Agrippan scepticism is a form of radical scepticism named after the ancient philosopher credited with presenting it (Empiricus 252). It is a general sceptical strategy; it may be applied to any argument or claim whatsoever, and is not limited to claims that are epistemological, metaphysical or otherwise in nature. This broad applicability gives Agrippa's Trilemma a great deal of force.

Agrippa's Trilemma presents us with five sceptical procedures: Discrepancy, Relativity, Infinity, Assumption and Circularity (Williams, Problems of Knowledge 61-62). The first two procedures, Discrepancy and Relativity, are general methods of disagreement and do not provide us with radically sceptical arguments. Discrepancy simply says that we may disagree about almost anything; if someone feels like being argumentative, for example, she may disagree for the sake of disagreement. Relativity says that any argument one makes may be accused of being applicable only to the one presenting it; an argument may be convincing only to the arguer personally, or for her culture, political affiliation, and so on. Discrepancy and Relativity present reasons why we may always oppose an argument, but they do not give us radical scepticism; the fact that we may disagree does not force us to admit that no argument is any better or worse than any other. Disagreement leaves room for correctness and incorrectness, varying levels of support, and proofs, even though some of us might not accept them. But once we recognize this, we acknowledge that we may be asked of any claim put forward what our reasons are for holding it. If we are to avoid the criticism that an argument we make may be disagreed with or made relative to our personal position, we must acknowledge that an explanation of why we hold our claim to be true may be asked of us.

It is here where the Trilemma of Infinity, Assumption and Circularity become apparent. We have made a claim, and the sceptic rightly asks us why we hold it; how do we know it to be true? If the reason we hold our claim to be true is based on an assumption, our claim is shown to be empty because of its dogmatism. In order to avoid the pitfall of dogmatism, we will cite some sort of evidence to show how we know our claim to be true. The sceptic may now ask why we hold this to be true. If we cite another assumption, then we have not furthered our position at all and our claim remains dogmatic. This is the Assumption portion of the Trilemma.

If we make a new claim in support of the previous claim, this too will need to be supported. If we continue making new claims, we embark on a vicious regress of justification. This is the Infinity portion of the Trilemma. Finally, if we cite an earlier claim in support of the one in question, we fall into the trap of circular argumentation. This is the Circularity portion of the Trilemma.

It follows that for whatever claim we make none of our reasons for holding it will be good enough; if we are not dogmatically assuming support we are either citing earlier claims as support or providing novel reasons. If we cite earlier claims as support for our argument, then it is circular. If we cite novel reasons as support, we are led down an infinite regress of justification. Every claim may be challenged with Disagreement or Relativity, and every challenge may be responded to in only three ways, none of which improves our position. The result is that not only can every claim be questioned, but no claim is ever any more or less justified than another; this is why Agrippa's Trilemma is a radical form of scepticism. If we are to avoid being dogmatic with assumptions or circular with our arguments, then any claim we make must cite novel support. As such, any claim we make is doomed to an infinite regress of justification; every claim is equally (un)justified.

A striking feature of Agrippa's Trilemma is that it applies to any argument or claim. Aristotle's response to scepticism was to work from self-evident premises in order to deduce true claims. But the radical sceptic may then ask: why do you hold that self-evident premises will bring us to true conclusions? If we are assuming it, our Aristotelian claim fails. Our options are novel support or circularity, and we embark on Agrippa's Trilemma again. Foundationalist arguments fall prey to Agrippa's Trilemma as well. When the Foundationalist presents the claim that all knowledge is justified by the senses, we may ask if that is an assumption. A familiar pattern occurs.


Humean and Cartesian scepticism ask whether or not we can know anything about the "external world." We seem to see tables and people and trees in front of us, but do we know that they exist, independently of what we believe? This sort of radical scepticism about the external world is what this paper will be focused on.

Descartes' path to scepticism is likely a familiar one, but it is worth surveying. Descartes begins by doubting anything that could be doubted, in order to assess all his supposed knowledge (Descartes 89). His goal is to find out which, if any, of his beliefs could be considered knowledge; his goal is a broad assessment of all of his beliefs as knowledge (Williams, Unnatural Doubts 23). In order to accomplish this, Descartes chooses a belief that he and most anybody would consider a paradigm of knowledge: that he knows that he is sitting at his desk in front of a fire (Descartes 90). The conditions for knowing in this case are as optimal as one could hope for; Descartes has a clear view of everything he observes, the lighting is good, he is in close physical proximity to what he is observing, and so on. It is in this way that we are to take it that Descartes' conclusions apply to all of knowledge (Stroud, The Significance of Philosophical Scepticism 9).

In this sense, Descartes' belief that he is sitting in front of the fire with a piece of paper is a best case belief; you could not ask for better circumstances or more justification. It is for this reason that Descartes' evaluation of his knowledge is considered an assessment of all knowledge of the external world in general; if Descartes fails to have knowledge in this case, then clearly knowledge in more difficult circumstances of less obvious things is not possible (Stroud, The Significance of Philosophical Scepticism 9-10). The paradigmatic nature of the knowledge that Descartes doubts gives it its general application to all our knowledge of the external world, because it is the best position one could be in for claiming to have knowledge. As such, if Descartes does not have knowledge in this situation, then it is safe to say that no one has any knowledge of the external world in any situation.

Descartes' initial conclusion is, of course, that he does not know that he is sitting in front of a fire with a piece of paper. His reason for his conclusion is that he does not know if he is dreaming. Descartes has had dreams that he believed at the time were real experiences. Obviously, any belief Descartes formulated while dreaming would not be knowledge; dreaming that one is sitting in front of a fire with a piece of paper does not give one knowledge that one is sitting in front of a fire with a piece of paper, even if one is actually in such a situation. The difficulty is that Descartes could very well be dreaming all of his observations yet not know it; if he is dreaming, then none of his current observations will count as evidence that he is not dreaming, because any observations could be the same whether or not he was dreaming and he would not know the difference. The possibility that Descartes is dreaming is one he cannot rule out, and it is a possibility that would deny Descartes knowledge; if Descartes cannot know whether or not he is dreaming, he cannot rule out the possibility that his observations do not lead to knowledge.

