Acta Analytica

, Volume 28, Issue 1, pp 111–125

Against the Minimalistic Reading of Epistemic Contextualism: A Reply to Wolfgang Freitag


    • Department of PhilosophyNorthern Illinois University

DOI: 10.1007/s12136-012-0153-5

Cite this article as:
Ashfield, M.D. Acta Anal (2013) 28: 111. doi:10.1007/s12136-012-0153-5


Several philosophers have argued that the factivity of knowledge poses a problem for epistemic contextualism (EC), which they have construed as a knowability problem. On a proposed minimalistic reading of EC’s commitments, Wolfgang Freitag argues that factivity yields no knowability problem for EC. I begin by explaining how factivity is thought to generate a contradiction out of paradigmatic contextualist cases on a certain reading of EC’s commitments. This reductio results in some kind of reflexivity problem for the contextualist when it comes to knowing her theory: either a knowability problem or a statability problem. Next, I set forth Freitag’s minimalistic reading of EC and explain how it avoids the reductio, the knowability problem and the statability problem. I argue that despite successfully evading these problems, Freitag’s minimalistic reading saddles EC with several other serious problems and should be rejected. I conclude by offering my own resolution to the problems.


Epistemic contextualismFactivityReflexivity problemKnowability problemStatability problemEven-handedness

1 The Factivity Problem for Epistemic Contextualism (EC)

Factivity is the very plausible principle that knowledge of a proposition p, in any context, implies the truth of p. Indeed, truth is usually taken to be a necessary condition of knowledge, and Peter Baumann declares it “weird if not crazy to deny the factivity of knowledge. Whatever knowledge is, it is factive. Nothing is a concept of knowledge in a broad sense if what it is a concept of isn’t factive.”1 So consider a contextualist version of factivity, letting Ξ be the set of all possible contexts, Σ, the set of all possible epistemic subjects, Τ, the set of all possible times, Πe, the set of all contingent empirical propositions, and ‘Kx(S, t, p)’ express the sentence, “S knows in context x at time t that p”:2
$$ {\text{Factivit}}{{\text{y}}_{\text{c}}}:\forall x \in Ξ, \forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:{{\text{K}}_x}\left( {S,t,p} \right) \to p. $$
How does factivity create a problem for EC? Let l be a low-standards context, and h be a high standards context, let S1 and S2 be two epistemic subjects, let t be some time, and let hands stand for the proposition S1 has hands.3 Consider the following:
  1. (1)

    Kl(S1, t, hands).

  2. (2)

    ¬Kh(S2, t, hands).

  3. (3)
    Kh(S2, t, (1)); or equivalently Kh(S2, t, Kl[S1, t, hands]).
    • (1) can be used to yield the following instance of Factivityc:

  4. (4)

    Kl(S1, t, hands) → hands.

Since there is no reason to suppose one cannot know that knowledge is factive in a high standards context, assume S2 knows (4) in her high standards context:
  1. (5)

    Kh(S2, t, [Kl(S1, t, hands) → hands]).

But now consider a contextualist version of the very plausible principle of epistemic closure, a principle to which contextualists are already committed:4
$$ \begin{array}{*{20}c} {{\text{Closure}}_{{\text{c}}} :\forall x \in \Xi ,\forall S \in \sum ,\forall t \in T,\forall p \in \Pi _{{\text{e}}} :{\left[ {{\text{K}}_{x} {\left( {S,t,p} \right)}\& {\text{K}}_{x} {\left( {S,t,{\left[ {p \to q} \right]}} \right)}} \right]}} \\ { \to {\text{K}}_{x} {\left( {S,t,q} \right)}.} \\ \end{array} $$
Closurec maintains that knowledge is closed under known entailment, given that the context of knowledge ascription remains constant.5 Substituting (1) for p and (4) for p → q, we get the following instance of Closurec:
  1. (6)

    (Kh(S2, t, Kl[S1, t, hands]) & Kh[S2, t, (Kl[S1, t, hands] → hands)]) → Kh(S2, t, hands).

By adjunction of (3) and (5), and modus ponens from (6), we derive:
  1. (7)

    Kh(S2, t, hands).

But (7) contradicts (2), so it looks like the factivity of knowledge yields a reductio of EC.

While (1) and (2) jointly constitute a paradigm case for EC, Anthony Brueckner and Christopher T. Buford argue that (3) should be rejected, because EC is not committed to the possibility of such asymmetrical knowledge ascriptions where one epistemic subject, S2, lacks knowledge of a proposition, but can correctly ascribe knowledge of that proposition to another epistemic subject, S1.6 Indeed, if S2 knows that knowledge is factive (as we assumed earlier in order to derive (5)), then she knows that truth is a necessary condition of knowledge and should therefore refrain from ascribing knowledge to S1 of a proposition that she does not herself know to be true in her high standards context. I think this is right, but Baumann, Elke Brendel, Crispin Wright,7 and others seem to disagree, so let’s see what the problem is supposed to be.

