[ Acknowledgements | Introduction | Part One | Part Two | Part Three | Appendix ]
The most that follows, I have tried to show, from the true premisses of Hume's argument for inductive scepticism, is the incurable invalidity of all inductive inferences. Inductive fallibilism (9) therefore, if I am right, exhausts Hume's positive contribution to the philosophy of induction. In the present century, when Hume's reputation has been raised to an immense height, far above where it stood for nearly two centuries, many writers have credited him with contributions to the philosophy of induction which go far beyond this. If my account of Hume's argument is correct, all these claims must be erroneous [1]; and it is interesting to observe, as one frequently can do in conversation with contemporary philosophers, that when Hume is claimed to have proved some important thesis about inductive inference, the thesis almost always contracts, under pressure, to one of non-deducibility, i.e. to inductive fallibilism [2].
If, however, we consider the historical effects of Hume's argument for inductive scepticism, then it will appear that our intellectual debts to it are much greater than they would seem to have been just from what was said in the preceding chapter about the importance of inductive fallibilism.
At the present time, as I have already said, inductive fallibilism has become a commonplace. But less than 100 years ago, it was not only not so, but rather the contradictory thesis was a commonplace, if anything. Thus the author of a manual of `inductive logic' belonging to the predominant school of Mill writes in 1876 that `we can hardly conceive men of science commonly speaking of the most firmly established generalizations of mechanics, optics, or chemistry, simply as conclusions possessing a high degree of probability' [3]. Nor is this an exceptional remark: Fowler certainly speaks for the prevailing philosophy of science, and for the educated common sense, of his time. Even Hume himself, as we have seen, admits that `one would appear ridiculous who would say, that 'tis only probable the sun will rise tomorrow, or that all men must die; tho' 'tis plain we have no further assurance of these facts, than what experience affords us' [4]. Because `'tis [...] certain, that in common discourse we readily affirm, that many arguments from causation exceed probability, and may be received as a superior kind of evidence'.
It will hardly need to be emphasized that what Fowler found hardly conceivable is now so much a commonplace that it is the opposite which is difficult to conceive; and that what Hume was sure would `appear ridiculous' to his contemporaries does not appear so at all to ours. With us, in fact, as the titles of innumerable books and articles testify, the difficulty is to mention induction without mentioning probability in the same breath. There has been, then, in the last 100 years, a remarkable conversion to the belief in the thesis of inductive fallibilism. What brought it about? Inductive fallibilism, as I have said, could be learnt, at any time, from Hume's argument for inductive scepticism. But from what, at this time, has it been learnt, as a matter of historical fact?
With a partial exception to be mentioned later, it is from philosophers that inductive fallibilism is learnt, both by philosophers and by others. And a philosopher at the present time is most likely to have learnt inductive fallibilism from his own reading of Hume, and from Hume's argument for inductive scepticism in particular [5]. Consequently, one could answer the historical question posed above, by saying that Hume's argument for inductive scepticism is the source of twentieth-century inductive fallibilism; and that answer would be true. Yet it would also be very strongly suggestive of something that is not true. For Hume's argument about induction was very far from having this effect either immediately or by a direct channel of influence. Indeed, quite generally, Hume's positive influence on the philosophy of science during the eighteenth and nineteenth centuries was extremely limited. His inductive scepticism, as has already been said, attracted no adherents, and scarcely even any notice, during that time. Even his inductive fallibilism made little lasting headway against the prevailing Newtonian over-confidence. This, and the odium theologicum under which all his philosophy labored, combined to hold Hume's influence and reputation far below what they should have been; especially in England, where both these factors operated more strongly than elsewhere. Now, however, no reputation stands higher, and Hume looms over the philosophy of induction like a colossus. Inductive fallibilism, in particular, prevails all round us. By what indirect channels, then, has it reached us?
There appear to be two main channels to be distinguished. Especially in the second half of the eighteenth century, Hume was a powerful influence on the general intellectual life of France, and one place where his inductive fallibilism, in particular, undoubtedly struck root, was among the writers on probability in that country around 1800 and shortly thereafter. Hume is sometimes supposed to have been a critic---even, by Keynes, for example, an `effective' [6] critic---of the `classical' theorists of probability. This supposition I believe to be groundless [7], but if he was, his criticisms completely escaped their notice. His inductive scepticism, of course, met the same disbelief there as everywhere else. His inductive fallibilism, on the other hand, was welcome and important to the probability theorists; and it is not hard to see why.
