral terms accord with these conditions. Non-sororal polygyny is, in fact, the only social determinant we have discovered that favors bifurcate collateral terminology. With it are commonly associated distinct terms for Fa, FaBr, and MoBr, for Mo, MoSi, and FaSi, and for the several members of all comparable trios of relatives.

Theorem 5: Non-sororal polygyny tends to be associated with kinship terminology of the bifurcate collateral type.
The theorem is validated by the data assembled in Table 20, wherein the coefficients of association are exclusively positive in sign and reliability attains the five per cent level or better for three of the six trios, despite the fact that patrilocal residence and patrilineal descent again exert their countervailing influence.

Table 20
  Non-sororal polygyny Other Forms of Marriage
Trios of Relatives Bifurcate Collateral Terms Other Terms Bifurcate Collateral Terms Other Terms Statistical Indices
  Q X2
MoSi-Mo 25 19 34 51 +.33 2
FaBrWi-Mo 16 11 28 29 -.30
FaBrDa-Si 12 33 18 69 +.16
MoSiDa-Si 11 22 20 68 +.06
BrDa-Da 23 17 28 50 +.41 20
WiSiDa-Da 7 9 12 16 +.02

Theorem 5 is further validated in Table 21, which corrects Table 20 by eliminating all societies with exogamous unilinear kin groups. These, as will be demonstrated under Theorem 19, tend to produce bifurcate merging terminology, thereby counteracting the influence of non-sororal polygyny in promoting bifurcate collateral nomenclature. Table 21 reveals a consistent increase in the magnitude of coefficients of association, and the indices of reliability, though lowered in two instances by the reduction in the total number of cases, in one instance attain the maximum level of confidence at one tenth of one per cent.

Table 21
  Non-sororal polygyny Other Forms of Marriage
Trios of Relatives Bifurcate Collateral Terms Other Terms Bifurcate Collateral Terms Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 21 9 15 39 +.54 1000
MoBrWi-FaBrWi-Mo 9 6 7 21 +.55 10
FaSiDa-FaBrDa-Si 4 27 2 54 +.60 2
MoBrDa-MoSiDa-Si 5 27 2 54 +.68 5
SiDa-BrDa-Da 11 16 12 38 +.35 2
WiBrDa-WiSiDa-Da 4 9 3 14 +.22

Monogamy should not of itself exert any special influence upon kinship terminology, and analysis of the data shows that it tends to be associated with terms consisteflt with the prevailing rule of residence (see Theorems 6 to 12). The isolated monogamous family, however, is definitely associated with nomenclature of the lineal type (see Theorem 14). Tests of the four theorems concerned with marriage forms have indicated that these exercise an appreciable influence on kinship terminology, although the demonstration has in several cases fallen short of being conclusive.
The various cultural rules prescribing the place of residence of a married couple constitute a third and extremely important group of social equalizers and differentials. To recapitulate, a rule of residence is called “matrilocal” when a married couple normally live with or near the wife's parents, “patrilocal” when they reside with or near the husband's parents, “matri-patrilocal” when the prevailing rule is matrilocal for a year or two and thereafter shifts to patrilocal, “hilocal” when residence is optionally matrilocal or patrilocal depending upon circumstances or personal preferences, “avunculocal” when the couple live with or near the husband's maternal uncle, and “neolocal” when they establish a new household without reference to the location of the families of either. Any rule of residence affects the proximity of particular categories of relatives, operating sometimes as a social differential by regularly separating spatially the members of two kin-types, sometimes as a social equalizer by regularly bringing them into proximity with one another or by separating them from Ego in like degree.
Patrilocal, matri-patrilocal, matrilocal, and avunculocal residence, classed collectively as unilocal rules, tend to bring into physical proximity a group of relatives linked to one another by a single line of descent, i.e., those related either through males or through females, and to segregate them from relatives through the opposite sex. Hence these residence rules, in contrast to bilocal and neolocal residence, operate as a social differential to support the inherent distinction of bifurcation.

Theorem 6: In the presence of patrilocal, matri-patrilocal, matrilocal, or avunculocal residence, separate terms tend to be applied to relatives of the same generation who are linked to Ego through connecting relatives of different sex.

The data assembled in Table 22 decisively validate this theorem. Coefficients of association are uniformly high and positive, and the probabilities of the results occurring by chance are minimal.

Table 22
  Unilocal Residence Neolocal or Bilocal Residence
Pairs of Relatives Different Terms Same Terms Different Terms Same Terms Statistical Indices
  Q X2
FaSi-MoSi 157 49 20 15 +.41 20
FaBrWi-MoBrWi 96 33 14 11 +.39 5
FaBrDa-FaSiDa 136 60 13 24 +.61 1000
MoBrDa-MoSiDa 129 59 11 24 +.65 1000
BrDa-SiDa 143 33 16 13 +.56 1000
WiBrDa-WiSiDa 48 17 8 7 +.42 2

Matrilocal residence brings together a group of women who are related to one another through women, and thus tends to minimize the inherent distinction of collaterality and to favor the extension of kinship terms from lineal relatives to the collateral relatives who reside with them. In some instances, as in the case of WiSiDa and Da, it also tends to override the inherent distinction of affinity. These anticipated effects are expressed in Theorem 7. Avunculocal residence creates alignments of relatives which tend to be intermediate between those resulting from patrilocal and from matrilocal residence. Since the cases are too few in number (eight in our total sample) to justify separate statistical treatment, they are grouped with those of matrilocal residence in the following theorems and tabulations.

Theorem7 : In the presence of matrilocal or avunculocal residence, terms for primary relatives tend to be extended, within the same generation, to their collateral relatives through females.

The data in Table 23 confirm Theorem 7 by positive coefficients of association of fair consistency and considerable magnitude and by moderate indices of reliability.

Table 23
  Matriolocal or Avuncular Residence Other Rules of Residence
Pairs of Relatives Same Terms Different Terms Same Terms Different Terms Statistical Indices
  Q X2
MoSi-Mo 34 12 103 92 +.46 20
WiSi-Wi 6 33 21 158 +.16
MoSiDa-Si 40 5 139 48 +.47 10
WiSiDa-Da 13 4 40 35 +.47 5

As a corollary of Theorems 6 and 7, which show that matrilocal and avunculocal residence promote bifurcation and merging respectively, the same factors should produce bifurate merging terminology.

Theorem 8: Matrilocal and avunculocal residence tend to be associated with kinship terminology of the bifurcate merging type.

The data compiled in Table 24 validate Theorem 8 rather decisively. The coefficients of association are uniformly positive, relatively high, and quite consistent with one another, and the degree of reliability is generally good.
Patrilocal residence with its matri-patrilocal variant should logically produce bifurcate merging terminology precisely as do matrilocal and avunculocal residence. Theorem 6 has shown that it supports the inherent distinction of bifurcation. Since it brings into proximity such relatives as FaBrWi and Mo, FaBrDa and Si, and BrDa and Da, it should also serve to override the inherent distinction of collaterality between these relatives, and thus produce merging. That patrilocal residence has these effects is unquestionable, but they tend to be obscured by the strongly associated factor of nonsororal polygyny, which exerts a contrary influence. While patrilocal residence of itself favors bifurcate merging terminology, the polygyny which commonly accompanies it inhibits merging (see Theorem 4) and favors bifurcate collateral nomenclature (see Theorem 5). In formulating a theorem, however, this countervailing influence must be ignored.

Table 24
  Matriolocal or Avuncular Residence Other Rules of Residence
Pairs of Relatives Same Terms Different Terms Same Terms Different Terms Statistical Indices
  Q X2
MoSi-Mo 34 12 103 92 +.46 20
WiSi-Wi 6 33 21 158 +.16
MoSiDa-Si 40 5 139 48 +.47 10
WiSiDa-Da 13 4 40 35 +.47 5

Theorem 9: Patrilocal and matri-patrilocal residence tend to be associated with kinship terminology of the bifurcate merging type.

The testing of Theorem 9 presents certain difficulties. Matrilocal and avunculocal residence have already been found to exert an identical influence (see Theorem 8), and one which is not appreciably counteracted by non-sororal polygyny. If, consequently, societies with these rules of residence were included in the computations of data, they would tend to hide the influence of patrilocal residence. Table 25 therefore omits all cases of matrilocal and avunculocal residence, comparing patrilocal only with neolocal and bilocal societies. Only three trios of relatives are included-those in which the second member is brought into proximity with the third by patrilocal residence. Aunts and maternal cousins reflect the same influence, but only in derivative fashion, and the wife's nieces are commonly too remote to be significantly affected. Table 25 supports the theorem decisively with positive coefficients of association that are high, consistent, and extremely reliable.

Table 25
  Patrilocal or Matri-patrilocal Residence Neolocal or Bilocal Residence
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
MoBrWi-FaBrWi-Mo 42 56 5 19 +.48 10
FaSiDa-FaBrDa-Si 87 67 9 40 +.62 1000
SiDa-BrDa-Da 73 64 8 36 +.67 1000

Theorem 5 has shown that non-sororal polygyny tends to promote bifurcate collateral terminology. We have also noted that a high degree of association exists between this type of polygyny and patrilocal residence. Hence it follows as a corollary that ~atrilocal residence should tend to be associated with bifurcate collateral terminology, even though it is not to be regarded as a cause of the latter. This anticipated relation may be formulated as a theorem.

Theorem 10: Patrilocal and matri-patrilocal residence, because of their association with non-sororal polygyny, tend to be accompanied by kinship terminology of the bifurcate collateral type.

The data compiled in Table 26 validate this theorem with positive coefficients of association which are consistently positive in sign, though the indices of reliability are relatively low. In judging the significance of these results, the reader should not forget that they are attained in the face of the strong countervailing influence demonstrated in Theorem 9.

