cultural contexts of ethnic differences
The objective of this article is to contribute to the development of analytical devices for dealing comparatively with cultural differences made relevant in systems of interaction. First, the strengths and limitations of a leading social anthropological perspective on ethnicity are considered. Thereafter, certain aspects of ethnicity in two so-called multi-ethnic societies, Trinidad and Mauritius, are described and contextualized analytically. Finally, a general classification of inter-ethnic contexts is suggested. The criterion suggested to distinguish contexts in this respect is the varying cultural significance of ethnicity. The proposed classification is not necessarily incompatible with certain other attempts to compare ethnic phenomena.
Although my point of departure is a concept of ethnicity which is relational and processual, that is what I call a formalist view, I nevertheless insist that an understanding of the scope of cultural differences must supplement an appreciation of the formal features of ethnicity - whether they are to be localized to the level of the social formation or to the level of interaction. In order to achieve this, I shall draw on Wittgenstein's (1983 ) concept of language-games.
An implication of the analysis of ethnicity, briefly discussed at the end of the article, is that a sensible concept of culture must depict culture both as an aspect of concrete, ongoing interaction and as the meaning-context for the very same interaction.
Conceptualizing ethnicity: A critical review
One of our most widespread errors as social theorists, and a difficult one to eradicate, is the reification of concepts. As regards concepts relating to ethnicity, the temptation to do so is highly understandable. Notably, any conceptual reification of "ethnic groups" as social (and sociological) entities possessing certain fixed cultural and organizational characteristics is positively encouraged by folk taxonomies and popular assumptions almost anywhere in the world. Academic reactions against such reifications began, with respect to research on ethnicity, in Leach's study of Kachin politics (1954), and in the studies of urbanization in the Copperbelt in the 1950s (among others, Mitchell 1956; Epstein 1958; Mayer 1971 ). An explicit and very influential statement was Barth's (1969:10-11) criticism of earlier writers' identification of ethnicity as a property of "cultural groups", their axiomatic view of culture as a fixed, monolithic entity, and their subsequent cultural determinism. Implicitly, Barth also rejected the concern of some cultural anthropologists with "national character" as simplistic, tautological and misleading. Focusing on types of social relations and processes between and within ethnic groups, Barth (1969), Eidheim (1971) and others developed a set of formal concepts for dealing with interpersonal ethnicity without initially reifying either a concept of ethnic group or a concept of culture. Their approach, although sometimes poorly understood, has greatly influenced contemporary studies of ethnicity. Through the work of Barth, Eidheim and others, the focus of ethnic studies shifted from group characteristics to properties of social process. From then on, formal concepts were available enabling students of ethnicity to discard unsatisfactory strategies of empiricist "butterfly collecting", to replace substance with form, statics with dynamics, property with relationship and structure with process.
The formalist approach, conceptualizing ethnicity as a type of social process in which notions of cultural difference are communicated, enables us to view ethnicity comparatively, and to account for ethnic phenomena without recourse to crude conceptions of "cultures" and "peoples". It has moreover proven more flexible, and capable of higher theoretical sophistication when dealing with complex contexts, than a related approach in which ethnicity is reduced to a kind of stratification system, or in which ethnic process is virtually by definition reduced to group competition over scarce resources (Despres 1975; Cohen 1974b). Such reduction prevents full understanding of the discriminating characteristics of social systems where the communication of cultural differences is essential to the reproduction of the system.
For all its merits, the formalist approach associated with Barth (1969) has two important limitations preventing a satisfactory comparative understanding of ethnicity.
First - and this is nowadays a common criticism (O'Brien 1986; Wolf 1982; Worsley 1984; Fardon 1987) - it is in principle ahistorical. Its very useful, highly abstract comparative concepts such as ethnic boundary (Barth 1969), dichotomization/complementarization (Eidheim 1971), symbolic form and function (Cohen 1974a), and so on, indispensable in accounting for ethnicity on the interpersonal level, divert analytic attention from the wider social and historical context and thus implicitly disregard processes taking place beyond the grasp of the individual agent. For one should never neglect, or even "bracket", the fact that ethnicity is always a property of a particular social formation in addition to being an aspect of interaction. Variations on this level of social reality, moreover, cannot be accounted for comprehensively through studies of interaction, no matter how detailed they may be. For instance, ethnicity involving a modern national state is qualitatively different from ethnicity activated in a neighbourhood because a state and an individual are different kinds of agents. In addition, the context of interaction is constituted prior to the interaction itself and must therefore form part of the explanation of interpersonal processes. This implies that we ought to investigate the historical and social circumstances in which a particular ethnic configuration has developed, and a subsequent localization in time, place and social scale of the ethnic phenomenon in question must follow. A concept of power distinguishing between individual and structural power is essential here. Moreover, these findings are bound to influence our analysis, and should not be bracketed, even - or perhaps particularly - if the ultimate goal is a reduction of social process to a formal comparative model of ethnicity. On the other hand, historically bounded studies of ethnicity and related phenomena (e.g. Anderson 1983; Smith 1986) usually fail to account for the reproduction of identity on the level of interaction, and have limited comparative scope.
Secondly, and partly by implication, it can be misleading to consider ethnicity simply as an "empty vessel" or a system of arbitrary signs, or a form of deep grammar. Certainly, the "critical focus of investigation" ought to be "the ethnic boundary that defines the group, not the cultural stuff it encloses" (Barth 1969:15, italics in the original) - that is, ethnicity is a property of relationship, not "the sound of one hand clapping", to paraphrase Bateson. It is further doubtless correct that ethnic distinctions can persist despite insignificant differential "distribution of objective [cultural] traits" (Eidheim 1969:39), and that the symbolic articulation of cultural difference can frequently be seen to change in form and content, historically and situationally. Nevertheless, the cultural specificities or differences invoked in every justification of ethnic differentiation or dichotomization may (or may not) have a profound bearing on the experiential nature of ethnic relations themselves. This implies that the medium is not necessarily the message, and that the differences themselves, which represent a level of signification conventionally glossed over by the formalists, should be investigated, and not only the form of their articulation. In other words, if there are contextual imperatives for the production of ethnic signs - and it would be foolish to suggest otherwise, then the contexts in question must be understood along with the acts of inter-ethnic communication.