Other formulations of the argument have been presented, such as brain-in-a-vat scenarios, Cartesian demons, and so on. Each has the same general form and conclusion: we cannot know anything about the world because there are certain possibilities that cannot be ruled out by observation, and the presence of those possibilities denies us knowledge of the external world. The result is radical scepticism; we can never know anything about the world, because there are certain conditions on knowledge that cannot be met.


Worth noting in contrast with Cartesian and Humean style scepticism is the argument from error (Williams, Unnatural Doubts 135-136). Descartes himself considers a form of it in the Meditations (Stroud, The Significance of Philosophical Scepticism 8). Descartes notes that, from afar, a tower may look round when it is in fact square. In such a case we do not know whether or not the tower is round. This would seem on the face of it to be a sort of sceptical argument.

Arguments from error are less persuasive than Cartesian or Humean style radical scepticism. The reason is that the argument from error does not rule out the possibility of knowledge; it merely points out that in some situations standards for knowledge may be difficult to meet. We may not know whether the tower is round or square from a distance, but when we are up close, we can clearly observe the shape of the tower. The argument from error does not rule out this possibility. If the argument from error is considered a form of scepticism, it is definitely not radical scepticism; it does not make it impossible for us to have any knowledge of the external world whatsoever. Cartesian and Humean style scepticism, by contrast, rule out the possibility of ever having knowledge of the external world, and it is for this reason that Cartesian and Humean scepticism are of more interest than the argument from error. At worst, the argument from error encourages us to set high standards on our knowledge, but not impossible standards.


Radical scepticism seems to clash badly with our normal beliefs about knowledge. We believe and act in everyday situations as if we know all sorts of things, and we seem to have no problems at all. Attributions of knowledge are consistently made, and we accept or reject knowledge claims for many reasons.

Scepticism says that all those claims and attributions, even those that are seemingly irrefutable in everyday circumstances (that I am sitting at a computer typing, for example,) are taken with just as little reason as the claim that I am actually a brain in a vat. The question then is: why do we accept the sceptical conclusion? Its being clearly incompatible with our every day knowledge claims seems like a prima facie reason for rejecting it as a theory of knowledge, or at least for considering it dubious.

The difficulty is that scepticism seems to use the very everyday concepts of knowledge that we consider it to be in conflict with (Stroud, The Significance of Philosophical Scepticism 70). That we do not know anything as a result of observations we have during dreams is perfectly reasonable, as is the contention that often when we dream we do not know we are dreaming. To take one more step, and say that we might be dreaming right now without knowing it, and as such we are not justified in holding our knowledge claims about the external world as true, seems natural too. In this sense radical scepticism has an intuitive edge; radical scepticism's conclusions seem to be drawn from aspects of knowledge that we all accept. As such, its contrast with our everyday beliefs about our knowledge seems not to be a reason to hold sceptical arguments to be false or dubious, but rather as a reason to hold that our everyday beliefs simply are not what we thought them to be. Radical scepticism is drawn from everyday aspects of knowledge, and as such its conclusions are based on intuitive premises. As such, radical scepticism seems to display a dark secret about our everyday knowledge; even by its own standards, our everyday assertions of knowledge are never justified. What we thought was knowledge we can in fact never know at all.

What are we to think of this clash between our everyday assertions of knowledge and radical scepticism's conclusions? We continue to believe all the things we believed before, even after learning of sceptical arguments. We also act on those beliefs; to doubt that the car that I see ahead of me exists while I am driving down the highway at 100 kilometres an hour would clearly result in a disastrous choice. It also seems perfectly reasonable to act in such a way, at least by our everyday standards, even though sceptical arguments tell us that no piece of knowledge is any better-held than any other. To the sceptic, whether or not you slam on the brakes to avoid the car ahead of you, you are making an equally unjustified choice based on your knowledge. It seems that, although sceptical arguments are derived from intuitive premises, they nonetheless seem to be at odds with everyday knowledge claims and actions.

Hume's answer for this was essentially a psychological thesis; we believe and act as if we know certain things because it is in our very nature as humans to do such things. We cannot avoid it, because it is how we are made (Hume 110). As Quine said, "The Humean predicament is the human predicament." (Quine, Epistemology Naturalized 72) We understand the sceptic's conclusions, but they seem to fade away the moment we are left to make a decision based on our supposed knowledge in normal circumstances. The sceptic's conclusions are profound, remarkable and persuasive, but they cannot hold up to our firm convictions about the way our knowledge works, even if those convictions are ultimately unfounded.

This response seems reasonable, but it is also unsatisfying. There is of course the somewhat ironic point that Descartes' proposal that it is human nature to act as if we know certain things is unfounded under sceptical lights; if we take the sceptic seriously, we do not even know that claim. But further difficulties seem more worrying. If the sceptic is correct, even our most cherished and prized discoveries are lowered to the level of the merest speculation. Any claim about the world cannot be considered knowledge more than any other claim; those who believe that the world is flat have just as much, or little, reason for claiming knowledge as those who claim it is another shape. Knowledge seems to have been rendered an unattainable goal, and the result is that none of our claims are what we thought they were.

The sceptical problem remains. To appeal to a psychological fact about us as humans to explain why we act as if we have knowledge in light of sceptical claims seems almost to amount to an appeal to ignorance; we do not know anything, but we can still act as if we do because there is nothing else we can do. This is a bleak and unsatisfying conclusion, and "knowledge" being relegated to a byproduct of human nature seems to run counter to what we take to be important about it.

