Wittgenstein on Scepticism and Certainty

Written by AC Grayling.

Wittgenstein's 'On Certainty' (hereafter OC) is a collection of provisional notes, recording a journey not an arrival . But it is not difficult to see an intended destination for the journey, nor is there anything obscure about the territory being travelled. Yet OC has some surprising and unexpected features. For one thing, it recapitulates certain old attitudes in Wittgenstein, harking back to, but making different use of, Kantian traces in the 'Tractatus', here in the form of a roughly sketched (and possibly naturalistic) anti-realism similar in striking respects to Kant's empirical realism. For another thing it appears to represent Wittgenstein's acceptance, at last, of philosophy's legitimacy as an enterprise. In all his earlier work he explicitly premissed the claim that philosophy is a spurious enterprise, arising from misunderstandings about language. In OC he takes a central, traditional philosophical problem–the problem of scepticism and knowledge–and tries to formulate a refutation of scepticism, and a characterisation of knowledge and its justification. And he does this by engaging with another attempt to do so, namely, Moore's.

In order to evaluate the ideas it contains I shall therefore take OC at face value–as an unfinished enquiry, the ideas in which nevertheless strongly indicate the finished theses it works towards–and proceed as follows.

First, there are two main themes in OC, which are, at the least, not comfortably consistent with each other. One is a reply to scepticism, and as such contributes recognisably to the theory of knowledge. Indeed it is a reinvention almost from scratch of views familiar, and usually more fully argued, elsewhere in philosophy, of a broadly foundationalist stamp. In this respect it carries forward, or unfolds, themes already suggested in the Philosophical Investigations (henceforth PI). Alongside the first theme–or more accurately, wrapped round it as a vine about a tree–is the other, not comfortably consistent, theme, a relativistic one which undermines the claims constituting the first theme. After stating each theme I discuss the tension between them, suggest the best way out of it, and indicate how OC itself, and materials from PI, affords Wittgenstein's own different basis–a fudged one–for resolving the tension.

Wittgenstein's conceptions of doubt, certainty and knowledge, his persistent conflation throughout OC of contingent propositions with those he identifies as 'grammatical' propositions, and his revealing conflation of scepticism with idealism, are central to understanding the themes of OC, and I discuss them in their due places, concluding with an overall evaluation.


My exegetical task is effected by suitably anatomising OC. The view I shall call OC1 and which constitutes a version of a foundationalist refutation of scepticism, and therefore a contribution to the theory of knowledge, has two components, the first of which is that scepticism is answered by appeal to the fact that beliefs inhere in a system, and the second of which is that this system of beliefs rests on foundations which give those beliefs their content. Here are some passages exemplifying the first component of OC1 (all emphases are Wittgenstein's):

83. The truth of certain empirical propositions belongs to our frame of reference (WR249).

88. It may be for example that all enquiry on our part is set so as to exempt certain propositions from doubt, if they are ever formulated.

94. But I did not get my picture of the world by satisfying myself of its correctness; nor do I have it because I am satisfied of its correctness. No: it is the inherited background against which I distinguish between true and false. 105. All testing, all confirmation and disconfirmation of a hypothesis takes place already within a system ... The system is not so much the point of departure, as the element in which our arguments have their life.

162. I have a world picture. Is it true or false? Above all it is the substratum of all my enquiring and asserting (WR252).

341. The questions that we raise and our doubts depend on the fact that some propositions are exempt from doubt, are as it were like hinges upon which those turn.

Here are some passages exemplary of the second component of OC1:

103. And now if I were to say "It is my unshakeable conviction that etc.", this means in the present case too that I have not consciously arrived at the conviction by following a particular line of thought, but that it is anchored in all my questions and answers, so anchored that I cannot touch it.

162. I have a world picture. Is it true or false? Above all it is the substratum of all my enquiring and asserting (WR252).

411. If I say 'we assume that the earth has existed for many years past' (or something similar), then of course it sounds strange that we should assume such a thing. But in the entire system of our language-games it belongs to the foundations. The assumption, one might say, forms the basis of action, and therefore, naturally, of thought.

512. Isn't the question this: 'What if you had to change your opinion even on these most fundamental things?' And to that the answer seems to me to be: 'You don't have to change. That is just what their being "fundamental" is.'

