Ethics Discussions at PEA Soup: Derek Baker & Jack Woods’ “How Expressivists Can and Should Explain Inconsistency,” with a critical précis by Mark Schroeder

39 Replies to “Ethics Discussions at PEA Soup: Derek Baker & Jack Woods’ “How Expressivists Can and Should Explain Inconsistency,” with a critical précis by Mark Schroeder”

1. We would like to thank the editors of PEA Soup for selecting our paper for discussion and especially Mark Schroeder for writing a critical précis. We greatly appreciate the chance to explain in more detail our interpretative decisions, and to clarify the relation between logical and semantic inconsistency in our positive proposal.
   We’ll start with the concern that considerable space is dedicated to criticizing views other than Schroeder’s. In the paper, we consider multiple possible interpretations of Schroeder`s argument, and even possible changes one might make to the argument to avoid our objections. We do not believe that Schroeder is committed to all of these possible interpretations and so in this sense we are criticizing claims he never makes. But this is standard philosophical practice. We tried our best to make sure that we did not criticize Schroeder himself for faults with interpretations, but undoubtedly there are points at which we weren’t careful enough.
   One reason we thought it necessary to entertain so many interpretations (besides simple thoroughness) is that many of Schroeder’s core ideas are effectively spread over two distinct books, and we remain uncertain on the precise relation between the two. The notion of disagreement receives minimal discussion in Being For (the phrase itself doesn’t appear at all, though the concept is clearly discussed on page 40)—and none of this discussion relates it to van Roojen’s problem. The later book does discuss intrapersonal disagreement at length and explicitly in the context of van Roojen’s problem—but makes no mention of B-type inconsistency. Our sections 2 and 2.1 can effectively be seen as addressing a self-contained version of Being For, whereas section 3 is addressed to Being For read in the light of the later book. Again, our apologies if Schroeder thought this unnecessary, but we honestly thought it more illuminating and less presumptuous on our part to consider both versions. We’d add that by addressing a range of interpretations, we achieve wider scope for our argument. We don’t intend to simply object to the very specific position Schroeder currently holds, but to illustrate problems with lines of argument that someone attracted to Schroeder’s basic framework yet willing to change parts of it might (and anecdotally do in conversation) push.
   In any case, we’re happy that Schroeder clarified which interpretation was the correct one. We’re also happy that this turned out to be the interpretation we spent the most space—effectively the entire second half of the paper—addressing.
   We do have worries that we misunderstood Schroeder on one important point. He claims that our term discordance is a broader notion than what he meant by inconsistency of attitude. But we explicitly state that we intend to use “discordance” to mean inconsistency of attitude, and explain that the terminological difference is only motivated by a desire to avoid the pervasive confusion of attitudinal inconsistency with semantic and logical inconsistency without cumbersome adjectives. Perhaps this is just an oversight on his part, but based on some things Schroeder says, it may be that substantially more presuppositions were built into the terms “A-type” and “B-type inconsistency” than we realized, and thus our use of “discordance” could not be charitably interpreted as picking out the same relation. We would like to hear more from Schroeder about this, but it may be that we pursued the wrong argument in trying to demonstrate the existence of B-type forms of inconsistency, and that really our position is one of extreme skepticism about the existence of any form of inconsistency of attitude, in Schroeder’s sense. But we’d prefer to hear him explain his position in more detail before we take a stand.
   Note though that the targets of Schroeder’s argument, Gibbard to the side, have worked with a notion of discordance (or inconsistency, or conflict, or whatever) that is not obviously to be identified with intrapersonal disagreement. We feel that intrapersonal disagreement is best thought of as explicating inconsistency of attitude, much like we suggest you might do with contravention of functional role or interpretational incoherence. Any of these three might also do to explicate Schroeder’s notion of inconsistency-transmission and thereby explain why some, but not all, combinations of mental states are discordant. Some explication we take to be an explanatory must in giving an account of inconsistency of attitude. If we are all fellow travelers here, so much the better for everyone. It`s the right view.
   However the objection should be stated in the technical vocabulary, half the paper is directed against what Schroeder tells us is (more or less) the correct interpretation of his position, and we believe it would be most philosophically fruitful to focus on the issues raised there, which also comprise our positive contributions to the debate. We think we’ve said enough to defend our interpretative decisions, and we’d like to move to what we see as the substance of the disagreement.
   We’ll start by admitting our faults. We regret that our position did not come across as clearly as possible in our paper. What we do is account for what we call a broad notion of inconsistency of which pragmatic, semantic, and logical inconsistency are all instances. Sentences are broadly inconsistent when they express discordant states of mind. We also hold that broadly inconsistent sentences express discordant states of mind (partially) in virtue of the meanings of the composite expressions and the grammar of the sentences involved.
   We took the distinction between logical and non-logical (semantic and pragmatic) inconsistency to be easy to fix. We merely invoke the standard Tarskian solution of attributing logical inconsistency to grammatical structure and the meaning of a few choice words. Nothing fancy here, just good old (logical) common sense. And this picture is completely continuous in the sense that Schroeder has in mind. Pick a larger set of logical words, expand the class of logical inconsistencies; pick a smaller set, constrain it. Bob`s your uncle.
   For example, taking Schroeder`s example of doxastic logic, if we treat K (the knowledge ’’operator’’) as a logical operator, then
   KA & ~A
   comes out as logically inconsistent on our account (given the veridicality property KP=>P for any P). Why? Well, in virtue of the grammar and the meaning of K, &, and ~, the above sentence is guaranteed to be broadly inconsistent—the meaning of A is irrelevant. Nota Bene—this explanation has nothing to do with truth.
   The problem of pragmatic inconsistency we took to be the problem of accounting for cases in which the meaning of the expressions and the grammar of the asserted sentences suffice to generate discordant attitudes, but where what was asserted could nonetheless be true. So we had to discuss truth. We pull pragmatic inconsistency away from semantic by imposing on semantic inconsistency the constraint that—loosely—semantic inconsistency is context-invariant in the sense that it cannot be reasserted no matter what the context. There are many details to be worked out here; we focus on the case of Moore-paradoxicality as it is an exemplar and the source of van Roojen`s problem. This may have led to an account of pragmatic inconsistency which is insufficiently general. In fact, an objection from Mike Ridge in conversation worries us in this regard. Happily, this is an excuse for more work (that one of us is currently engaged in).
   We do wonder if some confusion about logical-semantic continuity may have stepped in here, due to a failure on our part to more clearly identify what is explanandum, what explanans. One of the key differences between semantic inconsistency and pragmatic inconsistency is that at least one of a set of semantically inconsistent sentences must be false, whereas this is not so in the case of pragmatic inconsistency. The target phenomenon, remember, is the Moore-paradoxical utterance ‘It is raining but I don’t believe it is raining’. This sentence sounds like a contradiction, but the speaker hasn’t literally contradicted himself, because both conjuncts could be true. But how can we account for this, in a non-ad hoc way, armed only with an undifferentiated notion of broad inconsistency? We appeal to the role of truth as a device of reassertion. The Moore-paradoxical sentence can be reasserted by a third-party without discordance, thus a third party can assert that the Moore-paradoxical sentence is true.
   Notice this means, though, that truth isn’t doing any work explaining why the Moore-paradoxical sentence lacks semantic inconsistency. Rather, the relation between truth and the two varieties of inconsistency is the assumed data point, for which an explanation must be given. The explanation of why some inconsistencies are literal and others merely pragmatic is done in terms of reassertion. Truth is a derivative property, for our account of semantic relations as much as for logical relations.
   An expository oversight on our part, which Schroeder calls our attention to, is that we do not explicitly state that the same non-reassertability is a necessary condition on logical inconsistency as well. We thought this redundant, as logically inconsistent sentences in the ordinary sense of logic would be inconsistent when jointly reasserted. Schroeder points out, however, that if our logical vocabulary is expanded to include terms for various attitudes, this may no longer be true. Now we’re not entirely sure about this. It seems to us that his example only works if you also add ‘I’ to your logical vocabulary, and also add in that ‘I’ changes the inferential role played by the attitudes taking it as an argument, and so we have some inclination to say, simply “for God’s sake, man, don’t do it! There be monsters.”
   But addressing the issue more directly, we don’t think we would be forced to treat ‘Either I believe that P or ~P’ as a logical truth, whatever you add to vocabulary. A third party could coherently say ‘what they said was false’. Notice that the reassertion condition on truth serves as a device for finessing the indexical component—sly, eh?—so as to avoid complications involving updating the ‘I’ to ‘s/he’. Even if we treat `believe` as a logical operator, we can deny that the sentence is a logical truth for this reason.
   There is still the interesting question of whether a logical system including both the logic of indexicals and an epistemic logic might cause problems for our framework. We don’t really know, though we suspect the problems here are problems of detail, not substance. (We also have substantial misgivings about the claim such systems have to being genuine logics—though this is tangential to Schroeder’s basic concern with continuity, which we think correct.) Obviously the challenge would be to show that we make the right predictions here: ‘I am here now’ is a logical truth in the logic of indexicals; Schroeder’s offending sentence is not. This is worth thinking about, but it will probably not be on our radar for a while. So far, all we’ve done is suggest a framework. We would like to use this framework to give a more concrete account of basic logical vocabulary—that is, we would like to solve the negation problem and original version of Frege-Geach. Only after this will we address extensions. We wouldn’t be surprised if we run into problems (Schroeder very admirably concludes Being For by noting problems his semantics faces with extensions), but if the place where difficulty finally arises is an attitude-operator-including logic of indexicals, then our account has made substantial progress.
   Have the authors “appeal[ed] to notions of inconsistency that are analogous to their account of semantic inconsistency, rather than their account of ‘logical’ inconsistency” in setting out the above clarifications? The “rather than” presupposes a lot. We’re explicitly making elements of our account of semantic inconsistency elements of our account of logical inconsistency, in part because we were already implicitly committed to it (unless we want a disjunctive account of truth!), and because, in any case, it helps us to achieve the sort of continuity Schroeder regards as a desideratum—a desideratum we’re glad he pointed out, because we find it attractive as well. It is, after all, the hallmark of the Tarskian approach to logical consequence we appeal to in our paper.
   An earlier draft of the paper briefly discussed the vagueness of where logicality ends and other forms of a priori inconsistency begin, and in the context of expanding logical vocabulary. It was cut for space constraints and because the discussion was too loose. Schroeder’s précis—with its focus on continuity—has helped us clarify the issues for ourselves and give them a sharper statement. We’re very appreciative of that. We are also happy that Schroeder has clarified what he takes the disagreement to be, and that it is fairly close to our own diagnosis. We took him to be committed to a psychological analog of logical inconsistency; he clarifies that it is an analog of semantic inconsistency. And given his way of capturing continuity, the difference in those commitments is likely very narrow.
   Our account may, of course, lack continuity between pragmatic and semantic inconsistency, due to invoking context-insensitivity; but this is not objectionable as there was never any presupposition of clean continuity here to begin with.

