Explaining the Instrumental Principle

Jonathan Way University of Southampton

[This is a preprint of an article whose final and definitive form will be published in the Australasian Journal of Philosophy]

ABSTRACT: The Wide-Scope view of instrumental reason holds that you should not intend an end without also intending what you believe to be the necessary means. This, the Wide-Scoper claims, provides the best account of why failing to intend the believed means to your end is a rational failing. But Wide-Scopers have struggled to meet a simple Explanatory Challenge: why shouldn’t you intend an end without intending the necessary means? What reason is there not to do so? In the first half of this paper, I argue that the Wide-Scope view struggles to meet the Explanatory Challenge because it takes the principles of instrumental reason to have unlimited application – to apply to all agents, in all circumstances. I then go on to offer an new account of these principles. The new account is very much in the spirit of the WideScope view, and shares its central advantages, but lacks its unlimited application. This view should, therefore, find the Explanatory Challenge more tractable. In the second half of the paper, I argue that this prediction is confirmed. If the requirements of instrumental reason apply only when a means is, or is believed to be, necessary for your end, then plausible independent claims about reasons, rationality, and intentions, explain why failing to intend the necessary means to your ends is a rational failing.

1

1. Introduction

All summer long you have intended to see the exhibition at the national gallery. But as the summer has rolled along, you have never quite found the time to do so. Now it is the last day left, before the exhibition ends. Unfortunately, the new semester is looming, and you had rather hoped to spend the day preparing for the upcoming classes. Since you do not have time to do both, you face a choice – either go to the exhibition today, or drop your intention of seeing the exhibition at all. What you should not do – and what it would be irrational to do – is to maintain your intention to see the exhibition, while refraining from forming the intention to go today. Examples of this sort suggest that means-end incoherence – intending to E, believing that M-ing is necessary for E-ing, but failing to intend to M – is a form of irrationality.1 This in turn suggests that an account of practical reason should recognize an instrumental principle – a requirement against means-end incoherence. There has recently been considerable debate about how this principle is to be formulated, and what its grounds are. The debate begins with the failure of the following, perhaps initially tempting, principle:

NS

If you intend to E and believe that M-ing is necessary for E-ing, you should intend to M.

The problem with NS is that it implies that you can make it the case that you should intend to M, simply by intending to do something for which you take M-ing to be a

1

This claim is slightly qualified at the end of this section.

2

necessary means. Since M is arbitrary, this means you can make it the case that you should intend anything at all. This is very implausible.2 The most influential solution to this problem is the Wide-Scope view.3 According to the Wide-Scope view, the principle violated in means-end incoherence is:

WS

You should not [intend to E, believe that M-ing is necessary for E-ing, and not intend to M].

What WS requires is simply that you avoid means-end incoherence. So, since there is more than one way to avoid means-end incoherence, WS does not imply that if you intend to E and believe that M-ing is necessary for E-ing, you should intend to M. If you are means-end incoherent, you can also come to comply with WS by dropping the intention to E or by dropping the means-end belief. So WS does not have the implausible implications of NS. WS is a far more plausible formulation of the instrumental principle than NS. However, NS is not the only alternative to WS. Consider also the Disjunctive view:

D

If you are means-end incoherent, then either you should intend to M or you should not intend to E or you should not believe that M-ing is necessary for Eing.4

On this view, means-end incoherence guarantees that you have an attitude you should not have, or that you lack an intention you should have. So when you are means-end 2

For versions of this objection see, e.g. Bratman [1987: 23-27] and Broome [1999: 409-10]. See, e.g. Broome [1999], Dancy [2000: ch. 2], Darwall [1983: 15-17, 43-50], Wallace [2001]. 4 For defence of D, see Kolodny [2007: 251-2], [2008: 452-3]. 3

3

incoherent, there will always be a particular move you should make, to resolve the incoherence. The Wide-Scope view need not accept this. According to the WideScope view, the problem with means-end incoherence is precisely in the means-end incoherent combination. Taken individually, each of the elements of the combination may be unobjectionable.5 This is an important advantage of the Wide-Scope view over the Disjunctive view. Consider the example we began with. In this case, it is – or at least could be – false that there is one particular way in which you should resolve your means-end incoherence. You might form the intention to go to the exhibition today. But you might equally well drop the intention to see the exhibition at all. Either of these moves would be perfectly acceptable. Nor are such cases unusual. In many cases our ends are merely permissible – for example, because there is an alternative which is equally, or incommensurably, worth pursuing, or because there is an alternative which is supererogatory. In such cases, it is not the case that you should drop the end, and nor need it be the case that you should intend the means, or drop the means-end belief. The Disjunctive view appears to rule such cases out. The Wide-Scope view has a further important advantage over the Disjunctive view. The Disjunctive view struggles to capture the sense that means-end incoherence is a distinctive kind of irrationality. This concern is best brought out by considering cases in which you intend to do something that you should not intend to do. Suppose, for instance, that you should not intend to cheat on the test, but intend to do so anyway. And suppose that you are means-end incoherent with respect to this intention – although you know that cheating requires stealing a copy of the test, you lack the nerve to do this. Intuitively, two things have gone wrong here – you have an intention 5

This kind of structure should be familiar. For example, it may be okay to take drug 1 and okay to take drug 2, but definitely not okay to take both drug 1 and drug 2.

4

you should not have, and you then fail to intend the means to carrying it out. You are both misguided and incompetent. The Wide-Scope view gets this right, because WS is a distinct principle, holding over and above the fact that you should not intend the end. But the Disjunctive view gets it wrong. According to the Disjunctive view, all that is wrong is that you have an intention you should not have. You make no further mistake in not intending the means to your end. I take the advantages of WS over NS and D to provide a good prima facie case for the Wide-Scope view.6 Nonetheless, the Wide-Scope view faces serious challenges of its own. In particular, Wide-Scopers have struggled to meet a simple Explanatory Challenge: why should you comply with WS? What reason is there not to do so? It is this challenge which I consider in this paper. In section two, I briefly review the problems the Wide-Scope view faces meeting the Explanatory Challenge. In section three, I argue that Wide-Scopers struggle to meet this challenge because of a certain feature of WS – its unlimited application. In section four, I go on to offer an alternative account of the instrumental principle. The account offered is very much in the spirit of WS, and shares its central advantages, but lacks its unlimited application. This account should thus find the Explanatory Challenge more tractable. In sections five, six, and seven, I argue that this prediction is confirmed, by outlining an explanation of the formulation of the instrumental principle I offer. Before moving on, two points of clarification should be noted. First, it is not quite accurate to say that means-end incoherence, as defined so far, is a form of irrationality. Failing to intend what you take to be the necessary means to an intended end is only irrational if you also believe that intending the means is necessary for

6

Of course, there are other alternatives to the Wide-Scope view that I cannot discuss here. See, e.g. Raz [2005], Finlay [2009].

