According to some philosophers, there can be no normative reason for an agent to perform an action unless she can be motivated by – or come to act for – that reason (Bond 1983: 7; Brandt 1979; Goldman 2009; Korsgaard 1986; Williams 1981, 1989; Wong 2006: 540). This claim is of both substantive and theoretical interest. First, if true it places substantial constraints on the sorts of things which can provide reasons for agents – it seems to rule out, for example, reasons being provided by considerations which are ungraspable by relevant agents. Second, this claim has been prominently deployed in an argument for a particular view about the nature of reasons, namely that they are always grounded in elements of the subjective motivational set of the agent whose reasons they are (Williams 1981).

It is not surprising, therefore, that the claim has generated considerable philosophical controversy. In particular, Schroeder (2007) has suggested an apparent counterexample of an agent who has a normative reason, but is incapable of coming to act for this reason. Herein I argue that this proposed counterexample fails, given the availability of plausible ‘expansive’ accounts of what it is to act for a normative reason. I first elaborate the details of Schroeder’s example to provide a scenario where it seems, after all, possible for the agent to act for his normative reason (§§II–IV). Building on this case, I offer two substantive accounts of acting for a normative reason according to which the protagonist in Schroeder’s example can act for his reason (§§V-VI). I then argue that these accounts are consistent with one of the key motivations for the initial claim, namely the thought that reasons are for reasoning with (§VII). The conclusion is that the apparent counterexample serves to refine, not refute, the view that there is an interesting necessary connection between normative reasons and the possibility of acting for those reasons.

I.

What is the connection between normative reasons and motivation? A normative reason is one that in some way justifies an action: it is a consideration in favour of that action (Scanlon 1998: 19). Consider the following view of the connection between such reasons and motivation:

(1) ‘If it is true that A has a [normative] reason to Φ then it must be possible that he should Φ for that reason.’ (Williams 1989: 39; cf. 1981: 106)

(1) is underspecified. One thing we need to know to evaluate it is the type of possibility involved. (1) is sometimes understood as employing a notion of rational possibility, where A Φing for a reason is rationally possible just in case A would Φ for that reason were she to rationally deliberate (Shafer-Landau 2003: 172). I discuss this interpretation below (§§III-IV). But there is a weaker interpretation. On this view, the relevant possibility is logical, so (1) claims that A has a reason to Φ only if the idea of A Φing for that reason involves no contradiction. Yet even this claim has an apparent counterexample. Consider:

Nate loves successful surprise parties thrown in his honor, but cannot stand unsuccessful surprise parties. If there is an unsuspected surprise party waiting for Nate in the living room, then plausibly there is a reason for Nate to go into the living room...But it is simply impossible to motivate Nate to go into the living room for this reason – for as soon as you tell him about it, it will go away. (Schroeder 2007: 165, cf. 33–34, 125–128.)

Nate has a reason to go into the living room. In order to motivate Nate to go into the living room for this reason it needs (it seems) to be the case both that there is a surprise party waiting for Nate there and that Nate recognise that there is such a party. But a party cannot, as a matter of conceptual necessity, be both a surprise for someone and known to them. Hence Nate cannot be motivated to go into the living room for this reason. Yet it is still true that Nate has this reason. So (1) is false.

II.

But this is perhaps too quick. I noted that one way in which (1) is underspecified concerns the type of possibility deployed. But another relevant point of contention concerns what it is for A to Φ for a normative reason. This raises the possibility that we might reply to the apparent counterexample by denying that it is impossible for Nate to go into the living room for his normative reason. Some commentators have labelled reasons such as Nate’s ‘elusive reasons’, since they are ‘considerations which are reasons for some agent but would cease to be reasons if the agent became sufficiently aware of them’ (Ridge and McKeever 2012: 112). Elusive reasons are reasons which are at existential peril of agents reasoning with them. My more general claim therefore, will be that there are some plausible accounts of acting for normative reasons according to which it is possible for agents to act for elusive reasons.

To see how this might work, it is necessary first to consider the logical form of normative reason statements. Whenever A has a normative reason to Φ there is some feature, F, of A’s situation and her options that provides (or constitutes) her reason. Thus a perspicuous form of reason statements is ‘F is a reason for A to Φ’ (Schroeder 2007: 15–21). Nate, for example, has a reason to go into the living room and that there is a surprise party waiting for him there provides this reason.