2. My signature can be added to the above.

3. One correction, 3rd paragraph, the italicized “disagreement” should read “intrapersonal disagreement.” Sorry about the confusion.

4. A (possibly clarificatory) question to Derek and Jack: I am not sure your proposal succeeds in demarcating illogicality even if the general strategy is legitimate. To be sure, no logically consistent sentence will come out as logically inconsistent according to your proposal, because syntactic inconsistency is, for you, a necessary condition for logical inconsistency. (A discordance is a case of logical inconsistency if it corresponds to a logically inconsistent sentence.) But I think your proposal is too stringent, misclassifying some logical inconsistencies, because of the two ways you offer for construing discordance of attitudes. The first invokes “strong interpretive incoherence”. But the (interpretive) principle of charity enjoins us to avoid ascribing obvious inconsistencies, and allows for the ascription of subtle logical errors. The rationale, made explicit by Richard Grandy’s (kindred) “principle of humanity”, is that we cannot explain how someone could believe both p and ~p. But it is all too easy to explain how someone could harbor very subtle inconsistencies, for instance, believe – on the basis of a subtly erroneous proof – a false mathematical proposition which is inconsistent with axioms he accepts. It is, similarly, easy to see how someone can fail to notice that two beliefs are subtly logically incompatible. The explanation in both cases is that we are logically imperfect: deduction is hard. But then, interpretive incoherence draws the line in the wrong place, failing to classify as discordant complex logical inconsistencies. Correlatively, logical inconsistency isn’t a special (formal) kind of discordance.
   Your second explanation invokes “functional role”. But again (although here I am even less sure), it seems to me that it draws the line in the wrong place vis-à-vis illogicality. If the role of belief is to represent the world correctly, then the failure of a massively erroneous consistent belief system will be much greater than that of an inconsistent one which is “close to the truth”, i.e., has a very favourable ratio of truth to falsity. The fact that the latter is (logically) guaranteed to be imperfect is simply irrelevant to its (admittedly imperfect) success in representing our world. That a belief is bound to fail in every world doesn’t reflect badly on its achievement in ours, doesn’t add to its actual failings. So on this way of drawing the line (in which discord is gross representative failure), illogicality doesn’t come out as always discordant.

5. Derek & Jack,
   I think I have a worry that might be the same as Ruth’s first worry or related to it.
   One way to worry about the adequacy of an expressivist explanation of logical inconsistency is to worry that it doesn’t generate a well defined notion of logical inconsistency because there are other kinds of tension or “discordance” (such as Moore paradoxicality) of which logically consistent sentences and attitudes are subject. If the expressivist explains logical inconsistency just in terms of such tension, that complaint goes, she will be liable to find such inconsistency where there is none. Derek and Jack propose to handle this by defining a general notion of discordance to explain the pressure to avoid inconsistent attitudes, but then to hive off logically inconsistency as a special case by noting a further feature shared by logically inconsistent judgements. That feature is, if I understand the view, some shared syntactic feature of the sentences used to express the attitudes. This would allow the logically inconsistent to be distinguished from the merely discordant but not logically inconsistent by these semantic features.
   It seems to me that the logically or even semantically contradictory is not a subset of the discordant as characterized in the paper.
   As a first pass, the paper gives us two handles on discordance. Discordant attitudes are such that we have trouble interpreting one agent as holding both. (402) And discordant attitudes held by the same person cannot play the functional role the attitudes should play in a person’s mental life. (402-3).
   But there are logically inconsistent sets of beliefs that don’t meet these criteria. Take any large set of inconsistent beliefs which is such that no smaller subset generates the inconsistency. Or to make things vivid take a set of claims of the sort that generate a premise paradox – X1, X2, X3, . . . Xn, and “At least one of X1-Xn is false.” The set is inconsistent, but I’m not at all sure it is discordant. I have no trouble understanding someone who accepts such a set. In fact I think it is rationally required to accept such a set given enough reason to believe each one of X1-Xn (if the set is large enough, as when someone writes a book for which the standard preface disclaimer makes sense). This is (as I think of it) the flip side of the Harman point that the paper is at pains to credit. Logic is one thing, good reasoning another. Not only does good reasoning require more than logic requires, it also sometimes allows accepting inconsistent conclusions.
   Nor does it seem that these beliefs are prevented from playing their relevant functional roles by their inconsistency. They may not play them completely, in that we should probably not use them all together as premises in certain arguments. But we can get along well enough reasoning with them so long as we pay some attention to the fact that they are jointly inconsistent.
   So it seems like these sentences (and the beliefs they express) are inconsistent, but not discordant.
   You might think I shouldn’t be so confident they’re not discordant. For, don’t I feel some tension in accepting the sentences that make up the preface paradoxical set? I agree, I do, but that’s just to say I think they’re inconsistent and that is one way of being in tension, since inconsistent sentences cannot all be true. I take it that the project here is to explain the nature of inconsistency via some independent notion of discordance. And in this case, to the extent that I agree they are discordant it is only in the catch-all sense that groups any kind of inconsistency under this heading. But if it is to do explanatory work it seems like we need a core notion.
   Anyway, the basic idea here is just to worry about extensional adequacy for the theories explanation of logical inconsistency from the opposite direction from the one addressed in the paper. Instead of worrying about whether every instance of discordance is also one of logical inconsistency (which the present proposal is designed to address) I worry about whether every instance of logical inconsistency is one of discordance.
   Thanks!

6. Dear Ruth Weintraub, thanks for the questions.
   Obviously some of the problem here is that a more detailed account of discordance is called for. As we acknowledged in the paper we weren’t doing more than pointing to directions that could plausibly provide explication. That said, here’s how we think the two accounts should be able to answer your worry.
   Discordance on our account is basically a form of irrationality, and all actual agents are going to be irrational to some extent. It will never be the case that all of an agent’s attitudes are playing all of their functional roles perfectly. Sentences are broadly inconsistent for us because they evince some specific form of discordance. I could have guessed there was some respect in which your psychology was incoherent—you are human after all—but I never would have guessed you had inconsistent beliefs about the rain until you said what you said.
   The worry about the functional role story seems to be this: it may be that my psychology could become less discordant overall by accepting more logical inconsistency. Sure. But what you are doing here, we’d argue, is trading off discordance in one place for discordance somewhere else. The discordance somewhere else may ultimately undermine rationality less significantly, but it is still a departure from rationality. So we don’t think there’s a problem with saying that logically inconsistent sentences express attitudes that are guaranteed to contravene each others’ functional roles—even if acquiring those attitudes improved the functional performance of other attitudes in the psychology.
   On the Davidsonian account, the worry is more that small mistakes don’t count against interpretability. Agreed. The one of us who likes the Davidsonian picture more (Derek should confess to it) is of the opinion that Davidson tends to draw overly strong conclusions from his ideas, and its not clear that these overly strong conclusions really follow. As you point out, Davidson says that we could never interpret someone as believing p and believing ~p. This sounds false, and we don’t even know that it follows from the Principle of Charity, unless that Principle is taken in its strongest possible form. It seems there is a very plausible weaker version, which just says there is a minimal level of rationality we must attribute to a person before they cease to count as a thinking person. So the thought is this: any attitude has lots of connections to other attitudes and behavior—some minimal number of which must be there before the attitude can be attributed to an agent. But these connections are fungible. The state isn’t responsive to the right kind of evidence—so is it still a belief? Well, does it make up for it by influencing action and other beliefs in the right way? The idea is that any failure to manifest one of these constitutive, fungible connections is a form of discordance, because these are the sorts of failures that can ultimately add up to uninterpretability. But small failings won’t necessarily be a problem at all (all actual agents are guilty of them all the time). So yes, failing to reach the right answer to a very difficult math problem won’t prevent us from interpreting an agent, but it does involve a failure to manifest some of those connections on which interpretability is based, if that makes sense.
   Note that the example of at least one of my beliefs is false (one of Jack`s favorite examples, by the way) is a case in which the inconsistency is not logical by our account since the grammar and syntax and logical expressions do not guarantee discordance—only the additional meaning does so. Given our reassertion condition, we can also deny that it is a semantic inconsistency. But, so noted and given a distinction between local and global discordance, we may be committed to calling it a subtle pragmatic inconsistency (though there are some complications here we can talk about). If this means that we, qua human beings, are doomed to some irrationality, then we need our account to involve a bit of idealization. But we needed that anyways.
   Derek and Jack

