7. Category and complement1

There are many forms of dualism. Here I wish to explore the relationship between the dyadic categories encountered in compositions of pervasive parallelism and other binary categories utilised, in Indonesia and elsewhere, to organise dual forms of social organisation. In this regard, it is useful to recall that Lévi-Strauss’s 1956 article ‘Do Dual Organizations Exist?’ was written to honour J. P. B. de Josselin de Jong and that it served as a brief, if somewhat belated, recognition of Dutch research on Indonesian dyadic structures. Lévi-Strauss’s intention in the article was to draw a comparison between American-Indian and Indonesian forms of dual organisation, yet the focus of his comparison was curiously incongruent, since it involved a comparative analysis of the specific social structures of the Winnebago and the Bororo on one hand, and a constructed model of an Indonesian-type social structure on the other. This Indonesian-type model, based on five binary oppositions, was characterised by three positive features: non-residential marriage classes, prescribed marriage, and an opposition between the sexes, thus supposedly resulting in a system of moieties distinguished as male and female in association with asymmetric or generalised exchange.

This model is intriguing but its derivation is difficult to fathom. Moieties of a sort occur throughout Indonesia, but they are not invariably designated as male and female and their function is rarely to regulate marriage. In some Indonesian societies, specific categories of men and women are defined as strictly marriageable, but these categories do not constitute marriage classes nor are they coincident with a particular clan structure. Systems of asymmetric prescriptive marriage do indeed occur in Indonesia, but they are by no means universal. Indeed such systems constitute a minority in a region where marriage is overwhelmingly non-prescriptive. In western Indonesia the Toba Batak offer what is considered to be a ‘classic’ example of asymmetric prescriptive marriage, whereas in eastern Indonesia, societies with similar such systems are found scattered and interspersed among societies with other forms of marriage, primarily on the islands of Flores, Timor and Sumba, and on some of the islands of the Moluccas. Thus, although it is certainly possible to discern the various elements of Lévi-Strauss’s model, their combination conforms to no known Indonesian society. Hence it is reasonable to question the relevance of the model for the comparative analysis of dyadic structures in Indonesia. Yet to dismiss this model as irrelevant would be to ignore its relation to (and possible derivation from) the more influential model of eastern Indonesian social structure developed by F. A. E. van Wouden in the doctoral dissertation entitled ‘Types of social structure in eastern Indonesia’, which he wrote under the direct supervision of J. P. B. de Josselin de Jong in 1935 (subsequently translated and published as van Wouden 1968).

Van Wouden, whose work Lévi-Strauss alludes to in his article, attempted to disentangle an accumulation of disparate ethnographic evidence from eastern Indonesia. Like Lévi-Strauss, van Wouden regarded marriage as the ‘pivot’ for social organisation, whose categories provided the basis for an all-embracing cosmological classification. He also noted that ‘ordinary’ (MBD/FZD) cross-cousin marriage and ‘exclusive’ (MBD) cross-cousin marriage represented ‘two opposed systems of affinal relationships between groups’ (1968:90). Since van Wouden thought that these types of marriage formed the foundation for the dualistic and triadic patterns of classification that were so evidently interwoven in the cosmologies of eastern Indonesia, he was obliged to construct a model that reconciled them. In his model, which was intended to represent the original form of Indonesian social organisation, van Wouden opted for exclusive cross-cousin marriage but in a closed chain of relationships among an even number of clans. By the logic of this model, if the clans are patrilineal there must be an equal number of latent matrilineal groups, resulting in a ‘double-unilateral’ (or double-unilineal) system. The limiting case required four clans that would ideally produce a four-clan or ‘double two-phratry system’. In terms of the model, as van Wouden noted, ‘dual organization…is not required by the system, but can very well accompany it’ (1968:88).

In retrospect, although it is possible to comprehend both van Wouden’s and Lévi-Strauss’s models, it is difficult to resuscitate the intellectual ambience that once made these models so compelling. Both models now seem stunningly simplistic. Both are constructed on a simple set of binary oppositions and are thus implicated in a dualism of the sort they are intended to illuminate. Moreover, both models share the same Durkheimian inheritance that ultimately derives classification in general from the categorisation of social forms. As a result, both models focus primarily on the products of classification rather than the processes of classification. For this reason in particular, neither model now offers an appropriate starting point for the study of dyadic structures.

Eastern Indonesia: Three ethnographic cases of dual structures

In eastern Indonesia (and especially in the Lesser Sundas and the Moluccas) comparative research has developed considerably in recent years. When van Wouden wrote his dissertation he had to draw together and attempt to make sense of perhaps a hundred scattered reports of varying lengths and reliability. He had not a single ethnography of note to guide his speculations. Today, however, there are at least a dozen substantial ethnographies on the region and at least another dozen studies now in preparation.2 The picture that emerges from these studies is somewhat different from the one that van Wouden sketched.

