ARTICULOS

Contextualism and Testimony

Contextualismo y testimonio

Leandro De Brasi

Departamento de Filosofía
Universidad Alberto Hurtado
ldebrasi@uahurtado.cl

---

Abstract

It has recently been argued that Subject-Sensitive Invariantism conflicts with an attractive testimonial principle of transmission. One might think, given the differences between Subject-Sensitive Invariantism and Contextualism, the latter would not be susceptible to related objections. I argue however that some testimonial principles with much intuitive appeal give us a related though different reason to reject Contextualism. In short, while some argue that Subject-Sensitive Invariantism seems to sometimes make testimonial knowledge too easy to come by, I argue that Contextualism seems to often make testimonial knowledge too hard to come by. Contextualism then does not enjoy an advantage over Subject-Sensitive Invariantism when it comes to capturing the transmissive role of knowledge.

KEY WORDS: Epistemic Contextualism; Subject-Sensitive Invariantism; Testimonial Knowledge; Defeaters.

Resumen

Recientemente se ha argumentado que el Invariantismo Sensible está en conflicto con un principio testimonial de transmisión atractivo. Uno podría pensar que, dadas las diferencias entre el Invariantismo Sensible y el Contextualismo, este último no sería susceptible a objeciones relacionadas. Sin embargo argumento que algunos principios testimoniales de gran atractivo intuitivo nos dan una razón, relacionada pero diferente, para rechazar al Contextualismo. En pocas palabras, mientras que algunos argumentan que el Invariantismo Sensible parece, a veces, hacer que el conocimiento testimonial sea demasiado fácil de obtener, argumento que el Contextualismo parece, a menudo, hacer que el conocimiento testimonial sea demasiado difícil de obtener. El Contextualismo entonces no goza de una ventaja sobre el Invariantismo Sensible cuando se trata de capturar el rol transmisor del conocimiento.

PALABRAS CLAVE: Contextualismo epistémico; Invariantismo Sensible; Conocimiento Testimonial; Canceladores.

It has recently been argued that Subject-Sensitive Invariantism conflicts with some commonly accepted maxims about the social role of knowledge (Schaffer 2006) and in particular our testimonial practice (MacFarlane 2005). One might think, given the differences between Subject-Sensitive Invariantism and Contextualism, the latter would not be susceptible to related objections. I argue, however, that some testimonial views with much intuitive appeal give us a related though different reason not to prefer Contextualism. More specifically, any plausible principle concerning the transmission of knowledge via testimony and a plausible norm of testimonial assertion seem to give us such a reason. In short, while some argue that Subject-Sensitive Invariantism seems to make testimonial knowledge too easy to come by, I argue that Contextualism seems to make testimonial knowledge too hard to come by. If this is correct, Contextualism does not enjoy an advantage over Subject-Sensitive Invariantism in this regard, as it is sometimes thought (Schaffer 2006).
The paper proceeds as follows. I first introduce some plausible testimonial principles that will help us set the issues (§1) and present the main elements of the Contextualist and Invariantist approaches (§2). Given this background, I say enough to give you a sense of an alleged problem that some have raised for Subject-Sensitive Invariantism and to see why it does not apply to Contextualism (§3), before considering the case against Contextualism (§4). Finally, I offer some closing remarks (§5).

1. The Transmissibility of Knowledge

Let me introduce some prima facie plausible testimonial views that we will exploit later on when making the case against Contextualism. First, there is a genuine phenomenon concerning the transmission of knowledge by testimony and, whatever principle this transmission obeys, such principle will involve a no-defeater condition:

[N-D] Knowledge is transmitted from speaker to hearer only if the hearer has no undefeated defeaters for the testimony (and this is not the result of negligence or incompetence).

It is important to note that the no-defeater condition is meant to deal with cases where the hearer has or should have a reason to doubt that knowledge is being transmitted, say due to incompetence or insincerity of the speaker, and that any plausible principle concerning the transmission of knowledge via testimony will include some such condition.¹
Also, although there might not be a constitutive knowledge norm of assertion in general (Maitra 2011), there seems to be a regulative knowledge norm of testimonial assertion:

[A] Testify that p only if you know that p.