7. Hello Derek, Jack, Mark (and everyone else).
   There is something puzzling to me about the dispute here. All parties seem to accept that a set of sentences is inconsistent iff all of its members cannot be jointly true. (That’s B&W’s third data point on p. 394 and it’s reiterated in the first paragraph of S’s precis.) Then the debate is about how expressivists should make sense of this and, more specifically, how they might do so in a way that allows us to mark a distinction between sentences that are logically inconsistent, semantically inconsistent, and inconsistent in some other way.
   What is puzzling me is that I would have thought that we need a notion of semantic inconsistency, indeed even a stronger notion of logical inconsistency that doesn’t turn on truth. If you have two PhD supervisors and one says
   (1) Submit this paper to Ethics now
   and at the same time the other says
   (2) No, revise it more first
   then haven’t they disagreed with each other by saying something semantically inconsistent? (‘Submit now’ and ‘revise more first’ seem incompatible in virtue of their meanings.) Moreover, imagine further that the the second supervisor had instead said
   (3) No, do not send this paper to Ethics now
   then hasn’t she said something logically inconsistent with (1)? Sentences (1) and (3) are inconsistent because of the logical form of these sentences and the meaning of the logical particle ‘not’.
   Indeed we can mix and match declaratives and imperatives with logical relations. Someone who says:
   (4) If it’s raining, bring an umbrella
   (5) It’s raining
   (6) Don’t bring an umbrella
   has said something that I’d want to class as logically inconsistent, but it doesn’t look like this is something we could explain as A-type inconsistency of mental states.
   If this is right (and we don’t want to say that imperatives are truth-apt), then clearly we’re going to need something other than the realist explanation of semantic and logical inconsistency (e.g. in terms of when sentences can be conjointly true) in our overall theory of how language works. (Btw, this is a point made well by Charlow in a recent PhilStudies paper.)
   Maybe that’s just a way of expressing surprise that anyone would have thought the realist explanation of logical and semantic inconsistency is going to be the only possible model. But I’m also skeptical that the best non-realist explanation of this is going to come from purely psychological properties of the mental states canonically expressed by the relevant sentences rather than something having to do with the kinds of moves they can be used to perform in different conversations. (It’s not even clear that imperatives canonically express one kind of mental state, since they can be used to command, advise, suggest, entreat, forewarn, invite, offer, etc.)

8. Hi Mark,
   Thanks for joining in. We’d start off by saying that as a methodological principle in offering any semantic theory we should start with simpler cases, trust our intuitions there, come up with theories for those simple cases, and then assume that the more complex cases display the same principles, even if it’s not that intuitive in those cases. (Derek once delivered a paper doing very standard action theory with means-end reasoning about potentially conflicting obligations—and received the question, how would he make sense of the goals of Robespierre in the framework he was using. Maybe not the fairest test for the framework.) So part of our answer will simply be to dig in and say that in simple cases of logical inconsistency of judgment there is discordance—and that seems intuitive and obvious—so we should assume the same principles are in play in the more complicated cases.
   Gil Harman’s point that logic isn’t the same as reasoning isn’t, we think, completely dispositive. As we noted above, there may be cases in which trading off irrationality in one place for another improves overall rationality. That doesn’t mean there wasn’t a tradeoff. So we can acknowledge it would be insane if people were to actually try to bring about complete logical consistency of their beliefs, and that they would probably end up less reasonable people overall from such an attempt, while still maintaining that logical inconsistency of belief is still a rational failing.
   Is this just table thumping? Yes (though we’ll try some non-table-thumping in a moment). But there’s probably not a lot more we can do here, given that this is a foundational assumption of the theory, which, as we said in the paper, we don’t have a satisfying account of, only some directions you could go in to build an account. (We don’t feel terrible about this either. Realist semantics will have to table thump at a similar point.)
   Two points to try to make it intuitive that even with really large belief sets a subtle logical inconsistency is still a form of discordance. (1) You acknowledge that they are in tension. But you say that this is explained by the fact that they are logically inconsistent, not the other way around. But simple logical inconsistency shouldn’t be sufficient to introduce tension. As Being For points out, I can wonder whether p and wonder whether ~p without tension. Desires with logically inconsistent contents may introduce a form of tension, but not a rational one. I can entertain p and entertain ~p. So it’s got to be logical inconsistency plus something else. And what’s the something else? (2) A very plausible explanation of why subtle inconsistencies among very large belief sets are unproblematic is because we have limited working memory, and so we are always reasoning with a subset of those beliefs that are jointly consistent. But it sounds really weird to say that the belief set is not irrational when that belief set would cause problems for me if I were to get smarter.

9. Hello Derek and Jack,
   Thanks for participating in this forum. I’d like to understand your project better and so I’m going to ask for some clarification about some ways that you describe your project in the paper.
   Early on, you say you are sympathetic to the project of using an unreduced notion of attitudinal inconsistency to account for how noncognitive attitudes can be inconsistent with each other (p. 392). I take it that what you call “discordance” is just such a kind of attitudinal “inconsistency” or conflict, in a general sense of ‘conflict’ and ‘inconsistency’.
   Then, you say you’ll explain the inconsistency of sentences in terms of the discordance of the states of mind those sentences express (p. 396). This sounds like what some would call a psychologistic or mentalistic explanation of the inconsistency of sentences, and so a psychologistic explanation of certain logical or semantic properties of sentences (e.g. their inconsistency relations).
   But then, at the end of your paper, you say “Rather than explaining the logical properties of sentences by holding that they inherit those properties from the attitudes expressed, the expressivist should hold that the logical properties of the normative attitudes are inherited from the sentences that express them …” (421). You also seem to want to overtly avoid a psychologistic explanation of logic (and semantics, too?).
   I find this a little confusing. Doesn’t the inconsistency relation that holds between two sentences count as a logical property of those sentences? If so, why wouldn’t you agree that you are trying to explain the logical properties of sentences (e.g. their inconsistency) as something inherited from the attitudes expressed (i.e. from their discordance)?
   Also, an additional question about discordance: suppose one were to think to herself the thought expressed by ‘it’s raining but I don’t believe it is raining’. Does that person thereby have a discordant state of mind?
   I expect you will say ‘yes’, since you want discordance to be a general kind of conflict. But then I wonder whether the Moorean sentence that I just used to articulate the relevant thought shouldn’t come out as inconsistent on your view, since it expresses a discordant state of mind and seems guaranteed to do so once we understand what it says.
   Again, I’m just asking for clarification here. I know you talk about these issues a lot in your paper, but I am looking for a simple and straightforward answer to these questions rather than trying to pick on the details of your ideas. Maybe once I feel more confident that I understand your ideas I will try to ask you some harder questions.
   Thanks!

10. Matt, thanks. We agree about semantic and logical inconsistency. Note that the official explanation of logical inconsistency does not involve truth. At all. Likewise, though we use truth to separate semantic from pragmatic inconsistency, this is because we focus on declaratives. All we really need are devices of disquotation which allow us to reiterate content asserted/commanded/etc by others without simply mimicking their utterance. Which we have or can introduce for many bits of language. So we’re onboard with the idea that a nice account of logical and semantic “inconsistency” should be similar if not identical for various moods—in fact, it’s a virtue we believe our careful separation of the property of mental states (discordance) from the property of sentences (inconsistency) offers a plausible route to.
    As to the connection between use and other accounts of expression, we have taken no particular stand, though our discussion suggests a more conventional account of expression. Something to be dealt with in further work—personally, I’m somewhat skeptical though sympathetic of overly use-oriented accounts of semantics.

11. Andrew, a quick point (and more tomorrow since it’s getting quite late this part of the world): what we mean is that the logicality of logical inconsistency—what makes it a case of logical as opposed to semantic or pragmatic inconsistency—is a feature of the sentences or the like with which the discordant attitudes are expressed. The word “inherited” might be a little misleading here as we just mean to point out the only sense in which a case of discordance can be logical—by virtue of the syntactic and meaning-of-logical-expressions features of the relevant vehicles of expression.

