Which is the better theory, David McNaughton's internalist moral realism or some form of non-cognitivism?
On McNaughton's view, values are an objective and observable component of the external world. They are, moreover, intrinsically motivating: if one forms a clear and accurate conception of how the world is, then this fact alone will necessarily motivate one to act in accordance with the objective values. Insofar as one fails to do so, this must be due to their inaccurate conception of the situation.
The non-cognitivist, by contrast, sees value as something we project onto the world. It is not some pre-existing thread in the fabric of the cosmos. Values are invented rather than discovered. Non-cognitivists have traditionally understood our moral attitudes to be closer to desires than beliefs: the point of morality is not to describe the world, but to change it.
A serious problem for simple non-cognitivism is that our language clearly treats moral claims as truth-apt, whereas non-cognitive attitudes cannot be true or false. For some moral claim M, we might say, "It is true that M," or "I believe that M," or "If M, then X". These sentences become nonsensical if M lacks cognitive content, and is merely an emotive expression such as "hurrah for *!". It makes no sense to say that 'Hurrah for *!' is true. The non-cognitivist may respond by adopting a minimalist theory of truth, whereby "'M' is true" is just a long-winded way of affirming "M". But this strategy seems insufficient to account for the more substantial symptoms of propositional content, such as aptness for belief or embedment within conditional (if-then) statements.
It is an obvious fact of moral life that we can be mistaken in our moral beliefs. But for the simple non-cognitivist, there is nothing for us to be mistaken about (and he doesn't grant that moral attitudes are beliefs in any case). He effectively grants us moral infallibility, though at the cost of justificatory nihilism. There is no reason to prefer one moral position over another -- they are all equally non-justified. Moral progress becomes impossible. The moral convert has merely updated their preferences; they weren't mistaken before, and their new set of moral attitudes is no better justified now. This is an implausible and unattractive conception of morality.
However, the non-cognitivist can avoid these problems by moving towards 'quasi-realism'. By imposing certain constraints on what can count as an appropriate moral attitude, the quasi-realist leaves room for moral progress. On this view, there are even moral truths, though they are constructed out of dispositional facts, rather than discovered out there in the world as McNaughton would have it. A sophisticated non-cognitivist could define moral truth in terms of, for example: (1) the hypothetical desires of an "ideal spectator" who is fully informed and impartially sympathetic towards humanity; or (2) the universal prescriptions of a consistent self-interested agent; or else (3) the maximally coherent and unified set of desires (or evaluative beliefs) that would result from a reflective equilibrium process. These options could also lead to reductive ethical naturalism, which externalists (myself included) would prefer. But to retain a non-cognitive aspect to the theory, we must add that moral language (properly understood) also expresses a particular attitude or commitment to the value of human well-being.
This sophisticated version of non-cognitivism co-opts much of the intuitive appeal of moral realism, but without the implausible metaphysics. McNaughton's theory, by contrast, is problematic on both ontological and epistemic grounds.
McNaughton would have us believe that we interact with irreducible moral entities that exist out there in the world. But given the more parsimonious account provided by the non-cognitivist, McNaughton's theory violates Ockham's Razor by multiplying entities beyond necessity. We have no need for such a bloated ontology. Worse, he claims that the objective values are such that their mere apprehension is enough to motivate us; but the thought of entities with such powers strikes us as unacceptably "queer", as Mackie famously objected. McNaughton's theory posits entities of a bizarre kind quite unlike the rest of our ontology. Moreover, it is far from clear how we could come to apprehend or "observe" these values in the first place. Non-natural values do not seem to be the sorts of things that we could causally interact with. The basis of our moral knowledge remains entirely mysterious on McNaughton's account.
Does internalist moral realism offer any advantage over sophisticated non-cognitivism that can make up for these flaws? McNaughton claims that only internalist moral realism can adequately account for "the authority of moral demands". But this is not so. If we adopt the third variety of sophisticated non-cognitivism described above, the immoralist (by definition) fails to possess or act on the maximally unified and coherent desire-set. This leaves them open to rational criticism. This form of constructivist non-cognitivism can understand moral demands to be demands of reason - and you don't get more authoritative than that.
Thus we find McNaughton's internalist moral realism to be quite unmotivated. The intuitive aspects of it can be just as well provided by a sufficiently sophisticated non-cognitivism, and without the metaphysical baggage. Constructivist non-cognitivism is the better theory.