This norm, which seems to regulate our testimonial practice, concerns only what is necessary to testify felicitously and can explain the usual challenges, excuses and recriminations to testimonial assertions. For example, it can explain why answering some testimonial inquiry when not knowing leaves one open to criticism unless one qualifies the answer.²

2. Contextualism and Subject-Sensitive Invariantism

Before considering the problems Subject-Sensitive Invariantism (hereafter SSI) and Contextualism seem to face, let me explain the main differences between these approaches.
Contextualism is the view that the epistemic standards that a subject must meet in order for a 'knowledge' attribution to be true vary according to the conversational context of the attributor.³ These contextually determined standards for how strong the subject's epistemic position with respect to p must be can vary depending on features like the awareness of error-possibilities and practical stakes of the attributors (Cohen 1999; DeRose 1999). So, 'know' picks out different epistemic relations relative to different attributor contexts.
However, as an Invariantist account, SSI holds that the same epistemic relation is picked out by 'know' in all conversational contexts. But, SSI claims that this unique knowledge relation is sensitive to the awareness of error-possibilities, practical stakes and so on of the subject (at the time). So, what is required to satisfy the one relation depends partly on the sort of non-traditional factors that Contextualism exploits (Hawthorne 2004; Stanley 2005).⁴
So, while SSI agrees with Classical Invariantism that 'know' picks out, pace Contextualism, a unique knowledge relation, SSI agrees with Contextualism that those non-traditional factors, such as the awareness of error-possibilities, practical stakes and the like, have, pace Classical Invariantism, a role to play in it.⁵

3. Trouble for SSI

Having introduced the key components to the approaches, let me briefly turn to the alleged trouble for SSI.⁶ John MacFarlane (2005) and Jonathan Schaffer (2006) seem to make a telling point against SSI if one accepts the following principle:

[T] Knowledge is transmitted from speaker to hearer if the hearer accepts that p on the basis of the speaker's sincere testimony that p, the speaker knows that p and the hearer has no undefeated defeaters for the testimony.

This is an attractive principle and it is sometimes claimed that it, or something very much like it, is widely held (MacFarlane 2005, p. 133), not least because much knowledge seems to be transmitted from one person to another in this manner.
Assuming [T] , SSI seems to run into trouble when there is a mismatch of practical stakes and the like. more specifically, when the speaker's stakes are low and the hearer's stakes are high.⁷ Suppose that speaker S is in a low-stakes situation with respect to p, which enables her to know that p on her evidence. And suppose that hearer H is in the same evidential position with respect to p as S is, but does not know that p because she is in a high-stakes situation. Then, given that testimony transmits knowledge in accordance with [T], H should be able to acquire knowledge that p from S's testimony. After all, S knows that p and, crucially, H must regard S as knowing that p: H cannot deny that S knows that p, given SSI. However, it is counter-intuitive that H could gain knowledge this way: H has the same evidence for p as S has and this evidence was, ex-hypothesi, not strong enough for H to know.
In other words, MacFarlane and Schaffer point out that the audience can gain high-stakes knowledge off the speaker's low-stakes evidence, as long as there are no defeaters for its acceptance. But this, Schaffer says, is "epistemic cheating" (2006, p. 97). And it is certainly counter-intuitive that the audience could gain knowledge this way, since:

[someone] subject to high epistemic standards should not be able to escape their stringent demands by getting her knowledge second-hand from someone in more forgiving circumstances, whose epistemic position may be no better than her own! (MacFarlane 2005, p. 134).

