Would you believe that von Fintel & Gillies might be wrong? =========================================================== [Comments on a paper by Ben Lennertz, APA Pacific, April 20, 2011] Kai von Fintel and Anthony S. Gillies What von Fintel & Gillies said ------------------------------ In our papers "CIA leaks" and "*Might* made right" (henceforth, MMR) we discussed dialogues like this: [Alex is aiding Billy in the search for her keys:] (1) Alex: You might have left them in the car. (2) Billy: You're right. Let me check. (3) Billy: No, I still had them when we came into the house. We argued that Alex has no right to straightforwardly assert any epistemic claim other than the one that is grounded in her own evidence. But Billy responds as if Alex had put into play a stronger claim. In (2), Billy isn't merely saying that Alex is right that it is compatible with Alex's evidence that the keys were left in the car. Billy is saying that it is compatible with their joint evidence that the keys are in the car. In (3), Billy is clearly not saying that it is false that it is compatible with Alex's (limited) evidence that the keys are in the car. The theorists that we call CIA agents argue from data like these that we need a radically new conception of the meaning of epistemically modalized sentences: they are semantically sensitive not just to the two standard Kaplanian parameters context and index, but also to an "assessor" or "judge" parameter. Thus, a new kind of proposition is introduced and Billy plugs himself into the assessor parameter and judges Alex's proposition as true or false, as the case may be, relative to his own epistemic state. We argued, instead, that the response types in (2) and (3) and others along the same lines reveal a significant amount of indirection: Billy's response does not target the proposition strictly expressed by Alex but has a target that is indirectly salient after Alex's utterance. We discussed three kinds of indirection. 1\. An almost normative reproach "[A] *might*-claim might be open to reproach, retraction, or disavowal, as soon as it becomes clear that the speaker was relying on assumptions that were not reliable --- or, more generically, when it becomes clear that her epistemic state was not as it should have been --- even though it is perfectly true that the prejacent was compatible with what the speaker knew. Blofeld could be said to be wrong not because his *might*-claim is false but because he is relying on misleading evidence, never a good move in his business." (CIA leaks, pp. 83-84) [See also Stephenson 2007, pp. 516-517] 2\. Targeting the prejacent It is often possible to respond to the prejacent proposition (the proposition embedded under the modal) even though that proposition is obviously not asserted. This is a fairly general mechanism that arises not just with modals: (4) A: I think it's raining out. B: No, it isn't/No, it can't be. 3\. Other salient propositions Billy is still looking for her keys: (5) Alex: People often put their keys next to their telephone. Billy: [checks next to his phone & finds the keys] You were right! Thanks. Billy is not necessarily endorsing Alex's generic claim about what people often do. Instead, it seems he is applauding Alex's implicit suggestion that Billy should check next to his phone. At the end of "CIA leaks" and in MMR, we explored in detail one idea of how additional propositions can be "put into play" by an epistemic "might" claim. We suggested that Alex's utterance in (1) did not just assert the egocentric claim that she has grounds for asserting, it also "put in play" stronger propositions (without actually asserting them) in a way that makes them available for Billy to react to. The picture is that Alex's utterance is ambiguous by design and puts into play an array of variably strong epistemic modal claims, the weak egocentric one and stronger group readings (in particular, the group of Alex and Billy). Given these mechanisms, we argued that there is no call at all to move to a relativist semantics of epistemic modals. Lennertz' argument ------------------ Lennertz argues that occurrences of epistemic "might" embedded in belief reports constitute a problem for us. Here is his central example: (BEL DIS) Fred: George believes that the Dodgers might have won last night. Sandy: That's true. But he's wrong. Lennertz declares that the following are "our intuitions" about this case: (i) what Sandy says is true (Sandy knows both that George has the relevant belief and that the Dodgers lost), (ii) Sandy is talking about George's mental states, and (iii) Sandy is talking about the same state in the first and second sentences of her response. We certainly agree with (i) but we need to look at (ii) and (iii) much more closely. Given the amount of indirection that responses like "he's wrong" are exploiting, it is not obvious at all that what Sandy's second response targets is the same mental state as the first sentence or that it targets a mental state in the