¹ For objections along these lines, see Feldman, Richard, “Skeptical Problems, Contextualist Solutions,” Philosophical Studies, 103 (2001): 61–85CrossRefGoogle Scholar; Hawthorne, John, Knowledge and Lotteries (New York: Oxford University Press, 2004)Google Scholar; and Stanley, Jason, “On the Linguistic Basis for Contextualism,” Philosophical Studies, 119 (2004): 119–46CrossRefGoogle Scholar. For replies, see Cohen, Stewart, “Contextualism, Skepticism, and the Structure of Reasons,” Philosophical Perspectives, 13 (1999): 57–89Google Scholar; Cohen, Stewart, “Contextualism Defended: Comments on Richard Feldman's ‘Skeptical Problems, Contextualist Solutions,’” Philosophical Studies, 103 (2001): 87–98CrossRefGoogle Scholar; and DeRose, Keith, “The Ordinary Language Basis for Contextualism and the New Invariantism,” Philosophical Quarterly, 55 (2005): 172–98.CrossRefGoogle Scholar

² Views along these lines include views that employ a contextually determined set of relevant alternatives. Proponents include Keith DeRose, David Lewis, and Mark Heller. The primary target in what follows is DeRose.

³ See DeRose, Keith, “Solving the Skeptical Problem,” Philosophical Review, 104 (1995): 1–52, at p. 36.CrossRefGoogle Scholar

⁴ Nozick, Robert, Philosophical Explanations (Cambridge, MA: Harvard University Press, 1981), pp. 172–76.Google Scholar

⁵ I assume that the conversants are cooperating and that none of them are protesting the effect that mentioning the BIV hypothesis has on raising the standards.

⁶ According to DeRose, conditionals like (2) are “true regardless of how high or low the standards for knowledge are set” and to deny such a conditional is to embrace an “abominable conjunction.” See DeRose, , “Solving the Skeptical Problem,” pp. 27–29.Google Scholar

⁷ This is a first approximation. Some more clauses must be added to deal with odd cases. A better approximation is: If (i) S knows that p, (ii) S knows that if p then q, (iii) S forms the belief that q on the basis of her beliefs that p and if p then q, and (iv) S is generally competent at modus ponens reasoning, then S knows that q.

⁸ We have just such a case. George tracks the truth of the claim that he has hands, since he believes that he has hands in all the nearest possible worlds where he has hands, and he does not believe that he has hands in the nearest possible world where he does not have hands (e.g., a world where he recently had an accident). However, he does not track the fact that he is not a BIV, since in the nearest possible world where he is a BIV he believes that he is not a BIV.

⁹ Notice, though, that a situation in which the target belief falls short of the knowledge mark is not sufficient for being an effective sceptical scenario, for a situation that is just baldly described as one in which the subject fails to know is not effective. But this should not lead us back to the falsity account since a situation in which the target belief is false need not be effective either, for, similarly, a situation that is baldly described as one in which the subject's belief is false is not effective either. I develop a complete account in “Effective Skeptical Scenarios” (ms.). For other discussions of the ignorance account, see Hetherington, Stephen, “Gettieristic Scepticism,” Australasian Journal of Philosophy, 74 (1996): 83–97CrossRefGoogle Scholar, and Winters, Barbara, “Skeptical Counterpossibilities,” Pacific Philosophical Quarterly, 62 (1981): 30–38CrossRefGoogle Scholar. DeRose himself is open to the ignorance account, noting that the target belief's being false on a scenario is not necessary (and, he says, only “usually sufficient”) for that scenario to be effective (“Solving the Skeptical Problem,” p. 18).Google Scholar

¹⁰ Epistemic internalists might claim that beliefs based on such experiences are justified; but they should still agree that they are not warranted and therefore do not count as knowledge.

¹¹ It is at this point that the move from the falsity account and closure to the ignorance account and SLP is important. The falsity account will not do, since it delivers the wrong result that the BIV hypothesis is not well suited to target the three-objects belief. It delivers this result because on the BIV hypothesis, there exists at least George's brain, a vat, and a computer, enough to make the three-objects belief true. In addition, this is where the move from closure to SLP is important, since SLP, not closure, gets us from the claim that on the BIV hypothesis George fails to know that he is not a BIV to the claim that George does not know that there are at least three objects.

¹² In §10 of “Solving the Skeptical Problem,” DeRose argues that there is another thing that makes the sceptic's utterance of (2) true. As he puts it, this is the fact that George is in at least as good an epistemic position to know that he is not a BIV as he is to know that he has hands. So too, though, George is in at least as good a position to know that he is not a BIV as he is to know that there are at least three objects. For reductio, assume that this is not so. Assume, that is, that George is in a better position to know that there are at least three objects than he is to know that he is not a BIV. This claim, along with DeRose's claim that George is in at least as good a position to know that he is not a BIV as he is to know that he has hands, entails that George is in a better position to know that there are at least three objects than he is to know that he has hands. But this last claim is not independently plausible, for, as we have seen, the BIV hypothesis can be as effectively (or ineffectively) used to cast doubt on whether George knows that there are at least three objects as it can to cast doubt on whether George knows that he has hands. In addition, a person who knows that there are at least three objects knows this on the basis of perception. But if perception equips one to know that there are at least three objects, it also equips one to know that one has hands. So, if one has what it takes to perceive, and thereby know that there are three objects, one surely has what it takes to perceive, and thereby know, that one has hands.

¹³ An anonymous referee suggested that George might think differently about the similarity ordering. From George's point of view, a world where he is a BIV is radically different from the actual world, even more so than a world where he is not envatted but in which there are less than three objects. This is contrasted with the third-person point of view, from which George's being a BIV is not so radically dissimilar from the actual world. The suggestion has it that “I know that there are at least three objects” is true when uttered by George, but “George knows that there are at least three objects” is false when uttered by someone else. This is not independently plausible since it means that if a contextualist were to assess George's self-ascription (which, of course, is made in George's context) and then make her own assessment, upon hearing George utter “I know that there are at least three objects,” she would say something true if she were to say “what George just said is true, but he doesn't know that there are at least three objects.” The latter is clearly false.

¹⁴ One might include among the tenets of sensitivity-contextualism not only the contextualist thesis and the sensitivity thesis, but also a claim about what features of sceptical hypotheses make them well suited to target beliefs. If all of this is packed into sensitivity-contextualism, then this third tenet, as I have argued, had better endorse the ignorance account. DeRose seems not to have written a claim of this third kind right into sensitivity-contextualism.

¹⁵ One might protest that this is unfair. Instead, we should think of the view as contending that the difficulty in knowing a proposition is determined by two things: the proposition's modal distance and the resources of the relevant epistemic creature. On this view, how difficult it is to know a given proposition is partly determined by the kind of epistemic creature that one is. This is surely plausible, but it is not enough to handle cases of this last kind, for here the two propositions are equally modally distant, yet specifying fixed epistemic resources for the creature—for example, perception and counting—would imply that the two propositions are equally difficult to know using the specified resource. But this is not so.

¹⁶ Versions of this article were presented at the 2000 Meeting of the Canadian Philosophical Association and the 2003 Eastern Division Meeting of the American Philosophical Association. For useful comments and discussion, thanks to Bryan Belknap, Tim Black, Albert Casullo, Phil Hanson, Jim Stone, two especially helpful referees from this journal, and the editor, Eric Dayton.