Reading Wittgenstein's On Certainty as a Whole: An Interpretive Picture

Metaphysics & Epistemology
1

Reading Wittgenstein’s On Certainty as a Whole: An Interpretive Picture

I’ve been reading and writing on Wittgenstein’s On Certainty since 1980, and what follows is my interpretive picture of the work as a whole. I want to note that Wittgenstein left no title for these notes, which were edited and published posthumously by G.E.M. Anscombe and G.H. von Wright, so the title is entirely my own. I’m working as an independent scholar, someone who’s lived with this text for forty-five years rather than reporting the consensus of a department. Forum members who know my work will recognize familiar ideas. Those who don’t are welcome to engage the ideas on their merits, which is the only standard that matters in this thread.

On Certainty is a notoriously difficult text, unfinished, written in fragments across the last eighteen months of Wittgenstein’s life, and resistant to the kind of systematic summary that academic philosophy tends to reward. Part of its difficulty, I want to suggest, comes from a source that hasn’t received enough attention. Wittgenstein uses the word “certainty” in at least four distinct senses throughout the text, and he moves between them without always flagging the transition. Much of the confusion in the secondary literature, and much of the productive disagreement, traces back to readers talking past each other because they’re tracking different uses of the word.

The four senses I want to work with are these:

Subjective certainty is the psychological state, conviction, the feeling of being sure. This is what Wittgenstein is consistently pushing against in OC. Certainty in this sense isn’t what he’s after, and distinguishing it from the others is the first clarifying move the text requires.

Hinge certainty is the bedrock sense, the non-epistemic certainty of propositions that stand fast for us, not because we’ve examined and verified them, but because they form the enabling conditions of any examination or verification at all. This is the heart of OC, and the sense most directly tied to Wittgenstein’s own concept of hinge propositions.

Epistemic certainty is the defeater-resistant standing a belief can have within a practice of justification, the legitimate, practice-indexed use of “I know” that Wittgenstein contrasts with both Moore’s overclaiming and the skeptic’s underclaiming. This sense is operative throughout OC even though Wittgenstein doesn’t name it as such.

Absolute certainty is the certainty of logical and mathematical necessity. Wittgenstein’s treatment of this sense in OC is the most complex and in some ways the most unresolved. He’s clearly aware that mathematical certainty has a different character from empirical hinge certainty, but the relationship between them is something the text gestures toward without fully settling.

A few things I want to be clear about from the start. Wittgenstein doesn’t explicitly draw these four distinctions himself. The taxonomy is mine, offered as a reading tool rather than a doctrine hidden in the text. What I’m claiming is that these distinctions are required by the text, that the confusions Wittgenstein is working against, and the moves he’s making, only become fully visible once you recognize that certainty is doing different work in different places throughout OC. The taxonomy doesn’t impose a system on a text that resists system. It provides a map for navigating terrain that Wittgenstein himself knew was unmapped.

I also want to be clear about the scope of this thread. My aim isn’t to limit myself to strict exegesis of OC. I’ll be working in three registers, what the text directly supports, what it implies but doesn’t fully develop, and where the distinctions lead when taken further than Wittgenstein himself goes. I’ll try to signal which register I’m in at each stage, because those are genuinely different kinds of claims and deserve to be evaluated differently.

In what follows I’ll take up each sense in turn, working through the relevant passages in OC and then developing what each sense implies beyond the text. I’ll start with subjective certainty, the sense Wittgenstein is arguing against, because clearing that ground first makes everything else easier to see.

2 Likes
2

Post 1: Starting at the Beginning, OC 1-4

The opening of OC looks like a conclusion, and that’s probably deliberate. Witt brings us directly into an interrogation that assumes we already know the crux of the argument. If you’re not familiar with Moore’s Proof of an External World and A Defense of Common Sense, the opening remarks can feel somewhat daunting, so a brief word about the context is probably in order.

Moore argued that he could prove the existence of an external world by holding up his hands and saying, here is one hand, and here is another. He took this as a paradigm case of certain knowledge, something so obvious that it couldn’t seriously be doubted. Witt’s response to this isn’t to agree or disagree with Moore’s conclusion. It’s to question whether Moore’s point makes sense, whether “I know…” is doing any work in this context.

OC 1 opens with, “If you do know that here is one hand, we’ll grant you all the rest.” That’s not a concession. It’s a challenge. Witt is saying that the interesting question isn’t whether Moore’s conclusion is true, but whether Moore knows in the sense he claims. The word “if” is carrying a lot of weight in the sentence.

In OC 2 Witt makes the point that’s easy to miss but crucial to what follows. He says that from its seeming to me, or to everyone, to be so, it doesn’t follow that it is so. It’s a simple observation but very important. Moore’s certainty that here’s a hand is grounded in how things seem to him, is the conviction that no doubt is possible. But Witt is pointing out that the strength of his conviction, even if everyone shared it, doesn’t establish that Moore knows anything. Seeming isn’t knowing in the epistemic sense. This is precisely why subjective certainty, the mere feeling of being sure, can’t do the epistemic work Moore is asking it to do.

This is where I’ll introduce the first of the four senses of certainty, because it’s already implicit in these opening remarks even though Witt won’t develop it until later.

