The Embedding Objection: Part V "The Objection from Pragmatics"

This is the fifth of a series of posts about the different embedding difficulties that, as a family, are thought to present the most pressing objection to expressivism and the different kinds of expressivism toward which each difficulty is most forcefully directed. The first post explained what I take expressivism to be. The second post distinguished four main kinds of expressivism: Simple non-truth-evaluable expressivism (e.g., Ayer's emotivism), Simple minimalist expressivism (e.g., Blackburn's projectivism), Complex minimalist expressivism (e.g., Stevenson's emotivism), and Complex robust expressivism (e.g., Hare's prescriptivism, my Expressive-Assertivism). The third and fourth posts discussed The Objection from Truth Ascriptions and The Challenge from Incomplete Semantics respectively. In this post and the next, I describe a different kind of embedding objection, what I will call "The Objection from Pragmatics." My articulation of this objection is different from any that I have seen before, though it arises from thinking about an embedding difficulty with which many people are probably familiar. From what I have been able to tell so far, The Objection from Pragmatics is the most pressing of all embedding difficulties for expressivism, and I'm not yet sure how even my own kind of Complex robust expressivism can respond adequately to it. It will take a little while to get to the objection, so please bear with me. Perhaps you will have some suggestions for me at the end?

Expressivists claim that ethical sentences like 'Feeding the hungry is the right thing to do' work very much like exclamations or imperatives. That is, their proper and literal use is the performance of either a direct expressive or a direct imperative illocutionary act. However, it is clear that there are a great many complex sentences that can embed ethical sentences, but not exclamations or imperatives. For example, (1) and (2) are "acceptable," "grammatical," "natural," or what have you, but none of (3)-(6) are natural.

1. If feeding the hungry is the right thing to do, then I'll donate to Oxfam at the next opportunity.
2. It is not the case that feeding the hungry is the right thing to do.

3. If terrific for feeding the hungry, then I'll donate to Oxfam at the next opportunity.
4. If feed the hungry, then I'll donate to Oxfam at the next opportunity.
5. It is not the case that terrific for feeding the hungry.
6. It is not the case that feed the hungry.

The objection that usually comes to mind at this point is the following: (i) according to expressivism, the proper, literal use of an ethical sentence is the performance of a direct expressive or a direct directive illocutionary act; (ii) however, sentences that are paradigmatically used to perform these acts--exclamations and imperatives respectively--are not embeddable in some contexts in which it is acceptable to embed ethical sentences; (iii) therefore, expressivists have a hefty burden to explain the key difference between ethical sentences (on the one hand) and exclamations or imperatives (on the other) that allows ethical sentences, but not exclamations or imperatives, to be embeddable in such contexts. Some Simple expressivists have defended their theories from this objection by taking a two-part approach. The first part is to point out the obvious grammatical differences between exclamations and imperatives on the one hand and ethical sentences on the other; ethical sentences are indicative sentences, and only sentences in the indicative mood are embeddable in such contexts. (Simple expressivists then incur an additional burden of explaining why ethical sentences are in the indicative mood, and not in the moods paradigmatically used to perform expressives or directives, but I leave that issue aside.) The second part is to explain the point of having in natural languages complex sentences that embed ethical sentences. For example, Blackburn claims that the point of having complex sentences like (1) in natural languages is to be able to articulate one's commitment to the obtaining of certain relations between one's attitudes (Ruling Passions, pp. 71-75). That is, the point of having sentences like (1) in natural languages is to be able to articulate one's commitment to the obtaining of certain relations between one's approval of feeding the hungry and one's belief that he or she will donate to Oxfam at the next opportunity. If a speaker properly and literally utters (1), then the speaker expresses (i) her commitment to believe that she will donate to Oxfam at the next opportunity were she to come to approve of feeding the hungry, (ii) her commitment to disapprove of feeding the hungry were she to come to disbelieve that she will donate to Oxfam at the next opportunity, and so on. (Gibbard tells a similar story: the point of having sentences like (1) in natural languages is to be able to articulate one's commitments to the obtaining of certain relations between one's decisions (rather than one's approval) and one's beliefs (Thinking How to Live, Ch. 3).) A complex expressivist theory might well take up this approach and claim that the point of having sentences like (1) in natural languages is to be able to articulate one's commitments to the obtaining of certain relations between, on the one hand, one's belief that feeding the hungry is the right thing to do and one's approval towards feeding the hungry, and on the other hand, one's belief that one will donate to Oxfam at the next opportunity.