5

taking the means.7 For brevity, I shall continue to omit this condition in statements of putative instrumental principles. However, it is worth bearing in mind that it is meant to be implicit, and in the footnotes, I shall observe a couple of points at which the condition is particularly important.8 Second, one might wonder whether the WideScope view is really a genuine alternative to NS and the Disjunctive view. For on certain assumptions, both NS and D entail WS.9 However, whether or not we accept these assumptions, it is certainly not the case that WS entails either NS or D. I thus take the distinctive claim of the Wide-Scope view to be that WS is true while NS and D are false.10 It is this claim, after all, that allows Wide-Scopers to avoid the problems of NS and D.

2. The Explanatory Challenge

In ‘Why be Rational?’ Niko Kolodny argued that Wide-Scopers cannot answer the question ‘why should you comply with WS?’11 The basic idea here is quite easy to see. The natural way to explain why you should comply with WS is to show that complying with WS is either necessary or sufficient for doing something else you should do. But WS cannot be explained in either of these ways. You can comply with WS without thereby doing anything else you should do. For instance, you comply with WS by intending the means to wicked and crazy ends. You do not thereby do

7

Some would insist on the stronger condition that you believe that now intending to M is necessary for M-ing (e.g. Setiya [2007: 668]). This stronger condition can also be assumed, if it is felt necessary. 8 See n. 26 and n. 35. 9 Consider the view that ‘ought’ is a sentential operator, and that ‘you ought that p’ is true just in case ‘p’ is true at all of a certain set of possible worlds. On such a view, ‘necessarily, if p, you ought that q’ entails ‘necessarily, you ought that not [p and not-q]’. 10 Of course, Wide-Scopers need not deny that many instances of NS and D are true. 11 Kolodny [2005: 542-7]. These arguments are repeated and embellished in Broome [2005], Kolodny [2007], [2008], Southwood [2008] and Way [2010].

6

anything else you should do.12 And although this is perhaps less clear, it is also plausible that you can violate WS without failing to do anything else you should do. We saw evidence for this when considering the Disjunctive view – means-end incoherence seems not to guarantee that you have any attitude you should not have, or lack any attitude you should have.13 These difficulties may move us to consider various less obvious ways of showing that you should comply with WS. For instance, we might try appealing to the value of the disposition to comply with WS. Or we might consider whether compliance with WS is in some sense constitutive of being an agent with intentions. Alternatively, we might wonder whether it is simply a mistake to look for reasons to comply with WS. Perhaps WS is simply a basic normative principle, and so not the sort of thing we should expect to find other reasons to comply with. However, each of these options is problematic. The value of being disposed to comply with WS would not explain why, as WS claims, you should avoid means-end incoherence in every case. It can, after all, be good to have a disposition that should not always be exercised. The same point undermines perhaps the most plausible version of the idea that WS is constitutive of intention, according to which it is the disposition to comply with WS that is necessary for having intentions.14 Finally, the strategy of taking WS as a basic normative principle also seems unpromising. One way to bring this out is to observe that WS covers only a very limited territory even within the general field of instrumental rationality and – a little more broadly – 12

Objection: complying with WS is necessary for being rational, which is plausibly something else you should do. Response: to be rational is, in part, to comply with rational requirements, such as WS. The claim that you should be rational thus presupposes, and so cannot explain, the claim that you should comply with WS. 13 Although note that even if D is false, complying with WS may be necessary for something else you should do. For instance, complying with WS might be necessary for having beliefs you should have [Wallace 2001, Setiya 2007]. 14 Furthermore, Kolodny [2008] makes a strong case that the disposition to comply with WS is neither instrumentally valuable nor necessary for having intentions.

7

requirements of coherence. This field includes the rationality of taking non-necessary means to your ends, having consistent intentions, intending to do what you believe you should do, and having consistent and coherent beliefs. If we follow the suggested strategy, we will likely end up positing basic normative principles covering each of these, at the least. But taking this path provides no illumination of what seems to be in common between these requirements. All we will have is a list of plausible sounding principles.15 Although these difficulties have received a lot of attention, I think there is still a tendency to underestimate their significance. In particular, a common reaction is to take them to show only that we need to distinguish between reasons and rationality. Although the terminology varies, the core idea here is quite familiar. While rationality depends on an agent’s mental states, especially her beliefs, reasons are facts of which an agent may be unaware. The standard example here is of Bernard Williams’ agent who justifiably believes that he has gin in his glass, when in fact it is petrol. This agent, it is suggested, would be rational to (intend to) take a sip, although he has no reason to do so (at least not on account of his false belief).16 Reasons and rationality come apart. If we accept this kind of picture, we might reformulate WS as a principle about what rationality requires, rather than about what you should or have reason to do [cf. Broome 2005]:

WSRR

You are rationally required not to [intend to E, believe that M-ing is necessary for E-ing, and not intend to M].

15

But see Hussain [unpublished] for defence of this sort of strategy. See Williams [1981: 102] for the example. For appeal to this sort of example to make this sort of point see e.g. Parfit [2011: 34], Schroeder [2009: 228-9], Setiya [2007: 659], Way [2010: 220], and Wedgwood [2003: 201-2]. But see Gibbons [2010] and Lord [2010] for challenges to this standard line. 16

8

And so understood, we might think that it is not so surprising that we cannot always find reasons to comply with the instrumental principle. The problems finding such reasons thus do not undermine the basic Wide-Scope idea that rationality requires means-end coherence. They serve only to remind us that we need to distinguish between reasons and rationality. However, while I am happy to accept the distinction between reasons and rationality, I think that this move can obscure the fundamental challenge these problems pose. As I see it, the challenge facing WS is one of explanation. What we wanted to know was why we should comply with WS. We were looking for reasons to comply only because, quite generally, we explain why you should do something by offering reasons to do that thing.17 What the problems rehearsed above suggest is thus that there is no way to explain WS. But if that is the right way to think about the force of these problems, the basic worry is only relocated, and not avoided, by the move to WSRR. For we can just as well ask why means-end coherence is rationally required, as we can ask why we should be means-end coherent. Nor is it clear that explaining why means-end coherence is rationally required will be substantially easier than explaining why we should be means-end coherent. For notice that Williams’ example does not merely motivate a distinction between reasons and rationality. It also suggests a picture of the relation between them. It is plausible to think that we should explain why it is rational for Williams’ agent to

17

I grant that this does not appear to be the motivation of some of those who have looked for reasons to comply with WS. For Broome [2005], for instance, we should think of rationality as a source of rational requirements, in something like the way in which etiquette and the rules of chess are sources of requirements. On this view, WSRR is presumably to be explained by appeal to this source. Broome’s motive for looking for reasons to comply with rational requirements seems to be to capture an intuitive distinction between sources like rationality, morality and prudence, on the one hand, and etiquette and the law, on the other. As I argue elsewhere though [Way 2009], capturing this distinction does not require that there is necessarily reason to comply with rational requirements.

