Epistemological Scepticism

Forms of Scepticism

Philosophy contains many kinds of scepticism, which we can categorise along the following dimensions:

1. Subject-matter.

2. Scope, or extent.

3. The strength, or corrosiveness, of the scepticism.

    1. The nature of the sceptical argument.

Firstly, as regards the subject-matter of the sceptical doubt, not all scepticism concerns the central epistemic concepts. Thereís scepticism about inductive inference, for example, and about meaning and rule-following. (These wonít concern us much for the moment). Our concern is with those varieties of scepticism which attack one or more of the epistemic concepts that weíve been using, such as knowledge, belief, justification and understanding. These are versions of epistemological scepticism.

2. Some scepticisms are universal or global in scope; they take a concept and deny the existence of any instances of it. Thus they may deny the existence of any knowledge, any justification, any understanding, etc. Others, more limited in scope, are partial or local, denying only that we have knowledge (or justified belief) about some limited domain, such as mathematics, other minds, the future, the past, or the external world.

3. Sceptical conclusions differ in strength along two dimensions. The first is the dimension of modality. Some sceptics claim that there are no actual examples of epistemic concept E. They might claim, for example, that thereís no knowledge, or no certainty, or no justifiction. But a more radical scepticism would claim not just that there are no examples of concept E, but that there can be no such examples. This is to claim not merely that knowledge doesnít exist, (a deficiency we might think we could remedy by adopting some more stringent procedures of inquiry) but that it couldnít exist.

Along another dimension, the strength of a scepticism is inversely proportional to the strength of the epistemic achievement that it disputes. Thus, if one holds that there is or can be no such thing as absolute certainty, for example, one has only a weak form of scepticism, since absolute certainty is pretty demanding. But if one holds that there is or can be no such thing as knowledge, one has probably moved to a significantly stronger grade of scepticism, since the conclusion seems more destructive. If, at the extreme, one denies either the existence or the possibility of either justified belief or of understanding, one is entertaining perhaps the most corrosive scepticism.

Scepticism, Dogmatism, and Fallibilism

Some think scepticism is a good thing because itís the antidote to Ďdogmatismí. This is just a misconception. Dogmatism is the tendency to hang on to oneís beliefs either when one doesnít have enough evidence to support them, or in the face of countervailing evidence. Its real antidote is fallibilism, the recognition that oneís beliefs could always be wrong. But fallibilism does not amount to scepticism. Fallibilism allows one to believe things, scepticism demands that we believe absolutely nothing. Please donít confuse the two!

The Argument from Error

There are several types of familiar sceptical argument. Perhaps the most important is the argument from error, which is often the only sceptical argument that writers on epistemology get around to talking about, although it has more than one form. But the existence of rival sceptical hypotheses is also important. Itís the argument from error which we will begin with.

This argument is motivated by the suggestion that you donít know that a proposition is true if you could be mistaken about it. The idea is that, just as with the moral of the Gettier counterexamples, lucky guesses and happy accidents donít count as knowledge: if itís accidental that the belief is true, it could easily have been false, so presumably only the most reliable of reliable procedures should secure knowledge. Knowing must somehow rule out not just actual error, but the threat of error itself.

The sceptic insists that the threat of error is serious. Such a threat might arise from one of several sources. Firstly, the sceptic might seek to cast doubt upon our cognitive resources. Our cognitive resources include sense experience, memory, and reasoning. Our sceptic might have empirical grounds for holding that any one or more of these faculties is somehow deficient, or that we are incapable of exercising them efficiently. Secondly, she might focus on alleged weaknesses in the relation between the mind and its objects, or on alleged peculiarities of the objects which we purport to know. Thus, for example, one might be sceptical about our knowledge of mathematics, if the objects of mathematics are thought to be timeless abstract Ďideasí such as numbers, or propositions. Or one might worry about the mindís access to abstract entities, or about its access to other minds, for example. And thirdly, the sceptic might argue that the phenomenon of evidential gaps, that is gaps between the evidence for a hypothesis and that hypothesis itself, indicate that alternative hypotheses are both compatible with the evidence and incompatible with the original hypothesis. Such cases arise in scepticism about induction, about other minds, and about the past, for example.

