Review of Methods in Sociology: A Critical Study by Charles A. Ellwood
Gardner Murphy
CHARLES A. ELLWOOD. Methods in Sociology: A Critical Study. Durham, N. Car.: Duke University Press, 1933. Pp. 214. $1.50.
Though there are many ways of classifying sociologists, one classification which specially interests the outsider is that which brings out their differences in the basic definition of their science. Sociologists have tended in recent years to divide sharply on the question of the importance of quantitative methods. A large and eloquent group makes the observation and manipulation of quantities the hall-mark of all scientific enterprise; in particular, they evidently mean that scientific sociology must treat persons as objects in time-space and that their behavior must be conceived after the manner of mechanics. Their antagonists have contended that social experience, social motives, social values must be given a central place in sociology if the external and observable time-space aspects of social con-tact are to be understood; counting such social contacts as pushes and pulls will get nowhere unless we know whether the pushes and pulls are hostile, playful, accidental, etc. The numbers and formulae derived from such data will have no meaning except in relation to social reality as we know, feel, and will it.
The distinction thus drawn lacks subtlety. There might be quantitatively minded sociologists who were chiefly interested in counting or measuring certain aspects of social experience, and devising quantitative laws of such experience; and there might be other sociologists devoted to an orthodox Watsonian behaviorism but lacking all interest in measurement of social interaction. In point of fact, however, behavioristic and quantitative methods are usually found together, and it is only among those who insist on the importance of social experience that the quantitative approaches are slighted or given a subordinate rôle.
Psychologists cannot help being involved in this dispute, since some of the earlier formulations of behaviorism accelerated the rise of statistical sociology, promising that a way would shortly be found to state social experience in terms of explicit or implicit muscular response--or even promising to systematize the domain of objective behavior study in such fashion that the problem of social experience could be dispensed with. The promise was assumed to be capable of such prompt fulfillment as to justify the belief that it had already been fulfilled. Psychologists are unfortunately responsible also in some measure for the prevalence of the view that quantitative and qualitative methods are mutually opposed; the belief that because experience shows qualitative variation, it must on that account be refractory to all mathematical approaches.
( 262)
Ellwood's book seeks to make more explicit the claims of the "qualitative" group, finding a useful though subordinate rôle which may be played by the statisticians. The two most general sociological methods are, he believes, the historical and the psychological, the historical being essential in under-standing the context or setting in which any social phenomenon appears, and the psychological giving the meaning which it has for those who participate in it. From this point of view, the cardinal virtue of the sociologist is thorough scholarship—scholarship in the sense of knowledge of all that has been observed and written in relation to one's chosen problem, and the capacity to reorganize and apply observational material in the formulation of hypotheses. For larger social trends a knowledge of history may be the chief requirement; for shorter time spans, as in the case study of individuals, emphasis will be on psychology. It is unlikely that anyone ignorant of his history or his psychology will be able to devise vital or significant hypotheses. On the other hand, the freshness and the acumen of the sociologist are by no means guaranteed by mere familiarity with the necessary observational material.
Quantitative methods, Ellwood evidently believes, are not applicable to history; they add something to the case study; but they are chiefly significant in special fields where problems of analyzing individual experience can be neglected; for example, studies of population trends and the influence of geography and climate upon material culture. Even where quantitative methods are valid, the laws to which they give rise are mere generalizations regarding special social trends and may at any time cease to have meaning as society changes. The curve for population growth or for the development of a new kind of material culture is, for example, useful in describing conditions in a given era and place, but it is common knowledge that these curves permit but little prediction, simply because new interests, tastes, inventions, and ideas change the basic trends and twist the curve into a new shape.
All this is stated lucidly. The necessity of a thorough analysis of qualitative problems is well described, and the demonstration that measurements are valuable only when the datum to be measured has been clearly defined is cogently made. Very little of the argument is new (the amount of quotation and citation of authorities is rather burdensome), yet to have this brief and clear statement of a viewpoint is most useful.
The book betrays, however, little awareness of the reasons why this type of argument seems inadequate to psychologists. Scanning their own science, psychologists know how easy it has always been to formulate vivid and appealing hypotheses, to "think out" the experiences and the values inherent in a situation, and how tremendously hard it is to give these hypotheses a testing so unambiguous as to convince their colleagues.