first place. Embedded epistemic modals - Anchoring to attitude holder -------------------------------------------------------- Before turning to (BEL DIS) and Sandy's response in particular, we should bring in some background on embedded epistemic modals. As we have mentioned already, epistemic modals on their own are quite underdetermined in their meaning. While the egocentric reading might be a default that is always available, other readings are clearly available. Examples include the cancer test case of DeRose: (4) I don't know whether John might have cancer. The doctors know but they won't tell us until Monday. In (4), the reading is arguably one that targets the knowledge of a group encompassing the speaker and other more well-informed individuals. DeRose argued that epistemic modals always include the speaker's evidence but may include further individuals and their evidence. There are examples that look like a truly "exocentric" reading is available, such as the surprise party example of Egan, Weatherson, and Hawthorne: [Context: Ann is planning a surprise party for Bill. Unfortunately, Chris has discovered the surprise and told Bill all about it. Now Bill and Chris are having fun watching Ann try to set up the party without being discovered. Currently Ann is walking past Chris's apartment carrying a large supply of party hats. She sees a bus on which Bill frequently rides home, so she jumps into some nearby bushes to avoid being spotted. Bill, watching from Chris's window, is quite amused, but Chris is puzzled and asks Bill why Ann is hiding in the bushes. Bill says:] (5) I might be on that bus. As Stephenson points out, once embedded, such exocentric readings become unavailable: (6) #I think I might be on that bus. Stephenson devises an analysis that predicts that embedded epistemic modals are always anchored to the attitude holder. -- FN: This insight actually goes back to Antinucci & Parisi 1971, who state as a condition on epistemic "must", which they essentially analyze as meaning that X CAUSES the fact that it is BOUND to be the case that Y BELIEVES Z, that the first argument of BELIEVES must be the speaker. In their footnote 1, they add: "Actually, the precise condition on the first variable of BELIEVE X Y is that it must be coreferential with the subject of the immediately dominating performative declarative verb (whether this verb is used performatively or not). Thus in a sentence like John says that Harry must have gone this variable is coreferential with the subject of the verb 'to say'." -- Actually, DeRose's example, which is after all a case of an embedded "might", shows that, rather than *identifying* the anchor with the attitude holder, the mechanism has to ensure that the attitude holder is *included* in the group whose evidence is at stake. Embedded epistemic modals -- Collapse? -------------------------------------- Recently, Anand & Hacquard and Kratzer have argued that embedded epistemic modals usually do not give rise to a two layered modal meaning. (7) George believes that the Dodgers might have won. A straightforward iterated modality analysis would give a semantics such as the following to (7): in all of the worlds w' compatible with George's belief system in w, the evidence available to George in w' does not rule out a Dodgers win. The suggestion is that, rather, (7) has a single layered analysis: the Dodgers won in some of the worlds w' compatible with George's belief system in w. Anand & Hacquard and Kratzer propose different mechanisms to achieve that goal, but the differences need not occupy us here. In Stephenson's set-up, pretty much the same result is obtained: (7) means that in all of George's belief worlds, the epistemic alternatives of the judge (identified with George by Stephenson's mechanism) include worlds where the Dodgers won. Stephenson argues that "to believe something is to believe that one knows it, which means that the epistemic alternatives of a person’s doxastic alternatives are the same as the person’s doxastic alternatives." Thus, she ends up with the same meaning as Anand & Hacquard and Kratzer. There has not yet been a thorough investigation of this hypothesis and this is not the place for us to conduct one, except to note that since it reduces the level of embedding of the non-modal prejacent, it may lend credence to the idea that the prejacent is available for approval/rejection. What does Fred claim? --------------------- We think that it is reasonable to assume that Fred's sentence "George believes that the Dodgers might have won last night" simply claims that it is compatible with George's information state that the Dodgers won last night. We do not think that our MMR approach should be carried over in the sense of making Fred's sentence multiply ambiguous and having Sandy respond to one of the secondary meanings of Fred's sentence. So, to some extent we agree with Lennertz that our MMR approach cannot explain his embedded epistemic modal. But there are two caveats to this concession (and they are big ones). 