Skepticism regarding a subject matter is the view that knowledge about the subject matter is not possible. Many subject matters have come under skeptical attack. It has been argued, for example, that it is not possible to obtain knowledge about the external world, about as-yet-unobserved states of affairs, and about minds other than one's own. This entry will focus upon skepticism about knowledge of the external world.
The Cartesian Skeptical Argument
The following skeptical argument is suggested by Descartes's first Meditation. Consider the skeptical hypothesis SK: There are no physical objects; all that exists is my mind and that of an evil genius, who causes me to have sense experience just like that which I actually have (sense experience representing a world of physical objects). This hypothesis, says the skeptic, is logically possible and incompatible with propositions implying the existence of the external world, such as that I have hands. The skeptic then claims that (1) if I know that I have hands, then I know that not-SK. To justify premise (1), the skeptic points out that the proposition that I have hands entails not-SK, and he asserts this closure principle: If S knows that ϕ and S knows that ϕ entails ψ, then S knows that ψ. The skeptical argument's other premise is that (2) I do not know that not-SK. To justify this premise, the skeptic points out that, if SK were true, then I would have sense experience exactly similar to that which I actually have. Because my sensory evidence does not discriminate between the hypothesis that SK and the hypothesis that not-SK, this evidence does not justify me in believing not-SK rather than SK. Lacking justification for my belief that not-SK, I do not know that not-SK. From (1) and (2) it follows that I do not know that I have hands. A similar argument may be given for each external-world proposition that I claim to know.
Those who think that minds are physical in nature may well balk at the skeptic's claim that the evil-genius hypothesis is logically possible. Accordingly, the skeptic will replace that hypothesis with this updated version of SK: I am a brain in a vat connected to a computer that is the ultimate cause of my (thoroughly unveridical) sense experience.
To see how the foregoing pattern of skeptical reasoning may be extended to other subject matters, let the target knowledge claim be that there are minds other than my own, and let the skeptical hypothesis be that the complex patterns of bodily behavior that I observe are not accompanied by any states of consciousness. The analogue to premise (2) will in this case be supported by the claim that, if the skeptical hypothesis were true, then I would have behavioral evidence exactly similar to that which I actually have.
Denying the Logical Possibility of SK
Let us consider two radical responses to the Cartesian skeptical argument. The evil-genius and vat hypotheses both depend on the assumption that the external world is mind-independent in such a way that it is logically possible for sense experience to represent there to be a physical world of a certain character even though there is no physical world, or at least no physical world of that character. An idealist denies this assumption of independence. The idealist maintains that facts about physical objects hold simply in virtue of the holding of the right facts about sense experience, then denies that skeptical hypotheses such as SK are logically possible: any world in which the facts of sense experience are as they actually are is a world in which there is an external reality of roughly the sort people take there to be. Thus premise (2) is false: I know that not-SK in virtue of knowing the necessary falsity of SK.
The second radical response to the skeptical argument rests on a verificationist constraint on the meaningfulness of sentences. Like the idealist, the verificationist holds that the sentence "I am a victim of thoroughgoing sensory deception" fails to express a logically possible hypothesis. Given that the sentence fails to express a proposition for which sense experience could in principle provide confirming or disconfirming evidence, the verificationist counts the sentence as meaningless. Because the sentence expresses no proposition at all, it does not express a proposition that is possibly true.
The antirealist puts forward a similar view, maintaining that one's understanding of a sentence's meaning consists in a recognitional capacity manifestable in one's use of the sentence. Suppose that the conditions under which a sentence X is true transcend people's powers of recognition. Then one's understanding of X's meaning could not be identified with one's grasping of X's recognition-transcendent truth conditions (because such a grasping could not, in turn, be identified with a manifestable recognitional capacity). This conception may be applied to sentences that allegedly express skeptical hypotheses. If people cannot detect the obtaining of their truth conditions, then what is understood when skeptical sentences' meanings are understood must be something other than their truth conditions. Grasping such sentences' meanings must instead consist in grasping the detectable conditions under which they are warrantedly assertible. Thus, it would turn out that an allegedly problematic skeptical hypothesis fails to make any coherent claims about putative conditions in the world that outstrip the human capacity for knowledge.