Whatever the source of the errors, then, the sceptic insists that we must take the possibility of error very seriously indeed. The argument then proceeds: You have made mistakes in the past. And nothing in your present situation tells you reliably that that situation is not one in which you are mistaken. So since you obviously did not know in those past cases where you were mistaken, you canít reasonably claim to know in the present instance. The conclusion of the argument can be generalised to any situation, yielding the conclusion that we have no knowledge at all, in fact that no knowledge is possible.

The argument, as Jonathan Dancy points out, tacitly relies upon a universalizability principle, to the effect that a judgment that this is a situation in which I know that p commits me to making a similar judgment in relevantly similar circumstances. If I recognise that I have even once wrongly claimed to know that p, then I should never claim knowledge of any proposition unless I can actually point to a relevant difference between that case and this one.

But the fact that I have in the past made actual errors is unnecessary, strictly speaking. The mere possibility of error will suffice: even if I have never actually been deceived in this type of epistemic situation, the ever-present possibility of error will be enough to generate possible situations in which one wrongly thinks one is in possession of knowledge. Universalizability means that we have to be sure that our present epistemic predicament is not one of these possible cases.

Rival Sceptical Hypotheses

Not all the most famous and challenging sceptical arguments are versions of the argument from error. Instead, some introduce what I shall call rival sceptical hypotheses which are intended to be completely incompatible with our ordinary beliefs about our situation. The most famous are Descartesí idea that one might at any given moment be dreaming, his Ďevil geniusí hypothesis, and, more up-to-date, the idea that one might be a brain in a vat. (The well-known film The Matrix presents a related scenario. For philosophical reactions to it, see the 'philosophy' section on its website:  http://whatisthematrix.warnerbros.com/). Each is supposed to show that there is a rival hypothesis which explains the observable phenomena just as well as our more ordinary beliefs. The sceptic then hopes to unsettle us by inquiring as to why we should rationally prefer our hypothesis (our ordinary account of the situation) rather than hers, or neither. And the sceptic wonít be satisfied by an answer like "Well, we were just brought up to believe it"!

Here we mustnít make the mistake of thinking that the sceptic is saying that his sceptical possibility is the case. Sheís just saying that no-one can rule out the possibility that we might be in that predicament. She challenges us to point to some feature of our hypothesis which rationally recommends it over her own, more bizarre, possibility. Itís obviously important here that the situation the sceptic describes is a genuine, coherent possibility, and not a nonsense, so the sceptic has to be a little careful. But the imagination of sceptics is fruitful, and they donít normally get caught out that easily. (The brain-in-a-vat scenario, though, may be a nonsense. It depends on whether brains (rather than the just organisms that have brains) can be credited with thinking).

Behind the postulation of the rival scepticial hypothesis is a sceptical argument, the general form of which is as follows:


(1) You donít know that youíre not in the sceptical predicament (dreaming, being deceived by an evil demon, a brain in a vat, etc).


(2) If you were in that predicament, nothing in your experience would reveal to you that you were, since the experience would be identical to that you would otherwise have.


But (3) if you donít know that youíre not in the predicament, then you do not know anything else (about the external world, at least).


If you claim to know that p, where p is some empirical proposition which conflicts with the supposition that you are in the sceptical predicament, then from the fact that if you know that p you know that you are not in the predicament, we can conclude that since you donít know that you are not in the predicament, you donít know that p! Underlying the argument is a general principle, the principle of closure for knowledge, which Dancy formulates as:

PCk: [Kap & Ka(p --> q)] --> Kaq.

If this principle holds generally, it seems to show that "since you donít know that you are not a brain in a vat you cannot know any proposition p of which you know that if p were true, you would not be a brain in a vat" (Dancy, p.11). One way to attack the argument from rival sceptical hypotheses would be to undermine this principle.

The Threat of Self-Stultification

Some conceivable sceptical positions are so corrosive that theyíre self-stultifying. For example, any argument whose conclusion is that understanding is impossible, or even contingently non-existent, is hoist with its own petard: if understanding doesnít exist then we canít understand either the argument for that conclusion, or its conclusion. Self-refutation of this form is a whirlpool that sceptical arguments must always skirt. Itís sufficient to dispel at least some of the more extreme forms of scepticism. Those who argue that thereís no knowledge mustnít claim to know their conclusion, and those who claim that justified belief is unattainable mustnít mind if we feel that theyíre not justified in believing their argument. Thus scepticism certainly has its risks, as well as its (alleged) payoffs. Scepticism about understanding, at least, can only be local.

