h is “culture”, but it is culture viewed in a narrow sense. Culture, rather than consisting of the material and behavioral phenomena of a society, is viewed as consisting only

INTRODUCTION

Cultural anthropology in America during the past decade and a half has been strongly influenced by the distinction drawn by Goodenough (1956a) between ethnography and ethnology and by the approach to ethnography that has its roots in the same soil as this distinction. This approach is represented in the work of Conklin (1964, 1967), Goodenough (1956a, 1957, 1964b), and Frake (1961, 1962) as well as others whom we shall discuss shortly.

The basic theoretical construct in this approac

of the cognitive organization of these phenomena (cf. Tyler 1969: 3). Thus, the theoretical position of this approach is that culture is a code (cf. Goodenough 1957: 168) and the models and methodology to be used in its elucidation are those of linguistics.
For instance, Frake in attempting to define ethnography in terms of this new approach writes:

The discipline [ethnography] is akin to linguistics; indeed, descriptive linguistics is but a special case of ethnography since its domain of study, speech messages, is an integral part of a larger domain of socially interpretable acts and artifacts. It is this total domain of “messages” (including speech) that is the concern of the ethnographer. The ethnographer, like the linguist, seeks to describe an infinite set of variable messages as manifestations of a finite shared code, the code being a set of rules for the socially appropriate construction and interpretation of messages (1964a: 132).

This approach to anthropological inquiry has been variously called the “new ethnography”, “ethnoscience”, “ethnosemantics”, “structural ethnography” (cf. Hymes 1964a), and “cognitive anthropology” (cf. Tyler 1969). However, none of these terms, in my opinion, adequately identify the main intellectual currents in this approach, and, consequently, I propose here to refer to those who deal with the cognitive aspects of a culture largely, but not necessarily exclusively, in terms of the techniques of structural linguistics as “cognitive structuralists”.

However, our purpose here is not to review the whole literature of cognitive structuralism, which has now become large. Instead we will focus primarily on those writers who have attempted to explain and justify the epistemological basis of cognitive structuralism, and our critique of this approach will be largely but not exclusively in terms of the logic of the linguistic models that the cognitive structuralists use as well as in terms of the internal logic and consistency of their theory.

THE DISTINCTION BETWEEN ETHNOGRAPHY AND ETHNOLOGY

When the people of the Earth all know beauty as beauty,
There arises (the recognition of) ugliness.
When the people of the Earth all know the good as good,
There arises (the recognition of) evil.
Tao-te-ching, Book One, trans. by Lin Yutang (1948)

One of the earliest publications in the genealogy of cognitive structuralism is Goodenough’s monograph on Trukese property and kin groupings. He explicitly introduces the use of a linguistic model for cultural description:

... a culture is as susceptible of rigorous analysis and description as is any language. The demonstration of this proposition is, in fact, a long-range objective towards which the present study was undertaken as an exploratory step. The writer feels that the recent advances in methodological and conceptual rigor in descriptive linguistics present a serious challenge to descriptive ethnography; for patterned verbal behaviors, the basic data of linguistics, are but one branch of patterned behaviors in general, which are the basic data of ethnographic analysis ... (1951: 10).

Goodenough at the same time raised questions as to the value of numerical data such as are collected for purposes of cross-cultural comparison; and he reached a significant conclusion on which he was to elaborate in subsequent publications. While the methodology of comparative inquiry produces a summary at one point in time of the results of many decisions, Goodenough argued (1951: 11), it does not elucidate the cultural rules whose application results in that situation.
Another early significant paper in the genealogy of cognitive structuralism was Conklin’s discussion of Hanunoo color categories. This paper demonstrated the need for more rigorous methods, devoid of contamination from the socio-cultural system of the investigator, in the analysis of specific semantic systems. Thus, Conklin states:

Color terms are a part of the vocabulary of particular languages and only the intracultural analysis of such lexical sets and their correlates can provide the key to their understanding and range of applicability. The study of isolated and assumed translations in other languages can only lead to confusion (1955: 340).

However, this distinction drawn between that which has sometimes been called the “inside” view of culture versus the “outside” view did not receive extensive treatment until Goodenough’s seminal article on residence (1956a) in which he explicitly contrasted the methodology of ethnography with ethnology.

Both Goodenough and Fischer had worked on Truk and had taken household censuses. But Goodenough’s classification of residence patterns differed significantly from that produced by Fischer. In attempting to explain how this came about, Goodenough concluded that the difficulty arose as the result of using cross-cultural concepts, entities, and categories in investigations of particular sociocultural systems. He states:

Concepts used for comparative purposes, moreover, must be based on criteria which are independent of any particular culture....

It is, therefore, a procedural fallacy to use these concepts as a basis for classifying the residence choices of individual members of a society. They do not choose on the basis of criteria which are outside their culture, which exist only in the heads of the anthropologists. They choose on the basis of criteria which are provided by their particular culture and which may be quite different -- indeed probably are -- from those used by the anthropologist in classifying their culture (1956a: 20).

Therefore Goodenough concludes:

... what we do as ethnographers is, and must be kept, independent of what we do as comparative ethnologists. An ethnographer is constructing a theory that will make intelligible what goes on in a particular social universe. A comparativist is trying to find principles common to many different universes.... Although they operate at different levels of abstraction, both ethnographer and comparativist are engaged in theory construction. Each must, therefore, develop concepts appropriate to his own level of abstraction . . . (1956a: 37).

This important distinction between the “ethnographic” and the “ethnological” points of view, although phrased in a stimulating and thought-provoking form, was not entirely new at the time Goodenough published this article. Pike (1954), also using a linguistic model, had distinguished “etic” investigations with cross-cultural concerns and validity from “emic”, i.e., system-specific investigations, that focus on the distinctions that are structurally significant for a specific sociocultural system. What was new, however, and significantly so, was the conclusion drawn by Goodenough from this distinction. He did not believe the difficulties that he found in residence classification in the Trukese data could be solved by the normal progress of anthropological science as it was then indicated. Instead, he viewed the discrepancy in the two sets of data as calling for new theory and new concepts for describing specific sociocultural systems.

However, before proceeding with our analysis, certain observations might be usefully made here. Goodenough’s article (1956a) has attracted wide interest and favorable comment. And it has had a significant, if not revolutionary, impact on American anthropology. For example, Naroll generalizes the distinction Goodenough made in this article into what he calls “Goodenough’s Rule” (1967). This response appears to have arisen because Goodenough, by presenting a confrontation between two sets of apparently incompatible data, demonstrated for the first time that this trouble spot in anthropological science appeared insolvable by normal procedures. And consequently to resolve this anomaly, a new approach using new conceptual tools was called for. It is also interesting to note that while cognitive structuralism has come under attack (see Berreman 1966; Coult ]968; and Harris 1968 for example), there has not been any full scale attempt to explain within the terms of standard anthropological theory the discrepancy Goodenough noted between his data and Fischer’s.

At this juncture I think it may also be illuminating to point out with regard to the history of science that the early cognitive structuralists stem from the same environment that is considered by some to have produced the acme in cross-cultural research methods, Yale University. Consequently, since the conceptual framework of the cognitive structuralists and those engaged in cross-cultural research developed in a similar environment, it may have been necessary to overdraw the distinction between ethnography and ethnology to establish the necessary contrast. Certainly Goodenough’s distinction in his 1956a article seems to be overdrawn, particularly in light of his later articles. However, before discussing these, we shall examine some of the implications of the ethnography-ethnology distinction and the approaches used by others who have incorporated this distinction into their thinking.

The Paradox of Extreme Relativism

The distinction that Goodenough has made between what may be referred to as system-specific models and ethnological models is an important and far-reaching one. However, carrying this distinction to an extreme with a disregard to the interplay between ethnological investigation on the one hand and the data derived from ethnographical research and its attendant concepts can result in what might be called the paradox of extreme relativism: How can any system be described without reference to any other system?

This issue of cultural relativism is an unresolved one in the field of cognitive structuralism, the implications of which will concern us in one form or another throughout the remainder of this essay. And while there is no monolithic view on cultural relativism among the cognitive structuralists, they, nevertheless, all tend towards the more extreme end of the continuum, at one pole of which are found those who maintain that an abstraction of any item from a cultural system to compare it cross-culturally renders the data meaningless (see Berlin and Kay n.d.: 32 for a discussion of this extreme position) and at the other, those anthropologists who are concerned solely with transcultural analysis and regularities. In spite of their position on cultural relativity, one feature common to all cognitive structuralists, as we shall point out, is the difficulty of maintaining a consistent system-specific stance in their ethnographic analyses.

We shall return to this problem of extreme cultural relativism later on in this essay and deal with it in depth after we have considered various related issues in cognitive structuralism. But first to set the stage for our inquiry into the epistemological basis of cognitive structuralism, we shall present Hymes’ view of cultural relativism.
Hymes has raised objections to any distinction between system-specific models and ethnological or universal models which would lead to an extreme position on relativism. He points out:

Thus, not only is an understanding of universals impossible on the view of a radical structural relativism that would hold consistently that in language “there are only differences,” but some features of individual systems in turn would be less intelligible, less deeply investigable (1964a: 19).

