Philos Stud (2016) 173:525–546 DOI 10.1007/s11098-015-0505-2

The reformulation argument: reining in Gricean pragmatics Zachary Miller1

Published online: 16 June 2015  Springer Science+Business Media Dordrecht 2015

Abstract A semantic theory aims to make predictions that are accurate and comprehensive. Sometimes, though, a semantic theory falls short of this aim, and there is a mismatch between prediction and data. In such cases, defenders of the semantic theory often attempt to rescue it by appealing to Gricean pragmatics. The hope is that we can rescue the theory as long as we can use pragmatics to explain away its predictive failures. This pragmatic rescue strategy is one of the most popular moves in philosophy of language, philosophical logic, and formal semantics. In this paper I argue that this strategy fails whenever the predictive failures at issue can be recast in epistemological or metaphysical terms. This general ‘‘reformulation argument’’ undermines a wide variety of pragmatic rescue attempts. Keywords

Pragmatics  Gricean pragmatics  Semantics/pragmatics distinction

1 Introduction A semantic theory aims to make predictions that are accurate and comprehensive. Sometimes, though, a semantic theory falls short of this aim, and there is a mismatch between prediction and data. In such cases, defenders of the semantic theory often attempt to rescue it by appealing to Gricean pragmatics. The hope is that we can rescue the theory as long as we can use pragmatics to explain away its predictive failures. This pragmatic rescue strategy is one of the most popular moves in philosophy of language, philosophical logic, and formal semantics. In this paper I argue that this strategy fails whenever the predictive failures at issue can be recast in epistemological or metaphysical terms. This general ‘‘reformulation argument’’ undermines a wide variety of pragmatic rescue attempts. & Zachary Miller [email protected] 1

Philosophy Department, The University of Oklahoma, Norman, OK, USA

123

526

Z. Miller

2 The semantic/pragmatic distinction In this section I introduce and illustrate the semantic/pragmatic distinction. In formal semantics, one of the central goals is to explain why an assertion conveys the information that it does and why an assertion enables audience members to draw certain inferences.1 To illustrate with an example, an assertion of (1) enables audience members to infer (2): (1) (2)

Jordan flew to South Korea and she tried kimchi. Jordan flew to South Korea.

The standard explanation of this fact relies on the conventional and truth-conditional meaning of and. It is part of the meaning of and that the truth value of an entire conjunction depends on the truth value of each conjunct. If each conjunct is true, then the conjunction is true. If either conjunct is false, then the conjunction is false. Thus, if the conjunction Jordan flew to South Korea and she tried kimchi is true, the meaning of and guarantees that Jordan flew to South Korea is also true. An assertion can convey many pieces of information solely by virtue of its meaning, in the conventional and truth-conditional sense. We say that the content of the assertion semantically entails these pieces of information.2 But semantic entailment isn’t everything. In addition, a typical assertion of (1) enables audience members to infer (3): (3)

First Jordan flew to South Korea and then she tried kimchi.

This temporal information does not seem to come from the truth-conditional meaning of the assertion. This temporal information demands a pragmatic, as opposed to a semantic, explanation. In this case, the pragmatic explanation appeals to the fact that a cooperative speaker narrates events in the order in which they occurred. When someone asserts Jordan flew to South Korea and she tried kimchi, audience members assume that the speaker is being cooperative, and conclude that first Jordan flew to South Korea and then Jordan tried kimchi.3 Paul Grice pioneered the idea that the rules of conversation help determine the inferences that audience members can make.4 We can mean more than we literally

1 Following the dominant trend in linguistics and formal semantics, I will use the term inference in an idealized sense. As Harman (1986) has emphasized, when it comes to the rationality of the inferences we make, many factors come into play. I focus on those factors concerning meaning and conversation. 2

There are many ways of characterizing semantic entailment. One might, for example, adopt a dynamic approach, according to which one claim semantically entails another just in case the assertion of the former claim necessarily creates a context in which it would be appropriate to assert the latter claim. See van Benthem (1996) for a discussion of different ways of characterizing semantic entailment in dynamic terms. For our purposes in this paper, a simple static characterization in terms of conventional and truthconditional content suffices.

3

If you believe that we do not assert sentences and only ever assert propositions, then feel free to replace each instance of ‘‘asserting S’’ with ‘‘assertively uttering S’’.

4

Grice presented these ideas in his William James Lectures, which he delivered at Harvard in 1967, and which were reprinted in Grice (1989).

123

The reformulation argument: reining in Gricean pragmatics

527

say, and we often accomplish this feat by taking advantage of the rules of conversation. Using Grice’s terminology, an assertion of Jordan visited South Korea and she tried kimchi conversationally implicates that first Jordan flew to South Korea and then she tried kimchi. According to Grice, the rules are, roughly, that speakers should strive to say things that are true, informative, and relevant, and that speakers should strive for clarity of expression.5 When it comes to explaining why an assertion conveys a certain piece of information, we now have two distinct explanatory strategies to work with. First, we could develop a semantic explanation, which would show that the piece of information is contained within the meaning of the assertion, and is thus semantically entailed. Second, we could develop a pragmatic explanation, which relies on the presumption that the speaker is being cooperative and the fact that cooperation requires adherence to a set of rules. According to a pragmatic explanation, the piece of information at issue is conversationally implicated.6 Let’s illustrate the semantic/pragmatic distinction with a second example. An assertion of (4) enables audience members to infer (5): (4) (5)

Most spicy foods contain capsaicin. Some spicy food contains capsaicin.

Here we should offer a semantic explanation. An assertion of Most spicy foods contain capsaicin conveys that some spicy food contains capsaicin because of the conventional and truth-conditional meaning of the quantifier most. According to this meaning, the claim Most As are B is equivalent to the claim More than half of all As are B. If more than half of all spicy foods contain capsaicin, then some spicy food contains capsaicin. In addition, a typical assertion of (4) enables audience members to infer (6): (6)

Not every spicy food contains capsaicin.

Here a semantic explanation is inappropriate. This information is not part of what is conventionally and truth-conditionally asserted. The claim that Every A is B entails More than half of all As are B. Because every entails most, most should not entail not every. If it did, then by the transitivity of entailment, every would entail not every. So in this case, a pragmatic explanation is called for. When more information would further the purpose of a conversation, cooperative speakers, all else equal, prefer to make more informative, as opposed to less

5

See Grice (1989: 26–27).

6

At this juncture, two points should be briefly addressed. First, there is a large literature on how exactly the semantic/pragmatic distinction should be drawn. I hope to avoid making any controversial remarks on this front. For an in-depth, and admittedly opinionated, discussion of this literature, see King and Stanley (2005). Second, Grice, and subsequent theorists, have thought that there are, even beyond semantic entailment and conversational implicature, additional ways for an assertion to convey information. An assertion might presuppose a piece of information, for example, or it might conventionally implicate it. In this paper, I will confine my attention to semantic entailment and conversational implicature, bracketing presupposition, conventional implicature, and other ways of exchanging information.

