Published online by Cambridge University Press: 01 January 2020
Roughly a half-century ago, R.M. Hare gave us a potent argument against attempts to account for the meaning of moral language in non-normative or ‘descriptive’ terms. The argument relies on the idea that in order to have genuine moral disagreement, we have to be talking about the same thing. Real disagreement requires agreement in meaning: if the words we use in our disputes mean different things, then we're just talking past, and not to, one another. Using this simple observation, Hare argues that if the meaning of an evaluative word such as ‘good’ were primarily descriptive, then groups with sufficiently different standards for applying ‘good’ wouldn't be able to enter into a real evaluative disagreement. But these disagreements are possible. Hence, he concluded, it's the evaluative meaning of ‘good’ that's primary — and any descriptive account is bound to fail because it doesn't capture the crucial element of endorsement that's central to normative language.
1 This kind of argument comes up in several places in Hare's work. See The Language of Morals (Oxford: Clarendon 1952) and ‘A Reductio ad Absurdum of Descriptivism,’ in Essays in Ethical Theory (Oxford: Clarendon 1989). Similar considerations seem to be behind Allan Gibbard's noncognitivism, especially in Wise Choices, Apt Feelings (Cambridge: Harvard University Press 1990), ch. 1.
2 Of course, their conclusions aren't uncontroversial. The ethical naturalist, though, tends to like externalist views of content and reference, for fairly obvious reasons. See Richard Boyd, ‘How to be a Moral Realist,’ in Essays in Moral Realism, G. Sayre-McCord, ed. (Ithaca: Cornell University Press 1988) for discussion.
3 T erence H o rgan an d Mark T imm o ns, ‘New W ave Mora l Rea li s m Meet s Moral Tw in Earth, Journal of Philosophic Research 16 (1991) 447-65; ‘Troubles for New Wave Moral Semantics: The Open Question Argument Revived,’ Philosophical Papers 21 (1992) 153-75; ‘Troubles on Moral Twin Earth: Moral Queerness Revived,’ Synthese 92 (1992) 221-60; ‘Copping Out on Moral Twin Earth,’ Synthese 124 (2000) 139-52
4 For other responses, see Geoffrey Sayre-McCord, ‘“Good” on Twin Earth,’ in Philosophical Issues 8: Truth, E. Villaneuva, ed. (Atascadero: Ridgeview Press 1997) and David Copp, ‘Milk, Honey, and the Good Life on Moral Twin Earth,’ Synthese 124 (2000) 113-37.
5 See David Brink, ‘Moral Realism and Skeptical Arguments from Disagreement and Queerness,’ Australasian Journal of Philosophy 62 (1984) 111-25 and Moral Realism and the Foundations of Ethics (Cambridge: Cambridge University Press 1989).
6 See David Brink, ‘Moral Realism and Skeptical Arguments from Disagreement and Queerness’ and Moral Realism and the Foundations of Ethics; and Richard Boyd, ‘How to Be a Moral Realist.’
7 I should note a bit of unease about the compatibility between the views of Brink and Boyd, but this doesn't affect my argument, and I'm not convinced it affects Horgan and Timmons in a significant way.
8 See Hilary Putnam, ‘The Meaning of “Meaning,”’ in Mind, Language and Reality (Cambridge: Cambridge University Press 1975).
9 I’m taking the central issue to be the preservation of disagreement, which may seem odd, given that the focus of Horgan and Timmons (‘Troubles on Moral Twin Earth’) concerns the explanation of moral-nonmoral supervenience relations. There's no conflict, though, because the argument about supervenience goes roughly like this: NMR typically appeals to an ‘innocence by association’ strategy in order to show that moral supervenience is no more problematic than is the supervenience of the mental, economic, political, etc. on the physical. But the moral case is special, Horgan and Timmons continue, because, in all of these cases, the supervenience needs to be explained, and the explanation that works for the mental, the biological, and so on doesn't work in the moral case. They conclude that the ‘innocence by association’ strategy is not enough. The Moral Twin Earth case is supposed to show that the semantic constraints that serve as part of the best explanation of other cases of supervenience won't work to explain moral supervenience.
