Honest reflection on classical skeptical arguments leaves one with two conflicting conclusions: (a) that there is something undeniably cogent about skeptical arguments, and (b) that we nevertheless know more or less what we take ourselves to know. The two conclusions are in obvious tension with each other, since the skeptic is claiming precisely to undermine human knowledge: he is asserting that, in the light of his arguments, we do not have the knowledge we take ourselves to have. For example, given the possibility that I am a brain in a vat, I do not know that there is a table in front of me. The skeptic denies ordinary knowledge, claiming to prove that no such knowledge is possible. In order to fend off skepticism, the defender of “folk epistemology” is apt to argue that the skeptic’s arguments are fallacious, having no weight—say, by arguing that we can rule out the brain in a vat hypothesis. But I think that both conclusions are true: skepticism does show something epistemologically disturbing, but it does not show that I don’t know what I take myself to know. The problem I have, then, is to reconcile the two conclusions: to give skepticism its due but prevent it from devastating human knowledge. Skepticism does show something important and surprising, but not that we lack ordinary knowledge. In what follows I propose a way to reconcile these two apparently conflicting conclusions.
Skeptical arguments take many forms, but they all tend in the same direction, viz. that human beings overestimate their epistemic credentials. If we never made any claims to knowledge or justification, exhibiting habitual epistemic modesty, and if this were part of common sense, then the skeptic would have nothing to argue against. We would regard him as saying nothing we don’t already believe. He would be like someone telling us in urgent tones that we don’t know anything about remote unobservable parts of the universe or the state of things before the big bang. But epistemic humility is not our habitual stance—we like to boast about how much we know. We don’t underestimate our knowledge; we overestimate it. This has been the complaint of skeptical thinkers since Socrates and Montaigne, as well as many others. We suffer from misplaced epistemic pride. Thus the skeptic points to the multiple sources of error inherent in perception, memory, and reasoning. We make many perceptual mistakes, often stemming from perceptual illusions; we are frequently wrong about other people’s minds; we hardly ever predict the future correctly. Getting things wrong is a daily occurrence, and yet people persist with their epistemic pride. The skeptic takes this familiar fact and elevates it into a general denial of knowledge (save, perhaps, certain privileged kinds of knowledge, as with the Cogito). Just as we make many small errors (and not so small ones, e.g. the geocentric theory), so we might be making some very big errors—about the external world in general, about the existence of other minds, about the future resembling the past. We might be dreaming or be brains in vats or be surrounded by automata or have existed for only five minutes or be about to encounter a blue emerald. The skeptic wants to know how we can rule these possibilities out; and if we can’t, why we make the cocky epistemic claims that we do. Why are we so confident and complacent, given that error presses in on every side? Isn’t the entire human race like those tiresome know-it-alls who actually know nothing? We seem constitutionally oblivious to the possibility of error, going on as if we are epistemic gods. The skeptic calls people on their epistemic arrogance, attempting to instill epistemic caution. His theme is that we are far more fallible than we are apt to suppose. Human belief constantly overestimates itself.
It is surely clear that there is a good deal to this critique: we are chronically overconfident when it comes to knowledge. When was the last time you met someone who was reluctant to claim to know this, that, or the other? We are braggarts and buffoons when it comes to knowledge—we just can’t help ourselves. But should we conclude from the skeptic’s salutary admonitions that we don’t know the ordinary things that we take ourselves to know? For isn’t it an equally compelling thought that knowledge is, in a great many cases, extremely easy to acquire? You wake up in the morning and look around; in a flash you acquire a large quantity of knowledge—that you are lying in bed, that it is daytime, that it is sunny outside, that you have just been asleep. And it is the same story for the rest of the day: effortless acquisition of countless bits of knowledge flooding your mind—thousands of items of information crowding in. This is not the result of arduous and fraught inquiry but arises from simple sensory perception plus memory—and anyone can do it, with no training or expertise required. Knowledge, we feel, is a piece of cake, a foregone conclusion, an inescapable consequence of being conscious. Children and animals have lots of it too. I exist, therefore I know. To be sure, we get things wrong once in a while, but in the vast majority of cases we get things right, and hence we know many things. So we are inclined stoutly to maintain, and all the skepticism in the world does not deter us. Are we simply being irrational, refusing to face up to the skeptic’s reasonable complaints? What, he will insist, if we are brains in vats? Then what becomes of our supposedly easy knowledge?