2 The Fluid View, the Contextualist’s Knowledge of EC, and the Problem of Reflexivity

In “Elusive Knowledge,” David Lewis posits a rule of attention that seems to rule out the possibility of resisting a rise in the semantic standards of ‘knows’ once a skeptical error-possibility is mentioned,8 resulting in what Nathan Howard calls the fluid view of epistemic context shift.9 On the fluid view, shifts in epistemic context are induced by conversational features, such that the mere mention of an error-possibility makes it salient in a context of ascription and drives up the semantic standards of ‘knows.’10 In addition, adjustments in semantic standards are asymmetrical: given our tendency to allow higher standards to trump lower standards and the time it takes to forget that standards have been raised, it’s easier to raise standards than to lower them.11 Since Lewis thinks that epistemological contexts are necessarily demanding contexts, because “when we do epistemology…we make knowledge vanish,” he anticipates a potential “problem of reflexivity” for the contextualist when it comes to knowing her own theory.12

3 The Knowability Problem

Assuming the fluid view, if we also assume that the contextualist knows EC is true in virtue of paradigmatic examples like (1) and (2)13 (call this the knowledge assumption), and that thinking about examples like (2) always involves thinking about an error-possibility (call this the error assumption), then there’s reason to think that the contextualist is committed to (3) as well. For every time the contextualist thinks about her knowledge of EC, including examples like (1) and (2), she thinks of some error-possibility in relation to (2) and the mere thought of it drives up the semantic standards of ‘know.’14 She thereby finds herself in a high-standards context in virtue of this error-possibility, and because of the asymmetry of standards adjustments she cannot immediately escape this high-standards context. If she still knows the theory of EC, including (1) and (2), in her high-standards context, then she’s in exactly the kind of position in relation to (1) that’s described by (3), that is Kh(S, t, (1)). Given Factivityc and Closurec, with (1), (2), and (3) in play we have everything we need for the reductio. Knowing the theory of EC, therefore, makes the theory inconsistent with itself. No doubt, some would rather say that, in her high-standards context, the contextualist does not know EC is true. But either way, it looks like EC is unknowable on the fluid view: this is the knowability problem.

4 The Statability Problem

The problem of reflexivity has been construed slightly differently by Brueckner and Buford. On their construal, the contextualist never finds herself in the kind of position described by (3). If, for example, she finds herself in a very high-standards skeptical context in virtue of the error-possibility salient to (2)—say, a brain-in-a-vat hypothesis—then she’s in no position to agree that there are such things as epistemic subjects, in which case the contextualist is in no position to make an asymmetrical knowledge ascription like (3).15

Brendel has shown that this result can be generalized over ∀x, y ∈ Ξ, where the set of error-possibilities salient in x is not identical to the set of error-possibilities salient in y. Suppose our contextualist is telling her familiar contextualist story about some epistemic subject S who visits a zoo. Let n be the context of a normal zoo visit, let m be the context of a zoo visit where it is a salient possibility that the black-and-white striped equines in the zebra enclosure are cleverly disguised mules, and let zebra express the proposition These animals are zebras. As the contextualist explains, S knows in context n that These animals are zebras is true, which yields the following example of (1): Kn(S, t, zebra). But she explains that because of the salient error-possibility, S does not know that These animals are zebras is true in context m, which yields the following example of (2): ¬Km(S, t, zebra). Our contextualist is now well on her way to explaining her theory as a whole; she explains the effects of mere mention and asymmetrical standards adjustments and is just about to make generalizations about what epistemic subjects can and cannot know in different contexts. But notice that, on the fluid view, once she mentions the cleverly disguised mule error-possibility in relation to (2), she finds herself in context m. Consequently, she cannot knowledgeably state anything that relies on the truth of These animals are zebras including the example of (1). She cannot correctly assert anything about low-standards knowledge indexed to context n while she’s in the higher-standards context m, which prevents her from using this paradigmatic case to make the generalizations necessary for articulating EC as a whole. The contextualist will encounter the same problem with all contextualist cases. Thus, “even much less remote skeptical worries than brains-in-vats can deprive us from knowing the main claim of contextualism.”16

As Baumann explains, “What the contextualist could not do (in her more demanding context)…is to concede that [examples like (1) are] true…The contextualist would have to retreat to something like “It is possible that [(1)].””17 So a contextualist can have low-standards knowledge of examples like (1), but, on the fluid view, she cannot correctly assert that knowledge in the higher-standards context of the EC theory as a whole. She has what Mylan Engel Jr. calls unspeakable knowledge of EC.18 In the epistemological context of EC, the contextualist can only suggest the mere possibility of (1), so she cannot knowledgeably state her theory as a whole: this is the statability problem. Brueckner and Buford bite this unstatability bullet, maintaining that a true theory need not be knowledgeably statable.19

5 The Minimalistic Reading of EC

In an effort to clarify what is actually at stake in the debate, Wolfgang Freitag characterizes EC as a response to global skepticism (GS), which maintains that the standards for knowledge are so high that it is impossible (or nearly impossible) to have any knowledge of empirical, contingent propositions.20 He argues that EC’s response to GS relies solely on its rejection of invariantism – this is EC’s defining feature. Freitag defines GS as follows:
$$ {\text{GS}}:\forall x \in Ξ, \forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:{ }\neg \diamondsuit {{\text{K}}_x}\left( {S,t,p} \right). $$
GS maintains that it is impossible for anyone in any context at anytime to know anything. GS follows from three assumptions: Invariantism, High Standard, and Skepticismh:
  1. 1.

