On the Pragmatic Explanation of Concessive Knowledge Attributions

The Southern Journal of Philosophy (2009) Vol. XLVII

On the Pragmatic Explanation of Concessive Knowledge Attributions Hagit Benbaji Ben-Gurion University, Israel Abstract On Lewis’s reading, fallibilism is the contradictory view that it is possible that S knows that p, even though S cannot eliminate some remote scenarios in which not-p. The pragmatic strategy is to make the alleged contradiction a mere pragmatic implicature, which is explained by false conversational expectations. I argue that the pragmatic strategy fails.

Consider the following claim: (1) I know that Harry is a zebra, but it’s possible that Harry is a painted mule. This concessive knowledge attribution 1 sounds contradictory. 2 Yet, David Lewis argues, it follows straightforwardly from fallibilism about knowledge. On Lewis’s reading, fallibilism is the view that it is possible that S knows that p, even though S cannot eliminate some remote scenarios in which not-p. Lewis concludes that fallibilism is contradictory, even “mad”—S alleged to know, yet has not ruled out the possibility of error. “Even if you’ve numbed your ears, doesn’t this overt, explicit fallibilism still sound wrong?” (Lewis 1996, 420).3 Some fallibilists deny that concessive knowledge attributions sound odd, let alone contradictory, maintaining that they simply

Hagit Benbaji is Assistant Professor of Philosophy at Ben-Gurion University of the Negev, Israel. Her research and teaching interests include topics in philosophy of mind, metaphysics, and philosophy of perception. Among her publications are papers in Synthese (“Constitution and the Explanatory Gap”), in Canadian Journal of Philosophy (“Token Monism, Event Dualism and Overdetermination”), and in European Journal of Philosophy (“Is Thomas Reid a Direct Realist about Perception?”).

225

225-238Benbaji2.pm

225

9/9/09, 11:08 AM

Hagit Benbaji

reflect the fact that knowledge does not require certainty. Spoken with the right intonation, they claim, such statements can even come across as perfectly reasonable. Accordingly, Nozick has no qualms accepting the claim that someone can know that Harry is a zebra but does not know that Harry is not a painted mule. Nevertheless, assuming, as I do in this paper, that concessive knowledge attributions seem contradictory poses a crucial challenge to fallibilism: How can one be a fallibilist without contradiction? The semantic strategy agrees with Lewis that concessive knowledge attributions are contradictory but dismisses the challenge altogether, by denying that they follow from fallibilism. The reason being, as Jason Stanley (2005) explains, that fallibilism is a thesis about evidence—namely, that evidence for a knowledge claim need not necessarily imply what is known—whereas (1) has nothing to do with evidence. Yet even Stanley must admit that fallibilism implies some statements of the form of (1), such as, “I know that Harry is a zebra, but I’m not certain that Harry is not a painted mule.”4 That is, even holders of the semantic strategy must numb their ears at some point. Since some concession of the form of (1), or a very similar one, does follow from fallibilism, it seems better to deny that any of them are contradictory. That is why a pragmatic strategy has been advanced recently. Trent Dougherty and Patrick Rysiew (2009) 5 agree with Lewis that concessive knowledge attributions follow from fallibilism but deny that they involve contradictions. Their strategy is to make the alleged contradiction a mere pragmatic implicature, which is explained by false conversational expectations. Thus, the content of concessive knowledge attributions is not contradictory; on the contrary, it expresses precisely the truth of fallibilism, yet the fact that they usually sound odd does not raise any challenge to fallibilism. The purpose of this discussion is to refute the pragmatic strategy. In what follows, I show why pragmatists believe that utterances of (1) follow from fallibilism (section 1), how they explain their odd-sounding nature (section 2), why this explanation fails (section 3), and why it must fail (section 4).

