Sometimes a second-order claim can undermine the first- order claim that it's about. Consider Moore's paradox: "It is raining, but I don't believe it." What makes this sentence paradoxical and seemingly contradictory, although formally it is consistent, is that the second half of the claim undercuts the speaker's right to assert the first half. If the speaker does not believe it is raining, he has no right to assert that it is raining. There are other examples along similar lines: "It is raining, but there's no reason to think it is raining"; "It is raining, but I don't know whether it is raining." These statements, while not strictly contradictory, are self-undermining: the second-order claim in the second conjunct undermines the assertion of the first-order claim in the first conjunct.
What conditions are required for a speaker to have a right to assert something? The above examples point to at least three such conditions: the claim must be true; the speaker must believe the claim; the speaker must have reason to believe, or be justified in believing, the claim - and, whether this is redundant or a distinct, fourth condition, the speaker must know the claim. These conditions, then, generate at least three (or four, depending on how you count the last) distinct ways a second-order claim might undermine a first-order claim. Perhaps there are other conditions that a claim must satisfy as well. Nothing turns on these being the only such conditions, although I do insist that there are at least these three conditions.
If this is correct, then a well-defined question can be posed for any given meta-ethical theory: Does the theory undermine normative ethics? That is, does the second-order theory undermine our ability to make moral claims, claims such as "Honesty is a virtue"; "Tom should return that book to the library"; "Death is bad"? At least some possible meta-ethical theories have this property: consider the view that moral claims assert propositions that no one ever knows to be true. This would be a claim of meta- ethics, not of normative ethics, but it clearly undermines one's ability to assert a theory of normative ethics. Equally clearly, some meta-ethical theories do not have this property: consider Moore's theory that ethical claims are claims about a non-natural property that we have intuitive awareness of. Whatever else may be said about this view, it does not undermine normative ethics.
The moral realist holds that moral claims assert propositions that are at least sometimes objectively true. What is meant here by the qualifier "objectively" is that their truth does not depend on beliefs, feelings, or other attitudes of observers towards the things evaluated. This provides a reasonable interpretation of the notion of the objectivity of ethics. Assuming the correspondence theory of truth, this view entails the usual slogans about values being 'part of reality' or 'part of objective reality.'
Let us use the blanket term "anti-realism" for any view that denies this. There are, then, four possible forms of anti- realism:
(1) Moral claims do not assert propositions.
(2) Moral claims assert false propositions.
(3) Moral claims assert propositions that are neither true nor false.
(4) Moral claims assert propositions whose truth depends on some attitudes of observers.
Each of (1)-(4) is incompatible with moral realism, and the denial of (1)-(4) entails moral realism, so these are the only four possible forms of anti-realism. (I leave aside the possibility of holding one of these four views with respect to some moral claims, and one or more of the others with respect to other moral claims. If my arguments succeed against each of (1)-(4), then they will also succeed against any such hybrid view.)
Consider some simple examples of forms of anti-realism:
(i) "x is good" just means "I like x."
On this theory, some moral claims will be true. For example, I like chocolate ice cream, so if I say, "Chocolate ice cream is good," my statement will be unproblematically true. If someone else, who does not like chocolate ice cream, says the same words, however, his statement will be false - again, unproblematically false.
This theory counts as a form of anti-realism, however, because the truth of the claim depends upon the attitudes of the speaker towards chocolate ice cream, and this would be true for any moral claim on the present theory. Thus, the goodness of chocolate ice cream would fail to be 'objective.'
(ii) A moral claim is an expression of emotion on the part of the speaker. "Adolf Hitler was evil" is comparable to, "Adolf Hitler--boo!"
On this theory, moral claims do not assert propositions. It is perhaps misleading then to call them moral claims at all, but there seems to be no neutral terminology, so let us just resolve to understand "moral claims" to refer to those things that are commonly thought of as moral claims, whether or not they genuinely are claims.
(iii) Moral claims purport to identify properties of objects (properties of 'goodness' or 'rightness') of a kind that in reality no object has. In reality, only the speaker's emotions or instincts or the conventions of his society prompt him to attribute these properties to things.
