¹ See Mill, John Stuart, Utilitarianism, Indianapolis, 1979 [1861], p. 18Google Scholar. The footnote is one of two footnotes of any substantial length in the entire work, but does not appear in all contemporary versions of the book (it was added in Mill's second edition).

² Ibid., p. 18.

³ Bentham, Jeremy, An Introduction to the Principles of Morals and Legislation, New York, 1948 [1789], p. 97Google Scholar.

⁴ Ibid., p. 99.

⁵ Bentham also gives an example in which, on his account, the consequences are intentional though the act is not intentional ‘throughout’. See Bentham, p. 83.

⁶ Ibid., p. 84.

⁷ Oddly, Bentham characterizes the direct/oblique distinction as a distinction falling only within the category of intended consequences, but it seems straightforward and plausible enough to extend that distinction to intended actions themselves. If the realization that my action falls under a particular description – an insult, say – motivates me to perform the action, then the action is directly intentional under that description. If, by contrast, I am aware that my action falls under that description but am not at all motivated by that fact, then the action is only obliquely intentional under that description.

⁸ Collected Works of John Stuart Mill, ed. Robson, John, Toronto, 1961–1991, xxxi. 252 fGoogle Scholar.

⁹ In the footnote in question, Mill actually says that the intention is ‘What the agent wills,’ making an intention an external state of affairs rather than a mental state, but presumably he would simply distinguish intentions qua what is intended from intentions qua acts of intending.

¹⁰ It may require the stronger condition that you see it as, to some degree, within your control.

¹¹ Hume, David, A Treatise of Human Nature, ed. Bigge, L. A. Selby, revNidditch, P. H., 2nd edn., Oxford, 1978 [1739], bks. I and IIGoogle Scholar.

¹² Thanks to Jonathan Dancy for useful discussion here.

¹³ Though there are other ways of reading this sentence (for some discussion, see Dancy, Jonathan, ‘Mill's Puzzling Footnote,’ Utilitas xii (2000), 220)Google Scholar, I think the account offered in the text makes the most sense, all things considered.

¹⁴ This interpretation receives further support from a passage in Mill's A System of Logic: ‘Now what is an action? Not one thing, but a series of two things: the state of mind called a volition, followed by an effect. The volition or intention to produce the effect, is one thing; the effect produced in consequence of the intention is another thing; the two together constitute the action.’ (Mill, , A System of Logic, London, 1875Google Scholar, bk. 1, ch. Iii, sect. 5.)

¹⁵ Jonathan Dancy advocates a similar reading of the tyrant case. See Dancy, 221.

¹⁶ An analogy might help make this point more clear – there are different ways we might go about individuating books. Suppose the only thing on my desk is a copy of Hume's Treatise of Human Nature. On one way of individuating books, there are three books on my desk, since the Treatise is divided into three books. On another way of individuating books, there is only one book on my desk, since the Treatise is itself just one book. Either way of individuating books seems reasonable, so long as we are clear about which we are using – nothing much needs hang on which way we go. However, Mill is going to make whether an agent performs an action which is right (or which is wrong) hang upon individuating actions in terms of intentions, but not in terms of motives. Since it is quite significant whether we end up saying that an agent performed a right action (or that she performed a wrong action), we need an account of why we should individuate actions, and hence attribute lightness and wrongness, as Mill suggests.

¹⁷ Mill, , Utilitarianism, p. 7Google Scholar.

¹⁸ Mill seems to suppose that the morally required is a proper subset of the expedient, but this would require some argument. For present purposes, I overlook this subtlety. For useful discussion, see Skorupski, John, John Stuart Mill, London, 1989, p. 324Google Scholar.

¹⁹ For a discussion of some problems with Mill's claim that duties involving rights are all duties of justice, see Lyons, David, Rights, Welfare, and Mill's Moral Theory, New York, 1994Google Scholar.

