Semiotics and Culture

Erik Schwimmer

Historically, semiotics and anthropology have developed independently. Semiotics was first conceived by philosophers and linguists, without any thought of solving problems arising out of man's cultural diversity. Conversely, anthropology and the study of culture have very diffuse origins. Those who contributed to its earliest development were sometimes philosophers, sometimes historians, geographers, biologists as well as sociologists, psychologists, while we cannot ignore the important amateur contributions of travellers, poets, artists, missionaries. Perhaps the most illuminating recent interpretation of the history of anthropology is that of K.O.L. Burrid ge, Encountering Aborigines (1973), in which anthropology is shown to be essentially a product of theological and philosophical questioning arising out of the European renaissance. Burridge thus, very appropriately, describes anthropology in a manner consistent with anthropological method: not as an area of objective knowledge as much as a mental activity inaugurated by certain occidental cultures in a specific stage of their development, an activity serving purposes specific to those cultures and to a specific period in time.

In considering the relationship between semiotics and culture, Burridge’s book is thus a useful starting point. It traces the history of western man’s reflections and investigations in one particular, but pivotal, cultural field, that of the Australian aborigines. It traces the history of the questions westerners asked about Aborigines, and the relation between these questions and the basic religio-philosophic and political issues of the centuries under consideration. For Burridge, this historically and environmentally determined set of questions (not a system but a set), this unorganized agglomeration, is the sum of anthropology.

Not all anthropologists would agree with such a view; nor do I entirely agree with it. Few, however, would fault Burridge’s description of the inventory. It leaves open several controversial questions: does the inventory form a system? If so, is it basically biological? Basically semiotic? Or again, is it an ideological system, without a counterpart in objective reality?

Our first question should be: taking Burridge’s inventory as our starting point, how is it related to semiotics? Secondly, if we survey the chief contemporary theories of anthropology, i.e., the various attempts inductively to make a system out of the inventory, or deductively to make a system using the inventory, how do such systems relate to semiotics? Our last, and perhaps most fruitful, question should be: what anthropological studies have opened up perspectives for semiotics?

Burridge (page 153) summarizes his inventory in a diagram in which anthropological activity comprises six stages of rationalization. At the first stage, we have those made by indigenous informants, who interconnect the inventory of actions and events of their own culture into a “homemade model.”At the second stage, the outside observer “has attempted to understand them in the only terms he has available: his own.“ At the third stage, the two initial models are related together: homemade models and investigators’models are made to “interact and modify each other.” When an anthropologist talks about his “data,” it is the rationalizations of this third stage that he is, according to Burridge, mostly referring to.

When Burridge outlines the three further stages of anthropological activity, his presentation becomes a little more controversial as it is anchored in the British functionalist school of social anthropology—it bases its theories exclusively on descriptive generalization. It is not the purpose of the present paper to confront this method of analysis with the axiomatic theory of some schools of anthropology (cultural ecologists, followers of Popper) or the transformational theory of some other schools (French structuralists, some “cognitive” anthropologists). Let us therefore merely say that, by one method or another, the “data” have to be critically examined for coherence and intelligibility, after which they become the subject of a “monograph.”^1. This monograph must then be subjected to comparison with other monographs so as to establish relationships between the various existing monographs and express understanding of cultural phenomena in a language that is apposite to a plurality of cultures, and—at least in principle—for all cultures. At some stage, such a language should be formalized and objectivized into a set of propositions about what Burridge calls “a priori relations, capable of containing all possible kinds of empirical relations.” The approaches of the principal schools have been summarized in the accompanying diagram.

Stages of Rationalization in Ethnography

First stage:

Indigenous informants report actions and events, and construct explanatory theories, and sometimes homemade models (to interrelate such theories): Text I

Second stage:

Initial theories of investigator: Text II

Hard data: observations and measurements by investigator: Text III

Third stage:

Synthesis of stages 1 and 2. Rationalization of the DATA as homemade models, investigator's models, and hard data “interact and modify each other”

Fourth stage:

Analysis of the DATA which are accounted for in the form of a MONOGRAPH, containing models which make the data coherent and intelligible, in the framework of a THEORY which might be, for instance: functionalist, or axiomatic, or transformational.

Fifth and Sixth Stages:

At what points in this enterprise do we find a relation with semiotics? Burridge’s first two stages—the construction of the homemade and the investigator’s models—are comparable to the starting point of a semiotic investigation, for both are in fact largely “texts.” The basis of the homemade model is a text written down, ideally, at the dictation of an indigenous informant, and the investigator’s model is based on his notebook of observations. The construction of the homemade model out of informants’ verbal statements is fundamentally a semantic exercise. The notebook will normally also contain a list of events recorded, objects described, and hypotheses by the field worker as to relationships between statements, events, and objects. Such hypotheses often lead to the recording of further statements, events, and objects, and notebooks should ideally indicate the logic of the investigation. One of the characteristics peculiar to such ethnographic texts is the intertwining of empirical and logical elements. From one viewpoint, informants’ statements and investigator’s observations are a disconnected set of notes; from another they are full of unstated logical presuppositions, as the statements are in general answers to questions asked (explicitly or implicitly) by the investigator. The observations are likewise answers to unstated questions, many of which are to be found in works of anthropological theory, or even in fieldwork manuals. The investigator’s notebooks, therefore, are texts suitable for semiotic analysis, and one may say that when Burridge talks of an investigator’s “model,” he assumes that the investigator has carried on his own semiotic analysis on the text to obtain his model.

