A theory of caste must offer a way of ordering the facts in such a way that it does not diminish the significance of some or ignore others. It must also be comparative. Caste organization is found in some parts of South Asia but not all. Equally, structural parallels may be found in many other parts of the world and one should not therefore assume that the defining characteristics of caste are unique to Hindu communities or to the ideology of Brahmanism. What is needed is a theory which explains why all of the traits associated with caste are found together where and when they are, whether in South Asia or elsewhere.
Various theories of caste are reviewed in this chapter before coming to the conclusion that one of these makes much more sense of the historical and ethnographic evidence than the others. Most theories depict castes as arranged in a ladder-like vertical order. Sociologists have tended to emphasise this ‘stratification’, regarding the ideological and ritual manifestations of caste, such as the pervasive concern with purity and impurity, as epiphenomenal. Anthropologists have generally avoided this error but have faced other intractable problems. Some see caste as a recent colonial ar- Social Evolution & History, Vol. 1 No. 1, July 2002 140–170
2002 ‘Uchitel’ Publishing House
tifact, others as an ancient indigenous category. Many are heavily influenced by the ideological reductionism of Dumont’s theory of Hindu society. On the other hand, Dumont’s analysis raises so many problems that some have attempted to retreat from theory and restrict their studies to ethnographic description.
The argument here is that caste results from an uneasy stalemate between the pull of localised lineage organization and the forces of political, ritual and economic centralization encapsulated in monarchical institutions. Caste systems are the product of a certain degree of centralization which involves the organization of ritual and other services around the king and dominant lineages. The central institution is (as Hocart suggested) the kingship. The removal of Hindu kings in India with the advent of colonialism does not negate this thesis, for it is a specifically western view of kingship which allows for only one monarch within a territory. Kingship (and the configurations of castes associated with it) was always reproduced at the courts of lesser chiefs, and is still replicated today in the households of members of dominant castes.
There is a considerable gulf between the ways in which sociologists and anthropologists typically look at caste1. Sociologists tend to regard caste, in a regrettably truncated fashion, as an extreme form of social stratification which is a kind of sociological opposite of class: ‘Open and closed stratification systems are sometimes described by the terms class and caste’ (Chinoy 1967: 178)2. Frequently equated with Hindu social organization through the misleading epithet ‘the Indian caste system’, sociologists usually home in on the fact that where there is caste, ‘an individual’s social position is fixed at birth and cannot be changed’ (Giddens 1989: 735). When sociologists use the concept to refer to situations outside of India, this is normally where racial segregation occurs, as in the southern states of the USA following the abolition of slavery or the more contemporary situation in South Africa3. What is common to both cases, it is argued, is the cultural insistence on maintaining the ‘purity’ of certain groups by establishing inviolable boundaries, particularly with reference to intermarriage.
Such a perspective is, however, of limited value. In the first place, the kind of ethnic separation referred to above normally concerns a very small number of groups, usually two or three, whose opposition to each other is crude and straightforward. In a typical caste system, however, it is not unusual for a score or more castes to be bound together in very complex relationships4. Moreover, the kind of ethnic cleansing which we associate with former Yugoslavia is undoubtedly also motivated by a concern for ‘purity’, as is the Mediterranean ‘honour and shame’ complex, but we would never think of describing either as a caste system.
To characterise caste systems baldly as forms of stratification is also to leave out most of what is really intriguing about them. In particular it tells us nothing about the seemingly endless flow of ritual and ritual prohibitions which preoccupy all members of caste-organized communities in their everyday activities, for example those revolving around the preparation and consumption of food. Nor does it account for the existence of untouchability, an institution which causes many egalitarian Westerners, and increasingly Indians themselves, to condemn caste out of hand as a barbarous institution which has no place in the modern world.
A further difficulty with the stratification approach is the implication that all castes are arranged, one above the other, in a relatively unambiguous way like a football league table. This is often coupled with a somewhat distorted view of the two Indian concepts most closely associated with the word ‘caste’, namely varna and jāti.
The varna consist of four categories, each ranked differently in terms of social honour. Below these four groupings are the ‘untouchables’, those in the lowest position of all. The jati are locally defined groups within which the caste ranks are organized... Those in the highest varna, the Brahmins, represent the most elevated condition of purity, the untouchables the lowest. (Giddens 1989: 213).
