Concessive Knowledge Attributions and Fallibilism

17
Concessive Knowledge Attributions and Fallibilism clayton littlejohn University of Texas at San Antonio Lewis thought concessive knowledge attributions (e.g., ‘I know that Harry is a zebra, but it might be that he’s just a cleverly disguised mule’) caused serious trouble for fallibilists. As he saw it, CKAs are overt statements of the fallibilist view and they are contradictory. Dougherty and Rysiew have argued that CKAs are pragmatically defective rather than semantically defective. Stanley thinks that their pragmatic response to Lewis fails, but the fallibilist cause is not lost because Lewis was wrong about the commitments of fallibilism. There are problems with Dougherty and Rysiew’s response to Stanley and there are problems with Stanley’s response to Lewis. I’ll offer a defense of fallibilism of my own and show that falli- bilists needn’t worry about CKAs. Introduction Fallibilists think it’s possible for you to know that p is true even if your evidence for believing p does not entail that p is true. 1 Since it seems that most of what we believe about the external world is based on fallible grounds, infallibilism seems to lead to skepticism. Lewis thought that the fallibilist faced a serious difficulty because he thought concessive knowl- edge attributions (CKAs) were overt statements of the fallibilist’s view and thought they were contradictory. Consider: (1) I know that Harry is a zebra, but it might be that Harry is just a painted mule (1996: 550). 1 Dylan Sabo offered wonderfully helpful comments on an earlier draft of this that was presented at the Pacific Division meetings of the APA and I want to thank him for his efforts and thank the audience for their questions. I also want to express my gratitude to an anonymous referee for this journal for extensive written comments. Finally, thanks to Keith DeRose, Trent Dougherty, Janice Dowell, Leo Iacono, Jonathan Ichikawa, Aidan McGlynn, and Andrew Moon for helpful corre- spondence on these issues. CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 603 Philosophy and Phenomenological Research Vol. LXXXIII No. 3, November 2011 Ó 2011 Philosophy and Phenomenological Research, LLC Philosophy and Phenomenological Research

Transcript of Concessive Knowledge Attributions and Fallibilism

Page 1: Concessive Knowledge Attributions and Fallibilism

Concessive Knowledge Attributionsand Fallibilism

clayton littlejohn

University of Texas at San Antonio

Lewis thought concessive knowledge attributions (e.g., ‘I know that Harry is a

zebra, but it might be that he’s just a cleverly disguised mule’) caused serious

trouble for fallibilists. As he saw it, CKAs are overt statements of the fallibilist

view and they are contradictory. Dougherty and Rysiew have argued that CKAs

are pragmatically defective rather than semantically defective. Stanley thinks that

their pragmatic response to Lewis fails, but the fallibilist cause is not lost because

Lewis was wrong about the commitments of fallibilism. There are problems with

Dougherty and Rysiew’s response to Stanley and there are problems with Stanley’s

response to Lewis. I’ll offer a defense of fallibilism of my own and show that falli-

bilists needn’t worry about CKAs.

Introduction

Fallibilists think it’s possible for you to know that p is true even if your

evidence for believing p does not entail that p is true.1 Since it seems that

most of what we believe about the external world is based on fallible

grounds, infallibilism seems to lead to skepticism. Lewis thought that the

fallibilist faced a serious difficulty because he thought concessive knowl-

edge attributions (CKAs) were overt statements of the fallibilist’s view

and thought they were contradictory. Consider:

(1) I know that Harry is a zebra, but it might be that Harry is

just a painted mule (1996: 550).

1 Dylan Sabo offered wonderfully helpful comments on an earlier draft of this that

was presented at the Pacific Division meetings of the APA and I want to thank

him for his efforts and thank the audience for their questions. I also want to

express my gratitude to an anonymous referee for this journal for extensive written

comments. Finally, thanks to Keith DeRose, Trent Dougherty, Janice Dowell, Leo

Iacono, Jonathan Ichikawa, Aidan McGlynn, and Andrew Moon for helpful corre-

spondence on these issues.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 603

Philosophy and Phenomenological ResearchVol. LXXXIII No. 3, November 2011� 2011 Philosophy and Phenomenological Research, LLC

Philosophy andPhenomenological Research

Page 2: Concessive Knowledge Attributions and Fallibilism

Lewis thought that contextualists could be infallibilists without being

skeptics, but the rest of us non-skeptics need to defend fallibilism from

his charge that overt statements of fallibilism are contradictory. In this

paper, I want to look at two defenses of fallibilism.2 Rysiew (2001) and

Dougherty and Rysiew (2009) think Lewis is right in saying that

(1) captures the commitments of fallibilism. They think that (1) is prag-

matically inappropriate, but not semantically defective. Stanley (2005)

thinks Lewis is right in saying that (1) is contradictory, but claims that

Lewis is wrong about the fallibilist’s commitments. Stanley thinks

fallibilists do not have to say that CKAs can express true propositions.

I think there are problems with both defenses of fallibilism. After argu-

ing that neither defense succeeds, I’ll offer a sketch of a defense that

should avoid the difficulties that arise for these views.

