"Discussion of Lévi-Strauss's paper 'Social Structure'" |
| ([1952] 1962) in An appraisal of anthropology today. Edited by Sol Tax. University of Chicago Press 1953 |
Participants in this discussion were Boyd, Brew, Fred Eggan, Greenberg, Clyde Kluckhohn, Kroeber, Robert Murdock, Margaret Mead, Nadel, Shapiro, Washburn, Willey
Dr. Eggan mentioned that the term "process" does not appear often in the inventory papers being discussed. I must confess that, as a foreigner, I had difficulty in understanding exactly what was meant by the opposition between "process" and "result," and, although I was not quite clear what was meant by "process," I assumed that it was something different from result and for that reason, I left results completely out of my paper.
This is to be regretted, because in this symposium we have no paper assessing what we actually know--the results. This is a very, serious gap in our program, since cultural anthropology . . . is certainly one of the most important fields, in which dozens and dozens of anthropologists are working. We are incurring the risk that we will not know what we have achieved.
Dr. Lévi-Strauss's paper presents certain difficulties, the reverse of the difficulties which Dr. Lévi-Strauss a moment ago mentioned--that he had with the concept of process. Although anthropologists in different countries are commonly talking about the same thing, they frequently use different terminology . . . and much of the difficulty that some readers have had with Dr. Lévi-Strauss's paper revolves around . . . the use of certain terms and forms of expression which may initially rub an American the wrong way.
Dr. Lévi-Strauss cites Dr. Kroeber with respect to the use of the term "social structure." "The term, 'structure,"' as Dr. Kroeber says, "appears to be just a yielding to a word that has a perfectly good meaning but suddenly becomes fashionably attractive for a decade or so, like 'streamlining,"' and so on. I would agree with Dr. Kroeber here. I do not lay much stress on the words that are used . . . what we mean by those words is important. Dr. Lévi-Strauss has been good enough to pass on to me a letter which RadcliffeBrown wrote commenting on his paper, and I shall read a few of these comments.
"As you have recognized," says Radcliffe Brown "I use the term 'social structure' in a sense so different from yours as to make discussion so difficult as to be unlikely to be profitable. While for you, social structure has nothing to do with reality but with models that are built up, I regard the social structure as a reality. When I pick up a particular sea shell on the beach, I recognize it as having a particular structure. I may find other shells of the same species which have a similar structure, so that I can say there is a form of structure characteristic of the species. By examining a number of different species, I may be able to recognize a certain general structural form or principle, that of a helix, which could be expressed by means of logarithmic equation. I take it that the equation is what you mean by 'model.' I examine a local group of Australian aborigines and find an arrangement of persons in a certain number of families. This, I call the social structure of that particular group at that moment of time. Another local group has a structure that is in important ways similar to that of the first. By examining a representative sample of local groups in one region, I can describe a certain form of structure.
"I am not sure whether by 'model' you mean the structural form itself or my description of it. The structural form itself may be discovered by observation, including statistical observation, but cannot be experimented on.
"You will see that your paper leaves me extremely puzzled as to your meaning. In dealing with Australian kinship systems, I am really only concerned with arriving at correct descriptions of particular systems and arranging them in a valid typological classification. I regard any genetic hypothesis as being of very little importance, since it cannot be more than a hypothesis or conjecture."
One brief comment on the last point: I feel that this position . . . is the greatest source of weakness in Radcliffe-Brown's work. With respect to social organization or social structure, we must always consider a structure moving through time, and our attention should be focused on a dynamic moving and changing equilibrium. . . . The diachronic aspect is essential to an understanding of the synchronic aspect.
To return to the first part of Dr. Radcliffe- Brown's comments, I think I understand what Dr. Lévi-Strauss means when be distinguishes social structure from reality. . . . As I see it, structure is an organization or a framework that has some permanence. If one makes an observation at a given moment . . . one has merely a number of concurrent elements that happen to be found together, as, for example, in a photograph. It is only when one observes numerous instances and sees what is constant and enduring and repetitive and what adheres to what that one discovers structure; so that RadcliffeBrown's finding structure in what he observes at a given moment is adopting a concept of structure which, to me, is relatively meaningless and useless.