However, I think this kind of expressivist defense from this objection does not meet its explanatory burden, because the objection is actually more forceful than it first appears.

Notice first that this expressivist response requires both parts of its defense. For the natural reply to the second part of the defense, i.e., to the claim that the point of having sentences like (1) in natural languages is to be able to articulate one's commitment to the obtaining of certain relations between one's approval of something and one's belief that something is the case, is simply: well then why can't exclamations (and imperatives) be antecedents of conditionals? If the point of such constructions is to be able to articulate one's commitment to the obtaining of certain relations between one's approval and one's beliefs, and if exclamations (and imperatives) are paradigmatically used to express one's approval or disapproval, then surely exclamations (and imperatives) should be able to be antecedents of conditionals (and complement clauses of negations, etc). It is at this point that the expressivist can fall back on the first part of her defense: ethical sentences are in the indicative mood, and only sentences in the indicative mood are embeddable as antecedents of conditionals (and as complement clauses of negations, etc). And here we come, I think, to crux of The Objection from Pragmatics: why is it that only indicative sentences are embeddable as antecedents of conditionals (and as complement clauses in negations, and as first disjuncts in disjunctions, etc.)? An initial response might be: "Because those are the only constructions that are grammatically correct." But this answer seems to get matters backwards. Surely, the conventions of a natural language, including its grammar and syntax, develop as a result of human beings' interaction and need for coordination, and not the other way around; it would be odd to think that beneficial human interaction and coordination can be constrained by something like the grammar and syntax of a natural language, which is conventional in origin.

Here's my diagnosis of why exclamations and imperatives cannot appear as antecedents of conditionals (and as complement clauses in negations, etc): given their semantics, there could be no useful conversational purpose for having a grammatical construction in natural languages with exclamations or imperatives as antecedents of conditionals (and as complement clauses of negations, etc), and so we should not expect to find such constructions developing in natural languages (and, hence, such constructions are ungrammatical in natural languages). If this is right, then The Objection from Pragmatics seems to be quite pressing for all expressivists. Here's is my formulation of the objection:

(7) There are particular, systematic, grammatical constructions, such as conditionals with exclamations, etc, that do not appear in natural languages. (empirical observation)
(8) The best explanation for (7) is that, given the semantics of exclamations and imperatives, there could be no useful, conversational purpose for having such constructions in natural languages (otherwise, such constructions would have developed in natural languages).
(9) Therefore, if there is no useful, conversational purpose for having a particular, systematic, grammatical construction in natural languages, then no such construction will develop in natural languages (Generalization of the assumption used in (8))
(10) If expressivism is true, then the semantics of ethical sentences are, in all important respects, like those for exclamations and imperatives. (Implication of expressivism)
(11) If the semantics of ethical sentences are, in all important respects, like those for exclamations and imperatives, then there should be no grammatical constructions, such as conditionals with ethical sentences as antecedents, in natural languages ((7)-(9), argument by analogy of ethical sentences with exclamations or imperatives).
(12) There are constructions, such as conditionals with ethical sentences as antecedents of conditionals, in natural languages. (empirical observation)
(13) Therefore, expressivism is false ((10)-(12))

The strength of this objection rests squarely on the truth of (8) and on what it means for the semantics of ethical sentences to be like those for exclamations or imperatives "in all important respects," as (10) states. In the next post, I'll try to defend (8) in more detail and show that there may be only one important respect in which ethical sentences need be like exclamations or imperatives in order to for this objection to gain traction. I'll then see how well the different kinds of expressivisms can defend themselves from this objection.

Comments

[Michael Cholbi]

Dan- Your post really gets to the crux of the embedding objection, I think. My only idea is this: Ayer (and others too, for all I know) make much of the distinction between an ethical sentence's reporting an attitude and its expressing an attitude. But couldn't expressivists avoid the Objection from Pragmatics by saying that ethical sentences can be embedded and therefore act like truth-functional components of complex expressions because ethical sentences are (re-phrased) reports of attitudes rather than expressions of them? That would look something like this: " If feeding the hungry is the right thing to do, then I'll donate to Oxfam at the next opportunity" amounts to:
If I have a positive pro-attitude toward feeding the hungry, then I predict that I'll donate to Oxfam (where the antecedent is true if I can sincerely report having such an attitude). That's sloppy and ill-formed, I know, but is there a reason for expressivists not to take this route? Would they then not be genuine expressivists? They would still need an account of why ethical sentences operate grammatically as reports instead of as the attitude-expressions that generate those reports. Does that seem at all promising?

