¹ For Sidgwick lumped with later non-naturalists, see, for example, Frankena, William K., Ethics (Englewood Cliffs, 1973), pp. 102 f.Google Scholar; Brink, David O., ‘Moral Motivation’, Ethics, cviii (1997), 9 Google Scholar.

² Since Sidgwick, unlike Moore, relies on pointing out problems with specific analyses, he neither can nor does claim to defeat all possible analyses. He sometimes admits that a future analysis might succeed (FC 480, IO 89; contrast FP 107 f.). (Bare parenthetical numbers in the text refer to Sidgwick, Henry, The Methods of Ethics, 7th edn. (Indianapolis, 1981)Google Scholar. When necessary, I abbreviate this as ME. Abbreviations for other Sidgwick works: FC = ‘Some Fundamental Ethical Controversies’, Mind, xiv (1889)Google Scholar; FP = ‘The Establishment of Ethical First Principles’, Mind, iv (1879)Google Scholar; I0 = ‘Is the Distinction Between “Is” and “Ought” Ultimate and Irreducible?’ Proceedings of the Aristotelian Society, n.s., i (1892); VB = ‘Verification of Beliefs’, Contemporary Review, xvii (1871); LK =Google Scholar Lectures on the Philosophy of Kant and Other Philosophical Lectures and Essays, ed. Ward, James (London, 1905)Google Scholar; LE = Lectures on the Ethics of Green, T. H.,. MrSpencer, H., and Martineau, J., ed. E. E. Constance Jones (London, 1902)Google Scholar.

³ This position is quite standard. It is held, for example, by Price, Reid, Alexander Smith, W. G. Ward, and Grote before Sidgwick, and Moore, Prichard, Ross, Broad, and Ewing after Sidgwick. The argument directed against the voluntarist is even more common – Cumberland, Cudworth, Shaftesbury, Hutcheson, and Whately give it before Sidgwick. For much of the history, see Prior, Arthur N., Logic and the Basis of Ethics (Oxford, 1949)Google Scholar.

⁴ '[T]he fundamental difference remains that the distinction between “truth” and “error” in our thought about what is, is held to depend essentially on the correspondence, or want of correspondence, between Thought and Fact; whereas, in the case of “what ought to be”, truth and error cannot be conceived to depend on any similar relation except on a certain theological view of duty” (I0 91).

⁵ Gibbard, Allan, Wise Choices, Apt Feelings (Cambridge, MA, 1990), p. 33 Google Scholar; Smith, Michael, The Moral Problem (Oxford, 1994), p. 31 Google Scholar. See also Pettit, Philip in Three Methods of Ethics (Oxford, 1997), pp. 105–10Google Scholar and Pettit, and Jackson, Frank, ‘Moral Functionalism’, Philosophical Quarterly, xlv (1995)Google Scholar.

⁶ These are common non-aturalist claims. See (for example) Moore, G. E., Principia Ethica (Cambridge, 1903), pp. 16 f.Google Scholar; Ewing, A. C., Ethics (London, 1962), p. 79 Google Scholar; Russell, Bertrand, ‘The Elements of Ethics’, Russell, , Philosophical Essays (London, 1966)Google Scholar, sect. 4. For the role of examples, see also Smith, , Problem, pp. 55, 163 f.CrossRefGoogle Scholar

⁷ Contrast Smith, , who objects that supervenience is a mystery for non-naturalists because, on their view, ‘all we can say about non-natural moral properties a priori is that they are simple properties’. Problem, p. 24 Google Scholar.

⁸ This account of rationality is taken from the first edition of the Methods (London, 1874), pp. 25 f. As Schneewind notes, Sidgwick is less explicit in later editions, but this is ‘the ideal of practical rationality which always haunted him’. See Schneewind, J. B., Sidgwick's Ethics and Victorian Moral Philosophy (Oxford, 1977), p. 235 Google Scholar.

⁹ It is hard to gloss ‘metaphysically distinct’. I intend the picture critics of non-naturalism present (described in the text below). One gloss is that metaphysically distinct properties are independent, in that each could exist without the other. But nonnaturalists do not believe this. Another gloss is that metaphysically distinct properties can depend for existence on the presence of base properties but are not identical to (or perhaps even constituted by) the base properties. Since I shall argue that Sidgwick could accept even identity, I leave aside the issue. For identity and constitution in the context of naturalism, see Brink, , Moral Realism and the Foundations of Ethics (New York, 1989), pp. 157–9CrossRefGoogle Scholar; for non-naturalism, see pp. 160n 12, 165n 16.