In Dr. Lévi-Strauss's paper, he uses structure" and "social structure" in a number of different senses. . . . He uses the terms "model"; to him the social structure is a model rather than the reality; it is what you see persisting and repeating, He uses "model" in the sense of a descriptive model sometimes. . . . One goes out and observes a society and, through one's researches, builds up a model of that social system. This is not what one finds in any moment. This is what comes out of a great deal of observation.
Dr. Lévi-Strauss also points out that the people studied will themselves have a model of their own system, and this model will often be different from the model that the anthropologist constructs. We have an excellent example in Evans-Pritchard's work on Azande witchcraft and religion, in which Evans-Pritchard found it necessary to go below the conceptions which the Azande had, in order to construct a model which would explain Azande religion.
Another case would be that of Lloyd Warner in his work on American class structure. Americans traditionally deny the existence of a class structure. The scientist determines that Americans have a class structure, in that the behavior of Americans can be understood only in terms of a class structure. Thus one constructs a model that explains, to the maximum extent possible, the phenomena that one observes.
There are also models that one constructs in interpretation. There are historical models; when one attempts to understand culture changing over time, one builds a model which organizes the data of history, maintaining the data themselves, as Professor Kroeber has pointed out, and holding them together in a meaningful configuration. The scientist also constructs models in scientific theory, in which case he commonly rejects the phenomena after he has used them and preserves only the tested theory.
Dr. Lévi-Strauss makes a distinction between mechanical and statistical models, and here he touches upon something of the utmost importance in social science. He himself mentions that sociologists are primarily concerned with statistical models. . . .
A point of argument between sociologists and anthropologists is with respect to statistical and mechanical models. . . . I might give an example from Dr. Lévi-Strauss's paper, in which he takes our marriage regulations in a modern Western society and points out that the prohibitions regarding marriage conform to a mechanical model, that certain specific relatives are excluded, and we can always count on these incest taboos prevailing. But with respect to permitted marriages, there is an enormous range of possibility as compared with a society in which one must marry one's mother's linked brother's daughter, or another woman who is a substitute therefor. In such a case, you would have a mechanical model fully explaining and accounting for the behavior.
To illustrate the distinction and the importance of this for sociology and anthropology: In our work on social organization in Truk, we found that the natives gave us the rule of residence in marriage as bilocal or ambilocal. . . . In considering all marriages, by a census of households and working through genealogies, we found that about 85 per cent of all marriages today and over the past are matrilocal and about 15 per cent patrilocal. There you have the statistical model. . . .
Our objective was to convert the statistical model into a mechanical model, so we studied the cases of patrilocal marriage, in order to find under precisely what conditions they occurred, and we were able to determine that patrilocal residence takes place only under very specific circumstances, namely, when there is not a large enough number of matrilineally related women to maintain a functioning matrilocal extended family, in which case the woman, on marrying, goes to live with her husband, his sister, mother, and so on. In other words, we were able to reduce statistical model to mechanical model.
Ordinarily, sociologists think that they have gone far enough when they have constructed statistical models. Anthropologists, I think, are more sophisticated than sociologists, in that their ideal is to convert statistical into mechanical models. This point has been specifically stated in Dr. Lévi-Strauss's paper.
Professor Murdock talked about structures moving through time. Is the concept of co-tradition related to this, for example?
In relation to some models, I may say I do not view the interrelation of fields as one of taking up words and carrying them across. I think structure in anatomy is a very different thing from structure in social systems, because the structure gets back to different ways of defining structure.
But, taking up the models used in modern genetics-Sewall Wright's models, we will say, of how evolutionary systems work-one would have to have information on the mating system, the amount of inbreeding, classes, breeding isolates, population size. These would be what one would have to have from the student of society in order to apply Wright's statistical models, in order to know what was going on in these societies genetically.
Likewise, if the social anthropologist has an idea, say, about the amount of inbreeding, if a society is fairly closely inbred, this would throw off Dr. Boyd's gene frequencies, and therefore he would not be able to tell the social anthropologist whether the system had operated the way the social anthropologist thought it bad. if a closed system of inbreeding is the ideal of a society, then this will change the gene frequencies in a perfectly predictable, definite way, provided that people did what they tell you they did. Now the chances are that they did not, but this now can be made a matter of observation and not bunch.
It would not be the gene frequency, but only the phenotype frequencies.
