Wittgenstein’s notes, collected as On Certainty, are a gold mine of ideas for philosophers concerned with knowledge and scepticism.1 But in my view, Wittgenstein’s approach to scepticism is still not well understood. Obviously, a short essay is no place for an exhaustive treatment of Wittgenstein’s anti-sceptical ideas. Instead, I shall present a reconstruction of a particular argument that I call ‘Wittgenstein’s Refutation of Idealism’. This argument is developed in the first sixty-five sections of On Certainty, although there is a later passage (OC 90) that must also be considered.
To appreciate this argument, it is essential to be clear about its target. It is evident that Wittgenstein’s thoughts on scepticism are prompted by Moore’s ‘Proof of an External World’ (Moore 1939) and ‘A Defense of Common Sense’ (Moore 1925). But these well-known papers differ in an important way. In his ‘Proof’, Moore’s topic is external world scepticism in its most general form: his aim is to prove, in defiance of the sceptic and idealist, that external objects, defined as ‘things to be met with in space’, really do exist. By contrast, in his ‘Defense’, Moore undertakes to defend a body of rather more specific beliefs: that the earth has existed for many years past, that he has never been far from its surface, and much else besides. To be sure, Wittgenstein is deeply interested in both of Moore’s projects. Nevertheless, in the sections that I shall be considering, he is concerned – mainly and perhaps even exclusively – with Moore’s ‘Proof’. Wittgenstein’s argument is about external world scepticism, and some of its essential points are specific to scepticism of this type.
What I have just said goes against a very natural reading of Wittgenstein’s text. To explain why, let me distinguish two forms of scepticism, Agrippan and Cartesian. Agrippan scepticism is centred on the problem of the regress of justification. Knowledge differs from mere true belief by being justified. But in making my grounds for a given belief explicit, I enter a further claim which will need grounds of its own. The sceptic can now ask me to produce these, and so on indefinitely. My initial claim, the sceptic will say, has involved me in a vicious regress of grounds for grounds for grounds … If at some point I dig in my heels and refuse to play along, I will be making a brute assumption; and knowledge cannot be based on a mere assumption. If at some point I repeat myself, I will be reasoning in a circle. Keep trying to say something new, say nothing, or repeat something already said: there is no fourth option. Yet none of the available three produces knowledge.
Cartesian scepticism differs from Agrippan in the central role it assigns to sceptical hypotheses: hypotheses that are wildly at variance with our ordinary beliefs but which seem extraordinarily difficult to rule out. The paradigm Cartesian problem is Descartes’ problem of our knowledge of the external world; and the paradigm sceptical hypothesis is that our experience is controlled by an Evil Deceiver, so that the external world, as we ordinarily conceive it, does not exist. In particular, there are no physical objects. If the external world is the physical world, there is no external world. Now Wittgenstein clearly advances the following views:
Noticing these aspects of his thoughts, it is natural to suppose that Wittgenstein’s anti-sceptical reflections are directed primarily towards the Agrippan problem. The Agrippan sceptic insists that knowledge can be founded only on prior knowledge. This is false, since basic certainties are not items of knowledge. The regress is thus blocked. But because of the heterogeneity of our framework judgements, solutions to Cartesian scepticism follow automatically. Our framework provides a setting in which we can come to know all sorts of things about the external world, the past, and so on.
Call this account of Wittgenstein’s anti-sceptical strategy ‘the Framework Reading’.2 Its central idea is that Wittgenstein sees both Agrippan and Cartesian scepticism as falling to a single master argument. Once we understand how the frame of our epistemic practices is constituted, we have no need for a specific response to a problem like that of the external world. Indeed, it is unclear what form such a specific response could take or what it could accomplish. But Wittgenstein does give such a response. His approach to Cartesian scepticism is much more subtle than the Framework Reading implies.
In its fine structure, Wittgenstein’s refutation of idealism is complex, with numerous ideas in play throughout. Nevertheless, it can usefully be divided into three main phases: the problem phase, the diagnostic phase and the therapeutic phase. What marks the transition from one phase to the next is an admission that the problem of scepticism has not yet been explored in sufficient depth or presented in the proper light.
The Cartesian sceptic asks whether we know that there are any physical objects at all. Moore counters by first inquiring what we mean by ‘physical objects’, concluding that we mean ‘things to be met with in space’. He then gives his proof. Holding up his hands in good light (while making certain appropriate gestures), he announces: ‘Here is one hand and here is another’. It follows that at least two physical objects are known to exist. Moore is confident that his proof is a good one. The conclusion follows logically from the premises, and he knows the premises to be true.
It is clear to Wittgenstein that Moore’s proof is completely ineffective. Moore wants to assure the sceptic that there are physical objects. He does so by insisting that he (Moore) knows that two such things exist, rather as I might assure a friend that he has not missed the last train to the city, since I know that there are at least two evening departures. But the sceptic isn’t seeking reassurance and, in any case, Moore is in no special position to give it. Moore has misunderstood the kind of response that scepticism demands. Thus:
If you do know that here is one hand, we’ll grant you all the rest. When one says that such and such a proposition can’t be proved, of course that does not mean that it cannot be derived from other propositions; any proposition can be derived from other ones. But they may be no more certain than it is itself . …
From its seeming to me – or to everyone – to be so, it doesn’t follow that it is so.
What we can ask is whether it makes sense to doubt it.
