1. Epistemological Philosophy of History

Epistemological philosophy of history has four sources. It arose from: 1) the rejection of German historism; 2) the rejection of speculative philosophies of history; 3) the attempt to offer a satisfactory reconstruction of historical explanation, based on the premises of the covering-law model (CLM); and 4) different forms of Collingwoodian hermeneutics. The epistemological nature of these four pillars of traditional Anglo-Saxon philosophy of history will be obvious to everybody. Historism and speculative systems were rejected because it was thought that they did not satisfy the epistemological criteria for historical knowledge. The CLM and Collingwoodian hermeneutics, on the other hand, attempted to discover the nature of these epistemological criteria. In the remainder of this section, I shall discuss each of these four components of the epistemological tradition and complete the picture with an assessment of their strengths and weaknesses.

Except for F. H. Bradley's The Presuppositions of Critical History (1874),^[2] it might be argued that Anglo-Saxon philosophy of history as we know it

[2] See W. H. Walsh, "Bradley and Critical History," in A. Manser and G. Stock, eds., The Philosophy of F. H. Bradley , Oxford, 1984.

today begins with M. Mandelbaum's The Problem of Historical Knowledge (1938). Here, Mandelbaum even steals a march on Collingwood, although the latter had, of course, been doing a great deal of work in the field since the 1920s. The significance of Mandelbaum's first work has often been overlooked. It seems likely that the conclusions Mandelbaum reached there left indelible marks on the epistemological tradition. At the time when Mandelbaum wrote his book, German historism had drifted into the so-called "crisis of historism."^[3] With his famous but usually misunderstood dictum that it is the historian's task "not to pass judgment on the past, nor to teach lessons for future use, but only to show how the past has actually been," Ranke had urged historians to consider the past only from the perspective of the past itself. An ethical relativism confusing the (time-bound) popularity of ethical norms with their (time-independent) applicability was mistakenly inferred from Ranke's injunction. Thus, when Mandelbaum found German historism in its state of self-inflicted destruction, the picture he drew of it, not surprisingly, did little to recommend historism to Anglo-Saxon philosophers. Historism became synonymous with a poor and obscure response to the challenges of ethical relativism.

The net result has been that Anglo-Saxon philosophy of history had from the very beginning isolated itself from one-and-a-half centuries of profound and penetrating thinking about the writing of history. This is even more regrettable because historism was not only the fountainhead of a sizable part of all historiography produced since the beginning of the last century but also because it possessed an awareness of the practice of history so conspicuously lacking in Anglo-Saxon philosophy of history. Owing to the intellectual disorientation in Germany after the Hitler period, German philosophers and historians—with a few exceptions, such as J. Rüsen, T. Nipperdey, or H. Lübbe^[4] —felt little inclination to formulate a modern and self-assured defense of historism. Georg Iggers's book—so very well-informed and erudite—codified the communis opinio that historism had been a regrettable phase in philosophy of history which now fortunately belonged to the same past it had always studied in such an erroneous and dangerous way.^[5]

It is characteristic of their almost contemptuous dismissal of German

[3] Mandelbaum described the attempts made by Simmel, Rickert, Scheler, and Troeltsch to counter relativism as a "set of failures": M. Mandelbaum, The Problem of Historical Knowledge , New York, 1938, 174. For a comprehensive German statement of the problem, see K. Heussi, Die Krisis des Historismus , Tübingen, 1932.

[4] J. Rfisen, Für eine erneuerte Historik Zur Theorie der Geschichtswissenschaft , Stuttgart-Bad Canstatt, 1976; T. Nipperdey, "Historismus und Historismuskritik heute," in Nipperdey, Die Funktion der Geschichte in unserer Zeit , Stuttgart, 1975, 82-95; H. Lübbe, Geschichtsbegriff und Geschichtsinteresse , Basel/Stuttgart, 1977.