10 Here's another way of supporting the claim that a difference in reference must be accompanied by some difference in practice. According to Boyd's view of reference, ‘Roughly, and for nondegenerate cases, a term t refers to a kind (property, relation, etc.) k just in case there exist causal mechanisms whose tendency is to bring it about, over time, that what is predicated of the term t will be approximately true of k (excuse the blurring of the use-mention distinction).’ Such mechanisms include investigative procedures, deference to experts, and so on. When these relations obtain, ‘we may think of what is said using t as providing us with socially coordinated epistemic access to k’ (Boyd, 116). Those causal mechanisms will include features of our practice, for example, the existence of a body of fairly stable judgments about particular cases, a methodology for handling disputes, and so on. All of these features of moral (or twin-moral) practice have an impact on the reference of our moral terms on a view like Boyd's. If twin-moral practice were identical to our own in every respect, we'd end up, at the end of the day, saying exactly the same things, given that other factors will be held constant. Hence there must be some difference in the initial positions that accounts for the difference in end-of-the-day theory.
11 For example, they suppose that the following claim is true: There is indeed a unique family of functional properties that causally regulates the moral judgments and moral statements of human beings in general, despite the fact that humans widely disagree among themselves about matters of morality (‘Troubles on Moral Twin Earth,’ 245). Later in the paper I'll argue that the concession is something we have reason to believe anyway. But why do Horgan and Timmons concede it? They present the Moral Twin Earth argument as a ‘knockout punch’ to NMR, regardless what happens with actual-world disagreement. This suggests that they think their argument is an additional problem for the realist. If I'm right, this is false. This conclusion is interesting for another reason: it focusses attention to the issue of disagreement and univocity here on Earth, which is where the realist wants it.
12 As always, there are some additional nuances here. Why can't some small, subtle difference in twin-moral practice ensure the difference in reference that Horgan and Timmons need? (After all, they need only to establish that twin-moral terms refer to different properties, not that they refer to drastically different properties.) If the small difference in question is between our current Earthian moral practice and some variant, we'd need to hear why this small change is enough to force the difference in reference, when, we're supposing, the realist can account for the wide variety of different positions represented in past and present moralizing here on Earth. The realist's (assumed) ability to cope with the wide scope of real variation should make us confident that a small change from our current theorizing is not enough to raise problems. (That is, as long as the view of reference in question is tracking roughly the same phenomena that determine meaning — or else we'll have a counterintuitive view of meaning, and a failure to respect the moral inquiry problem, which is discussed below. In that case we have reason to suspect that the view of reference we're working with is one that's ill-suited to the realist's needs.) Surely, though, we can imagine a case where we have one scenario that prompts judgments about shared meaning that's just subtly different from one that doesn’t. Think of a sorities series of possible worlds arranged so that there's only a very small degree of difference (in the relevant respects) in the (purportedly) moral practices on those worlds. Somewhere in the series there would be a point at which ‘right’ stopped referring to rightness and began referring to some other property. Yet, if we said ‘right’ was univocal at one world, consistency pressure forces us to acknowledge that it's univocal at a world almost exactly like it. True enough, but it's plausible to think that our intuitions get weaker and less significant as the cases become more and more different from our home world. Here, it's not clear that (a) we'd have any intuitive reaction at all or (b) what intuitions we did have would count for much.
A final w rinkle: the final part of this section of the paper will argue that the realist should use the notion of an idealized, end-of-the-day moral theory in addressing issues about univocity. If this is the right strategy, then there's an additional problem, because practices that look the same (or very close) now might end up diverging by the end of the day (it's more plausible to assume that they’re causally independent communities). Then we’d have a case where our intuitions clearly stood in opposition to the realist's semantics. There are a few responses available to the realist. First, we’d have to hear more about how suitably idealized moralizers would start with the same raw material, use the right process of investigation, and end in different places, since we'd think, initially, that this wouldn't happen. But it might turn out that way, perhaps because the seemingly trivial differences between our starting points ended up mattering a great deal. In this case, I think the considerations raised by my second argument against Horgan and Timmons (in section III) would apply. Roughly, the response would be this. We should respect the views of suitably improved thinkers in epistemically superior conditions. Hence, we have reason to think that their moral views are better than ours. We also have reason to think that they (the idealized moralizers) would reject the idea that the idealized twin-moralists mean what they do by terms like ‘right.’ Hence we discard the intuitions.
This reply works only if the two end-of-the-day theories are suitably distinct. If they’re quite similar, yet different enough to force the realist to admit there's a difference in reference, this reply won't cut any ice; we, along with our idealized, better-informed selves, would think that the other camp's terms do mean the same thing. In order to head off this challenge, we’d need some arguments in normative ethics — we need to show that it's implausible to think that two similar theories are both maximally coherent and stable end-of-the-day views. I think this is implausible, but giving convincing arguments for this will have to wait for another day. If, contrary to my suspicions, this is indefensible, the realist has another move available: in a case like this (with subtly different limit theories), we might think that a term like ‘right’ is used in different senses, between which our current use remains ambiguous. (Compare the subjective and objective senses of ‘right.’) Conceptual refinement is a plausible outcome of further inquiry. (Thanks to Nick Sturgeon for making this point and suggesting this example.)
