¹ Such obligations are often called ‘prima facie’, but this technical term is misleading and avoidable, since the indefinite article ‘a’ already signals that a moral obligation might be overridden.

² For example, Nietzsche, Friedrich, ‘The “Improvers” of Mankind’, The Twilight of the Idols, Collected Works of Friedrich Nietzche, ed. Levy, O., 18 vols., Edinburgh, 1909–1913, xvi. 44–9Google Scholar; and Mackie, J. L., Ethics, Harmondsworth, 1977, pp. 15–49.Google Scholar

³ It is not clear whether Descartes's demon could deceive me about necessary truths or about the cogito, but that does not affect my point here.

⁴ If ‘rule out’ seems too strong, substitute ‘exclude’ or ‘have good, adequate evidence against’. The terms do not matter. Also notice that one need not actually go through any process of ruling out a hypothesis in order to be able to rule it out or in order to be justi fied. Thus, justifiedness here is a state rather than a process, and neither the general nor the moral sceptic begs the question against externalists.

⁵ One might expand the antecedent of (2) so that it also requires that I believe and/or am justified in believing that p entails q. More premises would then have to be added to ensure that these new requirements are met. Such additions would not affect my basic argument.

⁶ This relation between kinds of justification is explained in my ‘Moral Skepticism and Justification’ in Moral Knowledge? New Essays in Moral Epistemology, ed. Sinnott-Armstrong, W. and Timmons, M., New York, forthcoming, § 5.2.Google Scholar

⁷ Many contractarians would add ‘under certain conditions’, such as impartiality, but that creates extra problems about why nihilists and sceptics must accept that those conditions (rather than none or some alternative) are the right ones for justifying moral beliefs.

⁸ Sturgeon, Prom Nicholas, ‘Moral Explanations’, in Morality, Reason, and Truth, ed. Copp, D. and Zimmerman, D., Totowa, New Jersey, 1984, p. 63.Google Scholar

⁹ For some more detail, see my ‘Moral Skepticism and Justification’, Moral Knowledge?, § 6.

¹⁰ This kind of view in ethics is associated with Foot, Philippa, Virtues and Vices, Berkeley, 1978.Google Scholar

¹¹ This label is misleading because the etymology of ‘cognitivism’ suggests that it is about knowledge, whereas it really is directly about speech acts or truth values. None the less, I bow to convention here.

¹² Hare, R. M., Moral Thinking, Oxford, 1981, pp. 52–7.CrossRefGoogle Scholar On p. 186, Hare admits that nihilism is coherent on his theory. His only argument against it (in his ch. 11) is that adopting nihilism would have bad effects.

¹³ For a clear discussion, see Blackburn, Simon, Spreading the Word, Oxford, 1984, pp. 189–96.Google Scholar

¹⁴ If I can be said to have an attitude towards these innumerable trivial acts, then this is because ‘attitude’ often refers to an evaluative belief. But expressivists want to analyse moral judgements as expressing something other than beliefs, so they cannot use this cognitive notion of an attitude to specify what is expressed toward these innumerable trivial acts.

¹⁵ One such suggestion is to analyse (i) as expressing withholding disapproval. However, withholding disapproval seems to be either a negative condition (of not disapproving) or a positive act (of fighting an inclination to disapprove). I will argue against the idea of expressing purely negative non-disapproval in the text. The latter, positive kind of withholding is not the kind ofthing that can be expressed, since it is an act. Moreover, there need not be any positive act of fighting an inclination to disapprove in order for one to believe (i). Finally, if the operator operates on disapproval rather than on breaking the law, as in ‘Withhold: Disapproval of breaking the law’, then it will be subject to the same problems as (v).

¹⁶ A different attempt to solve these problems, by Allan Gibbard, is criticized in my ‘Some Problems for Gibbard's Nonn-Expressivism’, Philosophical Studies, lxix (1993), 297–313.CrossRefGoogle Scholar

¹⁷ The problem of negation might seem to be solved by analysing ‘X violates a moral obligation’ as ‘X violates a justified moral prescription (or expression)’. This kind of analysis was presented first, to my knowledge, by Fogelin, Robert, Evidence and Meaning, New York, 1967.Google Scholar It has been developed recently by Copp, David, Morality, Normativity, and Society, New York, 1955.Google Scholar However, ‘justified’ is itself a normative term. If this normative term is also analysed in terms of prescription or expression, then this proposal does not really solve the problem of negation. And if ‘justified’ is not analysed in such ways, then prescriptivism and emotivism as general theories have been given up.

¹⁸ Moral sceptics might add that, according to idealizing constructivism, one cannot be justified in believing that an act violates a moral obligation unless one can be justified in believing that all ideal people would disapprove of an act; and that is impossible. This response parallels the sceptical point that, according to phenomenalism, one cannot be justified in believing in external objects, unless one can be justified in believing the many counterfactuals that constitute such objects; and that is impossible. My argument in the text works even if this additional sceptical argument fails.

¹⁹ Harsanyi, John C., ‘Morality and the Theory of Rational Behaviour’, reprinted in Utilitarianism and Beyond, ed. Sen, A. and Williams, B., Cambridge, 1982.Google Scholar

²⁰ Rawls, John, A Theory of Justice, Cambridge, Mass., 1971, see p. 111.Google Scholar

²¹ Gert, Bernard, Morality, New York, 1988, see p. 209.Google Scholar

²² Blackburn, Simon, ‘Securing the Note: Moral Epistemology for the Quasi-Realist’, Moral Knowledge?, § 4.Google Scholar

²³ Ibid.