Social Science Methodology as the

Ideology of Bureaucratic Authority

by Alasdair MacIntyre

(From Through the Looking Glass: Epistemology and the Conduct of Enquiry, ed. M. J.Falco, University Press of America, 1979; reprinted in The MacIntyre Reader, ed. K. Knight, NotreDame Press, 1998)

Charles Perrow has remarked of conflict in organizations that `The matter has been little studied beyond anecdotal or descriptive case studies' (Complex Organizations, p. 159). What we find in the literature in fact is a curious discrepancy between theoretical writing on organizations which all too rarely recognizes the fact of conflict except as a marginal phenomenon, and some descriptive studies which have identified conflict as widespread not only in industrial and political situations, but also in such institutions as hospitals and universities. The very ubiquity of conflict suggests that it cannot be merely a marginal phenomenon; yet the writers on theory use schemes of interpretation which insistently relegate it to that category. I want to suggest in this paper a perspective from which this discrepancy becomes intelligible by arguing that the social life of formal organizations is simultaneously ridden with conflicts and sustained by devices which enable participants not to recognize this fact. I want to argue further that the kind of theoretical writing which Perrow criticizes is itself in part an ideological expression of that same organizational life which the theorists are attempting to describe. One key element in this ideology of organizations turns out, so I shall argue, to be the conventional methodology of the social sciences. One key function of this ideology is to sustain bureaucratic authority. A necessary preliminary is to begin by providing a better understanding of the notion of conflict. To this task therefore I turn first.

I. Three types of conflict and the concept of ideology

Perrow in the passage already cited considers the type of conflict which arises in the competition of different groups within a formal organization [54] for such goods as security, power, and autonomy. The examples he gives include sales and production departments in firms, doctors, nurses and administrators in hospitals, and faculty and administration in universities. Of such conflicts, he writes, `To reduce, contain or use these conflicts is the job of the administrator' (Perrow, p. 159). Clearly administrators do generally succeed in such tasks and equally clearly this is one reason why theorists are able to see conflict as secondary and marginal in the life of organizations. But why are administrators able to succeed as easily as they do? Because, so I want to suggest, the participants in these conflicts themselves already share a view of organizations according to which conflict is secondary and marginal and act in accordance with that view. Consider, in order to point a contrast, a very different type of conflict, that involved in war.

What does a military strategist of genius - as against one of mere talent - a Napoleon or a Lee, a Liddell Hart or a Guderian, do? He makes himself and his effects as unpredictable as possible. By innovating in tactics, in organization, and in weapons' technology and use, he disrupts what has hitherto been taken to be the necessary structural order of a war or a battle. In so doing he treats the de facto regularities which have been found to hold in past wars and battles in a quite different way than do the lecturers at the Staff colleges of the nations whom the military geniuses defeat. The Staff colleges of such nations are often the unhappy heirs of some earlier genius. The Prussian General Staff codified the practice of Frederick the Great into a series of rules which they treated as the natural laws of war; Napoleon understood their codification, constituted himself a counter-example to their law-like generalizations and defeated them at Jena and Auerstadt. In 1861 the Union generals all knew what Napoleon would have done; they too believed themselves the possessors of the natural laws of war. Lee, who knew that they knew what Napoleon would have done, constituted himself a counter-example to their generalizations and defeated them.

Military history thus makes it clear that one very good way to defeat another nation would be to make sure that it used a certain set of textbooks in its military academies. By contriving that your opponents understood their actions, your actions and regularities embodied in military transactions between the two parties in one particular way, by inducing them to categorize what occurs on battlefields in one way rather than another, you would have made them available as future victims on the battlefield. In fact defeated nations, when they are defeated at the hands of the textbooks, are so defeated by reason of their own contrivances and not by those of the enemy. But it is clear that victorious generals have sometimes adopted the right textbook: Guderian's reading of Liddell Hart is a striking case in point. Hence there is a battle of textbooks, a battle of understanding and of concepts before there is a battle of military force, and one of the best ways to win the latter is to have won the former first.

There is an important parallel between this truth about military [55] conflict and an unnoticed truth about conflict in organizations. Those engaged in conflict within organizations normally understand those conflicts in much the same terms as Perrow describes them. They see themselves as having interests which lead them to want or to try to secure a greater share than their competitors of such goods as power, security, autonomy - and money (curiously unmentioned by Perrow). But in understanding their conflicts in this way they have already accepted some one particular way of understanding their situation, some one particular way of identifying interests, some one particular concept of reward or of power; and it may well be the case that by accepting this particular way of understanding their situation they have rendered themselves manageable and to this extent defeatable by others. They have already been defeated in the conflict of understanding and concepts as a preliminary to being defeated in a conflict of interests.