All of these ethnographies, without exception, confirm the prevalence and importance of dyadic structures. Yet the sheer variety and diversity of these structures militate against any conception of a single institution of ‘dual organization’. Nor is it simply the variety of these dyadic structures among the many different societies of the region that makes it difficult to apply the classic models of dual organisation; more difficult still is the application of this concept to the diversity of dyadic structures within any single society. To illustrate what I mean by this, I will describe in outline form the dyadic structures of three neighbouring island societies on which I have done fieldwork.

The first case is that of Savu. This island has one of the two societies in eastern Indonesia with a bilineal social organisation that might be considered to resemble van Wouden’s double-unilateral model. Together with the tiny offshore island of Raijua, Savu is composed of five ceremonial domains. Each of these domains is, in turn, composed of named, localised, patrilineal clans (udu), which are often further divided into lineages (kerogo), all of whom recognise the same ‘origin’ village. Crosscutting the particular allegiances of these localised groups is an island-wide system of ranked matrimoieties: Hubi Ae (‘The Greater Blossom’) and Hubi Iki (‘The Lesser Blossom’). These moieties, ‘The Blossoms’, are further subdivided into Wini (‘Seeds’). On Savu, however, there is neither a terminological nor a systematic rule of marriage governing relationships between patrilineal or matrilineal groups. Instead there is a marked tendency, for reasons of status, for the occurrence of internal marriages within each matrimoiety and intermarriage among high-ranking patrilineal groups.

Each ceremonial domain has its own lunar calendar and a native priesthood to conduct ceremonies in sequence throughout the year. The arrangement of the lunar calendar, the cycle of the ceremonial year, the organisation of the priesthood, and the allocation of ritual duties to specific priests and clans are all based on a series of interrelated dyadic structures. In the domain of Liae (see Fox 1979), the ceremonial year consists in an opposition between the planting season and the lontar-tapping season, the period of ritual silence and the period of gongs and drums, the time when the Deo Rai from clan Gopo and his priestly council, Ratu Mone Telu (‘The Three Male Priests), preside and the time when the Apu Lodo from Napujara and his council, the Ratu Mone Pidu (‘The Seven Priests’), hold sway. During the high ceremonial season that marks the transition between these two seasons, in the month of Bangaliwu Gopo, the Deo Rai takes precedence; in the following month of Bangaliwu Rame, the Apu Lodo takes precedence.

The progress of each lunar month is also conceived in terms of a set of oppositions: waxing and waning, east and west, life and death, above and below. And during each month a complex ritual dialectic assigns lineages and clans as opposing groups with specific ritual functions. Thus, for example, in the great cockfighting ritual of the month of Bangaliwu Rame, one lineage of Napujara, the Apu Lodo’s clan, joins Gopo, the clan of the Deo Rai, in ceremonial opposition to the clan Nahai, which is joined by an opposing lineage of Napujara. These groups assemble at different village sites, form groups known as Ada Mone (‘The Male Group’) and Ada Rena (‘The Female Group’), and then position themselves at the ‘upper’ and ‘lower’ ends of an enclosure on the top of the hill of Kolarae where they conduct ritual combat with their fighting cocks. Other clans take sides with one or the other group, or divide internally into opposing lineages. The essential point to be made here is that this particular configuration of opposing clans holds only for Bangaliwu Rame; other configurations based on different categories occur in other months of the year. The configuration that I observed and described for Bangaliwu Rame in 1973 is neither fixed nor unchanging but, by common understanding, it is recognised to be, in large part, the result of the internal historical dynamics of the development of the clans of Liae. In effect, on Savu there is no single set of concordant dyadic structures, but rather a proliferation of such structures, each fitted to a particular purpose.

The second case is that of the domain of Thie on the island of Rote. Unlike Savu, Rote has only patrilineal or, more precisely, patronymically ordered descent groups; maternal affiliation, however, is acknowledged for three generations, but this acknowledgment does not form the basis for a coherent matrilineal line of descent. Nor is there any terminologically prescribed rule of marriage.

The island was traditionally divided into 18 domains and, despite administrative consolidation since 1968, these domains retain their role as primary communities of orientation (Fox 1977). For centuries, each domain, under its own separate ruler, developed distinctive traditions and, except for a few royal and high noble inter-domain marriages, each domain has remained largely endogamous, following distinct rules and customary practices. A common set of basic cultural categories is evident throughout the island, but the social application of these categories varies from domain to domain (Fox 1979).

