1 Howard Sobel, ‘Useful Intentions,’ in H.J. Sobel, Taking Chances: Essays on Rational Choice (Cambridge, MA: Cambridge University Press 1994) 237-54, at 237. Thanks to Howard Sobel, Bruno Verbeek, Francois Schroeter, and anonymous referees for helpful comments on the issues raised by this paper, the OSPRO Committee of the University of Newcastle for their approval of the sabbatical leave during which it was written, and the Research School of Social Sciences, Australian National University, for financial support during this period.

2 To explain this assumption, consider the idea that your desire to mow my lawn and your belief that to do so you must refuel the mower is a reason for you to refuel the mower. First, it may be that only desires of a certain type are reasons for action, so that your desire to mow my lawn is a reason only if, for example, it would survive some reflective examination process. Second, strictly speaking, your reason to refuel the mower is not just your desire to mow my lawn, but rather the combination of that desire and your belief about what you must do to achieve this. And third, this combination is a complete reason for you to refuel to mower, in the sense that your reason to refuel the mower is just the combination of belief and desire, and does not include reference to any other mental state.

3 R. Audi, ‘Intending,’ Journal of Philosophy 70 (1973) 387-403

4 W. Davis, ‘A Causal Theory of Intending,’ American Philosophical Quarterly 21 (1984) 43-54

5 See M. Bratman, Intention, Plans, and Practical Reasoning (Cambridge, MA: Harvard University Press 1987) and A. Mele, ‘Intention, Belief, and Intentional Action,’ American Philosophical Quarterly 26 (1989) 19-30, at 20-4 for non-reductive accounts.

6 See J. Mintoff, ‘Buridan's Ass and Reducible Intentions,’ Journal of Philosophical Research 26 (2001) 207-21.

7 D. Gauthier, ‘Assure and Threaten,’ Ethics 104 (1994) 690-721, at 707; cf. Bratman, 34, on intentions as framework reasons.

8 This paragraph assumes that desires, preferences, and intentions are the only hypothetically possible candidates as reasons for action, but (as a referee pointed out) the argument of the paper does not really depend on excluding the possibility of external reasons. If we assume instead that desires, preferences, intentions, and external reasons are the only hypothetically possible candidates as reasons for action, then it would follow instead that an intention makes an action rational if and only if it makes it that one weakly prefers some action to any alternative, or makes it that one has a new external reason for action. But intentions cannot make it that one has a new external reason for action, and so it still follows that we only need to focus on how an intention can make it that one weakly prefers some action to any alternative. I shall therefore persist with the assumption in the text.

9 In his ‘Judicial Can’t,’ Nous, forthcoming, Shapiro considers the question of how the adoption of rules makes actions rational which otherwise would have been irrational. After endorsing the view that rules operate as constraints on action (31ff.), and rejecting the idea that they do this through rendering agents unable to act otherwise, he suggests that they do this through rendering agents unable to will otherwise, through repressing contrary reasons or disabling inhibitions against following the rule (37ff.). My description of the alternative understanding owes much to Shapiro's discussion, though Shapiro himself does not explicitly address the question of how the adoption of intentions can make actions rational. Thanks are also due to Francois Schroeter for independently suggesting this alternate understanding.

10 As we are about to conclude, this is the only one of the roles Bratman identifies which the putative intention does not play. I assume for the sake of argument that the relevant mental state shares enough of the other features of intentions still to count as an intention.

11 See D. Davidson, ‘Actions, Reasons, and Causes,’ in D. Davidson, Essays on Actions and Events (Oxford: Clarendon 1980) 3-19, who writes more generally of ‘pro-attitudes.’

12 G. Kavka, ‘The Toxin Puzzle,’ Analysis 43 (1983) 33-6

13 This general point is also made by D. Gauthier, ‘Commitment and Choice: An Essay on the Rationality of Plans,’ in Ethics, Rationality, and Economic Behavior, F. Farina, F. Vannucci, and D. Hahn, eds. (Oxford: Clarendon 1996), 218 and by T. Pink, The Psychology of Freedom (Cambridge: Cambridge University Press 1996), 126ff., and also applies to Reductive accounts.