By thus explaining the anthropologist’s fieldwork activity to semioticists, I am glossing over the unsystematic nature of most notebooks anthropologists have actually compiled. Partly they are unsystematic, no doubt, because fieldwork is a taxing activity and in practice the ethnographer has little time to think about academic niceties while he is recording data in the field. Partly also, they are unsystematic because it is only in the last few years that anthropologists have been able to look upon their fieldwork activity from an explicitly semiotic point of view. Indeed, this may have been an innovation of the ethnoscientists and cognitive anthropologists of the 1960s.^2. Even these last mentioned did not have the benefit of our present knowledge of semiotics; if we regard as semiotics only those operations explicitly derivable from Peirce or Saussure, there is little “semiotics” in anthropology: it developed independently.^3.

A very rigorous distinction must always be made, in analysis, between informants’ statements and investigator’s observations. The former are analyzable semantically—they are linguistic phenomena—whereas the latter, insofar as they describe actions and events, are analyzable in parts as semiotic texts. They have, however, a further dimension, namely a totally objective quality independent of the viewpoint of the investigator. If I describe, for instance, material culture, supporting my description by diagrams and photographs, and by measurements made with the aid of instruments, observer’s bias may for practical purposes be disregarded. Similarly, even in the description of events certain elements must be regarded as totally objective (timing, for instance, if I look at my watch).

Summarizing, we may say that in the first two stages of rationalization, we find semantic analysis of informants’ texts and semiotic analysis of investigators’ texts, but we must recognize that texts of the latter type contain many statements which are, at least in principle, unquestionable objective records about the culture under investigation. One may add that unless many such statements were recorded, the field worker would not be recognized as an anthropologist by his colleagues, and if he was trying to write a doctoral thesis, he might well fail his thesis examination because of deficiencies in his “data.” I am making this point in order to stress that a certain kind of objectivity is a strict formal requirement for entry into the profession, and therefore an integral part of the discipline of anthropology. This objectivity aspect is as important in anthropology as in medical semiotics where Shands (1975) uses the term “symptom” for informant’s (patient’s) statement and “sign” for objective facts about the patient’s condition. In contrast, a purely semiotic investigation of the field-notes (in terms of Eco’s semiotic theory) could well be made—could be made just as well—if there was not a single objective fact in the notebook.

The difference between semiotics and anthropology becomes clearer in what Burridge calls the third stage of rationalization, in which the homemade and the investigator’s models are made to “interact and modify each other.” For it is at this stage that we notice that there are discrepancies between informants’ statements and our objective measurements. Informants may state some rule of behavior which is often contravened in our record of events, or there may be conflicts in descriptions of the same events, offered by different informants. Our demographic data, our physical measurements, our seemingly unequivocal observations may call into question elements of the informants’ model, while again the informants’ model may lead us to question passages in our field record. Very often, the discrepancy is not merely due to error, but to our imperfect understanding either of events or of informants’ statements. At this third stage, therefore, an important semantic and semiotic activity is taking place, namely the critical examination of certain indigenous concepts when the investigator finds inconsistencies in his record due to his limited understanding of such concepts; and of explanations of behavior when these do not account for the established facts. At the same time, however, the third stage of rationalization does much to refine the objective records, the degree of conformity of the records to the facts about the culture. Very often, at this stage, discrepancies are discovered which require the systematic collection of a very objective type of data to test hypotheses. Thus, in my own investigation of plant emblems among the Orokaiva, I found that my theory about them did not explain the facts. I decided, in order to clarify the situation, that I should ask every adult in the village to give me a complete list of his/her plant emblems and of the parental plant emblems, including the third generation as well wherever possible. There is no doubt that I thus increased the degree of objectivity of my investigation. One cannot collect every fact imaginable, but one has to collect facts in an area of confusion as long as there is any chance that further facts will lessen that confusion. Again, this objectivizing activity is not part of a strictly semiotic investigation; it has to do with the descriptive requirements of a kind of natural science.

When we come to the establishment of coherence and intelligibility required in an anthropological monograph, we find that the culture we attempt to describe has to be interpreted simultaneously in two seemingly contradictory ways: as a system of signs, intelligible in its conceptual coherence, and at the same time as a viable economic, political, and social system, with contradictions which are managed by an appropriate superstructure. This kind of interdependence is formulated differently by the various schools of anthropology, but there will be little disagreement that anthropological intelligibility in principle incorporates physical as well as mental elements. The cultural ecology school and some narrowly Marxist anthropologists might hold that the mental is always and everywhere an epiphenomenon of the physical, but even so both kinds of elements (whatever their relationship) form an integral part of their work.^4.