    Invariantism: \( \forall x,y \in Ξ, \forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:\,□ \left[ {{{\text{K}}_x}\left( {S,t,p} \right) \leftrightarrow {{\text{K}}_y}\left( {S,t,p} \right)} \right] \).

Invariantism says that necessarily, whatever can be known in one context can be known in any context.
  1. 2.

    High Standard: Ξh ≠ Ø;

The High Standard assumption claims that there exists at least one high-standard context of knowledge ascription.
  1. 3.

    Skepticismh: \( \forall x \in {Ξ_h},\forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:\neg \diamondsuit {{\text{K}}_x}\left( {S,t,p} \right) \);

Skepticismh is the claim that in any high standards context, it is impossible for anyone at anytime to know anything. Jointly, these three assumptions entail GS. The standard pre-contextualist response to this triad was to deny the High Standard assumption, but EC denies the assumption of Invariantism.21 Very high standards operate in special skeptical contexts (Ξh), while lower, more relaxed standards are at work in ordinary, not-so-demanding contexts (Ξl). This non-invariantist claim, Freitag contends, is what makes EC theories distinctively contextualist. He calls this defining commitment the compatibility thesis, which is just the negation of Invariantism:
$$ {\text{Compatibility}}:\neg \left( {\forall x,y \in Ξ, \forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:\,□ \left[ {{{\text{K}}_x}\left( {S,t,p} \right) \leftrightarrow {{\text{K}}_y}\left( {S,t,p} \right)} \right]} \right). $$
Compatibility states that it is not the case that necessarily whatever can be known in one context can be known in any context. So it is possible that one might not know something in one context but still know it in another context, which blocks the derivation of GS from the High Standard and Skepticismh assumptions. Freitag thinks that EC is first and foremost a response to GS, and it is Compatibility that provides “the distinctively contextualist answer to the problem of skepticism.”22 Given Freitag’s stated assumption that Ξ = Ξh ∪ Ξl, Compatibility also entails:
$$ {\text{Low \;Standard: }}{Ξ_l} \ne \emptyset . $$
Low Standard says that there exists at least one low standards context of knowledge ascription. As Freitag summarizes, given the compatibility thesis, “the impossibility of knowledgeh does not conflict with the possibility of knowledgel. In this way, scepticism can, in a sense, be reconciled with our ordinary understanding of knowledge ascriptions as being true at least sometimes.”23

6 The Problems of Reflexivity Revisited

So far, we’ve seen that while (1) and (2) jointly constitute a paradigm case for EC, given Factivityc and Closurec, they generate a contradiction when they are jointly known. Factivityc and/or Closurec can only be jettisoned at the risk of wild implausibility. So proponents of the fluid view of EC, who take examples like (1) and (2), mere mention, asymmetrical standards adjustments, and the assumption that epistemological contexts are necessarily demanding contexts, to be essential commitments of EC, face some kind of reflexivity problem. Either EC as a whole cannot be known, so EC suffers from a knowability problem, or EC cannot be knowledgeably stated, in which case EC suffers from a statability problem.

Freitag’s minimalistic reading of EC’s commitments offers a retreat, dissolving the factivity-generated reductio and any resulting reflexivity problem by avoiding most of the commitments of the fluid view of epistemic context shift, including mere mention, asymmetrical standards adjustments and the assumption that epistemological contexts are necessarily demanding contexts. The contextualist can thereby retain Factivityc and Closurec, while enshrining the Compatibility thesis as EC’s distinguishing feature. Thus, Freitag declares that, “Compatibility is all that a contextualist has to claim in order to block the argument to skepticism.”24 So, consider Freitag’s analysis that EC has a knowability problem (KP) if and only if:25
$$ {\text{KP}}:\exists p \in \Pi _{{\text{e}}} :{\text{EC}} \to {\left[ {p\& \neg {\text{K}}_{x} {\left( {S,t,p} \right)}} \right]}. $$
EC has a knowability problem if and only if the theory of EC entails the truth of some empirical proposition p and also entails that p cannot be known in the context of EC. That is, EC has a knowability problem if and only if it commits the contextualist to the truth of a Moore-paradoxical statement of the form “p, and I don’t know that p”.26 But as Freitag argues, there is no such empirical proposition to which EC is committed (in this adapted quotation, I modify Freitag’s wording to conform to my style of presentation, but no substantial changes have been made):