(OC 1–2)
The sceptic thinks that he has found reason to question whether we know anything whatsoever about the external world. If his reasons for doubting are coherent, his doubts cannot be met by presenting particular examples of the kind of knowledge that is in question generally. But do his doubts really make sense?
Wittgenstein never wavers in his conviction that they do not: the sceptic’s doubts are wholly illusory. This is another reason why they cannot be met with a proof. If the scruples of the sceptic or idealist are incoherent, then so are the reassurances of the realist. No proof is possible because there is nothing to prove. This means that a response to scepticism cannot be dialectical: that is, it cannot take the form of showing that the sceptic is wrong, proving what he doubts. Rather, it must be diagnostic and therapeutic. It must identify the conceptual misunderstanding that gives rise to the illusion of sceptical doubt; and it must explain why the sceptic fails to see the illusion for what it is.
While Moore’s proof is deeply misconceived, it has considerable diagnostic interest. By reflecting on it, we can discern a number of important features of the logic of ‘doubt’ and ‘know’. We can also find clues to a deeper diagnosis of sceptical worries. Wittgenstein makes a number of interconnected suggestions.
The first is that ordinary doubts are essentially linked to the possibility of their being resolved. Thus:
If e.g. someone says ‘I don’t know if there’s a hand here’ he might be told ‘Look closer’. – This possibility of satisfying oneself is part of the language-game. Is one of its essential features.
(OC 3)
More than this, doubts do not arise as easily as the sceptic is apt to imagine. Wittgenstein writes:
[W]hat about such a proposition as ‘I know I have a brain? Can I doubt it? Grounds for doubt are lacking! Everything speaks in its favour, nothing against it. Nevertheless it is imaginable that my skull should turn out empty when it was operated on.
(OC 4)
The sceptic often argues as if the bare possibility of falsehood were a ground for doubt. But that is not at all how we normally proceed.
Wittgenstein’s second point is that Moore’s proof is not just ineffective: it involves a misuse of the expression ‘I know’. Wittgenstein asks:
Can one enumerate what one knows (like Moore)? Straight off like that, I believe not. – For otherwise the expression ‘I know’ gets misused. And through this misuse a queer and extremely important mental state seems to be revealed.
(OC 6)
Wittgenstein’s immediate concern is with the misuse itself, rather than with what it seems to reveal. The things Moore assures us that he knows are not ordinarily the objects of knowledge claims. Indeed, they are not ordinarily expressed in claims of any kind. Rather, knowledge and certainty are shown in practice, in the way I act: ‘My life shows that I know or am certain that there is a chair over there …’ (OC 7). This certainty has nothing to do with ‘making sure’ (OC 9).
Simple as they are, these points raise the question of whether Moore’s assurances are so much as intelligible.
I know that a sick man is lying here? Nonsense! I am sitting at his bedside, I am looking attentively into his face. – So I don’t know, then, that there is a sick man lying here? Neither the question nor the assertion makes sense. Any more than the assertion ‘I am here’, which I might yet use at any moment, if a suitable occasion presented itself … ‘I know that there is a sick man lying here’, used in an unsuitable situation seems not to be nonsense but rather seems to be matter-of-course, only because one can fairly easily imagine a situation to fit it …
(OC 10)
Stringing together meaningful words in a grammatically acceptable way is insufficient to guarantee a meaningful utterance: a speaker’s words must also be contextually appropriate. If they are not, it will be up for grabs whether he understands the words he comes out with, whether his utterance is a statement or just a verbal tic. We fail to see this because, even when an utterance fails to be contextually appropriate, we can easily imagine a situation to fit it. It therefore seems not to be nonsense even though (in its actual context) that is just what it is. Moore’s proof is an instance of this phenomenon:
… one thinks that the words ‘I know that …’ are in place where there is no doubt, and hence even where the expression of doubt would be unintelligible.
We just do not see how very specialised the use of ‘I know’ is.
(OC 10–11)
The use of ‘I know’ is specialised because it is linked with doubting. But as the earlier remarks on doubting revealed, Moore’s propositions are normally not doubted at all. Moore has therefore failed to enter his knowledge claims in a suitable situation. They may seem to be intelligible but in truth they are nonsense.
Wittgenstein’s initial remarks on the scope of ordinary doubt might give the impression that only the legitimacy of the sceptic’s ‘doubts’ is in question. This is not so. At issue is their intelligibility. It is not just that we do not doubt the things that Moore would like to say he knows: the question is whether we understand what it would be to doubt them. This is where the point that doubts too need grounds comes back into play. To express a doubt about a claim is to suggest that the speaker may have made some kind of mistake. If I cannot say what mistake – if I cannot specify how he might have gone wrong – then no intelligible doubt has been expressed.
Just as entering a doubt implies the possibility of saying what mistake might have been made, so entering a knowledge claim implies the possibility of saying how one knows. This will often mean being able to give appropriate grounds or evidence. Thus, an expression of doubt, implying the possibility of a mistake, can be met with an explanation of how one knows, an explanation that will show that no mistake was in fact made. The symmetry in the intelligibility requirements for doubting and knowledge claiming – the need to be able to say what mistake might have been made or how one knows – makes plain why the (in principle) possibility of resolving doubts (by explaining how one knows) is built into the language-game as one of its essential features.
The possibility of imagining suitable contexts for Moorean assurances is not the only source of their deceptive air of intelligibility. To explain why, Wittgenstein picks up the suggestion that Moore’s misuse of ‘I know’ seems to reveal a queer and important mental state. What makes the ‘mental state’ of knowing queer and important is that ‘know’ is factive: that is, ‘I know that there is a hand here’ entails ‘There is a hand here’.
