Wittgenstein on Self-Knowledge: Nonsense, Certainty and Grammar


Consider Wittgenstein's claim

(C) "I can know what someone else is thinking, not what I am thinking." (PI, II.xi, p. 189) In the following, PI denotes the Philosophical Investigations, and BB the Blue and Brown Books.
What are we to make of this bizarre "drop of grammar"? My aim here is to unpack this remark and Wittgenstein's related assertions in the Investigations and in the Blue Book, such as the claim
(D) "It can't be said of me at all (except perhaps as a joke) that I know I am in pain." (PI, §246)
A first (trivial) observation to make is that interpretted naïvely, these statements (C) and (D) of Wittgenstein's just seem wrong. For on a naïve reading, Wittgenstein is endorsing what is sometime called (Temkin, p. 98) the non-cognitive thesis of avowals - namely, if A is the subject of an avowal p then A cannot know that p. Here an avowal is a first-person present-tense psychological statement, such as the expression of a sensation ("I have pain") or a thought ("I fear clowns"). So we are faced with the pressing question: Does Wittgenstein deny, in one way or another, the possibility of self-knowledge? And is this denial defensible?

The first task is to contextualize Wittgenstein's remarks, and in particular to determine what sort of denial, if any, Wittgenstein is really making; I take this up in section II. (Caution is certainly called for here -- as indicated by Wittgenstein's coy rhetorical question "What gives the impression that we want to deny anything?" (§ 305) from another passage of the Investigations, albeit one not directly related to the matter at hand.) If the impossibility of self-knowledge is an empirical proposition - that is, a statement of fact about our ability to know ourselves - then it is paradoxical to a crippling degree. I begin section II by pointing out some textual reasons why the empirical interpretation is clearly flawed. This leads to the conclusion that a more reasonable interpretation of Wittgenstein's assertions - and certainly closer to Wittgenstein's intent - is as grammatical propositions.

In section III, I move on to the question of the precise nature of the grammatical denial that Wittgenstein does (in fact) make, which involves a particular usage of the expression "I know ..." to deny uncertainty. This necessitates a discussion of Wittgenstein's principle of contrastive sense (a.k.a. "significant contrast") in relation to the expression of certainty. Ultimately I argue, following Temkin and Cook, for the view that Wittgenstein denies the possibility of self-knowledge only in following very weak sense: Wittgenstein holds that there is no fact about sensations and thoughts to the effect that I know I have them and what they are. Instead, Wittgenstein argues that sentences like "I know I have pain" are (true) grammatical propositions about the language game in which "I know that p" is used to express a posteriori certainty concerning p. What Wittgenstein denies is that "I know..." is used in such sentences in a manner parallel to the way it is used in empirical propositions like "I know he has pain." This is the sense in which assertions such as (C) must be taken.

Finally, I argue in part IV that this conclusion of Wittgenstein's is not particularly troubling. For Wittgenstein's "analysis" (blasphemy to call it such!) applies only to a very particular usage of the sentences "I know what I am thinking", "I know I am in pain", etc. The worry that Wittgenstein seems to exclude grammatically certain prima facie reasonable propositions is assuaged by the fact that our vast set of language games actually encompasses a much broader range of uses. This accounts for the problematic intuition that Wittgenstein is somehow forbidding us from saying something we would very much like to say. In fact, he is largely leaving well enough alone.


Let us begin by contextualizing somewhat the "offending" Wittgensteinian claims (C) and (D). Two locations in the Investigations are relevant. The first comes at the beginning of the famous private language argument.

In what sense are my sensations private?--Well, only I can know whether I am really in pain; another person can only surmise it.--In one way this is wrong and in another nonsense. If we are using the word "to know" as it is normally used... then other people very often know when I am in pain.--Yes, but all the same not with the certainty with which I know it myself!--It can't be said of me at all (except perhaps as a joke) that I know I am in pain. What is it supposed to mean--except perhaps that I am in pain? ... The truth is: it makes sense to say about other people that they doubt whether I am in pain; but not to say it about myself. (§ 246, emphasis added)
The second relevant passage is in part II of PI, in the midst of a discussion of grammar and nonsense, especially with regard to the phrase "I know...":
"I know what I want, wish, believe, feel,... " (and so on through all the psychological verbs) is either philosophers' nonsense, or at any rate not a judgment a priori.

"I know.. " may mean "I do not doubt.." but does not mean that the words "I doubt..." are senseless, that doubt is logically excluded.

