In this paper we survey some main arguments for and against epistemological contextualism. We distinguish and discuss various kinds of contextualism, such as attributer contextualism (the most influential version of which is semantic, conversational, or radical contextualism); indexicalism; proto-contextualism; Wittgensteinian contextualism; subject, inferential, or issue contextualism; epistemic contextualism; and virtue contextualism. Starting with a sketch of Dretskes Relevant Alternatives Theory and Nozicks Tracking Account of Knowledge, we reconstruct the history of various forms of contextualism and the ways contextualists try to handle some notorious epistemological quandaries, especially skepticism and the lottery paradox. Then we outline the most important problems that contextualist theories face, and give overviews of their criticisms and defenses as developed in this issue.

Externalism about knowledge commits one to a modest form of contextualism: whether one knows depends (or may depend) on circumstances (context) of which one has no knowledge. Such modest contextualism requires the rejection of the KK Principle (If S knows that P, then S knows that S knows that P) - something most people would want to reject anyway - but it does not require (though it is compatible with) a rejection of closure. Radical contextualism, on the other hand, goes a step farther and relativizes knowledge not just to the circumstances of the knower, but to the circumstances of the person attributing knowledge. I reject this more radical form of contextualism and suggest that it confuses (or that it can, at least, be avoided by carefully distinguishing) the relativity in what S is said to know from the relativity in whether S knows what S is said to know.

Skeptics try to persuade us of our ignorance with arguments like the following: 1. I dont know that I am not a handless brain-in-a-vat [BIV]. 2. If I dont know that I am not a handless BIV, then I dont know that I have hands. Therefore, 3. I dont know that I have hands. The BIV argument is valid, its premises are intuitively compelling, and yet, its conclusion strikes us as absurd. Something has to go, but what? Contextualists contend that an adequate solution to the skeptical problem must: (i) retain epistemic closure, (ii) explain the intuitive force of skeptical arguments by explaining why their premises initially seem so compelling, and (iii) account for the truth of our commonsense judgment that we do possess lots of ordinary knowledge. Contextualists maintain that the key to such a solution is recognizing that the semantic standards for knows vary from context to context such that in skeptical contexts the skeptics premises are true and so is her conclusion; but in ordinary contexts, her conclusion is false and so is her first premise. Despite its initial attractiveness, the contextualist solution comes at a significant cost, for contextualism has many counterintuitive results. After presenting the contextualist solution, I identify a number of these costs. I then offer a noncontextualist solution that meets the adequacy constraint identified above, while avoiding the costs associated with contextualism. Hence, one of the principal reasons offered for adopting a contextualist theory of knowledge–its supposedly unique ability to adequately resolve the skeptical problem – is undermined.

Mylan Engel’s paper (2004) is divided into two parts: a negative part, criticizing the costs of contextualism and a constructive part proposing a noncontextualist resolution of the skeptical problem. I will only address the constructive part here. The constructive part is composed of three elements: (i) a reconstruction or reformulation of the original skeptical argument, which draws on the notion of epistemic possibility (e-possibility), (ii) a distinction between two senses of knowledge (and two corresponding kinds of e-possibility): fallibilistic and infallibilistic, and (iii) an argument which tries to hoist the skeptic by their own petard, namely the closure principle (CP). As I will argue, there are two ways to understand Engel’s anti-skeptical argument. Only in one interpretation does the argument depend on the proposed reconstruction of the skeptical argument in terms of e-possibility. But this version of the argument is unsound. More importantly, the skeptic has a strong prima facie objection at her disposal, which applies to both interpretations of the argument. If this objection is valid, Engels argument does not hold. But once it is invalidated, his argument is superfluous.

Contextualists often argue from examples where it seems true to say in one context that a person knows something but not true to say that in another context where skeptical hypotheses have been introduced. The skeptical hypotheses can be moderate, simply mentioning what might be the case or raising questions about what a person is certain of, or radical, where scenarios about demon worlds, brains in vats, The Matrix, etc., are introduced. It is argued that the introduction of these skeptical hypotheses leads people to fallaciously infer that it is no longer true to say that the relevant person knows. The author believes that that is a better explanation of the so-called intuition that the person does not know than the contextualists who claim that raising these skeptical hypotheses changes the standards that determine when it is true to say S knows that P. At the end he raises the possibility that contextualists might defend their view on pragmatic rather than skeptical grounds by arguing that the standards of evidence rise when more is at stake in a practical sense.

David Lewis, Stewart Cohen, and Keith DeRose have proposed that sentences of the form S knows P are indexical, and therefore differ in truth value from one context to another. On their indexical contextualism, the truth value of S knows P is determined by whether S meets the epistemic standards of the speakers context. I will not be concerned with relational forms of contextualism, according to which the truth value of S knows P is determined by the standards of the subject Ss context, regardless of the standards applying to the speaker making the knowledge claim. Relational contextualism is a form of normative relativism. Indexical contextualism is a semantic theory. When the subject is the speaker, as when S is the first person pronoun I, the two forms of contextualism coincide. But otherwise, they diverge. I critically examine the principal arguments for indexicalism, detail linguistic evidence against it, and suggest a pragmatic alternative.