123

528

Z. Miller

informative, assertions.7 Suppose that we are in a conversation about how many spicy foods contain capsaicin. When someone asserts Most spicy foods contain capsaicin, we can conclude that he or she does not believe that every spicy food contains capsaicin. Given the purpose of this conversation, if the speaker had believed this more informative alternative, then he or she would have asserted it.8 Hence, either the speaker has not taken a stand with respect to the more informative alternative, or the speaker believes that the more informative alternative is false. If we assume that the speaker is opinionated about the topic, then we can dismiss the idea that the speaker has failed to take a stand with respect to the more informative alternative, and we can conclude that the speaker believes that the more informative alternative is false. At the core of a pragmatic explanation is the idea that speakers tend to be cooperative and therefore adhere to certain rules concerning how a conversation should take place. Crucially, one cannot develop a pragmatic explanation without referencing a speaker, an audience, a conversation, and the rules of conversation. It is this fact that causes trouble for many attempts to develop pragmatic explanations of linguistic phenomena.9

3 Pragmatic explanations as rescue attempts In this section I describe an argumentative strategy that I will call the pragmatic rescue strategy. The pragmatic rescue strategy is one of the most widely used moves in philosophy of language, philosophical logic, and formal semantics. The main claim of this paper is that this strategy is extensively overused. In the next section I develop the reformulation argument, which shows that, in a wide range of cases, the pragmatic rescue strategy is misguided. A semantic theory makes claims concerning the conventional and truthconditional meaning of an expression. In the previous section, we relied on the 7

We need to be careful here. Green (1995) shows that there is no conversational rule demanding that speakers, all else equal, make maximally informative assertions. Even with the all else equal clause, speakers need only prefer more informative assertions when the additional information is required by the purpose of the conversation.

8

You might worry that one can believe a proposition yet not feel confident enough in its truth to assert it. Assuming a knowledge norm for assertion and a knowledge norm for belief, such a scenario could not happen if the relevant person were adhering to these norms. If you prefer a stronger norm for assertion or a weaker norm for belief, then the problematic scenario could arise even if the relevant person were adhering to these norms. In such a case, I suspect you could reformulate my example in slightly different terms to bypass the worry. At any rate, this issue is orthogonal to the main argument I will develop later in this paper.

9

In linguistics, there is a trend towards conceptualizing pragmatics as involving covert syntactically realized operators. These operators, by way of convention, transmit ‘‘pragmatic’’ information. I have in mind the work of Chierchia (2004), Chierchia (2006), and Fox (2007), among others. On this conception of pragmatics, a pragmatic explanation need only involve an operator and this operator’s conventional role; it need not reference a speaker, an audience, a conversation, and the rules of conversation. In this paper, I will be setting aside such grammatical approaches to pragmatics. There are many foundational worries one might raise about this project, but here is not the place to raise them. For my present purposes, I will merely focus on traditional pragmatics, and the use of traditional pragmatic explanations in philosophy. Such explanations are ubiquitous.

123

The reformulation argument: reining in Gricean pragmatics

529

traditional semantic theory of and in order to give a semantic explanation of why an assertion of Jordan flew to South Korea and she tried kimchi conveys the information that Jordan flew to South Korea. Similarly, we relied on the traditional semantic theory of most in order to give a semantic explanation of why an assertion of Most spicy foods contains capsaicin conveys the information that some spicy food contains capsaicin. A semantic theory should capture our judgments concerning which inferences are acceptable and which are not. Let’s refer to intuitively acceptable inferences as intuitive inferences, and let’s refer to intuitively unacceptable inferences as counterintuitive inferences. We can evaluate a semantic theory along two dimensions. First, how many intuitive inferences does the theory correctly predict? In this case, the more the better. Second, how many counterintuitive inferences does the theory incorrectly predict? In this case, the fewer the better. It would be a point in favor of a theory if it predicts that an assertion of A and B enables us to infer A. It would be a strike against a theory if it predicts that an assertion of A or B enables us to infer A. So with semantic theories there are two kinds of predictive failures. It can fail to predict an intuitive inference, or it can incorrectly predict a counterintuitive inference. When a theory gives rise to a predictive failure, this mismatch between prediction and fact can be dealt with in one of two ways. First, the theory can be revised in such a way as to avoid the predictive failure and bring the theory into alignment with the facts. Second, the pragmatic rescue strategy can be used to show that the predictive failure is something that we can live with after all. How exactly does the pragmatic rescue strategy work? Suppose that, intuitively, an assertion of A conveys the information that B. But also suppose that a semantic theory fails to predict this intuitive inference. This is, on the face of it, bad news for the semantic theory. Employing the pragmatic rescue strategy in this case would involve arguing that an assertion of A conversationally implicates that B. This would give us a pragmatic explanation that accounts for the intuitiveness of the inference. Or suppose that a semantic theory predicts a counterintuitive inference. Suppose that a semantic theory predicts that an assertion of C conveys the information that D, even though an assertion of C appears to do no such thing. Employing the pragmatic rescue strategy in this case would involve arguing that after an assertion of C, an assertion of D would violate the rules of conversation and thus be an inappropriate thing to say. This would give us a pragmatic explanation of the counterintuitiveness of the inference. The pragmatic rescue strategy enjoys a long and distinguished history. For practically every debate concerning the meaning of an expression, there are multiple pragmatic rescues on offer. The main claim of this paper is that, in a wide range of cases, the pragmatic rescue strategy is misguided. I develop a general argument, called the reformulation argument, in support of this claim.

123

530

Z. Miller

4 Illustrating the reformulation argument: the material conditional In this section I introduce the reformulation argument against the pragmatic rescue strategy. I introduce it by way of a case involving Grice’s notorious material conditional analysis of the indicative conditional.10 The material conditional analysis of the indicative conditional is a semantic theory concerning the indicative if… then… construction. According to the theory, an indicative if… then… statement is true just in case either the antecedent (the if… clause) is false or the consequent (the then… clause) is true. This material conditional analysis is simple and truth-functional, and it makes a number of correct predictions. For example, it validates modus ponens: A and if A, then B together guarantee the truth of B. It also validates the so-called direct argument: A or B guarantees the truth of if not A, then B. But the material conditional analysis appears to get some things wrong. Historically, its predictive failures have been referred to as the paradoxes of material implication. For example, according to the material conditional analysis, the truth of (7) guarantees the truth of (8): (7) (8)

Oswald killed Kennedy. If Oswald didn’t kill Kennedy, a Martian did.

