Disjunctivism, Skepticism, and the First Person *

One can sometimes tell what someone else feels or thinks by seeing and hearing what he says and does. It is very common for philosophers to interpret this idea so that ‘what he says and does’ is taken to allude to a basis for knowledge of what the person feels or thinks. The thought of a basis here has two elements. The first is that the basis is something knowable in its own right ... knowledge of the basis could have this status—be knowledge—independently of the status of what it is a basis for. The second is that judgments about what the person feels or thinks emerge as knowledgeable in favourable cases because of an inferential relation in which they stand to the basis. The notion of a criterion, as used by Wittgenstein in connection with this sort of knowledge, is often interpreted on these lines. 1

intellectual project, which Descartes founded-the project of "modern" epistemology. To be gripped by skepticism, of this form, is to be in the grip of this project. But a central thought of this essay is that this project is responsive to a more general epistemological ambition.
Understanding the project in this way makes it possible, not merely to isolate what it is about the project that generates the skeptical consequence, but to point to the possibility of a position that dismantles the project, and in dismantling it avoids the consequence, but without therein abandoning the general ambition to which the project is responsive. That enables a path to be trod between the Scylla of the modern project, and its attendant skepticism, and the Charybdis of contemporary epistemology, which also rejects the project, but in rejecting it, equally abandons the general ambition. Insofar as this general ambition is not kept in view, it will be tempting either to understand McDowell's disjunctivism as a contribution to the modern project, or to try to understand it as a contribution to contemporary epistemology.
And both will drain disjunctivism of its anti-skeptical import: the first, because insofar as it is understood as a contribution to this project, it is powerless to avoid Cartesian skepticism; and the second, because as contemporary epistemology has abandoned the general ambition, when any contemporary position speaks of "knowledge" it changes the subject, and as such is incapable of engaging with Cartesian skepticism.
4. The plan is as follows. The essay will begin by presenting the project of epistemology at a very general level ( § §5-14). That will allow the project of modern epistemology to be seen as a specific version of this general project-one that inevitably leads to skepticism ( § §15-20).
And it will show that, insofar as disjunctivism is understood as a contribution to this more specific project, it is powerless to resist the skeptical conclusion ( § §21-24). But the project of modern epistemology is incoherent. And the essay will bring this out, by pointing up an assumption that holds the project in place, but which the project cannot sustain ( § §25-34).
This assumption is a certain way of understanding the first person. The project must understand the first person in this way; and yet by its own lights it cannot understand the first person in this way. And without this way of understanding the first person, the project of modern epistemology falls to the ground. But the project of epistemology remains, insofar as it is understood in the very general terms with which this essay began. And insofar as disjunctivism is understood as a contribution to this purified project, it yields a satisfactory response to skepticism ( § §35-40). If the modern project is taken for granted, then the antiskeptical import of disjunctivism will remain obscure. Only by dismantling this project, in the way explained here, can the fundamental insight in disjunctivism emerge. *

The founding text of modern epistemology is Descartes' Meditations on First Philosophy.
This begins by describing something that happened to the author of the Meditations at some point in his life: "Some years ago I was struck by the large number of falsehoods that I had accepted as true in my childhood, and by the highly doubtful nature of the whole edifice that I had subsequently based on them." 5 This sentence uses the first person. And that is no accident, because the experience that it reports-"the experience of falsehood"-is at once part of its author's autobiography, and something that anyone is in a position to report by using "I". The sentence exhibits the unity of particularity and universality that characterizes the first person. And this unity is fundamental to the Meditations as a whole. Although it is possible to read the Meditations as autobiography, its author intends it to be read in a different way-not by disregarding its use of the first person, but by exploiting the universality that "I" contains. This is something that G.E.M. Anscombe notes, in her discussion of one of its central arguments: "The first-person character of Descartes' argument means that each person must administer it to himself in the first person." 6 And this character is central to the project of modern epistemology inaugurated by Descartes.
