iDIDJ Australia Didgeridoo Cultural Hub

A BUNDLE OF STICKS: THE DEBATE OVER YOLNGU CLANS.

Article from:Journal of the Royal Anthropological Institute
Article date:September 1, 2000
Author: KEEN, IAN

A critique of the clan model as applied in northeast Arnhem Land ethnography has, in turn, drawn a number of critical responses. This article defends the original critique, and takes up points raised in the responses. According to that critique a mismatch between elements of the elan model and Yolngu constructs related to identity, country, and ancestors has generated anomalies. An account of metaphors and other tropes both in Yolngu constructs and anthropological concepts and models supported this contention. This article discusses key points at issue: enclosure and boundaries of groups, descent, the homology of segments, taxonomic levels, corporateness, sets, and networks, the enduring nature of patrifilial identities, and the power of processual models to deal with anomalies. The debate has implications for related models in other regions, and for the nature of anthropological description, translation, and generalization.

In two articles and one chapter of a book, I developed a case that the clan model employed in Yolngu (Murngin) ethnography is misleading (Keen 1994; 1995; 1997); here I take up some of the issues raised in responses to that critique (Morphy 1997; Morton 1997; Sutton 1999; Williams 1999).These issues are significant because so many other analyses are predicated on the existence of clans: of the marriage system (e.g. Keen 1982; Shapiro 1981; Warner 1937), ecology and territoriality (Peterson 1972), land tenure (Williams 1986), ritual (Morphy 1984; Warner 1937), art (Morphy 1991), and so on. If the clan model is found wanting, then those analyses may require some recasting.

These are concerns of regional ethnography, but the debate touches on wider issues. The concept of 'clan' has been applied in ethnographic writings about Aboriginal society and culture elsewhere in Australia, and scholars working in other regions of the world have called related analytical categories into question (Barnes 1980; Kraus 1998; Schneider 1984; Wagner 1974). The debate has still wider implications for the nature of anthropological description, translation, and generalization.

By the early 1 980s, something of a consensus had emerged among scholars engaged in research in northeast Arnhem Land. According to this consensus, Yolngu clans (or sibs) are named, exogamous, patrilineal descent groups, divided between two patri-moieties (Dhuwa and Yirritja). Each clan owns a number of usually non-contiguous 'estates' in land and waters. Sites on each estate have associations with a variety of ancestral and other beings. Stories, songs, and rituals recount and re-enact ancestral events, represented in the form of painted and sculpted designs and sacred objects. Each clan owns its body of sacra, as well as land and waters as its property, and is a corporate group. Clans vary in internal structure from single lineages to more complex clusters of lineages, in some cases divided among land-holding sub-groups, each associated with a distinct clan estate. Each clan is a member of several cross-cutting sets or aggregates (within the moiety) on the basis of shared songs or other elements of rituals , and of language. Clans are also linked by sharing the same language or dialect denoted by the word for 'this' (dhuwal, dhuwala, djinang, dhangu etc.), in several cases across the moieties. Minor details of vocabulary and the like mark distinct clan 'tongues' (matha). Marriages link clans in complex patterns of alliance and quasi-kin relations. Each clan may be divided not only into distinct patrilineages and sub-lineages, but differentiated by the matrifilial links of its members.

The earlier ethnographies, particularly Warner's (1937), arranged these groups and categories into a taxonomic hierarchy of tribe, phratry, and clan. Later schemes did away with the tribe level but retained the idea of clans as structurally equivalent segments of Yolngu society, some divided into subgroups. The phratry level became a structure of cross-cutting sets or aggregates linking each clan to many others through its membership of a number of such sets on a number of bases: shared language, wangarr ancestors, ceremonies, public songs, and so on. [1]

The structures described in this way remain central in Yolngu social life. This combines a money economy based on transfer payments, the sale of art and craft, and monies related to a mining lease, with elements of a hunting and gathering mode of subsistence; and life in townships with smaller 'homeland centre' communities on traditional lands.