For this reason, I believe that the sharp contrast between sceptical conclusions and everyday attributions of knowledge does point to a defect in sceptical reasoning. In order to show this, we must address the issue of the intuitive nature of sceptical arguments, in order to show that they are not in fact as intuitive as some might take them to be. If this is so, then the clash between everyday knowledge and scepticism should be damning for scepticism.



Two papers by G.E Moore attempt to challenge the radical sceptic: "Defense of Common Sense," and "Proof of an external World." In "Defense," Moore lists a series of things he knows are true: that he has never been far from the Earth's surface, that there are other people, and so on (Moore, A Defense of Common Sense 33-35). Moore's proof of the external world is short and simple, but its implications and impact are not quite so simple (Stroud, The Significance of Philosophical Scepticism 84). His proof proceeds as follows: raising one hand, Moore says, "Here is one hand," and raising the other, he says, "and here is another" (Moore, Proof of an External World 73). The two hands are external objects, that is, things that are to be met in space, and it follows from the fact that there are two external objects that externals objects exist (Moore, Proof of an External World 66).

Moore's arguments are difficult to come to terms with in light of radical scepticism. Someone persuaded by radical sceptical arguments is likely to say that Moore has missed the point; these so called facts, which Moore so confidently reports, are the very things that the sceptic contests, and the very thing that the sceptical argument is said to show we cannot be justified in holding true. To lay them out as disproof of the sceptical argument is like responding to the sceptic's conclusion that we do not know of the existence of external objects with a stubborn, "Yes we do." This will not do for most philosophical arguments.

But dismissal of Moore's arguments ought not to be quite so quick. For one thing, as Moore points out, his proof contains all the features of a good proof. First, its premises are different from its conclusion; that is the conclusion that "External objects exist," could be true even if "Here is one hand," and "Here is another," are not. Second, Moore takes it that the premises are true; the conclusion is drawn validly from true premises.

An obvious difficulty is that, if the sceptic is correct, then Moore cannot claim that he knows his premises to be true; sure, they might be true, but if we do not know whether or not they are, how can we say the conclusion is true? If the sceptic is correct, then Moore must be wrong.

But it is still not quite so simple as that. Moore takes his argument to be doing the very opposite of what the sceptic claims; Moore is giving us examples of true things that he knows, and proofs that he knows those things, and thus concludes that the sceptic must be wrong. The coin has two sides, so to speak; Moore's contention is that he knows these things, and if he knows these things the sceptic must be incorrect. Both the sceptic and Moore cannot be correct at the same time, and so the sceptic must concede defeat. Which side are we to take?

Here we see an interesting facet of radical scepticism brought to light by Moore's argument; there seems to be a severe clash between Moore's argument and the sceptical argument. The two butt heads so to speak, yet neither seems to win an undisputable victory. This is made apparent by the intuitive nature of both arguments. Sceptical arguments seem to play on our intuitions about knowledge; that is, they are convincing because they seem to use our common conceptions of knowledge to show us that we in fact have no knowledge. Descartes might say: if we are dreaming we do not know anything about the world from our experiences, and when we are dreaming we sometimes do not know we are dreaming. We could be dreaming right now and not be able to tell that we were dreaming from our experiences alone, and if we cannot rule this case out, we cannot say that we know that external objects exist from our experiences of them. The argument is convincing to most everyone, but the conclusion seems absurd.

Moore on the other hand uses our intuitions as well; we do know all kinds of things! We know where we live, that our houses exist, that we have friends and family who are human beings. It is absurd to say that we do not know these things, because every part of our lives involves us knowing these things and drawing conclusions from such knowledge. If we did not know such things, we would be paralyzed by lack of knowledge. Should I go to the store and pick up some milk? I would, but I do not know whether the store exists. Moore's conclusion that we know about the existence of many external objects is perfectly reasonable to any sane person. Yet Moore's argument seems unsatisfying somehow, even despite the fact that it is the sort of thing we could do at any time in everyday life.

There is a persistent clash here between the sceptical argument and common, everyday use of "knowledge." Both seem to be convincing in their own ways. The sceptical argument is persuasive and has a drastic and surprising conclusion, yet its persuasiveness seems to disappear in everyday life, where its drastic conclusion falls away; we would not, maybe even could not, actually doubt the existence of the entire world and remain reasonable people. Moore's arguments, on the other hand, hit on the everyday examples of knowledge that we take to be obvious. We act as if we know all the sorts of things Moore takes us to know, and if someone were to tell us in everyday circumstances that we did not know such things, we would probably take them to be crazy, or joking. I could not seriously doubt that I am sitting in a chair now; it is the most obvious piece of knowledge I could give, and if I doubted it I might as well doubt that I can walk on the ground or talk to people or eat food or do anything at all that we do in our everyday lives. Yet Moore's argument still seems philosophically hollow; scepticism's argument is quite convincing, and if it is correct, then Moore cannot even use his premises, let alone make his conclusion.

We are left in a unique situation: we have a convincing sceptical argument with an absurd conclusion, and a somewhat ham-fisted argument, with a perfectly reasonable conclusion. Which are we to accept? If you have affinities towards one or the other, you may find either convincing. But I take it to be far from obvious at this point that scepticism prevails unquestionably due to Moore's assertions, or that Moore has shown scepticism to be impossible. I do believe that both cannot coexist; one or the other is correct in its conclusion. There must be something wrong with one, or even both, of the arguments or conclusions, but it will be a difficult thing to show.


Austin's approach to the problem of scepticism is to examine closely the way we use words like "knowledge." Once we understand how such epistemic terms are used in everyday situations, we will come to see that the sceptic misuses language, and as such cannot come to the conclusion that our everyday epistemic concepts lead us into radical scepticism; the sceptic cannot mean what he says he means, because his terms misrepresent the way language about knowledge is used (Williams, Unnatural Doubts 140).

