I have started in on a paper in which I show how
expressivists can use a certain kind of semantic theory. The semantic theory, which originated with Kirk
Ludwig, is based on the story Kirk and I tell here about the semantics for
nondeclaratives and, in particular, a semantics for exclamative sentences. So, for example, we provide, in part, a
semantics for ‘Congratulations!’, one that explains what is contributed to more
complex sentences like ‘If you won the race, congratulations!’. (In short, what exclamatives contribute to
the more complex sentence is their sincerity conditions, that is, conditions
that must obtain in order to warrant the application of the predicate ‘is
sincere (relative to a speaker and a time) in English’.) Since most expressivists think ethical sentences work very much like exclamatives, I claim that expressivists can use the theory, with appropriate modifications, of course. My question is not, in particular, about the
theory—I might write something up on that in a future post—but about something more
general, namely, what is at stake if expressivists can provide a logic that is
based, in the first instance, on sentential content or implication rather than
on attitudes. So…
I’ve been writing up this paper, and trying to make a big
deal about how I think it can provide expressivists with a "logic of
content," and so there is no need for expressivists to derive their logic of
content first from a logic of attitudes, as, say, Gibbard does. My problem is that I’m not really sure, after
all, what the big deal is. Suppose I’m
right that this semantic theory can provide expressivists with a logic that is,
in the first instance, a logic of sentential content rather than of
attitudes. So what? What is so important about being able to do
that? Can anyone shed any light for me on what might be at stake in this debate?
(I thought I might list several suggestions here, but I think I’d rather leave the question open for now.)
Dan,
I may have misunderstood your question, so forgive my thick-headedness. But isn’t this the idea: that the ‘logic of attitudes’ view doesn’t deliver at least two things, first, logical inferences between normative claims that are on a par with logical inferences between non-normative claims; and second (and perhaps as a result), the possibility of genuine practical irrationality (say, by deliberating to p and deliberating to something that implies not-p).
Hi Robert. The thick-headedness is mine, believe me. I think what is leading to my confusion is that I’m making the following assumption:
(A) The relation between certain kinds of sentences and the attitudes they are conventionally used to express is so tight that if one has a logic of attitudes, there has to be workable logic of sentential content, and if one has a workable logic of sentential content, there has to be a workable logic of attitudes.
So, I’m imagining someone saying to me: “So you can provide a logic of sentential content. So what? Gibbard has one too; he just derives his from his logic of attitudes. Why should I care so much that you can provide a logic of content *merely* because it is, in the first instance, not a logic of attitudes?”
So, here is one response, which I think is in the background of your comment. If (i) expressivists have no currently workable logic that is, in the first instance, a logic of content, and (ii) expressivists do not have a workable logic of attitudes from which to derive a logic of content, then providing any kind of workable logic that expressivists could use would be important.
Dan,
that’s really interesting. Couple of small points. First, what would you say about the use of exclamatives as antecedents of conditionals? ‘If congratulations!, then you will be happy’ doesn’t seem obviously meaningful in a way ‘If adultery is wrong, then God will punish Clinton’ is. I’m sure you have some story about this.
I’m also slightly skeptical whether sentences can have meaning that is independent of the attitudes they are used to express.
Anyway, I’m not sure how this is related but there is a way in which expressivists could try to get the logic of sentential content for free. I think Horwich has been arguing for this. The way to do this is to go minimalist about the content of the ethical sentences. So, the contents of moral sentences are given by their truth-conditions and the truth-conditions are expressed by the disquotational schemas. If this is right, the expressivist gets to use the ordinary, classic logic as the logic of the sentential content.
Jamie is probably better much in explaining what is wrong with this. But, if I remember the problem correctly, this kind of a view would leave all the important questions unanswered and answering to those questions will get us necessarily to the logic of attitudes. The expressivist has to utter (or write down) the sentences of her theory including the complex disquotational schemas. The deflationist account of the meaning of the sentences will not help to account for the meaning of these linguistic acts – the utterances. That meaning must depend on the attitudes which the sentence is used to express. And, then we get on to the problem that it does not seem like positive attitudes are expressed in the embedded claims. To account for what is going on in the complex claims we then seem to be pushed to the logic of attitudes a la Blackburn and Gibbard.
