Published in The Southern Journal of Philosophy, Summer 2004, 42:2, pp. 261-272.


Contextualism and Semantic Ascent


Contextualism is the view that the truth-conditions for K-sentences (i.e., sentences of the form “S knows that P” or “S does not know that P”) “vary in certain ways according to the context in which they are uttered.”[1] (1992, 914)  Contextualism’s intended purpose is to say something important about the nature of knowledge generally as well as to provide a response to the problem of skepticism. Here I will focus on the version of contextualism put forth in a series of recent papers by Keith DeRose.  I will argue that some of the recent refinements that his contextualism has undergone create serious problems for the theory.


DeRose’s Contextualism

Much of what DeRose has to say about knowledge is informed by reflection on Nozick’s “sensitivity” requirement.  Nozick thought that a belief is knowledge only if the following subjunctive conditional holds: if P were false you would not believe that P. The most popular way to understand the sensitivity requirement (or any modal claim for that matter) is in terms of possible worlds. To say that your belief is sensitive is to say that in the nearest possible world where P is false, you don’t believe that P.[2]

            DeRose replaces Nozick’s notion of sensitivity with the notion of epistemic strength.  Epistemic strength is also a function of how well the subject’s belief tracks the truth across a range of possible worlds.  “The further away one can get from the actual world, while still having it be the case that one’s belief matches the fact at worlds that far away and closer, the stronger a position one is in with respect to P.”  (1995, 34)  Nozick thought that a belief counts as knowledge only if it tracks the truth all the way out to the world in which P is false.  One could say that, for DeRose, knowledge is true belief that is “strong enough”.  How strong does a true belief have to be to count as knowledge?  That’s where context comes in.  How good an epistemic position S must be in with respect to P to make an utterance of “S knows that P” true will depend on such things as the importance of being right, the mentioning of a possibility and the consideration of a possibility in thought.  These kinds of “conversational features” determine the size of the sphere of possible worlds within which a subject’s belief must track the truth.  Contexts in which the epistemic standards are low will be contexts in which the sphere is rather small and so the subject’s belief must track the truth across only a small number of possible worlds.  Thus in a low standard context, a belief may not need to be sensitive to count as knowledge.  Contexts in which the epistemic standards are high will be contexts in which the sphere of relevant possible worlds is much larger.[3]

            Here are a few cases to illustrate how epistemic standards can be raised in the course of a conversation: 

Case 1.  DeRose and his wife are driving home on a Friday afternoon.  They plan to stop at the bank and deposit their paychecks but traffic is bad and the lines in the bank are very long.  So DeRose suggests to his wife that they make the deposit on Saturday morning since it’s not all that important that they make the deposit immediately anyway.  His wife says, “Maybe the bank won’t be open tomorrow.  Lots of banks are closed on Saturdays.”  He replies, “No, I know it’ll be open. I was just there two weeks ago on Saturday.  It’s open until noon.” 


Case 2.  This is just like Case 1 except now it is extremely important to deposit the paychecks because DeRose has just written a large and very important check.  If the checks bounce it spells big trouble for the DeRoses and the bank is not open on Sunday.  DeRose’s wife reminds him of these facts and adds, “Banks do change their hours.  Do you know that the bank will be open tomorrow?” 


DeRose claims that even though his epistemic position is the same in both cases, it is true to say in Case 1 that he knows that the bank will be open tomorrow but if he were to say this in Case 2 he  would be speaking falsely.  The difference is that the standards for knowledge in Case 2 are higher because it is much more important that he is right. 

Case 3.  DeRose and his kid are at the zoo.  The kid asks, “Daddy, do you know what that is?”  DeRose replies “Yes, I know what that is.  It’s a zebra.”


Case 4.  This is the same as Case 3 except after DeRose claims to know that the animal is a zebra, Dretske, who happens to be standing next to DeRose, says, “How do you know it’s not a cleverly painted mule?”  DeRose replies, “Well, I guess I can’t rule out that possibility.”  “So then you don’t know that it’s a zebra?”  “Well, I guess not.”