Hymes maintains that Jacobson’s conception of distinctive features is a “brilliant indication of how structural perspective and principle can be maintained at the levels of both individual systems and the universe of systems” (1964a: 19). Not only, Hymes states, can features of individual systems be at once structural and universal, but novel structural features and relations can be discerned on the level of universals. He concludes (1964a: 20) that it is difficult to maintain “a thoroughgoing, consistent balance between ethnology and ethnography, the comparative and the descriptive” and that there is constantly the danger of “slipping into a static mold, dividing the two, and perhaps assigning priority to one, whereas the two must develop in a sort of dialectic relation”.

The issue of radical relativism stems from the view taken by the cognitive structuralists on a priori categories, and further facets of this issue are revealed when the position of the cognitive structuralist on induction and apriorism is analyzed. However, before doing this, let us consider Goodenough’s revised position on ethnography and ethnology and the use of the phonetic-phonemic distinction as an analogy to explicate the contrast between ethnology and ethnography.

Goodenough’s Revised Position

In an article published in 1964 Goodenough takes the revised position that the conceptual levels of ethnography and ethnology are not completely disparate but are in fact interrelated. However, in this he apparently views the goals of ethnological inquiry in different terms than previously. He writes:

Constructing models of kinship phenomena is profitable at both levels of cultural anthropological analysis, the ethnographic and the ethnological. In ethnography there is need to construct for each society’s kinship system a model that most accurately and elegantly describes it. Ethnology, on the other hand, requires us to construct models that pertain to all kinship systems, in terms of which the many different ethnographic models can be compared. What these more general models look like, insofar as they have an empirical base, depends on the kind of models we construct in the particular cases of ethnography (1964b: 221).

This later view of the ethnography-ethnology distinction thus may not contradict Goodenough’s earlier statement so much as it reflects a further refinement in his thinking. Goodenough (1956a) originally made no distinction between comparative studies and ethnology but used these terms interchangeably. However, by 1964 Goodenough (1964a, 1964b ) has distinguished comparative studies from ethnology. The approach which he sometimes referred to as “ethnology” in his original 1956a article has now been rephrased as “comparative studies”, and he writes: “Consideration of typology and its attendant problems made it clear that classifications appropriate to comparative study are on a different conceptual level, serving different purposes, from the categorical distinctions that make the best sense of phenomena within a particular society” (1964a: 9).

In contrast to comparative studies, ethnology in this revised view is concerned with the construction of universal models from the system-specific models that are produced by the new approach to ethnography. These new models will not only provide procedures by which specific cultural systems may be described, but also they will provide the tools by which the cognitive structuralist will be able to make profitable cross-cultural comparisons.

It is also important here to note that Goodenough uses the term “comparative studies” in a very restricted sense. He is referring to those studies that employ the concepts and taxonomies of institutions such as family, descent, marriage, etc., that have developed over the years of anthropological inquiry. They have failed to be universally applicable because they carry a cultural load from the anthropologist’s own culture in contrast to the conceptual tools of the new ethnography, which are claimed to be culture-free and universal. That Goodenough does not eschew the comparative method as such is shown in his view of the “new ethnology” (as specified above) and by his use of the Lapp kinship system to elucidate a distinction in the American system (1964a).

In conclusion, the comparative method must thus be distinguished from “comparative studies” used in this restricted sense. That this is not done in the discussions of the ethnography-ethnology distinctions frequently leads to what appears to be contradictions and misjoined arguments. For example, certain inquiries using a comparative methodology are rejected out of hand without first attempting to discover whether the concepts used belong to the genre of “comparative studies” or not.

The Phonetic-Phonemic Model

To describe the relationship between the ethnographical and ethnological (most frequently in its comparative-studies sense but occasionally also in its universal-model sense) a linguistic analogy has been used in which phonetic procedures are equated with ethnology. Thus, the distinction between ethnography and ethnology in terms of this discourse is frequently phrased as an “etic-emic” distinction (see Pike 1954 for its original use; also Sturtevant 1964). For example, Goodenough in his 1964a discussion uses this linguistic model to explicate the distinction between ethnography and comparative studies. He maintains that the classification appropriate to comparative studies must be applicable to all societies, while the categories for ethnographic analysis must allow for the most economical statements of behavioral events within some one society as a universe in itself (1964a: 9). This distinction, he writes, has its counterparts in linguistics:

There, the analyst has a phonetic classification of sound types (international phonetic alphabet) with which he can describe the phonology of all languages and thus render them comparable. But the distinctions in this phonetic classification are not congruent with the distinctions of speech sound that are significant in any one language.... An object of descriptive analysis is to apply rigorous procedures by which to move from the general phonetic classification of initial observation and transcription to a new one--a phonemic classification--which most adequately and elegantly accounts for the meaningful acoustical distinctions in the speech of the native speakers ( 1964a: 9-10).

However, the equation of phonetics with comparative studies and phonemics with the new ethnography is not valid within the logic of the discourse that Goodenough uses. While it is true that phonetics and the categories of the comparativist share one feature, their cross-cultural relevance, phonetics contrasts with the categories of the comparativist in a more important feature. The cognitive structuralists specifically state that the concepts of the comparativist do not provide procedures for the development of the system-specific models that are the concern of the new ethnography. But the phonetic grid does just this in conjunction with the concept of a phonemic system. It provides procedures which lead to the isolation and description of the phonemics of a specific language. Consequently, since phonetics shares two features with the universal models of the cognitive structuralists, universal relevance and procedures leading to descriptions of system-specific distinctions, it seems more appropriate to equate the phonetic level with universal models rather than with the categories of the comparativist.

To summarize, the ethnology-ethnography distinction is frequently equated with the phonetic-phonemic model, but the referent for “ethnology” in these comparisons is commonly not made clear. Where “ethnology” is used to refer to comparative studies and the phonetic label used in a somewhat pejorative sense, I have argued that this linguistic model is completely inappropriate. Instead this linguistic model can only be appropriately applied when “ethnology” is used to refer to the “new ethnology” which is concerned with the building of universal models and is in a dialectical relationship with the “new ethnography”.

This brings us to a discussion of the position taken by the cognitive structuralists on induction and a priori categories, which, when held in an extreme sense, leads us to what I have called “the paradox of the cultureless ethnographer”.

THE PROBLEM OF INDUCTION AND A PRIORI CATEGORIES

One of the basic tenets of the cognitive structuralists is that scientific discovery procedures are based on the logic of induction, and, consequently, prior category assumption must be eschewed. For example, Conklin writes:

In ethnography, significant categories and relations are derived from intracultural analysis; they are not determined by the application of a previously designed typological grid. Prior category assumption is ruled out, and, instead, we try to base our work on such concrete realities as a local group of people and the kinds of objects and events the members of this group treat as culturally significant . . . (1964: 26).

Keesing uses a similar position as a basis for criticizing Nelson’s comments (1965) on Befu’s analysis of Japanese social structure. He writes:

I assume that when we analyze the social organization of a particular people, our immediate goal is a description which will permit productive statements about social interaction (Goodenough 1951, 1957). That is, it will tell us who will act in what capacity in what situations -- defining acts, capacities (i.e., roles or social identities), and situations as they are perceived and categorized by the actors themselves.
Given this assumption, prescriptive definitions of “descent”, “kindred”, etc., are completely out of place in an ethnographic description. We should isolate the categories and relationships which are culturally relevant and describe them rigorously. When we have done this, what we label them assumes an appropriately lesser importance (1966a: 475).

It is useful to discuss here in detail Keesing’s criticism of Nelson as by so doing we can illustrate certain of the issues raised by cognitive structuralism. First of all, within the logic of the ethnology-ethnography distinction, a dismissal of “prescriptive definitions” cannot be entertained until it has been ascertained whether the definition was advanced for comparative or ethnographical purposes. That this is seldom done in the criticism by the cognitive structuralists of the work of other anthropologists suggests that they hold the theoretical position of the comparative ethnologist to be untenable, rather than being merely unproductive. In this view, the methods of comparative ethnology are invalid because purposeful behavior is analyzed in terms of its meaning within a theory, however covert, of the comparativist rather than in terms of its system-specific meaning. But in dismissing the methods of the comparativist in terms of their being based on theory meaning rather than their being unproductive, the cognitive structuralists have placed themselves in a position of being against all anthropological inquiry based on theory meaning. I shall return to the import of this problem later in my discussion. However, what the cognitive structuralists have failed to realize in their quarrel with the comparative ethnologist is that their argument is not really over the distinction between behavior viewed in terms of theory meaning or system-specific meaning; or between native categories and a priori categories, for as I discuss below it is impossible to engage in any inquiry without a theory or a priori categories. Instead their disagreement fundamentally is over the levels and types of concepts used in scientific inquiry.

