A Logic of Social Science Thought
[Read before the Wellington Branch, Royal Society of New Zealand, September 10, 1946; received by Editor, December 5, 1948; issued separately February, 1949.]
Whenever we turn to questions of sociological method, we find scarcely a book, treatise, or paper is being published whose author does not emphasize the youth of the social sciences, compared with the other sciences, preferably the natural sciences in which problems and solutions, methods, observational and operational devices and techniques appear to be so much more mature and settled. One becomes somewhat tired of the tune: the youth of the social sciences being made the excuse for the implication of its shortcomings. If we consider the grounds on which first principles in science are put forward, I doubt whether the social sciences are any worse off than, for instance, physics. The revolutionary development, however, of the first principles of physics of recent years did not in the least, prevent physics being operated by the ordinary physicist as a practical science. Moreover, there is reason to distrust this plea of youth-fulness if the author, as his theme unfolds, has opportunity to refer statements made and theorems propounded to students of social problems who wrote over a hundred years ago—to wit, Comte—to over two thousand years ago—to wit, Aristotle. I am rather inclined to be less optimistic and, instead of pleading the youthfulness of the social sciences, frankly admit their backwardness.
I do not intend to examine in detail whether the social sciences are—always in comparison with the natural sciences—young or backward, or both. But I propose to go at some length into the question why social science has not in the past, and does not at present, permit of applicability to practical problems of social life with the same ease and assuredness which strikes the layman about physics, chemistry, and biology, for example, in the production of aeroplanes and atomic bombs, and in artificial insemination. What I have to say will be nothing new—I have made no invention, and no discovery to report; I shall, metaphorically speaking, try to put two and two together. But my problem is slightly more involved than an elementary arithmetical operation.
A last remark by way of preface: As my task will be a critical survey—logic, in the strict sense of the word, being, as a French historian put it, le bon usage de l'esprit(1)—and as I wish to have my thesis critically discussed, I was faced with the necessity of choosing references which alone will enable you to check up on the argument. I had at first selected from among my notes which have accumulated over a number of years the thoughts of three or four men(2)—an American philosopher, a British historian, a Swedish economist, and if time would permit, a German sociologist. None of the works of the first three was published before 1936, so it would
be safe to claim that our discussion concerns itself with present, not with past problems of social science thought. Even after this selection, however, I found the topic unmanageable, at least for the treatment which can be bestowed on it in one short evening. I had to abandon a comparative study of three or four authors and to limit myself to the survey of one special theory of social inquiry. Hence my title: a logic of social-science thought(3).
Let me commence by examining the following statement:—
A genuine problem is one set by existential problematic situations. In social inquiry, genuine problems are set only by actual social situations which are themselves conflicting and confused. Social conflicts and confusions exist in fact before problems for inquiry exist. The latter are intellectualisations in inquiry of them “practical” troubles and difficulties.
This statement contains one term—existential problematic situation—the meaning of which may not be self-explanatory. At least, if it is not used in a purely adjectival sense of something that exists. For reasons of economy, however, I must here by-pass the problem of philosophy that is hidden behind the term existential(4).
I select three implications of the statement which are directly to the point:
(1) For inquiry into a social problem, our author postulates as conjoint conditions the observational attainment of facts and appropriate operational conceptions, and thus links empirical research with social theory. The conjointness of conditions, without giving pride of place to either empirical research or theory, emphasizes that social theory need not be dead speculation but is a legitimate intellectual pursuit to the extent to which theoretical conceptions are nessary tools in the process of scientifically resolving social problems.
While nobody will quarrel with a consequent stricture upon a vast amount of what usually goes under “social theory” but is actually the expression of a good deal of wishful thinking garbed in scientific language, it should all the same be kept in mind that the tools of theoretical thought are not being shaped ad hoc to satisfy the requirements of specific problems. On the contrary, its very independence from specific application enables theory to fulfil its part in the bargain of conjoint observational attainment of facts and appropriate operational conceptions. The real difficulty only begins here: it has been asserted, for instance, that the rationalistic or conceptual school of sociology, although paying attention to facts, places its entire emphasis upon concepts, in such a manner that facts are directly subsumed under “principles,” and principles are regarded as fixed norms that decide the legitimacy or illegitimacy of existing phenomena(5). And further, that this school of thought, and the approach to social problems for which it stands, has dominated the past history of social science. This may be true, broadly speaking, of Auguste Comte's Cours de Philosophie Positive and even more of his Système de Politique Positive; but if we think of Durkheim or of Sumner or of Hobhouse—to mention only three outstanding names in the history of sociology—I doubt if we can so lightly condemn its past history.
(2) A second point will further stress the difficulties with which the suggested approach to social problems is beset. It is not only a practical difficulty—which could be overcome by appropriate treatment and the use of appropriate methods—but it is a difficulty of theoretical import that, in comparison with physical phenomena, we observe a more intricate constitution and a much wider field of social phenomena. This, our author points out, is a product of the “interacting intervention of the physical conditions with the human factor.” Perhaps you dislike this way of putting it as much as I do; but the fact remains that besides what has been called the “cultural matrix” of every inquiry, and more particularly and to a higher degree of social inquiry, social phenomena cannot be understood without an understanding of physical conditions and the mutual conditioning of physical factors and the human factor. And here be it noted that physical conditions comprise anything from biological conditions to geographical and climatic environment. The necessity in the social sciences of referring to these physical conditions—whether you deal with a question of industrial management or conduct research into a culture—is something every social worker is, and every social scientist should be, aware of. Without aiming at a comprehensive statement of this issue, I should like to mention two special aspects of the matter.
The first I cannot express better than by quoting,
This consideration is fatal to the view that social sciences are exclusively, or even dominantly psychological.
I hardly need labour this point, but only say that if we adhere to this view, it may offer a reconciliation between two extremes: a psychological school which tends to treat social phenomena as a summation of individual human sensations and emotions, responses and reactions, on the one hand(6); and a view like that, for example, of the late R. G. Collingwood, on the other hand, who does not admit psychology as a social science at all because its subject matter is exclusively the phenomena of the human soul, but not of the mind(7).