[5] G. G. Iggers, The German Conception of History [1968], Middletown, 1984.

historism that Anglo-Saxon philosophers of history—otherwise so sensitive to fine terminological distinctions—never even bothered to make a clear distinction between historism and what Popper called historicism.^[6] There is a strange story about Popper's rejection of historicism. He obviously had in mind what Walsh was to define a few years later as speculative philosophies of history.^[7] What Popper criticized was mainly the pretension on the part of some speculative philosophies to predict the future by extrapolating from the past to the future in one way or another. Since historians are usually interested in the past and not in the future, Popper's criticism did not succeed in presenting speculative philosophies as an illegitimate form of what historians legitimately try to do. Not only did the historicists' claim that they could interpret the past in a superior way survive Popper's onslaught relatively unscathed, but it has even been shown by B. T. Wilkins in his detailed analysis of the last chapter of Popper's The Open Society and Its Enemies that Popper actually believed speculative systems to function in historiography as "searchlight theories" and that they are therefore indispensable for all historical interpretation.^[8] This idea was to be elaborated on with vigor and perspicacity by Fain and Munz.

Another strategy in the attack on speculative systems has been to accuse them of being metaphysical. Speculative systems, it was argued, cannot be tested in the way "ordinary" historical interpretations of the past can be tested. Marx's claim that all history is the history of the class struggle is as unverifiable as its equally metaphysical counterpart that all history is the history of class cooperation. However, one can agree with Walsh that both speculative systems and "ordinary" historiography attempt to define "the essence" of part of the past and therefore cannot be distinguished one from the other by means of criteria which distinguish metaphysical claims to knowledge from verifiable ones.^[9] Once again, though philosophers tried to reject speculative systems, they could not find conclusive arguments against them.

It is therefore not surprising that the failure to discredit speculative systems effectively formed one of the first cracks in epistemological philosophy of history. As early as 1972—when the CLM still reigned supreme in Anglo-Saxon philosophy of history—Fain made an ingenious attack on the Humean notion of causality underlying most of the arguments in favor of the CLM.^[10] He pointed out that, contrary to Hume's theory of causality, in historiography the relation between what is called a cause and what is

[6] See the introduction to K. R. Popper, The Poverty of Historicism , London, 1957.

[7] W. H. Walsh, An Introduction to Philosophy of History [1951], London, 1967, 16.

[8] B. T. Wilkins, Has History Any Meaning ? Hassocks, 1978.

[9] See W. H. Walsh, Metaphysics , London, 1963, 172ff.

[10] H. Fain, Between Philosophy and History , Princeton, 1970.

called its consequence is not external but internal. Apart from purely formal conditions, history also has its material requirements regarding what is to be counted as a cause. Something that fits in the CLM is often not considered by historians to be an acceptable cause. According to Fain, speculative systems define these material requirements. They identify in the historical past a number of layers of historical events and phenomena having the same ontological nature; and having the same ontological nature makes events causally relatable. It did not become clear from Fain's book, however, whether the guidance provided by speculative systems on our journey through the past should be seen as an addition to the Humean causal model or as a replacement for it. In a very readable book, Munz has developed ideas very similar to those of Fain, although he strove quite explicitly for a reconciliation between speculative philosophy and the CLM.^[11] The final outcome of the debate has been that we look at speculative systems in the way we look at extramarital sex: it is practiced by many, is supposed to be both natural and exciting, but is nevertheless not exactly according to the proper rules.

This, however, has not been the central issue. The debate in epistemological philosophy of history has always been dominated by the controversy between the adherents of the CLM and the defenders of the legacy left by Collingwood. In the course of my exposition, it will become clear that, contrary to appearances, the two parties have more in common than they have separating them. It is ironic that the origins of the debate, as well as its justification, are found outside philosophy of history proper. This will become clear if we imagine a list of academic disciplines, arranged according to the ease with which they fit the positivist scientific model (I use the term positivist here in a general, untechnical sense). At the top of the list we shall find (theoretical) physics, then chemistry, biology, geology, the social sciences (starting with economics), and—finally at the end of the list—we come to history. The general background to the debate between the CLM advocates and the Collingwoodians has always been the question as to whether—from a methodological point of view—there is a point, as one moves down the list, at which things really become quite different. In other words, it was not historiography, per se, but the thesis of the unity of science that was the real issue in the debate. Not surprisingly, philosophers of a positivist bent who accepted this thesis found in history a marvelous challenge to their ingenuity. It was believed that if the scientific nature of historiography could even be demonstrated (by declaring one CLM-variant or another valid for historiography), the positivist's claim as to the unity of all scientific and rational inquiry would have been substantiated. Consequently, a great number of philosophers, most

[11] P. Munz, The Shapes of Time , Middletown, 1977.

of them interested in history not so much for its own sake but because of its quality as a peculiar fringe area, pounced upon the problem of historical explanation in the attempt to adapt it to the requirements of the CLM.