12. OK, I look forward to hearing more from you guys tomorrow. In the meantime, I’m going to hazard some more comments and raise a couple more issues for you to respond to.
    I think I understand Jack’s clarification: your claim about ‘logical properties’ on p. 421 is meant to be about the feature that distinguishes logical inconsistency from semantic or pragmatic types of inconsistency.
    What you actually write there in the paper is rather misleading. I’ll repeat it here for convenience: “Rather than explaining the logical properties of sentences by holding that they inherit those properties from the attitudes expressed, the expressivist should hold that the logical properties of the normative attitudes are inherited from the sentences that express them …” (421).
    What you really mean (and ought to have articulated more carefully) seems to be this: the only sense in which the discordance of attitudes can be called an instance of ‘logical inconsistency’ is that they are attitudes expressed by logically inconsistent sentences, and those sentences are in turn counted (by the Woods and Baker theory) as logically inconsistent because they are guaranteed to express discordant attitudes just in virtue of their logical vocabulary.
    I assume you will agree with this way of clarifying the quotation. I have two further points to make about it.
    First, you don’t seem to be responding directly to the worry that expressivists conflate distinct types of conflict or inconsistency. You mention literature raising such worries in fn 17. One way of articulating the relevant worry here is that it is one thing to disapprove of one’s own pattern of approvals and disapprovals and quite another thing to contradict oneself. They are different types of psychological conflict. (Blackburn at one point seemed to suggest that disapproving of one’s own approval/disapprovals could help explain the logical inconsistency of sentences, and critics responded that that conflated the relevant types of conflict.) You want to say that all psychological conflict is discordance, and that the differentiation into categories like logical or semantic conflict is primarily a distinction between sentences and not primarily a distinction amongst psychological features. But wasn’t at least part of the worry precisely that there are relevant psychological distinctions in the ways that one’s state of mind can be in conflict? Your response seems to simply deny that there is or should be a worry here in the first place, since you think there is just one type of psychological conflict (discordance). Am I understanding you rightly?
    My second point pertains to the first half of that quotation from p. 421 that I repeated just a moment ago. You make it sound like you are rejecting the psychologistic order of explanation. But, in fact, I think you want to continue to endorse it. Can you clarify?
    What I mean is that you continue to endorse the idea that the inconsistency holding between two sentences (eg. ‘murder is wrong’ and ‘murder is not wrong’) derives from the discordance or psychological conflict holding between the attitudes expressed by those sentences. You also signal (in the final two sentences of the paper) that you would like to eventually develop your ideas into a compositional semantics that uses “attitude-theoretic functions for the logical operators” (p. 424). I would categorize your claims here as part of the psychologistic approach to doing semantics and/or logic.
    Yet, in some places you seem to want to avoid that approach, such as a footnote (#18) in which you mention Charlow’s article as offering “a different way for expressivists to avoid commitment to a psychologistic semantics.” That seems to imply that you take yourselves as offering a way for expressivists to avoid psychologistic semantics. Is it possible that this is another place where your descriptions of your project are misleading?

13. Hi, Matthew.
    Though on a natural view that I accept, imperatives are not true or false, nevertheless they can be satisfied or unsatsified. Imperatives are inconsistent with one another when they cannot all be satisfied. Whatever we say about incompatibility of truth can be generalized straightforwardly to deal with incompatibility of satisfaction.
    I have been trying to argue for some years that expressivists should associate declarative sentences with genuine contents – the kinds of thing that can be the objects of attitudes like belief and desire, and the primary bearers of truth and falsity. These are the theoretical roles that are associated with what we call ‘propositions’. What expressivists need, I’ve argued, is a surprising view about what propositions are like, and a surprising view about what attitudes like belief and desire are like, to go with it. Although I did not use the word ‘proposition’ in that way in Being For, the view that I described there is like that, and I have described it in that way in other work, and I have described other intriguing expressivist frameworks that take this form in other work, as well. The dialectic about A-type and B-type inconsistency was an early, over-hasty, attempt to establish this conclusion.
    If expressivists take my suggestion, then they have propositions to work with. Sentences are inconsistent when the propositions that they express are inconsistent; mental states have nothing to do with it. Similarly, once we have propositions, we can abstract away to get entities – call them ‘edicts’ – that give us a proposition when supplied with an agent. Imperatives can be associated with edicts, and are inconsistent just in case the edicts that they express are not jointly satisfiable; that is, just in case supplying any agent yields inconsistent propositions. Again, no work is being done here by mental states; mental states were just a useful scaffold for helping us to get a grip on why the expressivist’s conception of propositions is so different from familiar conceptions, such as the possible worlds conception, or those we can get by reading King or Soames.
    You seem to be suggesting, as Charlow was, that we can work backwards from the kind of inconsistency among imperatives to the kind of inconsistency that expressivists need. Charlow thinks this because he rejects my claim that imperatives are inconsistent just in case they are jointly unsatisfiable. He thinks that there are imperatives that have ‘allowing’ force, to complement those that have ‘demanding’ force. I don’t see how this is so. If there were, then just as we can distinguish imperatives between those that issue demands, requests, orders, and suggestions, we would have to distinguish corresponding imperative that issue demand-allowings, request-allowings, order-allowings, and suggestion-allowings. I don’t think there are any such things. When you want to give permission to do something, you don’t say, ‘Do it’, you say, ‘you may do it’. But this sentence is not an imperative; it is a deontic modal sentence.
    Indeed, it is no surprise that Charlow needs to assume this strange thing in order to get leverage out of this idea; the very same dialectic that expressivists face about conflicting attitudes can be run in terms of conflicting imperatives. Instead of claiming that attitudes only intrapersonally disagree when they are the same attitude toward inconsistent contents, I would say that imperatives are only inconsistent when they express incompatible edicts. Just as an expressivist can hope to evade the worry by positing additional attitudes and postulating that they bear the right intrapersonal disagreement relations to one another, Charlow can posit different speech acts performed by imperatives. I don’t see anything relevantly different about the dialectic, here.
    Since I don’t think there are any such things as allowing imperatives, I think imperatives do only conflict when they are jointly unsatisfiable. So it’s backwards to start with whatever happens with imperatives and try to explain what happens with moral sentences. Instead, we should start with declaratives and generalize to imperatives, in the way that I indicated. This has the further virtue that all of the work in explaining inconsistency of imperatives is done by their contents, and none by the mental states expressed – if imperatives count as expressing mental states at all, which we need not assume that they do, as I see it.

14. Andrew,
    I think some of this debate is being lost in a difference in our idiolects. Could you say something about what you mean by ‘psychologistic’ or about what type of psychologism you find worrisome? Since we identify logical properties as grounded in non-psychological facts about the meaning of certain expressions, our view is not psychologistic in the usual sense, even if the property guaranteed to be instantiated on their basis is a property of psychological states.
    Likewise, we explicitly avoid using the term ‘inconsistency’ for properties of attitudes since it gets confused with formal properties of sentences and the like. As we point out early in the paper, attitudes with inconsistent contents are not thereby inconsistent in the intended sense–consider fearing or hoping–so there is no obvious formal property of attitudes that tracks the target notion. We introduced discordance to account for that bad-making property, whatever it may be (and offered a few alternative explanations.) We went so far as to use to indicate discordance when quoting others since we were worried about terminological confusions such as thinking that inconsistent attitudes are simply attitudes with inconsistent contents. Given these early stipulations, it seemed clear enough to us to say what we did as a summary of the extensive discussion that preceded what you quoted. In short, we said that logical properties, such as there are, of a case of discordant attitudes are derivative from the relevant syntactic properties of vehicles in an effort to parallel what we want to vehemently deny–that the logical properties of sentences are derived from some unspecified logical properties of attitudes themselves. Logic is formal, so logicality needs to lean on formal structure. Sentences have this, but not all potentially discordant attitudes do.
    If it helps, think about an analogy to unstructured propositions in a possible worlds semantics. Sentences have logical properties on such a view because they denote, by means of the meaning of the logical expressions and syntactical features of the sentences involved, various sets of possible worlds such as the empty set. But the propositions themselves do not—they have far too little formal structure. But if the sentences didn’t denote they would have no logical properties at all, being mere strings. Our view is completely analogous. Sentences have logical properties because of their syntactical features, but they wouldn’t have any such properties if they didn’t express mental states capable of standing in discordance relations.
    

15. Hi Mark (Schroeder),
    I have a question for you regarding what you just said about propositions and expressivism. It’s not an objection to what you said there, so much as request for clarification. If I understand you right you think that expressivists need propositions, even if they have an unorthodox position on them. But one of the main concerns of our paper (which we discuss towards the beginning), is that propositions by themselves don’t necessarily solve the Frege-Geach problem, unless those propositions have logical properties. Otherwise you lack resources to distinguish logical entailment from a posteriori entailment. Now the most obvious picture on which propositions will have logical properties, though, is a Fregean one. And if expressivists have to embrace Fregeanism about propositions, what’s the point of being an expressivist?
    So presumably it’s propositions plus something else that solves Frege-Geach, but what’s the something else? We thought the most natural answer is formal properties of the language. But then, our thought was, once you have allowed that properties of the language can account for the logicality of certain conflicts, it is hard to see what in principle objection there is for forms of expressivism you find objectionable. Admittedly, you may be in a better position to offer a more straightforward compositional semantics. And such accounts may die in the details. But it’s not clear why in principle they couldn’t work.
    So what kind of propositions are you taking on, how do help solve Frege-Geach, etc?

16. Point of clarification to the above: I may have implied that Fregeanism was the only way to get enough structure in one’s propositions, which isn’t correct. But if propositions are structured enough to have logical properties in Mark’s expressivism, we’d like to know what his picture of them is.
    “Main concern” is also probably better read as “big inspiration.”