It may seem that the application of this self-refutation argument is rather narrow, and that it can affect only the most extreme and uncompromising forms of scepticism. But one can show, I think, that the honest sceptic simply canít resist the pull of her own arguments, and that they will inevitably pull her into one of these most extreme and self-refuting forms of scepticism. Take someone whoís sceptical about knowledge. Knowledge, it is generally thought, is true belief plus some feature Ďxí, where x is some feature such as justification, evidence, certainty, causation of the right kind, counterfactual dependence, reliable method, etc. (Note that even sceptics agree to this much, since they must think they know what knowledge is in order to deny or doubt its existence).

So knowledge contains three ingredients: truth, belief, and x. The motivation for scepticism about knowledge is the thought that (at least) one of these ingredients isnít available. But presumably its motivation is not the thought that truth isnít available (thatís not scepticism, itís nihilism), or the thought that belief isnít available (since itís precisely the sceptic who claims that belief is all-too-available!). So the idea must be that itís the third ingredient, Ďxí, that we canít attain. This reveals the sceptic about knowledge as a sceptic about Ďxí. Scepticism about x, however, can itself be analysed, depending on what x is. If x is some external feature such as causation, counterfactual dependence, or reliability, the sceptic must insist that knowledge must involve more. Sceptics canít rest content with externalist analyses of knowledge. Even if they allow that an external feature plays some role in knowledge, theyíll insist that itís not its externality that makes it unattainable. The scepticís idea is to cast doubt not on the constitution of the world (on this issue, presumably, she frames no hypotheses), but on our powers to represent it. So it must be the internal aspect of the x-feature that the sceptic is challenging. And this, crudely, amounts to whether the belief is justified, warranted, certain, etc. In other words, a sceptic about knowledge uses something like a Ďjustified true beliefí analysis of knowledge, and then claims that although belief is available, justified belief isnít. So a sceptic about knowledge has been revealed as, at heart, a sceptic about justified belief. But this conclusion administers the coup de grace to the sceptic: both scepticism about justified belief and scepticism about understanding are, as we have already shown, self-refuting.

Itís possible to reach this conclusion in an even shorter way, with a single step: since understanding something is knowing its meaning, scepticism about knowledge is itself (the manifestly self-refuting) scepticism about understanding!

Dancy, on the other hand, remarks that "the sceptic should be entirely unworried by such arguments" (Introduction to Contemporary Epistemology, p.17). He construes the scepticís argument as hypothetical in form, as being of the form: if I know this, then I know that knowledge is impossible. So if I know anything I know nothing. The import of the sceptical argument is then to expose a paradox within the epistemic concept. Note that this argument forces the sceptic into holding that the epistemic concept she has in mind (certainty, justified belief, understanding, etc.) is not just contingently uninstantiated, but logically impossible. And the sceptic should worry that this argument means that she has nothing to assert; she canít express her doctrine in any assertible proposition that isnít self-refuting, and so she canít communicate it to others, or, indeed, to herself. (So, after all, Cratylus, who asserted nothing, but was content to sit in the corner and waggle one finger, was the most consistent sceptic).

But the sceptic has already fallen into a trap here. The more she does to convince us that what she calls Ďknowledgeí is impossible, the more she has to do to convince us that her analysis of knowledge answers to our concept of knowledge. If we ignore our standards for correct commonsense judgments about which cases count as cases of knowledge, weíll have no means of specifying the concept weíre talking about. This is because "the concept is pinned down by [what we take to be the correct] judgements which we make using it. Itís by reference to our judgements using the concept that we try to construct our analysis of the concept. So if we produced an analysis according to which most of our judgements using the concept were mistaken, there could be no grounds for saying that we had correctly analysed that concept... [A]n analysis which had a very bad fit with the judgements we make using the term would be for that reason a bad analysis" (Fisher & Everitt, p.24). The scepticís analyses of knowledge fail to correspond at any point to the judgements any of us make using the term; so they fail to be plausible analyses of our concept of knowledge.

Swallowing the Conclusion?

Is the conclusion of these sceptical arguments, that we never can have knowledge, one which we could live with? Could we just say, "So much the worse for knowledge, we never had any"? I hope to show that this is not a feasible response. So far weíve framed the argument from Rival Sceptical Hypotheses in terms of the concept of knowledge, but itís supposed to work just as well (or should that be Ďjust as badlyí?) if we run it in terms of justified belief. It would then go as follows:

(1) You have no justification for thinking that you are not in the sceptical predicament.

