Version classiqueVersion mobile
OpenEdition Books

Understanding History

 | 
Jonathan Gorman

Chapter 3. Reality

Texte intégral

1The three theoretical questions drawn from the between historians at the disagreement Philadelphia conference were these: Can historians provide knowledge about reality? What would be a proper justification for the cliometric approach to history? Is answering the question “was slavery profitable?” a way of achieving knowledge, so that, once answered, the issue is disposed of once and for all? We are currently dealing with the first of these questions. We argued in the last chapter that knowledge claims need to be rationally justified, and we conditionally defended the view that we should rely on empiricism as our criterion of rational support.

  • 1 One does not have to adopt an empiricist justification of science. See note 14.

2We also remarked in the last chapter that it is irrational to appeal, with the positivist, to physics and similar sciences to set the standard for knowledge unless one has reason to think that they do succeed in achieving knowledge for themselves. Still, it is widely believed that the natural sciences do achieve knowledge, and that they display this achievement through precise explanation and successful prediction, which are warranted on the basis of an independent empiricist standard.1 Historians, in seeking knowledge, also seek to explain, and it was the supposed “scientific” nature of the cliometric approach which was at issue for many historians at the Philadelphia conference, as described in chapter 1. There is an empiricist theory of scientific explanation and prediction which has been developed particularly for history, and this will be introduced and examined in this chapter.

  • 2 So called by W. H. Dray, in Laws and Explanation in History, Oxford University Press, 1957, p. 1.
  • 3 C. G. Hempel, “The Function of General Laws in History,” Journal of Philosophy vol. 39, 1942, p. 3 (...)

3The theory is known as “covering law theory,”2 and was first expressed for history by Carl G. Hempel in 1942 in his article “The Function of General Laws in History.”3

  • 4 Hempel in Gardiner, op. cit. p. 345.
  • 5 Hempel in Gardiner, op. cit. p. 347, note 1. There are theories of causation different from Hempel (...)

4Hempel claims that the explanation of an event of some specific kind E consists in indicating its causes, and, further, that “the assertion that a set of events — say, of the kinds C1, C2,..., Cn — have caused the event to be explained, amounts to the statement that, according to certain general laws, a set of events of the kinds mentioned is regularly accompanied by an event of kind E.”4 There is, he makes clear,5 no difference between displaying the cause of some event and showing that universal scientific laws cover that event. A scientific law thus expresses a regularity which occurs in the world, and an event is “covered” by that law in so far as its occurrence can be deduced from that law and from statements which assert the occurrence of the “causes.”

5Minimally, then, an explanation of E has to have the following form:

Whenever C1... Cn then E;
C1... Cn;
Therefore, E.

  • 6 This is also M. G. White’s position. See his Foundations of Historical Knowledge, New York: Harper (...)

6The statements in this explanation, this deductive argument, must be empirically warranted; and this necessity expresses Hempel’s empiricism. On Hempel’s view, anything that goes into historical writing which falls short of this model of explanation or which goes beyond it is in principle just so much waffle, and is to be rejected as a contribution to knowledge. It cannot be enough to say “historians explain differently” or some such, as we shall shortly see; in any event, historians frequently offer the “causes” of things. Nevertheless, Hempel recognizes that historical writing hardly ever takes the form of a valid deductive argument, and, rather than hold that the historian is too silly to make his position properly explicit, he regards most historical writing as being a form of rhetoric which is appropriate for sketching to an uninformed reader the bones of knowledge which are buried within.6

7Hempel’s covering law theory of explanation is merely asserted in his 1942 article. But it is not derived from an arbitrary approval of science; it would not be worth consideration if it had been. Empiricism, we remarked in chapter 2, has been perhaps most famously expressed by David Hume, and it is Hume’s understanding of empiricism which lies behind Hempel’s theory of explanation and which plausibly drives us towards it. This is why one cannot simply assert, in response to Hempel, that historians explain differently, for, if Hume’s empiricism is right, we have little alternative but to adopt a standard of explanation which embodies that empiricism, and Hempel’s is an elementary way of expressing such a standard.

8Empiricism says that knowledge comes only from experience, and in chapter 2 we briefly characterized the various senses through which we experience. But what, exactly, are we given in experience? Experience is experience of something, and the full expression of the empiricist position will require a general account of the world we know by this means.

  • 7 David Hume, A Treatise of Human Nature, Selby-Bigge ed., Oxford: Clarendon Press, 1888, Bk. 1, Pt. (...)

9Hume’s empiricism was stronger than that which we have so far presented: experience, for him, was the sole source, not only of knowledge, but of all our beliefs and ideas and imaginings. Our experiences he called “impressions” and the thoughts and imaginings about them he called “ideas.” His empiricism was therefore expressed in the following way: all ideas come only from impressions.7

10Hume made a claim about impressions and ideas which he thought was directly revealed by experience. He split impressions into “simple” and “complex,” and the ideas that come from them he split in exactly the same way. Simple impressions or ideas are such that one cannot break them down any further, and Hume had in mind a particular colour, taste or smell.

  • 8 For a strong argument by a monist empiricist see F. H. Bradley, The Principles of Logic, Oxford Un (...)

11This claim about the nature of experienced reality we may call atomism. Atomism (and there are other irrelevant uses of this word, which should be ignored) is the view that many things exist, rather than just one thing. It may seem obvious that many things exist, but the opposite position — monism — is plausible if you reflect that the many everyday objects and occurrences of which we are aware may not exist in total independence from each other, but exist rather as parts of some larger whole. It would be a mistake, in the monist’s view, to consider individual events or objects in isolation from their place in the world. By contrast, an atomist like Hume considers that there are many things, ultimately simple things, which are what they are independently of whatever else exists, and which may be combined with each other in order to make more complex things. But, whichever view you take, you should note that experience does not just give us atoms.8 Atomism is an a priori way of characterizing experience.

12Notice that “atoms,” in the sense a physicist might give to that word, are not what are involved here. The fundamental particles with which the physicist deals (if they exist at all) may not exist independently of each other. Compare being red in colour with occupying an area. It is impossible to conceive being red as existing independently of occupying some area, and it is impossible to conceive an area existing independently of having some colour, bland or polychromatic though it may be. Again, can one be a king independently of whatever else exists? Obviously not; subject citizens must also exist. Maybe the physicist’s atoms are somewhat like this, too.