Attacking Premise (1)
Premise (1) has come under attack by those who think that the skeptic has succeeded in stating a hypothesis that is genuinely logically possible and not known to be false. On this strategy the closure principle is denied. This opens up the possibility that I know that I have hands even though I do not know that not-SK. For example, one may deny closure by maintaining that knowing that ϕ requires knowing only that the relevant alternative hypotheses to ϕ do not obtain. Skeptical hypotheses, it is then said, are not relevant alternatives to the propositions involved in ordinary knowledge claims.
Another way of denying closure is to hold that S knows that P if and only if (i) S correctly believes that P, and (ii) S would not mistakenly believe that P if P were false. To satisfy the tracking condition (ii), S must not mistakenly believe that P in the possible worlds in which P is false that are most similar to the actual world, according to the standard semantics for counterfactuals. (Robert Nozick adds the further tracking requirement that S believes that P in the possible worlds in which P is true that are most similar to the actual world.) Now suppose that some hypothetical normal believer S satisfies these conditions with respect to the proposition that he has hands (S correctly believes that he has hands and would not mistakenly believe that he has hands in the no-hands possible worlds most similar to his world, worlds in which, say, he has lost his hands in a terrible accident). Then S knows that he has hands. But in all the possible worlds in which not-SK is false (SK worlds), S mistakenly believes that not-SK (he mistakenly believes that he is not in a vat). So S does not know that not-SK, even though this proposition is entailed by the proposition that S has hands. This is a counterexample to the closure principle.
Attacking Premise (2)
Let us turn to antiskeptical strategies that do not challenge premise (1) and that accept that SK is indeed logically possible. On these strategies, premise (2) is attacked. For example, Kant tried to show via a transcendental argument that, in allowing knowledge of certain key features of one's own mind, the Cartesian is already committed to the possibility of knowledge of the external world. Kant argued (in "Refutation of Idealism" in Critique of Pure Reason ) that, in order to have knowledge of one's own temporally-ordered inner states, one must also have knowledge of spatial objects outside one's mind, whose temporal ordering is related to that of one's inner states. A prima facie difficulty for the Kantian strategy is that arguing for a connection between knowledge of one's mind and knowledge of the external world seems to require the assumption of verificationism or idealism, which would render superfluous the rest of the transcendental argument.
The inference to the best explanation strategy relies on the idea that, even if two incompatible explanatory hypotheses are equally supported by the available evidence, I am still justified in rejecting one hypothesis if the other offers a better explanation of the evidence. It might be maintained that the ordinary hypothesis that the world is roughly as I take it to be offers a better explanation of my sensory evidence than does SK, in virtue of its greater simplicity. Thus, I can justifiably reject SK. The proponent of this strategy needs to specify the respect in which SK is more complex than the ordinary hypothesis and to make it plausible that hypotheses that are complex in the specified way are less likely to be true than simpler ones.
Another way to attack premise (2) is to adopt a reliabilist theory of knowledge, according to which knowing that ϕ is a matter of having a reliably produced true belief that ϕ. If reliabilism is correct, then in arguing that I do not know that not-SK, the skeptic would have the difficult burden of showing that there is in fact some flaw in the belief-producing mechanism that yields my belief that not-SK (thereby precluding that belief's amounting to knowledge).
Let us return to the skeptic's defense of his premise (2). To validate the premise, the skeptic needs to appeal to an epistemic principle that is (apparently) distinct from the closure principle. This is the underdetermination principle:
(UP) If S's evidence for Φ does not favor Φ over a competing incompatible hypothesis Ψ, then S is not justified in believing Φ.
The skeptic maintains that one's perceptual evidence would be the same regardless of whether SK holds or not-SK holds. By (UP), then, one's perceptual evidence fails to justify one in believing that not-SK. Hence, one does not know that not-SK.