599. To say: in the end we can only adduce such grounds as we hold to be grounds, is to say nothing at all.

OC1 thus states that scepticism gets no purchase because our beliefs inhere in a system (the first component) which rests upon foundations (the second component), which latter non-negotiably constitute the conditions upon which our beliefs have contentÑand which therefore constitute the conditions even for doubting, which, therefore again, cannot take the foundations for their target. The justification for the foundations is thus effected by a "transcendental argument" : restated, it is that foundational beliefs (expressed by what Wittgenstein calls, in senses of 'logical' and 'grammatical' special to OC, logical or grammatical propositions; see e.g. 51, 56-8) are what make the system possible, and it is within the system that claims to knowledge and challenges of doubt are alone intelligible. A clever encapsulation of the transcendental argument is given at 248: 'I have arrived at the rock-bottom of my convictions. And one might almost say that these foundation-walls are carried by the whole house.'

The view I shall call OC2 and which is not comfortably consistent with–perhaps, indeed, undermines–OC1, is to be found in paragraphs 65, 95-9, 166, 174, 192, 211 (WR254), 253, 256 (WR257-8), 307, 336 (and compare 559)–and perhaps also in paragraphs 5, 33, and 607. Here are some exemplary passages:

65. When language-games change, then there is a change in concepts, and with the concepts the meanings of words change.

95. The propositions describing this world-picture might be part of a kind of mythology ...

97. The mythology may change back into a state of flux, the river-bed of thoughts may shift.

99. And the bank of the river consists partly of hard rock, subject to no alteration or only to an imperceptible one, partly of sand, which now in one place now in another gets washed away, or deposited.

166. The difficulty is to realise the groundlessness of our believing.

256. On the other hand a language-game does change with time.

336. But what men consider reasonable or unreasonable alters.

OC2 is relativism. Relativism is the view that truth and knowledge are not absolute or invariable, but dependent upon viewpoint, circumstances or historical conditions. What is true for me might not be true for you; what counts as knowledge from one viewpoint might not do so from another; what is true at one time is false at another. Paragraph 97 arguably shows that the relativism implicit in this aspect of OC is of a classic or standard type. Its presence in OC is entirely consistent with its presence elsewhere in the later writings: one remembers the lions and Chinese of PI. What was left open in those earlier relativistic remarks was the degree of strength of the relativism to which Wittgenstein was committed. OC2 constitutes a claim that the framework within which claims to knowledge and challenges of doubt equally make sense is such that its change can reverse what counted as either. That is classically strong relativism.


To get a good feel for the tension between OC1 and OC2, compare 103 (where a given belief 'is anchored in all my questions and answers, so anchored that I cannot shake it') with 97-9 ('the river-bed of thoughts may shift'); 494 with 256; both 512 and 517 with any of the relativistic remarks cited, for example 559; and any of the relativistic remarks with 317 and 599, which latter is worth repeating here: 'To say: in the end we can only adduce such grounds as we hold to be grounds, is to say nothing at all'.

At one level, OC1 and OC2 can of course be so interpreted as to make them consistent. One can postulate foundations that are historically and in other ways parochial to the discourse under consideration, consisting in beliefs and principles which are basic in the OC1 sense for a given discourse, and not justified independently of it; but which are not immutable or absolute but as vulnerable to change, even if more slowly and circumstantially, as any of the ordinary beliefs comprehended in the framework. This precisely seems to be the import of Wittgenstein's river-bed metaphor: the river-bed is only relatively stable with respect to the water flowing over it, because it is worn away with time, and shifts its course.

But a relativistic foundationalism renders OC1 superficial as a response to scepticism, because so construed it does not begin to meet the really serious problem scepticism poses, and of which Wittgenstein is perfectly aware (see e.g. 14-16). To see what that is, one must retrace some steps.

