I'm in the midst of writing for a general audience about what I'm calling the 'Asymmetry Challenge' for (pure, noncognitivist) expressivism. The Challenge is to jointly solve the Sentential Mood, Truth-Aptness, and Asymmetric Embedding problems. These three problems are often recognized individually, though I think that jointly solving them is more difficult than many appreciate.
The source of the Asymmetry Challenge lies in the expressivist view that the moral sentences we typically use to express desire-like states or to prescribe behavior have important features that other kinds of sentences lack, even though these latter kinds of sentences paradigmatically express desire-like states or prescribe behavior. Moral sentences, unlike these other kinds of sentences, are declarative, truth-apt, and embeddable into a wider array of complex, linguistic constructions. But what feature could moral sentences have that these others sentences lack—again, especially when moral sentences function so much like these other types of sentences—that warrants this asymmetry? I'm hoping some of you may have suggestions or questions that might help me as I write my way through the Challenge.
English sentences and those of other natural languages are of a variety of kinds:
Declaratives: 'Insulting others is common', 'The coffee is imported', 'The desk is made of mahogany';
Imperatives: 'Do your work', 'Take two asparin', 'Have fun';
Interrogatives: 'Are you doing your work?', 'When are you coming home?';
Exclamatives: 'Hooray!', 'Thank you!', 'Yes!'; and
Optatives: 'How I wish you were here', 'If only I were with you'.
These various sentential constructions are conventional devices that allow us to perform different sorts of acts: declaratives appear to provide a conventional means of representing; imperatives and interrogatives appear to provide a conventional means of directing people to do things; and exclamatives and optatives appear to be conventional devices for expressing certain types of desire-like states. Thus, natural languages like English have grammatical constructions that serve as conventional devices allowing us to serve these various functions.
Expressivism holds that moral sentences provide a conventional means of expressing desire-like states—preferences, decisions, etc.—or of directing people's behavior. Here is Blackburn saying so eloquently:
"That is, it is only through understanding the activities associated with particular linguistic transactions that we understand the words used in conducting them. Amongst the activities involved in ethics are these: valuing, grading, forbidding, permitting, performing resolves, backing off, communicating emotions such as anger or resentment, embarrassment or shame, voicing attitudes such as admiration, or disdain or contempt, or even disgust, querying conduct, pressing attack, warding it off. … When we voice our ethics we have a distinct conversational dynamics. People are badgered. Reproaches are made and rejected. Prescriptions are issued and enforced. Resentments arise and are soothed. Emotions are tugged. The smooth clothing of statements proposed as true or denied as false disguises the living body beneath" (Ruling Passions, p. 51)
But all of this raises three difficult problems for expressivism. The first is to explain why moral sentences are declarative. For expressivists claim that moral sentences are conventional devices that we use to express desire-like states or to direct behavior. And natural languages have perfectly good grammatical constructions that provide a conventional means of performing these activities—exclamatives and imperatives. Thus, one would expect moral sentences to be exclamative or imperative. But of course, moral sentences wear the "smooth clothing of statements"; they are in the declarative mood. Why? Why should natural languages have developed to "disguise" some of our most important human actitvities? As far as I can tell, expressivists have never explained this asymmetry very well. Indeed, I'm unsure whether expressivists have ever explained this asymmetry at all. This is the sentential mood problem. (It sometimes appears that resolving this problem is part of what Blackburn has called the "quasi-realist" project. For example, even recently Blackburn has said that the "functional pluralist still has to confront the smooth propositional surface of ordinary discourse and thought: the fact that, as Wittgenstein said, 'the clothing of our language makes everything alike'" (Practical Tortoise Raising, 2010, Oxford (Oxford U. P.): p. 2). But Blackburn also goes on to describe this project as resolving the Frege-Geach Challenge, never coming back, as far as I can tell, to the sentential mood problem.)
The second problem is to explain why moral sentences are properly evaluable as 'true' or 'false'. For, again, expressivists claim that the primary function of moral sentences is to express desire-like attitudes or to direct behavior. But so also do, respectively, exclamative and imperative sentences, which are not properly evaluable as 'true' or 'false'. What feature could moral sentences have, but exclamative or imperative sentences lack, which could account for such an asymmetry? This is the truth-aptness problem.
The third problem is to explain why moral sentences and nondeclaratives are unequally embeddable in complex, linguistic constructions. For once again, expressivist claim that moral sentences, like exclamatives and imperatives, are conventional devices used to express desire-like states or to prescribe behavior. And since moral, exclamative, and imperative sentences all serve such similar functions, one should expect these kinds of sentences to embed in the same sorts of linguistic constructions. They do not. For example, moral sentences, but not exclamatives or imperatives, are embeddable as:
Antecedents of conditionals:
• If insulting others is wrong, I will refrain from insulting my little brother.
* If down with insulting others, I will refrain from insulting my little brother.
* If let's refrain from insulting others, I will refrain from insulting my little brother.
Complements of attitude verbs:
• I believe that insulting others is wrong.
* I believe that down with insulting others.
* I believe that let's refrain from insulting others.
Within the scope of modal operators:
• It is possible that insulting others is wrong.
* It is possible that down with insulting others.