The critique of the clan model

The clan model seems convincing for some purposes some of the time, but it generates anomalies. Warner's more elaborate taxonomy gave rise to all sorts of inconsistencies (Keen 1995: 507), but the simpler schemes are not immune either. In several cases anthropologists have differed about the application of the terms 'clan' or 'sib'. Shapiro (1981: 121-2), for example, counts Dultji-Warramiri and Mirtamirta Warramiri as separate 'sibs', while Warner (1937: 39), Keen (1978: 25), Morphy (1984: 6), and Williams (1986: 64) list Warramiri as a distinct clan or matha. Warner (1937: 43-6) classifies the name Djambarrpuyngu as a language name shared by several groups, Keen (1978: 23) takes Djambarrpuyngu as the name both of a 'clan-aggregate' and a clan, while Morphy (1984: 6; 1991: 48) counts Djambarrpuyngu as a clan, and Rudder (1977: 167-74) takes it as the name of a cluster of groups with distinct countries. For Warner (1937: 43) the name Djapu denotes the language of a single clan, for Williams (1983: 64) it is t he name of a clan with localized sub-groups, whereas Morphy 1984: 6; 1991: 48) classifies 'bottom' Djapu and 'top' Djapu as separate clans.

I have argued that we need to change the terms in which we discuss Yolngu identity in relation to country and ancestors, for the simple reason that there is a mismatch between elements of the anthropological meta-language and Yolngu constructs related to identity country, and ancestors. An account of the embeddedness of metaphors and other tropes both in those Yolngu constructs and anthropological concepts and models supported this contention. Much of human social action, I suggested, consists in the enactment of constructs based on metaphor, in the extended senses of the term (Lakoff & Johnson 1980), as a special case of the enactment of cultural constructs more generally. Tropes enter some social-structural constructs as 'enacted metaphors'; a familiar example is the 'position' of lecturer in anthropology, which draws on a spatial trope (Keen 1995: 504). Constructs for the anthropological description of social life also draw on metaphors. In their descriptions ethnographers often substitute terms in the an thropological meta-language for indigenous concepts, as well as using terms with no indigenous equivalents. In anthropological descriptions such terms are substituted for, the actors' enacted constructs, which may be based on quite different tropes. Substitutions of this kind are at best partial paraphrases (Turner 1979) and at worst wholly misleading (Keen 1995: 505).

The kinds of tropes used to frame Yolngu groups and group relations, and to describe relations among individuals and groups, are quite different from those upon which anthropological constructs in the clan model depend. Concepts such as lineage, clan, descent group, and corporate group depend on images of equivalent segments, external boundaries, and levels in a taxonomic hierarchy. These constructs go together with concepts of land and country, which also involve spatial metaphors of enclosure and boundaries, and which imply taxonomic hierarchies of small bounded places of one type contained within larger ones of a different type (1995: 502, 505-6).

None of these tropes, I suggested, fit Yolngu modes of patri-group identity and relations, which involve images drawn from the human body and plants, and beliefs about ancestral journeys and traces. Far from being constituted by enclosure within boundaries or related in a taxonomic hierarchy of group and sub-group, Yolngu identities, like their concepts of place, extend outwards from foci. Connections among such identities are not those of enclosing sets, but are those of open and extendable 'strings' of connectedness. Our descriptive language needs to reflect those tropes more closely (1995: 502).

The 1995 article looked in some detail at the constitution of ancestral countries, patri-group identity and land-holding, and extended identity: it examined the metaphors, body-imagery, and other tropes involved in that constitution. Rather than distinguishing between 'clans' and 'clan-sets' of various kinds, it distinguished between least inclusive ba:purru or mala ('groups') (e.g. Liyagalawumirr), and 'strings' of these groups (e.g. Ma:tjarra). They are least inclusive in the sense that they are not usually divided into other ba:purru identities, although they may be divided into groups distinguished from one another on the basis of lineage or affiliation to distinct places (e.g. Liyagalawumirr Guruwanawuy; Liyagalawumirr people of Guruwana). Yolngu apply the word mala, 'group', to both least inclusive

ba:purru and the more inclusive and extended ba:purru identities formed from congeries or 'strings' of the former. For the most part it is the least inclusive groups that anthropologists have called 'clans'. I referred to mala or ba:purru of various degrees of inclusiveness as 'patrifilial groups' or just 'groups'. These terms can denote groups of diverse structure which have in common, among other things, the fact that a person takes the ba:purru identity of his or her father.

The article concluded that rather than make unacknowledged substitutions of constitutive terms in the meta-language for indigenous terms, it does seem possible in principle to describe the images that others use, and the ways in which they use them, within the limitations of the language at one's disposal. Figures of network, focus, and extension, and 'strings of indeterminate length' may be more appropriate for Yolngu constructs of identity than those implicit in the use of clan and related concepts (1995: 520).