There is one other point about the use that Dr. Lévi-Strauss has made of the conception of models that 1 think is likely to confuse some American workers. . . . Professor Lévi-Strauss has specifically drawn on cybernetics in a different way from the way that a group of us who have been working closely with people at M.I.T. have been drawing on some of the same material. We are including . . . the model in the engineering sense, that is, the model that is built by a set of engineers to a set of specifications, whether those specifications are drawn from a whole ecological system, as, for instance, the model of Ashby, or whether it is a computation machine that is built on a set of specifications about human memory or human capacity to sort, translated into mechanical operations.
A whole group of people, some anthropologists and some not, have been influenced by these experiments . . , and the term "mechanical model" therefore is changing its general communication meaning.
In my paper I refer to living models as opposed to these machine models that are built according to a limited set of abstractions. . . . Professor Lévi-Strauss and Professor Murdoch are talking about what I call a "living model," except that this living model that we are coming back to is actually really identified persons. So when we talk about a living model we want to build either an experimental situation . . . or look at a living community with specified human components, each one of which is really identified. . . .
Unless we can keep these three or four points clear, I think there will be ultimately some confusion. As I understand Professor Lévi- Strauss's mechanical model, as applied, say, to a social organization situation, it would not contain identified persons but would be a picture of a person who might be a man with two sisters, two brothers, two aunts, and so forth, but he need not be an actual one.
Going through these machine models is a way of learning. We want to keep our human components identified, so we can carry our unanalyzed variables as we proceed from one spot to another. . . . If you abstract from your known social system and build a mechanical model in that sense, you can only put in what you have been able to extract and you throw away the rest.
What substitutes do the archeologists have for this?
Well, in the depths of archeology, the term "model," as used by Dr. Lévi-Strauss and Dr. Murdock, is completely analogous to our use of the term "type," as opposed to the tangible reality of the artifacts.
T: The archeologist obviously deals with artifacts. . . . Any interpretation
has to be an abstraction. It cannot become a living model, since there
is no way of getting at the people
themselves, I once thought of the parallel
more in terms of Max Weber's ideal type, which is an abstraction based
on considerable reality, and then attempting to apply that ~nce again
to the archeological interpretations.
I would like to say that the co- tradition is obviously a constructin old- fashioned language, historical reconstruction or an interpretation of the data.
Yesterday, I rashly said that I never used the word "model." I find that Professor Lévi-Strauss quoted me as having used "model," particularly in connection with a small-scale model of comparative balances. I was using it in a naive sense, as you have a model railroad.
I have a brief list of four-and, perhaps, now, with the living model, five -different senses in which the word can be and has been used by Professor Lévi-Strauss. . . .
One kind of a model is a machine built to specifications on which you can study more easily, exemplify, and illustrate more easily rather complicated conditions. A variation of that would be a machine constructed according to specifications in a different discipline which you find illustrative and useful in your own discipline. I do not think that the difference between mechanical and statistical applies to that at all, because lots of statistical observation may go into your construction and specification of the machine.
A second way Professor Lévi-Strauss uses "model" is: model equals norms. When he speaks about cultures and unconscious models and people having a different appreciation of what be calls models of his own society, what he really means is that we observe a certain society, from which, through essentially statistical observations, we derive a certain norm of behavior in, say, marriage rules. The people have their own idea of what the marriage rules are, which they may quite often put down in absolute terms. But the observer may find that the things which are claimed never occur in fact.
A third meaning is . . . the ideal type. An example of that is found in Professor Lévi-Strauss's paper, where he says, "For instance, the model of, let us say, a patrilineal kinship system does not in itself show whether or not the system has always remained patrilineal, or has been preceded by a matrilineal form, or has by any number of shifts been preceded from patrilineal to matrilineal and conversely." This seems to me to be an exact counterpart of Max Weber's ideal type, that is to say, the analysis of a situation in which you find certain implications which, for the specific purpose of demonstration, you put into ideal form, leaving out certain variations . . . . Max Weber has produced German bureaucracy as an ideal type; you can always argue, of course, that it is not a true one; it is only an idealization.
Now there is my final, fourth point, where Professor Lévi-Strauss contrasts model with reality, and here, he says that structure is a model and not reality. I am not going to talk about reality; I am going to quote Morris Cohen, who says that discussions on reality belong in religion. So far as we are concerned, we have a phenomenal world from which we abstract to varying degree. Everything is reality or not reality, whichever way you look at it. It depends on the level of abstraction.