What Moore is appealing to is something that looks like subjective certainty, the feeling of being sure, the psychological conviction that here is a hand and no further doubt is possible. And Witt’s response in these opening sections is already pushing back. Subjective certainty, the mere feeling of being sure, doesn’t settle anything epistemologically. The fact that Moore feels certain, that his certainty is unshakeable as a psychological matter, doesn’t establish that he knows in any epistemic sense.

OC 3 sharpens this point. Witt writes that if Moore is wrong about having two hands, then he’s also wrong about everything he says about them, and so on. The point is that Moore’s certainty isn’t just about his hands. It’s embedded in the way we engage with the world, a framework of assumptions and practices that he isn’t questioning when he holds up his hands. This is the first glimpse of what I’d call hinge certainty, though Witt won’t use that language yet. The certainty Moore is actually trading on isn’t subjective certainty at all. It’s something deeper, something that belongs to the framework within which doubting and knowing take place.

In OC 4 Witt says that the question of whether Moore knows or only believes that he has two hands is connected to the question of whether the statement “I have two hands” could ever be false for Moore. This is an important point. He’s already pointing toward the idea that some propositions function differently from ordinary empirical claims, that they operate differently in relation to evidence, doubt, and justification.

What I find striking about these four opening remarks, and what I think is often missed, is that Witt is already operating with at least three of the four senses of certainty. He’s pushing against subjective certainty as a philosophical foundation. He’s gesturing toward hinge certainty as what’s actually doing the work in Moore’s examples. And he’s also invoking epistemic certainty by pressing on whether Moore’s use of “I know…” is legitimate within the argument he presents.

The text opens, in other words, not with a question but with a diagnosis. Moore has confused these senses of certainty, and everything that follows in OC is in some way an unpacking of that confusion and a working out of what a clearer picture would look like.

I’ll explain subjective certainty in more depth in the next post, working through the passages where Witt is most explicitly arguing against it and showing why the confusion between subjective certainty and the other senses generates the philosophical problems he’s trying to dissolve.

3

Post #2 OC 5-7

OC 5 makes the point that’s easy to read past but it’s worth slowing down to carefully read. Witt says that whether a proposition can turn out false after all depends on what I make count as determinants for that proposition. That’s not a claim about logic or about the world. It’s a claim about the grammar of our epistemic practices, about how we determine what counts as evidence for or against a proposition. So, in OC 5 we’re moving away from the idea that propositions have a fixed relationship to truth and falsity independent of a practices function.

OC 6 connects directly back to Moore and introduces something that runs through the entire text. Witt asks whether one can enumerate what one knows the way Moore does, and his answer is, he believes not, because otherwise the expression “I know” gets misused. And then he adds something odd, that through this misuse a queer and extremely important mental state seems to be revealed. That mental state is, I think, precisely what I’ve been calling subjective certainty, the feeling of conviction that Moore’s enumeration of certainties seems to invoke or report. This connection becomes clearer when we get to OC 42, where Witt draws a sharp distinction between the mental state of conviction and knowledge itself. He says that what is known must be true, but that knowledge isn’t a mental state at all. It’s a standing within the practice of epistemology. Conviction, by contrast, is a genuine mental state, a psychological matter, but it isn’t epistemic. Witt’s word “queer” in OC 6 is telling. He isn’t saying the mental state doesn’t exist. He’s saying it creates a philosophical illusion by making it look as though “I know” is reporting an inner state of conviction rather than making a move within a practice. Moore’s mistake isn’t just logical. It’s grammatical. He’s misusing “I know” in a way that conjures a psychological state and then mistakes that state for philosophical bedrock.

In OC 7 Witt says that his life shows that he knows or is certain that there is a chair over there, or a door, and so on. He gives simple examples: telling a friend to take that chair, or to shut the door. Not his assertions, not his arguments, but his life, what he does. That’s a striking formulation and I think it’s doing something important. The certainty Witt is pointing to here isn’t expressed in propositions or justified by evidence. It shows itself in action, in the ordinary unreflective way we navigate the world and direct others within it. Nobody says “I know there is a chair over there, and here is my justification for that claim” before asking a friend to sit down. The certainty is displayed in the acting itself. This connects directly to what I’d call the nonlinguistic level of hinge certainty, the certainty that lives in how we act rather than in what we say. And it stands in sharp contrast to Moore’s gesture of holding up his hands and announcing “I know.” Moore is trying to make explicit and propositional something that Witt is showing us works precisely because it doesn’t need to be made explicit or justified in any epistemic way.

4

Post 3: OC 8-10

OC 8 is worth slowing down on because I think it’s one of those places where the unfinished character of the text shows. Witt says the difference between the concept of knowing and the concept of being certain isn’t of any great importance at all, except where I know is meant to mean I can’t be wrong. He gives the example of a law court, where I am certain could replace I know in every piece of testimony. The courtroom example holds up because in that context both expressions are doing the same practical work. The witness is committing to the truth of what they report, and the practice/context disciplines that commitment regardless of which phrase is used. But as a general claim, this sits uneasily with the larger argument of OC. The distinction between I know as an epistemic standing within a practice and I’m certain as a report on one’s psychological conviction, what I’ve been calling subjective certainty, is precisely the distinction that drives much of what follows. Witt’s diagnosis is that Moore, in lifting I know out of ordinary life and putting it to philosophical use, is left with nothing but his conviction that he can’t be wrong, even though Moore wouldn’t accept this characterization. Moore thought he was stating something obviously true, not reporting an inner state. But Witt’s point is that the philosophical context strips I know of the ordinary practical role that gives it its force, and what remains, whether Moore knows it or not, is subjective certainty doing work it isn’t equipped to do. Much of OC is an unpacking of why and how that move fails. So, OC 8 works for the narrow case Witt has in mind, but it would need significant qualifying to fit his later remarks, particularly OC 42 where he’s explicit that knowledge is not a mental state.