F’s – for example that there is a surprise party waiting – are sometimes called simply ‘reasons’, but may also be described as ‘reason-giving features’ or ‘reason-givers’. They are sometimes taken to be (sets of) facts or true propositions (Ridge and McKeever 2012; Schroeder 2007: 20; Skorupski 2009: 115) but here I take them to be intensional facts, that is, facts (parts of the fabric of the world) under a particular mode of presentation. On this view, substitution of co-referring terms in the F-position of a normative reason statement does not guarantee preservation of truth-value. So, for example, that I promised to carry your bags is an (‘agent-relative’) reason for me to carry your bags, but the same fact described in microphysical terms, for example, is not. In this respect at least (and possibly others) normative reason statements are intensional contexts.Footnote 1

Now, what is impossible in Nate’s case is that he recognises the relevant reason-giving feature. This is because that feature is specified partly in terms of its not being recognised. This makes Nate a counterexample to (1) only if acting for a normative reason requires such recognition, i.e. only if the following holds:

(2) Where it is true that A has a normative reason to Φ and A Φ’s, A Φ’s for that reason only if A recognises the reason-giving feature which provides that reason.

If (2) is false then the acknowledged fact that Nate cannot recognise the relevant reason-giving feature does not entail that he cannot act for this reason. So Nate is a counterexample to (1) only if (2) is true.

But (2) is false. One reason for thinking this is that there are cases where an agent has a normative reason to Φ, Φ’s and seemingly Φ’s for that reason, even though he does not recognise the relvant reason-giving feature. At least, such cases have many similaries with paradigm cases of acting for normative reasons, and this provides prima facie support for the view that in these cases the agent is acting for the relevant reason.Footnote 2 For example, consider the following case, elaborating Nate’s predicament:

LeTrain is a generally trusted and reliable judge of Nate’s reasons. Nate knows this about LeTrain and in fact believes that LeTrain is a better judge of his own reasons than he is. There is at this moment a surprise party waiting for Nate in the living room. LeTrain says to Nate: ‘You have a reason to go into the living room’. Though LeTrain recognises what the relevant reason-giving feature is she does not share it with Nate. A credulous though consciously deliberating Nate takes LeTrain at her word, forms the belief that he has reason to go into the living room (without having any belief about what the reason-giving feature is) and on that basis goes into the living room. Party time.

Here there is a strong case to be made for the view that Nate goes into the living room for the reason that there is a surprise party waiting for him there. One relevant feature of the case in support of this view is that Nate’s going into the living room is counterfactually dependent on the presence of this reason: if there was not a surprise party waiting, LeTrain would not say what she does, and Nate would not go into the living room. Such a condition cannot be sufficient for acting for a reason, of course, but there are other features of the case that add to it. Most notably, Nate’s going into the living room is dependent on a sequence of mental processes that can plausibly be considered an instance of non-defective reasoning, that begins with the cognition of the relevant reason-giving feature, together with the recognition that this feature is reason-giving for Nate, and that ends with a deliberate intentional action, flowing (in a non-deviant way) from an intention based on that reasoning. These considerations create a strong presumptive case for the view that, in this example, Nate goes into the living room for the reason that there is a surprise party waiting for him there.

At this point my opponent might reply: In the LeTrain case, Nate cannot act for the reason that there is a surprise party waiting for him, since acting for a reason requires that reason to be ‘the light in which’ one acts. But this reply is either confused or question-begging. It is confused if it simply conflates motivating with normative reasons. The former are supposed features of an agent’s situation which that agent takes to ‘to make his φ-ing right and hence to speak in favour of his φ-ing’ (Alvarez 2010: 35), hence they are the ‘positive light’ in which agents act. The latter are (or involve) features of the world that justify or provide considerations in favour of acting, and need not be cognised by the relevant agent. That these are distinct is shown at least by the fact that an agent’s motivating reasons need not reflect her normative reasons: the light in which she acts may be misleading or illusory (cf. Williams 1981; Ridge and McKeever 2012: 130). Given these definitions, it is of course true that acting for a motivating reason requires that reason to be ‘the light in which’ one acts, but unless more is said this tells us nothing about the distinct class of normative reasons. Alternatively, then, the current reply is question-begging if the claim that acting for a reason requires that reason to provide ‘the light in which’ one acts is taken to apply directly to normative reasons – for so interpreted this is simply claim (2), which is the point at issue.Footnote 3