6. To bring this out, imagine three books, each of which is primarily distinguished by how its author intends it to be read. The author might express his intention, in each case, as follows: (1) "The reader is to come to understand that I am like this"; (2) "The reader is to come to understand that everyone is like this"; (3) "The reader understands (as he would put it) 'I am like this'". Book (3) is like book (1) but unlike book (2) in that it is written in the first person rather than the third-but whereas book (1) seeks to communicate an understanding that is to be received in the third person, by the reader coming to understand (as he would put it) "The author is like this", book (3) invites the reader to articulate in the first person the very selfunderstanding that it seeks to articulate. By contrast, book (3) is like book (2) but unlike book (1) in that the understanding that it seeks to articulate is universal rather than particular-but whereas in the case of book (2) this is because it propounds a generalization by using "everyone", in the case of book (3) it is because it invites every one of its readers to articulate this understanding by using "I". A text in epistemology in the tradition inaugurated by Descartes is in these respects akin to book (3), rather than to books (1) or (2). 7. That brings out the central way in which this tradition differs from contemporary epistemology. A typical text of the latter is akin to book (2), in that it seeks to communicate an understanding that says that anything of a certain kind-a "subject", or a "self"-who satisfies a certain condition-which it goes on to specify-exemplifies a certain propertywhich it signifies by the word "knowledge". And it is quite possible that this understanding concerns a "subject", or a "self" who does not himself have this understanding: one of the doctrines of this kind of epistemology is that those it concerns (as it likes to say) "need not even have the concept of knowledge". By contrast, any instance of the self-understanding that a text in modern epistemology seeks to articulate will be had by the very one that it concerns: it is not possible that an instance of this understanding concerns someone who does not himself have this understanding. It is interesting that much contemporary epistemology is written, either in the first person, or by using a pronoun that is naturally taken to issue the same invitation-such as "one", or even "you". In a text of contemporary epistemology, these devices cannot be read as issuing this invitation; at best, they function as a plangent reminder of the kind of understanding that it is not seeking to articulate. * 8. I have just spoken of articulating an instance of self-understanding; and I might equally have spoken of self-knowledge, or self-consciousness: at the present level of generality, these come to the same. To understand the project of epistemology, it is first necessary to understand what the idea of articulating an instance of self-understanding comes to. 9. At the most general level, to articulate an instance of self-understanding is to express a thought. And this could be any thought: it does not matter what it is about.
10. That might seem wrong. It is tempting to think that to articulate an instance of selfunderstanding is to articulate a thought that is about something of a certain kind, a "self", or a "subject". As it is natural to say: "self-knowledge is knowledge of oneself". And it is tempting to think that this explains self-knowledge as knowledge of a certain kind of thingsignified here by the reflexive pronoun. But at the present level of generality, nothing like that is included in the idea of self-knowledge. It is tempting to think that it is not excluded either-that, when the descent is made to a more specific level, the articulation of an instance of self-understanding will at least include the expression of a thought about a "self". This is the position of modern epistemology. And it is the source of its difficulties-as will be seen.
But to bring these difficulties into focus, and to bring out the contrast between modern epistemology, and the project of contemporary epistemology, it will help to remain at the present level of generality, at least at the outset.
11. The project of articulating an instance of self-understanding could be described as that of articulating an understanding of a thought "from within". To articulate such an understanding just is to express the thought that the understanding is an understanding of. This contrasts with the project of articulating an understanding of a thought "from outside". To articulate such an understanding is to express, not the thought that the understanding is an understanding of, but a distinct thought that is directed towards this thought. There will later be cause to return to the putative project of articulating an understanding of a thought "from outside" (see § §30-31). But the very idea that articulating an instance of self-understanding is articulating an understanding "of" a thought is liable to mislead here, because it is liable to suggest that the understanding itself consists in a thought that is directed towards a further thought: the first thought is the understanding, and the second thought is the thought that the understanding is "of". But the idea of the thought that the understanding is "of" is merely the idea of the thought that the articulation of the understanding expresses. We might say that the thought, and the understanding "of" the thought, are the same. But it is better to avoid the misleading word "of" altogether, by calling the understanding "self-understanding". 12. At the present level of generality, the project of epistemology is a species of the project of articulating an instance of self-understanding, in that it seeks to express a thought. But as it is a species of this project, this is not all that it seeks to do. It seeks, not merely to express a thought, but to express a thought in a manner that reveals it to be true. This is what makes it distinctively epistemological. Thoughts are expressed in language, specifically in sentences.
And to use a sentence is to grasp the thought that the sentence expresses. The idea of expressing a thought in a manner that reveals it to be true is to be understood in this light. To express a thought in a manner that reveals it to be true is to use a sentence of which the following holds: to use this sentence is to grasp not merely the thought that it expresses, but the truth of this thought. And insofar as the truth of this thought is revealed, so is that of every thought that is understood to follow from the truth of this thought: to reveal the truth of the first thought is to reveal the truth of every thought that is understood to follow from it. At the present level of generality, the idea of knowledge just is the idea of the capacity to express a thought in a manner that reveals it to be true. The idea of skepticism about knowledge in general just is the idea that there is no such thing as this capacity. And the idea of skepticism about knowledge of the truth of thoughts of a specific sort is the idea that there is no such thing as the capacity to express a thought of this sort in this manner.