9

intend to take a sip by appealing to the case in which, because there is gin and tonic in his glass, there is a reason to take a sip. And this suggests a general framework on which the rationality of actions and attitudes is determined by appeal to what there is reason to do in cases in which we have corrected for mistaken beliefs, and perhaps also for other epistemic shortfalls. However, if we accept a framework of this sort, then while we can expect reasons and rationality to come apart when an agent’s epistemic circumstances are less than ideal, we can also expect them to converge, when epistemic circumstances are ideal. But the problems finding reasons to comply with WS did not turn on imperfections in agents’ epistemic circumstances. So much the same problems that arose in finding reasons to comply with WS will arise in trying to explain WSRR.18 I thus take the real lesson of these problems to be not simply that we need to distinguish between reasons and rationality, but that both WS and WSRR resist explanation.

3. A Diagnosis

I now want to offer a diagnosis of why the Wide-Scope view struggles to meet this Explanatory Challenge. We can start by observing something surprising about the Wide-Scope view, which can be brought out by way of some Kantian terminology. Kant held that instrumental rationality is to be understood in terms of hypothetical imperatives: imperatives, or oughts, which apply conditionally on your ends. The contrast is with categorical imperatives, which are supposed to hold unconditionally. This difference meant, Kant argued, that categorical imperatives, such as the 18

I develop this argument in more detail in Way [2010: 220-2]. Compare also Kolodny [2007: 248], Setiya [2007: 657-63], Schroeder [2009: 231-2].

10

requirements of morality, raise explanatory challenges that the requirements of instrumental rationality do not:

[T]he question as to how the imperative of morality is possible is undoubtedly the only one requiring a solution. For it is not at all hypothetical; and hence the objective necessity which it presents cannot be based on any presupposition, as was the case with the hypothetical imperatives. [1785: 4: 419]

Since hypothetical imperatives are ‘based on a presupposition’, we can explain a hypothetical imperative by appealing to its presupposition. But categorical imperatives leave us no such resources. We should therefore expect categorical imperatives to be more difficult to explain than hypothetical imperatives. As Mark Schroeder has observed, Kant’s point here gives us insight into the problems the Wide-Scope view faces in meeting the Explanatory Challenge.19 For notice that, if WS or WSRR is correct, Kant turns out to have been wrong that instrumental rationality is to be understood in terms of hypothetical imperatives. WS, for instance, does not say that you should avoid means-end incoherence if you have a certain end. Rather it says that you should avoid means-end incoherence unconditionally. And so Kant’s point predicts that there will be difficulties in explaining WS – as we have found.20 It is important to notice that Kant’s point has nothing essential to do with the fact that the ‘presupposition’ of a hypothetical imperative is an end. Kant’s point turns 19

Schroeder [2004: 349, n.20], [2005b: 9]. See also Schroeder [2005a]. On some views, Kant himself accepted the Wide-Scope view [Hill 1973 and, perhaps, Hampton 1998: ch. 4]. If so, we should reject the standard understanding of the difference between hypothetical and categorical imperatives appealed to in the text. However, this would not undermine the crucial point that hypothetical imperatives, standardly understood, are more amenable to explanation than categorical imperatives. Rather, it would show that Kant was wrong to think that instrumental rationality is any easier to understand than the requirements of morality (Hampton [1998: 165-6] draws precisely this consequence). In any case, Schroeder [2005a] makes a strong case that Kant was not in fact a Wide-Scoper. 20

11

on a difference between oughts with limited application – oughts which apply only to some agents, or in some circumstances – and oughts with unlimited application – oughts which apply to all agents, in all circumstances. When there is something which someone, in some circumstance, ought to do, we can appeal to idiosyncratic features of that agent, or that circumstance, to explain why this agent ought to do that thing. And when there is something which everyone, in some circumstance, ought to do, then we can appeal to idiosyncratic features of this circumstance to explain why everyone ought to do this thing, in this circumstance. But, if there is something that everyone, in all circumstances, ought to do – when an ought has unlimited application – no such idiosyncratic features will do. So there will always be fewer resources to appeal to, to explain an ought with unlimited application. But WS is an ought with unlimited application – it says that everyone, in all circumstances, ought to avoid means-end incoherence. That is why the Explanatory Challenge is difficult.21

4. The Intermediate-Scope View

I now want to introduce an alternative view of the requirements of instrumental rationality. This view accepts the distinction between reasons and rationality, and makes claims about both. The view begins with the:

Reasons Claim*

If M-ing is necessary for E-ing, then you have conclusive reason not to [intend to E and not intend to M].

21

We might worry about this diagnosis on the following grounds. On certain views of the logic of ‘ought’ (or of ‘rationally required’), it follows from the claim that ‘necessarily, if p, you ought to A’ that ‘necessarily, you ought to [A, if p]’ (cf. n. 9). Given such a view, it will be no more difficult to explain a claim like WS than a claim like NS or D. However, since the Wide-Scoper’s distinctive claim is that WS is true while NS and D are false, this form of explanation is not open to Wide-Scopers. Given this point, it may perhaps be more accurate to say that what makes the Explanatory Challenge difficult for Wide-Scopers is the underived unlimited application of WS.

12

This claim will require qualification in the next section, but we can see its initial plausibility by considering the sort of case with which we began.22 When you discover that today is the last day to see the exhibition, you are faced with a choice – go today, or give up your intention to see the exhibition. What you should not do is continue to intend to see the exhibition but not form the intention to go today.23 The Reasons Claim* does not tell us that something is wrong with all cases of means-end incoherence, since it does not apply in cases where the means-end belief is false. However, given the sort of connection between reasons and rationality suggested by Williams’ petrol and tonic example, it is natural to take the Reasons Claim* to support the:

Rationality Claim

If you believe that M-ing is necessary for E-ing, then you are rationally required not to [intend to E and not intend to M].