Take one key example, that of industrial work and its rewards. The dominant way of understanding such work under capitalism - and not only in America - is that whereby workers, management and investors all share in the distribution of what is jointly earned, in order that each gets as much as possible, and what matters is that as much as possible be produced. All three groups therefore have a common interest to which their particular interest ought to be subordinated. On this view men are primarily consumers and they work in order to be able to consume. On this view so long as each group recognizes the common interest it is rational for the others to do so; only when the other groups have selfishly relinquished that interest is it rational to be selfish oneself. So normal is this view among us that it seems natural too. But from another perspective it is precisely unnatural.

On this rival view, what is essentially human is rational activity, and consumption exists to serve activity and not to be served by it. We ought to eat in order to work, not vice versa. The classical expression of this thought is Aristotle's, but all artists, most professors and some socialists believe it too. Only sentimentalists believe that work ought or can be always interesting, but in an order where work serves consumption it is bound to be always uninteresting. On the first view my fundamental interest as a member of one group is in how large a share of the product of work I consume; on the second view I can have no fundamental interest in the continuance of an order that represents work, interest and rewards in the way that the first view does. It is clear that if the first view is universally or even just very widely held, the concept of interest employed will be such that conflicts over interests will be local, manageable and, if the managers are sufficiently adroit, marginal; but if the second view were ever to be held by even a substantial minority of workers, then conflict between them and the managing and investing classes would be endemic, central and possibly interminable. But the managing and owning classes do not have to fight this particular battle over interest and privilege as fiercely as they might otherwise, because they have already [56] for the most part won the battle over how interest and privilege are to be conceptualized and understood.

Or consider another closely related example. Modern liberal politics is dominated by a conception of the political process as one of bargaining between interests. Political morality consists in the observance of certain legally enforceable restrictions upon conduct; morality in general is relegated to private life. There is largely lacking any conception of political life as being the pursuit of a common good which transcends all partial interests and which can be realized by the individual only through his participation in political life. There cannot be, for our dominant effective notion of the common good is merely that of an artifact compounded out of individual and partial interests as a result of the bargaining process. Aristotle's Politics precisely insofar as it is a successful conceptualization of the Greek city-state fails to comprehend our mode of comprehension of politics and therefore our politics.

Yet matters are more complex than I have hitherto suggested. Human nature being what it is, you may turn your gaze insistently away from Aristotelianism, as the modern bourgeois world does, bit perhaps you can never rid yourself of its categories of interpretation entirely. In both the examples which I cited, certain non-Aristotelian modes of understanding and classification have prevailed. Politics, morality. law, economics and justice are all categories understood in a particular way. (In the modern world's understanding, for example, the notion of a just price makes no sense; justice belongs in one realm and the price mechanism in another.) Concepts such as interest, power, liberty (see C. B. Macpherson's Analysis in Democratic Theory: Essays in Retrieval) are understood in highly specific ways. (Notice incidentally that the concepts of interest and competition between rival groups presupposed in Perrow's analysis are the counterpart to my economic and political example.) But questions which cannot be intelligently asked within the dominant conceptual scheme still get asked. The relationship of the good citizen to the good man is an essentially Aristotelian question; and when so understood it is a question about the structure of community, about the distribution of certain dispositions (virtues) in a systematic way within the entire community. This is the question which haunted the Watergate inquiries; but because the conventional idiom would not allow it to be asked, moralizing took the place of inquiry. Instead of looking for corruption in the structure of community (or lack of it), it was looked for in the heart of Richard Nixon. (Not that it could not be found there.)

What I am suggesting is that unresolved conceptual conflicts partially determine the form of some overt conflicts; that it may be that the ubiquity of that conflict which organization theorists take to be marginal stems from the deeper unrecognized conflict in modes of understanding. Between the three types of conflict which have been identified - the first being those competitive conflicts between different interests within socially established organizations in good working order which Perrow [57] acnowledges, the second being military conflict, and the third being those conflicts of conceptualization and understanding on which I am laying emphasis - the resemblances are as important as the differences. The second and the third resemble each other and differ from the first in that they are `unmanageable' and they characteristically end in victory for one party, defeat for the other. But in the case of the third type both victory and defeat are usually tacit and unrecognized. Indeed part of what makes victory victory and defeat defeat consists, in this type of case, in their being unrecognized.

