Scepticism and Irrationalism

1.1 Introduction

The scepticism to which I will try to provide an answer in this book is neither Academic scepticism, which said that there is but one thing one can know, namely that one can know nothing else, nor Pyrrhonian scepticism, which said that one cannot even know that, nor Cartesian scepticism, which said that even logic and mathematics are dubious, but Humean scepticism. Humean scepticism allows that each of us can have a good deal of egocentric knowledge about our own beliefs, feelings, and perceptual experiences. It also allows that logical truths can be known. But it denies that one can progress by logical reasoning from perceptual experience to any genuine knowledge of an external world, if there is one. Since its target is not all knowledge, but only knowledge of the external world, Humean scepticism is not self-undermining; it does not exclude the possibility that one can know (perhaps on logical grounds) that there can be no genuine knowledge of an external world.

Humean scepticism seems to follow from the conjunction of the following three propositions, which I will number since I will often be referring back to them in the course of this book:

(I) there are no synthetic a priori truths about the external world;

(II) any genuine knowledge we have of the external world must ultimately be derived from perceptual experience;

(III) only deductive derivations are valid.

I will also refer to these three propositions as, respectively, the anti-apriorist thesis, the experientialist thesis, and the deductivist thesis. They entail that for any factual statement h to constitute knowledge, there must exist true premises e that report perceptual experiences and from which h is logically derivable. But if h speaks about the external world and e speaks only of perceptual experiences, h goes beyond e and therefore cannot be logically derived from e.

1.11 Antisceptical Strategies

I will now attempt to list the main strategies that have been adopted in response to this simple and seemingly cogent argument. These strategies fall into two groups: those that deny one or other of the above three propositions, and those that try to cope with scepticism while retaining all three. I will begin with the former.

Apriorist strategy: deny proposition (I), modifying proposition (II) accordingly, and claim that there are certain truths about the structure of the world (for instance, that the Law of Causality reigns throughout it) that are not analytic but can be known a priori, being necessarily true. This of course was the essence of Kant's answer to Hume.

Transcendental argument strategy: modify proposition (II) and claim that there are certain principles (for instance, a Principle of Induction) that are not themselves established in the ordinary inductive way but can be proved by an argument of the form: This principle must be true if scientific knowledge is to be possible; but scientific knowledge exists; therefore this principle is true'.

Conjecturalist strategy: deny proposition (II) and claim that scientific knowledge about the external world is not derived from experience but is essentially conjectural and is only negatively controlled by experience. This of course is the essence of Popper's answer to Hume.

Nondeduetivist strategy: deny proposition (III) and claim that an inference does not have to be deductive for it to be valid.

I now turn to strategies that do not (or need not) involve the denial of any of those three propositions.

Probabilist strategy: retain proposition (III) but add that probability logic, or the logic of partial entailment, is a legitimate extension or generalisation of classical logic, and that in the case of a hypothesis h that is not entailed by evidence e we may be able to establish, with the help of probability logic, that h is more or less strongly confirmed by e. This of course is the essence of the solution of the problem of induction offered by Keynes, Jeffreys, Reichenbach, Garnap, Hintikka, and others.

Phenomenalist strategy: close the alleged gap between perceptual experience and hypotheses about the external world by claiming that physical objects just are clusters or sequences of actual or possible perceptions. Mach took this position; so did Russell in his (1914) and Carnap in his (1928).

Vindicationist strategy: vindicate the inductive method by showing, not that it will work well, but that, if any method for predicting the future will work at all, then an inductive method will work at least as well as any other. Reichenbach, Feigl, and Salmon have taken this position.

Pragmatist strategy: concede that there is no purely epistemological answer to scepticism, but add that we may nevertheless accept and reject hypotheses rationally, in the light of our utilities.

Naturalist strategy: concede that there is no epistemological answer to scepticism, but add that scepticism is of only academic interest because nature has endowed us with a cognitive machinery that is too robust and vigorous to be affected by it. Thus we can make our peace with scepticism, acknowledging it in philosophical moments and forgetting it when we hang up philosophy and go out to dine or play backgammon. This was, of course, Hume's "answer" to his own problem. Pascal had previously taken a rather similar view.

Of course, not all of these strategies are mutually exclusive. The transcendental argument and apriorist strategies may be combined, as they were by Kant; so may the probabilist and phenomenalist strategies, and so may the nondeductivist and naturalist strategies. However, I would like to insist that two strategies that are often bracketed together, namely the nondeductivist and probabilist, should be regarded as quite distinct. There are versions of probabilism that come rather close to nondeductivism; I am thinking of those subjective or personalist versions that allow us to distribute initial probabilities according to taste. But the kind of probabilism with which we will be concerned tries to provide a nonarbitrary and impersonal measure of both initial and relative probabilities. A system of probability logic that is to achieve this has to be perfectly explicit. The contrast with nondeductivism could hardly be more glaring. The latter does not try to generalise classical logic: it postulates a peculiar, nondeductive relation between statements whereby e can provide inductive proof for h without entailing it; and far from being perfectly explicit, nondeductivists maintain a sealed-lips policy concerning the nature of this mysterious relation.

There...