Expressivism is a bipartite theory. It holds, first, that ethical sentences lack truth conditions (are not truth apt, truth assessable etc.). Secondly, it holds that ethical sentences do not express beliefs; instead they express, but do not report, certain distinctive pro and con attitudes. (Nor, if it comes to that, do they posit, ascribe or presuppose these attitudes in the wide sense that these sentences are capable of truth but could not realise that capacity unless the attitudes existed.) The second clause is supposed to underpin the first, semantic claim. If ethical sentences expressed beliefs (rather than attitudes), or reported (rather than expressing) attitudes, they would automatically be true or false (setting aside complications due to vagueness that are by the way here). For they would then be true just when the beliefs they expressed were true, or the attitudes they reported obtained. The reason, according to expressivists, that ethical sentences lack truth conditions is precisely that they express attitudes.
Also, the fact that ethical sentences lack truth conditions is seen by expressivists as the key to explaining the famous persistence of moral disagreement. Expressivists argue that if I say that X is right and you say that X is wrong, we might be in genuine moral disagreement, as opposed to talking past each other, even if we were in complete agreement as to the facts, including the facts about people's attitudes (including our own). But then, they argue, `X is right' and `X is wrong' cannot serve the role of making claims about how things are, and thereby cannot have truth conditions, because we, ex hypothesi, would agree on how things are. Indeed, this is arguably expressivists' main argument for their view. It is, for example, the argument with which Gibbard launches Wise Choices, Apt Feelings.4 And it is a principal reason why they insist that their view is an advance on subjectivism. They argue that if the sentences reported our attitudes--or our attitudes after due deliberation, or our attitudes towards our attitudes, or the properties that prompt the relevant attitudes after negotiations with our fellows, or ...--we would not be in disagreement. We would simply be saying that we have different attitudes, or different attitudes after deliberation etc.[5]
Our principal contention is that the doctrine that ethical sentences express attitudes does not deliver a satisfactory explanation of why they lack truth conditions, but we will also argue that, in any case, expressivism does not help with the persistence of moral disagreement problem. Our arguments turn, in the main, on rather general considerations to do with language and not on matters local to ethics. This is the sense in which we agree with Horwich and Geach's conclusion while disagreeing with their paths to it.
Our reasons for holding that expressivists do not offer a satisfactory explanation of why ethical sentences lack truth conditions are the focus of the next few sections. A final section argues for the claim that expressivism's denial of truth conditions to ethical sentences does not help with the problem of moral disagreement.
How might the view that ethical sentences express attitudes deliver a satisfactory account of why ethical sentences lack truth conditions? What is needed is an account of what it is for an ethical sentence to express an attitude which is plausible in itself and which entails that the sentence lacks truth conditions. The basic thrust of the argumentation to come is that it is very hard to fulfil these two conditions together; it is a case of one without the other.
There are many examples of what we might loosely call external signs of various psychological states, where these signs do not have truth conditions. Ayer's emotivist version of expressivism gives the example of expressing boredom. A yawn expresses boredom but is not truth valued. However, yawning is not an intentional act whereas the uttering of an ethical sentence typically is. There are, of course, cases where the claim that something is wrong is `forced out of one' in somewhat the way that a yawn is. But they are the exception, and, as far as their status as intentional actions is concerned, do not appear to differ from shouts of `It's a tiger' when you suddenly realise that it is a tiger that is charging out of the bushes; and such shouts have truth conditions (we hope that they are false). Again, it is sometimes observed that wincing might be described as an expression of pain and yet obviously lacks truth conditions. But wincing is not typically intentional.
There are, though, plausible examples of intentionally delivered external signs of psychological states that lack truth conditions. Expressivists often give as examples words like `Boo' and `Hurrah'. They do not have truth conditions and are not used to make claims about how things are, and yet their production is typically intentional. The suggestion often made is that `X is good' is very roughly like `Hurrah', and `X is bad' is very roughly like `Boo'. In particular, expressivists argue, these sentences are like the relevant exclamations in not having truth conditions and so in not making claims about how things are.
