Chapter 2. Knowledge

1The philadelphia conference demonstrated how theoretical problems are involved in the practice of history. From the disagreement between historians exemplified in chapter 1 we drew three theoretical questions:

1. Can historians provide knowledge about reality?
2. What would be a proper justification for the cliometric approach to history?
3. 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?

2The concept of the profitability of slavery, in the third question, is to be interpreted in the “economic” sense (rather than the “accounting” sense), as explained in the last chapter. This notion of “profitability” is central to the cliometric approach exemplified in chapter 1, and this concept and its relationships will be explained in chapters 4 and 5. Notice that the concept will be so understood that, if the answer to question 3 is “yes,” then the cliometric approach to history will be justified, and the justification will supply the answer to question 2. Furthermore, “justification” is so understood that, if the cliometric approach (or any other approach) to history is justified, then question 1 is also answered in the affirmative.

3It may look as if we should turn immediately to question 3, since everything else appears to depend on it. But that would be a mistake. Notice that, if the answer to question 3 were to turn out to be “no,” then nothing at all could be concluded about the answer to question 1. Furthermore, we would need some idea of how historical knowledge is to be achieved, and how an approach to history could be justified, before we could begin to answer question 2 or 3. We therefore now turn to the first question, whether historians can provide knowledge about reality. The fundamental issues of knowledge and justification will be dealt with in this and the following chapter.

4Central to the questions derived from the Philadelphia conference was the question of whether historical knowledge is even possible, and the philosophical theory of historical knowledge is intended to answer this. But, in the face of so bald a question, the most expert philosopher will flounder. We need to have some idea of what is to count as knowledge, and of what might justify a claim to have achieved it.

5“Knowledge” is a noun; it seems reasonable to suppose, then, that it is the name of something. But where do we look for that thing? Perhaps in libraries or in computer memories? If we allow that knowledge exists here, then it would mean that, for example, knowledge would continue to exist in the rubble of ruined libraries long after the end of humankind.1 Many philosophers would feel uncomfortable with this result, because they would find it odd to suppose that knowledge could exist independently of people knowing (or at least of knowers knowing, if you think that there are other beings in the universe). “Knowledge,” it would be held, consists essentially in what people know, and it is knowing, considered as a verb, which we should attend to. We should understand “knowledge,” then, not as the name of some entity, but as shorthand for the many cases of people knowing things.

6There are four kinds of knowing, although it ought to be said that philosophers do not like recognizing too many kinds, and often seek to show that one kind is really a part of another (that way, it gets explained). The first kind is knowing that something is the case, that some statement is true. “Statement” can itself be a technical word in philosophy, but we need not worry about that here, for it means for us only the kind of sentence that can be true or false. “The cat sat on the mat” and “over 25 per cent of males were managers” are statements; “Do this” is not a statement but some sort of command; “Is Fogel right?” is a question. One needs no more than the most basic grasp of English grammar to know a statement when one sees or hears it.

7The second kind of knowing is knowing how to do something, such as riding a bicycle.2 The third kind is knowing what something is like, such as what it is to be a slave.3 The fourth kind is knowing about something, such as the state of slavery in the ante-bellum South.4

8Statements are the kind of sentence that can be either true or false. They cannot be both true and false; and let us now suppose, as it is standard to do, that only statements can be true or false. Statements are linguistic expressions of some sort, and it is impossible to make sense of truth and falsity attaching to anything but some sort of linguistic expression. The first kind of knowing, knowing that something is the case, is understood by its relationship with statement-truth: to know that something is the case is to know it to be true, and the “it” here is a statement. We may similarly understand the other three kinds of knowing better if we examine their relationship with statement-truth.

9The second kind of knowing, knowing how to do something, has a quite different relationship from the first kind with statement-truth. Knowing that one has to push the pedals, and so forth, in order to ride a bicycle, is not the same as knowing how to ride a bicycle. One might know any number of statements here (all true, of course), and still fall off. Conversely, one might be able to ride with ease, and yet be unable to express in the form of true statements what was involved: one might be quite unsure, for example, whether one was pushing or pulling on the handlebars; indeed, an attempt to attend to what one was doing here, in order to formulate some true statements, might of itself cause one to fall off.

10Thus, knowing how to do something is not in any way a matter of knowing that certain things are the case, and hence it is not a matter of having within one’s mental grasp any true statements at all. Statements (whether true or false) simply have nothing to do with it: such knowledge is not a linguistic matter. Truth is irrelevant to this kind of knowledge. Historical knowledge is expressed in linguistic ways — for example, “that slavery was an efficient system” — and it is these linguistic expressions which we assess when we assess knowledge claims in history.