… ‘I know’ seems to describe a state of affairs which guarantees what is known, guarantees it as a fact. One always forgets the expression ‘I thought I knew’.
(OC 12)
This remark probes more deeply into Moore’s confusion. Moore thinks that he can report (quite straightforwardly and intelligibly) on his current mental state (it is one of knowing). At the same time, because ‘know’ is factive, he thinks that in offering his reports he establishes what the sceptic claims to doubt. But Moore’s ‘reports’ are just knowledge claims, and the truth of such claims certainly does not follow from his making them (OC 13). On the contrary:
It needs to be shown that no mistake was possible. Giving the assurance ‘I know’ doesn’t suffice. For it is after all only an assurance that I can’t be making a mistake, and it needs to be objectively established that I am not making a mistake about that.
(OC 15)
Can Moore reply that he doesn’t just know: he knows that he knows? No. The thesis that knowing implies knowing that one knows is just another way of saying that ‘I know’ means ‘I am incapable of being wrong’; and whether I am incapable of error ‘needs to be established objectively’ (OC 16). When a knowledge claim is entered, the issue is not the claimant’s mental state but his epistemic status: his right to enter a claim in a way that implies immunity from error.
Can Moore earn this right by ‘establishing’ that he has hands? Perhaps not, if ‘establish’ means ‘give grounds’. Grounds need to be antecedently more certain than the proposition for which they are cited as grounds (OC 1) and, in the case of Moorean judgements, it is not clear that anything meets this condition. Naturally, there is no concession to the sceptic here. Questions of establishing one’s right to make a knowledge claim only arise where doubt is intelligible, which brings us back to the question of whether, in the case of Moorean judgements, mistakes are possible. Wittgenstein continues:
Suppose now I say ‘I am incapable of being wrong about this: that is a book’ while I point to an object. What would a mistake here be like? And have I any clear idea of it?
(OC 17)
It is tempting to treat this question as rhetorical, the implied answer being ‘No, I do not have any clear idea’. Given this answer, we can conclude that, in connection with Moorean judgements, talk of knowledge and doubt are equally out of place. As Marie McGinn puts it, Moorean judgements do not ‘embed in epistemic contexts’. They lie outside the scope of knowledge, evidence, justification and doubt. They belong to the framework within which epistemic claims can be entered: the framework within which such claims become intelligible.
The ideas just sketched figure prominently in what I have called the Framework Reading. On this account of Wittgenstein’s approach to scepticism, the essential moves are made very early. Of course, there is a lot of detail to be filled in. But the crucial idea – that scepticism is turned aside, not by propositions that are known to be true, but by judgements that it does not make sense to doubt – is in place. However, what we need to notice is that Wittgenstein himself does not present matters in this light. Even if Moorean judgements do have this special status, we cannot answer the sceptic merely by pointing this out. Far from being all over bar the shouting, the argument has barely begun.
Wittgenstein makes this very clear:
‘I know’ often means: I have the proper grounds for my statement. So if the other person is acquainted with the language-game, he would admit that I know. The other, if he is acquainted with the language-game, must be able to imagine how one may know something of this kind.
The statement ‘I know that here is a hand’ may then be continued ‘for it’s my hand that I’m looking at’. Then the reasonable man will not doubt that I know. … Nor will the idealist; rather he will say that he was not dealing with the practical doubt which is being dismissed, but there is a further doubt behind that one. … That this is an illusion has to be shown in a different way.
(OC 18–19)
These important passages throw into relief what I regard as so misleading about the Framework Reading: it makes Wittgenstein’s response to scepticism too direct. As a result, it loses sight of an essential point: that comments on the logic of ordinary doubting and knowledge claiming will cut no ice if we are in the grip of the illusion that there is a special kind of philosophical doubt, purporting to call epistemic ordinary procedures into question.
This point is worth elaboration. Marie McGinn claims that Wittgenstein’s key insight is that, because Moorean judgements do not embed in epistemic contexts, we do not stand in an ‘epistemic relation’ to such judgements. Accordingly, Moore’s insistence that he knows such things to be true and the sceptic’s attempt to doubt them both misfire. But no such conclusion is yet available. The most that has been shown is that such judgements are not ordinarily treated as either supportable by evidence or open to question. However – and Wittgenstein is well aware of this reply – our indulgent attitude is merely a reflection of practical exigencies. We have to take lots of things for granted if we are to get on with life. But in the context of philosophical reflection, where practical considerations are set aside, we can put ourselves into an epistemic relation with the most banal everyday certainties. Indeed, we can come to appreciate that we always stand in such a relation, even though for practical purposes we may ignore the epistemic demands that this relation imposes.
This is why I call the opening phase of Wittgenstein’s refutation of idealism ‘the problem phase’. The intent of his opening remarks is not to answer the sceptic in a direct way – however provisional and incomplete the answer – but rather to alert us to the peculiar and problematic character of philosophical doubt. The question ‘Have I any clear idea of what it would be to be mistaken with respect to a Moorean judgement?’ is not answered. To be sure, that the answer will be negative is implied by the claim that the idea of a doubt behind ordinary doubt is an illusion. But we are given the warning that this claim must be established ‘in a different way’, i.e. by doing more than rehearse the characteristic features of ordinary doubting and knowledge claiming.