13There are, then, many things, according to this Humean view. Furthermore, they are causally related. The usual supposition here is stronger than the claim that some causes exist, and different from the claim that every event has a cause, and it is this: that the only relation that exists between all the things that exist throughout space and time is a causal relation. The universe consists of many things in causal relationship with each other, and this is all there is to it. The people the historian deals with are part of this universe, of course. This strong causal claim we may call mechanism.

14The view lying behind the covering law theory of explanation thus involves the three elements now identified: empiricism, atomism and mechanism. The view is that the world consists of many things in causal relationship with each other, a world which we are supposed to know about on the basis of experience.

15As empiricists, we want to know what causation is. We seem to have an idea of it: we talk and think of one thing causing another, and so forth. Given Hume’s empiricism, we know that this idea of causation must have its source in experience.

16There are two routes by which ideas come from impressions. First, by direct “copying,” which is true for all simple ideas and their associated impressions. “Copying” is an unclear notion, but it is intended to express, for example, the relationship between the “idea” conveyed in the statement “This pen is blue” and the impression or experience consisting in the fact that this pen is blue. This works not only for statements: thus the idea “blue” which we have is a copy (we can imagine blue perfectly well) of the impression of blue, which we have when we see a blue thing.

17The second route by which ideas derive from impressions is through the imagination. In imagination we operate with ideas which have already been given to us from experience, and we can add them together to make more complex ideas. We can also split up complex ideas into their simple parts. Complex ideas may or may not represent complex impressions.

18So far as the idea of causation is concerned, there must therefore either be some impression of which it is a copy, or else it must be built up by the imagination out of ideas which do derive directly from experience. All we need do is attend to the relevant impressions or experiences, and we will then know all there is to know about the idea of causation and how we got it.

19Suppose that our idea of causation is a direct copy of an impression. All that we will need to do, then, is examine our experiences of those cases called causation and see which feature they have in common. Thus, suppose that I drop the blue pen on my desk and noise results. It was my dropping it that caused the noise. This is one case of causation. Again, consider the blackboard that is somehow caused to remain attached to the wall of my room, rather than drop to the floor as it would otherwise do. This is a second case of causation. Both cases alike being cases of causation, there must be some feature which they have in common, some feature the experience of which gives rise to the idea of causation (given the supposition in the first sentence of this paragraph).

20And yet a little reflection makes plain that there is no relevant feature that they have in common. At the level of ideas, we characterize both of the examples as cases of causation, but at the level of impressions they have nothing relevant in common. It follows at once that our idea of causation cannot be a direct copy of some feature of the world. It must therefore be a complex idea built up by the imagination, for this is the only alternative. Note that it does not follow from this that the idea of causation is “merely imaginary”: the idea of causation may be a proper summary of a number of simpler ideas which are well founded in experience and which are associated with each other for very good reason. Our next step, therefore, is to follow Hume and break up — analyse — the complex idea of causation into its simpler parts. The empiricist’s job here will only be complete when, firstly, he has shown to our satisfaction what the simpler parts are, and, secondly, he has demonstrated how these parts are connected with experience in a direct way.

21Hume argued that the complex idea of causation involves three elements. The first is that of contact between the things that are causally related. If we are to say correctly that C causes E, then it must be the case (it is necessary) that C and E are in contact. More, however, is required (it is not sufficient), and we shall come to that extra in a moment. Is Hume right in his claim here? Certainly it is easy for him if he is right, for the idea of contact is one which we can readily derive from experience. No doubt there are many cases where one thing is in contact with another and we regard causation as in some way operating through the contact: say I throw a brick at a window and it causes it to break. But is contact necessary: is causation impossible without it? Is “action at a distance” (an expression familiar in Hume’s time) impossible? Surely not; we believe that the sun causes the flowers to grow, and magnets can pull things, without there being necessarily any contact. There might be contact, of course: physicists might discover little streams of particles making contact across an apparent gap. But it is not for philosophers to say that physicists must discover this. We do not know such things a priori: the relationship between cause and contact is not a matter of one idea containing another.

22“Contact” has here been used in its normal sense of spatial, physical contact. But Hume also means contact in time: there is no temporal gap possible between the operation of the cause and the operation of the effect. This claim directs our attention to a feature of our causal talk: it is not objects that cause other objects, it is, as Hempel insisted, rather events that cause other events. We do not say that bricks cause broken windows: rather, that it is the throwing of bricks that causes the breaking of windows. Throwings are not objects, but events. Earlier we noted the supposition that the world consists of many things causally related. The “things” here, given that they are causally related, have to be events and not objects. And yet this does not seem to be the whole story, for my blackboard is still attached to the wall of my room, and whatever causes it to remain there is continually operating. It seems inappropriate to call this operation an “event,” and it seems more appropriate to call it a “state of affairs,” although by this time we may be uncertain exactly what “event” means.

23What is causally connected with what now becomes a puzzling matter, but it is a matter of some importance for history. We may say that every event has a cause, but is the Second World War an event? Whatever it is, is it the kind of entity that can have a cause, and what kind of entity could count as its cause? Is the world of the historian really one in which “causes” — whatever they are — operate? It is not easy to make sense of the notion of contact in time (let alone contact in space) in the case of history. Often philosophers use the word “conditions” to refer to causally effective states of affairs which operate over a long term. Thus European nationalism may have been a part of the cause of the First World War, even though it was a condition which existed for a long period of time beforehand.

24Contact in time and space is therefore not a wholly intelligible notion. Our difficulties with it derive in part from serious problems with our everyday notions of time and space. Consider a problem raised by the classical Greek philosopher Zeno. Imagine Achilles and the Tortoise having a race. Achilles beforehand, knowing how fast he can run, agrees to give the Tortoise a start of some yards. They both start at the same time. In order to catch the Tortoise, Achilles has first to reach the point from which the Tortoise started. After a short time he reaches it. In the time he takes to get there, the Tortoise moves a further short distance. Achilles has next to make up this distance. In the time he takes to do so the Tortoise moves a further short distance. Achilles has next to make up this distance. In the time he takes to do so the Tortoise moves a further short distance. Achilles has next to make up this distance. In the time he takes to do so the Tortoise moves a further short distance, and so on ad infinitum. Thus Achilles can never catch the Tortoise.

  • 9 Advanced study of causation may usefully begin with E. Sosa ed., Causation and Conditionals, Oxfor (...)

