Jump to ContentJump to Main Navigation
When Truth Gives Out$

Mark Richard

Print publication date: 2008

Print ISBN-13: 9780199239955

Published to Oxford Scholarship Online: September 2008

DOI: 10.1093/acprof:oso/9780199239955.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 14 December 2017

(p.166) Appendix II: Relativism and Contextualism about Knowledge

(p.166) Appendix II: Relativism and Contextualism about Knowledge

When Truth Gives Out
Oxford University Press


The relativism defended in Chapter 4 is hardly shocking. It doesn't, for example, entail that thinking or saying makes it so. In placing the relative within the range of variation allowed by our practices of accommodating and negotiating meaning, it confines it to an arena in which we recognize that there may be legitimate differences of opinion, different opinions each of which, in the right situation, may be correct. My relativism is so boring that one might wonder whether it's of much interest. Does it give us some traction on any philosophical issue?

It does: it provides an effective answer to objections to contextualism about knowledge ascriptions. Indeed, it increases contextualism's plausibility.1

According to the contextualist, the relation expressed by ‘knows’—and thus what is said by a sentence like ‘George knows that he has hands’—depends upon the ‘standards of knowledge’ supplied by the context of the sentence's use. A representative version of the view, which I'll presuppose in what follows, has it that

  1. (a) knowledge that p requires being able to rule out scenarios in which p is not true, where ruling out a scenario generally requires giving good evidence that it doesn't obtain;

  2. (b) what scenarios one needs to rule out in order to know varies across contexts.

What determines when one needs to be able to rule out a scenario in order to know? Many contextualists point to the salience of a scenario: simply raising the possibility that (say) our current experience is a product of imagination and not of perception seems to put us in a position where we aren't entitled to claim to know p (p a claim justified only if our experience is veridical) unless we can give reasons to think that our experience is not merely a product of our fancy. Though it doesn't figure very much in the contextualist literature, it seems that our interests and purposes in ascribing knowledge can also contribute to determining whether one needs to be able to rule out a scenario before knowing.2

Why should salience of counter‐possibilities and our interests be factors in determining whether someone has knowledge? Well, the first seems required by (a) above. Knowing p just is, in part, being in a position, when someone says ‘Well, for all we know, q (and if q then not p)’, to say ‘Don't worry about q, because . . . ’. As for the dependence of (p.167) knowledge on interests: We use what we know to figure out what we ought to do. Thus, the standards we need to meet in order to count as having knowledge bear on action: the more flexible we need or want to be in acting, the less certainty we can demand for the beliefs on which we base our calculations of expected benefit. Our interests and purposes help shape how we manage this trade‐off between what we require for knowledge and how flexible we are in action, by helping us decide what sorts of contingencies we ought to be able to rule out before using a belief as a basis for deciding how to act.

That these things are factors in determining whether someone has knowledge is, if not a truism about knowledge, at least a truth. And, so far as I can see, contextualism about knowledge best makes sense of them.3 This alone makes it an attractive philosophical position. It also offers insight into what is going on when we consider skeptical arguments to the effect that since for all we know q (q being some doubt‐casting claim, such as that we are brains in a vat), we can't know p (p being some claim we would normally claim to know, such as that we have hands). Such arguments are unsettling in part because (a) before we confront them we are quite confident that we know the relevant p, but (b) afterwards we uneasily think that perhaps we don't know it. If contextualism is correct, then these reactions are to be expected because they are correct: until the salience of the doubt‐casting scenario is raised we do know the relevant p; after it is raised, we don't.4