While it would be foolish to assert that the clan model is inapplicable to Aboriginal social organization anywhere in Australia, it has been shown to be wanting in several regions. For example, Sutton (1978) showed that the concept of dialectal tribe (Birdseil 1953; Tindale 1974) was far from universally applicable across Australia. A few years later Myers (1986) demonstrated the inapplicability of the orthodox Radciffe-Brownian approach to local organization to the Western Desert. I have examined the contrast made by several scholars between, on the one hand, the rather individualistic, flexible, and inclusive social organization and relations to country in the Western Desert and, on the other, ownership of land by more exclusive patrilineal descent-groups in the semi-desert, Arnhem Land and Cape York Peninsula (Keen 1997). Drawing on the earlier critique, I concluded that among Warumungu, Warlpiri, Kaytej, and Alyawarre people at McLaren Creek, countries are mosaics of ancestral significance that can be di vided and recombined; shared identity is not always agreed and unambiguous; and patri-groups are not ordered in a taxonomic hierarchy of different degrees of inclusiveness.

Both Morphy (1997) and Williams (1999) have defended (and in Morphy's case elaborated) varieties of the clan model as applied to Yolngu social organization. Morton (1997) found the critique useful for his reanalysis of Arrernte land-holding; he was not persuaded, however, that the concept of corporation should be abandoned. Sutton (1999), who examined the persistence of patrifilial identities in the Daly River region and in Cape York Peninsula, agreed, and he also argued that the term 'clan' should be retained as a modal type of organization. In what follows I consider several issues raised by these responses in turn.

Enclosure and boundaries

The original argument stated that terms such as 'boundary' have been used metaphorically to express unambiguous criteria for membership of groups, such as common ancestry, unilineal succession, unilineal descent, and filiation. I suggested that the idea of defining a group or set in terms of membership criteria, or inclusion within boundaries, is incompatible with Yolngu discourse and practice. Patrifiliation, commonly taken as a basis of group boundaries in descent theory, is not a rigid criterion of ba:purru identity (Keen 1995: 519). There are other ways of gaining a ba:purru identity, especially being brought up by a man who is not one's father but who is one's mother's husband. In inheriting a particular ba:purru identity a person also inherits a cluster of cross-cutting extended ba:purm identities. Further, some individuals represent another ba:purru as identical to their own in order to assert ownership of the former, either as an individual or collectively.

In response, Williams quotes her earlier work which distinguishes boundaries of social groups from factors that can affect relations between groups. For varying purposes 'boundaries are always capable of short-term and long-term definition of the basis of differing criteria of inclusion and exclusion', a capacity that facilitates change over time, and that does not preclude disagreement about particular inclusions and exclusions (Williams 1986: 223-4; 1999: 134). Without examples it is not clear what Williams intends here; the passage could imply that groups are clearly defined, but multiple in type and cross-cutting in membership, and that some differences are temporary and ad hoc.

The problem with the use of a trope such as 'boundary' for social analysis is its lack of precision. In some ways Yolngu patreifilial groups appear to be very clearly defined: each is marked by many names, possesses its own body of mythology, is known by particular wangarr and other beings, possesses a cluster of places with ancestral and other significance, owns and (in many cases) performs a corpus of songs and dances, and makes the related painting, sand-sculptures, and sacred objects.

However, it is not uncommon in my experience for people to assert that two groups are really 'the same', especially where succession to the land and sacra of a deceased or moribund group is at stake. It is also common for individuals to claim multiple identity, through patrifiliation, spirit conception, and being 'grown up' by a mother's husband. Morphy (1991: 50) provides evidence of patrifilial groups temporarily merging their identity, even to the extent of sharing rights in land. Furthermore, the ways in which people extend names of their own group outwards to label others with similar attributes is strongly reminiscent of conceptions of country in which the identity of focal places extends outwards to include many small places (Keen 1995: 509; Morphy 1984: 26; Rudder 1997: 182). It is for these kinds of reasons that I suggested that the enclosing image of boundary is inappropriate.

Descent

Certainly filiation creates identities and differences in northeast Arnhem Land. If one's father is of the Yirritja moiety and one's mother is Dhuwa, then one is Yirritja. If the father is Daygurrgurr ba:purru (patrifilial group) then the children are Daygurrgurr, and not of their mother's group. But my use of the expression 'patrifilial group' is contentious.