One says "I know" where one can also say "I believe" or "I suspect"; where one can find out. ... It is possible to imagine a case in which I could find out that I had two hands. Normally, however, I cannot do so....

I can know what someone else is thinking, not what I am thinking. It is correct to say "I know what you are thinking", and wrong to say "I know what I am thinking." (A whole cloud of philosophy condensed into a drop of grammar.) (PI, II.xi, pp. 188-189, emphasis added)
We should start by eliminating the naïve empirical reading of the two bolded sentences, but to do so we must first clarify the nature of the assertion in each case.

Recall that in the private language argument Wittgenstein is concerned to dispense with the idea that there could be a "language which describes my inner experiences and which only I myself can understand" (§ 256). In dealing with the question of whether "only I can know whether I am really in pain" in § 246, one thing Wittgenstein does is dispatch with the possibility that ordinary language used to talk about sensations is private in the requisite sense. Wittgenstein observes that his interlocutor cannot be concerned with the ordinary sense of knowledge when he professes that only I can know whether I am in pain, for in fact it is very often that others know this, at least in the ordinary sense (e.g. when I tell them so without lying). And indeed, the interlocutor's exclamation in response - that the sort of knowledge at stake in this question involves a greater degree of certainty than ordinary knowledge - seems basically correct. We should take "I know I'm in pain" to mean something like "I am certain that I'm in pain."

If this is indeed the operative sense of "I know..." then the empirical interpretation of (C) and (D) is simply implausible. If I can be certain of anything, surely I can be certain that I am in pain (when I am). Similarly, if I think that p then I am certain that I think that p. So in what sense is it "incorrect" to say of me that I know what I think? If anything, it seems more "incorrect" to say in this empirical sense that I know what others think. After all, can I be certain of it? But as Wittgenstein asks (PI, II.xi, p. 191), "Am I less certain that this man is in pain than that twice two is four?" No! we are meant to respond. We can be equally sure of these things, given the proper evidence. That is not to say that there could be no uncertainty (as for example, when another man keeps his sensations bottled up, with no outward sign), but in correct circumstances there is none (as for example, when I see the man writhing in agony in a fetal position on the floor). So in the empirical sense, the accurate replacement for (C) is, "I can know what another is thinking, and I can know what I am thinking." Of course, the empirical interpretation, as exegesis of Wittgenstein's text, is ruled out by his rhetoric. As Stanley Cavell remarks (The Claim of Reason, p. 100) in treating the suggestion that I know I am in pain as a joke, Wittgenstein hardly indicates that I do not know (in the sense of being certain) this, but rather that the relationship between me and my pain is a deeper one than mere certainty.

So, we can safely conclude, Wittgenstein is manifestly not making an empirical claim about what I can and cannot know. The alternative is that his claim is grammatical; we must clarify its precise nature, however. Granting that certainty is what is at stake, it is clear that the argument condensed into the last sentence of § 246 is an appeal to the principle of contrastive sense: "A doubts p" is contrasted with "A knows p" in the sense of "A is certain that p". Wittgenstein defends the thesis that it makes no sense to say that I am uncertain whether I have pain, and therefore it makes no sense to state the negation of this. This, then, is the essence of Wittgenstein's denial: he rejects the notion that it is grammatical at all to say, for example, that I know I am in pain in order to deny uncertainty concerning this point. (Importantly, and we will return to this in section IV, other uses of "I know I am in pain" may be perfectly permissible.) To drive this point home, Wittgenstein compares the case of knowing I am in pain with knowing my intentions in § 247. The sentence "Only you can know if you had that intention" is offered as a grammatical statement concerning the use of "intention" (as well as "know"). Similarly, "I know I am in pain" as an expression of certainty is either nonsensical or else it must be grammatical. In the latter case, the content of the assertion is that it would be nonsense for me be uncertain whether I am in pain. Comparing the situation with the case of "Every rod has a length" in § 251, Wittgenstein is quite explicit that such grammatical sentences can be picked out by the inconceivability of their negation - a clear statement of the principle of contrastive sense.