25And yet we know perfectly well that Achilles does catch the Tortoise. This paradoxical result shows that our argument must be wrong, but where is the mistake? We rely on nothing more than our everyday notions of space, time and motion in achieving the unacceptable result. Zeno used this sort of paradox as part of an argument for the unreality of space, time and motion. Certainly he was right to do this in a way: space, time and motion as ordinarily conceived are in some way unreal, which is to say that we must have to some extent at least simplistic and incorrect conceptions of them. Plainly space and time do not “add up” to infinity in the way that we might ordinarily suppose. Maybe our understanding of causation is as bad. Or worse. Considerations of this sort enter into an advanced theory of causation.9

26The second element which Hume found to be a part of our complex idea of causation was the idea of temporal priority. He took the view that, when we say that one thing causes another, we mean, among other things, that the cause comes before the effect in time. Put briefly, temporal priority is necessary (although not sufficient) for causation. Just as in the case of contact, it is easy for Hume here if he is right, for temporal priority is just the kind of thing that we can readily experience. But is he right? Again, there are difficulties. There are two ways in which Hume might be wrong: firstly, if causes can come after their effects in time; and secondly, if causes and their effects can be contemporaneous.

  • 10 This is A. C. Danto’s question, See his Analytical Philosophy of History, Cambridge University Pre (...)

27It does seem odd to suppose that effects could come before their causes in time, but is it impossible? Could you, for example, travel back in time to the ante-bellum American South? (Would historians be better placed than they are if they could do this sort of thing?)10 On your way back to that time, whatever was causing you to move would be causing you to move backwards in time, and thus the effect of the time machine’s operation would be temporally prior to the cause. It is not clear whether this is an impossible idea or not, but the answer is by no means obvious, and it follows from this that it is by no means obvious that temporal priority is necessary for causation. If this is so, then we have no good reason to suppose that temporal priority is an essential part of our idea of causation.

28Remember that Hume will also be wrong if causes can be contemporaneous with their effects. Such a case is readily available: my blackboard continues to be attached to the wall of my room, and whatever is causing it to remain there is still operating, and obviously operating at the same time as the effect it is bringing about. It does seem, therefore, that Hume’s claims about some of the parts of the idea of causation are mistaken, even though we may recognize that, in many cases, contact and temporal priority may happen to be involved. In summary, it is not the case that contact or temporal priority are necessary for causation.

  • 11 David Hume, A Treatise of Human Nature, Selby-Bigge ed., Oxford University Press, 1888, p. 173, ru (...)

29Hume, however, pre-empts the problem here by regarding the third element of the idea of causation as by far the most important. “Tis chiefly this quality,” he said, “that constitutes the relation.”11 This third element of the idea of causation he claimed to be the idea of “necessary connection.” When one thing causes another to happen, it is “necessarily connected” with it, it makes it happen. It is not just that the effect happens coincidentally following upon the cause.

30This seems very plausible — more plausible than the suggestions about contact and temporal priority — but it raises a different kind of difficulty for this empiricist position: from what impression is the idea of necessary connection derived? We have already looked at two cases of causation (the dropped pen and the blackboard) and found that they have no feature in common. Just as they have no feature in common to support the idea of causation — no feature which the idea of causation can “copy” — so they have no feature in common to support any other idea, including the idea of necessary connection.

31There is no point looking further, and a quite general conclusion becomes clear to us. It is that, while experience may inform us that something does occur, it cannot inform us that it must. Consider two otherwise identical experiences, differing only in the respect that, while the first just happens, the second happens necessarily. How different are the experiences? There is no difference at all, and there could not be any difference. Since we are supposing that experience is all we have to go on, it follows that there is no such thing as necessary connection. And yet we have the idea of it! Do we give up Hume’s empiricism?

32The empiricist cannot admit that necessity is a part of the world, and thus he must not so much explain our idea of necessary connection, as explain it away. This Hume does in a most valuable move. The idea of necessary connection, he says, is nothing more than a strong expectation on the part of the mind for the event we call the effect, given the event we call the cause. The feeling we have that the window just has to break, given the brick travelling towards it (which is the “idea” of necessity), is nothing more than the strong expectation that the window will break. What has to be explained, then, is where the mind gets this expectation from.

33The answer Hume gives is that, out of mere habit, the mind moves from the idea of the event called the cause to the idea of the event called the effect. When the cause occurs, the mind then comes to expect the effect. What has now to be explained is the habitual movement of the mind from the one idea to the other. This is a matter of habitual experience: in our experience, the one impression (we call it the cause) is “constantly conjoined” with the other impression we call the effect. This tells us that the essence of causation, its central feature, is that of constant conjunction, which is the regular association of things we regard as causally related. Causal talk is nothing more than a way of expressing the regularity which occurs in nature. When we search for the causes of things we are searching for the regularities of experience. There is nothing else to it, as Hempel stressed.

  • 12 Ibid., rule 4.

34How, exactly, should we express this “regularity” or “constant conjunction”? If C and E are constantly conjoined, then we may wish to say this: that, whenever you have C, then you have E; and, whenever you have E, then you have C.12 The second part of this, however, is mistaken. It may be that whenever you throw a brick at a window it will break, but it is not the case that whenever a window is broken, a brick must have been thrown at it. We allow that there is more than one way of causing a window to break. Constant conjunction or regularity is properly expressed in statements of the form “Whenever C, then E.” These statements express a claim about the world, and are true insofar as the world is that way. Such statements are therefore synthetic and not analytic.

35“Whenever C then E,” therefore, is the way to express a regularity of nature, and our explanations of what happens must make explicit reference to such a statement. In consequence, we arrive at Hempel’s position that an explanation of E must look like this, and be warranted on the basis of experience:

Whenever C1... Cn then E;
C1...Cn;
Therefore, E.

36Can such an explanation be warranted on the basis of experience? We have seen that one non-empirical and a priori assumption has already been made in arriving at the present position: that of atomism.

37Regrettably, there is a crucial ambiguity in statements of the form “Whenever C, then E,” an ambiguity which raises a further problem for this kind of empiricism. “Whenever” refers to time, to all times, but what does “all” mean here? All the times in our experience? Let us try this and see where it leads. If it means only all the times in our experience, then, when one says “The cause of cancer is virus X operating in conditions Y,” then one means this: “Whenever I observed virus X operating in conditions Y, cancer followed.” One means nothing more. One is not saying that if you now have virus X operating in conditions Y, then you are justified in expecting cancer. One is not saying that there is any point in looking for virus X operating in conditions Y, for one is not suggesting or even hinting that the regular association which one has observed between virus X in conditions Y and cancer is a regularity which exists beyond one’s experience. One’s meaning is limited to one’s past experience alone. But why waste money on research into the causes of cancer if all that one is looking for is a summary of past experience, with no implications for the future whatever?