A second aspect of this problem is perhaps more serious because lack of consideration of it has stultified, or at best rendered inconclusive, so many research projects in the social sciences, and will continue to do so if we cannot overcome this difficulty. It is closely connected with the postulate of conjointness of theoretical conceptions and empirical observation of social phenomena. I wish to refer to the fact that, perhaps because of the greater intricacy and complexity of social phenomena and social problems, once certain directing conceptions have been evolved from their application in a specific research, they tend to become general currency while their particular frame of reference is neglected, and are taken for granted as some kind of general principle. On the lowest level of abstracting general conceptions from factual information, a case in point is the stereotypes of public opinion to which Walter Lippmann devoted the best part of a very good book(8). On the higher level of theoretical thinking in the social sciences, I would mention as an instance which has evoked much discussion the positive correlation which supposedly exists between the religious
attitude of Calvinism and the early capitalist economy. Erich Fromm, in his Fear of Freedom,(9) uses this as a general principle by which he wishes to support his psychological character analysis of Luther and Calvin: what he takes for granted as “the relation of the new religious doctrines to the spirit of capitalism “is rather a far cry, to put it mildly, to the very specific conceptions, like that of the calling, which Max Weber suggested in his research on economic ethics of world religions, and which were still further modified in the fine work of R. H. Tawney. Tawney's argument culminated in the remark, “like the author of another revolution in economic theory, Calvin might have turned on his popularizers with the protest, ‘I am not a Calvinist,”’(10); that does not fit very well with a research project which aims at its results through a character analysis of Calvin, and—and this is the point—at the same time employs as a general principle the researches of Weber and Tawney. In a similar manner, I believe, F. Hayek (Road to Serfdom) has taken over, generalized, and at the same time over-emphasized and emptied, of their original significance the conception of certain, trends in modern politics as denned by the French historian, Elie Halévy, in his paper, L'ère des tyrannies.(11)
On the other hand, there are the theoretical conceptions which by other authors were intended to attain the status of directing conceptions and guiding principles for further research, but which somehow did not “sell”—as the four wishes in Thomas's and Znaniecki's work on the Polish Peasant. These did not manage to become general theory, possibly because as purely psychological categories they were not competent to grasp social reality, perhaps because they were put forward at a time when behaviourism was still the leading if not the fashionable school in social psychology(12).
And finally I must mention in this connection the trouble to which Gunnar Myrdal, in his American Dilemma, went in order to avoid any implication of his accepting theoretical conceptions without their having been derived from, and checked against, the subject matter of his own work. The upshot of it was that he wrote a chapter—“The American Creed”—which constitutes the only legitimate frame of reference in his survey of the American dilemma in the negro question. I have dwelt on these cases at some length in order to make explicit the difficulties which I contend have to be solved before we can as it were covenant with a logic of social science thought. And I beg you to keep in mind these questions raised—for more than raise these questions I realize I have not been able to do—and to remember their bearing on the solution of our main problem.
(3) There is a final point which it is necessary to examine if we wish fully to grasp the import of the statement from which we started. “Social conflicts and confusions exist in fact, before problems for inquiry exist.” Again, comparing the state of affairs in the physical sciences with that which obtains in social-science inquiry, we will find that in the former the subject matter has mainly been prepared by prior inquiry. The material in social science, however, our author urges, exists chiefly in a crude qualitative state. The
problem has first to be. solved by means of what methods it can be converted into “prepared” material which may be subjected to scientific treatment.
We may describe in shorthand as “analytic discrimination” the work necessary to convert a problematic situation into a set of conditions forming a definite problem. The analogy of medical diagnosis will help to indicate the nature of this work. What this means in social science, I may perhaps make clear if I give you a case in its crude qualitative state just as it comes from the newspapers:
Wharf labourers and railwaymen strike in order to have their wages adjusted to the rising price level, but by doing so delay U.N.R.R.A. shipments which, so the policy of their organisations determines, should at the greatest possible speed and in the greatest possible amounts reach the war-devastated countries.
There is undoubtedly a confused and conflicting situation behind this story, setting a social problem of the first magnitude, the solution of which would be worth the best efforts of the social scientist. But the work of analytical discrimination required to make a problem like this scientifically tractable is largely omitted in administrative and political practice—with the inevitable result that methods to resolve the problematic situation are proposed and acted upon without any clear conception of the material in which projects and plans are to be applied and are to take effect. You will note that if we take our author at his word, it is the persons directly engaged in the management of practical affairs (whether President Truman or the secretary of the Seamen's Union) whom he charges to act in such a manner that the consequences of their actions conform to the requirements of analytic discrimination, or diagnosis, or scientifically sound treatment. Hopeless though this sounds if held against a big problem like the story from the newspaper cutting, there is no doubt that it is quite logical: you expect a similar thing from the architect in the housing department in his field of activity, or from the pharmacist who dispenses social security prescriptions. Then, we ask, why not construct the activity of the administrator on the lines of the same scientific logic?
By arraying before you all these difficulties and by making straight for the point at which the consequences of a consistent pragmatism become quite clear, I wish to do no more than penetrate to the problem which looms largest in the modern logic of social-science thought. If the complaint about the backwardness of the social sciences may have a history of three hundred years or more, the solution which I now propose to discuss is modern, in fact twentieth century, so much so that we will lose very little by leaving even John Stuart Mill's logic of the social (or moral) sciences where he has propounded it in the sixth book of his system. As our main proposition, I submit now a statement which has been taken from the same author with whose definition of social problems I began. It expresses succinctly the point to which he was leading. The sentence reads as follows:—
The experimental method is the only legitimate procedure in the solution of social problems, as it is the only legitimate procedure in all scientific research.
I propose to discuss this proposition in such a manner that it can be subjected to strictly logical criticism. I should perhaps say here that I am not concerned with the daily usage of “experiment,” as for instance in the following cases which come from my file of newspaper cuttings, all under the same date line:—
Professor Laski at the recent Labour Conference suggested that after a long period of mutual distrust and suspicion British Labour should make right now an experiment in friendship with Russia;
“The great allied operation against Europe was an experiment unprecedented in the history of the world” (General Eisenhower, or more likely his public relations officer);
The biggest experiment in marketing ever attempted, is how Mr. Arthur, formerly of Dunedin, described the plan to extend the wool-selling season from five months (November to March) to nine months (September to May).
