1. Introduction

Imagination and Convention is a provocative and stimulating book. When asked to give an example of a clear case where philosophy has made indisputable progress, one of the few candidates that comes to mind is something like a broadly construed Gricean theory of conversational implicatures: something along the lines of Grice’s theory has to be right, that is any theory that essentially involves inferential processes, Relevance Theory included.¹ Nevertheless, in their book, Lepore and Stone make a pretty good case against anything that resembles a Gricean theory. They argue extensively that there is no real theoretical use for the notion of conversational implicature. It is not just that they disagree with a Gricean account of conversational implicatures, they deny the phenomenon altogether. They also offer a plausible alternative to a Gricean theory of linguistic communication: one that relies heavily on linguistic conventions and imagination, rather than on pragmatic mechanisms. A broadly construed Gricean theory of conversational implicature is no longer indisputable.

In a sense, Gricean theory and Lepore and Stone’s theory are radically different. The former theory allows a simplification of semantics by complicating pragmatics, while the latter seeks to simplify pragmatics by complicating semantics. Of course, each theory argues for principled reasons to go one way or another. Grice’s core insight is that a conversation is a rational activity where communication is the goal and, as such, rational participants to a conversation are cooperative in their contributions -they tend to follow the Cooperative Principle and derived Maxims. A Gricean would use this to explain many linguistic regularities. If there is a fact floating around that could be used to explain certain phenomenon, it is a good idea to make use of it. We can get a lot of mileage from the fact that conversations are a rational activity. Speakers tend to exploit the Cooperative Principle to communicate information that goes well beyond the literal meaning of the sentences they utter. This is communication grounded in pragmatic mechanisms. Let’s consider a few examples:

Grice would think that the literal meaning of sentences (1)-(3) is quite simple. In (1) it is just a question regarding the possibility of having French Toast, in (2) it is just a classical conjunction without reference to a time order, and in (3) it is that the object in question looked red. The semantics of these sentences is not at all fancy. It is just what gets compositionally determined by the subsentential constituents and their order -no ambiguities or hidden variables come into play. However, the information that normally gets conveyed by utterances of these sentences is quite different. If you utter (1) in a context where a waiter is taking your order, the information that gets across is a request to get the French Toast. Normally, an utterance of (2) conveys the information that first the Oil prices doubled and then demand of consumer goods plunged. Finally, an utterance of (3), in the right kind of situation, conveys the information that the object in question wasn’t really red. Grice would think that all these are conversational implicatures -they are calculable on the basis of the literal meaning of the sentences uttered and the assumption that the speaker is following the Cooperative Principle. I will spare you the well known details about how calculations of this kind would go. It suffices to point out that, according to Grice, in this kind of case interlocutors grasp what the speaker wants to convey based on pragmatic reasoning -their linguistic knowledge only plays a modest role.

Lepore and Stone can grant that a conversation is a rational activity and that oftentimes participants are cooperative. However, they do not think that one should explain cases like (1)-(3) in terms of conversational implicatures. There is another kind of fact floating around and it is better to explain those cases on its basis. We have many conventions, and many of the things that we do should be explained on the basis of our knowledge of those conventions. In most countries it is a convention to drive on the right, and it is in good part because Ana knows about this convention that she is now driving on the right. According to ^{Lewis (1969)}, a convention is a regularity of a certain kind. Roughly, when a community reliably solves a coordination problem in the same way -i.e., whether to drive on the right or the left- and in the community there is a mutual expectation that everyone solves the problem in that way, then there is a convention to solve the problem in that way. As there are social conventions, there are also linguistic conventions. For instance, “red” could have meant yellow or blue, but it is a linguistic convention in our linguistic community that it means red -we have coordinated on assigning that meaning for that word, and we expect everyone else in our community to do so. The same goes for every other word. Assume compositionality and you get that the meaning of sentences is also determined by our linguistic conventions. Lepore and Stone argue that linguistic conventions go far beyond those responsible for assigning meaning to linguistic expressions -for instance, there are conventions regarding intonation, tense, coherence relations, and informational structure. As such, it is to be expected that many linguistic regularities should be explained on the basis of our knowledge of those conventions.

