This paper discusses an important puzzle about the semantics of indicative conditionals and deontic necessity modals (should, ought, …): the Miner Puzzle (Parfit 1988; Kolodny & MacFarlane 2010). Rejecting modus ponens for indicative conditionals, as others propose, seems to solve a version of the puzzle, but it fails to solve a close variant of it. In fact, I prove that the puzzle arises for a variety of sophisticated analyses of the truth-conditions of indicative conditionals. A comprehensive solution requires rethinking the relationship between relevant information (what we know) and practical rankings of possibilities and actions (what to do). I argue that (i) relevant information determines whether considerations of value may be treated as reasons for actions that realize them, (ii) incorporating this normative fact requires a revision of the standard semantics for modals expressing weak deontic necessity (should, ought), but not for modals expressing strong deontic necessity (must, have to), (iii) an off-the-shelf treatment of weak deontic necessity, due in part to von Fintel & Iatridou (2008); Sloman (1970), which distinguishes “basic” and “higher-order” values, and interprets weak deontic necessity modals relative to both, is well-suited to this task. The prominence of normative considerations in our proposal suggests a more general methodological moral: semantic analysis of normative language demands that close attention be paid to the underlying normative phenomena.

In lieu of all that, please enjoy some Joan Baez covering Pete Seeger.

[youtube=http://www.youtube.com/watch?v=Lz_eJqQCCig]

Anyway, if you’re interested, you can find the abstract on my webpage, under the “Talks” heading.

Here is the abstract. I’ll be trying to expand this draft into a more substantial piece in the coming month or two, so comments are especially welcome.

Abstract: We clarify and expand on the connection, suggested by Andrew Alwood in his recent commentary (Analysis, 2010) on Mark Schroeder’s Being For (2008), between recent linguistic work on conventionalized illocutionary force and noncognitivism. We show one way to leverage this work in resolving a version of the embedding problem for noncognitivism. We also argue that it supplies some rather strong pressure for noncognitivists to abandon “pure” varieties of their meta-ethic for a “hybrid” variety. We close with some remarks on how to understand the relevance of imperatives to theorizing about the meaning of normative language.

On an unrelated note, I have also posted a pre-preprint of “Restricting and Embedding Imperatives” (Amsterdam Colloquium 2009, M. Aloni and K. Schulz Eds., pp. 223–233) on my webpage.

This probably makes me seem much more productive than I actually am

I think there’s a pretty wide consensus among linguists on what we might call a specific, inflationary notion of assertion. Generally speaking, inflationary notions of assertion involve an answer to the question what does an assertion that p typically communicate? that goes beyond merely that p. For instance, on one inflationary notion of assertion (Stalnaker’s), if someone asserts that it might by raining, they propose the proposition that it might be raining (a set of possible worlds where it might be raining) for addition to the Common Ground. On a deflationary notion of assertion, the question what does an assertion that p typically communicate? is answered simply by that p. The deflationary notion is, as you’d expect, weaker than the inflationary notion: nonfactualists and factualists about epistemic modality alike will agree that epistemic modals are apt for assertion in the former sense, but may disagree about whether they are apt for assertion, on whatever inflationary notion of assertion is your favorite. (I could be more rigorous about all of this setup, but I won’t.)

Okay, so here’s the sort of appeal to linguists that I’m interested in. (Note: I think there’s something a bit glib about it, although I’m not quite sure why.) If there’s anything like a consensus view about how to model assertion among linguists (I think it’s fair to say there is), it’s something rather close to Stalnaker’s. But actually I can make the argument I want with something weaker. I think there’s a pretty wide consensus among linguists about a certain kind of inflationary notion of assertion: the contents of felicitous assertions typically (modulo certain puzzling kinds of sentences, like potentially informative necessary truths) characterize (although they aren’t necessarily identical to) contingent possible-worlds propositions.

Now consider a nonfactualist about some kind of language K (epistemic modals, normative claims, whatever). According to the nonfactualist about K, a sentence of type K does not have a proposition as its semantic value; it doesn’t function, as a matter of its semantics, to represent the world as being one way or another. So such sentences, if they characterize possible-worlds propositions at all, can characterize only the entire universe (or the empty set). According to the linguists’ consensus, inflationary notion of assertion, something characterizing this sort of possible-worlds proposition is typically not assertable (unless it is one of the puzzling kinds of sentences, which, we’ll suppose language of type K is typically not). We can trust the linguists about which inflationary notion of assertion is best suited to modeling the ordinary exchange of information in conversation. So, for the nonfactualist about K, sentences of kind K are assertable only in a deflationary sense (or some inflationary sense that is not of legitimate empirical interest).