If we replace in these examples, the like of which can easily be found in the daily press by the score, the word “experiment” by “action” or “attempt,” little if anything would be lost in contents of the message. Why, then, say “experiment in friendship,” “experiment in strategy,” “experiment in wool marketing,” ? T I do not suggest just because it sounds rather a mouthful. It would seem that people speaking vaguely of experiment really mean to convey that the greater the attempt, and the more largescale the action undertaken, the greater the possibility of its being beyond the control of the persons who are responsible for this attempt or that action. There is a chance of failure, of the situation getting out of control. However, should this be implied by the use of the word “experiment,” it would be precisely what has been excluded in a true experiment—an experiment requiring complete control over its outcome in order to record and register it in a manner which permits scientifically significant and reliable communication about it. So much of common usage which is not my concern here.
In the definition which our author gives of experimental method, every measure of policy put into operation is in the nature of an experiment, for two reasons:—
It represents the adoption of one out of a number of alternative conceptions as possible plans of action;
Its execution is followed by consequences which are observable, so that they may serve as tests of the validity of the conception acted upon.(13)
These two reasons are sufficient, he continues, to entitle the social scientist “to take advantage of the experimental method.” There is no doubt that they must be considered essential for any scientific approach to a practical situation which being conflicting and confused is problematic. It will be found on further examination how difficult it proves to be in the social sciences adequately to fulfil the requirements stated in these two elements. The second
will occupy us at a later stage. As regards the first, “adoption of one out of a number of alternative conceptions as possible plans of action,” I am too forcefully reminded to withhold it from you of the criticism of Gladstone's policy of expediency.(14) The critic holds responsible for this policy the true disciples of Bentham, and then he continues:
This is significantly expressed in a formula which they took over from Robert Peel and which, despite all the ridicule with which it met, sticks to them: I see, they say, three ways before me; the first I cannot go for these; the second not for those reasons; therefore I must choose the third. Only too often they succeed by this in distracting their audience from the first premiss, the inquiry whether there might not be ten, instead of only three, possible ways among which to choose.
The two elements of experimental method—deliberate choice of action and exhaustive observation of consequences—having been stated, our author denies it to be essential to the nature of experiment
that social phenomena do not permit of the controlled variation of sets of conditions in a one-by-one series of operations.
In other words, regarding social phenomena, and therefore in the research into social problems, it is a fact that you can make your experiment only once; that is, after you have adopted one set out of a number of alternative conceptions as possible plans of action, you have made your choice, you cannot, as it were, go back on it, start all over again, adopt another conception, perform a new operation, and so forth repeatedly. Our author, in the same sentence, claims that this fact does not render the experimental method inapplicable in the social sciences. From this we may infer as a third proposition that controlled variation of sets of conditions in a one-by-one series of operations is not an essential element of experimental method.
Let us consider this point more closely. First of all, I am not sure if the scientist, the chemist for instance, would be prepared to follow the logician if he thus defines experiment. For the scientist, the place where he carries out his experiments is the laboratory in which the requirement of a one-by-one series of operations on similar material can be satisfied in such a manner that controlled variations of conditions or of sets of conditions are possible, and the consequences following upon each operation can be observed and recorded, and thus the validity of conceptions be tested. This method has been distinguished as the “pure, or laboratory, or direct, experiment,” from other forms of research techniques which are loosely comprised by the term experiment; such as the uncontrolled or indirect experiment, the ex post facto or retrospective experiment, controlled observational studies, trial-and-error methods.(15) Indeed, in examining the literature on method, we find a bewildering variety of meanings given to the term “experiment”; Talcott Parsons, to gibe only one example, suggests what he calls mental experiments by constructing objectively possible courses of events besides the actual course which the event has taken, and by a rather remarkable short-circuit of reasoning maintains that experiment is nothing but the comparative method.(16) We may concede, with the Oxford English Dictionary, to call “experiment” any “action or operation undertaken in order to discover something
unknown, to test a hypothesis, or establish or illustrate some known truth” —which, however, would not be very meaningful, and, besides, somewhat out of focus for our purpose as long as we continue in the pragmatic view. The term may become clearer by elimination: we eliminate the archaic usage of “the action of trying anything, a test, a trial” (which if the O.E.D. were right would be the death warrant of the common usage). We eliminate further the sense of experiment as a “procedure adopted in uncertainty whether it will answer the purpose”—which comes very near the extension of the wool-marketing season from five to nine months. This last meaning, we may note in passing, is supported by O.E.D. by a sentence of Francis Bacon, “It is good… not to try experiments in States.”(17) Skipping two hundred years of scientific history, we find Jeremy Bentham putting on record that his work in the moral sciences was “an attempt to extend the experimental method of reasoning from the physical branch of sciences to the moral” (moral being equivalent to our social).(18) Another century and a-half, and the recently published Dictionary of Sociology defines experimental method as “that branch of induction which attempts to confirm or disprove some tentative conclusion by repeated observations or demonstrations.” There is no end in quoting specimens of the scientific usage of experiment and experimental method, some clearly metaphorical, some only intelligible from their specific historical setting, some too wide and too vague as concepts to be operationally useful.
To return therefore to our defined problem of experimental method: I would admit that to an extent it is possible to devise even in social science controlled variations of sets of conditions, although we find that educationists and industrial psychologists in the staging and carrying out of experiments are more often than not satisfied with one set only of conditions, and at once proceed to observing and comparing the results of their experiment with their control group. More directly: if an intelligence test is given, it would be possible and necessary to apply under varying conditions varying tests to different groups, and compare the results of each of these tests among themselves and with the control group. Or, if you test conditions of efficiency in factory production, you would have to choose a number of different groups of workers to whom tests of varied character are applied, and so forth. Procedures such as these may turn out to be more involved, to require a widening of the field of research, and in its turn may necessitate a more elaborate system of first-hand records and correlation of results, and may be more expensive than the carrying out of one test, or the recording of several case histories. Suggestions on similar lines I find have been made in a recent paper in the American Journal of Sociology whose author(19) at the same time considers adequate for most purposes the less involved—he says: less “mathematically sophisticated”—statistical techniques of correlation of the type worked out some thirty years ago by Pearson and Toennies. Repetitive studies are also advocated by the social psychologist Kimball Young, who condemns it as “the crying disgrace of sociology, as of the other social sciences
as well, that they have failed to resume the ancient responsibility of all scientists to take seriously one of the oldest canons of science: the duplication of controlled observation.”(20)
Why, then, I persist, is it nevertheless true that “social phenomena do not permit of the controlled variation of sets of conditions in a one-by-one series of operations”? It was this stubborn fact which compelled our author explicitly to refuse to pattern his definition of experimental method after the type of what we have classified as pure and direct and laboratory experiment. We cannot explain away this fact—that much I hope to have made clear so far-but if we can show its why and wherefore we may reach an understanding of our proposition of logic of social-science thought. I can perhaps offer an answer which I hasten to add is neither complete nor free of flaws, but which I would ask you to give a hearing, since I believe it will advance the discussion a step further. We may even arrive at some judgment whether or not this step is, as I suggest it is, decisive for the whole logic of social-science thought.