The cultural differences referred to in ethnic interaction, then, cannot always be reduced to its form without a loss of analytic comprehension. Since culture is such a difficult term to handle analytically, and since one of the main insights from formalist studies of ethnicity is that culture cannot be treated as a fixed and bounded system of signs, it is tempting to reduce or disregard this level of social reality in description and analysis. The most common (tacit) reduction of culture has consisted in showing how ethnic signifiers may change due to changes in context, thereby indicating that the signifiers themselves are really arbitrary, and that the fundamental aspect of ethnicity is the very act of communicating and maintaining cultural difference. This is the position advocated by Leach (1954), who emphatically states:
"Culture provides the form, the 'dress' of the social situation. As far as I am concerned, the cultural situation is a given factor, it is a product and an accident of history. I do not know why Kachin women go hatless with bobbed hair before they are married, but assume a turban afterwards, any more than I know why English women put on a ring on a particular finger to denote the same change in social status; all I am interested in is that in this Kachin context the assumption of a turban by a woman does have this symbolic significance. It is a statement about the status of the woman." (Leach 1954:16)
This type of argument has been very illuminating, but it is unsatisfactory in the end because the cultural context of an act of communicating distinctiveness may, as correctly assumed (and experienced) by non-anthropologists, make a systematic difference in inter-ethnic encounters. At a certain point in the analysis of ethnicity, where recognized cultural differences shape or prevent meaningful interaction, or where power asymmetry distorts discourse, it becomes impossible to neglect substantial features of social, cultural, historical contexts. Although the formal relationship between say, the Canadian state and Mohawk Amerindians may be similar to that between say, the Botswana state and Basarwa (San) people, the social and cultural significance of the respective relationships differ because of important differences in the cultural contexts referred to in the ongoing invocation of differences. This implies that formal modelling of ethnicity may miss the point not only because it leaves out aspects of ethnicity which are important to the agents, but also because it disregards the potentially varying importance of cultural differences in the articulation of ethnicity.
Handelman's (1977) typology of ethnic incorporation, ranking ethnic groups or categories from the socially very loose to the socially very strongly incorporated, has similarly limited explanatory power. It is misleading insofar as it treats ethnic categories or groups as analytical entities. This will not do: it is necessary to account for the production and reproduction of ethnicity in a less abstract, less static way in order to understand its concrete manifestations. Any detailed analysis of ethnicity must therefore take into account the varying cultural significance of ethnicity, not only cross-culturally, but also intra-culturally and perhaps most importantly, intra-personally. Different inter-ethnic contexts within a society, which may or may not involve the same sets of persons, have variable significance in relevant ways. Ethnicity, as a source of cultural meaning and as a principle for social differentiation, is highly distributive within any society or set of social contexts involving the same personnel. Its varying importance, or varying semantic density, can only be appreciated through a comparison of contexts, which takes account of differences in the meaning which are implied by the acts of communicating cultural distinctiveness which we call ethnicity.
It is my contention, therefore, that the cultural contexts of ethnic differences should not be ignored in description and analysis. This perspective carries a further, and obvious, ethical implication: it entails taking the representations of ordinary "lay" agents seriously. For the people whom we designate as members of ethnic groups tend to disagree with anthropological accounts dealing with themselves. An example of such disagreement is the ongoing conflict between Saami (Lappish) students and non-Saami staff at the anthropology department of the university of Tromsø, Northern Norway. Many of the students would prefer that their struggle for cultural identity and social autonomy were not dubbed by anthropologists as manipulation with symbols, "mechanisms of boundary-maintenance" or "processes of metaphorization and metonymization" which can ultimately be reduced to political strategy (cf. Grønhaug 1975; Thuen 1982). Native views cannot replace analysis, but they should not be reduced to universal form either. In sum, if ethnic signs are seen as metonymic-metaphoric signifiers for ethnic difference, then we should pay some attention to the signified - which can be studied on two levels. On the one hand, ethnic signs signify the communication of cultural difference, which has been studied thoroughly by social anthropologists - either as competitive strategies, or as a technique for the maintenance of a cultural identity or a way of life, or both. On the other hand, ethnic signs refer to systematic distinctiveness which is in part being reproduced outside of the acts of communicating distinctiveness.
This level of signification has normally been bracketed by formalists, with whose general approach I agree. Their comparative analytical models are clearly superior to earlier theorizing about ethnicity; they are simple and comparative, they account for empirical processes, and they have great explanatory power regarding the reproduction of social and cultural discreteness. I shall merely propose to supplement this set of concepts with concepts enabling us to compare the forms of distinctions reproduced, and to distinguish between contexts where the kind of cultural difference at work varies qualitatively. This can be done without reifying a concept of "culture", provided that the units for analysis are not groups or individuals but contexts of interaction.
The discussion concerns fundamental issues in social anthropology, and I am not able to pursue them much further here. Nevertheless, let me summarize the discussion so far: Modern social anthropological studies of ethnicity have tended to overstate their case in contrasting the virtues of their formal, processual model for analysis with the vagueness and reifying tendencies of substantivist approaches. Their sometimes unacknowledged semiotic bias (notably, the view of ethnic symbols as arbitrary signs) prevents them from elucidating the varying impact of significant cultural difference and the content of meaning in interaction (cf. Wilden 1980:352n). For the sake of comparative models of interaction, they have largely disregarded the study of ethnicity (or ethnicities) as specific sets of features in the contemporary world (cf. Giddens 1987:165 for a similar point with regard to the study of organizations).The formalist direction in ethnic studies therefore needs to be supplemented by a consideration of two theoretical points: First, ethnicity is a property of a social formation and an aspect of interaction; both systemic levels must be understood simultaneously. Secondly, ethnic differences entail cultural differences which have variable impact cross-culturally, intra-culturally and intra-personally, on the nature of social relations.