Table 26
  Patrilocal or Matri-patrilocal Residence Other Residence Rules
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 58 100 20 63 +.29 10
MoBrWi-FaBrWi-Mo 30 68 13 31 +.03
FaSiDa-FaBrDa-Si 19 135 6 75 +.49 2
MoBrDa-MoSiDa-Si 23 119 5 75 +.49 2
SiDa-BrDa-Da 37 100 14 60 +.23 2
WiBrDa-WiSiDa-Da 16 33 4 25 +.50 5

Bilocal residence produces an alignment of kinsmen very different from that resulting from any rule of unilocal residence. It brings some collateral relatives through males and some through females into proximity with lineal relatives, and thus tends to counteract the inherent distinctions among all three. The overriding of both collaterality and bifurcation produces terminology of the generation type, including the extension of sibling terms to all cousins.

Theorem 11: Bilocal residence tends to be associated with kinship terminology of the generation type.

This theorem is conclusively validated in Table 27. The coefficients of association are high in value, positive in sign, extremely consistent, and almost maximally reliable.

Table 27
  Patrilocal or Matri-patrilocal Residence Other Residence Rules
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 58 100 20 63 +.29 10
MoBrWi-FaBrWi-Mo 30 68 13 31 +.03
FaSiDa-FaBrDa-Si 19 135 6 75 +.49 2
MoBrDa-MoSiDa-Si 23 119 5 75 +.49 2
SiDa-BrDa-Da 37 100 14 60 +.23 2
WiBrDa-WiSiDa-Da 16 33 4 25 +.50 5

Neolocal residence, such as characterizes our own society, results in the spatial separation of lineal relatives from all collateral relatives irrespective of the sex of the connecting relative. It consequently emphasizes the inherent distinction of collaterality. It minimizes the inherent distinction of bifurcation, since it creates no residential distinctions between collateral relatives through males and those through females. In combination, these two effects produce terminology of the lineal type.

Theorem 12: Neolocal residence tends to be associated with kinship terminology of the lineal type.

Table 28, which tests this theorem, omits the data for a wife's nieces. These relatives are marginal under neolocal residence and, as in our own society, elementary kinship terms are commonly lacking for them. The theorem is conclusively validated by high, positive, consistent, and reliable coefficients.

Table 28
  Neolocal Residence Other Residence Rules
Trios of Relatives Lineal Terms Other Terms Lineal Terms Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 5 11 18 206 +.68 100
MoBrWi-FaBrWi-Mo 3 6 18 125 +.55 2
FaSiDa-FaBrDa-Si 7 11 19 197 +.74 1000
MoBrDa-MoSiDa-Si 7 9 19 188 +.84 1000
SiDa-BrDa-Da 6 10 19 177 +.70 100

Forms of the family constitute a fourth group of social equalizers and differentials. They differ somewhat from the forms of marriage, since monogamous, polyandrous, and polygynous families may either stand alone or be aggregated into larger extended families. The polyandrous family is too rare for statistical treatment. The influence of the polygynous family has already been shown in Theorems 2 to 5, and the isolated monogamous family will be separately considered in Theorem 14. The several types of extended familiespatrilocal, matrilocal, avunculocal, and bilocal-depend upon the prevailing rule of residence and exert the influences upon kinship nomenclature already set forth in Theorems 6 to 11. Over and above the effect of spatial proximity or segregation produced by residence rules, however, extended families introduce the new factor of social participation, since their mem hers not only live nearby but also belong to the same familial group. This should tend to enhance their similarity and to augment the effect of common residence.
However, when societies are classed according to the presence or absence of particular types of extended families, others with the same rules of residence exerting similar influences must be treated as negative cases. This necessarily results in the lowering of coefficients despite the effect of participation. To exclude all cases with similar rules of residence would not so much test the influence of extended family organization as show the contrast between its effects and those of neolocal and bilocal residence, for which after Theorems 11 and 12 no further demonstration is required. We have discovered no way of formulating and testing theorems to cope with this problem, and must therefore be content with summarizing the results of various computations. Matrilocal and avunculocal extended families have been found to be strongly associated with bifurcate merging terminology, and bilocal extended families with generation terminology, but in both cases by coefficients slightly lower in average magnitude than in the tests of the comparable residence rules (see Theorems 8 and 11). With patrilocal extended families bifurcate collateral and bifurcate merging terminology predominates, as anticipated from Theorems 9 and 10. In short, the results conform precisely to theoretical expectations, even though technical difficulties prevent the formulation and testing of the latter in the form of theorems.
Clans constitute a fifth group of social equalizers and differentials. Like the rules of residence that determine their form, they influence kinship terminology through the factor of proximity. Like extended families, they also exert an influence through participation. Furthermore, since clans involve a recognition of unilinear descent as well as of unilocal residence, they exert an effect upon terminology comparable to that which will be demonstrated below for unilinear consanguineal kin groups. All of these influences supplement and reinforce each other. Clan membership increases the similarities between lineal and collateral relatives in the same line of descent, and thus promotes merging. The inherent distinction of bifurcation is supported, since persons related to Ego through intervening relatives of opposite sex are spatially and socially segregated. Hence clans tend to produce kinship nomenclature of the bifurcate merging type.

Theorem 13: Clans, whether patrilocal, matrilocal, or avunculocal in type, tend to be associated with bifurcate merging kinship terminology.

Since patri-clans involve the complicating factor of polygyny, this theorem is first tested with respect to the influence of matrilocal and avunculocal clans. Table 29, from which wife's nieces are omitted because of paucity of data, provides substantial validation in this partial test.

Table 29
  Matrilocal or AvunculocalClans Patrilocal Clans or NoClans
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 7 9 75 118 +.10 2
MoBrWi-FaBrWi-Mo 6 6 48 67 +.16
FaSiDa-FaBrDa-Si 15 2 90 99 +.74 100
MoBrDa-MoSiDa-Si 15 0 77 99 +1.00
SiDa-BrDa-Da 9 4 84 89 +.41 2

With respect to the influence of patri-cians we encounter once more the countervailing factor of non-sororal polygyny, which has been found to favor bifurcate collateral rather than bifurcate merging terminology (see Theorem 5). When only patrilocal residence is involved, polygyny is sufficient, in conjunction with the inherent distinctions of collaterality and bifurcation, to override its influence statistically (see Theorem 10). When, however, participation in the same compromise kin group is added to proximity through patrilocal residence, the effect of polygyny is submerged, and patri-cians are found to be definitely associated with bifurcate merging terminology. Table 30, wherein Theorem 13 is subjected to a second partial test, again provides substantial confirmation.

Table 30
  PatrilocalClans Other Clans or NoClans
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 31 38 49 88 +.18 5
MoBrWi-FaBrWi-Mo 24 23 30 50 +.27 5
FaSiDa-FaBrDa-Si 41 23 62 78 +.38 20
MoBrDa-MoSiDa-Si 33 24 59 76 +.27 5
SiDa-BrDa-Da 36 23 58 68 +.30 10
WiBrDa-WiSiDa-Da 12 11 18 27 +.16 2

Neither Table 29 nor Table 30, of course, adequately reflects the influence of clans, for when types of clans are tabulated separately, societies possessing other types exerting an identical influence are necessarily counted as negative cases. This defect is corrected in Table 31, which compares clanless tribes with those possessing clans of any type. Here the confirmation of Theorem 13 is not merely substantial but conclusive.

Table 31
  Clans Present Clans Absent
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 38 47 44 80 +.33 2
MoBrWi-FaBrWi-Mo 30 29 24 44 +.31 5
FaSiDa-FaBrDa-Si 56 24 49 77 +.57 1000
MoBrDa-MoSiDa-Si 48 24 44 76 +.55 1000
SiDa-BrDa-Da 45 27 49 66 +.39 20
WiBrDa-WiSiDa-Da 13 14 18 24 +.08

Now that the influence of extended families and clans has been disposed of, it becomes possible to examine the bearing of the uncomplicated nuclear family upon kinship terminology. This is obscured when families are aggregated into larger kin groups of either the residential or the compromise type. Bifurcate collateral terminology tends to result when nuclear families are combined through plural marriages into non-sororal polygynous families (see Theorem 5) or are grouped by common residence into patrilocal extended families. Bifurcate merging tends to appear when nuclear families are linked together by sororal polygyny (see Theorem 3), or are united by common residence into matrilocal or avunculocal extended families or into clans of any type (see Theorem 13). Generation terminology tends to occur when they are aggregated into bilocal extended families. Is there any form of nomenclature which is typical of instances where the nuclear family stands in majestic isolation, as it does in our own society, uncompounded into larger kin groups?
A partial answer to the above question is provided by the prevailing rules of residence (see Theorems 6 to 12). Nevertheless, one type of kinship terminology appears to be especially characteristic of the isolated nuclear family, as we may designate an independent nuclear family in the absence of clans. This is the lineal type of nomenclature. The very isolation of the nuclear family operates as a social differential to favor separate terms for lineal and collateral relatives, and at the same time operates as a social equalizer, either through immateriality or through equivalent lack of proximity, to minimize the inherent distinction between collateral relatives through different sexes.

Theorem 14: In the absence of clans and of polygamous and extended families, the isolated nuclear family tends to be associated. with kinship terminology of the lineal type.

In Table 32, which tests this theorem, terms for wife's nieces are omitted because of the infrequency with which they exist or are reported in societies with isolated nuclear families. The theorem is validated by coefficients of association that are positive in sign, high in value, mutually consistent, and reliable in the majority of instances at the five or the one per cent level of confidence.