I must interpolate here that the experiment our author is thinking of is of a type which differs from the examples I have just given. He thinks, as I would infer from his own example of legislative measures, in terms of governmental activities in dealing with social problems that come under government jurisdiction. He is quite outspoken in his stress upon any such activity being essentially in the nature of an experiment. If we follow this suggestion which from a pragmatist point of view I need not further qualify, then the question which I raised and the answer which I shall try to give now would be only the more pertinent.
The chief reason for a direct experimental method of controlled variation of sets of conditions being impossible of execution in the social sciences is that you can neither discard nor replace the material to which your experiment has been applied. While you have, in natural sciences, any number of amounts of the unit of the substance or material with which to repeat experiments by controlled variation of sets of conditions—whether plants, chemical substances, or guinea pigs—you have one substance, and one only, to which to apply your social-science experiment. After it has been applied, the substance has undergone such changes as were effected by the experiment, that is, the substance is now loaded with the consequences of the experiment, and to revert the process by nullifying the consequences, and undoing the changes is not possible. Nor can you scrap the material with which you have experimented—in social science you have no guinea pigs, except metaphorically speaking.
I can deal here only briefly with the supplementary question whether this is true also if you consider not governmental activities of legislation or administration or business management, but experiment of the small-scale social-research type, such as those of education or industrial psychology which I mentioned earlier. In them, we have, as compared with social problems, that is, compared with conflicting and confused social situations which have been operationally defined and are about to be solved by the
so-called experimental method, undoubtedly a closer approximation to scientific laboratory conditions of experiment. But even there it remains an approximation. Without a great deal of assumptions before and reservations after the experiment, we cannot, even in research projects of this kind, know whether the experimental group or groups and the control group are in all their relevant, features alike or sufficiently alike to warrant any of the conclusions which we seem to be entitled to draw from the result of such a test or experiment. The problem of all-controlled experiment—to determine the variables, and to exclude all irrelevant factors and control all relevant factors—remains even here subject to one unsurmountable obstacle. For we do not know in advance which factors (provided we succeed in a satisfactory factorial analysis of a social-science problem by thorough analytic discrimination) are relevant, and which among the relevant are variable; and if we assume relevance and variability, it is only in a rather vague and often haphazard manner, which at its best is hypothetical but which in the experience of most social scientists is scientifically unsatisfactory. If we have to administer tests of an experimental kind to our subject matter to find out about the relevancy of certain factors, and the variability of others, these first tests will be haphazard if not uncritical, and therefore biased and unscientific. It becomes clear that we are only moving in a vicious circle.
While putting forward this heresy, I am not unmindful that the refinement of techniques, more particularly of statistical methods, has been substantial in recent times. But even the best and mathematically most perfected statistical method does not prevent its being applied wrongly, nor will it ever solve the fundamental problem of approach.(21) I have for reasons of economy to omit here the old question of calculating social phenomena, which I suggest could profitably be discussed by this section on another occasion.
Now, social-science methodologists have certainly not been blind to this problem and just brazenly insisted on the applicability of experimental method to their subject matter. In order to save the experimental method they have drawn support from the rules of the classical logic of scientific induction by which the solution of a problem is first approached by a hypothesis, or even a “hunch,” then the hypothesis tested in an attempt to verify it, and finally from the results of the verification procedure a general statement of the solution is formed. The experimental method would then presumably work within defined limits, and in this manner a valuable amount of control over the experiment itself be gained. What may be gained, however, in control is to some extent, if not in inverse ratio, lost by the institution of a hypothesis which often materially changes, often garbles the terms of the initial problem.(22) But even if we admit the trichotomy of inductive scientific procedure in hypothesis—verification—proposition, it would not really affect the issue, which is experiment as an empiristic method in social science. The forming of the hypothesis is partly a matter of what I have described earlier as analytic discrimination in the forming of a problem by intellectualisation
of a confused and conflicting situation; and it is partly a matter of evaluation on which something will have to be said later on.
I may now have reached a point where the objection seems inevitable that by denying the applicability of experimental method to the solution of social-science problems, we would sacrifice what little progress there has been made in the pursuit of social inquiry; that in my search for the reasons of the backwardness of the social sciences I have, as it were, turned the clock still further back. I have mentioned before, in fact it was part of the argument, that this dilemma has led to a modification of what in the social sciences may be considered to be an experiment. These modifications have perhaps pushed back the difficulty, but not solved it, for in a problem of logic one cannot compromise. It will then be incumbent upon me to supplement the negative answer which I have given so far, by a positive answer.
I shall now try to give this second part of my answer. You will recall that we were in agreement when I insisted with our author on the fact that social phenomena do not permit the controlled variation of sets of conditions in a one-by-one series of operations. I had come to the conclusion that by adopting one set out of a number of alternative conceptions as possible plans of action, and by acting on it as legislators or administrators, we irrevocably commit ourselves to one course and change our substance. Our material has been subjected to the impact of one actual operation out of several possible ones, and it will never be the same as it was before the operation was performed.
Now, if this fact made us wary of speaking of experimental method, we can, on the other hand, describe this operation as an event of history. Whether anti-strike legislation is passed to resolve certain types of defined or definable situations of social trouble; whether a government official carries out his assignment to remove economic tensions by administrative measures as (in the American phrase) “chief trouble shooter”; whether a judicial or child-welfare decision is made in an individual case of delinquency: these activities are always events of history—history of labour relations, history of industrial and business relations, history of a case of maladjustment to its social environment, and so forth. I would be prepared to go even further, and say that the administering of an educational test is an event in the history of a school career, and that the activities of the industrial psychologist in changing experimentally the conditions under which factory production takes place, are events in the history of industrial production.
