At the Wisconsin Metaethics conference, Michael Smith presented a question that I hadn't considered before, and which I'll think about more in the future: Which is more fundamental, deontic predicates or evaluative predicates? Can all of one be defined in terms of the other? I don't know what the best way to put the deontic/evaluative distinction is, but the notions of obligation, warrant, and "the thing to do" (a nice expression I hear a lot at Michigan) are on the deontic side. Evaluative notions are more likely to attach to the goodness of a state of affairs than to guide an agent in belief, action, or emotion. As an externalist and realist about the goodness of pleasure, I'm employing an evaluative notion that I regard as irreducible to deontic terms. Smith also thinks that evaluative terms are fundamental, though he thinks that there are a lot more reductive steps before the evaluative comes out as the winner. Let's see if I can remember how this goes: He'd want the goodness of pleasure to be analyzable in deontic terms -- we all have normative reasons to generate pleasure. But this in turn is analyzable in evaluative terms -- being motivated to generate pleasure is good. There's one more deontic step -- one ought to make oneself into the type of person who is motivated to generate pleasure. The clinching evaluative step is that being someone who wants to make himself into the kind of person who is motivated to generate pleasure is good. This fits nicely with part of his view in The Moral Problem, but I'm dubious about whether it works. I don't want to grant him all these reductive moves, but someone on the deontic team who granted these steps might be able to push for more. In any case, it gives a good view of how the deontic/evaluative game could be played.
David Sosa wrote a paper called "Consequences of Consequentialism" that tried to capture a lot of deontic talk in terms of the goodness of states of affairs. It was quite good, and I wonder if Smith has looked at that.