Table 32
  Isolated Nuclear Family Compounded Family Forms
Trios of Relatives Lineal Terms Other Terms Lineal Terms Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 7 23 15 177 +.56 20
MoBrWi-FaBrWi-Mo 4 11 14 110 +.41 2
FaSiDa-FaBrDa-Si 9 23 15 170 +.62 100
MoBrDa-MoSiDa-Si 9 20 15 162 +.70 100
SiDa-BrDa-Da 6 22 21 148 +.30 2

A sixth and extremely important group of social determinants includes the various types of consanguineal kin groups-bilateral groups such as the kindred and the deme, unilinear groups such as the lineage, the sib, and the moiety, and bilinear kin groups such as the section. Among these, anthropologists have paid least attention to bilateral kin groups. They have, for example, rarely made a special effort to discover kindreds, and they practically never report their absence. In testing theorems concerning the kindred, therefore, it becomes necessary to compare societies in which it is reported with those in which it is not reported instead of the preferable comparison between those in which it is present and those in which it is absent. This doubtless reduces the magnitude and reliability of coefficients, since presumably a number of tribes in which the kindred is present but unreported are included among the negative cases.
The kindred unites in a social group all of Ego's near relatives regardless of collaterality or bifurcation. It should therefore be expected to operate as a social equalizer to override these inherent distinctions and produce kinship terminology of the generation type.

Theorem 15: Bilateral kindreds tend to be associated with kinship terminology of the generation type.

This theorem is validated in Table 33 by uniformly high and consistent positive coefficients of association with a degree of reliability that is quite satisfactory under the circumstances.

Table 33
  Kindreds Present Kindreds Unreported
Trios of Relatives Generation Terms Other Terms Generation Terms Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 7 23 15 177 +.56 20
MoBrWi-FaBrWi-Mo 4 11 14 110 +.41 2
FaSiDa-FaBrDa-Si 9 23 15 170 +.62 100
MoBrDa-MoSiDa-Si 9 20 15 162 +.70 100
SiDa-BrDa-Da 6 22 21 148 +.30 2
WiBrDa-WiSiDa-Da 6 22 21 148 +.30 2

The influence of demes upon kinship terminology presents a special problem, for it appears to be significant only within Ego's own generation. When societies with endogamous demes are compared with other bilateral societies they reveal a practically identical distribution of generation, lineal, bifurcate merging, and bifurcate collateral terms for all relatives of the first ascending and first descending generations. With respect to cousins, however, they show a markedly stronger preponderance of generation terminology, despite the parallel influence of kindreds in many of the bilateral societies with which they are compared. This tendency shows up even more strikingly when they are compared with all other societies, as is done in Table 34. But here, too, no effect is observable on either the first ascending or the first descending generation; tests for the relevant kinsmen reveal low coefficients with either plus or minus signs, indicating that association is purely random.
Why should endogamous demes exert such a marked influence on kinship terminology within Ego's own generation but none whatsoever on other generations, especially since every other type of consanguineal kin group can be statistically shown to affect kinship terms on all generations in similar manner and to approximately the same degree? While we cannot answer this question definitely, we offer a tentative suggestion in the hope that others may be able either to verify it or to propose an acceptable alternative theory.
With the exception of the deme all consanguineal kin groups, whether unilinear or bilateral, divide the local community into members and non-members. Participation and non-participation provide a basis for terminological distinctions on all generations. The deme, however, is coextensive with the community, and hence cannot produce any differentation in kinship terms among its members. Under these conditions family structure becomes peculiarly important. Between generations, the strongly functional parent-child relationships stand out in especially strong relief, with the result that avuncular and nepotic relatives tend to be denoted by distinctive kinship terms. In the societies of our sample which possess endogamous demes, as a matter of fact, the kinship terms for aunts, nieces, uncles' wives, and wife's nieces accord with theoretical expectations on the basis of other factors.
Within the same generation, however, the absence of segmentation in the community removes one of the commonest grounds for differentiating siblings from corsins, and deme organization itself acts as a definite social equalizer to favor the extension of kinship terms from the former to the latter. Moreover, persons of the same generation tend nearly everywhere to be more closely associated in economic and social activities than are persons of different generations, so that family distinctions are more readily overridden than in the first ascending or first descending generation.
The only factual support for this highly tentative hypothesis that we have been able to discover comes from the kinship terminology for parallel cousins in societies with unilinear descent.
In patrilineal societies FaBrDa but not MoSiDa is necessarily a member of Ego's kin group, and in matrilineal societies MoSiDa but not FaBrDa. Yet the overwhelming majority of unilinear societies also extend sibling terms to MoSiDa in the former and to FaBrDa in the latter. Theorem 1 has offered a partial explanation, but an examination of Table 15 will show that it accounts for by no means all of the cases. Omitting instances of double descent, which necessarily affiliates both parallel cousins with Ego, our sample reveals that MoSiDa in societies with exogamous patrilineal kin groups,or FaBrDa in matrilineal tribes, is called by the same term as sister in 100 societies and by another term in 29. Of the latter, 24 have general non-sororal polygyny, which accounts for the differentiation (see Theorem 4). These facts suggest a widespread tendency, not completely explained by the principle of analogy, for sibling terms to be extended to cousins in the absence of specific social differentials. Since, in the presence of demes, this tendency is reinforced by a definite social equalizer, the preponderance of generation terms for all cousins in deme societies may conceivably be accounted for.
Inasmuch as this hypothesis involves assumptions not specified in the formulation of Postulate 1, it must be phrased as a proposition rather than as a theorem.

Proposition 16: In societies with endogamous demes, sibling terms tend to be extended to both cross and parallel cousins.

Table 34, which tests this proposition, omits nine societies for which deme organization seems fairly likely, though not definitely established. Since the coefficients of association are not only positive and consistent but unusually high and reliable, the proposition must be regarded as decisively validated.

Table 34
  Endo-demes Present Endo-demes Absent
Trios of Relatives Generation Terms Other Terms Generation Terms Other Terms Statistical Indices
  Q X2
FaSiDa-FaBrDa-Si 9 4 47 164 +.77 1000
MoBrDa-MoSiDa-Si 9 5 48 151 +.71 100

Patri-demes and matri-demes, even though the societies for which they are reported are too few for reliable statistical analysis, raise no special problems. In distribution of types of kinship terminology, societies with patri-demes are almost exactly intermediate between those with endogamous demes and those with patrilocal clan-communities, as is consistent with the hypothesis that the three types of organization constitute a normal sequence of development. From deme through patri-deme to patri-clan, lineal terminology decreases steadily for all relatives and generation terms for cousins, whereas bifurcate merging terms increase markedly for all relatives, bifurcate collateral terminology remaining relatively constant.
Unilinear kin groups of the consanguineal type affect the terminological classification of kinsmen through similarities and dissimilarities produced by social participation. Proximity is not a factor since their members do not live together except for those who may be aggregated by a rule of residence. The alignment of relatives varies somewhat with the rule of descent but not with the size of the kin group. Lineages, sibs, and phratries with the same rule of descent bring into association the same categories of relatives, and can therefore be treated together. Moieties have a like effect, but create certain additional similarities which will require special theorems to demonstrate. The influences of patrilineal and matrilineal descent are strictly parallel. They sometimes coincide, as when both distinguish FaSi from MoSi. Sometimes they are independent; thus only patrilineal descent directly associates BrDa and Da, and only matrilineal descent WiSiDa and Da. Never, however, do they run counter to each other. Even where they appear independent, they commonly produce similar results, through analogy or otherwise, as in the case of parallel cousins. In this indirect fashion nearly every influence exerted on kinship terminology by one unilinear rule of descent tends also to be exerted by the opposite rule. For this reason it becomes possible in most instances to treat matrilineal and patrilineal kin groups together and to differentiate them only from cases of bilateral descent.
If unilinear kin groups are present in a society, but do not regulate marriage or are endogamous, they fail to produce regular alignments of kinsmen, and their effects upon kinship terminology are minimized. For example, it is customary among many Islamic peoples 77, including the Kababish and Kurds of our sample, for a man to marry his FaBrDa, who is a member of his own patri-sib. To the children of such a marriage the relatives through both parents belong to the same sib, and indeed are often the same persons. Inherent distinctions such as bifurcation lose their meaning in such a case, and social equalizers and differentials do not produce the expected alignments of relatives. Similar confusion results when violations of exogamy are common. For this reason, in the formulation and testing of theorems on unilinear descent, a society will be classed as unilinear only if its kin groups are exogamous, or reveal a definite tendency toward exogamy. Unilinear societies all of whose kin groups are endogamous or non-exogamous will be classed with bilateral societies. It should perhaps be pointed out, however, that this rarely makes a difference of more than a few points in the magnitude of coefficients.
Membership in the same exogamous unilinear kin group operates as a social equalizer to override the inherent distinction of collaterally and produce the phenomenon of merging. Since the principle of analogy normally extends the merging tendency to parallel relatives outside as well as within the prevailing line of descent, it is possible to formulate a theorem which applies to both unilinear rules of descent at the same time.

Theorem 17: In the presence of exogamous matrilineal or patrilineal lineages, sibs, phratries, or moieties, terms for lineal relatives tend to be extended, within the same sex and generation, to collateral kinsmen who would be affiliated with them under either unilinear rule of descent.

In Table 35, which tests this theorem, the usual comparison between FaBrWi and Mo is omitted since these two kinsmen are not necessarily affiliated under either matrilineal or patrilineal descent except in the presence of moieties. The low values and reliability of the coefficients for parallel cousins and siblings are due mainly to the fact that bilateral kin groups normally affect them in much the same way as do unilinear kin groups (see Theorem 15 and Proposition 16). What provides conclusive confirmation of Theorem 17 is the occurrence of high, positive, consistent, and maximally reliable coefficients of association for the comparisons of MoSi with Mo and of BrDa with Da.