11While historical knowledge may require the exercise of certain skills for its achievement, it does not consist in the exercise of those skills. The philosopher R. G. Collingwood adopted a position which might be thought contrary to the conclusion here: he held that history was essentially the “re-enactment of a past experience” or the “rethinking of a past thought.”5 But even where such a skill or activity is so essential as to be constitutive of history, the expression, in the present, of the thought or experience involved has to be linguistic in form. It was not, for Collingwood, historical writing, the expression of knowledge, which was constituted by “re-thinking a past thought,” but historical reality itself.

12Less easily excluded from our concerns is the third kind of knowing, knowing what something is like, such as what it is to be a slave. Claims of this sort are often viewed as being essentially derived from some privileged experience. Thus some say that one cannot really know what it is to be hungry and poor unless one has experienced it, or that men (say) can never really understand (i.e., know what it is to be) women. Such experiences as will support such knowledge are available only to a special group, and not publicly available to all.

13Notice that the privilege claimed here is a strong one. It is not merely that an ordinary kind of privilege is involved, such as the privilege of those few who have been able to observe a total eclipse of the sun from an aircraft flying above the equator. The privilege we are concerned with entails, rather, that the acquired knowledge of “what something is like” is not transmissible or communicable to those without the experience, not because of some fondness for secrecy, but in principle.

14“Only women know what it is to be women” is not like “Women know the answer (but aren’t telling).” Women couldn’t tell. If knowing what it is to be a woman is expressible in terms of statements at all, then it is communicable. A man could know what was involved, simply through being able to grasp what the relevant statements were; having grasped their meaning, then he would also have grasped what it was for them to be true, even though he was not in a position to experience the facts involved for himself. Women might know that certain things were so; once men knew that they knew it, then they would know those things for themselves. But if men cannot grasp such knowledge, then it follows that women cannot express their “knowing what it is like” in the form of statements — not even to each other. Knowing what something is like would be essentially private.

15Knowledge of what something is like, conceived in this way, would be impossible to express, and thus historians would have no choice but to avoid it. The difficulty here, however, is that it is plausible to say, firstly, that historians ought to be in the business of doing this impossible thing, by making intelligible to their readers “what it was like” to be, say, a mediaeval peasant or a black slave. Secondly, it is surely plausible to hold that historians and even authors of fiction are able to express to us “what it was like” to be a mediaeval peasant or a black slave; indeed, many novelists do this sort of thing rather well.

16While great literary expertise may perhaps be required on occasion, such knowledge does seem to be communicable, from which it follows that it cannot be derived from privileged experience, but rather from experience that all can share in some sense. The reader of a historical account may be able to share with the historian the experience of, say, being a slave, without either reader or historian ever having been slaves. This kind of knowledge, then, requires an empathy with the experience of others, a sharedness of understanding. Such empathy is required for historical knowledge, and is in fact achieved. Nevertheless, since it is expressible, it must be expressed in statements, which must of course be true. We may conclude, therefore, that knowing what it is to be something, knowing what something is like, has to be — in the form that historians must be concerned with — expressible in terms of knowing that something is the case.

17The fourth kind of knowing is knowing about something, such as the state of slavery in the ante-bellum South. “Knowing about” has many uses. “Knowing about riding bicycles” implies more than merely being able to ride some one bicycle, or even any bicycle that is offered; it implies also that one knows a lot of true statements about riding bicycles. Although “knowing about” does not imply unlimited knowledge of the relevant matters, it implies far more than “knowing how” does, or just one instance of “knowing that.” “Knowing about,” however, is of wide meaning in that sometimes it does, and sometimes it does not, imply something about “knowing how.”

18Historians know about things, but there is no implication here that they are thereby exercising some skill, akin to riding a bicycle (although skills may have been required to achieve that knowledge). Rather, what they know is that a number of things were the case. But not a small number of things: knowing “about” something implies a comprehensiveness of things known, but only of relevant things. Historians’ accounts of historical matters such as the American Civil War express knowledge about their subject matters only in so far as they contain some comprehensive set of statements which severally express cases of knowing that certain things occurred. You do not know about the Second World War, in the sense historians need, when you know only that, for example, it began in 1939. You need also to know that a great many other things happened, and what they were. The things that need to be mentioned will depend on the question that is asked: “Was slavery profitable?,” if answered, will provide knowledge about the profitability of slavery, and will include a comprehensive set of true statements relevant to the issue.6

19Our four kinds of knowledge, and the kinds of understanding they permit, are all relevant to history, then; but all in effect collapse into the single notion of “knowing that something is the case.” The “is” here does not of course exclude the past tense. Our understanding of historical knowledge should henceforth be understood in this way.

20Now that we have an outline of what is to count as knowledge, we can deal with the question of what might justify a claim to have achieved it.

