Sociological Analysis and the "Variable"[1]
Herbert Blumer
University of California, Berkeley
MY aim in this paper is to examine critically the scheme of sociological analysis which seeks to reduce human group life to variables and their relations. I shall refer to this scheme, henceforth, as "variable analysis." This scheme is wide-spread and is growing in acceptance. It seems to be becoming the norm of proper sociological analysis. Its sophisticated forms are becoming the model of correct research procedure. Because of the influence which it is exercising in our discipline, I think that it is desirable to note the more serious of its ,shortcomings in actual use and to consider certain limits to its effective application. The first part of my paper will deal with the current shortcomings that I have in mind and-the second part with the more serious question of the limits to its adequacy.
SHORTCOMINGS IN CONTEMPORARY VARIABLE ANALYSIS
The first shortcoming I wish to note in current variable analysis in our field is the rather chaotic condition that prevails in the selection of variables. There seems to be little limit to what may be chosen or designated as a variable. One may select something as simple as a sex distribution or as complex as a depression; something as specific as a birth rate or as vague as social cohesion; some-thing as evident as residential change or as imputed as a collective unconscious; some-thing as generally recognized as hatred or as
* Presidential address read at the annual meeting of the American Sociological Society, September, 1956.doctrinaire as the Oedipus complex; some-thing as immediately given as a rate of news-paper circulation to something as elaborately fabricated as an index of anomie. Variables may be selected on the basis of a specious impression of what is important, on the basis of conventional usage, on the basis of what can be secured through a given instrument or technique, on the basis of the demands of some doctrine, or on the basis of an imaginative ingenuity in devising a new term.
Obviously the study of human group life calls for a wide range of variables. However, there is a conspicuous absence of rules, guides, limitations and prohibitions to govern the choice of variables. Relevant rules are not provided even in the thoughtful regulations that accompany sophisticated schemes of variable analysis. For example, the rule that variables should be quantitative does not help, because with ingenuity one can impart a quantitative dimension to almost any qualitative item. One can usually construct some kind of a measure or index of it or develop a rating scheme for judges. The proper insistence that a variable have a quantitative dimension does little to lessen the range or variety of items that may be set up as variables. In a comparable manner, the use of experimental design does not seemingly exercise much restriction on the number and kind of variables which may be brought within the framework of the design. Nor, finally, does careful work with variables, such as establishing tests of reliability, or inserting "test variables," exercise much restraint on what may be put into the pool of sociological variables.
( 684)
In short, there is a great deal of laxity in choosing variables in our field. This laxity is due chiefly to a neglect of the careful reduction of problems that should properly precede the application of the techniques of variable analysis. This prior task requires thorough and careful reflection on the problem to make reasonably sure that one has identified its genuine parts. It requires intensive and extensive familiarity with the empirical area to which the problem refers. It requires a careful and thoughtful assessment of the theoretical schemes that might apply to the problem. Current variable analysis in our field is inclined to slight these requirements both in practice and in the training of students for that practice. The scheme of variable analysis has become for too many just a handy tool to be put to immediate use.
A second shortcoming in variable analysis in our field is the disconcerting absence of generic variables, that is, variables that stand for abstract categories. Generic variables are essential, of course, to an empirical science—they become the key points of its analytical structure. Without generic variables, variable analysis yields only separate and disconnected findings.
There are three kinds of variables in our discipline which are generally regarded as generic variables. None of them, in my judgment, is generic. The first kind is the typical and frequent variable which stands for a class of objects that is tied down to a given historical and cultural situation. Convenient examples are: attitudes toward the Supreme Court, intention to vote Republican, interest in the United Nations, a college education, army draftees and factory unemployment. Each of these variables, even though a class term, has substance only in a given historical context. The variables do not stand directly for items of abstract human group life; their application to human groups around the world, to human groups in the past, and to conceivable human groups in the future is definitely restricted. While their use may yield propositions that hold in given cultural settings, they do not yield the abstract knowledge that is the core of an empirical science.