Do we have here the needed model of how an intentionally produced sentence might express, in some good sense, an attitude and yet fail to have truth conditions? The trouble, in brief, is that words like `Boo' and `Hooray' could have had truth conditions, and the conditions under which they would have had truth conditions are those that plausibly apply in the case of ethical sentences. We could, for example, have used `Boo' simply to mean that the producer of `Boo' does not like what they are here and now acquainted with, in which case `Boo' would have had truth conditions; it would have been true in S's mouth at t iff S did not like what S was acquainted with at t. The reason `Boo' does not have these, or indeed any, truth conditions does not lie in the fact that `Boo' lacks the right syntax--it would be strange if syntax alone stopped us making claims about how things are--but rather in the contingent, mutable fact that `Boo' has a rather erratic, highly personal connection to our feelings. The situation is like that with the word `but'. It is standard doctrine that `P and Q' and `P but Q' have the same truth conditions, but it might, of course, have been the case that their truth conditions differ.[6] It might have been that the contrast signalled by `but' was part of the literal meaning of `P but Q', in which case the existence of a contrast would have been a necessary condition for the truth of `P but Q'. The reason it isn't is that the convention of using `but' when there is a contrast is not sufficiently entrenched and clear cut. (And, as you would expect if this is right, the minority who think that `but' does affect truth conditions are those who insist that the convention is sufficiently well-entrenched and clear cut.)
Why does the convention need to be so well-entrenched and clear cut? This is a controversial issue in the philosophy of language, but the essential point is reasonably uncontroversial. When we produce sentences to say how we take things to be, there will typically be a whole web of beliefs lying behind their production. The problem is to isolate the ones that matter for what is literally said, in the sense of determining truth conditions, by using the sentences. For instance, typically, an English speaker only says `Snow is white' when they believe that there is someone around to hear what they are saying, but that is not part of what they intend to convey and is not a necessary condition for what they say being true. It is the nature of the complex set of conventions for the use of the sentence that serves to filter out the `wrong' candidates. We are using the phrase `well-entrenched and clear cut' to mark out the conventions that do the trick.[7]
We can now say why examples like `Boo' and `Hurrah' are of no use to expressivists. Expressivism is not the view that there is only a rough, unclear, insufficiently-entrenched-for-truth aptness connection between ethical words and the attitudes that lead to their production. Expressivists think that ethical terms are well-suited for the task of discussing the attitudes they hold ethical sentences express; if the ethical terms were not well-suited, expressivists would have trouble telling us what their books and articles are about. But this means that they cannot say that the reason ethical sentences lack truth conditions is the same as the reason `Boo' does, namely, that we have not settled the usages precisely enough.
You might, of course, advance a version of expressivism that was a kind of subjectivism-except-that-we-speak-loosely view. This would be an expressivism that holds that all that stops subjectivism being true is that `right' and the other ethical terms are rather like `but', and it would secure the doctrine that ethical sentences lack truth conditions. But then the very moment that you make suitably precise in words the attitudes you wish to discuss and place at the centre of your account of ethics, you automatically give ethical sentences containing them truth conditions. This highlights the difference between this, bizarre style of expressivism, and standard expressivism. Standard expressivists do not hold that once they have made clear in words which attitudes are central in ethical theory on their view, ethical sentences thereby acquire truth conditions. They do not hold that they can only be precise at the cost of refuting themselves.