21When is a claim to knowledge justified? A number of relevant considerations were presupposed in chapter 1. Firstly, that knowledge has something to do with the avoidance of mere opinion. Secondly, that knowledge has something to do with the elimination of background attitudes adopted in advance of proper research. Thirdly, that knowledge has something to do with the avoidance of evaluative judgements. Fourthly, that knowledge has something to do with the avoidance of irrational considerations. These issues can be clarified briefly here.

22In the first consideration, “opinion” is a vague word. Some opinions can be well-founded, in other words, properly supported by rational argument. It would not matter to the achievement of historical knowledge if opinions in this sense were involved. Plainly, however, “mere” opinion, opinion which has no rational basis, is to be avoided, and this is expressed in the fourth consideration in the last paragraph.

23The second consideration is similarly subordinate to the fourth: it would not matter to the achievement of historical knowledge if background attitudes which were rationally well-founded were involved. We only need to avoid such things where they are not rationally supported. Thus anyone who argues that historical knowledge is impossible because we cannot avoid the intrusion of some background opinions and attitudes is mistaken. What he has to do, if he wishes to pursue his argument, is show that those unavoidable background opinions and attitudes are impossible to support rationally. The “neutral observer,” whom we introduced in chapter 1 to decide between different background beliefs, is, of course, the completely rational person. Perhaps there is no such person? That would not matter, so long as we knew what it was to act in a rational way, so long as rationality here is possible.

24The third suggestion, that knowledge has something to do with the avoidance of evaluative judgements, is also subordinate to the fourth. The foundation for evaluative judgements is part of the subject matter of ethics. Our understanding of morality, of right and wrong, has suffered more than most theories from the supposition that its subject matter is a mere matter of opinion. So widespread is this belief that it is often thought that one need do no more than show that some discipline — history, for example — is “evaluative,” if one is to show that it is subjective, a mere matter of opinion. Much of moral philosophy, however, is concerned with the question of the objectivity of judgements of value. If judgements of value can be rationally supported then it would not matter if historians made such judgements.

25Historical knowledge, therefore, is that which is rationally supported to be such. Our question “is historical knowledge possible?” thus comes down to the question: can the claims which historians make in their writings be rationally supported? Whether we ought to proceed, in our Philadelphia example, by adopting an econometric approach similarly requires parallel rational support.

26“History,” the historian J. B. Bury announced in 1902, “is simply a science, no less and no more.”7 “But,” commented G. M. Trevelyan on this, “history has no properly scientific value.”8 What we have achieved so far is the reduction of the theoretical question about the possibility of historical knowledge to a question about the rational support for historical claims. Some philosophers, following Bury’s assumption, hold that the only way of rationally supporting a claim to knowledge is by means of science. This belief that science is the only way of achieving knowledge is called positivism. (The word “positivism” has been used in many different ways in the history of ideas. Be careful, if you meet it outside this book. The meaning given here is the standard one in the theory of knowledge.)

27“Science” here is taken to refer directly to the natural sciences, such as physics. The question whether history does or can achieve knowledge then becomes the question whether history does or can share the kind of precise explanation and successful prediction which, it is supposed, are achieved by the natural sciences.

28But note that it is irrational to appeal to physics and suchlike to set the standard for knowledge unless one has reason to think that they do succeed in achieving it for themselves. Positivism is a very weak claim, when it has no attached standard for what it is to be scientific, and when it offers no reason why we should regard science as a way of achieving knowledge.

29Examining Bury’s claim that history is a science would only be useful if we meant by this to check history against the standards set by an independent theory of knowledge. “Science” here would then be the middle man, and is more efficiently cut out from the argument. There is no justifiable commitment in advance to positivism, no commitment in philosophy of history to the philosophy of the natural sciences. It is as well that this is so, for much recent work on the philosophy of science does not appeal to independent theories of knowledge but rather to the history of science.9 It would be a flatly circular argument to attempt to justify history by an appeal to science, and then attempt to justify science by an appeal to history.

30Independent of science or of history, the theory of knowledge — “epistemology” — offers two basic, but competing, accounts of how knowledge claims are to be rationally justified: empiricism and rationalism. First we shall deal with empiricism, a theory perhaps most famously expressed by David Hume, whose Treatise of Human Nature was first published in 1739. The essence of empiricism is the belief that knowledge can come only from experience.

31It should be noted that there is no essential connection between positivism and empiricism. One may be a positivist and not an empiricist: one may believe that sciences like physics achieve knowledge, and are the only way of doing so, and yet not think that science derives knowledge only from experience. Or one might be an empiricist and not a positivist: thus one might think that knowledge comes only from experience, but that science concentrates upon only one kind of experience, and is not the only way of achieving knowledge.

32In principle we would not need to defend any reliance upon empiricism here. In principle we could use empiricism as a hypothesis in the theory of knowledge, being entirely neutral about whether it would work, and leaving its plausibility to be decided by how much sense it made of our knowledge-acquiring activities and of our explicit contributions to knowledge. Nevertheless, this pudding-in-the-eating test apart, empiricism does have considerable initial plausibility, and we shall next proceed to a conditional defence of it.