The Rationality Claim does tell us that something is wrong with all cases of meansend incoherence. The means-end incoherent agent fails to combine her intentions in the way that rationality requires, given her means-end belief. Since the Reasons Claim* and Rationality Claim are wide-scope with respect to the combination of intentions, and narrow-scope with respect to the means-end fact or belief, we may call the combination of these claims the Intermediate-Scope view. The Intermediate-Scope view has many of the advantages of the Wide-Scope view over NS and the Disjunctive view. Unlike NS, the view does not imply that

22

Throughout the paper, I asterisk claims which require revision or rejection. Previous putative instrumental principles have included the implicit condition that you believe that intending to M is necessary for M-ing (see section one). Here the implicit condition is instead that intending to M is necessary for M-ing. 23

13

merely intending an end ensures that you should – or are even rationally required – to intend the believed means to that end. And unlike the Disjunctive view, the view does not rule out cases of mere permissibility, or deny that means-end incoherence is a distinctive form of irrationality. Furthermore, the view has these advantages over NS and the Disjunctive view by very similar means to the Wide-Scope view. It has these advantages because, like the Wide-Scope view, and unlike NS or the Disjunctive view, the Intermediate-Scope view sees the instrumental principle as a constraint on a combination of attitudes – on intending the end without also intending the believed necessary means.24 However, there is a crucial way in which the Intermediate Scope view differs from the Wide-Scope view. Neither the Reasons Claim* nor the Rationality Claim have unlimited application. Instances of the Reasons Claim* apply only to agents for whom a means is necessary for an end, and instances of the Rationality Claim apply only to agents with the relevant means-end belief. So if the diagnosis of section three is correct, we should expect the Intermediate Scope view to be in a better position to meet the Explanatory Challenge. We have already seen some evidence that this is the case. If reasons and rationality are connected in the way suggested by Williams’ petrol and tonic example, it looks plausible that the Reasons Claim* can help to explain the Rationality Claim. Here is one way this explanation might go. Derek Parfit [2011: 34], among others, suggests that reasons and rationality are connected in the following way:

24

These similarities might lead us to call the Intermediate-Scope view a version of the Wide-Scope view, as I did in Way [2010]. Note also that some will think that the Rationality Claim entails WSRR (cf. n.9). I leave this issue open.

14

Rational Requirement*

You are rationally required to A if you believe some things whose truth would give you conclusive reason to A.

If Parfit is right, we can explain the Rationality Claim in terms of the Reasons Claim*. This is because if you believe that M-ing is necessary for E-ing then you have a belief whose truth, according to the Reasons Claim*, would give you conclusive reason not to [intend to E and not intend to M]. So Rational Requirement* implies that you are rationally required not to [intend to E and not intend to M] – as the Rationality Claim asserts.25 The move from the Wide-Scope view to the Intermediate Scope view should look promising, then. Nonetheless, the Intermediate-Scoper’s claim to meet the Explanatory Challenge is unsatisfactory as it stands. First, as I shall explain below, both the Reasons Claim* and Rational Requirement* are subject to counter-examples. Second, even if suitably modified versions of these claims do explain the Rationality Claim, we may seem only to have pushed the Explanatory Challenge back a step. A qualified version of the Reasons Claim* will still postulate a strong reason against intending the end without also intending the necessary means. And so we will still face the task of explaining this reason. The rest of this paper aims to meet these challenges. In sections five and six, I argue that both the Rationality Claim and a suitably qualified version of the Reasons Claim* are true if the following, much weaker, claim is true: 25

Some will object to Rational Requirement* on the grounds that it allows irrational beliefs to ground rational requirements. Suppose, for example, that paranoia leads you to irrationally believe that your lunch has been poisoned. If your lunch was poisoned, you would have most reason to intend to throw it away. But many will find it implausible that you are thereby rationally required to intend to throw your lunch away [cf. Wedgwood 2003: 203; contrast Parfit 2011: 112-7]. I want to remain neutral on this issue. Readers convinced by the objection can substitute ‘rationally believe’ for ‘believe’ in Rational Requirement*, and accept a similarly restricted version of the Rationality Claim.

15

Basic Claim If M-ing is necessary for E-ing, there is a reason not to [intend to E and not intend to M].

In section seven, I sketch an explanation of the Basic Claim. I thus hope to show at least that there are grounds to be optimistic that the Intermediate-Scope view can meet the Explanatory Challenge.

5. Explaining the Reasons Claim

There is conclusive reason to A if the reasons to A outweigh the reasons not to A. Thus – since one way this can happen is if there is a reason to A and no reason not to A – we can move from the Basic Claim to the Reasons Claim* if the following is true:

No Defeaters*

If M-ing is necessary for E-ing, then there is no reason to [intend to E and not intend to M].

In this section, I shall argue that a qualified version of No Defeaters* is true. Since we are assuming the Basic Claim, this will show that a qualified version of the Reasons Claim* is true. It will be helpful to begin by thinking in general about when there are reasons to [intend to A and not intend to B], where A and B are act-types. This will allow us to assess what reasons there might be to [intend to E and not intend to M], when Ming is necessary for E-ing. I suggest that there are two main possibilities.

16

Type 1 Cases

There is a reason to [A and not B] – that is, to A without also B-ing. If you intend both to A and to B you will both A and B. So you have a reason to [intend to A and not intend to B]. For example, you have a reason to change lanes without accelerating. If you intend both to change lanes and to accelerate, you will do both. So you have a reason to [intend to change lanes and not intend to accelerate].

Type 2 Cases

There are benefits of [intending to A and not intending to B] other than that you will thereby A without B-ing. These benefits give you a reason to [intend to A and not intend to B].

If these are the only possibilities, then No Defeaters* will be false if there are situations in which M-ing is necessary for E-ing but you have a reason to E without M-ing (Type 1) or there are other reason-providing benefits of [intending to E and not intending to M] (Type 2).26 The first possibility can, I think, be ruled out. When M-ing is necessary for Eing you cannot E without M-ing. So, on the assumption that ‘reason’ implies ‘can’, you cannot have reason to E without M-ing.27 The second possibility is more problematic. It can be illustrated by a variant of Kavka’s [1983] toxin puzzle. In Kavka’s example, an eccentric billionaire offers you $1m if you intend, at midnight tonight, to drink an unpleasant toxin at noon tomorrow. To get a Type 2 case, we need only add a further condition on winning the

26

There may seem to be a third possibility. In cases in which (now) intending to M is not necessary for M-ing, reasons to E may be reasons to [intend to E and not intend to M]. However, as noted in n. 23, we are restricting our attention to cases in which (now) intending to M is necessary for M-ing. 27 For defences of ‘reason’ implies ‘can’ see Streumer [2007] and Vranas [2007].