Another possible response to the claim that Nate goes into the living room for the reason that there is a surprise party waiting for him there suggests that although there is a normative reason for which Nate goes into the living room, it is not that there is a surprise party waiting in the living room. This response needs to do two things. First, since it accepts that there is a reason for which Nate acts, it needs to give an account of what this reason is. Second, it needs to show that, in light of this alternative reason, it is not the case that Nate acts for the reason that there is a surprise party waiting in the living room. The simplest way of doing this would be to show that the alternative reason is incompatible with the surprise party reason. A more complex way would be to argue that given we have specified an alternative reason, we lose any argument we might have had for thinking that Nate acts for the surprise party reason. I can think of at least three candidates for what Nate’s alternative normative reason might be. But (as I will now argue) none of these accounts can can deliver the second (required) stage of the response.

First, Nate’s reason for going into the living room might be that there is a reason to go into the living room There are several problems with this view. It seems to require that whenever F is a normative reason for A to Φ then the fact that F is a normative reason for A to Φ is itself a – complementary – reason for A to Φ (after all, if Nate acts for the normative reason that there is a reason to go into the living room, then there must be such a normative reason, cf. Ridge and McKeever 2012: 132). But then, of course, the fact that F is a normative reason for A to Φ is a reason for A to Φ is also a reason for A to Φ; and so on. Each normative reason grounds an infinite number of normative reasons for the same action: a case of ‘Too Many Reasons’ if ever there was one (cf. Schroeder 2007), since surely some actions are supported by a less-than-infinite number of reasons! Second, even putting this problem aside it does not follow from the admission that Nate acts for the reason that there is a reason to go into the living room, that he does not act for the reason that there is a surprise party waiting for him there. Nor does the admission that he acts for the reason that there is a reason to go into the living room undermine the previous (presumptive) case for thinking that Nate does act for the surprise party reason. In other words, this version of the response fails to deliver on the second necessary stage of a response of this kind: it fails to provide any reason to think that it is not the case that Nate acts for the reason that there is a surprise party waiting in the living room.

Consider, second, the view that Nate’s (only) normative reason for going into the living room is that LeTrain – a reliable judge of Nate’s reasons – said ‘You have a reason to go into the living room’. This view is suggested by the more general claim that reliable testimony regarding one’s normative reasons to act is itself normative reason to act. There are at least two problems with this view, highlighted by Ridge and McKeever (2012: 118), and a further problem with its application to the case of Nate. First, when Nate goes into the living room on the basis of LeTrain’s advice, it would make sense for him to remark: ‘Before coming into the living room, I had no idea what reason(s) I had to come in here’. But on the current suggestion this remark is false, since Nate did know at least one of his reasons: namely that LeTrain said ‘You have a reason to go into the living room’. Second, suppose that although LeTrain is generally a reliable judge of Nate’s reasons, we face a particular case in which he is not. He advises Nate to go into the living room and Nate obliges, but no surprise party awaits, only a half-eaten pizza. Nate remarks: ‘LeTrain got is wrong: there was no reason for me to come in here’. Again, according to the current suggestion, Nate’s remark is false, but this seems implausible (cf. Ridge and McKeever 2012: 118–120). A third problem for the current interpretation of Nate’s case is that even accepting (contra the previous two points) that LeTrain’s advice is a normative reason for which Nate acts, this (again) does not rule out that Nate also acts for the normative reason that there is a surprise party waiting in the living room, nor does it in any way undermine the previous argument for that claim.

Third, then, consider the suggestion that Nate’s reason for going into the living room is that going into the living room would promote worthy ends which LeTrain’s claims about my reasons are tracking. The thought would be that to say that Nate considers LeTrain a reliable judge of his (i.e. Nate’s) reasons is to say that Nate thinks that LeTrain reliably judges which ends are worthy of Nate’s pursuit, and reliably claims that Nate has reason to perform acts which promote those ends. We can presume, too, that Nate desires to promote these ends. Therefore Nate’s underlying (normative) reason for going into the living room is that doing so would promote the worthwhile ends which LeTrain’s judgements about Nate’s reasons track (cf. Ridge and McKeever 2012: 133–134). This is an innovative suggestion, but once again it fails to be dialectically effective since it fails to deliver the second stage that responses of this type require. Accepting that the above is one of the (normative) reasons for which Nate acts does not preclude also accepting that Nate acts for the (normative) reason that there is a surprise party waiting in the living room, nor does it undermine the previously given presumptive case for that view.