This becomes more clear in analyzing the problem of “prescriptive definitions” and “prior category assumption”. If we equate ethnographic description with the description of a specific language, then Keesing’s admonishments against prescriptive definitions would be equivalent to maintaining that the concept of the phoneme, or the morpheme, for instance, are not amenable to discussions of definition, which in fact has been far from the case (see Chao 1934; Swadesh 1934; and Twaddell 1935).
But this concern over “prior category assumption” and “prescriptive definitions” stems from the position that the cognitive structuralists take on induction: only by induction can valid scientific inference be drawn. This is the basis of Keesing’s criticism of Nelson. Also, Goodenough specifically writes: “Ethnographic description, then, requires methods processing observed phenomena such that we can inductively construct a theory of how our informants have organized the same phenomena.... [A phonetic transcription] is the raw data from which statements describing language may be induced...” (1957: 167-168).
However, this approach would appear to imply a simplistic view of induction based on enumeration. Within recent years, the inductive method has come under strong attack by Popper (1961) and by Jarvie (1964). The logic of scientific inquiry is phrased by them in deductive terms with the falsibility of hypotheses as the cornerstone. Interestingly enough, Goodenough’s description of how a linguist proceeds in the discovery of phonemes is not inductive in character but fits this hypothetico-deductive method (see Braithwaite 1955; Salmon 1966). Goodenough states:

A linguist arrives at a statement of them [phonemes] not by direct observation, but by testing various hypotheses which he formulates about them until he finds a hypothesis which fits the acoustical phenomena as he has noted them and which also provide him with a model for producing acoustical phenomena himself which result in predictable responses by others (1956b: 195-196).

Thus, the position of the cognitive structuralists is that the anthropologist should record the distinctions of a society as perceived and categorized by the members of that society themselves, and the investigator will distort ethnographic reality if he approaches a society with any a priori concepts. And yet, such concepts are just what the linguist does use. While it is true that the properties of a specific phoneme can only be described in terms of the system in which it is found, the concept of the phoneme cannot. It is the concept of the linguist, not that of the system being investigated (Twaddell 1935; Roger Brown 1956). This is not to deny that the phoneme described is a functioning category in the language being analyzed, and this is in fact the crucial point in terms of the argument over concepts.

Thus, the problem is not that the only valid scientific discovery method is based on induction, for this turns out to be a chimera; or that we must eschew apriorism, for the position taken here is that this is in fact impossible. Rather than avoiding apriorism the task of the scientist is to make the nature of his apriorism more explicit so that it can be tested for adequacy and productiveness. Consequently, instead of throwing our lot in with the inductive method, what we need to do at this stage in the evolution of anthropological inquiry is to focus on developing abstract analytical concepts. By this I mean concepts that carry no cultural burden from any sociocultural system (beyond that of why the question was posed in the first place). Instead they are constructed from universal aspects of behavior and include the observational procedures by which the cultural features of the target society can be identified and described. We shall return to discuss the nature of such concepts further after we have considered the problem of apriorism in greater detail and presented those solutions to this issue that have been devised by the cognitive structuralists.

The Paradox of the Cultureless Ethnographer

Leach reaches conclusions with regard to comparative ethnography and apriorism similar to that of the cognitive structuralists. His point of view is that we have not yet devised universal, scientific concepts but are still relying on the premises and concepts which are inherent in the English-language pattern of thought (1961a: 27). But his corrective to this problem is not based on a linguistic model in which system-specific distinctions are king. He argues that we should devise mathematical models for analyzing our data. Leach’s position on scientific concepts warrants further discussion, and we will return to it later.

However, even Hempel who propounds a reconstructed inductionism admits that it is impossible to eliminate a priori concepts in scientific discovery (1966: 112-115). But we do not have to turn to philosophers of science to find support for this conclusion that apriorism is impossible to eradicate. Within the discipline of anthropology there is a theoretical position which has the same entailment but which strangely enough lies full in the historical stream which leads up to the cognitive structuralists. This is the Sapir-Whorf hypothesis.

Thus, this brings us to a further paradox in the ethnography-ethnology distinction which is related to the paradox of extreme relativism. This latter paradox results from maintaining that the ethnographer should not use any a priori concepts and at the same time maintaining that every sociocultural system has its own idiosyncratic cognitive style which determines for its members how they segment, absorb, and explain raw experience. Therefore, to systematically investigate and describe any specific system in terms of the new ethnography required by the ethnography-ethnology distinction presupposes a cultureless ethnographer.

Resolutions to the Problem of A Priori Concepts and the Problem of the Cultureless Ethnographer

In the original phrasing of the distinction in which ethnography is contrasted with comparative ethnology, there is no apparent resolution to this paradox of the cultureless ethnographer. However, in the rephrasing of the distinction in which ethnology is equated with the building of universal models from system-specific ones, a partial resolution of this paradox and the problem of a priori concepts lies in the methods used to construct the universal models.

First, inquiry should proceed from the particularistic to the universal level rather than vice versa as the paradox of the cultureless ethnographer at first would seem to suggest. New universal models can then be built which are based not on the observations of the culture-bound ethnographer, but on the distinctions which are discovered to reside in a number of sociocultural systems (cf. Goodenough 1964b: 237).

Once there has been sufficient data to build new universal models, these can in turn be used as a means of discovery. Goodenough points out: “One purpose of model construction at the general or universal level is to provide investigators with a kit of things to look for when making empirical studies at the particularistic or ethnographic level” (1964b: 236).

Universal models that consist of a “kit of things to look for” would seem to imply that for any cultural domain the range of world cultures contains a finite number of distinctions. And this brings us back to the problem of how one defines a cultural domain. The entailment of the cognitive structural position is that the cultural domain for any society can only be defined in system-specific terms. This is the thrust of Conklin’s perceptive argument with regard to the domain of kinship (1964). Schneider (1965) also makes a similar point. Now a “kit of things to look for” would not provide the conceptual tools that are necessary to define the boundaries of a cultural domain in any particular society since by definition it would include items both from within as well as beyond the system-specific cultural domain.

However, there is yet another approach to these universal models that are to be built from the experience of many ethnographic investigations that would seem to resolve this boundary problem. This approach focuses on the procedures used to determine the system-specific models rather than on the possible content of these models.

Thus Goodenough writes:

At this level [the ethnographic], different kinship systems require different models to describe them. The common denominator is not to be found in any particular aspect of the formal content of the various systems, but in the operations by which we construct a model of what a system’s content is, by which we conceptualize it in terms most appropriate to the phenomena (1964b: 221).

The third but related approach to the resolution of the paradox of the cultureless ethnographer and the problem of a priori concepts has also emphasized the development of culture-free field methods and discovery procedures rather than a universal grid. And these are based on those utilized by the members of the target society themselves. Thus Frake (1964a) states that the ethnographer must discover how a person in a given society finds out from one of his fellows what he knows and this in general involves discovering the queries that are put and the responses that are elicited by them from native actors. And Metzger and Williams (1966) in order to eliminate a priori categories have focused on eliciting procedures in the development and use of native frames of reference.

Whether these approaches to the resolution of the problem of a priori concepts and the paradox of the cultureless ethnographer are sufficient or not is at the present state of development hard to assay. We shall, however, return to the problem that a priori concepts poses for the ethnographer later on in this paper and discuss their value.

PROBLEMS IN DEALING WITH ACTIONS SYSTEMS

We have pointed out that the distinction between ethnology and ethnography has had an important and seminal influence on cultural anthropology in America (see Colby 1966 particularly for a review of this influence). And we have analyzed how this distinction entails certain paradoxes that are not easily resolved. We have also discussed how the cognitive structuralists do not eschew ethnology, as long as it is based on a particular type of universal model. However, those that adhere to this distinction are not themselves able to maintain an “ethnographic” stance consistently in the analysis of their data.

This is not only true for material analyzed by means of componential analysis, which we shall discuss shortly, but it is also true for those analyses by the cognitive structuralists that occur within the conceptual framework of social structure or social organization. This, of course, should not surprise us for when we deal with the theoretical construct of social structure we focus on action systems, and, while the cognitive structuralists have been explicit on how to handle the cognitive features of various cultural domains, they have not specifically dealt with the procedures necessary for a description of the distinctive features of action systems.

These difficulties in dealing with behavior arise because the cognitive structuralists have tended to view behavior as an epiphenomenon, a sometimes poor and inconsequential reflection of cognitive reality. This position apparently stems from two sources. First, the linguistic model used provides a paradigm for this. While the phonology of a language can be described and the grammatical rules detailed, what a person says on any particular occasion cannot be predicted. Secondly, Goodenough has argued that in focusing on behavior in the analysis of kin groupings and in ignoring the cultural principles organizing this behavior in favor of their culture-bound concepts, anthropologists have failed to identify nonunilineal descent groupings and have in fact forced some of the empirical evidence for these into a unilineal mold (cf. Goodenough 1955, 1961). Thus he writes:

anthropologists have been talking about two different orders of reality as if they were part of the same order.... One of the phenomenal order of observed events and the regularities they exhibit.... An observer can perceive this kind of statistical patterning in a community without any knowledge whatever of the ideas, beliefs, values, and principles of action of the community’s members, the ideational order.... The phenomenal order is a property of the community as a material system of people, their surroundings, and their behavior. The ideational order is a property not of the community but of its members. It is their organization of their experience within the phenomenal order, a product of cognitive and instrumental (habit formation) learning. The ideational order, unlike the statistical order, is nonmaterial, being composed of ideal forms as they exist in people’s minds, propositions about their interrelationships, preference ratings regarding them, and recipes for their mutual ordering, as means to desired ends.... Thus, the phenomenal order of a community, its characteristic “way of life”, is an artifact of the ideational order of its members (1964a: 12).