Strangely enough, even from the point of view of positivist philosophy of science, the battleground for the controversy had been chosen in such a way that the philosophical significance of the debate could never be more than marginal. Philosophers of science, whether they were neopositivists, adherents of Popper, of Kuhn, or of whatever other philosophical denomination one might think of, were never interested in explanation as such but in theory and concept formation. Nevertheless, during the CLM debate it was rarely, if ever, asked whether being in conformity with the CLM would in itself be sufficient to elevate history to the status of a science; nor was it asked whether something analogous to theory and concept formation might not also be found in historiography. Raising the latter question would have advanced the birth of the narrativist tradition in philosophy of history by some twenty years. No doubt, the fact that the controversy between the CLM proponents and the Collingwoodians naturally centered on the not very illuminating problem of the existence of covering laws prevented a more timely "takeoff" of the narrativist approach.

Nevertheless, the CLM debate has been well worthwhile, if only because its apparent lack of resolution made philosophers of history aware of a number of unsuspected characteristics of historiography. There were even positive results. At the time of the debate—and this was surely no coincidence—both historians and philosophers of history^[12] advocated a rapprochement between history and the social sciences. Suggestions like those of Joynt and Rescher^[13] that history should be seen as a kind of applied science and the historian as a "consumer" rather than as a "producer" of socioscientific laws placed the CLM in an optimal position to mediate between history and the social sciences. Conrad and Meyer's famous article in 1957 on the relation between economic theory and economic history^[l4] —generally regarded as having triggered the New Economic History—is a striking illustration of the fruitfulness of the CLM for

[12] Most influential has been D. S. Landes and C. Tilly, History as Social Science , London, 1973; the relation between history and the social sciences has become the most hotly debated topic in German philosophy of history.

[13] C. B. Joynt and N. Rescher, "The Problem of Uniqueness in History," History and Theory 1 0960): 158; and in G. H. Nadel, ed., Studies in the Philosophy of History , New York, 1965, 7. The locus classicus of the CLM is, of course, C. G. Hempel, "The Function of General Laws in History," Journal of Philosophy 39 (1942). For an exposition of the statistical variant of the CLM, see C. G. Hempel, Aspects of Scientific Explanation , New York, 1968, 380ff.

[14] A. H. Conrad and J. R. Meyer, "Economic Theory, Statistical Inference, and Economic History," in A. J. Conrad and J. R. Meyer, eds., The Economics of Slavery , Chicago, 1964, 3-30.

actual historical practice. Some form of CLM is clearly presupposed by both counterfactual analysis and model building in economic history.^[15]

Within the epistemological tradition, the CLM has been attacked from both the "inside" and the "outside." CLM defenders themselves were quick to recognize that there was little in actual historical practice that was in accordance with the requirements of the CLM. Moreover, it proved depressingly difficult to produce a historical law which was both valid and interesting. To meet this problem, a number of statistical-inductive variants of the original nomothetic-deductive CLM were developed. But, even then, difficulties remained. It could be argued that M. Scriven's and M. White's proposal to reduce the role of covering laws to a mere justification of the historian's choice of a specific event as a cause, instead of that of a general premise in a deductive argument, has been the most successful strategy in the history of the CLM and its subsequent metamorphoses in refuting the charge of empty schematism and inapplicability.^[16]