We have seen, then, that the three alternative suggestions for the reason for which Nate acts (in the case of LeTrain) do not put pressure on the core claim, viz. that (one) of the normative reasons for which Nate acts is that there is a surprise party waiting in the living room. There is also, in addition to the presumptive case given above, a further argument that can be made in support of this claim. For suppose one attempts to deny that Nate acts for the reason that there is a surprise party waiting in the living room. Then consider the explanation of Nate’s action in terms of the (intensional) fact that there is a surprise party waiting in the living room. What type of explanation is this? It does not seem to be merely causal, for this ignores the way in which Nate’s action was the result of a recognisable chain of non-defective reasoning (albeit one not completely ‘owned’ by Nate). It is also not an explanation in terms of the italicised sentence being Nate’s motivating reason, since (as noted above) the fact that there is a surprise party waiting in the living room is not ‘the light in which’ Nate acts.Footnote 4 But if this explanation is neither merely causal nor explanation in terms a motivating reason, what type of explanation is it? This tricky question is answered if we adopt the view that the explanation of Nate’s action in terms of this fact is a normative reason explanation, that is, a case where Nate acts for the normative reason that there is a surprise party waiting in the living room.

So suppose we accept the view that Nate goes into the living room for the reason that there is a surprise party waiting for him there. But Nate does not recognise the relevant reason-giving feature. So (2) is false. The moral is that when acting for reasons, the burden of reliably and deliberatively connecting those reasons to action need not always be borne by the agent’s own recognition of the relevant reason-giving features. It can also be borne, for example, by the trusted testimony of others. From time to time, we can let LeTrain take the strain. Williams (1981: 107) adds a further type of case: when an agent recalls that there is some reason for her to Φ, but cannot remember what the reason-giving feature is. In effect, this is a case where the agent’s past-self takes the role of LeTrain. It is perhaps also a case where it is more intuitively compelling to hold that the agent acts for the relevant reason.)

As we saw, Nate is a counterexample to (1) only if (2) is true. But the case of LeTrain shows that there is a plausible view of what it is to act for a normative reason according to which (2) false. Hence, if this view is correct, Nate is not, after all, a counterexample to (1). What this suggests is that in assessing (1) one must pay careful attention to the issue of what it is to act for a normative reason. I return to this issue in §V.

III.

In §§I-II, I interpreted (1) as involving logical possibility. But we can also understand (1) to involve rational possibility, where A Φing for a reason is rationally possible just in case A would Φ for that reason were she to rationally deliberate. Williams is often taken to hold this view (Shafer-Landau 2003:172; Schroeder 2007: 6–7). Note already one oddity of formulation: we do not say that such-and-such is nomologically possible just in case it would happen given the actual laws of nature, only that it could happen given those laws. A further problem is that this formulation seems to preclude the possibility that rational deliberation be consistent with two or more incompatible actions (as in the case of Buridan’s Ass). Perhaps we should say, instead, that A Φing for a reason is rationally possible just in case it could be that A Φ’s for that reason after rational deliberation, i.e. that there is a rational version of A who Φ’s for that reason.Footnote 5

Is Nate a counterexample to this version of (1)? He would be, if it were not rationally possible for him to go into the living room for the reason that there is a surprise party waiting for him there. We saw above that a case can be made for the view that it is logically possible for Nate to go into the living room for this reason, since he does so in the case of LeTrain. The issue, therefore, is whether this is also a case where Nate has rationally deliberated. And it seems that it is. For relying on the reliable testimony of others is a paradigmatically rational way of deliberating. Thus, if the above argument is correct, it is both logically and rationally possible that Nate goes into the living room for this reason. Hence Nate is also not a counterexample to (1) when the relevant possibility is understood as rational (pace Bedke 2010: 42).

IV.

If the above arguments are correct, Nate’s reason – and the class of elusive reasons more generally – is less problematic than some suppose. But it is still instructive. Consider again the view that A has a reason to Φ only if it could be that A Φ’s for that reason after rational deliberation. This view is still underspecified, since we need to know what does and does not count as rational deliberation. I take Williams’ term ‘practical reasoning’ (1981: 104, 110) to be synonymous with ‘rational deliberation’ here. When it comes to specifying this notion, Williams provides a list:

A clear example of practical reasoning is that leading to the conclusion that one has reason to Φ because Φing would be the most convenient, economical, pleasant etc. way of satisfying some element of [one’s subjective motivational set]…But there are much wider possibilities for [rational] deliberation, such as: thinking how the elements of [one’s subjective motivational set] can be combined, e.g. by time-ordering; where there is some irresoluble conflict among the elements of [that set], considering which one attaches most weight to...; or, again, finding constitutive solutions, such as deciding what would make for an entertaining evening, granted that one wants entertainment (1981: 104).