A treatment of the relationship between the systemic level of interaction and the systemic level of social formation, necessary in the final analysis when the validity of ethnicity as a comparative concept is to be assessed (cf. Fardon 1987), falls outside of the scope of this article. The ethnographic examples and contexts to be discussed below illustrate the second theoretical point; namely, that the cultural differences which are confirmed in the communication of ethnic differences, vary between contexts which may otherwise be comparable, and that this variation should be understood in accounts of ethnic processes.
Ethnic differences and language-games
The problem of accounting for "actual" cultural differences in ethnicity is a subtle and difficult one. The communication of cultural difference can be observed and described, but the cultural differences referred to in these acts are themselves elusive. They cannot be measured. Since culture is not a fixed property of persons or groups, the differences cannot be identified as "cultural traits" of the agents either. I have nevertheless argued the need to distinguish between kinds of cultural difference invoked in ethnic contexts, where the distinctions are confirmed and reproduced. Such differences, I suggest, can be identified through a careful interpretation of these contexts. I propose to use the Wittgensteinian concept of language-games to distinguish, in a formal way, between inter-ethnic contexts where the degree of shared meaning is variable. In his initial discussions of language-games in the Philosophical Investigations, Wittgenstein defines a simple language-game as "a primitive language" (Wittgenstein 1983 , I: §7, cf. §2). With this he refers to an intersubjective field delineated to a particular context, which is reproduced by the individuals interacting. To engage in social interaction thus amounts to playing a language-game, that is the conventional application of certain concepts or the enaction of certain rules which define a particular version of the world as relevant. The rules of a language-game, when they are eventually internalized, thereby constitute a cognitive matrix of varying complexity, through which one perceives and interprets events taking place in the world. If the agents interact according to different such sets of concepts or rules in a given situation, this means that they propose to play different language-games; in other words, their respective delineations of the relevant meaning-structures of the world differ in this regard. They have acquired different knowledge about the world, and they reject each others' proposed rules (cf. Wittgenstein 1983, I: §47).
Some inter-ethnic encounters can be interpreted as contexts where such differences are expressed, and they can therefore not be reduced to competition over scarce values, since a competitive relationship necessarily entails prior agreement over what is to count as a value. The concept of language-games, as it will be used here, indicates the local, contextual character of culture seen as the production and reproduction of shared meaning. As I shall show, such a concept also makes it possible to distinguish between forms of cultural differences as expressed in ethnicity, without lapsing into essentialism or statements about the ontological properties of "cultures". This I will now demonstrate through an examination of different contexts in two societies where ethnicity is important and of varying significance as regards degrees of shared meaning in interaction.
Trinidad and Mauritius
The area of Trinidad is about twice that of Mauritius, but they sustain populations of comparable size (slightly over 1 million). Both are tropical islands and former British colonies where the French influence has been (and in the case of Mauritius, still is) substantial. Both experienced plantation slavery on a large scale until the 1830s, and subsequently received substantial numbers of Indian immigrants under the colonial indentureship scheme. The demographic compositions of the islands have important similarities: the main ethnic categories, as depicted in national statistics and in folk taxonomies, are blacks, Indians, Chinese, Europeans and culturally ambiguous categories of phenotypically "mixed" people. Indians, most of them Hindus and Muslims, are the most numerous in Mauritius; whereas blacks, who are as a rule Christian, are about as numerous as Indians in Trinidad. There are several ethnic subdivisions which may be relevant; for example, the distinction between Hindus and Muslims can in both societies be important, and there are ethnic categories which may be relevant in Mauritius but not in Trinidad (for instance, the category of Tamils) and vice versa. In popular representations and in public discourse, the most important distinction is nevertheless that between Indians and blacks. There are, of course, also non-ethnic social classifications which can be relevant; two such distinctions, important in both societies, are those obtaining between "the middle class" and "the working class" (local terms), and between rural and urban people.
Trinidad, an oil-rich island, is wealthier than Mauritius, and is to a greater extent than the latter integrated into wider systems of exchange. Both societies are parliamentary democracies, and both are changing rapidly, economically and institutionally. Ethnicity plays an important part in daily interaction in both societies.
Ethnicity in institutional politics
In both the island-states, most political parties from the post-war period have been organized along ethnic lines, and have derived their support from an ethnic base. However, ethnicity is rarely made an explicit issue in public, political contexts. Unlike class and gender, for example, ethnic distinctions are not officially acknowledged as fundamental in the organization of the societies, where prevailing ideologies and official discourse are, for all their mutual differences, vaguely nationalist. Whenever arguments based on ethnicity are invoked by a politician, other politicians publicly react in a hostile way. Likewise, it is normally considered rude to accuse politicians of following ethnic strategies; there are subtle and cunning techniques available to this end, not least because of the indexical character of ethnicity. Ethnicity is nevertheless frequently a master variable in Trinidadian and Mauritian non-politicians' accounts of politics. Two remarkably parallel examples will serve to illustrate this.
In Mauritius, a development scheme for sugarcane smallplanters was launched in the mid-eighties, cutting taxes and providing loans for smallplanters. In Trinidad, a similar scheme for would-be small businessmen began in 1989. Both of these governmental programmes were criticized along similar lines by members of politically nondominant ethnic categories. In Mauritius, non-Indians noted that most of the smallplanters were Indian, and that there were indeed many poor blacks for whom nothing was done. In Trinidad, non-blacks claimed that only blacks would benefit from the incentive programme, both because it was believed that most small entrepreneurs were black, and - more importantly - because blacks were thought to enjoy better connections with the authorities. Although such arguments never appeared unmasked in the media (except in letters to the editors), they enjoyed wide support.