1. Why Utterances of (1) follow from Fallibilism According to the received view of fallibilism, it is the thesis that evidence for a knowledge claim need not imply what is known: One can know that p though one’s evidence for knowing that p does not entail that p.6 Why does (1) follow from this “near universal” understanding of fallibilism? The answer depends on how one understands the notion of “possible” in (1). It seems uncontroversial that “possible” in (1) denotes epistemic possibility, that is, in contrast to a logical or metaphysical possibility, whether or not a certain proposition is 226

225-238Benbaji2.pm

226

9/9/09, 11:08 AM

On the Pragmatic Explanation of Concessive Knowledge Attributions

an epistemic possibility depends on the epistemic situation of the cognizer at that time. More specifically, it depends on what the person knows, believes, or has evidence for. Here are the three most obvious candidates for defining “epistemic possibility”: we can define it to be any possibility that is consistent with what the subject believes, or with what she knows, or with what she has evidence for. All three work equally well in many intuitive cases of epistemic possibilities, such as the following example from De Rose (1991, 3). John has some symptoms indicative of cancer. The doctor tells him that there is a test such that if its result is negative, cancer is ruled out; if positive, then John may or may not have cancer. The test is done, the results are negative, and the doctor says to a colleague, “It’s not possible that John has cancer.” At the same time, John’s wife (Jane) says to her brother, “It’s possible that John has cancer; they’ve run a test on him that may rule out cancer, but I don’t know whether there are any results yet.” Both the doctor and Jane, De Rose argues, speak truly. 7 What the doctor believes, knows, or has evidence for is not consistent with the claim that John has cancer, which is why the claim that John has cancer is not a claim it is possible for him to uphold. What Jane believes, knows, or has evidence for at that time is consistent with the claim that John has cancer and is thus an epistemic possibility for her. But of the three, which epistemic status provides the best account of such possibility? That we cannot simply relativize the notion of epistemic possibility to that of belief follows immediately from the existence of beliefs that are less than fully confident. I believe that Harry is a zebra, Harry seems to be a zebra, but I’m not sure about it. I’ve heard something about the manager of this zoo, who from time to time enjoys removing the zebra from its enclosure and replacing it with a painted mule. There is no oddity in claiming: “I believe that Harry is a zebra, but it’s possible that Harry is a painted mule.” This might suggest that an epistemic possibility is one that is inconsistent with beliefs that are certain, or known to be true. Thus, a knowledge-based account suggests itself. After all, when I claim to know, as Ayer says, “I claim the right to be sure” (quoted in Williams 2001, 54). Whenever I justifiably express certainty, a knowledge-claim can be substituted for it. There do not seem to be circumstances under which an attribution of knowledge that p would be true, yet the possibility of not-p could be upheld by the attributer or knower. Not surprisingly, then, the standard account of epistemic possibility is formulated in terms of knowledge: (Knowledge)

It is possibleA that p is true if and only if what A knows does not, in a manner that is obvious to A, entail not-p. (Stanley 2005, 128)8 227

225-238Benbaji2.pm

227

9/9/09, 11:08 AM

Hagit Benbaji

Given the knowledge-based account, (1) is equivalent to (1*): (1*)

I know that Harry is a zebra, but what I know does not entail, in a manner that is obvious to me, that Harry is not a painted mule.

This is plainly contradictory.9 But it does not follow from fallibilism; fallibilism merely says that “one can know that p, on the basis of evidence that is logically consistent with not-p” (2005, 128). Dougherty and Rysiew (2009) resist this account of epistemic possibility, for they believe that (1) is not contradictory. Some utterances of (1) are perfectly in order, so the knowledge-based account imputes to actual utterers of (1) “any unexplained gross failures of intelligence or linguistic competence.” Instead, they accept Lewis’s original intuition that “S’s epistemic possibilities are just those possibilities that are uneliminated by S’s evidence…. [T]he uneliminated possibilities are those in which the subject’s entire perceptual experience and memory are just as they actually are” (1996, 552–53). Accordingly, Dougherty and Rysiew suggest the following account of epistemic possibility:10 (Evidence) It is possibleA that p is true if and only if A’s evidence for p does not, in a manner that is obvious to A, entail not-p.11 On the evidence-based account of epistemic possibility, knowledge leaves some epistemic possibilities open. When people utter concessive knowledge attributions they are rightly expressing the fact that knowledge is compatible with evidence that doesn’t entail what’s believed; (1) perfectly expresses the truth of fallibilism, that is, it says that my evidence for knowing that Harry is a zebra is consistent with its not being a zebra, that is, with its being a painted mule.