* It is possible that let's refrain from insulting others.
Within the scope of negations:
• It is not the case that insulting others is wrong.
* It is not the case that down with insulting others.
* It is not the case that let's refrain from insulting others.
What feature could moral sentences have, which exclamative or imperative sentences lack, that could account for such asymmetry, especially when, according to expressivists, they serve such similar functions? This is the asymmetric embedding problem.
So the Asymmetry Challenge for expressivism is to jointly solve the sentential mood, truth-aptness, and asymmetric embedding problems. Doing so is difficult. To get some sense of the pressure this challenge puts on expressivism, consider how tempting it must be for expressivists to respond to the truth-aptness problem by claiming that moral sentences are truth-apt, but not so for similarly functioning exclamatives or imperatives, simply because moral sentences are declarative. But such a response simply forces into the open the sentential mood problem: why are our moral sentences declarative when they function so much like the perfectly useful exclamative and imperative constructions we have at our disposal? Furthermore, the expressivist's answer to the sentential mood problem better not be in this case, on pain of begging the question, that moral sentences are declarative because the declarative mood is the construction reserved for sentences that are truth-apt; for their declarative form was in this case invoked to explain their truth-aptness. (See also Dreier's 'Bob is Hiyo' objection.)
Likewise, consider another tempting response to the asymmetric embedding problem: neither exclamatives nor imperatives can embed in these contexts because the resulting sentences are ungrammatical; only the embedding of declaratives, one might respond, results in grammatical sentences. As before, this response simply forces into the open the sentential mood problem: why are moral sentences declarative when they function so much like exclamative or imperative sentences? But second, it raises the question why these contexts only accept declarative sentences? After all, it is not as if exclamatives and imperatives cannot embed in complex grammatical constructions. For exclamatives and imperatives in fact embed widely into a variety of such constructions, such as:
Consequents of conditionals
• If you have completed Part 1 of the exam, then well done!
• If you have completed Part 1 of the exam, then move on to Part 2.
• Good night and good luck!
• Take two asparin and call me in the morning.
• Well done or have you once again reverted to cheating?
• Wear a warm shirt or bring your sweater.
What feature do declaratives, including moral declaratives, have which permits them, but neither exclamatives nor imperatives, to embed as antecedents of conditionals or complements of attitude verbs and modal and negation operators, especially when moral sentences function so much like exclamatives and imperatives?
The most powerful strategy for meeting the Asymmetry Challenge is to identify a feature that moral sentences have, but that exclamatives and imperatives lack, that can solve the sentential mood, truth-aptness, and asymmetric embedding problem in one feel swoop. Gibbard could be read as deploying this sort of strategy. The strategy requires a certain view about 'true' and 'false', namely, that such words are vehicles for registering our agreement and disagreement respectively. If I evaluate as true the sentence 'Insulting others is common', I am agreeing with the belief expressed by that sentence, the belief that insulting others is common. Likewise, if I evaluate as true the sentence 'Insulting others is wrong', I am agreeing with the decision expressed by that sentence, the decision to avoid insulting others. Three assumptions drive such a view: first, all sentences are vehicles for expressing mental states; second, sentences that are truth-apt are those that express mental states with which it is possible to agree or disagree; third, sentences that express mental states with which it is possible to agree or disagree are those for which the declarative mood is reserved. Thus, although moral sentences share with exclamative and imperative sentences the feature of being conventional devices for expressing desire-like states, they, on Gibbard's view, express decisions, and decisions have in common with belief-like states that with them it is possible to agree or disagree. I can agree or disagree with a decision to refrain from insulting others, just as I can agree or disagree with a belief that insulting others is common. Thus sentences that express decisions, like sentences that express representational beliefs, are evaluable as 'true' or 'false', since we use 'true' and 'false' to register agreement or disagreement; and they are in the declarative mood, because the declarative mood is reserved for sentences that express agree-able or disagree-able states. Although Gibbard says nothing explicitly about the particular complex linguistic contexts driving the asymmetric embedding problem, a reasonable extension of this strategy would suggest that these linguistic contexts require sentences that express agree-able or disagree-able mental states and, hence, that these linguistic contexts require declarative sentences.
I think this strategy fails to jointly solve the sentential mood, truth-aptness, and asymmetric embedding problems. For in whatever sense moral sentences might be thought to express decisions, such decisions are also naturally expressible by imperative sentences. For example, in whatever sense 'Insulting others is wrong' might express the decision that we refrain from insulting others, that decision is also naturally expressible by imperatives, such as 'Let's refrain from insulting others', 'Don't insult others', 'Avoid insulting others', and the like. (Gibbard himself recognizes this.) But imperative sentences are not truth-apt. Furthermore, solving the truth-aptness problem this way fails to solve the sentential mood problem; for if imperatives, like moral sentences, express decisions, why are moral sentences not imperative? Similarly, if imperative equally express decisions, then according to the strategy on offer, imperatives should be embeddable in the problematic linguistic contexts that drive the asymmetric embedding problem, contrary to fact. Thus, this strategy fails to solve the sentential mood, truth-aptness, and asymmetric embedding problems. Expressivism seems unable to jointly solve these, and therefore, fails to rise to the Asymmetry Challenge.