I would have said :
Due to the fact that there are an infinite number of sentences, meanings must be compositional. In particular, they must embed in various ways: in logical connectives, in propositional attitudes, inside quotations, and so on. It is not clear that expressivists have any kind of compositional semantics, and if you can come up with one, they’ll be very happy.
One constraint on this project is that the embedded propositions must not express what the unembedded propositions express (except in certain cases, like conjunctions, which also need explaining). This may be easier if you talk about contents than about attitudes.
A second constraint on the project is that it has to show why certain propositions have relations of consistency, inconsistency, and entailment. This is easier to do by talking about the truth-conditions of the propositions (which is or is part of the content) than by talking about the attitudes of accepting those propositions but _not_ talking about their content.
Heath,
Did you mean to say this?
One constraint on this project is that the embedded propositions must not express what the unembedded propositions express (except in certain cases, like conjunctions, which also need explaining). This may be easier if you talk about contents than about attitudes.
Normally the antecedents of conditionals had better express what they do when unembedded or you get a fallacy of equivocation. I thought that part of Geach’s point was that if all you said about moral statements was that they express a certain non-cognitive attitude when unembedded you run into a problem explaining why the arguments which then embed them in conditionals don’t equivocate since someone can accept the whole conditional without having the attitude expressed by the antecedent. (I don’t think that Geach thought this the end of the argument, but he was pessimistic about being able to come up with a story about what they meant when embedded that would secure the expected entailments.) So as I read it, the desideratum is precisely that they express the same thing when embedded as when not.
But maybe I’m missing your point.
Dan, by a ‘logic’, do you mean something completely characterized by a consequence relation?
We know what the logic is of normative sentences; well, maybe we don’t, but whatever it is, it’s common to expressivists and realists. Your stuff about sincerity conditions makes me think that you are explaining the logical consequence relation. A (the?) standard way to do this is to find something that the relation will preserve. So maybe you’re going to show that the consequence relation is sincerity-preserving?
That would be helpful, I think, for expressivism.
Jussi:
when the truth conditions in a theory are minimalist, they are not useful in explanations. That’s the short answer.
Mark:
Heath said that embedded sentence don’t express what they express unembedded. They will presumably have the same content. (Ordinary descriptive sentences don’t express our beliefs when they occur in antecedents, since we don’t have to believe the content of the antecedent to sincerly assert the conditional.)
Hi Jussi.
I’m not sure I’m following how we are supposed to be pushed to the logic of attitudes. If we’re talking about Jamie’s “Expressivist Embeddings and Minimalist Truth,” I took Jamie’s point there to be, essentially, about compositionality: in order to have a story about implication, validity, inference, etc, we first need some story about compositionality, whether that story is, first, about the compositionality of complex sentences or, first, about the compositionality of complex attitudes; but a minimalist view like Horwich’s doesn’t give us any story about compositionality and, so, adds little to the debate about the problem of embedding. Maybe I’m missing something?
I do have a story about this, though, unfortunately, it contains too much detail to discuss here on the blog. At bottom, the explanation is pragmatic. You can find a hint of the explanation here. I must say, though, that I don’t have a good explanation to give on behalf of noncogntive expressivists. I think your example points to a significant objection for such views, one that, I think, has yet to be fully explored. If ‘Congratulations, John, for winning the race!’ and ‘Adultery is wrong’ are each used conventionally to directly express only noncognitive attitudes, why wouldn’t Gibbard think that *(1) is perfectly meaningful if he thinks (2) is perfectly meaningful?
*(1) If congratulations, John, for winning the race, then God will punish Clinton.
(2) If adultery is wrong, then God will punish Clinton.
And if *(1) is perfectly meaningful on Gibbard’s account, then why don’t we find such constructions in the language? (Of course, one wants to say, “Because *(1) is ungrammatical, while (2) is grammatical.” But, presumably, our grammatical rules are the result of finding certain kinds of conventions useful, and if constructions like (2) are useful, so too, on Gibbard’s account should *(1) be useful.) Noncognitive expressivists might have something to say, but I’m not sure what it could be.