2. Why Utterances of (1) Sound Odd Let us assume, with the pragmatic account, that (1) may express a truth and, moreover, that it may just express the truth of fallibilism. Why, then, does it sound odd? The answer to this question is what makes the strategy a version of pragmatism. Since (1) is not contradictory but rather the opposite, as it may express the truth of fallibilism, the only way to explain away the oddity of (1) given the evidence-based account of epistemic possibility is to explain it away pragmatically.12 Pragmatists rely on Grice’s (1989) Co-operative Principle (CP)—make your conversational contributions maximally relevantly informative. In everyday speech, “It’s possible that q” usually serves (pragmatically) to imply that the speaker does not 228

225-238Benbaji2.pm

228

9/9/09, 11:08 AM

On the Pragmatic Explanation of Concessive Knowledge Attributions

know whether or not q is true and, therefore, does not know that not-q. Here’s how this works: if … you (think you) know that not-q or that you know that q, then other things being equal you shouldn’t say (merely) “It’s possible that q.” For to do so would be to violate the maxim of Quantity—you’d be giving, and represent yourself as having, less information than you in fact (think you) do. (Rysiew 2001, 493)

When the speaker claims, “It’s possible that Harry is a painted mule,” the hearer infers (pragmatically) that the speaker doesn’t know that Harry is a zebra (for otherwise, the speaker would surely have said so). But the antecedent of (1) states that the opposite is, in fact, the case. Hence someone who utters (1) seems to be contradicting herself.13 This explanation cannot be the whole story. The strategy is based on the claim that when I say, “I know that p,” I do not imply that I can rule out all not-p possibilities.14 I merely imply that I can rule out the relevant ones. The relevant not-p possibilities are fixed by what we (normal) humans take to be the likely counterpossibilities to what the subject is said to know. This account of relevant alternatives is fairly invariant over contexts; what is considered to be relevant does not change with the interests of the knower (Rysiew 2001, 488). In the case of (1), all speakers would agree that the possibility that Harry is a painted mule is irrelevant. All speakers, then, would agree that I can know that Harry is a zebra, despite not being able to rule out the possibility that Harry is a painted mule, so (1) is true. Why is it, then, that when the speaker says “it’s possible that Harry is a painted mule,” the hearer infers that he does not know that Harry is a zebra? If it is clear to all speakers that “Harry is a painted mule” is an irrelevant possibility, why is it that uttering it pragmatically implies that the speaker doesn’t know that Harry is a zebra?15 Proponents of the pragmatic strategy offer the following reply. Whenever someone raises the possibility that Harry is a painted mule (by uttering it), it becomes salient (Rysiew 2001). Once this possibility is “in the air,” it implies not just (what the evidence-based account states) that not-p has a nonnegligible probability on total evidence, but that one has some real grounds for supposing not-p, so that one is not confident that p. Similarly, Dodd suggests that an utterance of “not-p is possible” implicates that not-p “is an epistemic possibility that the members of the context of utterance need to take seriously” (Forthcoming, 14; see also Dougherty and Rysiew 2009). But if that’s so, one should not claim to know that p, “since in so doing one of course represents oneself as (confidently) believing that p” (Dougherty and Rysiew 2009).16

229

225-238Benbaji2.pm

229

9/9/09, 11:08 AM

Hagit Benbaji

Now suppose that the hearer has no such false expectations with regard to the possibility in question; it is completely clear to her that “Harry is a painted mule” has merely a negligible probability on total evidence. If it is really clear in this case that the speaker has no special ground for doubting that Harry is a painted mule, if, that is, it is obvious that it is just the usual observation that this claim (Harry is a painted mule) is consistent with the evidence for knowing that Harry is a zebra— why mention it? In this case, (1) sounds odd because it states the obvious. Thus, the pragmatic explanation suggests the following dilemma: Either the possibility raised in the second clause of (1) is taken to express a real doubt, a significant concession, or it is not. On the first alternative, (1) is odd since its second conjunct (“it’s possible that …”) leads the hearer to infer that the speaker does not know that Harry is a zebra, contra to what the speaker asserts in the first conjunct of (1). On the second alternative, in which the doubt is not considered to be significant, (1) is odd since it is utterly trivial—there is no reason to say it. Either way, the explanation for the oddity of (1) is purely pragmatic.

