1 McConnell, T.C. ‘Moral Dilemmas and Consistency in Ethics,’ Canadian Journal of Philosophy 8 (1978) 272–3CrossRefGoogle Scholar; Hare, R. M. Moral Thinking (New York: Oxford University Press 1981), 27–8CrossRefGoogle Scholar; and Conee, E. ‘Against Moral Dilemmas,’ Philosophical Review 91 (1982) 87–8CrossRefGoogle Scholar

2 Since the context is moral, the adverb ‘morally’ should be understood to qualify every use of ‘ought’ in this paper. Nonetheless, my arguments can be extended to uses of ‘ought’ in non-moral contexts.

3 I will use the term ‘alternative’ so that there is an alternative that the agent ought to adopt when the agent ought to do some act of a certain kind (O(Ex)Kx)) and when the agent ought not to do any act of a certain kind (O-(Ex)Kx).

4 A better definition of moral dilemmas is that they are situations where there are non-overridden moral requirements for an agent to adopt incompatible alternatives. This revised definition still implies (1), because moral requirements are one kind of moral reason, and, as I will argue below, an agent morally ought to adopt any alternative which there are non-overridden moral reasons for that agent to adopt. I discuss the need for these revisions in ‘Moral Dilemmas and Incomparability,’ American Philosophical Quarterly 22 (1985) 321–2Google Scholar.

5 See, for example, my ‘A Solution to Forrester's Paradox of Gentle Murder,’ Journal of Philosophy 82 (1985), 162–8CrossRefGoogle Scholar.

6 Grice, H.P. ‘Logic and Conversation,’Google Scholar in Davidson, Donald and Harman, Gilbert eds., The Logic of Grammar (Encino, CA: Dickenson Publishing Company 1975) 67Google Scholar.

7 McConnell, T.C. ‘Moral Dilemmas and Requiring the Impossible,’ Philosophical Studies 29 (1976) 409CrossRefGoogle Scholar

8 Conee, 87

9 McConnell, Cf. ‘Moral Dilemmas and Consistency in Ethics,’ 272–3Google Scholar; and Conee, 87-8.

10 Although this definition is used by most advocates of this argument, other definitions of ‘permitted’ do not support (4.12). For example, if ‘ought’ is defined in terms of moral reasons, including moral ideals, but ‘permitted’ is defined in terms of moral requirements, then ‘permitted’ does not imply ‘not ought not,’ because there might be no moral requirement not to adopt an alternative, even if there is a moral ideal not to adopt the alternative. Nonetheless, the argument from ought and ought not can be reformulated in terms of moral requirements, if moral dilemmas are defined in terms of moral requirements, as in n. 4.

11 I criticize some such arguments in ‘Moral Dilemmas and Incomparability.’

12 A similar explanation of why ‘ought’ seems to imply ‘permitted’ is given by Foot, P. ‘Moral Realism and Moral Dilemma,’ Journal of Philosophy 80 (1983) 385CrossRefGoogle Scholar.

13 Grice, 69

14 Grice, 67. J. Searle also uses this maxim of quantity to show a related but different conversational implicature in ‘Prima Facie Obligations’ in Practical Reasoning, ed. Raz, J. (New York: Oxford University Press 1978), 88Google Scholar.

15 Grice, 74

16 Such conversational considerations also explain why speakers seem to disagree when one says that an agent ought to do something and another says that the agent ought not to do it. They cannot both be correct if both refer to overriding moral reasons. Even if they are interpreted so that both can be correct, they still seem to disagree, because each denies what is conversationally implied by what the other says. Nonetheless, they can explicitly cancel these conversational implications. If they do, they did not really disagree after all.

17 E.g. Lemmon, E. J. ‘Deontic Logic and the Logic of Imperatives,’ Logique et Analyse 8 (1965) 47Google Scholar; and Trigg, R. ‘Moral Conflict,’ Mind 80 (1971) 46Google Scholar.

18 Hare, 27-8

19 I give one such counterexample in’ “Ought” Conversationaly Implies “Can”,’ Philosophical Review 93 (1984) 252-4. See also my’ “Ought to have” and “Can have”,’ Analysis 45 (1985) 44-8.

20 Fraassen, Van ‘Values and the Heart's Command,’ Journal of Philosophy 70 (1973) 13CrossRefGoogle Scholar

21 Cf. n. 5.

22 McConnell, ‘Moral Dilemmas and Requiring the Impossible,’ 412Google Scholar

23 Even if one rejects (CL), one could substitute a weaker principle, such as ‘(□(A->B)&◊ A) → (OA->OB)’ where ‘◊’ symbolizes logical possibility. This principle does not imply ‘O(A&-A) →OB’ for every value of ‘B.’ However, this escape is ad hoc in the absence of any independent reason to reject (CL), and it does not fit any standard formal semantics for deontic logic.

24 A defender of moral dilemmas can replace the agglomeration principle with a slightly weaker principle: [(OA&OB)& ◊(A&B)] → O (A&B). This qualified principle does not take us from ‘OA&O-A’ to ‘O(A&-A),’ so it cannot trouble defenders of moral dilemmas, if they deny that ‘ought’ implies ‘can.’ However, this principle is ad hoc in the absence of independent justification, and it is subject to the objections that follow in the text.

25 Cf. n. 19.

26 Different arguments against agglomeration are given by Williams, B. ‘Ethical Con sistency,’ reprinted in Problems of the Self (London: Cambridge University Press 1973), 180Google Scholar ff.; and Marcus, R. B. ‘Moral Dilemmas and Consistency,’ Journal of Philosophy 77 (1980) 134CrossRefGoogle Scholar.

27 I have benefitted from helpful comments by Bob Fogelin and Ruth Marcus.