Lepore and Stone argue that the proper explanation of cases (1)-(3) should be couched in terms of linguistic conventions, not conversational implicatures. Let’s take a brief look at how the explanation goes. They argue that example (1) is not an instance of a conversational implicature, because in the purported Gricean derivation there is an concealed appeal to the following linguistic convention: “as a matter of fact, the sentence she uttered is associated with a request”. (^{Lepore and Stone 2014, p. 99}). According to them, this shows that the communicative effects in this example are not grounded on Gricean cooperative reasoning: instead, linguistic conventions are getting the work done.² Evidence of this, they argue, is that nondetachability doesn’t hold -for this to hold, the conversational implicature has to survive when the sentence uttered is substituted by one with the same meaning.³ As they argue, an utterance of

(1′) Am I able to have the French Toast?

isn’t an indirect request to have the French Toast. But it should be, if we were dealing with a case of conversational implicature in (1). Now, if the Gricean account is wrong here, it becomes plausible to think that we are dealing with a case of ambiguity: “can” is ambiguous between a possibility reading and a request reading. Part of the appeal of a Gricean explanation in terms of conversational implicatures is that when it works it minimizes the appeal of positing ambiguities. However, when it doesn’t work, an explanation in terms of ambiguity looks quite plausible. Notice that ambiguity is a matter of convention: we have a linguistic convention where “can” has two possible readings. So the difference between (1) and (1′) is that only the former has a request reading.

Lepore and Stone are quite skeptical that case (2) is one where the temporal order is a matter of conversational implicature. This is so because, again, they think that a covert appeal to a linguistic convention appears in the purported Gricean derivation (^{Lepore and Stone 2014, p. 117}). As before, they claim that good evidence for this is that non-detachability doesn’t hold in this case. Consider the following equivalent sentence:

(2′) Oil prices have doubled and demand for consumer goods has plunged.

Here, they argue, the bit of information about first oil prices doubling and then consumer goods plunging doesn’t get across. This is good evidence that we are not dealing with a case of conversational implicature. Rather, they argue that temporal order is a matter of logical form. According to them “logical form [. . . ] determines a specification of a set of temporal relations among described eventualities” (p. 116). As before, if the example cannot be explained in terms of conversational implicatures, positing ambiguity feels quite plausible. So the view is that there is a way of disambiguating the logical form of a conjunctive statement, as in (2), such that it corresponds to a narrative (where the first conjunct precedes the second in temporal order) and there is a different way of disambiguating the logical form where this is not the case (as in (2′)). Hence, the explanation goes, given our linguistic conventions a conjunctive statement can have at least two logical forms specifying different temporal relations among the events described.

Lepore and Stone argue that the proper explanation of (3) should be couched in terms of intonation, rather than Gricean cooperative reasoning. The details of their account are numerous, but the main point is easy to state. When a speaker utters (3) to express limited agreement she does it with certain intonation. As a matter of fact, that intontation is conventionally associated in English with limited agreement -a different intonation would be used to express, say, surprise rather than any kind of disagreement. So the fact that the speaker manages to communicate limited agreement has to do with knowledge of linguistic conventions rather than pragmatic mechanisms. The point is quite straightforward: as plausible as a pragmatic account of (3) might look, if there is already a convention that explains the case, the pragmatic story cannot get off the ground.

So far we have considered some examples that traditionally have been taken to be cases of conversational implicatures, and we have looked at Lepore and Stone’s alternative explanations in terms of conventions. Now, there is another class of examples that have been explained in terms of conversational implicatures and that resist being accounted for in terms of conventions: metaphors, hints, sarcasm, irony and humor. This is significant, because unless Lepore and Stone can offer an alternative account of those examples, their whole project loses some traction: from a theoretical perspective, positing lots of ambiguities -as they do- feels less attractive if conversational implicatures are an open possibility.