This simple description, while accurately reflecting the most common conceptions of caste both among Hindus and among those who comment on them, contains a number of misrepresentations. In the first place, the concept of varna refers to four basic social functions (usually defined as priesthood, kingship, generating wealth, and providing service) and not to kinds of social groupings. Jātis, which are groups based on kinship and marriage, are not subsets of varnas, any more than English people with the surname ‘Smith’ are a subset of people who work as smiths.
Secondly, it is problematic to say that Brāhmans are the highest varna for a number of reasons. There are thousands of Brāhman castes (jātis) whose members continually dispute each others’ status. If some of them enjoy greater prestige than others, as they themselves always state, the criterion for this must be something other than assimilation to the Brāhman varna. Another difficulty is that those Brāhmans who are priests are often seen as subordinate to those whose rituals they perform and whose status they thereby legitimate: paradigmatically, the kings and landowning nobility who are identified with the ksatriya varna. Priests are often portrayed as vessels or scapegoats who, through ritual, take away the sin, evil and death of their patrons. Parry’s work on the Mahābrāhmans of Benares (especially 1980, 1986) provides one of the best examples of this and is supported by the work of Raheja (1988a) and Levy (1990) among others5. In any case, only a minority of Brāhmans actually work as priests so it is not self-evident that the status of all Brāhmans is determined by this particular function, as is often claimed.
The implication that there is an automatic correspondence between varna and jāti is also unwarranted because there is often a certain amount of dispute about which varna a particular jāti should be identified with. Parry’s representation of the caste hierarchy in Kangra in north-west India is a particularly good illustration of this. Here, he says, a number of castes whose varna is perceived by others as śūdra see themselves as either ksatriya or brāhman (Parry 1979: 110, Table 14). Such discrepancies between self-representation and representation by others are a common feature of caste-organized communities.
In short, then, the conventional portrayal of caste, as one might find it in a standard textbook of sociology, or, one should add, as one might find it in the accounts of many Hindus themselves, is riddled with problems. Anthropological theories of caste have also run into a number of difficulties. There is widespread agreement that the most important theoretical statement on the subject in recent times has been Louis Dumont’s Homo Hierarchicus (1980) and a glance at any bibliography on caste will immediately reveal that the influence of this book has been unparalleled. Yet Dumont’s theory has been the focus of sustained criticism for more than a quarter century, to the point where it is no longer clear how much, if any, of it remains tenable.
Some anthropologists have tried to sideline the theoretical problems by sticking to what they can actually observe on the ground during prolonged periods of fieldwork, as if ethnographic description and theoretical abstraction belong to mutually exclusive zones. At an international workshop on ‘Caste Today’ held in London in 1993, for example, the implications of recent ethnographic and historical research for the theory of caste were markedly absent from virtually all of the workshop’s deliberations6.
Where theory did appear, as in the question of the relation between caste and colonialism, the central issue was not confronted. Some claimed, following the deconstructionist fashion, that caste was a relatively recent colonial artefact deriving from the classificatory obsessions of census-makers. Others argued, or implied, that caste is an ancient Hindu principle enshrined in classical Hindu texts. It cannot, of course, be both. In other publications, three of the contributors to Caste Today have starkly illustrated the differences of opinion about this crucial fact regarding the explanation of caste:
Caste has existed for thousands of years: Sanskrit literature provides us with irrefutable proof of this.
(Deliège 1993: 9; my translation from the French).
Despite the many changes which historians are now beginning to bring to light, a certain correspondence prevailed for nearly two thousand years between the actual division of society into castes and subcastes and what people considered to be right, proper and desirable from the social point of view... In the traditional order, caste was an integral part not only of Hindu law but also of Hindu religion. (Béteille 1992: 16).
It is increasingly clear that colonialism in India produced new forms of society that have been taken to be traditional, and that caste itself as we now know it is not a residual survival of India but a specifically colonial form of civil society. (Dirks 1992: 59).
Whichever of these views happens to be true, it should be obvious that it is impossible to divorce sociological theory from ethnography and history: even to introduce the word ‘caste’ is to imply a particular way of cutting up the world which relies on certain theoretical presuppositions. More obviously still, if anthropologists working in different regions are to compare their findings, there is no option but to turn to theory. As in the study of any social institution, to be comparative is to be theoretical and to be theoretical is to be comparative.