On this view, moral claims will be either false or perhaps neither true nor false. The claims do have propositional content and are assertive in nature, but what they assert is never the case.
Notice that it is the first part of this view - that moral claims purport to identify certain properties which in reality do not exist - that makes this a form of anti-realism. The second part, concerning how people come to make these claims, is not by itself a form of anti-realism. The view that people's moral judgements are generally unconnected with the moral reality, so to speak, even though there is a moral reality, is a perfectly consistent, even if not a widely held, meta-ethical theory, and would be classed as a form of moral realism. There may not be much point in holding that there are objective moral truths if one is not going to use them to explain why people hold the moral views they do, and there may be other objections to this view besides. I only wish to make clear that such a position is possible and would be a form of realism, and therefore the second half of the theory stated above does not by itself entail anti- realism.
Our characterization of moral realism and anti-realism draws some support from its correct classification of these cases. The philosopher who wishes to avoid moral realism, then, must endorse at least one of (1)-(4), or be charged with inconsistency (or at least indecisiveness). And herein lies the dilemma. Each of (1)- (4) is problematic in its own way; each will prove very difficult to defend.
Consider the view
(1) Moral claims do not assert propositions.
This seems false on a casual examination of a moral claim. "Bob was a good man" has the same grammatical form as "Bob was a tall man." It contains a subject and a predicate, with the verb in the indicative mood - prima facie marks of an assertion. We should not assign too much weight to such superficial considerations, but I think they do establish at least some presumption of the assertive nature of moral claims. The anti-realist owes us some explanation of how and why he thinks (1) is the case. What can be said, then, to render (1) plausible?
Moral claims seem to be accompanied normally and perhaps always by certain emotions. The person who insists that slavery is morally wrong feels a certain kind of indignation - moral indignation - at the thought of the practice of slavery. The person who says that Mother Theresa is a good woman typically feels another emotion - a kind of approval and satisfaction - towards Mother Theresa. Perhaps what moral claims do is only to give expression to these feelings. "x is good" is comparable to "Hurray for x" in this respect, although perhaps the two sentences express somewhat different emotions, or express them in different ways. Now sentences like "Hurray for x," "Congratulations on your Nobel Prize," or "Ouch" do not assert propositions. They do not assert that the speaker feels a certain way about something; they just serve to express his feeling.
Alternatively: Moral claims also generally seem to serve a function of regulating behavior. When I accept the claim, "x is right," this generally has a more-than-accidental correlation with my doing x. If I accept that x is right but fail to do x when given the chance, I am guilty of some kind of inconsistency. Furthermore, when I assert that x is right, I generally want other people to do x as well. Perhaps, then, moral claims are something like a sophisticated kind of imperative. "x is right" is comparable to "Let x be done," and "x is good" comparable to "Let x be pursued." Now sentences like "Turn that music down" and "Don't jump in the water" do not assert propositions. They do not assert that the addressee is going to perform a certain action; they recommend or propose that he perform it. Likewise, perhaps moral claims serve to recommend courses of action rather than to describe them, although perhaps they do so in an unusual way or under special conditions.
Now both of these views, as well as any other version of (1), are undoubtedly in conflict with how we ordinarily think of moral claims. There is no doubt that we do ordinarily think of moral claims as assertions and treat moral claims like assertions. If someone says, "Adolf Hitler - yay!" an appropriate response might be "Boo!" but not "That's false," nor even, "I disagree." But if he says, "Adolf Hitler was a good man," then "That's false" becomes appropriate. Likewise, in response to "Let us support Adolf Hitler," one cannot say, "That's not true," although one can certainly say, "Let us not support Adolf Hitler."
Moral claims can also enter into logical relations. One can say, "I should return this book to the library; therefore, I can return this book to the library." This may not be a very useful inference (since it would be difficult to know the premise unless one knew the conclusion), but at least it makes sense. But it does not make sense to say, "Hurray for x; therefore, I can do x." This makes as much sense as to say, "Ouch; therefore, it's raining outside." And, pace Professor Hare, "Let us do A; therefore, we can do A" is just bad English.