Williams describes Yolngu clans in various ways: as patrilineal land-owning groups; as groups that continue in time through agnatic succession; as recruited through a process of serial patrifiliation or patrilineally; as the 'aggregate form of serial patrifiliation'. Members hold a lineage ideology; or an ideology of patrifiliation. Each clan has a myth of named founding ancestors, although its membership is assumed to extend further back into a time that merges with the creative period of the ancestors (Williams 1983: 98-9; 1986: 95; 1999: 135-6).

Scheffler (1978: 521) argues that Warlpiri 'patrilines' and 'patrilodges' are not descent groups or patrilineages because they are not defined with reference to ancestors. The same is true of Yo1ngu ba:purru or mala. Neither serial patrifiliation nor descent is sufficient to define a Yolngu patrifilial group. Many are clusters of lineages either unconnected by a common apical ancestor, or whose apical relations are a matter of dispute (or at least difference of opinion) (Keen 1994: 68-71). The identity of some Yolngu lineages is constituted through the inheritance of names, which unit a man's descendants but differentiate the succeeding branches (Keen 1994: 68).

The genealogical conceptualization of relations among groups may impose a segmentary structure, as in Evans-Pritchard's (1940) classic account. But Yolngu lineages are not elements of wider, 'aliquot' segmentary structures (Shapiro 1981). Of course, genealogical relations can be taken to imply 'levels' differentiating parents and children and so on, as well as 'segments' constituted by the children of siblings and of different wives of the same man. [2] However, certain features of Yolngu kinship, such as polygyny, the differential reckoning of kin relations through men and women, marriage between MBSD and FFZS (who are in a 'mother'-'child' relation), and a son's inheritance of his father's junior wife, work against a clear generational structure.

Yolngu representations of lineages have an organic, branching structure rather than a segmentary one. Rudder (1997: 159) reports Yolngu depictions of family and lineage connections by means of images of plants, with the ancestors at the roots and the children as branches or vines. There is a case, then, consistent with the spirit of the earlier paper, for rejecting the term 'descent', for Yolngu conceive (in these images) of ancestors as below the living rather than above, and the growth of the plant from below or out from the core. [3]

Sutton (1999: 29-30) suggests that where unconnected lineages subscribe to a common Dreaming ancestor and homeland, Aboriginal totemic, country-holding patri-groups may be said to have a notion of common ancestry. In northeast Arnhem Land many groups share major creative wangarr ancestors such as the Djang'kawu Sisters, and so are undifferentiated by 'descent' from the wangarr, although they may be differentiated by other wangarr as well as other kinds of beings associated with their countries, and by 'tongue'.

The expression 'patrifilial group' does seem useful, for it can apply both to least inclusive groups, where a descent ideology does indeed apply (although not necessarily an ideology of common descent), and extended groups which are congeries of lineages and groups connected by the common identity. Even where they share a common wangarr ancestor, it would be strained to refer to this extended group as a patrilineal descent group, although a person does inherit all those interconnected identities from his or her ba:pa (F), ma:ri'mu (FF/Z), and ga:thu (FFF/Z).

Another reason for preferring the more flexible term 'patrifilial group' is that so-called clans are not homologous elements of a segmentary organization.

Homologous segments

An important trope both in Durkheimian segmental organization and of segmentary lineage systems, I pointed out, is that of a larger whole divided into ideally homologous parts (Keen 1995: 506). Even if it is not embedded in a complex taxonomic hierarchy of levels of groups, such as tribe, phratry, clan, and lineage, the concept of 'clan' implies a type of organization with at least some degree of uniformity among its instances. Its use presupposes that the anthropologist can specify criteria for its application, and that these criteria fit known cases. For example, in Morphy's view (1991: 47) Yolngu clans are patrilineal descent groups where members acknowledge common ancestry, hold rights in land in common, and have the same body of sacred law (madayin). However, not all groups counted as clans in the literature are characterized by an ideology of common descent (Keen 1994: 68), and rights in land are not always unambiguously 'common'.

As Williams remarks, 'one should not assume that [land-owning groups] are everywhere identical in structure, size, derivation of names, and possession of a unique language or dialect' (1986: 94). [4] While I do not know of a comprehensive review, the variation in the structure of named patrifilial groups appears to include:

A group with distinct lineages all associated with one main country; but the group is 'looking after' the country of another group with the same wangarr ancestors, whose members have died Out (Liyagawumirr).

A group with distinct lineages distributed between two countries, the people of a third country having died out (Liyagalawumirr).