17

$1m: that you do not also intend the necessary means to drinking the toxin. You then seem to have a reason to intend the end of drinking the toxin without also intending the means. That is a counter-example to No Defeaters*. Some may be tempted to respond by denying that the $1m prize provides a reason for this combination of intentions. A number of philosophers insist that, in the original toxin puzzle, the $1m prize is at most a reason to want to intend to drink the toxin, and to bring it about that you have this intention.28 Presumably, these philosophers would also deny that the $1m is a reason to intend to drink the toxin without intending the necessary means, in our modified version. However, a more concessive response is available. We can grant that Type 2 cases are counterexamples to No Defeaters* but insist that – in a sense to be explained – they are examples of reasons of the wrong kind. They thus do not threaten the following restriction of No Defeaters*:

No Defeaters

If M-ing is necessary for E-ing, then there is no reason of the right kind to [intend to E and not intend to M].

The distinction between reasons of the right and wrong kind is a way of marking the very similar contrasts we see between two kinds of reasons for intention, belief, and other attitudes. In the case of intention, the right kind of reasons to intend derive from reasons to perform the action intended. The wrong kind of reasons derive from other benefits of intending – the $1m prize in the original toxin puzzle is an example. In the case of belief, the right kind of reasons to believe are evidential, the wrong kind of reasons are pragmatic – Pascal’s wager is a familiar example of the latter. These

28

For example, Pink [1996: 147-52], Shah [2008: 14-16].

18

distinctions share some common marks. First, and most strikingly, there is a clear asymmetry in the ease and way in which you can respond to the two kinds of reason. For example, you can form intentions directly on the basis of reasons for action. But you cannot in the same way decide to drink the toxin because so intending will earn you $1m, at least when you recognize that there is no good reason to drink the toxin. Similarly, while you can believe on account of the evidence, you cannot in the same way believe because believing will have pragmatic benefits. Second, there are types of overall evaluation to which only the right kind of reasons seem relevant. Of central relevance here is the fact that only reasons of the right kind are relevant to rationality. For example, the benefits of believing in God are not themselves enough to make it rational to believe in God. Nor does the $1m prize make it rational to intend to drink the toxin. It seems to be rational to intend only if the action intended is, as things stand, rational to perform.29 The distinction between Type 1 and Type 2 cases of reasons for a combination of intentions looks to be a further illustration of this distinction. Type 1 cases look like cases of the right kind of reasons. Type 2 cases look like cases of the wrong kind of reasons. Type 2 cases are thus no threat to No Defeaters. Of course, No Defeaters does not support the Reasons Claim*. However, if the reason specified in the Basic Claim is of the right kind, No Defeaters does support the:

29

The terminology of the ‘right’ and ‘wrong’ kind of reasons is most familiar from discussions of fitting-attitude accounts of value (e.g. Rabinowicz and Rønnow-Rasmussen [2004]). In that context, it is used to mark a distinction between types of reasons for valuing. In extending this terminology to reasons for belief and intention, I am following Hieronymi [2005] and Schroeder [2010], among others.

19

Reasons Claim

If M-ing is necessary for E-ing, then the right kind of reasons not to [intend to E and not intend to M] outweigh the right kind of reasons to [intend to E and not intend to M].

And the Reasons Claim, I suggest, is all that the Intermediate-Scope view should endorse. For just as Type 2 cases count against No Defeaters*, they also count against the Reasons Claim*. So nothing stronger than the Reasons Claim is plausible, in any case. What is more, since our central interest is in the irrationality of means-end incoherence, and since only reasons of the right kind are relevant to rationality, we need not worry that there may be reasons of the wrong kind in favour of means-end incoherence. We have now seen how to explain the Reasons Claim, granted the truth of the Basic Claim. In the next section, I explain how this argument can be reapplied to explain the Rationality Claim.

6. From the Reasons Claim to the Rationality Claim

The Rationality Claim follows from the Reasons Claim along with the following, slightly modified, version of Parfit’s connection between reasons and rationality:

Rational Requirement**

You are rationally required to A if you believe some things such that, if they are true, the right kind of reasons to A outweigh the right kind of reasons not to A.

20

Unfortunately though, Rational Requirement** is subject to counter-example.30 Consider the following variation on Williams’ gin and petrol example. Suppose Williams’ agent does indeed have gin in his glass, and suppose that because of this he has most reason to take a sip (and so most reason of the right kind to intend to take a sip). But suppose that he also believes, falsely, that he will be drinking wine later, and is concerned about mixing his drinks. It might then be irrational for him to take a sip, although he believes something which is true and which makes it the case that his reasons of the right kind favour intending to take a sip.31 We therefore need a more plausible view of the connection between reasons and rationality. In the rest of this section, I sketch such an account, and then argue that the new account still allows us to move from the Reasons Claim to the Rationality Claim. As a first step, we need to distinguish between objective and subjective reasons. Objective reasons are what I have thus far meant by ‘reasons’ – facts which count for or against an action or attitude, and which together determine what you should do. As a heuristic – not a definition – we can say that subjective reasons are considerations which it is rational for an agent to treat as objective reasons in deciding what to do. For example, in Williams’ original case, it is rational for the agent to take the content of his belief that there is gin in his glass into consideration in deciding what to do. However, it is not true that there is gin in his glass, and so this

30

The same example shows, mutatis mutandis, that Rational Requirement* is false. Like that principle, Rational Requirement** also raises worries about rational requirements being grounded in irrational beliefs (cf. n.25). As before, such worries may be met by restricting the antecedent of the principle to rational beliefs. 31 This problem could be avoided by saying that what you are rationally required to do is what you would have most reason to do if all your beliefs were true. However, this move is highly problematic. First, if you have inconsistent beliefs, it is not possible for all your beliefs to be true. Second, certain cases involving ignorance (e.g. Jacob Ross’s ‘three-envelope’ case) are often thought to show that what you are rationally required to do can come apart from what you have most reason to do even when all your beliefs are true. See, e.g. Schroeder [2009: 243-5].