Why do we study history, or, less sweeping, how do we study history? By observing the consequences which follow upon certain events. Consequences which are considered the product of changes in the composition, structure, character of the group of people which is the subject of our study, or of the institutions by which this group of people lives. By considering events which are produced as consequences of our own choosing, which (to repeat) represent as policies enacted the adoption and execution of one out of a number of alternative conceptions as possible plans of
action, and by observing the changes consequent upon these events, we have reversed the cause-effect relationship, and are from now on concerned with means-end relationships. This point is decisive. A long book which R. MacIver wrote on Social Causation could contribute very little to the problems of the social scientist because MacIver did not consider the reversal from the cause-effect aspect of social problems to the means-consequences relationship.(28) I leave no doubt that it is one of the great achievements of the advocacy of experimental method in the social sciences to have made this point. This is particularly true of our author whose logic of inquiry in the social sciences I have discussed at such length. He says:
Causal propositions (in the sense of propositions whose content is a relation of conditions that are means to other conditions that are consequences) are involved in every competently conducted inquiry. To bring about, to make, to generate, is to effect, and that which serves this purpose is a cause in the only legitimate sense of the word.
This may sound revolutionary to the historian as well as to the social scientist who approaches any historical question in the accepted sense of historical inquiry.(24) But we do in fact no more than invoke the teleological principle of Aristotelian logic. If we prefer to call it pragmatic, let us by all means do so, especially since it cuts us off from all inquiry which is purely speculative. It should, however, be understood that this is essentially also the position taken up by R. G. Collingwood, who, in opposing what he calls the “scissors and paste” brand of historical inquiry insists that “historical problems arise out of practical problems.”(25)
From this point of view, that is, the historical point of view, it may be useful to reconsider the preliminary postulates or praecognita for social inquiry, which I mentioned before discussing the experimental method. Let me briefly recapitulate these postulates, viz. (1) the conjointness of observational attainment and appropriate operational conceptions, and (2) the postulate of social-science research being conducted within its cultural matrix, and with proper attention being paid to the technological and physical climate of social problems, are the stock-in-trade of the historian who has grasped that the conceptions and theorems of philosophers and scientists of a period are conditioned by the social and cultural environment of their own times. It is in fact the essence of what we may call the “sense of history” that the interpretation and appraisal of theories like those of Plato or Descartes or Newton rests upon an understanding of the conditions obtaining in their own times. The same must then be true of sociological research into conditions of our own times, once we admit that history is understood to include the present. For it is absurd “to suppose that history includes events that happened up to yesterday, but does not take in those occurring to-day.”(26)
The other two points (3) the cautioning against taking conceptions for granted by giving them general currency while they have evolved from a particular and specific investigation; and (4) the requirement of analytic discrimination in order to convert a confused and conflicting situation into a set of conditions forming
a definite. problem,(27) can both be made to fit in with a logic of social science thought which considers social problems as historical problems.
Before I try to enlarge upon certain aspects that follow from the position which we have now reached, I must add a second point which I feel has been especially well made by the advocates of the experimental method. The claim which our author made—that the enactment of every policy is logically, and should be actually, in the nature of an experiment—would, if it were acted upon, be bound to discourage laxity and discontinuity in discriminative observation of the consequences of policies adopted. The working social scientist is only too aware of these shortcomings. In an experiment, observation of the consequences, is imperative, essential, and of paramount importance. In practical social work, we see very little, and that only in rough and vague approximation, that is known, for example, about the effect on the standard of living of certain taxation legislation;(28) about the consequences for an industrial policy of full employment of pegging of prices; or about an anticipated improvement in scholastic achievements through the adoption of a new teaching syllabus. Who among social practitioners, whether teacher, social worker, economist, judge, can claim to be free of laxity and discontinuity in observing discriminately the consequences of policies enacted? Departmental specialisation goes a long way to explain both laxity and discontinuity of observation of consequences, by limiting responsibilities. And I would like to believe that the argument for experimental method in the social sciences has been motivated by the desire to counteract the consequences of specialisation.(29) But we have to consider this question above all as an eminent problem of practice of which every one of us will have a tale to tell. In social case work, for instance, it has always been one of the most difficult, but often sadly neglected tasks of training social workers how to do so simple a thing as writing a report of factual observations in a specific case in such a manner as to justify conclusively recommendations made or actions to be taken.
Observation is being made with every possible care and discrimination in those few cases which are being termed expressly, and because of their being isolated cases, “experiments”; take the Feilding Community Centre,(30) or the Peckham experiment(31) or experiments with a backward class.(32) There is no doubt that the treatment as though they were experiments of such special cases will enable the observer to arrive at some reasoned statement of the type of policy which would have to be pursued if such-and-such consequences are to result from action taken. But these cases are few and far between, they will often be found to have been instituted not as administrative measures, or to have been undertaken not as a matter of governmental policy. The very rareness of social experiments would suggest that we are as yet far removed from a state of affairs in which it can be claimed that the solution of social problems is performed according to rules which are valid in experimental method.
I believe that there are some practical and some theoretical reasons for this imperfection of the social sciences; and it is these reasons (which are ultimately responsible for the backwardness of the social sciences) to the consideration of which some further remarks must be devoted. Let me deal first with the theoretical questions. I see four main questions lack of an appropriate solution of which continues to obstruct progress in the social sciences. The first two I shall only outline:
(1) One is the fallacy of taking facts as finished and over with, which our author says is more serious in social than in physical inquiry. Once we admit that social facts are inherently historical, it follows that they are “sequential courses of change.” We would destroy the very quality which makes them social, if we removed and isolated them from the history to which they belong. The wider problem which is here indicated is that of “social change,” or “social dynamics.” As far as I can see, despite many attempts no satisfactory theoretical conceptions have yet been found to cope with it, let alone to find a general formula. The more sedate type of treatment, like that of MacIver in his text book, Society, as well as the adherents of the brand new “operational sociology” like W. S. Dodd and George Lundberg are equally far from supplying a competent formula in what has been called “the grand manner.”(38) It is nevertheless possible in every special inquiry to fulfil several more special requirements of logic which would permit us to cope with the vast problem of change as such.
(2) A second question is closely related with the fallacy of taking facts as finished and over with. This is the misconception that “the facts are just there and need only be observed,” or in the cruder form which it often takes in political oratory, that “the facts speak for themselves.” Let me here only repeat that taking conditions as they are is either a truism or a fallacy; a truism if we wish to say that existing conditions are the material of our observation, and a fallacy if they are taken to be fixed whilst they are changing and in process. The truth is that facts do not speak for themselves, but on the contrary have to be cross-examined, and this process of cross-examination should in fairness be held as publicly as any examination of witness before court.