Lepore and Stone understand figurative and evocative language in terms of open-ended invitations of a certain kind. For example, a metaphor is “an openended invitation to find insights from seeing one domain, analogically, as another”. Hints, irony, and humor require a different kind of imaginative engagement. Here is a crucial aspect of their views regarding figurative and evocative language: “But none of these imaginative processes involves anything like grasping some indirect or ulterior proposition that the speaker and the audience coordinate on, again in the technical sense at issue here” (p. 262). This is one of the main ways in which their treatment of figurative and evocative language is different from one coached in terms of conversational implicatures. In the case of conversational implicatures speaker and interlocutor want to coordinate on some bits of information -and it is plain to them that that information isn’t determined by the literal meaning of the sentence uttered. The view under discussion here is different. There is no specific information that speaker and interlocutor have to coordinate on: the objective of, say, a metaphor is not to convey a specific proposition, rather it is an open-ended invitation to use our imagination in a certain way. This view fits many instances of metaphor very well.

Conversational implicatures have been used to explain a wide range of phenomena. Lepore and Stone have good arguments to the effect that we should understand such phenomena either in terms of conventions or in terms of imagination. I have a few worries regarding this project. It is not clear to me that Lepore and Stone have ruled out the possibility of there being conversational implicatures. They have shown that there are far fewer conversational implicatures than we thought there were. In section 2, I argue that there are instances of figurative and evocative speech where speaker and interlocutor coordinate on a specific proposition: this cannot be understood either in terms of knowledge of conventions or in terms of open-ended invitations to use our imagination in some way. Perhaps it is better to understand these cases in terms of conversational implicatures. In section 3, I make some critical observations regarding their treatment of ambiguity.

2. Imagination

I agree that there are instances of figurative and evocative speech that are better understood in terms of open-ended invitations. When Matt Groening says “Love is like a snowmobile racing across the tundra [ . . . ]” it seems like a mistake to think that there is a specific proposition or set of propositions that the author wants us to reason our way to by assuming that he is following the Cooperative Principle.⁴ It feels more like some sort of open-ended invitation. Perhaps there are legitimate ways of interpreting the metaphor that the author hasn’t even considered. This suggests that we are not dealing with a case of conversational implicature. However, there are instances of figurative and evocative speech where the speaker wants to convey a specific bit of information. In this kind of case, one cannot understand the utterance properly without getting the specific bit of information that the speaker wants to convey.⁵

Here is the kind of case that I have in mind: I’ve told you that Olivia has accepted an invitation to my dinner party. Knowing that you are good friends with her I ask: “Is it very likely that she will show up to the party?” Here’s your reply: “Picture a feather lying on the sidewalk. It may fly one way or another depending on the wind, dogs, cars, or what have you.” Now, this may not be an exquisite metaphor, but there is a very specific bit of information you want to get across with it. If we don’t coordinate on it, I will be missing your point. Your contribution to the conversation is this:

(i)It’s not very likely that Olivia will show up to the party.

Notice that I can’t get to (i) only based on my knowledge of linguistic conventions -disambiguation doesn’t seem to play a mayor role here either. But, at the same time your answer to my question is not an open-ended invitation to do anything. Of course, I could go off to reflect about how our lives and choices depend on random factors outside of our control. That, however, is not what you intended me to do. You were trying to put (i) on the conversational record, and you succeeded. Yet, according to Lepore and Stone, that could not be done by way of using a metaphor.

Let’s think a bit about what the conversational record is and why it is important for Lepore and Stone to restrict what goes in there. The notion of conversational record is based on the notion of conversational score in ^{Lewis 1979}. Thus, we can think of it as “a definite, precise, and circumscribed inventory of the contributions that interlocutors have made to the conversation” (2014, p. 234). As Lewis points out, the role that the conversational record plays in a conversation is similar to the role a scoreboard plays in a baseball game. The scoreboard is affected by what happens in the game and it affects what happens in the game: each strike gets registered in the scoreboard and if you get three strikes on the scoreboard you are out. Similarly, the conversational record is affected by what happens in a given conversation and it also affects what can happen in that conversation: it is affected by keeping track of the interlocutor’s contributions to the conversation, and it affects it because many of the expressions we use get interpreted based on information in the conversational record.