Whether there are any recognized schools of anthropology who teach, or have ever taught, that mental elements in culture, i.e., systems of signs, determine its techno-economic infrastructure is most doubtful. Marvin Harris imputed such a viewpoint to Lévi-Strauss, and Robert Murphy (1972) appears to support such an imputation. For anyone who has taken a look at what the French structuralists are actually doing in anthropology, and who has observed the great similarity between their practical preoccupations and those of their colleagues across the Channel, the Harris-Murphy account of French “mentalism” and “Hegelianism” is almost a fairy tale. There is today very little difference between the French and British way of collecting field data, in other words their operations in what Burridge calls the first three stages of rationalization. These schools view infrastructure as an integral part of the data that have to be made intelligible and coherent in the preparation of monographs.

A more subtle question, relating to “mentalism,” might be raised in connection with some highly important American schools of anthropology, described variously as ethnoscience, cognitive anthropology, the new ethnography, or ethnography of communication. It is these schools that led Harris to give a central position to a distinction between the “etic” and “emic” study of culture, a distinction which was developed within the ethnoscience school itself, after a linguistic analogy (cf. Pike 1966). I do not myself think that the “etic”-“emic” distinction was intended by the ethnoscience school as the basis of any general theory of culture; their dominant preoccupation was plainly methodological.

The problem they raised is one which semioticists will have no difficulty in understanding. It is quite independent from any general question as to whether we may equate culture with semiotics, in other words, whether culture can be exhaustively described as essentially reducible to a system of signs. We have already seen that anthropologists do not normally work on such an assumption. If therefore Frake, Conklin, Lounsbury, Burling, Metzger held such a view it would have been somewhat revolutionary. Certainly nobody would have taken notice unless they had clearly stated it.

What they did say, and in much detail, was that anthropological fieldwork was greatly deficient in methodology. While they did not reduce anthropology explicitly or implicitly to semiotics, they did emphasize that semiotics is an important component of anthropology, i.e., that all field workers, as part of their task, are preoccupied with the analysis of semiotic domains in culture. Yet this tended to be done with variable degrees of rigor. Even British anthropologists of the Oxford school, traditionally most precise in the delineation of indigenous systems of thought, tended to think of the “understanding” or “translating” of such systems in anthropological terms as an art rather than a science. I doubt whether the work of the various schools of the new American ethnography was essentially better than that of the Oxford school, but it was certainly more systematic in describing its method, hence more “teachable” under the conditions of the American university system. Typical examples of “new ethnography” are analyses of color categories, “categories of eating,” the diagnosis of disease, the use of kinship terminology. Rules for the conduct of specific rituals such as weddings or offerings to supernaturals were painstakingly elicited and checked against behavior by a systematic question and answer method. The methods of eliciting data and analyzing the resulting texts were basically semiotic. Yet these practitioners could not and would not claim that they were thus analyzing the whole of culture; their method applied only to specific domains.

Certain theoreticians belonging to this school, of whom I would mention especially Goodenough (1970) and Keesing (1971), went one step further and believed that culture could be equated to a summation of a large but finite number of semiotic domains. As they argued, the domains can be analyzed so as to intersect, or in practice always do intersect, so that as the number of domains increases, and especially if one chooses the domains with such intersections in view, finally somehow the whole of culture can be accounted for. I must confess that such a mechanical construction of culture out of semiotic domains feels to me like a deus ex machina theory; I would not be inclined to defend it against Marvin Harris' assaults.

By no means all ethnoscience practitioners have followed the lines of Goodenough and Keesing; I shall summarize their various attempts below. Thus Berlin (1970, 1971) attempted no less than “a universalist-evolutionary approach to ethnographic semantics.” Having analyzed a specific semiotic domain (plant taxonomies) in various cultures, he posited some universal characteristics of plant taxonomies and some variations that would hold universally in accordance with various stages of evolution. I have seen one detailed test of the theory, in some British Columbia Indian cultures, and this test did not support the theory very well (Nancy Turner, n.d.). Nonetheless, it is clear that once semiotic domains have been analyzed (which is necessarily done by “emic” methods), the resulting model can serve almost any theory of culture the investigator happens to hold. In anthropology, semiotic analysis is a useful, often indispensable, method, but it does not presuppose or entail any specific theory of culture.

We have seen that there is little support for a reduction of anthropology to a domain of semiotics. It is true that Lévi-Strauss, in 1960, made a suggestion to this effect, building on an earlier stray remark by Saussure (Lévi-Strauss 1973: chap. I), but as over the last decade or so semiotics developed in France as a serious discipline (dominated by figures like Greimas, Barthes, Kristeva), it soon became clear that the approaches were fundamentally different. As Lévi-Strauss was, during the 1960s, mainly preoccupied with myth analysis (Lévi-Strauss 1964, 1966, 1968, 1971), it is there that we can see the diverging of the paths, especially with regard to the treatment of text corpora. While Greimas (1970), Bremond (1973), and the others analyzed their texts purely from evidence internal to the closed corpora, using ideally all the evidence the corpora contained, Lévi-Strauss and others of his school (as well as Hymes [1971] and a few others in the United States) introduced issues of culture history, ecology, social relations, establishing contradictions an ethnologist would find in the cultures. Such contradictions would enter freely into the interpretation of the texts. Clearly Greimas and Lévi-Strauss have entirely different ideas as to the kind of intelligibility that should be produced by analysis.^5. It seems therefore most convenient to abandon Lévi-Strauss’s earlier suggestion and distinguish between anthropological intelligibility (as in Lévi-Strauss) and semiotic intelligibility (as in Greimas).