On DeRose’s view, it is true to say in Case 3 that DeRose knows that the animal is a zebra, while in Case 4 it is true to say that he does not know this.   In neither of these cases is DeRose able to rule out the possibility that the animal is a cleverly painted mule.  But in Case 4, the mentioning of this possibility raised the epistemic standards beyond the ones that were in place in Case 3.  DeRose’s epistemic position was strong enough for knowledge given the standards in place in Case 3 but not strong enough in Case 4.

Case 5.  DeRose is driving by himself through the countryside.  He looks out the window and sees a big red barn.  He says to himself, “I know that’s a barn”.  He revels in this knowledge for a second and then continues to drive on thinking about other interesting things.


Case 6.  This is the same as Case 5 except after saying to himself “I know that’s a barn” he recalls a philosophy paper he once read about there being fake barn faćades on the sides of some American highways.  He says to himself, “Well, I guess it could just be a fake barn faćade.  So I guess I don’t know that’s a barn.”   


Again, contextualism says that is true to say in Case 5 that he knows while in Case 6 it would not be true to say this.  Let’s call the person who utters a K-sentence an attributer of knowledge and the referent of the ‘S’ in a K-sentence the subject of knowledge.  Case 6 is similar to Case 4 in that the standards have been raised so that the relevant K-sentence is false.  The difference is that in Case 6 the subject and attributer are the same person.  This is an example of the standards being raised by silently considering an alternative possibility.[4]

            These pairs of cases are designed to show that sometimes it seems correct to say of a subject that he knows that P while other times it seems incorrect.  And this is true even though there is no change in the subject’s evidence, his confidence, the truth value of P, or the conterfactual persistency of the subject’s belief that P.  Contextualism, if true, would explain why we are inclined to assert opposing K-sentences in these cases.  In low standard contexts, “S knows that P” means S knows relative to low standards that P while in high standard contexts “S knows that P” means S knows relative to high standards that P.[5] 

            One of contextualism’s main selling points is its way of responding to skepticism.  As DeRose sees things, the central argument for skepticism is as follows:

(SA)    1.  You don’t know that not-H.

2.     If you don’t know that not-H then you don’t know that O.

3.     Therefore you don’t know that O.


Let H be any skeptical hypothesis; such as the hypothesis that you are a brain in a vat (BIV).  Let O be any ordinary proposition that you might claim to know; such as the proposition that you have hands.  Since being a BIV analytically entails not having hands, the argument says, you don’t know that you have hands because you can’t rule out the possibility that you are a BIV.

            Some philosophers have sought to block SA by denying its second premise.  Dretske and Nozick both claim that the problem with SA is that it implicitly involves commitment to the idea that knowledge is closed under known deductive implication.  This “Closure Principle” says that if S knows that P and S knows that P entails Q then S knows that Q.   It would appear that the skeptic who employs SA is implicitly appealing to this principle.  He is claiming that since you don’t know that H is false and, given closure, you would be able to know that H is false if you knew O, you do not know O.  Denying closure gives these philosophers a way to deny 3 while assenting to 1.  You can know that you have hands even though you don’t know that you’re not a BIV.  DeRose finds this line of reply objectionable because it makes commitment to what, on it’s face, looks patently absurd.  It involves “embracing the abominable conjunction that while you don’t know that you’re not a bodiless (and handless!) BIV, still, you know that you have hands.”  (1995, 28)  Beyond that counterintuitive consequence, some find denying closure objectionable because it seems to block our ability to apply deductive reasoning and arrive at new knowledge.

One traditional Moorean response to arguments like SA is just to say that the conclusion is obviously absurd so one of the premises must be false.  Take your pick.  If you agree that denying 2 is unacceptable you must reject 1.  While DeRose will ultimately agree that we can and often do know that were not BIV’s he does not favor such a quick and dirty solution.  DeRose thinks that a good solution to skepticism ought to explain why so many people are compelled by SA and this kind of response, by itself, doesn’t do that.   

Contextualism provides a way to explain the intuitive plausibility of SA without giving in to skepticism completely. 