The important distinction being drawn here is between asserting that I do not, in fact, have any uncertainty concerning p and that it would be nonsense for me to have uncertainty concerning p (because of the very grammar of our language). Wittgenstein makes this point quite eloquently in the Blue Book:

Consider the case in which we should be inclined to answer the question "Are you sure that it is this you wish?" by saying: "Surely I must know what I wish". Now compare this answer to the one which most of us would give to the question: "Do you know the ABC?" Has the emphatic assertion that you know it a sense analogous to that of the former assertion? (BB, p. 30)
Surely not, we are meant to answer his question. The former is an expression of the illogicality of uncertainty for grammatical reasons, the latter is an expression of a lack of uncertainty. A sentence expressing self-knowledge of an avowal (e.g. "I know what I wish"), insofar as it is concerns certainty, is always of the former sort. At this juncture it is important to make a textual observation. Wittgenstein seems to contradict himself when he says in § 247 that "here 'know' means that the expression of uncertainty is senseless", and in PI, II.xi, that "'I know...' may mean 'I do not doubt...' but does not mean that the words 'I doubt...' are senseless" (p. 188). The explanation is that the latter passage is expressing a sort of normative judgment concerning the use of 'I know...'. In fact, as will be discussed in section IV, one can in fact use the grammatical sense of "I know I am in pain" in some circumstances, as when explaining the meaning of "knowing" or "pain" to someone unfamiliar with the words. However, phrasing this as "I know I am in pain" instead of "It doesn't make sense to say that I an uncertain whether I am in pain" certainly causes needless confusion, since the first phrasing masks the grammatic proposition as an empirical one in the manner of § 251. Thus Wittgenstein is perhaps justified in decrying such usage as nothing more than an empty source of philosophical puzzlement, and this may explain the strong (initially so troubling) phrasing of his assertion (C).


We thus see that Wittgenstein's main claim in (C) and (D) is that "I know what I think", "I know I am in pain" and the like, as expressions of certainty, are grammatical propositions, and thus senseless when construed otherwise (in that they say absolutely nothing about thoughts or pains). He seeks to establish this by the principle of contrastive sense, putting forth "I am uncertain that p" as a contrast to "I know that p" as it is used here. There are several implicit claims he makes. The first is that being certain really is the operative sense of "knowing" in these contexts, and (relatedly) that "I am uncertain that p" is a correct contrast to "I know that p". The second is that such contrasts are in fact illogical; i.e., sentences like "I am uncertain whether I am in pain" and "I am uncertain what I think" are logically impossible.

As mentioned in section II, certainty is brought into the question of the privacy of sensations by the interlocutor, who in this case speaks for the classical Cartesian view that sensations are private. (See Temkin's article for a discussion of the relationship between Wittgenstein's view and the Cartesian one.) As Cook (p. 285) emphasizes, insofar as Wittgenstein is concerned to refute the Cartesian view that, for example, "Only I can know I am in pain", he is free to presume the essential Cartesian premiss that there is a genuine use of "I know p" to assert certainty concerning p; this is essentially the point of the interlocutor's exclamation "Yes, but all the same not with the certainty with which I know it myself!" at § 246. But beyond such logical expediency, examples can be adduced in support of this linguistic claim. Cook gives the example of "Yes I know it's raining; I'm looking out the window!" said over the telephone to a skeptical friend, to illustrate (convincingly) that "I know p" can be an expression of certainty. Stated in these terms, Wittgenstein's aim is to argue that avowals of which I am the subject cannot (in our language) be substituted for p in this usage.

So much for certainty. We now turn to the slightly subtler question of whether the principle of contrastive sense is actually operative in this situation. Wittgenstein is correct in claiming that sentences such as "I am uncertain whether p" (for an avowal p as above) are always inherently meaningless, but the case is not entirely clear cut. The intuition behind Wittgenstein's claim is quite clear - it is the same as that underlying the Cartesian cry that my knowledge of my own pain is more certain than another's knowledge of it. That is, consider a purported counterexample in which a needle has been jabbed into my arm, and I feel it as an unpleasant sensation. If I were to claim uncertainty as to whether I am in pain, the natural response is to assume that I simply do not know how "pain" is used, that I am not a full participant in the language game of pain reports. A true counterexample, in contrast, must be of the sort Wittgenstein considers in § 288, in which I say something like "Oh, I know what 'pain' means; what I don't know is whether this, that I have now, is pain." Call thise sort of utterance an expression of descriptive hesitancy, a term due to Temkin. Wittgenstein shrugs his shoulders at this point and writes such instances off as "queer reaction(s) which we have no idea what to do with." Temkin (p. 107), however, offers the good example of borderline cases between pain and tickle as apparently legitimate cases of such descriptive hesitancy. (Such cases are familiar to me, for example, having received vaccinations as a child from a pediatrician skilled at the art of distraction. Caught up in his banter, by the time I registered the sensation in my arm as being caused by a needle, I was through with the ordeal.) Do such instances undermine Wittgenstein's argument? Can one simply shrug one's shoulders at them?