Two Defenses of Fallibilism

According to Rysiew (2001), fallibilists should say that CKAs can

express true propositions. The reason (1) sounds contradictory, he

says, is not that it’s semantically defective but that it’s pragmatically

inappropriate to assert (1). He says that in ordinary speech, ‘It might

be that �p’ pragmatically imparts that the speaker does not know

that p is true (2001: 493). Thus, the speaker who asserts (1) pragmati-

cally imparts that she knows something she claims she does not

know.

Rysiew is right on this point. (1) is pragmatically inappropriate.

Consider:

(2) Harry is a zebra, but it might be that Harry is just a

painted mule.

This sounds contradictory. However, we know that (2) could express a

true proposition. Suppose the speaker did not know that Harry was

not a cleverly painted mule but Harry was in fact a zebra. Given some

standard assumptions about epistemic modals, (2) would express a true

2 The authors that I’ll be criticizing here (i.e., Dougherty, Rysiew, and Stanley) are

fallibilists about knowledge who defend invariantist accounts of knowledge and

evidence ascription. Like them, I think CKAs don’t force us to give up fallibilism

or invariantism about knowledge and evidence ascription to avoid skepticism. Still,

there are contextualist views worth considering that won’t be considered here

because they are outside the scope of this paper. One view that I think is attractive

combines invariantism about knowledge ascriptions with contextualism about epi-

stemic possibility statements. This view is a possibility if, as I shall argue, there’s

more to epistemic necessity than knowledge. I think the view would do a nice job

accommodating some of the data presented below and would undercut some argu-

ments for contextualism about knowledge attribution.

604 CLAYTON LITTLEJOHN

Page 3: Concessive Knowledge Attributions and Fallibilism

proposition under these conditions. Whatever account we give to

explain why (2) is pragmatically inappropriate should predict that (1) is

(at least) pragmatically inappropriate.

Rysiew gave the fallibilists what they wanted (i.e., an explanation

as to why (1) seems contradictory that does not entail that (1) is a

contradiction), but Stanley (2005) argues that he did not give them

what they needed. Rysiew didn’t explain how (1) could express a

true proposition because he offered no semantic account of epistemic

possibility statements. The problem was not just that the account

was incomplete. As Stanley noted, given a standard account of

epistemic possibility statements, Rysiew’s defense of fallibilism failed.

Consider:

(EPk) p is epistemically possible for S iff �p isn’t obviously

entailed by something S knows.

If open epistemic possibilities are due to ignorance, (1) entails:

(3) I know that Harry is a zebra, but I do not know that

Harry is not just a cleverly disguised mule.

Assuming the speaker is not exceptionally dim, it seems (3) couldn’t be

true. It follows from (�3) and (EPk) that (1) is false.

Stanley did not think that the difficulties facing fallibilism were

nearly as serious as Lewis did. Rather than try to explain how (1)

could express a true proposition, he argued that the fallibilist does not

have to say that (1) could express a true proposition.3 Remember that

the fallibilist says that it doesn’t follow from the fact that S knows p

that S’s evidence for believing p entails that p is true. Unless there is an

entailment from (1) to (�4), the fallibilist could deny (1) and say that

their view is properly expressed as follows:

(4) I know that Harry is a zebra, but my evidence for believing

that Harry is a zebra does not logically entail that Harry is

not a painted mule.

To my ears, (4) doesn’t sound the slightest bit odd.4

3 I think Stanley is right that the fallibilist does not have to say that (1) can express

a true proposition, but I think there are CKAs that fallibilists cannot deny do

express true propositions.4 Note that even if a subject’s evidence consists of all and only the propositions the

subject knows, it doesn’t follow from the fact that a subject knows p that the sub-

ject’s evidence for believing p entails p. Williamson need not deny (4).

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 605

Page 4: Concessive Knowledge Attributions and Fallibilism

Dougherty and Rysiew (2009) have responded to Stanley’s criticism

of Rysiew’s (2001) defense of fallibilism. They say that the fallibilist

has to say that CKAs such as (1) can express true propositions, so they

aren’t satisfied with Stanley’s defense of fallibilism since he concedes to

Lewis that CKAs are contradictory. To meet Stanley’s challenge, they

try to explain how (1) could express a true proposition. To do this,

they recommend replacing (EPk) with this:

(EPe) p is epistemically possible for S iff �p isn’t entailed by

S’s evidence (2009: 127).

The thought is that by replacing (EPk) with (EPe) they can say that

there is no entailment from (1) to (3). Thus, they can agree that (3) is

false but insist that (1) is true.

I don’t think that they’ve met Stanley’s challenge. They’ve only

pushed the problem back a step. There is nothing wrong with their

explanation as to why we should not assert (1), but they need to sever

the connection between (1) and (3) and they can do this only if (EPe) is

a viable alternative to (EPk). I’m not convinced that (EPe) is viable or

a real alternative.

The first problem is with (EPe) itself. Dougherty and Rysiew want

to replace (EPk) with (EPe), but, ‘It must be that p’ is true only if p is

true. That is to say, ‘must’ is factive.5 If you accept (EPk), it is obvious

how you would account for the factivity of ‘must’. A speaker truthfully

asserts, ‘It must be the case that p’ only if the speaker knows p.