Up to the middle of the eighteenth century, the classical theory of probability had remained almost entirely embedded in the consideration of games of chance [8]. That is to say, while the principles of probability could be, and often were, detached from that limited context, and stated quite generally; and the auxiliary mathematics to which the theory had given rise, for example the algebra of combinations, also could be, and often was, stated without even any mention of probability, as it should be; still it was almost always from `problems in play' that the statements of probability, in particular the numerical assessments of particular probabilities, were drawn, and applications provided for the principles. According to inductive fallibilism, however, the non-demonstrative character possessed by most of the inferences which men make in connection with games of chance is shared by all those inferences which men make from the observed to the unobserved in empirical science. If that is true, games of chance, from being almost the whole of the area of application of the theory of probability, become only a small and comparatively unimportant part of that area.
Among the theorists of probability, where the influence of Hume's writings about inductive inference was strong, but his scepticism rejected, one would expect, therefore, that inductive fallibilism would be seen as offering an enormous acquisition of new territory: as bringing within the area of application of the principles, and the mathematics, of probability nothing less than the whole of empirical science. In particular one would expect inductive fallibilism to bestow a novel importance on `Bayes's Theorem', or Laplace's `Sixth Principle' [9] of probability: that principle which was described as enabling us, given the `a priori' probability of a certain causal hypothesis and the `direct' probability of certain observations relative to the hypothesis, to deduce the `inverse' probability of that causal hypothesis relative to those observations. The direct probabilities, of course, would be supplied by the `Theorems' of Bernoulli.
And that, of course, is precisely what did happen. It was the novel claim of the school of Laplace, in the last quarter of the eighteenth century and the first quarter of the nineteenth, that, as Laplace himself writes in the famous Essai, `the entire system of human knowledge' [10] belongs to the province of the theory of probability. Accordingly, as historians of the theory of probability have remarked, there is hardly any inference, inductive or otherwise, the probability of which the writers of this period seem to have thought it beyond the powers of the inverse principle, and Bernoulli's Theorems, to assess [11]. The often-quoted dictum of Quetelet is the confident epitome of this great expansion of claims made on behalf of the theory of probability: `L'urne que nous interrogeons, c'est la nature' [12]. It will be evident that that dictum would simply not be possible except where the prevailing philosophy of science contained, as an absolutely settled tenet, inductive fallibilism.
And it was Hume who furnished the Laplacean school with its philosophy of science. This fact has been remarked upon by other writers, for example von Wright [13]. Nor were the writers of this school themselves backward in acknowledging their debts to Hume; as can be seen from such a manual of the classical theory of probability as that of Lacroix, first published in 1816 [14]. Less than half of this book is occupied with the problems in play which were the staple of the older theory of probability, and all the rest of it bears the impress, constantly acknowledged, of Hume's philosophy of induction (without, of course, his scepticism). This section of the book takes in the whole field of inductive inference under the heading of `Détermination de la probabilité des causes (ou des hypothesès) par les observations' [15]: which determination, Lacroix believes, can in every case be made by Bayes's principle with the aid of Bernoulli's Theorems. The probabilities which these together yield will preserve us, on the one side, from Hume's inductive scepticism; and, on the other, from that dogmatism and inductive over-confidence which is `infiniment plus dangereuse que la première' [16], because we are inclined to it both by reactionary philosophies [17] and by our natural habits of mind [18]. The theory of probability thus provides `un moyen très spécieux de réfuter les excès du scepticisme, sans recourir à ces principes posés a priori [...]' [19]; a `doctrine moyenne, que l'on pourrait nommer scepticisme gradué [...]' [20].
The school of Laplace, then, taking their philosophy of science from Hume, claimed the whole field of inductive inference for the theory of probability. But that means that, for them, the fundamental thing to be learnt from Hume was the discovery that inductive inferences even at the best resemble, in their degree of conclusiveness, the kind of inference which is characteristic, not of mathematics or logic, but of games of chance.
