Sides1

Dumont 1953,1966

Lounsbury 1964

Kay 1966a,b

Trauttman 1981

Schneider 1985

WHITE Douglas R. (UC Irvine) and HOUSEMAN Michael (Paris EPHE)

SIDEDNESS: 60 MILLION STRONG?

American Anthropological Association 2001

Abstract. Sidedness is a concept of dual organization at the behavioral level in marriage networks, discoverable by network analysis. Through examples in South Asia, Australia and elsewhere, sidedness in marriage networks is shown to be much more widespread than moiety organization. Sidedness may instead self-organize via network interaction into a statistical pattern that "goes without saying" and that emerges from the convergent effect of a whole series of "diametricalizing" tendencies (political rivalries, ceremonial dualism, parallel lines of inheritance or transmission of status, emergent patterns in marriage choices, etc). Failure to detect this type of dual organization in marriage behavior may result from the way ethnographers associate dual organization with hereditary moieties.

Marriage network analysis of ethnographic studies in South Asia shows Dravidian network sidedness as a complex emergent structure in which marriage behavior matches an egocentered dual organization of kinship terminology (Houseman and White 1998), long thought to lack a matching two-sided structure at the behavioral level. Analysis of marital networks in present-day "settled" Australian Aboriginal populations (Houseman 1997) suggests another pattern of sidedness in the absence of any terminological dual organization. The existence of a such flexible but sided behavioral structures in marriage choices has gone virtually undetected among ethnographers of Dravidian-language societies in South Asia or of "settled" groups of Aboriginal Australians.

The phenomenon of sidedness may be widespread among speakers of Dravidian languages, who number over 160 million (a high outer limit for sidedness in South Asia: Kannada, Tamil, Tulu, Gondi and Telugu linguistic subfamilies in India and Sri Lankai numbering 30M, 44M, 1.5M, 2.5M and 53M respectively). Analysis of marital networks in present-day "settled" Australian Aboriginal populations (Houseman 1997) suggests another pattern of sidedness in the absence of any terminological dual organization.

The realm of kinship and marriage, what one practitioner once called the "nude" of anthropology, is much neglected, seen as a prime example our the imposition of Western values and categories -- the preeminence of blood ties for example -- onto other realities. Because of our very human tendency to project our own cultural presuppositions, it was fairly easy for Schneider (1985) to demonstrate how anthropologists have projected assumptions about the priority of biological assumptions. Kinship logics and terminologies that embody assumptions about rank or marriageability are particularly difficult. In the case of Natchez kinship idioms, for example, American Anthropologists following Swanton (1911) projected the existence of social categories (the exogamous ‘Honored’ class of Natchez nobility) thought to reproduce themselves biologically, when in historical fact Honored status was a rank for men only, there were no Honored women, no Honored descent groups and no Honored social class. The error remained undetected for eighty years, and was finally demonstrated by careful textual analysis of the historical sources (White et al. 1971).

Dravidian kinship terminology is another case in point. Lewis Henry Morgan (1871) lumped Dravidian with Iroquois under the term “Classificatory” kinship, and Lowie (1928) lumped them under “bifurcate merging.” Not until Lounsbury (1964a), a century after Morgan, were Iroquois and Dravidian terminologies clearly differentiated. Yet Lounsbury’s work has not been assimilated into anthropology to the present day (GTT: 10, 340, 379). Kay (1965, 1967) is accused by Viveiros de Castro (1998:340) of imputing a mistaken calculus which inputs the same crossness index to ego in any B/Z pair on the basis of a either a patrilineal or matrilineal descent rule. (There is a simple variant of Kay’s calculus which inputs the opposite crossness index to ego in any B/Z pair). Viveiros de Castro (341) disqualifies terminological analyses of Dravidian which make “crossness a property coordinate across generations, and alliance a global and sociocentric relation” by the criterion that this “seems not to be true for the Dravidian case.” This detaches the cart from the horse without further examination of the possibility of any relation whatsoever.