13. It is worth bringing out how the idea of self-understanding that is internal to this project differs from the idea of self-understanding presupposed by contemporary epistemology. It belongs to this branch of epistemology to reject the KK thesis: that if someone knows something, then he knows that he* knows it-where the function of the star is to indicate that the knowledge reported by the consequent is self-knowledge. 7 The standard objections to this thesis all assume that the thesis is an instance of the (absurd) idea that, if someone knows something, then he knows something else: the knowledge reported by the consequent is knowledge of the truth of a further thought, distinct from the thought that constitutes the knowledge reported by the antecedent. This assumption is (to my knowledge) never made explicit, let alone argued for; but because the knowledge reported by the consequent is selfknowledge, the effect of the assumption is that self-knowledge is itself pictured as knowledge of the truth of a further thought. This explains why contemporary epistemology shows no interest in the possibility of self-knowledge: what reason could there be to think that the possibility of this further knowledge has any bearing on the possibility of the knowledge that the antecedent reports-short of the truth of the (absurd) KK thesis? But self-knowledge does not consist in a further thought. For someone to have the self-knowledge with which the project of epistemology is concerned is, simply, for him to have the capacity to express a thought in a manner that reveals it to be true. His having this capacity is not a condition over and above his having the knowledge that the antecedent of the thesis reports: it is nothing other than his having this knowledge. 8 And for this reason, it is possible to cast the idea of the kind of self-understanding with which the project is concerned in terms of the KK thesis properly understood: if someone knows something, then he knows that he* knows it-simply in that he has the capacity to expresses it in a manner that reveals it to be true. Contemporary epistemology thinks that the way to reject skepticism is to reject KK. But to reject KK properly understood is to reject the very idea of knowledge, whereas to reject-or endorsewhat contemporary epistemology understands by KK has no bearing on skepticism at all.
14. Contemporary epistemology misunderstands the KK thesis because it takes for granted that the self-knowledge reported by the consequent is knowledge to the effect that a certain property is exemplified by a "self", and as such is knowledge of something other than what is 7 I owe the device of the star to Hector-Neri Castañeda, "He: A Study in the Logic of Self-Consciousness" (1966), in his The Phenomeno-Logic of the I: Essays on Self-Consciousness (Indiana: Indiana University Press, 1989). 8 So understood, KK is not a target of the "anti-luminosity" argument. This argument turns on the principle that if at one time someone knows X, then at a very slightly later time X equally obtains. Whatever plausibility this principle has rests on the idea that someone's knowing X, and X, are distinct conditions: it would be absurd to claim that if at one time X obtains then at a very slightly later time X equally obtains. And it is an assumption of the argument that someone's knowing that he* knows something, and his knowing it, are examples of someone's knowing X, and X respectively, and as such are themselves distinct conditions. But properly understood KK says that someone's knowing that he* knows something, and his knowing it, are not distinct conditions at all. His knowing that he* knows something, and his knowing it, are each simply a matter of his having the capacity to express a thought in a manner that reveals it to be true. (For the "anti-luminosity" argument, see Timothy Williamson, Knowledge and its Limits (Oxford: Oxford University Press, 2000).) known in the knowledge reported by the antecedent-which may not concern a "self" at all (and, even if it does, is not to the effect that this "self" exemplifies the property that figures in the content of the knowledge reported by the consequent). But at the present level of generality, the idea of self-knowledge internal to the KK thesis is not the idea of knowledge of a "self". The office of the formulation that constitutes the consequent of the thesis-and, specifically, the office of the star on the pronoun-is to show that the knowledge that the consequent reports is self-knowledge. But at the present level of generality, this is merely to show that the knowledge reported by the consequent consists in the capacity to express the thought that figures as the content of the knowledge reported by the antecedent in a manner that reveals this thought to be true. And this capacity just is the knowledge reported by the antecedent. The idea of knowledge is the idea of self-knowledge, in that it is the idea of a capacity to express a thought in this manner. At the present level of generality, this is all that the idea of self-knowledge internal to the KK thesis comes to.
15. Within modern epistemology, the case for skepticism comes into focus once a descent is made, from the present level of generality, to a level at which self-understanding does concern a "self", or a "subject". And this level was operative earlier in this essay, when an instance of self-understanding was said to concern someone-the very one who has the instance of self-understanding at issue. At this level, the articulation of an instance of selfunderstanding is not merely a matter of expressing a thought; it is also a matter of identifying the one who thinks the thought as its thinker. Suppose that the thought is expressed by the sentence "p". Although one way to articulate an instance of self-understanding is to use this sentence, its use does not fully articulate the instance of self-understanding, because the thought that its use expresses does not itself identify the one who thinks the thought as its thinker. Fully articulating the instance of self-understanding requires using the distinct sentence "I think p": the thought expressed by this use contains the thought expressed by "p", and as such equally expresses this thought; but it also goes beyond this thought, because it also identifies the one who thinks it as its thinker. And it is in this light that modern epistemology is to be understood.
16. The experience of falsehood with which Descartes begins the Meditations specifies the kind of thoughts with which modern epistemology is primarily concerned: it is primarily concerned with thoughts that are capable of being true, or false. And "p" expresses such a thought. But a use of "I think p" does not. Insofar as someone uses this sentence, its use expresses the thought expressed by "p". But his use of this sentence does not merely express this thought. It equally expresses a further thought that contains the thought expressed by "p". So, it expresses two thoughts: the thought expressed by "p", and a further thought that contains this thought. In what follows, I will call the second of these thoughts "the thought that a use of 'I think p' expresses", and the first "the thought that a use of 'I think p' contains". And the crucial point, for present purposes, is that the thought that a use of "I think p" expresses cannot be false. It is self-verifying. This is for the following reason. In using "I think p", the user thinks the thought that his use of this expresses. And as this thought contains the thought expressed by "p", in thinking this thought, he equally thinks the thought expressed by "p". The thought that his use of "I think p" expresses is true just in case he thinks the thought expressed by "p". So, the thought that his use of "I think p" expresses is true. Insofar as "I think p" is used at all, the thought that its use expresses is true. It is incapable of being false. And in using "I think p", its user understands this: he grasps the thought that his use of this sentence expresses, and in grasping it, he grasps that it is incapable of being false. In using "I think p", therefore, he grasps the truth of the thought that his use of this sentence expresses. And as such, his use of this sentence may be said to reveal the truth of this thought.