In the domain of Thie (and in one other domain, Loleh) a system of marriage moieties has developed. Of Thie’s 26 clans, 14 are assigned to the moiety of Sabarai, of the Manek or ‘Male Lord’, and 12 are assigned to the moiety of Taratu, of the Fetor or ‘Female Lord’. In his dissertation, van Wouden devoted special attention to Thie, seeing in the domain a phratry system and, on the basis of hints in one source, even a possible eight-class marriage system. When examined in more detail, however, this dual organisation dissolves in a variety of disparate dyadic structures (see Fox 1980). Each moiety is divided into major clans (leo inak) and minor clans (leo anak), whose status conforms to that of ‘noble’ and ‘commoner’ in other domains. The group of noble clans in each moiety is further subdivided into various ancestral groupings with specific political functions, while the minor clans within the moiety of Sabarai form a ritual group associated with the powers and fertility of the earth. These minor clans also form a separate marriage unit that may marry either with Taratu or with the major clans of Sabarai. Thus, at a further level of specification, an apparent dual organisation becomes a triadic structure.3

This moiety system dissolves still further, since one clan on each side is exempt from following any moiety rules—in one case because members of the clan originate from an offshore island, and in the other case because the clan’s ancestor is said to have arrived late at the ceremonial gathering at which the moiety system was established. These justificatory explanations further highlight an essential conceptual feature of the moieties. They are not conceived as a primordial structure but rather the reverse: as a formal historical ordering of an untidy process of clan formation undertaken at the behest of one of the later rulers of the domain.

As on Savu, in Thie, history is given due recognition. The assignment of groups to a particular dyadic segment is thus considered to be contingent on past events. This becomes clearer when one examines the various ceremonies that were once performed by the moieties. The major clans of both Sabarai and Taratu performed their own origin celebrations. Four minor clans, two from each moiety, also performed individual celebrations; however, the most important ceremonies were performed by the minor, ritually powerful clans of Sabarai. These clans, of which two were associated with the east and three with the west, were obliged to lead an annual ritual combat. At a ritual feast for all the clans, a rice-pounding block would suddenly be tipped over and those who found themselves to the east of the block joined the two clans of the east and those to the west joined the three clans of the west, thus arbitrarily obliterating all other dual structures. Thus, again as on Savu, in Thie there is, despite initial appearances, no single, all-embracing dual organisation, but instead a host of particular dyadic structures.

The third case is that of the Atoni of West Timor. Dualism among the Atoni, based on such categories as male and female, inside and outside, and ‘wife-giver’ and ‘wife-taker’, has been amply described for the structure of their domains, their descent groups and their houses (Cunningham 1964, 1965, 1966; Schulte Nordholt 1971, 1980). Unlike the Savunese and Rotenese, the Atoni do have a terminological rule of symmetric marriage, though by preference they tend to arrange particular marriages in an asymmetric fashion. Despite this rule, however, the organisation of descent groups defies simple description. In general, domains on the island of Timor are larger than those on Rote or Savu and settlement is more scattered. Moreover, the continual migration of different segments of named descent groups throughout the island has produced a heterogeneous structure within each domain. Thus, a domain whose traditional structure was founded on an idealised set of relationships among certain leading clans actually consists of myriad local relationships among minimal descent groups. Narratives of the wanderings of the ancestors acknowledge these formative processes, all of which create a situation not unlike that on Rote and Savu, where there exist multiple dyadic structures but no classic dual organisation.

Faced with the ethnographic situation in eastern Indonesia, certain conclusions can be drawn. Clearly, the classic models of dual organisation appear inappropriate, for there is no single organisational form for the variety of dyadic structures that is to be found either within any one society or among the numerous different societies of eastern Indonesia. Furthermore, since the variety of dyadic structures implies an absence of a formal concordance between social and symbolic forms, the study of dual classification risks becoming a typological enumeration of dual structures. Hence I would argue that what is needed is a further study not of the products of classification, but of the processes of classification.

Such a study must be undertaken at two levels: first, at a general, abstract level that focuses on the features that seem to underlie processes of dual symbolic classification, and then at a categorical level that focuses on the way in which specific sets of categories form complex systems of symbolic classification. The first level allows comparison with other societies throughout the world, while the second level can only relate to a reasonably defined ethnographic field of study: to historically related societies that share common linguistic categories. Whereas the first level necessarily requires a degree of formalism, close attention to the second level leads ultimately to an intimate examination of related metaphors for living. Since the two levels are related, however, I shall attempt to examine processes of dual classification in relation to the ethnography of eastern Indonesia, focusing on five features of dual symbolic classification systems in eastern Indonesia: 1) parallelism, 2) recursive complementarity, 3) categorical asymmetry, 4) category reversal, and 5) analogical crossover.4

The processes of dual classification in eastern Indonesia

Parallelism

Roman Jakobson has argued that the principle of parallelism is implicated in all poetic statements. Jakobson’s notion of parallelism is broadly relevant to an understanding of symbolic statements in general (Fox 1977:59–60) and, as I have argued, to ritual performances as well (Fox 1979:169–71). Here, however, I will confine myself to the strict form of parallelism known generally as ‘canonical parallelism’, since virtually all the societies of eastern Indonesia use some form of canonical parallelism for the expression and transmission of ritual knowledge.5 Generally this parallelism takes the form of a ritual language in which all or most semantic elements are paired in dyadic sets, structured in formulaic phrases, and expressed as couplets or parallel verses (Fox 1971, 1974, 1975).

As an example of this form of canonical parallelism, I present a brief Rotenese ritual composition by the chanter L. Manoeain of the domain of Ba’a.