¹⁰ Schneewind writes that Sidgwick ‘does occasionally speak, especially in the earlier editions, of “qualities” of rightness and goodness, but this is not to be construed as implying a theory of the sort later put forward by Moore or Ross about the ontological status of what is known when we know that an act is right or good. Any theory Sidgwick has on this matter remains implicit. The basic [normative] notion itself is not … a platonic entity visible to our mental gaze; it is not an introspectively discoverable mental entity; it is not a concept derived from perception of a unique non-natural property. It is the constraint imposed by reason, which in matters of theory is familiar enough through the laws of logic, as it bears on activity and sentience’ Sidgwick's, Ethics, pp. 205, 222.Google Scholar

¹¹ Warnock, G. J., Contemporary Moral Philosophy (London, 1967), p. 14 CrossRefGoogle Scholar.

¹² Nowell-Smith, P. H., Ethics (Harmondsworth, 1954), pp. 41, 48Google Scholar.

¹³ Ayer, A. J., ‘On the Analysis of Moral Judgements’, in Philosophical Essays (London, 1954), p. 242 Google Scholar.

¹⁴ Smith, , Problem, p. 25 Google Scholar.

¹⁵ Brink, , Realism, pp. 162 fGoogle Scholar.

¹⁶ Some later non-naturalists clearly rejected the inference from failure of non-normative analysis to non-natural properties. See Raphael, D. D., Moral Judgment (London, 1955), pp. 35 f., 42 f.Google Scholar; Ewing, , Second Thoughts in Moral Philosophy (London, 1959), p. 52 Google Scholar.

¹⁷ For the latter, see, for example, Frankena, p. 103.

¹⁸ For evidence of a revival, see the Sidgwickian positions in Scanlon, T. M., What We Owe to Each Other (Cambridge, MA, 1998)Google Scholar, ch.1 and Parfit, Derek, ‘Reasons and Motivation’, Proceedings of the Aristotelian Society, suppl. vol. lxxi (1997), 121 fGoogle Scholar.

¹⁹ See, for example, Putnam, Hilary, Reason, Truth and History (Cambridge, 1981), pp. 206–8CrossRefGoogle Scholar; Harman, Gilbert, The Nature of Morality (New York, 1977), p. 19 Google Scholar; Brink, , Realism, pp. 165 f.Google Scholar; Boyd, Kichard, ‘How to be a Moral Realist’, in Essays on Moral Realism, ed. Sayre-McCord, Geoffrey (Ithaca, 1988)Google Scholar; Railton, Peter, ‘Naturalism and Prescriptivity’, Social Philosophy and Policy, vii (1989). Simon Blackburn finds it ‘unclear what problems’ the property-identity-without-synonymy approach solves. Surely the attraction, for naturalist and non-naturalist, is that one can avoid metaphysically distinct properties while rejecting meaning equivalences. Given Blackburn's metaphysical parsimony, he should see the appeal. For the charge, seeGoogle Scholar ‘How To Be an Ethical Anti-Realist’, in Blackburn, , Essays in Quasi-Realism (New York, 1993), pp. 180 fGoogle Scholar.

²⁰ Smith, , Problem, pp. 48–58. For the platitudes, see pp. 39–41Google Scholar.

²¹ Sidgwick insists that even empiricists must admit the need for intuitive knowledge of universal rather than particular judgements, such as the judgement that nature is uniform (VB 589 f.).

²² Similarly, Moore writes that ‘when I call [some] propositions “Intuitions”, I mean merely to assert that they are incapable of proof; I imply nothing whatsoever as to the manner or origin of our cognition of them’ (Moore, p. x). For the same separation of intuitionism from perception, with helpful discussion of other objections to intuitionism, see Brink, , Realism pp. 107–13Google Scholar; Audi, Robert, ‘Intuitionism, Pluralism, and the Foundations of Ethics’, in Moral Knowledge?, ed. Sinnott-Armstrong, Walter and Timmons, Mark (New York, 1996)Google Scholar.

²³ For Sidgwick's epistemology, see Schneewind, , pp. 56–61 or my Rational Egoism (New York, 1999), pp. 63–71 Google Scholar.

