Grice’s Razor Allan Hazlett Metaphilosophy, forthcoming

Abstract: Grice’s Razor is a principle of parsimony which states a preference for linguistic explanations in terms of conversational implicature, to explanations in terms of semantic context-dependence. Here I propose a Gricean theory of knowledge attributions, and contend on the basis of Grice’s Razor that it is superior to contextualism about ‘knows’. Keywords: Grice, knowledge, contextualism

One of Paul Grice’s major contributions to the philosophy of language was his contention – central to his theory of conversational implicature – that many of the various and complex ways in which we use our language can be explained by appeal only to the fact that we are rational beings, with an interest in mutual understanding and communication. What is not required, Grice maintained, are complex, inelegant, and unparsimonious proposals about the meanings of words. Call this principle of preference for pragmatic explanations, over semantic complexity, Grice’s Razor. Here I apply Grice’s Razor to the issue of the semantics and pragmatics of knowledge attributions. In §1 I present and explain Grice’s principle. In §2 I discuss a skeptical Gricean theory of knowledge attributions, recently defended by Jonathan Schaffer as an alternative to contextualism about ‘knows’. In §3 I propose and briefly defend two additional principles of preference for linguistic theories which, I maintain, complement and constrain Grice’s Razor. Finally, in §4 I present and defend my preferred anti-skeptical theory of knowledge attributions. Gricean theories of knowledge attributions are notable, as opposed to contextualism, for not taking the problem of philosophical skepticism to be (in some sense) a case of equivocation. Grice’s Razor should be welcomed by those

1

philosophers with an interest in not taking philosophical problems in general to be (in any sense) cases of equivocation. I consider these matters in §5. 1. Grice’s Razor, sympathetically formulated Consider: (1) Juliet is the Sun. (2) Juliet is a human being. And consider Romeo’s utterance of (1) outside the Capulet mansion, and his utterance of (2) in a later context, in which an interlocutor is under the impression that ‘Juliet’ names Romeo’s pet guinea pig. Romeo, who believes that no human being is an immense extra-terrestrial body, can appropriately utter both (1) and (2), despite the appearance of contradiction. Why? The correct answer is that (1) is not true, but merely assertible in context – Romeo’s utterance of (1) is metaphor. Following Grice, we can explain the propriety of (1) as follows. Juliet, assuming that speakers, including Romeo, generally conform to a maxim of Quality (‘Do not say what you believe to be false’), and assuming that speakers, including Romeo, generally know that their interlocutors generally assume that speakers generally conform, concludes, upon hearing Romeo’s utterance of (1), that Romeo is flouting the aforementioned maxim – he is not so crazy as to have actually mistaken her for the solar orb – and infers that Romeo would flout said maxim – i.e. intentionally violate it, with the intention that she Juliet realize this – only if he were trying to get her to recognize that he loves her very immensely, in a way such that she is as important to him as the Sun is to the Earth. And so Juliet comes to believe just that. Romeo’s utterance is proper – it does not mislead, but rather elegantly informs, his interlocutor – even though it is false. (See Grice 1989, 34, and Davidson 1978.) Compare this with an incorrect answer: that the word ‘Sun’ is semantically ambiguous – denoting in some contexts the solar orb which illuminates the Earth, and in others some salient person of emotional significance to the speaker. This incorrect answer aims to explain the ‘puzzle’ of why (1) and (2) are both correctly 2

assertible semantically – both are true, and hence both (ceteris paribus) assertible. But this is wrong. The Gricean explanation, on which ‘Sun’ is univocal and Romeo’s utterances of (1) and (2) really are contradictory, is preferable – it is simpler and proceeds by appeal only to the independently plausible theory of conversational implicature (among other virtues). To posit ambiguity – new semantic machinery – when we can explain the case by appeal only to Grice’s pragmatic theory would be gratuitous. This sort of reasoning appeals to a principle of parsimony that Grice called the “Modified Occam’s Razor.” “Senses,” Grice contended, “are not to be multiplied beyond necessity.” (Grice 1989, 47, see also 291) For this reason, an explanation in terms of the pragmatic theory of conversational implicature is better, ceteris paribus, than an explanation in terms of semantic ambiguity – one that ‘multiplies senses’. The appeal of this principle is perhaps most obvious in the following kind of case. Some philosophers had proposed that ‘and’ is semantically ambiguous – meaning sometimes the logical connective (‘&’) and meaning sometimes the temporally loaded ‘and then’. For example, Urmson wrote: In ordinary discourse the connectives often have a richer meaning [than in logic]; thus [5] ‘he took off his clothes and went to bed’ implies temporal succession and has a different meaning from [6] ‘he went to bed and took off his clothes’. (Urmson 1956, 9-10) Grice criticized this proposal as semantically profligate. A typical utterance of (5) does not imply that he took off his clothes and then went to bed, but it does implicate as much. This explains the appearance of invalidity in the move from a typical utterance of (5) to (6). And the reason, Grice contended, that a typical utterance of (5) implicates that the clothes came off first is that it is normally assumed that in general speakers conform (and know that it is normally assumed that they conform) to the maxim of Manner: ‘Be orderly’, i.e. recount event in the order that they occurred. (See Grice 1981, 185-6) Our examples so far have a common form. Two appropriate utterances, apparently contradictory, whose appropriateness is to be explained pragmatically (preserving the intuition that the two utterances are indeed contradictory). Are all

3

such cases to be explained in this way? Compare these cases, and the proposed solutions, to the following. Consider: (3) All banks contain treasure. (4) This bank does not contain treasure. And consider a treasure-seeker’s utterance of (3), in the context of a conversation concerning what sorts of institutions would be appropriate for a robbery, and her utterance of (4), in a context in which she and an accomplice have been digging at the edge of a river. Why are (3) and (4) assertible, in context, despite the appearance of contradiction? An incorrect answer is that ‘bank’ is univocal, that the aforementioned utterances really are contradictory, and that (3), though false, is assertible because of some special pragmatic rules that govern ‘bank’. (See DeRose 1999, 200, and 2002, 172-6.) The correct answer is that the imagined utterances of (3) and (4) are not contradictory: they only appear that way because ‘bank’ is semantically ambiguous. What this shows is that the Modified Occam’s Razor only cuts when our pragmatic explanation of the phenomenon in question is independently plausible – only when we would not need to posit any ‘special rules’ to deal with the case in question. Suppose ‘bank’ is univocal, i.e. that the treasure-seeker’s utterances of (3) and (4) are contradictory. There is no way, just by appeal to Grice’s theory of conversational implicature (for example), to explain how it could be appropriate to utter both. So a lesson: a pragmatic explanation is preferable to a semantic explanation, but only when it is independently plausible, when it needs to posit nothing new. In other words, only when it itself is parsimonious. This seems to capture Grice’s idea, but recall that his own formulation spoke only against the ‘multiplication of senses’. What exactly is prohibited here? Imagine a defender of ‘Sun’ ambiguity who said: “I agree that senses should not be multiplied, but I do not claim that ‘Sun’ has more than one sense. It has one sense, but it is akin to an indexical like ‘I’ or ‘here’. Depending on context it picks out different individuals. Its sense, we could say, is a function from contexts to individuals – but the function it picks out is the same in all contexts.” Now I do not know if Grice would have said that his Modified Occam’s Razor 4