(Perhaps ‘It must be that p’ says more than just that the speaker knows

p, but it seems to entail that p is the case.) The factivity of ‘must’ is

explained in terms of the factivity of ‘knows’.6 Dougherty and Rysiew

say that p can be epistemically necessary for S even if S does not know

p. Given that they deny that knowledge is necessary for epistemic

necessity, it is far from obvious how they will account for the factivity

5 Hawthorne (2004: 28) claims that ‘must’ is factive.6 Suppose the speaker asserts, ‘Cooper has been looking for clues for hours. He

must have found something by now’, it does seem that the speaker imparts some-

thing about the evidential basis of the belief Cooper has found a clue. Even if

‘must’ involves some sort of evidential meaning component, it could still be the

means by which we express that some proposition is epistemically necessary for us

(i.e., that its falsity is ruled out by our evidence or something we know). See Fintel

and Gillies (2007) who offer us this example. I know that the ball is in one of three

boxes. If I look and see that it is not in the first and then look to discover that it

is not in the second, my assertion ‘The ball must be in the third box’ does not

assert something logically weaker than the bare assertion ‘The ball is in the third

box’, but it does indicate something about the inferential nature of the belief that

the ball is in the third box.

606 CLAYTON LITTLEJOHN

Page 5: Concessive Knowledge Attributions and Fallibilism

of ‘must’ or epistemic necessity. At a minimum, they would have to

show that the falsity of a proposition is sufficient for excluding that

proposition from a subject’s evidence.7

The second problem with their proposal is that you can sever the

connection between (1) and (3) by replacing (EPk) with (EPe) only if

facts about an individual’s evidence are not constrained by facts about

what that individual knows. Suppose, for example, that you agree with

Williamson (2000) and identify a subject’s evidence with that subject’s

knowledge:

(E=K) S’s evidence includes p iff S knows p.

(EPe), (E=K), and (�3), entail (�1). If CKAs can express true propo-

sitions, we have to give up one of these claims.

To their credit, Dougherty and Rysiew (2009: 127, n. 5) acknowledge

that their solution works only if we deny (E=K), but I don’t think

they appreciate the severity of the problem they face. Bracketing the

difficult question as to whether or not evidence is propositional, there

do seem to be some plausible objections to (E=K). Consider two.

First, (E=K) implies that if one subject knows p but some other sub-

ject fails to know p for purely Gettierish reasons we can say that p is

part of the first subject’s evidence but not the second no matter how

similar these subjects and their epistemic situations might otherwise be.

That seems counterintuitive. It seems that if we send two subjects on

drives through the country on subsequent days showing the first real

barns and showing the second some real barns and some fakes, it

seems the first might know she’s seen a barn while the second might

believe that she’s seen a barn on essentially the same grounds. We

might stipulate that all of their beliefs are true and experiences are

veridical. I just don’t see that the second subject lacks evidence the first

has. It seems that some explanations as to why some subject’s belief is

Gettiered when some very similar subject’s belief is not will appeal to

7 If they say that there are no false propositions included in someone’s evidence,

we’ll see that they face additional difficulties having to do with evidence ascription

and proper concession. Whereas concession is ‘cheap’, denying that someone has

evidence is not nearly so cheap. It isn’t hard to create a context in which someone

might (properly) concede ‘It might be that there are no hands’ but it seems that in

such contexts we can still say that among the propositions included in someone’s

evidence is that they have hands. If the proposition I have hands is included in the

speaker’s evidence, it seems that if the speaker says ‘It might be that there are no

hands’, the speaker says something false on their account. What they have to do to

avoid this problem is retreat and say that propositions such as the proposition that

I have hands cannot be included in someone’s evidence. I’ll say more about why I

think this is a bad move below.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 607

Page 6: Concessive Knowledge Attributions and Fallibilism

extra-evidential considerations. Second, (E=K) implies that we can

acquire new evidence by means of competent deduction.8 While it

seems that competent deduction allows us to extend our knowledge

and transmit old justification to support new beliefs, it doesn’t seem

implausible to suggest that it does this without enabling the subject to

acquire new evidence in the process (e.g., it seems implausible to think

that extending our knowledge to include the deductive consequences of

propositions known on the basis of inductive inference the deductive

consequences of propositions known on the basis of induction thereby

become part of our evidence).

These objections might give us good reason to reject (E=K), but

they are perfectly consistent with the claim that immediate or non-

inferential knowledge of some proposition’s truth suffices for that prop-

osition’s inclusion in your evidence:

(IKSE) If S has non-inferential knowledge that p is the case, p

is included in S’s evidence.9

I think (IKSE) is attractive. It avoids the difficulties that we saw arise

for (E=K) and it seems intuitive. Those who reject (IKSE) have to show

that someone could know p non-inferentially and still not overcome

whatever obstacles stand in our way when we try to acquire new

evidence. If we think of a subject’s evidence as consisting of things the

subject can properly treat as reasons for belief, denying (IKSE) amounts

to denying that someone can properly treat that which they know with-

out inference as a reason for believing further things.10 On the assump-

tion that S knows p only if S justifiably believes p, (IJSE) entails (IKSE).

(IJSE) If S is non-inferentially justified in believing p, p is part

of S’s evidence.