Table 35
  Exogamous Unilinear Kin Groups Present Present Exogamous Unilinear KinGroups Absent
Pairs of Relatives Same Terms Different Terms Same Terms Different Terms Statistical Indices
  Q X2
MoSi-Mo 106 53 31 51 +.51 1000
FaSiBrDa-Si 125 29 64 22 +.19 2
MoSiDa-Si 114 31 62 23 +.15 2
BrDa-Da 107 37 34 43 +.57 1000
WiSiDa-Da 38 25 15 15 +.21 2

Membership in different unilinear kin groups operates as a social differential to support the inherent distinction of bifurcation and to favor differentiation of kinship nomenclature between kin-types in different lines of descent.

Theorem 18: In the presence of exogamous matrilineal or patrilineal lineages, sibs, phratries, or moieties, separate kinship terms tend to be applied to comparable relatives of the same generation who are linked to Ego by connecting relatives of different sex.

This theorem is decisively validated in Table 36 by positive and consistent coefficients of association that are unusually high both in numerical value and in reliability.

Table 36
  Exogamous Unilinear Kin Groups Present Present Exogamous Unilinear KinGroups Absent
Pairs of Relatives Different Terms Same Terms Different Terms Same Terms Statistical Indices
  Q X2
FaSi-MoSi 122 35 52 29 +.35 20
MoBrWi-FaBrWi 83 26 25 19 +.48 20
FaSiDa-FaBrDa 124 24 25 60 +.85 1000
MoBrDa-MoSiDa 117 22 24 61 +.86 1000
SiDa-BrDa 116 18 49 28 +.57 1000
WiBrDa-WiSiDa 42 10 14 13 +.59 20

As a corollary of Theorems 17 and 18, bifurcate merging terminology is to be expected in the presence of exogamous consanguineal kin groups of either unilinear type.

Theorem 19: Exogamous matrilineal or patrilineal lineages, sibs. phratries, and/or moieties tend to be associated with kinship terminology of the bifurcate merging type.

This theorem is tested for matrilineal kin groups in Table 37, for patrilineal kin groups in Table 38, and for both combined in Table 39. It is strongly confirmed in the first partial test and rather less strongly in the second, owing to the contrary influence of non-sororal polygyny which, as previously noted, is highly associated with patrilineal descent. In both partial tests, of course, cases of the opposite unilinear rule are treated as negative cases, with a consequent probable reduction in the magnitude and reliability of coefficients. Table 39, which corrects this defect, is validated by positive, high, and consistent coefficients of association which in four of the six cases are found reliable at the maximum level of confidence (one tenth of one per cent).

Table 37
  Exogamous Matrilinear Kin Groups Present Exogamous Matrilinear Kin Groups Absent
Pairs of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 37 28 58 115 +.46 100
MoBrWi-FaBrWi-Mo 24 15 39 74 +.51 100
FaSiDa-FaBrDa-Si 49 12 76 97 +.58 1000
MoBrDa-MoSiDa-Si 49 10 64 98 +.76 1000
SiDa-BrDa-Da 37 15 76 82 +.45 100
WiBrDa-WiSiDa-Da 12 6 22 37 +.54 10

The hypothesis incorporated in Theorem 19 has previously been offered in slightly variant forms by both Lowie 78 and Kroeber 79. Lowie ascribes bifurcate merging to exogamy, Kroeber to unilinear descent per se. Lowie's hypothesis differs from that of the present author only by the inclusion, along with exogamous unilinear kin groups, of patri-demes and matri-demes, which are also characterized by exogamy. Kroeber's version, on the other hand, differs by the inclusion of non-exogamous unilinear kin groups. In a previous article 80, based upon a preliminary analysis of 221 of the 250 societies covered in this work, the present author reached the conclusion that Kroeber's version of the hypothesis enjoyed a slight statistical advantage over Lowie's. With the additional coverage of the present volume and the correction of a few errors, this modest advantage has evaporated. As Table 40 indicates, the three forms of expression of the hypothesis emerge from a new statistical analysis with almost identical support. The figures given are the totals for the six trios of relatives covered in Tables 37, 38, and 39.

Table 38
  Exogamous Patrilineal Kin Groups Present Exogamous Patrilineal Kin Groups Absent
Pairs of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 47 45 48 88 +.22 5
MoBrWi-FaBrWi-Mo 38 39 24 49 +.33 10
FaSiDa-FaBrDa-Si 66 33 59 76 +.44 100
MoBrDa-MoSiDa-Si 54 35 59 73 +.31 20
SiDa-BrDa-Da 56 34 57 64 +.30 20
WiBrDa-WiSiDa-Da 20 19 13 24 +.32 2


Table 39
  Exogamous Patrilineal Kin Groups Present Exogamous Patrilineal Kin Groups Absent
Pairs of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 47 45 48 88 +.22 5
MoBrWi-FaBrWi-Mo 38 39 24 49 +.33 10
FaSiDa-FaBrDa-Si 66 33 59 76 +.44 100
MoBrDa-MoSiDa-Si 54 35 59 73 +.31 20
SiDa-BrDa-Da 56 34 57 64 +.30 20
WiBrDa-WiSiDa-Da 20 19 13 24 +.32 2

Cross-cousin terminology of the Crow type, in which generation levels are overridden by calling FaSiSo by the same term as FaBr (or Fa) and FaSiDa by the same term as FaSi, while MoBrSo is terminologically classed with BrSo (or So) and MoBrDa with BrDa (or Da), has frequently been noted to be particularly characteristic of matrilineal societies 81. This is accounted for by Postulate 1 since, under matrilineal descent, FaSi and her children are members of the same consanguineal kin group; the fact of such participation presumably operates as a social equalizer to overcome the inherent distinction of generation and produce an extension of kinship terminology from the first ascending generation to lower generations within the same kin group. Once this happens the reciprocal terms for BrSo and BrDa are extended to the other cross-cousins, the children of MoBr, both through analogy and by virtue of the fact that such relatives are all children of the men of Ego's own matri-lineage or matri-sib. Under double descent, the tendency to produce Crow terminology for cross-cousins is neutralized by the opposite influence of patrilineal descent (see Theorem 21). The tendency is also minimized if the matrilineal kin groups are not characterized by exogamy.

Table 40
  Unilinear Descent Bilateral Descent
Theorist Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
Kroeber 447 314 94 277 +.615 1000
Lowie 458 332 84 257 +.617 1000
Murdock 435 288 108 301 +.616 1000

Theorem 20: In the presence of exogamous matri-lineages, matrisibs, or matri-moieties, unless exogamous patrilineal kin groups are also present, kinship terms for FaSi tend to be extended to FaSiDa, and those for BrDa to MoBrDa.

This theorem is conclusively validated in Table 41 by similar high positive coefficients of association with maximal reliability. Parallel reasoning accounts for cross-cousin terminology of the Omaha type, in which the children of FaSi are classed with SiSo and SiDa, and the children of MoBr with the siblings of one's mother. Omaha nomenclature is normally found in societies with patrilineal descent. Under this rule, MoBr and MoBrSo are members of the same kin group, as are MoSi and MoBrDa, and the similarity resulting from their social participation favors the extension of kinship terms across generation lines. The reciprocal terms for SiSo and SiDa tend to be extended to the opposite cross-cousins, the children of FaSi, partly by analogy and partly because paternal cross-cousins and sororal nieces and nephews are all children of the women of Ego's own patri-sib or patri-lineage.

Table 41
  Exclusive Matrilineal Descent with Endogamy Patrilineal, Double, or Bilateral Descent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
FaSiDa-FaSi 21 29 8 181 +.88 1000
MoBrDa-BrDa 15 33 7 181 +.84 1000

Theorem 21: In the presence of exogamous patri-lineages, patrisibs, or patri-moieties, unless exogamous matrilineal kin groups are also present, kinship terms for MoSi tend to be extended to MoBrDa, and those for SiDa to FaSilla.

This theorem is decisively substantiated by the data in Table 42.

Table 42
  Exclusive Patrilineal Descent with Endogamy Matrilineal, Double, or Bilateral Descent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
MoBrDa-MoSi 21 68 3 144 +.87 1000
FaSiDa-SiDa 24 66 6 143 +.79 1000

Most of the cases in Tables 41 and 42 of the occurrence of Crow or Omaha terminology with other than the expected type of social organization are accountable on other grounds. Thus the Pawnee and Siriono have Crow terms with matrilocal residence; the Manus, Pentecost, and Yuchi have Crow terms with double descent; the Fox have Omaha terms with non-exogamous patri-sibs, and the Takelma with patrilocal residence. The only genuinely negative cases are the Bachama, Koranko, and Seniang, who have Crow terms with patrilineal descent and patrilocal residence.
Only about one third of the exclusively matrilineal societies in our sample are characterized by Crow terminology for cross-cousins, and only about one fourth of the patrilineal societies have Omaha terminology. This reflects the fact, already noted, that the inherent distinction of generation, which must be overridden to produce them, is the most resistant, or has the strongest relative efficacy, of all such differentials. To overcome it presumably requires both time and the full elaboration of unilinear institutions. This accords with the hypothesis advanced by White 82 that Crow or Omaha terms tend to appear only when a system of unilinear kin groups is fully developed “and comes to exert its influence more and more upon the social life of the tribe.” Evidence to be presented in Chapter 8 will lend strong confirmation to this theory.
The present author suspects that the amitate and the avunculate, the special relationships with a paternal aunt and a maternal uncle respectively, may prove to be important contributory factors in the development of Crow terminology in the one case and of Omaha terminology in the other. He has done personal field work in two matrilineal societies with Crow terminology, the Haida and the Trukese, and in both of them the relation with a paternal aunt is exceedingly important. Among the Haida 83, for example, the father's sister has special and highly important functions to perform at every crisis in an individual's life—at birth, at puberty, at marriage, at potlatches, in sickness, and at the funeral ceremony. If she is not alive to perform them, her daughter acts in her stead; if no FaSiDa is still living, a FaSiDaDa plays the vital roles. In other words, the functions are inherited in the female line, as is natural under the prevailing social conditions. The functional equivalence of these several relatives, irrespective of generation, presumably reinforces the similarities that flow from participation in the same kin group and thus accentuates the tendency to apply the same kinship term to all of them.
The role of exogamous moieties as determinants of kinship nomenclature has been stressed by a number of authorities, particularly by Rivers 84. Theoretical analysis and careful scrutiny of the data, however, indicate that this emphasis is largely unwarranted. Moieties appear to produce precisely the same effects upon terminological classification as do other unilinear consanguineal kin groups. Most of their allegedly unique consequences, such as the application of identical terms to FaBrDa and MoSiDa, can be and are produced in the absence of moieties by unilinear descent coupled with extension by analogy (see Theorem 1). Moieties merely exert the same influences more directly and more sharply. Their presence enhances the tendency to merge lineal and collateral relatives, strongly accentuates the inherent distinction of bifurcation, and increases the incidence of bifurcate merging terminology, all of which have been found characteristic of unilinear groups in general (see Theorems 17, 18, and 19). These influences can be summed up in a single theorem.