According to contextual analysis, when the skeptic 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 skeptic 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 conversation.  (1992, 917)


By introducing skeptical hypotheses into the conversation the skeptic can manipulate the standards for knowledge in a way that requires my belief that I am not a BIV to be sensitive for it to count as knowledge.  Since this belief is not sensitive, the skeptic creates a context in which the first premise of SA is true.  DeRose sees the second premise as a comparative conditional.  Suppose I were to utter “If Texas is not big, then North Carolina is not big.”  One strong reason for assenting to this conditional is the comparative knowledge that Texas is at least as big as North Carolina or, in other words, that North Carolina is no bigger than Texas.  Given this comparative fact, the conditional is true no matter how high or low we set the standards of bigness.  According to DeRose, the conditional “If you don’t know that you are not a BIV then you don’t know that you have hands” ought to be understood comparatively as well.  It is true because you’re in no better position to know that you have hands than you are to know that you are not a brain in a vat.  If you don’t know the latter then you don’t know the former and this is true no matter how high or low the standards for knowledge are set.

If the skeptic is successful in raising the epistemic standards he creates a context in which the first premise of his argument is true.  Since the second premise is true no matter where the epistemic standards lie, he thereby creates a context in which his argument is sound.  This explains why skeptical arguments seem so compelling when they are presented.  But the common folk need not fear for the skeptic’s high standards are not the ones normally in place.  Even though the skeptical argument is sound when presented, most of our commonsense K-sentences are true as well. 


The Warranted Assertability Objection

We tend to assert positive K-sentences in Cases 1, 3, 5 and outside philosophical discussions of skepticism and we tend to assert negative K-sentences in the opposing cases. DeRose believes that contextualism is superior to opposing positions because it is the best explanation for these linguistic data. Our tendency to make these seemingly inconsistent assertions is attributed to a shift in their truth-conditions.  “Invariantism” is the name of one position that opposes contextualism.  The invariantist claims that the truth conditions for K-sentences do not vary with conversational context.  Peter Unger (1975) is someone who holds such a position and he is also a skeptic.  Of course, a skeptic could simply dismiss DeRose’s data as philosophically irrelevant and say that whenever we assert a positive K-sentence, we inappropriately misapply the term ‘know’.  But Unger’s defense of skepticism is interesting in that he seems to accept the linguistic data and he seems to see explaining them as a serious burden for the skeptical invariantist. 

Unger explains our tendency to assert what he takes to be false K-sentences by pointing out that sometimes it is useful to say false things.  It suits the purposes of the good people at Planter’s to refer to the environments within their peanut cans as “vacuums”.  This is not because ‘vacuum’ has a different meaning depending on whether you work in a physics laboratory or at a peanut cannery, but because “the interior of those cans, surrounding the nuts, is near enough to being a vacuum for the purposes at hand.”  (1975, 52)  Likewise, while Unger says that all of our utterances of positive K-sentences, are in fact false, it suits our purposes to make them.  Thus there is no need to postulate a semantic shift. 

 DeRose puts the objection this way.  Our linguistic dispositions do not reflect a shift in the truth-conditions of K-sentences but only a shift in their warranted assertability conditions.  Since we know that for many instances of P, a shift in the warranted assertability conditions does not reflect a shift in the truth-conditions, why should we think that it does when P is a K-sentence? 

The idea that the rules which govern when it is appropriate to assert certain claims can differ from the rules which govern the truth conditions for those claims dates back at least as far as the work of H.P. Grice.  As anyone who has taught a course in formal logic knows, people often have trouble understanding how the wedge could be a translation of the English word “or”.  No ordinary speaker of English would say that he wants a hamburger or some fries if he wants a hamburger and some fries; even though, if the wedge means “or”, any situation in which someone wants a hamburger and some fries is also a situation in which he wants a hamburger or some fries.  Some might draw the conclusion that the wedge is not a correct translation of “or”.  But another alternative is available.  The reason why asserting that one wants a hamburger or some fries in situations where he wants both might seem inappropriate is not because it is false but because of a conversational rule that says: assert the stronger.  If someone asks whether I have seen Steve lately, I say something true if I say that I have seen him sometime in the last 5 years.  But saying this is misleading and conversationally inappropriate if my basis for saying it in this context is my knowledge that I saw him 5 minutes ago.    Something similar might be true of the above case involving “or”.  Perhaps saying that one wants a burger or some fries seems incorrect not because it is false but because it violates a conversational maxim. 