Baumann claims to be in possession of an empirical truth to which EC is committed, namely, hands. If EC indeed entailed hands, EC would be surprisingly empirical. Not only would the truth of EC be dependent on the most contingent vagaries of life, but it would also open up unexpected paths for falsifying EC: just chop off S’s hands (before t)! Baumann holds that the contextualist is committed to hands because it follows, via Factivityc, from (1). But EC is committed to (1) even less than to hands. I would expect a normal subject to knowl she has hands, provided, of course, she has any. But such knowledgel is no part of the contextualist theory…Indeed, EC may be true in a world with only handless epistemic subjects. So (1) is not part of EC either.27

And a little earlier he says (again, with some merely stylistic changes):

If p is an empirical proposition, the example of (2) concerning p follows from Skepticismh for Skepticismh is usually conceded by contextualists, but it is worth emphasizing that it is no part of EC as such. EC does not entail Skepticismh, though EC may be irrelevant for epistemology if Skepticismh is false…A contextualist may consistently reject Skepticismh, and she may even do so without compromising the anti-skeptic force of her position. However, if Skepticismh is denied, Compatibility is not needed to undermine the skeptical argument [for GS] and EC is a lot less attractive. The concession to skepticism in the form of Skepticismh is therefore a vital part of the interest of EC, but not part of its content. The attribution of a commitment to (2) to EC is unwarranted.28

Since EC entails the truth of no particular empirical proposition p while also entailing that p cannot be known in the context of EC, KP comes out false,29 so EC has no knowability problem.

It would be surprising to find a contextualist who did not believe that many low-standards empirical knowledge ascriptions are true, or that high-standards empirical knowledge denials are true. But as Freitag explains, “to motivate their positions, contextualists often describe situations in which, according to their view, the epistemic agent knowsl some proposition p*, although she does not knowh it. Such a view does not, however, entail the unconditional commitment towards knowledgel of proposition p*.”30 This kind of explanation strikes Baumann as a reluctance to take one’s own contextualism seriously and apply it to concrete cases,31 but he is mistaken. The contextualist’s belief that a low-standards empirical knowledge ascription is true in a concrete case does not entail that EC is committed to the existence of a truth-maker in the external world for that low-standards knowledge ascription. It simply means that, in her low-standards context, the contextualist believes such a truth-maker exists. Likewise, the belief that a high-standards empirical knowledge denial, with respect to a given empirical proposition, is true in a concrete case does not entail that EC is committed to the non-existence of a truth-maker for the empirical proposition in question. It simply means that, in her high-standards context, the contextualist is not sufficiently confident of its existence.

Without commitments to specific instances of (1) and (2), mere mention, asymmetrical standards adjustments, or the assumption that epistemological contexts are necessarily demanding contexts, (3) is also avoided by the minimalistic reading, and without (1), (2) or (3), the factivity problem dissolves. So the contextualist has no problem of reflexivity: neither a knowability problem nor a statability problem. He concludes that “a contextualism that restricts itself to its distinctive anti-sceptic aims does not have the knowability problem.”32

7 Against the Minimalistic Reading of EC

As elegant as Freitag’s solution may appear, problems remain for the minimalistic reading of EC. Freitag thinks contextualists can be content with the assertion that Invariantism is false. But all that assertion actually boils down to is the mere possibility that some epistemic subject, at some time, may know some empirical proposition in some context, and not know that proposition in another context; that is all the distinctively contextualist thesis actually maintains:33
$$ {\text{Compatibility}}*:\exists x,y \in Ξ, \exists S \in Σ, \exists t \in {\rm T},\exists p \in {Π_{\text{e}}}:\diamondsuit \neg \left[ {{{\text{K}}_x}\left( {S,t,p} \right) \leftrightarrow {{\text{K}}_y}\left( {S,t,p} \right)} \right]. $$
Indeed, the minimalistic reading of EC leaves the theory very thin, and this is even more apparent in Freitag’s alternate formulation of EC:34
$$ {\text{EC}}*:\forall x \in Ξ, \forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:\diamondsuit \left[ {{{\text{K}}_l}\left( {S,t,p} \right) \& \neg {{\text{K}}_h}\left( {S,t,p} \right)} \right]. $$

So to be clear about what it is Freitag is proposing, the minimalistic reading of EC represents a retreat to the mere possibility that the contextualist story accurately describes the world; the minimalistic reading of EC leaves the contextualist story a mere possibility rather than an unknowable or unstatable reality. But, the fact that EC can survive only as a mere possibility remains a threat to its plausibility for several reasons.

Firstly, for contextualists who accept the B modal system, it’s not even clear that the minimalistic reading of EC can be maintained.35 Suppose that in every possible world, the contextualist story (ECS) is necessarily true if it is true at all:
  1. 1.

    □(ECS → □ECS) (Premise).