There is a general difficulty for expressivists here. The difficulty is not merely that there is a crucial disanalogy between words like `boo' and `but', on the one hand, and `good' and `right', on the other. Words and phrases like `good', `ought to be done', `just', and `morally bad' are conventional signs which we learnt when we learnt English, and there is a serious problem for holding that certain words and expressions are conventional signs and yet generate sentences that lack truth conditions, a problem which goes back to Locke's observations about voluntary signs.[8]
Locke observes that, because it is contingent and fundamentally arbitrary that we use the words we do for the things we do use them for, our ending up with the conventions or arrangements we have in fact ended up with is to be understood in terms of our, explicitly or implicitly, entering into agreements for the use of these words for these things.[9] However, entering such agreements requires that we know what it is that we are using the words for. As Locke puts it, `Words being voluntary signs, they cannot be voluntary signs imposed by him on things he knows not. ... they would be signs of he knows not what, which is in truth to be the signs of nothing.'[10] In other words, because the word `square' is a voluntary, agreed-on conventional sign of the property we use it for, the explanation of how it is that we come to use it for that property requires that, on certain occasions, we take something to be square, and use this fact to found the convention of using the word `square' for that property.[11] In like fashion, any explanation of how we English speakers came to use the voluntary sign `good' for the attitude we do use it to express, according to expressivists, must allow that we recognise the attitude in question in us. For, to follow Locke, we could hardly have agreed to use the word for an attitude we did not recognise and failed to believe we had, for that would be to use the word for `we know not what'. But that is to say that expressivists must allow that we use the word when we believe that we have a certain kind of attitude. And that, in turn, is precisely to concede truth conditions to `That is good'; it will be true when the corresponding belief is true. Here is a way to put the point. The attitudes that expressivists hold are expressed by the use of ethical terms in sentences had better be attitudes we recognise as those it is correct, according to the conventions of English (or French or ...), to use the ethical words for, and it had better be that we use the ethical words just when we believe that we have the attitudes. Would it be better, according to expressivists, to use them when we thought we had quite different attitudes, or had no opinion at all on the matter? How then, runs our challenge, can our uses of the ethical terms in sentences like `That is right' fail to be reports of those attitudes? The same goes for the word `boo', of course. If, as a minority hold, the convention governing the use of `boo' is `good' enough for it be correct according to the conventions of English to use `boo' just when we believe that we do not approve of what we are booing and seek to convey this, `boo' would have truth conditions. Locke's point means that expressivists face a dilemma: if the conventions governing `boo' are enough like those governing our use of ethical terms, `boo' has truth conditions, but then it is useless as a model for understanding their view for that very reason; if the conventions aren't enough alike, `boo' lacks truth conditions, but then it is useless as model by being disanalogous in a crucial respect.
The view that ethical terms serve to report attitudes is compatible with the view that they express the attitudes. Indeed, a good way to express an attitude is to report that you have it. It would be absurd to say to someone who says `I dislike George a good deal' that they failed to express their dislike because they reported it. But if `That is right' and the like are reports as well as expressions of the attitudes, then they have truth conditions, contrary to the central claim of expressivists. Interestingly, Ayer half-sees the crucial difficulty in his discussion of a `complication' concerning the crucial distinction between expressing and asserting.
The distinction between the expression of feeling [his term for the relevant attitudes] and the assertion of feeling is complicated by the fact that the assertion that one has a certain feeling often accompanies the expression of that feeling, and is then, indeed, a factor in the expression of that feeling. Thus I may simultaneously express boredom and say that I am bored, and in that case my utterance of the words, "I am bored", is one of the circumstances which make it true to say that I am expressing or evincing boredom. But I can express boredom without actually saying that I am bored. I can express it by my tone and gestures, ... , or by an ejaculation, or without uttering any words at all.[12]
Ayer sees that saying that one has a certain attitude counts as expressing it but observes, rightly, that there are ways of expressing an attitude that do not involve saying that you have it. But his examples are ejaculations and gestures, and cases that do not involve `uttering any words at all'. This hardly helps with examples like the sentences `This is good' and `That is wrong', which do involve words, are not typically ejaculations, and, moreover, are part of the voluntary convention we English speakers entered into concerning what words stood for what when we mastered the language.
Blackburn also notes that one way of expressing an attitude is to say that one has it.[13] This is in the course of explaining that his view is not that expressing attitudes, on the one hand, and reporting them, describing them or saying that you have them, on the other, are incompatible; it is simply that you can have the former without the latter. We agree. The problem is that the cases where you can have the former without the latter differ in the crucial respect from the way we use ethical words and sentences.
In discussion, five objections have been pressed by expressivists at this point. Some expressivists have objected that the production of a sentence like `That is wrong' is much more like an exclamation--or ejaculation, as Ayer puts it--than we are allowing. They argue that we contemplate some situation and simply find the words `That is wrong', as it might be, coming to our lips. The attitude `outs' itself without any conscious cognitive processing as such, and so, they argue, there is no truth assessable claim being made when we produce the sentence. What happens is almost like wincing.