In both cases, ethnicity created a division in public opinion; the ethnic boundaries, dormant in many social contexts, were activated. The wealthy, the poor and the destitute would rally together under one set of arguments or the other. The common view was that Mauritian Indians would, almost by definition, support the smallplanter incentive while blacks would oppose it; similarly, it was widely held that Trinidadian blacks would either be indifferent to or support the small business scheme whereas Indians would either be indifferent to or oppose it on the grounds that it would strengthen the presumed dominant position of blacks in Trinidadian society. The point here is not whether "all" or "most" blacks or Indians really acted in a certain way or held certain views; there is certainly variation in these regards, but we should note the fact that ethnic distinctions in this way became relevant in public discourse on a political matter. And the language of public discourse does not always take account of the empirical variations behind the stereotypes.
Moving to a higher systemic level, it has been documented extensively that institutional politics is largely organized along ethnic lines in both Trinidad and Mauritius (cf. Ryan 1972 for Trinidad; Simmons 1983 for Mauritius); this pertains both to voting and to the internal organization of parties and political organizations. As I have argued elsewhere (Eriksen 1988), it is impossible for a politically oriented individual to disregard the ethnic dimension in politics in Mauritius (this holds true for Trinidad also, although in a slightly different way), even if one might prefer to do so. In both societies, moreover, a standard reply to questions about ethnicity is that ethnic conflicts are "created by the politicians". This is wrong. Ethnicity not only plays an important part in non-political social fields in both societies, it is more fundamental outside the realm of institutional politics. This will be demonstrated below.
Several important contexts of political ethnicity in contemporary Trinidad and Mauritius, regarded as the ethnically based attempt to win political power, can profitably be interpreted as an integral part of the political systems. The same sets of rules are subscribed to by all involved in routine politics, and there is a wide consensus over values and modes of discourse. In other words, cultural differences are in themselves unimportant in these contexts; their importance lies in the creation of options for politicians and parties to draw upon such differences in their quest for popularity and power. The formal congruence of ethnicity among politicians of different ethnic membership is complete: the political culture, or language-game, is homogeneous as it is being confirmed in ongoing, institutionalized political life; there is no relevant cultural difference informing the pattern and meaning-content of institutionalized interaction between politicians. The rules which guide this form of interaction are under normal circumstances cross-ethnically uniform both among politicians and among voters, and the cultural differences influencing ethnic relations must therefore lie outside the realm of institutional politics.
In these multi-ethnic political contexts, then, ethnicity is constituted by a system of arbitrary signs in the sense that political process might have had identical formal properties in the absence of ethnic classification; substantive issues might have been different, parties and voting patterns would have been different and so on, but the set of rules constituting the political system would have been the same, all other things being equal. In other words, agents taking part in institutional politics in Trinidad and Mauritius are obliged to follow the rules of a shared language-game, and insofar as ethnicity is relevant in these contexts of politics, cultural difference is communicated through a shared cultural idiom; a shared language-game.
This should not be taken to imply that politics as such presupposes a shared language-game, neither in the societies in question nor elsewhere. The examples merely illustrate a type of contexts. Indeed, a hundred years ago, the incommensurability in political culture between Indo-Trinidadians and the colonial government was evident. Weller (1968) thus writes about the wave of wife-murders in the Indo-Trinidadian community during the late nineteenth century. In a court case cited by Weller (p. 65 ff.) the defendant says, "I have killed my wife. What is that of your business? I didn't kill anybody else's wife." Clearly, the defendant represents juridical principles incommensurable with those of the polity. There is here disagreement over the rules which constitute the judicial system as a field of interaction. The respective language-games are incommensurable; they refer to different versions of relevant social reality.
Language-games in the labour market
On the societal level, both Trinidad and Mauritius have traditionally sustained ethnically correlated divisions of labour. In both societies, agricultural labour has for over a century been culturally and statistically associated with Indians, the civil service and industrial labour with blacks, management with whites and coloureds, and commerce with whites and Chinese (in Trinidad: Chinese, Syrians and Portuguese). With a few exceptions, the correspondences between folk assumptions and the actual division of labour have been fairly close; the labour markets of both Trinidad and Mauritius have, nevertheless, changed radically in this respect since World War II. In both societies, the blacks were in colonial days perceived as being culturally closer to the European rulers than the Indians, and were therefore preferred in the colonial civil service.
In accounting for the presumed cultural proximity between blacks and Europeans in colonial times, one can point to shared language (French or English, as the case may be) and religion (varieties of Christianity), family organization (the nuclear family as ideal) and by extension, an assumed shared, modern perspective on the individual (encompassing, among other things, an assumed sympathy for the virtues of bureaucratic organization). Indians were in both societies depicted, by the colonial elites and others dominating public discourse, as illiterate pagans and particularistic, "clannish" (Trinidadian expression) schemers whose first priority was always the extended family. They were also treated for generations with indifference; they have not traditionally been part of the "Hegelian" system of opposites constituting the Black-Brown-White socio-cultural system in these colonies (cf. Fanon 1952), which created a mutual feeling of familiarity among those included. Although the division of labour has changed, the stereotypes remain strong.
The point to be made here is that the stereotypes referred to normative aspects of perceptions of cultural differences and can therefore not be reduced to arbitrary signs subjected to free manipulation. It is doubtless true that the perceived cultural closeness of blacks and Europeans provided for easier cooperation in the workplace, and that most blacks in both societies were in colonial times more "modern" than most Indians: they were better educated, and they spoke English and/or French. In other words, from the perspective of an employer, there might conceivably have been sound reasons for recruiting blacks rather than Indians for a variety of jobs. One might of course criticize the employers in question for not wishing to shed their stereotypes and consider aspects of individuals instead of stereotypic ethnic categories; however, there were perceived cultural discontinuities between the categories "blacks" and "Indians" which were relevant in the labour market. Language-games routinely reproduced by Bhojpuri-speaking Indian immigrants, and attributed to them as ethnically distinguishing "cultural features", were seen as being incompatible with the demands of the labour market. Skills relevant in one language-game, for instance in a rural village, were dismissed as irrelevant in another, dominant language-game. The cultural differences referred to are, incidentally, comparable to cultural differences currently being reproduced as ethnic (and class) boundaries in European cities, where immigrants and their children are regarded, by prospective employers and by public servants, as culturally inept or handicapped because some of their language-games are defined as incompatible with certain requirements of national society.