14 But, I have been asked by Bruno Verbeek, might they not be weak-willed? And if weak-willed, could they not intend to A but in full consciousness be disposed (due to their weakness) not to A? They may indeed be weak-willed, but this is no problem. First, the weakness of such a person does not consist in a failure to act as they intend (for, to repeat, they have no intention to A), but rather consists in a failure to intend as they judge (since, presumably, they judge that they ought all-thingsconsidered to do A). Second, even if they really did intend to A, they are hardly rational, since they are nevertheless disposed to act against what they judge they ought all-things-considered to do. So it would still follow Sobel has failed to show how a rational agent could use the Second and Third Ways to make actions rational by intending them.

15 She takes her intention to be evidence for, rather than (say) entailing, a judgment that A is favored by her present desire-belief reasons, since she will be aware that her original deliberations might have been mistaken. However, she takes her intention to be sufficient (subject to new information), rather than (say) one piece of evidence amongst others, since if it were just one piece of evidence amongst others then she would need to consider all the other evidence before acting, and so would forego the benefits of having adopted the intention in the first place.

16 This account must presume the agent prefers to do what she has evidence to believe is favored by her desire-belief reasons. For the distinctive feature of this account is that an action, A, is made rational by her evidence that A-ing is favored by her desire-belief reasons, but (on the assumptions we have made in this paper) such evidence will provide reason to A only if accompanied by an appropriate desire or preference, and the appropriate preference seems to be a preference to do what she has evidence to believe is favored by her desire-belief reasons. 17 This problem is even worse if (contrary to Pink's claim) a decision to A does not determine that one will A, but only raises the expectation somewhat that A will be performed. For then one will not be certain (but only raise the expectation somewhat) that one will perform the preliminary means to A, and so not be certain (but only, etc.) that one will incur increased benefits by doing A and increased costs by not doing A — even assuming there are such benefits and costs. The increase in one's desire to A will be correspondingly smaller.

18 It should be noted that, after reflecting on Bratman's example involving the decision to take the 101, Pink ends up claiming that intentions provide what he calls ‘rationality-constitutive requirements’ for action. He says: ‘Rationality-constitutive requirements just specify the proper functioning of our capacity for applying end-derived [=desire-belief] requirements…. One example of such is the requirement that if we intend the end, we should also intend the means’ (Pink, 134, emphasis in original). However, saying that an intention together with a belief about means requires some action is very close, it seems to me, to saying that they together are a complete normative reason for that action, albeit a reason of a different type from what Pink calls end-derived requirements. See Bratman, 34, on a similar distinction between framework and desire-belief reasons, and J. Mintoff, ‘Are Decisions Motive- Perpetuating?’ Analysis 59 (1999) 266-75 on intentions as motive (and reason) generators.

19 A defender of one of the challenged accounts (a referee has pointed out) could claim that the analysis of what is to count as an intentional act should refer only to causes (or motivating reasons), in particular causation by intentions, and not to (normative) reasons. An act might be intentional if done with an intention, if it had a cause in the form of an intention, even if that intention is not itself a reason for that action and there is no other motive that is a reason. But what analysis could such a defender give of what is to count as acting with an intention? Consider the following cases: (i) I turn right at Page Mill because I intend to take the 101 and believe I must turn right to do so; (ii) I turn right at Page Mill because I intend to take the 101 and believe I must turn left to do so. In both cases my action is caused by an intention, but, presumably, in case (i) I act with an intention (and so intentionally) while in case (ii) I do not (if such a case is even possible). What is the difference? The obvious suggestion is that in case (i) my intention and belief together are a reason to turn right, while in case (ii) they are no reason at all to do this — but not if (as our defender claims) intentions are not reasons for action. Another suggestion is that in case (i) I believe that what I do is a means to, or a part of, or etc., what I intend, while in case (ii) this is not so — but it seems the plausibility of this suggestion, and how the defender ends up filling out the ‘etc.,’ will depend implicitly on the idea that if an action is a means to, or a part of, or etc., some action we intend then that is a reason to perform that action. And so the onus is on the defender of one of the challenged accounts to explain what it is for an action to be done with an intention without explicitly or implicitly relying on the idea that intentions are reasons for action. If they cannot, then while an act might indeed be intentional if done with an intention, one acts on an intention only if that intention is a reason for that action.