Two groups with distinct countries and overlapping sacra, forming a single named group with a common 'tongue'. Leaders of one dominated the religious life of the other (Daygurrgurr Gupapuyngu and Birrkili Gupapuyngu).

A group with distinct 'top' and 'bottom' (up- and down-river) groups with distinct countries, each rather autonomous (Warramiri, Djapu, Ganalbingu. Dhalwangu) (see Keen 1978: 27,29, 81, 86, 90; Morphy 1984: 6; 1991: 48; Shapiro 1981: 121-2; Peter Toner pets. com.)

The Djambarrpuyngu groups, of which Rudder has made an extended study, are more complex still. Rudder was given varying accounts of the number and identity of constituent Djambarrpuyngu groups: one man listed three, another eight, another four. His respondents specified attributes of Djambarrpuyngu groups: language, names of the groups, 'surnames', countries, location ('top' or 'bottom', i.e. up- and down-river), sacra, and invocations called out in ceremonies (Table 1). Putting these accounts together, Djambarrpuyngu groups sort most clearly with regard to the main estate or country, but form cross-cutting clusters with reference to ancestral beings, songs, and names, and relations with non-Djambarrpuyngu groups. Rudder's respondents did not sort Djambarrpuyngu into named sub-groups or 'clans' on the basis of a major shared ancestor. The most clan-like cluster of groups was those with Shark ancestor.

The Djambarrpuyngu case cannot be resolved merely by more rigorous application of the term 'clan'; to force these complexities into the clan model is to block further investigation of ba:purru organization and identity. We might ask, do Djambarrpuyngu people with Shark ancestor co-operate with each other in ways and to a degree not shared with other Djambarrpuyngu? What are the implications of sharing Shark ancestor for rights in and the control of land?

Taxonomic levels

Clan models, I pointed out, involve a taxonomic hierarchy which entails grades, orders, or classes, ranked at levels one above the other (Keen 1995: 506). Items at one level are taxonomically equivalent (named species for example, or sub-clans), and nest within items at a higher level (such as genera or clans). As I pointed out earlier, Warner's scheme distinguished three main levels: tribe, phratry, and clan, while later schemes omit the upper level. Williams, for example, writes of clans or matha ('tongues', see below) which join others to form manikay mala ('song-groups') and madayin mala ('sacra groups') (1986: 65); some clans are divided into named, localized sub-groups (1986: 64).

The semantics of mala and ba:purru do not fit the semantics of taxonomic hierarchical classification and the related class semantics, for in the usages of many Yolngu people several mala may make a mala and several groups with distinct ba:purru identities may also have a common ba:purru identity. Thus Ma:tjarra is the name of a ba:purru or mala ('group') made up of a string of constituent mala or ba:purru including Manyarrngu, Liyagawumirr, Guyula, and others. This is not consistent with a model of Yolngu groups as embedded in a hierarchy of more inclusive but different types of group, for the more inclusive and the less inclusive groups are all ba:purru or mala (Keen 1995: 519). [5]

This argument needs to be modified to the extent that Yolngu habitually identify a group by pairing names, and identifying each name as that of a distinct aspect, such as 'Djambarrpuyngu matha, Ngaladharr mala' (Djambarrpuyngu tongue, Ngaladharr group). This practice led Berndt (1965: 79; 1976: 20) to posit a structure of intersecting mala and matha categories. At the intersection of one named mala and one named matha lay a 'matha-mala pair'; every person was a member primarily by patrifiliation of one such pair. However, I showed that Yolngu do not consistently identify a given name with a given aspect (group, tongue, mala or ba:purru identity); nor do they consistently pair the same names to denote the same group (Keen 1995: 508).

A related point is that people with a common (and least inclusive) ba:purru identity sometimes disagree about the structure of their group and the relation of divisions of the group to places. Although not incompatible with an ideal of segmentary structure, I pointed out, these disagreements mean that it is not possible consistently to analyse more inclusive units down into elementary units such as 'local descent groups' (Berndt & Berndt 1977: 40) or land-holding sub-groups (Keen 1978; Williams 1986: 64). Inconsistent even with an ideal of a hierarchical taxonomy of segmentary groups are the context-dependent and cross-cutting connections of a group with several 'strings' (and clusters) of groups with the same sacra (Keen 1995: 519).