13 I f this objection were just the familiar line that any natural properties would fail to play the right sort of action-guiding role, simply in virtue of being natural properties, realism would be dead in the water, though the requirement itself would look suspiciously question-begging. What I have in mind, in talking about the practical role problem, is just the thought that the realist needs to do more to show how his account allows for moral terms to play their distinctive role — and questions about how reference is fixed are relevant to the practical or evaluative importance of these terms.
14 The idea that our end-of-the-day moral theory is part of what determines the correct account of moral properties is a fairly popular one. See Brink, ‘Moral Realism’ and Moral Realism; Sayre-McCord, ‘“Good” on Twin Earth’; and Nicholas Sturgeon, ‘Moral Explanations,’ in Morality, Reason, and Truth, D. Copp and D. Zimmerman, eds. (Totowa, NJ: Rowman and Allanheld 1985). See also Georges Rey, ‘Concepts and Stereotypes,’ Cognition 15 (1983) 237-62 for a related suggestion about concepts generally.
I should also note that Horgan and Timmons wouldn't find this objectionable, since they assume that, at least on the Brink-Boyd view, the functional properties to which our moral terms refer are those described by the moral theory that's arrived at via applications of a coherentist methodology. Though they don't say much about the connection between (a) the properties that causally regulate our moral practice and (b) the properties described by an idealized moral theory, it seems that they're happy to grant that the methodology we use in arriving at (b) will be a way of getting at (a).
15 This point is made in Allan Gewirth, ‘Positive “Ethics” and Normative “Science,’” Philosophical Review 69 (1960) 311-30.
16 See Nicholas Sturgeon, ‘Moral Explanations’; Peter Railton, ‘Moral Realism,’ Philosophical Review 95 (1986) 163-207; and Brink, Moral Realism for discussions of this point. Michael Smith in The Moral Problem (Cambridge, MA: Blackwell 1995) also discusses convergence and its role in normative inquiry.
17 Think, for example, of Shelly Kagan's arguments in The Limits of Morality (Oxford: Clarendon 1989) concerning the instability of ‘hybrid’ views that seem, initially, to be quite plausible. As another example, think of the efforts of Peter Singer, Peter Unger, and other radical moralists to convince us that we’re deeply confused about what morality requires.
Reflections on these considerations might tempt us to think that there just aren't all that many coherent and stable end points when it comes to moral theorizing, because views that combine consequentialist and deontological elements will collapse under pressure. Furthermore, we might think that once we admit some consequentialist element into our moral theorizing — say, the idea that there's a pro tanto reason to promote the good — we'll have a hard time keeping ourselves from ending up with a straightforward consequentialism once we subject our views to critical scrutiny. If there are few stable endpoints, and if the consequentialist elements in our thought are deeply rooted and hard to contain, then all sorts of initial positions will end up producing Tc on reflection. If so, then surface differences between these starting points, which may look quite significant, are irrelevant to the question of shared reference that occupies us here.
18 Think again of the ease with which a radical moralist like Singer can use our seemingly innocent intuitions and principles to drive us to conclusions we find utterly bizarre. Like Peter Unger, I think that this is evidence for the claim that our ordinary thought is confused — though I don't follow him past that point. See Peter Unger, Living High and Letting Die (New York: Oxford University Press 1996) in particular.
19 Thanks to an anonymous reviewer for insisting on the importance of this point.
20 Keep in mind, too, that it is essential to Horgan and Timmons’ argument that Moral Twin Earth and Earth have different maximally stable and coherent end-of-the-day views. This is crucial because this difference is the means by which they force the realist to admit that ‘right’ refers to different properties on different planets.
21 It may look as though I'm cheating by relying on an overly robust idealization. In this particular case, that doesn't seem right, since, after all, our idealized moralists have already considered something like the twin-moralists’ theory Td, and rejected it. We're assuming only that the idealized theorists have ironed out the wrinkles of their views and come to stable consensus. It seems odd that moralists might purge themselves of the deontological elements of their moral thought, commit themselves fully to Tc, and then turn around and find the justifications for Td compelling enough to change their minds.