(2) If you were in that predicament, nothing in your experience would give you reason to believe that you were, since the experience would be identical to that you would otherwise have.

(3) But if you have no reason to believe that you are not in the sceptical predicament, then you have no reason to believe anything else incompatible with that supposition.

The conclusion of this version of the argument is that there is (almost) nothing we are justified in believing (the exceptions are things like the cogito, and, perhaps, a priori truths). Again, you might think that you could live with this too. But youve now fallen into the scepticís trap. The scepticís conclusion is not that we are never completely justified in believing anything. Itís that we are never justified to any degree. This means that the first premise of this version of the argument is far more corrosive than the first premise of the knowledge version. The new first premise is to be understood as saying that you have no justification whatsoever for thinking that youíre not in the sceptical predicament. This sounds implausible. But the sceptic will support it by saying that any reason for believing that you are in a real world rather than a sceptical predicament must be given in terms of a difference in your experiences. And, ex hypothesi, in this case there is no such difference.

The conclusion of the argument is that weíre never justified in believing anything. If weíre never justified in believing anything, to any degree at all, that means that with respect to any proposition p and competitors to it, you can never have reason to favour one over any of the others. You can never have reason to favour the hypothesis that the sun will rise tomorrow over the hypothesis that it wonít, or the hypothesis that you have two legs over the hypothesis that you have 17, or 1765 legs. And if you can never have reason for believing one over the other, then itís irrational to do so. To be epistemically rational would consist in suspending belief about all empirical propositions.

Norman Malcolmís Anti-Sceptical Argument

Iíve already suggested that scepticism must avoid self-refutation, and that it conspicuously fails to do so. But there are other ways of dealing with scepticism. One short way with scepticism consists in holding that it is senseless. Norman Malcolm argued thus.

The sceptic about, e.g. other minds, complains that one never has good evidence as to the contents of another personís consciousness. But if this complaint is to be meaningful, then the sceptic must be able to coherently describe a situation which would count as our having direct evidence of another personís consciousness, a situation that does not in fact obtain. This could be a kind of situation that humans can never be in, but it must be possible for some conceivable being to be in that situation. If we ask the sceptic to do this, she might say that what would count as, for example, direct awareness of the contents or existence of another personís consciousness would be having the very same experiences as that other person. If one could do this, she might say, she would allow that we have direct awareness of the existence and the contents of other minds. Or, in another example, she might say that for a proposition about a material object to be known with certainty would be for one to have performed an infinite number of checks, all of which are confirmed.

Malcolm claims that the problem with these suggestions is that they do not describe coherent possibilities. The idea that one might be having numerically the same toothache as someone else (not just a qualitatively identical one) is self-contradictory. Note that telepathy doesnít count, since thereís still my experience of the toothache and yours. We donít have numerically the same experience, or thought. Likewise, the idea that one might perform an infinite number of tests in a finite time is self-contradictory too:

It is self-contradictory to say with regard to a given proposition, p, that certain people began at a definite time, t1, to make verification tests, and at a certain definite time, t2, they ceased to make tests, and between t1 and t2 they performed an infinite number of tests.

Malcolm sums up this charade as follows:

What the [epistemological sceptics] have done is to invent a self-contradictory meaning for the expression Ďknowing for certain that another person is having an experienceí. They then announce regretfully, that we can never really know for certain that other persons have experience, and that they are not really automatons. It should be clear that the meaning which [sceptics] give to that expression is an artificial one. It would be absurd to maintain that when in ordinary life a person says "Itís quite certain that Joan is not pretending. I know for certain that she really is in great pain", what he means is that he is having numerically the same pain that Joan is having. And it would be absurd to maintain that his statement is in any other way self-contradictory. (ibid., pp.24-5, my emphasis).

In fact the scepticís whole scenario is a philosophical farrago. Itís not that she has set an unattainably high standard, a standard which even our best efforts then fail to reach. Itís that the scepticís standard is no standard at all: itís attainment is logically impossible. Sheís changed the meaning of the expression Ďcertaintyí. Failing to Ďattainí her false standard is no failure at all: "One can fall short of an ideal", Malcolm says, "only if it makes sense to speak of attaining the ideal. But the philosophers have defined Ďcertaintyí in such a way that it does not make sense to speak of attaining certainty. And therefore it does not make sense to speak of failing to attain certainty" (p.22, my emphasis). The sceptic has attached a self-contradictory meaning to a perfectly ordinary expression.