33The claim that experience is the only source of knowledge does not prevent the empiricist from recognizing that, for example, we are born with knowledge of how to suckle. Recall that empiricism, like rationalism, is offered as a theory of “knowing that.” It is in this sense that the empiricist says that experience is the only source of knowledge. Moreover, “experience” covers a lot: everything of which you are consciously aware, independent of your thoughts: that which comes to you by means of touch, taste, smell, hearing and sight. If there are other senses, then these too are included; and indeed there are other senses of a kind: one may feel ill, or conscious of oneself as moving, or even angry, although it is not obvious that these experiences come to us through any of the five main senses. All these experiences are, in some way, given to us, and we seem to be passive receptors of what we sense. Once we have our eyes open, there does not seem to be much we can do to alter what is seen. We are not aware of having much option about it, although certainly there is a wide latitude of choice of how to describe what we experience. With experience understood in this properly large sense, we may move to a second point: apart from experience, where else could knowledge come from?

34Perhaps we are just born with knowledge, or have it revealed to us directly by God? Empiricism does not have to deny our being born with innate beliefs, or our having something revealed to us by God. It attacks such suggestions in their weak place: given that you are born with, or had revealed to you, some beliefs, what makes you suppose that they constitute knowledge? No matter how effectively God, or your genes, are working in your epistemological interests, why should the rest of us suppose that your privileged sources are giving you knowledge rather than merely false beliefs? God would not mislead, you might say; and yet one could only tell this — discover that He was not misleading you — by checking the beliefs with which you were supplied against some independent standard: experience, which we can all share.

35One essential feature of knowledge is brought to our attention here. A person who makes a claim to knowledge has to defend it to the satisfaction of others, by appealing to public standards that all can share. Experience, as encapsulated in the language we learned to share in talking about a shared world, is pre-eminently the right kind of thing for this role. Private and privileged sources for knowledge claims, no matter how splendid, are worthless, unless others can be persuaded, by good objective reason, to go along with them. “Knowledge,” an ordinary English word, we were taught to use correctly as children, and it is a matter for public checking.

36But need public standards that all can share necessarily involve only experience? What about logic? Is not that a source of knowledge? The rationalist philosopher (not the same as the rational philosopher, which the empiricist would also claim to be) holds that reason, which we all share and yet is not derived from experience, is a possible source — or even, as René Descartes claimed in his Discourse on Method of 1637, the only source — of knowledge. The rationalist claims, at the minimum, that we do have innate knowledge: some things we know to be true simply because it is self-evident to us that they could not be false. We can use these things as premises in our arguments, the validly reached conclusions of which we can claim to know.

37Certainly we can rely upon logic, and where we know that our arguments are valid, we can claim properly to know the conclusions of them if we know the premises. More about logical reasoning will be said in chapter 3. The empiricist, arguing against rationalism, attacks not so much the validity of such arguments as the rationalist truth of their premises. The empiricist’s view is that the premises the rationalist wishes to use, if they say anything that can be known at all, must get their truth from some connection with experience.10

38In order to be clearer about the claims of empiricism and rationalism, and about related issues of knowledge and existence which are relevant to history, we shall next spend a little time introducing some terminology originally derived from Immanuel Kant, who produced the famous synthesis of empiricism and rationalism in his Critique of Pure Reason of 1781. The words he introduced were, firstly, “analytic” and “synthetic,” and these technical words are now standardly used to mark the two ways in which statements can be true.

39“Analytic” and “synthetic” exhaust the possibilities: for any true statement, the claim is that it is either analytic or synthetic. An analytic statement is true simply in virtue of the meanings of the words in it. For the moment, we shall not look at historical examples of analyticity or syntheticity, in case the past tense obscures the issue. Consider, instead, the statement “a bachelor is an unmarried man.” This is a true statement, and its truth derives entirely from what the words in it mean. Given that “bachelor” means what it does mean (“unmarried man”), then it is bound to be true, for one is, in effect, saying that an unmarried man is an unmarried man. It could not be false, and its truth derives entirely from its meaning.

40By contrast, a synthetic statement is true in virtue of some fact about the world, in virtue of reality actually being that way. The statement “This pen is blue” is true because the pen I am referring to is blue, and it is that fact which makes the statement true. However, the relationship between a synthetic statement and the fact which makes it true is not so pure as the relationship between an analytic statement and the meanings it involves. This is because some element of meaning is involved in accounting for the truth even of a synthetic statement. The meanings of the words involved in an analytic statement are entirely sufficient for supporting its truth, but the fact involved in supporting a synthetic statement is not sufficient for ensuring its truth, for more is required about the meaning. Thus “This pen is blue” would not be true if the pen I am referring to is not blue, but it would also be false if “blue” had some different meaning, such as “black.”