There are several unresolved dilemmas with respect to this idealistic position. First of all, it is only through behavior (linguistic behavior, symbolic behavior, etc.) that we map the ideational world of the subject. But, since behavior is only an epiphenomenon, an artifact that has played us false, how can we be sure that the ideational world defined by such behavior is valid? Of course there are levels of behavior involved here, but what these are, how they are interrelated, remains to be spelled out.

Furthermore, the argument for the primacy of the ideational world is built upon another logical fallacy. Goodenough has argued that the concepts used in comparative studies and other anthropological inquiry have been culturally contaminated by the ideological concerns of the western world (cf. Goodenough 1970). But if anthropological methodology has failed to isolate the natural units of a society because of its ethnocentric bias, of course there should be a disjunction between the cognitive organization of the members of a target society and the social units exogenously determined. We should not expect otherwise. But this does not then provide us with a charter to focus entirely on cognitive organization as the only true ethnographic reality. What this does mean is that we have to devise better methods of isolating the natural social units of a society (cf. Appell n.d.), and only then can we test the relationship between cognitive organization and social organization productively.

However, let us look at how the cognitive structuralists have in fact dealt with empirical materials of social action. Keesing, for example, in his description of Kwaio kindreds (futalana) states, in a similar vein to other cognitive structuralists, “I construe the structure, and component units, of a culture to be unique in the same sense that the structure of a language is unique” (1966b: 346). But his concern is not with the Kwaio semantic field that includes the “kindred” as an element. Instead, Keesing is concerned with the kindred as an element of social organization for he writes to begin his article: “Few aspects of social organization has provoked as much recent controversy as the personal kindred . . .” (1966b: 346). Even so, in dealing with this Kwaio category of futalana, Keesing states that his approach is still to be “ethnographic” in the sense defined by Goodenough, and he writes:

The best descriptive theory of Kwaio society I have so far devised treats these classes of kin, defined in terms of nuclear individuals, as important analytical units.... In such a description, the best gloss for this cultural category would seem to be “kindred” (1966b: 347).

But Keesing also states:

Kwaio society could also be described without using a “kindred” concept. The same empirical phenomena can be successfully accounted for, though so far with somewhat less economy and more redundance. Such a description would simply specify rules for determining the kinship distance between any given individual and each of his relatives.... Futalan would then refer to the collectivity of these kin and could perhaps best be glossed as “his relatives” This may ultimately turn out to be the most effective way of ordering the data. Thus on the ethnographic level, whether Kwaio society “has kindreds” or not boils down to question of descriptive economy, not of “fact” (1966b: 347).

The entailment of the ethnography-ethnology distinction, it seems to me, is that on one level of analysis the functioning social isolates of the target society certainly must be identified and their interrelationships detailed. And while Keesing has given us two different ways of viewing the Kwaio futalana, he has not told us how the Kwaio themselves view this social category. For example, we do not know whether the Kwaio view the futalana as merely a social field in which the relations of the members of that field to a propositus are solely diadic; or whether the futalana is a functioning element in Kwaio social organization. By a functioning element I mean here that the members of this category have the recognition of a shared moral order with regard to a particular propositus. And thus, even though the members of this category individually hold identical rights or obligations in a particular cultural domain with respect to a common propositus, they recognized that they share these with the result that various members, or a significant proportion of them, come together at the pertinent times to form social groupings in order to discharge more adequately their individually held obligations (see Appell 1967a).

The difficulty in applying linguistic models to action systems is also illustrated in Goodenough’s work on the Lakalai hamlet. This interesting contribution on the social structure of the Lakalai hamlet was published in 1962 well after Goodenough had made his conceptual innovation on the use of linguistic models for cultural description, and in his latest work (1970) he continues to use this material without resolving the problems that I have previously raised (cf. Appell 1967a). He has not, in my opinion, described the cognitive organizing principles for this grouping as he has failed, like Keesing, to isolate the system-specific features of the social isolate. For it is not clear whether the kin organization described is the result of kinship being a necessary criteria for hamlet membership or only the unintended outcome of a small scale society.

LINGUISTIC MODELS AND ACTION SYSTEMS

The application of linguistic theory and models to culture has produced significant and perhaps unprecedented advances in our understanding of the cognitive organization of cultural systems. But the use of a linguistic paradigm in dealing with social behavior raises questions as to its appropriateness, a problem which to my knowledge has never been satisfactorily faced by the cognitive structuralist. Certainly, the linguistic paradigm includes the assumption of automatic behavior at certain speech levels, that is behavior carried on without the conscious knowledge on the part of the subject, and in addition, the lack of organized community-wide sanctions. Furthermore, I would question whether many aspects of social structure such as “village”, “descent group”, and so on occur with sufficient frequency to form a corpus of social behavior large enough for the type of analysis that the linguist does to speech behavior. And finally, social behavior differs from speech behavior not only in that it is less automatic, occurs less frequently, and has organized sanctions, but also in that the members of a target society frequently have very clear-cut “theories” as to how their society works and use these “theories” to guide behavior. These “theories” are outlined most frequently and explicitly at times of jural disputes, and therefore, the features of the system-specific isolates of a target society can often be best determined in such an environment.

Finally, if the cognitive structuralists are to be concerned with social action, methods will thus have to be devised to link a specific “descriptive theory” at least to the level of behavioral reality and even perhaps to the deeper level of psychological reality. And specifically with regard to the level of psychological reality, techniques will have to be employed to establish that the formal models actually distinguish between what is in the heads of the members of a target society and what is in fact only representative of what is in the head of the observer. This problem of the psychological reality of the analysis has been recognized in componential analysis and various tactics have been devised in attempts to deal with it. Nevertheless, it still remains one of the most difficult, unresolved problems for the cognitive structuralist.

Issues With Respect to the “Kindred” and Other Concepts in Social Anthropology

But there are still other problems that have to be solved before a linguistic model can be productively used to deal with social action.

The cognitive structuralists have expounded against a priori categories. However, I have taken the position here that prior category assumption is not necessarily inappropriate but that the type of a priori category used is. For example, the concept of the phoneme is based on the assumption that all languages may be productively described by the use of such an a priori category. One of the difficulties in the position of the cognitive structuralist is that no one has systematically investigated the formal properties of the concepts used by the comparative ethnologist and contrasted such properties with those concepts that the cognitive structuralists believe are more productive. Thus, they have not spelled out in detail the distinctive features of concepts that are acceptable in the light of their theoretical position. For example, Keesing in his criticism of Nelson’s use of “descent” and “kindred” states that ethnographic investigations should focus on “roles or social identities” and “situations as they are perceived and categorized by the actors themselves” (1966a: 475). However, are not “roles” and “social identities” cross-cultural concepts? But most importantly, Keesing offers no explanation as to how these concepts differ from the concept of a kindred.

The distinction must thus be drawn between prior categories that are structured in terms of content and those that involve observational procedures for eliciting the content of any specific system. One of the characteristics of the concept of the phoneme is that it involves operational procedures for isolating the phonemes of a specific language, and it seems to me it is perfectly plausible to consider the kindred at a similar conceptual level as the phoneme (see Appell 1967a for such an attempt) rather than as an ideal type. But the cognitive structuralists have not, nor have they made an analysis of the concept to show why it cannot be so considered.

Thus, Goodenough (1970) seems to have fallen into the same trap of typologizing that he has accused the comparativists of being ensnared in. He has elaborated a typology of kindred-like units which appear to be based on similar ethnocentric conceptions of the social order. Even if I have misinterpreted him and this turns out to be the basis of an etic grid for kindreds, Goodenough still has not provided this etic grid with a systems theory, comparable to that of the phoneme, which, along with the necessary observational procedures will enable the anthropologist to isolate the system-specific, natural units of any society. And therefore it can only have limited productiveness.

Thus to view the kindred in an analogous manner to the phoneme, we would first have to construct a theory of social systems at the same level of generalization that exists for languages, and at this level such a theory would account for all social systems, not just a limited range of societies such as unilineal ones or cognatic ones (also cf. Leach 1961a). In constructing such a theory a society may be conceived of, at least in part, as being composed of social units and the relationship between such units with regard to scarce goods and services, as I have suggested elsewhere (Appell 1965, 1967a, 1967b, 1968, n.d.). The social units of such a model would comprise the persons, networks, groups, etc., that are recognized by the society in question as forming functioning social entities in various social realms. Such system-specific social entities I have termed “social isolates”.
However, it is not my purpose at this point to construct a theory of social systems but to analyze further the position taken against apriorism by the cognitive structuralist and the other entailments of the ethnography-ethnology distinction.

The Problem of Terminology and Misjoined Arguments

It has been argued, as an entailment of the ethnography-ethnology distinction, that once the social entities of a particular social system have been discovered and described, the term that is applied to such entities is relatively unimportant, for “one can put whatever labels one wishes on the categories isolated, just as in linguistics once a phoneme has been isolated and described the assignment of an alphabetical symbol to it is a matter of convenience” (Goodenough 1956a: 29; see also Keesing 1966a and 1966b for the same argument).