It is scarcely odd then that this type of conflict has largely gone unrecognized by social theory. With that lack of recognition has customarily gone a lack of recognition of the importance to be attached to certain other features of social reality, all of which tend to make that reality unmanageable and therefore such that if and when they do appear on the scene, they are apt to be treated by the manager and by his social scientific advisors as ghosts to be exorcised. The first is a feature of the relationship of understanding and interpretation of social reality to that reality, a relationship in at least one crucial point very different from that of the understanding of nature to physical reality.

If widespread brain lesions resulted in the loss of all our beliefs about atoms and molecules so that not a trace of such concepts remained in our thought or our language or was embodied in our practice, there would still be atoms and molecules, just as there are now, and nothing that is now true in particle theory would then be false. If similar lesions resulted in the loss of all our beliefs about representative government so that not a trace of that concept remained in thought or language or was embodied in practice, there would be no such thing as representative government any longer. Beliefs about the concepts of physical realities are always secondary to those realities; the physical world does not require us to have any particular beliefs about it or concepts of it, for it to exist. But with social reality it is quite otherwise. Social practices, institutions and organizations are partially constituted by the beliefs and concepts of those who participate in, have transactions with and attitudes towards them. Why do I say only `partially'? Because it is of course the belief-informed and concept-informed activities that are practices, institutions and organizations, and activities have other important relevant characteristics. But mere behavior by itself, abstracted from beliefs and concepts, is meaningless; confronted with uninterpreted or uninterpretable behavior we are forced into the position of the protagonists of Kafka's novels.

In any type of practice or institution of any complexity, the modes of interpretation that constitute the practice will not always be entirely coherent internally nor consistent with one another: the patients' understanding of the doctor-patient relationship and the doctors' understanding of that relationship, which together give form to their material transactions, are notably not necessarily at one. To borrow a useful metaphor, although one which can also be misleading, their relationship embodies [58] or may embody an argument about sickness, health, expertise, drugs and many other topics. Each attempts to win this argument in part by casting the other into a role which fits the dramatic forms suited to their own side of the argument. As with the doctor-patient relationship, so with the parent-child, the ruler-ruled, the professor-student and so on. These forms of tension and argument manifest themselves at the level of vocabulary in the essentially contestable character of the concepts which are deployed in such arguments. Examples of essentially contestable concepts include those of art, politics, party, education, disease, natural science. It is of the essence of such concepts that actual or potential debate arises over their application. What counts as a central case exemplifying the concept and what as a marginal case are always open to question. Hence in this area certain key generalizations remain debatable too. If such concepts as those of marriage, tragedy, and education are essentially contestable - as they are - then what are and what are not genuine counter-examples to generalizations about marriage, tragedy, and education will be debatable in a way in which the question of what are or are not genuine counter-examples to generalizations about aminoacids, men over six feet tall, or lugworms, is undebatable. Moreover there is no way of understanding the subject matters to which such concepts apply without employing them. What objectivity requires in the study of such subject matters is an awareness of the contestable and argumentative character of what is going on.

Just because of those characteristics which social practices in general share with arguments of a certain kind - if you like, what I am saying is that arguments of a certain kind are among the paradigmatic social practices - they embody in themselves a certain unpredictability. It is of the nature of debates, arguments, battles and other conflicts that in the centrally important cases their outcome is unpredictable.

Conflict, arguments, contestability, unpredictability; it is striking how much more often these appear in our lives than in our social theories. But there is of course good reason for this. There are historical periods in which these are open and admitted features of social and political life, especially those in which a dramatic move is being made from one fundamental conceptualization of social action to another. Professor W. H. Adkins (Merit and Responsibility) has described how the concepts of Greece's heroic age continued side by side and in tension with those of the polis in fifth-century Athens. In Burnt Njal the concepts of Icelandic paganism strive to retain their hold in a social world already nascently Christian. In such periods rival interpretations of social life manifestly constitute social life. But for many historical episodes this is not so. Rivalry and conflict at the level of conceptualization are more usually latent.