Of course one could propose extended conventions for the use of "false" such that you can call an imperative that you oppose the following of "false," and one could similarly propose extended usages of "entails" and "therefore." This would perhaps enable one to continue talking in the same way as those who do not hold (1), but it would be quite misleading. The fact would remain that one's theory was at odds with common sense, which treats moral claims straightforwardly as assertions.
Our problems worsen once we turn to more complex claims involving moral terms: "Unless the verdict was unjust, Simpson didn't kill his wife." That is an intelligible claim - even, I think, a true one. But, "Unless boo for the verdict, Simpson didn't kill his wife"? Compare also, "Unless a terrible miscarriage of justice occurred, Simpson didn't kill his wife." How shall we read this? "Unless boo! on something that happened, Simpson didn't kill his wife"? Or, "Unless let us not have things of a sort happen that did happen, Simpson didn't kill his wife"?
Perhaps some fancy logical footwork could be done to render these into intelligible-sounding sentences - though the difficulties are far from trivial. But notice that no such trouble has to be taken with the sorts of sentences that really are, uncontroversially non-assertive. Sentences like "Turn the music down" and "Congratulations on your Nobel Prize" do not cause us these kinds of problems. They do not require to be made intelligible in contexts like "If turn the music down, then the music is very loud," or "If congratulations on your Nobel Prize, then you got a Nobel Prize." We simply do not talk like that. (Though we can say, "If the music is too loud, then turn it down.")
Now perhaps this does not prove that moral claims assert propositions, since more work could perhaps be done on the above problems, and it's conceivable that moral claims work differently from all other non-assertive utterances. But all the linguistic evidence supports their being assertions. Moral claims act just like assertions of propositions, and the old wisdom says, if it looks like a duck, walks like a duck, quacks like a duck . . .
In response to these problems, it is natural to propose a return to
looking at moral claims as assertions, but assertions about a
speaker's attitudes towards the objects of evaluation. Thus,
instead of regarding "x is good" as expressing a feeling
towards x, we might regard it as reporting a feeling
towards x. Or instead of regarding "I ought to do A" as an
imperative addressed to myself to do A, we might regard it as
reporting a desire or resolution to do A. In either case,
moral claims will assert propositions that will sometimes be true:
If I do have the required feeling about x, then my statement, "x is
good" will just be true. Or, on the other theory, if I do have the
desire or intention to do A, then it will unquestionably be true
for me to say, "I ought to do A." Alternatively, we might regard
moral claims as reporting the attitudes of a community
towards the object of evaluation. "x is good" may report that the
speaker's community approves of x. Thus, we have
(4) Moral claims assert propositions whose truth depends on some attitudes of observers.
This kind of theory has the advantage of allowing us to treat moral claims as propositions, allowing us to make straightforward sense of sentences like, "Unless the verdict was unjust, Simpson didn't kill his wife." (I say we can make sense of such sentences, though they might not any longer appear true.) It also allows us to treat some moral claims as true and others as false.
Any such theory also has at least three problems, however. First, it makes it, if not impossible, very odd to disagree with someone about a moral issue. If a person says, "Abortion is wrong," I may be inclined to disagree with him, without questioning his sincerity. I may say, "That's not true." But it would be most impertinent of me, if a man reports that he feels a certain way about abortion, to insist, "That's not true" - to insist, in other words, that he does not feel the way he says he does.
Perhaps I may still say I disagree with him in the sense that I have a different, and opposite emotion about abortion. But if this is a form of 'disagreement,' it is one that leaves open the possibility that both parties may be correct - that is, that what each says may be true. And however tolerant I may be of differing opinions, I do not think that what a Nazi asserts when he says, "Adolf Hitler was a great man" is true. We all of us feel that what he asserts is certainly not true.
Some of the difficulty may be remedied by holding moral claims to report the standards of a community, rather than an individual. In that case, at least, I may disagree with the individual who says abortion is wrong, without accusing him of lying. I just think he has misconstrued the standards of our community. However, this shifts my difficulty to that of disagreeing with the community. I could not, on this view, hold, "My community approves of abortion, but abortion is wrong." Yet this certainly seems like a consistent position. Statements of this form have frequently been true throughout history. For example, it would have been true for a Southerner in the early nineteenth century to say, "My society approves of slavery, but slavery is wrong." The present form of anti-realism would have to regard such a position as a contradiction.