17. Hi, Derek.
    I’m not sure what you think the Frege-Geach Problem is, or why you think that it is a consequence of what I said that propositions must have logical properties. What I think the Frege-Geach Problem is – at any rate, the most general problem in the vicinity – is how to give an adequate account of the meanings of complex sentences. Adequate in what sense? In that it must give complex sentences the right kinds of meanings to be an ingredient in a complete explanation of why complex sentences have all of the properties that they do, which they have in virtue of their meanings.
    Why is logic important for this task? It’s because one way in which complex sentences are formed is with logical words like ‘and’, ‘not’, ‘or’, ‘all’, and ‘some’. But logical relationships between sentences hold in virtue of the meanings of logical words. To say this, as you have been emphasizing, is just to repeat an orthodox platitude about logic. So getting logical relationships right is one important test – not the only important test, but an obvious and central important test – of adequacy for an account of logical words, and hence for an account of the complex sentences formed by those logical words.
    Does it follow that logical properties must be properties of propositions? No more than someone who thinks that propositions are sets of possible worlds must think that propositions have logical properties. On the contrary, logical relationships are properties of sentences. They hold of sentences in virtue of their meanings, but not only in virtue of their meanings – also in virtue of their structure. And not in virtue of every fact about their meanings – only in virtue of the meanings of the logical words. So I will say – as I think should anyone else – that two sentences are logically inconsistent just in case they are inconsistent in virtue of their form – that is, just in case any constant choice of meanings for their non-logical words would make them inconsistent. See page 112 of Being For.
    I agree that propositions by themselves do not solve the Frege-Geach Problem, but I don’t think for the same reasons. It is easy to find natural assignments of sentences to things that behave in interesting ways like propositions but fail to have an adequate account of what it is to believe those propositions. But this isn’t because propositions don’t suffice to account for logic – I don’t think that anyone ever thought that genuinely logical, as opposed to broadly semantic – inconsistencies are supposed to be explained in a way that does not appeal to sentential structure. It’s because they give bad accounts of what it is to believe certain propositions, or in general because they make other bad predictions about which properties complex sentences will have – either not giving us enough resources to explain properties that they should have, at least partly in virtue of their meanings, or predicting that they will have other properties that they in fact do not have.
    One feature that I do not like about your treatment of the distinction between semantic and pragmatic inconsistency, is that you are allowing an equivalence from within the agent’s perspective. What makes the difference is what other agents can say about her. But now compare two inferences, considered as inferences between the thoughts expressed by two sentences, rather than inferences between the sentences themselves. The first is from ‘P’ to ‘~Q’, and it is licensed because ‘P’ is inconsistent with ‘Q’. The second is from ‘I don’t believe that P’ to ‘~P’. If there is something equally bad, from a rational point of view, about thinking both that P and that I don’t believe that P, as there is about thinking that P and that Q, then why aren’t the same inferences rational? Don’t tell me that it’s because third parties could say different things about the sentences used to express those thoughts; I’m worried about the thoughts themselves, independently of how they are expressed.
    Because I think the Frege-Geach Problem is about providing at least the right resources for explaining all properties of sentences that are due to meaning, and not just about logic, I think it’s just as bad to fail to be able to explain the differences between which inferences are rational, as it is to fail to explain which sentences are inconsistent – even if we are not explaining inconsistency in terms of the rationality of inferences.

18. Derek and Jack,
    I would like to “pester” you again, because I am not persuaded by your reply (to me and to Mark van Roojen). I am willing to grant (but don’t quote me, because it is not my *considered* opinion) that inconsistency (and failure of deductive closure) is irrational. But I thought you were requiring particularly *severe* failures of rationality as necessary (and sufficient?) for discordance, so don’t take every irrationality to be a case of discordance. So I still don’t see how you manage to draw the line in the right place. Believing something on the basis of a subtly fallacious proof renders you inconsistent (because the conclusion is inconsistent with the axioms which you accept), but far less irrational than (say) you would be if you jumped to a conclusion on the basis of weak evidence (which, I take it, isn’t a case of discordance).
    As to functional role, I agree that an inconsistent belief system doesn’t perfectly represent the world. But *no* human belief system does. So only *egregious* functional failings can be taken to render a belief system discordant. And, again, the criterion will draw the line in the wrong place (vis-à-vis inconsistency), because inconsistent belief systems can do a fairly good job, better than do many consistent ones. And (I’m repeating myself here, but I am not persuaded by your reply) that an inconsistent system is *guaranteed* to represent imperfectly doesn’t exacerbate its failure. What matters is its *actual* success, no? Analogy: that my car won’t function well in arctic conditions doesn’t count against it (when I consider whether to replace it). What matter is how it functions in my (Mediterranean) locale.
    Ruth

19. Hi Mark and Matthew:
    I agree with Mark that expressivists need an account of propositions. However, I disagree with what he says later – or at least, I think it is a more tendentious claim than it might at first seem: “Sentences are inconsistent when the propositions that they express are inconsistent; mental states have nothing to do with it.” On the view I develop in *Impassioned Belief* We do explain meanings in the first instance via thought expressed, but it turns out that for certain mental phenomena, if we are talking about types rather than tokens, then the mental state/proposition distinction (or, perhaps better mental event/proposition distinction) may encode a false dichotomy. I defend a modified version of Soames’s account, on which propositions just are cognitive event types. I also argue that there are ways of adapting my view to avoid reliance on that particular conception of propositions (and, indeed to combine it with slightly adapted versions of all of the leading accounts of propositions), but that is the view I find most attractive. On that view, it is at least misleading to say that inconsistency is explained in terms of propositions and mental states have nothing to do with it. And I don’t think the point is entirely unimportant, since it is important to a view worthy of the name expressivism that it in some way explaining meaning via state of mind expressed.
    – Mike

20. Hi again (mostly directed at the exchange between Mark and Matthew):
    I should have explicitly said that once we have explained how to associate proposition contents with declaratives in some way that is expressivism-friendly (I have one strategy, Mark has another), I agree with him that we can use propositional contents to explain the meanings of imperatives in terms of satisfaction conditions. It is true that we do sometimes *use* imperatives to give permissions (e.g. “have an apple”) but it is not so obvious to me that this is part of the semantics for imperatives, as opposed to a pragmatic phenoemena. After all, we can also use declaratives to give permissions in the right context, “I think you’d enjoy one of these cookies I brought” (while holding out the jar of cookies).
    One last thing – near the end of Matthew’s original post he says the following:
    “But I’m also skeptical that the best non-realist explanation of this is going to come from purely psychological properties of the mental states canonically expressed by the relevant sentences rather than something having to do with the kinds of moves they can be used to perform in different conversations. (It’s not even clear that imperatives canonically express one kind of mental state, since they can be used to command, advise, suggest, entreat, forewarn, invite, offer, etc.)”
    A few observations about that remark.
    First, notice that even if expressivists do explain the meanings of every sentence type directly in terms of state of mind expressed it does not follow that they are committed to a one-to-one correspondence between a given sentence type and a given state of mind, as the parenthetic remark requires to have its intended dialectical force). Just as inferentialists can explain meanings as functions from conversational contexts to possible moves in a language game (as you suggest here), expressivists and other ideationalists can explain meanings as functions from conversational contexts to states of mind expressed. What is good for the inferentialist goose is good for the expressivist gander, after all.
    You might then worry that we do more with language than express states of mind, but (a) expressivists don’t have to deny that either, in general – they just want to argue that the most important/distinctive use of e.g. moral language is to express states of mind, and there at least, we should explain meaning in this way (they may want to explain a lot more than that via state of mind, but they don’t have to qua local expressivist) – in this sense inferentialism is actually compatible with a local old fashioned expressivism on one way of understanding that view, and (b) note that this is a rather different worry than the one originally expressed.
    Second, once we have established propositional contents in some way via explaining the meanings of indicatives – perhaps by privileging their use as assertions – inferentialists also do not have a monopoly on the idea that assertion is “downtown”!), expressivists can then use propositional contents to explain the meanings of imperatives in the way that Mark suggests – via satisfaction conditions by invoking propositions again. In that case, again, we are not in any way driven to explain the meanings of imperatives directly in terms of state of mind expressed – the connection to state of mind expressed will be much more indirect, and only used to earn the right to talk of satisfaction conditions.
    Just trying to avoid some false dichotomies here, I guess. I’m an ecumenical kind of guy…

21. Ruth,
    ‪ So, I agree that there is a difference between how well a belief system (as a whole) satisfies its functional role and (local) inconsistency, but contravention is something altogether different. Though this is a place where Derek and I subtly disagree (he prefers interpretive incoherence, I prefer functional role-type stories). On my view, contravention of functional role does not always lead to a bad result in actual cases. It can be quite good to be discordant, in a sense. But there is still a palpable sense in which the discordant attitudes aren’t doing what they are designed to do, even if the result turns out to be better overall for what the entire system of beliefs is designed to do. (Derek adds that he thinks interpretative incoherence can do the same thing but that if it can’t he’ll convert to my position.)
    By the way, apologies for some delays in responding. Due to our Voltron-like nature and separation by 6 hours, composing joint replies takes a bit. And when a commitment should be solely attributed, we’ll mention it, as just above.