All sorts of information gets registered in the conversational record (i.e., order of salience, standards of precision, borders of permissibility, questions under discussion). What’s important for the current argument is that the conversational record also keeps track of certain commitments:

the conversational record must track the COMMITMENTS of the interlocutors to the information that characterizes what they are doing in the conversation. [ . . . ] Take assertion. When a speaker asserts something, one effect is that she has made a public commitment to the proposition she has advanced. Normally, the other parties in the conversation must address this assertion in turn, by agreeing with it, disputing it, asking for clarification, or raising prerequisite questions on related issues whose answers are understood as preconditions for interlocutors’ judgment on the matter. (^{Lepore and Stone 2014, p. 248})

Thus, if an assertion is accepted, the asserted proposition makes it into the conversational record. Also, it is fair to say that some instances of figurative and evocative speech do not add information to the conversational record -this is in line with the view that they are open-ended invitations of some sort, rather than assertions that trigger commitments.

However, it seems clear that, in the example above, when you answer my question with “Imagine a feather lying on the sidewalk. . . " you are quite committed to (i) -the proposition that it’s not very likely that Olivia will show up to the party. This, of course, unless you explicitly cancel your commitment. As with assertions, I could challenge (i), or I could agree with you, or perhaps I could follow by asking for some details. After your metaphor I could immediately follow with: “How do you know? Has she missed many of your parties?" This is a pretty good sign that you are committed to (i).

Furthermore, after your answer to my question, (i) is readily available to settle pronoun disambiguation. Suppose that we both take for granted that we want to introduce Carla and Olivia at the party. I could follow your metaphor with:

(ii) Carla will be disappointed. They should get in touch by email, since it is not very likely that she will show up to the party.

It wouldn’t be so clear that “she” refers to Olivia unless (i) had just been added to the conversational record -if it were not in the record, it would be tempting to interpret the pronoun as referring to Carla. This is good evidence to the effect that (i) made it into the conversational record.

We have coordinated on a proposition, and all you did was to offer a metaphor. We didn’t coordinate by disambiguating or by simply relying on our knowledge of linguistic conventions. It seems that cases like this cannot be explained either by imagination or by convention. Some sort of Gricean explanation in terms of conversational implicatures looks quite appealing.

This is not an isolated case. Let’s consider another example to get clear on what the problem is. Grice thought hints are paradigmatic cases of conversational implicatures. Lepore and Stone think they are something else. On their view, how to interpret a hint is an open-ended matter. Let’s consider one of their examples. Just as a student has handed in his exam the examiner says:

(iii) You might want to look at your answer to problem three again.

Lepore and Stone’s point is that even if the examiner has some specific correction in mind, she has chosen not to be explicit about what the correction is: it could be that there is a mistake, or that the enumeration is wrong, an obvious spelling mistake, or something else. Moreover, they think that after uttering (iii) the examiner is committed to very little:

Does she explicitly commit herself to there being an error in the students solution? Or, to its incompleteness? or to the student’s misnumbering his solution? Maybe she wants him to realize that he forgot to do it altogether. Maybe she wants to call his attention to general considerations: Students should always review solutions to problems on exams, for example. Such interpretive possibilities seem open-ended. (^{Lepore and Stone 2014, p. 188})

The examiner is not committed to a specific bit of information, so there is nothing to coordinate over, so nothing makes it into the conversational record. I won’t object to this. However, after a small tweak, the example delivers a very different result. Suppose that it is common ground between the student and examiner that the student always makes one and only one spelling mistake in his exams. The student is quite visibly obsessed with the fact that he hasn’t been able to correct this problem. Before the exam the student tells the examiner that he is determined not to have a single spelling mistake this time. An hour later, just as the student hands in his exam, the examiner utters (iii). The student immediately looks for a spelling mistake in the right place. Sure enough, there it is.