The truth of (7) guarantees the falsity of (8)’s antecedent. And, according to the material conditional analysis, the falsity of (8)’s antecedent guarantees (8)’s truth. But surely we do not want the truth of (7) to guarantee the truth of (8). This is a predictive failure for the material conditional analysis. Grice attempts to explain away predictive failures such as this one by way of pragmatic rescue. According to Grice, the truth of (7) does guarantee the truth of (8). But, even though (8) is true, I cannot felicitously assert it. For if I were to assert the conditional, I would misleadingly implicate that, for all I know, Oswald might not have killed Kennedy. A pragmatic story as to how this implicature gets generated runs as follows. According to the material conditional analysis, the claim Oswald killed Kennedy is strictly more informative than the conditional If Oswald didn’t kill Kennedy, then a Martian did. This is because the claim Oswald killed Kennedy is strictly more informative than the disjunction Either Oswald killed Kennedy or a Martian did, and, according to the material conditional analysis, the disjunction Either Oswald killed Kennedy or a Martian did is truth-conditionally equivalent to the conditional If Oswald didn’t kill Kennedy, then a Martian did. Now assume that it is conversationally important whether Oswald killed Kennedy. That is to say, assume that the conversation’s purpose requires me to provide this information if I am in a position to do so. Given this assumption, if I know that Oswald killed Kennedy, then I should just assert this instead of asserting the strictly weaker conditional. Given that I asserted the less informative claim, I must not have been in a position to assert the more informative claim. And I would 10

See Grice (1989: 58–85).

123

The reformulation argument: reining in Gricean pragmatics

531

not be in a position to assert this more informative claim because, for all I know, Oswald might not have killed Kennedy. But there is good reason to think that this pragmatic rescue fails, regardless of exactly how the pragmatic derivation is spelled out. We can see this by reformulating the problematic inference at the level of rational commitment. Suppose that I believe that Oswald killed Kennedy. If the material conditional analysis is right, this means that I am rationally committed to believing that if Oswald didn’t kill Kennedy, then a Martian did.11 But I am not rationally committed to this. In fact, it appears quite rational to believe that Oswald killed Kennedy while disbelieving that if Oswald didn’t kill Kennedy, then a Martian did. Furthermore, there is no way for a pragmatic explanation to help. My beliefs and rational commitments have nothing to do with speakers, audiences, conversations, or the rules of conversation. If we assume that we want a similar explanation both at the level of assertion and at the level of rational commitment, then the pragmatic story is not the right story about what is going on at the level of assertion, since it cannot apply at the level of rational commitment. There are precedents for the idea that, when it comes to the level of rational commitment, pragmatics cannot rescue the material conditional analysis: But this move to pragmatics [endorsed by Grice and others in defense of the material conditional analysis] seems to be wide of the mark if the problem we are interested in is an epistemological one.…If holding u commits us to w, then if we accept that u it is clear that w is one of our epistemic commitments regardless of the pragmatic facts about potential utterances of w. Gillies (2004: 591) And there also are precedents for the idea that, because of this failure at the level of rational commitment, pragmatics cannot rescue the material conditional analysis even at the level of assertion: But the difficulties with the truth-functional conditional [i.e., the material conditional analysis] cannot be explained away in terms of what is an inappropriate conversational remark. They arise at the level of belief. Edgington (1995: 245) The scope and importance of this line of thought has been drastically underestimated. It can be generalized along two dimensions. First of all, there is no need for us to limit our attention to conditionals. What we have is a general way of reformulating linguistic problems as epistemological ones.12 Secondly, there is

11

At least as long as we make the assumption that rational commitment is closed under logical consequence. If this assumption is too strong for your tastes, practically any reasonably weakened principles along these lines will do the work that I need it to do. Furthermore, one might worry that our intuitions in these cases track conditional beliefs, as opposed to beliefs in conditionals. To address this worry, we could assume that a belief in a conditional implies a conditional belief. In this context, this assumption will not be problematic. 12 Speaks (2008) applies something similar to the epistemological reformulation strategy to names, definite descriptions, and quantifier domain restriction.

123

532

Z. Miller

no need for us to limit ourselves to epistemology. We just need to reformulate the problematic inferences in non-conversational terms. Epistemology is non-conversational, but there are other non-conversational domains. I will illustrate this with the case of free choice permission.

5 Generalizing the reformulation argument: free choice permission Suppose that we are discussing Jordan’s powerlifting hobby. An assertion of (9) typically enables audience members to infer (10) and to infer (11): (9) Jordan can bench press 100 pounds or squat 200 pounds. (10) Jordan can bench press 100 pounds. (11) Jordan can squat 200 pounds. This inference pattern is modeled on the so-called paradox of free choice permission.13 This name is a bit of a misnomer for two reasons. First of all, the puzzle is only ‘‘paradoxical’’ in the sense that the traditional semantic theory of possibility fails to predict these intuitive inferences. Secondly, the inference pattern arises, and is just as puzzling, for all varieties of possibility, not just permission. As Klinedinst (2007: 262) puts it, ‘‘the same puzzle crops up across modalities[…] but for historical reasons—it was first noticed in work on deontic logic (Ross, 1941)—it is known as the ‘puzzle of free choice permission’’’. And as Geurts (2010: 108) puts it, ‘‘we observe free choice inferences with non-deontic modals’’. I will focus on ability modality in this section.14 Why does the traditional semantic theory of possibility fail to predict these inferences? In brief, because the traditional semantic theory treats possibility as existential quantification (over ‘‘accessible’’ possibilities). And existential quantification does not distribute over disjunction—Ax(Fx _ Gx) entails neither AxFx nor AxGx. Just because there is a bar or a coffee shop downtown, it does not follow that there is a bar downtown. Could the proponent of the traditional theory of possibility employ the pragmatic rescue strategy? Perhaps an assertion of (9) conversationally implicates (10) and (11). This has proven to be a popular approach.15 But this pragmatic approach is bound to fail, regardless of how the details are spelled out. We can apply the reformulation argument here by taking the problematic inferences and reformulating them in epistemological terms. My belief

13 The earliest discussions of free choice permission are in Ross (1941) and von Wright (1951). Since then, the puzzle has attracted a great deal of scholarly attention. For some of the canonical treatments, see Kamp (1973), Zimmerman (2000), Kratzer and Shimoyama (2002), and Fox (2007), among others. 14 I will do this in order to keep things simple. Deontic modality arguably supervenes on speech acts, beliefs, intentions, and other mental states in tricky ways. Using deontic modality as a case study would be unnecessarily complicated. 15

It has been defended in Kratzer and Shimoyama (2002), Alonso-Ovalle (2006), Schulz (2005), Aloni and van Rooij (2007), Fox (2007), and Geurts (2010). Some of these approaches are quite sophisticated, and go beyond the Gricean framework with which most philosophers are familiar. These complexities need not concern us here.