As I said, I want to acknowledge elements of truth in both points of view, so I need a way to reconcile skepticism and common sense (which is not to say all of common sense). The way I propose to do that is to exploit the distinction between knowing something and being able to defend the claim that one knows. I want to suggest that the skeptic is right that we cannot defend our habitual claims to know against his arguments, but that this does not show that we don’t know—and we might very well know what we take ourselves to know. In other words, we have indefensible (to the skeptic) knowledge. The general point is that from the fact that I cannot defend a claim to know it does not follow that I do not know—and I might very well know that which I cannot defend. Before coming to the case of knowledge specifically, I will explain the distinction I have in mind by reference to other cases that are less controversial.
Consider any fact about a person, such as having two legs. The obtaining of this fact is obviously not at all the same thing as being able to defend the claim that one has two legs. The skeptic might successfully argue that no one can defend the claim to have two legs against various skeptical possibilities (such as being a brain in a vat), but it does not remotely follow that one does not have two legs. The fact is one thing; defending the claim that there is such a fact is quite another. An animal might have two legs and yet be unable to defend any claims, not even being endowed with language and the concept of justification. This is blindingly obvious: things can have properties and not be able to defend the claim that they have these properties. Moving to cognitive capacities, consider perception and memory: a subject could easily be able to see and remember while not being able to defend the claim that she can see and remember. A skeptic that tried to demonstrate that we never see or remember anything, by arguing that we cannot defend our habitual claims to see and remember, would be barking up the wrong tree; that would be a palpable non sequitur. An inability to defend the claim that one sees and remembers has no consequences whatever for the question of whether one does in fact see or remember. It shows merely that one cannot (justifiably) claim to see or remember not that one cannot do either of these things. Facts about persons and self-ascriptions of such facts are quite distinct. The same point applies to possessing true beliefs and being able to defend the claim that one possesses true beliefs: the former fact could obtain without one having the ability to defend a belief in that fact against skepticism (or anyone else). The capacity to have true beliefs in no way depends upon the capacity to defend the claim that one has the capacity to have true beliefs: these are just different capacities.
But how does it stand with the concept of knowledge? The first point to make is that knowledge and the ability to defend a self-ascription of knowledge are not the same thing: an animal or child could have knowledge and not even be able to understand skeptical challenges, let alone respond to them. Only an extreme and unwarranted appeal to the “linguistic turn” could justify conflating knowledge itself with the defensibility of claimsto knowledge: knowledge itself is not the same as claiming knowledge—which is a speech act not a cognitive state of a person. Claiming to know adds something to merely knowing. Claiming is asserting, and asserting is itself a claim to knowledge—that one knows the truth of what one asserts. If I assert that I know that p, then I claim to know the content of my assertion, viz. that I know that p. Thus self-ascriptions of knowledge are claims to know that one knows—but that goes beyond simply knowing. And it is by no means obvious that knowing entails knowing that one knows. The latter is reflective knowledge that requires having the concept of knowledge, but it is not at all obvious that one can know only if one has the concept of knowledge (again, animals and children provide counterexamples). From the fact that one does not know that one knows it does not follow that one does not know. So if one cannot defend the claim to know against the skeptic’s challenges, that shows only that one does not know that one knows. The assertion that one knows might still be true, even if one cannot defend it, since that would require that one knows that one knows. Connectedly, to make a claim to knowledge is to make a claim on another person—that he or she should believe what one does oneself believe. If I say to you assertively “It’s raining outside,” I expect you to form that belief, and indeed I instruct you to form it. I thus incur a responsibility to you—and I can be criticized if my belief turns out to be false and my claim to know unfounded. But I incur no such responsibility just by having knowledge; that does not imply that I am inviting you to share my belief. I have to assert what I know before I take on the associated responsibility. So being able to defend my claim to know is a very different matter from merely knowing, silently and privately. The skeptic may be right that knowledge claims are indefensible by the person who makes them, but it does not follow that they are false. For, first, such claims involve a claim to know that one knows; and, second, such claims involve incurring interpersonal responsibilities—neither of which follows from knowledge per se. Defending a claim to know is just far more demanding than the mere possession of knowledge—it is going much further out on a limb.
