On Friday, Paul Boghossian gave a talk on epistemic rules. (I think it's the same paper that's in pdf form here.) On the way to discussing the Kripkenstein stuff, he claimed that rules are normative propositions and not imperatives. Being a cognitivist, I was generally sympathetic to his view at the time, though I thought some of his arguments didn't do as much to establish his thesis as he thought.
I've been thinking about rules more, though, and I'm seeing some reason to regard them as imperatives. Here are two ways we talk about rules that resemble the way we talk about imperatives.
-We usually don't say that rules are 'true' or 'false', just as we don't say that imperatives are true or false. Propositions, however, are different.
-It's natural to talk about following rules and following imperatives. It's a little less natural to talk about following propositions.
Even if this and other things convince us to regard rules as imperatives, though, I don't know how much it's going to help noncognitivists in ethics and other domains, because these imperatives might be best regarded as derivative from normative propositions anyway. (It'd be something like this -- accepting an imperative commits one to accepting a normative proposition, and the truth of the normative proposition determines the goodness of the imperative.) All the worries about embeddings, etc, are still out there when you're talking about non-truth-evaluable things. So it's open to the cognitivist to just say, "Well, we're going to need something truth-evaluable to explain embeddings and moral reasoning and the intuition that we need some substantial notion of moral truth to ground moral discourse. So assuming that Gibbard and Blackburn and those guys can't deliver all that, we can concede rules to the noncognitivist and still win."
Against what I've written just above, I suppose the noncognitivist could try to turn the tables by explaining the acceptance of normative propositions in terms of the acceptance of imperatives. This just seems kind of weird, though, because normative propositions are going to require truth-makers, and now the noncognitivist is going to have to posit normative properties or tell some kind of story about why we keep holding on to our normative propositions when there aren't any properties out there. So it looks like the noncognitivist can't have anything to do with genuine normative propositions.
In an amusing but irrelevant side note, one of our first-year graduate students has affectionately nicknamed Boghossian "Bog Hoss".