But most of the objections to the CLM came from the disciples Collingwood won some twenty years after his premature death. Henceforth, when referring to this tradition, I shall use the term analytical hermeneutics, which has been suggested by F. Olafson. A short terminological digression is in order here. It is useful to distinguish between a German (or continental) hermenutical tradition, from Schleiermacher to Gadamer or Derrida—and beyond—and Anglo-Saxon hermeneutics, from Collingwood on.^[17] The former has as its paradigm the interpretation of texts (preferably biblical, juridical, or literary), and the latter the explanation of intentional human action. It must be emphasized that the aims of these two forms of hermeneutics are quite different: German hermeneutics tends to see the past (that is, the text) as something given and urges us to take a step back, as it were, in order to find out about its significance; Anglo-Saxon hermeneutics moves in exactly the opposite direction, by urging us to try to discover new historical data (that is, the intentions behind human action). German hermeneutics wants us to choose a vantage point outside or above the past itself; Anglo-Saxon hermeneutics requires us to penetrate ever deeper into the past. Characteristically, German hermeneutics—especially Gadamer^[18] —is largely indifferent to the so-called mens auctoris,

[15] P. D. McClelland, Causal Explanation and Model Building in History, Economics, and the New Economic History , Ithaca, 1975; M. G. Murphey, Our Knowledge of the Historical Past , Indianapolis, 1973.

[16] M. Scriven, "Truisms as the Grounds for Historical Explanations," in P. Gardiner, ed., Theories of History, new York, 1959; M. White, Foundations of Historical Knowledge , New York, 1965, chap. 3.

[17] I elaborated on this distinction in my Denken over geschiedenis; Een overzicht van moderne geschiedfilosofische denkbeelden , Groningen, 1984.

[18] H. G. Gadamer, Truth and Method , trans. G. Barden and J. Cumming, New York, 1986.

whereas "analytical hermeneutics" has no other objective than to reconstruct it. German hermeneutics shares with the narrativist tradition—to be dealt with in the next section—a synthetic approach to the past; Anglo-Saxon hermeneutics is openly analytical—a fact which may justify Olafson's choice of terminology. German or continental hermeneutics has deeply influenced today's literary criticism and, via literary criticism, has recently found its way into the narrativist tradition within Anglo-Saxon philosophy of history.

The epistemological nature of analytical hermeneutics is particularly pronounced. As has been demonstrated by Van der Dussen in his dissertation and by Meiland in an admirable little book,^[19] Collingwood's reenactment theory was originally an answer to the epistemological question as to how historical knowledge is possible (in a nutshell, the answer can be summed up as follows: historical knowledge is possible because by reenacting the thoughts of the historical agent, these thoughts are brought into the present and can then be investigated here and now). The same is still true of Dray's action rationale explanation, since this model is supposed to define which epistemological criteria have to be fulfilled before we are allowed to say, "I now have the explanation as to why x did a. "

Although analytical hermeneutics went through a difficult period in the 1950s, a series of monographs on Collingwood, written in the 1960s by Donagan, Mink, and Rubinoff,^[20] rapidly tipped the balance between the CLM and analytical hermeneutics in favor of the latter. Analytical hermeneutics underwent a number of transformations in the course of time. Collingwood's still rather crude reenactment theory gave way to Dray's rationale explanation, which was to be refined, in its turn, by the intentional explanation and the so-called "logical connection argument" (LCA), which will be described later on. The practical inference to be reconstructed by the historian was analyzed with an ever-increasing sophistication. However, most philosophers of history nowadays agree that further refinement of the scheme of practical inference will inexorably be subject to the law of diminishing returns. That may explain why some philosophers of history have recently become attracted to Collingwood's as-yet-undeveloped logic of question and answer^[21] —a pronounced contextualist theory of history quite irreconcilable with the propositionalism of his earlier reenactment theory.^[22]

[19] J. W. Meiland, Scepticism and Historical Knowledge , New York, 1965, chap. 3; W. J. van der Dussen, History as a Science: The Philosophy of R. G. Collingwood , The Hague, 1981, 157ff.

[20] A. Donagan, The Later Philosophy of R. G. Collingwood , Oxford, 1962; L. O. Mink, Mind, History and Dialectic: The Philosophy of R. G. Collingwood , Bloomington, 1969; L. Rubinoff, Collingwood and the Reform of Metaphysics: A Study in the Philosophy of Mind , Toronto, 1970.