21

consideration is not an objective reason to take sip. Rather, it is a subjective reason to do so. The distinction between objective and subjective reasons is the analogue, at the pro tanto level, of the distinction at the overall level between what you should do and what you are rationally required to do. At both levels, we can distinguish between a normative relation which turns on the facts of an agent’s situation, and a normative relation which turns on the agent’s beliefs about her situation. It seems clear that if we are to draw the latter distinction, we should also draw the former. However, once we have the notion of a subjective reason, it is also very natural to think that what you are rationally required to do is determined by the balance of (the right kind of) subjective reasons – just as what you ought to do is determined by the balance of objective reasons. For example:

Rational Requirement

You are rationally required to A if the subjective reasons (of the right kind) to A outweigh the subjective reasons (of the right kind) not to A.

This account is not subject to the counter-example to Rational Requirement**. In the modified case, Williams’ agent plausibly has a subjective reason to take a sip, since he believes that the glass contains gin. But he also has a subjective reason not to take a sip, because he believes that he will drinking wine later. And although more would need to be said about the weight of subjective reasons in order to establish this, it is plausible that the latter reason outweighs the former, so that he is rationally required not to take a sip.32

32

See Schroeder [2009: 242-5] for further motivations for the sort of picture just sketched.

22

We have now seen why Rational Requirement is a plausible improvement on Rational Requirement**. I shall now argue that it still allows us to explain the Rationality Claim. The explanation to be given is structurally parallel to the explanation of the Reasons Claim. One way for the subjective reasons to A to outweigh the subjective reasons not to A is for there to be a subjective reason to A and no subjective reason not to A. So we can explain the Rationality Claim by defending the following two claims:

Subjective Basic Claim

When you believe that M-ing is necessary for E-ing, you have a subjective reason (of the right kind) not to [intend to E and not intend to M].

No Subjective Defeaters

When you believe that M-ing is necessary for E-ing, you have no subjective reason (of the right kind) to [intend to E and not intend to M].

Since subjective reasons are considerations which it is rational to treat as objective reasons, it is natural to expect these claims to be true if, as we are assuming, the Basic Claim is true and if, as I argued in section five, No Defeaters is true. I shall now argue that this expectation is confirmed.33 A quick way to argue for the Subjective Basic Claim and for No Subjective Defeaters is to appeal to the account of subjective reasons offered by Parfit [2011: 35] and Schroeder [2009: 233]:

33

Again, some readers will want to substitute ‘rationally believe’ for ‘believe’ in these principles, and in SubjectiveSUF and SubjectiveNEC (see below), so as to yield a similarly restricted version of the Rationality Claim (cf. n.25). Whether or not we should make these revisions is a question that requires further attention.

23

Subjective*

You have a subjective reason to A just in case you believe something which, if true, would give you an objective reason to A.

Subjective* implies the Subjective Basic Claim. For if you believe that M-ing is necessary for E-ing, you believe something which, by the Basic Claim, would give you an objective reason not to [intend to E and not intend to M], if it is true. And No Subjective Defeaters seems to follow too, since, if the argument of section five is right, there is nothing you can believe which, if true, would be an objective reason (of the right kind) to [intend to E and not intend to M]. Unfortunately, Subjective* is open to similar counter-examples to Rational Requirement**. For example, suppose you believe that you promised to A, and also believe that this promise was coerced. It is at least arguable that it is not rational to treat a coerced promise as a reason; if that is right, the content of your belief that you promised to A is not a subjective reason to A. However, if it is true that you promised, but not true that you were coerced, then you do satisfy the right-hand side of Subjective*. Subjective* appears to be false. I do not have a general solution to this problem. For present purposes though, we can work around it. I shall offer one sufficient and one necessary condition for subjective reasons. And I shall show that these conditions are enough to defend the Subjective Basic Claim and No Subjective Defeaters. The above counter-example to Subjective* rests on the possibility of a consideration which is a reason to do something in one circumstance, but not a reason to do that thing in other circumstances. However, not all reasons are of this sort –

24

some considerations are reasons to A in every circumstance in which they occur.34 So the problem does not threaten the following sufficient condition for subjective reasons:

SubjectiveSUF

If you believe something which, if true, would give you a reason to A in any circumstance in which you can A, and you do not believe that you cannot A, then you have a subjective reason to A.

However, according to the Basic Claim, in any circumstance in which M-ing is necessary for E-ing, there is a reason not to [intend to E and not intend to M].35 Thus SubjectiveSUF and the Basic Claim are enough to support the Subjective Basic Claim. We can take a similar approach to defending No Subjective Defeaters. Any subjective reason of the right kind to [intend to E and not intend to M] would have to derive from a subjective reason of the right kind to [E and not M]. However, consider the following necessary condition on subjective reasons:

SubjectiveNEC

You have a subjective reason to A only if you do not believe that you cannot A.

SubjectiveNEC is an analogue for subjective reasons of ‘reason’ implies ‘can’. It seems a plausible condition. Subjective reasons are considerations rational agents take into account when deciding what to do. But when deciding what to do, rational agents 34

More exactly, some considerations are reasons to A in every circumstance in which you can A. This qualification is required since we are assuming that ‘reason’ implies ‘can’. Note also that even holists can allow that some reasons are of this sort [Dancy 2004: 77-8]. 35 Or more exactly, any circumstance in which M-ing is necessary for E-ing and (now) intending to M is necessary for M-ing. See n. 23.

25

consider only options they believe available to them. So when deciding what to do, rational agents do not take into account considerations favouring actions they believe they cannot do.36 SubjectiveNEC implies that No Subjective Defeaters is true. If you believe that M-ing is necessary for E-ing, you believe that you cannot [E and not M]. So by SubjectiveNEC, you have no subjective reason to [E and not M]. And since the right kind of reasons to intend derive from reasons to act, you have no subjective reason of the right kind to [intend to E and not intend to M], either. I have now argued that No Subjective Defeaters is true, and that if the Basic Claim is true, the Subjective Basic Claim is true, too. I have also argued that what you are rationally required to do is determined by your subjective reasons. Taken together, these claims imply that if the Basic Claim is true, the Rationality Claim is true.

7. Explaining the Basic Claim

In the previous two sections, I have argued that if the Basic Claim is true, we can explain both the Reasons Claim and the Rationality Claim. This is a significant result, for the Basic Claim is very weak. If meeting the Explanatory Challenge requires only that we explain the Basic Claim, then we should be much more confident that the Explanatory Challenge can be met. Nonetheless, there is an argument that purports to show that the Basic Claim cannot be true. The first task of this section is to dispel this argument. Doing so will also point us towards an account of why the Basic Claim is true.

36

Note that this is not to say that a rational agent will never believe that she has a reason to do something she believes she cannot do. It is not necessarily irrational to deny that ‘reason’ implies ‘can’. However, even if you deny this, it is irrational to decide to do what you believe you cannot do.