²⁰ Mill, , Utilitarianism, p. 47Google Scholar.

²¹ Filling in the details of this account in a way that makes the resulting view plausible is no trivial task. For the question naturally arises as to whether we should call an action wrong on the grounds that an ideal set of rules would prohibit it, and call for punishment of the agent, or whether some non-ideal set of rules is the relevant one. If we favour ideal rules then Mill's analysis warrants a charge of implausible utopianism. Perhaps in an ideal world there would be a rule requiring nations never to go to war but this would hardly make it plausible to suppose that in the actual world, where no such rule is in play, that it would always be wrong to go to war. Fred Berger has made a similar point against the idealized reading; see Berger, Fred, Happiness, Justice, and Freedom, Berkeley, 1984, p. 112Google Scholar. Interestingly, structurally similar charges are often brought against Kant; see Hill, Thomas E. Jr, ‘Kant's Utopianism’, Akten des 4. Internationalen Kant-Kongresses, Mainz, Dordrecht, Holland, 1972, pp. 307–15Google Scholar, repr. in his Dignity and Practical Reason, Ithaca, 1992. The basic problem in each case seems to be a failure to distinguish what Rawls refers to as ideal theory from non-ideal theory. If, however, we are not to analyse rightness and wrongness in terms of ideal rules then it is unclear which of the many possible non-ideal rules are the appropriate ones and why they are appropriate in the case at hand. The only obviously salient candidate would be whatever rules are actually in play, but that would be unduly conservative, making it impossible to criticize existing practices on moral grounds (as opposed to grounds of expedience). Further, switching away from rules which would be ideal in the circumstances seems to lose the sense in which one really ought to be punished, since the ‘ought’ is neither the maximizing ‘ought’ of expedience nor the ‘ought’ of morality, if we assume those are distinct 'ought's. Mill seems to need a middle-ground position here, according to which we always take some of our present practices to be justified to some significant degree though any one of them could in principle be called into question. For useful discussion of this sort of holistic interpretation of Mill on this point, see Skorupski, pp. 318–20. The problem is a very tangled one, and I lack the space here to try to resolve it. My suspicion, however, is that a greater emphasis on the rules of one's own conscience, over which one has relatively direct control, would make it considerably easier to defend the ideal conception without falling prey to objections of utopianism. In this respect, I am in full agreement with David Lyons, who also argues that greater emphasis should be placed upon the sanctions of conscience, though I disagree with the details of his proposal (see the following footnote).

²² For a defence of this way of sorting out what is analytic and what is not in Mill, see David Lyons, Rights, Welfare, and Mill's Moral Theory. In elaborating this view, however, Lyons attributes a view to Mill which seems to me to be unwarranted. Lyons distinguishes himself from D. G. Brown, who claims that Mill's view was that it is analytic that any action which is wrong is one which merits some sort of external sanction. See Brown, D. G., ‘Mill on Liberty and Morality,’ Philosophical Review, lxxxi (1972)Google Scholar. I fully agree with Lyons on this point. Lyons, however, goes on instead to attribute to Mill the view that ‘wrongness is conceptually connected with justified guilt feelings, but only contingently or synthetically connected with external sanctions’ (Lyons, p. 130). The text instead seems to suggest that Mill did think there was an analytic connection between wrongness and external sanctions, though certainly not the strong one advocated by Brown, according to which the appropriateness of external sanctions is necessary for an action's counting as wrong. Rather, Mill seems to have thought that it was a conceptual truth that the appropriateness of external sanctions was sufficient, though not necessary, for an action to count as wrong. More specifically, he held that the concept of wrongness is disjunctive (though the disjuncts obviously have something in common – we need not suppose it is disjunctive in some non-natural way): an action is wrong if and only if either (a) external sanctions (legal or non-legal) ought to be applied to it, or (b) internal sanctions ought to be applied to it. Mill's chapter V definition is itself explicitly disjunctive in form (to be wrong is to be such that one ought to be punished ‘in some way or other for it’). It is also true, as Lyons no doubt would emphasize, that cases in which external but not internal sanctions are very rare indeed, but they do not seem to be impossible, and we can even imagine circumstances in which they might be relatively common. I see no reason to suppose that Mill would hesitate to call such at least possible actions wrong, given his explicit definition of wrongness, anyway.