41It is common to shorten the explanation of these terms in the following way: “analytic” means “true by virtue of meaning,” and “synthetic” means “true by virtue of fact.” However, “synthetic” really means this: “(given that the words mean what they do) true by virtue of fact.” But there is a little more. We have introduced here ways in which statements are true, but of course statements can also be false. A statement which is false by virtue of meaning, such as “all bachelors are married,” we call self-contradictory, for the obvious reason that if you make this statement then you are contradicting yourself. However, a statement which is false by virtue of fact — such as “this pen is blue,” said of a black pen — we still call “synthetic.” One can only admit that this may be unhelpful, but plead that philosophers usually leave it to other people to discover whether synthetic statements are true or false. Notice, incidentally, that definitions of words are analytic.

42We may illustrate with historical examples the application of these terms. “The slave Frederick was black” is synthetic; it is true in virtue of fact. But what fact? Here a problem arises.

43First we must introduce the concept of objectivity. It is common to contrast “objective” with “subjective,” and these are words very often used to express the epistemological questions about history. “Objective” suggests something “real” and “out there,” beyond what is merely in our minds. “Subjective” suggests something which is just in our minds, and hints at the merely imaginary. Sometimes people look at the matter as if knowledge were a kind of picture in our heads, a picture of reality outside. The picture becomes knowledge, rather than mere imagination, just so far as its representation of reality becomes more exact.

44This very simple way of looking at the matter is called realism, because it supposes that reality is indeed “out there,” and existing independently of us and our understanding of it. This belief is to be contrasted with its opposite, idealism, which takes the view that what we call “reality” is something which depends for its existence upon our own minds. At first sight most people think idealism to be a quite implausible position, but if you reflect that there is no way that we can appreciate reality except through the medium of our own minds, then you may come to grasp its plausibility. The idealist, of course, argues for more than its mere plausibility; he holds that we can simply make no sense of the existence of something independently of its existence as known by us, as mediated through our minds. We are never in a position to know about something independently of our knowing about it.11

45This apparently abstract metaphysical issue between realism and idealism may be too abstract for everyday experience, but it is important for the theory of history, since it is obvious that we cannot see the past. We get at the past by means of historical research, but we are never able to look up from our historical writings to compare what we have written with reality, to see if we have got it right. Reality is always present, and any claims we might make about the past seem to be quite uncheckable. The idealist claims that the only knowledge we have of the past is the knowledge expressed in the writings historians construct. Thus the past which historians write about has no existence except in their writings. R. G. Collingwood, mentioned earlier, had a position like this.12

46Since there is no historical object “out there” which historical writing is about, and which would make that writing, if correct, “objective,” you might think that the idealist metaphysician is committed to holding that history cannot be objective. This is not so, however. For, if the idealist is correct, knowledge cannot be conceived to consist in the matching of what we say with some independent external reality at all. The very idea of trying to conceive knowledge in that way is impossible. We can hardly complain, then, if we “fail” to achieve this impossible thing, this nonsensical thing, and we must therefore understand knowledge in a different way. We already have the beginnings of a different approach when we talk about knowledge being based upon rational support. We do not yet know what “rational support” consists in, but we know — if we are idealists — that it cannot consist in representing some independent external reality. It must lie in something like, for example, valid reasoning from present evidence. That way we could perhaps guarantee truth, even though we could never observe the past that we were writing about. And once we have achieved truth, we must admit that we have achieved objectivity also.

47A simple-minded realist might think that the idealist is committed to holding that history cannot be objective, for the realist allows, by contrast, that there is some independent external reality which is correctly pictured by our writings, if they are true. But even the realist is no better off here, for while there may be some such external reality, it is certainly the case that, so far as history is concerned, we have no way of observing it. Thus in practice truth has to be guaranteed in some other way, perhaps in valid reasoning from present evidence. That way we could perhaps guarantee truth, even though we seem regrettably unable to obtain independent verification from the past that we were writing about. And once we have achieved truth, we must admit that we have achieved objectivity also.

48Thus both idealist and realist metaphysicians find themselves forced to say that historical knowledge, if it is possible, is not made so by any kind of copying of reality.

49Given this issue between realism and idealism, perhaps there is no such thing as the past fact that Frederick was black. (This is why a historical example was not used to introduce the notion of syntheticity in the first place.) Perhaps there is no such thing as a past reality for the statement to be true of. Perhaps; but even so, note that it does not follow that “The slave Frederick was black” is analytic. One easy test for analyticity is this: for any statement, try denying it. If you can do so without contradicting yourself, then it is not analytic. (Do not confuse contradicting yourself with merely being wrong.) Since every true statement is either analytic or synthetic, then one which is not analytic must be synthetic. Our historical example, by this test, is synthetic. This is hardly surprising: the statement obviously says something about reality, in some sense of “reality.”