26

The Basic Claim says that when M-ing is necessary for E-ing, there is a reason (of the right kind) not to [intend to E and not intend to M]. However, we have been assuming that reasons (of the right kind – henceforth I assume this qualification) to intend to A depend on reasons to A. It may be thought that a corollary of this purist view is that reasons against intending to A depend on reasons against A-ing. And it seems to follow from this apparent corollary that any reason against [intending to E and not intending to M] must derive from a reason against [E-ing and not M-ing]. Unfortunately though, it is not at all clear that there are any such reasons. For when M-ing is necessary for E-ing, you cannot [E and not M]. So any reason against [E-ing and not M-ing] would be a reason that you cannot fail to comply with. But at least according to many philosophers, there cannot be reasons that you cannot fail to comply with.37 So it seems as if the Basic Claim cannot be true. My response to this argument shall be to challenge the assumption that reasons against intending to A must derive from reasons against A-ing. I shall argue that proponents of purism about reasons for intending should in fact deny that all reasons against intending derive from reasons against acting. If this is right, the argument against the Basic Claim collapses. A central motivation for purism about reasons for intending is the appeal of the following claim, linking permissible intention to permissibly acting as intended:

Link

It is permissible now to intend to A later only if, in the absence of relevant new information, it will be permissible to A later.38

37

See Lavin [2004] for extended discussion and references. ‘Permissible’ here is not a moral notion. Rather, it is simply the dual of the overall ‘should’. I shall also use ‘okay’ to express this notion. 38

27

The intuition behind Link is that the central point of intending is to move us to do things that are worth doing. If Link is false, it is not clear how intentions could serve this purpose. And Link supports purism, since if there are reasons to intend other than reasons to act, those reasons could make it okay to intend even when it is not, as things stand, okay to act.39 The same intuition that supports Link also supports the idea that reasons to act provide reasons to intend only when intending makes a difference to whether or not you will act as intended. If you are already certain to A – for instance, because A-ing is a side-effect of something else you intend to do – reasons to A do not provide you with reasons to intend to A. This suggests that we should understand purism about reasons for intending in the following way:

P+

R is a reason (of the right kind) to intend to A just in case R is a reason to A and intending to A promotes A-ing.

And if all reasons against intending depend on reasons against acting we have:

P-

R is a reason (of the right kind) not to intend to A just in case R is a reason not to A and intending to A promotes A-ing.

However, we now have a problem. P+ and P- have been motivated by the appeal of Link, and the underlying intuition that the point of intending is to move us to do things that are worth doing. But P+ and P- do not in fact vindicate Link, or the underlying intuition. To see this, consider a case in which intending to A does not

39

See Pink [1996, especially chs. 4 and 5], for extensive discussion.

28

promote A-ing. In such a case, P+ and P- imply that there is no reason to intend to A, and no reason not to intend to A. So in such a case, P+ and P- imply that it is okay to intend to A. And P+ and P- imply this quite independently of the reasons for and against A-ing, and so quite independently of whether it is okay to A. So P+ and P- do not guarantee Link. We can avoid this problem by denying P- and so allowing that considerations other than reasons not to A can provide reasons against intending. If there are reasons of this sort, then when intending to A does not promote A-ing, these reasons will tip the balance, and so it will not be okay to intend to A. Such reasons would not threaten Link. Link is secure so long as the permissibility of A-ing is a necessary condition on permissibly intending to A. But to deny P- is not to reject this necessary condition. Rather, it is to place a further necessary condition on permissible intention. On a view which accepts P+ but denies P-, it is okay to intend to A only if there is good enough reason to A, intending to A promotes A-ing, and there are no sufficiently weighty further reasons against intending to A. What might these further reasons be? To answer this question, it is helpful to recall some of the features of intention emphasised by Michael Bratman [1987]. When you intend to A you are disposed to reason towards A-ing. This requires you to monitor your prospects for A-ing, and to be disposed to take suitable means as and when required. At least in many cases, intending also involves being disposed to reason on the assumption that you will A – for example, not deliberating further about whether to A, refraining from planning to do anything incompatible with A-ing, and so on.

29

Bratman and others have emphasized that there are great benefits to having a capacity for attitudes of this sort. The capacity to settle ahead of time what we will do allows us to take advantage of good conditions for deliberation, and facilitates interand intra-personal co-ordination [Bratman 1987: especially chs. 1 and 2; Holton 2009: ch.1]. The point I want to make is that intending nonetheless involves a kind of cost to the agent, on account of these features. That intending to A requires you to monitor your progress towards A-ing, and to be disposed to take suitable means if required means that intending is a way of tying up psychological resources – resources which are limited, and for which there is always another use. This means, I suggest, that there is always some reason against intending to do something. Furthermore, it seems plausible that reasons of this sort are relevant to rationality, and so are of the right kind. Rational agents display a sensitivity to these costs of intending – that is why rational agents do not in general intend to do things they expect to do anyway, or things they know they will not do even if they intend to.40 Rational agents do not knowingly waste the use of psychological resources intending involves. So there seem to be good grounds for accepting a view which is purist with respect to reasons for but not against intention. Accepting such a view undermines the argument that the Basic Claim cannot be true. We need not think that any reason against [intending to E and not intending to M] must derive from a reason not to [E and not M]. It might be that when M-ing is necessary for E-ing, there are costs of [intending to E and not intending to M] that provide reasons against that combination of intentions.

40

Note that we cannot explain why rational agents do not intend to do such things simply by noting that there are reasons to intend to A only if intending to A promotes A-ing. This only explains why there is no reason to intend to A. But it can be okay to do something if there is no reason to do it, so long as there is also no reason not to do it.

30

In fact, we have already seen enough to see that this is plausible. I have suggested that the use of psychological resources in intending means that there is always some cost of, and so reason against, any intention. These costs will also provide some reason against [intending to E but not intending to M] unless the lack of an intention to M cancels out these costs. But clearly that is not so – the lack of an intention to M does not stop someone who intends to E from monitoring his progress in E-ing or from disposing him to take means as required.41 And so it seems plausible that the use of psychological resources in intending will, quite generally, provide some reason against [intending to E and not intending to M], when M-ing is necessary for E-ing. And so the Basic Claim will be true.42

8. Conclusion

In the previous three sections, I have tried to show how the Intermediate-Scope view can meet the Explanatory Challenge. The Intermediate-Scope view makes two main claims:

Reasons Claim

If M-ing is necessary for E-ing, then the reasons of the right kind not to [intend to E and not intend to M] outweigh the reasons of the right kind to [intend to E and not intend to M].