The second apparent kind of generic variable in current use in our discipline is represented by unquestionably abstract sociological categories, such as "social cohesion," "social integration," "assimilation," "authority," and "group morale." In actual use these do not turn out to be the generic variables that their labels would suggest. The difficulty is that such terms, as I sought to point out in an earlier article on sensitizing concepts,[2] have no fixed or uniform indicators. Instead, indicators are constructed to fit the particular problem on which one is working. Thus, certain features are chosen to represent the social integration of cities, but other features are used to represent the social integration of boys' gangs. The indicators chosen to represent morale in a small group of school children are very different from those used to stand for morale in a labor movement. The indicators used in studying attitudes of prejudice show a wide range of variation. It seems clear that indicators are tailored and used to meet the peculiar character of the local problem under study. In my judgment, the abstract categories used as variables in our work turn out with rare exception to be something other ,than generic categories. They are localized in terms of their content. Some measure of support is given to this assertion by the fact that the use of such abstract categories in variable research adds little to generic knowledge of them. The thousands of "variable" studies of attitudes, for instance, have not contributed to our knowledge of the abstract nature of an attitude; in a similar way the studies of "social cohesion," "social integration," "authority," or "group morale" have done nothing, so far as I can detect, to clarify or augment generic knowledge of these categories.
The third form of apparent generic variable in our work is represented by a special set of class terms like "sex," "age," "birth rate," and "time .period." These would seem to be unquestionably generic. Each can be applied universally to human group life; each has the same clear and common meaning in its application. Yet, it appears that in their use in our field they do not function as generic variables. Each has a content that is given by its particular instance of application, e.g., the birth rate in Ceylon, or the sex distribution in the State of Nebraska, or
( 685) the age distribution in the City of St. Louis. The kind of variable relations that result from their use will be found to be localized and non-generic.
These observations on these three specious kinds of generic variables point, of course, to the fact that variables in sociological research are predominantly disparate and localized in nature. Rarely do they refer satisfactorily to a dimension or property of abstract human group life. With little exception they are bound temporally, spatially, and culturally and are inadequately cast to serve as clear instances of generic sociological categories. Many would contend that this is because variable research and analysis are in a beginning state in our discipline. They believe that with the benefit of wider coverage, replication, and the co-ordination of separate studies disparate variable relations may be welded into generic relations. So far there has been little achievement along these lines. Although we already have appreciable accumulations of findings from variable studies, little has been done to convert the findings into generic relations. Such conversion is not an easy task. The difficulty should serve both as a challenge to the effort and an occasion to reflect on the use and limitations of variable analyses.
As a background for noting a third major shortcoming I wish to dwell on the fact that current variable analysis in our field is operating predominantly with disparate and not generic variables and yielding predominantly disparate and not generic relations. With little exception its data and its findings are "here and now," wherever the "here" be located and whenever the "now" be timed. Its analyses, accordingly, are of localized and concrete matters. Yet, as I think logicians would agree, to understand adequately a "here and now" relation it is necessary to understand the "here and now" context. This latter understanding is not provided by variable analysis. The variable relation is a single relation, necessarily stripped bare of the complex of things that sustain it in a "here and now" context. Accordingly, our under-standing of it as a "here and now" matter suffers. Let me give one example. A variable relation states that reasonably staunch Erie County Republicans become confirmed in their attachment to their candidate as a result of listening to the campaign material of the rival party. This bare and interesting finding gives us no picture of them as human beings in their particular world. We do not know the run of their experiences which induced an organization of their sentiments: and views, nor do we know what this organization is; we do not know the social atmosphere or codes in their social circles; we do not know the reinforcements and rationalizations that come from their fellows; we do not know the defining process in their circles; we do not know the pressures, the incitants, and the models that came from their niches in the social structure; we do not know how their ethical sensitivities are organized and so what they would tolerate in the way of shocking behavior on the part of their candidate. In short, we do not have the picture to size up and understand what their confirmed attachment to a political candidate means in terms of their experience and their social context. This fuller picture of the "here and now" context is not given by variable relations. This, I believe, is a major shortcoming in variable analysis, insofar as variable analysis seeks to explain meaningfully the disparate and local situations with which it seems to be primarily concerned.
The three shortcomings which I have noted in current variable research in our field are serious but perhaps not crucial With increasing experience and maturity they will probably be successfully over come. They suggest, however, the advisability of inquiring more deeply into the interesting and important question of how well variable analysis is suited to the study of human group life in its fuller dimensions.
LIMITS OF VARIABLE ANALYSIS
In my judgment, the crucial limit to the successful application of variable analysis to human group life is set by the process of interpretation or definition that goes on in human groups. This process, which I believe to be the core of human action, give a character to human group life that seem to be at variance with the logical premise of variable analysis. I wish to explain a some length what I have in mind.