speaks against this proposal, but it seems to me that the motivation behind the Modified Occam’s Razor is to rule out precisely this kind of bizarre, inelegant, and ugly semantic theory. So what I aim to do here is to formulate a principle in the spirit of Grice’s prohibition against ‘multiplying senses’, but one that speaks as well against the sorts of semantic profligacy that don’t strictly speaking posit multiple meanings or senses for an expression. What we need is a principle that prohibits those proposals that, in some intuitive sense, unnecessarily complicate the semantics for a given expression or class of expressions. Ambiguity itself need not involve the ‘multiplication of senses’ (which we get in the case of two words that are pronounced the same way, or in the case of one word with two different meanings). There are various sorts of ‘semantic complexity’ that do not. Semantic complexity, in this sense, is required for indexicals, as it is for context-sensitive gradable adjectives such as ‘tall’ and ‘rich’, much discussed by defenders and critics of epistemological contextualism. (This is so regardless of whether one claims that there are multiple senses of ‘rich’ or whether one claims that ‘rich’ denotes a function from individuals and contexts to truth-values.) Consider another example, from Grice: The phrase “French citizen” standardly means “citizen of France”, while the phrase “French poem” standardly means “poem in French”; but it would be a mistake to suppose that this fact implies that there are two (indeed more than two) meanings or senses of the word “French.” The word “French” has only one meaning, namely “of or pertaining to France”; it will, however, be what I might call ‘context-sensitive’; we might indeed say, if you like, that while “French” has only one meaning or sense, it has a variety of meanings-in-context[.] (Grice 1988, 193) But this is clearly semantic complexity of the sort relevant to our concerns here. It would be no more legitimate to posit multiple senses for ‘Sun’ than to posit one sense with multiple ‘meanings-in-context’. If we wish to capture the idea that we should not complicate our semantics unless we have to (“beyond necessity”), we need to capture the idea that we should not posit any of these various kinds of semantic context dependence unless we have to. These, then, are paradigms of semantic complexity – cases in which context plays a role in 5

determining the semantic value of an expression – : (i) mere homonymy, in which two expressions, with different meanings, are coincidentally pronounced or written the same, e.g. Grice’s example of ‘vice’,1 (ii) polysemy, in which one expression has two different meanings, e.g. ‘bank’ (with a common etymology for both uses), and ‘tall’ and ‘rich’ (perhaps), and (iii) paronymy, in which one expression behaves differently, semantically, depending on context, but without a ‘multiplication of senses’, e.g. indexicals, ‘tall’ and ‘rich’ (perhaps), and ‘French’. What do all these varieties of semantic complexity have in common? Simply this: that in all these cases we can easily imagine pairs of contexts in which seemingly contradictory sentences could appropriately be uttered, where the assumption of semantic complexity will explain the appearance of contradiction and explain the appropriateness of both utterances in terms of their truth. Consider: ‘I am in the grip of a vice’ (in answer to a question concerning my character), ‘I am not in the grip of a vice’ (in answer to a question concerning whether I am constrained by any clamps), ‘Payton is tall’ (compared to the average person), ‘Payton is not tall’ (compared to his NBA colleagues), ‘I am sitting’ (said by someone seated), ‘I am not sitting’ (said by someone standing), etc. As all these cases show, the Gricean objection is not an objection to semantic complexity as such, but (as the cases of ‘Sun’ and ‘and’ show) an objection to the positing of semantic complexity when an independently plausible pragmatic explanation of the appropriateness of apparently contradictory utterances is available. Following Wayne Davis, I will call the stronger principle which captures this objection – not just the objection to needless senses, but the objection to needless semantic complexity – Grice’s Razor. (See Davis 1998, 19.) Importantly, the positing of semantic complexity is ceteris paribus to be avoided not merely because this will keep the number of posited senses down, but because pragmatic explanations in terms of conversational implicature are explanations in terms of general features of language, rather than explanations in

1

In the US we refer to certain clamps with ‘vise’, not ‘vice’ as in the UK.

6

terms of the features of particular expressions.2 It is this that distinguishes explanations in terms of the semantic complexity of certain expressions (as above with ‘bank’, ‘vice’, ‘tall’, and ‘I’) from explanations in terms of conversational implicature. Given all this, here is the principle I aim to defend and apply: (Grice’s Razor) If E1 and E2 are putative explanations of some linguistic phenomenon, and T is the rest of linguistics, prefer E1 to E2, ceteris paribus, if (i) E2 involves, to a greater extent than E1, the postulation of semantic complexity, and, (ii) E1 does not involve the postulation of pragmatic phenomena not postulated by T. Three points. First, semantic complexity comes in degrees, and it is doubtful that there is any expression that is not to some degree semantically complex. A relatively non-context-sensitive expression like ‘having a mass of one kilogram’ falls on the low end of the scale, while expressions like ‘tall’ and ‘I’ fall on the high end. Second, as noted above, Grice’s Razor does not rule out semantic complexity – it just rules it out when a pragmatic explanation can explain everything that needs explaining. Third, the principle, as stated, relies on a distinction between the semantic and the pragmatic. For our purposes here the domain of the semantic is truth-conditions; this is why an explanation of acceptable but apparently contradictory utterances in terms of semantic complexity is an explanation on which both utterances are true – the truth of both utterances explains their assertibility. The pragmatic concerns the assertibility and unassertibility of sentences, in general. But here I will only be concerned with explanations in terms of Grice’s theory of conversational implicature. A ‘pragmatic phenomenon’, then, for the purposes of this paper, can just be taken to be any of the contentions that Grice makes about which Maxims speakers

2

Grice maintained that if we can “allow [the various uses of a word] to be variants under a single principle,” then “that is the desirable thing to do.” (1989, 291) As Davis notes, multiple senses “imply specific conventions governing the use of particular words,” whereas explanations in terms of conversational implicature, for example, appeal only “to the general rules of rational, cooperative behavior represented by the Cooperative Principle.”(1998, 19) See also Bennett [Conditionals].