This seems to us a highly implausible account of our moral psychology. It is a commonplace that we argue about and reflect on which ethical judgements to make. We do not typically `give out' with the first thing that pops into our head. This indeed is a point that Blackburn is explicit about when he talks about `taking all possible opportunities for improvement of attitude'.[14] Moreover, even if some relatively simple kind of `outing' story were true, this would not support the view that ethical sentences lack truth conditions. Face recognition is mostly not a conscious cognitive process, which is why verbal descriptions of suspects are often of little use to the police. It is a process of recognition that issues forth in judgements like `That's Harry Morgan' in ways which we folk do not properly understand (maybe some cognitive scientists know the underlying mechanism, but it is not a piece of folk knowledge). Nevertheless, such judgements have truth conditions--obviously. In the same way, even if ethical sentences are prompted by an ability to recognise certain attitudes in a way not amenable to rational reconstruction and debate at the folk level, they would still have truth conditions; they would simply be claims about the recognised attitudes.
The second objection is much more concessive. It allows that belief about the nature of our attitudes must accompany the production of ethical sentences--how else could we be following the conventions for the use of the ethical vocabulary if we do not have the relevant beliefs? It grants our point that expressivists should not hold that we are ostriches when we give voice to our attitudes. It insists, though, that the sentences do not express beliefs about the attitudes. We have the beliefs, that is what saves us from being ostriches, but we are not expressing them.
This is hard to believe. When we follow the conventions for the use of the ethical vocabulary, we seek to conform to them, and know, or at least very much hope, that those we are communicating with are seeking to conform to them. We regulate our linguistic behaviour under the presumption that our fellows are doing the likewise as part of our membership of a common linguistic community. But this is precisely to be guided by belief and thereby to express it; it is much more than a matter of being accompanied (by chance?) by belief.
The third, fourth and fifth objections are that there must be something wrong with our argument from Locke--because (i) we know that there is a distinction between expressing and reporting belief, and, it is objected, expressivists can simply apply this distinction to attitudes to get what they need; because (ii) imperatives lack truth conditions, yet are the product of voluntary agreements; and because (iii) indicative conditionals lack truth conditions, yet are the products of voluntary agreements. These three objections call for separate sections.
Expressivists often suggest that we can apply this distinction to other psychological states, including especially the `ethical' attitudes, and that, when we do, we get the account they need of the sense in which ethical sentences express attitudes. They observe that we can distinguish the doctrine that `X is right' reports a certain pro-attitude to X, from the doctrine that it expresses that pro-attitude to X. The first view is subjectivism; the second, they claim, is expressivism.[15]
How does this help with the task of finding a suitable sense of `express' which ensures that `X is right', and the like, lack truth conditions? That is, how does it serve to elucidate a sense of `express' such that if `X is right' expresses an attitude instead of reporting one, it follows that `X is right' lacks truth conditions? Sometimes, it almost seems as if the argument is supposed to run as follows: if `X is right' reports an attitude, `X is right' has truth conditions; therefore, if it expresses an attitude instead, it lacks truth conditions. But this would, of course, be to commit a fallacy (a version of denying the antecedent). What we need to do, rather, is ask about the sense in which `Snow is white' expresses a belief, apply this sense to the claim that `X is right' expresses an attitude, and see what we get.
The trouble for expressivists is that, although there is an important difference between expressing and reporting belief, it is plausibly a difference in what is reported. It is not a difference between reporting something and not reporting at all. When you express your belief that snow is white by producing, in the right context, the sentence `Snow is white', you are reporting the content of what you believe. This is how the sentence gets to be true iff the belief is true. Therefore, if we take the distinction as drawn for beliefs and apply it to attitudes, we get that `X is right' expresses a certain pro-attitude iff `X is right' reports the content of the attitude. And this is not at all what expressivists are after. First, it makes `X is right' out to have truth conditions, namely, those of the content; and, secondly, it is very implausible in itself. The relevant content will something like that X happens--for that is what we are favourably disposed towards, according to expressisivists, when we assert that X is right--and that is very different from X being right (unfortunately).