17. This sets the agenda for modern epistemology. A use of "I think p" reveals the truth of the thought that it expresses. As such, the thought that it expresses is something that its user knows. But the same is not true of "p": if the thought that this sentence expresses is to count as something that its user knows, then there must be a way of expressing it that equally reveals it to be true. A use of "p" does not do this. And nor does a use of "I think p": it reveals the truth of the thought that it expresses, but it does not reveal the truth of the thought that it contains. The search is on, then, for a way of expressing the thought expressed by "p" that does reveal its truth.
18. It is at this point that a distinctively Cartesian assumption comes onto the scene: the sought-for manner of expression must be the same as a use of "I think p", not only in that it must contain the thought expressed by "p", but in that the thought that it expresses must equally be self-verifying. It a question, however, how it can be. To reveal the thought that "p" expresses to be true, it is not enough simply to express it; it seems that its expression must somehow reveal it to be in a nexus with something else, and through revealing it to be in this nexus reveal it to be true. And it seems that this "something else" cannot simply be another thought. On the contrary: it seems that it must be something given to the senses: only through being revealed to be in a nexus with this can the truth of the thought be revealed. It is tempting to think that a sentence beginning "I perceive …" is a form for expressing a thought that bears on what is given to the senses. And as such, it is tempting to think that "I think p because I perceive …" is a form for revealing the thought expressed by "p" to be in a nexus with what is given to the senses, and as such for revealing it to be true. But given the Cartesian assumption, it can do this only insofar as the thought expressed by a use of a sentence beginning "I perceive …" is self-verifying: only then can the thought expressed by "I think p because I perceive …" be self-verifying, and only insofar as this thought is selfverifying can the Cartesian assumption be satisfied. So, the question arises as to how the thought expressed by a use of a sentence beginning "I perceive …" can have this feature.
19. And it is at this point that it is tempting to fall into The Myth of the Given. It might seem that a sentence beginning "I perceive …" can express a thought that is self-verifying, insofar as the meaning of whatever takes the place of the dots is determined simply by what is given to the senses at the time of its use. To work with an easy example: suppose that the word "red" takes their place. And suppose that it is possible to explain what is meant by this word at the time of its use simply by pointing to what is given to the senses at the time. As it is tempting to say: "By 'red' I mean this"-where the bare demonstrative manifests the act of pointing to what is given that purportedly serves to explain the meaning that the word "red" bears at the time, and in so doing expresses how what is given is given to the senses at the time. It seems that there is no possibility of what is expressed by a use of "I perceive red" being false, if the meaning that "red" bears at the time of its use is explicable in this manner.

The conception of meaning that this presupposes is notoriously problematic-it is
Wittgenstein's target in his remarks on private ostensive definition, and Wilfrid Sellars famously awarded it the title of a myth. But, for now, there is no need to bring out these difficulties. 9 It is enough to note that no attempt at revealing the truth of what is expressed by "p" which conforms to the Cartesian assumption can succeed. No expression of a thought that is incapable of being false can reveal the truth of a thought that is capable of being false. This is because the expression of one thought can reveal the truth of another only if the truth of the first entails the truth of the second. But if the truth of the first does entail that of the second then, insofar as the second is capable of being false, so is the first. Given the Cartesian assumption, it is impossible to know the truth of thoughts that are capable of being false.

It is at this point that McDowell's disjunctivism enters the fray. It rejects the Cartesian
assumption that what is expressed by the sought-for manner of expression must be selfverifying. Consider a use of "I perceive p". A use of this sentence expresses a thought that contains the thought expressed by "p". And the truth of the thought that it expresses consists in the successful exercise of a fallible capacity that is possessed by the subject that the thought serves to identify. This capacity is fallible because it admits not only of successful but also of unsuccessful exercises: when it is successfully exercised, the thought expressed by a use of "I perceive p" is true; when it is not, this thought is false. The hope is that, insofar as this thought is true, expressing it will not merely express the thought that it contains-the thought expressed by "p"-but in so doing will reveal this thought to be true.