The Godelier, Trautmann and Tjon Sie Fat (1998) variant of current kinship orthodoxy is a general formulation whereby a kinship terminology does “not exist in a vacuum but are influence by the marriage rules of the societies in which they are found,” and hence is best studied as a “kin term system” (GTT:5). The Dravidianate (the “Dravidian system” or Trauttman’s “Type A” kinship terminology, to identify it as an analytic types rather than a historical type associated with linguistically Dravidian societies) is defined as “the union of kinship terminology with a marriage rule, in this case the rule of cross-cousin marriage (GTT:7). Under this rule, a brother and sister may not marry, but their children should marry” (GTT:7). Crossness plus the presence or absence of a rule of cross-cousin marriage defines the Dravidian versus Iroquois “systems” respectively (GTT:15).

Anne-Christine Taylor, however, is an apostate of the Trautmann faith regarding the Dravidianate. Her study of seven groups in the Jivaro-Candoa language family having characteristic features of the Dravidian type shows ideal marriage rules that vary from the “elementary formula of local exchange partners among “close” people to the complex formula of distant marriages with the “father’s father’s sister’s daughter’s daughter within wider and wider areas, which constitute political units” (GTT:13). “The widening is accompanied by the emergence of … a threefold distinction of true consanguines, … affinal consanguines, and true affines.” She shows that referential terminology and the language of address may polarize with respect to crossness categories, effecting a discourse of negotiation of identities and diacritics among peoples “whose kinship systems and languages are, at a deeper level, unitary’ (GTT:14).

Taylor would seem to be asking for some kind of synthesis needed between the GTT orthodoxy and some kind of deeper extension of Lévi-Straussian theory of alliance, not at the level of abstract models but more concretely grounded in ethnography. Structuralism, however, by emphasizing affinity (and therefore consanguinity) in the analysis of Dravidian (Dumont 1953), failed to dislodge the “spell of descent.” “All current discussion on crossness takes as its point of departure the classic essay by Louis Dumont (1953)” (GTT:18), and Tjon Sie Fat’s elegant algebraic analysis of key features of Dravidian terminology is no exception. Unconsciously, Viveiros de Castro, Trautmann, Tjon Sie Fat, and others, agree on shared ‘diagnostic’ features of Dravidian that are consistent only with certain notions of descent that remain implicit. The argument is obscured and obstructed because these authors argue that Dravidian societies and the Dravidian system are fundamentally cognatic, and hence they can dismiss the descent-bias in their approach by simply ignoring descent as a variable.

We have been credited by the editors of GTT (p. 14) with developing

“a new method of representing Dravidian-like kinship systems that visually renders both empirical cases and ideal-type models -- the "statistical" and "mechanical" models of Lévi-Strauss's well-known discussion (1969:xxvii ff.) -- in the same register, so that they can be directly compared. [Houseman and White] introduce into the discussion of Dravidianate systems a rigorous graphic modeling of actual marriage networks that is based on their conception of "sidedness," which occupies a middle ground between the egocentric structures of the kinship terminology and sociocentric structures of social groups, such as moieties and section systems. Their elegant new instrument has a number of promising applications in kinship studies.”

Dravidian terminology as defined in the GTT orthodoxy corresponds to what we call viri-sidedness, and emphasis on virilocal groups, lacking a patrilineal rule of descent, is precisely what we find in the majority of Dravidianate societies. They take as the canonical model of Dravidian the terminological form that is statistically more common. In doing so, they have ignored a second form of the Dravidianate that is consistent with uxori-sidedness. Further, in areas of the world where the Dravidianate is common, there are cognatic systems of reckoning that are consistent with contrasts between uxori- and viri-sidedness. The distinguishing kinship terminologies of the Dravidianate in Lowland South American societies, as we show in our (Houseman and White 1998), vary in precise correspondence with viri- versus uxori-sidedness in marriage practices. Yet the GTT orthodoxy equates Dravidian terminology with kinship terminologies consistent only with the former. If anthropology were just at the point of discovering descent rather than new cognatic principles such as sidedness, this would be analogous to asserting that the only form of unilineal descent was patrilineal, sliding under the rug matrilineal descent because it was statistically infrequent. The problem is much trickier, however, because Dravidian kinship terminology in ego’s generation is consistent with both viri- and uxori-sidedness. Scheffler’s (1971) diagnostic for (TJF 1990:111-112) is correct in that it focuses only on ego’s generation. When applied to the first ascending or descending generations (TJF 1998), however, this equivalence does not apply.