Hi, Dan.
I think that’s an interesting question. A lot turns on what you think it takes to amount to a ‘logic of content’. As I see it, the reason to think that expressivists are committed to a so-called ‘logic of attitudes’ is that they take mental states to be the semantic values of sentences, and so are forced to treat sentential connectives as operating on mental states. Hence they are committed to a ‘logic of attitudes’ in the sense that the logical properties of sentences must be explained by properties of the attitudes that they express.
I think it’s important to keep in mind the reason why reason why the ‘logic of attitudes’ has been generally less satisfactory (I would say, ‘a failure’) than orthodox semantic views. This is that on existing expressivist views – Blackburn, Gibbard, and Horgan and Timmons, for example – expressivists effectively commit to a vast range of primitive attitudes, whose inferential relationships with one another are taken as primitive, rather than explicable in terms of the logical relationships among their contents.
But this turns out not to be an essential feature of expressivism, even once we fix the expressivist idea that the semantic values of sentences are to be understood as mental states. I’ve shown, for example, how to construct an expressivist semantics for a simple language that does not have these kinds of defects. It still appeals to a ‘logic of attitudes’, in some sense, but it does so in an explicable way, which explains the inferential relations between different attitudes by appeal to ordinary logical relations among the contents of those attitudes.
I think the idea you have is quite different. The things you are calling ‘contents’ – within the Ludwig-Boisvert program – are not the contents of the attitudes that are expressed by sentences; nor are they truth-conditions for the sentence. They are what I believe you call ‘satisfaction conditions’, and there, just as with the formal items that Gibbard’s semantics assigns to sentences, I think everything turns on how we are to interpret the formalism. So it depends on what satisfaction conditions amount to.
I take it that your semantics is going to generate satisfaction conditions for complex sentences by standard sorts of recursion clauses, given satisfaction conditions for atomic sentences. One of the prominent features of your and Ludwig’s account, however, is that it assigns very different kinds of satisfaction conditions to different kinds of sentences. For example, ordinary atomic descriptive sentences get truth-conditions, imperatives get compliance conditions, and exclamatives, on which you’re suggesting modeling moral sentences, get sincerity conditions.
So your semantics can assign ‘if smiling is wrong then I’m a walrus’ a hybrid sort of satisfaction condition – namely, I take it, the set of circumstances in which it is not the case that I am both opposed to smiling and a walrus. This is not a truth-condition for the sentence, because the sentence does not mean, ‘if I am opposed to smiling, then I’m a walrus’. But it is not a sincerity condition for the sentence, either. My sincerity does not turn on whether I am a walrus, but only on whether I believe I am. So I think that one of the problems for understanding the relationship between the view you are suggesting and other expressivist views, is going to be understanding exactly how we are to interpret these kinds of hybrid satisfaction conditions. What is being satisfied? What does it mean for it to be satisfied? I really don’t know.
Suppose, however, that you moved to a semantics on which all sentences are assigned sincerity conditions. Then the project would be to assign the right sincerity condition to ‘if smiling is wrong, then I’m a walrus’. Intuitively, the right answer to that is simply going to be the answer to what it is to think that if smiling is wrong, then I’m a walrus. And that’s what standard expressivist ‘logic of attitudes’ approaches are trying to do. In fact, as I’ve argued, probably the best way for expressivists to understand what ‘express’ means, in their theory, is in terms of a semantics which works by assigning sincerity conditions to sentences, normative and descriptive alike.
(On this score, I think that even if you adopted this move, your semantics would clearly assign the wrong sincerity conditions to complex sentences; this is easy to observe in the case of disjunction.)
Ack! It’s hard to keep up, around here. In response to Heath, Mark, and Jamie: I think Heath is correctly onto one of the desiderata for an expressivist semantics, but whether he has described it correctly depends on what we mean by ‘express’. I take it that the desiderata for what expressivists mean by ‘express’ should be decided, not by how it is used in ordinary English, but by whatever way of understanding it leads to the most promising version of expressivism. And I’ve argued that it should be understood in such a way that normative sentences do express the same attitudes even when embedded. But on other conceptions of the expression relation – for example, on that of Gibbard from Wise Choices – this won’t be correct.