3. Why the Pragmatic Strategy Fails Both pragmatic explanations, I contend, do not work. Let us adopt what Crispin Wright called “the standard of quotidian common sense”: take appearances as veridical unless there is a reason to do otherwise (Wright 2005, 243). When I raise the possibility that Harry is a painted mule in the standard quotidian situation, I cannot utter the knowledge claim, according to the pragmatists: Uttering it would imply that I have special grounds for doubting that Harry is a painted mule (perhaps the manager of the zoo told me something about Harry), so that I am not confident that Harry is a zebra. But when the possibility that Harry is a painted mule is raised to salience, the reaction, “It looks like a zebra, everything is normal, and this is good enough for knowing that it is one” does not pragmatically imply that I have a special ground for doubt that Harry is a painted mule. Mentioning that this is a quotidian perceptual situation makes it clear that I do not have real grounds for supposing that Harry is a painted mule. It follows that if I put my cards on the table and make explicit my evidence, the supposed pragmatic implicature should be canceled. It is odd so long as the hearer expects that I have some special evidence, for example, that I heard something about Harry from the manager of the zoo and therefore I am able to doubt whether it is painted, so I have a real ground for doubt. As long as I do my best to avoid such expectation, the following should not sound odd at all:

230

225-238Benbaji2.pm

230

9/9/09, 11:08 AM

On the Pragmatic Explanation of Concessive Knowledge Attributions

(2) I am standing just in front of Harry, it looks like a zebra, and I have no reasons to think otherwise, so I know that it is; yet, it’s possible that Harry is a painted mule. Statement (2) sounds no less odd than (1). But utterances of (2) cannot raise false expectations, because they explicitly state what the evidence is, so the pragmatic explanation does not apply to it. The pragmatic strategy predicts that the oddity of (1) arises from expectations on the part of the hearer about the knower’s evidence, expectations that are pragmatically implicated by the possibility of doubt (“but Harry might be a painted mule …”). This prediction is testable. The test is to state the nature of the evidence in question and see if the oddity disappears, as it should according to the pragmatic strategy. But the oddity remains. In order to reinforce my claim that the oddity cannot be cancelled by making explicit the fact that the knower does not have any special evidence that gives rise to the possibility of doubt, let us consider a modified version of (2): (2*) Harry is a zebra, but it’s possible that he is a painted mule though I have no reason to believe that. The second clause states explicitly that I have no reason to doubt that Harry is a painted mule (I’ve heard nothing special about Harry, I’ve seen nothing, etc.). Thus, utterances of (2*) cannot give rise to false conversational expectations. Yet (2*) is odd. How about the second alternative, that (2) (or (2*)) sounds odd not for raising false expectations, but simply for being obviously or trivially true? This explanation is not satisfactory, as Dougherty and Rysiew admit. After all, the assertion that the sun is going to rise tomorrow is obvious and, presumably, out of the blue, but not self-defeating as is (2). To highlight the selfdefeating nature of concessive knowledge attributions, Hawthorne remarks that “we can respond to such an assertion, not by complaining that too much has been said, but by noting that the claim is self-defeating: ‘well, if there is a chance that not-p, then you don’t know that p, do you?’” (2004, 25). Dougherty and Rysiew’s response to Hawthorne is that this complaint (“you don’t know that p, do you?”) assumes that there is some real ground for doubt that not-p (2009, 8–9). But this reply is inapplicable to (2). Statement (2) makes explicit that the chance is not “significant,” because it explicitly states what the evidence is (it looks like a zebra, and there is no reason to think otherwise). The fact that we can plausibly respond to an utterance of (2) with— “so you don’t know that p, do you?”—shows that the source of the oddity is not that it is trivial. I conclude that pragmatists fail to explain the oddity of concessive knowledge attributions. My objection to the pragmatic account utilizes the general strategy against pragmatic explana231

225-238Benbaji2.pm

231

9/9/09, 11:08 AM

Hagit Benbaji

tions, namely, that an uncancelable implicature is not an implicature.17 Pragmatists explain that (1) is odd because, when the possibility is made salient, or serious, or grounded, and we affirm our knowledge in the face of it, we let the audience expect that we can rule it out, or that we have some serious ground for believing this possibility. If that were the right explanation, then all one would have to do in order to cancel the expectation is to state what one’s evidence is. Make explicit the evidence for knowing that Harry is a zebra, and the false implicature would immediately disappear. Alas, it does not. That the oddity does not go away by making the evidence explicit indicates that it is not due to some expectations of the hearer;18 rather, it is due to the semantic of knowledge-claims. There is of course, a last resort for the pragmatists, as Lewis predicted: deny that (2) is odd at all—“numb your ears.” I end by showing why this is too late for pragmatists.