1. Sa Lepa-Lai nunun	The Waringin tree of Sa Lepa-Lai
2. Ma Huak Lali-Ha kekan	And the Banyan Tree of Huak Lali-Ha
3. Keka maba’e faluk	The Banyan has eight branches
4. Ma nunun mandana siok.	And the Waringin has nine boughs.
5. De dalak ko sio boe	These are the nine roads
6. Ma enok ko falu boe.	And these are the eight paths.
7. Fo dala sodak nai ndia	The road of wellbeing is there
8. Ma eno mamates nai na.	And the path of death is there.
9. De suli malamumula	Therefore watch with care
10. Ma mete makananae.	And look with attention.
11. Ndanak esa dulu neu	One branch points east,
12. Ma boso musik ndia	But do not follow that
13. Te fiti-ngge ledon dalan ndia	For this is the road of the sun’s fiti-ngge
14. Ma telu-ta’e bulan enon ndia.	And the path of the moon’s telu-ta’e.
15. De fiti-ngge fiti-fiti	The fiti-ngge thrusts and thrusts
16. Ma telu-tae tati-tati.	And the telu-ta’e chops and chops.
17. De nggelo lesuk nai ndia	The neck breaks there
18. Ma ladi puk nai ndia.	And the thigh snaps there.
19. Ndanak esa muli neu	One branch points west,
20. Boso musik ndanak ndia	Do not follow that branch
21. Te nitu hitu dalan ndia	For this is the road of seven spirits
22. Ma mula falu enon ndia	And this is the path of eight ghosts
23. De mate nituk nai ndia	The death of the spirits is there
24. Ma lalo mulak nai ndia.	And the decease of the ghosts is there.
25. Ndanak esa ki neu	One branch points north,
26. Boso musik ndanak ndia	Do not follow that branch
27. Te pila bii-late dalan ndia.	For this is the road of the red ‘goat’s grave’ spider.
28. Ma modo bolau enon ndia.	And this is the path of the deadly green spider.
29. De peta-aok nai ndia	The swelling of the body is there
30. Ma hina-talek nai na.	And the festering wound is there.
31. Ndanak esa kona neu	One branch points south,
32. Boso musik ndanak ndia	Do not follow that branch
33. Te manufui tela dalan ndia	For this is the road of forest fowl
34. Ma kukuha na’u enon ndia.	And this is the path of four-talon grass bird.
35. De o leno kada telas dale	You only wander within the forest
36. Ma o pela kada na’u dale	And you only turn within the grass.
37. Te ndana esa lido-lido lain neu	But one branch goes forward to Heaven
38. Ma dape-dape ata neu.	And goes straight to the Heights.
39. Na musik ndanak ndia	Then take that branch
40. Te dala sodak nde ndia	For this is the road of wellbeing
41. Ma eno molek nde ndia	And this is the path of peace
42. Fo nini o mu losa kapa sula soda daen	To bring you to the buffalo-horn land of wellbeing
43. Ma mu nduku pa-dui molek oen.	And to the flesh and bone water of peace.
44. Dae sodak nai ndia	The land of wellbeing is there
45. Ma oe molek nai na	And the water of peace is there
46. Fo o hambu soda sio	For you will find the wellbeing of nine
47. Ma o hambu mole falu	And you will find the peace of eight
48. Ma dua lolo ei	And with legs outstretched
49. Ma kala ifa lima	And with arms cradled on the lap
50. Fo ifa limam no limam	Cradle your arms upon your arms
51. Ma lolo eim no eim.	And stretch your legs over your legs.

This composition shows the power of parallelism to create a cosmology, simply and effectively: the world-waringin as the tree of life with its branches as paths leading in different directions. The composition is seemingly traditional: the parallelism is impeccable, and standard formulae (as, for example, ‘the buffalo-horn land of wellbeing/the flesh-and-bone water of life’) are strictly maintained. Moreover, the cosmology created by the imagery of the composition accords with a common tree-of-life cosmology found throughout Indonesia. The fact is, however, that this cosmology does not conform to the standard cultural cosmology of the Rotenese, which, based on imagery of the island as a creature laid out lengthwise, assigns entirely different values to the directions, giving priority to the south and east over the north and west (see Fox 1973:356–8). The blind chanter ‘Old Manoeain’, who recited this composition for me, was one of the leading Protestant ministers on Rote and was renowned for his use of Rotenese ritual language in his sermons (see Fox 1983). His composition, as far as I can determine, is a personal attempt to create a kind of Christian cosmology. The essential point is that it offers an alternative cosmology—another possible world—using the same dualistic linguistic resources that are regularly used to express and uphold the standard cosmology.