Theorem 22: Exogamous moieties tend to be associated with kinship terminology of the bifurcate merging type.
This theorem is conclusively validated in Table 43.

Table 43
  Exogamous Moieties Present Exogamous Moieties Absent
Pairs of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 15 8 79 136 +.53 20
MoBrWi-FaBrWi-Mo 14 5 50 83 +.63 100
FaSiDa-FaBrDa-Si 20 3 105 105 +.74 100
MoBrDa-MoSiDa-Si 21 2 92 109 +.85 1000
SiDa-BrDa-Da 16 5 97 92 +.50 10
WiBrDa-WiSiDa-Da 8 4 26 43 +.54 5

Exogamous moieties also have the effect of aggregating in the same kin group certain affinal and consanguineal relatives who are not ordinarily brought together by sibs or lineages. Among these are MoBrWi, WiMo, and FaSi; FaBrWi, FaWi, and MoSi; WiBrDa, SoWi, and SiDa; WiSiDa, Willa, and BrDa. In all of these cases, cross-cousin marriage exerts a parallel influence, and does so more strongly (see Theorems 26 and 27), so that the effect of moieties is revealed by statistical coefficients of only modest magnitude and reliability. Among tertiary relatives, only the terminological equating of parallel cousins and WiBrWi appears to reflect primarily the influence of moieties.

Theorem 23: In the presence of exogamous moieties, kinship terms for WiBrWi tend to be the same as those for female parallel cousins.

This theorem is substantially validated in Table 44.

Table 44
  Exogamous Moieties Present Exogamous Moieties Absent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
FaBrDa-WiBrWi 7 4 10 44 +.77 100
MoSiDa-WiBrWi 7 5 10 41 +.70 20

The only really peculiar effects of moieties upon kinship structure appear in connection with double descent. The concurrence of matrilineal moieties and patrilineal kin groups gives rise, as was seen in Chapter 3, to so-called “marriage classes,” including such bilinear kin groups as sections and subsections. These groups operate as social equalizers favoring the extension and differentiation of kinship terms, in a manner analogous to lineages, sibs, and moieties.

Theorem 24: In the presence of bilinear kin groups, kinship terms tend to be extended to all relatives who belong to the same section or subsection.

Full validation of this theorem would carry us far afield into a discussion of social organization in aboriginal Australia, to which continent, except for a limited region in Melanesia, the conditions in question appear to be confined. Interested readers may turn to the summary of the Australian data by Radcliffe-Brown 85 for the evidence to substantiate the theorem. They will find that within a section, or subsection, the extension of kinship terms is carried to such an extreme that even such highly resistant inherent distinctions as sex and generation are frequently overridden.
For present purposes the demonstration will have to rest on a single tabulation, that of terms for Mo and SoWi in Table 45. Different as these two relatives are, not only in generation but in nearly every aspect of Ego's functional relationships with them, they are nevertheless called by the same kinship term in three of the five societies in our sample which meet the conditions of the theorem and for which information is available, namely, the Arunta and Murngin of Australia and the Ranon of the New Hebrides. The explanation lies in the single common similarity that the two relatives belong to the same section; they are matrilineally but not patrilineally akin to Ego. The Kariera and Pentecost have sections—though dubious in the latter case—and yet differentiate SoWi from Mo, while the Iatmul of New Guinea and the Tallensi of West Africa equate these relatives in terminology but are not reported to have double descent or bilinear kin groups. The data in Table 45 support Tileorem 24 by a coefficient of association of extraordinary magnitude and reliability.

Table 45
  Sections Present Sections Absent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
SoWi-Mo 3 2 2 196 +.90 1000

A seventh group of social equalizers consists of a variety of special rules governing marriage. Consideration has already been given to the influences exerted on kinship structure by monogamy (see Theorem 14), sororal polygyny (see Theorems 2 and 5), non-sororal polygyny (see Theorems 4 and 5), and exogamy (see Theorem 19). There remain to be considered the customs of sister exchange and preferential mating.
Where the regular method of obtaining a wife is for two men to exchange women, each taking a sister of the other as his spouse, coincidence occurs in certain kin-types. As Rivers 86 has pointed out, such relatives as MoBrWi and FaSi, WiBrWi and Si, and WiBrDa and Silla become the same persons under a rule of sister exchange, and are consequently likely to be designated by identical terms.

Theorem 25: When the normal mode of marriage is by the exchange of sisters, the same kinship terms tend to be applied to MoBrWi and FaSi, to WiBrWi and Si, and to WiBrDa and Silla.

This theorem is substantially validated by the data in Table 46.

Table 46
  Sections Present Sections Absent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
MoBrWi-FaSi 8 2 57 99 +.75 20
WiBrWi-Si 6 1 13 45 +.91 100
WiBrDa-SiDa 8 2 29 56 +.77 20

Rules of preferential mating fall into two groups-those governing primary or first marriages and those governing secondary or subsequent marriages. Of the former type, by far the most common is cross-cousin marriage, i.e., preferential mating with a FaSiDa or a MoBrDa. Cross-cousin marriage is called symmetrical when either of these relatives is a preferred spouse, asymmetrical when only one is eligible or preferred. When a large proportion of marriages within a society are between cross-cousins, coincidence becomes frequent with respect to certain kin-types, with consequent extension of kinship terms 87. When marriage with a FaSiDa is preferred, a man's paternal aunt becomes his mother-in-law, her daughter is either his wife, his wife's sister, or his brother's wife, and his sororal niece is his daughter-in-law, and kinship terminology is likely to conform to these equivalences.

Theorem 26: Under a rule of preferential marriage with a FaSilla, the same kinship terms tend to be applied to FaSi and WiMo, to FaSiDa and Wi and/or WiSi andjor BrWi, and to SiDa and SoWi.

The data in Table 47 confirm this theorem by high and consistent positive coefficients of association which attain an exceptional level of reliability.

Table 47
  FaSiDa a Preferred Spouse FaSiDa Not Preferred
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
FaSi-WiMo 14 20 13 139 +.76 1000
FaSiDa-Wi 6 28 3 159 +.84 1000
FaSiDa-WiSi 10 18 5 137 +.88 1000
FaSiDa-BrWi 9 17 5 138 +.81 1000
SiDa-SoWi 8 18 9 124 +.72 100

Parallel extensions occur when the preferred union is with a MoBrDa. Under this rule a man's maternal uncle's wife becomes his mother-in-law, her daughter becomes his wife or his sister-in-law, and his wife's brother's daughter becomes his daughter-in-law.

Theorem 27: Under a rule of preferential marriage with a MoBrDa, the same kinship terms tend to be applied to MoBrWi and WiMo, to MoBrDa and Wi andjor WiSi and/or BrWi, and to WiBrDa and SoWi.
This theorem is conclusively validated in Table 48. Other preferential primary marriages have comparable effects upon kinship nomenclature. Thus the term for WiMo is extended to FaSiDa and MoBrDa among the Lesu, where the preferred union is with the daughter of a female cross-cousin, and to MoMoBrDa and MoFaSiDa among the Arunta, where the favored mates are second cousins who are the daughters of these women. Such cases, however, are too sporadic for reliable statistical testing.

Table 48
  MoBrDa a Preferred Spouse MoBrDa Not Preferred
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
MoBrWi-WiMo 21 17 10 88 +.83 1000
MoBrDa-Wi 7 39 3 145 +.79 100
MoBrDa-BrWi 14 23 7 127 +.83 1000
WiBrDa-SoWi 12 22 5 130 +.87 1000
SiDa-SoWi 8 12 5 50 +.74 100

Preferential rules governing secondary marriages, such as the levirate and sororate, have been advanced by many authorities as determinants of kinship terminology. Certain grounds for skepticism regarding such theories were presented earlier in the present chapter. Moreover, it will be shown in Chapter 9 that secondary marriages are largely determined by the forms of social organization, including kinship structure, instead of the causal relationship being the reverse 88. Since theories asserting the influence of preferential secondary marriages are subject to these objections, and are not derivable from Postulate 1, they cannot be presented as theorems. They may, however, be formulated as propositions in order to subject them to test.
The levirate, it is argued, by bringing about the substitution of BrWi for Wi, creates an equivalence between such relatives as FaBrWi and Mo, FaBrDa and Si, BrWi and Wi, and BrDa and Da, and thus favors the use of the same classificatory terms for the relatives of each such pair of kin-types. While the argument is logical enough, the necessary infrequency of secondary as compared with primary marriages greatly reduces the probability that they can operate as an effective social equalizer in other than exceptional instances. This is borne out by the tests of Proposition 28.

Proposition 28: In the presence of the levirate, kinship terms tend to be extended from .Mo to FaBrWi, from Si to FaBrDa, from Wi to BrWi, and from Da to BrDa.