22. Mark,
    We think we may have triggered an unintended implication—we did not mean to imply that you thought that propositions have logical structure, merely that that route won’t work (though, of course, we did mean to imply that some have thought this.) But your ensuing suggests that you agree with us on our main point and are more concerned with the extension of our solution to cases outside of the logical properties of sentences. Our understanding of Frege-Geach(-Searle) is stated, by the way, in the final paragraph of the paper (we agree with your characterization. And our discussion of Gibbard in the paper makes the point that the extension to a compositional story beyond logical connectives is not at all obvious. But future work and all that.)
    We agree that the claims we’ve been making about logic are boring and orthodox, and we assume everyone if explicitly asked about it would endorse them. Nevertheless we pull out a lot of quotes from various authors that we think fit badly with that orthodox picture—attributing logical properties to attitudes and propositions, for example. We don’t think we’re taking these out of context, and we suspect that somewhere along the way an implicit assumption that expressivists were committed to psychologism snuck in. We’d like to clear that up.
    The point about reasoning is an interesting one. Here’s what we roughly think (as this goes beyond our paper). Obviously the inference from “I don’t believe that p” to “~p” is fallacious in the sense of not instantiating a good implication pattern. (and, of course, it’s fallacious anyways because not believing p is not the same as believing ~p. But let that pass.) But since we haven’t offered a theory of reasoning, we’re not clear what would suggest we are committed to this. We don’t tie discordance or to an inference-licensing property, and so acknowledging that the Moore-paradoxical agent is discordant doesn’t force us to say anything about inference, at least not by itself.
    Not only would we insist on a gap between discordance and the rationality of particular inferences. We would insist on a gap between full-fledged logical or semantic inconsistency and inference. This is the familiar point that the logical or semantic inconsistency of p and q does not license adding ~q to our beliefs if we believe p. We might need to drop p, we might need to do nothing, we might need to go whistle Dixie because belief hygiene is stressful. A proper theory of rationality and inference is not going to be so tightly tied into notions of inconsistency, the specification of which was our main concern in the paper. We worry as well that any plausible theory of reasoning is going to be holistic and very complicated.
    (As an aside, you do tie inconsistency with inference-licensing properties, if we haven’t misunderstood you. Presumably the Gil-Harman points above are not something you regard as problematic for your view. So we’d be curious how inference-licensing works in your system—in case we need to steal it.)
    Now we do think that whatever story we end up telling about reasoning and inference will assign a lot of work to the language of expression (the very beginnings of a story would be IV.C. of our paper), and you may find this objectionable. We’d be interested to know more what the objection is, but our thought is that of course once you learn a language properties of that language can change how you reason about the subject matter. This must be true or what would be the point of formalization? So at present we’re fine with cutting off chains of reasoning with the thought that “the truth of p has nothing to do with the truth of q”—that seems entirely respectable even with our take on truth. (Along these lines, we indicate in section IV.C. that we’re attracted to a picture of belief on which the contents are linguistically tagged. We haven’t worked this out yet, but we think this would be attractive for reasons not even especially tied to expressivism.)
    This ties into our response to Ruth. Discordance is a distinctive property mental states can have, but the result of which is not to rule out that there can be many other reasons for not updating in various ways. Both patterns of implication you lay out involve discordant attitudes. But there are other problems—adding in the belief that ~p whenever I realize that I don’t believe p will lead me to fail to accord my beliefs to the evidence, for example. In short, meaning is not sufficient to explain which patterns of inference are rational and which are not.
    One concession we should make—to you and Ruth and we suspect a few other people—is that we may have overlooked the difference in two kinds of logical failings: failings of inconsistency and failings of closure (and of course their non-logical but semantic analogs). This may force us to modify our account of inconsistency some: inconsistent sentences will express discordance that results, roughly, from a surfeit of attitudes; while discordance that results from lack of attitudes may find “expression” in failure to accept implications or entailments. This is something we need to think about. So thanks for bringing it to our attention.
    VOLTRON (or Jack and Derek)
    

23. Mike,
    Thanks for the comments. We have no arguments against expressivists using propositions; we just think it would be so badass if they didn’t. We want to see if we can pull it off.

24. Sorry, this is mainly directed at Mark re. imperatives:
    I totally agree that expressivists (and everyone else!) should associate declarative sentences with genuine contents – the kinds of thing that can be the objects of attitudes like belief and desire, and the bearers of truth and falsity. One can of course hope, fear, prefer, etc. that stealing is wrong; and in that case there better be something that one hopes, fears, prefers, etc. (Personally I’ve always thought those contents should be explained in terms of their inferential roles rather than their representational purports.)
    What I meant to gesture at in invoking imperatives was merely that I think we should also associate genuine contents with imperative sentences, though I suspect they’re going to have to be of some different type than we associate with declarative sentences, e.g. they’re not going to be bearers of truth and falsity (and I’m going to say that they have interestingly different inferential roles from declarative contents). Nonetheless, it seems that there’s going to be something that it is for these contents to be inconsistent. And that’s enough to warrant our thinking that not all inconsistency, not even all logical inconsistency has to be explained in terms of impossibility of conjoint truth.
    Although some normative declaratives seem to me like they are inferentially related to imperatives, I don’t think that means the expressivist should model inconsistency of normative declaratives on the inconsistency of imperatives. Whatever we say about permission-style imperatives like “Have an apple!”, I’m with you here: you cannot hope, fear, doubt, prefer, etc. something imperativial, but you can hope, fear, doubt, prefer, etc. inconsistent things, including inconsistent normative things. However, the fact that (logical) inconsistency is not in general a matter of impossibility of conjoint truth made me skeptical of the assumptions that seemed to be in the background of the disagreement with B&W. Jack’s comment above makes me think I may have misunderstood what these assumptions are.
    For what it’s worth, I think your account of the inconsistency of imperatives is going to have to be more complicated. “Bill, kiss Lucy!” and “Lucy, don’t be kissed by Bill!” are not mutually satisfiable but it’s not obvious that they are inconsistent, at least not logically inconsistent.

25. And this one is addressed at Mike re. imperatives:
    I think expressivists (and everyone else!) should agree that one of the important things we do with moral language (and all other language!) is to express our minds; and we should also all agree that this not all we do with language. My skepticism about trying to explain all of the semantic and logical properties of sentences (my focus above was imperatives but I think this is also true of declaratives and interrogatives) in terms of properties of psychological states is that this just pushes the bump in the rug: what then explains why those psychological states have the intentional and logical properties that they do? Maybe sometimes it has to do with their representational purport, but I’m inclined to think that this cannot be the most general answer (because many mental states don’t have representational purport).

26. Matthew: Yes, I agree with pretty much all of what you just posted, apart from the intended rhetorical focus of “pushes the bump in the rug.” It is an important part of the broader expressivist/ideationalist project to explain how different states of minds have the contents they do have. And I agree of course that the story will not always be a representational one. Of course, there is a good question whether we should take mind or language (or neither) to be fundamental when explaining contents, but I don’t see why the expressivist should be seen as pushing a bump from the linguistic side to the mental side of that equation any more or less than the inferentialist should be seen as pushing that bump from the mental side to the linguistic side. Ex ante, which order of explanation (or whether to take a no priority view) should be entirely up for grabs. I fear we are now in danger of going into issues we’ve discussed in person endlessly though so perhaps we should take this offline!

27. “What then explains why those psychological states have the intentional and logical properties that they do?”
    Well, for logical properties, we split the explanation in two parts. What explains why the sentences are inconsistent is that the sentences are guaranteed to express the psychological state of discordance (in virtue of meaning and grammar). What specifies the type of guarantee (semantic, logical, pragmatic, etc.) and explains why it is expressed are properties of the language itself. Seems to us this nicely splits the horns of your worry.
    Psychological states do not have logical properties directly. This is a category mistake. Or so we have claimed. (although you can easily define a derived sense of it as we mentioned above).

28. Matthew,
    I think ‘Lucy, kiss Bill’, is like ‘Lucy, Bill is a nice guy’ and like ‘Lucy, is Bill a nice guy?’. The ‘Lucy,’ part simply makes clear who is being addressed, and no more plays a role in what it takes for it to be satisfied than in the declarative version it plays a role in what it takes for it to be true, or in the interrogative version it plays a role in what is being asked.
    As I said, I think imperatives express properties, in the sense that properties can be extracted from propositions by abstracting away from some individual. The imperatives ‘Bill, kiss Lucy’ and ‘Lucy, don’t be kissed by Bill’ express the properties lambda(x)(x kisses Lucy) and lambda(x)(x is not kissed by Bill). These properties are jointly satisfiable, so I agree that they are not inconsistent commands. There is something else wrong with them.
    I actually think that this supports the view that imperatival contents are property-like rather than proposition-like (and hence truth-evaluable), though I also think we can also find support for this view by looking at embedded imperatives, just like we can get support for view about the semantic value of interrogatives by looking at embedded interrogatives. Embedded interrogatives are headed in English by ‘whether’, ‘if’, or wh-words; embedded imperatives are non-finite clauses. There is more to say about this than fits in a blog post, but I think we’ve talked about this before.

29. VOLTRON,
    I think you’ve misunderstood what I have written about the ‘inference-licensing property’ if you think that it runs afoul of Harman’s point. Harman’s point, as you say, is that if you find yourself believing P and believing P->Q, it may be that the right thing for you to do is to conclude Q, but it may be that the right thing is to give up your belief in P->Q, and it may be that the right thing is to give up your belief in P. The inference-licensing property does not say, despite its name, that if you believe P and believe P->Q, you are license to infer Q. It says that if you believe P and believe P->Q, you are rationally committed to believing Q, and something is going wrong so long as you are paying attention to the question and you persist in the state of non-closure.
    As I read you, you are granting that closure norms are something that you still need to explain, and agreeing that your explanation of them will not come out of your account of inconsistency. My point was just that you took pains in your paper to try to address what you took the lesson of the distinction between semantic and pragmatic inconsistency to be. According to you, the lesson is that we must make this distinction, and so you responded by offering a way to make it. My suggestion is that there are other lessons that we might take from this distinction.