The question is whether the following proposition made it into the conversational record.

(iv) There is a spelling mistake in the student’s answer to question three.

It seems that the examiner is committed to it: the student could accept or challenge that the mistake is there. Furthermore, if she didn’t want to communicate that proposition, she should have said something like: “Spelling is flawless, but you might want to look at your answer to question three again.” Also notice that it would be quite natural for the student or the examiner to follow (iii) with:

(v) I cannot believe it’s there!

Where, of course, “it” refers to the spelling mistake. This is good evidence that after the utterance of (iii) both of them are coordinated on (iv). If so, (iv) got into the conversational record.

It seems like we can use figurative and evocative speech to add a proposition to the conversational record. However, Lepore and Stone are strongly committed to this not being possible.

We have assumed that the information that interlocutors draw from figurative and evocative speech stays “off the record”. This assumption is crucial to our argument. The record is an abstraction, constituted by rules that interlocutors implicitly or explicitly agree to. Only coordination can update the record. (2014, p. 262)

What I have tried to show is that sometimes the commitments one acquires by using figurative and evocative speech are not that different from the ones one acquires by asserting plain sentences. Interlocutors sometimes manage to coordinate on a proposition by way of using figurative and evocative speech. I do not see any principled reason to keep this information off the conversational record.

After considering our two examples it is rather clear why sometimes figurative and evocative speech feels more like an open-ended invitation to use our imagination in certain ways, whereas in other cases it feels more like an attempt to coordinate on a specific proposition. This difference depends heavily on how specific the question under discussion is. I follow ^{Roberts (1996)} in thinking of the goal of discourse as an attempt to answer the question under discussion, which can also be thought of as the immediate topic of discussion. For my two examples to get off the ground it is important that the questions under discussion are as specific as this:

Notice that my examples are set up in such a way that it is clear that the goal of discourse is to answer either (a) or (b), depending on the example. Now, the point is that it is because these questions are so specific that the interlocutors can coordinate on an answer even if the input is the particular metaphor or a hint in the examples above. Of course, the idea is not that given any question under discussion that is specific enough one can communicate a particular proposition regardless of the metaphor or hint one uses. The view is rather that given some very specific questions under discussion, some metaphors or hints can be used to communicate something quite precise.

Suppose the questions under discussion were as broad as this:

If as an answer to the first question you were to utter the metaphor above (“Imagine a feather lying on the sidewalk. . . ”), then I can agree that your contribution feels more like an open-ended invitation to use my imagination in a certain way. In this kind of context, there is no way to coordinate on “It’s not very likely that Olivia will show up to the party” based on your metaphor. Furthermore, it wouldn’t be clear at all how you expect me to use that metaphor to extract a specific answer to (c). In fact it would be unreasonable for you to want me to attempt such a thing.

It is not even clear how to use that metaphor to answer any of (c)’s subquestions. We say that p? is a subquestion of q? just in case any answer to q? entails an answer to p? So Did you invite Maria to the party? is a subquestion of Who did you invite to the party?. Sometimes when we don’t have a complete answer to the question under discussion, we try to at least answer some of the subquestions of the question under discussion -that is what a cooperative interlocutor would do. Suppose the question under discussion is Who is coming to the party? If you don’t have the full list of guests, you might still try to answer a subquestion (Are Maria and Lucy coming to the party?) by saying Maria and Lucy are coming to the party. Notice that if the question under discussion is (c) and someone offers the metaphor as an answer (Imagine a feather lying on the street. . . ), it is not even clear how to use it to answer any of the subquestions -furthermore, it is not obvious which of the subquestions the speaker is addressing with her metaphor. Thus, it is not a surprise that in this kind of case we cannot coordinate on a proposition that makes into the conversational record.