( 263) Ellwood tells us repeatedly that the validation of hypotheses lies in further observation, but he lays down no canons to tell us how these observations can be forced to verify or refute. In historical criticism, or even in biography, methods of work are now so definite that a single scholar may at times offer an hypothesis which all must sooner or later accept, but it is an open secret that in social psychology and sociology the master key of critical method has not yet been forged. We simply have not progressed far enough to know how to use new observations in the verification of hypotheses which have once been formulated. A glance over the last thirty years will show how easy it is to construct systems of psychology, and how impossible to construct one which will stand solidly against the testing which new experiments permit. And systems of sociology which have been thought out since the time of Comte have convinced their own originators year by year and decade by decade, have been abandoned, revived, and abandoned again, depending upon moods, economic conditions, fads in other sciences, and so on. Sociologists once worked out, in the comfort of their private studies, theories of human nature based on the seven primary instincts of the year 1908; they have now worked out, with equal coherence, systems based on the reflexes of the newborn and Watson's experiments on Albert in 1920. The output of sociological theories of this sort shows no tendency to decline. By failing to explain how sociological hypotheses are to be tested, Ellwood fails to touch the point which would chiefly interest the psychological reader.
Moreover, if it be true that the realities with which we deal are immediate social experiences, and if these experiences are so different from one man to another that the same social fact has a different meaning for the two, how can one man's hypothesis ever be tested by the second? The arguments offered by Ellwood and the group with which he wishes to affiliate himself show, if valid, that the testing of hypotheses must take on as many forms and give as many different results as there are attitudes of observing or attitudes in the testing of observations. If the inner world of private experience and motive is the true heart of the sociologist's problem, how is the sociologist's own personal world of motives and values to be exempted? How, with all the obvious differences in the emotional needs and wishes of social scientists, is anything like a coherent, acceptable, scientific body of data to be gathered and systematized? One may respect Hobbes, or Spencer, or Sumner, as an artist with many flashes of insight for which we may well be grateful, but the notion of science involves a corporate, social enterprise in which different workers can agree—at least as to major laws—and it must be evident that the thinking-out of sociological systems has as yet given no promise that a science is to be expected from that quarter.
( 264)
Difficulties, however, arise, not only in finding a body of significant systematized data upon which sociologists can agree, but even in basing a single sociological system upon irrefutable facts. We have agreed with Ellwood for the sake of argument that questions of social behavior are basically questions of motive or value. But how does one ever know that one has plumbed the depths of the soul and set the motives in order as they really are? Psychologists admit, for the most part, that they know very little about the bases of social behavior even of themselves and their friends, those with whom they are in hourly contact; yet many sociological writers wish to define the motives of those who live thousands of miles away or who have been dead for a hundred years. They can tell us in detail of the psychology of the Italian Renaissance, the French Revolution, and the Chartist Movement, though they cannot grasp the psychology of the grocer's boy, or even that of their statistically minded opponents. Ellwood's argument is that sociology is based upon psychology, that is to say, a scientific psychology of motives; but there is no scientific psychology of motives. There are only fragments of data which the future will certainly regard as trivial. Let us by all means use all the psychology we have, even if much of it is the rough-and-ready psychology of uncontrolled observation, but let us not confuse this enterprise with the construction of the fundamentals of a science.
If Ellwood means that sociologists must do the best they can with the best results of a guess-work psychology of motives, no one can quarrel with him, for, indeed, there are urgent problems upon which it is better to guess than to remain silent. There is not only no harm in the venture of guessing about society, provided one recognizes that it is a guess; it may even suggest problems which can be tested by a reliable technique. It is very human and natural that psychologists and sociologists (and economists and anthropologists) who are working with these uncertain data should want the honorific title of "scientists," since in the twentieth century what is exact is more honorable (except among ultra-modern physicists). Yet it is equally natural that they should at the same time want the right to develop the groundwork of their chosen disciplines without embarrassment from measurements until they have some notion of what may profitably be measured.