8 Bird (2004) defends the idea that there can be inferential evidence. It suffices for

my point that there are cases where S can come to know p via inference without p

thereby getting into S’s evidence. If, however, the reader is disinclined to revise

(E=K) in the way I’m suggesting, this is bad news for Dougherty and Rysiew

because that means they could not say both that we have knowledge of the exter-

nal world and that such knowledge figures in the formulation of a CKA.9 For the purposes of this discussion, we don’t need an account of evidence. Any

account of evidence that incorporates (IKSE) will cause problems for Dougherty

and Rysiew. This is not the place to defend the account, but my own preference is

for the following view: p is part of S’s evidence iff p is the case and S is non-infer-

entially justified in believing that p is the case.10 If you think of evidence as something like a reason you have to rely on in trying to

justify further beliefs, you shouldn’t deny (IKSE). It seems odd to deny (IKSE)

even if you deny this.

608 CLAYTON LITTLEJOHN

Page 7: Concessive Knowledge Attributions and Fallibilism

Like (IKSE), I think (IJSE) is rather attractive. Again, given the gloss

on evidence above, those who deny (IKSE) and so deny (IJSE) have to

say that it can be epistemically improper for you to treat p as a reason

for believing further things even if they concede that you’re epistemi-

cally justified in believing p.

To see why (IKSE) and (IJSE) creates difficulties for Dougherty

and Rysiew, let p be the proposition that S has hands. Suppose S

believes this, in part, because of her veridical experiences of her hands.

Suppose that S’s friend asks her to consider various skeptical scenar-

ios. It seems she could speak truthfully by saying, ‘It might be that

there are no hands, only hand appearances’. If S could truthfully say

‘It might be that there are no hands’, it follows from (EPe) that her

evidence does not include the proposition that she has hands. It fol-

lows from the fact that this proposition is not included in her evidence

and (IKSE) that she cannot know non-inferentially that she has

hands. According to (IJSE), would also have to deny that she can be

justified non-inferentially in believing that she has hands if we said

this. I’d prefer to retain (IKSE) and (IJSE) if possible and I’d prefer

not to be forced to say that the scope of non-inferential knowledge

and justification is so limited that it excludes propositions about the

external world.11 I think Maher’s (1996) suggestion that external world

propositions known immediately by observation is included in your

evidence is an attractive one. We’ll see below that to accommodate

some data about proper concession, we’ll see that Dougherty and

Rysiew’s fallibilist has to deny this view as well as deny (IKSE) and

(IJSE) if this fallibilist is going to say that the scope of non-inferential

knowledge is broad enough to encompass direct knowledge of the

external world.

Essentially the same problem arises for Stanley. Recall that Stanley’s

fallibilist thinks that CKAs such as (1) do not express true propositions.

He does think that (4) is true. Consider two propositions: (i) that S has

hands; (ii) that there is at least one hand. It seems that having enter-

tained the possibility that she might be hallucinating, S might properly

say, ‘It might be that there are no hands, only hand appearances’. If that

11 As an anonymous referee noted, not everyone holds the view that our knowledge of

the external world is non-inferential. That is true and I do not wish to say here that

this sort of view is mistaken. Many contemporary foundationalists (e.g., Feldman

(2004) and Pryor (2000)) are willing to say that our evidence includes propositions

pertaining to things external to us (e.g., propositions ascribing shape and color

properties to objects that are the semantic values of perceptual demonstratives) and

my point is only that it would be better if we had an account of epistemic possibility

and necessity that did not force us to abandon such views in favor of a more classical

foundationalism that takes our evidence to consist of propositions having to do with

sense-data.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 609

Page 8: Concessive Knowledge Attributions and Fallibilism

were to express a true proposition, it follows from this that Stanley’s

fallibilist would have to concede that she has no evidence that would

obviously entail that there is at least one hand. This, in turn, suggests

that Stanley’s fallibilist would concede that her evidence does not

include the proposition that she has hands. It follows from this conces-

sion and (IKSE) that Stanley’s fallibilist cannot claim to have non-

inferential knowledge that she has hands. It follows from (IJSE) that

Stanley’s fallibilist cannot claim to have non-inferential justification that

she has hands.

In brief, the problem is this. It seems that for any proposition, p,

such that p is a proposition about the external world (i.e., one whose

truth-value isn’t determined solely by facts about a subject’s non-factive

mental state), the introduction of a skeptical hypothesis into discussion

can make it proper for a subject that we would otherwise credit with

non-inferential knowledge that p to concede that it might be that p is

mistaken. On Dougherty and Rysiew’s view as well as Stanley’s view,

we have to say that such a subject’s concession is proper only if this

subject has no evidence that obviously entails that this proposition is

true. Barring any contextualist maneuvering on their part about

evidence ascription, this forces them to adopt the view that we

cannot have these propositions about the external world as part of our

evidence.12 Either they have to say that such propositions cannot be

known non-inferentially because some sort of classical foundationalism

is true or that we cannot know such external world propositions non-

inferentially because the skeptic is right. Myself, I think that we know

lots about the external world non-inferentially, our evidence includes

loads of propositions about the external world (e.g., propositions we

know non-inferentially to be true on the basis of veridical experience

taken at face value), and we shouldn’t be forced to retract either of

these claims just because it is easy to get concessions out of us. I think

the fallibilists might reasonably worry that the defenses offered thus far

on their behalf concede too much to accommodate intuitions about

proper concession of possible error.

A Repair

Where do we go from here? A promising strategy for the fallibilist is to

revise (EPk) as follows:

12 An anonymous referee asked whether a contextualist view about evidence ascription

might avoid these worries and I think that this contextualist view could avoid these

worries. I don’t have any problem with such a contextualist view, but I don’t think

it’s needed to deal with the problems that arise for Dougherty and Rysiew or the

problems that arise for Stanley.