38Obviously this is wrong. When we look for the causes of cancer we are looking for things which, if they occur in the future, will bring about cancer. The point is to know in advance of the occurrence of cancer that certain conditions will bring it about. You only have the cause, so far as our ordinary understanding of this notion is concerned, when you are able to have some element of prediction. This is a part of what “cause” means. We will only allow “whenever C then E” to ground causation (as Hempel wishes it to do) if the “whenever” in it refers, not merely to all the times in our experience, but to all the possible times there are, so that we would know, for example, that if virus X operating in conditions Y occurred, then cancer would occur. Causation means nothing less.

39The crucial ambiguity in “whenever C then E” has now been explained, and we have unambiguously chosen the sense where “whenever” covers all possible times, and not just all the times in our experience. This sense of “whenever” is often called the “universal” sense, and it is contrasted with the weaker “accidental” sense. The regularity involved in our causal talk involves the universal sense, and Hempel made clear that it was “universal” laws that operated in his theory of explanation. But this is at a cost to the empiricist approach. Hume had two problems: first, he had to break up the complex idea of causation into the parts of which it is made, and, second, he had to show how those simpler parts derived directly from experience. Hume’s first two suggested parts of the idea of causation, contact and temporal priority, were clearly not essential parts of the idea of causation, although, given them, we could recognize their connection with experience well enough. The third idea, necessary connection, suffers from the opposite defect.

40It is perhaps plausible enough to hold that the chief feature of the causal link is regularity, but how does that idea derive from experience? It should now be clear to you that, if the accidental sense of “whenever C then E” is involved (where “whenever” means “at all the times in our experience”), then the empiricist has no difficulty; but, since it is the universal sense which is involved (where “whenever” means “at all possible times”), then the empiricist faces a severe problem. It is open to us to ask Hume this: from where do we get our idea of a possible but unexperienced event?

41Our and Hempel’s problem, however, is not quite Hume’s, that of finding out what experience our “idea” of a possible but unexperienced event derives from. Unlike Hume, we and Hempel can hold this idea innate, if we wish; our problem is rather how we can come to know that a universal statement — “whenever C then E” — is true, when we have only experience to go on. If we cannot know on the basis of experience that such statements are true, then we cannot know of the regularities of association which (such statements claim) characterize the reality which experience “gives” us. And then we have no empirical reason to believe that the world is a regularly ordered place, and we are left with no account of what that reality is which experience is experience of, and left with no account of how one thing can be the cause of another. We don’t appear to have been given anything by experience here.

42How, then, in principle, might one justify a claim to know that the world is a causally ordered place, that certain statements of the universal form “whenever C then E” are true, when one has only experience to go on? Experience, at best, gives us merely “whenever C then E” in the accidental sense. On the empiricist approach, the only method of certain justification, in default of experience, is deductive reasoning.

43It is logic which sets for us the standards of deductive reasoning, and is the ultimate rational check upon what we say in our arguments. It is appropriate here to make some remarks about logical reasoning.

44People often argue. Commonly their interest in arguing is to persuade their opponent of some point of view. If one is particularly charismatic then one may be able to persuade people easily, but one does not need to be charismatic to do this: one could, for example, just threaten. Years ago a child might be beaten into being persuaded to believe that 2 + 2 = 4, but 2 + 2 does equal 4 regardless of what steps are taken to make one believe it, The best reason for believing it is that it is true, but an appeal to reason is often not an effective means of persuasion. A rational person, however, is one who is persuaded by means of reason.

45One rational way of demonstrating that something is true, we have seen, is by pointing to that experienced fact which makes it true, but this is not a kind of argument, and logic is essentially concerned with rational arguments. When an argument is correct we say that it is “valid,” and when it is incorrect we say that it is “invalid.” An argument is split up into “premises” and “conclusion.”

46The premises of an argument are those beliefs that are assumed to be true for the purposes of the argument. The conclusion is what logically follows from them. “Following logically from” is a powerful notion: if one accepts the premises of a valid argument, and yet denies the conclusion, then one is contradicting oneself. To contradict oneself is to state both that something is, and is not, the case. Note that an argument might be valid even though the premises are false. Thus the following argument consists of nothing but false statements, but is valid:

Charles de Gaulle was Japanese;
All Japanese are tigers;
Therefore, Charles de Gaulle was a tiger.

47This is valid because, if one accepts the premises to be true (which one may do for the “sake of the argument”), then one is forced to accept that the conclusion is true. Otherwise one is contradicting oneself. The following is an invalid argument consisting of nothing but true statements:

Charles de Gaulle was French;
2 + 2 = 4;
Therefore, all French are mammals.

48Given the premises, one is not driven to accept the conclusion. One would not be contradicting oneself in holding the premises true and the conclusion false; one would simply be wrong. (Construct for yourself an example of a valid argument with true premises and conclusion.)

  • 13 For a non-empiricist introduction to logic, see R. Bradley and N. Swartz, Possible Worlds, Oxford: (...)

49Note that it is contradiction which lies behind the power of a valid deductive argument. If the premises are true, then the conclusion must be true, and this “if-then” statement is analytic for a valid deductive argument. If one is an empiricist, then this “must be true” of deductive necessity will be interpreted as all other analytic statements are interpreted: as known a priori but empty of all knowledge content. The use of deductive reasoning, by an empiricist, does not commit him to rationalism by a hidden route. It follows, on the empiricist view, that there can never be more meaning in the conclusion of a valid deductive argument than could be found in the premises.13

50If, in default of experience, we use logic as our standard for justifying the claim to know that the world is a regular place, then our problem becomes, in effect, how it is possible for us to deduce the universal statement “whenever C then E” from the accidental statement “whenever C then E.” No argument can be valid if there is more meaning in the conclusion than there is in the premises, and it follows at once that it is impossible to deduce the universal from the accidental here, impossible to deduce causation from experience.

51And yet, you may say, in spite of this supposed impossibility, and however irrelevant universal statements may be to historical writing, it remains the case that there are universal statements which we know to be true — for example, Newton’s first law of motion, which states: if a body is not acted upon by some externally impressed force, then it will continue in a state of rest or of uniform motion in a straight line. Obviously, Newton was not able to deduce, in a logically valid argument, from the limited range of his own experience, a conclusion which covers us today, some three hundred years later. And yet he did it all the same, it appears!

  • 14 There is a wealth of other assumptions which might be made about the nature of science. See, to be (...)