  2. 2.

    ◊ECS → ◊□ECS (1, Possibility Distribution).

  3. 3.

    ◊□ECS → ECS (B Theorem).

  4. 4.

    ◊ECS → ECS (2,3, hypothetical syllogism).

  5. 5.

    ◊ECS (Freitag’s assertion).

  6. 6.

    ECS (4,5, modus ponens).

Given Premise 1, on the B modal system, Freitag’s minimalistic reading of EC entails not merely the possibility of the contextualist story, but the truth of the contextualist story, so the contextualist who accepts B is committed to something stronger than Freitag’s minimalistic reading. Freitag might deny Premise 1, arguing that the contextualist story is only contingently true—true in virtue of contingent facts about how “knows” works. However, since Freitag thinks Invariantism would be a necessary truth, if it’s true at all,
$$ {\text{Invariantism}}:\forall x,y \in Ξ, \forall S \in Σ, \forall t \in {\rm T},\forall p \in {Π_{\text{e}}}:\,□ \left[ {{{\text{K}}_x}\left( {S,t,p} \right) \leftrightarrow {{\text{K}}_y}\left( {S,t,p} \right)} \right], $$
his own construal of the Invariantism assumption militates against such a response. So it looks like the minimalistic reading can be maintained only by rejecting B, a cost many philosophers will find unappealing. Freitag admits that contextualists might want to claim more than his minimalistic reading requires,36 but it turns out that, unless they reject the B modal system, they’re actually committed to a stronger claim.

Secondly, although Freitag’s minimalistic reading of EC technically avoids the knowability and statability problems as described earlier, the contextualist story still suffers from a knowability and statability problem, of sorts, on this view. Freitag’s clever move is to define the theory of EC as a thin claim about the mere possibility that the contextualist story is true, but notice that one still cannot know that the contextualist story is actually true: call this the lesser knowability problem.37 Under these conditions, contextualists who take knowledge to be the norm of assertion will find the contextualist story unassertable: call this the lesser statability problem. Admittedly, this situation is an improvement over the previous knowability and statability problems, but it still leaves EC with little appeal as an epistemic theory.

Thirdly, though it is possible that the Invariantism assumption is false, it seems equally possible that the Invariantism assumption is true, and that the contextualist story does not accurately describe the world. The minimalistic reading of EC offers no assurances against this possibility. Moreover, if we were to find good independent evidence that Invariantism were true (e.g., Brendel’s case against the indexicality of “knows”),38 an invariantist of a certain kind, like a neo-Moorean or a Nozickean, could be in a position to consistently know his theory, provided other theoretical commitments satisfied the conditions for knowledge. The minimalistic reading of EC does not afford the same luxury to the contextualist. Even if the contextualist were furnished with similar evidence that Invariantism is false, she still could not consistently know the contextualist story to be true. In this respect, EC is most similar to strong versions of skepticism, since strong versions of global skepticism entail that nothing, including the theoretical claims of global skepticism, can be known.39

Fourthly, and most devastatingly, even if Invariantism is false, there may still be no context of knowledge ascription where standards are low enough for epistemic subjects like us to succeed in having knowledge, or making true knowledge ascriptions. Though Freitag assumes that Ξ = Ξh ∪ Ξl, it may be the case that Ξ = Ξh ∪ Ξmh, where Ξmh is the set of moderately high-standards contexts, wherein it is still nearly impossible for any thinker like us to know anything at any time. Being as thin as it is, the minimalistic reading of EC offers no assurances against this form of global skepticism: “contextualism as such is silent about (1) the standards for knowledge operative in different contexts and (2) whether these standards are, or can be, fulfilled by epistemic subjects.”40 So I wonder as Baumann does, “What is the attraction of contextualism if one cannot (at least as a contextualist) coherently say or think that knowledge attributions made in a lower context are in fact true? Only that [it] might be possible? The kind of contextualism that results would be a very much weakened one and not very attractive.”41 In fact, Freitag’s minimalistic reading weakens EC so much that it no longer fulfills what he identifies as its distinctive anti-skeptical aim. For even though the Compatibility thesis blocks the derivation of GS from the High Standard and Skepticismh assumptions, we’ve just seen that it still leaves the door open to global skepticism.

Finally, even if the minimalistic reading fails in its purported anti-skeptical aim, Crispin Wright states that EC’s principal attraction lies in its even-handed approach to the debate between skepticism and folk epistemology, in trying to make space for both in their own respective ways.42 Freitag similarly claims that on his minimalistic reading of EC, “scepticism can, in a sense, be reconciled with our ordinary understanding of knowledge ascriptions as being true at least sometimes.”43 Yet, the minimalistic reading cannot deliver on this promise of even-handedness. The minimalistic reading’s similarity to strong versions of global skepticism in their unknowability, its leaving the door open to global skepticism, and its lack of any commitment to low-standards knowledge in concrete cases all point to a decided leaning away from folk epistemology in favor of skepticism. Not only does the minimalistic reading fail to fulfill its distinctive anti-skeptical aims, it also fails to be even-handed.