My thesis, then, is that the skeptic is right about knowledge claims but wrong about knowledge per se. He is right that such claims are not ultimately defensible against his arguments, but he is wrong to infer that we lack the corresponding knowledge. We should not, given his arguments, go around forthrightly claiming knowledge, but we might nevertheless have quite a lot of it. Similarly, we should not go around forthrightly asserting that we have two legs, but we might still have two legs—these are perfectly compatible propositions. The skeptic’s arguments apply most persuasively to epistemic speech acts but not to knowledge itself—or any other personal facts. This concedes a lot to the skeptic, admittedly, but it does not concede his most serious allegation—that we simply don’t know what we take ourselves to know. Remember, the skeptic alleges not just that we cannot assert that we know; he also maintains that we positively don’t know. He insists that the proposition that we know ordinary things is false, not merely that we are not entitled to make such knowledge claims. I am agreeing with the latter (but see below) but I resist the implied conclusion that we don’t know ordinary things. I am not then asserting (so far at least) that we do know; I am merely asserting that the skeptic has not shown that we don’t. The non sequitur easily arises from the way the skeptic formulates his challenge: he characteristically asserts, “You claim to know ordinary things but let me show you that this claim is indefensible”; and having done that he infers that we don’t know those things. But his conclusion doesn’t follow from his premises—not, at least, without further argument (see below). Questions about knowledge claims are one thing; questions about knowledge are another. Disputing our knowledge claims is not ipso facto disputing our knowledge.
When I say, “we cannot defend our knowledge claims to the skeptic” the last three words are important. The skeptic presses quite radical doubts that are not pressed in normal life, and these doubts are hard to quell. But I am not saying that no defense of a knowledge claim to anyone is possible. If I claim there is a table in my study and someone challenges me to defend this, I can reasonably reply that I was just in my study and saw a table. This is enough to answer an ordinary doubt, say one prompted by the fact that the furniture is currently being moved out of my house (so the table might have been removed already). Defenses of knowledge claims are defenses to someone, and different types of doubt can be raised by different people; what works to satisfy one kind of doubter may not work for another. So I can defend my knowledge claims to some people in some contexts, even if I cannot defend them to a skeptic. My claims are not indefensible tout court, but only relative to a particular sort of persistent doubter. My point so far has just been that an inability to defend knowledge claims to a skeptic is not enough to show that we lack knowledge, though an inability to defend such claims to anyone in any context might well be incompatible with knowledge.
What would it require to successfully defend knowledge claims to a skeptic? The way to do that would be to show that one cannot be mistaken about the claim in question—that the claim admits of certainty. Classic examples would be truths of elementary arithmetic or self-ascriptions of mental states: if one believes it, then it must be true. In such cases (assuming they exist) we can know that we know, because we know that error is not possible in these cases. Thus I can assert “I know that I am in pain” even to the most resourceful skeptic. But in other cases we cannot argue in this way, because the knowledge is not certain—error is conceivable. What the skeptic is pointing out, in effect, is that most knowledge claims are not like the special cases, which are defensible even by his high standards. I think we should concede this point to the skeptic, however reluctantly, but then stubbornly insist that nothing much follows about knowledge itself. Thus the existence of knowledge is compatible with the indefensibility of knowledge claims, which is what I set out to establish. The skeptic is pointing out that many knowledge claims fall short of the kind of knowledge claim exemplified in the special limited class of certainties, and thus are indefensible by his high standards; and this is something we did not appreciate until he pressed his skeptical case. All right, we can reply, but still you have not shown that we lack knowledge in these cases—we no more lack knowledge in these cases than we do in the special cases. We do not lack knowledge, in a perfectly uniform sense, just because we cannot always defend our claims to knowledge. I can know there is a table in front of me in exactly the same sense that I can know I am in pain—the asymmetry of defensibility of the respective knowledge claims is neither here nor there. Knowledge doesn’t fail to be knowledge just because it isn’t certain knowledge.