Why should there be such difficulty in providing a theoretical explanation of caste when an enormous amount of ethnographic and historical evidence on the subject has been produced over the last forty years or so? There are, I believe, a number of interrelated reasons for this which can be summarised as follows:
1. Most theories of caste appear to involve an unjustifiably arbitrary selection of evidence7. Since it appears to many that it is the facts themselves which are inconsistent, a common approach has been to ignore those elements which are awkward and to present the allegedly overwhelming picture suggested by those facts which are retained. The objection to this is that the awkward facts still remain, even if they are hidden.
2. While Dumont’s Homo Hierarchicus has been criticised endlessly, many continue to argue, or imply, that his basic premises are undeniable (see, eg., Deliège 1993). Given the difficulty of constructing an alternative theory using the same premises, or of finding alternative premises from which to begin, the majority of recent commentators on caste have preferred to concentrate on particularistic historical or ethnographic studies. However, this retreat from theory has not been any less problematic since, whatever is presented, some selection or prioritising of material is inevitable: it is impossible to write about caste without betraying an endorsement of one theoretical position or other.
3. The claim is often made that caste is unique to either India or Hinduism, thus making comparison with other social forms elsewhere difficult, if not altogether undesirable.
4. Caste is seen by some as being quintessentially a traditional village phenomenon while others see it as a colonial invention which exists more in the minds and classificatory needs of imperialist foreigners than it does as an ethnographic or historical reality. Apart from being mutually contradictory, both of these perspectives are too limited in their purview to provide an adequate explanation of the range of phenomena which are associated with caste.
The remainder of this article will explore these problems. I will argue that a consistent theory of caste is not as elusive as it has often been made to appear, and that the difficulty is not with the facts, but in finding a way of ordering them which does not diminish the significance of some or ignore others.
Discarded theories of caste
Before coming to Dumont’s theory of caste, which deserves special consideration, let me briefly mention four other approaches to the subject which should be rejected. In fact all of them have been dismissed many times before but they have an insidious way of resurfacing and two of them are still in popular circulation. (By ‘popular’ I mean both in the minds of many who practise caste and in the minds of many of those who study them).
The first theory which we can dispense with is that caste is a product of race, although an important qualification must be added before we throw out the idea altogether. The connection between caste and race was made by some of the earliest outside commentators on India who related it to the Aryan invasion of India and to the fact that members of Brāhman castes were often light-skinned while members of peasant and low castes were often dark-skinned or displayed other ‘aboriginal’ traits8. There is some basis in reality for this observation, but it is of limited value for an explanation of caste. Over the centuries certain populations were undoubtedly subjugated by others and the historical remnants of this can still be seen clearly in many places. But for the most part there is no obvious connection between caste and racial characteristics and in any case, the complexity of caste systems cannot be explained by a marker as crude as race. The often made comparison with racial segregation does not explain why one should need such a plethora of castes in any locality.
The second theory, though false, is still widely believed. This is the idea that there is an inherent connection between caste and occupation which explains how caste systems work in general9. There are two reasons why this connection is not a sufficiently good guide. The first is that it is never the case that all members of a given caste perform a particular occupation. A very common situation is that a particular caste is associated with a particular occupation: for example, being a priest or barber or potter, but that many (perhaps even all) members of the caste are agricultural labourers, or work in nearby towns as clerks or rickshaw drivers or whatever, or are employed by the military in some more or less distant garrison.
So while a couple may be referred to as Mr. and Mrs. Potter (i.e., they are designated as members of the local Potter caste), they may in fact have never made a pot in their lives. This will not deter others from saying about them: ‘Our sons cannot marry their daughters because they are Potters and we are Tailors [for example]’ even though the Tailors in question have never sewn cloth but are in fact minor civil servants in a local government office, or members of a musical band performing Hindi film music at weddings and other social occasions. And, by extension, the same statement will be made by members of other castes. The general principle, then, is that a Mr. Potter need not be a potter, a Mr. Brāhman need not be a priest, and so on.
The second reason why the correspondence between caste and occupation is a poor guide to the working of caste is that many people who perform the same occupation belong to quite distinct castes. This is particularly obvious with relation to two classes of work: agricultural labour and priesthood. Given that the vast majority of the population of the Indian subcontinent work on the land but are divided into thousands of castes, it is obvious that this occupation does not, by itself, provide an indicator of caste membership. It is also evident that many groups perform priestly activities but see themselves as completely separate from each other and different in caste status. (For examples at either end of the Indian sub-continent, see Levy 1990 and Fuller 1984). In fact it is quite common for some priestly groups to regard others performing ritual activities as kinds of Untouchables.