A third way--in addition to the appeal to exclamations and to the distinction between expressing and reporting beliefs which we have just discussed--that expressivists often put their view is by saying that ethical sentences belong to the general category of commands and recommendations.[16] This suggests that they might elucidate the needed sense of expressing by saying that it is like the relation between the desires and attitudes that lead to an order or a recommendation: expressisivists might say that my `ethical' pro-attitude to X stands to `X is right' as my desire that the door is shut stands to the order `Shut the door'. And, as orders and recommendations lack truth conditions, we now have, expressivists might well suggest, an account of the sense in which ethical sentences express attitudes which ensures that ethical sentences lack truth conditions.
The trouble for this suggestion is that it is very plausible that orders (and recommendations, but we'll frame the points that follow in terms of orders) have truth conditions in the relevant sense. It is not to the point that it is `crook' English to talk of orders as true or false. For, first, our topic is not what is or is not good English, and, second, if it were, we could close the debate with the observation that it is good English to talk of ethical sentences as true or false. The issue about truth conditions, as it bears on our discussion, concerns what we are doing when we produce the words `Shut the door' in the right circumstances (production in a play or a philosophy of language tutorial doesn't count). But any competent English speaker knows that producing the words `Shut the door' in the right circumstances is ipso facto to command that the door be shut; that's what is being done. But that is to say that it makes no difference whether I say `Shut the door' or I say `I command that the door be shut'--a point which is independently plausible. However, the latter obviously has truth conditions. It follows that `Shut the door' is true in S's mouth at t just if S did indeed command at t that the door be shut.
We can reach the same conclusion via Locke's point. Ordering is something we learn to do, and something others recognise us as doing through our shared mastery of the voluntary conventions that settle when we are, and when we are not, ordering. The possibility of this rests on our knowing when we are ordering. If the expressivist suggestion under discussion in this section is correct--that to use ethical words is ipso facto to order--this must be because we know that this is the case. It had better not be a dark secret that this is what we are doing. But then expressivists must hold that we produce ethical words when we know that, by doing so, we are ordering. As we said a moment ago, to produce the words `Shut the door' in the right circumstances is to order that the door be shut, and, because it is a matter of fact whether or not it is ordered that the door be shut, this means that `Shut the door' is true just if it is ordered that the door be shut. All this makes good intuitive sense. Surely a good way to order is to say that you are--if someone asks you to tell them what you are you doing when you are ordering some course of action, a good answer is to say that that is precisely what you are doing. It makes no difference whether I say `Shut the door' or `I order you to shut the door'.
Of course, producing the sentence `Shut the door'--and, equally, the sentence `I order you to shut the door'--in the right circumstances is to order you to shut the door; the sentences are performatives. But, as many have observed against J.L. Austin, this does not stop them having truth conditions; instead, it makes it easy to make them true.[17]
There is, of course, a major point of difference between the two sentences--`Shut the door' and `I order you to shut the door'. Inserting them in the antecedents of conditionals produces very different results. `If I order you to shut the door, then I order you to do something' is fine, but `If shut the door, then I order you to do something' is nonsense. But this does not tell us that `I order you to shut the door' is not a performative when it appears unembedded; it tells us that, unlike `Shut the door', it ceases to be a performative when it appears in the antecedent of a conditional.
Many have argued that indicative conditionals lack truth conditions. Be they right or wrong, it would be too quick a refutation of their position simply to cite Locke. It would, that is, be too quick to point out that indicative conditionals are sentences we produce as a result of our learned mastery of a natural language, and then conclude without further ado that they have truth conditions on the ground that they must, being convention-governed, express beliefs.[18] However, it would not be too quick to infer that indicative conditionals serve to express, in some general sense, a speaker's epistemic state. That does indeed follow from Locke's point. When we learn to use the indicative conditional construction, we enter a voluntary, implicit agreement to use the construction; and, unless this agreement is to do `we know not what', it must be to use the construction when our epistemic state is thus and so.
As many have observed, the (surprising) peculiarity about indicative conditionals is that the epistemic state that underlies their production seems to be a ratio of credences rather than the credence in any given proposition as such: what matters for the correctness of producing `If A then B' is the ratio of the credence of the conjunction of A and B to the credence of A.[19] What is important for us is that the peculiarity does not consist in their production being underwritten by something that is not epistemic, something that is not fully determined by facts about credences of speakers.