50Thus “true in virtue of fact,” for the idealist, cannot mean “true in virtue of independent past fact” when he is dealing with statements in the past tense, and this has to be understood in some more complex way. The idealist often speaks of synthetic truth as being assured by the place of a synthetic statement in a vast web of beliefs, the acceptability of which depends upon its internal coherence, consistency and comprehensiveness, and the whole of which expresses (mind-dependent) reality. A notion of coherence very close to this will be developed in the last chapter.

51People can get very muddled when they think about idealism, realism, empiricism and positivism.13 Empiricism and positivism have already been distinguished. Still, many historians think that empiricism, realism and positivism go essentially together. Partly this is due to the various ways in which all these words have been used in the history of thought. But, even given the explicit standard meanings introduced in this chapter, quite a few historians have thought that idealism and positivism were natural opposites. This is not so, and it is worth spending a moment to see why.

52Positivism asks us to judge by the standards of science, without specifying what this means; realism and idealism are theories about the nature of reality; empiricism is a theory about the foundations of knowledge. It is easy to be a positivist idealist: one simply takes an idealist view of all reality, including that part of reality which is the subject matter of the natural sciences. One then goes on to assert that the standard of knowledge which is exemplified in natural science is the standard which is to be used for all knowledge, including history. Additionally, just as it is perfectly possible to be a positivist idealist, so it is perfectly possible to be an empiricist idealist. As an empiricist, one holds that one’s knowledge comes only from experience, but what that experience is experience of is answered by an idealist conception of reality.

53One final source of possible confusion: do not suppose that to be a scientist is necessarily to be a positivist, or a realist, or an empiricist. Many scientists believe in God; some believe that science is a tool, and not a description of an independent world; and however strongly their theories happen to derive from experimental work, it may not be the case, and they need not believe, that experience is the only possible source for knowledge.

54Let us next turn to exemplify analyticity. Notice that, while “The slave Frederick was black” is synthetic, “The slave Frederick was owned by somebody” is analytic, for being owned by somebody is part of what “slave” means. “Frederick was owned by somebody,” however, is synthetic. These are simple examples. Some are more difficult to recognize. Is “All human beings are mortal” synthetic or analytic? Is “Blacks are human beings” synthetic or analytic? Is “The loss of jobs led to unemployment” synthetic or analytic? Compare this last with “The freeing of the slaves led to unemployment.”

55It is obviously not always easy to tell whether some statements are synthetic or analytic. But do not think that your inability to judge some cases is due to lack of clarity in these technical terms. On the contrary, by trying and failing to apply them we learn that there are many words the meanings of which we are not absolutely clear about. The lack of clarity lies not in the notions of analytic and synthetic but in the words to which we try to apply them. Thus, does being a human being mean being a mortal thing? If it does, then claiming that all human beings are mortal is to say something analytic, while to deny it is to contradict yourself. But would one be contradicting oneself in supposing that some human being (you only need one) was immortal? You can just decide the issue, up to a point, so long as everybody else knows what you mean. The value of dictionaries lies in telling us what most people mean by certain words.

56“Blacks are human beings” is problematic in a different way from “All human beings are mortal.” “Blacks” does not mean “black objects” but black human beings, and thus denying that blacks are human beings is to contradict yourself. The statement is therefore analytic; and yet, is one really saying here nothing which relates to the world? Attend to when you might make a statement like “Blacks are human beings.” Most likely you would do this in order to make a moral judgement, a recommendation of a certain kind, a plea for better treatment, for example. But moral judgements, recommendations and pleas are more like commands and questions, in some ways, than like statements. If they are not statements, then “analytic” and “synthetic” are not intended to apply to them. Whether moral judgements are statements — the kinds of thing that can be true or false — is a major issue of moral philosophy.

57There are two more technical terms originally derived from Immanuel Kant’s work. These are “a priori” and “a posteriori,” literally translated from Latin, respectively, as “from the first” and “from afterwards.” You may often come across statements being described as “a priori” or “a posteriori” in works of philosophy. But there is an unfortunate unclarity about these descriptions: properly speaking, it is not statements that are “from the first” or “from afterwards,” but the way that they are known. In this, we have a contrast between the “a priori/a posteriori” pair of terms, and the “analytic/synthetic” pair of terms. The former refers to the ways in which statements are known. The latter refers to the ways in which statements are true. These are different things.

58A statement which is known a priori is known from the first: as soon as you know what the statement is, the moment that you understand it and grasp its meaning, then you know it to be true. (It may take time to grasp the full meaning of a statement.) We have already met an example of a statement which is known in this way: it was the analytic statement “A bachelor is an unmarried man.” Such a statement is known a priori: we know it independently of experience.

