Fortes, and Calhoun following him, conceptualise relations to agnatic ancestors in terms of descent, and relations to nonagnatic ancestors in terms of cognatic kinship. Note the conceptual confusions that result rom attempting to deal with what is ancestor-focused (and hence descent in Scheffler's (1966) and my sense) as if it were ego-focused, a matter of cognatic kinship. First, the term 'nonagnate' is recurrently used by Calhoun as if it referred to kinship, where it can only appropriately refer (as 'agnate' can) to descent status. Second, Fortes and Calhoun recurrently contrast 'agnatic' (or 'patrilineal') with 'matrilateral'. Yet 'matrilateral' serves as a label for nonagnatic relationships only if we continually change the ego who is our point of reference. For a Tale man's nonagnatic ancestors include many that are not matrilateral (in relation to him): his FM's lineage ancestors; his FFM's lineage ancestors. They may have been 'matrilateral' ancestors to somebody (his F, his FF)- but are patrilateral to him. The literature on 'complementary filiation' is similarly flawed with references to 'matrilateral' and 'mother's brothers' - even though a close reading of Fortes makes clear that what he speaks of as 'mother's brothers' include members of 'grandparent' lineage, some of whom are patrilateral from ego's point of view (see Keesing 1970a: 766). These problems disappear when we conceptualise descent categories - agnatic, cognatic, nonagnatic - with reference to the ancestors that are their focus.

Much that Fortes describes as Tallensi complementary filiation in fact reflects nonagnatic descent status vis-à-vis the ancestors of an agnatic lineage. The nonagnates who attend a Tale sacrifice comprise an array of C's (that is, B's who are not A's), defined by their common descent status vis-à-vis the ancestors receiving sacrifice, not an array of individuals separately enacting ties of ('complementary') filiation to their 'mothers' brothers':

Extra-clan kinsfolk from far and wide are present. I have counted as many as thirty, ranging from the full sister's son of the new lineage head to a distant 'sister's son', whose matrilateral ties with the clan went back to the founding ancestor of his maximal lineage (Fortes 1949: 150).

What the thirty nonagnatic participants has in common was their descent status vis-à-vis the ancestors receiving sacrifice, not a common kinship relationship to members of the sacrificing lineage. Fortes was prevented from viewing the thirty agnates as sharing a common descent

Fortes's theoretical preconceptions were reinforced by Tale idioms in which the mother's brother-sister's son relationship serves as a prototype of 'all matrilateral-sororal kinship ties' (1949: 283) - that is, what I am calling relations of nonagnatic descent, as well as cognatic kinship. Tale also talk idiomatically about nonagnatic descent as if it were matrilateral (ma dcsam is lit. 'mother begetting'). But such idioms do not obscure the conceptualisation of chains of successive filial links - descent in Scheffler's and my sense:

when a person says of another ..., speaking of lineages, ti d)yabame - 'we (my lineage) have begotten (born) ... their lineage', what is meant is that the latter can trace a line of descent to the ... former's lineage. A man refers to a sister's son (aheng) or to a classificatory aheJ in these terms (Fortes 1949: 19, my emphasis).

Nonagnatic descendants need not comprise an undifferentiated category (a point on which my 1970a paper was insufficiently explicit). Kwaio and Tallensi nonagnatic descendants related through a chain of all-male filial links, except for the last or penultimate link (the children or grandchildren of female agnates), have a specially marked status within the nonagnatic category. Some rights and interests apply only to descendants with such marked descent status.

So conceptualised, the Kwaio and Tallensi systems still appear to be more similar than Fortes's description of Tallensi would suggest. However, there is no doubt that Tallensi place greater jural emphasis on agnatic ties in defining lineage membership. Tallensi more rigidly differentiate status of agnates and nonagnates, viewing agnatic descent as a connexion of deep moral as well as jural import. They do not allow de facto transformation from nonagnatic to agnatic status by virtue of childhood residence and social commitment, as Kwaio do. But for purposes of comparison I continue to deem it important to distinguish conceptual constructions from their realisation in social groups, to distinguish cultural categories from the rights and duties that may, in a particular society, be defined in terms of them. Scheffler's 1966 argument remains compelling. Let me illustrate with reference to Solomon Islands ethnography.