In Table 49, which tests this proposition, the six societies with bilinear kin groups are omitted, since sections or subsections make the levirate practically inevitable.

Table 49
  Levirate Present Levirate Absent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
FaBrWi-Mo 46 32 20 26 +.30 5
FaBrDa-Si 84 23 49 16 +.09
BrWi-Wi 20 79 6 51 +.37 5
BrDa-Da 53 38 42 18 +.21 2

The sororate, similarly, is alleged to effect a terminological equivalence between MoSi and Mo, MoSiDa and Si, WiSi and Wi, and WiSiDa and Da.

Proposition 29: In the presence of the sororate, kinship terms tend to be extended from Mo to MoSi, from Si to MoSiDa, from Wi to WiSi, and from Da to WiSiDa.

Table 50, which tests this proposition, omits two groups of tribes those with bilinear kin groups, since these make the sororate almost automatic, and those with preferential sororal polygyny. Theorem 2 has already indicated that sororal polygyny has the effects upon kinship terminology here attributed to the sororate, so that the inclusion of cases of the former would give a fictitious impression of the influence of the latter.

Table 50
  Sororate Present Sororate Absent
Pairs of Relatives Same Term Different Terms Same Term Different Terms Statistical Indices
  Q X2
MoSi-Mo 38 35 35 29 +.04
MoSiDa-Si 53 14 48 16 +.12
WiSi-Wi 9 59 5 54 +.24
WiSiDa-Da 18 16 13 10 +.07

Analysis of Tables 49 and 50 indicates that Propositions 28 and 29 lack statistical support. Coefficients of association are low, are inconsistent with each other, and in three of the eight computations are negative in sign. This is, incidentally, the first time that negative coefficients have been encountered, except for one instance in an uncorrected test. Even more significant is the low level of reliability. In no instances does the x2 index of reliability attain the ten per cent level of confidence, and of the only three coefficients that attain even the fifty per cent level one is negative in sign. Both propositions, therefore, must be considered invalidated. If either the levirate or the sororate has any genuine influence on kinship terminology, it is comparatively slight and is probably confined to very special circumstances.
The assumption that the levirate and sororate produce merging, invalidated above, led Sapir 89 to the further hypothesis that these rules of preferential secondary marriage may be responsible for the phenomenon of bifurcate merging terminology. Though its basis is now undermined, the theory may be formulated as a proposition and tested.

Proposition 30: The levirate and sororate tend to be associated with kinship terminology of the bifurcate merging type.

This proposition is subjected to test in Tables 51 and 52.

Table 51
  Levirate Present Levirate Absent
Pairs of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 39 64 23 40 +.03
MoBrWi-FaBrWi-Mo 38 36 8 34 +.64 100
FaSiDa-FaBrDa-Si 52 51 30 32 +.04
MoBrDa-MoSiDa-Si 49 49 25 34 +.15 2
SiDa-BrDa-Da 50 39 31 28 +.05
WiBrDa-WiSiDa-Da 16 26 8 12 +.04

The preponderance of low values among the coefficients of association in Tables 51 and 52, the occurrence of one zero coefficient and one with negative sign, and the very low indices of reliability concur in indicating that Proposition 30 is invalid. A possible individual exception may be noted in the case of the wives of uncles, for whom bifurcate merging terminology is fairly strongly associated with the levirate. On the whole, however, it appears clearly unsafe to invoke the levirate or the sororate in explanation of kinship phenomena.
Theories which ascribe kinship terminology to other preferential secondary marriages deserve only passing notice, for they are of even more dubious validity. Although Crow terminology for crosscousins, for example, has been attributed to the influence of preferential mating with MoBrWi, our data show that such preferred marriages occur with every major type of cross-cousin terminology, and are no more common with Crow than with Iroquois nomenclature. Preferential secondary unions with WiBrDa, which have similarly been regarded as the source of Omaha terminology, are actually found in only a minority of the Omaha societies of our sample, and in a substantial number such marriages are expressly forbidden. The conclusion is therefore inescapable that preferred secondary marriages of whatsoever type are not to be included among the significant determinants of kinship terminology.
The demonstration of Postulate 1 is now complete. From it have been derived 26 separate theorems, which have been tested in 155 different computations. Of these, 35 are preliminary or uncorrected. Analysis of the coefficients of association and the indices of reliability for the 120 final computations indicates that the postulate is decisively validated.
The fact that all 120 coefficients are positive in sign is of the utmost significance. It fulfills to the maximum the rigorous requirement of the postulational method that all theorems, without exception, be found to accord with the facts. It indicates consistency among the several tests of the same theorem, an important criterion of reliability. Even when the specific indices of reliability are low for each separate test of a theorem, the fact that they all yield coefficients of positive sign enormously decreases the probability that the numerical distributions can be due to chance alone.

Table 52
  Sororate Present Sororate Absent
Pairs of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 31 48 22 42 +.16 2
MoBrWi-FaBrWi-Mo 22 51 10 32 +.39 5
FaSiDa-FaBrDa-Si 38 36 31 31 +.03
MoBrDa-MoSiDa-Si 36 36 28 33 +.08
SiDa-BrDa-Da 35 35 28 28 +.00
WiBrDa-WiSiDa-Da 13 18 4 13 +.40 2

The magnitude of the coefficients of association themselves is likewise significant. Five sixths of the total of 120 have numerical values higher than +.30. The mean is + -54; the median coefficient, +.55. The distribution, which approximates a normal frequency curve, is as follows:

  • 3 coefficients in the range from + .91 to +1.00
  • 13 coefficients in the range from + .81 to + .90
  • 16 coefficients in the range from + .71 to + .80
  • 16 coefficients in the range from + .61 to+ .70
  • 20 coefficients in the range from + .51 to + .60
  • 19 coefficients in the range from + .41 to + .50
  • 13 coefficients in the range from + .31 to + .40
  • 11 coefficients in the range from + .21 to + .30
  • 5 coefficients in the range from + .11 to + .20
  • 4 coefficients in the range from + .01 to + .10
  • 0 coefficients in the range from —1.00 to .00

Most significant of all, probably, is the very high level of reliability revealed by the x2 indices. Only for Theorems 2, 3, and 4 do these fall short of providing, by themselves, genuinely substantial confirmation. Since the x2 index of reliability is a function of the size of the sample, as well as of its distribution, low indices are not to be regarded as significant when the number of cases is small. The distribution of the 119 indices of reliability—omitting the one inappropriate case where Q is + 1.00—according to their values in relation to the number of cases is shown in Table 53.

Table 53
Level of Reliability More than 200 Case 100 to 200 Cases Less than 100 Cases Total
1000 (.001) 25 12 1 38
100 (.01) 8 5 3 16
20 (.05) 10 5 3 18
10 (.10) 3 1 2 6
2 (.50) 10 4 6 20
Less than 2 2 5 5 12

Although the above results are perhaps unprecedented in social science, they by no means do justice to the actual possibilities. It must be remembered that multiple factors are operative in every instance, but that in most of our theorems we have isolated only a single factor for analysis. If several factors are taken into consideration at the same time, the magnitude of the coefficients rises appreciably, and usually also their reliability. This can be demonstrated for the social system in our own society. Two of its elements have been individually tested-neolocal residence under Theorem 12 and the isolated nuclear family under Theorem 14. If we combine these two factors and add two other characteristic features of our own social structure—strict monogamy and the absence of exogamous unilinear kin groups—we arrive at the results shown in Table 54. Comparison with Tables 28 and 32 will reveal how the combination of several factors has raised both the values and the reliability of the coefficients. Similar results are obtainable from an exceedingly large number of similar combinations which it would be tedious to detail here.

Table 54
  Societies Having Monogamy, Isolated Nuclear Families, Neolocal Residence, and No Exogamous Lineages or Sibs Societies Having Other Combinations of Elements of Social Structure
Trios of Relatives Lineal Terms Other Terms Lineal Terms Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 5 3 16 196 +.91 1000
MoBrWi-FaBrWi-Mo 3 1 22 107 +.90 100
FaSiDa-FaBrDa-Si 5 3 19 188 +.92 1000
MoBrDa-MoSiDa-Si 5 2 19 178 +.92 1000
SiDa-BrDa-Da 5 3 20 165 +.86 1000

Postulate 1 having been validated, it will be appropriate to consider briefly the relative influence of the various social factors with which types of kinship terms have been found to be significantly associated. The relative efficacy of the several groups of social equalizers and differentials cannot be deduced from the postulate nor inferred directly from the validation of the theorems, but must be independently determined. Neither the values nor the reliability of the coefficients in the tables can be taken as a direct measure of the efficacy of the factor tested unless its independence of other factors is established. Very few of the factors operate as independent variables. Patri-clans, for example, are always associated with both patrilocal residence and patrilineal descent, which exert an identical influence. When they are found to coexist with particular kinship phenomena, this demonstration of itself gives no indication of how much of the indicated effect is to be attributed to the residence rule, to the rule of descent, or to the factor of clan participation itself. Moreover, patri-clans are strongly associated with non-sororal polygyny, which exerts an opposite influence, yet tests of individual theorems cannot indicate the extent to which this counteracts specifically either the residence, descent, or participational factors. Only when one factor can be demonstrated to be associated with all others in a purely random manner—a condition even remotely approached only in the case of sororal polygyny—can the magnitude and reliability of coefficients be taken as a rough indication of its efficacy.
A complete analysis of the relative efficacy of social equalizers and differentials would require more space than is available, and is probably not necessary. We shall therefore confine ourselves to a consideration of the relative weight to be assigned to three groups of factors which have been found particularly influential—forms of marriage, rules of residence, and descent or participation in consanguineal kin groups.
That the relative efficacy of marriage forms is lower than that of descent or kin group affiliation is suggested by the markedly higher values of coefficients of association and indices of reliability in tests of the theorems concerned with the latter (Theorems 15 to 24) than in those dealing with the former (Theorems 2 to 5). Since patrilineal descent and non-sororal polygyny tend to produce different types of kinship terminology, i.e., bifurcate merging and bifurcate collateral respectively, the distribution of these types when both factors are present should shed light on their relative efficacy. A special tabulation of the incidence of such terms for aunts and nieces, omitting cases of double descent, reveals that bifurcate merging occurs approximately 50 per cent more often, and this is confirmed by the results in Table 38 under Theorem 19. Since sororal polygyny exerts an influence on kinship terms identical with that of unilinear descent, the cases in which only one of the two factors is present provide another opportunity for comparison. They are analyzed in Table 55. Here bifurcate merging terminology is found associated with sororal polygyny in slightly less than half of all instances of bilateral descent, whereas it is associated with unilinear descent in slightly more than half of' all instances where sororal polygyny is absent. Only in the case of cousins, to be sure, are the results statistically reliable, but here unilinear descent is shown to be definitely more effective than sororal polygyny. All of the above facts point in the same direction, i.e., toward the superior relative efficacy of rules of descent when compared with forms of marriage.