Notice that Malcolmís argument is apparently directed only at those forms of scepticism which hold that an epistemic concept is contingently uninstantiated. Malcolm presupposes that if the scepticís complaint is to be meaningful, the sceptic must be able to coherently describe a situation which would count as having direct evidence of another personís consciousness, albeit a situation that does not in fact obtain. But to presuppose this is just to presuppose that such evidence, or justification, or knowledge, is possible. We saw that some radical sceptics will seek to deny this very presupposition, claiming that knowledge is impossible. Malcolm would almost certainly accuse them of putting a philosophical misconstruction upon our ordinary language.

The debate now becomes an issue of whether the sceptic has not simply changed the meanings of familiar words like Ďknowledgeí, Ďcertaintyí, Ďjustificationí, etc. Is it that the sceptic is systematically misusing these words? If she persistently uses the word Ďjustificationí in some peculiar way, to mean something different from the rest of us, why take her complaint seriously? Could it be that an appeal to meaning is enough to refute the sceptic?

Logical and Empirical Possibility

In the course of his critique, Malcolm introduces an important distinction between logical and empirical possibility. He admits that what the sceptic points to are logical possibilities. Bertrand Russellís sceptic about the past, for example, holds that it is possible that the world was created 5 minutes ago, with all memories and false records intact. A modern techno-sceptic claims that you canít rule out the possibility that youíre a brain in a vat. Malcolm says: these are logical possibilities. To say that something, p, is logically possible is merely to say that the statement Ďpí is not self-contradictory. But he then goes on vigorously to deny that these logical possibilities are genuine empirical possibilities. He says, "I hold that it is an absolute confusion to say that because it is logically possible that the world was created five minutes ago, therefore there is some possibility that the world was created five minutes ago" (p.27). He explains the concept of empirical possibility by giving familiar examples:

The doctor says, "With your leg in that condition it would be impossible for you to march in the parade to-day; but perhaps in three days you will be able to get about all right". The doctor certainly does not mean that the statement, "you will march in the parade to-day", is self-contradictory. And he certainly does not mean by his second statement, merely that it is logically possible that you will get about all right in three days. (ibid., p.28).

The concepts of logical and empirical possibility are distinguished by two features. Only empirical possibility admits of degree, since logical possibility is obviously an all-or-nothing matter. So, when speaking of empirical possibilities one can say "thereís some possibility that p", "itís barely possible that q", "thereís a slight possibility that r", etc. And, secondly, only empirical possibility admits of tense, since logical possibility is tenseless. With empirical possibilities one can say things like "there was a possibility of it, but there isnít any more". Confusing the two concepts of possibility leads to misunderstanding the nature of certainty. From the fact that something, p, is logically possible sceptics conclude, wrongly, that thereís some possibility of p; and from this they conclude that itís not absolutely certain that not-p. This is their favourite argument for the conclusion that no empirical statements are certain, but if Malcolm is correct it is entirely invalid.

What would happen if we did change the epistemic concepts in the way the sceptic wants us to? We still have a practical need for words answering to the old epistemic concepts. And the new ones, by contrast, are useless: thereís almost nothing substantial and informative one can say with them!

Pragmatic Reasons for Resisting Scepticism

Iíve agreed with Malcolm that sceptics arenít interested in our current epistemic concepts, as they should be, but are in fact trying to replace those concepts by new ones. Well, what would happen if we did change or replace them in the way the sceptic wants us to? What would be gained?

Using the new sceptical epistemic concepts, we would no longer be able to say of ourselves, or others, that we know anything, are certain about anything, have reason to believe anything, or are justified in believing anything. In fact, these would be the only substantial and informative things one could say with them! Nevertheless, we would still have a need for words answering to the old epistemic concepts. Why? Because our practical interest in those concepts is in making distinctions that are enduringly important in the stream of human life (any human life). We need to distinguish between what people know, and what they merely believe, between people who know things, and people who only believe them, between people who believe things on the basis of evidence, or sufficient evidence, or reasons, or good reasons, and people who donít. Our old epistemic concepts make exactly these distinctions. The scepticís new concepts, by contrast, can make none of them. These new concepts are typical philosophersí creations that are, for practical purposes, absolutely useless. The usefulness of the old epistemic concepts is the fundamental argument in their favour.