IIIIIIII

Schneider’s criticism of kinship studies, then, has not helped to avert the confusion over kinship systems. Unlike the GTT orthodoxy who edited the book in which our first study of sidedness was published, Jean-Luc Jamard, writing in L'Homme (2000:735-736), caught the spirit of our analysis (in translation):

“Michael Houseman and Douglas R. White, using a variety of informatic tools, collaborate to show the emergent properties of a network of marriages that are effective through their dynamic aspect in the pratique -- behavioral practices -- of matrimonial alliances, where they find observed regularities that are not a simple effect of a terminological logic and rules of marriage. These constitute, at the level of practice, a sort of primary behavioral regularity [encodage], of a complex order. This is precisely demonstrated in that the two researchers, in the course of their analysis, are able to detect a structure of sidedness [structure à coté], or bipartite network where a pair of supersets of marriages, connected by agnatic and uterine parental links [which do not imply a principle of unilineal descent], operate so as to organize network configurations of marriage alliances across a range of societies in lowland Amazonia. The authors succeed in creating an empirical sociology of high quality that takes the first steps towards a conceptual and theoretical advance towards a sort of grounded theory (Glaser and Strauss 1967) based on facts established methodologically through carefully controlled working hypotheses."

We want to suggest what a model of a complex, practice-grounded alliance structure looks like. To do so, we have taken a particularly straightforward example, one that is easily grasped on an intuitive level. It is a certain type of bipartitioning of the marriage network that we call "sidedness." This bipartitioning is formally similar on a gross level to that of moieties, which has intuitive familiarity. (Indeed, one would be surprised and suspicious if global properties we would propose were totally foreign from alliance models that have been proposed before.) However, this familiarity with respect to certain features (2-ness) will allow us to better see the ways in which it is quite different.

What are network structures and what do they look like?

Kinship networks are time-linked, ordered by generations of succession from parents to children. Unlike kinship terminologies, the contain no “zero-point” or ego. They are multi-ego and not egocentric. Genealogical connections are partial networks, in that they are only one of many relations among egos. They are also partial because of limits on knowledge about genealogical connections. Nonetheless, analysis of the structure of simple, partial genealogical networks, so long as they provide extensive coverage of the marriages that take place within a society, that is, among its members, with some degree of closure, can reveal much that is of interest in terms of statistical and structural regularities, and about changes in such regularities over time. The network perspective implies that structural properties are taken to be dynamical – emergent and changing –rather than static. Since network data are necessarily partial, a common critique is that there is no criterion for adequate sampling, and that the investigator faces an insurmountable problem of where cut or stop at the “boundary” of the network. The criterion of “some degree of closure,” however, provides a means of evaluating both whether a sample is adequate and a boundary condition for evaluating both marriages within the group and marriages that may potentially cross an invisible boundary of potential outmarriage. Whenever a child is born there is a potential that they may marry “outside.” The boundary between “inside” and “outside” is, however, an emergent one: some marriages have an outermost form of “closure” when one can trace within the network a path in which bride and groom were already connected by prior generalogical and marriage links. Such connections prior to marriage define a relation of structural endogamy among all the actors involved. In the cycle defined by a path of prior connections, plus the new marriage, every link contributes to the “closure” of structural endogamy. New marriages that contribute to structural endogamy in this way are called “relinking” marriages, and the cycles of links that constitute the structural endogamy often overlap with other such cycles, adding to the size of the structurally endogamous group. The network perspective gives rise to a certain type of formalism and a certain type of language that simply describes how relinking and structural endogamy occur, in any society.

Formally, a marriage cycle occurs in a network of kinship and marriage occurs whenever any two people marry who are already related or indirectly linked through ties of marriage and common ancestry. A block is a maximal set of nodes (in this case, couples) in which every pair of nodes is connected by cycles. All such blocks and their connections form the core of a network. Networks, blocks and cores are models of marriage behavior in relation to genealogical linkages. Yet, while all models are abstracted representation, the network model retains an isomorphism between actual actors and their relations and nodes in the graph of the network and their relations. The network model, then, is resistant to any kind of perceived necessity on the part of the analyst to model kinship and marriage by reduction to algebraic structures. It not a reduced set of variables, and it retains the global generative potentialities of a structured whole. By way of isomorphic mappings, the network model can retain the particularities of observed behaviors and of embedded narrative discourse as they change over time.