A third theory which runs into difficulties is favoured by some comparative sociologists. This revolves around the idea that caste is fundamentally about dominance and exploitation and any talk of pollution concepts as the criterion for distinguishing between castes is mere ideological obfuscation. On this argument, Brāhmans are at the top and Untouchables are at the bottom not because of their respective degrees of purity, but because of their respective degrees of material or economic power. The work of Gerald Berreman is one of the best examples of this kind of argument though there is a similar, if more subtle, version in the writings of the Indian sociologist André Béteille10. The reason that this explanation does not hold is simply that economic power and ritual status do not always coincide. There are Brāhman castes whose members are very poor and there are castes which are seen as low or even Untouchable whose members are in fact quite wealthy in comparison with the rest of the population11. There also frequently appears to be some kind of in-built mechanism which inhibits any easy conversion of acquired wealth into higher caste status. Against this it must be said that conquering groups always enter a local caste system with ksatriya, or kingly, status even if they were previously regarded as ‘barbarians’ or outside the pale of the local caste system altogether.
Perhaps the strongest argument against the idea that there is a straightforward correspondence between caste status and economic position is that if this were so, it is doubtful if we would notice anything distinctive about caste systems or have any reason for using the word ‘caste’ in the first place.
A fourth theory which has little to recommend it is the kind of argument advanced by E. R. Leach (1960) in an essay entitled: ‘What do we mean by caste?’ Leach argued that in caste systems every group has its place, and the element of competition among them had been largely removed, resulting in harmonious, integrated systems. While caste systems are often well integrated, this does not mean that competition for status is lacking12. On the contrary, the institution of hypergamy in north India, which is widespread among landowning castes, is nothing other than a competitive marriage strategy where the goal is to improve one’s status at the expense of others. Other examples of refusing to accept one’s place include the construction of false genealogies (Shah and Shroff 1958), name changing in order to make it appear that one ‘really’ belongs to a higher caste (Rosser 1966), moving to another locality in order to assume a new identity (Caplan 1975), and conversion from Hinduism to Christianity and Buddhism (Isaacs 1964, Juergensmeyer 1982). In short, the idea that castes always meekly accept their status position is unfounded.
What remains of Dumont’s theory of caste?
Another argument against seeing caste as a form of social stratification is that this perspective smuggles in a modern, individualistic, ‘Western’ set of values which is inappropriate where traditional, ‘holistic’ values prevail13. This is one element of Dumont’s position which seeks to restore the indigenous values of caste to their rightful place. Dumont claims that two oppositions form the ideological basis of caste. The first of these is between purity and impurity which, he argues, is manifested most clearly in the location of Brāhmans and Untouchables respectively at opposite poles of the caste system. The second opposition is between (ritual) status and (secular) power and this shows itself, he states, most clearly in the respective positions of the brāhman and ksatriyavarnas in the ancient Vedic texts, or in the respective positions of priest and king (or priestly caste and dominant caste) on the ground. There is, says Dumont, an ideological disjunction such that power is subordinated to status, the king to the priest, because it is through religious values that the whole system gains its meaning, and the Brāhman priest is the repository of these values.
At first sight this way of looking at things seems appealing because it appears to account for the endlessly repeated alleged ‘fact’ that Brāhmans are the highest caste. However, on closer inspection the concept of Brāhmans being supreme because of their superior priestly purity (Dumont 1980: 56) is difficult to sustain. As I have already mentioned, there is a widely held belief that priesthood is a defiling activity, the idea being that through rituals which are performed for a patron the officiant takes away the patron’s impurities leaving the patron pure, at least momentarily. Some have argued that it is really inauspiciousness which is being absorbed by the priest rather than impurity and that one ought to draw a distinction between these two concepts (Raheja 1988a). I will not pursue this argument here though the concept of inauspiciousness is indeed more apposite in explaining how caste relations are structured. The reasons for this will become clearer in the discussion of Hocart's theory of caste below.
The fact that there are many different kinds of priest, not all of whom are Brāhman by caste, presents another problem. Men of Barber caste, for example, often perform ritual activities for lower castes which are strictly analogous to those performed by Brāhmans for higher castes (Parry 1979: 59). And, as Raheja (1988a: 20) has pointed out, affines are sometimes called to perform priestly duties where a Brāhman, or a priest from another caste, cannot be found.