[Dan Boisvert]

Hi Michael. Thanks for the comment. Yes, the view you describe would no longer be an expressivist view, at least as I have defined it to include all the major expressivists. Expressivism, as I see it, is the view that (E) is true:

(E): If a speaker utters an ethical sentence and uses all the words in the sentence properly and literally, then the speaker performs a direct illocutionary act whose sincerity condition requires the speaker to have a pro- or con-attitude whose directedness is toward a suitable object of intentionality whose identity is made clear by the conventions governing the use of the sentence or its constituent terms

(I explain this in some detail in the first post on the embedding objection.) The illocutionary acts that fit this bill are expressives and directives. On the view you are suggesting, a proper, literal use of an ethical sentence is an assertive that describes or reports one's attitude, and not an expressive or directive. So, this option does not seem to be open to one who wants to defend expressivism.

[Jamie]

why is it that only indicative sentences are embeddable as antecedents of conditionals (and as complement clauses in negations, and as first disjuncts in disjunctions, etc.

Dan, I'd like to spend a little time getting straight on the data before I make up my mind about the explanation.

First, imperatives can definitely appear as first disjuncts.

Shut up or I'll have to get rough.

Second, what exactly do you mean by the complement clause of a negation?

And third (this is a tricky point that I've been saving up), what exactly is 'the antecedent of a conditional'?

[Dan Boisvert]

Hi Jamie,

First, imperatives can definitely appear as first disjuncts.

Shut up or I'll have to get rough.

Your certainly right. Even exclamations can be first disjuncts in some disjunctions:

Congratulations on winning the race or have I spoken too soon?

When I wrote up these example constructions, I was conflating The Objection from Pragmatics with another embedding objection. Thanks for pointing out the slip.

,Second, what exactly do you mean by the complement clause of a negation?

I just mean complements of 'It is not the case that ______.'

And third (this is a tricky point that I've been saving up), what exactly is 'the antecedent of a conditional'?

Well, I'll give you my exact definition of 'antecedent', but I'm not sure it will help you much, since the definition is a bit theory-laden (it's based on a generalization of a "material" theory of conditionals) and most of the important questions that you are likely to have in mind will probably depend more on what exactly counts as a conditional, rather than on what counts as an antecedent. In any event, here's my definition of 'antecedent':

For any conditional C that is composed, in part, of two sentences, S1 and S2,

S1 (S2) is an antecedent of C iff (C is satisfied (fulfilled, or whatever term we want to use for our most general kind of bivalent evaluation) iff S1 (S2) is unsatisfied (unfulfilled, etc) or S2 (S1) is satisfied (fulfilled, etc))

(I may be guessing a bit about what is driving this question, but let me just say for the record that I would not regard your example above, 'Shut up or I'll have to get rough', as a conditional whose antecedent is 'you don't shut up'. I would take this sentence at face value to be a disjunction and, so, not to have an antecedent. Likewise, I would not regard similar kinds of conjunctions to be conditionals that have antecedents. For example, I would not regard 'Leave now and I won't call the police' to be a conditional whose antecedent is 'you don't leave now'.)

Your general, overall, question is this: what's the data that we are supposed to be working with? From what I can tell, it looks like the data is relatively minimal. It looks like the constructions that are at issue in this objection are those we discussed in a previous post. Indicatives are embeddable in (a)-(e).

(a) If ____, then p
(b) that ____ (including 'it is not the case that')
(c) whether _____
(d) _____ therefore p
(e) _____ because p

However, imperatives are not embeddable in (a)-(d), and exclamations are not embeddable in (a)-(e). There are sure to be more constructions, but I don't think too many more.

[Jamie]

Hm, this is getting complicated.

I think I'll give you my opinion without support right now, and then toss out some support later when I think about which bits can stand independently without too much song-and-dance.

I boldly conjecture that (8) is not likely to be true. I don't think that the best explanation for unembeddability of certain sentence types into certain contexts is that there would be no useful conversational purpopse served by the resulting sentences. I can see the temptation, but I think it's not the best explanation (not because I have a better one, but because there's a problem with the explanation).