The point that I want to make is that things are very different when the question under discussion is (a) (“Is it very likely that Olivia will show up to the party?”). In this case it is pretty clear how to get an answer to this question based on the relevant metaphor. A Gricean has a story to tell about this. I’m not confident we can explain this case by relying heavily on our knowledge of conventions or open-ended invitations to use our imagination in a certain way. This case does look a lot like a case of conversational implicature -for instance, cancellability and non-detachability hold without a hitch.

Similarly, if the examiner were to answer to question (d) with (iii) (“You might want to look at your answer. . . ”), it would feel more like an open-ended invitation for the student to figure out what’s the issue with his answer to question three. However, as an answer to question under discussion (b), the hint is more than enough to coordinate on a specific proposition, as I have shown. It is not clear to me how to explain this case by relying heavily on our knowledge of linguistic conventions or open-ended invitations to use our imagination in a certain way. This too looks a lot like a case of conversational implicature.

^{Szabó (2016)} makes a similar point to the one I’m making here. He considers Grice’s letter of recommendation case and argues that it cannot be accounted for in Lepore and Stone’s framework. Recall, the case is one where a professor writes the following letter of recommendation: “Dear Sir, Mr. X’s command of English is excellent, and his attendance at tutorials has been regular. Yours, etc.” In this case Szabó grants that “there is a certain slack in what is being communicated here” (^{Szabó 2016, p. 9}). What he means is this. Perhaps the letter reader can understand the letter writer to mean that the student is a bad philosopher, or that he doesn’t meet the standards of the grad program he’s applying to, or that the writer believes that the student is not a good philosopher, so on and so forth. At the end Szabó thinks that the letter writer meant all those things. I can get on board with that, but I think it is harder to argue that letter writer and reader coordinated on a particular proposition that made it into the conversational record in this kind of case.

Furthermore, if the letter writer can be interpreted as communicating a range of propositions, it is not that crazy to take her as offering an open-ended invitation to be understood in any of those ways. This is more so once we consider that in the recommendation letter case it is not even clear what the question under discussion is. Perhaps it is whether the candidate is qualified for the graduate program he’s applying to, but it could also be whether he would be a good fit for the department, or whether his research is interdisciplinary enough, or what have you. Search committees are often interested in knowing much more than whether the candidate is qualified for the graduate program in question, and it is often very difficult to know exactly what they are looking for.⁶

What is nice about the cases I consider is that, relative to them, it is much easier to argue that speaker and interlocutor coordinate on a particular proposition that makes it into the conversational record. Furthermore, as we have seen, it can be shown how that proposition can be immediately deployed to resolve anaphora ambiguity, strongly suggesting that those cases shouldn’t be understood in terms of open-ended invitations of some sort.

In their reply to critics, ^{Lepore and Stone (2016)} offer a reply to Szabó’s objection. It would be natural for them to offer the same reply here. Their central point is that linguistic conventions mark certain utterances as replies to a given question under discussion. So, when you utter the metaphor above (“Imagine a feather lying on the street. . . ”) a linguistic convention marks your metaphor as an answer to the question under discussion. I’m happy to grant that there is a convention involved here -although, like ^{Szabó (2016)}, I’m not convinced it’s a linguistic convention, rather than a social one.

Perhaps the convention in play here is a social convention where we help solve problems had by people we cooperate with. Here’s the social convention in action. You are helping your neighbor find her dog, Owen. If you find Owen in your backyard, you may bring him to her -that would be a way to solve the problem. It’s easy to think of a question under discussion as a problem -the problem of knowing whether the world is this way or that way -and of an assertion as an attempt to solve that problem. Thus, it makes sense to think that an answer to a question is marked as a social convention. It is very important for Lepore and Stone that the convention is linguistic rather than social. Here is why. If the convention is linguistic, then the effects that we see in our examples are tightly connected to our knowledge of linguistic meaning and logical form, much more that a typical Gricean explanation. Lepore and Stone don’t get to claim that advantage if the convention is social.