²⁴ Christine Korsgaard suggests a different connection: because non-naturalists find normative concepts unanalysable, ‘they think that our understanding of the normative concept does not enable us to pick out its objects, and that we must therefore have recourse to a sense or a faculty of intuition that functions like a sense.’ (See Kant's, ‘Analysis of Obligation: The Argument of Foundations F’, Monist, lxxii (1989), 331.) This explanation cannot fit Sidgwick, however, since he features neither a sense-like faculty nor indeed any special process for applying normative concepts to the non-normative world. Similarly, Frankena writes that for non-naturalists, basic moral judgements ‘can only be known by intuition; this follows, it is maintained, from the fact that the properties involved are simple and non-natural’ (p. 103). Sayre-McCord writes that ‘because moral properties are taken to be nonnatural and not discoverable by normal empirical investigation, this account falls back on “moral intuition” (which is no less mysterious) to explain how it is we come to know about morality’ (3 f.)Google Scholar.

²⁵ For example, see Monro, D. H., Empiricism and Ethics (Cambridge, 1967), pp. 81–6Google Scholar; Warnock, pp. 15 f.; Blackburn, , Spreading the Word (Oxford, 1984), p. 188 Google Scholar; Hare, R. M., ‘The Practical Relevance of Philosophy’, in Essays on Philosophical Method (London, 1971), pp. 112–14;Google Scholar Moral Thinking (Oxford, 1981), p. 71 Google Scholar; Smith, ‘Should We Believe in Emotivism?’, Fact, Science and Morality: Essays on Ayer's, A. J. Language, Truth and Logic, ed. Macdonald, Graham and Wright, Crispin (Oxford, 1986), pp. 295 fGoogle Scholar. For a recent statement of the motivation objection raised against Sidgwick in particular, see Freeman, Samuel, ‘Contractualism, Moral Motivation, and Practical Reason’, Journal of Philosophy, Ixxxviii (1991), 291 Google Scholar.

²⁶ See Smith, , Problem, p. 61 Google Scholar.

²⁷ In the first edition of the Methods, Sidgwick suggests an externalist account of the tie between accepting a moral claim and motivation: ‘[W]e need not ask whether a mere cognition can act upon the Will and prompt to action. It is enough if it be granted that there exists in all moral agents as such a permanent desire… to do what is right or reasonable because it is such: so that when our practical reason recognises any course of conduct as right, this desire immediately impels us… towards such conduct’ (p. 27). This account would not satisfy those who criticize non-naturalism on grounds of motivation, however; nor, it seems, did it satisfy Sidgwick, given the different explanation suggested by the final edition.

²⁸ Smith, Problem, ch. 5.

²⁹ For a similar construction, see Schneewind, pp. 225 f., who notes that Sidgwick himself offers this sort of analysis in the first two editions of the Methods. Sidgwick rejects Fowler's analysis of ‘right’ in terms of ‘good’ at FC 482 f., but Fowler's analysis, unlike that proposed here, eliminates all use of ‘right’ or ‘ought’ in the specification of ‘good’. For a proposal similar to Sidgwick's, see Ewing, , ‘A Suggested Non-Naturalistic Analysis of Good’, Mind, xlviii (1939)CrossRefGoogle Scholar.

³⁰ Smith, , Problem p. 186 Google Scholar.

³¹ Smith, , Problem, p. 56 Google Scholar.

³² Smith, , Problem, pp. 21–4Google Scholar. Smith describes intuitionism as the view that ‘we possess a special faculty of moral intuition which enables us to see that… claims to the effect that there are certain relations between natural properties and non-natural moral properties, are self-evident’ (‘Should We Believe?’, p. 295, emphasis removed; also p. 290, Problem, p. 19). Frankena has the same view: non-naturalists claim to know by intuition that, for example, ‘what is pleasant or harmonious is good in itself; or … that one man ought to be just, kind, and truthful toward another man’ (p. 103).

³³ Mathematics is similar. The most plausible candidates for self-evidence are propositions stating relations between items in the mathematical realm, rather than connections between the mathematical realm and the world (such as the claim that our world is Euclidean). Sidgwick writes that ‘I have tried to show how in the [axioms] there is at least a self-evident element, immediately cognisable by abstract intuition; depending in each case on the relation which individuals and their particular ends bear as parts to their wholes, and to other parts of these wholes. [These principles] present themselves as self-evident; as much (e.g.) as the mathematical axiom that “if equals be added to equals the wholes are equal”’ (pp. 382 f.; see also p. 507).