In certain instances, e.g., the working-out of the idea of the "culture area" in anthropology, it has been necessary to define a concept carefully and to apply it in many situations before attempting to apply any quantitative tests. The concept of culture areas was put forward almost simultaneously with several other anthropological concepts which have now been abandoned. When first offered, this doctrine of culture areas was the crudest sort of idea. Only through further observation, largely of a rough
( 265) quantitative sort, has the validity of the idea been established. The quantitative testing in such cases is rarely a matter of "mathematics" (as in Clements' work) ; it consists mostly in noting resemblances and differences in the cultural elements of different ethnic groups. There are, indeed, some anthropologists who seem unaware that the grouping of such similarities and differences is the first step in all quantitative procedure.
But the mathematical method usually proceeds, of course, very much further than this simple grouping and counting of observations. There have been in recent years several brilliant analyses of quantitative data following upon thorough study of the qualitative sociological facts which are to be set in order. I refer, for example, to Rice's detailed statistical studies of the applicability of this notion of culture areas in his investigations of political attitudes (Quantitative Methods in Politics, 1928). There has been some discussion and some criticism of these studies by Rice, but, for the most part, the qualitatively minded group does not seem to have noted just what Rice was doing and how his work differed from the typical manipulation of numbers in population studies or the study of price indices in relation to the business cycle. A concept worked out in the course of anthropological field studies, was refined, thought through, and rendered applicable to American political life, and then, instead of being committed to the pages of a sociological treatise, was given a gruel-ling statistical testing involving some thousands of correlation studies. A time will come when insistence on the minimizing of quantitative methods will seem reasonable only when one has acquainted oneself in detail with the nature of such enterprises and sets out to discuss in detail what has been done.
It is granted, of course, that sociology may in certain cases be able to verify hypotheses by methods other than those recognized in other sciences. It is, for example, quite possible that in the study of leadership certain qualities will be found significantly related to one another and to certain social consequences though no true form of measurement of such qualities has been suggested. Even here, however, one suspects that trouble arises in part from ignorance of the present state of statistical method. To make any generalization whatever about leadership involves quantitative method. To say that certain traits belong together usually or always, or that they are followed by certain social consequences usually or always, is to involve oneself in a mathematical generalization which can avoid the appearance of being mathematical only by remaining vague. To say that one quality goes with another quality is to make a statement which is useful just in so far as we are able to state how widely the rule holds and the conditions which make these two phenomena hang together; the first step in this process is to make observations under various conditions, testing the
(266) variations in the results. I am not, of course, referring to product-moment correlations. There are many cases in which the interdependence of variables may be stated quantitatively though no variable can be independently measured. I refer to such commonplaces as the method of mean square contingency, especially devised to treat of the kind of problem which the sociologists have here in mind. Ellwood does not discuss such methods, or any of the many existing methods by which it is possible for the sociologist to treat quantitatively the relations which exist between qualities. If, of course, we had to measure every aspect of a phenomenon before the relation of this phenomenon to other phenomena could be stated quantitatively, we should have to postpone systematic quantitative sociology for decades, perhaps centuries. In none of the quantitative sciences which now exist, however, has it ever been found necessary to insist on such a criterion. In fact, it has been the rule rather than the exception that the quantitative interrelations between phenomena should be stated before the phenomena themselves are so stated. Psychology has witnessed for a hundred years a steady succession of such developments. Measurements of interrelations between phenomena have often led to a direct measurement of phenomena which had first defied all mathematical treatment. To give but two illustrations from contemporary work, the quantitative delineation of Gestalt psychology by Köhler and the mathematical conception of intelligence which has developed since Spearman's two-factor theory point to the constant improvement in definition of our problems as we work from the mathematics of interacting processes to the mathematics of the individual psychic act.
In summary, it seems to the reviewer that Ellwood's book states in cogent form the necessity of careful observation and a thinking-out of qualitative hypotheses regarding human nature and society before one attempts to collect numerical data. At the same time Ellwood seems to have no suggestion as to how laws of social interaction are to be verified without the use of quantitative methods. It is to be hoped that his next book will present the positive program of sociology as he sees it, with special reference to the problem of verifying sociological hypotheses.
GARDNER MURPHY
Columbia University
New York City