As others have noted, one interesting thing about this list is its distinctive instrumentalist flavour: all these types of deliberation are directed towards the promotion or satisfaction of elements of the agents’ existing motivational sets. (Hence one type of counterexample to Williams is provided by apparent cases of acting for reasons that do not have this flavour, cf. McDowell 1995; Milgram 1996: 210–212 and Wong 2006: 549.) But another interesting feature of Williams’ list, and of most subsequent interpretations of it, is the strong implication that practical reasoning is essentially private: that is, a way of coming to conclusions about one’s reasons that does not involve the input of other individuals. The example of LeTrain shows this implication to be misleading. There are ways of rationally deliberating (and ways of coming to the conclusion that Φing would help satisfy some element of one’s motivational set) that rely on the reliably testimony of others, and which can therefore work without revealing the relevant reason-giving features (compare Bedke 2010). Understanding rational deliberation this way, the constraint on reasons provided by (1) is considerably loosened. In particular, that constraint can now allow reasons such as Nate’s.

V.

In §II I argued that Nate (in the case of LeTrain) acts for the reason that there is a surprise party waiting in the living room. But what general account of acting for a normative reason does this conclusion suggest, and what are the alternatives? In this section I set out some possibilities.

As we saw above, the logically perspicuous way of presenting a reason is a triadic relation, R, holding between an agent A, action Φ and reason-giving feature F. We can represent this as: R{A, Φ, F}. Consider the natural view of what it is to act for such a reason:

A Φ’s for the reason R{A, Φ, F} iff.

  1. (i)

    F

  2. (ii)

    F is a normative reason for A to Φ

  3. (iii)

    A believes that: F and A regards F as a reason for A to Φ

  4. (iv)

    The intentional attitudes involved in (iii) are appropriately sensitive to (i) and (ii)

  5. (v)

    The intentional attitudes involved in (iii) cause, in a non-deviant way, A’s Φing.Footnote 6

There are, of course, many details that need filling out and about which sensible philosophers might disagree (Ridge and McKeever 2012: 129). For example, a complete account requires saying what it is to regard something as being a reason, what it is for such an attitude to be appropriately sensitive to such reasons (counterfactual dependence is an obvious candidate) and what makes for a non-deviant causal connection between the relevant attitudes and the action. But I can afford to remain neutral on these issues, since my aim is to draw attention to a distinct dimension by reference to which different accounts of acting for a reason can be distinguished.

The natural view is natural insofar as it reflects the simple thought that acting for normative reasons requires an appropriate sensitivity to those reasons (cf. Broome 2007: 351; Ridge and McKeever 2012: 136). More specifically, it spells out this sensitivity in terms of a two stage process: first, the agent’s intentional attitudes being appropriately sensitive to those reasons; second the agent’s actions being appropriately dependent on those intentional attitudes. More precisely still, it takes the first stage of this process to involve cognisance of the relevant reason-giving feature as well as cognisance that it this feature is reason-giving. In other words, the natural view captures the thought that acting for normative reasons requires a deliberative sensitivity to those reasons.

If we accept the natural view then Schroeder’s original example of Nate is indeed a counterexample to (1). Nate’s reason in this case can be represented as: R* = R{Nate, go into the living room, that there is a surprise party waiting in the living room}. And condition (iii) cannot be met, since Nate cannot believe that there is a surprise party waiting in the living room.

But, as the discussion of LeTrain shows, there is at least one further, more expansive, account of what it is to act for a normative reason:

A Φ’s for reason R{A, Φ, F} iff.

  1. (i)

    F

  2. (ii)

    F is a normative reason for A to Φ

  3. (iii’)

    EITHER A believes that: F and A regards F as a reason for A to Φ OR A believes that: there is a reason for A to Φ.

  4. (iv)

    The intentional attitudes involved in (iii’) are appropriately sensitive to the facts involved in (i) and (ii)

  5. (v)

    The intentional attitudes involved in (iii’) cause, in a non-deviant way, A’s Φing.