It is a central feature of established social orders that they tend to embody in their social institutions denials of the centrality of conflict, argument, contestability and unpredictability in human life; and it is a [59] central feature of most social theories that they share this characteristic with most social orders. Nor should this parallelism surprise us. On the view that I have taken, social practices are themselves embodied interpretations of social practices. Social theories are unembodied interpretations which attempt to stand in the same relation to society that physical theories stand to nature. There will always be the possibility therefore that instead of mirroring the conflicting interpretations embodied in a practice, instead of achieving the objectivity involved in the acknowledgement of conflict, social theories will, in a way that serves the interests of one of the contending parties, deny by omission the facts of conflict, argument, contestability and unpredictability.

For it is characteristic of the adherents of rival social interpretations embodied in a complex social practice to deny the reality of rivalry in the interest of a claim that there is an incontestable underlying structure; social victory at this deep level is the achievement of inducing those who participate in the practice to agree in conceptualizing their activities in such a way that one of the contestable interpretations no longer appears contestable, but simply how things are - `the facts'. It is when a social theory serves this form of social practice that it functions as an ideology. Or rather, since the word `ideology' is available as nobody's property and for everybody's idiosyncratic meaning, that indicates how I am going to apply the word.

Notice three features of the notion of ideology as I am using it. The first is that in order to function as an ideology a theory must express a partial truth. Radically false theories could not so function. We therefore have to be careful about how we make a distinction between science and ideology. A view that is partial may mark progress in science as well as being available for ideological uses. Indeed since all our views in science are subject to a possible need for revision because of their partial character, as well as for other reasons, ideology ceases to be something that we can merely impute to others, exempting ourselves from the charge. Nonetheless, when a view is functioning ideologically, the mode in which it is held is likely to differ from that in which it is held by those who are treating it as part of science. For it is likely to be held in ways that put it in practice, although not necessarily in principle, beyond refutation.

Secondly, to call a view ideological says nothing about the motives or the intentions of those who advance it. Thus this account of ideology is not reductionist in the way that some theories of ideology have been. The proponents of such theories claimed that certain thinkers merely reflected the social order in their theorizing; but the present claim is that any such reflection is a function of theories, if they are ideological, over and above whatever other characteristics those theories may possess. Notice also that my account is incompatible with any basis-superstructure analysis of social orders.

Thirdly and perhaps most importantly, not only is my use of the concept of ideology different from the uses both of Mannheim and of Marx, it differs also from that of those modern political scientists, such as Robert Lane, who have sought to include ideology in their accounts as one more discrete empirical variable. The reason for this is clear. When we have reached the level at which we are engaged in handling social life as a, collection of discrete empirical variables, then all the work of interpreta tion and conceptualization is already behind us. We are already dealing with a fully interpreted world. One conceptualization rather than another is already presupposed. In the world thus understood, the beliefs held by s individuals or by groups of individuals may well function as just one more variable factor. But the beliefs, the interpretations, which constitute the ordering of the world to be investigated, which make the available range of variables what it is, these underlying constitutive beliefs, con cepts and interpretations will have disappeared from view. With them all too often there will also have disappeared from view argument, conflict, contestability and unpredictability as fundamental phenomena.

There will of course in the constituted investigated world be interpretations, arguments, conflicts and failures in prediction. But these will be understood within a context of regularity and order. They will be secondary phenomena. The notion of order will indeed be regulative for this kind of investigation and various concepts of system and structure will be the ideals to which empirical theory seeks to conform. In so seeking, indeed in its very acceptance of the constituted world as its starting point, empirical theory, both political and social, will indeed be ideological in a just the sense that I have described. Ideology then has two characteristic effects: it works to conceal the features of particular conflicts, of particular contestable concepts and situations, of particular unpredictabilities; and it does this by working to conceal conflict, contestability and unpredictability as such. Ideology is the mask worn both by particular dominant orders and by order itself. But it is also the mask worn by those critics of social orders who equally with its conservative defenders wish to deny any ultimacy to conflict, contestability and unpredictability. Ideology therefore can be a phenom enon of the Left as well as of the Right.