(3)(34) A third question is the problem of generalization. It differs not only in degree but in kind from the general problem of induction, once we remember the historical oneness in time and space of social phenomena. The logical problem is whether inference from one or some to all is permissible. All inference involves extension beyond the scope of already observed objects. This would require the observed objects, or, to be more correct, the observed facts about them, to be representative of the larger universe from which they are taken, and about which a general statement is to be made. Any other form of general statement, that is, one which cannot be termed a secure inference from specific observed cases which are representative, is in the nature of a hypothesis. While I am not sufficiently familiar with the history of logic to affirm that traditional logic has too often failed to
recognize the necessity of abstract hypotheses which will involve purely deductive relations of propositions, I believe that it is safe to say that there has been in recent years a growing realization in the social sciences of the hypothetical character of almost all general statements. Unfortunately, there prevails a confusion of terms which urgently needs clearing up. We have met earlier with hypotheses which stand at the beginning of a research project—there the term was used as shorthand for the formation of a definite problem of inquiry. Here we find hypotheses at the end and as the result of research and inquiry. It is probably more correct to term the latter kind, that is, general statements which are arrived at by an appropriate inquiry into a defined social problem, heuristic principles which remain subject to revision. In any case, the realization of this difficulty in the social sciences has been so marked that only the greater a need is being felt for generalizations which rest logically on acceptable inferences.
The recent development of statistical method is sustained by this impulse; in its mathematical foundation, however, the concept of probability has replaced the immediate aim of inferential generalization. I need not remind you that the concept of probability has engendered the “experimental” nature, in a sense, of social inquiry. Moreover, in the social sciences, we encounter a difficulty which has no counterpart in the natural sciences. For in the natural sciences it is possible to determine by prior experimental operations the one to be an exemplary specimen of all. This procedure, and the more complicated procedures which satisfy the notion of inference, presuppose prior experimental operations with the specimen. The details of the logical procedure of inference further reveal that samples are of an order different again from exemplary specimens. We speak of samples if the given thing is an element in a homogeneous continuum. In the example of a sample of grain it has been necessary first to institute homogeneity by mixing the universe in such a manner that a handful of grains taken from a bin will represent in proportional distribution all the constituent grains in the bin. This procedure has become so much an every-day affair without which large merchandizing is unthinkable that the logical character of the procedure presents no serious difficulty. In the social sciences where the material is men and institutions, not grains of wheat or their quantities, the experimental determination of the properties of the specimen, and the mixing of the universe to obtain fair samples are inapplicable. This fact has caused, and continues to cause, the social scientist considerable qualms whenever he subjects his material to tests which are designed in analogy to sampling techniques and inferential generalizations in the natural sciences. The difficulty cannot, I maintain, be overcome by improvements in statistical method. It can be allayed by the judicious employment of the results of other researches to check one's own results, by repetition, by significant correlation, and so forth. Even under the most favourable conditions, however, straight-out inferences to support true generalizations will remain rare in the social sciences; they will remain approximations only—that is the sense in which we speak
of probability, which is a probability of correct knowledge, not a probability of events.
The only way out of this dilemma that I can suggest is just as radical as my criticism of the third of the three constituent elements of the experimental method: to discard altogether the procedure of drawing inferences based on specimens and fair samples whenever it is possible to work with the whole of the universe; to have all members of the homogeneous continuum counted and tabulated and to present the results in such a manner that they are at once intelligible and controllable. A poor solution? Perhaps. We may, however, derive some consolation from the fact that to begin with the numbers involved are relatively smaller, and that much depends on the proper institution of a research project. In this respect I take it as a sign of healthy life in the social sciences that local and regional surveys have come to stay.
(4) The fourth question is that of evaluation. Let us consider the following proposition:—
It is a sound principle that all moral approbation and condemnation should be excluded from scientific inquiry. This principle is often wrongly converted into the notion that all evaluation should be excluded.
The conversion, our author states, rests on the fallacy that the moral blames and approvals are evaluative, whilst they are not so in any logical sense of evaluation. They are neither evaluations nor judgments because they assume that certain ends should or ought to be attained or not attained. Ends are thus excluded from the field of inquiry, and this exclusion contradicts the means consequences logic on which we settled earlier. Not only is judgment and evaluation required in order to establish a strict conjugate relationship of means and consequences—that is, means have to be adjudged as appropriate to attain defined ends, and ends have to be adjudged on the basis of available means by which they can be attained. Furthermore, to select out of the mass of data which are potentially available for observation that material which may be considered to be “the facts of the case” is adjudgment, appraisal, evaluation. “Without it, there is no guide for observation; without it, one can have no conception of what one should look for, or even is looking for. One ‘fact’ would be just as good as any other—that is, good for nothing in control of inquiry and in settlement of a problem.”
At this point, one sees with regret the author leaving the topic. Something more is required in a logic of social-science thought which purports to be a theory of scientific inquiry. For every social scientist has been exposed to a grave danger whenever it came to evaluation in the course of his research. Every step in the evaluation procedure whether in the selection of the relevant facts, or in the adjudgment of the conjugate relationship of means and consequences, is beset with the danger of biases, which perhaps more correctly should be called concealed or hidden valuations. This question of concealed valuations in the mind of the working scientist, and some answer how to overcome them, would seem to deserve some mention, if not a chapter to itself, in any logic of social-science thought. I cannot, at this stage, launch out into a discussion of the problem of objectivity in the social sciences in
all its ramifications. Not only would this open up to question the principle of pragmatism which I have endeavoured so far to follow as a presupposition in my inquiry. But I could neither adequately survey it without introducing the work of Max Weber on the objectivity of social-science knowledge, which is one of the earliest and still much discussed contributions to the sociology of knowledge, nor without making reference to epistemology and semantics. Fortunately, in raising the issue, I can at once limit it by referring to the authoritative treatment by Gunnar Myrdal to whom we owe an exhaustive and, to my mind, convincing discussion of the problem of concealed valuations. His survey is closer to the specific troubles of the mid-twentieth century social scientist than the five chapters on bias by Herbert Spencer,(35) and insofar more to the point. Myrdal's solution which I trust is psychologically sound is, first, to become aware of the danger of concealed valuations in its magnitude, and beware of anything which may “drive underground” valuations; and, second, to face them and, since it is humanly impossible completely to avoid valuations beyond the evaluative requirements of a specific problem, to introduce them as explicitly stated, specific, and sufficiently “concretized” value premises. The necessity of making explicit as value premises any possible bias in social-science research, is only the more urgent if we adopt the pragmatic view of social science, that is, if we wish to advance towards scientifically sound social engineering. Myrdal is very outspoken in his adverse criticism of the practice of driving valuations underground, instead of making them explicit as value premises. It is under this title that he attacks the utilitarian preconceptions of the “harmony” doctrine which he finds are so often made tacitly and always taken for granted. While his remarks are directed against the theorem of equilibrium, I gather that he attacks at the same time what he believes to be a typically American complacency about deepseated social conflicts and confusions like the one which he and his staff have surveyed. But what he has to say on this subject, although tucked away in a number of appendices, will beneficially stir any social scientist.