Let us simplify the model we are working with. A sceptic challenges us to justify a particular empirical belief, for example that there is a book on the table here before us. We respond, exploiting the same resource for doing so as OC does, by saying in effect that these circumstances are such and those words mean such that this is tantamount to a paradigmatic circumstance for using those words in these circumstances–that is, for claiming that there is a book on the table. The sceptic pushes his point, invoking considerations about non-standard perceptual phenomena and other psychological contingencies, including error; at which point we change gear and invoke countervailing considerations about the framework of the discourse (the system of beliefs constituting it; the 'conceptual scheme') by stating the assumptions upon which not just the claim, but also the challenge to it, make sense. And at this level of sceptical challenge, that has to be enough: justifications in ordinary discourse come to an end at this point.

But now the sceptic mutates; he becomes a different and bigger monster. He is no longer interested in hearing what we have to say about the book on the table, but in what we have to say about the framework, the system of beliefs. What justifies our acceptance of the framework, or (more weakly) our employment of it? What if there were another framework, or other frameworks, in which different assumptions led to different outcomes with these words and these circumstances? And so on. The sceptic, in other words, has adopted the habiliments of relativism. Relativism, indeed, is the ultimate form of scepticism, because it challenges us to justify, as a whole, the scheme within which mundane judgments get their content and have their life.

The answer which says: 'this is the scheme we have; it is a bare given that we have it', and which might–but this is a different thing–add, 'and of course there might be others', and–yet a further and a bigger step again–'we might never know what these other schemes are like or even that they exist', is unsatisfactory, at very least as the first response to relativism. One might end by responding with (in this emergency: preferably) the first step just described, after a long haul; but there are strong anti-relativist arguments to evaluate first, which, if they turn out to be plausible and stable in the face of challenge, provide a powerful way of blocking scepticism altogether. It would seem that for the argument of OC1 to work as a refutation of scepticism, this stronger recourse is required. Without it, knowledge and truth are concepts parochial to the scheme; they are not knowledge and truth but 'knowledge-in-the scheme' and 'truth-in-the scheme', as 599 says. (Think of being told that 'Jarndyce is a lawyer' is true "in" Bleak House' is the only kind of exemplification the concept of truth has.) There are good reasons for thinking that the assumptions constitutive of the framework have to be undischargeable. These reasons are drawn from seeing how the notion of alternative frameworks collapses under direct pressure, the terminus of the argument being that anything recognisable as a framework has to be identical in fundamental respects with the framework from which it is so recognised. The argument is adapted from a familiar one offered by Davidson, who writes of "conceptual schemes" rather than frameworks, and individuates them as sets of intertranslatable languages .

The argument, familiarly enough, is as follows. The relativist proposal is that there can be languages which we might not be able to recognise as such–which, that is, we cannot translate. But how do you recognise a language as such if you cannot translate it? The problem can be stated in terms of what the only plausible candidate for a criterion of languagehood can be, namely, translatability into a familiar idiom. Since language-use involves 'a multitude of finely discriminated intentions and beliefs' which we could not attribute to someone unless we could understand his speech, we can only recognise the presence of such intentions and beliefs if translation is possible. Moreover, if it turns out that there are differences between our and the alien's beliefs, this will be courtesy of a shared background of beliefs which makes the differences apparent. Differences are more meaningful when there are fewer of them; when there are few against a shared background of belief, the differences are more of opinion than conceptualisation–they relate to variances in the scheme's superstructure, which tolerates conflicts of view in (for characteristic examples) politics and taste, while still locating them in the same world. Since the cognitive foundations of the scheme have to be shared for these more entertaining differences to be possible, the conclusion is that conceptual relativism is incoherent. (Davidson takes this to mean that the very idea of a framework is empty because it implies what the argument denies, namely, the possibility of real conceptual diversity.)

Underlying such arguments, very interestingly, is an implicit commitment to the controversial view that possibility is an epistemic notion, that is, that possibility is conceivability. Something is a possible state of affairs (a possible past fact, a possible language or scheme) only if it is constructible from actual states of affairs (from what we know, from the language we speak). The intended contrast is this: on the idea that possibility is a purely logical notion, denoting mere absence of contradiction, the number of possible worlds there can be other than the actual world is infinite. But on the idea that possibility is an epistemic notion, denoting graspability in thought (or translatability into a familiar idiom), the number of possible worlds other than the actual world is limited by accessibility relations between them and it. But where there are such relations, the idea of a world being in some strong sense different from this one loses its grip. Standard ways of defining possible worlds involve redistributing truth-values over the propositions constituting the world-book of the actual, or increasing their number by adding other propositions consistent with them. But to do this is to redescribe this world, not–except by courtesy of the phrase–to create new worlds.