59A statement which is known a posteriori is known only “afterwards”; it is not enough merely to hear the statement, understand it and grasp its meaning, for some further step which leads to knowledge has to be taken. We have met an example of such a statement already: it was the synthetic statement “This pen is blue.” We know that this is true, but we did not know it a priori, for we had to take the additional step of looking. Such a statement is known on the basis of experience.

60We have said that the “a priori/a posteriori” pair of terms differs from the “analytic/synthetic” pair of terms, in that the former refers to the ways in which statements are known, while the latter refers to the ways in which statements are true. We have stressed this difference; and yet you will have noticed that the examples given were the same for “analytic” and “a priori,” and for “synthetic” and “a posteriori.” This was deliberate: it would certainly be misleading you to think that the pairs of terms only occasionally correlate. The empiricists would like you to think that the pairs were always correlated, however, and this we shall now examine.

61Are ways of knowing and ways of truth the same things? First, some elementary remarks about the relationship between knowledge and truth. The most straightforward relationship between them is that to know a statement is to know it to be true. You cannot know something false. But we need to be exact here. You may know perfectly well that this pen is blue, and thus know that it is false that it is black. But in knowing the latter you do not know a false thing but a true one, namely this: it is false that this pen is black. What you know here, the bit after the colon, is true. To know that something is the case is to know it to be true. This is what “know” means, in part; thus “To know that something is the case is to know it to be true” is analytic.

62Here, then, we have one firm link between knowledge and truth. But it is not obvious that there is a firm link in the opposite direction: while there cannot be knowledge (that is, “knowing that”) without truth, it may well be that there can be truths without knowledge. Whether or not there are truths without knowledge is a difficult philosophical issue — it is, in fact, the idealism-realism issue in one of its many forms — and there is nothing obvious about the answer.

63Ways of knowing and ways of truth perhaps diverge, then. We know that there are analytic a priori and synthetic a posteriori statements. We may see the further possibilities here by asking whether there can be statements which are both analytic and a posteriori, and whether there can be statements which are both synthetic and a priori.

64What would a statement which was both analytic and a posteriori be? It would be a statement which was true in virtue of its meaning, and yet which we did not know to be true even though we knew its meaning, for we did not know it a priori. Plainly such a statement would be impossible to have. The truth of an analytic statement consists in its meaning, and once one knows what the statement is and thus knows its meaning, one cannot fail to know its truth, for there is nothing else to know.

65The final consideration here is of a statement which is both synthetic and a priori. Such a statement would be true in virtue of fact; it would involve asserting something about the world which was true just because reality was like that. And yet we would know that statement to be true a priori, as soon as we understood it, without any need to appeal to experience; although we would recognize that, if we did appeal to experience, then we would find nothing to the contrary. (This would suggest that some statements could be known either a priori or a posteriori, in the inclusive sense of the word “or.”14) Notice that it would be inconsistent to suppose that the recognition of the truth of the statement derives from its meaning alone, for that would make it analytic. The rationalist here might speak of our knowledge of the statement being derived from “reason,” but “reason” here, while certainly not experience, must also not be understood as the absence of self-contradiction, for, again, this would make the statement analytic. “A priori” is a way of knowing, not a kind of truth.

66Now, one might be wrong in supposing that there are statements of this synthetic a priori kind, but would one be contradicting oneself in supposing it? It seems clear that one would not be. The possibility of synthetic a priori statements, or indeed of statements which may be known either a priori or a posteriori, is in itself enough to show that “analytic” is not equivalent to “a priori,” and “synthetic” is not equivalent to “a posteriori.” It is also enough to show that the “analytic/synthetic” pair of terms is not functioning in the same way as the “a priori/a posteriori” pair of terms: ways of knowing and ways of truth are in principle different things.

67We introduced the analytic/synthetic distinction and the a priori/a posteriori distinction in order to make clearer the issues between the rationalist and the empiricist. The empiricist takes the view that knowledge comes only from experience: in other words knowledge is always a posteriori. Thus, for him, there are no synthetic a priori statements. There are, however, both analytic and synthetic statements, on his view. Since analytic statements are known a priori, surely the empiricist must therefore allow that some knowledge is a priori? Not in any worthwhile sense. What do you know when you know an analytic statement? The answer, for the empiricist, is nothing, for they are empty. To say that a bachelor is an unmarried man has as much content as to say that A = A, which is, none at all. They do not enter into knowledge, therefore, trivially true as they are, and the empiricist thus keeps his claim that knowledge comes only from experience.

68It is worth noting in passing that, even if you disagree with the empiricist about the relationship to knowledge of analytic statements, you are not much better off in objecting to his overall position. Whatever they are, analytic statements are not synthetic, and it is synthetic statements — the ones which say something about reality (however “reality” is to be understood) — that we are interested in, for the purpose of analysing knowledge of the actual world of history. For our purpose, the part of the empiricist claim having to do with synthetic statements — that these must be a posteriori — is all we need to deal with.