II

I have already suggested that the characteristic procedures sanctioned by , the methodology of empirical social science may be ideological. Yet this' phenomenon suggestion is perhaps a little premature for we need first to recognize that the very existence of methodology as a subject in the social science curriculum is a suspicious circumstance. What is it actually doing in that curriculum? Some writers have understood its task, whether approvingly or disapprovingly, as being that of informing social inquiry with the norms and methods of the natural sciences. But we ought to note at once that neither physicists nor chemists nor biologists study anything called c `methodology'. So that the very inclusion in our curriculum of such a subject itself differentiates our practices from those of natural scientists. I have sometimes therefore been tempted to think of it as essentially a histrionic subject: how to act the part of a natural scientist on the stage of the social sciences with the more technical parts of the discipline functioning as do greasepaint, false beards and costumes in the theater. But of course actors in the theater always know that they are actors and so do their audience; whereas methodologists and their audience at least do not recognize their performances as dramatic.- What then is actually going on? At least three things.

One might be described as the equivalent of the helpful practical hints that you find at the beginning of some cookbooks. Advice on how different ways of wording questions in your survey will elicit different answers from your respondents is an example. Natural scientists have to learn how to see through a microscope and cooks have to learn how to judge the age of an egg. But neither cooks nor natural scientists elevate the acquiring of useful knacks to the status of a subject in their curriculum.

A second set of activities that go under the label `methodology' do indeed belong to a discipline in its own right: applied mathematics and statistics. The last thing that I want to deny is the essential part played by statistics and other branches of mathematics in all inquiry (I myself would be quite prepared to make the passing of examinations in statistics, calculus and computing science a prerequisite for graduate study in any social science and perhaps in the humanities as well). But it is important to understand what statistics can and cannot do for you. Its genuine power includes an ability to perform two central tasks for the social scientist; it can provide, under certain assumptions, a measure of how far an association between two sets of items is not random; and it can in providing such a measure provide evidence that warrants the elimination of certain hypotheses from consideration. Note that both abilities are negative and only negative.

The crucial truth that is often ignored when an insufficiently puritanical and ascetic view of the power of statistics is taken, is that from no set of purely statistical premises can any conclusion of a causal kind be validly derived. This holds even when, as with path analysis, we are able to give a temporal ordering to our variables. And it holds too if we reword it so that it runs: No set of purely statistical premises is adequate evidence for any genuine causal conclusion. This must be the case, for given any correlation between any two sets of items, no matter how strong - it may be perfect, if you wish the truth of the statement which asserts that correlation is equally compatible with the truth of any one out of an infinite set of statements which relate those items causally, and the truth of any one member of that set entails the falsity of all the others. Thus there is a logical gap between all statistical statements and all causal statements; a social scientist who offers statistical evidence for the truth of any genuine causal statement has performed the logical feat of select ing from an infinite set of statements consistent with that evidence just one. The social science journals are full of evidence that the vast majority of social scientists feel able to perform this logical and heuristic feat without ever telling us how they do it. How do they do it? We are fortunate in possessing at least one authoritative type of account in the conventional writings on methodology. I follow Herbert A. Simon's version (`Causation', pp. 352-4).

Simon of course acknowledges the truth that from statistical premises. alone causal conclusions cannot be validly derived in at least two places in his argument. He says this explicitly in distinguishing spurious from genuine correlations (Simon, p. 354); and he has earlier allowed that when we have expressed the relationship between variables in a set of algebraic equations, we have not thereby unambiguously provided a device for genuine causal ordering: `The causal ordering would be altered, of course, if before solving the equations the system is modified by taking linear combinations of them. This process is algebraically admissible . . .' (Simon, p. 353). The general form of Simon's solution to these is the general form, explicit or assumed, given by a number of writers. `Since the causal orderings among variables can be determined only within the context of a scientific theory a complete structure - it is only within such a context that spurious correlation can be distinguished from genuine correlation' (Simon, p. 354). What is `a scientific theory - a complete ' structure', according to Simon? It turns out to be a body of causal knowledge concerning a number of the relevant variables. But this of course at once involves us in circularity. We can only acquire the simpler kinds of causal knowledge if we already possess a more sophisticated kind. This a cannot be right. Technical sophistication has ended in mystification. What Simon has in fact reproduced with admirable clarity is an account of how p, in wellordered, well-established, natural sciences, where we do already ,' possess the requisite structures of theory, certain causal inferences are, made. But this cannot be how they are made at the theoretical, empirical levels at which social scientific investigation moves. So how do social scientists in fact move from statistical premises to causal conclusions?