7

generally conform to, and assume that others conform to, etc. (See Grice 1989, 24-37.) Why endorse Grice’s Razor? Grice’s Razor is a principle of parsimony, a consequence of the idea that simpler theories are better than more complex ones. It manifests its commitment to that idea in two ways. First, Grice’s Razor opposes the unwarranted postulation of entities – senses and meanings. Second, Grice’s Razor opposes theoretical redundancy – positing theory to explain something that can be explained by theory we’re already committed to. The theory of conversational implicature is a powerful theory, and Grice’s Razor just says that nothing should be posited to do work that the theory of conversational implicature can do itself. 2. Grice’s Razor, unfettered Contextualism about ‘knows’ is the thesis that ‘knows’ and cognates are semantically context-sensitive – whether ‘S knows p’ is true depends not only on S, p, and their relationship, but also on the context in which ‘S knows p’ is uttered. Typically this view is motivated by the fact that that different utterances of ‘I know Magritte painted this’ are intuitively appropriate or inappropriate depending in part on how high the ‘epistemic standards’ are in the context of utterance (a casual day at the museum vs. appraisal for Sotheby’s), or on what alternative possibilities are salient, relevant, or in that context’s ‘contrast class’ (whether Dali painted it vs. whether a master forger painted it). In particular, we are inclined to retract, and in some cases reverse, our knowledge attributions as standards go up or as alternatives are mentioned. Imagine that Martin and Maria are botany students preparing for an exam on the difference between a tree and a bush. Martin says, of a nearby tree, ‘I know that’s a tree’. Maria decides to harass him. ‘What if it’s a fake plastic tree, designed to look just like a real one?’ Martin doesn’t say anything. ‘Well,’ Maria says, ‘do you know that it’s a tree, or not?’ Martin replies: “I suppose I don’t know that it is.” Contextualists point out that Martin’s way of speaking in this story does not sound wrong – he has not 8

misbehaved, linguistically speaking. As above we are faced with two seemingly contradictory sentences uttered by Martin: (5) I know that’s a tree. (6) I don’t know that it is. We can identify, though, three things that an explanation of Martin’s behavior in this case needs to do: (i)

Explain why Martin’s utterance of (5) was proper.

(ii)

Explain why Martin’s reluctance to utter (5) was proper.

(iii)

Explain why Martin’s utterance of (6) was proper.

Call this the contextualist’s puzzle. The contextualist solves the puzzle by proposing that the semantic value of ‘I know that’s a tree’ depends on context – (5) is true when Martin utters it – which takes care of (i) – but Maria’s mentioning of alternative possibilities changes the conversational context, so that (5) would be false if Martin were to utter it – which takes care of (ii), and the fact that Martin’s utterance of (6) is true, so also (iii). Jonathan Schaffer has recently proposed a skeptical criticism of contextualism, which appeals to Grice’s Razor, along lines suggested (but not endorsed) by Grice.3 Grice’s sympathy with skepticism is evident in some more recent papers, for example in “Meaning Revisited” he writes: Some people [say] that the only things we know are necessary truths[.] Now I might want to say that those people are right, if what they meant was that strictly speaking the only things that can be known are necessary truths. However, this does not restrict us to supposing that people who talk about knowing other things are using the word “know” improperly: all it requires is that there should be some license to apply the word nonstrictly to things which in some way approach or approximate to the ideal cases. (Grice 1989, 301-2) But while Grice does not provide a theory of this “license,” Schaffer does. So it is his proposal that I will examine here is some detail. On his view our ordinary claims to know are in general false, because knowledge that p requires the elimination of all logical possibilities incompatible with p, and for most p and S 3

Schaffer 2004 and Grice 1989, 149-53 and 301-2. Note that Schaffer no longer endorses the view I discuss here – so for my purposes ‘Schaffer’ refers to Schaffer qua author of Schaffer 2004.

9

such that we ordinarily would say ‘S knows p’, S cannot eliminate all logical possibilities incompatible with p.4 Martin, for example, does not know that the plant in question is a tree, on Schaffer’s view, because (as Maria points out) he cannot eliminate logical possibilities incompatible with its being a tree. However, Schaffer contends that ordinary claims to know are pragmatically assertible because they are cases of hyperbole. He writes: “I know that I have hands” entails that I have eliminated every logical possibility of handlessness, which exaggerates the range of possibilities I can eliminate. […] The speaker who says, “The airplane is a mile long” will (if all goes well) be understood to mean that the airplane is large by the standards of the current context. Likewise the speaker who says, “I know I have hands” will (if all goes well) be understood to mean that he can eliminate the possibilities of handlessness in the current context. (Schaffer 2004, 140-1) Call this view hyperbole skepticism. The hyperbole skeptic reasons that since “[l]inguistic machinery should not be complicated without necessity,” (Ibid., 147) hyperbole skepticism is to be preferred to contextualism, which must postulate semantic complexity (specific to ‘knows’ and cognates) that we are not independently committed to. Hyperbole skepticism postulates no new phenomena, because on this view we can explain the propriety of our ordinary claims to know by appeal to Grice’s theory of conversational implicature, which can explain cases of hyperbole. Now I think this has to be wrong. It is wrong because it entails that I do not know that I have hands, and I believe that I can reject any view that entails that the way Moore rejected skepticism – I am more certain that I know I have hands than that any of the premises adduced in defense of skepticism are true. But this is not a problem with Grice’s Razor, it is a problem with Schaffer’s application of it – in this case ceteris are not paribus. Schaffer claims that they are, but as I argue below (§3) this is wrong – we need to propose some principles to stand alongside Grice’s Razor that state a prima facie

4

This conception of knowledge as requiring elimination of alternative logical possibilities clearly only applies to knowledge of contingent truths. There are no logical possibilities incompatible with the proposition that 2+2=4, for example, but this is something that I know. Thanks to my referee for drawing my attention to this.

10

preference for theories that are charitable and for those that vindicate our pretheoretical intuitions.5 But let us consider hyperbole skepticism in more detail. Cases of hyperbole, like cases of metaphor, are cases in which a maxim of Quality (‘Do not say what you believe to be false’) is flouted. These are cases in which one speaker A violates a maxim which speakers are generally assumed to conform to, where A intends her interlocutor B to recognize that A had violated the maxim, and furthermore to recognize that A intended for B to recognize that A had violated the maxim. B must then determine, by a procedure of rationalization, what A intended to communicate to B. If B gets it right, communication does not break down, despite the violation of one of the maxims.