For all these reasons, I think Freitag’s minimalistic reading of EC deserves to be rejected. Unless the contextualist can articulate her theory in such a way that solves the factivity-generated reductio of EC and the accompanying knowability problem, without retreating to an unstatable version of the theory or the mere possibility that the contextualist story is true, we are best advised to reject EC altogether. It seems that we must have either a robust version of contextualism, or none at all.

8 A Better Response to the Apparent Problems of Reflexivity

I think that the fluid view of EC is implausible. In particular, its commitment to mere mention seems indefensible. There is plenty of epistemological evidence that mere mention is incorrect,44 and contextualists like Keith DeRose and Stewart Cohen, have noted the linguistic evidence against it as well. For example, Cohen argues that,

Though skeptical considerations frequently lead to a strong upward pressure on the standards, the shift to a skeptical context is not inevitable. The pressure toward higher standards can sometimes be resisted. One device for doing this is adopting a certain tone of voice. So in response to the skeptic, one might say, “C’mon, you’ve got to be kidding – I know I am not a brain-in-a-vat!”. If this is the dominant response among the conversational participants, then everyday standards may remain in effect. In such a case, the speaker unmoved by skeptical doubt is not failing to adjust his ascriptions to contextually determined standards. Rather, such a speaker is managing to keep the standards from rising.45

Exercising veto power over an error-possibility in this way seems conversationally appropriate, and such linguistic behavior flies in the face of mere mention. Thus, even if the fluid view should prove to be unsalvageable from its apparent problems of reflexivity, the contextualist can always abandon it in favor of a more plausible version of EC.46 As such, the fluid view merely represents a worst-case-scenario for the contextualist.

However, even the fluid view can resist the problems of reflexivity discussed earlier, because its problems of reflexivity obtain only if both the knowledge assumption and the error assumption are true. That is, the contextualist only has trouble knowing her theory if we assume both that she knows EC is true in virtue of paradigmatic examples like (1) and (2), and that thinking about examples like (2) always involves thinking about an error-possibility. The knowledge assumption looks plausible to me, so I attack the error assumption.

To begin, it is not clear that thinking about high-standards knowledge denials necessarily involves thinking about error possibilities. It is even less clear that thinking about the theory of EC in general necessarily involves thinking about error possibilities, so I think the error assumption is on shaky ground. But let’s assume that the knowledge assumption is right, and the contextualist only knows EC is true in virtue of paradigmatic examples like (1) and (2). Let’s also assume that our contextualist is meditating on a familiar paradigmatic example involving an epistemic subject S who has hands and a brain-in-a-vat hypothesis. As the story usually goes, the contextualist recalls to herself that (1) in S’s low-standards context S is able to know S has hands (i.e., Kl(S, t1, hands)), but once a brain-in-a-vat hypothesis becomes a salient error-possibility for S, (2) S does not know S has hands in her now very high-standards skeptical context (i.e., ¬Kh(S, t2, hands)). But an error-possibility’s salience for S and its salience for the contextualist are two entirely separate matters. To see why, consider the linguistic distinction between use and mention.

I think we can distinguish between S’s use of an error-possibility in (2), and the contextualist’s mention of S’s situation in meditating on her theory. (Things get a bit confusing here because the jargon of ‘mere mention’ is a misnomer; it is actually the mere use of an error-possibility that has been supposed to invoke high-standards into a conversation context of knowledge ascription, not its mere mention.47) On such a distinction, the contextualist is not subject to the effects of mere use, because she never uses an error-possibility in the operative sense. To illustrate the point, as you read this, you know that you have eyes. You know that you’re reading a philosophical paper with those eyes. You know that the paper you’re reading is written in English.48 And you know all of these things, despite the fact that I’ve mentioned the brain-in-a-vat hypothesis and made reference to skeptical error-possibilities throughout my argument. Moreover, you’ve had no need to exercise veto power over these skeptical error-possibilities (if you accept that such a move is possible), because I have not attempted to use them. What all of this shows is that even in deliberately meditating upon the effects of such a very high-standards skeptical error-possibility like the brain-in-a-vat hypothesis, the contextualist is able to jointly know things like (1) and (2) and fulfill the requirement of the knowledge assumption, without ever thinking of an error-possibility in the operative sense. So it looks like the error assumption is false, and without the error assumption the fluid view faces no problems of reflexivity.

Now one might object that by introducing the distinction between mere use and mere mention, I’ve already betrayed the spirit of the fluid view. One might insist, for example, that the assumption that epistemological contexts are necessarily demanding contexts means that the contextualist’s mention of an error-possibility in her story leads inevitably, if inadvertently, to its use in her epistemological context. This doesn’t seem right to me, but no matter. One more response can be offered on behalf of all versions of EC, including the fluid view.