One might conceivably argue that one ought to proceed in steps by first analyzing texts by purely semiotic techniques, and then try a later stage to clarify why a particular culture, at a particular point in time, elaborated the system we have established by semiotic analysis. Thus one could divide neatly into a semiotic and a purely anthropological stage of a cultural investigation. In practice, however, it does not generally happen that purely semiotic techniques make even a specific domain of culture intelligible as a system.

Thus, Bulmer found in an ethnozoological study of the Karam of New Guinea (1967) that the carefully elicited Karam rules for the classification of animals did not predict the primary taxon in which an important game animal, the cassowary, should be included. The only way in which this anomaly could be made intelligible was by introducing data on the ecological, social, economic, and ritual levels of the culture. The symbol “cassowary” is an integral part of many Karam semiotic domains, and its position in one domain is determined in part by its position in the others. Certainly one can start in one domain, but ultimately the anthropologist cannot abandon his role of detective, following whatever clues, in any cultural domain, will clear up the problems raised by anomalies.

This brings us to the key question how we can conceive of the relationship of semiotic domains and culture. I have already referred to the cultural ecology school which views meaning as produced, fully and directly, by ecological and technological factors. In other words, in that school of anthropology it is not necessary to inquire how people think about what they do, as their actions, and their thoughts as well, are determined by ecological and technological factors alone. Clearly, this kind of theory is non-falsifiable as it is always possible to construct some techno-ecological explanation to account for any action. We need not consider it further here, as the cultural ecology school places semiotics outside the field of anthropology, except insofar as the production of meaning (ideology) can be apparently accounted for by attaching to it some techno-ecological explanation.

Other anthropological schools have rarely, whatever their theories, in practice managed to avoid semiotic analysis, and some have explicitly advocated it, though mostly not as the ultimate objective of their discipline. A crucial question that concerned anthropologists was how they should regard indigenous informants' statements of “belief.” One of the great ancestors of social and cultural anthropology, Tylor, held that “the beliefs of savages are the result of attempts to understand natural facts.” This view implied that the analysis of indigenous statements of such beliefs would yield valuable insight. Tylor’s view led to several decades of anthropology during which vast quantities of such statements of belief were collected, compared, compiled, and organized in works, the most famous of which has remained Frazer’s Golden Bough.

The British functionalist school was a reaction to this kind of exegesis. Radcliffe Brown, its chief theoretician, compared the mental processes underlying indigenous statements of belief to “those that are found in dreams and in art” (1933). Hence, he did not believe that by analyzing them he would learn anything about social reality, except inasmuch as myths provide “charters” for social action. Thus, functionalists considered it fruitless to ask what a myth meant, but useful to recognize what institution it justified, legitimized, was a “charter” for. If British social anthropologists have made a considerable contribution to our understanding of the “semiotics” of culture—and I think they have—it is because they collected excellent, very detailed data on social institutions, and their monographs sought to present these coherently and intelligibly. In constructing models of the charters of social institutions, they in practice analyzed semiotic domains, with punctilious attention to linguistic aspects and to comprehensive coverage of the domains. If this is true for Malinowski and Firth,^6. and even for Radcliffe Brown, as Leach has shown in a recent essay (1971), it is even more true for later British social anthropologists such as Fortes and Goody, and for the Oxford school of anthropology.

Fortes’s essay on Oedipus and Job (1959) is a good example of the way the British school both contributed to and took its distance from semiotic analysis. The essay reduces Tale religious thought to two basic contradictory statements: (1) we are punished because we are guilty; (2) we are punished even though we are innocent. This reduction is achieved after systematic semiotic analysis of a vast body of data. But then these findings are related to Fortes’s analysis of Tale kinship—his theory of “complementary filiation.” According to this theory, Tale society is organized around a basic contradiction between agnatic and uterine power. This contradiction between actual sources of power is expressed by a contradiction between sources of misfortune. But in the end, Fortes argues, the contradictions are necessary for the maintenance of and the functioning of Tale society. It is to make this, to him, basic point that he engages in his brilliant semiotic exercise. The point I wish to make is that though Fortes does not aim at semiotic analysis, he is very good at it, and some readers like myself find his analysis far more exciting than the conclusion to which it leads him.

The Oxford school is less concerned with the explanation of the functioning and maintenance of society and more programmatic about its semiotic objectives. Evans Pritchard proposed a theory according to which anthropology is essentially “translation,” by which he meant that the field worker's data are cast essentially in the conceptual system of the culture under study (as perceived by the investigator), and the task of analysis (leading to the “monograph”) consists in translating this conceptual system into a language comprehensible to a reader of a different (viz., the scientist’s) cultural background. This seems, on the surface, a purely semantic activity; or rather, insofar as the field worker’s data are records of indigenous behavior as well as verbal statements, a semiotic activity. But we cannot ignore a substratum of functional theory even in the works of the Oxford school: Evans Pritchard’s works about ecology, politics, kinship, and religion are interconnected by the assumption that each kind of institution serves the others, is determined by the others in its form and content. The work of translation to some extent serves the purpose of demonstrating this system of interrelations.