One might be tempted to accuse the cognitive structuralists of cognitive naivete, if the environment in which they are attempting to make this contrast were not known. While it is true that the assignment of a symbol to a phoneme is arbitrary and a matter of convenience -- and this may be extended to include types of social isolates as well-it is only arbitrary in theory. In a practical manner it is arbitrary and a matter of convenience only with regard to the first set of symbols used for a description of the first phonemic system. From then on the specific symbol used becomes an attribute of the category in conjunction with the related sounds which forms the category (cf. Roger Brown 1956: 278). Thus, Lévi-Strauss writes (1963: 91), “the linguistic sign is arbitrary a priori, but ceases to be arbitrary a posteriori”.

However, the cognitive structuralists are attempting to make the point within a particular environment. This environment involves the free-ranging arguments over the definition of conceptual tools such as “descent”, “the kindred”, etc., and their application. The cognitive structuralists are dissatisfied with such arguments and the apparent waste of time and energy entailed (cf. Conklin 1964: 29), but their dissatisfaction for these sterile arguments is for the wrong reasons. Their position is that the system-specific distinctions are paramount and that those used by the anthropologist distort ethnographic reality. But they tend to forget that in the linguistic model, the phoneme, for example, is also the concept of the investigator and has also been subject to scrutiny and redefinition. We should, therefore, not decry arguments per se over definitions, just ones that have been misjoined.
The arguments over the term “kindred”, for example, have been misjoined because any definition of the term should only be viewed as a nominal definition, not a real definition. But the disputants in the argument have reacted as if they were confronted with a real definition rather than focusing on its entailments and its productiveness (cf. Bierstedt 1959).

However, the cognitive structuralists want to avoid the paradoxical situation which arises when arguments take place as to whether the kindred does or does not exist in a particular society. For instance, if in a society we label a distinctive social isolate as a “kindred”, we are faced with the possibility that someone will argue that it is not a kindred. We are then left with the problem of a social isolate on our hands, which is recognized in the society as an important distinction but whose existence in the anthropological profession is denied, since there has been no alternative procedures provided for identifying “kindredlike” social isolates other than it is or it is not an example of the ideal type of kindred. Instead of these, the arguments should have focused on whether an operationalized concept of the kindred provides a useful and productive approach for the discovery of social isolates in any specific system.

SYSTEM-SPECIFIC MEANING AND COMPONENTIAL ANALYSIS

The problem of maintaining a consistent system-specific perspective is not only found in dealing with behavioral systems. It is also found in componential analysis. In building models of a society, the cognitive structuralists are concerned with describing behavior in terms of the meaning it has for the actor and not for the observer. Componential analysis was originally viewed as a procedure for discovering such system-specific meaning, and the elements isolated, it was maintained, had relevance only for the particular system. As such, these elements could not be compared cross-culturally just as, using the phonetic- phonemic analogy, the phoneme /p/ in Malay cannot be compared with the /p/ of English since they only have relevance in terms of the particular structure of the system in which they are found.

However, as Colby points out, the development of a componential analysis for any specific system has leaned heavily on distinctions found in other system (1966: 9).

This conclusion is substantiated by Goodenough’s discussion of the relevance of his model of the Truk kinship system (1964b:221): “The analysis showed that, in addition to the classic list of criteria for discriminating kinsmen first noted by Kroeber (1909) and later elaborated by Murdock (1949), there are others that also may enter into kinship systems . . .” .

However, within the logic of the discourse of the ethnography-ethnology contrast a distinctive feature characteristic of the cognitive structure of a specific system cannot be taken from such a system for cross-cultural use without unacceptable distortion. It is at the phonemic level, not the phonetic. Nevertheless, it still could have cross-cultural relevance, I would argue, if procedures were introduced to generalize it to the etic level, which is founded on different operations than the emic. This is certainly the logical entailment of using the linguistic analogy, and such an approach would appear to be one resolution to the dilemma of extreme cultural relativism. But the processes by which this can be done have yet to be explored by the cognitive structuralists. Thus, it might be argued, on the basis of the logic of cognitive structuralism as presently phrased, that, contrary to Goodenough’s claim, the feature he found in the Lapp system can have no relevance for the English system since the generalization of the Lapp feature to the “phonetic” level has not been explicitly made (1964b).

The position of the cognitive structuralists that componential analysis represented the cognitive world of the target society has been attacked by Burling (1964a). He pointed out that those who claim that componential analysis or comparative methods of semantic analysis can provide a means for discovering how people construe their world must first explain how to eliminate the great majority of logical possibilities and narrow the choice to the one or few that are “psychologically real.”

By 1965 Wallace observed: “The claim that a componential analysis represents a native speaker’s cognitive world is now often avoided...” (1965: 229). And he argued that componential analysis does not yield a unique description of a native cognitive system but instead the psychological validity of the analysis must be established by means independent of the mechanics internal to componential analysis itself. Thus he writes that the investigator:

needs to use additional techniques beyond those of the classic method of componential analysis. First, there must be techniques for identifying dimensions of classification and logical operations which are demonstrably real to the native speaker. Second, there must be techniques for demonstrating that a given logical operation or dimension of classification, however derived by the anthropologist, is not employed in the native speaker’s semantic calculus (1965: 232).

Wallace then presents the procedures that he and John Atkins used to identify the psychologically real dimensions of the Japanese kinship system and the psychologically real logical operations used. Romney and D’Andrade (1964) have also suggested means for testing for the psychological validity of a componential analysis.

Viewed in this light, any componential model that does not include procedures for linking it with the psychological reality of the target society has not truly replicated the cognitive structure of the target society and may be thus considered by some to represent an incomplete emic analysis. (We will shortly discuss further the import of linking formal models with empirical reality.) However, such untested formal models should not be discarded. Wallace (1965: 245-247) points out that they may indicate logical implications of the empirical system, points most likely to yield to cultural innovation or change, and therefore they are useful for predictive purposes.

Goodenough’s position on the psychological validity of componential analysis has evolved, and it would now appear that he considers it to be a false issue in the sense that there is no one true psychological reality for a target society. Thus, he writes:

The criteria by which one chooses one model over another, however, remain to be determined. The very fact that it is possible to construct more than one valid model of a semantic system has profound implications for cultural theory, calling into question the anthropological premise that a society’s culture is “shared” by its members . . . (1965: 259).

Goodenough thus raises a fundamental issue in anthropological inquiry that demands further attention. For at present our analytical constructs, with the exception of Wallace’s mazeway, presuppose the sharing of sociocultural material and do not generally include procedures for eliciting data on the degree to which behavior is actually shared by the members of the society we study.

By 1967 Goodenough had further elaborated on this position. He writes that there is no one “true” cognitive view for anthropological sciences to discover and describe, for people “do not necessarily share a common view but merely have the illusion that they do” (1967: 1207). Componential analysis instead provides “a formal model of the procedures by which people learn what others seem to mean by the words that they use”. Thus, with regard to his analysis of Lapp kinship he writes:

people who use their terms in the same way may still have somewhat different subjective views as to what the terms signify, and if the same person may have more than one view, any componential representation of what the term means, provided it leads us to use them denotatively in the same way others do, is ethnographically adequate.... it would be wrong to assume that the model of Lapp-kinship semantics presented here represents the way individual Lapps actually think about the signification of their kinship terms (just as it would be wrong to assume that the formal statement of a language’s grammar represents the way individual speakers think about that grammar). What the model represents is a pattern of usage, something each Lapp spends a considerable portion of his life learning to understand. Adequate representations of this usage are bound to help us share understanding with Lapps in the same way that Lapps share understanding with one another--and with the same limitations (1967: 1207-1208).

I find Goodenough’s position as represented here not entirely clear. He appears to maintain that the model produced by componential analysis does not represent the psychological reality of a target society; but that it does represent the cognitive processes by which the members of the target society learn to use its terms. These processes would also appear to be universal processes, used by the anthropologist as well; but the members of the target society are not able to articulate these in such a manner as is provided by componential analysis. But between the level of analysis which enables the anthropologist to use the terms correctly and the deeper level of universal psychological processes--so Goodenough’s position seems to indicate--lies an unexplored area of multiple, system-specific psychological realities that are not tapped by componential analyses.

Thus the issue whether or not the goal of componential analysis can or should be to replicate psychological reality or behavioral reality is an important one, but it incompletely defines the nature of the problem. The position taken by Hammer (1966) and D. Kaplan (1968) throws further light on the problem. Their view – if I do not do them too great an injustice by so summarizing it – is that a formal model by its very nature is a simplification of empirical reality, whether behavioral or psychological. Therefore, it cannot replicate empirical reality. Formal models must be first judged as adequate in their own terms, that is by using formal criteria. Following that, the degree of their linkage with empirical reality must be demonstrated. Hammer (1966) takes the position that it is immaterial whether formal models represent the “native cognitive structure”. Instead in explicating the linkage of formal models with empirical reality the issue really is the application, scope, and limitations of the model. D. Kaplan (1968) also points out that formal models cannot generate empirical conclusions, and he criticizes some of Lounsbury’s recent conclusions on these terms.