4. Why the Pragmatic Strategy Must Fail Dougherty and Rysiew insist that we often do utter concessive knowledge attributions, without the slightest oddity. This is their main reason for rejecting the knowledge-based account of epistemic possibility. They give the following examples: (a)

“Of course it’s possible that oil prices will fall dramatically over the next month, but we all know that that’s not going to happen.” (cf. Rysiew 2001, 497)

(b)

“I know they are going to lose but it’s possible that they will win, so I’m going to carry on watching just in case.” (cf. Hawthorne 2004, 24n. 60)

(c)

“We now know that there is a top quark: we have presented considerable evidence indicating its existence. Of course, our experiments are not conclusive, and there is a chance that our results will be overturned. We are confident, however, that they will be borne out.” (cf. Dougherty and Rysiew 2009, 10)

(d)

“I know no one will get confused about this, but make it explicit; it’s possible someone will.” (Dougherty and Rysiew 2009, note 11)

Now let me make clear that to my (overly) sensitive ear, (a)–(d) sound wrong. When such a concession is being made—“I’m going to carry on watching just in case”—it is usually clear that what the speaker has in mind is the tiny probability that the team will win. Yet once the possibility of winning is “in the air,” the speaker 232

225-238Benbaji2.pm

232

9/9/09, 11:08 AM

On the Pragmatic Explanation of Concessive Knowledge Attributions

cannot get it out of his mind—What if they actually win? How can I miss that? Once this possibility is salient to him, he cannot leave the game, which indicates that he does not take himself to know that they will lose. Had the speaker really known that they were going to lose, he would not have kept on watching the game. The same goes for the overly cautious referee in (d). Again, the referee is almost sure that no one will get confused, but she is bothered by the very small chance, indeed, almost zero, that someone will not understand it correctly—so much so, that she asks the author to make it explicit. This request, once again, defeats her claim to know. And so on—in all these cases once one reflects on the negligible probability that not-p, one cannot take oneself to know. To put Lewis’s insight in a paraphrase of Bernard Williams’s contentious remark, reflection destroys the attribution of knowledge. Dougherty and Rysiew, however, numb their ear; for them, casses (a)–(d) just state “the sober truths of fallibilism.” For the sake of argument, then, let me assume that they are right and that (a)–(d) do not sound odd, let alone contradictory. Rather than helping the pragmatic strategy, this assumption undermines it. If (a)–(d) do not sound odd at all, then a natural question suggests itself: Why does (1) sound odd? Assuming that (a)–(d) are perfectly proper, it is not true that by mentioning a possibility one makes it seem a serious one, namely, a possibility that destroys knowledge. In (a)–(d) the possibility of not-p is mentioned, and it is not taken as seriously. So why take “possible” in (1) to indicate a serious possibility? Recall the pragmatic explanation: Once this possibility is “in the air,” it implies not just (what the evidence-based account states) that not-p has a nonnegligible probability on total evidence, but that one has some real grounds for supposing not-p. But that there are correct utterances such as (a)–(d) shows that this is not always the implication, at least, not when it is conjoined with a knowledge claim. When a speaker utters the words “I got up early but …” the hearer might rightly expect the utterance to end with “I was late.” That is, the concession “but” usually indicates that what is expressed in the antecedent is very unlikely (e.g., “I was late”) given the first conjunct (“I got up early”). However, one might end the utterance in the following way: “I got up early but … I was there right on time!” Admittedly, this is not a regular way to state that I was not late, but the point is that the second conjunct cancels the implication of the first; it can no longer imply that I was late, when I affirm that I was not. Similarly, I agree that usually when we say “it’s possible that not-p” we implicate that we do not know whether or not p. But once we conjoin it with “I know that p,” the implicature should be canceled, and it should be understood to mean only that not-p is consistent with the evidence for knowing, as the implicature of “I got up early but …” is canceled once I clarify 233