Yet it is significant that Evans Pritchard and his school^7. placed anthropology among the humanities rather than the (social or other) sciences. The coherence and intelligibility this school pursued was essentially of an historiographical type. While functionalism, in its classical form, viewed the social system after the analogy of a biological organism, the Oxford school—and in a different way, perhaps also the American anthropologists Kroeber and Eggan—viewed it as a coherent intelligible formation produced in an environmental and historical context, and as such approachable only as unique and incomparable (except insofar as other such formations could be brought into an environmental and historical relation to it).

We are thus concerned not merely with one atemporal semiotic system (as exemplified by Evans Pritchard's books about the Nuer [1940, 1951, 1965]) but with the transformations of such a system over time, as exemplified by Evans Pritchard’s essays about the Azande and his book about the Sanusi of Cyrenaica (1949), and by Burridge’s books about millennial movements. Burridge’s views about anthropological theory, discussed in the opening pages of this essay, are another instance of the same kind of transformational model.

Not too far removed from this Oxford school are the theories of those who view culture as equivalent to information. This view is exemplified by the work of Gregory Bateson and by that of Claude Lévi-Strauss and his followers. While Lévi-Strauss in the end has become most fully identified with this type of theory, we must not forget that Bateson was the first to see its possibilities.

In his monograph about the Iatmul (Naven, 1936), he divided his analysis into two aspects of culture he called ethos and eidos. Ethos is defined as “a culturally standardized system of organization of the instincts and emotions of individuals” and eidos as “a standardization of the cognitive aspects of the personality of individuals.” Eidos is cognitive organization which pervades every aspect of the culture and gives support to every cultural activity. In later works, Bateson referred to eidos as a “codification system,” organized by contradictions in the culture by a process to which Bateson gave the name “schismogenesis.”For Bateson (1966, 1972), the “codification system” is the culture.

If the analysis of a culture can be reduced to analyzing the codification system, then indeed it almost seems that anthropology becomes a purely semiotic activity. Yet the test for such a proposition does not lie only in Bateson’s own later studies in psychiatry and metacommunication, but also in ethnographie work, viz., in the book Naven itself. From an anthropological viewpoint, the most vital concept in this book is schismogenesis, i.e., the location of a basic contradiction in the social system, as the various attitudes a man develops towards his maternal kin are not logically reconcilable. They must be acted out in ritual or neutralized in the fantasy world of myth. Thus Naven is not merely concerned with codification, but also with the ways man finds for coping with the basic contradictions in his life. And these contradictions arise, as Bateson shows very well, in all aspects of life, including the economic and sociological as well as the developmental. Thus, Bateson has not genuinely eliminated infrastructure from the study of culture. He has merely shifted infrastructure to the field of information itself, mapping out an “ecology of mind” and thus postulating a natural basis for symbolism. There is no theoretical charter for excluding infrastructure from the study of schismogenesis.

Both Bateson and Lévi-Strauss, in studying culture as an information system, took a deliberate step away from empirical reality by concentrating on discontinuities in culture. Both Bateson’s schismogenesis and Lévi-Strauss’ binary oppositions reduce the fluidities and continuities of everyday life into what one may call codable units, i.e., structures of which the elements are constituted by a series of either/or choices. The theory that culture is information really amounts to a proposition that culture is codable by some such method. I do not think that it is essential to the theory that the oppositions should be binary; there are many other possible ways of constructing a code—and Lévi-Strauss has used them too, e.g., triads, Kleinian groups, in Mythologiques (esp. 1968, 1971)—but binaries are the simplest form, and very pervasive in culture.

From a purely methodological viewpoint, it is not the use of binary oppositions and other oppositions that set Lévi-Strauss apart from many other contemporary anthropologists. More distinctive is his use of a structuralist comparative method, by the construction of families of models known as combinatories. The argument, put briefly, is that any given body of anthropological data is codable in many different ways. In principle, there is no way of deciding what is the “right” or “wrong” coding. In practice, three kinds of verification are possible. One may collect further data, and see whether they are consistent with the various codings that have been suggested. One may test the various codings for internal logical consistency, often revealing “anomalies” (such as in Bulmer’s instance of the cassowary referred to above) and thus opening up further areas of empirical investigation. Finally, and most importantly, one may test the codings—which, it should be remembered, are all “emic,” i.e., expressed in terms of indigenous categories of thought—on some assumption that they correspond to something in objective reality.

It is this third kind of verification that has made Lévi-Strauss a very controversial figure. The theoretical justification for it is highly complex and has been best analyzed, in my opinion, by Valerio Valeri in a little-known essay. Valeri explains: “Social reality always implies organization: it is a mediated system of signs. And signs can function in a communication system only insofar as they are finite in number. Otherwise, we could not have communication which is based on combination and recurrence: on repeatability as against unrepeatability, which is incommunicable” (1970:355). To put it briefly, social reality, like the natural world, is the product of a limited number of combinations, regulated by a code. The systems of thought reconstructed by anthropologists form part of this basic “combinatory” and in this sense have objective reality.