Dan – I think Gibbard (post-2003) can explain the difference between your sentences 1 and 2. His view isn’t just that normative sentences are expressive; it is that they express states of mind that can enter into disagreements with one another. ‘Congratulations, John, for winning the race’ may express some state of mind, but it doesn’t express one that can enter into disagreements, so it isn’t a candidate for his semantic theory.
Jamie, this is very helpful:
Finding something that the relation preserves is exactly what I think I’m doing. Thank you for helping me to get clear on this.
Heath: Thank you. If in fact it is easier to meet the two constraints by talking about contents rather than attitudes, then that would be a positive.
Mark S: Thank you so much for all of that. That was really helpful. I can’t possibly respond to all of it right now, but I’ll try to respond to some of it shortly.
Mark S, I’ll have more to say tomorrow. Here’s one comment for now, although it’s not very self-flattering:
If this is right, then I think I might finally be seeing the source of my confusion; for I’ve never taken expressivists as taking mental states to be the semantic values of sentences. Indeed, for Gibbard’s norm-expressivism, I’ve always taken the semantic value of normative sentences—that is, what they contribute to more complex sentences containing them–to be certain sets of possible worlds. (Which possible worlds is to be explained by the attitudes conventionally expressed by utterances of the normative sentences. This is the sense in which I thought his logic of sentences was derived from his logic of attitudes.) I thought that was the whole point of the “formal” part of his norm-expressivism. I guess I’ll have to go back over that section and look over your paper! Thanks.
Dan,
it’s little daunting to try to remember what Jamie was doing when he might be reading these. I might get this completely wrong but here it goes. I thought one point he made in that paper was that, in principle, you could use deflationist truth to account for compositionality but unfortunately you need complex sentences in order to introduce the deflationist truth. For this reason, if you want the theory to be explanatory at all, you will need some account of these complex sentences already in terms of what attitudes are expressed with them. But, if you have that account, then you can use it solve account for compositionality so that you wouldn’t need deflationist truth to do that anyway.
I’m not so much worried about the grammaticality of *(1) but rather of its having any sense. I know what meaning Gibbardian account would give for (2) but that doesn’t seem applicable to *(1) at least in an obvious way.
The idea of sincerity conditions is interesting. I wonder about contradictions. If ordinary indicative sentences have truth-conditions, then it is easy to say what’s wrong with, for instance, the sentence ‘snow is white and snow is not white’. The logical form of the sentence says that there are no circumstances where the sentence is true.
I assume that the sincerity-conditions of the moral complex claims would be a ‘sincerity-functional’ outcome of the sincerity-conditions of the simple moral sentences in the same way as the truth-conditions of complex claims are a truth-functional outcome of the truth-conditions of the simple non-moral sentences. If that’s right, then the sentence ‘torture is wrong and torture is not wrong’ would have sincerity conditions that were never satisfied.
However, given that we can have conflicting and contradictory attitudes, it would seem that we could sincerely use the contradictory conjunction on some occasions. I’m just starting to wonder whether the sincerity conditions can then do the work of the truth-conditions.
Jamie,
If that’s what Heath meant (as I think he probably did), then I’m on board. For some reason I considered and rejected that interpretation but I cannot any longer see what he said to make me reject it.
Hi, Dan. On Gibbard and semantic values: I think the answer is that it depends on how you think about it. ‘Semantic Value’ is a term of art. True, for Gibbard [1990], sentences get assigned to sets of world-norm pairs. True, for Gibbard [2003], sentences get assigned to sets of hyperdecided thinkers. But in both cases, once we unpack how those sets are to be interpreted, what they really are is a definite description for the mental state expressed by the sentence.
Take, for example, the sentence ‘grass is green’, which gets assigned by the hyperplanner semantics to the set S of hyperplanners who think that grass is green. Why these hyperplanners, and no others? It’s because the ones in S are the hyperplanners with whom you do not disagree, simply in virtue of being in the mental state expressed by ‘grass is green’. So they collectively define a description of that mental state: the mental state, M, whatever it is, such that all and only the hyperplanners in S are the ones who don’t disagree with anyone in M. So the formal objects assigned by Gibbard’s semantics are proxies. Once we interpret them, we see that they’re definite descriptions of mental states. The whole point of semantics, according to expressivism as advocated by Gibbard, is to assign sentences to mental states. (I learned the idea from Gibbard, after all.) That’s why I said that for the expressivist, mental states are the semantic values of sentences.