So it seems that either SSI or [T] has to go and if one finds the principle attractive, one has at least an all-other-things-being-equal reason for giving SSI up. But, of course, not everyone finds [T] attractive and some are willing to reject it (e.g. Lackey 2008). So there is room for the proponent of SSI to argue that her position is compatible with a properly formulated principle of the transmission of knowledge via testimony.
Nevertheless, even if we hold [T], given that SSI's particular commitment to subject sensitivity is not shared by Contextualism (which is instead committed to some sort of attributor sensitivity), one might think Contextualism is not susceptible to related problems and so hold that it is able to provide a socially fitting conception of knowledge (Schaffer 2006). With regard to the above case, once [T] is suitably contextualized by making reference to epistemic standards, we can see that the above SSI-specific issue does not arise:

[C-T] Knowledge-_standarda is transmitted from S to H if H accepts that p on the basis of S's sincere testimony that p, S knows-_standarda that p and H has no undefeated defeaters for the testimony.

Since at least S does not count as knowing given H's high standards, H does not gain high-standards knowledge according to Contextualized Transmission. So Contextualism can avoid the above epistemic cheating charge even if one accepts [C-T]. However, as we will now see, the problem for Contextualism is that, given this possible mismatch of standards, a defeater becomes available that ends up threatening the transmissive role of testimony given [N-D].⁸

4. Trouble for Contextualism

In order to bring the problem to the surface, the first thing to notice is that Contextualism has the consequence that one can say "S knows that p" and someone else can say "S does not know that p" (about the same S and p at the same time) and both say something true (Cohen 2000, p. 94; DeRose 2009, p. 3). This consequence is, of course, intended as a virtue of Contextualism, since it can then easily accommodate cases such as the well-known Airport and Bank examples (Cohen 1999, p. 58; DeRose 1992, p. 913).
In the Airport example, Mary and John want to know whether the flight stops at Chicago, because they need to make an important business contact at the airport and overhear someone ask a fellow passenger, Smith, if he knows whether this is so. After checking his itinerary, he replies "Yes, I know -it does stop in Chicago", but Mary, who is in a highstakes situation, wonders whether he really knows, since the itinerary could contain a misprint or they could have changed the schedule at the last minute. So Mary tells John "Smith does not know the plane stops in Chicago" and decides to check with the airline agent.⁹
This is a good illustration of the variability of knowledge attribution that Contextualism is meant to handle. As Cohen says,

When in response to a normal query, Smith consult his flight itinerary and says, "I know the plane stops in Chicago", intuitively, what he says is true. [We] readily allow that we can come to know things on the basis of written information contained in things like flight itineraries. But when we consider the situation of [...] Mary, intuitively, [she speaks] the truth when [she says], "Smith does not know the plane stops in Chicago" (2008, p. 422).