Most intriguingly, there is a category of Brāhmans, in north India referred to as Mahābrāhmans (literally ‘great’ Brāhmans), whose function is to perform funerary rituals. This is a particularly despised group because they are seen as being permanently defiled by the death pollution they take on themselves (Parry 1980: 94). The problem for a theory of caste which would have Brāhmans at one pole and Untouchables at the other is that the Mahābrāhmans are seen as both at the same time! Their ethnographers tend to dispel the ambiguity by arbitrarily assigning them either very high or very low status but this really amounts to a sleight of hand (see Quigley 1993: 81). Moreover, the same apparent paradox applies in a much more understated way to the members of all those castes, whether Brāhman or other, who work as priests. It is no coincidence that Brāhmans who follow some profession other than priesthood widely regard themselves as superior in caste status to Brāhmans who function as priests (see, e.g., Fuller 1984: 59). Dumont’s opposition between pure and impure, which he equates with the opposition between Brāhman and Untouchable, is, then, suspect though there is no denying that Hindus are constantly preoccupied by pollution concerns and this must be explained.
What of Dumont’s other opposition, between status and power, or between priest and king? The main difficulty with this concept is that it only works at the level of ideology, and then only if one is extremely selective and listens to certain voices and not others (see also Burghart 1978). Dumont’s thesis depends primarily on the idea that ‘power in India became secular at a very early date’ (Dumont 1980: 76) but this is simply not sustained by the evidence, whether from texts or from ethnography. Kings always retained a central position in the rituals they patronized and their functions were replicated on a lesser scale by the well-to-do members of dominant castes, a situation which has not changed with the collapse of Hindu kingship in the face of colonialism. The argument has been put succinctly by Raheja14:
Kingship no longer exists, but it has been, perhaps, replaced by the ritual centrality of the dominant caste as it has been described in the Pahansu [village of her fieldwork] ethnography. There, as in many of the textual traditions on kingship, the Brahman is hierarchically superior, yet the dominant landholding caste stands at the center of a complex ritual organization that permeates nearly every aspect of the everyday life of the village. (Raheja 1988b: 517).
Dumont is well aware that his theory does not apply to the observable facts on the ground: ‘In theory, power is ultimately subordinate to priesthood, whereas in fact priesthood submits to power’ (Dumont 1980: 71–2). But the solution to this apparent contradiction, he insists, is to understand caste as a structure of ideas, not as something which can be grasped by looking at particular territories (Dumont 1980: 154). The difficulty with this argument is that Dumont is simultaneously asking us to consider an actual place (India) which exists in space and time: we are asked to be empirical when it suits his theory and something less than empirical when the facts appear to contradict it.
In spite of these difficulties, there is a legacy from Dumont’s theory which students of caste ignore at their peril, and some recent alternative interpretations of caste such as those of Klass (1980), Hall (1985), and Baechler (1988) are seriously flawed in this regard. Among the most important of Dumont’s ideas are the following:
• The ‘holism’ of caste systems is in direct contrast to the individualism which is the dominant ideology of the modern West.
• One cannot talk about castes in isolation, only in relation to other castes. This means that one must have some kind of theory of a system of relations. Also, since this system is repeatedly found in a very large number of localities, there must be an underlying structure which gives rise to this.
• The relation between priest and king (or priests and dominant castes) is central.
• The connection between priesthood and inauspiciousness is also central.
• Given the endogamous character of castes, the role of kinship and marriage obviously requires careful attention.
The Basic Elements of Caste Organization
Perhaps the most widespread idea of all about caste is that it is exclusive to Hinduism and this is why it is found only in India or in places where Hindus have migrated. This idea is impossible to counter if, following Weber (1958: 29–30), caste is defined in terms of one’s position relative to Brāhmans. But virtually all of the institutions which one associates with caste are found in different degrees in other societies at different periods of history. The most obvious of these are:
• recruitment to one’s social position at birth;
• kinship organization in terms of lineages;
• differentiation between noble (or kingly) lineages and others;
• endogamy such that marriage tends to be within a restricted group of lineages;
• pervasiveness of ritual as a mechanism for structuring social relations;
• pollution concepts which place an ideological stress on the purity of women, or of lineages, or of kings, or of priests;
• monarchical institutions, whether material (palace complexes, monuments to kings and royal deities), social (courtly lineages and royal retainers), or ideological (royal rituals, chronicles);
• untouchability and scapegoatism.