In these sorts of cases, in

other words, if all goes to plan the flouting speaker does not violate Grice’s Cooperative Principle (‘Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged’). Imagine that Martin is telling Maria about his recent vacation. ‘It was incredible’, Martin says, ‘the plane was a mile long’. Martin has said something that he believes to be false, and so he has violated the aforementioned maxim of Quality. But Martin’s intention is for Maria to recognize that he has willingly violated said maxim, and to recognize that he intended for her to recognize that. Maria, knowing that no plane is literally a mile long, and knowing that Martin knows this as well, and assuming that Martin is following the Cooperative Principle, concludes that he must have willingly violated the aforementioned maxim of Quality, and that he must have intended for her to realize this. Maria’s next move is to try to figure out what Martin intended to communicate to her. And this is not difficult, given the

5

Schaffer, citing Chomsky, rejects such principles on the grounds that “[o]ur linguistic intuitions provide evidence for acceptability, and do not discriminate between semantic and pragmatic sources.” (2004, 146) Chomsky writes: “[W]e cannot in general know, pretheoretically, whether [some] deviance is a matter of syntax, semantics, pragmatics, belief, memory limitation, style, etc.” (1977, 4) But these remarks are, I think, beside the point – my defense of a principle that favors theories consistent with our pre-theoretical intuitions (below) does not run afoul of Schaffer and Chomsky’s claims.

11

context of conversation: what he intended to communicate was that the plane was especially long. The hyperbole skeptic says the same, mutatis mutandis, about Martin’s utterance of (5). Martin has said something that he believes to be false, and so he has violated the aforementioned maxim of Quality. But Martin’s intention is for Maria to recognize that he has willingly violated said maxim, and to recognize that he intended for her to recognize that. Maria, knowing that Martin has not eliminated every logical possibility incompatible with the proposition that the plant is a tree (and therefore that he does not know that the plant is a tree), and knowing that Martin knows this as well, and assuming that Martin is following the Cooperative Principle, concludes that he must have willingly violated the aforementioned maxim of Quality, and that he must have intended for her to realize this. Maria’s next move is to try to figure out what Marin intended to communicate to her. And this is not difficult, given the context of conversation: what he intended to communicate was that he has eliminated the possibility that the plant is a bush. Now some of the implausible aspects of hyperbole skepticism are on display here. Even if you think that skepticism is true, and that no one knows much of anything, you presumably do not think that everyone else – ordinary speakers who make appropriate knowledge claims, for example – also think that skepticism is true. But if Martin’s utterance of (5) is successful hyperbole then Martin and Maria must be skeptics. I do not want to put too much stock in this objection to hyperbole skepticism, though. In §3 I’ll defend principles which, if correct, prefer my Gricean account of knowledge attributions, presented in §4, to hyperbole skepticism. The real problem with hyperbole skepticism will be made clear then. There is something deeply and importantly right about the hyperbole skeptic’s criticism of contextualism. Contextualism, I contend, is heir apparent to a certain influential family of views about ‘knows’ which were popular (at least, more popular than they are now) in the middle of the 20th century. Paradigmatic of the views I have in mind are, for example, the following claims:

12

Doubt Principle: An utterance of ‘S knows p’ is meaningful only if it is not mutually obvious to all conversational participants that p, e.g. if some doubt about p has been expressed or suggested. Proof Principle: An utterance of ‘S knows p’ is true only if S can prove or otherwise justify p with a persuasive argument, in context. These both were endorsed, almost surely, by Wittgenstein and Norman Malcolm in their criticisms of Moore’s frequent claims to know in his lectures (Wittgenstein 1969 and Malcolm 1949 and 1977). ‘I know that’s a tree’, they argued, was meaningless in the context of those lectures (if Moore were to manage to point to a tree, through a window, say), although it would be meaningful and possibly true in the botanical context described above. The evidence for the two principles, though, was the fact that to say ‘S knows p’ in situations in which it is mutually obvious that p, or in situations in which S does not have any way of arguing for p, would be to deviate from the way ‘S knows p’ is ordinarily used. For example, concerning the Proof Principle, Austin observed: Whenever I say I know, I am always liable to be taken to claim that, in a certain sense appropriate to the kind of statement (and to present intents and purposes), I am able to prove it. (Austin 1946, 85) And similar remarks populate Wittgenstein 1969 (e.g. §243, §484, §504) and Malcolm 1949 (e.g. 214). But as Moore argued in a 1949 letter to Malcolm (Moore 1993, 214-5), it is fallacious to infer, from the fact that an utterance of a sentence, in a certain context, deviates from the way that the sentence is ordinarily used, that the sentence is meaningless when uttered in that context. Wittgenstein and Malcolm identify the meaning of an expression with its use; they end up with a theory on which ‘S knows p’ is context-sensitive: meaningful and perhaps true in some contexts, meaningless in others. It is this identification of meaning with use that Moore criticized, in the case of ‘know’, by claiming that Malcolm had conflated an utterance being senseless – in the sense of being a pointless use of an expression – with an utterance being senseless – in the sense of the utterance

13

lacking meaning.6 And it was the refusal to distinguish between the use of an expression and its meaning that Grice’s theory of conversational implicature was tailor-made to combat, both in general and in the case of ‘knows’; Grice mentions the dispute between Malcolm and Moore in his “Prolegomena” to Grice 1989. All this is just to highlight the plausibility of a Gricean account of our various uses of ‘S know p’ – as part of a general strategy of not confusing the inappropriateness or weirdness or pointlessness of an utterance with it’s being meaningless or false. Compare, of course, the contextualist’s claim about uttering (5) after Maria has ‘raised the standards’. To claim knowledge in that context would be odd, to say the least. But just as oddity is not meaninglessness, pace Malcolm, so oddity is not falsehood, pace contextualism. Again, Grice’s theory was a reaction to this mistake, embodied in ordinary language philosophies, of taking the oddity of uttering S in one context, but not in another, to be a matter of the meaning of S, i.e. the semantics of S. 3. Grice’s Razor, constrained Before presenting the Gricean theory of knowledge attributions that I prefer, I’ll propose and defend two principles to accompany Grice’s Razor in informing our preference for linguistic theories. Here they are: Charity: If E1 and E2 are putative explanations of some linguistic phenomenon, prefer E1 to E2, ceteris paribus, if: (i) E1, to a lesser extent that E2, entails that speakers systematically communicate falsehoods, or if: (ii) E1, to a lesser extent than E2, entails that speakers systematically make false utterances. Believability: If E1 and E2 are putative explanations of some linguistic phenomenon, prefer E1 to E2, ceteris paribus, if an epistemically virtuous subject, in your position – i.e. someone with the same evidence, background beliefs, etc. – would believe E1, rather than E2.