The proper use of these linguistic resources requires a minimal knowledge of at least 1000 dyadic sets. This entails highly specific knowledge of which nouns, verbs, adjectives and prepositional forms may pair to form ‘canonical’ dyads. Thus, for example, one must know not just that ‘north’ (ki) may pair with ‘south’ (kona), but that moi, which functions as either an adjective or a verb meaning ‘slick, smooth, to lick’, forms a canonical pair with keni, ‘shiny, glossy, to polish’; or that nafi, ‘sea-cucumber’, only forms a set with sisik, ‘mollusc’; or that delas, the dedap tree (Erythina spp.) must be paired with nitas, the kelumpang tree (Sterculia foetida); or that melu, ‘stomach-cramps’, only pairs with langu, ‘headache’; or that nggio, which describes the ‘creaking or scraping’ of tree branches, must be linked to ke, which refers to ‘annoyance or teasing’; and so on through hundreds of specific dyadic sets.

This pervasive parallelism, however, is entirely neutral. It may contribute to and sustain a thoroughgoing and highly particularised dualistic perception of the symbolic world, but on its own it is insufficient to constitute the kind of ordered dyadic structures associated with dual organisations and dual cosmologies. A dual cosmology—as indeed any dual organisation—is characterised not by a simple pairing of elements but by the analogical concordance of elements within pairs according to some criterion of asymmetry. The rules of parallelism provide no such criterion. Hence dyadic sets are essentially neutral pairs: one element is not superior to another and either element may precede the other in expression. Extra-linguistic criteria are required to transform parallel elements into the elements of a dual organisation or cosmology.

Moreover, systems of canonical parallelism of the kind that are to be found throughout eastern Indonesia are an overly rich resource of dual categories. Most dual organisations rely on a relatively limited number of categories. Only a selection of categories, rather than all the resources of canonical parallelism, would suffice for this function. In short, the canonical parallelism of the ritual languages of eastern Indonesia may account for the elaborateness of dualistic structures in the region, but it cannot explain them. Thus, the argument must be extended further.

Recursive complementarity

Dutch anthropology has long insisted on the importance of complementary dualism to the understanding of societies in Indonesia and elsewhere (see de Josselin de Jong 1977). The complementary categories denoted in these studies are a familiar feature of what is often called two-column analysis—ordered lists of general categories arranged as complementary pairs:

right left

day night

sun moon

east west

red black

The possibility of constructing such a table for almost any society and the occurrence of at least some common categories in all such tables limit the usefulness of this kind of analytic exercise. Two-column analysis hardly offers more than a beginning to an understanding of complementary dualism.

In eastern Indonesia the most important and recurrent complementary categories reflect a common Austronesian derivation and a historically shared inheritance of similar metaphors for living. These categories include a variety of directional and spatial coordinates such as north–south, east–west, inside–outside, back–front, right–left and upward–downward. Equally important are colour categories (white–black, red–gold–blue–green), categories for parts of the body (head–tail, or head–buttock), categories for persons and gender (elder–younger, male–female), botanical categories (unripe–ripe, trunk–tip, planted–harvested), and other categories for qualities (cool–hot, bland–bitter). Some of these categories are more than just symbolically associated: they are linguistically synonymous or even identical. Thus, left–right is, in some societies, synonymous with north–south. Similarly, botanical categories that provide the principal metaphors for growth and development may have colour or spatial connotations. The complementary colours green–gold may also be synonymous with the categories unripe–ripe, while the categories of trunk–tip imply a spatial–temporal notion of origin and extension.

The major point, however, is that this array of complementary categories represents a relatively small selection from the total resources of all possible canonical pairs furnished by the parallelism of ritual languages. The configuration of this select array of categories varies from one eastern Indonesian society to another, but in each society it constitutes what Needham has described as the set of ‘primary factors’ in that society’s symbolic classification (Needham 1978:12–13). Moreover, and more importantly, these categories serve as the ‘operators’ of the symbolic system—that is, as organising elements for the classification of other categories and qualities. In this regard, what is significant is the recursion of these categories—the way in which they may be applied successively in various contexts and at many levels of signification.

The categories of male–female provide an excellent illustration of recursive complementarity. On Rote, these categories (mane–feto) may apply to persons as well as to certain kinds of trees and plants, to political offices, and, in Thie, to opposing moieties; they may also be applied within descent groups to distinguish client lineages, or between descent groups to indicate wife-givers and wife-takers. These categories may also be applied to different gifts exchanged by wife-givers and wife-takers. At yet another level these categories differentiate between the sirih or betel catkin, which is always masculine, and the pinang or areca nut, which is feminine—objects mutually offered by both sides in ceremonial exchange.

Forth, a recent ethnography on the domain of Rindi in east Sumba indicates how this recursion of male–female categories is extended even further in differentiating articles of bride-wealth given by wife-takers. To begin with, all bride-wealth objects (which consist of horses plus gold, silver, or tin chains and pendants) are classified as masculine in opposition to the dowry goods (textiles, beads and ivory bands) given by wife-givers. Internally, however, ‘masculine’ goods are distinguished as male and female: horses as a category are male, whereas metal valuables are female. At a further level, horses are distinguished as male and female and should be given as a pair consisting of a stallion and a mare. The metal valuables are also distinguished as male and female: chains are considered masculine and pendants female, and these categorically feminine pendants are still further distinguished according to their decoration as male or female. Similar distinctions can be applied to the ‘feminine’ goods given by the wife-givers: textiles must include men’s cloths and women’s skirts, and so on (Forth 1981:360–1).