To begin his study of the way we use epistemic terms Austin, in a manner similar to Descartes, makes use of a paradigm case of a knowledge claim. His goal is to show when a knowledge claim is reasonably advanced, how it can be challenged and when it ought to be withdrawn. Austin contends that his proper analysis of our everyday use of language bears no sceptical conclusion.

The contention is that the sceptic violates our ordinary use of language, but the sceptical argument carries its intuitive weight precisely because it seems to gel with our common concepts of knowledge; the sceptic does seem to use our everyday concepts of knowledge. If Austin is to succeed he must not only show that his analysis is the most intuitive, and more likely correct, one, but must also show the sceptic to be incorrect. We must compare the Austinian and sceptical approaches to knowledge-claim analysis.

The sceptic asks how we know what we claim to know; for example, the sceptic might ask, "How do you know you are now sitting at your desk by a fire?" as a challenge to our claim to be sitting at a desk in front of a fire. The challenge is one that is meant to threaten our knowledge in general; if we cannot justify our knowledge claim we must withdraw it, and all other knowledge claims if the claim at hand is paradigmatic of all knowledge.

In this way, we must prove that we know what we claimed to know. The difficulty of course is that we seem to be incapable of doing this. I claim that I am sitting by a fire, the sceptic challenges my claim by asking how I know such a thing. In response to the challenge we offer grounds; I know I am sitting in front of the fire because I can see that I am. At this point the sceptic produces a general objection, applicable to any of our grounds; the sceptic will assert that I do no know that I am sitting in front of the fire because I could be dreaming all of my experiences and not know it. The sceptic has shown us not to have knowledge.

Austin's analysis appears to proceed in a similar manner, but does not produce scepticism. In Austin's example, someone claims that the bird they see in the field is a goldfinch. The sceptic challenges by asking how our would-be knower knows such a claim. At this point grounds for the claim are produced; the claimer says, "I know it is a goldfinch by its red head." Finally, the sceptic responds with a doubt; the sceptic might say, "But woodpeckers also have red heads." (Austin 83)

Austin's sceptic seems to have thwarted our would-be knower. Yet knowledge in general has not been threatened; the debate between the two is still open, and the possibility of knowledge being attained is still live. Further, the particulars of this case do not lead to a general scepticism; explaining that woodpeckers also have red heads does not leave out the possibility of further examination of the bird, nor does it imply that other knowledge claims will suffer from a similar deficiency. The important feature is that Austin's example of a knowledge claim proceeds according to the way actual everyday knowledge claims do, and in everyday knowledge claims general doubts, such as brain-in-vat and dream scenarios, are never brought up. Reasonable, everyday doubts do not act like sceptical doubts.

But, it appears that the sceptic's approach is the same as Austin's; in both analyses a claim is made, the claim is challenged, grounds are given in response, and a doubt is given in response to the grounds. What exactly makes Austin's everyday case superior to the sceptic's case? What is right with Austin's example, which appears formally similar to the sceptic's yet does not lead to scepticism, and wrong with the sceptic's case? What makes Austin's case more "everyday" than the sceptic's?

According to Austin the sceptic aims to show that the evidence we normally advance as grounds for a claim is not enough, because it does not exclude such possibilities as dreaming or brain-in-a-vat scenarios that show we do not know what we claim to know. But, Austin contends that there are conditions on knowledge claims that govern when evidence is adequate; these are conditions that we use in everyday knowledge claims, but that sceptical challenges violate (Austin 83-85).

The two conditions that Austin cites are what Williams calls the definite lack condition and the reasonable sufficiency condition (Williams, Unnatural Doubts 141). The definite lack condition says that we cannot reasonably challenge the evidence for a knowledge claim unless it can be specified in a definite way how the evidence is deficient. The reasonable sufficiency condition says that evidence is adequate if for present intents and purposes there is no room for alternatives within reason. If a knowledge claim or challenge violates either of these conditions it is unreasonable, and not in line with our everyday use of knowledge terms. We are now in a position to see how Austin and the sceptic's cases, which appeared formally similar, are actually different. Presumably Austin's non-sceptical case will stay within the bounds of the definite lack and reasonable sufficiency conditions, and the sceptic's case will violate them in some way.

Austin's conditions are meant to harm the sceptic's position. But the definite lack condition seems to be unproblematic for the sceptical argument; citing the possibility that we may be dreaming is a definite challenge, and one that we can all understand. We cannot seem to meet challenges like dream possibilities or brain-in-a-vat scenarios, but that does not make them vague or fuzzy or somehow lacking. In fact, that sceptical dream possibilities are such challenging philosophical questions seems to show that they are indeed definite challenges. Or at least definite enough that we can argue over them.

On the face of it sceptical challenges seem to meet the definite lack condition. We are left with the reasonable sufficiency condition if we are to refute the sceptic with Austin's tools. We must show that given some specific intents and purposes, there is no reasonable room for sceptical alternatives when making a claim that the sceptic challenges.

But now a problem is immediately apparent: for philosophical purposes, it seems the sceptic's challenges are reasonable. The sceptic's intents and purposes, namely to question all of our knowledge at once, make his broad doubts seem very relevant in philosophical discourse. As Michael Williams explains, this is because Austin ignores the sceptic's philosophical project. In Austin's goldfinch example there are many specific ways the knowledge claim may go wrong; for example, it could be a woodpecker, the knowledge claimer could be an inexperienced bird watcher, the area they are in might be devoid of goldfinches, and so on. Many practical questions may have to be met for the knowledge claim to succeed, but broader doubts, such as dream possibilities, do not seem to arise in Austin's case.