6

This criticism of the identification of the meaning of ‘know’ with the use of ‘know’ is cogently renewed, against a different instance of the identification of meaning and use (and the confusion over two senses of ‘meaning’), in Warnock 1962.

14

Consider Believability. Scientific theories must meet a number of criteria to be worthy of our acceptance. It is often said that simplicity and elegance are virtues that we look for in scientific theories, and that any theory that does not meet these criteria falls short of our goals for a scientific theory. Believability says that justified believability it is a virtue of scientific theories as well. In particular, since an epistemically virtuous subject prefers, ceteris paribus, to believe propositions that are consistent with what she already believes, Believaibility entails that a theory that is consistent with our background beliefs is ceteris paribus superior, qua scientific theory, to one that is not. Consider my belief that I know that I have hands. This is one of my background beliefs (relative to our inquiry here), one of the things that I believed before I started thinking about the semantics and pragmatics of knowledge attributions. An epistemically virtuous subject, with the belief that she knows that she has hands, would never believe hyperbole skepticism if there were an equally adequate theory that did not entail that she does not know that she has hands. Believability makes the further claim that if there were an equally adequate theory, and if that theory indeed did not conflict with your background beliefs, then you ought to prefer it, qua scientific theory, to hyperbole skepticism. Some philosophers might find Believability obvious. I hope this is so; it seems obvious to me. But, as I will argue below, if it is true then hyperbole skepticism is inferior to an anti-skeptical Gricean theory of knowledge attributions. Consider, then, Charity. I know of only one compelling argument in favor of Charity, and I do not want to suggest that it is overwhelmingly compelling, nor that it is novel – these I offer as reasons for why I’ll try to present the argument for Charity using as few words as possible. But I think it is important to consider whether Charity is correct, or not, because it seems to me that something like Charity would be the only way to keep Grice’s Razor from, as it were, running amok. Consider a view on which all of our ordinary talk is false, but not improper, because it’s all metaphor. And imagine that there’s an equally good view available on which some of our ordinary talk is true, and not metaphorical. The ‘all metaphor’ view runs afoul of clause (ii) of Charity; if Charity is right 15

than we should prefer the alternative view. What else would warrant preferring the alternative view, but Charity? The argument I have in mind is from Davidson: The general policy … is to choose truth conditions that do as well as possible in making speakers hold sentences true when … those sentences are true. [This is so because] widespread agreement is the only possible background against which disputes and mistakes can be interpreted. Making sense of the utterances and behavior of others, even their most aberrant behavior, requires us to find a great deal of reason and truth in them. To see too much unreason on the part of others is simply to undermine our ability to understand what it is they are so unreasonable about. (Davidson 1974, 153) Now I do not always find that I understand exactly what Davidson is getting at when he argues in this way; and the suggestion that there is just one thing that he is getting at strikes me as wrong, as well. But I think a correct argument can be extracted from this passage (and others with which the reader is probably familiar). The argument goes like this. Imagine that you are confronted by someone – or perhaps, not to prejudge the issue, some creature – who is producing certain behavior. Question: is this creature’s behavior linguistic? In other words, are the sounds (for example) that this creature is producing utterances in a language, or are they just meaningless (which of course does not mean purposeless or pointless) sounds? Suppose that, on the hypothesis that the sounds are utterances, you undertake an investigation of this creature’s behavior to try to learn its language – if it speaks one. And suppose that, after much research, all the evidence you have gathered is consistent with two theories of the creature’s behavior. Theory U (for ‘uncharitable’) says that the creature indeed speaks a language, but its utterances are mostly false, and furthermore do not even communicate truths – the creature’s utterances reveal a set of mostly false beliefs. Theory C (for ‘charitable’) says that the creature does not speak a language after all; the sounds it produces are produced by some other mechanism. Suppose Theory U and Theory C both can explain the creature’s behavior equally well, elegantly and simply. The Davidsonian then makes the following bold conjecture: Theory C is superior to Theory T. But here is why this conjecture is 16

warranted. Theory T misclassifies the creature’s behavior. The only thing that could warrant taking a creature’s behavior as linguistic in the first place would be some evidence (or mere assumption) that the creature in question is rational. But if this assumption is not, in the end, supported by our evidence of the creature’s behavior, it should be rejected. Compare this to a case in which you observe two people sitting on a basketball court discussing philosophy. Two theories; one says they are playing basketball, but that they are playing very badly, the other says that they are not playing basketball at all. The second theory is correct, and this is so even if there is a ball nearby, even if they roll is back and forth between them as they talk.7 The assumption – even though ‘supported’ by the evidence that the two are, after all, on a basketball court – that the two are playing basketball is eventually rejected. They simply are not succeeding enough at playing basketball to count as playing. The same holds, mutatis mutandis, of our imagined creature. It is not succeeding enough at producing utterances to count as producing utterances in the first place. The possibility of interpretation, and hence the possibility of communication, is contingent on the mutual assumption of the for-the-most-part rationality of the parties involved. This is implicit in Grice’s “myth” of radical interpretation in “Meaning Revisisted,” where an interpreting creature must – if communication is to take place at all – regard the creature to be interpreted “as trustworthy in one or another of perhaps a variety of ways.” (Grice 1989, 294) An assumption that can of course be challenged in all this is that rationality involves both speaking the truth and communicating the truth. If it only requires the latter, then clause (ii) of Charity is unsupported. Further considerations, Davidsonian in spirit, speak in favor of a connection between rationality and speaking the truth. But since I will not need to appeal to Charity below, in defense of my proposed alternative to hyperbole skepticism, I will not present them here.

7

Making their ‘play’ more and more elaborate will destroy the intuition that they are not playing basketball, but it will also erode the plausibility of the view that they are playing very badly.