Conversational contextualism states that the truth-conditions expressed by knowledge-attributing sentences vary relative to the context of utterance. This context is determined partly by different standards the person involved must meet in order to make the sentence true. I am concerned with the question of how these standards can be raised or lowered, and especially what happens to the standards and the conversational score when parties in a discussion push the conversational scores in different directions. None of the available options for an answer seems satisfying. I argue that this results from a misunderstanding of the characteristics of the situation at hand.

In this paper I will discuss Michael Williams’s inferential contextualism – a position that must be carefully distinguished from the currently more fashionable attributer contextualism. I will argue that Williams’s contextualism is not stable, though it avoids some of the shortcomings of simple inferential contextualism. In particular, his criticism of epistemological realism cannot be supported on the basis of his own account. I will also argue that we need not give up epistemological realism in order to provide a successful diagnosis of scepticism.

Any contextualist approach to knowledge has to provide a plausible definition of the concept of context and spell out the mechanisms of context changes. Since it is the dynamics of context change that carry the main weight of the contextualist position, not every mechanism will be capable of filling that role. In particular, I argue that one class of mechanisms that is most popularly held to account for context changes, namely those that arise out of shifts of conversational parameters in discourses involving knowledge claims, are not suited to the job because they cannot account for the genuinely epistemic nature of the context shift. A form of epistemic contextualism that defines the context through the structure of our epistemic projects is suggested. Context changes in this account are linked to changes in the background assumptions operative in our epistemic projects and the methods used to carry out our inquiries.

According to Antonia Barke’s version of contextualism, epistemic contextualism, a context is defined by a method and its associated assumptions. The subject has to make the assumption that the method is adequate or reliable and that good working conditions hold in order to arrive at knowledge by employing the method. I will criticize Barke’s claim that epistemic contextualism can provide a more satisfactory explanation or motivation for context shifts than conversational contextualism (in particular, David Lewis’s contextualism). Two more points of criticism will be presented, which are meant to show that epistemic contextualism presupposes epistemic internalism, and that (epistemic) contextualism leads to an implausible view about which parameters the special achievement that is constitutive of knowledge depends on. I suggest that, contra (epistemic) contextualism, knowledge is a more robust phenomenon that does not depend on whether anyone calls into question any assumptions or raises skeptical doubts in conversation or in his or her mind (as, for example, Fred Dretske’s account says). I indicate how this can be reconciled with the phenomenon that knowledge attributions are somewhat unstable and seemingly context-dependent.

A number of virtue epistemologists endorse the following thesis: Knowledge is true belief resulting from intellectual virtue, where Ss true belief results from intellectual virtue just in case S believes the truth because S is intellectually virtuous. This thesis commits one to a sort of contextualism about knowledge attributions. This is because, in general, sentences of the form X occurred because Y occurred require a contextualist treatment. This sort of contextualism is contrasted with more familiar versions. It is argued that the position: (a) yields a better solution to the lottery problem, and (b) may be grounded in a more general theory of virtue and credit.

John Greco has proposed a new sort of contextualism which exhibits a principled grounding in an agent reliabilist virtue epistemology. In this paper I will discuss Greco’s two main reasons in favor of virtue contextualism. The first reason is that his account of knowledge can be derived from a more general theory of virtue and credit. The second reason consists in the thesis that a virtue contextualist solution to the lottery problem is superior to standards contextualism. With regard to the first claim, I raise some questions concerning the status and the content of the crucial conditions for Greco’s theory of intellectual credit. With regard to the second claim, I try to show that his arguments do not succeed in establishing the superiority of virtue contextualism to standards contextualism. I close with some remarks on the relation among Greco’s virtue contextualism, the traditional approach to the theory of knowledge and the proper domain of contextualism.

There are many ordinary propositions we think we know. Almost every ordinary proposition entails some lottery proposition which we think we do not know but to which we assign a high probability of being true (for instance:I will never be a multi-millionaire entails I will not win this lottery). How is this possible – given that some closure principle is true? This problem, also known as the Lottery puzzle, has recently provoked a lot of discussion. In this paper I discuss one of the most promising answers to the problem: Stewart Cohen’s contextualist solution, which is based on ideas about the salience of chances of error. After presenting some objections to it I sketch an alternative solution which is still contextualist in spirit.

Do we need defeasible generalizations in epistemology, generalizations that are genuinely explanatory yet ineliminably exception-laden? Do we need them to endow our epistemology with a substantial explanatory structure? Mark Lance and Margaret Little argue for the claim that we do. I will argue that we can just as well do without them – at least in epistemology. So in the paper, I am trying to very briefly sketch an alternative contextualist picture. More specifically, the claim will be that although an epistemic contextualist should commit himself to epistemic holism he can nevertheless appeal to epistemic principles other than defeasible generalizations in order to provide his epistemology with a structure.

Standard contextualism about know comes in two versions. On the first, know is treated as an indexical. On the second, it is treated both as an indexical and as a context-shifter. On both versions, it is intended as a satisfactory theory about the ordinary meaning of know. But any satisfactory semantic theory about know must be able to account for the fact that know is a factive verb. In this paper, it is argued that standard contextualists face an insolvable dilemma : they had better opt for the second version, but doing so would deprive know of its factivity. It is concluded that standard contextualism is unable to deal with the factivity of knowledge and that for this reason, it cannot constitute a proper treatment of the ordinary meaning of know.