I have been rereading The Possibility of Altruism, and have been struck by (pace Nagel) the essential incoherence of cognitive judgment internalism, at least on a rationalist construal of cognitivism (e.g. as opposed to naturalism).

On such a cognitive view, the truth of a moral judgment -- e.g. that act A is right -- is determinable by some rational assessment, regardless of one's involvement in A.  That is, whether or not one is in a position to do A, it's rightness is something anyone can determine, analogous to the way one can determine that "5 + 7 = 12" or "P&Q --> P" are true.  It is a rational truth.

But in determining the truth of a moral judgment I have not thereby decided to do anything, even in cases where I am involved.  In this case (shame on me) I may not yet have decided to do what is right.  The argument will be that, insofar I have not decided to do A, I have failed to fully appreciate the rightness of A.  But we have seen that the determination of the rightness of A is a purely rational matter, one that is intrinsically distinct from, and independent of, any person's ability to do A or any person's decision to do A.  Whether I, being in position to do so, indeed choose A is, on the contrary, an empirical matter.

It should be clear that the kind of logical connection required for internalism cannot hold between a rational truth and an empirical truth; neither can imply the other.  However, it may be countered, the connection supposed by judgment internalism only supposes such a connection between e.g. my determining A is right and my determining to choose A.  These are both empirical matters.

Nonetheless, because A's being right and my choosing A are logically separate and distinct, the same holds for their respective determinations.  The process by which one determines the rightness of an act is separate and distinct from the process by which one determines what one will choose.  The first procedure is a rational determination of what ought to be done; the second is an empirical determination of what, in the end, one wants to do.

Phenomenologically, of course, the two can get folded together.  It is at least logically possible that I am such an extraordinary person that the only deliberative issue for me outstanding, in any dispute with others, is what is the right thing to do.  But this does not obviate the fact that such behavior would be based upon a substantive principle, viz. "Always do what is right."  Anyone who indeed knows what is right, and generally wants only to do what is right, can prove to themselves the lack of motivational necessity in that principle by committing a modest sin, e.g. by telling an irrelevant falsehood (with full apologies all round afterwards, of course).

The rational procedures of morality simply need not be dispositive in one's deliberative process.  The rightness of A is a rational matter.  Whether one decides to do what is right -- in this case, or in general -- is an empirical matter.  The idea of an inherent connection between the two is flattering but incoherent.