22. But it is hard to see how a use of "I perceive p" can reveal the truth of the thought expressed by "p". Let it be granted that a use of "I perceive p" expresses a thought that contains the thought expressed by "p", and that if the thought that this use expresses is true, then so is the thought that this thought contains. It is hard to see how a use of "I perceive p" can reveal the truth of the thought that it contains. That would require it to reveal the truth of the thought that it expresses. And it can no more do this than a use of "p" can reveal the truth of the thought that it expresses. Recognition of this point underwrites the Cartesian assumption. Disjunctivism proceeds as if the Cartesian assumption is some sort of disposable accretion. But on the contrary: it seems to be compulsory, if there is to be a capacity to express thoughts in a manner that reveals them to be true. The problem is that this assumption makes it impossible to know the truth of thoughts that are capable of falsehood.
23. It might seem that disjunctivism has a response to this difficulty. So far, disjunctivism has been presented as holding that the truth of the thought expressed by a use of "I perceive p" consists in the successful exercise of a fallible capacity. But this is only part of the story.
More fully, it holds that the successful exercise of this capacity constitutes this thought, not merely as true, but as a thought whose truth the subject who successfully exercises the capacity knows. He knows, not merely the truth of the thought expressed by "p", but the truth of the thought expressed by his use of "I perceive p". This is what it means to say that the capacity is, not merely fallible, but self-conscious. But how is this supposed to help? Insofar as he knows the truth of this last thought, he must be capable of expressing it in a manner that reveals it to be true. And his use of "I perceive p" does not do this. So, he needs to use a further sentence that does. "I know that I perceive p" might seem to be the obvious candidate. But the same difficulty will arise again at this level. And insofar as disjunctivism gives the same response to deal with this difficulty, and any subsequent difficulty of the same shape, it will endorse the misbegotten version of the KK thesis that was considered earlier. It will endorse the thesis that, if someone knows something, then he knows something else-in the shape of the truth of the further thought that arises at each subsequent level. And that is not merely absurd; it is hopeless to deal with the present difficulty. If a use of a sentence cannot reveal the truth of a thought contained in the thought that it expresses, on account of its not revealing the truth of the thought that it expresses, then it does not matter how many further thoughts are added-so long as what is added at each subsequent level is equally incapable of being expressed in a manner that reveals it to be true, the revelation of the truth of the first thought will not arrive, and this thought will not count as knowledge.
24. The project of modern epistemology leads inexorably to skepticism. And disjunctivism is powerless to stop it. It might seem that the only way to avoid skepticism is to give up the idea that knowledge is the capacity to express a thought in a manner that reveals it to be true.
But that is not a way to avoid skepticism at all. It changes the subject, and as such leaves whatever case there might be for skepticism in place: it might conceive of itself as "antiskeptical", but its effect will be "precisely what the skeptic most desires". 10 The only possible positions seem to be skepticism, and a position that understands (what it calls) "knowledge", not as the capacity to express a thought in a manner that reveals it to be true, but as a property that "subjects" or "selves" exemplify just in case they satisfy certain conditions, none of which amounts to the possession of this capacity. This is the position of contemporary epistemology. And it is not opposed to skepticism: because it understands (what it calls) "knowledge" in this way, it does not merely reject the KK thesis on the misunderstanding of what this comes to considered earlier; it rejects KK even when properly understood, because it rejects the idea that knowledge is the capacity to express a thought in a manner that reveals its truth. The reason why contemporary epistemology shows so little interest in skepticism is because it has abandoned the very idea of knowledge internal to the epistemological project, and in so doing has abandoned, not merely the project of modern epistemology, but the project of epistemology tout court: the similarity between "contemporary epistemology", and the epistemological project is merely verbal. (It is perhaps worth noting that there are those who have tried to understand McDowell's disjunctivism within the contemporary framework. 11 They present themselves as doing justice to what they think of as the central insight in disjunctivism, and as such, they are happy to employ the language of "fallible capacities"; but the position they defend is merely a version of the contemporary position, which is distinguished from other versions only by its refusal to give a reductive specification of the conditions under which the property they call "knowledge" is exemplified: their position equally rejects KK even when properly understood, and so equally changes the subject, and so is equally devoid of anti-skeptical import. 12 ) * 25. Skepticism seems to be inevitable. Contemporary epistemology passes it by. And so long as the project of modern epistemology remains in place, its inevitability is guaranteed. But the modern project is incoherent: exposing its incoherence will enable the insight in disjunctivism to emerge; and as it does, skepticism will fall to the ground.
26. The form for the articulation of an instance of self-understanding was given as the sentence "I think p". A use of this sentence was said to express, not just the thought expressed by "p", but a further thought that contains this thought. And this further thought was said to be self-verifying. That a use of "I think p" expresses a further thought is something that modern epistemology takes to be obviously true. But it is not obviously true.
It is hard to see how a use of "I think p" can express any thought over and above the thought expressed by "p". That is a central moral of G.E.M. Anscombe's great essay "The First Person". Extracting this moral from Anscombe's essay, and seeing how it bears on the project of modern epistemology, will expose the incoherence of this project.
27. It might seem obvious that a use of "I think p" expresses a thought over and above the thought expressed by "p". Suppose that "p" is "snow is white". That sentence refers to something in a certain way (through the word "snow"), and says something of this thing (through the words "is white"). In the same way, a use of "I think p" refers to someone in a certain way (through the word "I"), and says something of him (through the words "think p").