17

4. Grice’s Razor, reapplied Charity and Believability complement Grice’s Razor – we now have three ceteris paribus principles which should inform our preference for linguistic theories. I concur with Schaffer that hyperbole skepticism is superior to contextualism, even granting the correctness of Charity and Believability.8 But the theory that I will now present, meiosis dogmatism, is superior to hyperbole skepticism, given the correctness of Believability.9 We start with any antiskeptical analysis of knowledge, i.e. an analysis which provides necessary and sufficient conditions for the truth of ‘S knows p’ on which a great many of our ordinary knowledge attributions are true. I prefer a reliabilist account of epistemic justification, but any anti-skeptical account will do. Recall our tasks (i) – (iii), above. On any reasonable anti-skeptical analysis, I will suppose, Martin’s utterance of (5) was true; thus (i) is taken care of.10 What about (ii)? Several philosophers have recently defended Gricean accounts of the impropriety of knowledge attributions in ‘high standards’ contexts.11 The meiosis dogmatist will adopt an account of this sort, on which it would be improper to utter ‘S knows p’ in certain contexts because it would be misleading. Grice, in an anti-skeptical mood, writes: If I say I know that p, then perhaps sometimes there is a nonconventional implicature of strong or conclusive evidence (not mere thinking that p, with p true). (Grice 1989, 53) 8

This could be called into question. I assume that hyperbole skepticism’s superiority to contextualism, vis-à-vis Grice’s Razor, outweighs contextualism’s superiority to hyperbole skepticism, vis-à-vis Charity. (It is not easy to evaluate contextualism vis-à-vis Believability, because its only claims are about ‘knows’ and cognates, not about knowledge; see DeRose 2000.) 9 The name ‘meiosis dogmatism’ was suggested to be by Schaffer, and I accept it with this caveat: ‘dogmatism’ can mean two things that are not part of my view here. In one sense, to be ‘dogmatic’ is to improperly say something, where the impropriety stems from one’s failure to back up what one says with an argument. I have not done that here. In another sense, to be ‘dogmatic’ is to believe something without proper evidence, and again, the view I am proposing does not involve that. ‘Dogmatism’ is here a more elegant name for anti-skepticism. See Moore’s discussion of dogmatism at Moore 1959, 171. 10 Taken care, of given that he could tell the tree from a bush, and that to utter (5) was relevant and informative. 11 See Rysiew 2001, Brown 2005, and Pritchard 2005. Moore 1993 is a predecessor of these proposals, and Grice (as noted below) suggests a version of the proposal in Grice 1989, 52-3. See also the aforementioned reference to Moore and Malcolm (and the discussion of ‘remember’ that follows) at 1989, 5.

18

But what might generate such an implicature? If knowledge does not require strong or conclusive evidence, why would claiming to know suggest that one has strong or conclusive evidence? Patrick Rysiew has defended an answer to this question, which I endorse, and which I will present here as part of the meiosis dogmatist’s position. Begin with some observations about knowledge: I. Although knowledge that p does not require conclusive evidence for p, knowing often requires a certain degree of evidence for p.12 II. Although knowledge that p does not require the elimination of all logical possibilities incompatible with p, knowing that p often involves one’s discriminatory capacities, i.e. the ability to eliminate certain possibilities incompatible with p (a point stressed in Schaffer 2004).13 III. Although knowledge that p does not require being able to prove or argue cogently for p (a point stressed by Moore, e.g. at the end of “Proof of an External World”), knowing that p will often be accompanied by the ability to prove or argue cogently for p.14 The maxims of Quantity and Relation jointly enjoin us to make our contributions to a conversation relevant, informative, and not overinformative. The mutual assumption that speakers obey these maxims can generate conversational implicatures. For example, suppose that Martin and Maria cannot reach the cookie jar. ‘We need to find someone taller than us’, Martin complains. Maria considers the fact that their roommate Marcus is taller than they are, and could probably reach the cookies. But she knows that Marcus is at work, and she knows that Martin doesn’t know that. Maria refrains from replying with ‘Marcus is taller than us!’, because were she to utter that sentence, it would mislead Martin – he would infer, given the assumption that Maria is obeying the maxims of Quantity and Relation, that Marcus is home and available to assist them in getting the cookies. Therefore, to utter that sentence, in this context, would be improper because it would have a false implicature.

12

Often – perhaps always, depending on your view of epistemic justification. I consider it a live option to maintain that there are cases of knowledge without evidence, e.g. knowledge that my perceptual faculties are reliable. 13 Often, but not always, e.g. knowledge of necessary truths. 14 Often, but not always, e.g. basic perceptual knowledge.

19

But we can say the same, mutatis mutandis, to complete task (ii). Maria has just mentioned the possibility that the object they are looking at is not a tree, but a fake tree. Martin refrains from saying: ‘I know that’s a tree’, once this possibility has been made salient, for were he to say that, it would mislead Maria – she would infer, given the assumption that Martin is obeying the maxims of Quantity and Relation, that Martin is able to eliminate the salient possibility, i.e. she would infer that Martin has some information that she does not have, about the thing they are looking at – he has, say, gone up close and inspected it, to confirm that it is a real tree. Why would she infer this? Because knowing does indeed require, in this case, the elimination of certain possibilities incompatible with the object’s being a tree. Knowledge involves discrimination. Imagine that Martin claims knowledge, in this case, and assume that he is conforming to the Maxims of Quantity and Relation. The only way a claim to know, in this context, could be informative and relevant is if Martin is able to eliminate the possibility that the object is a fake tree. Since Martin cannot do any such thing, he refrains from claiming knowledge, for to do so would be improper because it would generate a false implicature. So in some contexts an utterance of ‘S knows p’ implicates that S can eliminate contextually salient alternatives to p. This is due to point II above. But given point I, we can explain Grice’s contention that in some contexts an utterance of ‘S knows p’ can implicate that S has conclusive evidence that p. Since knowing does crucially involve the possession of evidence, in a context where the question of whether S has conclusive evidence is the topic of conversation (for example), a claim that S knows would be informative and relevant only if S has conclusive evidence that p. And given point III, the same can be said, mutatis mutandis, to explain Austin’s observation, above. In some contexts an utterance of ‘S knows p’ would implicate that S can prove or cogently argue for p, e.g. contexts in which a proof or argument for p is what the conversational parties are in search of. This Gricean view is superior to the Proof Principle, discussed above. And finally, we can explain the oddity of ‘The hotel clerk asked me what my name was, and fortunately I knew the answer’ (from 20