123

The reformulation argument: reining in Gricean pragmatics

533

that Jordan can bench press 100 pounds or squat 200 pounds rationally commits me to the belief that Jordan can bench press 100 pounds and rationally commits me to the belief that Jordan can squat 200 pounds. If I believe that Jordan can bench press 100 pounds or squat 200 pounds, but fail to believe that she can bench press 100 pounds, then I am guilty of irrationality. How would a pragmatic explanation account for these facts concerning belief and rational commitment? The fact the free choice inferences can be reformulated in non-conversational terms shows that the pragmatic rescue strategy is misguided. Furthermore, in this case we can just as easily reformulate the problem in metaphysical terms. The mere fact that Jordan can bench press 100 pounds or squat 200 pounds guarantees that Jordan can bench press 100 pounds and guarantees that Jordan can squat 200 pounds. But whether or not a mere fact guarantees something typically has nothing to do with speakers, audiences, conversations, or the rules of conversation. Again, if we think that we need a similar explanation for each of these levels—the level of assertion, the level of rational commitment, and the level of mere facts—then a pragmatic story will have no role to play, since it cannot apply at the level of rational commitment or the level of mere facts. We are faced with three questions: (i)

Why does an assertion of Jordan can bench press 100 pounds or squat 200 pounds convey the information that Jordan can bench press 100 pounds and convey the information that Jordan can squat 200 pounds? (ii) Why is an agent who believes that Jordan can bench press 100 pounds or squat 200 pounds rationally committed to believing that Jordan can bench press 100 pounds and rationally committed to believing that Jordan can squat 200 pounds? (iii) Why does the mere fact that Jordan can bench press 100 pounds or squat 200 pounds guarantee that Jordan can bench press 100 pounds and guarantee that Jordan can squat 200 pounds? A pragmatic explanation could only address the first of these three questions. A pragmatic explanation takes advantage of the fact that there is a speaker obeying the rules of conversation in order to make things easier for his or her audience. When problematic inferences are removed from the realm of conversation, when they no longer have anything to do with assertion or discourse, there is no way to get pragmatic reasoning off the ground. The answers to these three questions should be very similar. Because a pragmatic explanation is out of place with regard to the second and third question, a pragmatic explanation is out of place with respect to the first question. Hence, if problematic inferences can be reformulated in non-conversational terms, e.g., epistemological or metaphysical terms, then they cannot be explained away pragmatically. Let me end this section by addressing the worry that context sensitivity poses a problem for the epistemological or metaphysical reformulation. Of course, in order to talk about a belief or a fact, we sometimes use context sensitive vocabulary, and there had better be a context of utterance whenever we use context sensitive vocabulary. But this does not show that such beliefs and facts are context sensitive; it is not even

123

534

Z. Miller

clear what this would mean. Nor does this show that such beliefs and facts are problematic in any way. To consider an analogy, I can talk about the local weather conditions. In order to know what exactly what I am talking about, you need to know how the context of utterance specifies the meaning of ‘‘local’’. But this does not make the local weather conditions context sensitive or problematic in any way.

6 The upshot Whenever we can reformulate problematic inferences in either epistemological or metaphysical terms, then a pragmatic rescue is out of place. A pragmatic rescue relies essentially on the fact that an assertion is made in a conversation with a speaker and an audience, and that there are rules concerning how a conversation should take place. This entire conversational structure is absent when we are considering the epistemological issues surrounding beliefs and the rational commitments that accompany them. Similarly, this conversational structure is absent when we are considering the metaphysical issues surrounding facts and what they guarantee. If problematic inferences emerge at these epistemological and metaphysical levels, then we cannot use a pragmatic explanation to resolve them at the level of assertion. This reformulation argument drastically undermines the flexibility and usefulness of the pragmatic rescue strategy, one of the most popular strategies in philosophy of language, philosophical logic, and formal semantics. I will consider two additional examples in this section to illustrate the reformulation argument’s scope. 6.1 Counterfactuals with disjunctive antecedents Consider counterfactuals with disjunctive antecedents. An assertion of (12) enables audience members to infer (13): (12) (13)

If Jordan were to drink tea or coffee, then she would feel energized. If Jordan were to drink tea, then she would feel energized.

Most semantic theories of counterfactuals fail to predict this intuitive inference. For example, the two most influential semantic theories concerning counterfactuals— those contained in Stalnaker (1968) and Lewis (1973), respectively—fail to predict this inference. To see why, consider a simplistic Stalnaker–Lewis theory of counterfactuals, according to which a counterfactual is true just in case the closest world at which the antecedent is true is also a world at which the consequent is true.16 And here closeness can roughly be thought of as similarity. Now suppose that

16

This is a serious simplification, and more nearly resembles Stalnaker’s theory. Lewis thinks that there need not be a unique closest world, and indeed thinks that there need not even be a closest world (e.g., in cases where we have an infinite sequence of worlds that keep getting closer and closer to the actual world). So Lewis has to complicate his semantics. These complications render Lewis’s theory substantively different from Stalnaker’s. For example, Lewis’s theory fails to validate conditional excluded middle, while Stalnaker’s does. But these differences are not relevant when it comes to the inference from (12) to (13).

123

The reformulation argument: reining in Gricean pragmatics

535

the closest world at which Jordan has tea or coffee is a world at which she has coffee but not tea. Perhaps she enjoys coffee and drinks it frequently because it provides her with so much energy, but never drinks tea because it makes her lethargic. In this case, according to the simplistic Stalnaker–Lewis theory, (12) is true; the closest world at which Jordan drinks tea or coffee is one at which she drinks coffee and feels energized. However, according to this theory, (13) is false; the closest world at which she drinks tea is one at which she drinks tea and feels lethargic, not energized. So the simplistic Stalnaker–Lewis theory fails to capture the intuitive inferential connection between (12) and (13).17 Some theorists propose explaining this inference as a conversational implicature, as opposed to a semantic entailment. See, for example, Bennett (2003: 168–170). But, regardless of how the details are spelled out, this move is undermined by the following observations. My belief that Jordan would feel energized if she were to drink tea or coffee rationally commits me to the belief that Jordan would feel energized if she were to drink tea. Similarly, the mere fact that Jordan would feel energized if she were to drink tea or coffee guarantees that Jordan would feel energized if she were to drink tea. This inference survives the move to the level of rational commitment and the move to the level of mere facts. Hence, it cannot be explained in conversational terms, since rational commitment and mere facts have nothing to do with conversation. 6.2 Ross’s paradox concerning obligation Consider Ross’s paradox concerning obligation.18 Suppose that we are discussing Jordan’s goal of getting to Springfield tonight. In this conversation, an assertion of (14) enables audience members to infer (15): (14) (15)

Jordan must take the A train or B train (given her goal). Jordan may take the A train (given her goal).

One aspect of Ross’s paradox is that the traditional semantic theory of necessity fails to predict this intuitive inference. Why does the traditional semantic theory of necessity fail to predict this inference? Because the traditional semantic theory treats necessity as universal quantification (over ‘‘accessible’’ possibilities). This contrasts with the traditional treatment of possibility as existential quantification. Universal quantification, existential quantification, and disjunction do not interact in the way they would need to in order to have Jordan must take the A train or the B train (given her goal of getting to Springfield) entail Jordan may take the A train

17 Loewer (1976) emphasizes just how devastating this problem is for the Lewisian theory of counterfactuals. 18 See Ross (1941). As with the paradox of free choice permission, this puzzle was first discussed with respect to deontic modality, even though it arises for all varieties of modality. As before, in order to avoid the distracting complexities of deontic modality, I will pick a different variety of modality. Namely, I will focus on teleological, or goal-directed, modality. This variety of modality concerns what one may or must do, given the challenge of achieving a certain goal.