610 CLAYTON LITTLEJOHN

Page 9: Concessive Knowledge Attributions and Fallibilism

(EPc) p is epistemically possible for S iff �p isn’t obviously

entailed by something S knows with certainty.13

According to (EPc), S’s concession, ‘It might be that �p’, is an

acknowledgement that she either doesn’t know p or that p isn’t certain.

If we replace (EPk) with (EPc), we can deal with the problem that

arose for Dougherty and Rysiew’s account. To sever the connection

between (1) and (3) and explain how (1) could express a true proposi-

tion, they had to say that it is harder to acquire evidence than it is to

acquire immediate or non-inferential knowledge. This is a problem for

those of us who think that nothing beyond non-inferential knowledge

of p’s truth is needed for p’s inclusion in our evidence. Instead, we can

say that more is required for epistemic necessity than is required for

knowledge. Thus, less is required for open epistemic possibilities than

ignorance or the lack of entailing evidence.14 If more is required for epi-

stemic necessity than knowledge, CKAs aren’t a problem for fallibilist.

If p is epistemically possible for S if either S doesn’t know p or X, the

intuition that S is proper to concede ‘It might be that �p’ only com-

mits us to saying that either S doesn’t know p or X. I think a plausible

candidate for ‘X’ is certainty, but the strategy requires only identifying

some condition that is not a necessary condition for knowledge that is

a plausible necessary condition for epistemic necessity.

We have good reason to think that (EPk) is really only a first

approximation towards an account of epistemic possibility. First, think

about cases of inductive knowledge. It seems odd to think that you

only have knowledge of future events when it is not epistemically possi-

ble that these events do not occur. Myself, I don’t doubt that our

13 We can distinguish statements of personal certainty (‘I’m certain that p’) from

statements of impersonal certainty (‘p is certain’) where the former seems to be

about the kind of confidence S has and the second has to do with the strength of

the speaker’s epistemic position. I’m assuming that both kinds of certainty are rele-

vant to assessing statements about epistemic possibility. DeRose claims that (1998:

69) that ‘It is certain that p’ is the dual of ‘It’s possible that p’ and suggests that

knowledge requires both kinds of certainty (2009: 186). I don’t think he’d be

opposed to (EPc) because my guess is that he’d not see much difference between

(EPk) and (EPc). An argument of Stanley’s (2008) discussed below suggests that

certainty isn’t required for knowledge. Below, I try to show that we can build on

Stanley’s observations to motivate (EPc). Stanley is sympathetic to a contextualist

treatment of ‘certain’. Combine his contextualist account of ‘certain’ with (EPc)

and deny (EPk), and we can concede to the contextualists that contextualism about

epistemic possibility statements is correct while remaining agnostic with respect to

their claims about knowledge attribution. In this way, we can try to undercut some

arguments for contextualism about ‘knows’.14 This allows us to keep our (IKSE) and (IJSE) while recognizing that a subject

could properly concede that pretty much any belief about the external world is a

belief she might be mistaken in holding.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 611

Page 10: Concessive Knowledge Attributions and Fallibilism

beliefs about the future constitute knowledge. I doubt that it would be

correct to say that it isn’t epistemically possible that these beliefs are

mistaken.

Second, think about conversations where skeptical hypotheses are

introduced. In such contexts, it seems proper to concede that we might

be mistaken in just about any belief about the external world. Now,

suppose that knowledge is necessary for warranted assertion and that

concessions (e.g., ‘It might be that I’m a BIV’) are really assertions. It

seems that given these assumptions and (EPk), the propriety of the

concession would depend upon whether the speaker knew herself to be

ignorant. But, it seems harder to know that you don’t know than it is

to know that it’s proper to concede that you might be mistaken. Given

(EPk), to assert knowingly that you might be mistaken, you either

know that you don’t believe p, that your belief about p is mistaken,

that the justification you have for your belief is insufficient, or that you

are in some sort of Gettier case. I doubt that you know one of these to

be true whenever you know that it’s proper to concede that you might

be mistaken. Thus, you either should think that concessions aren’t

really assertions, deny that knowledge is the norm of assertion, or say

(as I do) that in conceding that you might be mistaken you might only

be conceding that you are not completely certain.

Third, consider the contrast between (1) embedded and similar

embedded statements:

(5) I believe that I know that Harry is a zebra, but it might be

that Harry is just a painted mule.

(6) I believe that I know that Harry is a zebra, but he isn’t.

(7) I believe that I know that Harry is a zebra, but there’s no

reason for me to believe that he is.

(8) I believe that I know that Harry is a zebra, but I don’t

believe that Harry is a zebra.