52The problem of how you can deduce universal statements from the limited range of experiences we have is called the “problem of induction.” “Induction” is the old name for what scientists do in reasoning for their theories, and scientific theories, if we assume the atomistic and mechanistic assumptions made so far,14 are taken to consist primarily of statements in the universal sense of “whenever C then E,” statements which say something true about reality and are thus synthetic. The problem of induction exists precisely because scientific reasoning conceived in this way just could not be valid, which amounts to saying that it just could not be rational, if logical deduction from experience is our standard of rationality.

53Some philosophers used to suggest that we ought not to judge inductive reasoning by the standards of deductive reasoning, but it is not so easy as that. Deductive reasoning is not a game which you can choose whether to play or not: only the principles of logic, specifying deductively valid arguments, give you reason to believe with certainty the conclusion of an argument with true premises, for only then would you be contradicting yourself in denying the conclusion. Precisely because inductive reasoning is invalid deductively, it follows that, for all that your premises are true, for all that experience gives you, your conclusion may still be wrong. It is perfectly consistent with any argument or experience which Newton could have provided that his first law of motion would have been found to be wrong the moment he died; or, indeed, at any time before or since.

  • 15 For an elementary account of probabilistic explanation see C. G. Hempel, Philosophy of Natural Sci (...)
  • 16 See J. L. Mackie, Truth, Probability and Paradox, Oxford University Press, 1973.

54Do not think, as some do, that you can avoid the problem of induction by specifying only “probabilities.”15 It is doubtful if we know quite what we mean by saying this, and there are many kinds of probabilities,16 but consider the straightforward case: whenever C, then probably E. It does not matter how big or small the probability of E, given C, is — 1 per cent or 99 per cent — there is still an ambiguity in “whenever.” The probabilistic statement here is useless unless it is interpreted in the universal sense: at all possible times, if C occurs, then there is a probability of so much that E will occur. It still covers all possible times, however weakly those times are described, and this is necessarily more than experience can support: the problem of induction includes statistical arguments. In the accidental sense of “whenever,” of course, why rely on something as weak as probabilities? You experienced whatever relationships you experienced in their entirety, so you can be exact to 100 per cent. An accidental generalization is a mere summary of those experiences, no more than a list.

55It is plain, given empiricism, that we need valid deduction from experience if we are to believe a conclusion (given true premises) with certainty, for the possibility of doubt is only removed when it is contradictory to suppose the conclusion false. May we not have, though, “reason to believe” something which falls short of absolute certainty? Does “reasonableness” — that rationality which in chapter 2 we noted to be the foundation of knowledge — require that we accept something only when there is no possible alternative? May we not reasonably believe something when there is no actual alternative to it, and if we have no reason to think it wrong?

  • 17 Karl Popper, The Logic of Scientific Discovery, London: Hutchinson, 1968, ch. 1.

56One such alteration of the standard of “reason to believe,” and thus the standard of knowledge, was made by Sir Karl Popper in his widely influential contributions to the empiricist philosophy of science and of history.17 Popper noted an asymmetry about the logic, of the universal statement “whenever C then E.” No matter how many cases of C being conjoined with E one may observe, one is never in a position to conclude with certainty that “whenever C then E” is true. But one needs only one case of C occurring without E to conclude with absolute certainty that “whenever C then E” is wrong. To put it simply, we cannot prove or verify a universal statement, but it is logically possible to disprove or falsify it. Furthermore, falsifications are just the kind of thing we can experience: just as it is possible to experience individual cases of C and E conjoined, so it is possible to experience individual cases of C without E.

57Therefore, although we cannot be empirically certain that a universal statement is true, we might come to be empirically certain that it was false. Does not this come to the same thing, you may ask? Does not finding out that a statement is not false (they are not all false, after all) amount to finding out that it is true? These questions involve a misunderstanding. Certainly we may allow that “not false” equals “true.” But “not finding a statement to be false” does not mean “finding a statement to be not false.” If we do find a case of “C and not E”, then we know that “whenever C then E” is false. But if we do not find such a case, then it tells us nothing at all about this universal statement, not even that we have no reason to disbelieve it. This last requires a further assumption, as we shall see.

58Suppose you have a favourite theory involving some central claim that whenever C then E; for example, “whenever you have full employment you will have inflation.” You cannot prove that your theory is true, but then neither can your rival theorists verify their claims, so you are no worse off. All that you need to do is avoid finding out that your theory is false. There is one easy way to do this: keep your eyes shut. Obviously this will not do. You would only have some reason to believe “whenever C then E,” according to Popper, if you had tried hard to discover cases of C without E, but without success. The harder you had tried, the more reasonable it would be to believe that whenever C then E. Of course, this is a risky procedure for your favourite theory, since you might succeed in finding a case of C without E, and you would then know that your theory was false. Still, it is worth the risk, since an untested theory is worthless, Popper’s line of argument, however, assumes that a tested theory is worth something: that one has reason to believe a theory if one has no reason to disbelieve it. One places the burden of disproof upon experience, so that a theory is to be accepted unless experience shows it to be wrong. But why is the burden of proof this way round? The only reason for this is that, if the burden of proof is placed upon the theorist himself, it is too heavy a burden for him to bear, since the problem of induction prevents him from successfully carrying it. It is an arbitrary move if, in an argument between two people in which the burden of proof is on one of them, the one simply shifts the burden to the other!

  • 18 See essays by M. Black and P. Achinstein in P. H. Nidditch ed., The Philosophy of Science, Oxford (...)

59This shift in the burden of proof is warranted only if a further assumption is made, namely that reality consists of regularly ordered events and conditions, such that we can suppose our assertions of regularity correct except where we find to the contrary. This assumption, which underlies Popper’s work, is reminiscent of an older “solution” to the problem of induction, which assumed that the future resembles the past. This assumption was often thought to make good the validity of inductive arguments. There are some technicalities about whether, as an assumption, it could make induction work,18 but even if this assumption did make induction work it would not get us anywhere unless it was true. The empiricist has as difficult a problem as ever as to why we should believe a priori assumptions about reality like these. The mere fact that past futures have resembled past pasts (did they? how do you know?) is no guarantee that future futures will resemble future pasts.

60In spite of this, some empiricists have felt strangely comfortable in relying upon the empirically unwarranted assumption of similarity between past and future. But the problem for them is worse than merely one of worrying whether the future resembles the past; as we have seen, the problem is rather whether possible cases resemble actual ones, and the future-past relationship is merely a simple version of this, at best. The empiricist finds that he has to assume what he is supposed to be proving experience to give, namely, that the world consists of regularly ordered events and conditions.

  • 19 Notice that it need not be denied that some theory connects “whenever C then E” with other scienti (...)