[21] R. G. Collingwood, An Autobiography [1939], Oxford, 1970, chap. 5.

[22] See my "De Angelsaksische hermeneutiek en de deschiedbeoefening," in T. de Boer, ed., De filosofie van de mens—eh cultuurwetenschappen , Meppel, 1989.

The debate between the advocates of the CLM and the analytical hermeneuticists was hampered by the unexpected difficulty in identifying what it was that was at stake in the controversy. A notable exception was P. Skagestad, who in a brilliant book^[23] succeeded in translating the controversy (with Popper and Collingwood as the main antagonists) into an ontological issue. If Popper's third world (containing the thoughts of historical agents) ought to be stratified into an object-level and a metalevel, the CLM is to be preferred; if not, analytical hermeneutics is preferable. Relying upon Russell's theory of descriptions, Skagestad opted for the latter alternative.^[24] Usually, however, the issue was not stated so clearly. When hermeneuticists argued that they did not apply laws (since their explanation was based exclusively on the ascertainment of a fact, that is, what " I " would have done under certain historical circumstances), and CLM proponents answered that such an explanation always presupposed a covering law (namely that all rational persons would do what I believe I would do myself under such circumstances), the debate tended to degenerate into a rather fruitless controversy about the priority of the context of justification versus the context of discovery— to put it in Reichenbach's terms.^[25]

Dray's influential first book^[26] is a striking illustration of how difficult it apparently was to state with clarity the nature of the disagreement between the CLM and analytical hermeneutics. It has been noted by several commentators that Dray's criticism of the CLM and his defense of his action rationale explanation formed entirely different strands in his argument. It was as if Dray first had to transform himself into a reluctant advocate of the CLM before he was able to criticize the model so effectively. And in a later article of Dray's, the same division is even more pronounced.^[27] The net result of this course of events was, of course, that the CLM found itself in a relatively secure position. Its supporters could decide where the battle with their opponents was to be fought, and as long as the model did not succumb to disagreements among its own adherents, all criticism would, in practice, amount to a refinement and not a rejection of the model.

In a later phase of the debate, the logical connection argument (LCA) provided analytical hermeneutics with a better argument to prove its independence from the CLM. Following suggestions made by Wittgenstein in

[23] P. Skagestad, Making Sense of History: The Philosophies of Popper and Collingwood , Oslo, 1975.

[24] I expressed my reservations about Skagestad's argument in my "Een nieuwe syn-these?" Theoretische geschiedenis 6 (1979): 58-91.

[25] R. H. Weingartner, "The Quarrel about Historical Explanation," in R. H. Nash, ed., Ideas of History , vol. 2. New York, 1969.

[26] W. H. Dray, Laws and Explanation in History , Oxford, 1957.

[27] W. H. Dray, "The Historical Explanation of Actions Reconsidered," in P. Gardiner, ed., The Philosophy of History , Oxford, 1974.

his Philosophical Investigations (such as his "the human body is the best picture of the human soul"^[28] ), the LCA replaced the causal relation between motives and actions with a logical one. With one simple stroke the CLM, with its causal—and not logical—covering laws, had been expelled from the domain of the explanation of human actions. In order to prove the LCA, Donagan wrote that if an agent has the intention I and knows that action a may realize I , and action a is still not carried out, we shall have to conclude that the agent never seriously intended I . In other words, it is part of the meaning of having an intention that the relevant action will be carried out. As may be clear from this admittedly imperfect rendering of Donagan's version of the LCA, the LCA in its initial formulation seemed to achieve the union between intention and action almost by a feat of "magic."^[29] Later defenders of the LCA tried to remedy this. G. H. Von Wright thus argued that the antecedens and the consequence in a practical inference of the form (1) A intends to bring about p ; 2) A considers that he cannot bring about p unless he does a ; 3) therefore A resolves to do a )^[30] are analytical, since it is impossible to verify the consequence without verifying the antecedens, and vice versa. The deficiencies in Von Wright's argument were convincingly exposed in Rex Martin's Historical Explanation.^[31] Martin's book, hitherto unsurpassed in the development of the possibilities inherent in analytical hermeneutics, has up to now not received the attention it deserves.^[32] Martin's thesis was that the LCA is not a logical rule but a regulative rule, like the rule that every event has its cause. The function of such rules is to make a certain kind of inquiry epistemologically possible.