But before we consider Lounsbury’s approach, I believe the following observations can be made from our discussion of componential analysis. Psychological processes are infinitely more complex than behavioral reality, infinitely more difficult to investigate, and thus it is interesting to note that the most successful attempts in this regard have been done with highly literate subjects. Furthermore, the validity of the psychological-reality issue can be challenged on the grounds that there are in fact multiple psychological realities for any social domain in any society as well as on the grounds that it is a false issue, since all formal models so simplify empirical reality that a duplication of the psychological reality in any domain is in fact unattainable. However, componential analysis can with certain limitations replicate the usage of native categories and, therefore, can open the door to certain behavioral levels of system-specific meaning, if operations to link the formal model to the behavioral level of reality are used.

The entailment of this is that while formal analyses of published data or of data removed from field situations may help develop more powerful formal models, such analyses do not necessarily delineate system-specific reality. This can only be done in the field by constantly checking one’s model against behavioral events. Furthermore, such second-hand data may include “cultural pseudomorphs”, spurious cultural forms that combine ideological items from the culture of the investigator and the culture of the target society (Appell n.d.), particularly if the collection of the data was not done in terms of the goals of cognitive structuralism (see footnote 26).

THE EXTENSIONIST APPROACH OF LOUNSBURY

Up to this point we have not considered the work of Lounsbury for two reasons. First of all, he has not, to my knowledge, made any statements as to the relevance of his work to the ethnography-ethnology distinction or to most of the other epistemic issues in cognitive structuralism we have been considering. Secondly, following his early, ground-breaking contribution to the development of componential analysis (1956, 1964a), he has in recent analyses of kinship systems veered away from this to develop what has been called the “extensionist” method. And this new and powerful method has been viewed on one hand as complementary to componential analysis and on the other as contradictory to some aspects of the position taken by the cognitive structuralists. Thus at this point a discussion of Lounsbury’s recent approach will throw light on many of the issues of cognitive structuralism we have isolated.

Lounsbury describes his approach as follows:

We may consider that a “formal account” of a collection of empirical data has been given when there have been specified (1) a set of primitive elements, and (2) a set of rules for operating on these, such that by the application of the latter to the former, the elements of a “model” are generated; which model in turn comes satisfactorily close to being a facsimile or exact replica of the empirical data whose interrelatedness and systemic nature we are trying to understand. A formal account is thus an apparatus for predicting back the data at hand, thereby making them “understandable”, i.e., showing them to be the lawful and expectable consequences of an underlying principle that may be presumed to be at work at their source (1964b: 351).

The assumptions necessary for his method of formal analysis are explicitly stated by Lounsbury (see 1964b: 381-382, 1965: 149-152). First of all he makes the assumption that every kinship term has a primary referent and that this primary referent is the genealogically closest kin-type from the class of those covered by the term. Secondly, these classificatory kinship terms have, beside their primary referents, secondary referents consisting of more distant types of relatives. These secondary referents may be rephrased in terms of their primary referents by one or more genealogically-based equivalence rules. Thus, Lounsbury states his position is both Murdockian and Malinowskian. It is Murdockian in that he assumes the importance and near universality of the nuclear family; and it is Malinowskian in that he assumes that the relations of kinship and thus the basic meaning of kin terms derive from the primary relations that arise within the nuclear family.

However, Lounsbury is not concerned with developing a methodology for producing a “valid ethnography” in the terms of cognitive structuralism. Instead his interests lie in other directions. He hopes that his formal method “may eventually make it possible to do ‘controlled comparison’ with sufficient rigor so that the results can be meaningful” (1965: 181). He also hopes this method will lead to more adequate functional explanations. Thus, he considers his results to be relevant to problems of social structure, and he relates his extension rules to “social facts”, such as rules of succession (1964b: 382-383). Consequently, the discovery and statement of the extensionist rules for any society “constitute a necessary preparation for a proper sociological explanation. . .” (1965: 175).

One of the issues we have been concerned with is the relation of such formal models to the system-specific distinctions of a target society. There is evidence that Lounsbury believes his transformation rules do isolate system-specific distinctions for he refers to his rules as “semantic rules” and his method as “semantic analysis”. But whether they replicate behavioral reality or psychological reality is not completely clear. There is some indication, however, that they might in fact have relevance at the psychological level for Lounsbury indicates that componential analysis is a method for determining overt classificatory equivalences while his extensionist analysis is a method of discovering covert equivalences (cf. 1965: 151).

But it should be noted that Lounsbury has not become involved as yet in the issue as to whether such formal models are sufficient to replicate system-specific reality or whether other operations must be included to link the transformational analysis with empirical reality. D. Kaplan sums up the problem as follows:

Methodologically, a formal analysis is useful because it may reveal significant relationships among the variables of a system of which we might have been totally unaware. But the formally possible is not the empirically probable, and if it is scientific theory we wish to develop, the former is useful only to the extent that it leads us to the latter. Until a formal analysis is made substantive by being given an empirical interpretation, it can only remain an interesting logical exercise (1968: 247).

Certainly, if we consider that identical terminological systems are embedded in markedly different sociocultural systems, a serious question is raised as to just what level the methods of componential analysis and the extensionist approach do replicate in sociocultural systems. And for those cognitive structuralists who take an extreme position on cultural relativism, this is a crucial but as yet unanswered question.

Lounsbury’s Approach and the Etic-Emic Analogy

Additional light is thrown on Lounsbury’s position if we consider how his approach fits the phonetic-phonemic analogy, and a consideration of this will also further our inquiry into the nature of anthropological concepts. However, first it should be made clear that Lounsbury’s view of anthropological inquiry is hardly isomorphic with that of the cognitive structuralists. For example, even though the analytical techniques he has developed form an important part of the methodology of the cognitive structuralists, he does not take an extreme position on cultural relativism. In fact he suggests that cultural relativism rather than being considered as a doctrine should be in fact an empirical question for whatever cultural domain is under investigation (cf. 1965: 182). Thus he argues that contrary to what others have maintained, the use of genealogical concepts in analyzing kinship data does not impose an ethnocentric bias. For if it did, “how is it that the end product of all this wrongheadedness turns out to be an internally consistent, simple, and accurate account of the ethnographer’s collection of data. . ., rather than ending in total confusion and ‘utter incomprehensibility’ as we ought to expect?” (1965: 182).

Lounsbury’s answer to this question is:

The genealogical frame of reference was the correct one after all; the extensionist hypothesis was right in the first place; and the nuclear family (which is fundamental to both of these) is probably--in a certain minimal but basic sense--just about what Malinowski and Murdock have judged it to be. (To admit of this universality in essential elements of the nuclear family, however, is by no means to imply uniformity in the cultural interpretation and legal attributes of these elementary relationships in different societies.... In the foregoing analysis we have caught a glimpse of some of the many variables [the different equivalences, their context restrictions, the solidarities, their relative priorities, etc.] that can take on different values within this elemental frame when it is culturally interpreted and endowed with specific social attributes.) (1965: 182).

This also might be interpreted that Lounsbury believes he has, if phrased in terms of the phonetic-phonemic analogy, constructed the fundamental etic grid for kinship analysis.

ETIC GRIDS AND ABSTRACT ANALYTICAL SYSTEMS

However, let us now look more closely at these concepts and distinguish various levels. First, there are etic grids that pertain to specific sociocultural domains. These grids incorporate the whole range of possible distinctions found in human societies, and as such they are essentially an inventory of ethnographic data. Etic grids, however, differ significantly from other cultural inventories in that they are constructed from the system-specific discriminations found in the range of world cultures and are not composed of data derived from the use of culture-bound tools of analysis.

It is also important to note that the distinctions or grid items incorporated into any etic grid may be of two types: universal items or particular items. A universal item appears in all cultures; a particular item pertains to one or more cultural systems but it is not a discrimination found in the full range of societies. Thus, while the items in an etic grid are not all universal, the grid itself is universally applicable to all societies for the purpose of identifying system-specific discriminations.

Grid concepts are complementary to but have to be distinguished from what I here call abstract analytical systems. Phonemic analysis illustrates this latter type. Such analytical systems contain categories of discriminations that are found in all societies. They are composed of universal entities and the necessary relations between such entities, thus forming an integrated system. These conceptual entities are, in contrast to etic grid entities, formal in nature and thus devoid of any substantive matter. They carry no cultural content but instead contain observational procedures by which the content of any particular cultural system can be discovered and described.

Lounsbury’s extensionist method consists of just such an abstract analytical system: primary kin referents and operations to extend and transform these to account for the whole range of kin terms in any system. The basis, he maintains, for its universality lies in its genealogical frame of reference. Furthermore, Buchler (1964: 311-312) suggests that the extensionist approach is based on universal cognitive processes of encoding “kernals” plus appropriate transformations in order to produce meaningful combinations.

Lounsbury has also constructed an etic grid for this analytical system, although it is probably not yet complete. Thus, certain of his transformation rules are universal; others, he states, may have only a wide generality, such as his “step-kin” and “in-law” categories. And as such they do contribute to the etic grid.

THE RELEVANCE OF COMPONENTIAL AND EXTENSIONAL ANALYSES TO SPECIFIC CULTURAL DOMAINS

In contrasting his extensionist method with componential analysis, Lounsbury has written that the former is “the method of total class definitions; the other is a method of basic member definitions and supplementary rules of extension” (1964a: 1088). In the former method “differences of degree within the class of denotata of a term are of course recognized, but these are treated as ‘nondistinctive’ “ (1964a: 1088). And in general, the cognitive structuralists have considered these two methods to be complementary (cf. Goodenough 1967: 1208; Romney 1965).