In OC 9 Witt asks whether, in the course of his life, he makes sure he knows that here is a hand, his own hand. In OC 10 he pushes the point further, viz., “I know that a sick man is lying here? Nonsense! I am sitting at his bedside, I am looking attentively into his face.” Neither the question nor the assertion makes sense in this context, he says, any more than the assertion “I am here” would, though that’s a sentence one might use perfectly well if a suitable occasion presented itself. This continues the insight from OC 7 but it’s more focused now. The certainty Witt is pointing to isn’t something we verify or establish. It’s displayed in what we do, in sitting at the bedside, in looking into the face. To announce “I know there is a sick man lying here” in that context isn’t wrong. It’s idle. The words have no work to do because the certainty is already fully present in the action itself. This is hinge certainty operating at the nonlinguistic level, the certainty that lives in our engagement with the world rather than in any proposition we might assert about it. Witt closes OC 10 by asking whether 2 × 2 = 4 is nonsense in the same way, which signals that the relationship between this kind of certainty and mathematical certainty is already an idea he’s thinking about. He doesn’t pursue it here, but I’ll return to it later when the text warrants.

5

Post 4 OC 11-15

OC 11 seems simple but carries a lot of weight. Witt says, “We just do not see how very specialized the use of ‘I know’ is.” That single statement is doing something important. It’s a diagnosis of why Moore’s argument goes astray and why philosophy keeps making the same mistake. We treat “I know…” as if it were a general-purpose epistemic expression that fits any proposition in any context, when in fact it only does real work under specific conditions and within specific language-games. Moore’s mistake is to take “I know…” out of those contexts and deploy it as a philosophical instrument, as though saying “I know here is a hand” were the same kind of proposition as saying “I know the train leaves at three.” OC 12 brings this out by exposing the illusion that generates the error. “I know…” seems to describe a state of affairs which guarantees what is known (like something magical), guarantees it as a fact. That word “seems” is carrying most of the weight. The grammar of “I know…” makes it look as though it names a relationship between a person and a fact, a relationship so secure that error is ruled out. But Witt reminds us of something we tend to forget, i.e., the expression “I thought I knew.” If “I know…” really did guarantee the truth of what follows, then “I thought I knew” would be incoherent. The very existence of that ordinary expression shows that “I know…” doesn’t function as a guarantee. It functions as a claim within a practice, and like any claim within a practice, it can turn out wrong.

OC 13 takes the argument a step further by examining what actually follows from someone’s saying, “I know.” Witt points out that you can’t infer “It is so” from someone else’s utterance “I know it is so,” not even if you add that the person isn’t lying. Sincerity (sincerity of conviction) doesn’t close the epistemic gap. Someone can be perfectly sincere and still not know what they claim to know. Then he makes an interesting point, viz., from the proposition “He knows that there’s a hand there,” it follows that there is a hand there. But from his utterance “I know,” it does not follow that he knows it. The difference is between knowledge as a standing within a practice, where the grammar of "knows entails truth, and the act of claiming to know, which is just an assertion that may or may not be justified. This is precisely where Moore’s confusion lies. Moore thinks he’s making an epistemological point, establishing that he has knowledge by saying “I know here is one hand.” But what he is actually doing is expressing his conviction that he can’t be wrong. The two uses of “I know…” look identical on the surface. In a philosophical context, where there is no practice to discipline the claim, the epistemological use of “I know…” has nothing to grip onto. All that remains is Moore’s subjective certainty, his unshakeable feeling that he can’t be mistaken, which he confuses with epistemic standing.

OC 14 and 15 bring this line of thought to a head. “That he does know takes some shewing,” Witt says, and then in 15, “It needs to be shewn 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.” This is where the argument lands. Saying “I know…” is this context is just an assurance, a personal commitment, and in ordinary life that assurance has its place within practices that can test it, correct it, or withdraw it. But Moore is trying to use “I know…” in a context where no such testing is possible, where the assurance is free of any epistemic practice that could establish its standing. Moore’s conviction, no matter how powerful, cannot do the work of showing. And that is Witt’s point. Knowledge requires more than conviction. It requires a practice within which the claim can be evaluated. Where that practice is absent, “I know” doesn’t fail by being false. It fails by being idle.

6

Post 5 Summary of OC 1-15

Before we continue, I want to step back and take stock of what Witt has said in these opening fifteen remarks, and how they connect to the framework I’m developing in this thread.

The surface argument is about what Moore is claiming. Witt is dismantling Moore’s claim to know that here is a hand by showing that Moore’s use of I know doesn’t do the work Moore thinks it does. Moore thinks he’s making an epistemological point, proving something about the external world by pointing to what he knows with certainty. But Witt demonstrates that Moore is confusing two things that look the same on the surface, viz., his conviction that he can’t be wrong, and the epistemic standing that I know has when it operates within an epistemic language-game. Moore doesn’t see the difference (nor do many other philosophers). But in a philosophical context, where there is no justificatory practice to discipline the claim, the epistemological use of I know has nothing to grip onto. All that remains is Moore’s subjective certainty, his unshakeable feeling that he can’t be mistaken, which he confuses with epistemic standing. Witt isn’t saying Moore is wrong about having hands. He’s saying that Moore’s conclusion that he knows he has hands is idle and misleading.