The thing is, some of the contexts seem to have equivalents that are friendlier to imperatives. And, there are other kinds of sentences that do not embed into the resistant contexts, but those sentences are equivalent to other sentences that are happier embedding. Let me give just one example. As you say, sentential negation ("It is not the case that S") likes indicatives and not imperatives. But verb negation works for all sorts of sentences (even interrogatives), and verb negation is a near-equivalent of sentence negation. This is not the clearest example, but it is the simplest to state.

I like the "whether" context, by the way. Notice this about "whether": it can also take infinitive complements, and when it does the complement functions semantically like an imperative.

-Tell me whether to draw another card.
-Don't draw another.

What do you make of this?

[Dan Boisvert]

Hi Jamie. There are many things I'd like to say in response to your comment. Let me just stick with a few.

I don't think that the best explanation for unembeddability of certain sentence types into certain contexts is that there would be no useful conversational purpose served by the resulting sentences.

The thing is, some of the contexts seem to have equivalents that are friendlier to imperatives. And, there are other kinds of sentences that do not embed into the resistant contexts, but those sentences are equivalent to other sentences that are happier embedding.

I'm not sure what you have in mind as examples of your first point. For your second point, perhaps you have in mind examples like this:

-Here is John or are my eyes deceiving me?
-If here is not John or are my eyes deceiving me?

(I think I found examples like this in one of your papers, but I'm not sure which one.) At any rate, it looks like you are assuming the following claim:

If two sentences are logically equivalent, then those two sentences are pragmatically equivalent, i.e., the two sentences share all the same conversational or pragmatic characteristics.

This claim seems to be false. For example, the two sentences 'I took my shoes off and went to bed' and 'I went to bed and took my shoes off' are logically equivalent, but they do not share the same pragmatic or conversational features. This shows that there are pragmatic or conversational features that can attach to sentential connectives or to a resulting construction as a whole, and not just from the logic (or even just the semantics) of the sentences embedded within those constructions or the resulting construction as a whole. Conditionals and their logically equivalent disjunctions are also good examples. A conditional is logically equivalent to its corresponding disjunction, but since disjunctions and conditionals serve different purposes in a natural language, the two logically equivalent sentences will not share all the same pragmatic or conversational features. (Of course, this raises the question, what purposes do conditionals and disjunctions serve in natural languages, but I probably shouldn't get into that now.)

Let me give just one example. As you say, sentential negation ("It is not the case that S") likes indicatives and not imperatives. But verb negation works for all sorts of sentences (even interrogatives), and verb negation is a near-equivalent of sentence negation.

I'm afraid I'm about to make things more complicated. I think verb negation is a near-equivalent of sentence negation only for indicative sentences. I take sentence negation to represent the failure of the sentence's positive bivalent evaluation to obtain. (For example, the sentence negation for 'John is here' ('It is not the case that John is here') represents the failure of 'John is here' to be true.) Let's use 'Neg' as our sentential negation sign for a sentence in any mood. The two sentences

-Do not let him borrow your car.
-(Neg)Let him borrow your car.

are not near equivalents, as far as I can see. If the positive bivalent evaluation for imperatives is obedience, then the latter sentence claims that the obedience conditions for the resulting imperative do not obtain. That is, the sentence claims that the hearer did not obey the command. The former sentence commands the hearer to do anything other than let "him" borrow the car. So I don't think these two sentences are near-equivalents. Similar considerations apply, I think, to interrogatives.

Notice this about "whether": it can also take infinitive complements, and when it does the complement functions semantically like an imperative.

-Tell me whether to draw another card.
-Don't draw another.

I'm a bit puzzled here, but I'm puzzled by 'whether' contexts generally, so please bear with me. Certainly 'draw another card' is an imperative. But if the infinitive complement is 'to draw another card', then I'm not seeing how this is functioning like an imperative. Perhaps this raises another question: Why is it that only imperatives, but not indicatives, interrogatives, or exclamations can be embedded within infinitive contexts?