Inductive fallibilism returned, from the theory of probability to the arena of general philosophy, and also from France to England, in a series of waves; all minor, but each a little more insistent than the one before it. It is prominent, as part of an expressly Laplacean philosophy of science, in various writings of Sir John Herschel on probability in the 1850s and 1860s [21]; in W.S.Jevon's Principles of Science in 1874; and in Karl Pearson's Grammar of Science of 1892 [22]. In England it collided with the predominant philosophy of science of the J.S.Mill type, and it was in fact against the inductive fallibilism of Jevons that Fowler was vehemently protesting in the preface from which I quoted above. It is via a Laplace-Jevons channel, then, that inductive fallibilism came to be a definite element, though a subordinate one, in the philosophy of science which was current in the last quarter of the nineteenth century.
There is, however, a second and more important channel by which inductive fallibilism has reached us. Most philosophers now, as I have said, learn inductive fallibilism from Hume himself; but that in turn is possible only because of the influence on us of former members or near-members of the Vienna Circle. It is the Logical Positivist irruption into twentieth-century philosophy which has raised to their present pitch the reputation and influence of Hume's philosophy of science.
Now, in its turn, the inductive fallibilism of the Logical Positivists was no doubt derived in large part from Hume himself, perhaps partly via Mach. But it was not altogether so derived. A proof of a judgement of invalidity---that is, of the possibility of the premiss of an argument being true and its conclusion being false---is one thing when it is from premisses purely a priori: and that is what Hume provided, concerning inductive arguments. But for bringing a judgement of invalidity home to men's bosoms, there is nothing so effective as a proof of it by an actual counter-example, or what is taken to be such; a case, that is, of the class of arguments in question, in which the premiss is true and the conclusion false. Now, for men of the circle of Carnap, Popper, and Wittgenstein, the latter was what Einstein provided. For Einstein satisfied them, as well as most others competent to judge that some of the Newtonian generalizations were false in fact, notwithstanding the truth of all the previous empirical evidence in their favor; thus giving to philosophers and scientists a reminder, of the most striking kind, of the fallibility of even the best-confirmed of scientific generalizations.
It might seem, then, that we ought to say that at least half of the present currency of inductive fallibilism is owing, not to the influence of Hume's philosophy, but to the influence of Einstein's science (and, of course, of the continuing turbulent state of physics since 1905). And so we ought, were it not for the fact that Einstein himself several times tells us that it was Hume (and Mach, but in respect of original thought that means Hume), who woke him from his dogmatic slumbers; who taught him that no scientific generalization can be, and in particular the absolute character attributed by Newton to time is not, necessitated by any empirical evidence [23].
This is probably the more important channel. What the other (Laplace-Jevons) channel contributed, in net effect, was just that inductive fallibilism was a thesis of some currency in the philosophy of science before the world was astonished by the fall of the Newtonian empire. It merely ensured that then, when men looked around, as they will always do in a time of intellectual crises, for something in the philosophy of science which can afford some consolation after the event, and which would even have prepared them for the event if they had adopted it in advance, they found it ready to their hand, in a general inductive fallibilism.
Thus, if there is nowadays---and there is---a cooling jet of inductive fallibilism which plays constantly on scientific confidence, preventing it from overheating, that jet has reached us by two main channels. Both are indirect, and both traverse more European than British territory. But both take their rise in Hume's argument for inductive scepticism. That argument, therefore, as well as being a permanent potential check to scientific inductive over-confidence, can fairly be said to have actually administered the fallibilist corrective which, as a matter of historical fact, Newtonian over-confidence received near the beginning of this century.
It is worth observing, as a consequence, that in one respect Hume's philosophy of induction has acted in its own despite. For Hume thought that the inductive confidence of men could not be destroyed, or even weakened, by philosophical argument; but history, if my account has been correct, attests that Hume's argument for inductive scepticism has weakened scientific inductive confidence. It has not weakened it, of course, to the point of inductive scepticism; nor has it weakened organic inductive confidence at all, even to the point of fallibilism. But considering how recent, complete, and widespread, is the conversion to inductive fallibilism which that argument has effected, some apology is due to D.C.Williams, for the derision which greeted his suggestion, in 1947 [24], that inductive scepticism could conceivably spread from the philosopher's study to become a popular, and even a political, force.