If after seeing the zebra you raise the possibility that the zookeepers

painted a mule and put it in the zebra cage, it seems that I could speak

truthfully if I utter (5). When we embed these other claims where the

second conjunct denies that a condition necessary for knowledge

obtains, the embedding doesn’t seem to wash away the sin of asserting

(6), (7), or (8). If we take the effect of embedding these claims to be

that the speaker thinks that it is not altogether unlikely that the embed-

ded claims are true, the fact that we find (5) to be acceptable is some

612 CLAYTON LITTLEJOHN

Page 11: Concessive Knowledge Attributions and Fallibilism

indication that the defectiveness of CKAs is not due to the fact that

they express obvious falsehoods but rather something else. Given that

we know that (1) is at least pragmatically defective, we can help our-

selves to whatever pragmatic explanation strikes us as promising as to

why (1) shouldn’t be asserted.15

Fourth, consider the sort of cases that led Radford (1966) to say

that knowledge doesn’t require belief. Pressed for answers on a quiz

show, a contestant consistently gives the right answers and is pleasantly

surprised to discover that the answers she’s giving are correct. It seems

that as she’s doing this she might rightly think to herself that she might

be mistaken while someone at home might be right to say that she knew

the answers to the questions. I think one lesson from such cases is this.

Whereas it is relatively easy to get yourself in a position to properly

concede, ‘It might be that �p’, it is relatively harder to get yourself into a

position to properly assert that you don’t know that p. This would be

puzzling if (EPk) were true, but to be expected if (EPc) is true.

Here is a fifth and final piece of evidence to consider. As Stanley

notes, redundant conjunctions provide evidence for entailment.16 When

there is an obvious entailment between two propositions, it will seem

odd to follow an assertion of one by one obviously entailed by the first.

Consider:

(9) I know that Bill came to the party. In fact, he did.

(10) I know that Bill came to the party. Indeed, I have good

reason to think he did.

(11) I know that Bill came to the party. Not only that, I

believe he came to the party.

Contrast these with:

(12) I know that Bill came to the party. In fact, I’m certain

that he did.

Whereas (9)-(11) seem odd, (12) does not. Stanley’s hypothesis is that

ascriptions of knowledge entail ascriptions of beliefs, good reasons, and

claims about the truth of propositions said to be known whereas an

15 Dougherty and Rysiew (2009: 129) suggest that when you say, ‘It might be that

�p’ you pragmatically imply that there are genuine reasons for doubt that make it

inappropriate to self-ascribe knowledge.16 Stanley (2008: 38).

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 613

Page 12: Concessive Knowledge Attributions and Fallibilism

ascription of knowledge merely pragmatically imparts that the knower

is certain of the truth of what is known. The evidence is that entail-

ments cannot be reinforced, but we can reinforce information that

is pragmatically imparted.17 Of course, if knowledge doesn’t require

certainty, that won’t be news to most of us.

Now, consider:

(13) Bill must have come to the party. In fact, I’m certain that

he did.

(14) Bill must have come to the party. In fact, it is certain that

he did.

(15) It is certain that Bill was at the party. Indeed, Bill must

have been there.

These seem redundant in the way that (9)-(11) are. If they are, then

there are obvious entailments between ‘must’ and ‘certain’. As there is

no obvious entailment from ‘knows’ to ‘certain’, there is no obvious

entailment from ‘knows’ to ‘must’. The absence of evidence of an

entailment is not particularly strong evidence of the absence of

an entailment. It is suggestive, however. Why is it that we don’t find

the evidence for the entailment from ‘knows’ to ‘must’? Isn’t this pre-

cisely where we would expect to find it?

There are two objections to (EPc) to consider.18 Suppose a military

instructor gives this advice to soldiers in training:

(16) Before you walk into an area where there are lots of high

trees, if there might be snipers hiding in the branches use

your flamethrowers to clear away the foliage.19

Perhaps the objection is that it is not plausible to suggest that these

soldiers are being advised to fire if they don’t know with certainty that

the enemy is not hiding in the foliage. If that’s the worry, I’ll dig in and

say that I don’t think it’s implausible. Suppose Brown didn’t use his

flamethrower to clear away the foliage and a sniper picked off a member

of his platoon. I can easily imagine a commanding officer chastising him

by saying that he should have fired into the foliage since he wasn’t certain

that there weren’t snipers hidden there. Now, suppose there wasn’t a

17 Stanley (2008: 38). He credits this point to Sadock (1978).18 I owe both of these objections to an anonymous referee.19 Example taken from Egan, Hawthorne, and Weatherson (2005).

614 CLAYTON LITTLEJOHN

Page 13: Concessive Knowledge Attributions and Fallibilism

sniper. If Brown hadn’t fired, I can imagine a commanding officer asking

him why he didn’t use his flamethrower if he wasn’t certain that there

wasn’t a sniper hidden in the foliage. I can imagine Brown trying to

defend himself by saying that he had good reason to think that there were

no enemies in the foliage and adding that there turned out to be none

hidden there. I can imagine his commanding officer saying that this was

neither here nor there because he wasn’t sure that there weren’t snipers

hidden in the foliage. If he later said to his mates that he knew that there

was no one there, I can imagine them saying that that was neither here

nor there because they were told to fire unless they were certain and that

he knows he wasn’t certain at the time.

There is another aspect of this example that calls for comment. I’ve

suggested that there’s some evidence that supports the hypothesis that

epistemic necessity should be understood in terms of what’s known

with certainty rather than simply in terms of what’s known. Because

the focus has been on CKAs, I didn’t have to worry about cases that

suggest that the proper assessment of a claim involving an epistemic

modal can depend upon the epistemic position of groups of individuals.