However, there’s more going on beneath the surface than a critique of Moore. Witt is also beginning to sketch a particular picture, though it’s still in early form. In OC 7 and again in OC 9-10, he points to a kind of certainty that doesn’t live in propositions or assertions at all but shows itself in action, in sitting at a bedside, in telling a friend to take a chair. This certainty isn’t something we arrive at through our justificatory practices. It’s the background against which justification operates. I’ve been calling this hinge certainty, and it’s the concept that will become central as we continue. In OC 5 Witt draws attention to the grammar of our epistemic practices, to the fact that what counts as evidence and what counts as a mistake are determined within those practices rather than being fixed in advance by some philosophical standard. And in OC 12-15 he establishes something that ties his criticism of Moore to the positive picture, viz., that knowledge requires a practice within which claims can be evaluated. Saying I know is giving an assurance, and that assurance only has standing when there is a practice that can test it. Moore’s I know fails not by being false but by being separated from any such practice. Moreover, Moore is relying on his internal certainty (his conviction) whether he recognizes it or not.

Three of my four senses of certainty are already in play by OC 15. Subjective certainty, the feeling of conviction, is what Moore is trading on and what Witt is pushing against throughout these opening statements. Hinge certainty, the non-epistemic certainty of what stands fast and makes inquiry possible, it’s what Witt is gesturing toward in his remarks about action and about what it would even mean to be wrong. And epistemic certainty, the legitimate standing a belief can have within a practice of justification, is what Witt is about to develop more fully in the remarks that follow, though he has already laid the groundwork by showing what I know looks like when it’s doing epistemic work. The fourth sense, absolute certainty in the logical and mathematical sense, has surfaced only briefly, in OC 10’s question about whether “2 × 2 = 4” is nonsense in the same way as “I know there’s a sick man lying here.” That question remains open and will stay open for much of the text. My taxonomy isn’t Witt’s. He doesn’t draw these four distinctions explicitly. But I think they’re required by the text, and the confusions he’s working against only become fully visible once you see that certainty is doing different work in different places.

Moyal-Sharrock has argued that hinges are fundamentally non-propositional, which aligns with what I’ve been calling the nonlinguistic level of hinge certainty but doesn’t capture the full picture, since many hinges do operate within our language-games. Pritchard’s view that hinges are commitment-constituting rather than knowledge-constituting maps well onto the point I’ve been making that hinge certainty is the foundation of epistemic language rather than an epistemic move itself. Where I diverge from both is in insisting that we need all four senses of certainty to navigate the text, not just a distinction between hinges and knowledge. The opening fifteen remarks of OC are setting the stage for everything that follows. Witt has cleared the ground by showing what certainty isn’t, namely the subjective conviction Moore appeals to. He has begun pointing toward what certainty is at the deepest level, the background that shows itself in how we act. And in the remarks that follow, he returns to the question of what legitimate epistemic certainty looks like when I know is doing epistemic work within a language-game. That’s where we’ll pick up next.

7

Hi Sam, I want to thank you for this thread. You are very thorough and the posts are quite detailed.

What I see at this point is that Wittgenstein is exposing a number of subtle differences and nuances in the use of “know”, within the context of an important categorical division. The important division is between the subjective certitude of “I know” (Moore’s use), and the sense of objective knowledge, “we know”.

The two are shown to be related to each other through practise, and this is what causes the difficulties involving many subtle differences and nuances. The common knowledge “we know” is supported by standard practise and therefore crosses into, and may even form the base for the subjective certitude of “I know”. So for example, “I know that the earth revolves around the sun”, just like Moore knows that this is one hand, and the basis of this certitude is the common practise which demonstrates that this is objective knowledge, something “we know”.

The principal issue which Wittgenstein seems to point to, is that the use of “I know” to represent subjective certitude extends into all sorts of situations, not necessarily supported by the common practise of objective knowledge. Sometimes “I know” is supported mainly by personal experience, and memory. This produces a vast variance of degrees of objective support for the subjective certitude of “I know”. At the two extremes are “I know” supported solely by personal experience, and “I know” supported by objective knowledge, with all degrees of mixing in between.

8

Thanks for the kind words.

I want to push back on a few points, though, because I think the framing of subjective I know versus objective we know doesn’t quite capture what Witt is after in the opening remarks. Witt isn’t setting up a contrast between individual and collective knowledge. Moore’s mistake isn’t that he’s speaking for himself when he should be appealing to what we all know. The problem is that he’s confusing his conviction that he can’t be wrong with the epistemic use of I know has when it’s doing real work within the language-game of epistemology. And that confusion wouldn’t be fixed by replacing I with we. Witt makes this point early on in OC 2 when he says that from its seeming so to everyone, it doesn’t follow that it is so. Collective conviction is still conviction.