According to one response to this line of thought, experiences justify perceptual beliefs (such as that a cat is near) without providing evidence or reasons for these beliefs because evidence and reasons always come in the form of beliefs which inferentially justify other beliefs. Thus the skeptic cannot appeal to (UP) in the foregoing way. Some philosophers maintain that perceptual Perceptual experiences, some philosophers maintain, justify perceptual beliefs in virtue of having propositional content, although they are not themselves propositions. A visual perception, say, has the representational content expressible by the sentence "A cat is near," and accordingly justifies an associated perceptual belief's having that same content.
One problem for this view is that it is plausible to suppose that nonhuman animals have perceptual experiences with representational contents that are similar to those of humans (given the physiological similarities between the relevant perceptual systems). But the animals' perceptual representations do not possess propositional content. One may reply that experiences nevertheless justify perceptual beliefs by virtue of having nonpropositional representational content, such as that possessed by maps and pictures. This view is, in one way, less attractive than the propositional view, however, because it is easier to see how a belief-like state with propositional content can justify a perceptual belief than to see how a state with a nonpropositional content can perform the same justifying feat.
Further, both views about perceptual justification have the following difficulty. (UP) can be reformulated as:
(UP*) If S's putative justifier for Φ does not favor Φ over a competing incompatible hypothesis Ψ, then S is not justified in believing Φ.
Now the skeptic may hold that one's nonevidential, perceptual putative justifier would be present regardless of whether SK holds or not-SK holds. Thus, one is not justified in believing not-SK, as the skeptic originally claimed.
Against this, it has been held that the perceptual states that one has when not-SK holds differ in their intrinsic nature from those that one has when SK holds. On this view, the veridical perceptual states possessed by a normal perceiver are object-involving, in that objects such as cats are constituents of their perceptual contents. This view might be put forward as a direct realist answer to skepticism, according to which our awareness of external objects is not mediated by awareness of our own experiences. But such direct realism has little antiskeptical force: the skeptic may maintain that even if veridical experience, should it occur, involves direct awareness of cats, it is nevertheless possible that all of one's experiences are unveridical, none possessing an object-involving perceptual content. When the object-involving view is put forward a little differently, however, there is a greater payoff. A disjunctive view challenges the skeptic's use of (UP*). Unlike a veridical perceptual experience of a cat, a nonveridical perceptual state of a brain in a vat is obviously not object-involving. The two states, then, are not tokens of a single perceptual state type; there is no common factor between the states. Because it is not true that the same putative perceptual justifier would be present regardless of whether SK holds or not-SK holds, (UP*) cannot be used to show that one lacks justification for believing not-SK. Thus, on the disjunctivist approach, premise (2) of the skeptical argument is not adequately supported.
One may use considerations from the philosophy of language and the philosophy of mind to argue that SK is in fact false. According to semantic externalism, the Cartesian commits an error in attempting to construct thought experiments involving massive deception. The Cartesian naively assumes that, starting with a subject S of thought and experience who is ensconced in a normal external environment, we may hold fixed the contents of S's thoughts and the meanings of his sentences while varying (in thought) S's external environment in such a way that S's thoughts about his environment come out to be predominantly false. According to the semantic externalist, the Cartesian fails to realize that the contents of one's thoughts and the meanings of one's sentences depend in certain ways on one's external environment.
For example, Donald Davidson argues that, when we interpret a speaker's sentences as expressing various beliefs that he holds, we are constrained to attribute beliefs to him that are by and large true of the environment with which he interacts (Davidson 1986). This is because there is no rational basis for preferring one interpretation that finds him to be massively mistaken in his beliefs over another such interpretation. It is constitutive of beliefs and of sentential meanings that they are what are correctly attributed in correct interpretation, on Davidson's view. Thus, it follows from the nature of belief and meaning that, contrary to what SK states, one can never be so massively mistaken.