But as so many codings of any body of data are possible, how can we decide which of them have this objective quality and which may be purely subjective, such as those (to refer back to Radcliffe Brown) “found in dreams and in art”? A coding is objectively true only insofar as it can be shown to be part of the basic combinatory of social reality.

How can this ever be shown? First of all, it can never be completely shown. Thus, no model ever constructed by an anthropologist can be claimed to correspond completely to social reality. This is a qualification of the highest importance; it puts Lévi-Strauss in a category apart from the phenomenologists. But the truth of a coding can be partially tested by methods of “internal” and “external” comparison. If it can be shown that one particular coding of a cultural domain stands in a relation of transformation to models of other domains of the same culture, so that the models of all the domains form, as it were, a “family of models” between which relations of transformation hold, then such a coding is in principle to be preferred to a coding that seems unrelated to other domains in the culture. This assumption is not entirely an invention of Lévi-Strauss; we find it also in Ruth Benedict’s theory of “patterns of culture.” But Lévi-Strauss tends to give primacy to those levels of culture Marxists would call “infrastructure,” as generating basic contradictions which are apt to reverberate on other levels, even though we cannot—in his view—assumes priori that they will do so.^8. We can apply the same method to external comparisons, i.e., we can compare the coding of a domain in one culture with codings of the same domain in other cultures, starting with those which are geographically contiguous or historically related, but not necessarily confining ourselves to these. Where geographic and historical relations exist, few anthropologists will dispute the aptness of be lieving that a coding which emphasizes the relations existing within a culture area is in principle more credible than a coding totally obscuring such relations. Normally, the relations that are found are not of identity but of transformation, and such transformation can in principle be historically reconstructed. As soon as we extend the method beyond this, we have to rely on a theory such as that whereby intellectual categories are based on the natural structure of perceptive categories, i.e., their aesthetics.

As Valeri interprets Lévi-Strauss, “Logic is based on aesthetics, and aesthetics on properties of physical nature,” such as for instance the neurophysiology of the brain. Such a theory would justify using the structural comparative method irrespective of what the geographical and historical distance between different cultures may be.

I do not wish to examine this theory critically, as it has never really been systematically presented by Lévi-Strauss himself. It is a reconstruction by Valeri, and by others, and I assume that if Lévi-Strauss attached importance to such reasoning, he could have developed it more fully in his own writings.

Mercifully, Lévi-Strauss has not stretched it to its limits: he works preferably within a “combinatory” where historical or geographical relations can be assumed, and does not rely for essential demonstrations on the more remote relations he has sometimes explored. In practice it is unimportant to an anthropologist whether symbolism has a natural basis. If in the analysis of New Guinean mythology, someone suggests to me that a South American or Polynesian parallel may help me to interpret a particular myth, I do not reverently accept such a suggestion because of any supposed unity of mankind; nor do I throw it out as anathema. If the suggestion helps me to interpret a considerable body of New Guinean myths, and if that interpretation can be strongly corroborated from data internal to my culture area, I would call it helpful. If I find that numerous suggestions, say from Polynesian sources, are helpful in Melanesia, then this merely leads me to the thought that we ought to try finding relations between the culture areas of Melanesia and Polynesia. The theory of the unity of mankind does not provide me with a shortcut in empirical research. It is merely a speculation arising out of striking similarities in systems of thought of widely separated peoples. Though an anthropologist may sometimes give himself over to such speculation, he should not let it interfere with his work, which requires the analysis of diversity as well as similarity.

The use of “combinatories” is one of the most sophisticated methods in anthropology, where the relationship with semiotics is evident. It is not really, as some have assumed, merely an extension of Jakobsonian linguistics to the field of culture, but is closer to the transformationalism of Lamb and—with important reservations—that of Chomsky. The main difference between Lévi-Strauss and Chomsky, as I understand it, is that the latter believes in a method of discovering a universal “deep” structure by examining one particular language, whereas Lévi-Strauss uses a comparative method in which the “depth” of the structures discovered increases as the scope of the investigation spreads out over an increasing number of cultures. A “combinatory” covering only South and North America would therefore be less “deep” than a combinatory that would account for, say, Oceanic or European data as well. The difference in approach between the two scholars is directly related to their ideas as to the kind of knowledge provided by their investigations. Chomsky believes he can obtain exhaustive knowledge of at least one language, his own, whereas Lévi-Strauss distrusts the objectivity an anthropologist would achieve in studying his own culture; and as for other cultures, knowledge is always bound to remain very fragmentary, just like the knowledge of an astronomer of other planets.

I have dwelt on French structuralism at some length because of its complexity, but must refer also to highly important American schools of anthropology deeply concerned with semiotics. These schools have been unanimous in rejecting Lévi-Strauss’s use of combinatories. They have usually not attempted to develop new methods of external intercultural comparison.

The most interesting leading figures, such as Clifford Geertz, David M. Schneider, and Victor Turner, have tended to work intensively in single cultures and relate their results directly to universal propositions. They have not been specifically concerned with cultural diversity as an epistemological tool for discovering the nature of man. Geertz, in his theoretical statements, tends to relate cultural symbols to institutions which in turn are related to ecological variables (as in his work on “agricultural involution”). In that respect, little would seem to divide him from the British Oxford school, but he relies far more on psychological explanations (of religion, etc.), and on information theory (like Bateson’s idea of eidos). He does not appear to view his sophisticated analysis of symbolic action as an end in itself, but as a necessary preliminary to the analysis of aspects of social life where their role is important. When he calls his concept of culture “essentially a semiotic one” (1973) he seems to ignore much of his own previous work, or else may be using the term “essentially” a little loosely.^9.