Grice 1989, 5) by appeal to our mutual assumption of conformity to the maxim of Quantity. It is not false (or meaningless), in this context, to say ‘I knew the answer’, but it does suggest something false: that you had to appeal to some evidence (point I), or eliminate some alternative possibilities (point II), or whatever. So we have an alternative, as well, to Wittgenstein and Malcolm’s Doubt Principle – to claim knowledge that p implicates (sometimes) that there is some doubt that p; to recall knowledge that p (as in Grice’s example) implicates (sometimes) that there was doubt that p. This picture is attractive, but it leaves task (iii) unfulfilled. To utter (5), in this context, would be improper, but this does not mean that to utter (6) would be proper. (Consider: ‘Marcus isn’t taller than us’) We come, then, to the part of the proposal that gives meiosis dogmatism its name. Recall the hyperbole skeptic’s account of our ordinary claims to know as flouts of the maxim of Quality. Meiosis dogmatism says the reverse. It is our knowledge denials – for example Martin’s utterance of (6) – that are flouts of quality, not cases of hyperbole but cases of meiosis or understatement.15 Just as, according to Schaffer, ordinary claims to know overstate our epistemic position, on my view skeptical denials understate our epistemic position.16 The sort of meiotic flouting of Quality I have in mind occurs when, for example, Martin complains to Maria: ‘We never get out of the house anymore’. Martin knows that in fact they do sometimes get out of the house, and knows that Maria knows this, and intends that she recognize that he knows all this. Maria, knowing that they do sometimes get out of the house, and knowing that Martin knows this as well, and assuming that Martin is following the Cooperative Principle, concludes that he must have willingly violated the first maxim of Quality (‘Do not say what you believe to be false’), 15

Not all cases of meiosis are flouts of Quality. Some are flouts of Quantity, such as: ‘There’s a few drops coming down’, said during a downpour. But the cases of meiosis I have in mind here are flouts of Quality. Contrast this sort of meiosis with litotes (e.g. ‘It was no ordinary task’, which implicates, but does not entail, that it was an extraordinary task), cases of which are not flouts of Quality. 16 It seems to me that the distinction between hyperbole and meiosis is not a rigorous one. ‘The plane was a mile long’ is supposed to be hyperbole because it overstates the length of the plane; is it not also meiosis because it understates the shortness of the plane? The distinction is useful, though, for contrasting the view I propose with that proposed by Schaffer. Nothing much rides on it; what is crucial is the notion of a flout of Quality.

21

and that he must have intended for her to realize this. Maria’s next move is to try to figure out what Marin intended to communicate to her. And this is not difficult, given the context of conversation: what he intended to communicate was that they do not get out of the house as often as they used to. The meiosis dogmatist tells the same story, mutatis mutandis, about Martin’s utterance of (6). Martin knows that knows the object is a tree, and knows that Maria knows this, and intends that she recognize that he knows all this. Maria, knowing that Martin knows that it’s a tree, and knowing that Martin knows this as well, and assuming that Martin is following the Cooperative Principle, concludes that he must have willingly violated the first maxim of Quality, and that he must have intended for her to realize this. Maria’s next move is to try to figure out what Marin intended to communicate to her. And this is not difficult, given the context of conversation: what he intended to communicate was that he indeed cannot eliminate the possibility that the object they are looking at is a fake tree. Martin concedes, by uttering (6), that his belief that the object is a tree is, in some respects, epistemically lacking. Compare someone who denounces underperforming Joey Harrington with: “Joey Harrington isn’t a quarterback!” Harrington lacks salient features desirable in a quarterback, but he is, in fact, a quarterback – just a bad one. Similarly Martin’s belief that the object is a tree lacks salient features desirable in a belief, but his belief is, in fact, a case of knowledge – just not knowledge that involves the ability to eliminate the possibility that the object is a fake. Note that ‘know’ is stressed in Martin’s utterance of (6), just as ‘never’ and ‘quarterback’ are in our other examples of Quality-flouting meiosis. Was this just a quirk of my presentation of the contextualist’s puzzle? Consult your intuitions about an utterances of (6a) ‘I don’t know it’s a tree’ and (6b) ‘I don’t know it’s a tree’. To my ear (6a) sounds normal only if Martin has, for example, learned that there is a sort of bush that has the same sort of leaf that he had taken to be evidence that the plant was a tree, as in: Maria: ‘Remember, our professor said to be on the lookout for that bush with those tree-like leaves’ 22

Martin: ‘Right. I guess I don’t know whether that’s a tree’ The difference between (6a) and (6b), of course, is that it would be normal for Martin to say (6a) only if he had stopped believing that the plant was a tree – because he realized that he did not have enough evidence to conclude that it was. But in our imagined case, when Martin says (6b), he does not stop believing that the object in question is a tree. If he did, and on that basis said (6a), we would not judge this to be proper. We would, in fact, find it very strange. And that is why his utterance, in the story, is a flout of Quality – because he does not stop beliving that the object in question is a tree. I submit, therefore, that this is further evidence in favor of meiosis dogmatism (against both contextualism and hyperbole skepticism): that we find the contextualist’s puzzle plausible only when there is stress on ‘knows’ in the knowledge-denying utterance. This takes care, finally, of task (iii). Contrast this account of the propriety of (6) with the hyperbole skeptic’s account of the propriety of (5). Recall that the hyperbole skeptic must attribute to Maria (and Martin) the belief that Martin does not know that the object in question is a tree – i.e. if the hyperbole skeptic is right that ordinary knowledge claims are successful cases of hyperbole, then ordinary speakers are skeptics. Meiosis dogmatism has no such unhappy consequence. Maria, according to the meiosis dogmatist, knows that Martin knows; indeed it is the fact that it is mutually obvious to Maria and Martin that he knows that makes his meiotic utterance a success. The fact that the propriety of Martin’s linguistic behavior does not require the attribution of skepticism to him (and Maria) is a virtue closely related to meiosis dogmatism’s virtues vis-à-vis Believability. Since meiosis dogmatism is consistent with my background beliefs – that I know I have hands, for example – it is superior to hyperbole skepticism vis-à-vis Believability. Since both are equally virtuous vis-à-vis Charity and Grice’s Razor, I conclude that meiosis dogmatism is superior to hyperbole skepticism. All three theories we have considered – contextualism, hyperbole skepticism, and meiosis dogmatism – are equally virtuous in terms of clause (i) of Charity. Contextualism is superior to the two Gricean proposals, vis-à-vis clause (ii). I

23

contend that this virtue is outweighed by meiosis dogmatism’s virtues vis-à-vis Grice’s Razor and Believability. 5. Grice’s Razor and philosophical problems How is the contextualist’s puzzle related to the epistemological problem of skepticism? Opinions differ, even among those sympathetic with contextualism, ranging from the view that the truth of contextualism is irrelevant to the problem (Sosa 2000) to the view that contextualism solves the problem (DeRose 1995). One thing is clear: to the extent that contextualism solves the problem of skepticism, it solves it, in some sense, by declaring victory to both sides. The skeptic is right in her claim that ‘no one knows much of anything’, and the ordinary person is right in her claim that she ‘knows that insects have six legs’. This ecumenical aspect of the contextualist treatment of skepticism has attracted the ire (justified, in my view) of many epistemologists – see Feldman 1999 and Conee 2005, for example. The problem is this. The contextualist solution to skepticism highlights certain utterances that we (supposedly) intuitively judge are not improper: the ordinary person’s ‘I know that insects have six legs’ and the skeptic’s ‘No one knows anything about the external world’. The contextualist says that, despite the appearance of contradiction, these two utterances are not really contradictory. The anti-ecumenicalist simply asks at this point: what would be the consequence of employing this methodology across the board? Consider: •

For the problem of free will: The ordinary person’s ‘I came here of my own free will’ (when she was not coerced to appear) vs. the hard determinist’s ‘No one ever acts freely’. (The problem of free will)

•

For the problem of affluence and poverty: The ordinary person’s ‘You did the right thing by buying the book for your daughter’ (instead of spending the money on whiskey) vs. the maximizing utilitarian’s ‘You did not do the right thing; you should have given the money to Oxfam’.