The peoples of northern Malaita have the same general conceptual scheme as the Kwaio, who occupy the mountainous central zone of the island. For To'abaita, originally studied by Hogbin (1939), I quote from Frazer (1981):

The widest grouping associated with an estate includes all those people who claim descent from the founding ancestor, or the most distantly remembered ancestor of that territory ... The founding ancestor is usually buried

in one of the burial grounds in the territory he founded, and this is the focal point for sacrificial rites carried out ... by ... descendants ... This category [is] made up of all the known cognatic descendants of a common ancestor ...

Within [this] descent category two sub-categories of descendants are distinguished, those who are descended through all male links from the founding ancestor ... and those descended through one or more female links ... None of these categories or sub-categories ever comes together as a cohesive group, nevertheless it is membership of these categories which is the basis for the formation of discrete groups (1981:65).

Frazer's analysis makes clear that the subcategory of descendants 'through all male links', i.e., agnates, are accorded primacy in land rights and ritual. However, nonagnates have secondary rights and ritual interests in an ancestral estate, which can in some circumstances become primary. Nonagnatic affiliation to local descent groups is fairly common (although agnates have primacy in succession to priestly and leadership roles).

Ross's (1973) account for the Baegu speakers of the northern Malaita mountains adds some clarification and some confusion. He describes a Baegu distinction which corresponds to the To'abaita one between 'those who are descended through all male links ... [and] through one or more female links'. The terms futa ana wane, glossed as 'those joined by male linkages', and futa ana geni, glossed '[joined by] female linkages, are equivalent to Kwaio futa ani wane 'agnate' (lit. 'born of men') and futa ani geni nonagnate' (lit. 'born of women'). Yet Ross treats these concepts, which properly refer to modes of descent from ancestors, as if they referred to subcategories of ego's kindred.

Ross emphasises the agnatic ideology which in Baegu defines membership in ritual and land-owning corporations:

The Baegu recognise named patrilineal descent groups ... Membership ... is almost without exception on the basis of patrilineal descent ... Only ... those who are properly descended (that is, agnatically) should have the stewardship of the land and the responsibility for ... honouring its ancestors (1973: 138).

However, he notes a complementary nonunilineal mode of descent:

The ancestors, being the founding fathers, are closely associated with the patrilineal descent groups, and as such they form an agnatic supernatural community. Nevertheless, sacrifices to the ancestral spirits are not restricted to agnatic lines. Any descendant, those born through female links as well as of male, is entitled to offer sacrifices to the ancestor ...

A worshipper's status depends on ... a concept of cognatic or bilineal descent that operates simultaneously with agnatic ideology (1973: 143).

For Fataleka speakers who inhabit the zone adjoining Baegu to the southeast, we have fragmentary evidence from Guidieri:
l'organisation segmentair Fataleka comprend huit clans ... patrilinéaires nommés ... Les clans se décomposent en segments agnatiques (fuiwane, litt.: la racine de l'homme) nommés et non-localisés ...

En 1969 ... ces lignes étaient au nombre de cinquante-huit. Dans le cas où l'unité du clan est assurée par le principe juridique qui détermine les droits de transmission en distinguant les relations entre les parents - le dogme agnatique, ou par celui qui ... distingue en outre la ligne aînée de la ligne cadette - le dogme de primogéniture, et établit une hiérarchie entre elles, la perception de l'espace parental est claire (1972: 324).

A diagram (1972: 326) distinguishes two categories of kin in relation to ancestors: barawai dorana ('groupe de frères descendants par les hommes') and wane tofuli keni ('(les) hommes (qui) descendent (des) femmes').

Dans la succession d'un sacrificateur, par exemple, cette règle [i.e., the primacy of agnates] est toujours opérante. Les vicissitudes démographiques s'un groupe lignager peuvent naturellement assouplir ou contredire ce principe. Mais dans ce cas, la succession est considerée comme irrégulière ... (1972: 324).