Corporateness

A number of anthropologists have described Yolngu clans as corporate groups, or as groups with corporate property (Keen 1978: 96; Peterson 1972; Williams 1983; 1986). According to Williams (1986: 95; 1999: 134), who has written extensively on this matter, matha or clans are named groups that are corporate 'with respect to land', by the ownership of land, or by virtue of the joint interests that the members have in particular land and waters. They are corporate by virtue of holding the ritual property that constitutes 'title' to land (1983: 98-9). Elsewhere these views are qualified: matha ('tongue', clan) indicates maximum 'potential' ownership (1986: 63; 1999: 133); matha identity is 'secondary to the determination of control of the most sacred ritual objects, hence of land' (1999: 132-3), although what is primary to that determination is not clear. Because of the ranking of individuals and of lineages, members' interests are not identical (1999: 133). Williams defines 'corporateness' as follows:

a set of people whose identity as a set arises from their common interest in an estate in land and/or water such that the set has a jural existence distinct from that of its individual members. A corollary in the Yolngu case is the assumption that land and waters exist and people make themselves corporate with respect to particular parts of land and waters (Williams 1999: 136). [6]

Williams contends that with the 'appropriate alterations', the 'classic' incidents of the corporation in English law -- perpetual succession, the right to hold land, the right to a common seal, the power to make by-laws, and the capacity to sue and be sued (Stoljar 1973: 96) -- characterize Yolngu landowning groups (Williams 1986: 96; 1999: 131). I suggest that the last three would require so many 'appropriate alterations' to fit the characteristics of Yolngu patrifilial groups as to be unrecognizable. Perpetual succession and the holding of land have rather closer parallels.

Commentators working in other regions have not been persuaded about the inappropriateness of the term 'corporate' (Morton 1997: 109; Sutton 1999: 29). Yolngu groups 'both as targets and as variably achieved states' have the elementary requirements of a corporation in Sutton's (1999: 29) view, namely 'an enduring group recruited on definable and predominantly consistent criteria, the members of which possess a recognised common identity, and whose members possess common jural status in relation to something valued, in this case country, sacra and so on'. However, anthropologists working in northeast Arnhem Land have not been able to specify consistent criteria, or shown that there is in every case 'a common jural status'.

The anthropological concept of corporation seems most applicable to the framing of Yolngu groups, their countries and sacra, as complexes that endure beyond the lifetime of individuals (see for example Keen 1978: 96). It gives rise to a number of difficulties, however. One is the attachment of the term to 'clan', implying that typically it is a unified group of a general type that is corporate. A second is where two or more rather autonomous groups with the same ba:purru identity focus their lives on their own distinct countries. A third is where people with a common patrifilial identity, but of separate lineages or clusters of lineages, disagree about the distribution of rights to the group's several countries. As far as the enactment of rights is concerned such groups appear, in practice, to have rather distinct rights in country. [7] A fourth difficulty has to do with the distribution of rights among individuals with a variety of relationships to a ba:purru group. While they are not equivalent, and their relative weight is open to discussion and negotiation, rights in country are distributed among a group-centred network, through the father and father's father, through mother and mother's father, through mother's mother, through spirit conception, totemic links, and so on (Williams 1986). Moreover, there may be a difference between rights or powers claimed and the ability to put those claims into practice (Keen 1995: 519).

Sets, strings and networks

Later versions of the clan model modify the taxonomic hierarchy to the extent that each clan is included in a number of cross--cutting 'sets' or 'aggregates' of different kinds; relations among the attributes of member groups are conceived as a polythetic set (Keen 1978; Morphy 1977). Nevertheless, expressions such as 'phratry', 'clan-set', and 'totemic union' (Shapiro 1981) imply a number of entities at a lower order being included within the boundaries defining the higher order entity, namely the set (Keen 1995: 520). While Yolngu speak of extended group identities as 'groups' (mala), it is misleading to think of them as 'sets' of groups, I suggested. It is always possible for people to discover further ancestral connections and hence extend the list of related peoples. Extended common identity is not that of a set governed by criteria, but an open-ended and potentially extendable 'string' of relations of family resemblance stretching out from the relative focus of the speaker, following the logic of ancest ral and human journeys and exchange relations (Keen 1995: 520).

In an assessment of Strehlow's account of Arrernte (Aranda) land tenure Morton (1997: 109) also finds metaphors of enclosure and taxonomic hierarchy inappropriate, but he does think that network models are compatible with the existence of corporate groups, bounded territories, and nested hierarchies, and that groups are, as a matter of fact, connected by such networks. Moreover, networks can serve to form corporations and territories 'however temporary or contradictory these may turn out to be' (1997: 124 n. 17).