I’d also push back on the idea of a spectrum of degrees between subjective experience and objective knowledge. Witt isn’t arranging knowledge claims on a continuum from less to more objectively supported. He’s making a point about the grammar of I know, about what the expression is doing in different contexts. The distinction between what I’ve been calling hinge certainty and epistemic certainty isn’t a matter of degree. They’re different in kind. A hinge doesn’t become an epistemic claim by accumulating more support, and an epistemic claim doesn’t become a hinge by being widely shared. They operate at different levels entirely.

You’re right that practice is central, and that’s an important observation. But I think its role in Witt’s account is more fundamental than providing objective backing for subjective conviction. Practice (general comment: think of a practice as an established use) is what gives I know its meaning in the first place. Without a practice to discipline the claim, I know doesn’t have weaker standing. It has no standing at all. That’s precisely what Witt is showing about Moore’s use of “I know here is one hand.” The words are grammatically well-formed but they’re idle because they’ve been lifted out of any practice that could give them epistemic life.

9

It’s not that the practices underlying hinges are essentially nonlinguistic. Rather, words are themselves actions which belong to larger systems of actions forming our bedrock practices. Someone can conceivably be persuaded to see the world our way through our words.

10

That’s supposed to read “…in the acting itself.” I’ll edit that.

11

OK, I think I see where I’m having a problem. It is your use of “epistemic”, and related terms like epistemology. This is not Wittgenstein’s word, so I don’t understand it as a part of any system or structure in the text. It is your word, so it is a part of your interpretive structure. However, I don’t think I understand how you are using it.

I offered a distinction between the subjective knowledge of “I know” (knowledge derived solely from my personal experience), and the objective knowledge which I termed as “we know”. As far as I understand, the way I used “we know” is to signify “epistemic knowledge”, what “know” is doing in the language-game of epistemology.

But you reject my “we know”, and instead use “epistemic knowledge”. But to me, they both mean the same thing. Can you explain to me why you think that the epistemic use of “know” is anything other than the meaning of “we know”?

I think I didn’t make myself clear. I didn’t mean a spectrum of degrees from one extreme to the other. I meant that each claim of “I know…” contains elements of both, personal subjective knowledge unique to myself, as well as elements of objective, epistemic knowledge, what “we know”. What I mean by degrees is that some claims are weighted on the one side, while others are heavier on the other, but each instance of “I know…” has elements of both.

This is why it is so difficult to separate out distinctly different uses of “know”. Each time it is used, there is a mix of each, and we can’t say that it’s exclusively one or the other. And I agree that it is a difference in kind, as you say, but still, in each instance when someone says “I know…” there is a mixing of the two kinds inherent within the statement. We separate them in principle, right here in this discussion, but in actual practise it would be very difficult to make a clear division.

You call them “hinge certainty” and “epistemic certainty”, to me it’s “I know” and “we know”. Forget the difference of terminology, what I think is important is that each instance of “I know…” has aspects of both types embedded. This is why there is so much nuanced meaning in the use of “I know”. Never is it all one or all the other. Each instance has aspects of both and they can be judged as weighted more toward one type or the other.

Not to be an intentional problem causer, but I’d have to disagree with this. The weaker sense of “I know” must be the primary sense. This sense refers to what I believe, based on my experience. The stronger sense develops from practise, discipline, and this is basically justification. Personal experience, and what I believe, as a living being, is prior to justification which comes through communication. So that weaker, unjustified sense, which is more like “I believe”, is the primary sense, as first sense.

However, the discipline of epistemology, sets justification as a criterion, and thereby seeks to exclude that weaker sense, even though the weaker sense inheres within common language use. Moore’s use is that common use. But epistemology seeks to separate that practise from the epistemic practise, through that criterion of justification, leaving it as nothing more than “I believe”. The problem though, is as I explained above, the two always comingle, and since the weaker is prior it inheres deeper, and it cannot be completely jettisoned.

12

Post 6 OC 16-18

OC 16 says that “If I know something, then I also know that I know it” amounts to saying “I am incapable of being wrong about that.” But then he adds the crucial qualifier: whether I am so needs to be established objectively. This is the same point again but stated more forcefully. The claim to be incapable of error is not something that can be settled from the inside, by reflecting on my state of mind (i.e., one’s conviction). It has to be established within a language-game that has the resources to determine whether a mistake was possible. By language-game I mean the shared activity within which the concepts like I know have their established use and where there are familiar ways of checking, correcting, and challenging claims. Moore’s gesture of holding up his hands and announcing what he knows is an attempt to settle the matter from the inside, from his conviction, and that’s exactly what Witt is saying can’t be done.

OC 17 shifts the angle in an interesting way. Witt asks us to imagine saying “I’m incapable of being wrong about this: that is a book” while pointing at an object, and then he asks: what would a mistake here be like? Do I even have a clear idea of what it would mean to be wrong? This opens up something that runs through the rest of OC. For certain propositions, the difficulty isn’t that we might be wrong but that we can’t coherently describe what being wrong would amount to. That isn’t because we’ve achieved some special epistemic standing. It’s because hinge propositions function differently from ordinary empirical claims. They belong to the framework within which error and correctness have their meaning. When Witt asks what a mistake would look like, he’s not inviting us to try harder to imagine one. He’s showing us that for some things, the absence of a coherent picture of error tells us something about the kind of role the proposition is playing in our lives. This is another early gesture toward hinge certainty, toward the idea that some things stand fast not because we’ve checked them but because they are the background against which checking makes sense.