These considerations rule out relativism. They therefore rule out OC2. There is no other way of taking OC2 than as a seriously strong relativist argument ('the river-bed of thoughts may shift' ... 'a language-game changes with time'). In the ideal state of things, therefore, OC1's offer of a response to scepticism is elected to stand, and OC2 is ditched. But as the text of OC was left to us, Wittgenstein was developing arguments for both, so the next question is: is there any way they could be made to reconcile, further up the road where their parallels meet?


The destination available to Wittgenstein in the light of the tension between OC1's need for an anti-relativistic resource, and OC2's undermining of this, is one made familiar by his treatment of the request for justifications in PI. It is to say: justification must come to an end: "my spade is turned". In PI this seemed to offer a form of foundationalism in which the basis–the given, that which justifies itself by being what it is–is practice: and moreover shared practice, which in its essentially mutual character is constitutive of the content (so, in the case of language, the meaning) of what is based upon it. This indeed is Wittgenstein's resource: see 7, 92, 110, 116, 196 (WR253), 229, 559; and perhaps also 232, 219, 344 and 378. Does this do, as a somewhat fudging way out of the problem?

I think not, because the PI turned-spade thesis is considerably weakened in OC by the degree of relativism OC2 constitutes. Of course there are relativistic noises in PI: such claims as that we would not understand a speaking lion if we met one, and that we no more understand Chinese facial expressions than words, have that tendency, because they are premissed on the lack of the shared form of life which makes understanding possible. But these relativities could be reducible–nothing implies that we cannot gain entry to the alien forms of life, that is, that we can find ways of translating lionese remarks and Chinese expressions upon doing so. Reducible synchronic relativities look very like familiar cultural differences, and hence are superficially relative only. But the idea that the foundations of sense are themselves merely relative–that the bed and banks are in constant process of erosion–implies a greater insecurity. Consider a relativist thesis like Feyerabend's, say, in which change in assertion-conditions entails change of sense. A different way of calibrating thermometers on his view changes the meaning of "temperature". If the bed and banks of discourse were shifting over time, meanings would change with them. But we would be in the position of a speech community whose meanings are shifting without our realising the fact, because agreements remain. (The rules change, but we all keep observing them in common as they do so. This falls foul of Wittgenstein's own rule-following considerations. The whole community is in the dilemma of the solitary would-be language user, who cannot tell the difference between following the same rule again, and only thinking he is doing so.)

A different and better way out of the problem is to suppose that Wittgenstein might have developed his conflicting lines to the point where the conflict became intolerable–I would say: where he recognised the unhealthy mixing of contingent and framework propositions in his examples, which constantly seduced him into thinking relativistically: more on this shortly. And then he just might have preferred the strong anti-relativist argument available in the line he was himself taking in OC1 on the grounds of sense. For in that aspect of his discussion he in effect reinvented the strategy, as noted, of employing a transcendental argument to show that sceptical challenge is defeated by appeals to the framework. Why not therefore see that the transcendental argument militates equally against relativism?

But if one does not supplement the response to scepticism (OC1) by some such strategy, the exercise in OC is at best partial, at worst self-defeating, with the self-defeat stemming from acceptance of OC2. As OC stands, it stands defeated in just this way, for it only deals with scepticism at the lower, less threatening level, and fails to recognise that scepticism in its strongest form is, precisely, relativism.

There are hints in OC of an alternative better way out: namely, some version of naturalism–in Hume's, not Quine's, sense; that is, as appealing to natural facts about our psychological make-up (not, as in Quine, as appealing to the deliverances of current theory in natural science: although the latter form of naturalism takes itself to absorb the former). See 287: 'The squirrel does not infer by induction that it is going to need stores next winter as well.' This hint is strengthened by 505: 'It is always by favour of Nature that one knows something' and the paradigmatically Humean 277: '"I can't help believing ...".' If one re-reads the practice-cum-form-of-life entries in the light of these–a twist of the kaleidoscope–a plausibly naturalistic thesis comes fully into view.

../page 2