Figure 7.1: Male and Female Exchange Goods

By this principle of recursive complementarity, nothing is exclusively of one category; anything that is categorised according to one component of a complementary pair can potentially contain elements of its complement.6 A great deal of the symbolic elaboration of dualistic structures in eastern Indonesia involves playing with this principle of recursive complementarity: male contains female, female contains male; inside contains outside, and outside, the inside; black, white, and white, black. Similarly, wife-givers are also wife-takers, and a group that is classified as elder to one group may be younger to another.

This principle should not be confused with hierarchy, since it is not wholly systematic and it rarely achieves great taxonomic depth. Other factors affect the application of this principle. Thus it is essential to know in any society in eastern Indonesia which sets of complementary categories apply. In Rote, for example, the categories male–female can be applied to distinguish the two unequal halves of the traditional house. Their application, in this case, is coincident with the categories outside–inside. On the basis of this coincidence, however, to apply male–female categories to the ‘Spirits of the Outside’ and the ‘Spirits of the Inside’ would produce a confusion of categories. In effect, no single set of recursive categories is applied systematically throughout the culture.

Categorical asymmetry

Another feature of those recursive complementary pairs that serve as ‘operators’ for the elaborate dual structures is their asymmetry. In this regard, parallelism is entirely neutral, always consisting of undifferentially paired semantic elements. In Rotenese ritual language, just as there are no criteria to distinguish the verb ifa (‘to hold on the lap’) as, for example, somehow marked in relation to its pair, ko’o (‘to cradle in the arms or on the hip’), so too one cannot differentiate the directional dulu (‘east’) from its pair, muli (‘west’). Yet, when east and west are used as recursive complements, east is definitely superior to west. On Rote, where the conflict between traditional and Christian cosmologies has, it seems, prompted a conscious need to justify the traditional, there exist a number of aphorisms that serve as symbolic syllogisms to give ‘value’ to the directional coordinates. Thus, for example, Dulu nalu muli, te hu ledo neme dulu mai, de dulu ba’u lena muli: ‘East is as broad as the west, but the sun comes from the east, therefore the east is much greater than the west.’ All of the other recursive complements that I have mentioned are similarly distinguished in an asymmetric fashion, although not all are as consciously justified: male is ‘superior’ to female, inside to outside, head to tail, red–gold to blue–green.

This categorical asymmetry is, in some ways, similar to markedness, but whereas markedness pertains to linguistic levels that are largely arbitrary and unconscious, this asymmetry of complementary categories occurs in socially constructed symbolic systems that can be consciously manipulated. Moreover, the category that is ‘marked as superior’ functions as the equivalent of the unmarked category. Thus, in those contexts where one component of a complementary pair is required to stand for the whole, it is the category marked as ‘superior’. Hence ‘male’ may stand for the whole in regard to persons, ‘inside’ for the whole of various bound structures, or ‘head’ in certain contexts for overall authority or precedence. The very existence of such categorical asymmetry, however, creates the possibility of its inversion.

Category reversal

Category reversal refers to a change in the polarity of any set of complementary categories. This only occurs in special contexts when proper order is subverted and the ‘world is turned upside down’: outside becomes superior to inside, female to male, west to east, north to south, and so on, in terms of the standard dual asymmetries of the culture.7

In eastern Indonesia the contexts for inversion are considered ‘extraordinary’ even when they occur with annual regularity. On Savu, category reversal appears to be associated with the transition from one calendar year to the next; on Timor, reversal was crucial to headhunting, which invariably occurred in the dry season (McWilliam 1982); on Rote, minor reversals occur at all funerals, but major reversals are associated with the burial of those who have died a bad death (Fox 1973). More generally, however, some Rotenese characterise the whole period of Dutch dominance over the island as a time of partial inversion of the proper cosmic order. In the Rotenese cosmology, south (right), which is the direction of maximal spiritual power, is superior to north (left), which is the direction of sorcery and bad death. During the colonial period, however, the island was under the spell of the Dutch and the proper order was partially inverted. A symbolic syllogism recorded from the domain of Oepao at the turn of the century is reported as follows—Ona ba’u i boe, te hu Komponi nai i, de i ba’u lena ona: ‘The South (Right) is as great as the North (Left), but the Company (i.e. the Dutch government) is in the North, therefore the North is greater than the South’ (Jonker 1913:613). This period, which the Rotenese also describe as a time of native ‘ignorance’ and ‘left-handedness’, ended with the achievement of independence and the return of symbolic power to the south.

Analogical crossover

I use the term ‘analogical crossover’ to refer to another prominent feature of dual symbolic classification in eastern Indonesia. This feature is not an aspect of a single set of dual categories but rather pertains to the potential ordering of complex sets of such categories. It is specifically a property of historically developing systems in particular contexts.