Table 55
  Unilinear Descent and Non-Sororal Polygyny Bilateral Descent and Sororal Polygyny
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 67 73 5 7 +.12
MoBrWi-FaBrWi-Mo 48 52 5 3 +.29 100
FaSiDa-FaBrDa-Si 93 42 2 9 +.82 100
MoBrDa-MoSiDa-Si 82 43 2 9 +.79 100
SiDa-BrDa-Da 77 43 7 4 +.01
WiBrDa-WiSiDa-Da 23 27 4 0 -1.00

The higher values of coefficients and of indices of reliability in the tests of Theorems 6 to 12, as compared with those under Theorems 2 to 5, suggest that the relative efficacy of marriage forms is also lower than that of rules of residence. This is probably due, however, to the supporting influence of descent, since parallel rules of residence and descent are very strongly associated. A test case is provided by kinship terminology in bilateral societies with both patrilocal residence and general polygyny. With respect to aunts and nieces, at least, bifurcate collateral terms outnumber those of the bifurcate merging type by a ratio of nearly 5 to 1. Since the former are expected with non-sororal polygyny and the latter with patrilocal residence, the preponderance of the former when the opposing factor of unilinear residence is excluded indicates an appreciably higher relative efficacy for the marriage form than for the residence rule. The same interpretation is suggested by Tables 20 and 21 under Theorem 5. Unfortunately a comparison of the influences of sororal polygyny and unilocal residence cannot be made with statistical reliability. From such evidence as is available, nevertheless, we are left with the residual impression, admittedly highly tentative, that the relative efficacy of marriage forms is greater than that of residence rules.
The comparison of descent and residence is relatively easy. Since unilinear descent and unilocal residence have been demonstrated to exert parallel influences on kinship terminology, an analysis of the cases where only one of them is operative, excluding those where neither or both are present, should yield an indication of their relative efficacy. This is done in Table 56. The data here com- piled show that unilinear descent is associated with bifurcate merging terminology in approximately two thirds of the instances where unilocal residence is absent, whereas unilocal residence is associated with bifurcate merging in less than one third of the societies that lack unilinear descent. This indicates that the relative efficacy of unilinear descent, i.e., of exogamous unilinear kin groups, is appreciably greater than that of unilocal residence. It also suggests that descent is, in general, a more effective factor than residence, and this is confirmed by the somewhat higher coefficients and indices of reliability that ordinarily support the theorems concerned with descent (Theorems 15 to 24) as compared with those con- cerned with residence (Theorems 6 to 12).
Despite our tentative conclusion that rules of descent and the kin groups resulting from them rank highest in relative efficacy among the major groups of kinship determinants, followed by forms of marriage and the consequent family types, the influence of residence rules should not be too heavily discounted. It is partially obscured, for example, by the fact that non-sororal polygyny is rather more generally associated with unilocal residence than with unilinear descent. Moreover, the time factor (see Assumption 13) probably operates to the disadvantage of residence. As will be shown in Chapter 8, the rule of residence is normally the first aspect of a social system to undergo modification in the process of change from one relatively stable equilibrium to another, the last aspect to change being kinship terminology. Hence in systems undergoing transition, of which any large and random sample of human societies will necessarily include a considerable number, kinship terms will more often be consistent with the conservative rule of descent than the with progressive rule of residence.

Table 56
  Unilinear Descent and Non-Sororal Polygyny Bilateral Descent and Sororal Polygyny
Trios of Relatives Bifurcate Merging Other Terms Bifurcate Merging Other Terms Statistical Indices
  Q X2
FaSi-MoSi-Mo 67 73 5 7 +.12
MoBrWi-FaBrWi-Mo 48 52 5 3 +.29 100
FaSiDa-FaBrDa-Si 93 42 2 9 +.82 100
MoBrDa-MoSiDa-Si 82 43 2 9 +.79 100
SiDa-BrDa-Da 77 43 7 4 +.01
WiBrDa-WiSiDa-Da 23 27 4 0 -1.00

In intrinsic interest, kinship terminology cannot compete with sex, marriage, the family, or community organization. Hence the reader who has had the interest and persistence to follow the necessarily dry and monotonous presentation of statistical facts about kinship in the present chapter, as the demonstration of the postulational method has proceeded, deserves some satisfying reward. Perhaps he will find it, as has the author, in the realization that the data of culture and social life are as susceptible to exact scientific treatment as are the facts of the physical and biological sciences. It seems clear that the elements of social organization, in their permutations and combinations, conform to natural laws of their own with an exactitude scarcely less striking than that which characterizes the permutations and combinations of atoms in chemistry or of genes in biology.