Both Schneider and Turner have more specifically used a philosophical kind of psychological explanation as a basis for their theoretical work, thus replacing Lévi-Strauss’ concept of “combinatories” by a rival explanatory construct. At first sight, we find little to choose between their claim that sentiment logically precedes structure and Lévi-Strauss' argument, presented in attempted rebuttal (1971), that structure or cognition precedes sentiment. Precedence depends, of course, on the nature of the question under investiga tion. Symbols derive patently both from sentiments and from cognitive structure. Yet, it is clear that, having rejected the notion of combinatories, Schneider and Turner had to put something in their place, unless they were to remain imprisoned in the exegesis of semiotic domains in specific cultures. When they speak of sentiments, it is because certain experiences of the nuclear family are patently universal, hence objective facts usable in the interpretation of symbols in any culture.

Schneider’s work on kinship (see esp. Schneider 1967, 1968, 1972) and Turner’s on ritual (see esp. 1964, 1967 a b, 1969, 1971, 1973) consists, in practice, largely of a kind of semiotic analysis of unusual thoroughness, very systematic in identifying the different levels of meaning. But like other anthropologists, they seek an anchor in some kind of objectivity, a direct relation to the referent; the reality of the culture under consideration. Schneider has described his approach as one of “psychological reductionism”; moreover, he has argued that Lévi-Strauss is likewise a psychological reductionist inasmuch as he, too, rests his analysis not infrequently on certain universal experiences of the nuclear family. We cannot inquire here into the justice of this claim, nor into the question whether, theoretically, the method of “combinatories” is possible in the absence of “psychological reduction.” All that concerns us is that such reduction is basically a device for giving anthropology a dimension of objectivity in its analysis of symbol systems.

I have called the psychology of Geertz, Schneider, and Turner a “philosophical psychology.” I intended to suggest by this that it is not based on any scientific school of psychology but on general psychological assumptions about human nature. Other anthropologists, of course, have based their work on more specific theories of behavioral evolution, structure of the personality, perception, cognition, and so forth.^10. Piaget has provided a “genetic structuralist” model, on a psychological basis, which emphasizes the importance of the transformation of structures over time. In anthropology, one of its most interesting products has been the work of Bourdieu (1972) based on fieldwork among the Berber, specifically his essays on the sense of honor, the house as a reversed model of the universe, and patrilineal parallel cousin marriage. The emphasis on process in these basically structuralist investigations makes them akin to the “post-structural” movement now current at Oxford University.^11.

In this survey of different ways of utilizing semiotic methods for analyzing cultures, I have left out many highly important figures merely because their methodology and theory have to be viewed as an amalgam or synthesis of several directions I have already summarized. Edmund Leach and Rodney Needham are among them; they have affinities with the various British schools mentioned as well as with Lévi-Straussian structuralism. They also exemplify one kind of semiotics now beginning to become important in anthropology, namely that based on the philosophical investigations of Wittgenstein.^12.

We must finally consider the Marxist critique of the various methods whereby anthropologists have tried to relate semiotics to natural science. Classical anthropological analysis fully includes the concept that whatever statements informants make about their social or religious system or about the meaning of symbols are determined by the informants’ own best interests whether as individuals or as members of certain families or certain communities (as opposed to others). Raymond Firth has gone to the trouble of devoting an entire book to show the ideological basis of Tikopian myths in terms of clan interests (1967). If Lévi-Strauss believes that symbolism has a natural basis; he recognizes also as a matter of course that it is continually manipulated for ideological purposes.

The Marxist critique arises out of the argument that if infrastructure, in the last instance, determines history, then it also determines, in the last instance, the formation of symbol systems. In that case, symbol systems have meaning only insofar as they are seen as part of a total dialectical system related to infrastructure, or—to use Marxist terminology—if they are “totalized.” Now we have seen that all anthropologists tend to engage in devices of “totalization,” but that they do not all start from modes of production as explanatory constructs “in the last instance.”

Can we start on a Marxist basis and still inquire into the meaning of symbol systems in the cultures anthropologists study? Several French anthropologists have tried it, relying for their theoretical underpinning on philosophers such as Althusser, Maurice Godelier (1967, 1973) being perhaps the best known of these. Godelier has formulated a theory of economic anthropology which avoids the snares of being entirely “emic” or substantivist, and also of being entirely “etic” or formalist. He tends to obtain interesting semiotic analysis of the “emics” of economic systems while placing these in a vaster “etic” evolutionary scheme, based on varieties in modes of production. This method is supported by an argument whereby in non-literate societies the mode of production is such as to draw various other levels of structure (kinship, politics, religion) into the sphere of infrastructure from which they became separated during later evolutionary stages.

In practice, certain American anthropologists such as Sahlins (1972) and Murra (1972-1974) work along lines very similar to Godelier, though they have not felt the need to reconcile what they do with Marxist doctrine.