All sociologists—unless I presume too
( 686) much—recognize that human group activity is carried on, in the main, through a process of interpretation or definition. As human beings we act singly, collectively, and societally on the basis of the meanings which things have for us. Our world consists of innumerable objects—home, church, job, college education, a political election, a friend, an enemy nation, a tooth brush, or what not—each of which has a meaning on the basis of which we act toward it. In our activities we wend our way by recognizing an object to be such and such, by defining the situations with which we are presented, by attaching a meaning to this or that event, and where need be, by devising a new meaning to cover something new or different. This is done by the individual in his personal action, it is done by a group of individuals acting together in concert, it is done in each of the manifold activities which together constitute an institution in operation, and it is done in each of the diversified acts which fit into and make up the patterned activity of a social structure or a society. We can and, I think, must look upon human group life as chiefly a vast interpretative process in which people, singly and collectively, guide themselves by defining the objects, events, and situations which they encounter. Regularized activity inside this process results from the application of stabilized definitions. Thus, an institution carries on its complicated activity through an articulated complex of such stabilized meanings. In the face of new situations or new experiences individuals, groups, institutions and societies find it necessary to form new definitions. These new definitions may enter into the repertoire of stable meanings. This seems to be the characteristic way in which new activities, new relations, and new social structures are formed. The process of interpretation may be viewed as a vast digestive process through which the confrontations of experience are transformed into activity. While the process of interpretation does not embrace everything that leads to the formation of human group activity and structure, it is, I think, the chief means through which human group life goes on and takes shape.
Any scheme designed to analyze human `group life in its general character has to fit this process of interpretation. This is the test that I propose to apply to variable analysis. The variables which designate matters which either directly or indirectly confront people and thus enter into human group life would have to operate through this process of interpretation. The variables which designate the results or effects of the' happenings which play upon the experience of people would be the outcome of the process of interpretation. Present-day variable analysis in our field is dealing predominantly with such kinds of variables.
There can be no doubt that, when current variable analysis deals with matters or areas of human group life which involve the process of interpretation, it is markedly disposed to ignore the process. The conventional procedure is to identify something which is presumed to operate on group life and treat it as an independent variable, and then to select some form of group activity as the dependent variable. The independent variable is put at the beginning part of the process of interpretation and the dependent variable at the terminal part of the process. The intervening process is ignored or, what amounts to the same thing, taken for granted as something that need not be considered. Let me cite a few typical examples: the presentation of political programs on the radio and the resulting expression of intention to vote; the entrance of Negro residents into a white neighborhood and the resulting attitudes of the white inhabitants toward Negroes; the occurrence of a business depression and the resulting rate of divorce. In such instances—so common to variable analysis in our field—one's concern. is with the two variables and not with what lies between them. If one has neutralized other factors which are regarded as possibly exercising influence on the dependent variable, one is content with the conclusion that the observed change in the dependent variable is the necessary result of the independent variable.
This idea that in such areas of group life the independent variable automatically exercises its influence on the dependent variable is, it seems to me, a basic fallacy. There is a process of definition intervening between the events of experience presupposed by the independent variable and the formed behavior represented by the dependent variable.
( 687) The political programs on the radio are interpreted by the listeners; the Negro invasion into the white neighborhood must be defined by the whites to have any effect on their attitudes; the many events and happenings which together constitute the business depression must be interpreted at their many points by husbands and wives to have any influence on marital relations. This intervening interpretation is essential to the outcome. It gives the meaning to the presentation that sets the response. Because of the integral position of the defining process between the two variables, it becomes necessary, it seems to me, to incorporate the process in the account of the relationship. Little effort is made in variable analysis to do this. Usually the process is completely ignored. Where the process is recognized, its study is regarded as a problem that is independent of the relation between the variables.
The indifference of variable analysis to the process of interpretation is based apparently on the tacit assumption that the independent variable predetermines its interpretation. This assumption has no foundation. The interpretation is not predetermined by the variable as if the variable emanated its own meaning. If there is anything we do know, it is that an object, event or situation in human experience does not carry its own meaning; the meaning is conferred on it.
Now, it is true that in many instances the interpretation of the object, event or situation may be fixed, since the person or people may have an already constructed meaning which is immediately applied to the item. Where such stabilized interpretation occurs and recurs, variable analysis would have no need to consider the interpretation. One could merely say that as a matter of fact under given conditions the independent variable is followed by such and such a change in the dependent variable. The only necessary pre-caution would be not to assume that the stated relation between the variables was necessarily intrinsic and universal. Since any-thing that is defined may be redefined, the relation has no intrinsic fixity.