To my mind it is more reasonable to think that the convention is a social one. Let’s say that we do have a convention to help solve problems had by people we cooperate with. If so, a linguistic convention linking assertions to answers to questions would be a redundant one -the work is being done by the social convention already. If there is no such social convention, then we have to explain our patterns of helping people we cooperate with in terms of cooperative reasoning, but then we can offer the same explanation about the link between questions and assertions. That kind of explanation, however, looks more Gricean than anything else.

But let’s grant that the convention is linguistic. So there is a linguistic convention linking your metaphor to the question under discussion. There is still the question of how we manage to coordinate on the specific answer to the question (“It’s not very likely that Olivia will show up to the party”) on the basis of your contribution. A linguistic convention might tell us that your metaphor is an answer to the question under discussion, but in this case it says nothing about what the answer is.

A Gricean could tell a story about how coordination on a particular proposition happens here, and they would do so based on principles of rationality and the like. I don’t think it would be fair for Lepore and Stone to simply claim that in these kind of cases something short of a Gricean reasoning is required. If all that is needed to show that is to point out that there is a linguistic convention linking questions under discussion with assertion, Imagination and Convention would have been a very short book. But it is not, and for good reasons.

I would like to close this section by quickly presenting examples where instances of sarcasm and irony have clear contributions to the conversational record -and, as such, they are cases where speaker and interlocutor coordinate on specific bits of information. As before, there are many examples where Lepore and Stone’s analysis in terms of open-ended invitations is spot on. However, the cases I’m about to present feel more like instances of conversational implicatures. As such, the remarks made about metaphors and hints should be easy to translate to the examples below.

The following example shows how with an instance of sarcasm we can communicate a specific proposition and adjust the standards of precision of gradable adjectives in the conversational record. Timmy asks: “What’s a circle?” Hector grabs a napkin and carelessly draws something that resembles an egg. Then he says: “This is a circle”. With discretion, Natalia passes a note to Hector with the following message: “Right, and France is hexagonal”. The following seems plausible: Natalia communicated that what Hector drew is not a circle and, furthermore, changed the conversational record by rising the standards of precision -now one doesn’t get to say things like “Italy is boot-shaped” and “France is hexagonal”.

Finally, let’s take a look at an example where we can communicate a specific proposition and adjust the borders of permissibility in the conversational record with an instance of irony. This semester Carla and Natalia are teaching a grad seminar together. Carla is in charge of the first half of the semester and Natalia is in charge of the rest. Carla had been very permissive in her classes: students could get in late, eat during class, and they could even talk a bit among each other. Today is Natalia’s first class, and towards the beginning a student asks: “Can I eat my sandwich?” To which Natalia replies: “Of course you can. And while you are at it, have a beer. Hell, do you want me to bring in a TV so that you can watch a show while you feast?” The following seems plausible: with her ironic contribution Natalia communicated that the student cannot eat her sandwich and, by doing so, she shifted the borders of permissibility -now students cannot eat during class.⁷ Everyone in the classroom gets that.

Humor is another case that one might want to explain in terms of conversational implicatures. Lepore and Stone argue that that kind of analysis is misguided. I’m not a funny guy, so I won’t attempt to construe a counterexample. However, the recipe for one should be clear by now: find a very precise question under discussion that an instance of humor answers clearly, changing the conversational record one way or another.

3. Conventions

We have seen that Lepore and Stone’s strategy is to explain cases of purported conversational implicatures in terms of linguistic conventions or open ended invitations of some sort. As we saw, their explanations in terms of linguistic conventions involve postulating a myriad of ambiguities. It is up to interlocutors to disambiguate in the right way. I would like to add something to a worry that has been voiced by ^{Bezuidenhout (2016)}, ^{Horn (2016)}, and ^{Szabó (2016)}. It may very well be that at least in some instances of disambiguation a Gricean-like explanation is required. If so, this may provide a lot of traction for a theory of conversational implicatures -we shouldn’t postulate all those ambiguities, if the Gricean machinery is available.