However, there is some evidence that these two methods may not be complementary with respect to any particular cultural domain but that their relevance has to be established in each case. And this brings us back again to the necessity of establishing the linkage of a formal model with its empirical referents before it can be maintained it replicates in any fashion “behavioral reality” or “psychological reality”.

Bright and Bright (1965) found among certain California Indian tribes that a “center-oriented” type of classification more appropriately reflected the nature of the empirical data. They write:

We see, then, that a hierarchical model, which shows only the relationship of domination..., cannot account adequately for the Indian taxonomies. In a hierarchy, an item either is or is not a member of the class named by the next higher node. But there is no way of indicating, in a hierarchical tree, the situation where a specific term like Yurok tepo “fir, tree” ... can also be used as a generic term, thus including other trees which resemble the fir by being coniferous. In addition, there is no way of indicating when an item is classified in a certain way because it is “like” another item which is more central to the focus of the domain in question. Therefore, although our European hierarchic taxonomies can be represented for the most part by a branching tree..., the aboriginal taxonomies of northwestern California can be represented more faithfully by a kind of “sphere of influence” model (1965: 253).

Dentan (1970) also raises similar questions in dealing with taxonomies among the Semai of Malaya. He finds that labelled categories sometimes fade away at their boundaries so that instead of being defined by the presence or absence of a particular characteristic they are defined by the degree to which a characteristic is present. Dentan concludes that there are two major processes of learning, discrimination and stimulus generalization; and he suggests that componential analysis is an appropriate tool for dealing with those areas in which the learning has been by discrimination. For domains involving polysemic categories for which there appears to be an ideal type, Dentan suggests that the processes by which they are learned may be based on stimulus generalization; therefore in the analysis of such categories an extensionist model might be more appropriate.

The work of Berlin and Kay (n.d.) on color taxonomies also suggests that the concept of focal or ideal types with extensions out to blurred boundary zones may be more appropriate for the analysis of some domains while others may be more profitably studied in terms of boundary-identifying methods and hierarchical taxonomies.

Berlin and Kay’s (n.d.) work also indicates that the extreme position on cultural relativity needs to be reconsidered by the cognitive structuralist. And Keesing, after considering the results of Lounsbury’s approach and those of Berlin and Kay, concludes, in a departure from his previous position, that if we look for focal types “we might find that the semantic categories of different cultures are much more similar than comparison of their boundaries has yet revealed. Recurring categories in different cultures with common focal types may prove to be much more frequent than we have suspected, even though variant principles for extensions make the categories of different cultures non-isomorphic” (1968: 66).

We will discuss in more detail later the nature of cultural relativism implied in the ethnography-ethnology distinction. However, let us here, as the result of our review of the extensionist method, again draw attention to the following points. The methods that were used early in the development of cognitive structuralism and which were proclaimed to be culture-free and uncontaminated in that they represented the system-specific distinctions of the target society now do not appear to be so culture-free or as system relevant as they once did, without the introduction of supplementary procedures to anchor them to the empirical reality of the target society (also cf. Bright and Bright 1965: 258).

SYSTEM-SPECIFIC MEANING AND THEORY MEANING

Many problems remain to be resolved in the methodology of the cognitive structuralists. They have not yet dealt with the problem of social and cultural change. Instead all attempts to date at componential and extensional analysis have involved the implicit assumption of system stability and integration. Yet there is evidence that in situations of sociocultural change the terminological behavior of kinship systems may remain relatively stable while other types of behaviors associated with the terminological system may change relatively rapidly (cf. Lundsgaarde 1967).

Furthermore, the cognitive structuralists have not yet devised methods for handling functional ambiguity, as Harris (1968) has pointed out. Certainly not all domains have either neat internal boundaries or logically consistent rules for the transformation of core or type terms.

One of the more important unresolved methodological problems as we have pointed out has to deal with the level of empirical inquiry. Can the methodology of the new ethnography replicate behavioral or psychological reality? Is it enough to devise a formal model of system-specific discriminations so that correct behavioral predictions (Burling’s position) or anticipations can be made; or should the new ethnography strive to determine the system-specific psychological processes?

However, the issue of psychological reality may in fact be wrongly phrased. Before the demand for psychological relevance had arisen, Wallace (1961c) raised the question of the degree to which cognitive sharing occurs in any society. He concluded logically there need not be sharing of “cognitive maps” of behavior, a position that Goodenough also appears to subscribe to (cf. Goodenough 1967). Thus, before the problem of the psychological reality of formal models can be adequately dealt with, the degree to which cognitive sharing occurs has first to be resolved. Inasmuch as the nature of psychological reality has yet to be established and the methodology for ascertaining it devised, in this essay I shall restrict the term “system-specific meaning” from this point on to refer only to the behavioral level of reality.

There is yet another important epistemological problem that remains to be discussed. This is the relation between system-specific and theory-meaning explanations of behavior, and various aspects of this problem will concern us for the remainder of this discussion of cognitive structuralism. The term “theory meaning” is here used to refer to descriptions and interpretations of behavior in terms of its meaning or significance to the observer (A. Kaplan 1964) in contrast with its meaning or significance to the actor--system-specific meaning.

To establish the necessary contrast between the new ethnography and the old ethnography based on comparative concepts, the cognitive structuralists have focused on system-specific meaning and have not adequately explored the distinction between system-specific meaning and theory meaning or the relevance of the latter for ethnographic inquiry.
For instance, Frake writes:

First, it is not, I think, the ethnographer’s task to predict behavior per se, but rather to state rules of culturally appropriate behavior.... In this respect the ethnographer is again akin to the linguist who does not attempt to predict what people will say but to state rules for constructing utterances which native speakers will judge as grammatically appropriate (1964a: 133).

Frake’s dissatisfaction with prediction is elucidated in a further article. He writes:

This conception of a cultural description implies that an ethnography should be a theory of cultural behavior in a particular society, the adequacy of which is to be evaluated by the ability of a stranger to the culture (who may be the ethnographer) to use the ethnography’s statements as instructions for appropriately anticipating the scenes of the society. I say “appropriately anticipate” rather than “predict” because a failure of an ethnographic statement to predict correctly does not necessarily imply descriptive inadequacy as long as the members of the described society are as surprised by the failure as is the ethnographer (1964b: 112).

Frake has made an important point in drawing attention to the significance of predictive failure on the part of the members of a target society. But he has left in limbo, at least for the moment, any theory meaning interpretations of behavior which would permit the observer to predict events on the basis of the pansophic conceptual tools and knowledge he possesses but which the members of the target society do not possess.

We have thus introduced the distinction here between behavior viewed in its system-specific meaning and behavior viewed in terms of its meaning in a theory of the investigator, because it is obvious that a full understanding of any particular sociocultural system cannot be achieved through its own cognitive distinctions or through a universal model built from the distinctions found in any number of sociocultural systems (unless the cognitive distinctions found in the society of anthropologists are also included). For instance, the approach of the cognitive structuralists to ethnography does not lead to an understanding of the unintended and unrecognized entailments of a specific system, which is one of the major contributions of structural-functional theory (see Robert Brown 1963 and Jarvie 1964). Nor does this approach lead to an understanding of the process of adaptation of a specific sociocultural system to its ecosystem and the unintended entailments or limitations of its particular adaptation. The approach of system-specific meaning, for example, might not tell us how subincision may function to limit population growth in a specific environment, how divorce in a particular society may be related to brideprice, and how sorcery and witchcraft may serve in the management of conflict and tension.

Cognitive Structuralism and Social Anthropology

But these criticisms of the limitations in the approach of the cognitive structuralists, not unique to me, ignore the problem that the cognitive structuralist and the social anthropologist (in the British sense) are concerned with constructs that are on different conceptual levels. We have adumbrated this previously when we discussed the difficulty of the cognitive structuralists in dealing with action systems. Now that we have analyzed some of the analytical procedures of the cognitive structuralists and the entailments of these, it will be worthwhile to contrast the methodologies of the cognitive structuralists and the social structuralists further.

The cognitive structuralist, largely as a result of his concern with sociocultural phenomena in terms of their system-specific meaning rather than in terms of any theory meaning, deals with ideational systems, and these are viewed as being the property not of a community but of its individual members (Goodenough 1964a:11-12). The social anthropologist generally starts at the other end of the behavioral event and is concerned with those types of action systems that have relevance for the theoretical construct of a social system. As such, the construct of a social system tends to focus on observed behavior, but not entirely so, for it also implies the existence of jural rules and other constraints on behavior that form part of the shared ideational system of a community. And to the degree these rules and constraints are made explicit, the more valid the model is for any particular society.

Because the cognitive and social structuralists start from opposing points, the social structuralists tend to criticize the cognitive maps drawn by the cognitive structuralists as not being rooted in social reality. That is, the cognitive structuralists have not demonstrated the social distribution of the maps drawn and the degree to which they are actually shared by the members of a social system.

On the other hand, the cognitive structuralists tend to criticize the models of social systems constructed by the social anthropologists in that it is not clear in many cases to what degree they actually recognize and include the cognitive maps of their members.