²³ Crisp, Roger makes a very similar suggestion in his commentary on Mill's Utilitarianism (Oxford, 1998)Google Scholar. Crisp remarks that, ‘these claims about intention are best interpreted as advice about when to blame and praise agents, rather than as attempts to offer a “standard of morals”’ (Crisp, p. 122). Crisp does not, however, argue for this plausible interpretative suggestion. In effect, I am trying to vindicate Crisp's claim that Mill has in mind the appropriateness of blame and praise in this passage, though I would (on at least some readings) dissent from Crisp's claim that this is to be opposed to an attempt to offer a ‘standard of morals’. For on my reading Mill would hold that to give an account of when blame/praise is appropriate is, in one sense, to give a standard of morals. This standard is, admittedly, based on the principle of utility, the concept of a duty, and various empirical hypotheses and so is a derivative standard. So if by ‘standard of morals’ Crisp had in mind the ultimate standard of morality (and he probably does) then he is right that Mill is not trying to do that here, and we do not disagree. I labour this point only because I think it is important to make sense of the fact that Mill does claim in the text that the intention entirely determines the morality of the action.

²⁴ This is, at any rate, Jeremy Bentham's argument in Principles, ch. XIII, 10–11, against strict liability, and there is no obvious reason to suppose Mill disagrees with it. I do not mean to imply that the argument is sound. One might reasonably worry that it overlooks (1) the general deterrent effects of such punishments, and (2) the possible general benefit of people choosing not to engage in activities they cannot guarantee will not have consequences for which they might be held liable. Since my primary aim here is the exegetical one of making sense of Mill's view, I need not worry over whether Bentham's argument, which I strongly suspect Mill would endorse, is in fact sound.

²⁵ Mill, , Utilitarianism, p. 47Google Scholar.

²⁶ Ibid.

²⁷ Though this is perhaps a strained sense of ‘good motive’, given the dangerousness of moral fanatics. Thanks to Jonathan Dancy for discussion here.

²⁸ CW, ix. 461 f.

²⁹ Autobiography, CW, i. 51. Thanks to an anonymous referee for bringing this very helpful passage to my attention.

³⁰ Nor should one worry that he had not yet formulated that view, for chapter V was actually composed before chapter II. See Berger, Happiness, Justice, and Freedom.

³¹ It might be objected that my proposal blurs the distinction between act-evaluation and agent-evaluation, since calling an action right on my account amounts to claiming that agents ought to be compelled to perform such actions. If this is a problem, then I think it pretty clearly really is a problem for Mill, given his claims in Chapter V. In fact, though, there is still room for an important distinction between act- and agent-evaluation. Calling an action right is to evaluate the action, and to say that it the action is of a type which ought, in general, to be compelled. The point of telling someone an action is right is to get them to do it. Similarly, the point of telling someone what they did was wrong is to get them not to do it in the future, and to do so by making them feel guilty. By contrast, when we tell someone she is a bad person, as when we call someone a coward, we are telling her not that any particular action of hers is such that she should feel guilty about it, but that she should try to be a different kind of person. In this case, the emotion we aim to produce is shame, rather than guilt, and the main point is to criticize the motives which typically lead the agent to act, rather than to criticize some particular action of hers. For a classic discussion of the distinction between guilt and shame (and one which does not match exactly with my own), see Rawls, John, A Theory of Justice, Cambridge, MA, 1971Google Scholar, pt. III.

³² Many thanks to Geoffery Sayre-McCord and Jonathan Dancy for helpful comments and encouragement.