225-238Benbaji2.pm

233

9/9/09, 11:08 AM

Hagit Benbaji

that I got to the meeting on time. If we can interpret the speaker charitably on some occasions (as we do in (a)–(d)), why not all the time? Furthermore, the default assumption should be that had the speaker had “a serious doubt” in mind, he would have made it clear; he would not have been satisfied by just throwing it into the air (“it’s possible that Harry is a painted mule”). Rather, he would have shared his reasons for raising this bizarre possibility (“I heard that the manager of this zoo likes to switch the zebra with a painted mule from time to time.”). Given that no reason is suggested, by default (1) should be understood to say what (a)–(d) do, and thus it should not sound odd at all. The pragmatic strategy must fail, I submit, since once pragmatists numb their ears to some utterances of (1), such as (2) or (a)–(d), and insist that these utterances of (1) just express the truth of fallibilism, the reaction to the rest of them should be like the reaction to “I got up early but … I was there right on time!” That is, the hearer must hear such utterances as expressing just the hope (as in case (b), of the disappointed fan) or the fear (as in case (d), of the worried referee) that knowledge is fallible. The pragmatic strategy must fail, then, because it must acknowledge that some tokens of (1) (such as (a)–(d), or (2)), those in which the implicature is canceled, are, allegedly, perfectly proper, and hence, we should hear the rest as saying precisely what they try to say, namely, the truth of fallibilism. Pragmatists must numb their ears to some utterances of (1), such as (2), and (a)–(d). Otherwise, if they admit that (2) and (a)– (d) are odd, no explanation can be given for their oddity, since in these examples it is clear that the possibility in question is not serious, that it merely says that p is probable on total evidence; so Lewis wins—fallibilism itself is odd. But if (2) and (a)–(d) are perfectly acceptable, then the charity principle, Grice’s maxim of Quantity, or just plain common sense, should convince the hearer that by uttering (1), the speaker just intends to express the truth of fallibilism.

Notes 1 The term is Rysiew’s (2001, 493). Stanley also considers concessive knowledge attributions in the second and third person, in the form “S knows that p but it’s possible for him that q” (where p entails not-q). 2 This way of formulating concessive knowledge attribution complicates the discussion by bringing into the picture the issue of closure. We can have, instead:

I know that Harry is a zebra, but it’s possible that Harry is not a zebra. This, too, is something the fallibilist is seemingly committed to. It’s not just the implications of a fallible belief with respect to which he is fallible. He is (somehow or other) committed to the possibility of being

234

225-238Benbaji2.pm

234

9/9/09, 11:08 AM

On the Pragmatic Explanation of Concessive Knowledge Attributions mistaken about the belief itself. But the implication from “Harry is a zebra” to “Harry is not a painted mule” is taken to be obvious, and it is assumed in the literature, so accordingly, I stick to (1). 3 Lewis converts to contextualism, but my purpose in this paper is the pragmatic strategy. 4 Stanley 2008. 5 As well as Dylan Dodd (Forthcoming), Jeremy Fantl and Matthew McGrath (2009). 6 Some object to formulating fallibilism as a claim about evidence because basic beliefs are not based on evidence. Others object to it because it ignores the possibility of nonpropositional evidence (if there is any, it cannot entail anything). But it is the received view that most epistemologists endorse, and it’s the definition that Dougherty and Rysiew (2009) assume, so I follow this near-universal understanding of fallibilism in this paper. 7 I’m not sure whether I agree with regard to Jane. Perhaps the correct answer would be: “They have run the test, but I do not yet know the result. So it might not be possible. I simply do not know yet whether or not John has cancer.” 8 Stanley follows what De Rose (1991) refers to as “Moore’s principle” (MP): MP: Whenever a speaker S does or can truly assert, “it’s possible that p is false,” S does not know that p. 9

Stanley asserts: “if the speaker is sufficiently intelligent to infer from Harry’s being a zebra that Harry is not a painted mule, (1) will express an obviously false proposition” (2005, 128). 10 As Dougherty and Rysiew (2009) note, this is in the spirit of Kripke’s original claim that “Hesperus is not Phosphorus” is epistemically possible, though metaphysically impossible: “given the evidence that someone has antecedent to his empirical investigation, he can be placed … in a qualitatively identical epistemic situation, and call two heavenly bodies “Hesperus” and “Phosphorus,” without their being identical. So in that sense we can say that it might have turned out either way” (Kripke 1972, 103–4). Since the evidence one possesses (prior to the discovery) does not entail the identity of Hesperus and Phosphorus (one can have the same evidence, while the identity is false), their nonidentity is epistemically possible. 11 Alternatively, (Evidence)

