Print This Post Print This Post
Realism and Semantics
By Guy Kahane

J. L. Mackie did a great service to metaethics by distinguishing, as previous philosophers hadn’t, between semantics and metaphysics.* He pointed out that it’s one thing to show that our normative concepts refer to objective properties, quite another to show that anything out there actually corresponds to these concepts. Defending normative realism therefore turns out to be harder than previously thought: winning the argument about the semantics only takes you halfway.

By drawing this distinction, Mackie was famously able to draw attention to (and defend) a hitherto overlooked metaethical position: the error theory. On this view, the realist is granted the semantics but not the metaphysics: our moral and evaluative concepts do aspire to refer to ‘objective prescriptions’, but there is nothing out there that corresponds to what we are talking about when we make moral and evaluative claims.

Mackie’s moral semantics wasn’t simply cognitivist; it is perhaps more accurately described as ‘non-naturalist objectivist cognitivist moral semantics’. After all, subjectivists can also be cognitivist, and non-cognitivist can be objectivists. And if the correct moral semantics was cognitivist, objectivist and naturalist, Mackie’s own arguments for the error theory clearly wouldn’t go through — which is not to say that Mackie’s arguments are ultimately successful, or that such a semantics is incompatible with the error theory. In fact, strictly speaking, there is a sense in which the error theory might be compatible even with subjectivism and non-cognitivism, a point that Mackie overlooked.

I’ve always thought that Mackie and some of those who have followed in his footsteps have stopped short of fully appreciating just how far the error theory might take us. Mackie thought that the truth of the error theory may not really change much for first-order ethics because we have good pragmatic reasons to hold on to morality. But for this move to make sense, the ‘non-naturalist objectivist cognitivist moral semantics’, and the consequent error-theory, have to be partial. They must apply only to a specific sector of normative discourse, say, to morality, leaving intact some other, genuine source of practical norms. But for those of us who think that normative discourse as a whole presupposes ‘objective prescriptivity’, this way out is simply not available.

Mackie’s distinction expands the logical space of meta-ethics. The error theory is one new position that it makes available. In the rest of this post I want to explore another sector of the logical space opened up by the semantics/metaphysics distinction. What if things turned out to be exactly the reverse of Mackie’s view? That is, what if realists are right about the metaphysics but wrong about the semantics? This must be at least possible, if Mackie’s distinction makes good sense. It should be possible that our normative concepts are, say, subjectivist in nature, but that nevertheless there are objective values and norms out there. There is a certain difficulty in even stating this possibility because if our semantics is subjectivist, then what is metaphysically ‘out there’ isn’t really what we talk about when we talk about values and norms. Still, this point doesn’t rule this out as a genuine possibility. To see this, imagine that the realist is right about everything: our semantics is objectivist, and there really are objective norms. Imagine next that someone who was persuaded by Mackie’s arguments starts a new community somewhere far up the Amazon. He teaches them a specially devised dialect of English where the semantics is subjectivist. The possibility I am now considering is simply that we are in the same position as the members of this imagined community. How could we ever come to realise our predicament? Does this predicament even make sense, and if not, why?

We can approach these questions from a different direction. We saw that the realist makes two claims: one in semantics, one in metaphysics. But what is, on the realist view, the relation between the two? Do they match by accident?** If so, then it would be only by sheer luck that we avoid the peculiar predicament described above. But if the match between semantics and metaphysics is not a mere accident, what explains it? Presumably, the metaphysics would have to explain the semantics: we think in objective terms because there are objective norms to think about. But now the realist is saddled with a very peculiar causal claim. She must claim not only that there are these non-natural properties out there, but also that these mysterious properties somehow (benevolently?) make sure that we have the right concepts to think about them. As if realists didn’t have enough problems!

* To save words, I’ll not distinguish between language, words and statements, on the one hand, and thoughts, concepts and propositions. I’ll use semantics to refer generally to accounts of the above. And I’ll similarly mostly speak about normativity rather than more narrowly about value or morality.

** As will be quickly evident to some readers, the little argument that follows echoes a parallel problem about normative belief put forward by Sharon Street and Richard Joyce.


  1. 1. Posted by Jussi Suikkanen | January 9, 2008 6:56 pm


    couple of points. First, there are many people who want to be global error theorists. Second, it’s not clear that the secondary norms are introduced as a normative. Another way to read it is to read the latter part as a causal explanation – given the pragmatic advantages we have evolved in a way that we cannot get ourselves to give up the belief in morality even when we realise it is false. Yet another way to read the error theorist is to think that by saying that we ought not to give up the moral beliefs she just means that this is in our interests. That is not a normative claim but rather a reductive one.