30. Derek and Jack,
    It’s great to be having this conversation across so many different time zones at once. Let me try to push things in a different direction by asking you a new question, in the spirit of Ruth’s and Mark’s.
    What do you make of sentences like ‘nothing exists’, ‘there are no creatures with beliefs’, and ‘every belief ever formed has been irrational’. Is it possible to coherently attribute the beliefs expressed by these sentences to anyone? Is it possible for them to play the functional role of beliefs? It seems to me that each of these sentences exhibits, by itself, the pragmatically problematic nature of Moorean pairs like ‘P’,’I don’t believe that P’ and ‘P’,’it is irrational for me to believe that P’. If you don’t like my use of single sentences, we can use the pair, ‘P’,’it is irrational for anyone ever to believe that P’. So I suppose that each of these expresses a thought that is discordant.
    But I don’t see how appeal to what a third party can say about truth will help with these cases, because they don’t include indexicals like ‘I’. I constructed them by looking for non-indexical sentences that are strictly stronger than ones like those you consider. So my question is, how does your account deal with cases like these? Do we conclude that they are semantically inconsistent after all?
    If I were you, I would appeal to modals at this point. I would clarify that it matters not what it can be that a third party says, but what anyone can say about what could possibly be true. This helps with some of the examples – though no one can coherently think that nothing exists, anyone can think that it is possible that nothing existed. But now consider the pair, ‘necessarily P’,’necessarily, it is irrational to believe that P’. I take it that these are semantically consistent, if we are to avoid some sort of universal idealism. Given S5, to think that it is possible that both of these are true is just to think that they are both true. But that is incoherent in the same way as it is incoherent to accept the pair ‘P’,’it is irrational for me to believe that P’. But I wouldn’t think that this is a good argument against S5, nor a good argument for idealism.

31. Mike –
    I wanted to follow up on two things that you said. The first is that when I interpret your view, it makes most sense to me when something different plays the role of propositions than any of the interpretations that you mention. The way I interpret it is simple: take any view you like about what you thought propositions – the objects of ‘narrow’ belief – were supposed to be. Call those pre-propositions. Real propositions are functions from ideal advisors to pre-propositions. To believe a proposition is just to have a ‘narrow’ belief in the pre-proposition that is its value at your own ideal advisor. Now psychology can drop out of the semantics. It may or may not be that pre-propositions are mental state types, but I don’t think anything about whether the view is recognizably expressivist hangs on it.
    You mention that you think if mental states drop out, then we lose some of the claim to be expressivists, because it is important to expressivism that we can characterize meaning in terms of the mental states expressed. But here’s another way of looking at it. The view that I’ve just given of propositions is strange. They aren’t the right kind of thing to be the objects of ‘narrow’ belief – the state that we all thought belief to be, and of which off-the-shelf functional role accounts of belief attempt to theorize. It’s only once we understand what it is to believe these propositions that we understand how this could be an insightful characterization of meaning, and so mental states are what give us our grip on why it is a characterization of meaning. Can’t that be enough for the expressivist?

32. Hi Mark,
    Well the interpretation you put forward isn’t really my view as laid out in the book, though I suppose it is a variant of the core ideas that has certain virtues – putting to one side that you’ve glossed it in terms of ideal advisors rather than normative perspectives, but the structure would be the same of course. And I do discuss another strategy in the book for making sense of propositions within my framework that is similarly neutral (in the “What if Soames is Wrong?” section) – though it isn’t quite the strategy you’ve laid out here it is close.
    I don’t think this blog, which was meant to be more on VOLTRON’S ideas, is probably the best place to compare and contrast the virtues of these different ways of going. I suppose I just wanted to flag up that your comment about how its all propositions and mental states have nothing to do with it is, at the very least tendentious – and would still be tendentious even given the approach you lay out here since that approach is still entirely *open* to a Soames account of propositions (or pre-propositions, at least) at the end of the day.

33. Hey all,
    There are a lot of great questions here. Just wanted to apologize that we will be on hiatus for a while longer, due to timezone differences and not sleeping enough on the first day of discussion.
    Best,
    Derek

34. This is a new question to Derek and Jack:
    Although I agree that there are B-type discordances and that they are a “natural kind”, I am not persuaded by all of your examples.
    a) I am unsure about credential states (p. 17). True, ascribing probability 0.6 to p and 0.5 to ~p is incoherent (violates the probability calculus). Indeed, is a case of logical inconsistency if we (plausibly) view the probability calculus as a generalisation of logic to degrees of belief. But if we think (admittedly contentiously) that belief is degree of belief greater than 0.5, this violation will be a case of A-type discordance: believing p and ~p. Perhaps less contentiously, this combination of attitudes involves being at least somewhat inclined to believe both p and ~p. There is no short name for this mental state, but it is psychologically real nonetheless.
    b) You say that the view according to which liking and disliking someone is an A-type discordance is “naïve” (fn. 54). Now, I think the authors that you cite by way of support (or those that they cite) argue against stronger claims than the one Schroeder needs, e.g., that ALL intentional states are propositional (Grzankowski) or that propositional attitudes are the most basic (Montague). To be sure, it is preposterous to equate liking a person with loving a proposition, but to construe liking and disliking as A-type discordance it is enough to suppose that liking someone is having an attitude towards (possible) states of affairs which are described by propositions. So what is wrong with the (much more modest) suggestion that liking someone is desiring that she thrive and that one spend time with her (both of which are propositional attitudes)?
    Ruth

35. Hi Mark,
    Thanks, that clears things up a lot, and these are helpful suggestions on how to characterize the dispute. The shift of focus to issues of rational commitment is also a really useful and interesting. If we’re not mistaken, one of way of putting your concern is that we’ve taken the distinction between semantic and pragmatic truth as a data-point, but said nothing about why that distinction would be important, and our account of the difference gives no real sense of why it would be important. That’s fair enough. It is something we’ve been thinking about, but what we’re going to say here will be beyond anything we’ve worked out to our satisfaction. But in for a penny, in for a pound, eh?
    Digression – We are not engaging here with the useful point about norms of closure—primarily because we are still thinking it through! We regard such norms as difficult to specify in any precise way for many reasons. So, for example, believing that Socrates is wise and good rationally commits me to believing that Socrates is wise [or correspondingly for beliefs] seems clearly right. But that believing that truth is naïve rationally commits me to denying modus ponens, conditional proof, or some other incredibly basic inferential rule is nowise clear (on pain of Curry’s paradox). And the line between these notorious hard to specify. But note that the vagaries attending to norms of closure make our taking no serious commitments about it a wise, if unconscious, rhetorical move. We can still say that such cases involve logically inconsistent—since we have guaranteed local discordance by asserting the former and the negation of the latter—but leave open that cases of withholding are discordant and related questions about their rationality (local and global). End Digression
    So first of all, we prefer to think of the relations between mental states primarily in terms of rational pressure, rather than reasoning, inference, or rational commitment. And because we’re primarily thinking of discordance in terms of violations of coherence-type constraints, we think that usually rational-pressure will go both ways. To use your example, does accepting P and P→Q rationally commit one to accepting Q? We don’t know if we’d say no, but we’d prefer to put it that it puts rational pressure on the subject to accept Q—but the fact that she doesn’t currently accept Q also puts rational pressure on her to give up P or P→Q. Rational pressure in either direction can also be outweighed by the rational pressure coming from other attitudes, etc. So the process will be relatively holistic, with limits on the holism set in arbitrary places by things like the size of our “working memory.”
    Some of this rational pressure is easily categorized as related to reasoning, and especially inference. But a lot of the rational pressure is different. If I judge something is dangerous, this puts rational pressure on me to feel fear. But it is very odd to treat the feeling of fear as the conclusion to a sequence of reasoning (“X is dangerous, therefore I am afraid of X”? “…therefore I shall be afraid of X”?). We also know that fear can lead to the judgment that something is dangerous—and admittedly there’s reason to be suspicious of the rationality of that. But as good coherentists, we are committed to the rational pressure going both ways. And Damasio, for example, has presented empirical evidence that actually the pressure emotions exert on our judgment deserves to be called rational. Karen Jones has a nice functionalist account of the emotions which would support the same kind of position. But again, this would look terrible as a piece of reasoning (“I am afraid of X, therefore X is dangerous”).
    Two other issues here. First, we think that as expressivists we should start unsettled about whether the rational updating of the attitude-type in the light of other attitudes should be regarded as reasoning, and especially questionable if it is inferential reasoning. The expressivist, we think, should earn his right to this. We might proceed as follows: a process of rational updating of attitudes gets called inferential reasoning just in case it is expressible as an argument (valid or otherwise—because induction, etc.) whose premises and conclusion are assessable for truth and criticizable for falsehood. The last is key, as we think that forming an intention on the basis of a normative judgment, for example, deserves to be called reasoning, but not a form of inference—and this has something to do with the direction of fit (or other replacement property) of intention. This doesn’t tell us which pieces of reasoning are good and which are fallacious—but drawing that distinction in general terms is notoriously difficult anyway. (Ad hominem or arguments from authority are sometimes good; objecting to lack of deductive validity is often fallacious.)
    Second, about Moore-paradoxical beliefs and related phenomena: we unfortunately only cited Shoemaker, but we had in mind a tradition of understanding self-knowledge that regards it as not a purely epistemic problem—this would also include, for example, Victoria McGeer—whose “Is Self-Knowledge an Empirical Problem?” deserves a huge shout-out—and Richard Moran. First-order beliefs might put rational pressure of the coherence sort on the agent to form the appropriate higher-order beliefs. Higher-order beliefs may also put rational pressure on the agent to adopt the target attitude, though people are cagier about this. (Derek has unpublished worked insisting that they do, because they have a functional role, roughly, of maintaining mental hygiene.) Notice that, even if there is this kind of rational pressure, that doesn’t commit us to treating the relation as one of inference.
    Now, if this tradition is wrong, we can deny that false higher-order beliefs involve any discordance at all, and so deny the Moore-paradoxical sentence is a semantic inconsistency in roughly the way you do. But it might be right, and in any case we know de se attitudes do different functional work than their third-person counterparts, and we wanted to keep that from “infecting” the semantic picture—so we thought we needed to develop something to answer that.
    Okay, preamble out of the way, why is the semantic-pragmatic distinction important? So roughly we think that when you assert something, you express an attitude, and effectively somehow invite others to share that attitude, or else indicate that they are calling your assertion into question. So we start communicating to reach consensus, and to make the breakdowns of consensus explicit for all the reasons that could be useful. So far, pretty standard.
    Now ‘truth’ is a device for reassertion, quantification, etc. But as we point out, this allows us to use ‘truth’ as a device for context-stripping, and we think there are obvious reasons why assessing the context-free status of claims is going to be extremely important in reasoning. Grandiloquently, this allows us to move from the particular, localized, potentially parochial point of view and consider universal principles and assess claims on context-free merits, and that project has all sorts of payoffs. To give concrete examples: you present the Peter Singer’s arguments to your students, and they ask if you donate all your money to charity; you present the arguments for skepticism, and they ask if you believe the conclusion. Telling everyone you want to focus on what’s true is a nice way of shifting focus away from the instructor or, more generally, the speaker, and to general features of the problem that survive (and become potentially clearer) when we abstract away from the particulars of the assertion.
    So the picture will look like this: we start with attitudes and their rational relations. We get a way of expressing these attitudes, and the sentences uttered in that way get their meaning in part because of the rational relations between the attitudes. But we also notice that the language offers devices to abstract away from some of these rational relations; and this allows us to engage in certain forms of theoretical inquiry and, more generally, to give wider universality and general applicability to our thoughts. Once we become familiar with these advantages, we start more and more to update attitudes not just unreflectively, but now also on the basis of the ways we express them—that is, we engage in conscious reasoning.
    Note that on this picture, the line between semantic and pragmatic inconsistency will be slightly fuzzy—not a terrible commitment, we think. As long as we can slot particularly clear cases, as we tried to do in the paper, then the fight about the rest is an interesting theoretical one.
    We still owe you a response to the “idealism or S5” dilemma. Mike hit us with a version of that a while back. It’s a problem for us. But this post is long enough.
    Derek and Jack