Alongside the instances where interpretation is made by merely applying stabilized meanings there are the many instances where the interpretation has to be constructed. These instances are obviously increasing in our changing society. It is imperative in the case of such instances for variable analysis to include the act of interpretation in its analytic scheme. As far as I can see, variable analysis shuns such inclusion.
Now the question arises, how can variable analysis include the process of interpretation? Presumably the answer would be to treat the act of interpretation as an "intervening variable." But, what does this mean? If it means that interpretation is merely an intervening neutral medium through which the independent variable exercises its influence, then, of course, this would be no answer. Interpretation is a formative or creative process in its own right. It constructs meanings which, as I have said, are not predetermined or determined by the independent variable.
If one accepts this fact and proposes to treat the act of interpretation as a formative process, then the question arises how one is to characterize it as a variable. What quality is one to assign to it, what property or set of properties? One cannot, with any sense, characterize this act of interpretation in terms of the interpretation which it constructs; one cannot take the product to stand for the process. Nor can one characterize the act of interpretation in terms of what enters into it—the objects perceived, the evaluations and assessments made of them, the cues that are suggested, the possible definitions proposed by oneself or by others. These vary from one instance of interpretation to another and, further, shift from point to point in the development of the act. This varying and shifting content offers no basis for making the act of interpretation into a variable.
Nor, it seems to me, is the problem met by proposing to reduce the act of interpretation into component parts and work with these parts as variables. These parts would presumably have to be processual parts—such as perception, cognition, analysis, evaluation, and decision-making in the individual; and discussion, definition of one another's responses and other forms of social inter-action in the group. The same difficulty exists in making any of the processual parts into variables that exists in the case of the complete act of interpretation.
( 688)
The question of how the act of interpretation can be given the qualitative constancy that is logically required in a variable has so far not been answered. While one can devise some kind of a "more or less" dimension for it, the need is to catch it as a variable, or set of variables, in a manner which reflects its functioning in transforming experience into activity. This is the problem, indeed dilemma, which confronts variable analysis in our field. I see no answer to it inside the logical framework of variable analysis. The process of interpretation is not inconsequential or pedantic. It operates too centrally in group and individual experience to be put aside as being of incidental interest.
In addition to the by-passing of the process of interpretation there is, in my judgment, another profound deficiency in variable analysis as a scheme for analyzing human group life. The deficiency stems from the inevitable tendency to work with truncated factors and, as a result, to conceal or misrepresent the actual operations in human group life. The deficiency stems from the logical need of variable analysis to work with discrete, clean-cut and unitary variables. Let me spell this out.
As a working procedure variable analysis seeks necessarily to achieve a clean identification of the relation between two variables. Irrespective of how one may subsequently combine a number of such identified relations—in an additive manner, a clustering, a chain-like arrangement, or a "feedback" scheme—the objective of variable research is initially to isolate a simple and fixed relation between two variables. For this to be done each of the two variables must be set up as a distinct item with a unitary qualitative make-up. This is accomplished first by giving each variable, where needed, a simple quality or dimension, and second by separating the variable from its connection with other variables through their exclusion or neutralization.
A difficulty with this scheme is that the empirical reference of a true sociological variable is not unitary or distinct. When caught in its actual social character, it turns out to be an intricate and inner-moving complex. To illustrate, let me take what seems ostensibly to be a fairly clean-cut variable relation,
lamely between a birth control program and the birth rate of a given people. Each of these two variables—the program of birth control and the birth rate—can be given a simple discrete and unitary character. For the program of birth control one may choose merely its time period, or select some reason-able measure such as the number of people visiting birth control clinics. For the birth rate, one merely takes it as it is. Apparently, these indications are sufficient to enable the investigator to ascertain the relations between the two variables.
Yet, a scrutiny of what the two variables stand for in the life of the group gives us a different picture. Thus, viewing the program of birth control in terms of how it enters into the lives of the people, we need to note many things such as the literacy of the people, the clarity of the printed information, the manner and extent of its distribution, the social position of the directors of the program and of the personnel, how the personnel act, the character of their instructional talks, the way in which people define attendance at birth control clinics, the ex-pressed views of influential personages with reference to the program, how such person-ages are regarded, and the nature of the discussions among people with regard to the clinics. These are only a few of the matters which relate to how the birth control program .might enter into the experience of the people.'The number is sufficient, however, to show the complex and inner-moving character of what otherwise might seem to be a simple variable.