(p.168) II.2

I have been extolling the virtues of contextualism about knowledge. But the view is liable to the same sorts of objections concerning disagreement as is contextualism about the gradable adjectives. Contextualists hold that the truth of ‘I know that I have hands’ is determined by the standards for knowledge operative when the sentence is used. Consider a variant of Chapter's 4 example, The Report. In it, Didi uses ‘knows’ with low standards for knowledge, low enough, we may suppose, so that she qualifies as knowing that she has hands. Naomi uses it with high standards, high enough, we may suppose, so that Didi does not qualify as knowing that she has hands. It would seem that given contextualism about ‘knows’, Didi speaks truly if she says ‘I, Didi, know that I have hands’; Naomi speaks truly if she says ‘Didi doesn't know that she has hands’. But Didi says that she, Didi, knows that she has hands; Naomi says that Didi does not know that she has hands. Since each speaks truly, it follows that Didi knows that she has hands, and that Didi doesn't know she has hands. Contradiction.

This, of course, is just an epicycle on one of the objections to contextualism about wealth that we discussed in Section 4.45. If we embrace relativism, we can respond to it in just the way we responded to the original objection. Given the banality on which it rests—that (ignoring issues about tense) whenever someone assertively utters the English ‘Didi knows that she has hands’ they say that Didi knows that she has hands—the contextualist must deny that both Didi and Naomi spoke the truth. I can evaluate sentences such as ‘Didi knows that she has hands’, ‘It is true that Didi knows that she has hands’, or ‘What Didi said is true’ for truth. In doing this, I interpret ‘knows’ and ‘true’ using the standards of my context, to determine who knows what and what is true. No matter what those standards might be, they won't make Didi know something but not know it; nor will they make the claim that Didi knows she has hands true and not true.

It is a mistake—a mistake contextualists about knowledge have committed—to formulate contextualism so that it implies that what Didi says with a sentence such as ‘Didi knows that she has hands’ is true just in case according to Didi's standards of knowledge Didi knows that she has hands. What Didi says with this sentence is true—i.e. has the property I ascribe (here and now) with ‘true’—just in case Didi knows that she has hands—i.e. just in case she has the property I ascribe to her (here and now) with ‘knows that she has hands’. Once the contextualist accepts the banality that whoever utters ‘Didi knows that she has hands’ says that Didi knows that she has hands, he must use a relativized notion of truth to formulate contextualism. Contextualism about knowledge ought be formulated as the view that whether a use of ‘Didi knows that she has hands’ is valid—i.e. is true relative to the conversational context in which it occurs—turns upon the standards of knowledge supplied by that context. Once contextualism is so formulated the puzzle above disappears, in exactly the way it disappears for contextualism about wealth.

Jason Stanley claims that this response won't fly.6 The claim occurs in the course of an argument that once the contextualist embraces relativism he can't account for the (p.169) ‘factivity of knowledge’, the banality that whatever one knows is true. Stanley interprets the contextualist who is a relativist (henceforth, for brevity, the contextualist) as holding that truth is to be evaluated relative to a possible world, a particular time, and a ‘judge’. As Stanley puts it, the ‘circumstances of evaluation’ for a contextualist are made up of a world w, time t, and judge y.7 If one thinks of matters this way then the banality that one knows that p only if p is true may be put so:

  1. (F) x knows at t that p at <w,t′,y> only if p is true at <w,t′,y>.

This won't work says Stanley:

Suppose that John is in a low standards situation with some evidence for his true belief that p, and Hannah is in a high standards situation, where more evidence is needed than John possesses in order to know that p. Then, by relativist lights, relative to <w, t, John>, John knows that p, and relative to <w, t, Hannah>, John doesn't know that p . . . we have disagreement about whether John knows that p, but both parties are correct. . . . [It is] deeply implausible that John and Hannah each is merely lucky to be right. . . . if they are both correct, then John knows that he is right, and Hannah knows that she is right. That is, in the envisaged case, John knows that John knows that p, and Hannah knows that John doesn't know that p. A neutral observer can then point out that John knows that John knows that p, and Hannah knows that John doesn't know that p (as I have just done). If [the principle (F)] were correct, it would then follow that John knows that p, and John doesn't know that p. But that is a contradiction. So [ (F) ] is false.8