As formulated, (EPe) says that if we’re talking about what’s epistemi-

cally possible for S, we should focus on S’s evidence, (EPk) says that

we should focus on what S knows, and (EPc) says that we should focus

on what S knows with certainty. This case suggests these views each

need refinement. If the instructor happened to know with certainty that

there wasn’t a sniper in some tree and the soldier had no idea whether

there was a sniper in the trees, surely the instructor intended to convey

that the soldier ought to fire into the trees. The instructor’s epistemic

state is not what matters for assessing (16) and so we need to modify

our account of epistemic possibility accordingly. One way to do this is

to understand epistemic possibility relative to the epistemic positions

of members of some contextually determined group and relative to

contextually salient ways of settling some question:

(EPc*) S’s assertion, ‘It is possible that p’ is true iff (i) no

member of the contextually determined group knows

for certain something that entails �p in a way that is

obvious and (ii) there is no relevant way that members

of this group to remedy this.20

The example above suggests that the group needn’t involve the speaker

and (EPc*) seems to handle the case nicely.

20 Essentially, this is DeRose’s (1991: 594) view with a slight twist. He talks about

what’s known, I’ve shifted to talking about what’s known with certainty.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 615

Page 14: Concessive Knowledge Attributions and Fallibilism

This sort of view might help us make sense of the case above, but it

is thought to have trouble handling cases like this one.21 A captain says

to his crew at noon:

(17) The treasure might be right underneath us.

After hours of dives that turned up nothing, the captain checks his

charts again and discovers that his ship is not where he had intended it

to be when he used those charts to plot their course this morning.

He sees that the map says that the treasure is miles from their current

location. The captain then says:

(18) I was mistaken, the treasure isn’t here.

The problem that this case presents for (EPe), (EPk), and (EPc) is this.

If the captain speaks correctly in asserting (18), it’s natural to take this

as asserting that (17) is false. If (17) is false and it’s false that it’s

epistemically possible (for the captain) that the treasure is beneath

them, then it’s epistemically necessary that it’s not beneath them.

I don’t think we want to say that the captain’s evidence entailed that

the treasure was elsewhere or that the captain knew that the treasure

was elsewhere. To deal with this case, we could put clause (ii) from

(EPc*) to work, but it seems to be an inessential feature of the case

that there was something available that could have shown them that

the treasure wasn’t where they were looking for it. You get similar

problems if we imagine a stowaway hiding in a barrel who knows

where the treasure is, knows that the ship is nowhere near it, and hears

the captain assert (17). Our stowaway mutters to himself:

(19) He’s wrong. The treasure is nowhere near here.

That seems right.22 Our stowaway isn’t a way for the relevant group to

determine the location of the treasure. Our stowaway also doesn’t seem

to be a member of any contextually determined group. So, clause (i) in

21 The example is taken from Hacking (1967). In discussing this case, DeRose (1991:

587) thinks that the captain says something false if the captain says that it is

possible that we will find the treasure here, but notes that he doesn’t think that this

judgment is ‘‘obviously correct’’. Like DeRose, my intuitions about this case are

not that clear. I’m a bit more inclined to think that the captain’s judgment was

mistaken if the captain says ‘It is possible that we shall find the treasure here’ (the

wording of the original example), but less certain what to say when we use ‘might’

in place of ‘possible’.22 Our stowaway is an eavesdropper.

616 CLAYTON LITTLEJOHN

Page 15: Concessive Knowledge Attributions and Fallibilism

(EPc*) does some work in the sniper case, but it doesn’t seem to do

much work in this case.

There are two strategies for dealing with the stowaway. First, we

could go relativist.23 On one version of the relativist view, ‘It might be

the case that p’ is true relative to a centered world <w, t, i> iff it is

compatible with everything that is within i’s epistemic reach at t in w

that p. Let ‘p’ be the proposition that the treasure is beneath the ship.

At the time of (17)’s utterance, p was compatible with everything

within the relevant individual’s epistemic reach (i.e., the captain’s epi-

stemic reach initially) but at the time of (18)’s utterance, p was not

compatible with everything within the relevant individual’s epistemic

reach (i.e., the captain’s epistemic reach at the later time). At the time

of (17)’s utterance, p was not compatible with everything within the rel-

evant individual’s epistemic reach (i.e., the stowaway). Our relativist

makes use of the notion of ‘epistemic reach’ and that can be cashed

out in different ways (i.e., what the individual is in a position to know,

what’s consistent with that individual’s evidence, what the individual

can know for certain). In arguing that (EPc) is preferable to (EPk), I’ve

already explained why I think this notion of ‘epistemic reach’ is better

understood in terms of what’s known for certain rather than just

what’s known.