[Michael Cholbi]

Dan- As I suspected, my proposal that ethical sentences are reports of attitudes wouldn't quite meet your definition of expressivism. But it seems to me it runs afoul of the definition only slightly and appears to offer those attracted to expressivism a way of accounting for the 'realistic' surface grammar of moral languge. (I take the various embedding objections to be symptoms of this larger problem.) If I'm understanding your (E) correctly, then (E) would need to be modified in only one way in order for my 'reporting' proposal to fit:

(E): If a speaker utters an ethical sentence and uses all the words in the sentence properly and literally, then the speaker performs an INDIRECT illocutionary act whose sincerity condition requires the speaker to have a pro- or con-attitude whose directedness is toward a suitable object of intentionality whose identity is made clear by the conventions governing the use of the sentence or its constituent terms.

So I guess my question is, what independent reason is there for those who think that ethical sentences *depend* on expressions of attitudes to also subscribe to the claim that ethical sentences *are* expressions of attitudes simpliciter? Are there difficulties I'm not seeing in such a dual semantics for ethical language (an 'objective' grammar of reports underlain by a grammar of attitude expressions)? I take this to be the project pursued by Blackburn's quasi-realism. Are you supposing that such a project is too fraught with difficulties? (I'm asking sincerely; I don't know enough of the critical literature on quasi-realism to be able to make an informed assessment of its prospects.)

Another note: I would think that a view like Gibbard's, appealing to the evolutionary emergence of linguistic norms in tandem with accepted norms, might be able to explain how ethical sentences act like indicatives while still originating in attitude expressions.

[Dan Boisvert]

Hi Michael. I'm afraid I'm now a bit confused at what is supposed to be the view you are describing. Here are two views:

View 1: a proper, literal utterance of an ethical sentence is the performance of a direct assertive that describes the speaker as having a particular kind of pro- or con-attitude

View 2: a proper, literal utterance of an ethical sentence is the performance of a direct expressive illocutionary act and an indirect assertive illocutionary act that describes the speaker as having a particular kind of pro- or con-attitude.

I took your original comment to be describing View 1, which is sometimes called "subjectivism" or "subjective naturalism" or something like that. However, your last comment makes me think you were actually describing View 2.

In any case, let me say something about each of these views. The term 'expressivism' has traditionally been reserved for those metaethical theories that explain the practicality or action-guiding nature of sincere moral judgments in a unique way, namely, by holding that proper, literal utterances of ethical sentences directly do something other than, or in addition to, describing or representing the world—that is, other than or in addition to performing a direct assertive. And the embedding difficulties are thought to arise for expressivist theories because of this other or additional feature. So, the reason embedding difficulties are supposed to arise for Stevenson's emotivism, Ayer's emotivism, Hare's prescriptivism, Gibbard's norm-expressivism, and Blackburn's projectivism, are that they all hold that the proper, literal use of ethical sentences directly express (in the illocutionary act sense of 'express') pro- or con-attitudes (or, in Hare's case, directly direct a certain group of people to do something.) Since View 1 is not a view according to which a proper, literal use of ethical sentences does something directly other than, or in addition to, describing or representing the world in a certain way, View 1 is not an expressivist view, at least as the term has traditionally been used, and View 1 would not be subject any of the embedding difficulties, as far as I can see.

View 2 would certainly be an expressivist view, and would certainly be subject to the different embedding difficulties. The suggestion that a proper, literal use of an ethical sentence is, in addition to being the performance of a direct expressive, also the performance of an indirect assertive that describes the speaker as having a pro- or con-attitude might be one way that an expressivist could get around the various embedding difficulties. In particular, it might be one way they can say what is different about ethical sentences from exclamations or imperatives that allows ethical sentences to be embeddable as antecedents of conditionals and certain other contexts. As far as I know, no expressivist has a view like this, including Blackburn, who holds that we can embed ethical sentences in these grammatical constructions because we have "earned the right" to treat them *as if* they described the world, not that they in fact do (even indirectly) describe the world. Moreover, for Blackburn, what we have earned the right to do is to describe certain actions, people, groups, institutions, etc. as if they had a certain property, and not to describe ourselves as having certain pro- or con-attitudes.

So, if the view you are describing is View 1, then it would not be an expressivist view, at least as the term 'expressivist' has traditionally been used, and the view would not be subject to the various embedding difficulties, at least as far as I can tell. If the view you are describing is View 2, then it is an expressivist view, it is subject to the various embedding difficulties, and might provide a way for an expressivist to get out this embedding difficulty, though I know of no expressivist view like this.