At a high level of abstraction conversational implicatures and disambiguations don’t look that different. In both cases a speaker utters a sentence and interlocutors have to search for the right interpretation among some candidates. If it is a case of ambiguity, candidates are possible disambiguations. If it is a case of conversational implicature, candidates are propositions that the speaker could have meant. In the case of conversational implicature, choosing the right interpretation involves Gricean cooperative reasoning -at least according to the rational reconstruction of the cases. It is not that crazy to think that at least some instances of disambiguation also involve some sort of Gricean cooperative reasoning. Perhaps we have to assume that the speaker is following the Cooperation Principle and derived Maxims in order to get at the right disambiguation.⁸ The way Lepore and Stone write about disambiguation in this passage doesn’t seem completely unfriendly to this idea:

we have come to see DISAMBIGUATION as a powerful mechanism that inevitably connects utterance interpretation to speakers’ goals and points of view in conversation. The idea is that the rules of language fix meanings for sentences in context only under specific grammatical analysis. When a speaker makes an utterance, there is often ambiguity. If the speaker is cooperative, however, it will be clear how to resolve the ambiguity, and the resulting interpretation will be one that fits the accepted purpose or direction of the conversation. (^{Lepore and Stone 2014, p. 148})

Now, it wouldn’t be fair to read too much into their use of “cooperative” in this quote, but their account of disambiguation here does look fairly Gricean at first sight. However, for them it’s crucial that in disambiguating there is always appeal to a linguistic convention: that one of the meanings of a given term is such-and-such, that one of the logical forms of an ambiguous sentence is such-and-such, or what have you. For them this makes it clear that we are not dealing with a proper Gricean derivation, since the need to appeal to a linguistic convention shows that the derivation is not a calculation from what is said and principles of rationality alone.

However, I’m not quite sure about the force of this point. In a typical Gricean derivation one is allowed to use information of all sorts and not only information regarding what is said and principles of rationality. Let’s go back to the recommendation letter case. In that case the rational reconstruction assumes that having an excellent command of English and attending tutorials regularly are qualities completely orthogonal to whether or not the candidate is a good philosopher. Without this assumption the Maxim of Relation doesn’t kick in. If the qualities that make for a good philosopher can be invoked as part of a Gricean derivation, I don’t see why we cannot appeal, as a matter of principle, to a linguistic convention as well, so long as most of the weight is being carried by the Cooperative Principle or the like. To my mind, more work has to be done to show that nothing like a Gricean derivation is involved in at least some cases of disambiguation.

There is one more worry I want to consider regarding ambiguities in Lepore and Stone’s framework.⁹ According to them “Can I have the French Toast?” is ambiguous between a question and a request. Now, imagine a linguistic community that is just like ours except that for its members “can” doesn’t have the request reading -the relevant ambiguity is not there since that community lacks a convention assigning that meaning to that word. Could there be a single instance where a speaker in that linguistic community communicates an indirect request by way of uttering “Can I have the French Toast?”? It seems that there could be such an instance, specially if the waiter is in front of the client ready to take the order -if Lepore and Stone want to claim this is not possible, the burden of proof is on their side. So it seems that the ambiguity in “can” is not necessary in order to explain indirect requests of this form. Whatever plausible explanation we may want to give for that case could be used to argue that “can” is not really ambiguous in our linguistic community -of course, a Gricean explanation could be given. A possibility left open by this argument is that the indirect request reading was once a conversational implicature that became conventionalized. However, this possibility doesn’t favor Lepore and Stone’s project, since it involves granting that conversational implicatures have some work to do.

Notice that a clearly ambiguous word like “bank” passes this test without a hitch. Imagine a linguistic community that is just like ours except that “bank” doesn’t have the river bank reading. Could speakers in that community communicate that they are going to the river bank by way of uttering “I’m going to the bank”. It seems that they couldn’t. This strongly suggests that “bank” is ambiguous in our linguistic community.¹⁰