When a truth becomes very well and widely recognized, there is always at least some tendency to trivialize it, by incorporating the predicate of it into the very meaning of the words which denote its subject. In the last thirty years this tendency has become quite pronounced in connection with inductive fallibilism, and this is what prevents the word `inductive', or any other translation of Hume's `probable', applied to arguments, from being a completely satisfactory translation at the present time. The tendency has, as I have earlier remarked, been carried to its conclusion in Carnap's usage of `inductive', and we will shortly meet with an even more remarkable instance of the pressure of inductive fallibilism on our language. It is somewhat ironic to reflect that Hume, as we have seen, apologized for having, in the bulk of Book I Part III of the Treatise, called `probable' certain arguments which, although they are confessedly from observed to unobserved instances, are `in common discourse' allowed to `exceed probability, and [...] receiv'd as a superior kind of evidence'. Ironic, because Hume's inductive fallibilism has triumphed so completely over `common discourse', that to say `'tis only probable the sun will rise tomorrow, or that all men must die', has now become (just because `we have no further assurance of these facts, than what experience affords us'), not only not `ridiculous', but a commonplace, and is even in a fair way to being made a triviality!
Hume's argument was in essence from inductive fallibilism (9) and deductivism (6), to inductive scepticism (8); and that argument is valid because, although any one of these propositions is consistent with the negation of another, the first two of them are inconsistent with the negation of the third.
For the same reason, inductive fallibilism conjoined with the denial of inductive scepticism, requires the denial of deductivism. Now, I have tried in this and the two preceding chapters to show the following: that Hume's inductive scepticism has always been, and with minor exceptions still is, denied, while his inductive fallibilism has met with wide and deep acceptance during, though not before, the twentieth century. That being so, consistency requires the denial of deductivism. But the denial of deductivism entails (as we saw in Chapter 6 section (i)) that sometimes two arguments, though both invalid, are of different degrees of conclusiveness; which is the fundamental thesis of the twentieth-century theory of logical probability. If, therefore, our philosophical beliefs are consistent, the currency in the twentieth century of inductive fallibilism, combined with the almost equally widespread rejection of inductive scepticism, would have been bound to bring into being, and even into prominence, the theory of logical probability.
And that, of course, is precisely what has happened. The process of adjusting our philosophy towards consistency has been slow and confused, and still is, as we saw in Chapter 6 section (ii), very far indeed from being complete. But the theory of logical probability did come into being just when inductive fallibilism first struck wide and deep root among philosophers and scientists, viz. in the first quarter of this century. Its development has gone on pari passu with the entrenchment of inductive fallibilism. And that development has been chiefly at the hands of some of the most consciously Humean of all inductive fallibilists, viz. the Logical Positivists. All of which is exactly what was to be expected from men who are unwilling to accept inductive scepticism, who are for the first time compelled to embrace inductive fallibilism in earnest, and who wish to make their philosophy of science consistent.
But this means that, in a historical sense, the twentieth-century theory of logical probability itself is something we owe to the influence of Hume's argument for inductive scepticism. This effect, to say the least, was not one which Hume intended; yet undoubtedly he produced it. He compelled us to separate the concepts of the reasonableness, and the validity, of an argument, by pointing out in the clearest manner possible the intolerable sceptical cost, in the case of inductive arguments, of not doing so. He laid down his intellectual life, so to speak, for deductivism, in order that the theory of logical probability might live. Hume ought therefore to be regarded as the patron saint of the twentieth-century theory of logical probability.
It should not be thought this piece of homage to Hume's argument for scepticism is a contrived or hypocritical one. On the contrary, the effect which I have just described his argument as having in the present century is not even the first effect of the kind which that argument has had. Early in the nineteenth century in France, as we have seen, Hume was honored, because, qua inductive fallibilist, he was the conqueror of new worlds for the classical theory of probability. It is essentially the same causal sequence which has been re-enacted, only in the wider arena of general philosophy, early in the twentieth century; and the school of Carnap ought not to be more backward, in acknowledging their debt to Hume's argument for inductive scepticism, than was the school of Laplace.
If we consider the argument, `No one in New York is able to cure every illness in less than two minutes, all physicians are able to do so, so, there are no physicians in New York", we may well agree with what Professor Paul Edwards says about it in a well-known article [25]. Which is, that its `sceptical' conclusion is deduced from one premiss (the first) which is a truth everyone knows, and a second which everyone knows to be false. Or rather, if the word `physician' is used in the ordinary sense in `All physicians are able to cure every illness in less than two minutes', that premiss is obviously false. If not, that premiss is true, but only in the sense that a stipulative definition, however arbitrary, is true; in that case the second premiss is just an idiosyncratic `high redefinition' [26] of the word `physician'. For there are certainly two concepts---`physician' in the ordinary sense, and this concept with the additional requirement of being able to cure every illness in under two minutes---which no user of ordinary English is likely to confound; but which are confounded by someone who thinks that, by arguing as above, he refutes the common-sense belief in the existence of physicians in New York.