In the context of epistemology, discussion of truth vs. warranted assertability for assertions involving core epistemological concepts such as knowledge and belief, has received a good deal of attention.  One interesting way in which this distinction arises is within the context of discussions of Moorean assertions such as: “It is raining but I don’t know that it is raining”.  The proposition expressed by this utterance is possibly true, but it seems that any instance of asserting it is incorrect or inappropriate.  The puzzle is to determine the rule that explains why assertions such at this seem to clash even though they are not explicit contradictions.  As a solution to this puzzle, DeRose endorse the Knoweldge Account of Assertion (KAA).  According to KAA, the conditions for warranted assertability are equivalent to the conditions for knowledge.  “One is well-enough positioned to assert that P iff one knows that P.”  (2002, 180)[6]  

When one flat out asserts that P (as opposed to saying “I think that P” or “I’m pretty sure that  P” or “Maybe P,” etc.), one is representing oneself as knowing that P.  To put it more propositionally, one is representing it as being the case that one knows that P.  Thus, when one flat out asserts the first half of the Moorean sentence, one is representing it as being the case that one knows that it is raining.  Therefore, when one goes on to say in the second half of the Moorean sentence that one does not know that it is raining, one is saying something inconsistent—not with what one asserted in the first half of the sentence—but with what one represented as being the case by the first half.  (1991, 598)


KAA provides DeRose with a means to turn the invariantist’s claim that the warranted assertability conditions for K-sentences are a function of conversational context against him.  If it is true that you are only warranted in asserting what you know and warranted assertability conditions vary depending on context, then knowledge must depend on context also.  Given KAA, the invariantist’s objection turns out to be an argument in favor of contextualism. 


The Problem of Vanishing Knowledge

At this point an invariantist would either have to reject the knowledge account of assertion or find some other way of explaining or rejecting the linguistic data. While I think that the invariantist has plenty of room to make a plausible case for either one,[7] I won’t pursue that here.  Instead I want to show that what DeRose says in response to this objection conflicts with other things he says as he articulates his position.

            DeRose writes,


it can seem to many that contextualism renders knowledge unstable or elusive in the sense that it would make our knowledge come and go – be gained or lost – as conversational context changes.  And to many, this can seem a very problematic implication of contextualism, since it seems to them that knowledge in fact doesn’t appear and disappear due to changes in conversational context.  (2000, 7)


Consider the following dialogue:[8]

A:  Do you know what that is?

B:  Yes, I know it is a zebra.

A:  But you can’t rule out its being a cleverly painted mule?

B:  No, I can’t.

A:  So, you admit you didn’t know it was a zebra?

B:  No, I did know then that it was a zebra.  But after your question, I no longer know.


Let’s assume for the moment that contextualism is true and that the third line of the dialogue is an instance of someone successfully raising the standards for knowledge.  The first two lines of the dialogue operate in a context sufficiently low to make an utterance of “B knows that’s a zebra” true while the rest of the dialogue operates in a context that would make this utterance false.  As DeRose admits, the last line of this dialogue seems completely silly and absurd.  If contextualism countenances this kind of talk, it ought to be rejected.  But on the other hand, isn’t this sort of thing exactly what contextualism implies?  If the consideration and mentioning of alternative possibilities operate to raise the standards for knowledge doesn’t it follow that when we do this sort of thing, we are destroying our knowledge?  Look at what another contextualist, David Lewis, says in his paper “Elusive Knowledge”.

Maybe epistemology is the culprit.  Maybe this extraordinary pastime robs us of our knowledge.  Maybe we do know a lot in daily life; but maybe when we look hard at our knowledge, it goes away.  (1999, 221) 


Rather than admit that our knowledge “goes away” when alternative possibilities come into play, DeRose follows a suggestion Lewis makes at the end of that paper and argues that the problem can be avoided by semantic ascent.  Contextualists say that the truth-conditions of K-sentences depend on our conversational context.  Notice that this is different from saying that whether we know depends on our context.  Whether we know is a question of whether the truth conditions for “S knows that P” are met by S.  This will depend on S, his environment and the strength of his belief and not on the conversational context.  The conversational context determines what the truth-conditions for the sentence “S knows that P” will be.