So, from the fact that someone says correctly "Smith knows that the plane stops in Chicago" from one point of view, it does not follow that such utterance is also true from another point of view.
Focusing on Mary and how things are from her perspective, she is an eavesdropper or interceptor of information who is in a more demanding attributor context than Smith, and so she cannot agree with Smith, given her standards, that he 'knows'. Mary's case helps us appreciate that if Contextualism is correct then, given a knowledge norm of testimonial assertion [A], a defeater regarding the strength of the testimony offered becomes available in the transmission of knowledge. Let me explain.
Supposing Contextualism true, a contextualized knowledge norm of testimonial assertion requires the speaker to know-a that p when testifying that p, where a is the speaker's own standards [C-A_speaker].¹⁰ Also, assuming Contextualism, it is clearly possible that the standards that regulate the speaker's assertion are less demanding than the hearer's standards, just as it happens in the case of Mary and Smith. Moreover, when this sort of mismatch of standards occurs, the speaker can and is likely to exploit ways of 'knowing' -that is, 'knowledge'-yielding methods- that only serve standards lower than the hearer's, just as Smith does.¹¹
In this sort of case, the hearer should not accept the testimony since, given the hearer's standards, it is likely not to be sufficiently strong. Indeed, the hearer can easily find herself in an epistemic position that does not guard against a range of epistemic failures that she takes to be relevant and important when accepting this testimony. In other words, the speaker can easily not count as being competent according to the hearer's standards given the mismatch.
So there is a defeater regarding the strength of the testimony offered. In fact, even in cases where the speaker correctly takes himself to know that p given his own standards and sincerely tells the hearer that p, if the hearer's attributor context is stricter than the speaker's, then the hearer seems to have a reason for not accepting p since, from the hearer's perspective, this testimony is likely not to be sufficiently strong.
Of course, the Contextualist might welcome this defeater. They might say that in cases where there is a mismatch, knowledge cannot be so easily transmitted. And given any plausible principle of the transmission of knowledge via testimony, which will include a no-defeater condition [N-D], it seems that knowledge will not be easily transmitted in these cases. That, one might think, is exactly the Contextualist's point and, assuming Contextualism, that seems just right. So the Contextualist might be fine with the idea that there is such a defeater in those cases of mismatch. However, the problem for Contextualism is that it seems that the defeater should be defeated in all testimonial exchanges. let me explain.
I take it that this sort of scenario where there is a mismatch of standards is common enough given the Contextualist framework. After all, the standards vary depending on things like the attributor's practical stakes and awareness of possibilities of error. These factors can and usually do vary from person to person, as Airport illustrates. Indeed, the ability to capture this variability in ordinary knowledge attribution is meant to be part of the attraction of Contextualism and in fact its main motivation (DeRose 1999, 2002, 2009).¹² So, assuming Contextualism, cases where there is some such mismatch are the kind of situations that one might ordinarily find oneself in.¹³
Given this sort of case can easily enough arise within the Contextualist framework, adopting a default stance that does not require us to eliminate this possibility when accepting testimony does not seem an option. This is to say, it seems that we cannot simply assume that a knowledge-yielding method that serves only a lower standard is not being employed by the speaker unless we happen to pick out some reason for thinking otherwise. Instead, it seems that the possibility that the hearer is not drawing information from a speaker whose knowledge-yielding method is sufficiently strong given the hearer's standards ought to be ruled out in all cases of testimony. After all, given these mismatches are common enough within Contextualism, it would be epistemically irresponsible not to rule them out in every testimonial exchange.¹⁴, ¹⁵
If this is correct, there are two main ways the Contextualist (while retaining the spirit of the view) can suggest to deal with this testimonydefeating possibility concerning the above incongruence between methods and standards (I-possibility for short), depending on whether she requires the hearer or the speaker to rule it out:
i. One (more natural) strategy, which exploits the above contextualized knowledge norm of assertion [C-A_speaker], requires the speaker to testify according to the standards in place in his own context and the hearer to eliminate the Ipossibility. In other words, the hearer ensures that she accepts the speaker's testimony only if his epistemic position is strong enough for her context.
ii. The other strategy, which involves relativizing the knowledge norm of testimonial assertion to the hearer's context [C-A_hearer],¹⁶ requires the speaker to testify according to the standards in place in the hearer's context, hence the speaker eliminating the Ipossibility. In other words, the speaker ensures that he testifies only if his epistemic position is strong enough for the hearer's context.
Let us consider these strategies.