Of course, it would be wrong to suggest that nothing has been accomplished. Revealing how different sceptical doubt is from ordinary doubt may not refute the sceptic but it can and should shake our casual confidence that we understand him. A vivid awareness of the peculiarity of sceptical doubt should make us receptive to the thought that a diagnostic investigation is worthwhile.
The first step towards dispelling the illusion of a doubt behind ordinary doubt is to identify its source. In Wittgenstein’s eyes, the proximate source of the illusion is a specific misunderstanding of our conceptual-linguistic situation. Identifying this misunderstanding is the task of his argument’s diagnostic phase. However, before presenting his diagnostic suggestion, Wittgenstein addresses resistance to the idea that any diagnostic enquiry is necessary. Having stressed once more the utter ineffectiveness of Moore’s assurances, Wittgenstein remarks:
… My believing the trustworthy man stems from my admitting that it is possible for him to make sure. But someone who says that perhaps there are no physical objects makes no such admission. The idealist’s question would be something like: ‘What right have I not to doubt the existence of my hands?’ (And to that the answer can’t be: I know that they exist.) But someone who asks such a question is overlooking the fact that a doubt about existence only works in a language-game. Hence, that we should first have to ask: what would such a doubt be like?, and don’t understand this straight off.
(OC 23–4)
Again, Wittgenstein conspicuously declines to offer a quick-and-dirty refutation. He does not argue (as we might have expected) that, since the possibility of resolving doubts belongs to our language-game as one of its essential features, sceptical or philosophical doubt is an obvious non-starter. Instead, he suggests that we ask what a sceptical doubt would amount to, warning us not to assume that we already know. His intent is clear: the peculiar character of philosophical doubt is a not a refutation of scepticism, but it is an invitation to pursue a diagnostic enquiry. If a doubt about existence only works in a language-game, and if the game of philosophical reflection is distinct from that of ordinary doubting, we are entitled to ask how the philosophical game is to be carried on.
The legitimacy of this question is reinforced by further reflections on ‘going wrong’ or ‘making a mistake’, reflections that invoke some of Wittgenstein’s most characteristic ideas about meaning and understanding. These are that understanding involves mastering concepts; possessing a concept involves learning the use of a word; and what is acquired, when the use of a word is learned, is at bottom a practical ability. By this last claim, Wittgenstein means that we use words without guidance from explicitly formulated rules. This applies to epistemic concepts such as ‘doubt’ as much as to any others. One can be wrong about so simple a matter as there being a hand, or an elementary calculation; but in particular circumstances, error is impossible and doubt misplaced (OC 25). However, we cannot give a rule specifying what those circumstances are. If we tried to give a rule, it would involve the expression ‘in normal circumstances’. And although normal circumstances can be recognised, they cannot be precisely described (OC 27). The ultimate absence of rules make itself felt in the attempt to state rules. Wittgenstein concludes that ‘Practice in the use of the rule shows what is a mistake in its employment’ (OC 29).
It is no accident that the idea of understanding as something essentially grounded in linguistic practice should make an appearance at just this point. The first phase of the argument ended with the sceptic or idealist denying any concern with ‘practical’ doubt. We now see that talk of a non-practical doubt is dangerously equivocal. Of course, philosophical questions may be ‘impractical’ in that, to consider them, we must set aside pursuits like making a living or cooking dinner. That is, a philosophical question may be purely theoretical. But no question is ‘impractical’ in the sense of ‘intelligible in abstraction from all particular practices of enquiry’.
Wittgenstein makes the point by recurring to Moore:
It’s not a matter of Moore’s knowing that there’s a hand there, but rather we should not understand him if he were to say ‘Of course I may be wrong about this’. We should ask ‘What is it like to make such a mistake as that?’ – e.g. what’s it like to discover that it was a mistake?
(OC 32)
It is not as though we have some general-purpose concept of ‘making a mistake’ such that, in any circumstances whatsoever, and without any particular error-possibility in mind, we can intelligibly say ‘Maybe you are making a mistake’. So once more, what is the language-game of philosophical doubt?
The absolutely crucial feature of philosophical doubt has already been identified. It involves taking seriously the possibility that no physical objects exist. If there is a genuine possibility here, we can see why the idealist wants to know what right I have not to doubt the existence of my hands. We can also see why this doubt, once entered, may turn out to be irresolvable. I cannot resolve it Moore’s way, by examples. But it is not obvious what other way, if any, is available to me.
Is it really possible that no physical objects exist? Well, can’t we imagine that no such objects exist, and doesn’t this show that their non-existence is possible? This is the key question addressed in the diagnostic phase of Wittgenstein’s argument. Here is what he has to say:
But can’t it be imagined that there are no physical objects? I don’t know. And yet ‘There are physical objects’ is nonsense. Is it supposed to be an empirical proposition? …
And is this an empirical proposition: ‘There seem to be physical objects’?
‘A is a physical object’ is a piece of instruction which we give only to someone who doesn’t understand what ‘A’ means, or ‘what physical object’ means. Thus it is instruction about the use of words, and ‘physical object’ is a logical concept. (Like colour, quantity, …) And that is why no such proposition as: ‘There are physical objects’ can be formulated.
Yet we encounter such unsuccessful shots at every turn.