What provides them with a bridge, so I want to suggest, are not the sophistications of technique, but the presuppositions which must be accepted prior to the application of such technique. Conventional methodology, I am suggesting, already incorporates a highly particular and partial view of the social world, and in inducing the social scientist to adopt this: view the types of connections available for his or her study are greatly: reduced in number. Behavior is abstracted from the realm of conflicting; meanings, which can then be reintroduced afterwards in suitable reduced form as a separate variable or set of variables. The world that is allowed to appear to view is one that conforms to five canons.

First it must be composed of discrete, independently identifiable variables. This is essential for the application of statistical methods. Secondly, these variables must be identifiable in an evaluatively neutral non contestable way. This is essentially for the social scientist to be able to identify which are and which are not the true members of a given population in an undebatable way. Thirdly the conceptualization of the subject matter studied is a task for the social scientist in which what governs his conceptualization is not a relationship to the rival conceptualizations of the society which is being studied, but his own convenience. The doctrine of operational definitions is the crudest version of this canon and embodies a characteristic misrepresentation of the practices of natural science. (On this point see Hilary Putnam `A Philosopher Looks at Quantum Mechanics'.) Fourthly the goal is the construction of law-like generalizations; the model here is not natural science, but a particular philosophical view of natural science exemplified in such writers as Herbert Feigl and C. G. Hempel. (Very often social scientists insist that their own generalizations are probabilistic, when the difficulties in identifying genuine lawlike generalizations in the social sciences are brought to their notice. But this claim is ambiguous. If they mean that they are concerned merely with statistically significant relationships, then they are clearly not doing science at all; but if they are claiming that they either do or aspire to provide genuine law-like probabilistic generalizations - of the kind that occur in statistical mechanics, for example - then it needs to be pointed out that not only is their truth more difficult to establish than that of non-probabilistic law-like generalizations, but also - as we noted earlier that from statistical premises alone no genuine law-like probabilistic generalizations can ever be derived or even importantly confirmed any more than can non-probabilistic ones. Thus restriction to probabilistic conclusions would not alter the nature of the methodological enterprise.)

Fifthly the type of generalization sought is of such a kind that it will afford a lever for producing predictable change in social structures; the relationship of cause to effect and of causal knowledge to causes is such that knowledge is taken to be productive of manipulative ability. This is of course the link between textbook methodology and social-scientific expertise as a commodity available to government and to private corporations.

These canons select certain features of social reality for attention and omit others. They simplify both the world and the selection of hypotheses. They allow for conflict and for ideology only within contexts of defined regularity, that is, for conflict insofar as it is manageable, and for ideology insofar as it is someone else's and not shared by oneself. Inability to manage conflict or to cope with ideology at this level will always be diagnosable as a symptom of lack of manipulative ability, and that in turn as a lack of social scientific knowledge. This has one splendid implication for the funding of social scientific research: for if the application of our present social scientific knowledge is successful, then of course we are warranted in asking for further funding, but if its application is unsuccessful, then this too is precisely a sign that further funding and further research is necessary. Failure vindicates the project as surely as success does; and, given the actual record of applied social science, now prudent this is! But this kind of irrefutability by failure, is of course one of the t characteristic marks of ideology.

Methodology then functions so as to communicate one very particular vision of the social world and one that obscures from view the fundamental levels of conceptualization, conflict, contestability, and unpredictability , as they constitute and operate in that world. It thus has one of the two t= centrally important effects of ideology. Does it have the other? Does it, that is, not only conceal these features of the social world in general but does it conceal them in the interests of one highly specific view of social reality held by individuals with specific social characteristics, so that it more particularly works to conceal conceptualizations, conflicts, contestabilities and unpredictabilities of some one particular character? I shall argue that it does, that social scientific methodology functions as ' one of the ideologies of bureaucratic authority. But to argue this I shall first have to elucidate the character of that authority.

III Methodology and the ideology of bureaucratic authority

Molecules do not read chemistry books; but managers do read books on organization theory. Such books therefore, whatever the intentions of their authors, never merely describe; they provide models for future behavior, and if they become sufficiently successful texts in influential business schools, some original descriptive inadequacies may gradually disappear as organizational behavior conforms more closely to the books which managers read. Unfortunately the effects of their own books are not among the phenomena so far studied by organization theorists. But when we find that widely differing schools of organization theory agree on certain key points, we are perhaps entitled to treat this as a clue to a the more general ideology of managers. Among these points two seem to be of salient importance; they concern control and - not surprisingly - conflict.