This question of evaluation leads me to a question of social science practice. In our society, the social scientist is neither consulted in determining the ends of policies and the consequences in the means-consequences relationship which constitutes every scientific pursuit; nor is he free in devising, let alone in applying such means as would produce defined consequences. This is true, whether we imagine social scientists to be an independent body of research workers and teachers, or boards of industrial management, or commissions of inquiry, or governmental agencies. We all know that there are cogent reasons, historically and politically cogent, for this state of affairs. If you like, the particular logic of social-science inquiry which I have reviewed would seem to have revealed, by advocating the impossible, a situation which itself is conflicting and confused. Whether we bemoan or rejoice in this paradox, we cannot resolve it here and now. If, as an extreme case, we would imagine a society in which the social
scientist is called upon to operate under his own responsibility social means-consequences relationships, it would be tantamount to permitting him to make history. If that is unfeasible, then there will be limitations to the scientific activity of the social scientist. Yet I would not go so far as to say that these limitations render any pragmatically conceived activity of the social scientist useless.
Rather, after I have done my best to demolish the experimental method in the social sciences, it is now perhaps safe to refer the problem to the common sense of Sidney and Beatrice Webb, who are not afraid to talk about the use of experiment, and even speak of “the world sociological laboratory in which we live.”(36) They show how it is one person, or group of persons, the persons in authority as entrepreneurs, legislators, administrators, ministers, often having no consciousness of the experimental nature of their work and no scientific interest in its results—who do their jobs as unintentional experiments; and another person or group of persons—the social scientist investigator—who do the observation and verification of consequences by diligently following the administrator's plough. In other words, we have a case of division of labour, or rather of functions between the administrator and the investigator.
I can now summarize my thesis that the dividing line between these functions in the social sciences differs fundamentally from the distribution of functions between the experimenter and the practitioner in the natural sciences. And as corollaries to this thesis I submit that
not until there is a social science can there be that distribution of functions between the administrator and the social investigator which cancels the otherwise inevitable neglect of scientific observation of consequences; and
the attempt—or, if you insist, the experiment—of constructing the function of the social scientist methodologically in parallel to that of the natural scientist, leads into a blind alley.
These two facts have placed the social scientist between the devil of experimental method of the natural sciences and the deep blue sea of having no social function. They have certainly in the past been largely responsible for the backwardness of the social sciences. Even so, whether or not we persist in calling it the pre-scientific stage of the social sciences, we have to-day far advanced in the understanding of society and how it works: so far that Auguste de Comte would be amazed at the accumulation in facts and theories of physique sociale. And yet not far enough to consider it a matter of course that social scientists are employed by industrial enterprises and governments alike, in sufficient numbers and with sufficient training, for the solution of social problems. It should not be forgotten, either, that many of the solutions in the past were achieved without the support, and even in opposition to the public opinion and government of their day: from an abundance of examples I mention only the ten-hour-working-day bill or the history of prison systems which at its very beginning is connected with the name of Bentham. Last, but not least, the cost involved
in the scientific work of the social investigator is almost negligible in comparison to the often huge expenditure which is being incurred to enable the pursuit of experimental studies in the natural sciences. The desired movement in the social sciences by which their backwardness may be overcome is largely a question of organisation and endowment.
M. Berr (La synthèse en histoire) quoted by M. Ginsberg: Studies in Sociology (Methuen, London, 1932), p. 27.
John Dewey: Logio, The Theory of Inquiry (Holt, New York, 1938), chap. xxiv: Social Inquiry.
The Public and its Problems (London, Allen and Unwin, 1926).
R. G. Collingwood: Human Nature and Human History, in Proc. Brit. Acad., vol. xxii (1936), pp. 97–127.
An Autobiography (Oxford University Press, 1939).
An Essay in Metaphysics (Oxford, Clarendon, 1940).
G. Myrdal: An American Dilemma, 2 vols. (Harper, New York and London, 1944), app. 1 to 3 (pp. 1027–1070).
M. Weber: Die Objectivität socialwissenschaftlicher und socialpolitischer Erkenntnis (1904), in Gesammelte Aufsätze zur Wissenschaftslehre (Tübingen, 1922), pp. 146–214 (an English translation of which is about to be published).
The references and quotations in the following are from chap. xxiv of Dewey's Logic, except where specially noted.
Dewey, l.o., “The subject-matter of social problems is existential. In the broad sense of “natural,” social sciences are, therefore, branches of natural science.” For recent discussion of the problem, cp. A. N. Whitehead: Adventures of Ideas (Cambridge, 1933) part i: Sociological; K. Walker: Meaning and Purpose (London, 1944); Jules Monnerot: Les Faits sociaux nee sont pas des choses (Paris, Gallimard, 1946).
This assertion may be referred, more especially, to the various attempts to construct sociology as a normative science by establishing “laws” of social intercourse and association; and among these, chiefly the schools of thought which tried to pattern norms of social life similar to general concepts of jurisprudence. The search for social laws, however, departed also from an assumed parallelism to natural laws. In modern sociological theory, the problem has been converted into a theory of values; cp., K. Mannheim: Ideology and Utopia (London, Kegan Paul, 1936), p. 73: “Even granting for the moment that this conception had some merit, the existence of certain formal realms of values and their specific structure would be intelligible only with reference to the concrete situations to which they have relevance and in which they are relevant.”
E.g., the “individual approach” of F. H. Allport: Institutional Behaviour (Chapel Hill, University of North Carolina Press, 1933).