The informal policy, in certain state agencies, of debarring Indians from high public positions, practised in Mauritius until the late 1960s and still practised in Trinidad, might typically be interpreted and explained by social anthropologists as a way of retaining political power and control in society; it might be spoken of as a set of policies destined to retain ethnic hegemony. Such an account, tautologically true, is inadequate for two reasons. First, had the quest for power and control been the only reason for not hiring Indians, the latter would also have been discouraged from business ventures and other independent economic activity. This has certainly not happened in any of the societies; it was possible for Indians to purchase land freely and to set up shops already in the final decades of the nineteenth century. An examination of colonial records from the period (cf. Brereton 1979 for Trinidad; Allen 1983 for Mauritius) indicates that the economic ascendancy of Indians was not regarded as a threat by the elites - on the contrary, their presumed frugality and ethos of hard work were praised officially.
The second reason why an economistic analysis of the labour market is invalid, as already suggested, is that perceptions of cultural differences play a part in the organization of the labour market, so that their articulation cannot be reduced to political strategy. A few examples from the contemporary situation further illustrate the differences between kinds of cultural differences invoked in ethnicity.
A well educated and culturally self-conscious Indo-Trinidadian of my acquaintance resigned his job as a TV journalist. His reason for doing so was "that one was not allowed to be an Indian on TV". When asked why, then, there were after all quite a few Indians regularly on the screen, he replied that they were not "real Indians". He added, "For them [the broadcasting corporation] it's fine that you're Indian as long as it doesn't show." What did this mean? "For instance, you should never wear Indian clothes - you should always appear in shirt and tie, never in shirtjac. And if you make programmes, you should always remember that Trinidad is a black country. If you want to make a programme about something Indian, it's fine as long as you present it as something remote and exotic. To be an Indian in the media, you have to be a hundred per cent creolized." We might at this point arrive at the premature conclusion that my friend left his job because of disagreement over programming policy, and that his discontent was rooted in cultural difference - that he represented a language-game incompatible with that of the board. However, we need to go one step further. First, the cultural differences invoked are of a "weak" type: they do not include disagreement over the organization of media - the Indian might, in principle, just as well have been a black marxist (i.e., the distinguishing signs are arbitrary). Secondly, the Indians who remain TV journalists have not necessarily relinquished their ethnic identity; they have only suspended it in the job context. There is no incommensurability between my friend's stance, the stance of other Indian journalists, and that of the board of the broadcasting corporation; all they disagree about is the proper place for the communication of Indian-ness in Trinidadian society. They have a common language-game for communicating cultural difference, although their respective cultural contexts may constitute incommensurable language-games in other respects.
An expression of a stronger kind of cultural difference, also from the context of wagework, is evident in the attitudes of some black urban Mauritian women towards domestic work. One young woman explained the differences thus: "If you work for white people, they treat you reasonably. If you work for Indians, they might not pay you, they might make passes at you, they shout at you and treat you like a dog. Me, I'd never work for an Indian." This statement contains a great deal of negative stereotyping. It is supported by other women in her situation, and indirectly by wealthy Indians, who may make statements to the effect that "those black women are lazy, they never do their job properly and they always try to sneak out of the house before they're through." The disagreement can be traced to differences in perceptions of domestic organization. A housemaid in a Hindu family is in a position structurally similar to that of a doolahin, a daughter-in-law. The doolahin and the black maid are not family members, and it is therefore permissible to treat them as outsiders. Black women, on the contrary, are taught to expect that their mothers-in-law (and by extension, other outsiders) will treat them with respect, like adopted daughters. The lack of intimacy and the condescending treatment encountered in the Hindu family is therefore interpreted as a violation of their rights. The Hindu employer, on the other hand, regards the informality of his black maid as impertinent and as a sign of sloppiness. What he wants is a maid, not an adopted daughter. To the extent that there is disagreement over the rules constituting the relationship, the respective language-games are incommensurable.
Language-games, ethnicity and class
A yet different kind of conflict was related to me by a coloured middle-class Trinidadian who had been working in a development agency lending capital to cooperatives. His background, and the language-game of his organization, were "middle-class" and bureaucratic, emphasizing the virtues of investment and planning. The relevant language-game of his borrowers was that of black urban working-class Trinidad. His conclusion: "Most of the loans were never paid back. I was so shocked at the irresponsibility of those people! They would stake everything they owned, plus a considerable loan from us, and then they just didn't care if they went bust! After a few weeks, they might get fed up, they might not bother to work a lot, and would be unable to pay their loan. This happened all the time. I'm not saying that they're dishonest, I'm just saying that they're very, very irresponsible." My friend's experiences, while apparently lending some credibility to his stereotypic views of working-class blacks, may be interpreted as expressing a conflict between language-games. The agents involved in interaction had different expectations and disagreed concerning which values were to be defined as relevant. Their respective codifications of the cultural context which can be called the labour market, differed relevantly.
The example raises a problem with which I have not yet dealt properly, namely the relationship between different sets of criteria for social differentiation. The "ethnic" aspect of this situation is obvious; my informant calls attention to presumed aspects of black working-class culture. The situation also has a "class" aspect of which the participants are conscious. There are, of course, further contexts where incommensurable language-games are confronted and where the ethnic dimension does not provide a convenient label for its description. In the particular example referred to, the relationship between the ethnic dimension and the class dimension in the production of relevant differences is important. In the present discussion, this is nevertheless a minor point since the concept of ethnicity is not regarded as universally or ontologically significant, nor necessarily prior to other principles of social classification. It is easy to think of contexts in many societies, including the Trinidadian and the Mauritian ones, where the social classifications made relevant in the production of distinctions between individuals are not defined as ethnic by the agents. It is because ethnicity is empirically pervasive in these societies that it demands our attention. The presence of non-ethnic principles for social classification nevertheless indicates that the concept of language-games could profitably be used in analyses of cultural differences which are not codified as ethnic by the agents.