It is possibleA that p is true if and only if on A’s total evidence p has a nonnegligible probability. (Dougherty and Rysiew 2009)

12 Indeed, “Grice’s Razor” (Davis 2005) recommends that whenever it’s possible, it is preferable to explain away an oddity of a grammatical utterance pragmatically: “Other things equal, it is preferable to postulate conversational implicatures rather than senses, conventional semantic implicatures, or semantic presuppositions because conversational implicatures can be derived from independently motivated psycho-social principles.” 13 And he is indeed doing so, Rysiew argues, though he hastens to qualify that with the explanation that the contradiction lies only in the

235

225-238Benbaji2.pm

235

9/9/09, 11:08 AM

Hagit Benbaji speaker’s utterance of (1); the sentence itself is not self-contradictory (Rysiew 2001, 493). 14 Rysiew uses the notion of ruling out to mean Goldman’s indistinguishability, which might be equivalent to Lewis’s modal incompatibility between one’s experiential state and not-p. 15 Moreover, in contrast to Lewis’s account, Rysiew believes that a possibility does not become relevant simply by being raised. 16 There is another way in which an utterance of (1) can be misleading, even to the speaker, as it were. Rysiew (2001) suggests that whenever someone raises the possibility that Harry is a painted mule (by uttering it), it becomes salient. In such a situation, to answer the skeptic: “But I know that Harry is a zebra” pragmatically implies that I can rule out the possibility Harry is a painted mule. To avoid such an implication, I refuse to ascribe knowledge to myself. Once this possibility is “in the air,” and the speaker affirms his knowledge in the face of it, this pragmatically implies that the speaker can rule it out (Rysiew 2001, 490); hence, it is improper to conjoin it to the knowledge claim. This explanation, which takes the knowledge part do the key work, has been criticized by Dodd (Forthcoming). My criticism applies to the more recent suggestion of Rysiew and Dougherty (2009), which takes the possibility part do the key explanatory work. 17 See Halliday 2005, 397–98. 18 Or, at least, it is not due to the specific expectations suggested by Dougherty and Rysiew (2009). Nor is it due to the fact that (2) states the obvious.

References Brueckner, A. 2005. Fallibilism, underdetermination, and skepticism. Philosophy and Phenomenological Research 71:384–91. Davis, W. 2005. Implicature. Stanford Encyclopedia of Philosophy, ed. E. N. Zalta. http://plato.stanford.edu/entries/implicature/ De Rose, K. 1991. Epistemic Possibilities. Philosophical Review 100: 581–605. Dodd, D. Forthcoming. Confusions about concessive knowledge attributions. Synthese. Dougherty, T., and P. Rysiew. 2009. Fallibilism, epistemic possibility, and concessive knowledge attributions. Philosophy and Phenomenological Research 78(1): 123–32. Fantl, J., and M. McGrath. 2009. Advice for fallibilists: Put knowledge to work. Philosophical Studies 142:55–66. Goldman, A. 1976. Discrimination and perceptual knowledge. Journal of Philosophy 73:771–91. Grice, H. P. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press. Halliday, D. 2005. What explains our intuitions about knowledge ascriptions? Proceedings of the Aristotelian Society 105:377–86. Hawthorne, J. 2004. Knowledge and lotteries. Oxford: Oxford University Press. Kripke, S. 1972. Naming and necessity. Cambridge, MA: Harvard University Press. Lewis, D. 1996. Elusive knowledge. Australasian Journal of Philosophy 74:549–67.

236

225-238Benbaji2.pm

236

9/9/09, 11:08 AM

On the Pragmatic Explanation of Concessive Knowledge Attributions Rysiew, P. 2001. The context-sensitivity of knowledge attributions. Nous 35:477–514. Stanley, J. 2005. Fallibilism and concessive knowledge attributions. Analysis 65:126–31. ———. 2008. Knowledge and certainty. Philosophical Issues 18, Interdisciplinary Core Philosophy. Williams, M. 2001. Problems of knowledge. New York: Oxford University Press. Wright, C. 2005. Contextualism and scepticism: Even-handedness, factivity and surreptitiously raising standards. Philosophical Quarterly 55:236–62.

237

225-238Benbaji2.pm

237

9/9/09, 11:08 AM