Certainly we do know what to make of these cases, so ignoring them will not suffice. Instead, Temkin (p. 108) is correct in suggesting that in such pathological situations, "I am uncertain whether p" does not stand in suitable contrast to "I am certain (or I know) that p". Rather, the illogical contrasts are closer in form to "p, but I am uncertain whether p", in which form their nonsensicality is made manifest. In particular, when hesitating about whether to describe a sensation as a pain or a tickle, I actually am expressing one of the following two sorts of uncertainty (assuming I am a full participant in the language game of sensations and avowals). Either I find that my sensation so far leads me to avow both p and q, both of which seem to describe it well, or else I find that it leads me to avow neither p nor q, each of which seems somehow inadequate as a description. This is closer in spirit to a class of purported counterexample considered by Cook, namely when avowals are conflated with self-diagnoses:

Thus, a man might say, "Never mind the aspirin; I didn't have a headache after all. It was only this tight hat I've been wearing." (Cook, p. 294)
Perhaps a more cautious man might have expressed his predicament earlier not by saying he had a headache, but rather that he was uncertain whether he had a headache (or his hat was simply too tight). Nonetheless, a man might indeed speak as above, and this case obviously does not contrast with his saying "I am certain I have a pain in my head" or "Surely I know I have a pain in my head". Vacillation between possible diagnoses is not precisely analogous to vacillation between two potential avowals without diagnostic character. But the two cases share an important feature. The uncertain migraineur might have said "I have a pain in my head but I am uncertain whether I have a headache", but could not have said (sensically) "I have a pain in my head but I am uncertain whether I have a pain in my head". Likewise, the vacillating tickle victim might say "I have a tingling sensation in my arm but I am uncertain whether to call it a tickle" but not "I have a sensation in my arm just like tickling, but I am not certain whether it is a tickle I feel". All Wittgenstein is rejecting as nonsense are sentences of the latter sort; another example of Temkin's - rephrasing Wittgenstein's from § 288 - is "What I am experiencing now certainly feels like pain, but perhaps it isn't really pain at all" (Temkin, p. 108).

Such considerations suggest that, properly construed, Wittgenstein seeks to make no earth-shattering claims. His profession of innocence alluded to parenthetically above -- "What gives the impression that we want to deny anything?" (§ 305) -- is at least partly in earnest. When I know that I have pain with the absolute certainty which so moves the Cartesian, it is because I actually have the sensation of pain. In these circumstances Wittgenstein's rejection of my expression of uncertainty as to whether I have pain seems perfectly correct. Probably more putative counterexamples along the lines of § 288 (Temkin's descriptive hesitancy) or Cook's self-misdiagnosis could be contrived. But the Wittgensteinian response in every case would likely be along the lines of, "but that's not what I'm talking about!"

So far we have established that the principle of contrastive sense is "in full working order" as Wittgenstein applies it to the case of propositions expressing self-knowledge of avowals; a proper contrast has been set up and the contrasting propositions are reasonably seen to be nonsense. Wittgenstein's conclusion -- that expressions of self-knowledge of avowals are similarly nonsensical, at least if taken to "mean" that the subject does not (in point of fact) have any uncertainty concerning the avowal -- is now a formal consequence of this principle. As it operates here, the principle of contrastive sense simply states that when the possibility of error does not exist, a proposition is not empirical. The plausibility of such an application of this principle is clear by analogy: Wittgenstein's "The room has length" (BB, p. 30) is another example of a statement seen to be grammatical in this manner; the very fact that it makes no sense to say a room lacks length shows that "the room has length" is not an empirical fact about the room!


In this concluding section, I now turn briefly to the question of whether Wittgenstein's assertions (C) and (D), properly interpretted as saying no more or less than what Wittgenstein actually means to say, are really so objectionable in the first place. In "The Availability of Wittgenstein's Later Philosophy" (p. 90), Stanley Cavell remarks upon the "extraordinary idea", attributed to Wittgenstein, that we cannot know what we think and feel. What is extraordinary here, he makes clear, is the assertion (C) construed as an empirical proposition. We have seen already that it is not such. But Wittgenstein's assertions tend to cause a persistent worry even when they are construed as grammatical. Cavell expresses this in The Claim of Reason (in a delightful show of quasi-self-referentiality) as follows. After quoting (C) (actually, the two subsequent sentences from PI, II.xi; see the excerpt above), he complains:

I find this much less convincing... I sort of know what it means: it is not merely that I do not have to look at myself to see or to learn whether I am thinking; but that I do not, as it were, take stock of the contents of my mind and determine whether they are in there. And yet it would be correct to say: "What you say keeps not ringing true to me. I keep having this worry-- I can't quite put my finger on what it is. -- Ah yes, I have it, I know what I am thinking..." (p. 101, emphasis added)
The worry captured by Cavell is something like a generic concern for what happens when "a sentence is called senseless... a combination of words is being excluded from the language, withdrawn from circulation" (PI, § 500). In the case in question, how is it possible that there could be no such thing in our language as "knowing what I am thinking"?