This ‘reductio’ of (F) is just a variant of the objection discussed above. The contextualist will reply that whether a claim is true relative to a perspective (or judge or context) depends upon the standards operative from that perspective (that the judge is using, that the context provides); that the person to whom knowledge is ascribed may have different standards is neither here nor there. No matter what the standards of the ‘neutral observer’, they will make no claim both true and false. So the observer will not be able to observe that John knows q and Hannah not q. The neutral observer can of course observe that John's self‐ascription of knowledge (that he knows p) and Hannah's denial that he has that knowledge are both valid—true in the contexts in which each is made. But no inconsistency arises from this.9

(p.170) Stanley makes a second, more interesting objection to relativism about knowledge. To set the stage for it, recall that we often represent a proposition as a function from ‘circumstances of evaluation’—the things relative to which a proposition is true or false—to truth‐values. Possible‐worlds semantics represents a proposition as a function from possible worlds to truth‐values: The proposition that during 1982 Al was fickle gets represented as the function that maps a world w to truth just in case during 1982 Al was fickle in w. Those who think that propositions may differ in truth‐value not just across worlds but from time to time within a single world represent propositions as functions which map a world and a time to a truth‐value: the proposition that Al is fickle, to this way of thinking, should be represented as the rule which assigns truth to a world w and a time t just in case Al, at time t in world w, is fickle.

If you represent propositions as functions you'll say that a sentence is true at a circumstance X just in case the proposition the sentence expresses maps X to the true.10 The sentence ‘During 1982 Al was fickle’ is in fact true—true at the actual world—just in case the claim that during 1982 Al was fickle maps the way things actually are to truth. And this is the case just in case Al was, as things actually are, fickle in 1982. Note now that there are operators (like ‘possibly’ and ‘necessarily’) whose meanings ‘shift’ the circumstance that is relevant to whether what a sentence says is true:

It could be that S is true at world w just in case the claim that S is true at some world.

Necessarily, S is true at world w just in case the claim that S is true at every world.

One might argue—David Kaplan once did argue—that tenses are best understood as playing such a shifting role, and therefore propositions should be thought of as functions from worlds and times to truth‐values. After all, Kaplan observed, it seems that

It was true that S is true at a time t (in a world w) just in case the claim that S is true at some time earlier than t (in world w).


It will be true that S is true at a time t (in a world w) just in case the claim that S is true at some time later than t (in world w).11

If what a sentence says is to be evaluated for truth relative to circumstances that systematically vary with respect to a parameter P—a possible world, a time, a place, whatever—one expects that there will be operators whose role is to shift that parameter as do modals and (according to Kaplan) the tenses. After all, if our claims vary in truth‐value as the values of the parameter P shift—if a claim might be true at (a circumstance with) p, false at (one with) p*—then we will have reason to want to convey that a particular claim, while not true here at p, is true at p*, or that it isn't true at any P of a particular sort, etc. Conversely, one expects that if there aren't operators whose role is to shift parameter P, then quite probably parameter P doesn't play a role in individuating circumstances of evaluation. This observation brings us to Stanley's objection.

The relativist thinks that claims are true relative to a possible world and a set of standards (or kindred parameter). This implies that language provides expressions whose role is to shift this parameter. If there aren't such expressions, that is good evidence that truth isn't relative to that to which the relativist says it is relative. But, says Stanley,

it is hard to see how the . . . relativist about knowledge could countenance the existence of such operators, if she is to preserve her claim to be providing a more charitable account of our intuitions about knowledge ascriptions. Suppose O‐j is an operator on the judge feature, one that evaluates the content of the embedded sentence at the judge feature corresponding to j. Then, the . . . relativist would countenance the truth of claims such as ‘O‐j John knows that the bank will be open, but not O‐h John knows that the bank will be open.’ I cannot think of any natural language stand ins for such operators that are linked to some notion of truth, and would make it acceptable to countenance the truth of instances of these claims.12