69If empiricism is true, then “analytic” and “a priori” will in practice amount to much the same thing, and “synthetic” and “a posteriori” will in practice amount to much the same thing, and the differences between the two pairs will amount to something of merely abstract interest. Many philosophers who are empiricists thus use “analytic” and “a priori” interchangeably, and “synthetic” and “a posteriori” interchangeably, and it is no surprise that the important word “known” is usually missing before “a priori” or “a posteriori.” Even non-empiricists or rationalists forget sometimes that such interchangeable uses are only reasonable if empiricism is true. A number of historians, social theorists, and others who learnt their philosophy from empiricist writers contradict themselves as a result, desiring perhaps to express an anti-empiricist viewpoint while using terminology which commits them to empiricism.

70By contrast with empiricism, if rationalism is true, then there will be synthetic a priori statements. These will say something about reality, and yet will be known a priori, known as soon as one hears them, not in virtue of the meaning alone (for that would make them analytic), but on the basis of “reason” in some form. One candidate for a synthetic a priori statement is this: “every event has a cause.” The rationalist claim is that you know it a priori, and indeed you could not know it a posteriori, for no experience could support it. You cannot experience every event, and even if you could it would not help, for the statement “every event has a cause” holds, not merely that all the events in your experience have causes, but that all the events there possibly could be have causes. So you would have to experience possible events as well as actual ones, which is not a requirement you could meet. Not only that, but even the events which you are in a position to experience often do not appear to you to have causes. To put it another way, there is a vast amount you do not know about the world: there are many events of the causes of which you are ignorant. Thus experience often gives you events without, apparently, causes. Therefore experience in a number of ways does not support the claim that every event has a cause; and yet, the anti-empiricist would claim, you do know it all the same. You know it a priori. The cause must be there, in every possible case, even though you fail to find it.

71The empiricist has two lines of reply to rationalist claims that some statements are synthetic a priori. He can say that apparently synthetic a priori statements we do, in spite of appearances, know a posteriori after all; or — and this is the stronger move and more commonly made — that we do not know them. We may believe that every event has a cause, and even bet money on it in funding research, but it just is mere belief, which we may be wrong about. It is open to us, therefore, to doubt such beliefs. This is beneficial, since we should not appeal to what amounts as guesswork, as the empiricist characterizes the rationalist position here.

72The strongest argument against the rationalist at this point, although it is a rhetorical one (which does not prevent its being efficacious in directing philosophical research), is an appeal to the principle of the wedge. Once we admit some synthetic statements as contributions to knowledge on an a priori basis, then where are we to stop? We have no reason for picking out some claims rather than others. What makes “every event has a cause” different from “dragonflies are fairies in disguise”? The only statements which we can accept a priori as self-evidently true are analytic statements, according to the empiricist, and these are empty. The more plausibly the rationalist argues for the self-evidence, the necessary truth, of the statements he claims to know a priori, the more material he gives the empiricist for arguing that these statements are analytic after all, and thus empty.

73Rhetorical though it is, the principle of the wedge has the effect of placing the burden of proof upon the rationalist. “Burden of proof” is a simple notion; in our courts of criminal law it lies upon the prosecution. (Logic alone is neutral about where it should lie.) The issue is, what should be assumed to be the case, in default of sufficient proof one way or the other? Not to decide is itself to make a decision, a decision to leave things to take the course they would have taken without one’s intervention.

74When we propose to shift the burden of proof away from the empiricist, we are saying that we shall hold empiricism correct until we are proved to be wrong. If, by contrast, we were to place the burden of proof upon the empiricist, then we would be in effect asking him (and all of us) to accept any statement for the status of synthetic a priori until disproof was available. In the case of a synthetic a posteriori statement this is quite acceptable, for it amounts merely to testing it against experience, which may well provide disproof. Statements which fail such experiential test are quite properly to be rejected, and those which pass are quite properly to be accepted.15 But how could you demonstrate that “every event has a cause” or “dragonflies are fairies in disguise” are wrong? Not finding the cause in some particular case only shows that you have not looked hard enough. Not finding a fairy behind the dragonfly only shows how good the disguise is.

75To avoid sheer speculation, to avoid the wildest beliefs going unchecked, we conclude that we should place the burden of proof upon the rationalist. We do this by trying to make the empiricist programme work, falling back on non-experiential sources of knowledge only where we have no alternative. This, then, is our defence of proceeding on the basis of empiricism. We continue with the following claim: knowledge can come only from experience. We shall see, in the course of the argument, that we will indeed need to fall back on non-experiential sources of knowledge as empiricism is pushed to its limits.