To see another manifestation of this anti-Cartesian line of thought, consider Hilary Putnam's Twin Earth, a planet like Earth except for the circumstances that the clear, thirst-quenching liquid that the Twin Earthians call "water" is composed of XYZ molecules rather than H2O molecules. The Twin Earthians' term "water" does not refer to water, but rather to the liquid on Twin Earth with which they interact. Hence, my Twin Earth counterpart's word "water" does not have the same meaning as my word, and when the Twin Earthian says "Water is wet," it is not to thereby express the thought that I think when I think that water is wet. Similarly, the semantic externalist maintains that, when my envatted twin in a treeless world uses the word "tree" in thought, it is not to refer to trees. Instead, the brain in a vat refers to those entities in the external environment that play a causal role with respect to his uses of "tree" analogous to that played by trees with respect to normal uses of "tree" in a tree-filled world. These entities may be states of the computer that systematically cause the brain in a vat to have "tree-like" sense experience. When the brain in the vat thinks the sentence "A tree has fallen," he does not thereby mistakenly express the thought that a tree has fallen. Instead, he expresses a thought about computer states, which may well be true of his environment. In general, then, the brain in a vat is not massively mistaken about the world, contrary to what the Cartesian maintains.
We may use these considerations, together with the assumption that I have knowledge of the contents of my own thoughts, against premise (2) in the following way: I am now thinking that a tree has fallen; if SK is true, then I am not now thinking that a tree has fallen; thus, SK is false. This argument, however, is powerless against versions of the skeptical hypothesis on which the brain in a vat is indirectly causally linked to ordinary objects. If, for example, there are programmers of the computer who refer to trees, then it becomes plausible to suppose that the brain does so as well. Further, there is a prima facie problem as to whether I may claim knowledge of the contents of my own thoughts, given semantic externalism. Such knowledge seems to require independent knowledge of the content-determining causal environment in which I am located, knowledge the antiskeptical argument was meant to provide.
Ambivalence about the Skeptical Argument
Contextualism is a response to skepticism that is based upon a novel view of the semantics of knowledge-attributing sentences of the form "S knows that P." According to the contextualist, such sentences are like sentences of the form "X is flat." The truth value of the latter sort of sentence depends upon both (1) the shape of the pertinent object, and (2) contextually determined standards regarding contour. Relative to one conversational context (in which bicycle racing is under discussion, for instance), "The road is flat" can come out true; relative to another context (where inclined planes are under discussion), the sentence (concerning the same road) can come out false. Similarly, the truth value of an utterance of, say, "John knows that the bank is open this Saturday" depends upon both (1) John's epistemic situation (e.g., his evidential beliefs, his perceptual experience, whether the bank is indeed open), and (2) contextually determined epistemic standards (set by the interests, intentions, and expectations of the knowledge-attributing conversationalists).
Suppose that John's basis for claiming that the bank is open this Saturday is that he visited it on a Saturday a month ago. Suppose that my business partner and I wish to deposit a check this Saturday or some time the following week. Then my partner's utterance of "John knows that the bank is open this Saturday" may well be true, given John's epistemic situation and given the low stakes in our conversational context. Holding John's epistemic situation fixed, imagine a different case in which our business will go bankrupt if the check is not deposited on Saturday. In this case, my partner's utterance of "John knows that the bank is open this Saturday" may well be false, given the higher stakes in this context, in which evidence superior to John's may well be required for knowledge about the bank.
The contextualist claims that his view both (a) explains why the skeptical argument may seem compelling, and (b) implies that there is much ordinarily-attributed knowledge in the world. When skepticism and skeptical possibilities are under discussion, the conversational context is such that abnormally high epistemic standards are in place. Accordingly, an utterance of the argument's premise (2)—"I do not know that not-SK"—comes out true. According to the contextualist, utterances of the argument's closure-based premise (1) are true in all conversational contexts. Thus, relative to a skeptical context, an utterance of the argument's conclusion is true. However, in an ordinary, nonskeptical conversational context, the epistemic standards are lowered, and utterances of premise (2) are false. Thus, knowledge-attributions in ordinary conversational contexts are not threatened by the skeptical argument.