Why is ethnicity so important?
Like activities in politics and in the productive sector, family life and certain leisure activities in the two societies are routinely understood and codified in an ethnic idiom. However, the contexts of ethnicity encountered here may differ markedly from those reproduced in fields which are to a greater degree regulated by sets of formal rules. In routine politics, a shared language-game contains rules for competition over shared, scarce values; in the context of wagework, a similar competition is important although, as I have shown, not always sufficiently important to prevent the articulation of incommensurable language-games. It is nevertheless usually in informal contexts of interaction that ethnic differences can be regarded as expressions of incommensurable language-games.
Cultural differences between blacks and Indians are in both societies strongly articulated in matters relating to sexuality. The sexual ideologies of black men in Trinidad and Mauritius encourage promiscuity; to brag publicly of one's numerous achievements in this regard is an affirmation of black identity. In the ideology of Indian-ness, on the contrary, great value is placed on sexual purity in women, and the sacred character of matrimony is emphasized. In an Indian language-game, the supposed sexual prowess of black men is coupled with the widespread notion that women are unable to resist sexual advances. In this way, black men seem to represent a threat against the domestic supremacy of Indian men - and stories about faithless Indian women eloping with black men are so widespread in both societies as to be proverbial. When, in Mauritius, I asked black men about their views on extramarital sex, they might reply, giggling, that "it's not like in Europe" - meaning that it was a daily occurence. Indo-Mauritians, on the other hand, would usually be reluctant to talk about sex at all. Aids figures from Trinidad, incidentally, tend to confirm the folk assumption that blacks there on an average have a larger number of sexual partners than Indians: there is a striking overrepresentation of blacks in the official figures.
This kind of cultural difference is very important, even if practices do not necessarily conform with folk representations. The distinction suggests that varying representations of self and relevant others indicate, and reproduce, a relevant cultural difference as regards the most intimate of human relationships. Variations in the conceptualization of sexuality are in both societies indexically linked with ethnic labels. It is therefore widely assumed that inter-ethnic interaction in this area can lead to conflicts in the most personal of social fields. Despite generally cordial relations between people of different ethnic identity, intermarriage is rare in both societies.
The important point here is that what anthropologists regard as political ethnicity ("competition over scarce resources") cannot be fully understood unless an understanding of private ethnicity (immediate struggles) is first established. It is in the intimate contexts of family, close friendship and the like that the basic cultural contexts making up individual identity are reproduced. Only if one fully understands the reproduction of discrete, socially discriminating language-games at this level can one hope to understand why ethnicity can be fashioned into such a powerful political force within the unitary language-game of institutional politics. It is in such contexts that the language-games on which all communication of cultural difference feeds, are reproduced. Such contexts are also crucial in the transcendence of ethnic disctinctions; it is significant, thus, that popular national sentiment transcending ethnic boundaries in either society is perhaps never stronger than in contexts of international sports.
The formal systemic frameworks, in this case those of politics and labour, are thus fed with cultural distinctions on which they have a mitigating effect insofar as they represent shared desirable values, but which they neither autonomously create nor reproduce. Both Trinidad and Mauritius have recently (in 1986-7 and 1982-3 respectively) experienced concerted attempts to transcend the ethnic dimension in politics through the formation of broad nationalist coalitions. Following their rapid breakup (in Mauritius, the government lasted nine months, in Trinidad seven), the politicians and the electorate immediately fell back on an ethnic perception of politics, and its subsequent organization was related to such a perception - although not all the new alignments followed strictly ethnic lines. For instance, in Trinidad, the foreign minister Basdeo Panday was removed by the black-dominated government and replaced by another Hindu, Sahadeo Basdeo, who was nevertheless considered a less "rootsy" Hindu than the former. Ethnicity in this case proved empirically more fundamental than other principles of classification (in this case, nationalism). Ethnicity is in many contexts the single most important criterion for collective social distinctions in daily life; ethnic distinctions are rooted in perceptions of differences between lifestyles, and the others are held to represent lifestyles and values which are regarded as undesirable. As mentioned, cultural differences are sometimes activated in non-ethnic situations, such as rural/urban, middle-class/ working-class and male/female contexts. However, in these societies, one is never simply "male" or "middle-class": one is Indian male or Coloured middle-class. The ethnic dimension nearly always enters into the definition of a situation; it is an underlying premiss for all social classification. To the extent that agents routinely ascribe their own experiences of cultural incompatibility to ethnic differences, ethnicity also remains dominant as a principle for cultural differentiation. This, among other things, entails the maintenance of incommensurable language-games conceptually identified with ethnic differences.
Models of ethnicity and culture
What is it that we compare or ought to compare when we deal with different contexts of ethnicity? We should first bear in mind that discrete social phenomena are in themselves not comparable in any interesting respect. They may, however, seem to possess a family resemblance encouraging abstract conceptualization for the sake of comparison or generalization. Sceptical of some such conceptualizations, Fardon (1987) advocates severe constraints on the scope for comparison of ethnic phenomena. Less cautious or more ambitious theorists might claim that we compare, for example, mechanisms for boundary maintenance (and transcendence), structural properties of processes of incorporation, processes of symbolization in political competition, or minority strategies - all of which are considered phenomena with uniform and identifiable, formal properties. Major concerns have been problems of social and systemic integration in so-called plural societies, and to a lesser extent problems relating to the distribution of power in national states containing minorities. These concerns have dictated the development of concepts; granted that social reality itself is multi-faceted, our concepts about it depend on the questions asked.