41

This is especially clear in a case in which the agent does not know or believe that M-ing is necessary for E-ing. 42 The point here is similar to the suggestion made in unpublished work by John Brunero and by Niko Kolodny that there is reason against [intending to E and not intending to M] because this combination makes you more likely to take costly and ineffective means to E. (For example, intending to visit Chicago without intending to book a flight may make me likely to waste money booking a hotel). However, while I agree that in very many cases there will be reasons of this sort, I worry that this will not be so in all cases. It seems only to be contingent that intending an end risks taking costly means, while it is plausibly necessary that means-end incoherence is irrational. For this reason, I prefer to appeal to costs that are plausibly a necessary part of intending.

31

Rationality Claim

If you believe that M-ing is necessary for E-ing, then you are rationally required not to [intend to E and not intend to M].

I have argued that these claims are true if the Basic Claim is true, and that the Basic Claim is indeed true. My strategy has been to show how these claims are supported by independently plausible views, notably: the view that rationality is to be understood in terms of subjective reasons, which are in turn to be understood in terms of objective reasons, the principle that ‘reason’ implies ‘can’, and purism about reasons for, but not against, intending. Furthermore, the conditions of both the Reasons Claim and the Rationality Claim have played a crucial role in these arguments. I therefore take the results of the second half of the paper to support the diagnosis offered in the first half. The Explanatory Challenge to the Wide-Scope view is intractable because of the unlimited application of Wide-Scope principles.43

References Bratman, Michael 1987. Intentions, Plans and Practical Reason, Cambridge, MA: Harvard University Press. Broome, John 1999. Normative Requirements, Ratio 12/4: 398-419. Broome, John 2005. Does Rationality Give Us Reasons? Philosophical Issues 15/1: 321-37. Dancy, Jonathan 2000. Practical Reality, Oxford: Oxford University Press. Dancy, Jonathan 2004. Ethics without Principles, Oxford: Oxford University Press. Darwall, Stephen 1983. Impartial Reason, Ithaca: Cornell University Press.

43

I would like to thank Kevin Falvey and an audience at the University of Stirling for discussion of a very early version of this paper, and Errol Lord and Mark Schroeder for written comments on a more recent version. I am especially grateful to two anonymous referees for the Australasian Journal of Philosophy, whose detailed, challenging, and constructive comments lead to very many improvements.

32

Finlay, Stephen 2009. Against All Reason? Scepticism about the Instrumental Norm, in Hume on Motivation and Virtue, ed. Charles Pigden, Basingstoke: Palgrave McMillan: 155-178 Gibbons, John 2010. Things That Make Things Reasonable, Philosophy and Phenomenological Research 81/2: 335-61. Hampton, Jean 1998. The Authority of Reason, Oxford: Oxford University Press. Hieronymi, Pamela 2005. The Wrong Kind of Reason, Journal of Philosophy 102/9: 437-57. Hill, Thomas 1973. The Hypothetical Imperative, Philosophical Review 82/4: 429-50. Holton, Richard 2009. Willing, Wanting, Waiting, Oxford: Oxford University Press. Hussain, Nadeem unpublished. The Requirements of Rationality. Kant, Immanuel 1785 (1993). Grounding for the Metaphysics of Morals, trans. James W. Ellington, Indianapolis: Hackett Publishing. Kavka, Gregory 1983. The Toxin Puzzle, Analysis 43/1: 33-36. Kolodny, Niko 2005. Why Be Rational? Mind 114/455: 509-63. Kolodny, Niko 2007. How Does Coherence Matter? Proceedings of the Aristotelian Society, 107/1: 229–263. Kolodny, Niko 2008. Why be Disposed to be Coherent? Ethics 118/3: 437-63. Lavin, Douglas 2004. Practical Reason and the Possibility of Error, Ethics 114/3: 42457. Lord, Errol 2010. Having Reasons and the Factoring Account, Philosophical Studies 149/3: 283-96. Parfit, Derek 2011. On What Matters, Volume One, Oxford: Oxford University Press. Pink, Thomas 1996. The Psychology of Freedom, Cambridge: Cambridge University Press. Rabinowicz, Wlodek and Ronnow-Rasmussen, Toni 2004. The Strike of the Demon: On Fitting Pro-Attitudes and Value, Ethics 114/3: 391-423. Raz, Joseph, 2005. The Myth of Instrumental Rationality, Journal of Ethics and Social Philosophy 1/1. www.jesp.org. Schroeder, Mark 2004. The Scope of Instrumental Reason, Philosophical Perspectives 18/1: 337-64. Schroeder, Mark 2005a. The Hypothetical Imperative? Australasian Journal of Philosophy 83/3: 357-372.

33

Schroeder, Mark 2005b. Instrumental Mythology, Journal of Ethics and Social Philosophy, symposium 1. www.jesp.org. Schroeder, Mark 2009. Means-End Coherence, Stringency, and Subjective Reasons, Philosophical Studies 143: 223-248. Schroeder, Mark 2010. Value and the Right Kind of Reason, in Oxford Studies in Metaethics Volume 5, ed. Russ Shafer-Landau, Oxford: Oxford University Press: 2555. Setiya, Kieran 2007. Cognitivism About Instrumental Reason, Ethics 117/4: 649-73. Shah, Nishi 2008. How Action Governs Intention, Philosophers’ Imprint 8/5. Southwood, Nicholas 2008. Vindicating the Normativity of Rationality, Ethics 119/1: 1-30. Streumer, Bart 2007. Reasons and Impossibility, Philosophical Studies 136/3: 351-84. Vranas, Peter 2007. I Ought, Therefore I Can, Philosophical Studies 136/2: 167-216. Wallace, R. Jay 2001. Normativity, Commitment and Instrumental Reason, Philosophers’ Imprint 1/3. Way, Jonathan 2009. Two Accounts of the Normativity of Rationality. Journal of Ethics and Social Philosophy, www.jesp.org. Way, Jonathan 2010. Defending the Wide-Scope Approach to Instrumental Reason, Philosophical Studies, 147/2: 213-33. Wedgwood, Ralph 2003. Choosing Rationally and Choosing Correctly, in Weakness of Will and Varieties of Practical Irrationality, ed. Sarah Stroud and Christine Tappolet, Oxford: Oxford University Press: 201-29. Williams, Bernard 1981. Internal and External Reasons, in his Moral Luck, Cambridge: Cambridge University Press: 100-110.

34