As I said near the beginning, I agree with Brueckner and Buford, who argue that (3) should be rejected, because EC is not committed to the possibility of asymmetrical knowledge ascriptions where one epistemic subject, S2, lacks knowledge of a proposition but can correctly ascribe knowledge of that proposition to another epistemic subject, S1. Again, if S2 knows that knowledge is factive, then she knows that truth is a necessary condition of knowledge and should therefore refrain from ascribing knowledge to S1 of a proposition that she does not herself know to be true in her high standards context. But this need not result in a general knowability or statability problem for EC. The incompatibility of (2) and (3) implies that EC is not knowable or knowledgeably statable in high-standards contexts, but EC says that in such contexts, one cannot truly claim to know much of anything. So why should EC itself be knowable in a skeptical context? EC’s response to the skeptic is to say that this fact doesn’t render our ordinary low-standards knowledge claims false. Nothing about this response depends upon our being able to know or knowledgeably state that EC is true in high-standards contexts. To think otherwise is to misunderstand the spirit of the contextualist response to skeptical arguments, as Freitag does when he mischaracterizes EC as having a distinctively anti-skeptical aim. Crispin Wright’s talk of even-handedness seems much closer to the spirit of EC. Indeed, the fact that EC is knowable in low-standards contexts but unknowable in high-standards contexts is exactly what we should expect from the course plotted by David Lewis in “Elusive Knowledge,” whereby EC passes alarmingly close to both the rock of fallibilism and the whirlpool of skepticism.49 EC does not give the contextualist a way to prove skeptics wrong once she is on their skeptical turf. It gives her a way to acknowledge their right to that turf, while enabling her to see why she is under no rational obligation to join them there.50


Baumann 2008, p. 584. See Wright 2005, p. 242. Cf. Blamey 2008, Stjernberg 2009, and Hazlett 2010.


Freitag takes “S knows in context x at time t that p” to be synonymous with and less cumbersome than “‘S knows that p’ is true in context x at time t.”


Following standard notation, I use corner quotes (e.g., and ) to denote propositions and quotation marks to denote sentences.


E.g., Cohen 1988, p. 105; DeRose 1995, footnote 33; Lewis 1996, p. 564.


For ease of presentation, Freitag supposes that the contextualist accepts a stronger principle, which I will call, \( {\text{Closure}}_{{\text{F}}} :\forall x \in \Xi ,\forall S \in \sum ,\forall t \in T,\forall p \in \Pi _{e} :{\left( {{\left[ {{\text{K}}_{x} {\left( {S,t,p} \right)}\& {\left( {p \to q} \right)}} \right]} \to {\text{K}}_{x} {\left[ {S,t,q} \right]}} \right)}. \)

ClosureF implausibly maintains that knowledge is closed under entailment simpliciter, and no contextualist would accept it. However, Freitag is in the clear, because given Factivityc, ClosureF implies Closurec:

1. Kx (S, t, p) & Kx(S, t, [p → q]) (Assumption).

2. Kx (S, t, p) (from (1) by Simplification).

3. Kx(S, t, [p → q]) (from (1) by Simplification).

4. p → q (from (3) by Factivityc).

5. Kx (S, t, q) (from (2) & (4) by ClosureF).

6. Closurec: [Kx(S, t, p) & Kx(S, t, [p → q])] → Kx(S, t, q) (from (5) by elimination of Assumption).

So anything that can be derived using only Factivityc and ClosureF, can be derived using Factivityc and Closurec. See Freitag 2011, p. 278. My expanded presentation of the factivity-generated reductio simply avoids the implausible ClosureF principle altogether. My thanks to Wolfgang Freitag for explaining how this derivation is accomplished.


Brueckner & Buford 2009, p. 434; 2010, p. 487. Although (3) says nothing about ascribing knowledge, I follow Brueckner & Buford in assuming that knowing an epistemic subject S knows p involves mentally ascribing knowledge of p to S. See also Mylan Engel’s discussion of unspeakable and unthinkable knowledge, where he assumes that the conversational and semantic features of EC apply to one’s mental life: Engel 2004, p. 212.


See Baumann 2008 and 2010; Brendel 2005 and 2009; Wright 2005.


Lewis 1996, pp. 559–561. See Cohen 2001, endnote 10.


I benefitted from the presentation and discussion of Howard’s paper, “Wright’s Factivity Objection and DeRose’s Ordinary Language Basis for Epistemic Contextualism,” delivered at the Fifth Annual Midwest Regional Graduate Conference, DeKalb, IL, October 21, 2011. This “fluid view” is contrasted against a “static view” where epistemic contexts do not shift so easily.


Engel 2004, p. 208; Brendel 2005, p. 41.


Engel 2004, p. 208.


Lewis 1996, p. 566.


Freitag refers to (1) as the ‘truth commitment’ and (2) as the ‘no-knowledgeX commitment’ respectively. See Freitag 2011, p. 281–282.