But one may wonder whether all this makes much of a difference. Whether human actions are explained by means of covering laws, the LCA, a regulative rule, or the general rule that all rational people are disposed to act rationally, a general rule is required in all cases. We therefore have no reason to be very greatly impressed by the deviation from the CLM as proposed by the LCA and others. It is nice, of course, that the LCA reconciled historical explanation with the Wittgensteinian and Rylean condemnation of causal "ghosts in the machine," but that hardly had anything to do with the original disagreement between proponents of the CLM and of analytical hermeneutics.

[28] L. Wittgenstein, Philosophical Investigations [1953], Oxford, 1974, 1780.

[29] T. Kuipers, "The Logic of Intentional Explanation," in J. Hintikka and E Vandamme, eds., The Logic of Discourse and the Logic of Scientific Discovery , Dordrecht, 1986.

[30] G. H. Von Wright, Explanation and Understanding , London, 1971, 96.

[31] R. Martin, Historical Explanation: Re-enactment and Practical Inference , Ithaca, 1977, 174-175.

[32] A. Ryan's review in History and Theory 19 (1980): 93-100, failed to do justice to the book.

Besides, these later phases of analytical hermeneutics could even be seen as open or covert flirtations with the CLM. The original gap between Collingwood and Hempel is much wider than the later one between Von Wright or Martin and Scriven or, for example, between Murphey and other more recent defenders of the CLM. The present state of affairs in the debate should be seen as a movement toward a convergence or synthesis of the CLM and analytical hermeneutics rather than as the victory of the latter over the former. For example, within Von Wright's version of the LCA, the dividing wall between the two has become as thin as the dubious irreducibility of intentional descriptions of human actions to causalistic or physicalist descriptions of them. When Von Wright discusses the event of someone ringing a doorbell, this supposedly "irreducible" intentionalist component in an intentional description of that event is so forced and debatable that one may come to feel that even this thin dividing wall has collapsed already.

Martin's book forms an even more telling example of the convergence of the CLM and analytical hermeneutics. He divides the antecedens of the practical inference into a number of separate premises, roughly: 1) the agent finds himself in situation S , in which he wants to bring about some change; 2) certain alternatives to that end present themselves; 3) the realization of intention I seems to the agent to be the best option; 4) the agent believes that doing a will realize I ; and 5) the agent has no conflicting intentions and is physically capable of performing a.^[33] First, it should be noted that, in contrast to previous definitions of the practical inference, Martin is able to explain, thanks to premises l, 2, and 3, why S gives rise to intention I in the mind of the agent. He thus avoids that vicious circle between intention and action which reduced the resorting to intentions in all the previous definitions to a role reminiscent of Wittgenstein's wheel in the machine that is driven without driving anything itself. Second, this elaboration of the scheme of the practical inference permits Martin to claim a new role and status for the CLM; for it will be the task of covering laws to connect the premises of the antecedens. Take, for example, Caesar after his conquest of Gaul. We can conceive of a general law to the effect that generals in similar situations—that is, when they are confronted by incursions on the part of a neighboring country—consider alternative ways of changing this unsatisfactory situation (thus linking premises 1 and 2), another stating that generals will usually decide that such incursions must be stopped (the link between 2 and 3), and still another one claiming that generals usually conclude that carrying out an invasion of the neighboring country will be the best solution (the link

[33] Martin, Explanation , 78-79.

between 3 and 4) and so on. The CLM has thus been quietly absorbed into analytical hermeneutics.

We can establish that the debate between the CLM advocates and the analytical hermeneuticists has always been moving toward synthesis more than toward perpetuation of the disagreement. From the vantage point of the present, it is better to speak of "peaceful coexistence" between the two approaches than of an open war between them. Therefore, in the current phase of the debate in philosophy of history, it will be the similarities rather than the differences between the CLM and analytical hermeneutics that will strike us. The following five points sum up these similarities. When taken together, they define the most general presuppositions of epistemological philosophy of history.