123

536

Z. Miller

(given her goal of getting to Springfield). This is because Vx(Fx _ Gx) does not entail AxFx. Some theorists propose explaining this inference as a conversational implicature, as opposed to a semantic entailment.19 But, again regardless of how the details might be spelled out, this move is undermined by the following observations. My belief that Jordan must take the A train or the B train (given her goal of getting to Springfield) rationally commits me to the belief that Jordan may take the A train (given her goal of getting to Springfield). Similarly, the mere fact that Jordan must take the A train or the B train (given her goal of getting to Springfield) guarantees that Jordan may take the A train (given her goal of getting to Springfield). This inference survives the move to the level of rational commitment and the move to the level of mere facts. Hence, it cannot be explained in conversational terms, since rational commitment and mere facts have nothing to do with conversation.

7 Some conversational implicatures that are not affected The reformulation argument is remarkably powerful, but there are some pragmatic stories to which it does not apply. Many inferences are confined to the realm of conversation; they do not survive the move to epistemology or the move to metaphysics. The reformulation argument does not undermine any pragmatic stories concerning such inferences. In this section, by way of illustration, I consider three kinds of conversational implicatures that are not undermined by the reformulation argument.20 7.1 Particularized conversational implicatures ‘‘Particularized’’ conversational implicatures, ones that rely on one-off features of the particular conversational context, remain unscathed. Consider Grice’s famous handwriting case.21 According to a variant of this case, if a professor is asked to evaluate the philosophical ability of Jones, and all the professor says in response is that Jones has good handwriting, this implicates, by way of glaring omission, that Jones is a bad philosopher. The reformulation argument does not undermine this pragmatic explanation. This is because the inference at issue does not survive the move to the level of rational commitment or the move to the level of mere facts: my belief that Jones has good handwriting does not rationally commit me to the belief that he is a bad philosopher; the mere fact that Jones has good handwriting does not guarantee that he is a bad philosopher. Because the inference appears confined to the realm of conversation, there is nothing untoward about using a pragmatic explanation.

19

See, for example, von Fintel (2012).

20

This section is not intended to be an exhaustive list. Other kinds of conversational implicatures are unscathed as well. 21

See Grice (1989: 33).

123

The reformulation argument: reining in Gricean pragmatics

537

7.2 The exclusivity of disjunction Many generalized conversational implicatures are also unscathed.22 Consider the exclusivity of disjunction, i.e., the fact that asserting a disjunction such as A or B often conveys the information that not both A and B. The traditional semantic theory of or fails to explain this inference, and historically it has been explained pragmatically. Again, the reformulation argument does not undermine this pragmatic rescue. My belief that Jordan is asleep or at work does not rationally commit me to the belief that she is not both asleep and at work. Similarly, the mere fact that Jordan is asleep or at work does not guarantee that she is not both asleep and at work. The exclusivity inference does not survive the move to the level of rational commitment or the move to the level of mere facts. Hence, there is nothing untoward about using pragmatics to explain what is going on at the level of assertion. 7.3 Scalar implicatures involving quantifiers Or consider the scalar implicatures that arise with some.23 When I assert Some spicy food contains capsaicin, this typically conveys that not every spicy food contains capsaicin. The traditional theory of some fails to explain this inference, and historically it has been explained pragmatically. As before, the reformulation argument does not undermine this pragmatic rescue. My belief that some spicy food contains capsaicin does not rationally commit me to the belief that not every spicy food contains capsaicin. Similarly, the mere fact that some spicy food contains capsaicin does not guarantee that not every spicy food contains capsaicin. The inference does not survive the move to the level of rational commitment or the move to the level of mere facts. Hence, there is nothing untoward about using pragmatics to explain what is going on at the level of assertion.

8 Addressing five objections In this section I address five objections to the reformulation argument.

22 A generalized conversational implicature depends only on general facts that are common to all, or almost all, conversations. Accordingly, generalized conversational implicatures are much more robust than particularized conversational implicatures. Slight changes to the context of utterance tend not to disrupt generalized conversational implicatures, though they can easily disrupt particularized conversational implicatures. 23

A scalar implicature is one based on the principle that, when additional information is required by the purpose of the conversation, speakers should prefer more informative assertions to less informative assertions. We already encountered a scalar implicature when we considered an assertion of Most spicy foods contain capsaicin.

123

538

Z. Miller

8.1 Reformulations are still linguistic According to this objection, the epistemological and metaphysical reformulations still involve linguistic questions concerning assertions. Consider the case of free choice permission. The proponent of a pragmatic rescue thinks that we should develop a pragmatic explanation of why an assertion of (16) enables audience members to infer (17) and to infer (18). (16) (17) (18)

Jordan can bench press 100 pounds or squat 200 pounds. Jordan can bench press 100 pounds. Jordan can squat 200 pounds.

According to the reformulation argument, a pragmatic explanation of this type would fail to explain why the belief that Jordan can bench press 100 pounds or squat 200 pounds rationally commits one to the belief that Jordan can bench press 100 pounds and rationally commits one to the belief that Jordan can squat 200 pounds. A pragmatic explanation of this type would also fail to explain why the mere fact that Jordan can bench press 100 pounds or squat 200 pounds guarantees that Jordan can bench press 100 pounds and guarantees that Jordan can squat 200 pounds. But a possible rejoinder runs as follows. When we wonder why the belief that Jordan can bench press 100 pounds or squat 200 pounds rationally commits one to the belief that Jordan can bench press 100 pounds and rationally commits one to the belief that Jordan can squat 200 pounds, what we are really wondering is why an assertion of (19) enables audience members to infer (20) and to infer (21): (19) (20) (21)

Agent S believes that Jordan can bench press 100 pounds or squat 200 pounds. Agent S is rationally committed to the belief that Jordan can bench press 100 pounds. Agent S is rationally committed to the belief that Jordan can squat 200 pounds.

So what we are really asking is how do the constructions ‘‘S believes that…’’ and ‘‘S is rationally committed to…’’ function. Similarly, when we wonder why the mere fact that Jordan can bench press 100 pounds or squat 200 pounds guarantees that Jordan can bench press 100 pounds and guarantees that Jordan can squat 200 pounds, what we are really wondering is why an assertion of (22) enables audience members to infer (23) and to infer (24): (22) (23) (24)

It is a fact that Jordan can bench press 100 pounds or squat 200 pounds. It is a fact that Jordan can bench press 100 pounds. It is a fact that Jordan can squat 200 pounds.