Standard two-column analysis of complementary pairs consists in a simple analogical arrangement based upon the asymmetric valence accorded these categories. The result is an apparent concordance. As I have indicated, it is possible to change the valency of any or all of these categories. In eastern Indonesia this occurs periodically at ‘special’ intervals and I have termed this change in valency ‘category reversal’. But it is also possible to retain standard valencies and instead change the analogical association between sets of complementary categories. It is this that I term ‘analogical crossover’.

An illustration of this feature can be taken from the symbolic classification of the Atoni Pah Meto of Timor. As among the Rotenese, the categories male–female (mone–feto) form important recursive complements. In most relational contexts, male is superior to female. Mone, for example, is used to refer to the wife-giver and, in any local settlement, refers to the Atoni Amaf, the ‘Father’ or ‘Father Atoni’, who represents the founding lineage segment in the area. Those who have come after this founder and have become wife-takers are designated as feto. Similarly, the categories outside–inside form another important set of recursive complements; as a set, inside is considered superior to outside. Thus both inside and male are given a positive valency in relation to their mutual complements, outside and female. There is, however, a further factor that affects the analogical association of these pairs. Timorese folk etymology links the word for ‘male’ with a similar word meaning ‘outside’. The combination of these categories leads to analogical crossover:

Figure 7.2: Analogical crossover

Male, which is superior in certain contexts, is associated with the outside, which is inferior; and female, which is a subordinate category, is associated with the inside, which is superior. The logic of these categories produces a tension that pervades the traditional classification system of the Atoni, expressing itself in an ideal model of four political units represented as four ‘fathers’ grouped around a female centre. Like a four-pillared Atoni lopo, these male figures ‘support’ a female centre but remain subordinate by their outside position.

Schulte Nordholt, who has described the Atoni political system at length (1971), has provided another example of this kind of analogical crossover in the various permutations of relationships among the four clans in Bikomi and Miomafo. In their affinal relations, Bana and Senak are male, while Atok and Lahe are female. By one ritual division, Lahe and Senak are inside and hence female, while Atok and Bana are male and outside; but by another internal division, Senak represents the immobile centre, which is female but superior to Lahe, who represents the active male outside (Schulte Nordholt 1980). A similar form of analogical crossover seems once to have formed the basis of the sacred hegemony of the ‘matrilineal’ Tetun, whose centre was at Wehale (Fox 1983).

The possibility of analogical crossover exists in all systems where a variety of complementary categories is applied to the same groups, persons or objects. The interplay of these categories in different contexts allows the creation of multiple alternative perspectives, or at least marks the contexts where alternative perspectives may apply.

None of the societies of eastern Indonesia appears to have perfectly concordant symbolic systems in which all groups or objects can be classified by categories that define them unequivocally in relation to other groups and objects. The discrepancies between different sets of complementary categories, extended too far for coherence sake, can only lead to evident contradictions. In eastern Indonesian societies, evident contradiction is often avoided by a continual process of mixing metaphors. Instead of allowing any set of complements to be extended to the point of contradiction, other complements are introduced to develop the system.

A brief contrast between Rote and Timor may help to illustrate this point. Rote shares with Timor the same complementary categories of male–female and outside–inside (though in the case of Rote, mane, ‘male’, is not linked explicitly to a word for ‘outside’). Nevertheless, the same asymmetries and associations hold true for these categories, and the potential for analogical reversal of the kind that has become elaborated on Timor is evident in Rotenese symbolic classification. The two halves of the Rotenese house, for example, are classified as male–female and outside–inside, the male half being the outside half. On the other hand, in the political realm, the ‘Male Lord’ (Manek), who is the ruler of the domain, occupies the centre or inside and the subordinate ‘Female Lord’ is relegated to the outside. Among the Rotenese, the inconsistencies and potential contradictions in these realms are muted, but a further problem arises in the classification of spirits which are categorised as belonging to two groups: ‘Spirits of the Outside’ and ‘Spirits of the Inside’. ‘Inside’ in this case is in reference to the house and, by analogy with other associations within the house, these spirits might conceivably be considered as ‘female’. This, however, is not done. Instead, a series of other complementary categories (east–west, firstborn–last-born) is commonly applied, and by various associations (inside = west = last-born), the last-born male, who inherits the house, assumes the role of guardian of the ‘Spirits of the Inside’.

The point I wish to make is that the possibilities for analogical crossover in these systems are innumerable, but the ones that are given cultural attention may be relatively limited. The Timorese, as well as the Tetun of Wehale, seem to have fashioned one analogical crossover into a mystery on which to found a potent political ideology (Fox 1983).