To explain caste is to explain why and when all of these traits are found together when only some of them are found elsewhere. To begin with the premise that all of the defining characteristics of caste are unique to India or to Hinduism, as has been far too uncritically accepted from at least Weber on, is to be much too selective in deciding what it is that defines caste and to deny the possibility of any kind of fruitful positive comparison with other places and other eras.
Even if one were to avoid groups which have converted from Hinduism for political reasons, one could still find South Asians who practise caste but profess a different religious identity – Muslims in Pakistan (Barth 1960) or Buddhists in Nepal (Gellner 1992) for example. One must also recognise that there are Hindus who argue that caste is not part of their religion. And there are Hindus who claim to have nothing to do with caste but who clearly maintain its existence – for instance, by refusing to eat or intermarry with people they deem beneath them.
Béteille (1991, 1992) has argued that the Indian urban middle classes are moving away from caste and that ‘family’ is becoming the more important institution. Apart from the fact that he presents virtually no evidence for this claim, it is clear that there are still limits to who is regarded as an acceptable marriage partner. By and large Béteille seems to be referring to the intermarriage of members of Brāhman and dominant castes which formerly would have been separated territorially and politically but now find themselves thrown together in all of the institutions of a modern state. While the barriers to intermarriage may be coming down among these groups, this does not mean that they will marry into other castes which they continue to regard as inferior. In any case, to say that ‘family’ is now becoming more important is not of itself to indicate any sea change since family pedigree is precisely what caste has always been about. As Hocart put it: ‘Castes are merely families to whom various offices in the ritual are assigned by heredity’ (Hocart 1950: 20). I will return to this idea below.
It is obvious that caste continues to play a significant role in the contemporary political arena, but among educated people this is widely seen as shameful and completely inappropriate in a modern, democratic state. Béteille argues that ‘[it] may safely be assumed that in India today, everyone is prepared to speak publicly in support of equality, but none in support of hierarchy or inequality’ (Béteille 1991: 3). This conclusion flies in the face of a huge body of recent ethnographic material. While it is undoubtedly correct to point out that there is an increasing distaste for the values of caste among the educated middle class, the institution itself does not appear to be in decline. In the last two decades alone, scores of ethnographies from all over South Asia have made clear that caste continues to be the bedrock of social organization for hundreds of millions of people. While there is perhaps increasing ambiguity surrounding the ideology of caste, it would appear that inequality, far from being regarded as invidious, continues to be seen pervasively as normal, inevitable, even ‘natural’.
This is not to deny that there have been significant changes in the ways in which caste manifests itself during the past century. For example, where caste blocs appear today, uniting previously disparate groups, this is a modern development which would have been inconceivable in the not very distant past. A good illustration of this is found in the mass conversions to Buddhism stimulated by Gandhi’s contemporary and rival Ambedkhar15. Himself an Untouchable, Ambedkhar sought to raise the status of other outcastes by breaking their attachment to Hinduism. But this could never have been achieved without the apparatus of modern communication, transport networks, and education, the bases of which were laid down during the period of colonial rule. And it could not have happened if the British had not decapitated the myriad local systems of political allegiances which underpinned pre-colonial caste systems by establishing new foci of political legitimacy and identity. Given the opportunities to organize collectively, Untouchables and others did so with a vengeance. It is not that they lacked the will to improve their position under the old régime; more often than not, political and economic structures deprived them of the means, primarily literacy and mobility16.
This does not, however, mean that caste is paradigmatically a product of the traditional Indian village. One of the most entrenched fallacies in the literature on caste, this notion stems from a distorted colonial view of India as a land of village republics and from the fact that most detailed ethnographic work on caste has been done on village communities which are of manageable size for a field researcher working alone17.
The mistake has been fuelled by a frequent failure to consider the historical and ideological relationship between village and town, the former being, literally, a reflection of the latter. In an article which draws attention to this mirroring quality, Pocock argues that both in theory and in fact the traditional Indian city stands for completeness. It is a microcosm of the cosmos and provides the most complete expression of moral values and social order because it is the locus of maximum caste activity. It is also the place where the king is to be found, the main function of the king being ‘the maintenance of caste order’ (Pocock 1960: 66). Rowe echoes this when he writes that:
The distinguishing mark of a town or city in the ancient texts... was that only there did one find all the castes resident... The founder of a village or petty kingdom of several villages... fulfilled the same role but on a minor scale. And the village socio-economic system... reflected the arrangements of the city on a lesser scale. (Rowe 1973: 213).