I have been harping on the logical gap between the fact of knowledge and the ability to defend claims about that fact, comparing this gap to that between seeing and defending a claim to see (or remembering or having two legs). But it may be replied that knowledge is different from these cases: in the case of knowledge there is no such gap—if we can’t defend our claim to know, then it follows that we don’t know. Let us call this the “exception thesis”—the thesis that knowledge is an exception to the general rule I have cited. According to the exception thesis, it is impossible to know without being able to defend the claim to know, because knowledge is inherently a matter of defending one’s claim to know against objections and doubts—which is not true of seeing or remembering or having two legs. Why would anyone adopt the exception thesis? The answer is obvious: because knowledge requires justification and justification is something that can be offered to another person in order to defend one’s claim to know. To know one must have reasons, but then one should be able to share these reasons with someone else, including a skeptic. I suspect this kind of point is precisely what lies behind the easy transition the skeptic is apt to make from undermining claims to know to undermining knowledge itself, regarding the two as indissoluble. To repeat: since knowledge requires justification, one cannot know unless one can provide the justification to others that one must have in order to know in the first place. Here the fact does imply the ability to defend the self-ascription of the fact: the very thing that justifies the belief (and hence makes knowledge possible) is what can be cited to defend a claim to know. Thus knowledge is an exception to the general rule I cited.
A number of replies can be made to this point. First, the reasons I have for a belief may not be sufficient to repel the skeptic’s radical doubts, but they may be sufficient for knowledge. When we analyze knowledge as true justified belief (plus some) we do not thereby imply that the justification that is a necessary condition for knowledge is of a kind that is sufficient to silence the skeptic; it is enough that the justification is adequate by normal standards. The very meaning of “knowledge” is not such as to require that a person can know only if he or she has conclusive or certain justification; if we did stipulate that, we would make skepticism about knowledge true by definition. The skeptic is arguing that our usual justifications don’t satisfy his radical doubts, but this is not the same as saying that we simply have no justifications at all. And I may have a justification sufficient to answer a furniture mover’s doubts about my table without having anything adequate to answer skeptical doubts. So I might have a justification for my beliefs (and hence have knowledge) without being able to defend my knowledge claims to a skeptic by citing that justification. If so, the fact of knowledge does not imply an ability to offer convincing anti-skeptical defenses—though it does require some kind of justification. In practice, the skeptic is questioning whether I have certain knowledge, but this leaves open the possibility of uncertain knowledge—the kind based on justifications that don’t generate certainty. Such justifications come cheap.
Second, it is controversial to claim that all knowledge requires justification, as a matter of definition. What about beliefs that rest on no other beliefs? If I believe that I am in pain (and know it), I do not have a justification for my belief, since it rests on nothing more basic; but that does not prevent such cases from counting as knowledge. The justification condition for knowledge is usually inserted to rule out cases of true belief for badreasons, not to rule out cases of basic belief based on no reasons. It could therefore be as well formulated by requiring that the belief not be unjustified, i.e. based on reasons that are bad reasons for that belief. We want to rule out accidentally true belief by adding the justification condition, but we don’t want to rule out cases of basic belief that have (and require) no justification. In addition, as soon as we accept externalist accounts of knowledge, which emphasize things like reliability and tracking, then the conditions for knowledge involve factors to which the knower does not have special access. The knower must be a reliable tracker of the truth or be causally connected to the fact in the right way; but it is not required that she knows that she is a reliable tracker or is suitably causally connected to the facts (something that might be presented to the skeptic who questions her knowledge claims). Thus it is possible to know in virtue of such external conditions and not be able to produce a justification that might convince a skeptic. Suppose a person does know that p in virtue of reliability (or the right causal connection to the fact): it does not follow—and won’t in general be true—that she has a reason for belief that will silence a skeptic. The knowledge does not depend on having such convincing reasons, so it is possible to know and not be able to produce such reasons. Animal knowledge makes the point well, since that kind of knowledge cannot depend on being able to marshal reasons that might be offered to a skeptical challenge; it depends rather on the right kind of hook-up to the world. Defending knowledge claims is a sophisticated dialectical skill that is not required for humble first-order knowledge of reality. Perhaps the highly intellectualist approach to knowledge that requires reflective justification for all knowledge might lead us to accept the exception thesis, but that picture is by no means compulsory and rules out by definition (and stipulation) vast areas of human and animal knowledge. If we formulate our conception of knowledge in such a way that we can (we hope) defend knowledge claims to a skeptic’s satisfaction, then the result is apt to be that we end up not knowing many of the things that we usually take ourselves to know, since much knowledge is quite unreflective. It is better to accept that we cannot provide such a defense and then separate knowledge as such from the defensibility of claims to knowledge. The skeptic is thus given his due (his pound of epistemic flesh), but he is not awarded the grand prize—the proposition that we lack knowledge of ordinary matters of fact.