Now, suppose it were suggested that Hume's argument, `All inductive inferences are invalid, all invalid inferences are unreasonable, so, all inductive inferences are unreasonable', is open to the same objections. The suggestion would be that its sceptical conclusion, too, is deduced from premisses, the first of which (i.e., inductive fallibilism) is a truth that everyone knows; and the second of which (i.e., deductivism) everyone knows to be false. Or at least (it would be said), deductivism is obviously false if the words `reasonable inference' occur in it in their ordinary sense. If they do not, then deductivism is simply an idiosyncratic high redefinition of `reasonable inference'. For (the suggestion would run) there certainly are two concepts---`reasonable inference' in its ordinary sense, and this concept with the additional requirement of being valid---which no user of ordinary English is likely to confound; but which are confounded by someone who thinks that, by arguing as Hume did, he refutes the common-sense belief in the reasonableness of inductive inference.
It will be evident that this suggestion, if what I have said in this chapter and the two preceding ones is true, offers us a new way to pay old debts, and one with the well-known advantages which theft has over honest toil. But the suggested comparison between the two arguments would fail at every point.
Inductive fallibilism is so far from being a truth which everyone knows, that it is a truth which, at the `organic' level, no one knows. And even at the level of high scientific culture, the full recognition that scientific inference, like gambling and unlike the arguments of Euclid, contains an ineradicable element of risk, has been only a very late comer into consciousness.
Deductivism is not an idiosyncratic high redefinition of `reasonable inference'. It is a logico-philosophical thesis; and one of long, wide, and deep currency, at least among philosophers; which is still the unstated assumption behind much assessment of the conclusiveness of inferences, both by philosophers and by others; and which to this day has been expressly denied by almost no philosopher.
Still further from the truth, and even scandalous, would be the suggestion that the truth of inductive fallibilism, and the falsity of deductivism, are known to every one just in virtue of a command of ordinary language. Ordinary English does equip its users with two palpably distinct concepts, `physician' and `physician able to cure every illness in less than two minutes'; but it simply is not true that it also equips its users with two palpably distinct concepts, `reasonable inference' and `valid inference'.
It is indeed true, as Edwards says, that `part of the definition of "inductive inference" is inference from something observed to something unobserved'. But it does not follow from that, as Edwards says it does, and it is not true, that `it is a contradiction to say that an inference is both inductive and at the same time [...] deductively conclusive' [27]. That conclusion no more follows from that premiss, than it follows from an inference's being from the singular to the non-singular, say, or from the contingent to the non-contingent, that it cannot also be valid. To suppose that it does follow, is simply to testify to one's confidence in the truth of another premiss, which is needed to make it follow, but which as we have seen is by no means a trivial one, viz. inductive fallibilism. We see in Edwards, therefore, an instance even more remarkable than that which Carnap furnishes, of the tendency to mistake the truth of inductive fallibilism, for the recognition of which we are chiefly indebted to Hume's philosophy, for a trivial transcript of part of the meaning of the word `inductive'.
The extent (which is great) to which inductive fallibilism is known now to be true; the extent (which is much less) to which deductivism is now known to be false; and the extent (which is still small) to which these truths have become incorporated in ordinary language: all these, if what I have said is true, are due to the influence of Hume's argument for inductive scepticism. If, therefore, someone were to suggest that that argument is open to the same objections as the above sceptical argument about physicians, the situation would be as follows. That the man who taught everyone to see empirical science as incurably fallible, would be being blamed for affirming the truth which every one knew before, and knew in virtue of a mere command of ordinary language; and that the man who compelled us to separate the concepts of `valid' and of `reasonable inference', would be being reproached for neglecting this distinction which every user of English is alleged to be conscious of. We should have the exquisite irony that the effects of Hume's philosophical triumph over `common discourse' would be being mistaken for wisdom which ordinary language itself gives gratis to all its users. It would be a scandalous repudiation of debts fairly contracted, if a philosopher were to compare Hume's argument with that about the physicians.