There’s potential for some pretty big philosophical payoff here. This sort of argument seems to rule out, right off the bat, certain forms of nonfactualist expressivism about normative language: for instance, those on which normative and non-normative declaratives express assertions in the very same (inflationary) sense. There’s also a way to leverage these sorts of considerations in constructing a (defeasible) argument against nonfactualism about any kind of declarative sentence. Linguists tend to agree that declarative sentences are conventionally apt for assertion (in, I’d presume, the aforementioned inflationary sense). So we would expect the sorts of declaratives we see people using with regularity in communication to conventionally characterize non-trivial possible-worlds propositions. But it is, at first glance, a bit hard to see how they should be able to do that, for a nonfactualist about any sort of declarative. (This argument’s not strictly valid, but I think there are plausible ways of making it so.)

These arguments, as I said, seem a bit glib to me, although I’m not quite sure why. There’s something just off about this way to deferring to the linguists — I’m just not sure what it is.

Alwood’s suggestion, very roughly, is that we model a noncognitivist account of normative sentence-meaning on our account of imperative sentence-meaning. How is this supposed to help the noncognitivist with the Frege-Geach problem, as Schroeder has developed it? If Paul Portner’s work on imperatives is on the right track, then (i) imperative clauses have conventionalized directive force, (ii) clauses with conventionalized force can embed. In Alwood’s words:

The embedded imperative clause contributes to the meaning of the whole sentence. This inspires optimism that conventionalized force-indicators can meaningfully occur embedded without determining force-potential as they otherwise would.

I’m puzzled by this.

Talk of “conventionalized force-indicators” that “embed” makes it sound like Portner has argued that imperative clauses contain a syntactic element indicating conventionalized directive force. This is not Portner’s view (although Chung-hye Han defended a view like this in her 1998 dissertation). Portner’s view (see “The semantics of imperatives within a theory of clause types” [Proceedings of SALT 14, 2004] and “Imperatives and modals” [NALS, 2007]) is that major clause-types are conventionally associated with distinct kinds of force, but that this is not “written down anywhere in the grammar,” but rather “follows from general principles” linking semantic types to functional potentials. The sense in which imperatives have conventionalized directive force is precisely the sense in which indicative sentences have conventionalized assertoric force: conventionalized force is a function of the type of the semantic denotation of the clause.

If that is right, though, it’s not Portner’s work on imperatives, per se, that gives the noncognitivist a reason to be optimistic. It’s Portner’s work on the conventionalized connection between clause-type and force, in general, that is relevant.

But I have to wonder if there is really any reason for optimism here at all. Non-imperative, normative sentences do not carry any sort of distinctive syntactic marking. Certainly, they do not constitute their own clause-type; they are declaratives. There is an argument to be made (and I make it in my paper) that the noncognitivist is committed to the idea that the conventional force of a normative sentence is not assertoric. But, if a sentence’s clause-type determines the type of its semantic denotation, and the type of its semantic denotation determines its conventional force, then normative sentences have the same conventional force as non-normative declaratives. Yikes.

What’s the noncognitivist to do? I think the best move probably is to deny the idea that clause-type determines semantic type. A suggestive analogy. Nick Asher and Alex Lascarides have argued (“Indirect Speech Acts” [Synthese, 2001]) that the semantic object assigned by the grammar to an interrogative like (1) is distinct in type from the semantic object assigned by the grammar to an interrogative like (2).

(1) Can you pass the salt?

(2) Are you able to pass the salt?

The motivation for this is the idea that (1) has conventionalized directive and interrogative force (for which they offer some pretty compelling evidence in their essay), while (2) has only conventionalized interrogative force. The friendly suggestion to the noncognitivist is that normative sentences are more like (1) than (2): normative sentences have conventionalized non-assertoric (perhaps expressive) force, in addition to conventionalized assertoric force (which they have in virtue of their clause-type).

Of course, this stratagem raises a host of new (and, I think, rather troubling) questions. First, where is the linguistic evidence that normative sentences have conventionalized non-assertoric (perhaps expressive) force? (The evidence that Asher and Lascarides give re: (1) does not seem to me to extend to normative sentences.) Second, if normative sentences have conventionalized assertoric force (in virtue of their clause-type), what is it that they are in the business of asserting?

I don’t know the answers to these questions. That said, it does seem to me that Portner’s work paints a rather more complicated picture for noncognitivism than Alwood is letting on.

I’ll be giving a paper on–again, not surprisingly–imperatives. The basic gist of the paper (certainly not original to me) is that the pragmatics of conditional imperatives is hard because of a misconception about the nature of illocutionary force: that we represent the force of an utterance by applying a force-operator (like assertion) to a force-less content (e.g., a proposition). Illocutionary operators behave more like modal-operators, I suggest. In particular, they can take restriction arguments, and they can “embed” (albeit not freely).

The paper (titled “Restricting and Embedding Imperatives”) is pretty formal, but, if you’re interested, is available on my webpage.

Happy holidays!

It was an outstanding weekend, in no small part because both my brother Simon and good friend Anders were in town. If you’re interested, you can find my comments here (under the “Talks” heading).