Surely natural language is loaded with such operators. Consider

From John's perspective, relative to Hannah's standards, it's true for Bob, given low standards, if we adopt very strict standards for knowledge/for being rich/for being flat, then . . .13

Stanley anticipates such a response. He argues that since these sorts of operators can be acceptably applied to pretty much any sentence whatsoever, to say that they express (p.172) relative truth commits us to saying that pretty much any sentence whatsoever is only relatively true. The crux of his argument is this. Consider a sentence like

  1. (C) By the standards of chemists, the stuff in the Hudson River isn't water, and someone with no lab experience doesn't know that hydrogen is an element.

The effect of ‘By the standards of chemists’ on each conjunct must be the same, says Stanley. Its effect on the second conjunct is to require very strong standards for its truth. So it has the same effect on the first one. So, if ‘By the standards of chemists’ can shift the standards of truth for knowledge claims, it can do the same for other sentences. So every, or pretty much every, sentence will possess truth‐values only relatively, as we can always come up with an operator that will shift the standards for a sentence's truth in a way that changes its truth‐value from our own perspective.14

Let's allow for the moment that ‘By the standards of chemists’, when it's used to indicate relative truth, has the same effect on ‘the stuff in the Hudson River isn't water’ as it does on ‘someone with no lab experience doesn't know that hydrogen is an element’. What is the effect it has? As I see it, it shifts those standards whose variation can in fact effect the truth of a discourse. What are those standards? Well, not just any variation of standards or opinions will have an effect on truth. I have suggested that we find relative truth when we have an area of discourse where we recognize that our concepts can be developed in different ways, and when we are prepared to accommodate and negotiate over different ways of fleshing those concepts out in the course of a conversation. The notions expressed by words which are subject to the sort of conversational accommodation discussed by David Lewis—those expressed by gradable adjectives and, if contextualism about knowledge is correct, verbs like ‘knows’—are the primary examples of such notions.

Given that this is what operators of the form on X's standards work on when they are used to convey the relativity of a truth, there are genuine and severe limits to what can be relatively true. We do not, for example, recognize that it is up for serious negotiation as to when the Revolutionary War ended, or whether Ben Franklin was the first Postmaster General, or whether water condenses on cool surfaces.15 It is just plain false, absolutely false, that on very loose standards the Revolutionary War was still being waged in 1791. There is no agent or perspective which is such that it's true for that agent or from that perspective that Mo Vaughn was America's first Surgeon General.16 There are, perhaps, some people who think that Mo was the first. But there is no way of thinking of political offices such that, on that way of thinking of them, (it is true that) Mo was the first.

What about the claim that the stuff in the Hudson River isn't water? Well, we can certainly negotiate what is to count as water for certain purposes; we will accommodate various uses of ‘water’ if doing so makes sense. Is it really the case that by the standards of chemists the Hudson River isn't water? The question just isn't concrete enough to answer. Chemists, like everyone else, use different standards on different occasions, depending on (p.173) their interests and purposes. One would have to have a particular situation with particular people in mind, in order even to begin to consider the question.

A rough test for whether a use of an operator such as ‘On Kekulé’s standards' or ‘On Kekulé’s way of looking at things' is supposed to be capturing the fact that something is true relative to a certain perspective is whether it can be expanded with ‘—and those standards are one permissible way to look at the matter—’ or ‘—and his is one permissible way of thinking of things—’. I in fact doubt that by the standards of chemists for knowledge—when those standards are one permissible way of thinking about knowledge—those without lab experience don't know that hydrogen is an element. Be that as it may, no one with a wit of sense is going to say that there is any reasonable way of looking at history on which it is true that Mo Vaughn was a nineteenth‐century doctor. It is not true, no matter who X might be, that by the standards of X concerning office holding—X's standards for who holds what office being one permissible way of looking at office holding—it is true that Mo Vaughn was once Surgeon General.