This means that the most an expressivist could extract from the example of indicative conditionals is the possibility that ethical sentences express a subject's epistemic state, but do so without expressing the subject's epistemic state with respect to any particular proposition. And we have said nothing to rule out this possibility, though we would insist that the onus of proof does not lie with us. Moreover, a position of this kind would be contrary to the expressivist's view that ethical sentences express attitudes, non-cognitive ones, that is, of some special kind. One way to bring out the contrast is to note that a central part of expressivism is precisely the claim that the justification for asserting an ethical sentence does not supervene on the speaker's epistemic state, whereas it is the case that the justification for asserting an indicative conditional supervenes on the speaker's epistemic state.
The argument from the persistence of moral disagreement starts from the observation that two claims are independently plausible. The first is that if you utter the words `X is right' and I utter the words `X is wrong', we are in disagreement. It is not like my saying that I like X, and your saying that you dislike X. The second is that this disagreement might survive our agreeing about all the relevant facts concerning X, including all facts about psychological responses to X. An example might be the disagreement between us if one of us is a total utilitarian and the other an average utilitarian, in a case where the difference matters for whether or not X is the right thing to do. We might agree about the effects of X on, say, happiness; we might agree that only effects on happiness matter for rightness and wrongness; and we might know all about our difference over whether it is the effect on the average or the effect on the total happiness that settles what is right. Still, it seems, we could be in genuine disagreement--it seems, for example, wrong to infer that you and I must be giving the word `right' a slightly different meaning. But then, it follows, urge expressivists, that `X is right' and `X is wrong' do not make claims about the facts, about how things are, for we agree completely about how things are. But if they do not make claims about how things are, they lack truth conditions and must be construed along expressivist lines.
There is a huge literature devoted to this argument, and there is no doubting the interest of the issues it raises. Our concern here, though, is simply to point out that it is no kind of argument for the view that ethical sentences lack truth conditions, and no kind of argument for expressivism.
The key question that needs to be addressed is what is meant by disagreement in the argument. Often, what is meant when it is observed that if I produce S and you produce T, we are in disagreement is that S and T cannot be true together. This, though, cannot be what expressivists mean, because, for them, all ethical sentences cannot be true together as they cannot be true to start with.
However, if you say `I believe that Chicago will win the championship' and I say `I believe that Boston will win the championship', we count as disagreeing, despite the fact that the sentences we produce can be, and most likely are, true together. But if a difference in belief can count as a disagreement, why not a difference in certain sorts of attitude? So, expressivists can fairly count your assertion that X is right and mine that X is wrong as disagreement by virtue of the difference in our moral attitudes. And this, of course, is what they mostly do.[21] But if disagreement in moral attitude counts as disagreement, then the persistence of moral disagreement is no reason to favour expressivism over a subjectivism that holds that our assertions report our moral attitudes. The difference between expressivists and subjectivists over the semantics of ethical sentences is neither here nor there as far as our having different moral attitudes go. Indeed, almost every party to the debate in meta-ethics believes that if I sincerely assert that X is right and you sincerely assert that X is wrong, we must have different moral attitudes; so, if that counts as our disagreeing, as expressivists who are not eliminativists about moral disagreement must allow, almost every party to the meta-ethical debate can respond to the problem of moral disagreement simply by noting that a difference in moral attitudes can survive agreement over all the facts.
We think that the result that expressivism does not help with the problem of moral disagreement is no surprise. Moral disagreement, and, indeed, disagreement in general, is a psychological phenomenon. The production of sentences makes public our disagreements; it does not create them. Your disagreement with Hitler's moral outlook did not come into existence when you remarked on it in language. This means that a thesis whose essential distinction from other theses lies in a certain claim about the semantics of sentences cannot be in a privileged position when it comes to giving a good account of moral disagreement.
Frank Jackson (who is responsible for this draft) Philip Pettit Research School of Social Sciences [1] Paul Horwich, `Gibbard's Theory of Norms', Philosophy and Public Affairs, 22, 1, (1993): 67-78.
[2] P. T. Geach, `Assertion', reprinted in Logic Matters, Oxford: Basil Blackwell, 1972, pp. 254-69.