            So what should the contextualist say about cases like our dialogue above where the epistemic standards are raised in the course of conversation?  He should not say that any knowledge has vanished, been lost, or destroyed.  Instead he should operate on a metalinguistic level.  “B knows that’s a zebra” is true relative to the low standards for knowledge in place in the first two lines of the dialogue; while “B knows that’s a zebra” is false relative to the high standards for knowledge in place in the rest of the dialogue.  No knowledge has been created or destroyed.  B never had high standard knowledge and the low standard knowledge hasn’t gone anywhere, it’s just that its no longer picked out by utterance of the K-sentence. 

            DeRose thinks that K-sentences function in a way similar to sentences that employ indexicals such as ‘here’.  A week ago I was in Starke and now I’m in Citra.  While in Starke, I truly uttered “I am here”.  But I cannot truly describe my location a week ago by uttering “I was here” because I wasn’t here, I was in Starke.  In the same way, the content of K-sentences is fixed by the conversational features of the context in which they are uttered and not by the features of the time being talked about.  This is why it is inappropriate and untrue for B to say “I did know then but I don’t know now” in the earlier dialogue.  But, DeRose tells us, “B can say ‘My previous knowledge claim was true’” (1992, 925) just as I can say, “My previous utterance of ‘I am here’ was true.” 

            In what follows I will argue that given what DeRose has said so far, B can’t say that.  Then I will spell out the consequences of this fact. 


The Trouble with Semantic Ascent

Let’s modify the earlier dialogue in the way DeRose would have it go.

A:  Do you know what that is?

B:  Yes, I know it is a zebra.

A:  But you can’t rule out its being a cleverly painted mule?

B:  No, I can’t.

A:  So, you admit you didn’t know it was a zebra?

B:  No, my previous knowledge claim was true, but now if were to utter that same K-sentence it would be false. 


B attempts to handle the problem of vanishing knowledge in the way that DeRose recommends.  But, if we follow DeRose in endorsing KAA, the last line of this dialogue is no less inappropriate than that of the earlier one.  If B is warranted in asserting “my previous knowledge claim was true” then according to KAA, he knows that his previous knowledge claim was true.  So B knows that his earlier utterance of “I know that is a zebra” was true.  Given closure and the assumption that B knows that “S knows that P” entails P, B also knows “that is a zebra” was true at the time he uttered it.  Let’s assume the possibility that the zebra was removed and replaced with a painted mule does not obtain and is not a possibility under consideration within the conversation.  From the fact that B can appropriately assert “My previous knowledge claim was true” it follows that he knows that’s a zebra and the he can appropriately assert “that’s a zebra” even at the end of the dialogue.  But, ex hypothesi, this is a dialogue where the standards have been raised so that B can’t appropriately assert “that’s a zebra”.  Consider another dialogue.

A:  Do you know whether you have hands?

B:  Yes, I know that I have hands.  Here’s one, here’s another.

A:  But can you rule out the possibility that you are a BIV?

B:  Nope.

A:  So you don’t know that you have hands.

B:  That’s true now, but my previous utterance of “I know I have hands” was true.  From which it follows that I have hands.  Ergo, I know that I have hands.


            Of course, A could always challenge B’s inference from the claim that his previous K-sentence was true to the claim that he now has hands.  After all, B could have been envatted sometime during the conversation.  As with the penultimate dialogue, I shall assume that this possibility does not obtain and that it isn’t relevant to the conversation. 

It was said earlier that one of the putative advantages that contextualism has over a traditional Moorean response to skepticism is that the latter makes refuting the skeptic too easy.  But if what has been said so far is correct, it would appear that contextualism makes refuting the skeptic too easy.  All that you need to do is go metalinguistic. We can reformulate the basic idea of this dialogue in the following anti-skeptical argument.

(ASA) 1.  My utterance of “I know that I have hands” in ordinary contexts is true.

2.     If my utterance of “I know that I have hands” in ordinary contexts is true then

“I have hands” is true.

3.     “I have hands” is true.

4.     Therefore, I have hands.


DeRose cannot reject (1) since it is an essential tenet of the case for contextualism.  (2) is analytically true given that “S knows that P” entails the truth of P.  (3) of course follows by modus ponens, an inference employed in SA and not challenged by DeRose.  And (4) is an application of the T-Schema which I assume is not under question. So, once we learn the metalinguistic game it’s easy to refute skepticism even when the context is high. 