The answer to this question, and the salve to Cavell's worry, is that Wittgenstein does not mean to declare unintelligble, to withdraw from circulation, a combination of words per se, but rather a particular use of a combination of words. One obscure use -- the grammatical -- is certainly not ruled out by (C). Like the example of "Only you can know if you had that intention" from § 247, "(Surely) I know I am in pain (when I am)" is a perfectly valid proposition when regarded as a fact about grammar itself, as when teaching someone the concepts "pain" or "knowing". But there are other more mundane usages of "knowing" which are also exempt from Wittgenstein's critique. Indeed, we saw in section II that Wittgenstein's argument is not concerned with the "ordinary" sense of knowing or cognizing; viz., it is certainly the case that I am cognizant of the fact that I have pain (when I do), just as I am cognizant of the fact that others do (when they do and they tell me or moan or ...). So Wittgenstein's argument certainly does not exclude this ordinary sense of knowing from the language. And there are many other, less ordinary, senses of "knowing" which are also not excluded. Cook gives the example of an exaspirated "I know I'm in pain, but we can't afford a doctor" (p. 205), and Temkin offers the conciliatory "I know I'm in pain, coach, but it's important to me that I play in this game" (p. 103).

Arguably, Cavell's example falls into this class of exotic uses; the sentence "Ah yes, I have it, I know what I'm thinking" actually expresses an ability to (verbally) formulate the thought in question appropriately (which ability was presumably lacking until the sentence was uttered, or just before). As with the examples of the uncertain migraineur and the vacillating tickle victim from section III, this sort of worry is best tackled by considering contrasting sentences. If the contrast to Cavell's utterance, describing his "mental state" before his Ah yes! moment of epiphany, is indeed a senseless instance of "p, but I am uncertain whether p", then the utterance itself must be taken to mean something like "Ah yes, I have it! There exists a thought t such that I know (I am certain) I'm thinking t." This does not accurately capture the sense of Cavell's example, for if it did the contrast would spell out along the lines of "There exists a thought t such that I am uncertain whether or not I'm thinking t". Of course if there were a well-formed thought t, Cavell would know whether or not he was thinking it. The point of the example is that there is not such a well-formed thought; one imagines Cavell possessing a whirl of possibly contradictory mental representations ("thoughts") which resist his attempts at verbalization until the moment of epiphany. The proper contrast (for the example, but not for making Wittgenstein's argument go through) is the negation of the existential quantifier, rather than the negation of the assertion of certainty regarding the quantified variable.

The examples of section III illustrate that a genuine counterexample to Wittgenstein's assertion (C) must involve a sensation or avowal which suggests its identity to the subject in a definitive manner but the identity of which the subject nevertheless doubts. A lack of such definitiveness is crucial to Cavell's example, which indicates that there is little hope that it can be parlayed into a credible instance of a sentence one would like to use, but which Wittgenstein grammatically "forbids" one from using. Generalizing these remarks, the upshot is that Wittgenstein's assertions (C) and (D) are hardly proscriptive. An example of an utterance with definite sense which Wittgenstein seems to be declaring senseless must, of necessity, not really be what Wittgenstein is talking about. And this sensible conclusion is rather reassuring, according, as it does, with Wittgenstein's conception of philosophy as a merely descriptive enterprise.

Works Cited

Stanley Cavell, "The Availability of Wittgenstein's Later Philosophy." The Philosophical Review Vol. 71, No. 1 (Jan., 1962), pp. 67-93.

---, The Claim of Reason. Oxford University Press, 1979.

John W. Cook, "Wittgenstein on Privacy." The Philosophical Review, Vol. 74, No. 3. (Jul., 1965), pp. 281-314.

Jack Temkin, "Wittgenstein on Epistemic Privacy." The Philosophical Quarterly, Vol. 31, No. 123 (Apr., 1981), pp. 97-109.

Ludwig Wittgenstein, Philosophical Investigations. Trans. G.E.M. Anscombe. 3rd ed. Malden, MA: Blackwell, 2001.

---, The Blue and Brown Books. New York: Harper & Row, 1958.