Suppose the onus is on the hearer to defeat the defeater (i). The problem is that it is not clear that hearers can, most of the time, work out what the speaker's knowledge-yielding method is. After all, hearers are not normally in Mary's fortunate situation, which allows her to be aware of the knowledge-yielding method employed (Mary happens to see Smith checking the itinerary -hence Mary's case does not seem problematic). Although we can sometimes work out what the method is (hence enabling us to determine whether it satisfies our standards), much testimony one is offered is simply information the speaker already possesses: that is, information stored in memory -whose source the speaker might not anyway be aware of, as it is normally the case (and much of which might come from testimony).¹⁷ So it is normally very difficult for the hearer to be able to gather the information to work out whether the knowledgeyielding method exploited by the speaker satisfies her own standards.¹⁸
So, since the hearer needs to eliminate the I-possibility even when the speaker cooperates and testifies according to his own standards (as [C-A_speaker] suggests in this case) and even if there is no mismatch of standards, the above puts pressure on the feasibility of the very pervasive testimonial practice and on the point of having any such practice in the first place. minimally, it threatens to diminish the transmissive powers of testimony.
Importantly, this ruling out of the I-possibility is not just something that would be difficult for us to carry out in many cases, we also do not normally do it (and probably not even committed Contextualists try to do
it). After all, much testimony we normally accept comes from speakers that we do not know much about, including, significantly, their contexts and methods.¹⁹ That is, regardless of whether we can eliminate such possibility, we do not anyway do it.²⁰ So, given (i), a consequence of Contextualism is that we do not normally transmit knowledge through testimony since, even in those cases where there is no such incongruence, we do not (even if we can) eliminate its possibility.
The Contextualist then might opt for the second strategy (ii): to relativize the knowledge norm of testimonial assertion to the hearer's context [C-A_hearer]. The speaker's assertion that p should be regulated by the standards applicable to his audience, hence eliminating the Ipossibility. In this case, the speaker is taking on the burden of producing information that is sufficiently strong given the hearer's context.
But this strategy is also unsuccessful, and for similar reasons to the previous one. After all, just as the hearer does not normally consider this I-possibility, nor do the speakers (again, probably not even committed Contextualists ask their audiences what their stakes are before answering). In fact, speakers often do not know who their audiences are, let alone what their stakes are. And even if they were to consider this Ipossibility, they would face the (often intractable) complications of tracking the methods in memory cases to work out whether they satisfy the hearers' standards.
Importantly, this second option seems counter-intuitive. After all, if the hearer is aware of the fact that the speaker asserts according to his own laxer standards, it seems wrong for her to criticize the speaker. Would Mary, having requested the information directly to Smith, complain to him when he says he knows by checking the itinerary? The recrimination seems out of place: Mary is not entitled to resent Smith. Whether or not finding out the hearer's standards is a realistic target, it is not something we think the speaker ought to do.
So, given the failure of (i) and (ii), it seems that Contextualism sets conditions on testimony that jeopardise the transmission of knowledge and rules out much of what we take to be testimonial knowledge even in cases where there is no incongruence between the method exploited by the speaker and the standards of the audience. This is because there is a clear tension between what Contextualism requires for felicitous testimonial exchanges and the situation we normally find ourselves in with regard to the necessary information to satisfy such requirements: a situation in which the diversity regarding the epistemic standards at play in testimonial exchanges is opaque and hard to determine.
But couldn't the Contextualist attempt to elude this problem altogether by eradicating the I-possibility in testimony? Let me briefly consider one move: to modify the knowledge norm of testimonial assertion so that the speaker testifies according to the highest achievable standard [C-A_{highest-achievable}]. In this case, Contextualism would resemble a fixed-standard account in testimonial exchanges. now, aside from the fact that this might be, to borrow Derose's words from another context, an "ugly and unmotivated amalgamation of contextualism and classical invariantism" (2009, p. 108),²¹ it seems wrong that the appropriate standard for the speaker's self-regulation is the highest achievable one. After all, as Cohen himself notes (see above), we do not think that Smith does anything wrong when, after checking his itinerary, he testifies and we think that Smith's fellow passenger can come to know via Smith's testimony. So it seems that Contextualism would reduce to the wrong kind of fixed-standard account, since such standard would be too demanding.²²
Of course there might be other ways the Contextualist can respond to the above challenge but, as long as these responses are not forthcoming,²³ it seems that Contextualism, given [N-D] and [A], makes
knowledge through testimony too difficult to obtain and so it cannot account for the manifold ways in which we learn from each other. More precisely, the mismatch of standards that Contextualism allows for makes a defeater available in testimonial exchanges by introducing the Ipossibility that ought to be but is not usually ruled out, thus threatening Contextualism's ability to preserve the seemingly widespread phenomenon of testimonial transmission of knowledge.