So what we are really asking is how does the construction ‘‘It is a fact that…’’ function. This strategy struggles to keep everything linguistic. Instead of construing the epistemological and metaphysical reformulations as about epistemology and metaphysics, this strategy insists that these reformulations are really about linguistic

123

The reformulation argument: reining in Gricean pragmatics

539

constructions: ‘‘S believes that…’’ and ‘‘S is rationally committed to’’, in the case of the epistemological reformulation, and ‘‘It is a fact that…’’, in the case of the metaphysical reformulation. This objection depends on skepticism about our ability to direct our attention to specific questions. It seems that I have the power to wonder why a belief comes with certain rational commitments. But this objection insists that my efforts to direct my attention to this topic fail. This objection insists that what I am really wondering is how ‘‘S believes that…’’ and ‘‘S is rationally committed to’’ function linguistically. Similarly, it seems that I have the power to wonder why a fact guarantees something. But this objection insists that my efforts to direct my attention to this topic fail. This objection insists that what I am really wondering is how ‘‘It is fact that…’’ functions linguistically. I find this skepticism unpersuasive. Why think that our inquires are confined to linguistic questions? Not all epistemology is about epistemic language. Not all metaphysics is about metaphysical language. There is, though, a prima facie more plausible objection that is similar to this one. According to this similar objection, it is not that we are mistaken about the questions that we are asking; rather, it is that pragmatic considerations corrupt the intuitions that we use in attempting to answer these questions. I take up this objection at length in Sect. 8.4. 8.2 Beliefs are mental sayings, or we don’t have an intuitive grip on them24 This objection targets the epistemological reformulation. According to this objection, either a belief amounts to saying something to oneself, or it does not. If a belief does amount to saying something to oneself, then perhaps we can think of a belief as a conversation with oneself, and perhaps we can tell a Gricean story about such a conversation. On the other hand, if a belief does not amount to saying something to oneself, then who knows what it is—probably a theoretical entity to which our commonsense intuitions should not be applied. First of all, even if some beliefs amount to saying something to oneself, not every belief does. There is an ordinary sense of belief according to which I believe Napoleon always weighed less than 1000 pounds and I believe that he always weighed less than 1001 pounds and I believe that he always weighed less than 1002 pounds, and so on, for all natural numbers greater than 1002. But I have certainly not said every member of this infinite sequence to myself.25 24

Thanks to an anonymous referee for bringing this objection to my attention.

25

One might object that we do not really believe these things until we consider them and say them to ourselves. Perhaps, prior to mentally articulating a belief, we merely possess a disposition to form the belief. There are at least two responses. First, the view that we do not believe things until we consider them is at odds with the commonsense interpretation of many cases. If I am a committed vegan and firmly believe that we should not unnecessarily harm any animal, then, intuitively, I believe that we should not unnecessarily harm ducks, even if I have never explicitly considered whether we should unnecessarily harm ducks. It is contrary to commonsense to claim that I currently lack this belief, and that I instead merely possess the disposition to form this belief after considering the issue. Second, even if we accept that such cases involve dispositions to believe, we can recast my reformulation argument in terms of these dispositions. Just because I am disposed to believe that Oswald killed Kennedy, it does not follow that I

123

540

Z. Miller

Let’s use the term ‘‘unsaid belief’’ to encompass all beliefs that do not involve saying something to oneself. We can provide further evidence that commonsense recognizes unsaid beliefs. First of all, it is part of commonsense that sometimes we believe something but are not sure whether we believe it. Perhaps in hindsight it is clear that, all along, I believed that a certain student was cheating, even though for most of the semester I could not bring myself to consciously recognize this belief. This is a problem for the view that belief always involves saying something to oneself. If we sometimes do not know whether we believe something, but always know whether we are saying something to ourselves, then belief, at least sometimes, will not involve saying something to oneself. Secondly, it is part of commonsense that we can say things to ourselves that we do not believe. For example, I might say some motivational things to myself every morning on the advice of a psychologist, and I might do this precisely because I do not yet believe them. So there must something more to a belief than a mere mental saying. But what is this additional component? The obvious response is to say that a belief is a sincere mental saying. But how can we distinguish a sincere mental saying from an insincere mental saying? The obvious response here is to say that whether or not we believe what we are mentally saying determines whether a mental saying is sincere or insincere. But we cannot, on pain of circularity, hold both (i) that whether I believe what I am mentally saying depends on my sincerity and (ii) that whether a mental saying is sincere depends on whether I believe what I am mentally saying. Hence, a conception of belief according to which beliefs are mental sayings has serious defects. Of course, commonsense probably is defective in various ways, but surely we should only ascribe defective views to commonsense when we have a strong reason to do so. I know of no strong reason to ascribe to commonsense the view that beliefs are always mental sayings. Given (i) my examples concerning Napoleon, (ii) the commonsense view that sometimes we believe something but are not sure whether we believe it, and (iii) the defectiveness of the view that every belief is a mental saying, I maintain that commonsense recognizes unsaid beliefs. Given that unsaid beliefs are part of commonsense, we do not lose our intuitive grip on them. How exactly unsaid beliefs should be theoretically cashed out is beyond the scope of this paper, as it does not have an effect on my line of argument. For example, whether unsaid beliefs should be cashed out in functionalist terms is beyond the scope of this paper and orthogonal to my line of argument. Furthermore, and importantly, unsaid beliefs are subject to rationality, so possessing an unsaid belief brings with it rational commitments. Return to the Napoleon example. There is an ordinary sense of belief according to which I believe Napoleon always weighed less than 1000 pounds. This belief brings with it rational commitments. For example, given some minimal assumptions about my other

Footnote 25 continued am disposed to believe that if Oswald didn’t kill Kennedy, a Martian did. So the reformulation argument remains intact, in one form or another, regardless of how this particular debate plays out. Thanks to Wayne Davis for raising this last point.

123

The reformulation argument: reining in Gricean pragmatics

541

beliefs, this belief rationally commits me to the belief that Napoleon never weighed more than 2000 pounds. So unsaid beliefs (i) are ordinary beliefs on which we have an intuitive grip, and (ii) bring with them rational commitments. Lastly, unsaid beliefs do not involve a speaker, an audience, a conversation, or conversational rules. So it does not make sense to apply Gricean pragmatics to them. Now, with the notion of unsaid belief in hand, we can recast my epistemological reformulation test in terms of unsaid beliefs. When we ask whether the belief that Jordan can bench press 100 pounds or squat 200 pounds rationally commits one to the belief that Jordan can bench press 100 pounds, we are not asking a question that concerns saying Jordan can bench press 100 pounds or squat 200 pounds to oneself. Rather, we are asking a question that concerns the unsaid belief that Jordan can bench press 100 pounds or squat 200 pounds. 8.3 Pragmatics can appropriately apply to non-linguistic domains According to this objection, the epistemological and metaphysical reformulations really do concern non-linguistic domains, but pragmatics, or something close enough, can intelligibly be applied to these domains. Douven (2012), for example, has argued that belief and acceptance are subject to some pragmatic considerations. As he puts it: At the root of Grice’s thinking was the insight that when uttering a sentence we often convey more information than what is contained in the sentence’s semantic content. Due to this—Grice convincingly argued—even a patently true sentence may be unassertable. For, truth being solely a matter of semantic content, by asserting a true sentence one may still ‘‘suggest,’’ or ‘‘implicate,’’ something false, thereby possibly misleading one’s audience. I will argue that much the same holds for what we can accept: a sentence may be unacceptable even if its truth is patent to us. More specifically, the claim will be that whether a sentence is acceptable is not merely a matter of our having evidence for it, or good reasons for believing it is true, or anything of that sort, but also of whether it satisfies certain further conditions, conditions that are akin to (some of) Grice’s celebrated conversational maxims. Douven (2012: 35) In the case of assertion, pragmatic considerations keep the conversation running well; they facilitate information-exchange among the conversational participants. In the case of belief, according to Douven (2012: 38), we obey the Epistemic Hygienics principle, which demands that you ‘‘not accept sentences that could mislead your future selves’’. Using this principle, we could perhaps construct pragmatic explanations at the level of belief. If this were on the right track, then we could give similar pragmatic stories at both the level of assertion and the epistemological level of rational commitment. For the purposes of this paper, it is sufficient to note that even if Douven’s argument undermines the epistemological reformulation, it does not undermine the metaphysical reformulation. No pragmatic story could be told with respect to the metaphysical reformulation. The epistemological reformulations involve agents