The second objection to (4) is that, for any attitudes that might plausibly be identified as correlated with moral truths (e.g., a sentiment of approval, or a certain kind of desire), it is normally or very often the case that we hold the attitudes in question because of certain moral truths. For example, it very often is the case that I feel a sentiment of approval when I contemplate x, because x is good. But is x ever good because I feel a sentiment of approval towards it? This seems unlikely. Or again: I very often desire or intend to do A because A is the right thing to do. But is A ever the right thing to do because I desire or intend to do it? Perhaps some unusual circumstances might be imagined in which that would be true, but that is not normally the case. The reason I should not sexually harass my students is not that I don't want to or that it wouldn't make me feel good. If I don't want to and if it wouldn't make me feel good, the reason for this is that I believe it to be wrong. But then I couldn't have come to believe that only because of my desires or feelings.
An analogous objection applies to the theory that would rely on the attitudes of a society or culture to determine moral truths - it then could not be the case that the society disapproves of A because A is wrong.
The third objection to (4) is that it implies that if our attitudes were to change in certain ways, then the moral facts would change in ways that are counter-intuitive. Suppose that sentences of the form "x is good" are held to report an attitude of approval on the part of a majority of our society (the same argument will apply regardless of what attitude is substituted for approval, and which observers are substituted for the majority of society). Then it will follow that if we all took an attitude of approval towards Adolf Hitler, then Adolf Hitler would be good. This consequence follows - and remains counter-intuitive - regardless of whether it is possible for us to take that attitude towards Hitler. A similar argument shows that in theory, all the world's problems would be solved if only we could get most people to approve of everything that is presently bad. The bad things would not cease to exist; they would just become good. For example, it is at present bad that there are people starving to death in some parts of the world. But if we could get enough people to approve of famine and the attendant suffering and death, then the world would be improved, since one of the major problems would be solved. Yet this consequence is hard to accept. If anything, it seems that the imagined development would constitute a serious worsening of the world's situation. Again, this is true regardless of whether the imagined development is psychologically or nomologically possible.
One way of countering this objection would be to 'rigidify' evaluative terms - that is, one might propose that "good" always has for its extension what we actually approve of, even when we are discussing counter-factual situations in which we wouldn't approve of those things. Thus, the reason we don't have to say that if everybody approved of Hitler then Hitler would be good, is that we presently disapprove of Hitler, and we continue to disapprove of him even when we're thinking about that counter-factual situation.
One consequence that can not be escaped by this ad hoc maneuver, however, is that any sentence of the form, "If we approve of x, then x is good," where the conditional is indicative (not subjunctive), must be true. Thus, we get
If most people approve of Adolf Hitler, then Adolf Hitler was a good man.
If most people approve of starvation, then starvation is good.
and, following our earlier observation that unless the verdict in the Simpson trial was unjust, Simpson didn't kill his wife:
Unless we disapprove of the verdict in the Simpson trial, Simpson didn't kill his wife.
which seems particularly unacceptable.
It seems clear, then, that common sense dictates that the truth of moral claims does not depend upon the attitudes observers take towards the objects of evaluation.
The preceding considerations naturally lead the reflective anti-realist to the most radical but most sophisticated position. The above outlined positions deny that moral claims even make any purport to objectivity and can thus be refuted by appeal to linguistic intuitions. The anti-realist is best off admitting that moral discourse does purport to describe an independently- existing moral reality. Moral claims are assertions, and they attribute objective properties to things - that is, properties that would be objective if anything really possessed them - but in fact, no such properties have ever existed. Moral claims involve an habitual error. The error theorist may then go on to explain how we come to make this error (perhaps by mistaking reactions in us for properties of external things).
If the arguments of the preceding sections are sound, then the anti-realist is forced to this position. But here his dilemma is acute. A second-order theory that holds that all first-order claims of a certain class involve a definite error, as clearly undermines those claims as anything possibly could. Given the error theory, it would appear that every claim of normative ethics is straightforwardly false, for it attributes to something a property that nothing has. If I hold this view, I undermine my ability to make ethical claims. I cannot say, "The book is orange, but nothing is orange" or "The book is orange, but that's false." And nor can I hold that murder is wrong but that's false.