Williams contends that, for this reason, Austin's case does not even reflect on the sceptical case presented by Descartes. Austin's example is not a best-possible case for a knowledge claim, and as such it does not threaten the sceptic's case. The specific doubts raised in Austin's case are such that his example cannot be a best-possible case of knowledge claim, and as such Austin's case cannot be paradigmatic of all knowledge, which it would need to be if it were to challenge the sceptic's case. If we can doubt whether the bird is a goldfinch or a woodpecker, then we are clearly not in the best possible situation for knowledge, and our knowledge claim does not reflect badly on the rest of our knowledge. If Austin's case were the same as the sceptic's, no simple doubts could be raised; we would have the best possible evidence for the bird being a goldfinch. It is at this point that sceptical doubts may arise and threaten all of our knowledge. In effect, Austin's example changes the subject on the sceptic, then claims that his example reflects badly on the sceptical case, even though they are not actually comparable.

The result is that Austin's everyday examples of a knowledge claim being challenged are particular enough in their doubts that they cannot approach the sceptical argument; Austin's case misses the point so to speak, and the sceptic and Austin talk past each other. It is easy to see how this is possible; both Austin's case and the sceptic's seem to be formally identical. It is only once we note that Austin's particular doubts deny the possibility of his case being generalizable to all of our knowledge that we notice the difference between the two. Austin is right to say the goldfinch claim ought to be withdrawn in light of the doubter's claim that woodpeckers have red heads too, but if the knowledge claim fails in this particular way there is no way for the doubt to apply to all of knowledge; we are left at best with an argument from error, which is not radically sceptical. Austin's example cannot generalize to all of our knowledge, and so it is not the same as the sceptic's claim.

The sceptic on the other hand uses cases of knowledge claims in which there are no sources of doubt particular to the situation; any doubt we have about Descartes' knowledge claim that he is sitting in front of a fire will apply to any knowledge claim at all, because no particular doubts are left over to have. In fact, because the sceptic needs a best case scenario of knowledge claiming to get his case of the ground, he cannot have any particular doubts in his case at all; if a reply, such as Austin's, does make use of particular doubts, it cannot reflect badly, or at all, on the sceptic's position.

Austin's attention to detail initially seems to be the strength of his position; Austin analyses our everyday uses of knowledge claims with great insight and understanding, and he illustrates our everyday use of knowledge claims with accuracy. But, the sceptic never uses, and cannot use, detail in the first place. The sceptic's intents and purposes involve not particular doubts about knowledge claims, but rather broad doubts that can apply to all of knowledge. In this way, the sceptic's claim seems to meet Austin's reasonable sufficiency condition; in philosophical discourse, when the sceptic is trying to assess all of our knowledge at once, we see that broad, normally unreasonable, doubts about dreams and brains in vats become relevant. The sceptic's claim meets both of Austin's conditions, and Austin's anti-sceptical argument seems to fail.

We may be left with some misgivings about all of this. Austin's knowledge-claim analysis is after all more true to our everyday use of "knowledge"; in everyday situations we do not make people overcome general doubts about dreaming. If I asked someone, when he claimed to have seen a goldfinch, whether he knew it was a woodpecker or not and if he also knew whether he was dreaming, I would seem very unreasonable; the knowledge claim would not be threatened by the possibility of dreaming. Yet the sceptic's claim seems perfectly reasonable under philosophical considerations.

Why do we not choose Austin's conditions and cases over the sceptic's? Both Austin and the sceptic make intuitive arguments, but one applies to our everyday situation, the other only to philosophical arguments. The sceptic's conclusion is incredible, but he needs incredible circumstances, in which no particular doubts are available, in order to reach it. In this way, the sceptical argument might seem less intuitive than Austin's. But we are far from being able to show anything of the sort yet. At best we find both arguments intuitive in some way or another, with clashing conclusions from both; it seems reasonable to say Descartes knows he is sitting in front of a fire, but also reasonable to say that we do not know whether we are a brain in a vat or not. If we are to show that Austin's everyday examples ought to take precedence over philosophical intents and purposes, we must be able to show that there is some sort of unintuitive deficiency in the philosophical position. But for now we are still left with Austin's conditions, which the sceptic meets, and Austin's intuitive knowledge claim case that is nonetheless irrelevant to the sceptic's case, which is intuitive in its own right.


A problem of scepticism is that when we call into question all of our knowledge, we are left with no conceptual tools with which we may make any epistemic progress; any supposed fact or theory we may cite in order to show that we have knowledge of the external world is just as unknowable as the knowledge we are trying to prove. Quine's approach to this problem is to make epistemology the subject of science, specifically psychology. (Quine, Epistemology Naturalized 82) If we study knowledge just as we study other human phenomena we will see that there is a great deal of progress to be made in understanding how we know. Epistemology will no longer be a study of knowledge that is external to all knowledge, so to speak, and becomes a study of knowledge that is internal to science; we will move our epistemic position from one in which we have no resources at all, to one in which we have all of science and psychology to support and develop our ideas about knowledge. This is the Naturalized Epistemology referred to in the title of Quine's essay on the subject. Epistemology, once naturalized, moves away from philosophy and into the realm of psychology; where we once argued about the conditions for knowledge, we now simply study humans and human phenomena in the same way a psychologist does.

Under Quine's naturalized epistemology the sceptical problem takes on a new light. The only way we can experience objects is through our senses. For example, light bounces off of objects and hits our eyes, and we see an image that presumably looks like the external object. As Quine says, the problem is that, "... we know external things only mediately through our senses" (Quine, Word and Object 1). A familiar sceptical thesis now presents itself: we do not know whether our senses are being systematically deceived so that everything we experience is actually an illusion; so we do not know anything about the objects we experience. We can never know if our inferences from our sense experiences are correct. Quine tends to describe the problem in the language of underdetermination: our rich theories about the external world are underdetermined by the meager evidence we get from our senses. That is, we cannot determine what theory is correct given our limited sense input, so we may always be wrong. But the general sceptical conclusion that we do not know anything that we infer from our senses remains common between the classic sceptical argument and Quine's theory underdetermination argument.