Notice, however, that ASA does not conclude that I know that I have hands.  The other tools DeRose gives us make the K-Sentence easily obtainable.  Since DeRose asserts (1) in a skeptical context I assume that we are warranted in doing so.  I also assume that we are warranted in asserting (2) since we know it to be an analytic truth.  Given KAA, it follows that I know (1) and (2).  Given closure, it follows that I know (3) and (4) since they are simply deductive consequences of (1) and (2).[9] 

Perhaps DeRose could respond to this point by retracting the claim that B can correctly assert that his previous utterance of a K-sentence was true.  This would mean the standards of knowledge in place in a given conversational context are operative all the way up to the metalinguistic level.  DeRose could reject ASA by claiming that semantic ascent does not transcend the epistemic standards in place in a given conversational context.

But this comes at a heavy price.  For starters, it weakens the alleged analogy between ‘know’ and context sensitive terms such as ‘here’.  As we saw above, semantic ascent does give us the ability to transcend context with sentences such as ‘I am here’.  If ‘know’ functions in essentially the same way as ‘here’, why shouldn’t the same be true of it?  But apart from this worry there is an even more serious problem.  DeRose must rely on the metalinguistic move to handle another objection to contextualism. Lewis raises this other problem in the following passage.

But wait. Don’t you smell a rat?  Haven’t I by my own lights, been saying what cannot be said?  (Or whistled either.)  If the story I told was true, how have I managed to tell it?  In trendyspeak, is there not a problem of reflexivity?  Does not my own story deconstruct itself? (1998)


In other words, in the process of making his case the contextualist, by raising and discussing skepticism, puts himself in a high standard epistemic context.  But according to the contextualist, these are contexts in which we nothing or very little.  Much of the contextualist’s case rides on things which, according to his own views, are unassertable and inappropriate in his context. 

For example, look at what DeRose says in the following passage.

Consider my belief that I have hands.  I believe this at the actual world and it’s true.  What’s more, in the other nearby worlds in which I have hands I believe that I do.  There are also, at least in my own case, some alarmingly close worlds in which I don’t have hands.  These include worlds in which I lost my hands years ago while working on my uncle’s garbage truck.  In the closest of these not-P worlds, I’m now fully aware of the fact that I’m handless … My belief as to whether I have hands doesn’t match the fact in various worlds in which I’m a BIV, of course, but these are very distant.  (1995, 34)


Contextualism is often criticized for begging the question against the skeptic.  You can see from the above (if you haven’t already) why people would say this.  DeRose is here simply asserting that he has hands and that the worlds in which he doesn’t but believes he does are far, far away.  While I have sympathies for the objection that contextualism is question begging, I won’t pursue that line of argument here.[10]  Instead, I will simply remind everyone that this passage is being offered within a philosophical discussion of skepticism; a high standard context if anything is.  Given contextualism and KAA it is not just question-begging but inappropriate for DeRose to say these things.  Let me expose the problem with the following skeptical argument against contextualism.

(SAAC)          1.  You don’t know that you are not a BIV.

2.     If you don’t know that you are not a BIV, then you don’t know that contextualism is true.

3.     Therefore, you don’t know that contextualism is true.     


According to DeRose, (1) is true when uttered in skeptical contexts.  (2) can be read in the same way DeRose reads (2) in SA; as a comparative conditional.  Contextualism is a contingent thesis about our use of the term ‘know’.  DeRose is certainly in no better epistemic position with respect to a claim about how we use a word than he is with respect the claim that he has hands.  Therefore, the second premise of SAAC is true for the same reason that the second premise of SA is true.  And the conclusion follows deductively.  If skeptical hypotheses succeed in raising the standards to a point where we do not know and cannot even appropriately assert that we have hands, we cannot assert abstract and airy claims about out linguistic usage either. So if contextualism is true, it cannot be said.

This problem of “reflexivity” is also supposed to be handled through metalinguistic


maneuvering.  Here is Lewis again.   