Guidieri goes on to note the more inclusive conceptions of kinship/descent that operate in ritual.

Le contexte cérémoniel ... modifie considérablement la composition du groupe parental. Les membres se trouvent réunis et se reconnaissent mutuellement solidaires en fonction de relations parentales multiplies, directes et collatérales ... Le principe unilinéaire s'estompe, le nombre de participiants augmente, les groupes intermédiaries apparaissent (1972: 326).

Here 'filiation cognatique' and 'filiation utérine', as he calls them, come directly into play (Guidieri 1972: fig. 3, p. 326).
For Kwara'ae, immediately northwest of Kwaio, we have recent data from Burt. These suggest that the same categories and descent constructs used by the Kwaio provide the conceptual blueprint for Kwara'ae social relations; but that, as in Baegu and Fataleka, land rights, ritual relationships and group identification are more directly based on agnatic connexions to founding ancestors.

In theory full 'ownership' rights in land belong only to men descended in the male line from the ancestor who first established these

rights ... However, despite the emphasis which the system of ritual organisation placed on patrilineal descent, ... in practice many ... hold the land they now occupy through cognatic ties with the original 'owning' lineage. Men's rights in the land of patrilineages to which their female ancestors belonged were symbolised by the sacrifices made to the cognatic ancestors from whom they inherited ... Sons are likely to settle on their own patrilineal estates. However, they could sacrifice to individual ancestors on their mother's side, and cannot be denied the right to cultivate her land if they choose ... Men such as these [are] 'born of woman' of the lineage (Burt in press).

While these northern Malaita peoples are visibly operating with the same conceptual categories as the Kwaio,the importance of agnation in defining land and ritual interests, the possibility of nonagnates attaining the status of de facto agnates, the interests of ancestors in agnatic and nonagnatic descendants, and the role of nonagnates in ritual and sacrifice, all vary considerably. Moreover, the Kwaio and Kwara'ae systems appear to have become 'less agnatic' and 'more cognatic' on the basis of population shifts and demographic trends in the colonial period. I have suggested (Keesing 1970b) that the Kwaio conceptual system incorporates both an 'agnatic model' and a 'cognatic model'; and I have attempted to analyse formally the mechanisms whereby 'as population density increases, social structure approaches what is implied in the agnatic model ... [whereas] when population dwindles, more descent group affiliations become multiple, nonagnatic, or diffuse' (1970b: 1016).

An even more striking contrast appears when we examine the Lau speakers who inhabit the coastal lagoons adjacent to the mountains where Baegu is spoken. Lau Baegu are related dialects; and their conceptual/ideational systems are very close, though elaborated in different ways appropriate to their maritime and interior environments. Yet as social systems Lau and Baegu are strikingly different. Lau communities built on islets or coral platforms dredged from the lagoon floor, represent a 'bush' cultural pattern squeezed into radically compressed social space.

Whereas Kwaio, Kwara'ae and Baegu live in tiny, scattered and shifting homestead clusters, the male cores of Lau patrilineages are localised in densely crowded villages, often of several hundred people (Ivens 1930; Maranda 1974; Maranda & Maranda 1970). Apparently the Lau conceptualise agnatic, cognatic and nonagnatic descent, and define social categories of agnates, cognates and nonagnates with reference to founding ancestors. Yet in ritual and the definition of lineages and their rights, Lau speakers are

It is worth recalling here a similar phrasing from the Tallensi, who

insist on the equal importance of both parents in the procreation of a child. Though patrilineal descent is overwhelmingly important in the jural, economic, and ritual constitution of Tale society ... the two principles of patrilineal descent and maternal origin always work together. The two ideas are inseparable in native thought ... Whenever I have discussed these matters with informants they have generally begun by enumerating the things that make 'paternal parentage (ba dog-am)' the most important facts of one's life, and have usually concluded that, of course, 'maternal parentage (ma dogam)' is also of the greatest importance, though in a different way (Fortes 1949: 30)².