36. mea culpa
    read `semantic and pragmatic inconsistency’ for `semantic and pragmatic truth’ and replace the digression with
    Digression – We are not engaging here with the useful point about norms of closure—primarily because we are still thinking it through! We regard such norms as difficult to specify in any precise way for many reasons. So, for example, believing that Socrates is wise and good rationally commits me to believing that Socrates is wise [or correspondingly for beliefs] seems clearly right. But that believing that truth is naïve rationally commits me to denying modus ponens, conditional proof, or some other incredibly basic inferential rule (on pain of Curry’s paradox) is nowise clear. And the line between these notorious hard to specify. But note that the vagaries attending to norms of closure make our taking no serious commitments about them a wise, if unconscious, rhetorical move. We can still say that such cases involve logically inconsistency—since we have guaranteed local discordance by asserting the former and the negation of the latter—but leave open whether cases of withholding are discordant and related questions about their rationality (local and global). End Digression

37. Hi Ruth,
    One of the reason we gave so many examples is that we figured people would object to some, but no one could reasonably object to all. But both examples you worry about are near and dear to us. For your worry about credences, note that identifying them with beliefs is contentious, and in any case, even if that identification works, it is very likely that beliefs reduce to credences rather than the other way around. After all, assuming the identity of beliefs and credences, it seems that a credence of (say) .7 in p is identical to the belief that p. But the belief that p could be identical to a whole slew of possible credences in p. So the credence seems like the more fundamental state.
    But if we explain beliefs in terms of credences and say that putatively B-type credential discordance is explained in terms of A-type discordance of belief, then we have properties of beliefs underwriting properties of credences, but credences underwriting beliefs. But this is strange. If we explain beliefs in terms of credences, surely we want to explain discordance of belief in terms of discordance of credence. And the latter form of discordance is B-type.
    We are more troubled by your suggestion about liking and disliking, but not convinced. Can’t I like someone without wanting to spend more time with them or desiring they thrive? Phenomenologically, I think yes—though perhaps this says nasty things about me (Jack) as a person! The worry we moot in the paper for this sort of approach is that there is no determinate state of affairs that corresponds to liking and disliking, so this sort of reduction seems problematic. We could say, alternatively, that to like someone is to have a propositional attitude towards some state of affairs from a cluster of possible ones (spending time, them thriving, them not dying quickly, etc), but this seemed to us a rather implausible view.
    A more plausible alternative is to say that the discordance, such as it is, of liking and disliking is derivative from desires or other propositional states that are constitutively tied to liking and disliking. For example, maybe liking is not the same as, but does involve, certain characteristic desires from a certain pool. Then, if those are A-type discordant, maybe the discordance of liking and disliking one and the same thing is just a spandrel of the A-type discordance of the desires thereby taken on.
    A version of this paper (actually, a large part of Jack’s original contribution) took on these sorts of worries, but much of the discussion had to be cut for reasons of space. But roughly the argument went like so–the grammar of our affective attitude ascriptions is not surface-propositional. We thus need a convincing reason if we are to construe the described or attributed attitudes propositionally. The most plausible grammatical reasons for thinking that the grammar of the ascriptions is deep-propositional do not work (this was the comparison to want-want to have). There are reasons of indeterminateness to think that the interior propositional attitude very different between different cases so that a reduction will be very difficult—maybe I like donuts because they are delicious, but want them very far away, but maybe I like Fred because he’s good fun and want him present as much as possible—and so it is simply more plausible on balance that such desires help to explain why we like or dislike someone, but are not to be identified with the psychological state itself. Given this and the (we think) plausible claim that liking and disliking one and the same thing can be discordant even when the psychological states are (partially) explained in different ways, it seems we need an autonomous form of affective discordance.
    Note also that liking some things, such as suggestion of dancing tonight and the suggestion of not dancing tonight, is also a case where there is no discordance, though these would plausibly be discordant if this given a propositional reduction. But liking and disliking the suggestion of dancing tonight is clearly discordant—we make the similar point with respect to disapproval in the paper.

38. First, we want to thank everyone for the discussion so far. We’ve learned a lot and had a chance to work through some challenging discussion. So thanks!
    Second, on the Mark’s second point, it’s a problem. As you suggest, the first thing we did when Mike (Ridge) brought up similar cases is to point out that we could allow modality in the context. So the first slate of examples “Nothing exists”, “All beliefs are irrational”, etc. are not a serious problem for us for the reasons you note. They`re clearly pragmatically inconsistent on our taxonomy.
    We then jigged a bit to the left, pointing out that even “Necessarily, belief in p is irrational, but p” (scoping the box narrow) is only pragmatically inconsistent since this could be reasserted coherently by someone who didn`t believe they were in a p world of a world they thought was a p world. They`re mistaken, but it can be coherently asserted.
    Now, the example you mooted came/comes up—and it is a serious problem. What are we to say about “Necessarily, belief in p is irrational, but p” (scoping the box wide) Well, we have a few things we could say, though we are currently unsettled about it. In the spirit of half-assedness pleaded for by David Sobel, here are a few moves we could make.
    (1) We could plead recherché. Such sentences are at the very limits of our account, so if we have to slightly redraw the notoriously fuzzy semantic-pragmatic line, we will not worry too too much. Perhaps it is a bit of a case of spoils to the victor. Like a new pair of shoes, adopting an expressivist view will pinch a bit here and there. Anyways, we think that running together a seriously quasi-realist picture with an expressivist picture is already a mistake. Some of our intuitions might have to go overboard in order to have a satisfying picture of moral discourse (we say this, of course, only while wearing our committed expressivist hats.)
    Note that both of us are antecedently sympathetic to pictures of reasoning, meaning, and representation where this sort of cost is minimized, though we realize that others are working with different pictures. But the point that how much of a cost this is will be theory-relative remains—and we think some of the motivations for expressivism fit naturally with our sort of picture.
    (2) We could go God’s eye point of view/irrationality is a human concept. Since God presumably does not form beliefs on the basis of evidence, notions of irrationality simply do not arise for her. If she counts as a relevant context, then…but what about “Necessarily, belief in p is discordant, but p”…
    (3) We could come up with some bangin’ solution to this problem which currently escapes us. This is something we’re currently fiddling with. Note that we, thank god, only claim that necessarily being discordant when uttered by anyone in any context is a necessary condition on being a semantic inconsistency. This gives us space to wheel in additional features to mark it off as something which is not quite pragmatic—or potentially a different form of pragmatic—inconsistency.

39. I heard from Derek and Jack that the time zone differences and lack of sleep in the past couple of days are taking their toll. So we’ve agreed to call an official end to the discussion here, in the sense that Derek and Jack will probably not check back regularly anymore. The thread will remain open to comments, however, for those who want to keep discussing.
    Many thanks to VOLTRON, to Mark Schroeder, and to all who contributed comments for a really awesome robust discussion!