One problem for contextualism is that it is hard to coherently state the view. For example, I cannot now correctly say that Michael Jordan knows that he has hands, since I am currently involved in a skeptical (written) conversational context. What I must instead say is that neither I, nor anybody else, knows that he has hands. I cannot even justifiably say that some ordinary-context utterances of "Michael Jordan knows that he has hands" are true, relative to the low epistemic standards in effect in such contexts. This is because I, in my present context, do not know whether anyone has hands.
Another problem for contextualism is that it seems to imply that speakers are mistaken about the very meanings of their knowledge-attributing sentences. That is, suppose that I think that the skeptical argument is compelling and yet at the same time find its conclusion to be repugnant: it just can't be true that I do not know that I have hands. This means that I am failing to realize that the sentence stating the argument's conclusion is perfectly true as uttered in my current philosophical context. This betrays a misunderstanding of what my sentence means when used in the philosophical context.
Ayer, Alfred Jules. Language, Truth, and Logic. New York: Dover, 1952. Verificationism.
Berkeley, George. Three Dialogues between Hylas and Philonous. Indianapolis: Bobbs-Merrill, 1954.
Brewer, Bill. Perception and Reason. Oxford: Oxford University Press, 1999. Perceptual beliefs.
Brueckner, Anthony. "Semantic Answers to Skepticism." Pacific Philosophical Quarterly 72 (1992): 200–201. Semantic externalism.
Brueckner, Anthony. "The Structure of the Skeptical Argument." Philosophy and Phenomenological Research 54 (1994): 827–835. Underdetermination Principle.
Cohen, Stewart. "How to be a Fallibilist." Philosophical Perspectives 2 (1988): 91–123. Contextualism.
Davidson, Donald. "A Coherence Theory of Truth and Knowledge." In Truth and Interpretation, edited by Ernest LePore. Oxford: Oxford University Press, 1986. Semantic externalism.
Descartes, René. Meditations on First Philosophy. In The Philosophical Works of Descartes. Vol. 1, translated by Haldane and Ross. New York: Dover, 1955. The Cartesian skeptical argument.
Dretske, Fred. "Epistemic Operators." Journal of Philosophy 67 (1970): 1007–1023. Denying the closure principle.
Dummett, Michael. "What Is a Theory of Meaning? (II)." In Truth and Meaning, edited by G. Evans and J. McDowell. Oxford: Oxford University Press, 1976. Antirealism.
Goldman, Alvin. "What Is Justified Belief?" In Justification and Knowledge, edited by G. Pappas. Dordrecht, Holland: Kluwer, 1979. Reliabilism.
Huemer, Michael. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield, 2001. Perceptual beliefs; direct realism.
Kant, Immanuel. Critique of Pure Reason. Translated by N. K. Smith. New York: St. Martin's Press, 1965. Transcendental arguments.
McDowell, John. "Singular Thought and the Extent of Inner Space." In Subject, Thought and Context, edited by P. Pettit and J. McDowell. Oxford: Oxford University Press, 1986. Disjunctive view.
Nozick, Robert. Philosophical Explanations. Chap. 3. Cambridge, MA: Belknap Press of Harvard University Press, 1981). Denying the closure principle; tracking condition.
Putnam, Hilary. "Meaning and Reference." Journal of Philosophy 70 (1973): 699–711. Semantic externalism.
Putnam, Hilary. Reason, Truth, and History. Chap. 1. Cambridge, MA: 1981. Semantic externalism.
Schiffer, Stephen. "Contextualist Solutions to Skepticism." Proceedings of the Aristotelian Society 96 (1996): 317–333. Contextualism.
Stroud, Barry. "Transcendental Arguments." Journal of Philosophy 65 (1968): 241–256. Transcendental arguments.
Vogel, Jonathan. "Cartesian Skepticism and Inference to the Best Explanation." Journal of Philosophy 87 (1990): 658–666.
Anthony Brueckner (1996, 2005)