123

542

Z. Miller

(since they involve believers with rational commitments). Agents have goals, and there are principles that they should adhere to in order to further these goals, and some of these principles might (i) concern belief and (ii) resemble Gricean principles. But the metaphysical reformulations do not even involve agents. They just involve facts.26 Furthermore, Douven’s argument has obstacles to overcome, as he himself admits. It relies on a picture of belief according to which having a belief involves storing a structured sentential-like entity in some part of the mind. Whether this is the correct picture of belief is a contentious issue in the philosophy of mind. How this debate plays out will determine the viability of Douven’s argument. The less plausible this picture of belief, the less plausible Douven’s argument. 8.4 Pragmatics corrupts our intuitions about non-linguistic domains According to this objection, the epistemological and metaphysical reformulations really do concern non-linguistic domains, and it is misguided to apply pragmatics to these non-linguistic domains, but our intuitions concerning these domains are corrupted or infected by pragmatic considerations. According to this line of thought, our intuitions about the belief that Jordan can bench press 100 pounds or squat 200 pounds, and our intuitions about the rational commitments that accompany this belief, are corrupted by our intuitions about what an assertion of Jordan can bench press 100 pounds or squat 200 pounds would conversationally implicate. Furthermore, our intuitions about the fact that Jordan can bench press 100 pounds or squat 200 pounds, and our intuitions about what this fact guarantees, are similarly corrupted. Because pragmatic considerations corrupt these intuitions, and because the reformulation argument depends on these intuitions, the reformulation argument fails. First of all, it is important to note that this objection depends on a skepticism concerning the purity and reliability of our intuitions. This kind of skepticism threatens to undermine the methodology employed throughout most of philosophy. It is not a kind of skepticism that we should appeal to lightly. It may, in the end, turn out that our intuitions are generally unreliable.27 But this conclusion will only be drawn after an long and complex debate. And this is not a debate that I will attempt to adjudicate here. Secondly, and much more conclusively, this objection renders mysterious the fact that the reformulation argument only undermines some conversational implicatures. Some conversational implicatures left unscathed are discussed in Sect. 7. If our

26

If these facts concern agents, that is completely incidental.

27

Some proponents of experimental philosophy argue for this conclusion. Consider, for example, the following claim found in Swain et al. (2008: 140): ‘‘the results of [some experimental philosophy research] challenge the legitimacy of appealing to intuitions. Weinberg et al. (2001, 2003) revealed that epistemological intuitions vary according to factors such as cultural and educational background; Machery et al. (2004) document a similar cultural variation in semantic intuitions; and Nichols and Knobe (2007) have discovered that the affective content of a thought experiment can influence whether subjects have compatabilist or incompatabilist intuitions.’’

123

The reformulation argument: reining in Gricean pragmatics

543

epistemological and metaphysical intuitions are corrupted by our intuitions about assertions and the inferences that these assertions enable us to make, then what explains the fact that some inferences reappear in the epistemological and metaphysical reformulations and some do not. In other words, what could explain the difference between (i) the inferences that arise with respect to the paradox of free choice permission, counterfactuals with disjunctive antecedents, and Ross’s paradox, on the one hand, and (ii) the inferences that arise with respect to Grice’s handwriting case, the exclusivity of disjunction, and the scalar implicatures associated with some, on the other hand? 8.5 The cancelability diagnostic The reformulation argument provides us with a diagnostic that can help us determine whether an implied piece of information is a conversational implicature: if the implication survives reformulation in epistemological or metaphysical terms, then it is not a conversational implicature. But how does this diagnostic interact with the classic diagnostic of cancelability?28 According to the cancelability diagnostic, semantic entailments cannot be canceled by the addition of further information, but conversational implicatures can be. For example, after I assert (25), nothing I subsequently say (short of a full-blown retraction) can cancel the implication that some spicy food contains capsaicin. (25)

Most spicy foods contain capsaicin.

On the other hand, after I assert (25), I can subsequently assert (26) in order to cancel the implication that not every food contains capsaicin. (26)

Indeed, every spicy food contains capsaicin.

This is taken to be evidence that (25) semantically entails that some spicy food contains capsaicin, but only conversationally implicates that not every spicy food contains capsaicin. What should we do when our newly developed diagnostic, cast in terms of reformulation, conflicts with the cancelability diagnostic? That is to say, what should we do when a set of inferences (i) can be canceled and (ii) can be reformulated in epistemological or metaphysical terms? Because the inferences can be canceled, the cancelability diagnostic suggests that they are conversational implicatures. Because the inferences can be reformulated in epistemological or metaphysical terms, the reformulation argument suggests that they are not conversational implicatures. How should we resolve this tension? The answer is that we can only decide on a case-by-case basis. It is well-known that the cancelability diagnostic sometimes yields false positives.29 More specifically, the diagnostic yields a false positive when the addition of further information, instead of canceling an implicature, actually resolves some sort of ambiguity or indeterminacy in 28

Grice first discussed the cancelability diagnostic. See Grice (1989: 44–46).

29

See, for example, Sadock (1978).

123

544

Z. Miller

such a way as to highlight a typically dispreferred interpretation. For example, typically an assertion of (27) implies (28) and implies (29): (27) (28) (29)

Jordan didn’t have coffee or tea with breakfast. Jordan didn’t have coffee with breakfast. Jordan didn’t have tea with breakfast.

But these implications can be ‘‘canceled’’, as in (30): (30)

Jordan didn’t have coffee or tea with breakfast, but I can’t remember which she didn’t have.