Social structure is as real or as unreal as anything else, but on a higher level of abstraction. A great many more variable phenomenal details are ignored or dropped out. in that case, I do not think it is very satisfactory to call the social structure a model. It is the society or culture, if you like, looked at from a particular point of view, ignoring a number of variables. Two cases come to my mind; the first is the concrete individuality of the persons we see, the Toms, Dicks and Harrys of any anthropological field. . . .
The second variable we drop out is the qualitative character of actions, Whenever we construct a relationship, say, love, submission, subordination, we ignore the concrete modes of behavior out of which we construct that position picture of somebody being submissive to another person or people standing in reciprocal or symmetrical or asymmetrical relationships. I fail to see the difference between relations and social structure which Professor Lévi-Strauss emphasized, in terms that social relations are the raw material and social structure is something that is not raw material. . . .
It occurs to me that part of the difficulty arises from the fact that the notion of model has been extended from physics and chemistry to the human field. In physics and chemistry, the models first constructed, the important models of the nineteenth century, especially of a statistical kind, were constructed to explain things which were not visible; for example, the behavior of gases or, in chemistry, the structure of carbon rings. Physicists or chemists tried by imaginative effort and by a great deal of speculation and fitting of experimental data to construct an image of something that could not be verified directly by immediate vision, and so the notion of the model acquired a quality of abstraction and artistic construction which is of a quite different order from the kind of model that an archeologist or a historian or even a linguist sets up in order to describe things which are directly observable. These be has to justify by certain methods of testing which are different from the tests used in the chemical and physical field, where the mathematical formulation is important.
The problem, therefore, may be put in this way: In the social fields, there are processes which are hidden from us, which cannot be described adequately by simply putting down what you see before you. It is therefore necessary to construct a model of such a kind that we do not test the model by saving it has a one-to-one correspondence to what we see but, rather, that it permits us to deduce certain things which can then be verified.
What is deduced from the model to verify? In some cases the model constructed by chemists or biologists-for example, the model of chromosomes has been confirmed by powerful microscopic methods, and such a confirmation is a wonderful triumph and an encouragement to go on with the construction of such models. But unless we keep clearly in mind that there are at least two types of model function, depending upon the kind of objects we wish to describe, whether they are overtly given objects or are hidden objects and processes, then we will, I think, constantly oppose one another because of the strangeness of the model described by Dr. Lévi-Strauss in one case and the model described by a person who wishes to deal with immediately given social situations.
I think this discussion has established that, from the prehistoric point of view, we are dealing with a model which is based on inference. In archeology we can* never get anything but a small fraction of a culture, so the model we use and produce is an inferential model, outlined so clearly by Professor Schapiro.
I think Dr. Nadel pointed out beautifully the different senses in which the word "model" is being used, and I think for each of them a rather simple English expression could be substituted. For his first meaning, I would simply use the word "analog"; for his second meaning, I would use the word "ideal.". . . The third I would call pattern formulation"
That was Max Weber's ideal type.
I would call that "pattern formulation''; and then the fourth I would call a "sketch" or an "outline." I submit these as simple English expressions in place of the word 'I model" used in four or five different senses.
An important one, I think, has been left out by you and was mentioned by Professor Nadel, namely, all-or-some statements which are not quantitative, and those involving definite quantification, like "Sixty-two percent of the marriages are cross-cousin marriages."
Can we use the word "archetype" for model in archeology, for instance, as an abstraction? Unfortunately, the object itself seems to have a tendency to be different from the ones the model has established. There are many things in the object that are different from the model. The model is only an abstraction. It is a kind of archetype.
I think the word "statistical" has led to two confusions. In Dr. Nadel's remarks, he mentions statistical in the sense of the accumulation of data for the construction of a model. . . . That is a means of arriving at the model. In Dr. Lévi-Strauss's paper, I think he makes a confusion of the word "statistical" when he says that I try to construct mechanical models with the help of a statistical model. Well, I used statistics as a method of arriving at a mechanical model. It is not a statistical model because I used statistics. . . .
Isn't he thinking that your series of types is made up of a composite type rather than of individual types?
Well, that is a question of f act.