There is no reason to think that saying that operators like ‘On Niels Bohr's standards’ allow us to express judgments about relative truth leads to the view that ‘everything is relative’.


Contextualists about ‘know’ typically point to the gradable adjectives in explaining their position, suggesting that the verb's semantics is cognate to that of the adjectives. Stanley claims that ‘knows’ doesn't pattern semantically with context‐sensitive expressions to begin with.17 He makes the following observations in support of his claim.

(i) Modals are generally held to be context‐sensitive: even fixing the sense of a modal (to, say, physical possibility), its reading can shift within a conversation. An example: suppose A is discussing future developments in technology, and B doesn't realize this. We may have

  1. A: So I could fly from NY to London in an hour!

  2. B: Impossible—even the Concorde can't do that.

  3. A: I didn't deny that; I was talking about future technology.

We don't have such shifts with ‘knows’, as witnessed by the bizarreness of

  1. A: I know that's a zebra.

  2. B: Couldn't it be a painted mule?

  3. A: Uh, I guess.

  4. B: So you admit you were wrong when you said you knew it was a zebra?

  5. A: I never said I knew that it was a zebra.

Stanley argues that if ‘knows’ were context‐sensitive in the way the modal is, then its content would shift in this conversation, as B's observation effectively raises the standards for knowledge. But then, if contextualism is true, A's use of ‘I know that's a zebra’ (p.174) expresses different claims at the beginning and end of the conversation. So A's last utterance would, if contextualism were true, be acceptable.

(ii) The possibilities for propositional anaphora are different with ‘knows’ than with other contextually sensitive expressions. While

It's raining here. If I'd been inside, that/what I just said would still have been true.

is acceptable, this is not:

I don't know I have hands. . . . Hey, I can open the door; I know I have hands. But what I said earlier still holds.

(iii) Some expressions, such as ‘For all I know’, act like quantifiers over contexts. A use of

For all you know, I'm a robber.

seems true just in case in some context ‘epistemically possible’ for the addressee, the speaker of that context is a robber. Thus, ‘For all you know’ neutralizes the context‐sensitive ‘I’ so that it does not have its normal contextual interpretation. The interpretation of ‘know’ does not so shift within this operator. Even if you are unsure what standards for knowledge are operative in the context, and so are unsure whether Fred counts as knowing that he has hands, there's no felicitous reading of

For all you know, Fred knows that he has hands.

(iv) Gradable adjectives, modals, and quantifier phrases can have their standards for application or extensions shift within a sentence, as witnessed by

  • (T) That butterfly is small, and so is that elephant.

  • That field is flat, and so is that rock.

  • In Syracuse there are many serial killers and many unemployed men.

  • I can't bench‐press 100 pounds, but I could with regular training.

This isn't the case with ‘knows’, as witnessed by the infelicity of

  • (T′) Bill knows he has hands, but not that he isn't a brain in a vat.

  • Bill doesn't know that he's not a brain in a vat, but he knows he has hands.

It seems to me that rather than undercutting contextualism about ‘knows’, Stanley's observations actually make it more plausible. For in each case what Stanley observes about ‘knows’ is also true of gradable adjectives. Stanley's first point is, in effect, that when a modal such as ‘could’ is used by different speakers in ‘A could fly to Europe in a half hour’ to express different sorts of physical possibility, we treat the uses of the sentence as saying different things. He observes that we don't treat uses of ‘A knows that that's a zebra’ as saying different things even when it is clear that different standards of knowledge are presupposed by the user. The observation about ‘knows’ is correct. But exactly the same sort of thing is true of ‘rich’ and other gradable adjectives. The following dialogue is exactly as bizarre as the zebra dialogue in (i) above:

  1. A: He is rich [It is flat].

  2. B: He can't afford a house on the Vineyard [it's really sort of bumpy].

  3. A: I see your point.

  4. (p.175)
  5. B: So you admit you were wrong when you said he was rich [it was flat].