An Essay on Metaphysics, pp. 112 ff.
Walter Lippmann: Public Opinion (Harcourt Brace, New York, 1922).
First published 1941 (American title: Escape from Freedom), now in: International Library of Sociology and Social Reconstruction, ed., K. Mannheim (Kegan Paul, London), cp., pp. 43 ff.
R. H. Tawney: Religion and the Rise of Capitalism, 1926 (Pelican Ed., 1938, p. 107).
Read before the Société Française de Philosophie, 28 Nov., 1936 (Paris, Gallimard, 1938, pp. 213 ff.); English version in Economica, February, 1941.
cp. P. Sorokin: Contemporary Sociological Theories (Harper, New York and London, 1928), p. 658.
The complete statement of the second, element reads: “its execution is followed by consequences which, while not as capable of definite or exclusive differentiation as in the case of physical experimentation, are none the less observable within limits, so they may serve as tests of the validity of the conceptions acted upon.”
L. Bucher: Der Parlamentarismus wie er ist, 2nd ed. (Stuttgart, 1881), p. 129.
cp. E. Greenwood: Experimental Sociology, A Study in Method (King's Crown Press, New York, 1944).
T. Parsons: The Structure of Social Action (McGraw-Hill, New York and London, 1937), pp. 612, 743.
Essays xxiv Of Innovations.
Quoted from Graham Wallas: Human Nature in Politics, 3rd ed. (Constable, London, 1938), pp. 119–120.
T. A. McCormick (Chairman of the Department of Sociology, Anthropology, and Social Work, University of Wisconsin): Simple Percentage Analysis of Attitude Questionnaires, in Amer. Journ. Sociol., March, 1945, vol. L, pp. 390–395.
Amer. Journ. Sociol., May, 1945, vol. L, p. 498.
J. Dewey, l.c., p. 499: “All the techniques of observation employed in the advanced sciences may be conformed to, including the use of the best statistical methods to calculate probable errors, etc., and yet the material ascertained be scientifically ‘dead,’ i.e., irrelevant to a genuine issue, so that concern with it is hardly more than a form of intellectual busy work.”
Robert S. Lynd: Knowledge for What! (Princeton University Press, 1938), p. 126: “It has been remarked that it is almost impossible to overestimate the significance of sex in human affairs, but that this impossibility Freud has achieved. In much the same sense, one may say that it is almost impossible to over-estimate the significance of refined statistical procedures for social science, but that some current researches are achieving that impossibility. If social science has tended to acquire humility as it has shifted from a priori theorizing to empirical observation and analysis, it is by way of reassuming its lost assurance through its erudite devotion to statistical manipulation.”
cp. R. M. MacIver: Social Causation (Ginn and Co., Boston), p. 158 n., criticizing “operationalism”: “Intelligence is that which the intelligence test tests.” See also M. R. Cohen: Preface to Logic (Routledge, London, 1946), pp. 141 and 156.
R. M. MacIver: Social Causation, p. 320.
For cultural anthropology, see B. Malinowski: A Scientific Theory of Culture and Other Essays (Chapel Hill, 1944), p. 67; id. The Dynamics of Cultural Change (ed. by P. M. Kaberry, New Haven, Yale University Press, 1945), p. 5. To understand the argument of J. Dewey for “experimental method” it should be noted that ten years before the publication of his Logic he had in mind “a certain logic of method” which “is experimental in the sense that… policies and proposals for social action will be entertained subject to constant and well-equipped observation of the consequences they entail when acted upon, and subject to ready and flexible revision in the light of observed consequences” (The Public and its Problems, p. 202 f.). On the other hand, it is realized in his Logic that the analogy to experimental natural sciences as regards the observation of consequences is not complete (cf. above note 13).
An Autobiography (Pelican Ed.), p. 78: “We study history in order to see more clearly into the situation in which we are called to act.”
J. Dewey, l.c., p. 501 f. It must be emphasized that Dewey not only insists on the essential historicity of social phenomena but recognizes in it the distinguishing feature in comparison with physical facts. This supports the earlier contention that his logic of social inquiry could not but modify the notion of experiment in sociology. On “historicity,” see R. G. Collingwood, Proceed. Brit. Acad., l.c., pp. 102 ff. (106).
J. Dewey, l.c., p. 499, indicates the method which follows from this approach: “Facts have to be determined in their dual function as obstacles and resources,” and “Realistic social thinking is precisely the mode of observation which discriminates adverse and favourable conditions in an existing situation, adverse' and ‘favourable’ being understood in connection with the end proposed.” For an application of this method, see F. Toennies in Verhandlungen des Fünften Deutschen Sociologentages (Proceed. German Sociolog. Society, vol. v, 1927), pp. 15–24.
Cp. The Economics of Full Employment (Oxford Institute of Statistics, Blackwell, Oxford, 1944), p. 91 f., and A. E. C. Hare: Industrial Relations In New Zealand (Whitcombe and Tombs, Wellington, 1946). Technical Appendix: The Distribution of Income.
J. Dewey, l.c., 508, has mentioned here only the academic, not the practical problem when he deplores “the existing division of social phenomena into a number of compartmentalized and supposedly independent non-interacting fields, as in the different provinces assigned, for example, to economics, politics, jurisprudence, morals, anthropology, etc.”
A. E. Campbell: The Feilding Community Centre (New Zealand Council for Educational Research, 1945), pp. 10 ff.
J. H. Pearse and L. H. Crocker: The Peckham Experiment (Allen and Unwin, London, fifth impr. 1945).
E. A. Taylor: Experiments with a Backward Class (Methuen, London, 1946).
R. M. MacIver: Society (Farrar and Rinehart, New York, 1937), Book 3: Social Change; and G. A. Lundberg: Foundations of Sociology (Macmillan, New York, 1939), pp. 503, ff.
Of the inadequateness of the remarks in this paragraph on the logic of statistical method in the social sciences I am painfully aware. An outline, however brief, of the problem which it is intended to present by a more detailed treatment, had to be inserted here so that the consequences of a logic of experimental method in social science may become obvious.
Herbert Spencer: The Study of Sociology, 19th ed. (Kegan Paul, London, 1897), chap. viii to xii.
Sidney and Beatrice Webb: Methods of Social Study (Longmans Green, London, New York, Toronto, 1932), pp. 220 ff.