5. Concluding Remarks

To conclude, while some argue that SSI seems to sometimes make testimonial knowledge too easy to come by, I have argued that Contextualism seems to often make it too hard to come by. If these arguments are correct, neither SSI nor Contextualism seem to, for different reasons, fully capture the transmission of knowledge via testimony.²⁴ This means that Contextualism, contrary to what some think, does not seem to enjoy an advantage over SSI in accommodating the testimonial transmission of knowledge. So Contextualism's scorecard does not look better than SSI's in this regard.²⁵ Indeed, as long as the case against SSI is predicated on a specific transmission principle [T] and inasmuch as [T] is rejected by some, Contextualists, who are willing to accept (and even motivate their view by means of) a knowledge norm of assertion [A], face a stronger challenge since any plausible principle of
transmission will involve a no-defeater condition [N-D].
Nonetheless, regardless of where we stand with respect to SSI's alleged problems, there seems to be a way of arguing from some general and intuitive testimonial principles, as opposed to particular cases, against Contextualism.²⁶ As things stand, Contextualism does not seem to be able to do justice to the widespread phenomenon of testimonial transmission of knowledge. This failure gives us a reason not to prefer Contextualism.²⁷

Notes

1 This minimal requirement for the transmission of testimonial knowledge should be common ground between all parties, whatever epistemology of testimony they happen to favour: reductionist or otherwise. So the below case against Contextualism does not depend on that further and unresolved debate. And psychological and normative defeaters are the two kinds of defeaters standardly taken to be relevant: defeaters that make it epistemically irrational to continue to hold p. For more on these defeaters, see lackey (2008, pp. 44-45).

2 Invoking a knowledge norm might not be the only way to explain the relevant phenomena, but it is a prima facie straightforward and plausible one. There is no space to consider these issues, but notice that at least some Contextualists are sympathetic or committed to some such knowledge norm (e.g. Cohen 2004; DeRose 2009). In fact, DeRose (2002, 2009) argues for Contextualism by means of a stronger knowledge norm (i.e. assert that p iff one knows that p), and that argument is used as a central support for Contextualism. Here, however, we will exploit this weaker and more plausible knowledge norm to argue against Contextualism. So, if this is correct and one buys into a (strong) knowledge norm in order to support Contextualism, one also gets an argument against it.

3 Here I am concerned with the sort of Standards Contextualism championed by Stewart Cohen and Keith DeRose, and we shall be working with a generic version of this Contextualism.

4 I take non-traditional factors to be some non-truth-conducive factors that have not been thought to matter to whether one knows.

5 Classical Invariantism denies that the truth-value of knowledge attributions depends on either the attributor's or the subject's awareness of error-possibilities, practical stakes and the like.

6 It is not my aim here to defend their case against SSI, but merely to present it as a means to introducing the related but different case against Contextualism.

7 For ease of expression, I just talk about stakes. notice, moreover, that what matters here is whether one is in the low/high-stakes situation and so, given Contextualism, a corresponding epistemic standard will apply to the case.

8 Regardless of whether one accepts [T], any plausible principle explaining the widespread transmission of knowledge via testimony will require a no-defeater condition (see §1). In this way, the following criticism leveled against Contextualism does not depend on the correctness of [T].

9 See also Cohen (2000, p. 95).

10 So one is to assert that p only if one meets one's own contextually set standards for knowing that p. This is the most natural relativization to make the norm Contextualism-friendly (DeRose 2009, p. 99); see below for others.

11 These are methods by which we form (and maintain) beliefs that put us in the right epistemic relation to the proposition (given the context). I take it there are methods (say, checking with the airline agent) that can satisfy lower and higher standards (say, Smith's and Mary's, respectively) and, importantly, methods (say, checking the itinerary) that can satisfy lower standards but fail to satisfy higher standards, just as Airport illustrates. I also take it that the former sort of methods tends to be more costly in terms of resources to the speaker than the latter ones. But notice that, sometimes, one might exploit a method that satisfies higher standards in a low-standard context given one's relation to the subject-matter. Consider, for example, the case of experts, who are likely to exploit an expert (recognition) method that can satisfy the hearer's higher-standard, even if those standards do not apply to him. But I take it this sort of situation is not the norm.