This interesting discussion about my paper is a fine experiment in cultural linguistics. . . . In three respects, at least, there are some agreements and confusion which I think arose exclusively from linguistic problems.
I have already mentioned the process question. . . . So I shall lay that aside and pass on to another confusion in respect to Professor Radcliffe-Brown's letter, that is, about the word "genetic." There has been a complete misunderstanding between us on the word "genetic," for be takes it in a historical sense while I was using it in a purely logical sense. In the field of geometry, for instance, we make a distinction between two kinds of definition. If we want to define a circle, we may say that it is a pattern made up of points which are equally distant from another point which we call the center, and this is a very good definition. Nevertheless, it is not genetic because you cannot make a circle with the help of the definition. But if you define the circle as a pattern resulting from having a segment of a line revolve around one of its ends, this is a genetic definition because it tells you bow to make a circle. But it is entirely different from a historical definition of the way any given circle has to come into existence. Therefore, when I argue against Radcliffe-Brown, that his interpretations of the Australian kinship system are not genetic, I am reproaching him, not for failing to bring up the history of the Australian kinship system, but for not explaining bow they are made.
This is also the reason for the diff erent approach between Professor Murdock and myself, when I mentioned that he built up a mechanical model out of statistical data, because his approach is extremely different from a truly geometrical approach. It would mean considerable, for instance, to say that the theorem of Thales or Archimedes is true because it is verified in 60 per cent of the cases. I think, therefore, geometry is using mechanical models, which are entirely different from statistical ones.
The third question is the relation between model and reality, and this is also mostly a linguistic problem, because in English it is difficult to distinguish between reality and concrete reality. I do not know bow you could quite qualify it. In my mind, models are reality, and I would even say that they are the only reality. They are certainly not abstractions, as was suggested by Professor Nadel, but they do not correspond to the concrete reality of empirical observation. It is necessary, in order to reach the model which is the true reality, to transcend this concrete-appearing reality. . . . Of course, a model can be very close to concrete reality, or it can be very far from it. In his letter Professor Radcliffe-Brown takes a very nice sample, because a sea shell is an empirical reality which is very close to its model. Unfortunately, in the field of social science, we very rarely meet with this kind of concrete reality, which shows the model in a very apparent way. However, it seems to me that some approximation can be found in the field of linguistics. There are, in a vocabulary, certain categories of terms which are very close to the model; let us say kinship vocabulary or the terms for parts of the body or the terms for the color scheme. Here the model is, in some languages at least, quite apparent, but the fact that it is not apparent for all parts of the vocabulary and that it is necessary to go to a deeper level to reach it does not prove that it does not exist, and it does not prove that the model is an abstraction. It proves that the reality is more hidden in some cases than in other cases.
Now I was asked to explain what I call a "model." For me, a model is exactly what Professor Schapiro stated. The model is not the mathematical formula, and the model is not the result of direct observations. Perhaps the best thing to do would be for me to give a few examples. Even before succeeding in seeing chromosomes, the geneticists were already making maps of chromosomes and genes, and this was a model; and, although no physicist ever saw an atom, nevertheless he was able to build an image which did explain all the properties of the atom and could be verified.
Another example: It has been discovered quite recently that crystals have a spiral-like growth, and this can be seen with the electronic microscope, although no actual photograph of any given crystal shows a perfect spiral. All the spirals are incomplete, deformed, but, nevertheless, the spiral itself is the model which may help to explain all its properties.
The distinction between the living model and the engineering model is a practical distinction which can be of great use, but, nevertheless, I do not think it goes extremely far. For instance, I agree with ]Professor Nadel that a small railway is a model. If I see that for the study at hand I do not need cars which are painted or built exactly like actual cars, but I can just use plain wooden squares, then the model will be a simplified one. It will be quite satisfactory if it can explain all the facts I am trying to explain, and it will to some extent be an engineering model, a machine built to specifications.
When I was assigned this paper, I discovered much to my surprise that I bad no idea whatsoever of what social structure was, and that I bad written quite a deal on social structure without knowing what it was. . . .
I will try to show that it is at the same time more and less than is usually thought; less than usually thought because I insist upon the distinction between social structure and social relations. Social relations are what are truly observed. . . . But, on the other band, social structure implies a problem of a very different nature, and I felt obliged to go very far toward demography, because it is impossible to study structure without studying numerical properties of groups and, on the other hand, religion.