OC 18 is important because it’s the first place where Witt gives I know a legitimate role. He says that I know often means “I have the proper grounds for my statement,” and that if the other person is acquainted with the language-game, he would admit that I know. The other person must be able to imagine how one may know something of the kind. This I know is doing real work. It’s not some free-floating philosophical pronouncement. It’s anchored in a shared activity where both parties understand what would count as a proper justification. This is what I’ve been calling epistemic certainty, a standing within a shared language-game where the claim is accountable to recognized standards. It’s the use of I know that Moore’s philosophical deployment obscures. In ordinary life, I know earns its place by being answerable to the language-game it operates within. When I say, “I know the shop is around the corner” and the other person is familiar with the context, they can recognize the claim as having proper standing. Moore’s problem is that he tries to use I know without any such answerability, as though the words are enough, detached from any shared activity that could evaluate whether the justification is proper.

13

Post 7 OC 19-20

OC 19 continues from OC 18’s point about I know having legitimate epistemic use. Witt says that “I know that here is a hand” can be continued with “for it’s my hand that I’m looking at,” and a reasonable person will not doubt this. In an ordinary context, with ordinary grounds, the claim has standing and does real work. But then Witt introduces the idealist, who says he wasn’t dealing with this kind of practical doubt. The idealist would claim there is a further doubt behind the ordinary one, a doubt about whether the external world exists at all. Witt doesn’t try to answer this doubt. He says it’s an illusion that has to be shown in a different way. This is important. The idealist takes a perfectly good epistemic exchange and insists that a deeper philosophical question lurks behind it. But Witt is saying this supposed deeper question isn’t a genuine extension of doubt. It’s a grammatical confusion. The idealist has detached the word doubt from any shared activity/use where doubting gets its force. You can’t refute this kind of doubt with more evidence, because the doubt calls into question the very framework where doubting operates. That’s why it has to be shown to be an illusion rather than answered on its own terms.

OC 20 makes this concrete. Witt points out that doubting the existence of the external world doesn’t mean doubting the existence of a planet, which later observations proved to exist. Doubting whether Saturn exists is a real doubt with a clear place in our language-games. We know what it looks like, what would resolve it, and what would count as getting it wrong. The philosophical doubt about the external world is nothing like that. Then Witt asks a pointed question: does Moore want to say that knowing here is his hand is different in kind from knowing that Saturn exists? If the two are the same kind of knowledge, then discovering Saturn should count as proof of the external world, which is obviously absurd. If they are different in kind, then Moore needs to explain what that difference is, and holding up his hands doesn’t accomplish that. What Witt is showing is that Moore is treating something that belongs to the framework of our engagement with the world as though it were an ordinary empirical discovery like finding a planet. The two operate at entirely different levels, and Moore’s proof collapses because he doesn’t recognize this.

14

Feb 26

Post 8 OC 21

OC 21 is one of Witt’s sharpest critiques of Moore up to this point. Witt says Moore’s view comes down to the following: know works like “believe,” “surmise,” “doubt,” and “be convinced,” except that the statement “I know” can’t be a mistake, so the truth of Moore’s assertion follows from the utterance. But Witt is pointing out what Moore’s position commits him to. If Moore’s “I know” is supposed to do the philosophical work of proving the external world, then it has to mean something stronger than an ordinary knowledge claim. It has to mean that error is ruled out entirely. Otherwise, it’s just an ordinary assertion and it proves nothing philosophically. Moore is caught in a dilemma he doesn’t see. Either his “I know” means what it ordinarily means, in which case it has no special philosophical force, or it means something stronger, something approaching a claim that error is logically impossible, in which case he’s overclaiming.

Witt reminds us of “I thought I knew.” People say this all the time, and the fact that they do shows that for empirical claims like Moore’s, saying “I know” doesn’t by itself rule out error. The grammar of “know” leaves room for being wrong in cases like these, even if it doesn’t for deductive conclusions or logical truths. But suppose Moore is right that in his case error really doesn’t make sense, that it’s inconceivable to be wrong about having hands. If that’s so, then saying “I know” adds nothing. An assurance from even the most reliable person that he knows cannot contribute anything, because the certainty isn’t coming from the assurance. It belongs to the framework (hinges) which allows epistemic language to function. And that framework can’t be secured by any assertion, no matter how confident, because it’s what makes assertion possible in the first place.

This pulls together the two main threads running through the text so far. Moore is confusing his conviction with epistemic standing. And the kind of certainty where error is inconceivable isn’t something that can be established by asserting I know. That kind of certainty belongs to what I’ve been calling hinge certainty. It shows itself both in how we live and act (nonlinguistic acts) and in propositions (linguistic acts) that stand fast within our language-games, but assertion is the wrong move for securing it regardless of whether the hinge is linguistic or nonlinguistic. Moore’s error is trying to say what can only function as a hinge. The phrase “I thought I knew” has implications well beyond Moore’s case, particularly for the Gettier problem, and I’ll return to it later in the thread.

15

I think I should point out, that your talk of hinges is misplaced here. You are reading ahead to things talked about later, and applying your own interpretation back onto things talked about here at the beginning. So far, he has portrayed certainty as ‘impossible to be wrong’. And the common use of “know” does not ever guarantee this. That is the issue visited by Plato in The Theaetetus.