A different approach borders on psychology in that it focuses on postulated "individual needs" for an ordered social world and the prevention of cultural entropy; in other words, the "mechanisms of boundary maintenance" are moved from the collective to the individual level, and the main dimension for comparison would be the manifest techniques for solving "perennial problems" (see e.g. Cohen 1974a; Nash 1989). Both types of approach, which are not mutually exclusive, take the cultural differences on which ethnicity is based for granted, and study them only as signifiers in inter-ethnic interaction. Since the relevance of cultural differences communicated through ethnicity varies systematically in important ways, I have suggested that we distinguish inter-ethnic contexts on such a basis, without committing the error of identifying cultural differences with properties of individuals or groups. In its simplest form, such a model might distinguish between contexts characterized by (i) one language-game, (ii) overlapping language-games, and (iii) incommensurable language-games.
The notion of a shared language-game implies agreement over constituting and strategic rules of interaction, and of course it goes beyond mere verbal language (Wittgenstein 1983 , I: §7; cf. Bloor 1983: 137-159). The agents understand each other when they are playing the same language-game. My examples of ethnicity as activated in routine institutional politics is an example of this. When language-games are partly overlapping, there is agreement as to the form and content of only some relevant aspects of the interaction. The example of the Mauritian housekeeper in an Indian home illustrates this: the terms of employment were agreed upon, but not the rank of a black woman in an Indian home. When, finally, language-games are incommensurable, interaction is difficult and its regulating rules will normally be defined by the most powerful agent. Misunderstandings and highly different definitions of the situation will be common. A territorial dispute involving a national state and an aboriginal population would be a typical example of this. In the material presented in this paper, the example about the moneylending agency and the "irresponsible borrowers" illustrates the same point on a more local scale - while also reminding us that there are relevant cultural differences in any society which do not necessarily have to be described as ethnic.
The classification is intended for the comparison of specific contexts, not complete systems. Agents who misunderstand each other when they discuss religion may well get along well when they talk about football; whatever certain linguistic philosophers might claim to the contrary, language-games are rarely (never?) entirely discrete, and they are not static. Language-games expand and contract through interaction, but certain perceptions of differences beyond the "difference of being different" can remain despite intense interaction. Ethnic categories can therefore sometimes constitute what Gellner (1983:64) has called entropy-resistant classifications. Empirically, they have hitherto been entropy-resistant in important respects in both Trinidad and Mauritius, since intermarriage is exceptional. This may of course change.
Important dimensions of ethnicity which have been left out of my crude typology are formal properties of process as well as structural and individual power, aspects of social integration and historical specificities. Fardon's important criticism of ambitious comparisons of ethnic phenomena (Fardon 1987), has thus not been dealt with fully. The typology proposed (which can be elaborated in several ways) is an attempt to operationalize the cultural contexts of ethnicity which, I have argued, have varying and sometimes crucial importance in the social organization of ethnicity.
Ethnicity and the concept of culture
The aim of this article has been to explore some new possibilities for research on ethnicity - and, by extension, research on culture. I have argued that ethnicity, as correctly suggested by the formalists, can be reasonably regarded as the collective enaction of socially differentiating signs. These signs, however, are not arbitrary because, unlike linguistic signs, they are intrinsically linked with experienced, practical worlds containing specific, relevant meanings which on the one hand contribute to shaping interaction, and which on the other hand limit the number of options in the production of ethnic signs. I have argued that variation on this level of signification, where incentives and constraints on action are reproduced, should be investigated in comparative analyses of ethnicity. For only when all participants are involved in the same language-game, are the ethnic signs truly arbitrary.
Only formal models of ethnicity are capable of producing comparative analyses. These very models are constructed from social facts which are themselves incomparable, and which differ in important ways. For instance, processes of dichotomization can be identified both in interpersonal Saami-Norwegian relationships and in Zulu-Afrikaner relationships, but there are important differences between the contexts of the encounters, shaping the interaction and the content and form of articulation of stereotypes. Clearly, we cannot neglect these substantial differences. I have suggested, therefore, a form of theoretical triangulation whereby we apply different sets of formal abstractions to the same material; one might label it, slightly pretentiously, a multidimensional model of ethnicity. Models for the comparative study of boundary-maintenance and ethnic incorporation should definitely be retained as analytical devices. I have argued that we also need a set of abstractions for dealing with variations in the degree of shared meaning in a given context. For it is impossible to understand say, different problems of different Fourth World populations, and conflicts between European host populations and Third World labour migrants, without taking seriously the differing cultural contexts (and semantic density) of manifestly communicated ethnic differences. To conceptualize shared meaning as shared language-games, I have argued, can turn such cultural differences into a matter of analysis, not just one of description or policymaking.
I believe this kind of argument, for which I claim little originality, to have profound significance for anthropological research and theorizing about culture. Let me, therefore, outline a few implications by way of conclusion. First, textual models of culture have limited value insofar as culture is an aspect of diverse social practices lacking the unity of a text as well as its structure and its form of internal logic. However, to deconstruct culture into the varying meaning-aspects of concrete social relationships apparently deprives us of the possibility of grasping culture as an interrelated system of signifiers - and this presents us with a problem. On the one hand, culture is neither a property of a person nor an integrated symbolic system; this is never more evident than in multi-ethnic societies where several competing, although overlapping language-games may present themselves as potentially relevant. But on the other hand, culture is a thing, une chose sociale, insofar as it provides necessary cues for all meaningful action. The predicament entailed by our deconstruction of a much misused term can only be resolved if social relationships are understood contextually yet not reduced to their contexts. We must, in other words, arrive at an understanding of culture which makes it impossible to talk of, for example, Norwegian culture, or Mauritian culture, or for that matter of Indo-Trinidadian culture, while at the same time not reducing culture to idiosyncratic individual actions. I have suggested that culture be conceptualized as a language-game; a learnt and internalized context of shared meaning bounded spatially, temporally and situationally, yet related to other such games through rules of translation and conversion, or through shared or continuous practices, personnel or other carriers of information. Within such a framework, it is possible to account for the communication of cultural difference on every level right down to the dyadic encounter without either reifying "cultural groups" as analytical entities or reducing the relevant differences to arbitrary signs; that is, without reducing agency to structure, meaning to politics, and creativity to normative behaviour. This use of the language-game concept thus makes it possible to delineate the extent of shared meaning in relevant contexts in a society, without presupposing either the existence of "an integrated culture" or that of independent individuals developing their mutual understanding from scratch. For the rules of language-games are learnt, although they are frequently modified by the agents.