This is very different from a more orthodox semantic theory, according to which even if there is a correspondence between the formal objects of the theory and mental states, the formal objects of the theory are not to be interpreted as names for or descriptions of mental states.
Hi, Jussi. I think you got it exactly right at the end of your last comment. The truth-conditions for a conjunction are the intersection of the truth-conditions for the conjuncts, but the sincerity conditions for a conjunction can’t be the intersection of the sincerity conditions for the conjuncts. They have to be the conditions in which the speaker is in a different kind of mental state. Saying what that further mental state is, is the difficult part of the expressivists’ semantic project.
Just to chime in a bit on Mark S’s 5/4/07 4:33 pm comment. In even the first Gibbard book you can read him as giving two accounts of the mental states in question. One is the one when he says that you are expressing a complicated sort of approval of norms and so on. But you can instead put the emphasis on the world/norm pair apparatus where he describes the state in question as “ruling out” just those states. And this way of putting it makes it easier to explain how the more complex states of mind expressed by conditionals are supposed to be interpretted.
They are the mental state which “rules out” those sets of combinations of worlds with norms. As Mark says the apparatus is used to specify a state of mind (though obviously we can argue about whether the state of mind in question really is scrutible and really does have the properties proposed by Gibbard).
Everybody is interpreting me very charitably, and correctly.
Best wishes on the paper, Dan!
Jussi,
Right. So, deflationist theories like Horwich’s are not explaining compositionality—we’d already have to have an explanation of compositionality—and, so, they add little to the debate about the problem of embedding. I think we’re in agreement here.
Mark S:
Thanks. I had forgotten about that principled explanation.
Jussi and Mark S: You have a number of very good questions about my theory. I’m going to post something shortly about the theory and try to respond to several of your questions/worries.
Mark S. and Mark vR:
This has been very helpful in clarifying for me a number of things about Gibbard’s expressivist semantics (and logic). I had always thought that (to put things in Mark’s S.’s terminology) the “formal objects assigned by Gibbard’s semantics” for normative sentences would be interpreted as something like (1), not (2).
(1) The possibilities that are ruled out by a speaker S, at a time t, in uttering normative sentence N;
(2) The attitude of ruling out action A in circumstance C by a speaker S, at a time t, in uttering normative sentence N.
Here are some connected thoughts, some of which are lessons that I’ve taken from this discussion. If you have any further thoughts or corrections, expecially about (6), please fire away. (I might also have to make way in (5) for a view like Barker’s.)
(1) Logic is the study of the consequence or “follows from” relation.
(2) A standard way of explaining the consequence relation is in terms of the preservation of something (a content). If some important features of some set of sentences obtains, then that feature obtains—is preserved—in some specific other sentence.
(3) The standard way of explaining the consequence relation is in terms of the preservation of truth. If some set of sentences is true, then some specific other sentence is true. Alternatively (?), if the truth conditions for some set of sentences obtains, then the truth conditions of some other set of sentences obtains. So, on this account, the content of a sentence is specified by its truth conditions.
(4) Ethical sentences stand in logical relations.
(5) Noncognitive expressivists hold that ethical sentences do not have truth conditions. Complex or hybrid expressivists hold that there is more to the content of ethical sentences than just their truth conditions. Therefore, expressivists cannot explain the consequence relation in the standard way.
(6) Expressivists, instead, explain the consequence relation by pointing to the rationality (or irrationality) of holding certain combinations of mental attitudes expressed in uttering ethical sentences, including complex ethical sentences; holding certain attitudes rationally requires/forbids/permits holding certain other attitudes. So, holding or failing to hold certain attitudes is a (rational) consequence or (rationally) follows from holding others. This is the sense in which expressivists provide a “logic of attitudes.”
(7) There appear to be some significant problems with explaining the consequence relation by way of the “logic of attitudes” approach.