A similar picture is given in the case of he other variable—the birth rate. A birth rate of a people seems to be a very simple and unitary matter. Yet, in terms of what it expresses and stands for in group activity it is exceedingly complex and diversified. We need consider only the variety of social factors that impinge on and affect the sex act, even though the sex act is only one of the activities that set the birth rate. The self -conceptions held by men and by women, conceptions of family life, the values placed on children, accessibility of men and women to each other, physical arrangements in the home, the sanctions given by established institutions, the code of manliness, the
( 689) pressures from relatives and neighbors, and ideas of what is proper, convenient and tolerable in the sex act—these are a few of the operating factors in the experience of the group that play upon the sex act. They suffice to indicate something of the complex body of actual experience and practice that is represented in and expressed by the birth rate of a human group.
I think it will be found that, when converted into the actual group activity for which it stands, a sociological variable turns out to be an intricate and inner-moving complex. There are, of course, wide ranges of difference between sociological variables in terms of the extent of such complexity. Still, I believe one will generally find that the discrete and unitary character which the labeling of the variable suggests vanishes.
The failure to recognize this is a source of trouble. In variable analysis one is likely to accept the two variables as the simple and unitary items that they seem to be, and to believe that the relation found between them is a realistic analysis of the given area of group life. Actually, in group life the relation is far more likely to be between complex, diversified and moving bodies of activity. The operation of one of these complexes on the other, or the interaction between them, is both concealed and misrepresented by the statement of the relation between the two variables. The statement of the variable relation merely asserts a connection between abbreviated terms of reference. It leaves out the actual complexes of activity and the actual processes of interaction in which human group life has its being. We are here faced, it seems to me, by the fact that the very features which give variable analysis its high merit—the qualitative constancy of the variables, their clean-cut simplicity, their ease of manipulation as a sort of free counter, their ability to be brought into decisive relation—are the features that lead variable analysis to gloss over the character of the real operating factors in group life, and the real interaction and relations between such factors.
The two major difficulties faced by variable analysis point clearly to the need for a markedly different scheme of sociological analysis for the areas in which these difficulties arise. This is not the occasion to spell out the nature of this scheme. I shall merely mention a few of its rudiments to suggest how its character differs fundamentally from that of variable analysis. The scheme would be based on the premise that the chief means through which human group life operates and is formed is a vast, diversified process of definition. The scheme respects the empirical existence of this process. It devotes itself to the analysis of the operation and formation of human group life as these occur through this process. In doing so it seeks to trace the lines of defining experience through which ways of living, patterns of relations, and social forms are developed, rather than to relate these formations to a set of selected items. It views items of social life as articulated inside moving structures and believes that they have to be understood in terms of this articulation. Thus, it handles these items not as discrete things disengaged from their connections but, instead, as signs of a supporting context which gives them their social character. In its effort to ferret out lines of definition and networks of moving relation, it relies on a distinctive form of procedure. This procedure is to approach the study of group activity through the eyes and experience of the people who have developed the activity. Hence, it necessarily requires an intimate familiarity with this experience and with the scenes of its operation. It uses broad and interlacing observations and not narrow and disjunctive observations. And, may I add, that like variable analysis, it yields empirical findings and "here-and-now" propositions, although in a different form: Finally, it is no worse off than variable analysis in developing generic knowledge out of its findings and propositions.
In closing, I express a hope that my critical remarks about variable analysis are not misinterpreted to mean that variable analysis is useless or makes no contribution to sociological analysis. The contrary is true. Variable analysis is a fit procedure for those areas of social life and formation that are not mediated by an interpretative process. Such areas exist and are important. Further, in the area of interpretative life variable analysis can be an effective means of unearthing stabilized patterns of interpretation which are not likely to be detected through the
( 690) direct study of the experience of people. Knowledge of such patterns, or rather of the relations between variables which reflect such patterns, is of great value for understanding group life in its "here-and-now" character and indeed may have significant practical value. All of these appropriate uses give variable analysis a worthy status in our field.
In view, however, of the current tendency of variable analysis to become the norm and model for sociological analysis, I believe it important to recognize its shortcomings and its limitations.