(OC 35–6)
We have identified the proximate source of the illusion of the doubt behind everyday doubt. The idealist or sceptic wants to treat ‘There are physical objects’ as an empirical or factual statement. He wants to treat it as a hypothesis. It is neither. However, this is not because, as the Framework Reading has it, that it is a ‘framework judgement’, lying apart from the route travelled by enquiry, beyond evidence and justification, and non-factual because neither true nor false. ‘There are physical objects’ is neither true nor false because it is nonsense. And it is nonsense because ‘physical object’ is not the concept of a kind of object, like ‘unicorn’ or ‘planet’. ‘Physical object’ is a piece of logical or semantic vocabulary, thus unsuitable for formulating the empirical hypothesis the sceptic or idealist would like to express.
Wittgenstein does not elaborate, but the main drift of his thought is clear. We do not have a general-purpose concept of ‘object’ that swings free of our ability to refer to objects in the course of playing particular language-games. Rather, our understanding of ‘objects’ is implicit in our mastery of singular reference. In other words, ‘objects’ are what singular terms pick out. Central to our mastery of singular reference are practices of identification and re-identification: we know what an object is when we know what does and does not count as the same object. These practices of identification and re-identification sort ‘objects’ into broad logical categories. The criteria for being the same chair as the one I saw at the auction yesterday are different from those for being the same shade of grey as the one on the walls of my office. Someone familiar with particular practices of identifying and re-identifying can be taught to recognise these broad logical divisions (‘physical object’, ‘colour’, ‘quantity’), and this can short-circuit the teaching of future words. This is why no such proposition as ‘There are physical objects’ can be formulated. At most it could mean ‘We talk about tables, chairs, dogs, cats, etc.’. This is not at all what the realist intends to assert or the idealist to deny.3
There are echoes of Carnap here, but with a difference. No philosopher inclined to wonder whether there are really such things as numbers will be satisfied with a demonstration that there are two prime numbers between six and twelve, so that at least two such things are known to exist. This parallels the case of the idealist or sceptic, who will reject Moore’s assurances with respect to the existence of his hands. The sceptic or idealist wants to make an ‘external’ claim about certain referential practices considered as a whole. But the only such claims we can make are semantic, describing the games we play. (This is calculating; and this is talking about ‘physical objects’.) There is no vantage point from which we could find these games metaphysically wanting (or see playing them as a practical decision).
It is important to take note of the special character of this diagnosis. Not all propositions that lie apart from the route travelled by enquiry do so for the same reason. To see this, consider another proposition much discussed by Wittgenstein in the later sets of notes: ‘The earth has existed for many years past’. No one who doubted this proposition could engage in historical investigation or seek historical understanding. The same goes for the proposition that not all historical records are the product of fraud or deception. All our discursive practices involve such (typically tacit) commitments. They constitute the ‘riverbed’ along which enquiry flows (OC 95–9), the axis around which enquiry moves (OC 152), or the hinges on which it turns (OC 341–3).
Is ‘There are physical objects’ a hinge proposition, only more general than hinges specific to history or physics because common to all such enquiries?4 No. ‘There are physical objects’ is not a hinge proposition: it is nonsense.
We learn to think by learning to talk; and we learn to talk by being trained to make particular judgements about things around us. It is therefore inconceivable that there should be discursive beings who had not mastered ‘physical-object’ talk. For more specialised kinds of talk, this is not so. Wittgenstein writes:
[W]hat goes into someone’s knowing…history, say? He must know what it means to say: the earth has already existed for such and such a length of time. For not any intelligent adult must know that …
(OC 85)
… Men have believed that they could make rain. Why should not a king be brought up in the belief that the world began with him? …
(OC 92)
A definite conception of historical time belongs to what we take as common sense. But not everything that belongs to common sense is a precondition of the very possibility of rational thought. This confirms what I have claimed. Wittgenstein’s response to idealism, the problem addressed in Moore’s ‘Proof’, must be distinguished from his discussion of Moore’s ‘Defense’, which raises very different issues.
The source of the idea of a doubt behind the doubt has been identified and exposed as a piece of conceptual confusion. We might therefore suppose that Wittgenstein’s argument is complete. This is not how Wittgenstein sees things.
But is it an adequate answer to the scepticism of the idealist, or the assurances of the realist, to say that: ‘There are physical objects’ is nonsense? For them after all it is not nonsense. It would, however, be an answer to say: this assertion, or its opposite, is a misfiring attempt to express what can’t be expressed like that. And that it does misfire can be shown; but that isn’t the end of the matter. We need to realise that what presents itself to us as the first expression of a difficulty, or of its solution, may as yet not be correctly expressed at all. Just as one who has a just censure of a picture to make will often at first offer the censure where it does not belong, and an investigation is needed in order to find the right point of attack for the critic.
(OC 37)
This passage marks the transition from the diagnostic to the therapeutic phase of the argument. Let me note, however, that I am using ‘diagnostic’ and ‘therapeutic’ as convenient markers for two phases of Wittgenstein’s argument. I do not mean to suggest any deep methodological distinction. On the contrary, as paragraph 37 makes clear, Wittgenstein’s therapy will depend on further diagnosis.
Why is it an inadequate answer to the idealist and the realist to say that ‘There are physical objects’ is nonsense? In saying that it is not nonsense ‘to them’, Wittgenstein is not saying that it is not nonsense. Nor is he conceding that the idealist and realist have given it a sense: not a clear sense, anyway. The point is rather that these philosophers – all of us when we are in the grip of sceptical anxieties – will not recognise that it is nonsense. They (or we) think that ‘There are physical objects’ can be understood as an empirical hypothesis. They (we) suffer from an illusion of meaning, the source of which remains to be exposed.