If this view is correct, Nate can act for the reason R*, since Nate can satisfy the second disjunct of (iii’). He does so in the case of LeTrain. In §I I described this case as one where Nate acts for the reason that that there is a surprise party waiting in the living room. But I now take this to be shorthand for more perspicuous claim that Nate acts for the reason R*. It follows that if defenders of claim (1) follow the expansive account of what it is to act for a normative reason, the apparent counterexample of Nate need not trouble them.Footnote 7

Which account of acting for a normative reason should we prefer? This may not be the best question to ask. For we were only concerned with what it is to act for a reason insofar as this notion is part of claim (1), which links normative reasons to acting for such reasons. So a better question to ask is this: Given the initial motivations for accepting claim (1), do they suggest that we should accept the natural or more expansive view of acting for a reason? My somewhat hedged answer to this question is that one of the key considerations which supports (1) does not require its defenders to accept the natural view. Hence the defender of (1) can, whilst respecting this motivation, adopt the expansive view, thus avoiding the counterexample of Nate. I defend these claims in §VII. Before that, however, I want to consider whether there might be an even more expansive account of what it is to act for a reason.

VI.

The above discussion of what it is to act for a normative reason might prompt the following question: So long as an agent’s actions are appropriately sensitive to her reasons, does she really need to cognise those reasons in any way? That is, might an agent act for a reason of hers, even though she satisfies neither disjunct of (iii’)? One immediate worry is that removing this condition threatens to reduce the connection between the agent’s reasons and her actions to a merely causal (or counterfactual) one: the thought would be that if the connection between the agent’s reasons and her actions does not involve some kind of practical deliberation in some way based on those reasons (or the relevant reason-giving features), then all that is left is features of the world causing agents to act in certain ways. But this last thought is potentially misleading, for the connection between an agent’s reasons and her actions might involve practical deliberation of some kind, just not the deliberation of the agent who has the reasons. The case of LeTrain was a partial example of this, since some of what one might call the ‘deliberative burden’ for Nate’s action was borne by the deliberations of LeTrain. But there may be more extreme cases, where an external agent takes the whole burden. Consider:

Tate enjoys aimless walks. A key feature of these walks, and Tate’s enjoyment of them, is that they are aimless. On these walks, the route Tate takes and her walking pace are not in any way dependent on Tate’s conscious deliberation. One thing that gives Tate immense joy is when, on these walks, she encounters spider-webs. Normally, the sight of these webs fills Tate with dread, but when encountered on an aimless walk their intricate complexity fills her with joy. Tate takes an aimless walk, and in the glade to her right (and nowhere else nearby) are some spider-webs.

The presence of these webs, it seems, gives Tate a reason to go into the glade. But it seems impossible for Tate to act for this reason, even given the exapanded view of acting for a reason mentioned above. For as soon as Tate engaged in any conscious practical deliberation (even the deliberation of the sort Nate deploys in the case of LeTrain) her walk would no longer be aimless, and the reason to go into the glade would disappear. Or so it seems. Consider the following elaboration of Tate’s case:

Loco knows Tate very well. In particular he is well-acquainted with Tate’s penchant for aimless walks and spider-webs therein encountered. Loco also knows that when aimlessly walking Tate has a (unconscious, implicit) tendency to prefer well-lit paths. Loco has Tate’s best interests at heart and is a perfectly reliable judge of Tate’s reasons. Loco has been observing Tate’s walk and knows of the location and contents of the glade. Loco correctly judges that Tate has a reason to go into the glade and, knowing Tate’s implicit preference for well-lit paths, takes steps to make the path to the glade well-lit. As a result, Tate aimlessly goes into the glade. Joy!

In this case, does Tate act for a reason? More specifically, does she act for the reason R**, which we can represent as: R{Tate, go into the glade, that there are spider webs in the glade}? We can certainly fill in the details of the case so that there is a relation of counterfactual dependence between the reason and the action. But this kind of sensitivity to reasons is (again) not sufficient for acting for a reason. Yet the obvious candidate for what needs to be added to such sensitivity in order generate a sufficient condition is practical deliberation of some kind, more particularly deliberation based on those reasons. And this is exactly what Loco provides. So there is a case to be made for saying, at least, that it is not obvious that Tate does not act for the reason R**.

If so, this case suggests an even more expansive account of what it is to act for a normative reason:

A Φ’s for reason R{A, Φ, F} iff.