But, contends Quine, what we have missed about this sceptical argument is that it is a scientific argument; that is, it is an argument made from within science. It is a scientific, empirical fact that we see images of objects by the stimulation of our sense receptors by light, for example, bouncing off of objects, and that we formulate theories about the existence of objects from this meager data (Quine, The Natural Theory of Knowledge 2) (Quine, Epistemology Naturalized 83-84). As such, this phenomenon may be studied just like any other scientific one. In other words, we may study our knowledge of the world just as we study our culture, psychology, and so on: in a scientific, empirical manner. In this way, we can make just as much progress in epistemology as we have made in other scientific areas, by studying it in the field, so to speak. The sceptical problem is presented by science, through theories about light bouncing off of objects and making us experience images, and it so it can be answered by science as well, just like any other theory.

Stroud advocates that after Quine's naturalization, the epistemological problem becomes a scientific study of theory in the same vein as psychology, sociology, and so on, in that it is a perhaps imperfect, but developing field with a general direction of enquiry in which we have many resources to answer admittedly tough questions (Stroud, The Significance of Philosophical Scepticism 217). Every theory in science is underdetermined to some degree; in chemistry we must infer the existence of subatomic particles, in psychology we must infer that people make fairly reasonable decisions. But science seems to be no worse off because of this. By understanding epistemology as a part of science, we grant it the same benefit. Presumably, if we are to make an inference from experience of an image to knowledge of an object, such a move will be less troublesome once epistemology is part of science.

The difference between theories of chemistry and psychology and epistemology is that epistemology's theories are not explicitly stated as traditional scientific theories are; perhaps they have been assumed for generations, since before culture and science were established. This, of course, is another empirical fact that may be studied scientifically. The reason we hold on to it is that it has been a successful theory. We continue to use the theory of external objects for its pragmatic benefits, however vague that idea may be.

A difficulty in Quine's anti-sceptical naturalized epistemology seems to arise, however. Epistemology is a scientific field like any other, meaning we can study our knowledge empirically. This has the benefit of bringing epistemology down to earth so to speak; we study chemicals and people and societies and so on, so if we study knowledge in the same way, we will have all the resources of psychology at our fingertips to solve what has been traditionally an ethereal problem. But the problem of underdetermination seems to apply to all scientific theories, so that we never know from our experience whether any theory is correct (Stroud, The Significance of Philosophical Scepticism 218). In other words, the traditional sceptical problem seems to apply to all of science: if I do not know if I am dreaming, I do not know that what I experience is real, and it follows that I do not know whether my scientific theory of epistemology, or chemistry, psychology or sociology, is correct either. Broad underdetermination of theory by our available sense evidence seems to undermine all of science, meaning epistemology's sanctuary in science is no safer from scepticism than the traditional epistemological position. We are left in an Agrippan situation: every scientific theory is underdetermined, yet all we have is scientific evidence to back up our choice of theory. When faced with the problem of underdetermination we are left with the unattractive option of supporting our choice of scientific theory with scientific evidence, making science, and the epistemology it encompasses, circular. The other options are to delve into a regress of novel support, however unlikely that is, or unfounded assumption.

In this way, it seems that Quine's naturalized epistemology does not answer the traditional sceptic's challenge. We can perform an empirical, scientific study of knowledge and call it epistemology, but it will be no safer from sceptical arguments than it was before. We may always ask how we know that the empirical study is correct. Once all of our knowledge is in question, an appeal to any of it to answer the sceptic becomes circular, science included.

Quine finds this unsatisfying. He contends that our fear of circularity in science, and hence epistemology, is "logical timidity" (Quine, The Roots of Reference 2). Sceptical arguments come from science, Quine contends. If we argue that we do not know about external things because we experience them mediately through images, we must refer to a theory about light reflection and eyes and images and so on. If we use illusion as our basis for sceptical argument, we must have a scientific theory of how illusions occur in order for it to succeed, and if we use dreams in our sceptical argument, we must have a theory about dreams. If we recognize that all sceptical arguments come from science, we ought to see that we may use science to answer those sceptical issues. The challenge to our knowledge by scepticism comes from scientific knowledge we have; our doubts of knowledge come from knowledge, so we ought to be able to use that knowledge in responding to the sceptic.

Once again, Quine has put us in a position where we have a selection of scientific tools to use in responding to the sceptic. Epistemology is part of science, so we can use all the resources of science in explaining how we have knowledge. Traditional sceptical problems seem to threaten that position, but traditional sceptical problems too are part of science, and so may be questioned by scientific evidence. Sceptical doubts are scientific doubts, so epistemology is in a strong position to answer those doubts with science.

Yet we run into another difficulty. If our understanding of illusions, dreams and images of objects depends on theories in science, then it seems that using those theories to disprove our knowledge of science is a classic reductio ad absurdum: science presents us with theories, and if those theories cast into doubt all of our scientific knowledge, we have shown science to be absurd by its own lights. Quine himself admits that this is a valid argument available to the sceptic (Quine, The Nature of Natural Knowledge 68). We are left in a strange position. Quine wants to tell us that the sceptical argument is an overreaction, and that plenty of scientific evidence is available in answering it. But the sceptical argument seems to call into question the very scientific evidence we are to use in answering it. At best we are in the unenviable position of having a sceptical argument that is not supported by scientific evidence because all of science has been shown to be underdetermined. Either science is available to us in answering the sceptic, in which case the sceptic can use scientific evidence to perform a reductio ad absurdum, or scientific evidence is not available to the sceptic or the anti-sceptic, because we do not have any scientific knowledge at all, in which case the sceptic's work is done for him anyway.