12 So not only can the Contextualist accommodate these attributions of knowledge but ought to do so.

13 Or, as one might put it, the possibility that there is such a mismatch obtains in nearby possible worlds.

14 Of course we might think they are not common enough given an egocentric bias by which we evaluate others as though they shared our concerns (cf. Stanley 2005, pp. 99-102). But this is an interfering factor that we should be correcting for, not embracing. Especially given that it seems that we can overcome this bias, since there are thirdperson attributions, as the Contextualist acknowledges, where the subject's concerns set the standards (DeRose 2009, p. 239). For example, when I say "Smith ran to the gate because he knew he would otherwise miss the plane," it would be strange to employ standards not appropriate to the practical situation faced by the subject.

15 Moreover, the Contextualist might suggest that the low-high mismatch situations do not occur that often since the contexts are rather coarse-grained. of course the Contextualist would have to make a case for such contexts, where both speaker and hearer can be in the same coarse-grained context even if their awareness of errorpossibilities, stakes and so are different (and not just slightly different). This is certainly not the way Contextualists present their views, but it is, strictly speaking, open to them to do so. However, the Contextualist needs to be faithful to the data that motivates her view, so if our attributions change in a fine-grained way, as their cases seem to suggest, then this move is a non-starter.

16 [C-A_hearer] Testify that p only if you know-a that p, where a is the hearer's standard. notice, anyway, that the speaker's context still "calls the shots": it is her conversational purposes that call for employing standards that are appropriate to the practical situation of the audience (DeRose 1999, p. 191; 2009, p. 240).

17 I take it that memory preserves knowledge but not normally the grounds for that knowledge. This is of course good news for beings with finite cognitive powers like ourselves, but bad news for the Contextualist since it limits the trans-contextual exploitation of knowledge: we cannot in these cases trace back the standards that a particular item of knowledge satisfies.

18 One might think that certain social roles might help here by setting a specific standard. Most roles however seem to allow the speaker to exploit a range of standards.

19 Notice that I am not here suggesting that there is some general scarcity of information about the speaker in testimonial exchanges. This doesn't seem to be quite right (Kenyon 2013). The suggestion instead is that normally we lack the information to rule out the I-possibility and still accept the testimony.

20 Only in extreme cases, in which the stakes are significantly high, we are likely to try to make the appropriate extra checks to find out the speaker's method. Tellingly, these are also the cases in which we are less likely to rely on testimony.

21 Although by focusing on the idea that knowledge closes the inquiry and given that one can be aware that tougher contexts can deprive us of knowledge we would otherwise enjoy, the Contextualist might have the means to motivate this modification.

22 I have been ignoring a further but related difficulty for Contextualism in this paper that [C-A_{highest-achievable}] makes it a pressing matter: if the speaker's standards are higher than the audience's, there is a significant risk of the audience failing to gain the sought and otherwise attainable (by the audience's standards) knowledge. Given [C-A_{highest-achievable}], the point of having a testimonial practice would be much in doubt. In fact, ideally, we would want to eliminate this (reverse) high-low sort of mismatch (where the standards of the speaker are higher than those of the hearer) too to allow the testimonial practice to be as useful as possible.

23 And I see no promising strategy the Contextualist can exploit. For example, one might think that another response would be to adopt a "single scoreboard" position (DeRose 2009, pp. 128ff), which would allow the Contextualist to suggest that for any conversational context there is one common standard in place that applies to all parties, as opposed to multiple personal standards but no common standard. But it is not clear how a common conversational standard that might not match one's personal standard is meant to address the issue here considered. Consider the case where Smith, with personal standard Low and shared conversational standard High, testifies that p. What matters to Mary is that her informant shares her personally indicated standard since it is this standard that regulates his response. If Smith's personal standard is lower than Mary's, then Smith is likely to exploit methods that serve only standards lower than Mary's and so to provide information which is not sufficiently strong from Mary's point of view.

24 Indeed, their shiftiness seems to be in tension with the idea that there is a social pool of knowledge.

25 That is, Contextualism suffers from an explanatory deficit in this regard just as SSI seems to do. But whether this is a fatal deficit for Contextualism depends on how well it and its competitors do but not merely in comparison with this issue. We need to weigh the costs and benefits for each approach and see which one has the best scorecard. And Contextualism might end up being the victor. It might have the best scorecard, for all I have said here; but not because it can provide us with a socially fitting conception of knowledge (as some suggest): or, more specifically, because it can accommodate the transmission of knowledge.