This assumes that the conversational and semantic features of EC apply to one’s mental life. See footnote 6.


Brendel finds it unclear how a remote error-possibility contextually salient to (2), like the brain-in-vat hypothesis, could be an error-possibility for EC, but this objection makes sense of such a case. See Brendel 2005, p. 50.


Brendel 2005, p. 50.


Baumann 2008, p. 583.


Engel 2004, p. 212.


Brueckner & Buford 2009, p. 437.


Freitag 2011, 4.


For example, Keith DeRose asserts that, contrary to Invariantism, semantic standards vary context-to-context: “According to contextual analysis, when the sceptic presents her arguments, she manipulates various conversational mechanisms that raise the semantic standards for knowledge, and thereby creates a context in which she can truly say that we know nothing or very little. But the fact that the sceptic can thus install very high standards which we don’t live up to has no tendency to show that we don’t satisfy the more relaxed standards that are in place in ordinary conversations.” DeRose 1992, p. 917, my emphasis. See also DeRose 1995, pp. 4–5.


Freitag 2011, p. 277.


Freitag 2011, p. 278.


Freitag 2011, p. 281, my emphasis.


This analysis, which Freitag prosaically names “Condition,” is accompanied by an extensive exposition and defense, which I will not rehearse here. See Freitag 2011, pp. 279–281.


See Williamson 2001, pp. 26–27; Wright 2005, footnote 8; Baumann 2008, pp. 587, 595–598.


See Freitag 2011, p. 282.


See Freitag 2011, p. 281.


Freitag says that its falsehood can be intuitively demonstrated by noticing that Compatibility is consistent with

Omniscience: ∀x ∈ Ξ, ∀S ∈ Σ, ∀t ∈ Τ, ∀p ∈ Πe: [p → Kx(S, t, p)] (Freitag 2011, footnote 20).

It’s worth noting that it would be no virtue of EC for it to be consistent with this insane-sounding thesis that all true propositions are known in all contexts by all epistemic subjects at all times. See Williamson 2002, p. 271. However, Freitag’s alternate formulation of EC is actually inconsistent with this Omniscience thesis:

EC*: ∀S ∈ Σ, ∀t ∈ Τ, ∀p ∈ Πe: ◊[Kl(S, t, p) & ¬Kh(S, t, p)] (Freitag 2011, footnote 21).

Assume the possibility that subject S knows in low-standards context l, but not in high-standards context h, at time t that r.
  1. 1.

    ◊[Kl(S, t, r) & ¬Kh(S, t, r)] (Assumption).

  2. 2.

    Kl(S, t, r) & ¬Kh(S, t, r) (Assumption).

  3. 3.

    Kl(S, t, r) (from 3 by Simplification).

  4. 4.

    ¬Kh(S, t, r) (from 3 by Simplification).

  5. 5.

    Kl(S, t, r) → r (instance of FactivityC).

  6. 6.

    r (from 2,5 by modus ponens).

  7. 7.

    r & ¬Kh(S, t, r) (from 4,6 by Adjunction and discharging Assumptions).


So, in some possible world, there is some true proposition r that is not known in some high-standards context h, at some time t, by some subject S. In this possible world where the contextualist story is true, Omniscience is false.


Freitag 2011, p. 282.


Baumann 2008, p. 583.


Freitag 2011, p. 281.


This Compatibility* thesis is logically equivalent to Freitag’s Compatibility thesis.


Freitag 2011, footnote 21.


I owe this observation to Stephen Maitzen. Cf. DeRose 1991.


Freitag 2011, p. 281.


At best the minimalistic contextualist can know the contextualist story is true in some possible world. But, of course, to know it to be true in that possible world is to make it inconsistent and unknowable, or unstatable in that world.


See Brendel 2009, pp. 407–408.


See Freitag 2011, footnote 19.


Freitag 2011, p. 281.


Baumann 2008, p. 583.


Wright 2005, pp. 240–241. See Freitag 2011, footnote 2.


Freitag 2011, p. 278, my emphasis.


E.g., Engel 2004, pp. 210–213.


Cohen 2001, p. 93. See DeRose 2000 and 2004.


One should treat Freitag’s minimalistic reading as a cautionary tale: in restricting EC’s commitments to avoid the problems of reflexivity, the contextualist must guard against sweeping away too much.


My thanks to Brandon Williams for pointing this out.


Engel 2004, p. 204.


Lewis 1996, pp. 550, 566.


Much of the substance and phrasing of this final paragraph is owed to comments from Geoff Pynn, delivered at the Central States Philosophical Conference, St. Louis, MO, September 16, 2011.



I wish to express my appreciation to Mylan Engel Jr., Louis Gularte, Steve Maitzen, Geoff Pynn, Wes Skolits, Brandon Williams, and an anonymous referee for comments on earlier versions of this paper. I also want to thank Peter van Elswyk for encouraging me to look into the factivity problem.

Copyright information

© Springer Science+Business Media B.V. 2012