  1. (i)

    F

  2. (ii)

    F is a normative reason for A to Φ

  3. (iii”)

    Some agent, X, believes that F and regards F as a reason for A to Φ (where it is possible that A ≠ X)

  4. (iv)

    The intentional attitudes involved in (iii”) are appropriately sensitive to the facts involved in (i) and (ii)

  5. (v)

    The intentional attitudes involved in (iii”) cause, in a non-deviant way, A’s Φing.

This more expansive account allows that Nate (in the case of LeTrain) acts for his reason, but also that Tate (in the case of Loco) acts for her reason. It also avoids reducing the sensitivity required for acting for a reason to a merely causal (or counterfactual) relation between normative reasons and actions, since deliberative cognisance of those reasons is required (clause (iii”)), albeit not by the agent whose reasons they are.

VII.

Which of the three available accounts acting for a reason – the natural view, the expansive view, or the more expansive view – should we prefer? I noted above that this may not be a fruitful question to ask and suggested we should focus instead on the issue of which account best fits with the original motivations for claim (1). This section discusses one of those motivations.

It is sometimes suggested that an important truth about normative reasons is that they are for reasoning with (Lillehammer 2003: 42). This was the thought behind our reluctance to say that agents could act for reasons even though no one deliberates on the basis of those reasons. It also coheres with the idea that a reason which resists recognition as a reason is as nonsensical as the idea of an undetectably funny joke. (1) is one way of capturing this important truth, and this is one of the principal sources of its support.

But the issue here is this: If the suggested important truth is the key motivation behind (1), what does this tell us about account of acting for a reason which (1) employs? The answer would seem to be that the defender of (1) can afford to be non-committal between the natural, expansive and more expansive views, since all these views perfectly well respect the thought that reasons are for reasoning with. That is, they all require as a necessary condition for acting for a (normative) reason that some agent or other deliberatively cognise the relevant reason-giving features and their normative significance, and that this cognition guide the resulting action. In this way they all respect the thought that a reason which no agent can cognise is as absurd as the idea of an undetectably funny joke. But, understood generally, this necessary condition is met both in the case of LeTrain and in the case of Loco. The upshot is that if the important truth that reasons are for reasoning with is the key motivation behind (1), the apparent counterexample of Nate is no threat.

None of this is to say, of course, that the supposed important truth is a good motivation for claim (1), or that there might be other motivations which demand or support a more stringent account of what it is to act for a normative reason. It is just to say that the example of Nate (and other elusive reasons) will not be dialectically effective against those who adhere to (1) for this reason. But it is notable, further, that those who discuss claims like (1) move very quickly from the thought that reasons are for reasoning with to the more restrictive thought that reasons are for reasoning with by the agents whose reasons they are. Joyce, for example, takes reasons to be ‘genuine deliberative consideration[s] for those to whom the reasons are ascribed’ (2006: 194). Ridge and McKeever note: ‘It would be puzzling if there were a class of good [i.e. normative] reasons that an agent could never act for’ (2012: 112), and later that: ‘What would be problematic would be if there were genuinely good reasons which the agent could in no way access’ (2012: 135; see also Setiya 2014: 223). Likewise, Wong asks: ‘what point could a reason have if it is not capable of motivating the agent who has it’? (2006: 540, my emphasis throughout). But while it would indeed be puzzling if there were reasons for actions which no agent could ever perform (cf. Streumer 2007) it is not so obviously puzzling to suppose that there are reasons such that the agent whose reasons they are could in no way access them, so long those reasons could guide the actions of that agent via the practical deliberations and actions of other agents. Likewise, to answer Wong’s question, the point of reasons is to guide actions through a distinctive process that involves the deliberative cognition of those reasons, but it does not follow that this process must involve the agent who has the reasons deliberatively cognising them herself.

VIII.

The proposed counterexample of Nate is directed at a claim which posits a necessary connection between an agent’s normative reasons and some version of that agent acting for those reasons. It alleges that there can be cases where agents have normative reasons, but the relevant version of that agent cannot be motivated for those reasons. But, I have argued, this objection assumes an impoverished account of what it is to act for a normative reason. In particular, it assumes that acting for a reason requires the agent whose reason it is to cognise the relevant reason-giving feature. But this assumption is questionable. As the example of LeTrain shows, acting for reasons can involve leaning on the reliable testimony of others – in particular testimony which obscures the relevant reason-giving feature. The upshot is that a careful assessment of the supposed necessary connection between normative reasons and acting for those reasons requires an even more careful assessment of what it is to act for a normative reason.