The realist wants to say something correct: that there is nothing defective in the confident way we talk about tables and chairs and rocks and trees. But one cannot make this point by insisting that, contrary to the sceptic or idealist, there really are physical objects out there, as one might insist that there really are mountains on the Moon. (They are not just a trick of the light.) The realist’s way of asserting the legitimacy of everyday talk about physical objects misfires: it uses logical vocabulary to make what is intended to be an empirical claim about the world beyond language. This much, Wittgenstein thinks, has been shown. But this is not the end of the matter because the confusion that has been identified is not gratuitous, not the result of mere blindness to the contours of the conceptual landscape. Rather, our confusion is induced by ideas that have yet to come into view. Until we have identified these ideas, we will not have found the right point of attack.
Wittgenstein never tells us in so many words what he takes the right point of attack to be. But I think we can identify it with some confidence. Wittgenstein begins his diagnostic investigation by returning to knowledge in mathematics:
Knowledge in mathematics: Here one has to keep on reminding oneself of the unimportance of the ‘inner process’ or ‘state’. … What is important is how we use mathematical propositions.
(OC 38)
What sort of proposition is this: ‘We cannot have miscalculated in 12 × 12 = 144’? It must surely be a proposition of logic.–But now doesn’t it come to the same, as the statement 12 × 12 = 144?
If you demand a rule from which it follows that there can’t have been a miscalculation here, the answer is that we did not learn this through a rule, but by learning to calculate.
We got to know the nature of calculation by learning to calculate.
(OC 43–5)
To learn to calculate at all, hence to learn what calculating is, we must be trained to accept certain calculations as (normally) unquestionable. Someone who was in doubt as to whether 12 × 12 = 144 would not have learned to calculate. He would not know what calculating is and would not understand arithmetical statements. As Wittgenstein explains in some later remarks:
The truth of my statements is the test of my understanding of these statements.
That is to say: if I make certain false statements, it becomes uncertain whether I understand them.
(OC 80–1)5
And although there is no rule for distinguishing those cases in which error is impossible from those in which it isn’t, this is not a lack: ‘The rule is not needed’.
(OC 46)
The possibility of making or even imagining a mistake is severely constrained by the demands of making sense at all. This is a ‘logical’ point. However, we should not misunderstand this characterisation.
What sort of proposition is: ‘What could a mistake here be like!’? It would have to be a proposition of logic. But it is a logic that is not used, because what it tells us is not learned through propositions. –It is a logical proposition, for it does describe the conceptual (linguistic) situation.
(OC 51)
Wittgenstein’s ‘logic’ simply describes how judgements of various kinds are made, though even then not in a way that reduces our practical know-how to strict rules. Accordingly, logic does not guide our practices. Nor does it justify them. It is not their foundation. Nevertheless, logical investigation reveals something of great significance for scepticism. Wittgenstein comes to the point:
The situation is thus not the same for a proposition like ‘At this distance from the sun there is a planet’ and ‘Here is a hand’ (namely my own hand). The second can’t be called a hypothesis. But there isn’t a sharp boundary line between them.
So one might grant that Moore was right, if he is interpreted like this: a proposition saying that here is a physical object may have the same logical status as one saying that here is a red patch.
(OC 52–3)
Here we find the right point of attack. The reason why sceptic and idealist think that ‘There as physical objects’ is a hypothesis is that they are convinced that experiential knowledge – knowledge of coloured patches or ‘sense-data’ – is epistemologically prior to knowledge of physical objects. In fact, in their view, experiential knowledge is epistemologically basic: knowledge of sense-data is distinctive in its immediacy, certainty and immunity from error. In reporting on our sense-data, we can perhaps make verbal slips, but we cannot make mistakes. With this doctrine in place, judgements about physical objects look to be inferential. Perhaps our sense-data arise in deviant ways, as they would if we were victims of the Evil Deceiver or brains-in-vats. The commitment to the existence of physical objects looks like an empirical hypothesis: a particular explanation of the origin of our sense-data. How we might justify this ‘hypothesis’ is obscure.
Within the constraints of the doctrine of the priority of experience, Moore’s proof is a total failure. But Moore may have been on to something: there is no reason to accept the doctrine. Immunity from error, across a wide range of cases, is a feature of language-use as such.
For it is not true that a mistake merely gets more and more improbable as we pass from the planet to my own hand. No: at some point it has ceased to be conceivable.
This is already suggested by the following: if it were not so, it would also be conceivable that we should be wrong in every statement about physical objects; that any we ever make are mistaken.
(OC 54)
Sense-datum talk and physical-object talk are no different in this respect: they can have the same logical status. Taking this point to heart, we strike at the source of the feeling that ‘There are physical objects’ is a (risky) hypothesis. As this feeling fades, we can come to see the realist’s ‘hypothesis’ for the nonsense it is.
The sceptic will resist this argument. Because it is logically possible that ‘things around us’ do not really exist, which is inconceivable in the case of colours, there is a potential sceptical problem about external objects for which sense-data offer no counterpart. It follows that our knowledge of sense-data is intrinsically certain, in a way that knowledge of tables and chairs can never be.
Wittgenstein anticipates this response.
So is the hypothesis possible, that all things around us don’t exist? Would not that be like the hypothesis of our having miscalculated in all our calculations?