You are implying that he has offered a framework of hinges to guarantee this certainty, but that is not the case. Since you are reading ahead, I will remind you that it is important to note that Wittgenstein later dismisses this definition of “objective certainty” (to exclude the possibility of error), as itself impossible, therefore a useless definition. He then settles on a completely different definition of “objective certainty”, and this is the type of certainty that hinges are proposed as supportive of, not the useless definition of “certainty” which mean error is impossible.

16

This is the very important issue with the difference between “I know” in the subjective sense, and “I know” in the objective (what you call “epistemic”) sense. In the subjective sense, “I know” signifies an attitude of certitude. But we all know that the person saying “I know” in this sense, could be wrong.

So Wittgenstein proceeds under the assumption that the objective (epistemic) form of “I know” ought to be able to ensure the certitude implied by the subjective form, as certainty. However he quickly discovers now, as Plato did, that this is not possible. It is impossible to exclude the possibility of error, and unreasonable to expect such a thing.

Then he moves along to propose something else, as the underlying, essential feature of objective (epistemic) knowledge. This is more like, ‘what would be accepted by a rational human being’. This provides his new definition of “objective certainty”, after dismissing ‘to exclude the possibility of error’. But I will argue that it’s really misrepresented as a form of “certainty”.

17

A Note on Non-Linguistic Hinges

I want to pause to address a point that’s likely to come up as this thread develops. Some philosophers resist the idea of non-linguistic or pre-linguistic hinges, and since I’ve been drawing a distinction between nonlinguistic and linguistic hinges, it’s worth saying something about why I think the resistance is misplaced and what I’m actually claiming.

First, what I mean by a non-linguistic hinge. I’m not talking about habits or reflexes. Tying my shoes a certain way is a habit. Blinking when something comes toward my eyes is a reflex. Neither of these is a hinge. A hinge is a certainty that functions as a condition of engagement with the world, something so fundamental that if it gave way, it wouldn’t be like discovering I’d made a mistake. It would be a collapse of the framework within which mistakes, checking, and trusting have any meaning at all. Non-linguistic hinges operate at several levels, all of them prior to language. At the deepest level are certainties about the basic coherence and continuity of experience itself. That the world doesn’t restart from moment to moment, that objects don’t vanish when I stop looking at them, that experience holds together as something continuous rather than fragmenting into disconnected instants. A child doesn’t check that the world is still there when they open their eyes. They just open their eyes and the world is there. These aren’t conclusions drawn from evidence. They’re the background against which anything could count as evidence or good reasons. At another level are certainties about our engagement with other beings. That other people and creatures exist and act, that I am not the only center of activity in the world. This isn’t a conclusion drawn from observing behavior. It’s already in place before any observation gets started. And at a more concrete level are certainties about physical interaction with the world. That solid ground will support me, that objects will behave predictably, that chairs will hold my weight. These are the ones most often cited as examples, but they don’t exhaust the category. What all of these share is that they function as conditions of engagement rather than as conclusions. If any of them gave way, the result wouldn’t be a revised belief. It would be a collapse of the framework within which believing, doubting, and revising have their point.

The resistance to this idea usually takes one of a few forms. Some philosophers hear an appeal to something private or ineffable, which seems to run against Witt’s later philosophy. But non-linguistic hinges aren’t private. They’re displayed in public action and shared engagement with the world. When I sit on a chair without checking whether it will hold me, that’s not a private inner state. It’s something anyone can observe. When a child reaches for a toy without doubting its existence, the certainty is visible in the reaching. These certainties are public in the relevant sense even though they aren’t propositional. Others argue that the moment you identify something as a hinge, you’ve already brought it into language, so all hinges are really linguistic. But this confuses the certainty itself with our description of it. The certainty lives in the walking, in the reaching, in the unreflective opening of the eyes, not in any sentence about these things. We can describe non-linguistic hinges in language, but that doesn’t make them linguistic any more than describing a sunset in words makes the sunset a linguistic phenomenon.

Some of the resistance also traces back to how the debate has been framed in the literature. Moyal-Sharrock pushed the non-propositional reading so hard that it generated strong pushback from philosophers who rightly pointed out that something like “the earth has existed for a long time” has propositional form and functions as a hinge. But the pushback often went too far in the other direction, dismissing the non-linguistic level entirely. The distinction I’m drawing between nonlinguistic and linguistic hinges is meant to avoid both extremes. It acknowledges what Moyal-Sharrock gets right about embodied pre-linguistic certainty while recognizing that some hinges do take propositional form within our language-games. The two levels aren’t in competition. They operate together. Non-linguistic hinges form the deepest stratum, the bedrock of unreflective engagement with the world that we share with animals and pre-linguistic children. They range from the most basic certainties about experiential coherence up through our embodied confidence in the physical world. Linguistic hinges build on that foundation, operating at a more articulated level within our shared discourse. Both are genuine hinges because both function as conditions that make inquiry possible rather than as conclusions within inquiry. The difference is the level at which they operate, not whether they deserve to be called hinges at all.