Or so it might be natural to think. But this assumes that it is possible to explain "I" as a referring term. And this is not possible-as Anscombe's essay shows.
28. Anscombe begins her essay by pointing up two requirements on explaining "I" as a referring term that cannot be satisfied together. The first requirement is this. To explain any word as such a term, it is not enough merely to identify, first, a word, and, second, a thing, and to say of the word that it refers to the thing. Any referring term refers to its referent in a certain way: this way is the sense of the word, and any explanation of a word as a referring term must express its sense-as it might be put: it must show its sense-through referring to what it specifies as the referent of the word in the very way that the word refers to it, paradigmatically, through using this very word. Snow-bound trivialities supply an example: (1) "Snow" refers to snow.
This identifies the referent of a word through using this very word, and in so doing expresses, or shows, the word's sense. The second requirement, by contrast, is taken to be specific to certain words, paradigmatically to "I", "now", and "here". It says that any explanation of such a word must account for its indexical character-in the case of "I", for example, it must account for the fact that its referent always varies with its user.
29. It is easy to see that these requirements cannot be satisfied together. Consider an explanation of the following familiar sort: (2) On everyone's lips, "I" refers to himself.
That satisfies the second requirement. But as it generalizes over everyone, it does not refer to anyone. And as it does not refer to anyone, it cannot express the sense that "I" bears on anyone's lips. It might seem that instances of (2) will avoid this difficulty, as they will refer to someone. But suppose that (2) is instantiated for someone called "NN": (3) On NN's lips, "I" refers to himself.
And it might seem that (4) does specify the referent of "I" on NN's lips, through using the name "NN". But this name cannot express the sense that "I" bears on NN's lips. And this is because no name conforms to the present explanation of "I": insofar as "I" does conform to this explanation, no name can express the sense that "I" bears on anyone's lips. It is possible to bring into this explanation the idea that the sense of "I" on NN's lips is distinct from the sense of any word that does not have the indexical character determined by (2), by marking the reflexive with a star to indicate its specialness, and explaining this special reflexive as follows: for a word on anyone's lips to refer to himself* is for it to refer to himself (ordinary reflexive) in the way that "I" does on his lips. That would yield, in place of (2): (5) On everyone's lips, "I" refers to himself*.
Which in turn would yield, in place of (4): (6) On NN's lips, "I" refers to NN in the way that "I" does on NN's lips.
But this can no more express the sense of "I", on NN's lips, than (4) can. As it might be put: it contains one set of quotation marks too many: 13 whereas the problem with (4) is that it uses the wrong word, the problem with (6) is that it does not use the right one.
30. In this light, Anscombe rejects the second requirement. But she does not reject the first.
This contrasts with contemporary treatments of the first person, which take the second requirement for granted, and in so doing suppress the first. Its suppression is reflected in the widespread tendency to think that the argument of Anscombe's that I have just rehearsed rejects the explanation of "I" as an indexical on the grounds that this explanation is "circular".
Gareth Evans is representative here. He glosses what he takes to be her argument as follows: [Anscombe] argues that either 'John refers to himself' is simply equivalent to 'John refers to John' [cf. (4) above], in which case the observation is wrong; or else it means something which can be elucidated only by reference to the first person pronoun [cf. (6) above] in which case the observation is useless. 14 Although Evans says that the problem with the second of these accounts of "John refers to himself" is that it involves reference to "I", his way of dealing with this problem shows that he thinks that the problem is not that "I" is mentioned rather than used, but that "I" shows up in the account at all, for he goes on to give a different account of the special reflexive, which neither mentions nor uses the first person. This account claims that the truth of (5) consists in the fact that "I" is associated with a reflexive way of referring, which fixes it that the one on whose lips it refers is the one it refers to. That will satisfy the second requirement. And as such, it cannot satisfy the first. Consider what it would yield in place of (6): (7) On NN's lips, "I" refers to NN in a reflexive way that fixes it that the one on whose lips it refers is the one it refers to.