1. A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 82-4; “California Kinship Systems,” University of California Publications in American Archaeology and Ethnology, XII (1917), 389-90.
2. A.L. Kroeber, “Yurok and Neighboring Kin Term Systems,” University of California Publications in American Archaeology and Ethnology, XXXV (1934), 15-22; “Kinship and History,” American Anthropologist, n.s., XXXVIII (1936), 338-41.
3. Cf. A.L. Kroeber, “California Kinship Systems,” University of California Publications in American Archaeology and Ethnology, XII (1917), 389; “Yurok and Neighboring Kin Term Systems,” University of California Publications in American Archaeology and Ethnology, XXXV (1934), 22.
4. See A.L. Kroeber, “Zuni Kin and Clan,” Anthropological Papers of the American Museum of Natural History, XVIII (1917), 86-7, in which the author takes issue with Lowie on the interpretation of bifurcate merging terminology, which he ascribes to the influence of unilinear descent rather than exogamy.
5. A.L. Kroeber, “Kinship and History,” American Anthropologist, n.s., XXXVIII (1936), 340.
6. A.L. Kroeber, “Yurok and Neighboring Kin Term Systems,” University of California Publications in American Archaeology and Ethnology, XXXV (1934), 21-2.
7. R.H. Lowie, “Historical and Sociological Interpretations of Kinship Terminologies,” Holmes Anniversary Volume (Washington, 1916), pp. 298-300.
8. See G.P. Murdock, “The Common Denominator of Cultures,” The Science of Man in the World Crisis, ed. R. Linton (New York, 1945), pp. 138-41.
9. The only cultural processes are historical. The author cannot accept the contention o 'White that evolution is a separate cultural process. See L. A. White, “Kroeber's ‘Configurations of Culture Growth,’” American Anthropologist, n.s., LXVII (1946), 82.
10. E.W. Gifford, “A Problem in Kinship Terminology,” American Anthropologist, n.s., XLII (1940), 193-4. See also A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 83.
11. See R.H. Lowie, “Kinship,” Encyclopaedia of the Social Sciences, VIII (1932), 569.
12. S. Tax, “Some Problems of Social Organization,” Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), p. 6.
13. R.H. Lowie, “Relationship Terms,” Encyclopaedia Britannica (14th edit, London, 1929), XIX, 89.
14. P. Kirchhoff, “Verwandtschaftsbezeichmungen und Verwandtenheirat,” Zeitschrift für Ethnologie, LX1V (1932), 51.
15. Cf. W.F. Ogburn, Social Change (New York, 1922), pp. 200-80.
16. R.H. Lowie, Culture and Ethnology (New York, 1917), p. 173.
17. Even as late as 1914 Rivers could allege that “it is possible to infer with certainty the ancient existence of forms of marriage from the survival of their results in the terminology of relationship.” W.H.R. Rivers, Kinship and Social Organisation (London, 1914), pp. 58-9.
18. A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 82.
19. A.R. Radcliffe-Brown, “The Social Organization of Australian Tribes,” Oceania, I (1931), 427.
20. A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 84; “California Kinship Systems,” University of California Publications in American Archaeology and Ethnology, XII (1917), 389.
21. S. Tax, “Some Problems of Social Organization,” Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), pp. 19-20.
22. Cf. C.L. Hull, Principles of Behavior (New York, 1943), pp. 183-203.
23. Lowie, for example, can think of no alternative to the obviously defective hypothesis of Tylor and Rivers that the practice originated under some assumed prior moiety organization. See R.H. Lowie, “Family and Sib,” American Anthropologist, n.s., XXI (1919), 33.
24. E. Sapir, “Terms of Relationship and the Levirate,” American Anthropologist, n.s., XVIII (1916), 327-37.
25. A.R. Radcliffe-Brown, “Social Organization of Australian Tribes,” Oceania, I (1931), 429.
26. S. “Some Problems of Social Organization,”Tax, Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), p. 16.
27. For a parallel criticism of Radcliffe-Brown's universal “laws,” see R.H. Lowie, The History of Ethnological Theory (New York, 1937), pp. 224-5.
28. Cf. S. Tax, “Some Problems of Social Organization,” Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), p. 10.
29. W.H.R. Rivers, Kinship and Social Organisation (London, 1914), p. 19.
30. E. Sapir, “Terms of Relationship and the Levirate,” American Anthropologist, n.s., XVIII (1916), 327n.
31. R.H. Lowie, “Kinship,” Encyclopaedia of the Social Sciences, VIII (1932), 570.
32. W.H.R. Rivers, Kinship and Social Organisation (London, 1914), pp. 44-5.
33. Ibid., pp. 21-5.
34. E. Sapir, “Terms of Relationship and the Levirate,” American Anthropologist, n.s., XVIII (1916), 327-37.
35. See P. Kirchhoff, “Verwandtschaftsbezeichnungen und Verwandtenheirat,” Zeitschrift fiir Ethnologic, LXIV (1932), 53; B. W. Aginsky, “The Mechanics of Kinship,” American Anthropologist, XXXVII (1935), 450-1. For a criticism see R.H. Lowie, Culture and Ethnology (New York, 1917), pp. 144-50.
36. In bifurcate merging the father's brother is called “father” and the mother's sister “mother,” while separate terms are used for the mother's brother and the father's sister. See R.H. Lowie, “A Note on Relationship Terminologies,” American Anthropologist, n.s., XXX (1928 ) , 265-6.
37. B. W. Aginsky, “The Mechanics of Kinship,” American Anthropologist, n.s., XXXVII (1935), 450-1; “Kinship Systems and the Forms of Marriage,” Memoirs of the American Anthropological Association, XLV (1935), 34-5.
38. E. W. Gifford, “Miwok Moieties,” University of California Publications in American Archaeology and Ethnology, XII (1916), 186-8.
39. A. Lesser, “Some Aspects of Siouan Kinship,” Proceedings of the International Congress of Americanists, XXIII (1928), 571; “Kinship in the Light of Some Distributions,” American Anthropologist, n.s., XXXI (1929), 722-5.
40. R.H. Lowie, “The Omaha and Crow Kinship Terminologies,” Proceedings of the International Congress of Americanists, XXIV (1930), 102-3; “Kinship,” Encyclopaedia of the Social Sciences, VIII (1932), 571.
41. W.H.R. Rivers, Kinship and Social Organisation (London, 1914), pp. 29-42.
42. See S. Tax, “Some Problems of Social Organization,” Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), pp. 12-13.
43. Cf. L. H. Morgan, “Systems of Consanguinity and Affinity of the Human Family,” Smithsonian Contributions to Knowledge, XVII (1870), 478-9.
44. G.P. Murdock, “Bifurcate Merging,” American Anthropologist, n.s., XLIX (1947), 60-2.
45. E. B. Tylor, “On a Method of Investigating the Development of Institutions ” Journal of the Royal Anthropological Institute, XVIII (1889), 264.
46. W.H.R. Rivers, Kinship and Social Organisation (London, 1914), pp.
47. Cf. R.H. Lowie, Culture and Ethnology (New York, 1917), pp. 136-8; G.P. Murdock, “Bifurcate Merging,” American Anthropologist, n.s., XLIX (1947), 57-8.
48. R.H. Lowie, “Exogamy and the Classificatory System of Relationship,” American Anthropologist, n.s., XVII (1915), 223-39; Culture and Ethnology (New York, 1917), pp. 140-60. But cf. “Family and Sib,” American Anthropologist, n.s., XXI (1919), 33, where Lowie shifts his position.
49. A.L. Kroeber, “Zufii Kin and Clan,” Anthropological Papers of the American Museum of Natural History, XVIII (1917), 86-7.
50. Cf. A. Lesser, “Kinship Origins in the Light of Some Distributions,” American Anthropologist, n.s., XXXI (1929), 722-5.
51. R.H. Lowie, Culture and Ethnology (New York, 1917), pp. 151-4.
52. Characterized by identical terms for cross-cousins of both types.
53. L. A. White, “A Problem in Kinship Terminology,” American Anthropologist, n.s., XLI (1939), 569-70. Cf. Theorems 20 and 21 below.
54. See, for example, E. W. Gifford, “A Problem in Kinship Terminology,” American Anthropologist, n.s., XLII (1940), 193; A. Lesser, “Kinship Origins in the Light of Some Distributions,” American Anthropologist, n.s., XXXI (1929), 728.
55. M.E. Opler, “Apache Data concerning the Relation of Kinship Terminology to Social Classification,” American Anthropologist, n.s., XXXIX (1937), 208.
56. Cf. E.V. Huntington, “The Method of Postulates,” Philosophy of Science, IV (1937), 482-95; G.P. Murdock, “The Cross-Cultural Survey,” American Sociological Review, V (1940), 369-70.
57. G.U. Yule and M. G. Kendall, An Introduction to the Theory of Statistics (11th edit., London, 1937), pp. 44-5.
58. G.W. Snedecor, Statistical Methods (Ames, 1946), p. 199.
59. In an analysis of 221 of our sample societies, 43 per cent were found to apply different terms to ElSi and YoSi (male speaking), whereas, if extended sibling terms were excluded, the percentage of age differentiation was only 19 for BrWi, 15 for WiSi, 13 for MoSi, 9 for FaBrWi and MoSiDa, 8 for FaBrDa, 7 for BrDa, 5 for FaSiDa, 4 for FaSi and MoBrDa, 2 for SiDa, and less than 1 for all other categories of secondary and tertiary relatives.
60. Cf. N.E. Miller and J. Dollard, Social Learning and Imitation (New Haven, 1941).
61. C.L. Hull, Principles of Behavior (New York, 1943), pp. 183-203. Cf. also E. R. Hilejard and D. G. Marquis, Conditioning and Learning (New York, 1940), pp. 176-85.
62. C.L. Hull, Principles of Behavior (New York, 1943), p. 266.
63. A.L. Kroeber, “Classifieatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 78-81.
64. Cf. A.R. Radcliffe-Brown, “The Social Organization of Australian Tribes,” Oceania, I (1931 ), 429.
65. When comparisons were made between all generations, the percentage of identical terms dropped from 12.8 to slightly more than 4.
66. A.L. Kroeber, “Classificatory Systems of Relationship,” Journal of the Royal Anthropological Institute, XXXIX (1909), 79.
67. Nevertheless, failure to recognize this logical axiom has been responsible for innumerable fallacies in social theory, including instinctivism and most other psychological explanations of cultural phenomena, as well as the invocation of alleged universal sociological principles as causal factors.
68. R.H. Lowie, Primitive Society (New York, 1920), pp. 157-62.
69. It is not assumed, however, that all determinants of social structure are economic. Both Islam and Christianity, for instance, have demonstrably produced changes in marriage institutions in many places, with resulting modifications in social alignments and kinship terminology.
70. R.H. Lowie, Culture and Ethnology. (New York, 1917), p. 173.
71. W.G. Sumner, Folkways (Boston, 1906), pp. 5-6.
72. W.F. Ogburn, Social Change (New York, 1922), pp. 200-80.
73. W. E. Lawrence, “Alternating Generations in Australia,” Studies in the Science of Society, ed. G.P. Murdock (New Haven, 1937), p. 327.
74. S. Tax, “Some Problems of Social Organization,” Social Anthropology of North American Tribes, ed. F. Eggan (Chicago, 1937), pp. 19-20.
75. R.H. Lowie, “A Note on Relationship Terminolowes,” American Anthropologist, n.s., XXX (1928), 265-6; “Relationship Terms,” Encyclopaedia Britannica (14th edit., London, 1929), XIX, 84-6.
76. P. Kirchhoff, “Verwandtschaftsbezeichnungen und Verwandtenheirat,” Zeitschrift für Ethnologie, LXIV (1932), 46-9
77. See B.Z. Seligman, “Studies in Semitic Kinship,” Bulletin of the School of Oriental Studies, III (1923), i, 51-68, 263-79.
78. R.H. Lowie, “Exogamy and the Classillcatory System of Relationship,” American Anthropologist, n.s., XVII (1915), 223-89.
79. A.L. Kroeber, “Zuni Kin and Clan,” Anthropological Papers of the American Museum of Natural History, XVIII (1917), 86-7.
80. G.P. Murdock, “Bifurcate Merging,” American Anthropologist, n.s., XLIX (1947). 59-00.
81. Cf. R.H. Lowie, Culture and Ethnology (New York, 1917), pp. 151-2.
82. L. A. White, “A Problem in Kinship Terminology,” American Anthropologist, n.s., XLI (1939), 569-70.
83. See G.P. Murdock, “Kinship and Social Behavior among the Haida,” American Anthropologist, n.s., XXXVI (1934), 363-5.
84. W.H.R. Rivers, Kinship and Social Organisation (London, 1914), pp. 72-3.
85. A.R. Radcliffe-Brown, “The Social Organization of Australian Tribes,” Oceania, I (1930-Sl), 84-63, 206-46, 822-41, 426-56.
86. W.H.R. Rivers, Kinship and Social Organisation (London, 1914), pp. 44-5.
87. See W.H.R. Kinship and Social Organisation (London, 1914), pp. 21-5.
88. A parallel conclusion is reached in B.Z. Seligman, “Asymmetry in Descent,” Journal of the Royal Anthropowgical Institute, LVIII (1928), 534-5.
89. E. Sapir, “Terms of Relationship and the Levirate,” American Anthropologist, n.s., XVIII (1916). 327s-37.

Evolution of Social Organization

Evolution of Social Organization

The Regulation of Sex

The Regulation of Sex

Incest Taboos and Their Extensions

Incest Taboos and Their Extensions

Social Law of Sexual Choice

Social Law of Sexual Choice

Appendix A

A Technique of Historical Reconstruction