Generally, we need make our inquirers only idealized enough to consider fully the reasons for competing moral views, actual and imaginable. That a view doesn't have real defenders, or doesn't have many, shouldn't make it impossible for improved reasoners to think through its possible justifications. If so, then it's hard to see what new material a distinct group (whether idealized or not) would bring to the table.
For now, I'm ignoring the significant problems involving the idea of a limit of inquiry or a theoretical end of the day. This notion faces serious challenges, yet I'm confident that any appropriate and defensible construal of it can do the work I need it to do here.
22 Again, the judgment of equivocation is made by a speaker who's undergone this kind of idealization process. As I'll argue below, part of the reason that we think the Earth-Moral Twin Earth disputes are univocal is because both Tc and Td get at something true, as far as we can tell. That is, both of these theories get a part of our (current, unrefined) thinking right. But Horgan and Timmons don't account for the fact that our current moral theory is improved by the process that yields Tc; we get to consequentialism because we think, on reflection, this is the moral theory that fits best
23 I ‘m assuming that our concepts and the concepts of our improved selves are the same, or that our inquiry doesn't change the meaning of the terms involved. This strikes me as the most plausible way to interpret the case, but, of course, other readings are possible. One might think that the results of our moral inquiry are so radical that we've changed the very meanings of the terms. Though I won't say anything about this here, I'll note that Horgan and Timmons cannot take this line, because doing so undermines their argument. Their objection requires that the meanings of our terms remain constant through theory-change, because they need for the meanings of our terms now to be fixed by the end results of this normative inquiry. If the end-of-the-day-theory doesn't say anything about the current meanings of our moral terms, then a difference in end-of-the-day theory doesn't entail a difference in meaning between moral terms and twin-moral terms now.
24 This locution is suggested by Gibbard's attempts to describe a sense of ‘bland, flavorless’ endorsement that captures ‘the last ought before action.’
25 This example, and a similar point, appear in Copp, ‘Milk, Honey, and the Good Life.’
26 One might think that twin-moral discourse differs from other evaluative perspectives (those of etiquette, economics, prudence, and so on) in having some kind of overriding purport. That is, like moral discourse, and unlike any other, twin-moralizing demands absolute allegiance; its requirements — from its point of view — cannot be overridden. Surely it does, but, we might think, so does the financial point of view, the prudential point of view, and so on. One might claim that prudential reasoning (for example) never really conflicts with moral requirements, because a correct understanding of prudential requirements shows that they respect the dictates of morality. This just isn't plausible. When someone follows self-interest instead of morality, we think they're wicked, or weak-willed, or overwhelmed by temptation, not that they've failed to understand the demands of prudence. On the contrary, it seems that these people understand prudential demands all too well.
27 It may be that particular reasoners think that what's morally right settles what to do, and that non-moral reasons are no reasons at all, at least when they compete with moral considerations. And they might be right. But surely we need to come to this further conclusion about what to do by a slightly different means from the way we answered the question of what's right, since we have at least one additional question to settle, namely, ‘does it always make sense to do what's right?’
28 It's interesting to note that this kind of evaluation makes sense only from the third person; first-personally, the question of what to do is settled by one's commitments. I owe this point and the example to Don Hubin.
29 So, for example, we can imagine people who make their decisions by thinking about beauty, or their own interest, or some other kind of consideration. These people have moral failings, we might think, but it's not clear that they're incoherent.
30 Hare tried to identify moral claims by their formal features alone, and this led him to some implausible views. We might think that Hare was on to something but erred in thinking that moral discourse could be identified without appeal to its content. The combination of naturalistic realism about morality and expressivism about the last ought before action is a way of preserving the insight in Hare's work while still holding on to a thicker, more contentful picture of our moralizing.
31 Note, too, that the same sentence is often used to express contradictory recommendations in different contexts. I might say ‘it would be financially crazy to take this trip!’ to urge us to stay home, or I might say it as a way of encouraging us to throw caution to the wind.
32 I would like to thank Justin D'Arms, Don Hubin, Nick Sturgeon, and William Taschek for comments on earlier drafts, and for many helpful conversations. This paper has also benefitted from the comments of three anonymous reviewers, and I am grateful for their suggestions. I'm also indebted to Pamela Hieronymi, Tyler Hower, and Julie Tannenbaum for insightful discussions and for their charity to alien interlocutors. Research on this paper was made possible by support from the Ohio State University and the Charlotte Newcombe Foundation.