  6. A: I said no such thing.18

Similar points hold concerning propositional anaphora. True,

I don't know I have hands. . . . Hey, I can open the door; I know I have hands. But what I said earlier still holds.

is unacceptable. But so is

She's not rich. . . . Oh, yeah, I guess I do think that he’s rich, and she and he make about the same. So I guess she is rich. But what I said earlier still holds.

As far as operators such as ‘For all you know’ are concerned, Stanley himself points out that they do not have the sort of effect on gradable adjectives which they have on words like ‘I’; once again, ‘know’ and the gradable adjectives pattern together.

Finally, there are the examples (T) and (T′). These are allegedly examples in which the content of a contextually sensitive expression shifts across uses within a single sentence. In thinking about these examples we should distinguish two kinds of context‐sensitivity. The word ‘tall’ is context‐sensitive insofar as it requires a reference class—something is tall for a tree/building/two‐year‐old; nothing is tall simpliciter. Other adjectives, while perhaps not requiring a reference class in order to get assigned an extension, clearly accept one. I may call a field flat and obviously be comparing it to the fields in Bucks County, or call a pond flat with the clear intent that we evaluate what I say relative to the class of naturally occurring large bodies of water. A quite different sort of context‐sensitivity results from the fact that (even after a reference class is determined) candidates for the extension of a term such as ‘flat’ or ‘rich’ are ordered by the degree of the relevant property (flatness, wealth), with the extension of the term being determined by ‘how one draws the line’ within the ordering—such line drawing being something that can be done differently in different contexts.

Call the first sort of context‐sensitivity comparative, the second extensive. I think once we see the distinction, we also see that Stanley's first two examples are examples of comparative, not extensive, sensitivity. The first sentence, for example, is OK if understood so:

That butterfly is small for an animal, and that elephant is small for an elephant.

It's not OK—and not an example of what Stanley wants it to be an example of—if it's understood as (p.176)

That butterfly is small for an animal and that elephant is small for an animal.

But it is no part of the agenda of the contextualist about ‘knows’ to say that the verb is context‐sensitive in every way in which terms such as ‘flat’ or ‘rich’ are context‐sensitive. What the contextualist about knowledge holds is that the standards for knowledge accompanying a knowledge ascription are relevant to its truth‐conditions. As this talk about standards is usually fleshed out, differences in standards are, very roughly, differences in how reliable one is in distinguishing cases in which what one is supposed to know obtains from ones in which it doesn't. This is a kind of extensive sensitivity. And, in fact, if we look at sentences involving gradable adjectives which could be true only if we ‘draw the line’ differently for different uses of an adjective, we get the same feeling of bizarreness which Stanley notes accompanies sentences such as ‘He knows he has hands, but he doesn't know whether he is a handless brain in a vat’. Consider, for example,

This rock is flat, but that one isn't (said of two rocks which are more or less indistinguishable).19

In short: contrary to what Stanley alleges, our intuitions about (semantic) acceptability and bizarreness are exactly what one would expect, if the contextualist claim, that ‘knows’ has a semantics like the gradable adjectives, is correct.


(1) Examples of contextualist views about knowledge are Cohen (1999), DeRose (1995), and Lewis (1996).

(2) That knowledge turns on interests has been much discussed in recent anti‐contextualist literature—see, for example, Hawthorne (2004) and Stanley (2006). But there is nothing about contextualism that would rule out ascriber interests as influencing the truth of knowledge ascriptions. Au contraire.

(3) There are ‘invariantist’ views of knowledge which acknowledge them. The invariantist holds that the extension of ‘knows’—and thus what counts as knowledge—doesn't shift across contexts. Invariantists say that whether X knows p depends upon X's interests and purposes: if X's interests make it important that he be able to rule out a particular scenario before he may claim to know p but he can't rule the scenario out, then from nobody's perspective does X know p.