[3] See Frank Jackson, Michael Smith and Graham Oppy, `Minimalism and Truth Aptness', Mind, 103, 411 (1994): 287-302, for arguments against Horwich; and Frank Jackson, `Inference for Noncognitivists' in Dale Jamieson, ed., Singer and His Critics, Oxford: Basil Blackwell, forthcoming, for arguments against Geach. In a sentence, one problem is that the example of indicative conditionals tells us that we cannot rule out a priori the possibility that meaningful indicative sentences might lack truth conditions and might figure in valid inferences in some non-standard but perfectly acceptable sense of `valid'. We grant, of course, the prima facie plausibility of the view that ethical sentences have truth conditions.
4 See chapter 1, esp. pp. 11-18; see also Blackburn, Spreading the Word, p. 168, and chapter 1 of C.L. Stevenson, Ethics and Language, New Haven: Yale University Press, 1944.
[5] We will, in the main, simplify and frame our points in terms of simple versions of subjectivism and expressivism and omit the (here irrelevant) complications arising from versions that focus on attitudes after deliberation, second-order attitudes, communally-agreed attitudes, properties qua typical prompters of the attitudes, and so on.
[6] For one exposition of the standard doctrine, see Michael Dummett, Frege, London: Duckworth, 1973, p. 2. The account that follows shortly of why `but' signals a contrast without altering truth conditions is essentially the one Dummett offers on p. 86. Incidentally, as he points out, the contrast signalled need not be between `P' and `Q'.
[7] For detailed accounts, see H.P. Grice, `Meaning', Philosophical Review, 66 (1957): 377-88, David Lewis, Convention, Cambridge, Mass.: Harvard University Press, 1969, and Jonathan Bennett, Linguistic Behaviour, Cambridge: Cambridge University Press, 1976.
[8] John Locke, An Essay Concerning Human Understanding, book III, chapter 2.
[9] For more on this, see, e.g., Grice, `Meaning', Lewis, Convention, and Bennett, Linguistic Behaviour.
[10] Essay, book III, chapter 2, SS2. Our emphasis.
[11] Or, maybe, the word starts out meaning that there is something square in front of one. The issue of whether we start with words or (one word) sentences does not affect the points we make here; the key point for us is simply that we perforce start out with belief. According to Lewis, Convention, in order to found the convention, we need much more than the belief that something is square; we need a whole web of beliefs about beliefs including those of our fellow speakers of the language. For our purposes here, the crucial point is that we need at least to regulate our use of the word via belief.
[12] Language, Truth and Logic, p.109.
[13] Blackburn, Spreading the Word, pp. 169-70. In his view, you can, in addition, express attitudes by making claims about how things are other than the claim that you have the attitude in question.
[14] Blackburn, Spreading the Word, p. 198, our emphasis.
[15] See, e.g., Blackburn, Spreading the Word, p. 169, and Gibbard, Wise Choices, Apt Feelings, p. 84.
[16] See, e.g., Stevenson, Ethics and Language. Incidentally, it might be argued that the reason `boo' and `hurrah' lack truth conditions is not that the conventions governing their use are not `good' enough, but rather that they belong with commands and recommendations. In this case, we would not have different models for understanding the expressivist view, but the same model approached in slightly different ways.
[17] See esp. David Lewis, `General Semantics', reprinted in Philosophical Papers, Vol. I, New York: Oxford University Press, 1983. The view that a sentence cannot both be a performative and have truth conditions is prominent in J.L. Austin, How to do Things with Words, Oxford: Clarendon Press, 1962, see esp. Lecture I. Interestingly, Austin thinks that holding that performatives are not true or false `needs argument no more than that "damn" is not true or false' (p. 6). But, as we have seen, `damn' lacks truth conditions for reasons local to certain exclamations.
[18] For more on why this would be too quick, see Frank Jackson, `II--Naturalism and the Fate of the M-Worlds', Proceedings of the Aristotelian Society, Supp. vol. 71 (1997): 269-82.
[19] See, e.g., Ernest Adams, The Logic of Conditionals, Dordrecht: Reidel, 1975.
[20] Simon Blackburn has especially urged this on us.
[21] Ayer is a notable exception. In Language, Truth and Logic, around p. 110, he argues that we should be eliminativists about ethical disagreement.