But in this case we should not conclude that these inferences are conversational implicatures. The natural explanation for this ‘‘cancelability’’ runs as follows. These inferences typically hold because the preferred interpretation of (27) has the negation taking wide scope with respect to disjunction: :(coffee _ tea). This reading implies :coffee and implies :tea. But there is another, albeit dispreferred, reading whereby the negation takes narrow scope: :coffee _ :tea. This reading does not imply :coffee and does not imply :tea. In (30), the additional information resolves the ambiguity in favor of this dispreferred reading. The lesson is that the cancelability diagnostic can fail. Whenever the reformulation argument and the cancelability diagnostic yield different verdicts, we should see whether we can explain away the cancelability in terms of ambiguity resolution or indeterminacy resolution. The example in the previous paragraph illustrates this. There we can explain away the cancelability in terms of scopal ambiguity resolution. Hence, the cancelability of those inferences does not suggest that the inferences are conversational implicatures. In every case that I have looked at where the cancelability diagnostic and the reformulation argument yield different verdicts, there is a plausible story explaining away the cancelability in terms of ambiguity resolution or indeterminacy resolution.

9 Conclusion The reformulation argument shows that the pragmatic rescue strategy—one of the most widely used strategies in philosophy of language, philosophical logic, and formal semantics—is misguided in a wide range of cases. By recasting problematic inferences in non-conversational terms—by reformulating them in epistemological or metaphysical terms—we can demonstrate that these problematic inferences are not to be explained away pragmatically. The significance of this result is sweeping. We have a new reason to reject the pragmatic stories that Grice (1975) develops in an attempt to defuse the paradoxes of material implication. We have a new reason to reject the pragmatic story with which Bennett (2003) tries to accommodate counterfactuals with disjunctive antecedents. We have a new reason to reject the many pragmatic stories, such as the one found in Kratzer and Shimoyama (2002), that have been developed in the hope of rescuing the traditional theory of possibility from the problem of free choice

123

The reformulation argument: reining in Gricean pragmatics

545

permission. And, lastly, we have a new reason to doubt the prediction, found in von Fintel (2012), that there is a pragmatic story that can rescue the traditional theory of necessity from Ross’s paradox. Undoubtedly, there are other debates to which the reformulation argument can be applied. Acknowledgments I would like to thank Heather Demarest, Josh Armstrong, Wayne Davis, Andy Egan, Thony Gillies, Ernie Lepore, Lisa Miracchi, Martin Montminy, Matthew Stone, Brian Weatherson, two anonymous referees, and many others—including audiences at Rutgers University and Purdue University—for helpful comments on the ideas contained in this paper.

References Aloni, M., & van Rooij, R. (2007). Free choice items and alternatives. In G. Bouma, I. Kra¨mer, & J. Zwarts (Eds.), Cognitive foundations of interpretation. Amsterdam: Edita KNAW. Alonso-Ovalle, L. (2006). Disjunction in alternative semantics. University of Massachusetts at Amherst Dissertation. Bennett, J. (2003). A philosophical guide to conditionals. Oxford: Oxford University Press. Chierchia, G. (2004). Scalar implicatures, polarity phenomena and the syntax/pragmatics interface. In A. Belletti (Ed.), Structures and beyond (pp. 39–103). Oxford: Oxford University Press. Chierchia, G. (2006). Broaden your views: Implicatures of domain widening and the ‘‘logicality’’ of language. Linguistic Inquiry, 37(4), 535–590. Douven, I. (2012). The pragmatics of belief. Journal of Pragmatics, 42, 35–47. Edgington, D. (1995). On conditionals. Mind, 104(414), 235–329. Fox, D. (2007). Free choice permission and the theory of scalar implicatures. In U. Sauerland & P. Stateva (Eds.), Presupposition and implicature in compositional semantics. Palgrave studies in pragmatics, language and cognition. New York: Palgrave Macmillan. Geurts, B. (2010). Quantity implicatures. Cambridge: Cambridge University Press. Gillies, A. (2004). Epistemic conditionals and conditional epistemics. Nous, 38, 585–616. Green, M. (1995). Quantity, volubility, and some varieties of discourse. Linguistics and Philosophy, 18, 83–112. Grice, H. P. (1975). Logic and conversation. In P. Cole & J. Morgan (Eds.), Syntax and semantics (Vol. 3, pp. 41–58). New York: Speech Acts, Academic Press. Grice, H. P. (1989). Studies in the way of words. Cambridge: Harvard University Press. Harman, G. (1986). Change in view: Principles of reasoning. Cambridge: MIT Press. Kamp, H. (1973). Free choice permission. Proceedings of the Aristotelian Society, 74, 57–74. King, J., & Stanley, J. (2005). Semantics, pragmatics, and the role of semantic content. In Z. G. Szabo (Ed.), Semantics versus Pragmatics. Oxford: Oxford University Press. Klinedinst, N. (2007). Plurals, possibilities, and conjunctive disjunction. UCL Working Papers in Linguistics, 19, 261–284. Kratzer, A., & Shimoyama, J. (2002). Indeterminate pronouns: The view from Japanese. In Y. Otsu (Ed.), The proceedings of the third Tokyo conference on psycholinguistics (pp. 1–25). Tokyo: Hituzi Syobo. Lewis, D. (1973). Counterfactuals. London: Basil Blackwell. Loewer, B. (1976). Counterfactuals with disjunctive antecedents. Journal of Philosophy, 73, 531–537. Machery, E., Mallon, R., Nichols, S., & Stich, S. P. (2004). Semantics, cross-cultural style. Cognition, 92, B1–B12. Nichols, S., & Knobe, J. (2007). Moral responsibility and determinism: The cognitive science of folk intuitions. Nous, 41, 663–685. Ross, A. (1941). Imperatives and logic. Theoria, 7, 53–71. Sadock, J. (1978). On testing for conversational implicature. In P. Cole (Ed.), Pragmatics, syntax and semantics (Vol. 9, pp. 281–297). New York: Academic Press. Schulz, K. (2005). A pragmatic solution for the paradox of free choice permission. Synthese, 147(2), 343–377. Speaks, J. (2008). Conversational implicature, thought, and communication. Mind and Language, 23(1), 487–505.

123

546

Z. Miller

Stalnaker, R. (1968). A theory of conditionals. In N. Rescher (Ed.), Studies in logical theory. American philosophical quarterly monograph series (Vol. 2, pp. 98–112). Oxford: Blackwell. Swain, S., Alexander, J., & Weinberg, J. M. (2008). The instability of philosophical intuitions: Running hot and cold on truetemp. Philosophy and Phenomenological Research, 76(1), 138–155. van Benthem, J. (1996). Exploring logical dynamics. Stanford, CA: Center for the Stanford, Study of Language and Information (CSLI) Publications. von Fintel, K. (2012). The best we can (expect to) get? Challenges to the classic semantics for deontic modals. Presented in a session on Deontic Modals at the Central APA, February 17, 2012. von Wright, G. H. (1951). Deontic logic. Mind, 60(237), 1–15. Weinberg, J. M., Nichols, S., & Stich, S. (2001). Normativity and epistemic intuitions. Philosophical Topics, 29(1), 429–460. Zimmerman, T. E. (2000). Free choice disjunction and epistemic possibility. Natural Language Semantics, 8, 255–290.

123