Conclusions

Before I venture some remarks on the relationship between dualism and hierarchy in eastern Indonesia, it is essential to make clear what is being discussed when these terms are invoked. Dualism, for example, is defined in the Shorter Oxford English Dictionary as a ‘twofold division’ or ‘twofold condition’. Since this definition neither specifies the relations that may hold between dual entities nor makes precise the coherence that may pertain within a dyadic division, it does little to elucidate the complex classificatory phenomena that are generally referred to under the rubric of dualism, particularly in the ethnographic literature on eastern Indonesia. The same might be said of the use of the term ‘hierarchy’, which tends to be used to describe a variety of social phenomena.

Dumont, in his discussion of the Indian caste system, has attempted to give a more precise definition to the term ‘hierarchy’. He defines hierarchy as ‘the principle by which the elements of a whole are ranked in the relation to the whole’, and he goes on to link this ‘principle of hierarchy’ to a single ‘opposition between the pure and the impure’ (1970:66). In turn, this opposition is analytically interpreted as a relation between ‘that which encompasses and that which is encompassed’ (1970:xii).

As useful as this definition of hierarchy may be, an attempt to apply it in eastern Indonesia is problematic for several reasons. The societies of eastern Indonesia do not have the encompassing religious coherence that Dumont has attributed to India; for this reason, hierarchy cannot be described as a single principle nor identified with a specific opposition, such as pure and impure. In eastern Indonesia there is a variety of contending oppositions that are of considerable importance to the definition of hierarchy and it is not one opposition but the interplay among various oppositions that gives rank to elements of a whole in relation to the whole.8 In these terms, hierarchy is not a principle but an outcome, the result of the application of several principles. By the same token, it is equally problematic to base a conception of hierarchy solely on the analytic distinction between the encompassing and the encompassed. Apart from the fact that the logic of this distinction is, at times, elusive, this is not the only distinction by which ‘hierarchy’ can be generated.

It is possible, however, to derive another lead from Dumont. In a crucial passage in Homo Hierarchicus, Dumont refers to hierarchy as ‘an order of precedence’ (1970:75). Adapting this phrase for eastern Indonesia, we may consider hierarchy as consisting of various orders of precedence. The issue, then, is to examine the way in which these orders of precedence are categorically structured using prevalent dyadic resources.

The hierarchical use of dual categories depends upon the conjunction of two analytic features: recursive complementarity and categorical asymmetry. With these features, a single set of dual categories can serve as the operator to produce an ordered sequence or graded series. As an example, we may consider the set of dual relative age categories elder–younger, which in eastern Indonesia is generally relied on to define relationships between persons or among groups that are regarded as sharing some aspect of common descent (see Fox 1980:331). Thus, among the Rotenese, Timorese or Tetun, these same-sex categories can be used to distinguish a graded series of status segments within a clan, lineage or birth group. Since a segment in the younger category may be elder to another segment, this series may be represented as follows:

elder	>	younger
		elder	>	younger
				elder	>	younger

A similar series based on male–female may be used, as among the Timorese, to define an order of precedence between wife-givers and wife-takers:

male	>	female
		male	>	female
				male	>	female

In eastern Indonesia the complementary categories used to create ordered series are the same categories that serve as operators for the system as a whole. Besides elder–younger or male–female, such categories as wife-giver–wife-taker, left–right, inside–outside or trunk–tip are particularly prominent. In a remarkable analysis of the ceremonial domain, Wai Brama, of the Ata Tana Ai of central east Flores, E. Douglas Lewis (1982a) has shown how a single line of precedence based on ‘origin from the source’ constitutes the means for a coherent ordering of all clans and segments of the domain.9 In a somewhat different context, a single royal descent line serves as the ultimate line of precedence for the hierarchical ordering of lineages and clans in the domain of Termanu on Rote. In this case, however, the ordered series is not structured by a common set of complementary categories but by a precise and rigidly maintained succession of ancestral dyadic names, each of which—after the first name—utilises a component of the preceding name (Fox 1971:42–7).

This leads to the final point I wish to make in this essay. Dualism is a prevalent conceptual resource in eastern Indonesia and as such it may be used either as a major vehicle for the structuring of hierarchy or as a counter to it. In another essay (Fox 1979), I have examined in detail the political and economic differences between the hierarchical structure of the domain of Termanu and the moiety structure of the domain of Thie. In analysing these two domains on Rote, it would be impossible to claim that there was more use of dualism in one domain than in the other. Both domains give evidence of elaborate forms of dualism. In Termanu, however, dual categories are utilised socially to form lines of precedence that foster status rivalry, extend alliance relations, and perpetrate patterns of relative dependence. In contrast, in Thie the primary conceptualisation of the domain is a dichotomy into moieties that are associated as male–female. Thus, for Thie, dualism occurs at the highest order of social classification and, though qualified by other dyadic divisions, this primary dualism has systemic implications for the whole of the domain’s system of classification. Any tendency to form lines of precedence always confronts a primary duality that undermines it. As a result, in Thie a systemic dualism serves as a counter to hierarchy.

We may therefore conclude with the observation that it is not dualism per se that defines societies with so-called dual organisation but rather the use of dualism at a general, systemic level, which thus determines the parameters for other forms of classification.