(8) Therefore, if expressivists could explain the consequence relation in a standard way—in terms of the preservation of something—that would be a boon to expressivism.
In my previous comment, insert the requisite ‘musts’ in (2) and (3).
Dan,
that’s again very good. I’m not an expert of philosophy of logic so I’m way out of my depth here. But, I’m not altogether easy with the notion that logic is interested in the preservation of something and that for non-moral sentences this something just happens to be truth.
My shaky grasp of this is that our notion of truth plays a much more significant role in grounding logical relations between sentences. For instance, certain features of classic truth like the holding of the principle of bivalence is required for showing that some inferences in classic logic are in fact valid. Otherwise, we get intuitionist logic were the consequence relations are different.
Now, what this has to do with the logic of moral sentences is that on the intuitive level the rules of the logic for those sentences seem to be pretty much the same as the ones for the non-moral sentences. If the features of truth are part of what determines what inferential relations hold between non-moral sentences, then the features of whatever plays the role of truth in the logic of ethics would have to be isomorphic to the features of truth. If truth doesn’t come in degrees, then sincerity shouldn’t come in degrees either, and so on.
So, I guess I might think that there is a further requirement for the expressivist. It’s not only that there needs to be an account for what is preserved but also that it is preserved in the same ways because whatever is preserved shares the logical features of truth.
Hi, Dan. A couple of things. First, on Gibbard. On my understanding, ‘ruling out’ should not be understood in terms of an attitude of ruling out; that would be a constructive answer to what the mental states expressed by complex sentences are like, but I think it would amount to a version of the kind of ‘higher-order attitudes’ approach advocated in Blackburn’s Spreading the Word, which is deeply flawed, as I think Mark van Roojen showed in ‘Expressivism and Irrationality’.
So on my interpretation, by ‘ruling out’, Gibbard just means what he means in Thinking How to Live by ‘disagrees with’. His strategy is to pick out mental states in terms of which other mental states they disagree with.
I think this speaks to the generality of your claim (6), as well. Mark van Roojen also showed, in ‘Expressivism and Irrationality’, that irrationality was too general for expressivists to appeal to, in order to explain what they want to explain. What they need, is the relationship that Gibbard calls ‘disagreement’.
Lastly, I don’t think it’s beyond standard expressivist approaches to be understood as trying to explain the consequence relation in terms of the preservation of something – they are trying to explain the preservation of commitment. This idea isn’t so far out, either; you might think that this is exactly what relevance logicians are trying to explain the preservation of.
That said, I certainly agree in principle with your suggestion about the boon.
Mark S is being too nice to me, both in using my name and ‘showed’ in one sentence and also in not pointing out that I have been known to not make the distiction he does here between the different ways of taking “ruling out”. I have a question though about the contrast and I’m hoping he’ll respond. Mark writes:
On my understanding, ‘ruling out’ should not be understood in terms of an attitude of ruling out; that would be a constructive answer to what the mental states expressed by complex sentences are like, but I think it would amount to a version of the kind of ‘higher-order attitudes’ approach advocated in Blackburn’s Spreading the Word, which is deeply flawed . . .
Mark, you seem to say that it is in virtue of the fact that the attitude is one of ruling out that makes it higher order. But doesn’t that depend on what it is an attitude toward? I was thinking that an attitude which rules out a set of possible worlds is (assuming much else which is at least controversial) one way to characterize the attitude of believing what is true in all members of the set of remaining worlds. Gibbard in WCAF suggests extending that idea by talking of ruling out norms and world-norm pairs, where these are not themselves psychological entities. Whether this tells us enough to understand the relevant attitudes is of course open to debate.
But on this way of thinking the attitude of ruling out does not seem obviously to be an attitude towards an attitude. If the norms are themselves thought of as something psychological we do, of course, wind up with higher order attitudes, but that seems like it follows from taking the objects of the attitudes themselves to be attitudes.
In a way, the second Gibbard book makes that sort of move easier, insofar as plans can be thought of as a set of act-types, or some characterization of acts in circumstances, but they can also be thought of as psychological state types — the state of mind involved in planning to do X. Gibbard himself (on page 56 of THTL) says permitting is ruling out ruling out (and later as disagreeing with disagreement or something similar). That looks like a characterization of a higher order attitude even in the later book.