The invariantist holds that whether X knows p turns on the interests and standards operative in X's context. The contextualist, on the other hand, sees questions about knowledge turning on the interests and standards operating in the context in which the question Does X know p? arises. So for the contextualist it might be that while from X's perspective X doesn't know p, X does know p from some other perspective. On the invariant view, the interests of the person who asks Does X know that p? are irrelevant to the question's answer. Hawthorne (2004) and Stanley (2006) defend versions of this view.

The problem with such invariantism is that on it, knowledge isn't determined by one's epistemic situation. Suppose that Mary and Jeff have exactly the same evidence, acquired in exactly the same ways. Suppose that their methods of acquiring evidence are equally reliable, and that their situations in the world are relevantly similar. (If knowledge about birds is under discussion and one is, for example, in a place where there are birds that look like goldfinches but aren't, so is the other.) So the two are in exactly the same epistemic situation, so far as being able to rule out various challenges to whether it is the case that the bird in the garden is a goldfinch. Indeed, if you were to challenge each in a particular way (you observe, let us say, that the bird's wing seems to have a red band, which goldfinches don't have), each would respond, with the same justification, in the same way (each would, let us say, respond that it's not a red band but the setting sun's light). Surely the identity of their epistemic—that is, their evidential and justificatory—situations makes it the case that if one knows the other does. The contextualist can agree with this, for the contextualist would say that if we ask if the two know, the answer is determined by applying our standards of evidence to each one's claim to know. But the invariantist cannot agree, as it may be that Jeff's interests demand that he be able to rule out some challenge to knowledge which Mary's interests deem irrelevant.

(4) This is not to say that contextualism ‘solves’ all the problems which skepticism raises. In particular, one feels that once skeptical scenarios are introduced, there is a question as to whether we ‘really’ know the claims the skeptic says we don't know. Given what the contextualist says about knowledge, this would seem to be a question about whether the standards for knowledge the sceptic's challenge introduces ought to govern or be applied to our everyday claims to know.

(5) It is (a version of) an objection to contextualism by John Hawthorne mentioned there.

(6) Stanley (2006).

(7) Judges are not the parameter I've suggested the relativist needs, but that won't effect matters; let's just pretend for the nonce that truth is relative to the person judging, instead of a particular set of standards operative in a potential context of judgment.

(8) Stanley (2006: 146).

(9) In Stanley (2006) Stanley suggests that a relativist will respond that what John knows is that relative to his own standards he knows that p; mutatis mutandis for Hannah. He then observes that this response does not present the two as genuinely disagreeing.

The imagined ‘response’ is simply irrelevant to Stanley's original objection, which was to account for the factivity of knowledge while maintaining relativism. (F) does this perfectly well. (Notice that in the story accompanying Stanley's objection, whenever John knows that p is true relative to a world, time, and judge, so is It is true that p.) The relativist can and will say the things in Stanley's imagined response, but she will not identify the claim on which John and Hannah disagree with anything like the claim that according to John's standards he knows p. What they disagree about is simply whether John knows that p.

In an earlier draft of Stanley (2006) Stanley anticipates the response to his objection in the text. Against that response he wrote:

The relativist might reply that [A] from the standpoint of relativism, there is no observer z such that John knows that John knows that p relative to z, and Hannah knows that John doesn't know that p relative to z. In other words, [B] there is no one with respect to which [sic] the relativist account of the intuitions we have been discussing is correct. But to concede that the relativist explanation of the intuitions we have been discussing is false for everybody surely does not amount to a compelling defense of relativism. (Labels added.)

But why think that [A] amounts to [B]? Stanley seems to have been thinking something like this: (1) The relativist holds that John may know John knows that q, while Hannah knows John doesn't know that q. So (as knowledge implies truth), (2) it must be true from the relativist's point of view that John knows that q and John doesn't know that q. So (3) if relativism is correct, there must be some perspective relative to which John knows he knows that p and Hannah knows he doesn't.