61We have to know that this assumption of regularity is true, if the burden of proof is to be shifted as Popper wishes. Only then may universal statements which are successfully tested be admitted to knowledge. But even if, as Popper admits, scientific theories are only provisional in their acceptance (and thus perhaps are not to be admitted to knowledge), it remains the case that the regularity assumption has to be made, even to allow provisionally well-founded belief. This is because, without this assumption, the experienced non-occurrence of a case of C without E says nothing whatever, not even with some small probability, about the truth of “whenever C then E.”19 We find, therefore, that, just as the a priori non-empirical assumption of atomism was made on this Humean empiricist approach which Popper and Hempel have adopted, so the a priori non-empirical assumption of mechanism (interpreted as expressing regularity) has also to be made.

  • 20 Hempel in Gardiner, op. cit., p. 346.

62But a priori non-empirical assumptions do not end here. Hempel stresses that the “C” and “E” in a scientific explanation mark kinds of things.20 We can see why this is so if we suppose that, as historians, we wish to explain the occurrence of the American Civil War. This could be taken to be the in the conclusion of the following covering law argument:

Whenever C then E;
C;
Therefore, E.

63However, this “E” also occurs in the universal statement “whenever C then E,” but how can one talk of “whenever” an event of 1861-65, which only occurred once? “American Civil War” is a proper name for a specific instance (using ordinary English elementary grammar), whereas the “C” and have to be the sort of thing that can recur, so that the “whenever” in the universal statement can operate properly. Notice that the words “civil war” will do, for they form not a proper name but a description: they refer to kinds of thing, and kinds of thing can recur again and again.

64Let us see this in an example of an explanation which uses a historical “law.” This law is surely false but it is suitable for illustration.

Whenever a revolution occurs, a civil war occurs thereafter;
The American Revolution was a revolution;
Therefore, a civil war occurred after the American
Revolution.

65It is not empty to say that the American Revolution was a revolution. “Revolution,” in its use in the last sentence, is part of a proper name, and is not necessarily a correct description. Was the Hundred Years’War a war? Did it last for a hundred years? The situation has to be made explicit, and the second premise of the above argument does this. It says what kind of thing the American Revolution was.

66This example helps us to spell out more exactly the form of a covering law argument. The outline is:

Whenever C then E;
C;
Therefore, E.

67and the more exact form is:

  • 21 The argument form here is rightly ambiguous in that the event of kind E may or may not be the occu (...)

Whenever there is an event of kind C then there is an event of kind E;
The occurring event X is an event of kind C;
Therefore, there is an occurring event of kind E.21

68In summary, we have seen that the covering law theory of explanation requires us to spell out the application of our mechanistic assumption to the experienced world in the form of a deductive argument. The first premise of that argument is a universal statement, in the exact form: “Whenever there is an event of kind C then there is an event of kind E.” This expresses the constant conjunction of kinds of thing, and it is held true on the assumption that the world is a regular association of kinds of thing, and that the absence of regularity has, in spite of testing, not been found for these kinds C and E.

69The second premise of the argument is the statement: “The occurring event X is an event of kind C.” A particular experience is here being characterized as being an experience of a particular kind. It is just such a characterization of experience which is required for testing a universal statement, as mentioned in the last paragraph. Thus, to say that “the absence of regularity has, in spite of testing, not been found for these kinds C and E” is to say that no actual experience of events is characterizable as an experience of something of kind C occurring without something of kind E occurring. Both in testing the first premise, therefore, and in asserting the second premise, one is required to rely on being able to state truly that some particular event is also a particular kind of event.

70In being committed to the existence of kinds, the covering law theory is committed to no more than is any other approach to history. Although the route to this commitment lies through a particular empiricist interpretation of the beliefs in mechanism and atomism, it is a commitment which is shared by all who speak, regardless of metaphysics. The use of language is always more than the mere naming of things, and includes describing things, the application to them of kind-words.

71What is it to say that a particular kind exists? Here there are serious metaphysical disagreements. On Plato’s rationalist view, for a kind to exist is for a single object to exist, one which appears in a number of different places at a number of different times. The kind “revolution” (if it is a kind) is a single real object with a number of scattered appearances throughout history. (Plato called such objects “Forms,” and they have also traditionally been called “universals.”) But the empiricist will not accept this suggestion. For him, we must rely upon experience so far as we possibly can, and at best we experience the particular “appearances” of a kind separately, and not any contrasting single reality. He has to reject the Platonic view that we can know by reason alone that there are these special entities, and he similarly rejects the view that we can know by reason alone which they are.

  • 22 B. Russell, An Inquiry into Meaning and Truth, Harmondsworth: Penguin, 1962, p. 22. Hume regarded (...)

72Consider the word “dog,” said of dogs, which form a kind of thing, and about which universal statements may be made. “What we really have,” said the empiricist Bertrand Russell, “is a number of more or less similar noises which are all applicable to a number of more or less similar quadrupeds.”22 Being a kind is then understood to be some sort of organisation of similarities between experiences, and being committed to the existence of kinds is being committed to the existence of such similarities.

73While we might, in a single experience, be faced with a number of similar objects, and apply the word “dog” to them all indiscriminately to mark our recognition of their similarity in the relevant respects, this experience only teaches us the word “dog” as applied to itself. “Dog” means more than this: it applies in principle to an infinite number of possible objects, and not just to those actually experienced. This is an essential feature of descriptive words and thus of all talk. It is a particularly obvious point for the covering law theorist, for the “whenever” in “whenever C then E” could not be used if it were not so. Thus, when the empiricist is committed to the existence of kinds, he is committed to the existence of similarities between possible objects — that is, to the existence of similarities between objects not yet experienced. That kinds exist is not a view that can be derived from experience, but also which kinds exist cannot be derived from experience either, it follows.

74Just like the universal statement, the assertion that a kind exists is a claim that goes beyond that which experience alone can warrant. It follows that empirical verifiability of the claim is impossible. One might think that we could here make an analogous move to that made by Popper in his “solution” to the problem of induction, and look for an appropriate sort of falsifiability. This, however, is impossible. The analogy would require that experience could show two objects or events of claimed similarity to be dissimilar after all, but there is always a similarity between any two things, even if only a similarity in respect of being experienced by us.

  • 23 A simple introduction to these issues is given in P. G. Winch, The Idea of a Social Science and it (...)