This no more shows the sense of "I", on NN's lips, than (4) or (6) does. Articles and even books have been devoted to defending, or trying to criticize this explanation. 15 But they all miss the point, for they all suppress the first requirement. And in suppressing it, they reveal something about themselves-that they have abandoned the project of articulating an understanding of thought and language from within: of expressing the thought, or in using the element of language, which the understanding is an understanding "of". Any explanation that treats "I" as an indexical abandons this project, for what is expressed by such an explanation instantiated for NN-such as (4), (6) or (7)-is not what NN expresses in using "I". These instances of the explanation are all given, not from within what NN expresses in using the first person, but from outside. Anscombe takes it for granted that a philosophical understanding of thought and language is given from within. And this is why it does not occur to her to abandon the first requirement. (In this light, it is ironic that she is said to reject the explanation of "I" as an indexical because it is "circular". The truth is that she rejects it because it is not circular, as any explanation of an element of language that is given from within, and as such shows the element's sense, must be; (1) is an example. 16 ) 31. These contemporary treatments of the first person proceed, not from within what is expressed in using a sentence with "I" as subject, but from outside. Modern epistemology, by contrast, is a project of articulating self-understanding: it seeks to proceed, not from outside what is expressed in the uses of language that express the thoughts with which it is concerned, but from within. But insofar as modern epistemology appeals to the idea of "I" as an indexical to make out its assumption that "I" is a referring term, it abandons this project at a crucial point. It purports to be articulating self-understanding; but in taking it for granted that a use of "I think p" expresses a thought over and above the thought expressed by "p", it proceeds from outside whatever is expressed in using "I think p". The idea of "I" as an indexical can be taken for granted only by exiting the project of articulating selfunderstanding: from within this project, this idea can have no foothold. 15 The most significant examples are Ian Rumfitt, "Frege's Theory of Predication: An Elaboration and Defense, with Some New Application", The Philosophical Review, vol. 103, no. 4 (1994): 599-637;and James Doyle, No Morality, No Self: Anscombe's Radical Skepticism (Cambridge, Mass.: Harvard University Press, 2018). Rumfitt defends the explanation, whereas Doyle tries to criticize it by defending what he calls Anscombe's "circularity argument". It is no surprise that Doyle cannot sustain this criticism, and is led to conclude that there "does not seem to be any important consideration yet adduced [by Anscombe or Rumfitt] that would decide the issue [between them] either way" (p. 130). There could not be any consideration that would decide the issue between them, because Anscombe does not advance a "circularity argument", and Rumfitt does not engage with her actual argument, because he defends the explanation. 16 I am grateful to Sebastian Rödl for this way of putting things. 32. The project of modern epistemology rests on the assumption that "I" is a referring term: only given this assumption can it make good on its fundamental idea that a use of "I think p" expresses a thought over and above the thought expressed by "p". But it seems that this idea cannot be sustained from within this project, because it is a project of articulating selfunderstanding. And if that is right, then the project is incoherent: it takes for granted that "I" is a referring term-and yet it is unable to make sense of this assumption, by its own lights.
33. If the project is to be saved from incoherence, then it needs to explain "I" as a referring term from within what is expressed in using "I": it needs to give an explanation of "I" that not merely assigns a referent to the word but, in assigning its referent, expresses its sense. And it is far from clear that it can do this. To bring this out, suppose that the attempt is made to do it. It is natural to think that there are two basic ways for words to refer: through an object's being given to the senses in a certain way; or through concepts alone, as set against what is given. Frege thought that the second of these ways could work for numerals-but even he thought that this was an exceptional case. So, that leaves the first way. Insofar as a word refers in this way, its explanation must proceed through the use of a demonstrativespecifically, through something with the significance of: "W refers to this G", where "W" signifies the word, and "G" signifies the kind to which the word's putative referent is said to belong. This kind of explanation matches an explanation of the "'Snow' refers to snow" variety, in that it equally seeks to show the sense of the word; but it operates at a more fundamental level, in that it seeks to show the sense that the word bears in its most basic attachment to reality, in which its sense matches that of a demonstrative. As the use of a word develops, so does its sense, as the word frees itself from its basic attachment to the real, by being used with the same referent outside of situations in which its referent is available to be demonstrated. But the potentiality of explaining a word as a referring term through using a demonstrative belongs to any word whose reference is secured through an object's being given to the senses in a certain way. And for this reason, any explanation of the present form will show at least an aspect of the sense of the word that it explains. But there is no prospect of bringing "I" within an explanation of this form. Any explanation of this form presupposes that something is given to the senses in a certain way. And any way in which an object is given to the senses has a first-person character, in that it is expressible by saying something with "I" in subject-position, paradigmatically something beginning "I perceive …"-such as "I perceive this G". But this means that there is a use of "I" that any explanation of this form presupposes, and as such, cannot explain. The explanation's inability to explain this use comes out in the possibility of asking, with respect to anything that is given to the senses: "Am I this thing"? Or even, if the given object is called "I": "Am I I?" In the first of these questions, and in its first appearance in the second question, the word "I" is figuring in the use that the explanation presupposes-the use as subject in "I perceive this thing", or "I perceive I". And the sheer possibility of asking these questions shows that no explanation of this form is capable of explaining this use: insofar as the questions can be asked, what "I" expresses in the use that the explanation presupposes cannot be the sense of any expression whose reference is explicable through the use of a demonstrative. Insofar, then, as "I" is not equivocal, it is not possible to explain "I" as an expression that refers through an object's being given to the senses in a certain way.
34. Modern epistemology cannot make good on its assumption that "I" is a referring term.
And as it cannot, it cannot descend from the general level at which the project of epistemology was first introduced. But the very idea of the project of modern epistemology is the idea of an epistemological project that makes this descent. So, the project of modern epistemology lapses, and melts away. But the project of epistemology remains-at the most general level. And at this level, the project says that it is possible to know the truth of a thought that is capable of falsehood only insofar as it is possible to express the thought in a manner that reveals it to be true. That might still look as hopeless as ever, insofar as the thought is capable of truth and falsity. So, it might not be clear what consequence the demise of the modern project has for skepticism.