How serious of a problem is this? Consider such claims as
(a) Happiness is good.
(b) Honesty is a virtue.
(c) It is wrong to burn children just for the fun of it.
The anti-realist must disagree with such claims, not of course in the sense of asserting their contraries, but in the sense of holding them false. He would not say happiness is bad, but he must insist that happiness is not good. Yet surely, if those evaluations are either true or false, they are true, rather than false.
This objection may appear to border on begging the question. But what we have to ask ourselves is this: what arguments is the anti-realist able to offer against moral realism; and are the premises of any of these arguments more initially plausible than each of (a)-(c)? We must choose between rejecting (a)-(c) (along with all other moral claims), and rejecting the anti-realist's premises. Only if he can adduce some premises that are (jointly) more certain than any of (a)-(c) can he hope to convince us to resolve the dilemma in his favor.
Perhaps the anti-realist would deny he is committed to holding all first-order evaluations false. Perhaps moral claims contain one or more false presuppositions and are for that reason neither true nor false, just as "The King of France is bald" is neither true nor false because it presupposes but does not state that there is a King of France. Nevertheless, at least this much is clear: the anti-realist of the 'error theory' variety can not hold moral claims to be true. So whether or not he accepts the law of excluded middle and concludes that "Happiness is good" is false, he must at least maintain that it is not true. And from the point of view of maintaining our first-order moral discourse, this is no improvement. A claim that contains false presuppositions is as clearly unassertable as a claim that is false. I cannot say, "The King of France is bald, but there is no King of France." And nor can I say, "Murder is wrong, but there are no objective values," if the existence of objective values is presupposed in first-order moral claims.
Yet most anti-realists would like to go on using moral language in the same way they always did before they came to their philosophical conclusion - they would like to insulate their moral views entirely from their meta-ethical views. They would like to - and they generally do - continue to hold that it is wrong to lie to one's friends, that abortion is permissible, that they have a right to property, and so on. Indeed, it would be difficult to imagine doing without such beliefs, in much the same way that it would be difficult to imagine living with the consequences of epistemological skepticism. Perhaps the anti-realist will propose that we go on using moral language, but with different meanings attached to it (he cannot rationally propose that we go on saying the same things with the same meaning, for the reasons just discussed). Perhaps he would propose that we simply subtract out the false presuppositions contained in moral claims, and continue to use them to assert the rest of what such claims normally assert. But this is a dubious suggestion. What is left of the claim that the King of France is bald, after we have relinquished the assumption that there is a King of France? And what is left of the claim that murder is wrong, after we have relinquished the assumption that there are such things as right and wrong (assuming these to be the properties the error theorist denies exist)?
Perhaps the error theorist will propose that we use moral language in one of the ways that other anti-realists have mistakenly claimed we already do. Perhaps he will propose that we use moral language to express emotions, issue imperatives, or report emotions or desires. Still, he will not be able to go on saying the same things that realists do about normative ethics, for the reasons given in sections III and IV. For suppose he uses "x is good" to mean something like "Let us pursue x": then he will not be able to intelligibly say that "x is good" is either true or false (when he says it), he will not be able to use it as a premise in an argument, and he will not be able to say things of the form, "If x is good, then p" (where p is some proposition). In other words, his new application of moral language will be subject to all the constraints we argued would apply to us if the prescriptivist analysis of our actual use of moral language were correct.
A similar point applies to any of the other anti-realist analyses he might choose to take over. So the anti-realist cannot go on talking the same way about ethics as ordinary folk do. Moreover, no such dissimulation on the anti-realist's part can remove the fact that he is denying what we ordinarily believe when we make moral claims, nor can it relieve him of the burden of supporting his theory with some arguments that are more compelling, more certainly correct, than those ordinary beliefs that he rejects.
The moral subjectivist or anti-realist's final dilemma is this: Will he give up his philosophical, meta-ethical theory, or will he give up his common sense, first-order moral beliefs? He cannot have it both ways.