75Would it be wrong to say, either that mountains are, or that they are not, like triangles? Or that some mountains are more like triangles than they are like other mountains? What kinds are involved here?23 Do the mountains which are like triangles, and other triangles, all form a single kind? Plainly it is necessary to identify the respect in which similarity is being judged here: shape of outline, say, or massiveness. But “shape of outline” and “massiveness” here form yet further kinds, which again are not given by experience.

  • 24 This is at the basis of N. Goodman’s so-called “new riddle of induction.” See his Fact, Fiction an (...)

76Experience alone does not determine what kinds there are, and there are two choices open to us: either reality is organized into fixed kinds which we know on an unexperienced a priori basis (we can express this differently by saying that experience is of really existing kinds), which is much like Plato’s position; or we can recognize that there is a certain freedom as to which respects we pick out as at issue in our judgements of similarity.24

77There is little doubt that we have innate standards for judgements of similarity for observable matters, but this does not cover matters of theory, of that which goes beyond observational experience. The historical past, of course, lies beyond observational experience. There is, in a developing science or other mode of knowledge, a choice of how to classify; and yet it is not capricious. Although we may judge similarities and classify our experiences in many different ways, the atomist mechanist empiricist will hold that it is the job of a historian to classify things into kinds so that they can be covered by laws, so that the assumed real regularity of nature (including society) can be recognized and used to our advantage.

78The assumption to which we have been committed in the application of covering law theory is that the world consists of many regularly ordered events and conditions, things which are “constantly conjoined.” But we have now seen that only kinds of thing can be constantly conjoined, and therefore we are committed, in the application of covering law theory, to the existence of kinds of thing. The empiricist who is committed to atomism and mechanism is thereby committed to the claim that the world is regularly ordered, even though this claim is not derivable from experience; and he is further committed thereby to the claim that kinds of thing exist, and this claim too is not derivable from experience. Note here two different questions: whether kinds exist (to which we have our answer); and which kinds exist (to which we have no answer).

  • 25 This is not, of course, a reason why historians should seek to re-classify in a similar manner. Pa (...)

79It is sometimes noted that historians “fail” to be “scientific” in that they rarely provide historical laws. It is more interesting that historians do not provide developed ways of classifying the historical matters with which they deal, so that laws may in principle be found to cover them. If physicists still used the everyday notions of the classification of experience such as “table,” “tree” and so forth, it is highly unlikely that their theories would have had even the most rudimentary form of existence.25 Experiences had to be re-classified in ways that permitted theories involving assertions of regularity to get a grip, and this was done because it was thought, on the basis of the approach outlined so far, that theories involving regularities were the only kind of theories we could properly have. Notice that if we made different a priori assumptions from those involved in the belief in regularity, then it would be appropriate to seek different schemes of descriptive classification of experience altogether. Alternatives will be introduced in the next chapter.

80Thus the mechanism and atomism of which the covering law theory is an application state only that the world consists of many events and conditions in regular association with each other. They do not state what kinds of events and conditions exist, but covering law theory is inapplicable without an associated set of claims as to which kinds do exist, and indeed any linguistically expressible metaphysics or set of a priori non-empirical assumptions, any approach to knowledge about the world, requires such a specification. However, while mechanism and atomism do not in themselves provide a detailed metaphysics of kinds, they do provide strong constraints upon the range of possible theories here.

81Thus a certain range of conceptions of human nature is alone appropriate to answering historical questions, on this view. Human beings are either natural objects in their own right, or conglomerations of natural objects, or parts of natural objects, the activities of which objects are events and conditions in regular association with each other. A decision must be made, and somehow rationally argued for (for experience alone does not determine it), as to what are the basic kinds of thing in the world, including the peopled historical world.

  • 26 Karl Marx offered theories at this macrotheoretical level.

82Like most historians, a number of covering law theorists will object to the example above which purports to explain the occurrence of the American Civil War by reference to the American Revolution and a linking law of history.26 If the American Civil War is a large event, existing as a single real object in its own right, then it will be appropriate to explain it by seeking a law governing its association with other such objects. On the other hand, while the covering law theorist is not committed to any one metaphysics as to which kinds exist, he has to make a decision about which kinds exist anyway, and he may well reject the view that the American Civil War is a basic object of reality in its own right.

83It is open to the covering law theorist to adopt, as closely as he can, a “common-sense” view, like many historians, and say that the fundamental objects of reality are individual actions. The American Civil War will then be regarded as an aggregate of more fundamental things, each an individual action, and each individual action will require its own explanation in covering law form. The covering law theorist is therefore able to conceive of a historical account as a highly complex thing outlining a mass of regularly related events, all organized under the title “The American Civil War.” A title like “The American Civil War” is just a proper name given by historians to a number of events which are aggregated into the invented appearance of a single entity by the historians themselves.

84Again, it would be open to a covering law theorist to hold that only the cells of the body (the “most basic form of life”), or even just individual atoms, are the fundamental objects of reality, and that human actions themselves are properly to be seen as mere aggregates of the operations of such entities. Historians rarely adopt this view, but the past few years have seen a considerable increase in biographies of great historical figures written from a medical point of view, with explanations of actions cast in terms of a person’s physiological state.

  • 27 See the essays by Lukes and Mandelbaum in A. Ryan ed., The Philosophy of Social Explanation, Oxfor (...)

85The many disagreements that there are between historians, and between historians and other human scientists, can thus be conceived by the covering law theorist to be, not disagreements about whether to use covering law theory, but disagreements within this theory about the basic objects of reality. Sometimes, in the elementary philosophy of social science and history, these issues are expressed as a choice between “holism” (the view that social “wholes” like the American Civil War exist in their own right), and “individualism” (the view that individuals and their actions are the basic elements of reality), but it is not so simple as this.27 There are many more than two alternatives here, and it is not even possible to range them on a single scale, from less to more aggregation, or (put differently) from more to less reduction. Moreover, the issues of the metaphysics of kinds are not issues which arise only for covering law theory, and causation itself is a “kind” of thing, and may require to be differently understood, or excluded as a mode of understanding altogether. A decision on such matters is basic to history. Yet the decision must be rationally made, and experience alone, we have seen, will not do.

  • 28 On the general issue of the scientific nature of economics, see M. Blaug, The Methodology of Econo (...)

86We asked, in question 2 at the beginning of chapter 2, what would be a proper justification for the cliometric approach to history. Historians have to describe and explain the past world, given assumptions as to what that world is. Given the philosophical assumptions of empiricism, atomism and mechanism, as interpreted in this chapter, the historian will contribute to knowledge by explaining along the lines required by the covering law theory of explanation, and he will provide for himself, or derive from auxiliary sciences such as economics (or sociology, psychology, cell physiology or chemistry) the laws which he will need.28 Covering law theory leaves open the question of which laws to use, since this depends upon which particular metaphysics of kinds is adopted. A cliometric approach to history embodies a commitment to a particular metaphysics of kinds, and it will be justified as an approach to history insofar as the metaphysics involved is justifiable.