But (1) isn't implied by the view sketched above (or the view developed in Richard (2004), to which Stanley objects). The clear‐headed relativist will introduce and explain what it is for a claim to be true in a context (from a perspective, for a judge), in part by relating that notion to the notion of truth full stop—that is, the relativist will give the sort of account of truth for and truth given in Chapter 4. He will then say that the claims, John knows that q and John knows that John knows that q may be true for John while the claims John doesn't know that q and Hannah knows that John doesn't know that q are true for Hannah. This is the relativist's view, and it is perfectly effable. If the relativist is right, his view is not just true for him, it's true full stop (and thus ‘true for everyone’.)

(10) If we were being prissy here, we would say that a use of a sentence in a particular context at a particular world was true if and only if the proposition expressed in that context was true at the world. I suppress the fact that the context of use may affect what a sentence says when it's irrelevant to the issue at hand.

(11) See Kaplan (1989a, Sect 3). There is an extensive literature on whether the tenses are best understood as operators (on something temporally neutral) or as quantifiers (over times, or intervals, or some other temporally ordered entity). A critical summary of some of it is given in King (2003).

I should perhaps say that I do not (contrary to what I once thought) think that the tenses are circumstance‐shifting operators. Indeed, I don't think that the modalities are such operators; I think the whole kit and caboodle are (restricted) quantifiers. For discussion see Richard (forthcoming b).

(12) Stanley (2006: 150–1).

(13) Stanley says of the operator ‘According to John’, in ‘According to John, John knows that the bank will be open’ that it doesn't have anything to do with relative truth; it ‘just mean[s] that if asked, John would accept the claim that he knows that the bank will be open’. I imagine he might say the same sort of thing of some of the phrases on this list.

It's true that on one use, ‘According to John’, ‘From John's perspective’, and other expressions of this ilk are variants of ‘In John's opinion’. But there surely is another use of such expressions on which they have a meaning along the lines of ‘It's true for John’, ‘If we accept John's standards for wealth, we must agree that’, and so forth. (Relevant here is the discussion (in Chapter 5's technical digression) of the interpretation of prepositional phrases such as ‘for John’ and ‘To Mary’ in sentences such as ‘To Mary, it's a chair’ and ‘It's tasty for John’.)

(14) Stanley, 151–2.

(15) There is, of course, a kind of negotiation we allow over these things. We allow, when it's necessary to be very precise about dates, stipulation about the exact time an event begins and ends. But it's simply not up for grabs as to whether the Revolutionary War was still going on in 1796.

(16) Vaughn played for the Red Sox from 1991 until 1998.

(17) Stanley (2004; 2006: ch. 3).

(18) Stanley argues (2006: 55–6) that ‘rich’ really does pattern with ‘could’ and not ‘know’ here. He argues that the last remark in the ‘rich’ version of this conversation can be made felicitous by ‘adding background context’—e.g. by changing A's last remark to

  1. (A:) I didn't say that he is rich; I wasn't considering that level of wealth.

But, says Stanley, this sort of contextualizing won't render the last comment in the zebra conversation felicitous.

Frankly, I don't find that this sort of ‘contextualization’ makes A's last remark any more acceptable. Indeed, if I were B and A said that, I would probably say something like

  1. (B:) Of course you said that he was rich; it's the fact that you weren't thinking clearly about what it takes to be rich that made you say the stupid thing you did, you twit.

But then I am somewhat rude.

(19) Contextual variation in the interpretation of the modals seems to me more like comparative than extensive sensitivity. (The sorts of shifts involved don't seem to involve ‘drawing a line’ on an ordering of worlds, but rather selecting a set of worlds relative to which to determine possibility.) Insofar as this is so, variants of the remarks just made apply to ‘I can't bench‐press 100 pounds, but could if I trained’. This is also relevant, of course, to Stanley's first objection to assimilating the contextual sensitivity of ‘knows’ to that of the gradable adjectives.