This criticism is particularly appropriate in one sense, since there is a tendency in describing social systems to force the observed data into ideal types of social units such as unilineal descent groups, etc., which are used as if they had cross-cultural validity. The result of this is that the system-specific social isolates are either ignored or distorted. For example, Firth (1963) defines the kindred, cognatic stock, and other concepts so that it is not clear whether the social units he is concerned with are the hypostatization of the anthropologist or the members of the society being described. Radcliffe-Brown’s approach to corporate descent groups is similarly confused (1950: 41).

This distinction between the conceptual levels of the cognitive structuralist and the social structuralist throws further light on current arguments over the concept of descent. One position is exemplified by Goody (1961). He confines the term descent to eligibility for membership in kin groupings. The other position is that taken by Barnes (1962) and as a development out of the work of Davenport (1959) and Peranio (1961), that taken by Scheffler (1966). Barnes makes the distinction “between filiation as a mechanism of recruitment to social groups and to ascribed relationships and descent as a sanctioned and morally evaluated principle of belief” (1962: 6). Scheffler also views descent in any society as an ideational construct of that society which forms the basis of rules regulating the transmission of social status. The entailment of these views, if we accept the approach of the cognitive structuralists as to the relativity of cognitive systems, is this: elements in the cognitive structure of any society that are found in the domain of what we may for convenience refer to as “descent constructs” (Scheffler 1966) have relevance only for that specific system in a like manner that its phonemes have relevance only for the phonemic system of that society. As such, any element in the system of descent constructs can only be adequately defined in terms of its relation, its contrasts, to other elements in such a system. Consequently, the cross-cultural classification of social groupings derived from the isolation of a particular cognitive element in one such system, as for instance patriliny, can only distort ethnographic reality. This is well illustrated by the problems that have arisen in dealing with descent in New Guinea on the basis of African models (see Barnes 1962).

Thus, the conclusion can be drawn that, at the present stage of our study of sociocultural phenomena, such a cross-cultural typologizing of social groupings on the basis of cognitive elements has limited productivity in contrast to an approach involving analytical operations of universal relevance that are devoid of cultural loading but which instead provide the procedure for the delineation of system-specific content.

However, the social structuralists cannot be considered to be wholly wrong for they recognize the importance of social groupings for the organization of society, a universal element, and in particular corporate social groupings. Corporate social groupings appear to be almost universal phenomena and are found even in cognatic societies in which descent constructs are not used as a basis for establishing membership in social groupings (see Appell 1965 and 1968). But, as we have seen, the social structuralists made the mistake of dealing with such social groupings in terms of ideational constructs which primarily have relevance only for the specific system from which they are derived. If, instead, the ordering of phenomena pertaining to social groupings had focused more on universal processes such as the ownership of scarce goods rather than on descent, the social structuralists would not have been in such a vulnerable position when their models were applied to other societies. They also would not have been open to criticism for overemphasizing kinship to the disregard of other phenomena (see Leach 1961b).

System-specific and Pansophic Explanations for Behavior

One of the important fallouts from the position taken by the cognitive structuralist is not that system-specific meaning is king and theory meaning of doubtful value. Instead it is that each has its place in ethnographic inquiry. Furthermore, investigations at the system-specific meaning level must be kept separate but coordinate with those at the theory-meaning level, and both approaches are necessary for a full and productive ethnographic description. When explanations in terms of theory meaning are not distinguished from those in terms of system-specific meaning, or where the two levels are indiscriminately intermingled, confusion can only result.

It is therefore obvious, but it still needs explicit statement, that certain anthropological problems are more appropriately investigated at one level than at another. Those problems that depend on a rational model of human behavior for their explanation can, of course, only be dealt with in terms of situational logic (Jarvie 1964), the logic of the specific system under consideration. This is specifically the case in attempts at explaining any behavioral event that involves choice. Other types of explanation that are not concerned with or do not accept a rational model of human behavior, such as those which postulate a universal structure to the unconscious (see Freeman 1968), appropriately need not maintain a consistent system-specific stance. But even in those inquiries involving choice, system-specific explanations for a behavioral event certainly does not exhaust its ethnographic interest. Other levels of explanation for the event that depend on the investigator’s accumulated knowledge and total range of conceptual tools – pansophic explanations – need also to be considered.

This brings us to the observation that the cognitive structuralists have yet to come to grips with such psychological processes as rationalization and unconscious motivation. Thus, analyses of stated explanations for behavior do not necessarily identify real motives. This is not to say that such analyses do not deal with important cultural aspects, but their relevance for the cultural system may not be ascertained until real motives are determined by field techniques other than dealing with verbal behavior. Let me give a very simple example that may elucidate certain aspects of this problem. Among the Rungus of Borneo in response to the question as to why place of residence was changed, the reply will usually be to gain access to better swidden areas. However, quite frequently the real motive is to move out of the grasp of malevolent spirits that are causing illness within the family. To mention the real motive would alert these malevolent spirits and might draw them to the new domicile of the family.
Unconscious motivation may also produce culturally important but undiscriminated behavior. That is, unconscious motivation might not produce functioning or recognized categories of behavior within the society itself. While it might be argued, on the other hand, by those who use a linguistic model for cultural analysis that such a model does in fact isolate system-specific discriminations below the conscious level of awareness, these are nevertheless functioning categories. They are discriminations that evoke response even though they are below the conscious level of awareness; but I am concerned here with significant behavior that has yet to be discriminated at any level by the members of the cultural system. It is these types of psychological processes that can produce patterns of covert behavior. However, Frake’s concept of a “valid ethnography” does not encompass such behavior (1964b: 112). And Conklin’s statement that “we try to base our work on such concrete realities as a local group of people and the kinds of objects and events the members of this group treat as culturally significant . . .” (1964: 26) also seems to exclude covert behavior patterns.

The behavior of the eldest female in a sibling set among the Rungus of Borneo provides one example of the types of covert behavior to which I refer (see Appell 1965 and 1969a). The eldest female after her wedding accepts the sexual advances of her husband without initial protest significantly more frequently than do her sisters, and she resides in a village away from her mother also significantly more frequently than do her sisters. This behavioral pattern was discovered by a statistical analysis of Rungus behavioral events as there is no recognition in Rungus society that the behavior of the eldest female in the environment of relations with males differs or contrasts in any way from that of noneldest females.

Cultural Relativity and Scientific Procedures

As we have pointed out previously, there is a strong current of extreme cultural relativism in the work of the cognitive structuralists with but few exceptions. One aspect of this is expressed in the use of the phonemic model as a paradigm for their research. Consequently, it is worthwhile at this point to return again to this paradigm and analyze more closely its nature. We will thus attempt to summarize our views on the doctrine of cultural relativism.

In essence the phonemic model uses difference in meaning as a discovery procedure to isolate the elements of any specific system, and, it has a structure that depends on contrast between these elements. As such, any element in the system, it is maintained, cannot be torn from the system for comparison with equivalent elements in other systems. But not all analytical systems, not all theoretical constructs, dealing with sociocultural phenomena are structured in these terms. Thus, there appears to be no cogent reason why elements in such systems not based on difference in meaning or structured by contrast cannot be fruitfully compared cross-culturally.

Furthermore, the phonemic model has been oversimplified in its use as a paradigm for the study of cognitive systems and as a justification for the alleged noncomparability of such systems as well as other sociocultural phenomena. Not only have whole phonemic systems been compared but also elements from such systems. For instance, Wallace (1961a) examines the number of phonemes in various languages of the world in his study of universal cognitive processes. And historical linguistics also makes extensive use of the comparison of individual phonemes between related languages (also cf. Casagrande 1963).

But the relativity of cognitive systems is not only thrown into question by historical linguistics. The ethnographic evidence from highland Burma (Leach 1954) also casts doubts on an extreme position for cultural relativity. In highland Burma it appears that the cognitive systems of ethnic units can only be properly understood in terms of their contrasts with those of neighboring ethnic units.
Thus, because of the position taken by the cognitive structuralists on the noncomparability of sociocultural phenomena, the new ethnography in some respects can also be called the new relativism. And where this does apply, the criticisms of the old cultural relativism still hold (cf. Kluckhohn 1953: 520). For instance Wallace writes: “Indeed, a radical linguistic relativism would probably be, by its own axioms, not only incapable of proof but incapable of being described” (1961b: 142).

However, the problem of relativism lies not so much in its ultimate truth but in the approach taken by anthropologists towards this problem. There are instances where it appears to be considered as an article of faith, rather than a part of the conceptual frame (A. Kaplan 1964: 159-161) or “text” (Robert Brown 1963: 147-148) to system-specific experiments.

Thus, when I say that the cognitive structuralist has not made explicit the conceptual frame or text of his “experiments”, i.e., his system-specific investigations, I mean that he has not given an accurate statement of the conditions or qualifications under which his hypotheses are supposed to hold. He has not given an accurate statement of what variables are being considered as constants for the purpose of the investigation and the limits within which his hypotheses may be considered valid. Certainly cultural systems are not formed in a complete social and historical vacuum. And if they were, “If individual men or whole peoples dwelled alone in incommensurate worlds constituted only by their unique experiences or by those shared within the bounds of isolated communities, communication among men or among peoples would be impossible” (Casagrande 1963: 294).

On the other hand, if we accept W