First, both the CLM and analytical hermeneutics were relatively insensitive to the problems of actual historiographical practice. Beyond the New Economic History, historians did not have much reason to be interested in covering laws (or their application), and the explanation of the actions of individual historical agents studied in analytical hermeneutics is only a negligible part—and certainly not the most interesting part—of the historian's task. In fact, the adherents of both the CLM and of analytical hermeneutics looked at historiography from a viewpoint outside historiography itself. The theory of the CLM reads like a lecture on applied logic or science, and analytical hermeneutics like a chapter in a book on the philosophy of action.

Second, both are primarily, if not exclusively, concerned with historical explanation. The historian's universe, as seen through the eyes of both, looks very much like a piece of white paper speckled with an immense number of little dots, while it is the historian's explanatory task to connect these dots with one another as well as he or she can. But, that the historian's task is essentially interpretative—that is, to discover a pattern in the dots—had now been lost sight of. Precisely because of its epistemological concern with tying the historian's language as closely as possible to the past itself, philosophy of history was never able to spread its wings and to become a philosophy of historical interpretation.

Consequently, both the CLM and analytical hermeneutics focused their attention on the details and not on the totality of historical studies. The historian has to establish and explain individual facts and was therefore conceived of essentially as a kind of detective, as Collingwood said.^[34] Perhaps Collingwood's experience as an archaeologist (he was certainly not a historian in the proper sense of the word) goes a long way in explaining his preoccupation with the problem of why people did, made, or thought

[34] R. G. Collingwood, The Idea of History [1946], Oxford, 1970, 266ff.

certain things in the past; and it is undoubtedly true that his reenactment theory is well suited to the problem of how to study the artifacts from a remote past which has left no written tradition.

However, anybody even superficially acquainted with historiography will recognize that the explanation and description of individual historical facts form only a very minor part of what historians do. We admire great historians like Ranke, Tocqueville, Burckhardt, Huizinga, Meinecke, or Braudel, not for the accuracy of their descriptions and explanations of historical states of affairs, but for the panoramic interpretations they have offered of large parts of the past. Whichever way one tries to overcome the limitations of the CLM and of analytical hermeneutics, the scope of epistemological philosophy of history will invariably prove too narrow to account for such narrative interpretations of the past.

Third, in both its manifestations, the epistemological tradition demonstrates a lack of a sense of history that is quite astonishing for a philosophy of history. It seems to accept either tacitly or openly Hume's famous statement "that there is a great uniformity among the actions of men, in all nations and ages, and that human nature remains the same in its principles and operations."^[35] This insensitivity to historical change manifests itself in the CLM in the generality of the covering laws it uses, whereas analytical hermeneutics by necessity presupposes a similarity between the historian's thought and the thought and action of the historical agent studied by the historian.

Fourth, in neither of its guises—either the covering law model or Collingwoodian hermeneutics—has epistemological philosophy of history ever succeeded in its hope of bridging the gap between the historian's language and historical reality. The CLM failed in this respect because, for a variety of reasons, explanans and explanandum never matched in a satisfactory way. It is true that Danto has done much to narrow the gap between the two by pointing out that we always explain events under a certain description of them and that one of the historian's most fascinating tasks is therefore to describe the past in such a way that we can feed those descriptions into the machinery of the covering laws we have at our disposal. But even Danto had to admit that whatever success the historian may have on this score, an appreciable distance will always remain between the past in all its complexity and explanatory language.^[36]

A similar criticism can be leveled at analytical hermeneutics, but this kind of criticism is considerably more interesting. Analytical hermeneutics has been accused of not being able to account for those aspects of the past

[35] D. Hume, Enquiries Concerning the Human Understanding and Concerning the Principles of Morals , Oxford, 1972, 83.