(OC 55)
It would, which means that the ‘argument from differential certainty’ presents no new considerations. In fact, the argument assumes what it offers to prove: that sense-datum knowledge is privileged. Only given this assumption can we seem to ourselves to understand ‘There are physical objects’ as a (quite possibly false) hypothesis. But sense-datum judgements enjoy no special immunity from doubt. Rather, limitations on the intelligibility of error are a pervasive feature of discursive practice. Whether we are calculating, asking someone to fetch a chair, or pointing out a colour, in the right circumstances error is inconceivable.
A further factor in our feeling that talking about tables and chairs reflects commitment to a hypothesis is the assumption that we have an all-purpose concept of ‘existence’ whose application is clear in any context whatsoever (this parallels the mistake about the concept of an ‘object’). Thinking along these lines, we might be tempted to argue: ‘You know what physical objects are – tables, chairs, things like that – and you know what ‘exist’ means; so you understand the claim that physical objects might not exist’. Naturally, Wittgenstein regards this as another illusion.
When one says ‘Perhaps this planet doesn’t exist and the light-phenomenon arises in some other way’, then after all one needs an example of an object which does exist. This doesn’t exist, … as for example does…
(OC 56)
Of course, the idealist has an example. The table, which I seem to see, does not exist, as for example does the table-shaped brown patch (sense-datum). But what contrast is implied here? The sense-datum exists in a special way: it is immediately present to consciousness. And what is the hallmark of this immediate presence? That mistakes are inconceivable? We are getting nowhere. Logical investigation shows no trace of the epistemic cleavage between sense-datum judgements and judgements about physical objects. Rather, wherever we look, we find doubt gradually losing its sense. That is simply how things are: ‘The language-game just is like that’ (OC 56).
Wittgenstein’s argument is still not complete. The idea that knowledge of sense-data is epistemologically basic cannot be read off our ordinary way of doing things. To the contrary, it is in tension with the logic of ordinary epistemic practices. So what is the source of its appeal? Wittgenstein’s answer to this question is not given in his first set of notes, though the ground is prepared there.
As we saw, Wittgenstein thinks that one of the pitfalls of Moore’s attempt to enumerate things he knows is that his performance seems to reveal ‘a queer and extremely important mental state’. Thus:
… ‘I know’ seems to describe a state of affairs which guarantees what is known, guarantees it as a fact. One always forgets the expression ‘I thought I knew’.
(OC 12)
‘Know’ is factive: ‘I know that there is a hand in front of me’ entails ‘There is a hand in front of me’. At the same time, avowals of one’s current mental state have a special authority. In normal circumstances, a sincere avowal that I believe so-and-so is criterial for my believing it. You might on occasion question my sincerity but you would not, except perhaps in very special circumstances, suspect me of being mistaken. Normally, from my saying ‘I believe so-and-so’ you can infer that I that I do indeed believe so-and-so. According to Wittgenstein, Moore’s performance involves acting as if the same were true of knowledge.
Moore’s view really comes down to this: the concept ‘know’ is analogous to the concepts ‘believe’, ‘surmise’, ‘doubt’, ‘be convinced’ in that the statement ‘I know …’ can’t be a mistake. And if that is so, then there can be an inference from such an utterance to the truth of an assertion. And here the form ‘I thought I knew’ is being overlooked …
(OC 21)
Wittgenstein thinks that the way to avoid Moore’s misstep is to resist thinking of knowledge as a ‘mental state’:
One can say ‘He believes it, but it isn’t so’, but not ‘He knows it, but it isn’t so’. Does this stem from the difference between the mental states of belief and of knowledge? No. … One may for example call ‘mental state’ what is expressed by tone of voice in speaking, by gestures etc. It would thus be possible to speak of a mental state of conviction, and that may be the same whether it is knowledge or false belief …
(OC 42)
If ‘mental states’ are such that a special epistemic authority attaches to avowals or first-person reports of them, knowledge is not a mental state. Knowing is a not a matter of a claimant’s mental state but of his (or his claim’s) epistemic status: for example, whether or not the evidence backs him (it) up. And epistemic status is impersonal and public. There is nothing subjective about it.
That is the problem. If one thinks of knowledge as a mental state, thus as subjective, while recognising that ‘know’ is factive, one will suppose that a subjective or inner state can ensure that certain facts really do obtain. But how can a subjective state guarantee objective facts? It cannot. Thus, one is led to the view that the only facts that can be known, or ‘immediately’ known, are themselves subjective: facts about other mental states. Wittgenstein explains:
‘I know’ has a primitive meaning similar to and related to ‘I see’ (‘wissen’, ‘videre’). And ‘I knew he was in the room, but he wasn’t in the room’ is like ‘I saw him in the room, but he wasn’t there’. ‘I know’ is meant to express a relation, not between me and the sense of a proposition (like ‘I believe’) but between me and a fact. So that the fact is taken into my consciousness. (Here is the reason why one wants to say that nothing that goes on in the outer world is really known, but only what happens in the domain of what are called sense-data.) This would give us a picture of knowing as the perception of an outer event through visual rays which project it as it is into the eye and the consciousness. Only then the question at once arises whether one can be certain of the projection. And this picture does indeed show how our imagination presents knowledge, but not what lies at the bottom of this presentation.
(OC 90)
What lies at the bottom of this presentation is the idea of knowledge as a ‘mental state’. This idea is perhaps the sceptic’s most dangerous illusion.