³⁴ Strawson, , ‘Ethical Intuitionism’, Philosophy, xxiv (1949)Google Scholar. Ayer does not add the synthetic a priori possibility. See Ayer, , ‘Analysis’, pp. 236 fGoogle Scholar. Ayer also gives the (much criticized) argument that non-naturalists treat moral claims as descriptions; descriptions cannot be recommendations; moral claims are recommendations; hence non-naturalism is false (pp. 240–2, 244 f.). (Ayer's, argument is endorsed by Hare, The Language of Morals (Oxford, 1952), p. 30.)Google Scholar

³⁶ Toulmin, Stephen, The Place of Reason in Ethics (New York, 1950), pp. 19–21 Google Scholar. This argument seems to be endorsed by Hare, Language, p. 95.

³⁶ Toulmin, pp. 25, 28.

³⁷ Nowell-Smith, pp. 43, 46, 56–8.

³⁸ Ayer, , Language, Truth and Logic (New York, 1952), p. 106 Google Scholar. Smith and Strawson offer another variation: intuitionists provide no theory of error for their special faculty. See Smith, ‘Should We Believe?’, p. 295 and Strawson, p. 27. W. Hudson's main objection to eighteenth century intuitionism is, again, the absence of epistemic tests. See Ethical Intuitionism (New York, 1967), pp. 57–60 CrossRefGoogle Scholar.

³⁹ To my knowledge, Richard Brandt is the only mid-century critic to distinguish Sidgwick from those keen on a perceptual model. Unfortunately, his criticisms then point out disanalogies between plausible candidates for synthetic a priori status such as ‘if anything is red, it is spread out in space’ and synthetic a priori connections between natural and non-natural properties. He also finds no agreement on ethical claims, but does not consider the sort of axioms Sidgwick offers as passing an agreement test. See Ethical Theory (Englewood Cliffs, 1959), pp. 198–201 Google Scholar.

⁴⁰ Hare, , Language, pp. 96 f., 103Google Scholar.

⁴¹ Hare, , Language, pp. 104 f., 108Google Scholar.

⁴² Hare's argument fails even against its intended target. By‘knowing the criteria of application’, Hare means knowing how to sort. But I can understand redness without being able to sort things into red and not-red; say it is dark, or the light is red. And I can sort without understanding; say redness is correlated with some feature, such as the wavelength of the reflected light, and I sort on this basis. Of course the ability to sort under some conditions may be needed for understanding redness. You could test my understanding by asking me to sort things into red and not-red, in cases where I know that the light is white and adequate. But similarly, the ability to sort things under some conditions into good and not-good may be needed for understanding goodness. You could test my understanding by asking me to sort augers into good and not-good, in cases where I know the purpose of augers. The background information needed to move from understanding to criteria as one goes from object to object is much greater for ‘good’ than for‘red‘, but this does not establish a difference in kind.

⁴³ Hare, , Language, pp. 148 fGoogle Scholar. Hare makes the argument with increasing frequency in later works such as ‘The Argument from Received Opinion’, in Essays on Philosophical Method, Moral Thinking (pp. 69 f), and throughout Essays in Ethical Theory (Oxford, 1989)Google Scholar, Sorting Out Ethics (Oxford, 1997), andGoogle Scholar Objective Prescriptions and Other Essays (Oxford, 1999)Google Scholar. The argument is repeated by Gibbard, ch.l and pp. 33 f, and by Black-burn, , Spreading, p. 168 Google Scholar.

⁴⁴ At one point, Hare notes that ‘intuitionism’ avoids the argument against naturalism, but he rejects intuitionism (in many places) on the ground that it provides no method of resolving disagreements. (Ethical Theory, pp. 118 f.; for the rejection, see, for example, Ethical Theory, pp. 104–7, 118 f.) In Moral Thinking he writes, ‘I am, when I speak of intuition, neither attacking nor defending Sidgwick’ (p. vi). This leaves it unclear what objection Hare would raise against Sidgwick's reply to the no-disagreement argument.