Perhaps the best extant examples of this connection between caste and preindustrial urbanism are to be found in the social organization of the Newars of the Kathmandu Valley in Nepal, the seat of a complex, urban civilization for more than a millennium. One reason for this is that the ‘Valley’, in reality a circular bowl surrounded by Himalayan foothills, is extremely fertile and while the majority of the population have traditionally been agriculturalists, the area has always been capable of supporting a large population many of whom could be engaged in occupations other than producing food. As a gateway between India and Tibet, the strategic location of the Kathmandu Valley also made it an important trading centre. Over the centuries, the combined wealth of commerce and agriculture was converted into impressive temple, palace, and domestic architecture as well as being channeled into the competitive patronage of ritual which required a plethora of religious specialists.
Moreover, unlike India, Nepal was never colonized by either the Muslims or the British. In the small city-kingdoms of the Kathmandu Valley, caste has flourished for centuries among both Hindus and (Mahāyāna) Buddhists, with Hinduism, as Gellner (1992: 55) observes, being the religion of the rulers.
While there is not space in this short article to give even a potted description of the complex history and sociology of the Kathmandu Valley, one can point to some of the main features which underpin its caste organization18. These same features are at the basis of caste everywhere, even if not always in such a clear, or complete, form.
Any description of caste organization must rest on two preliminary observations. The first is that to be a member of a caste, one must first be a member of a lineage for it is groups of lineages which we recognize as castes by their agreement or refusal to intermarry, interdine, or perform certain rituals together19. The second observation is that caste systems are only possible given a certain measure of territorial centralization in regions which are capable of sustaining a complex form of social organization – river basins and other fertile areas are typical. One does not normally find caste systems in deserts or high mountains and it is no accident that in Nepal, for example, caste organization typically evaporates at approximately 4,000–6,000 feet above sea level where the land becomes progressively infertile and populations begin to depend on pastoralism to augment their meagre agricultural resources.
Caste, I would argue, results from a kind of uneasy stalemate. On the one hand, there is the pull of the lineage, institutionalised in the various ways in which lineage members exert control over one another – by restricting marriage choices, in the observance of periods of mourning when fellow lineage members die, by taking part in each other’s life-cycle rituals (marriage, caste initiation, and so on), and through periodic collective worship of lineage deities. On the other hand, there are the forces of economic, political, and ritual centralization, encapsulated in monarchical institutions. Among these, the most striking are the often spectacular palace-temple complexes and rituals such as Dasaĩ which are traditionally linked to royal power and which provide a means by which lineage and caste divisions can be transcended as communities come together in common devotion and celebration20.
If lineage principles were so strong as to inhibit any kind of developed centralization (as in the paradigmatic forms of segmentary organization), then clearly caste could not take root since it depends on a complex division of labour which must be based on something other than kinship. If, on the contrary, centralization were to be so effective that recruitment for the performance of various social functions was always based on principles other than kinship, then obviously caste could not flourish either. The weakening of caste among urban professionals in modern India reported by Béteille, though exaggerated, certainly indicates this.
I have already pointed out some of the dangers in representing the order of castes in any local caste system as an unambiguous perpendicular ladder with Brāhmans at the top and Untouchables at the bottom. Many of the difficulties in constructing an adequate theory of caste seem to me to be the result of being imprisoned by this ladder-like representation which one is led to almost inescapably if one starts with the notion that every caste can be said to be ‘higher’ or ‘lower’ than every other caste. An alternative representation which does not give rise to this problem focuses rather on the fact that caste systems are relatively centralized forms of political organization: Figure 1 gives a simplified model of this21.
Figure 1. The general structure of caste systems
This model is of course grossly simplified, and particularly in two respects. Firstly, it is not only dominant caste households which are able to attach other castes to themselves. To a greater or lesser extent, households in every caste will attempt to replicate this pattern by using their resources to employ members of other castes (or sometimes other, usually affinally related, lineages within the same caste) to perform various services for them. Obviously, the greater one’s resources, the greater will be one’s capacity to do this. But virtually no household is so poor that it cannot at least occasionally afford to retain others to perform specialised ritual functions – at funerals, weddings, or caste initiation ceremonies for example.