35. But as the assumption that "I" is a referring term fades away, a hitherto suppressed insight in disjunctivism comes to light. When that assumption was in place, it seemed clear that a use of "I perceive p" expresses a thought that goes beyond but contains the thought expressed by "p". Short of this assumption, however, the only thought that is expressed by a use of "I perceive p" is the thought expressed by "p": a use of "I perceive p" does not express a further thought that must itself be revealed to be true, in order to reveal the truth of the thought expressed by "p". And that removes an obstacle to understanding "I perceive p" as fit, not merely to express the thought expressed by "p", but to reveal its truth. 36. A popular word in contemporary epistemology is "factivity". And it is said that the locution "I perceive p" is "factive". But there are two things that it could mean to say this. It could mean that a use of "I perceive p" expresses a thought, over and above the thought expressed by "p", whose truth entails the truth of the thought expressed by "p". A lesson of skepticism is that "factivity", in this sense, is of no significance for the project of epistemology. But that is not the only thing it could mean. It could mean that a use of "I perceive p" expresses nothing over and above the thought expressed by "p", but in a manner that reveals this thought to be true. A use of "I perceive p" could not perform this revelatory function if it expressed a further thought over and above the thought expressed by "p", because then it would be necessary to use a further sentence to reveal the truth of this further thought in turn. But as it does not express any further thought, there is no obstacle to its performing this function. And this is of epistemological significance.
37. It is crucial that the revelation of the truth of the thought expressed by "p" is had in using "I perceive p". It is tempting to think that it is possible to think of someone who is using this last sentence, but without using the sentence oneself. And now it might seem obvious that his assertion of the sentence might be false, because the thought expressed by "p" might be false.
But that is to proceed from outside what he expresses in using the sentence, because it is not itself to use the sentence; and that is not the way of proceeding that characterizes the project of epistemology. Although modern epistemology is wrong to assume that "I" is a referring expression, it is right to assume that epistemological reflection has a first-person character.
To engage in this reflection is not to think of someone who is using a sentence beginning with "I"; it is to use that sentence oneself. And to use the sentence "I perceive p" is not to think of someone in some way expressed by "I"), and to say something of him (through what is expressed by "perceive p"). It is not to think anything of anyone-or at least, of anyone other than the one signified in "p" (if anyone is). To use "I perceive p" is to reveal the truth of the thought it expresses, which is nothing other than the thought expressed by "p".
38. This, then, is the insight in disjunctivism: there is a sentence the use of which does reveal the truth of the thought it expresses, even though this thought is capable of being false. And "I perceive p"-more generally: "I know p"-is an example of such a sentence. That insight, however, cannot come into focus when the assumption that "I" is a referring expression is in place. Modern epistemology takes this assumption for granted-but it cannot make good on it from within its project. Once this is seen, and the assumption abandoned, the insight in disjunctivism will emerge. And once it has emerged, skepticism will fall away. It is possible to express a thought that is capable of being false, but in a manner that reveals its truth.
39. And as skepticism dissolves, it is possible to understand more fully what the KK thesis comes to. The thesis says that if someone knows something, then he knows that he* knows it.
And now that the project of modern epistemology has lapsed, it is clear that the selfknowledge reported by the consequent does not ascribe a property to a "self", but consists merely in the capacity to express the thought that constitutes the content of the knowledge reported by the antecedent in a manner that reveals it to be true. What has emerged from the present reflection is that, insofar as the only thought expressed by a factive locution such as "I perceive p"-or more generally "I know p"-is the thought expressed by "p", this capacity consists in the capacity to use such a locution. At the most general level, knowledge is the capacity to express the thought expressed by "p" in a manner that reveals it to be true; and this capacity just is the capacity to say: "I know p". As skepticism falls away, it becomes clear that, at the most general level, the KK thesis comes to the following: if someone knows the truth of a thought expressed by "p", then he can express this thought by saying "I know p". And in this light, it becomes equally clear that the idea of knowledge is the idea of selfknowledge, not merely because knowledge consists in the capacity to express a thought in a manner that reveals it to be true, but because this is the capacity to use a factive locution that contains the first person. Knowledge is a self-conscious capacity, because it is the capacity to express a thought in this manner, through using such a locution. And it is fallible, insofar as this thought is capable of being false. 17 * 40. The result of the present reflection is that it is possible to know the truth of thoughts that are capable of being false. And that is enough to refute Cartesian skepticism. It is another matter whether a specific thought that is capable of being false is-in fact-true. But knowledge of the truth of such thoughts is not something that philosophy can yield.
Knowledge of the truth of such thoughts can be acquired only through knowledge of the truth of such thoughts: it cannot be acquired solely through knowledge of the truth of thoughts of some other kind, if such there be. And philosophy does not itself consist in knowledge of the truth of such thoughts. But it can remove obstacles that stand in the way of seeing how it is possible to know the truth of such thoughts. And that is what this essay has sought to do.