Culture is, like social structure, dual: it provides a necessary frame within which action can be meaningful, and it depends entirely on intentional action for its reproduction. Conversely, culture is activated in all human relationships, while simultaneously a condition for the very same relationships to be meaningful. Intentional agency is therefore a necessary component in cultural change. It remains only to add that ethnicity is similarly constituted - and that its significance in any situation cannot be taken for granted.
Acknowledgements. Several people have read earlier versions of this article and commented usefully. Harald Eidheim and Kim Johnson helped clarify the analytical perspective; Man's two anonymous referees commented extensively and critically on a number of relevant issues, and finally, Tim Ingold helped improve both the argument and my English. Fieldwork in Mauritius (1986) and Trinidad (1989) was funded by the Norwegian Council for Science and the Humanities (NAVF).
Allen, Richard (1983)Creoles, Indian Immigrants and the Restructuring of Society and Economy in Mauritius, 1767-1885. Unpublished Ph.D. thesis, University of Illinois
Anderson, Benedict (1983) Imagined Communities. London: Verso
Barth, Fredrik (1969) Introduction, in F. Barth (ed.), Ethnic Groups and Boundaries. Oslo: Universitetsforlaget
Bloor, David (1983) Wittgenstein: A Social Theory of Knowledge. Cambridge: Cambridge University Press
Brereton, Bridget (1979) Race Relations in Colonial Trinidad 1870-1900. Cambridge: Cambridge University Press
Cohen, Abner (1974a) Two-Dimensional Man. London: Tavistock
Cohen, Abner (1974b) The Lesson of Ethnicity. In Abner Cohen (ed. ),Urban Ethnicity. London: Tavistock
Despres, Leo ,ed. (1975) Ethnicity and Resource Competition in Plural Societies. The Hague: Mouton
Eidheim, Harald (1969) When Ethnic Identity Becomes a Social Stigma. In Fredrik Barth (ed.),Ethnic Groups and Boundaries. Oslo: Universitetsforlaget
Eidheim, Harald (1971) Aspects of the Lappish Minority Situation. Oslo: Universitetsforlaget
Epstein, A.L. (1958) Politics in an Urban African Community. Manchester: Manchester University Press
Epstein, A.L. (1978) Ethos and Identity. London: Tavistock
Eriksen, Thomas H. (1988) Communicating Cultural Difference and Identity. Ethnicity and Nationalism in Mauritius. Oslo: Occasional Papers in Social Anthropology
Fanon, Franz (1952) Peaux Noirs; Masques Blancs. Paris: Seuil
Fardon, Richard (1987) 'African Ethnogenesis'. Limits to the Comparability of Ethnic Phenomena. In Ladislav Holy (ed.), Comparative Anthropology. Oxford: Blackwell
Gellner, Ernest (1983) Nations and Nationalism. Oxford: Blackwell
Giddens, Anthony (1983) The Constitution of Society. Cambridge: Polity
Giddens, Anthony (1987) Social Theory and Modern Sociology. Cambridge: Polity
Grønhaug, Reidar (1975) Transaction and Signification. Typescript, University of Bergen
Handelman, Don (1977) The organization of ethnicity. Ethnic Groups, vol.1, pp. 187-200
Leach, Edmund R. (1954) Political Systems of Highland Burma. London: Athlone
Leach, Edmund R. (1961) Rethinking Anthropology. London: Athlone
Mayer, Philip (1971 ) Townsmen or Tribesmen. Cape Town: Oxford University Press
Mitchell, J. Clyde (1956) The Kalela Dance. Manchester: Manchester University Press
Mitchell, J. Clyde (1974) Perceptions of Ethnicity and Ethnic Behaviour: An Empirical Exploration. In Abner Cohen (ed.), Urban Ethnicity. London: Tavistock
Nash, Manning (1989) The Cauldron of Ethnicity in the Modern World. London: The University of Chicago Press
O'Brien, Jay (1986) Toward a Reconstitution of Ethnicity: Capitalist Expansion and Cultural Dynamics in Sudan. American Anthropologist, vol. 88, pp. 898-906
Okamura, Jonathan (1981) Situational Ethnicity. Ethnic and Racial Studies, vol. 4, pp. 452-463
Ryan, Selwyn (1972) Race and Nationalism in Trinidad and Tobago. Toronto: University of Toronto Press
Simmons, Adele S. (1983) Modern Mauritius: The Politics of Decolonization. Bloomington: Indiana University Press
Smith, Anthony D. (1986) The Ethnic Origin of Nations. Oxford: Blackwell
Thuen, Trond (1982) Meaning and Transaction in Sami Ethnopolitics. Mimeo, University of Tromsø
Van den Berghe, Philip (1975) Ethnicity and Class in Highland Peru. In Leo Despres (ed.), Ethnicity and Resource Competition in Plural Societies. The Hague: Mouton
Weller, Judith A. (1968) The East Indian Indenture in Trinidad. Rio Pedras: Institute of Caribbean Studies
Wilden, Anthony (1980) System and Structure, 2nd. ed. London: Tavistock
Wittgenstein, Ludwig (1983 ) Philosophical Investigations, tr. G.E.M. Anscombe. Oxford: Basil Blackwell
Wolf, Eric (1982) Europe and The People Without History. Berkeley: University of California Press
Worsley, Peter (1984) The Three Worlds. London: Weidenfeld & Nicholson
©Thomas Hylland Eriksen 1991