In 1971, I showed that the affiliation of nonagnates to descent groups represented the operation of a medical category of nonagnatic descendants: those whose mothers returned to their natal descent groups and raised their children there. For some purposes of descent reckoning, these women who raise their children in their (the women's) natal group are treated as if they were male links (Keesing 1971: fig 1). I call these links (from the standpoint of the children of these women) 'quasi-patrifilial'; and I class those descendants related through such links as 'quasi-agnates'³. The nonagnatic affiliants to Kwaio descent groups are mainly quasi-agnates. Technically, the distinction between agnates and quasi-agnates is a marked one (in the linguistic sense). In some contexts, mainly of ritual, the distinction is maintained and quasi-agnates are classed as nonagnates; in others, including land rights, feasting and residence, the distinction is neutralised and quasi-agnates are classed as if they were agnates. I suspect a similar conceptualisation operates among To'abaita, Baegu, Kwara'ae, and other northern Malaita hill peoples. The frequency of quasi-agnatic status is obviously a product of residence patterns and the outcome of custody claims, which in turn reflect the strength of agnatic vis-à-vis cognatic ideology (Keesing 1970b), as well as demographic circumstances. Peoples to their north appear to differ from Kwaio in part because of the lesser frequency of such quasi-agnatic linkages, in part because

What does all this have to do with the Tallensi? It indicates the inadequacies of stressing the differences between Kwaio and Tallensi on grounds that the Tallensi lineages are 'strictly agnatic', that relations between ancestors and their agnatic descendants are qualitatively different from their relations with nonagnatic descendants, and that the rights of Tale agnates and nonagnates in ritual and sacrifice are sharply distinguished. The Malaita systems represent a spectrum of relative importance placed on agnatic and cognatic descent in defining group membership, ritual status and relations with ancestors. Apparently all northern Malaita peoples recognise a principle of cognatic descent as defining the broad category of all descendants with interests in an ancestrally-created estate and important connexions to the founding ancestors. But in none of these systems is it a 'matter of indifference' whether 'connexions are matrilateral [sic] of patrilineal' (Calhoun 1980: 310). The category of cognatic descendants is, everywhere in northern (and southern) Malaita, internally differentiated, with separate status ascribed to agnates and nonagnates.

The Malaita evidence shows, I think ,the importance of distinguishing between descent constructs and categories, on the one hand, and the rights that are ascribed, and the groups generated⁴, with reference to them. Such an analysis may, as Calhoun says, be 'unsociological'. But if it allows us to compare effectively the spectrum of northern Malaita societies, to place within the same structural framework other Oceanic societies with similar descent constructs and categories (such as the Varisi speakers of Choiseul, only remotely related to the Malaita peoples, who similarly distinguish, within a cognatic descent category, between those 'born of men' and 'born of women' [Scheffler 1965]), and to make more coherent sense of Tallensi 'complementary filiation', then being 'unsociological' is surely a virtue.
Australian National University Roger Keesing

I am indebted to my colleague Michael Young for helpful comments.

¹ When Kwaio recorded the membership of 'lines', during and after the postwar Maasina Rule movement, they listed all the cognatic descendants of each group's founding ancestors. Apparently despite their greater emphasis on strict agnation other northern Malaira peoples did the same.

²2 Two points bear note. First, despite Tallensi idioms, note the inadequacy of 'maternal origin' as a term for nonagnatic descent, which can be either matri- or patrilateral. Second, despite the fact that the 'two ideas' of 'patrilineal descent' and 'maternal origin' are 'inseparable', Fortes goes to pains to separate them - so they exist, as it were, on different conceptual planes and are the foci or two different books. In Tale thought they seem simply to be two sub-varieties of dosam - 'descent'.

³ A few individuals who, because their agnates have died out or moved to Christian villages on the coast, have attached themselves to groups to which they are related through female descent lines.

⁴ As I argue in Keesing 1970b and Keesing 1971, actually membership in descent groups, as opposed to entitlement to membership, may be affected by circumstances of residence and life history; so that while descent groups may be generated with reference to descent category membership, they are in many 'unilineal' societies not strictly coterminous with unilineal descent categories.