Some theorists lay emphasis upon the motives of those over whom authority is exerted (Likert); some lay emphasis upon the premises from which subordinates argue (March, Simon); some assert and some ignore the importance of informal groupings within organizations. But in each case motives, premises or informal groupings are seen as means which a successful manager may employ to achieve ends which he himself, if he is sufficiently high in the hierarchy, has selected. The kind of knowledge which a manager has to have is causal, expressible in generalizations, and must provide him with an essentially manipulative ability. Noting this fact we can begin to understand an almost ritualized process in recent writing on organization theory in which those who begin by repudiating Weber tend to end by reinstating something very like the Weberian view of bureaucratic authority. It is easy to understand both phases of this ritual dance, for it is easy to see that the actual behavior of managers and administrators often does not conform closely to Weber's description of the wielding of bureaucratic authority. Anthony Downs (Inside Bureaucracy), I. M. Destler (Presidents, Bureaucrats and Foreign Policy) and a number of others have made that very clear. So that an initial repudiation of Weber becomes a stock part of the descriptive task. Yet the same evidence shows that when managers are put to the task of justifying their authority and their behavior in authority what is striking is the way in which it is to a large degree that in which they deviate from Weber's account that they feel called upon to explain and to justify. Hence the more ideal either the theorist's or the manager's description of management, the more Weberian. And it is the ideal which operates at the level of justification and which determines the forms of justification. Charles Perrow (Complex Organizations, p. 148) points out this combination of an explicit repudiation of Weber with an unconscious return to Weber in the case of March and Simon; theirs is far from the only case.

Suppose then that we sketch the elements of the ideal of bureaucratic authority; what we find corresponds strikingly to the canons which defined the world view of methodology. First the bureaucrat has to deal in discrete items which can be given an established and unique classification. Anyone and only those who satisfy criteria one to five qualify for welfare; cases of types three to seven must be dealt with by officers of at least grade four; actions involving expenditures of between n and m dollars must be reported under such-and-such headings. Thus the bureaucratic world is composed of discrete variables. Secondly the classificatory scheme which gives rise to, which in an important sense creates those variables, must itself be treated as non-contestable. The scheme has to be accepted independently of the evaluative viewpoint of particular individuals or social groups. Thirdly it is the bureaucrat who is free to create the classificatory scheme; it is he who, so to speak, operationalizes his concepts so that items will be handleable by him in his way.

We thus find that the first three methodological canons which I sketched all have their mirror image in bureaucratic practice. Or rather not so much in actual practice as in that idealized picture of actual practice in terms of which bureaucratic authority holds others and itself responsible. So also are the fourth and fifth canons mirrored in the same way. The relations between already classified items has to be ordered in such a way that the bureaucrat's rules correspond to a set of causal generalizations which warrant predictions: doing A will in fact produce consequences B and C. And these causal generalizations must not only warrant predictions, they must be translatable into recipes for producing effects; they must be levers for effective manipulation.

To justify is precisely to show that one possesses the kind of knowledge that will enable one to perform such manipulations at the appropriate level of the organization. Failure to justify authority will be either failure to impose the relevant classificatory order on the relevant material or to have elaborated the relevant rules and generalizations which will afford prediction, or to have failed to manipulate successfully in order to achieve the goals of the organization.

Two objections might be raised to this account of bureaucratic authority. The first is that, to use André L. Delbecq's terminology (`The Management of Decision-Making within the Firm', pp. 329-31), I have spoken only of that authority exhibited in what he calls routine decision-making, and have ignored what he calls creative decision-making and negotiated decision-making. To this my reply would be that success in the latter two is largely, even if not entirely, measured by the extent to which what formerly belonged in the areas of creativity (the innovative solution of hitherto unsolved problems) or negotiation (bargaining with labor unions, for example) is transferred to the area of routine activity and decision-making (the problem is solved and a formula for future procedure adopted, the strike is settled and a method for avoiding future disputes is laid down).

A second type of objection may be thought more crucial. Egon Bittner has argued (`The Concept of Organization', pp. 239-55) that what the theory of organization embodies are concepts which seem clear-cut, but are not. That of efficiency is perhaps the most notable example. What counts as bureaucratic effectiveness is open to interpretation in a way that Weber himself did not entirely understand. If Bittner is right - and I think that he is - an apparent clarity at certain points, both in organization theory and in administrative practice, is a fictitious clarity, and it might well seem that my account does not recognize this. But what matters for my argument is the form of justification of bureaucratic authority as that is employed both prospectively and retrospectively, both practically and theoretically. If key terms employed in the course of that justification are themselves interestingly opaque, that may be important in other respects - indeed it is - but it does not affect the substance of my present argument.