    I’ve talked about the view you introduce with few people. There is a way of reading Alistair McIntyre just as holding that view. He might be seen to be saying that the moral vocabulary once during the antiquity or middle ages really described an normative reality out there. Yet something happened and our use of the moral language degenerated from description to voicing our wants and dislikes. That this happened does nothing to the fact that the reality is out there – we just don’t talk about it.

  2. 2. Posted by Guy Kahane | January 9, 2008 9:33 pm

    Thanks Jussi, good points.

    You’re right about MacIntyre, I haven’t thought of that. It’s not clear how far MacIntyre wants to go. He certainly goes further than his precursor, Anscombe, who can be read as saying that atheism (which of course she did not herself accept) leads to a kind of error-theory with respect to some bits of moral discourse (I suspect that Anscombe’s view was an influence on Mackie, but I don’t know the history).

    The worry I was trying to voice is still not very focused, but although I don’t remember the relevant bits of After Virtue so well, I don’t remember MacIntyre really addressing it. I suppose that to some extent the situation he is describing is different, since the old, abandoned vocabulary, while no longer ours, is still in view, so there is a way back — presumably MacIntyre himself somehow found it, and I think that both MacIntyre and Anscombe hold that many theists held on to objectivism. This suggests another possibility not, I think, often considered in meta-ethics — that different meta-ethical views might give the correct semantics for the normative concepts of different people.

    About global error theories — I discuss this at length elsewhere, and here I just wanted to make a point about how far such a view MIGHT take us, not to argue it DOES take us that far. But just two quick points in response. (1) I myself was never persuaded by the view that we can just appeal to the irresistible force of nature and everything stays the same. This is essentially a very grand empirical claim which nobody has even tried to seriously defend. In any case this view would just predict what would happen if we came to belief the theory, or perhaps report what HAS happened to the author, but it would not describe any pragmatic reasons to hold on to normative discourse. (2) As I understand ‘it’s in our interests’, that’s as much an evaluative/normative claim as any moral statement!

  3. 3. Posted by Jussi Suikkanen | January 10, 2008 10:04 am

    About this:

    “This suggests another possibility not, I think, often considered in meta-ethics — that different meta-ethical views might give the correct semantics for the normative concepts of different people.”

    Walter Sinnott-Amstrong gave a nice paper on this sort of variantist views at the APA. He distinguished between many forms of such views and argued against them. The basic problem of course is one already recognised by Hare. We do seem to be able to enter moral disagreements with pretty much everyone, including the cannibals. Unless we share terms with the same meaning it is not clear how this could be possible.

  4. 4. Posted by Guy Kahane | January 10, 2008 2:02 pm

    Thanks again — not been to the APA…

    I suppose this possibility is naturally raised by the now familiar evidence about cross-cultural differences in intuitions about various basic philosophical concepts. The obvious point to raise here is, as you say, the old point against relativism that such conceptual variation would make genuine disagreement impossible — of course relativists have all sorts of responses to this move that can be tried in the ‘metaethical pluralism’ case.

    I’d be interested to hear what else Sinnott-Amstrong had to say about this but it seems to me that the disagreement point has limited force. Can we really use it to rule out pretty much a priori a scenario such as that described by MacIntyre? It seems to me that this way of trying to rule out metaethicla pluralism would drive us to see normative concepts as having very thin metaethical content — so that, for example, the claim to objectivity wouldn’t be genuinely part of the semantics but a redundant gloss on it. This of course is something many antirealists claim anyway.

    Just to go a little bit further in drawing the implications of the semantics/metaphysics distinction. One small point that emerges is that the distinction ALSO gives further work for the subjectivist/anti-realism. For him too winning the argument about the semantics couldn’t be the whole story, precisely because of the possibility we’ve been discussing.

  5. 5. Posted by Jussi Suikkanen | January 10, 2008 4:38 pm


    It’s true that Mackie made the semantics/metaphysics distinction and that it is useful in many ways. But, it should be noted that it is a controversial distinction that ultimately needs defending. Many anti-realists have good reasons to resist it.

    The traditional picture Mackie works with is that that reference comes before truth. You analyse the moral claims to figure out what they purport to refer to. You then investigate the world and this may disappoint you – you find that there is nothing corresponding there and thus you think your claims are false.