Wittgenstein closes the circle by returning to the point from which his argument took off: the peculiarity of Moore’s performance in ‘proving’ the existence of an external world. It now turns out that the misuse of ‘know’ involved in Moore’s attempt to answer the sceptic encourages or reflects the very conceptual misunderstandings from which sceptical doubts derive their power. Dialectically, this is a brilliant move. But it is also diagnostically profound.
The idea of a doubt behind ordinary doubt depends, in the last analysis, on the doctrine of the priority of experience over knowledge of the worldly objects. But the sceptic understands this doctrine in a very peculiar way. As I like to say, the sceptic is an epistemological realist: constraints on justification are not rooted in norms that we impose (and might modify) but are derived from the natural order of reasons, an order that holds independently of human artifice or convention.6
On the face of it, this is a strange metaphysical view. The doctrine of the epistemic priority of experience is normative: it is about what sorts of claims need to be supported by what sorts of evidence, if they are to amount to knowledge. But it is not easy for the sceptic to present it in this light for, so presented, it is apt to look like an imposition and a far from obviously reasonable one at that. Accordingly, the sceptic proceeds as though the doctrine were enforced by the epistemological facts. Judgements about the external world must be supported by experience – knowledge of one’s own sense-data – because in the last analysis such experiential knowledge is all we have to work with. That is just how things are; that we are condemned to work outwards from (subjective) experience is simply our epistemic situation, which appears on reflection to be a predicament. In any event, the priority of experience is not a justificational ideal that the sceptic imposes, but rather a constraint inherent in the human condition. Scepticism thus appears as a surprising discovery.
The imaginative picture that Wittgenstein describes shows how one might be an epistemological realist. Gripped by the idea of knowledge as a fact-guaranteeing mental state, we think that the reach of our capacity to guarantee the facts sets an outer boundary to what we can (immediately) know. In this picture, the limitation on our epistemic resources does not appear as a normative constraint but as a quasi-psychological limitation intrinsic to the process by which knowledge must be obtained. Unfortunately, thus pictured, the process is one whose reliability we will never be able to guarantee.
If we break with the idea of knowledge as a queer mental state, we can break with the sceptic’s epistemological realism. This is what Wittgenstein wants us to do. The alternative to epistemological realism is a pragmatic view of norms. The normative structure of doubting and justifying is implicit in practices of enquiry which, as human institutions, are subject to change. There is no immutable order of reasons for the sceptic to discover or exploit. Reminders to the effect that the rules of our language-games, thus the epistemic constraints implicit in those games, are not wholly beyond our control recur in Wittgenstein’s refutation of idealism.
Whether a proposition can turn out true or false after all depends on what I make count as determinants for that proposition.
(OC 5)
… [O]ut of a host of calculations certain ones might be designated as reliable once and for all, others as not yet fixed…
… But… this in only a decision for a practical purpose.
(OC 48–9)
It is, I think, significant that Wittgenstein’s first set of notes ends with a reminder that our language-games are to some degree malleable. Our language-games change as our conception of the world changes.
… A meaning of a word is a kind of employment of it. For it is what we learn when the word is incorporated into our language.
That is why there exists a correspondence between the concepts ‘rule’ and ‘meaning’.
If we imagine the facts otherwise than as they are, certain language-games lose some of their importance, while others become important. And in this way, there is an alteration – a gradual one – in the use of the vocabulary of the language.
(OC 61–3)
When language-games change, then there is change in concepts, and with concepts the meanings of words change.
(OC 65)
With such changes go changes in the normative epistemic structure of our games: the doubts that we recognise as reasonable (or even intelligible), what we may or must take for granted, and so on. As I read Wittgenstein, it is clear why he ends on this note. Epistemological realism, the idea of an immutable order of reasons, is scepticism’s ultimate source; and the pragmatic conception of norms, implicit in the idea of meaning as use, shows how we can refuse to accede to it.
1 It is worth remembering that the paragraph numbers of OC were introduced by the editors. They are not Wittgenstein’s.
2 I hope that the Framework Reading is not completely unrecognisable. Variants of it can be found in important studies of On Certainty by Marie McGinn and Avrum Stroll, as well as in some writings of Crispin Wright. See McGinn 1989, Stroll 1994 and Wright 1985.
3 My take on this aspect of Wittgenstein’s argument owes a lot to Robert Brandom’s discussion of referring to objects in Brandom 1994 (see p. 360f).
4 Wright 1985: 455. Wright is careful to distinguish general commitments from hinges as contextually-specific certainties, such as ‘I have two hands’ (assumed in normal circumstances). Nevertheless, he advocates a uniform treatment for ‘There is a material world’ and ‘The earth has existed for many years past’. This is just what Wittgenstein repudiates.
5 Notice that this argument depends on a straightforward attribution of truth to Moorean judgements. At the same time, Wittgenstein is tempted to deny that propositions belonging to the ‘background’ of enquiry are properly thought of as true or false. To go into this issue would take me too far afield. Let me just say what I take to be the source of Wittgenstein’s hesitation: he is torn between a deflationary view of truth (in the form of a redundancy theory) and an epistemic account (the true/false as what we can confirm/disconfirm). What these two approaches to truth have in common is that on neither can we explain why certain proposition ‘stand fast’ by saying that they ‘correspond to reality’. Incidentally, Wittgenstein is entitled to adopt deflationism. The argument connecting truth and meaning anticipates some well-known views of Davidson. Pace Davidson, these views are compatible with a deflationary conception of truth. See Williams 1999.
6 See Williams 1996.