What I am claiming then is that the canons of bureaucratic authority and those of conventional social scientific methodology are mutually reinforcing. Both mask fundamental conflict; and both are liable to make conflict a more marginal phenomenon than it is, or rather than it would otherwise be. By representing conflict as marginal and manageable by administration, both do indeed make it more manageable. We can indeed see the methodology of the social sciences as ideological in its particular as well as in its more general effects. It is important to recall at this point my thesis that a view can only be effectively ideological in function if it is partially true. There is a corresponding truth that a view can only be effectively ideological if up to a certain point it does open up limited possibilities for social control or social change. But the words to be underlined are `partial' and `limited'. It is the success of ideology that blinds us, not its failures; and it is precisely by them that we must refuse to be deceived. Hence when Johan Galtung (Theory and Methods of Social Research, p. 1) attempts to disarm philosophical criticism of conventional social scientific methodology as a product of `oversophistication' by asserting of methodology that it `works', he misses the point. It works all too well.

IV Postscript

My argument requires one additional point to be made if its import is not to be misunderstood. It moved from conflict to ideology and from ideology to methodology and from thence to bureaucratic authority. Its structure is such that so far conflict must appear to be the basic concept and consequently I may seem to be endorsing a Heraclitean view of social life in which conflict, rivalry and strife are the fundamental features of the social universe. To avoid this misleading appearance another concept would have had to be introduced, one whose absence seriously injures the present account, but one the introduction of which would have seriously complicated the presentation of my argument. The concept needed is that of a social tradition. Weber of course originally introduced the concept of bureaucratic authority precisely in order to contrast it with that of traditional authority. But his account of tradition was so inadequate that the point of the contrast was partly lost.

The most important social conflicts occur within traditions as well as between them. Such conflicts are conflicts between the various incommensurable goods which men within a particular tradition may pursue. A viable tradition is one which holds together conflicting social, political and even metaphysical claims in a creative way. The activities which inform a tradition are always rationally underdetermined; that is, we can specify no set of rules, no set of rational procedures, which are either necessary or sufficient to guide the activity informing the tradition as it proceeds. The complex of institutions in fifth-century Athens, those in medieval kingships, Christianity or Islam, and indeed natural science, are all examples of traditions in the sense intended. It is when traditions begin to break down that modern bureaucratic organizations characteristically arise. Traditional societies have always had formal organizations, but those in authority within those organizations always had to justify themselves against appeals to the authority of the tradition which the organization served. The Pope has never been able to avoid appeals against his own judgments to the doctrines of the Christian religion; physics itself is always a source of objections to members of the scientific hierarchy. But of course traditions cannot exist without embodying and re-embodying themselves in organizational forms, and organizations always do breed managers. Nonetheless, at most what the managers can control is the organization and not the tradition. Hence conflict in traditional societies survives all attempts at conflict management. The relation of tragic drama to the politics of the assembly in Athenian political life, the quarrel be tween royal law and church law in twelfth-century Europe, and the debates between the new and the old astronomy in the seventeenth century are paradigmatic examples. Hence also the much greater danger from ideology in modern formal organizations where traditions are always endangered.

The recognition of tradition therefore is a necessary key to understanding the relation of conflict, ideology and organization in political life. Unhappily that recognition has been fatally hindered by two facts. The first is the way in which only conservative writers on politics - a Burke or a Michael Oakeshott - have tended to take tradition seriously. The second is the fact that a prerequisite for understanding the role of tradition - and therefore the role of conflict - is a training in depth in philosophy and in history, just those subjects so often excluded from the political science curriculum in the name of the behavioralist revolution. What that revolution replaced them with was of course the study of methodology. Insofar as that study was simply the study of statistics and applied mathematics, who could but applaud its introduction? The American Voter will stand, if not perhaps forever, at least for a very long time as its memorial. But it still remains true that graduate students in political science for the most part know very little more mathematics than they did before. What the introduction of the study of methodology actually did was to drive out those very inquiries which could have revealed its ideological function. By its place in the curriculum, methodology has achieved yet another ideological dimension.