⁴⁵ Gibbard, p. 154. ‘Ethical intuitionists like Sidgwick and Moore seemed to think of normative judgment as a power to apprehend facts of a special kind, much as the senses involve a power to apprehend the layout of surrounding objects’ (p. 107; see also p. 317). Similarly, Frankena's main objection to non-naturalism is that moral claims prescribe (p. 105). According to Darwall, Frankena's defection from Sidgwickian non-naturalism was due largely to doubts about ‘a kind of perception that brings independently existing normative facts into view'. See Darwall, Stephen, ‘Learning from Frankena’, Ethics cvii (1997), 698, 702Google Scholar.

⁴⁶ See, for example, Stoljar, Daniel, ‘Emotivism and Truth Conditions’, Philosophical Studies, lxx (1993)Google Scholar.

⁴⁷ Gibbard, p. 318. For Gibbard on non-cognitivist objectivity, see Gibbard, part III. For a short statement of another attempt, see Frankena, p. 108.

⁴⁸ Schneewmd, p. 72.

⁴⁹ Sidgwick, notes ‘ordinary moral or prudential judgments’ (p. 25). I leave aside prudential judgements, since one might treat them as hypothetical imperatives and so as instances of (i). They are also not cases in which one is tempted to say that a recognition of reasonableness motivates without input from non-rational desire, and so are no test for HumeGoogle Scholar.

⁵⁰ John Deigh reads Sidgwick as inferring from failure of naturalistic analysis to motivation by reason without desire. He gives hypothetical imperatives as a counter-example: on Sidgwick's view, they cannot be analysed as descriptions of means-ends connections; yet hypothetical imperatives motivate through desire. On my reconstruction, however, Sidgwick does not make this inference. Failure of naturalistic analysis rules out Hume's tactic for handling apparent counter-examples to his contention that reason cannot motivate without desire. Failure of naturalistic analysis does not itself justify concluding that reason can motivate without desire. Deigh also objects that I cannot know, by introspection, that reason rather than calm passion motivates me. My reconstruction does not require this positive knowledge. It requires that I be able to rule out desire motivating in the manner of (i) or (ii); Sidgwick thinks this can be done by rejecting the analysis of moral judgements (i) and (ii) suggest. See Deigh, , ‘Sidgwick on Ethical Judgment’, Essays on Henry Sidgwick, ed. Schultz, Bart (Cambridge, 1992), pp. 242 f., 251–8Google Scholar.

⁵¹ Moore, p. ix.

⁵² The desire presumably counts as a rational rather than non-rational one, and so the proposal of II is anti-Humean according to Sidgwick's understanding of Hume.

⁵³ Moore, pp. vii–ix; see also pp. 20 f.

⁵⁴ Stevenson, C. L., ‘The Emotive Meaning of Ethical Terms’, Mind, xlvi (1937), 14 CrossRefGoogle Scholar.

⁵⁵ Ayer, , ‘Analysis’, p. 246 Google Scholar.

⁵⁶ Nowell-Smith, front cover.

⁵⁷ Hare, , Language, p. iii Google Scholar. For a more recent statement of the benefits of conceptual clarification, see Gibbard, pp. 9, 31–3. Hare himself, of course, has not repented.

⁵⁸ Sidgwick does think Spencer guilty of the confusion Moore finds. ‘[W]e must distinguish inquiry into the meaning of words from inquiry into ethical principles. I agree with Mr. Spencer in holding that “pleasure is the ultimate good”, but not in the meaning which he gives to the word “good”. Indeed if “good” (substantive) means “pleasure,” the proposition just stated would be a tautology, and a tautology cannot be an ethical principle.' But Sidgwick quickly moves on, noting that this ‘relates to form rather than substance’ (LE p. 145). In contrast, Moore treats Spencer's (apparent) reliance on definition as the root of Spencer's ‘utter confusion’(Principia, pp. 53 f, xv).

⁵⁹ See, for example, Brink's argument for objective utilitarianism inRealism, ch. 8.

⁶⁰ For more on the similarity to Rawls, see Schultz, , ‘Introduction’, Essays on Henry Sidgwick, p. 7 Google Scholar.

⁶¹ Ayer, , Logic, p. 102 Google Scholar.

⁶² I wish to thank Sissela Bok, Bob Bright, Roger Crisp, Brad Hooker, Carl Matheson, David Phillips, Tim Schroeder, Bart Schultz, John Skorupski, audiences at Manitoba andUtilitarianism 2000, and, especially, Joyce Jenkins, for helpful comments.