After this brief survey, can we say what perspectives anthropology and semiotics have opened up for each other? I think it will be clear that the contemporary anthropologist engages in semiotics a good deal of the time. It will also be clear that he does not do so very systematically. Finally it is clear that all anthropological schools have reasons why their study is not reducible to semiotics, but the reasons are all different and agree in nothing except a desire to totalize.

Umberto Eco (1972) has drawn a neat distinction between what semioticists do with texts, and what biologists (natural scientists) do with the reality to which the texts refer.^13. I have tried to show that Eco’s argument cuts through the middle of anthropology, dismembering the subject into biology and semiotics, while rejecting as a hopeless enterprise the attempt to relate them together. My survey has shown that if the enterprise is not wholly hopeless, it is very nearly so; we must therefore answer Eco with caution.

We have seen, in this essay, that numerous anthropological schools, though starting their investigations from the viewpoint of basically biological kinds of theory, were forced to build into their methodological equipment ever more sophisticated semiotic techniques. Conversely, those who started off with a basically semiotic or linguistic viewpoint, such as sociolinguists, cognitive anthropologists, ethnoscientists, in order to work on the “ethnography of communication,” have tended to become increasingly enmeshed in the kind of anomalies of which we saw one example when discussing Bulmer’s cassowary.^14.

Veron (1973) has put forward a similar argument about the relations between linguistics and sociology: both disciplines have the tendency to formulate fundamental problems that can be resolved only by the other. He asked how such questions, that seem to circulate by being tossed from the one discipline to the other, can ever be satisfactorily dealt with. They can be dealt with only in a universe of discourse of greater generality that would be simultaneously semiotic and biological. Veron finds this universe of discourse in what he terms “ideology”—the “social production of meaning.”

Though Veron elaborated this concept in a later paper (1974), it remains somewhat sketchy from the viewpoint of the social sciences. It may be usefully supplemented by a recent work by the sociologist Fernand Dumont (1974) who proposes a science of ideology, in which ideologies may be studied from two conflicting viewpoints: (a) from an objective natural-science viewpoint, as a “residue” from infrastructure; or (b) from the viewpoint of the subject, a construction of a meaningful ideational schema by which external world and self-identity are structured. He described these conflicting viewpoints as “un conflit des pratiques de l'interprétation: ce pourrait bien être, à tout prendre, une définition opératoire de la société, une définition de l’objet de la sociologie” (1974:171).

Such views of the concept ideology differ considerably from those proposed by Hjelmslev (1963) and Barthes (1964), who relegate ideology to a secondary “connotative” level of meaning, separable from a “denotative” level. Whether we prefer to follow Veron and Dumont or Hjelmslev and Barthes depends on our whole philosophy of society, but it seems to me also that the Hjelmslev-Barthes approach would lead the field anthropologist into serious practical difficulties. When I analyzed four explanations offered by the Orokaiva tribe of the Northern District of Papua to account for a disastrous volcanic eruption, I discovered (Schwimmer 1975b) that, in fact, the distinction between denotation and connotation was hard to make: denotation (objective explanation of the phenomena) is inseparable from connotation (ideological claims of tribal superiority and the like). All one could say is that some explanations the Orokaiva offered did greater violence to facts they ought to have known than did others; in other words, some explanations were more scientifically credible; but this is not quite the distinction suggested by Hjelmslev.

In the first part of the present paper we found that Burridge, in effect, treated the discipline of anthropology entirely as an historical phenomenon. In the second part of the paper I attempted to show that this history pertains partly to semiotics, partly to various forms of “totalization.” I am now arguing that these two constituents of the discipline can cohere only if we view anthropology as a study of processes of social production of meaning. Thus, the study of infrastructure is relevant to anthropology insofar as infrastructure produces meaning; and meaning systems are relevant insofar as they are related to social production.^15.

This view, if applied to the history of anthropology, leads directly to Burridge's implication that the discipline's development is inseparable from the reflections arising out of colonial relationships entered upon by European powers. When seeking the “meaning” of anthropology, Burridge actually made it manifest by presenting it as a process of social production. Is there any more satisfactory way of saying what anthropology means? Certainly, other theorists have been far more ambitious than Burridge; they have tried to provide anthropology with an ontological meaning, as though it were supposed to answer the question: What is man? But in making such ambitious attempts they failed to recognize that the question “What is man?” has no anthropological meaning, unless we also answer with the retort: Why do you ask that question?

Thus we are led back to the recognition that anthropology is an ideological science, in the sense of Veron and Dumont. It contains within it both semiotics and biology. A very specific example of the problems with which an ideological science can deal is offered by the many excellent studies made of religious movements over the last two decades. They contain a fine analysis of the systems of “meaning,” the doctrines, taught by the various movements involved; thus, they depend on semiotic techniques. But at the same time, they deal with objective reality insofar as it produced the movements—real economic and political contradictions, in a framework of social relations. They deal with structural transformations, and with the real genesis of these transformations. But they do not become “nomothetic” except in respect of methodology or epistemology: anthropology is in fact a non-science, except insofar as it reflects upon the ideological production of meaning.

It is for semioticists to say whether this makes anthropology generically different from other domains in semiotics, or generically the same.