24

•

For the issue of moral realism: The ordinary person’s ‘You ought to return that wallet to its rightful owner’ vs. the moral skeptic’s ‘It is never the case that anyone ought to do anything’.

The anti-ecumentalist complaint is simply this: it would simply not be any kind of solution to the philosophical problems mentioned if we were to say that ‘free’, ‘right’, and ‘ought’ are context-sensitive, and that despite the appearance of contradiction the two utterances, in these cases, are really compatible. It would, in short, be profoundly unsatisfying to any philosopher interested in these issues to be told that that is the ‘solution’ to these problems, and it would, perhaps more importantly, be unacceptable to the defenders of hard determinism, utilitarianism, and moral skepticism to learn that their views are not revisionary. Such views, if true, are striking philosophical discoveries, and if true reveal that ordinary people – believers in free will, partiality, and moral obligations, all – are mistaken in these beliefs. So, in any case, argues the anti-ecumenicalist. The contextualist, in these cases, would make the mistake of attributing to the revisionst a content (for her utterance) that she does not wish her utterance to have (namely, a content consistent with common sense). Does the Gricean make the mistake of attributing to skeptics an intention they don’t have, namely an intention to speak meiotically? I’ve made no claims about the pragmatics of claims made by philosophical skeptics here, and it seems quite right to say that their denials of knowledge are not meiotic. Their denials of knowledge are simply false, because they have made a philosophical mistake. There is no special burden on the anti-skeptic to explain their propriety. But what I want to point out, to conclude, is that a Gricean account of apparently contradictory utterances (whatever their relation to such-and-such philosophical problem) will not be an ecumenical one – on which said utterances are not really contradictory, on account of equivocation (of some sort). In this way the application of Grice’s Razor, to linguistic puzzles of philosophical interest, does not threaten to interpret philosophical disputes as less than genuine, on account of equivocation. To the extent that Gricean theory solves any philosophical problems (outside of the philosophy of language) – and I doubt that it does – it cannot do so without taking 25

a side. Consider Schaffer’s skepticism, and my proposed anti-skepticism. To this extent, a Gricean approach is importantly, and I think rightly, anti-ecumenical.*

*

This paper, which was first composed in 2004 at Via de Cerchi 7, in Florence, was presented in July of 2005 at a workshop on the semantics/pragmatics distinction before the CONTEXT 05 conference in Paris, and in March of 2006, at the Pacific APA meeting in Portland. I received much help from both audiences. Particular thanks are also due to Jonathan Schaffer, Jason Stanley, Baron Reed, and my referee at Metaphilosophy, for many valuable comments and suggestions.

26

Bibliography: Austin, J.L. (1946): “Other Minds,” reprinted in Philosophical Papers, third edition (Oxford, 1979), pp. 76-116 Brown, J. (2005): “Adapt or Die: The Death of Invariantism?,” Philosophical Quarterly 55, pp. 263-85 Chomsky, N. (1977): Essays on Form and Interpretation (North Holland) Conee, E. (2005): “Contextualism Contested,” in Steup and Sosa (eds.), Contemporary Debates in Epistemology (Blackwell), pp. 47-56 Davidson, D. (1974), “Belief and the Basis of Meaning,” reprinted in Inquiries into Truth and Interpretation (Oxford, 1984), pp. 141-54 --- (1978): “What Metaphors Mean,” reprinted in Inquiries into Truth and Interpretation (Oxford, 1984), pp. 245-64 Davis, W. (1998): Implicature (Cambridge) DeRose, K. (1995): “Solving the Skeptical Problem,” Philosophical Review 104, pp. 1-52 --- (1999): “Contextualism: An Explanation and Defense,” in Greco and Sosa (eds.), The Blackwell Guide to Epistemology (Blackwell), pp. 187-205 --- (2000): “Now You Know It, Now You Don’t,” --- (2002): “Assertion, Knowledge, and Context,” Philosophical Review 111, pp. 167-203 Feldman, R. (1999): “Contextualism and Skepticism,” Philosophical Perspectives 13, pp. 91-114 Grice, H.P. (1981): “Presupposition and Conversational Implicature,” in Cole (ed.), Radical Pragmatics (Academic Press), pp. 183-98 --- (1988): “Aristotle on the Multiplicity of Being,” Pacific Philosophical Quarterly 69, pp. 175-200 --- (1989): Studies in the Way of Words (Harvard) Malcolm, N. (1949): “Defending Common Sense,” Philosophical Review 58, pp. 201-220 27

--- (1977): “Moore and Wittgenstein on the Sense of ‘I Know’,” in Thought and Knowledge (Cornell, 1977) Moore, G.E. (1993): “Letter to Malcolm,” in Selected Writings (Routledge), pp. 213-16 --- (1959): “Certainty,” reprinted in Selected Writings (Routledge, 1993), pp. 171-96 Pritchard, D. (2005): “Contextualism, Skepticism, and Warranted Assertibility Manoevres,” in Keim-Campbell, O’Rourke, and Silverstein (eds.), Knowledge and Skepticism (MIT) Rysiew, P. (2001): “The Context-Sensitivity of Knowledge Attributions,” Noûs 35, pp. 477-514 Schaffer, J. (2004): “Skepticism, Contextualism, and Discrimination,” Philosophy and Phenomenological Research 69, pp. 138-55 Sosa, E. (2000): “Skepticism and Contextualism,” Philosophical Issues 10, pp. 118 Urmson, J.O. (1956): Philosophical Analysis: Its Development Between the Two World Wars (Oxford) Warnock, G.J. (1962): “Claims to Knowledge,” reprinted in Language and Morality (Barnes and Noble, 1983), pp. 43-54

28