It looks to me like you are not just complaining about making the attitude into a higher order attitude. You think it as also important that we not take ‘ruling out’ as meaning what we normally mean when we say things like, ‘Dan ruled out posting yet another reply to the thread he started and decided instead to start a new thread.’ Rather you think there is an important more technical sense of ‘rule out’, or ‘disagree’ that we can use to give a definite description of the relevant attitudes in phrases like ‘the attitude which rules out X-ing’ or ‘the attitude which disagrees with a plan to X’. And you think this technical sense is not the more ordinary sense of ‘rule out’ or ‘disagree’.
Am I getting that right, or am I reading too much into your comment?
Hi, Mark.
Hmmm. Let me say what I take to be the most general form of the main problem you raised in ‘Expressivism and Irrationality’ (and maybe you’ll disagree with me), and admit that you’re right that it might not depend on the attitudes being strictly speaking attitudes toward other attitudes.
I took the most general problem to be this. If ‘if murder is wrong then grass is green’ expresses some attitude toward some object, which is itself the kind of attitude that passes all of the right sorts of tests to be expressible by a normative predicate, then we should be able to construct a different sentence which expresses the same attitude, simply by predicating a predicate which expresses it of a term which picks out the object of the attitude.
For example, if for Blackburn the attitude expressed is B!(|B!(M)|&~|Bf(G(g)|), where the parentheses pick out the state of Booing murder and not believing that grass is green, then it can be expressed just as well by the sentence, ‘it is wrong to boo murder and not believe that grass is green’. The problem, then, is that since validity is explained in terms of mental states expressed, such a view can explain the validity of ‘murder is wrong’;’if murder is wrong, then grass is green’::’grass is green’ only by also explaining the validity of ‘murder is wrong’;’it is wrong to boo murder and not believe that grass is green’::’grass is green’. Since the same mental state that can be expressed by the conditional can also be expressed in other ways, any successful explanation is guaranteed to generalize too far.
If that’s the problem, then it doesn’t really hang on whether the objects of the attitude are themselves mental states (though Hyperplanners really do seem like just a device for reifying maximally decided states of mind, and the Wise Choices hybrid between norms, which are really mental states, and possible worlds, which are really their contents, always seemed unstable to me). So suppose that ‘if murder is wrong then grass is green’ expresses an attitude ofruling out certain ways of coming to be fully decided. Express it by RO(___). RO is an attitude that admits of disagreement, so it passes Gibbard’s test for being expressible by a predicate. So the same attitude that is expressed by the conditional could be expressed by ‘planning to blame for murder and not believe that grass is green is RO’. And then you’d have an explanation of the validity of the modus ponens argument only if you also had an explanation of the validity of the one which substitutes the ‘RO’ sentence for the conditional, which is clearly showing too much.
So I take the point that the problem about higher-order attitudes views is more general than ones that are strictly higher-order attitudes. But I do think this is an unsatisfactory way to understand Gibbard’s view. I’ve been taking it that the stuff about disagreement amounts to a further filling out of the notion of ‘ruling out’. That’s part of why I’ve been understanding Gibbard’s view, like Horgan and Timmons’, as non-constructive – as picking out the mental state expressed by complex sentences in terms of the inferential relations that it needs to have in order for the account to work, rather than by spelling out the what the state is actually like, and then showing that it has the right inferential relations. …which is what I think that a successful expressivist semantics would have to do – at any rate, that is the central premise behind the positive stuff in Being For.
I also realize that you had tried to apply your general argument in ‘Expressivism and Irrationality’ to Wise Choices, but that’s the part of your argument in that paper that I disagree with; it looks to me like what does the work in that argument is the details of Gibbard’s account of ‘expression’, not the rest of his view, and it looks like he thought the same thing, based on his response in Thinking How to Live.
Thanks Mark, that’s very helpful. And I don’t disagree with how you characterize the most general form of the problem, nor with your pointing out that it isn’t strictly the higher order nature of the attitudes that causes it. I also see how that motivates the different approach to the sort of thing Gibbard is trying to do.