It was with an argument of Russell, not of Hume, that Edwards compared the argument about physicians. He therefore may not be exposed to the reproaches I have conditionally made. Writing when he was, Russell should perhaps not have made, as he did, the deductivist assumption. Yet it comes as something of a shock to be reminded that Edwards was writing, in 1949, about a book published in 1912. That date was distinctly before the full tide of inductive fallibilism had arrived even at Cambridge: a fact of which anyone can satisfy himself by reading the opening pages of the first of Broad's famous articles of 1918 on `The Relation between Induction and Probability' [28]. One is entitled to wonder, then, whether Edwards's treatment of Russell was much less unfair than it would have been if applied to Hume. In any case, as Edwards was of course well aware, Russell was simply repeating the essentials of Hume's argument for inductive scepticism. Besides which, Edwards's article confessedly was just one application of a general method for liquidating debts which might be supposed to be owing to great philosophers, by portraying their relation to ordinary language as almost exclusively that of parasite to host. There are grounds, then, for thinking that Edwards did intend to compare the argument about physicians with Hume's argument about induction; and if so, then what was said conditionally above does apply to him.
[1] Some of the more extravagant claims made on Hume's behalf, by writers during the last fifty years, are collected and briefly discussed in the Appendix section (iii).
[2] Theses of non-deducibility, i.e. judgements of invalidity, can fairly be said to have been Hume's forte in general. In ethics, for example, the most important proposition which is associated with his name is `P(h, e) < 1 for all factual e and ethical h'.
[3] T.Fowler, The Elements of Inductive Logic (Oxford), preface to the third edition, p.xiii. This work `designed mainly for the use of students in the universities' (subtitle), pretends to no originality or to any other high intellectual distinction. But of course not its quality, only its representative character, is what concerns us here.
[4] Both this and the following quoted sentence are from the long paragraph on p.124 of the Treatise which was quoted in full in Chapter 2 section (iii) above.
[5] Along with Hume's variants of this argument. Cf. Appendix section (i).
[6] A Treatise on Probability, p.83.
[7] Cf. Appendix sections (ii) and (iii).
[8] Certainly up to the third edition, in 1756, of de Moivre's Doctrine of Chances, or a method of calculating the probabilities of events in play.
[9] Essai philosophique sur les probabilités (Paris, 1814), Ch.III.
[10] A Philosophical Essay on Probabilities, trans. Truescott and Emory (Dover, N.Y., 1951), p.2.
[11] Cf. Keynes, Treatise, Chs.VII, XVIII, XXX; Todhunter, History of the Mathematical Theory of Probability (Cambridge, 1865), Chs.XVI-XX.
[12] Quoted, e.g. by Keynes, Treatise, p.428. But I am unable to say where Quetelet wrote this.
[13] Cf. The Logical Problem of Induction (2nd edn., London, 1957), p.220 n.1.
[14] S.F.Lacroix, Traité élémentaire du calcul des probabilités. (My page-references are to the 3rd edn., Paris, 1833).
[15] Ibid. p.172.
[16] Ibid. p.181.
[17] Ibid.
[18] Ibid. pp.62--3, 181, 185.
[19] Ibid. p.299.
[20] Ibid. p.181. Lacroix's italics.
[21] e.g., the article on Quetelet in the Edinburgh Review, 1850. (This article is sometimes erroneously attributed to the father, Sir William Herschel. All of the articles on probability attributed to that writer by Keynes (Treatise, p.445), for example, are chronologically impossible. William Herschel died in 1822).
[22] I neglect, as marginally related to the philosophy of science, though in different degrees, the influence of Laplace on such disparate thinkers as Augustus de Morgan and Francis Galton. A reminder is perhaps also needed, when one mentions such a book as Pearson's, that as with Fowler on the other side of the question of inductive fallibilism, quality of thought in not in question, only historical fact.
[23] Cf. e.g. Albert Einstein: Philosopher-Scientist, ed. Schilpp (Harper Torchbook edn., New York, 1959), i.53.
[24] In The Ground of Induction, pp.15--20.
[25] `Bertrand Russell's Doubts about Induction', reprinted in Logic and Language (First Series), ed. Flew (Blackwell, Oxford, 1952).
[26] Ibid. p.60.
[27] This and the preceding quotation, ibid. pp.68--9. Edwards's italics.
[28] Mind, vol.27.
[ Previous: Chapter 7 | Next: Chapter 9 | Table of Contents ]