1. Our analysis is clearly concerned with the pragmatics of assertively uttering a sentence with a bare epistemic modal. We argue that such sentences are ambiguous in a particularly interesting way: a speaker can utter the sentence without resolving the ambiguity, thereby putting into play several propositions at the same time. It would appear obvious that *embedded* modals are beyond the scope of such an analysis, since the modalized sentence is not assertively uttered and thus a pragmatics of assertive utterance of modals is not applicable. If an analysis is by its very nature not applicable to a particular phenomenon, it can hardly be blamed for not making (correct or incorrect) predictions about it. 2. The overall point of our discussion in "CIA leaks" and MMR was that the array of responses that so impressed CIA agents are due to massive amounts of indirection. And we maintain that this is so also in Lennertz' case. We agree that group readings of George's "might" are not "put into play" by Fred, but that leaves open several other kinds of indirect targets. In what follows we explore several ways of understanding the second part of Sandy's response: "he's wrong", which appears appropriate even though it seems clear that George had all the right in the world to be in a belief state that left open the possibility of a Dodgers win. We suggest that two of them may be at work in Lennertz' case while another (which may be the one that Lennertz has in mind) is not. Strong readings of George's might? ---------------------------------- We have argued that when George says that the Dodgers might have won, he is thereby putting in play not just the egocentric proposition that a Dodgers win is compatible with his evidence, but also stronger propositions about what is compatible with the evidence available to groups containing him. There's one way in which this could carry over to Lennertz' case. If Fred's utterance was based on and/or reports an original utterance by George of something like "The Dodgers might have won", then by our story in MMR, George may thereby have floated group readings of that sentence. Given the general fact of indirection in these kinds of conversation, it is not inconceivable that Sandy is in fact targeting one of the group readings that were floated by George's original utterance (*not* by Fred's utterance, to be sure). Note that if this is what's happening in (BEL DIS), then contra Lennertz' intuition (ii), the second part of Sandy's utterance does not target a mental state of George's and thus, contra (iii), a fortiori not the same mental state as the first part of her utterance. Reproach? --------- Lennertz seems to hint that there is an element of reproach in Sandy's response when he writes "Sandy response is directed at George's hedge. He does have it, but he *is wrong to*" (our emphasis, see also his footnote 25). Saying that George is wrong to have that hedge basically amounts to saying that George is blameworthy. And certainly, "wrong" can easily carry that kind of flavor. In fact, we should note that we have encountered speakers who reject Sandy's response because of that flavor. In the case at hand, George cannot be blamed for thinking that the Dodgers might have won. He is not basing his assessment on misleading evidence, for example. So, we think that in as much as Sandy's response is appropriate, it cannot have the reproach reading. Prejacent Denial ---------------- We had another mechanism up our sleeves to explain some of the responses to BEMs. And that mechanism might be applicable to (BEL DIS): Sandy's response might be targeting the prejacent: the proposition that the Dodgers won. One might have thought that the proposition that the Dodgers won appears too far embedded in (BEL DIS), being underneath both "George believes that" and "might". But if the collapse theory is correct, we are dealing with only one level of embedding. In any case, we'd like to re-employ yet again a favorite example of ours: the Mastermind example. Pascal and Mordecai are playing Mastermind. After some rounds where Mordecai gives Pascal hints about the solution, Pascal says (8) There might be two reds. Mordecai, knowing the solution, turns to us and reports as follows: (9) Pascal believes that there might be two reds. He's right: there might be. He's right: there are. He's wrong: there can't be. He's wrong: there aren't. We take this to be suggestive evidence that the prejacent is available as a target for "he's right/wrong". If Sandy's response "he's wrong" targets the prejacent of Fred's claim, then again, neither of Lennertz' intuitions (ii) and (iii) are borne out: Sandy is not targeting one of George's mental states. Conclusion ---------- It's true that Lennertz believes that von Fintel & Gillies might have a problem with embedded epistemic modals, but he's wrong. On the other hand, he's right about many other things, including that embedded epistemic modals are a fascinating topic.