Secondly, the model might give the impression that each caste is an undifferentiated bloc when the reality is more complex. Different lineages and different households within each lineage may vary substantially in their wealth and this may have important ramifications in terms of internal status differentiations within the caste. For example, the formation of de facto sub-castes is not uncommon when certain lineages attempt to set up more or less exclusive marriage circles within the caste and begin to call themselves by a particular name which sets them apart from their erstwhile caste fellows22. Over time these groups may become recognised by some as separate castes even though their de jure claims to such status may be seen as rather shadowy by others.
A number of other qualifications should be borne in mind when looking at the simplified model depicted in Figure 1. For example, castes vary enormously in size. Typically, the largest castes are the dominant landholding groups and those who provide the bulk of the agricultural labour while other castes which provide specialist services are frequently very small. While the members of most castes, because of their attachment to the land, will see themselves as belonging to one caste system only, others, particularly merchants, may well function in more than one community and may thereby attain a certain degree of autonomy from dominant landowning castes.
It would be impossible to present on a single, flat page a model which showed either the ideological or physical distance of different castes from the centre. Some castes, or some lineages within certain castes, will legitimately claim to be nearer to the (royal) centre either in terms of their kinship pedigree (perhaps some among them have married into noble lineages), or in terms of the privileges they enjoy. The castes which supply the king’s priests will undoubtedly claim higher status than the castes which supply the farmers’ priests. Untouchables are frequently represented as being outside the community altogether and in fact must often live physically apart from other castes. This is because their primary function is to act as scapegoats and to take out pollution (i.e. whatever threatens social order) to beyond the community’s limits. Finally, it is also possible that there will be others somewhere in the vicinity (renouncers, independent sects, members of other ethnic groups) who cannot be accommodated easily within the local caste system.
In spite of these qualifications, the underlying structure of caste organization is as Hocart (1950) portrayed it. The central institution for comparative purposes, he claimed, is kingship (and its associated ritual). For Hocart, castes are essentially ‘families’ which perform hereditary functions in order to ensure that the king and nobles remain free from pollution (1950: 17, 20). Whereas Du-mont claims that the ideological pivot of caste is the priest’s purity, Hocart argues that it is the purity of the king and nobles. There is another, more subtle, but very significant, difference between the two views which explains Hocart’s reasoning.
In Dumont's view, the priest is highest because he is pure. Yet even if we were to restrict this claim to the realm of ideology, Hindu views concerning the purity of priests are extremely ambivalent. As one eminent Sanskritist has concluded, if a Brāhman does not wish to compromise his purity, the one thing he must not do is function as a priest (Heesterman 1985: 38). For Hocart, however, it is not that the king and nobles are pure but that they shouldbe pure because they provide a model for others. From this perspective, the priest is an instrument or vessel who facilitates the king’s kingship or the noble’s nobility: caste organization, which requires some members of all non-noble lineages to provide ritual services, is a continual striving to make this possible.
Where Dumont argues that the underlying structure of caste is in the ideology, in fact the ideological representations made by Hindus are irretrievably contradictory. The common structure underlying caste systems is rather to be found in constraints given by kinship on the one hand and kingship on the other, both of which are set against a particular material backdrop which allows a territorially limited kind of centralization to develop.
It is not only the historical experience of Nepal which belies the claim of a growing number of scholars that caste emerged out of the colonial experience of India23. The removal of Hindu kings in India heralded the superimposition of new forms of political authority and new forms of political association. But the traditional institutions of kingship, on which caste organization depends, have proved much more resilient and remain, as they have been for centuries, encapsulated in the political and ritual centrality of dominant lanowning castes. Here it is partly the Western conception of kingship which is at fault for it tends to characterise kingdoms as having a single monarch. The traditional Hindu view has always been much more elastic, allowing for ‘little kings’ and ‘great kings’ as reflected in such epithets as mahārājadhirāj, ‘great king of kings’, which is still today the title of the king of Nepal.
No-one has done more to elucidate the concept of ‘little kings’ than Dirks in his historical study of the former princely state of Pudukkottai in the Tamil-speaking region of southern India. He argues convincingly that ‘until the emergence of British colonial rule in southern India the crown was not so hollow as it has generally been made out to be. Kings were not inferior to Brahmans; the political domain was not encompassed by a religious domain’ (Dirks 1987: 4). Unfortunately Dirks confuses representations of caste in the colonial and postcolonial periods, particularly those of Weber and Dumont, with caste as it actually was under British rule and has been since Independence. Paraphrasing Dirks, one might say that even at the end of the twentieth century, kingship in India is not as hollow as it is generally made out to be.