    But, anti-realists can turn this picture around and say that truth drives reference and thus semantics metaphysics. This is a claim that is sometimes called the priority thesis. You start from the looking at which claims are correct within the discourse, i.e., true (understood in epistemic way) and read of the reference from these. This is in a way to become a deflationist about reference and the objects in the world (or at least the talk about them). This seems appealing for instance in mathematics. Some mathematical claims are proven and thus true. Some of these claims use referring singular terms. But, if the claims are true by the norms of the discourse, we can read what is out there and what the claims refer to from the true claims. From true mathematical claims we can read for instance that there are numbers. But, saying that there are numbers need not be seen going beyond the true mathematical claims. This is a way in which one could defend ethics without ontology to use Putnam’s terms.

  6. 6. Posted by Guy Kahane | January 10, 2008 5:50 pm

    Jussi, absolutely — on some ways of understanding the semantics/truth relation, something like Mackie’s error theory wouldn’t make sense. As early critics pointed out, it’s not terribly clear how Mackie himself thought that moral discourse could be sufficiently coherent to be capable of being false — of havine certain truth conditions that the world might fail to meet — yet without there being anything out there that could meet these conditions. (An obvious model here is atheism: does the atheist think that belief in God is incoherent, or that it is false, and does the latter imply that there could have been a God, but that, as it happens, there is none?) Anscombe was far more explict than Mackie on this (though not terribly persuasive).

    This was actually a worry raised by some (self-described) realists like McDowell. I’m not entirely sure I yet see how the anti-realist proposal you describe is really an objection to Mackie’s distinction itself, as opposed to a denial of his semantics. Could it be, on this view, that the semantics was really non-natural & objectivist & cognitivist yet the error theory would not go through because the semantics somehow ‘generated’ the metaphysics? I don’t yet see that.

    Mackie didn’t deny the coherence of antirealism as a possible semantics, he just denied that its the correct semantics of our discourse. But anyway, as I note in my post, something like his error theory might be compatible with various forms of antirealism, including, I think, the view you describe.

    We can define ‘error theory’ without making any reference to realism, simply as (1) ‘All our normative beliefs are false’ or as the stronger claim that (2) ‘All normative propositions are false’ (these are of course just rough statements). This formulation understand the error theory as a (sweeping) substantive claim. And they seem compatible with the kind of antirealism you describe. After all it might be that we are mistaken to think that our normative beliefs meet the standard of correctness presupposed by normative discourse, or even that nothing can meet it because e.g. it presupposes some metaphysical condition that can’t be met. I don’t see what, for example, prevents a noncognitivst theist from endorsing something like Anscombe’s claims about obligation.

  7. 7. Posted by Nick Shackel | January 28, 2008 6:07 pm

    Guy, I’m not sure quite what qualification ‘these are of course just rough statements’ is supposed to achieve, but surely the suggestion that all normative propositions are false is incoherent. Under that suggestion:

    1. It is false that beer is good and it is false that beer is not good.
    2. So it is not true that beer is good and it is not true that beer is not good.
    3. So beer is not good and it is not true that beer is not good.

  8. 8. Posted by Guy Kahane | January 28, 2008 6:52 pm


    Yes this is one complication I had in mind when I qualified the claim. There is some discussion of this problem — e.g. by Sinnot-Armstrong and Charles Pidgen. Alexander Miller in his intro book to Metaethics tries to get around it by defining the error-theory as claiming the falsity only of the atomic positive sentences of the area of discourse.

    This seems more a technical difficulty than an objection to the error-theory. If there is a broader lesson here than I think it might be the one I have argued for elsewhere on other grounds: the thought that normative nihilism is best understood as a substantive normative view, no different in principle from views like utilitarianism or rational egoism.

  9. 9. Posted by Jussi Suikkanen | January 29, 2008 11:50 am


    that argument seems a bit too strong for my liking. If it worked it would be successful for error theory about any domain – and not only against global error theory. And, surely we could be error theorists about some domain. Someone defending error theory could say two things.:

    1. She could deny that beer is not good is a normative claim and thus it can be true. You can do this by saying that not in the claim has a wide scope (It’s not the case that beer is good) rather than a narrow scope (Beer is not-good).

    2. Something like 2 can be formulated plausibly when it comes to vague predicates. Most of us want to accept that it is ok to say that: It’s not true that Fred is bald and it’s not true that Fred is not bald when it is indeterminate whether Fred is bald or not. And, if I remember this correctly, why this does not lead to a contradiction is that some step to a contradiction (which you don’t quite yet have in 3) from this requires bivalance (or the law of the excluded middle) which many views about vagueness reject.

Post a comment

Name: (required)

Email Address: (required) (will not be published)



(Spamcheck Enabled)

This work is licensed under a Attribution-NonCommercial-NoDerivs 3.0.