87But this covering law approach, with any associated selection of auxiliary science such as economics, needs a rational defence. We have defended empiricism against rationalism, and argued that we should proceed by trying to make the empiricist programme work, falling back on non-experiential sources of knowledge only where we have no alternative. We are now in such a position of fallback: it has become apparent in developing the Humean background to Hempel’s position that we have to make some assumptions about what the world is which we experience, and these assumptions are prior to our reliance on experience and not themselves derived from it. They are metaphysical, non-empirical, a priori assumptions.

88If we are to make sense of a shared world embracing historical time in particular we will have to have some kind of metaphysics, even if not an atomistic one; and it is clear that certain knowledge that the world is like this is not to be gained from experience alone. There must be, then, some knowledge claims — the metaphysical ones — that are not simply given in experience. While historians can provide knowledge insofar as they can base their claims upon experience, they will have to justify at the same time the conception they have of the nature of the reality which experience discloses. A conception of reality appropriate to the cliometric approach will similarly need justification. The question of the justification of these non-empirical a priori metaphysical assumptions is left until chapter 7.

Notes

1 One does not have to adopt an empiricist justification of science. See note 14.

2 So called by W. H. Dray, in Laws and Explanation in History, Oxford University Press, 1957, p. 1.

3 C. G. Hempel, “The Function of General Laws in History,” Journal of Philosophy vol. 39, 1942, p. 35. Reprinted in P. Gardiner ed., Theories of History, New York: The Free Press, 1959, pp. 344–356. The point of view involved was further expressed for history in M. G. White, Foundations of Historical Knowledge, New York: Harper and Row, 1969, and A. C. Danto, Analytical Philosophy of History, Cambridge University Press, 1968.

4 Hempel in Gardiner, op. cit. p. 345.

5 Hempel in Gardiner, op. cit. p. 347, note 1. There are theories of causation different from Hempel’s “regularity” conception. See, for example, R. G. Collingwood, “On the So-Called Idea of Causation,” Proceedings of the Aristotelian Society 38, 1937–38 p. 85; D. Gasking, “Causation and Recipes,” Mind 64, 1955, p. 479; J. L. Mackie, “Causes and Conditions,” American Philosophical Quarterly 2, 1965, p. 245; J. L. Mackie, “The Direction of Causation,” Philosophical Review 75, 1966, p. 441, and essays in T. L. Beauchamp ed., Philosophical Problems of Causation, Encino: Dickenson, 1974.

6 This is also M. G. White’s position. See his Foundations of Historical Knowledge, New York: Harper and Row, 1969.

7 David Hume, A Treatise of Human Nature, Selby-Bigge ed., Oxford: Clarendon Press, 1888, Bk. 1, Pt. 1, Sect. 1.

8 For a strong argument by a monist empiricist see F. H. Bradley, The Principles of Logic, Oxford University Press, 2nd ed. 1922, vol. 1.

9 Advanced study of causation may usefully begin with E. Sosa ed., Causation and Conditionals, Oxford University Press, 1975.

10 This is A. C. Danto’s question, See his Analytical Philosophy of History, Cambridge University Press, 1968.

11 David Hume, A Treatise of Human Nature, Selby-Bigge ed., Oxford University Press, 1888, p. 173, rule 3.

12 Ibid., rule 4.

13 For a non-empiricist introduction to logic, see R. Bradley and N. Swartz, Possible Worlds, Oxford: Basil Blackwell, 1979.

14 There is a wealth of other assumptions which might be made about the nature of science. See, to begin with, T. S. Kuhn, The Structure of Scientific Revolutions, University of Chicago Press, 1970, and R. Harré, The Philosophies of Science, Oxford University Press, 1972.

15 For an elementary account of probabilistic explanation see C. G. Hempel, Philosophy of Natural Science, Englewood Cliffs, N.J.: Prentice-Hall, 1966.

16 See J. L. Mackie, Truth, Probability and Paradox, Oxford University Press, 1973.

17 Karl Popper, The Logic of Scientific Discovery, London: Hutchinson, 1968, ch. 1.

18 See essays by M. Black and P. Achinstein in P. H. Nidditch ed., The Philosophy of Science, Oxford University Press, 1968.

19 Notice that it need not be denied that some theory connects “whenever C then E” with other scientific laws, so that, conditional on these other laws, a case of non-occurrence of C without E may affect the reason for believing “whenever C then E.” But this only transfers the issue to the problem of why we should hold true the other laws. In the end, the metaphysical assumption here has to be held true.

20 Hempel in Gardiner, op. cit., p. 346.

21 The argument form here is rightly ambiguous in that the event of kind E may or may not be the occurring event X itself. Given our interest, the occurring event of kind E is a different event Y, properly named the American Civil War. Again, it is not empty to say that the American Civil War was a civil war.

22 B. Russell, An Inquiry into Meaning and Truth, Harmondsworth: Penguin, 1962, p. 22. Hume regarded resemblance as basic.

23 A simple introduction to these issues is given in P. G. Winch, The Idea of a Social Science and its Relation to Philosophy, London: Routledge, 1989, particularly chapter 1.

24 This is at the basis of N. Goodman’s so-called “new riddle of induction.” See his Fact, Fiction and Forecast, Harvard University Press, 1955, and, for an updated discussion, J. L. Pollock, Contemporary Theories of Knowledge, London: Hutchinson, 1987.

25 This is not, of course, a reason why historians should seek to re-classify in a similar manner. Paul Veyne argues at length and with great plausibility that it is essential to the historical approach to ignore the general and describe in terms of the unique and singular. See his Writing History, Middletown: Wesleyan University Press, [1971] 1984.

26 Karl Marx offered theories at this macrotheoretical level.

27 See the essays by Lukes and Mandelbaum in A. Ryan ed., The Philosophy of Social Explanation, Oxford University Press, 1973.

28 On the general issue of the scientific nature of economics, see M. Blaug, The Methodology of Economics, Cambridge University Press, 1980, and on the philosophy of economics, C. Dyke, Philosophy of Economics, Englewood Cliffs, N.J.: Prentice-Hall, 1981.

© Les Presses de l’Université d’Ottawa | University of Ottawa Press, 1992

Conditions d’utilisation : http://www.openedition.org/6540