I have noticed that religion was not a problem listed in the field of this symposium, and I am surprised that religion was not introduced, except in a negative way by Professor Nadel. But I do not think it is possible to understand social structure without taking into account the fact that there are structures which, instead of being related to another structure, opening new correlations, are really related to all the structures taken together and help to close the social structure. If we had not had orders of religion, the social order could extend indefinitely, and there would be new correlations arising, one after the other. It is only because there is some religious structure in human society that it is possible to close up the social structure.
We must not confuse reality with substance, and I take my favorite example from the philosophers. You have a brick wall. O.K. You take the bricks out, one by one. Materially, you have destroyed nothing, but a form is gone. You can take it further, of course. You can take each brick and pulverize it, and you have still got all of the matter, all of the reality in the sense of substance, that you bad at the start, but only a damn fool would say you had not destroyed something.
I would like to link this to the question that Professor Schapiro asked; whether, in problems of culture, we have something analogous to the problems which physical scientists faced in the last century and still are facing, of creating inferential constructs which will help us to understand-understand in the sense of predict-what will happen, but which are drawn only indirectly from what is seen. Of course, to this question, I would answer, unequivocally, "Yes."
Henry Murray always says there are three orders of phenomena about people; one can say that there are some things an individual knows about himself and is prepared to tell you. Then there is a category of things that he knows about himself but which he is unwilling to tell you. Third, there is a series of propositions which are true, but which the individual cannot tell you, not because he is unwilling, but because be does not know them. And I think the same thing is true about cultures.
All things that we talk about, obviously, are real; a statement about something is just as real as the thing is. A model of a shell is just as real as the shell. It is a metaphysical question which we think we do not have to go into. But there is an important distinction. . . . Somehow it is different; a shell is not the same thing as the model of a shell. What is it? I think we must make a distinction between things and discourse about things; and to discourse about things, we have to have symbols, so we must have this relation of reference of a symbol to the thing. I think that is probably all that is involved. It brings up, I think, an important concept of symbolism and symbols, which we have not mentioned at all.
Not just things and discourse about things, please. Things, discourse about things, and forms, i.e., arrangements of things.
"Discourse" is a better name for it.
Professor Kluckbobn has introduced an important concept . . . that is, the word "inference." I limited myself to discussing models and structures of a descriptive or illustrative kind, since that was what Professor Lévi-Strauss was doing.
Now it has been pointed out that models can also be used to reveal hidden mechanisms which are explanatory of an observed regularity or constellation of phenomena. Here we are dealing with inferences from an observed effect in order to explain that effect. I think these are two entirely different approaches. Admittedly, as Professor Lévi-Strauss explained, we do not know yet what social structure is, but one thing I think we can say negatively: It is not an explanatory top category. It has not to do with forces or hidden mechanisms, which we infer and for which we then construct models. Structure is still a descriptive or illustrative or diagnostic model. I do not think it includes anything in the way of forces which have to be interpolated or inferred to account for existing constellations of facts.
In addition to the point that Professor Schapiro made, I think we have to consider the use of models as a method of communication between the sciences. One of the most important things in Professor Lévi-Strauss's paper is the discussion of which 'forms of mathematics are suitable for the number of likely observations in the sort of phenomena that he was discussing where we have little runs or short runs.
Professor Wiener, in his first book on cybernetics, claimed we could not handle social science mathematically because of the length of our runs. One of the advantages, therefore, of borrowing from the models, in the sense that Professor Schapiro was describing them from physics, back and forth between the sciences, is that it permits us to see whether we could use the mathematics that another science has developed to work with a particular kind of model. . . .
When I was trying to use mathematical methods for the study of kinship rules, I bad great difficulties with mathematicians, who all told me it was impossible, because there was no mathematical way to describe marriage, in the same way as Professor Schapiro the other day was saying that there is no mathematical way to show what form in art is.
Then a mathematician came and said this was irrelevant, because be did not care about marriage, he was interested only in the relationship between the forms of marriage. From his point of view, there were only the relationships. . . This is very important and has been brought up by the new qualitative approach of mathematics in topology or group theory, which I think is applicable to problems in social structure.
I want to add one thing for consideration during the recess. When you have different degrees of abstraction what does that do to the problem of deriving process? Does it make it more difficult or easier, or does it make it impossible?