It seems we are left in a lose-lose situation. If sceptical arguments are correct, even if they are part of science, they may make all of science unavailable to us. Naturalizing epistemology, and everything else, leaves our knowledge just as vulnerable as before, because all of science is vulnerable to the sceptical argument. Finding epistemology's sanctuary in science seems to provide us with many resources for answering the sceptic, but all those resources disappear if the sceptic's arguments are correct. If the sceptic is correct and we see that all of our theories are underdetermined, we cannot appeal to any of our scientific theories in answering the sceptic. If we do not know anything about the external world, we do not know if our theories are true, and if we do not know if our theories are true, we cannot use them to answer the sceptic.

But Quine's position may still be stronger than it appears, even after we admit the possibility of a reductio ad absurdum of science. Quine would insist that a reductio ad absurdum of science would itself be a scientific fact to be discovered empirically. Under a naturalized epistemology, the question of whether science is absurd will have to be studied scientifically, just as everything else is studied. In this way, Quine admits that we may find that science is absurd in the way a reductio ad absurdum would show it to be, but such a fact is to be discovered by science (Quine, Reply to Stroud 475). We may also find out that science works just fine; naturalizing epistemology leaves such questions up to science, and in order to dismiss all scientific discoveries we need a scientific reason to do so. After Quine's proposal there is no other way we can argue. The reductio ad absurdum is a possibility, but a scientific one, which will be vindicated or disproved by science itself.

It would seem, if Quine is correct, that we simply have to wait for the science to come in if we are to get a good answer to the epistemological, sceptical question of our knowledge. The psychologists will study as many humans as they can and come back to us with a verdict; over time our knowledge about our knowledge will strengthen, and we will come to find one theory or another works best. But at this point, things also seem to get strange. What does the psychologist know about epistemology? She can tell us about what is happening in our brain when we claim to have knowledge, or the way people act when they say they have knowledge, and so on, but it seems difficult to say that we would ever be satisfied with a psychological response to scepticism.

The impression is that Quine's naturalized epistemology, while avoiding the problem of scepticism, seems to be drifting away from epistemology altogether; Quine is unknowingly changing the subject. Barry Stroud stresses this point (Stroud, The Significance of Philosophical Scepticism 251). As he explains, Quine wants both to talk of epistemology as strictly a study of causal chains of events, in other words as a purely scientific study, and also as the study of how the meager input for our senses leads to the torrential output of our theories. The difficulty is that strictly naturalized epistemology, in which the only thing that is studied is causal chains of events from light hitting eyes to belief-events in humans, would leave out all possibility of underdetermination; there is no such thing as a cause underdetermining its effect. Either the cause preceded the effect, and the effect would not have happened without it, or it did not. As Stroud explains, there are no gaps in causal chains, as there are in chains of inference from meager input to torrential, theoretical output: "It makes no sense to say of one event (e.g., an impact at a sensory surface) that it 'underdetermines' another events (e.g. a coming-to-believe-something) that occurs later in the series" (Stroud, The Significance of Philosophical Scepticism 251).

Causal underdetermination then cannot be the sort of underdetermination that Quine would have a naturalized epistemology solve. Rather, it is informational underdetermination that Quine wishes to address; the problem of underdetermination is that the data we get at our sensory inputs underdetermines the information we are to surmise from that data. The sceptical question is concerned not with how we are caused to come to believe something, but rather with the truth of that belief given the sparse information we are given by our senses. This is not, and cannot, be a causal question, because once we are concerned with causal determination there is no question of underdetermination; underdetermination only applies in the case of inference from data to theory. As Stroud explains:

It is the truth or falsity of the content of the 'output' that Quine says is not 'determined' by the data or the sensory impacts; the relation of 'underdetermination' holds between one set of truths and another.... [Quine] asks how knowledge is possible, given that 'the only information that can reach our sensory surfaces from external objects' is 'meager' in relation to what we come to believe about those objects as a result of receiving that sensory 'information'. That gap is just what gives rise to Quine's epistemological problem. (Stroud, The Significance of Philosophical Scepticism 251)

If we are to drop all epistemological talk in favour of talk of causal chains, as a naturalized epistemology would have us do, then we abandon the possibility of even addressing the sceptical problem; causal chains leave no room for underdetermination. Naturalized epistemology allows us to avoid the problem of meager input underdetermining torrential output, but it does so at the cost of becoming irrelevant to the problem of scepticism. Far from being able to answer scepticism, naturalized epistemology becomes inert to it. As Stroud says, "[Quine] wants to avoid all questions of awareness. But he can do so only by avoiding all talk of the 'meagreness' of our 'input' relative to our 'torrential output' as well" (Stroud, The Significance of Philosophical Scepticism 252).

With nothing but causal explanations of physiological events, naturalized epistemology is left with no way to confront the sceptic, because it cannot address the sceptic at all. Naturalized epistemology may have its practical benefits, but it does not give us progress in the sceptical problem. Other approaches, which directly confront the sceptic, may prove useful.


Nozick's reply to scepticism is to deny the principle of epistemic closure, which he takes to be a central tool in the sceptic's argument (Williams, Unnatural Doubts 336). The principle of epistemic closure states that if I know a proposition p, and I also know that p implies q, then I know that q.