To make this concrete, consider how the levels relate. The continuity of experience, that the world holds together from moment to moment, is a non-linguistic hinge at the deepest level. No one formulates this as a proposition. It shows itself in every act of reaching, looking, and expecting. The certainty that other people exist and act is a non-linguistic hinge at another level, displayed in how a child responds to a parent’s voice before it has any words of its own. The certainty that the ground will hold my weight is a non-linguistic hinge at the level of bodily engagement, visible in how I walk without hesitation. Now compare these with a linguistic hinge like “the earth has existed for a long time.” This has propositional form. It can be spoken, and it looks like an empirical claim. But within our language-games it functions as part of the framework rather than as something we check or justify. It stands fast in our discourse the way the ground stands fast under our feet. The progression from the deepest non-linguistic hinges through to linguistic ones isn’t a hierarchy of importance. It’s a picture of how certainty is layered in our form of life, with the non-linguistic level forming the foundation on which the linguistic level builds.

18

I want to push back on a few points because I think there are some misreadings of what Witt is doing that are worth clarifying.

You say that Witt “proceeds under the assumption that the objective form of ‘I know’ ought to be able to ensure the certitude implied by the subjective form.” But Witt isn’t arguing under that assumption. He’s dismantling it. The whole thrust of the opening remarks is that subjective certainty, the feeling of conviction, can’t do epistemic work and was never in a position to do so. Witt isn’t looking for an objective version of “I know” that delivers what the subjective version promises. He’s showing that the subjective version was never promising anything epistemically meaningful in the first place. The direction of the argument is the opposite of what you’re describing.

The comparison with Plato is also misleading here. Witt isn’t reluctantly discovering that we can’t exclude the possibility of error, as though that were a disappointing result. He’s showing that the demand to exclude all possibility of error is itself confused, one that misunderstands the grammar of know. This isn’t skepticism. It’s a grammatical investigation into how our epistemic concepts actually function.

I also disagree with the idea that Witt moves toward something like “what would be accepted by a rational human being” as a new definition of objective certainty. Witt isn’t proposing a new definition at all. He’s describing how I know already functions within our language-games, where there are recognized procedures for checking and where claims are answerable to shared standards. That’s not a replacement theory of certainty. It’s an observation about ordinary epistemic life that was there all along but that philosophers keep overlooking because they’re chasing another idea.

Finally, your suggestion that epistemic certainty is “misrepresented as a form of certainty” concerns me because it assumes that real certainty must mean the total exclusion of error. But that’s exactly the conflation that causes confusion when reading OC. Epistemic certainty, the kind of standing a belief can have within a language-game where it’s supported by proper grounds and resistant to relevant challenges, is a genuine form of certainty. It’s just not absolute certainty in the logical or mathematical sense. Those are different senses of the word and keeping them apart is important to understanding what Witt is doing. Collapsing them into one and then concluding that anything short of absolute certainty isn’t really certainty at all is precisely the mistake the text is working to dissolve.

19

Yes, he is dismantling the certitude involved with “I know”, but that is at a slow pace, and it hasn’t progressed very far at this point in the text. At this point, he takes the subjective certitude, and tries to replace it with an objective form. So he moves from “I am certain that…” to “It is certain that…”.

He has dismantle subjective certitude, and now he looks to a further type, what it means to be objectively certain. However, he will proceed to show that “it is certain that…”, if it is assumed to mean that the possibility of error has been excluded, cannot be obtained. So this will be where the dismantling of “certainty” really starts.

He has shown that “I am certain that…” cannot exclude the possibility of error. So he inquires concerning “it is certain that…”, and proceeds to show that this does not exclude the possibility of error either. After this, he will proceed with another proposal for “objectively certain”, which does not exclude that possibility. So it becomes not explicitly a dismantling of certainty, but a compromising.

I disagree with this. He clearly describes how common use of “I know…” implies that the user believes oneself to be certain. That is what “I know this is a hand” adds to “this is a hand”. The speaker is expressing a form of certainty. However, we see that quite often when people use “I know” in this way, the speaker is in error. That kind of certainty cannot exclude the possibility of error. So, Witt looks for an objective form of “I know”, what you called epistemic knowing, and I called “we know”, questioning whether this objective form of knowledge could do what the subjective failed at, and this is to exclude the possibility of error. It seems like, “It is certain that…” ought to imply that the possibility of error has been excluded.

You can frame what he is showing, (that it is impossible to exclude the possibility of error), however you want. It might be framed as “a disappointing result”, or it might be a misunderstanding of “the grammar of know”. The different frames are not relevant, and there are other ways to frame it. I would say that it’s more about the grammar of “certainty” as implied by the title.

If you want, I can produce the quote where he redefines “objectively certain”. This is a little while after he dismisses “to exclude the possibility of error” as unreal because it’s impossible. However, this would be to jump ahead in the book, and that might not be appropriate at this time.

To me, it appears like you are getting distracted by your epistemic interests. The book is about “certainty”, it is not about “know”. He leads into the inquiry into “certainty”, through examples of “I know”, which seem to demonstrate some form of certitude. But “I know” is not the principal subject here, “certainty” is. So it’s actually you who is chasing another idea. You are looking for some sort of epistemology, thereby focusing on the word “know”. But Wittgenstein is looking for certainty, with very little concern for epistemology.

This is looking way too far ahead in the book now. However, I will point out that I think it is your interpretation, not mine, which leads to confusion. Look, you want to separate “epistemic certainty” from “certainty” in the “mathematical sense”. How could this proposal be anything other than the seeds of confusion? That is not Witt’s proposal, it is yours, and it is a recipe for confusion.

20

I’m not going to continue arguing with you. We just disagree and we won’t resolve these differences by continuing to argue.