[36] A. C. Danto, Analytical Philosophy of History , Cambridge, 1968, 220ff.

that cannot be reduced to the (conscious or unconscious) intentions of individual human agents. Take, for example, the stock market crash of 1929. Since it was nobody's intention to become poorer, the crash cannot be explained in terms of the intentions of the speculators involved. Most adherents of analytical hermeneutics have accepted this serious limitation to their theory with a certain equanimity.^[37] Von Wright is therefore an exception when he tries to refute the criticism that analytical hermeneutics is powerless when it comes to the unintended consequences of intentional human action. He takes as his example the origin of the First World War. According to Von Wright, each step taken by the Serbian, Austrian, German, or Russian government was the reaction to a previous step and can be explained by means of intentional explanation, by taking into account what diplomatic situation arose after each previous step. In this way there is nothing left that might give substance to the thesis of the unintended consequences of intentional action.^[38] Von Wright's argument can be countered as follows. Number all the successive practical deliberations of the several governments involved up to the outbreak of war: P_l . . . P_n . What, then, was the cause of the outbreak of war? Historians will rarely select P_n as the most likely candidate; they will prefer to say that each step in the series P_l . . . P_n contributed to the outbreak of war and was, therefore, part of the cause. Consequently, P_l , for example, was part of the cause, even though this practical deliberation did not intend to bring about the war. The language of the unintended consequences of intentional human action thus appears to be an essential part of the historian's language.

It is necessary to emphasize the following. Von Wright was correct insofar as his argument showed that only people and not superhuman forces make history, but he was wrong to infer from this that the historian's explanatory potential is exhausted with the appeal to intentional human action. The language of history permits the historian to see the past from a perspective different from that of the historical agents themselves, and it is purely and solely this change in perspective that gives rise to the thesis of the unintended consequences of intentional human action. This thesis is therefore not an ontological claim (the past contains both intentional actions and their unintended consequences) but a thesis concerning the autonomy of the historian's language with regard to the intentional actions of historical agents. As soon as it is conceded that the historian is not committed to the agent's point of view, the language of the unintended consequences can and will be used.^[39] In other words, analytical hermeneutics was bewitched by the epistemological dream of a complete parallelism be-

[37] Dray, Laws, 119; Martin, Explanation , 15.

[38] Von Wright, Explanation , 139ff.

[39] See chapter 3.

tween the historian's language (intentional explanations) and what was seen as the actual past (the practical deliberations of historical agents), and this dream seemed so real that it made philosophers of history completely blind to the realities of the writing of history. However, history is often shown or interpreted in terms of what has no demonstrable counterpart in the actual past itself. Thus neither the CLM nor analytical hermeneutics succeeded in achieving the epistemological goal of tying language to the world, of words to things. The CLM failed because historical reality proved to be too complex, and analytical hermeneutics failed because of its inability to account for the complexities of the historian's language. Obviously, the failure of analytical hermeneutics is more serious than that of the CLM. The latter can at least be transformed into a program for future historical research, whereas the failure of analytical hermeneutics is a failure to explain what historians have been doing already for several centuries.

Fifth, there is the epistemological nature itself of both the CLM and analytical hermeneutics. Here we discover an assumption which is so ubiquitously present, which seems so obvious and so innocuous, that it has hardly ever been paid any attention. According to this assumption, we can and should in all cases distinguish clearly among the following three levels: 1) that of the past itself; 2) that of the historical language we use for speaking about the past; and 3) that of philosophical reflection on how historians arrived at their conclusions and how these conclusions can be formally justified. Historical language is, to borrow Rorty's metaphor, the mirror of the past, and it is the essentially epistemological task of the philosopher of history to analyze how this mirror succeeds in showing us the past.

It is true that this scheme has always shown some cracks. For example, the troublesome problem of speculative philosophies seemed to blur the distinction between levels 2) and 3). In addition, historians were sometimes concerned about terms like continuity, discontinuity, order, or chaos. Obviously, the terms themselves belong to level 2); however, one may wonder whether they are only conceptual instruments for organizing our knowledge of the past or whether they also refer to aspects of the actual past. This insoluble problem suggested that the line of demarcation between the first two levels was not as clear as epistemological philosophy of history had always liked to believe. However, these problems—if recognized at all—went unheeded, like Kuhn's anomalies that are "set aside for a future generation with more developed tools."^[40] It was only after the publication of Hayden White's Metahistory that these "anomalies" were to take on a new significance.

[40] T. Kuhn, The Structure of Scientific Revolutions , Chicago, 1970, 84.