Nozick's contention is that the principle of epistemic closure is vital to the radical sceptical argument about knowledge of the external world. Every such argument must have a premise that uses epistemic closure to draw a conclusion about our knowledge from the knowledge we take ourselves to have (Nozick 172-178). For example, if I know that I am sitting in a chair, and I know that my sitting in a chair implies that I am not dreaming, then I know I am not dreaming. From here we take note that I do not know that I am not dreaming, and if I do not know that I'm not dreaming, then I cannot know that I am sitting in a chair; if I do know I am in a chair, then from the principle of epistemic closure it follows that I know I am not dreaming. Nozick's strategy is to deny the principle of epistemic closure, thereby blocking the sceptical argument from getting off the ground; it no longer follows from the fact that I do not know that I am not dreaming that I also do not know I am sitting in a chair.

Dretske takes a path similar to Nozick in responding to the sceptic, though it takes a different name. He bases his discussion on what he calls penetrating operators, but the main issue is nonetheless still closure (Dretske 1014). Dretske uses the example of going to a zoo and looking at a zebra (Dretske 1016-1017). We know that it is a zebra, yet we do not know that it is not a disguised mule. Dretske's way of describing this is to say that our knowledge fails to penetrate to the presuppositions of our knowledge; our knowledge of that being a zebra does not penetrate to knowledge about it not being a disguised mule. For Dretske, as for Nozick, closure fails; we know x without having to know what is implied by x. And, it must fail if we are to avoid scepticism, because if closure does not fail, then we are left in the position of having to know that what we think is a zebra is not a disguised mule, or a shape-shifting alien, or a clever painting, and so on.

A difficulty with this strategy is that the principle of epistemic closure seems to be perfectly reasonable. It is intuitive that if I know I am at home, and I know that being at home means I'm not at school, that I know I am not at school. Giving up the principle of epistemic closure seems to deny us knowledge that we take to be obvious. In this way, Nozick and Dretske's replies to scepticism amount to an epistemic concession to the sceptic; by denying closure we have blocked the sceptical argument and saved a great deal of our knowledge, but at the same time have denied ourselves a good deal of other knowledge. We may be able to say that we know things about the external world, but cannot say that we know things that are implied by closure; I know that I am sitting in a chair at home, and know that being at home means not being at school, but do not know from deduction that I am not at school, apparently.

Perhaps denying closure does not have as drastic an effect on our knowledge as radical scepticism does, but we have not gained very much ground through this denial. At best we have replaced the problem of scepticism about knowledge of the external world with a sort of scepticism about knowledge from closure. If we are to find a plausible and acceptable answer to scepticism, we should try not to give up knowledge that we find intuitive, because that is what the sceptical argument would take away from us in the first place.

Further, the principle of epistemic closure may not even be necessary for the sceptical argument. Michael Williams has argued that the sceptic does not in fact need closure at all to make his argument. Williams' contention is that Nozick and Dretske unknowingly confuse two questions when they attack closure in an attempt to stave off scepticism. One question is whether knowing a proposition p and that p implies q means knowing that q; this question about closure implies nothing about what it takes to know p. We may ask if knowing that I am at school and that being at school means not being at home leads to me knowing I am not at home, but this says nothing about what it takes to know that I am at school. The question of what it takes to claim knowledge of p might or might not remain open, depending on your epistemological position. If it does remain open, another question arises about whether we must know that we know p if we are to have knowledge of p. This question is not about closure; it is about a stronger epistemic principle. It is this stronger epistemic principle that the sceptic actually makes use of.

Knowing that we know p, for the sceptic, involves knowing the presuppositions of p. For example, take the familiar sceptical argument about dreaming: if I am to know that I am sitting in a chair I must know that I am not dreaming, so that I know that I know I am in a chair. It is a presupposition of knowing that I am in a chair, according to the sceptic, that I know I am not dreaming. This is perhaps a stronger principle than epistemic closure; it implies not that knowing p and that p implies q leads to knowing q, but that in order to know p at all, we must first know that q. What the sceptic really needs might be called the principle of epistemic presupposition (Hymers 224). The question of presupposition is, as Williams puts it: "must I first know that the relevant presuppositions hold in order to come to know that P?" (Williams, Unnatural Doubts 331) This is a different question from the question of closure. But the presuppositions of p are often, if not always, the same as the q given to us by closure. For example, take our earlier example of closure: if I know that I am at school (p) and that being at school means not being at home (p ⊃ q), then according to closure I know I am not at home (q). Epistemic presupposition makes use of the same propositions, but of different relations between the two; if I must know that I know p, then if I am to know that I am at school (p) I must first know the presupposition of p, that I am not at home (q). I must first know q before I can claim knowledge of p, where q and p are the same propositions in both the case of closure and of epistemic presupposition. In this way it is easy to see how the principle of closure was mistakenly taken to be required by the sceptic; they are very closely related. But, they are not the same thing. Once we make note of this, we see that a denial of closure is not only unintuitive, but ineffective against radical scepticism anyway.

Nozick and Dretske are not necessarily wrong to deny the principle of epistemic presupposition, because it does seem to put unreasonable constraints on our knowledge. It is a presupposition of water being a liquid that its temperature be above 0° Celsius. Epistemic presupposition implies that if you know that some water is a liquid, then you must first know that its temperature is above 0° Celsius. But this is strange; you know that the water is liquid when you notice that it flows, not given this observation after checking the temperature (Hymers 224). It is counterintuitive to imagine that when finding some liquid water we must know that the temperature is above 0° Celsius before we can know that it is a liquid; we know it is liquid, without necessarily knowing anything about the temperature.

Epistemic presupposition, which the sceptic really requires, is not the same as closure. By conflating the sceptic's use of closure with epistemic presupposition, Nozick and Dretske have suggested that we deny the intuitive principle of closure, while doing no damage to the sceptic's position. It seems reasonable to hold on to closure and look for more effective responses to radical scepticism. It also seems reasonable to deny something like the principle of epistemic presupposition, but we will have to have good reasons for doing so; the sceptic may still argue that the principle of epistemic presupposition is necessary, given sceptical considerations.