I could have had my say fair and square, bending no rules.  It would have been tiresome, but it could have been done.  The secret would have been to resort to ‘semantic ascent’.  I could have taken great care to distinguish between  (1) the language I use when I talk about knowledge, or whatever, and (2) the second language I use to talk about the semantic and pragmatic workings of the first language.  (Lewis, 1998, 238)


But if, as was suggested earlier, the high epistemic standards put in place by the skeptic chase us up the semantic ladder, then metalinguistic maneuvering does not give the contextualist anywhere to go.  The contextualist cannot make his case unless he does it metalinguistically.  So, if semantic ascent doesn’t get us out of our epistemic context, then the contextualist is, by his own lights, consigned to silence.

On the other hand, if semantic ascent can get us “outside” our epistemic context then contextualism is no improvement over the traditional Moorean response.  Refuting skepticism is easy.  And even apart from skepticism, it would seem impossible for anyone to raise the epistemic standards once we become familiar with the technique of semantic ascent.  Anytime someone raises the standards, we can just go metalinguistic and retrieve our knowledge.

There is at least one line of response available.  If DeRose were to accept the idea that the epistemic standards in place in a given context also operate at the metalinguistic level, he could abandon the idea that contextualism solves the skeptical problem and he could stop talking about skepticism altogether without giving up contextualism.  He could articulate the view and discuss it’s implications without making us think about any skeptical hypotheses.  Of course, this would involve relinquishing one of the main virtues of contextualism, but perhaps there is a case to be made for it’s plausibility independent of handling of skepticism.  So contextualism might be plausible someday, but only if we’re willing to keep our standards low. 








DeRose, K.  (2002).  “Knowledge, Assertion, and Context”.  The Philosophical Review, 111: 2,



--.  (2000).  “How Can we Know that We’re Not Brains in Vats?”.  The Southern Journal of

Philosophy 38, Spindel Conference Supplement, 121-148.


--.  (1995).  “Solving the Skeptical Problem”.  The Philosophical Review, 104: 1, 1-52.


--.  (1991).  “Epistemic Possibilities”.  The Philosophical Review, 100: 4, 581-605.


--.  (1992).  “Contextualism and Knowledge Attributions”.  Philosophy and

Phenomenological Research, 52: 4, 913-929.


Lewis, D.  (1999).  “Elusive Knowledge” in Skepticism: A Contemporary Reader, DeRose and

Warfield (eds.), New York: Oxford University Press.


Unger, P.  (1975).  Ignorance: A Case for Skepticism.  New York: Clarendon Press.


Yourgrau, P.  (1983).  “Knowledge and Relevant Alternatives”.  Synthese, 55 175-190.


[1] In what follows I will refer to statements of the form “S knows that P” as “positive” K-sentences and statements of the form “S does not know that P” as “negative” K-sentences.

[2] Typically, the relative “nearness” of possible worlds is understood this way:  W1 is nearer to the actual world then W2 iff W1 is more similar to the actual world than W2. 

[3] I gather that this is DeRose’s current take on things because he says in (2000) that we can know that we are not brains in vats even though my belief that I am not a brain in a vat is not sensitive.  Since this seems to conflict with the “Rule of Sensitivity” put forth in (1995), I assume that he has changed his mind. 

[4] Of course, this way of raising the standards will only work when the conversation is an internal one.

[5] For ease of exposition and because nothing turns on it here, I will talk as if there are only two epistemic standards: high and low.  But I don’t mean to suggest that contextualism is necessarily committed to such a simplistic view of things.

[6] DeRose leaves it open question whether there are also additional rules governing assertion.

[7] If I were in Unger’s position, I would object to DeRose’s description of it.  One can make the point about usefulness without saying anything about “warranted assertability”.

[8] A dialogue like this appears in Yourgrau (1983).

[9] A propositionalized version of ASA can also be given.

            (ASAP)             1.  S  knows relative to low standards that S  has hands at t is true.

2.     If S  knows relative to low standards that S  has  hands at t is true then S  has  hands at t is true.

3.     S  has  hands at t is true.

4.     Therefore, S has hands.

The same criticism can be given using ASAP.  Even though ASAP only proves that a subject has hands at t where t may be sometime before now, if skepticism is truly the tough problem that we think it is, ASAP should not even be able to prove that.

[10] In response to the allegation of begging the question, DeRose says, “if the skeptic is marshaling deeply felt intuitions of ours in an attempt to give us good reasons for accepting his skepticism, it’s legitimate to point out that other of our beliefs militate against his position, and ask why we should give credence to just those that favor him.”  (1999, 215)