26 As well as other shifty epistemologies. I believe the worry raised against Contextualism can also be successfully raised against Assessment relativism (MacFarlane 2011), provided the required changes are made. like Contextualism, this approach "holds that the truth of knowledge claims is sensitive to contextual factors, such as [...] how high the stakes are. For the relativist, however, the relevant context is the context from which the knowledge claim is being assessed, not the context at which it was made" (MacFarlane 2011, p. 536). So, in this case, 'know' is sensitive to the context of the assessor but given that both the speaker and the hearer also count as assessors regarding the epistemic strength of the testimony (consider again the case of Smith and Mary, respectively), the possibility of a mismatch is again introduced. After all, a knowledge-attribution might be accurate as assessed from one context but not another, given different standards apply. This assessment-variability then I suggest introduces the same kind of transmissibility problems that face the Contextualist.

27 I would like to thank audiences in the Northern Institute of Philosophy Early Career Conference, the Copenhagen Workshop on Deference, Testimony and Diversity, and the 7th Latin Meeting in Analytic Philosophy at the Institut Jean Nicod, and particularly to Federico Luzzi, Jessica Brown and Franck Lihoreau.

References

1. Cohen, S. (1999), "Contextualism, Skepticism, and the Structure of reasons", Philosophical Perspectives, 13, pp. 57-89.         [ Links ]

2. Cohen, S. (2000), "Contextualism and Skepticism", Philosophical Issues, 10, pp. 94-107.         [ Links ]

3. Cohen, S. (2004), "Knowledge, Assertion, and Practical reasoning", Philosophical Issues, 14, pp. 482-491.         [ Links ]

4. Cohen, S. (2008), "Ascriber Contextualism", in Greco, J. (ed.), The Oxford Handbook of Skepticism, Oxford, Oxford University Press.         [ Links ]

5. DeRose, K. (1992), "Contextualism and Knowledge Attributions", Philosophy and Phenomenological Research, 52, pp. 913-929.         [ Links ]

6. DeRose, K. (1999), "Contextualism: An Explanation and Defense", in Greco, J. and Sosa, E. (eds.), The Blackwell Guide to Epistemology, Oxford, Blackwell.         [ Links ]

7. DeRose, K. (2002), "Assertion, Knowledge, and Context", Philosophical Review, 111, pp. 167-203.         [ Links ]

8. DeRose, K. (2009), The Case for Contextualism, Oxford, Oxford University Press.         [ Links ]

9. Hawthorne, J. (2004), Knowledge and Lotteries, Oxford, Oxford University Press.         [ Links ]

10. Kenyon, T. (2013), "The Informational Richness of Testimonial Contexts", The Philosophical Quarterly, 63, pp. 58-80.         [ Links ]

11. Lackey, J. (2008), Learning from Words, Oxford, Oxford University Press.         [ Links ]

12. MacFarlane, J. (2005), "Knowledge Laundering: Testimony and Sensitive Invariantism", Analysis, 65, pp. 132-138.         [ Links ]

13. MacFarlane, J. (2011), "Relativism and Knowledge Attributions", in Bernecker, S. and Pritchard, D. (eds.), The Routledge Companion to Epistemology, London, Routledge.         [ Links ]

14. Maitra, I. (2011), "Assertion, Norms and Games", in Brown, J. and Cappelen, H. (eds.), Assertion, Oxford, Oxford University Press.         [ Links ]

15. Schaffer, J. (2006), "The Irrelevance of the Subject: Against Subject-Sensitive Invarianstism", Philosophical Studies, 127, pp. 87-107.         [ Links ]

16. Stanley, J. (2005), Knowledge and Practical Interests, Oxford, Oxford University Press.         [ Links ]

Received: April 29, 2013; accepted: May 7, 2014.