Here is a point that is often ignored: you can have a justification for a belief and then forget that justification; but if you knew to start with, you will preserve that knowledge. You had a justification but now it escapes you: yet still you know, intuitively speaking. So you can’t produce anything now that might halt the skeptic, even though you have knowledge. So long as the justification was once in your possession, you can have knowledge; it isn’t necessary for it to persist. We have all sorts of knowledge for which we once had producible reasons but no longer do (for memory fades)–which means that we cannot now defend our claim to know by producing a justification. And yet we still know. Producing justifications is proper in law courts and examination halls—here memory is paramount—but much of our knowledge is blissfully free of such obligations. We know but we can’t say why.
It seems to me, then, that the exception thesis is false. If so, we can bring knowledge under our general principle and say that the fact of knowledge does not necessarily confer the ability to defend the claim that this fact obtains. The fact of knowledge can obtain without an ability to defend self-ascriptions of that fact. Herein lies my attempt to reconcile the power of skepticism, on the one hand, with resistance to the typical skeptical conclusion that we lack everyday knowledge, on the other. The skeptic is right about our epistemic overconfidence, which is manifest in our normal linguistic practices, but he is wrong to infer that knowledge does not exist or is not possible. For all he has shown, we have all the knowledge we typically take ourselves to have—of the external world, other minds, the past, the future, and so on. Thus the skeptic is agreed to be cogent and disturbing, but he leaves much of what is precious to us intact. This is why skepticism is shocking to us when we first encounter it, by why we don’t tend to react by ceasing to believe anything, though we might well be chastened into greater epistemic modesty. 
I will end with a point about skepticism and the first-person. The skeptic always challenges us to defend ourclaims to knowledge—“How can you defend your claim to know that p?” Or we ask ourselves: “How can I defend my habitual knowledge claims?” But things look different if the skeptic speaks in the third-person: “How can you defend his knowledge claims?” Here we are quite ready to admit that he may not do much of a job of defending himself, but that we could do much better defending him. Even if he can’t properly defend his claims to know, we can defend them, pointing out his general reliability, causal connectedness, caution, and so on. In this case, we may think he does know perfectly well, despite his ineptitude at self-defense—he is no expert epistemologist. We thus separate knowledge as such from self-defense of knowledge claims quite easily in the case of another person. We should adopt the same latitude towards ourselves: we might not be very good at rebutting the skeptic, but that doesn’t mean that we don’t have knowledge—someone else (e.g. God) might see that we satisfy all the conditions for knowledge quite adequately, even though we have no insight into this. That is, the fact of knowledge might obtain quite definitively in our case, though we are not in a position to recognize and assert that fact. The skeptic trades on the latter weakness to argue that the former fact is not the case, but this is a non sequitur. The skeptic is right up to a point, but he overreaches when he denies ordinary knowledge. What he fails to appreciate is the possibility of indefensible knowledge.
 The skeptic might even change our linguistic practices, as well as our epistemic attitudes: we don’t go around loudly proclaiming our knowledge but temper our epistemic assessments so to admit the possibility of error. We go from saying, “I am quite certain that p” to “I am somewhat inclined to believe that p”. The latter could be true even though I know full well that p. It is not knowledge that is at stake but our attitudes towards knowledge.