Second, it’s possible to handle this case with something less exotic

than relativism. It’s hard to resist the relativist view if we’re operating

under the assumption that the propriety of (18) and (19) is due to the

fact that the captain spoke falsely in uttering (17). An alternative

construal of what’s happening is that the speakers who assert (18)

and (19) are negating the prejacent (i.e., that the treasure is right

beneath the ship) rather than asserting the negation of (17).24 If that’s

what is happening in this case, we don’t need to move to the relativ-

ist view. We can see that (EPc*) already has the resources for dealing

with it. The question as to whether we should prefer (EPc*) to some

alternative relativist view is a difficult one that I think needn’t be set-

tled here. For whichever view we prefer, we’ll need to say something

about what epistemic reach amounts to or which aspect of a relevant

group’s epistemic position matters to assessing a statement about

epistemic possibility. The success of the argument that purports to

show that there’s more to epistemic necessity than just knowledge

doesn’t depend upon taking a position in this debate between the

relativist and the relativist’s critics. To neutralize the threat CKAs

23 For development and defense of the view, see Egan (2007).24 See von Fintel and Gillies (Forthcoming) who suggest that this is one way to deal

with eavesdroppers without having to adopt the relativist framework.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 617

Page 16: Concessive Knowledge Attributions and Fallibilism

pose to fallibilism, all we have to show is that there’s more to episte-

mic necessity than just knowledge and we can show that this is

so even if the debate between the relativist and the relativist’s critics

continues.

Conclusion

The fallibilist should occupy a middle ground between Stanley’s view

and Dougherty and Rysiew’s view. Stanley is right that the fallibilist

doesn’t have to say that CKAs of the form ‘I know p, but it might be

that �p’ could express true propositions when the subject doesn’t have

evidence that entails, inter alia, that p is the case. However, as there are

cases where it seems that we have such evidence (e.g., where one sees

that one has hands and infers that there is at least one hand) where it

also seems appropriate to concede that one might be mistaken anyway

(e.g., where one considers skeptical hypotheses). So, I think he’s mis-

taken when he says that CKAs like (1) never express true propositions

and mistaken when he says that this is the line that the fallibilist must

take. Dougherty and Rysiew are right that the fallibilist isn’t necessar-

ily contradicting herself when she says, ‘I know p but it might be that

�p’ but they are wrong to suggest that the fallibilist would be con-

tradicting herself if she possessed evidence that entailed p. They can

only say that it is proper to concede the possibility of error in a wide

range of circumstances by saying that we have far less evidence than

we antecedently would have thought. We need to revise (EPk) to say

that more is required for epistemic necessity than knowledge. Maybe

(EPc) or (EPc*) doesn’t get every case right, but it’s a step towards

a view that hopefully does. Simply by making a move towards that

view and away from (EPk) and (EPe), we can see that CKAs are not a

serious problem for fallibilists.

References

Bird, A. 2004. Is Evidence Non-Inferential? The Philosophical Quarterly

54: 252–65.

DeRose, K. 1991. Epistemic Possibilities. Philosophical Review 100:

581–605.

—— 1998. Simple Mights, Indicative Possibilities, and the Open

Future. The Philosophical Quarterly 48: 67–82.

—— 2009. The Case for Contextualism. New York: Oxford University

Press.

Dougherty, T. and P. Rysiew. 2009. Fallibilism, Epistemic Possibility,

and Concessive Knowledge Attributions. Philosophy and Phenome-

nological Research 78.1: 123–32.

618 CLAYTON LITTLEJOHN

Page 17: Concessive Knowledge Attributions and Fallibilism

Egan, A. 2007. Epistemic Modals, Relativism, and Assertion. Philo-

sophical Studies 133: 1–22.

Egan, A., J. Hawthorne and B. Weatherson 2005. Epistemic Modals in

Context. In G. Preyer and G. Peter (ed.), Contextualism in Philoso-

phy. New York: Oxford University Press.

Fantl, J. and M. McGrath 2009. Knowledge in an Uncertain World.

New York: Oxford University Press.

Feldman, R. 2004. Foundational Beliefs and Empirical Possibilities.

Philosophical Issues 14: 132–48.

von Fintel, K. and A. Gillies 2007. An Opinionated Guide to Epistemic

Modality. In T. Gendler and J. Hawthorne (ed.), Oxford Studies in

Epistemology, Volume 2. New York: Oxford University Press.

—— Forthcoming. Might Made Right. In B. Weatherson and A. Egan

(ed.), Epistemic Modality. New York: Oxford University Press.

Hacking, I. 1967. Possibility. Philosophical Review 76: 143–68.

Hawthorne, J. 2004. Knowledge and Lotteries. New York: Oxford

University Press.

Lewis, D. 1996. Elusive Knowledge. Australasian Journal of Philosophy

74: 549–567.

MacFarlane, J. 2005. The Assessment Sensitivity of Knowledge Attri-

butions. In T. Gendler and J. Hawthorne (ed.), Oxford Studies in

Epistemology, Volume 1. New York:Oxford University Press.

Maher, P. 1996. Subjective and Objective Confirmation. Philosophy of

Science 63: 149–74.

Pryor, J. 2000. The Skeptic and the Dogmatist. Nous 34: 517–49.

Radford, C. 1966. Knowledge—By Examples. Analysis 27.1: 1–11.

Rysiew, P. 2001. The Context-Sensitivity of Knowledge Attributions.

Nous 35.4: 477–514.

Sadock, J. 1978. On Testing Conversational Implicature. Syntax and

Semantics 9: Pragmatics: 281–97.

Stanley, J. 2005. Fallibilism and Concessive Knowledge Attributions.

Analysis 65.2: 126–131.

—— 2008. Knowledge and Certainty. Philosophical Issues 18: 33–55.

Swanson, E. 2005. Something Might Might Mean. MS.

Williamson, T. 2000. Knowledge and its Limits. Oxford University

Press.

CONCESSIVE KNOWLEDGE ATTRIBUTIONS AND FALLIBILISM 619