I. THE TRADITIONAL VIEW

Analyses of the supererogatory have generally assumed three focal points, the conjunction of which I shall call ‘the traditional view’. Though some further limit actions that can properly be called ‘supererogatory’, rarely do any views not accept the basic principles captured in the following brief statement by Rawls:

The first principle noted by Rawls can be captured as follows:

Permissible not Required holds that one is never morally required – but is always morally permitted – to perform a supererogatory action.

However, Permissible not Required is clearly insufficient to capture what it means for an action to go beyond the call of duty. One would not, for instance, describe the permissible, but not required, action of double-knotting my shoes as supererogatory. The difference between saving one's friends by jumping on a grenade and double-knotting one's shoes is that the former and not the latter appears to have a comparatively positive moral valence. The former and not the latter is:

Note that Morally Good is a comparative claim. It holds that ϕ-ing is morally better than some relevant baseline. Of course, this baseline is not just any old action; actions can be especially morally good as compared to, say, the bombing of Hiroshima. But this doesn't mean they're supererogatory. Rather, Morally Good holds that supererogatory actions are not simply permissible, but have a particularly positive moral status in comparison to actions that, as it were, are merely in accord with one's moral obligations.

Consider now the third feature of the traditional view, also noted by Rawls. Many hold that one essential feature of the supererogatory is that supererogatory actions are supererogatory in part because they involve some non-trivial sacrifice to the agent. Though this claim is controversially strong, a somewhat weaker claim is surely an important aspect of the traditional view:

Take Urmson's soldier. One of the essential features of his action that renders it supererogatory is that it is sacrificial. But were it the case that this soldier could have saved his friends without sacrifice, he certainly would have been morally required to do so. If I am in a position to donate half my yearly salary to Oxfam International, but only at significant cost to my own well-being, doing so is supererogatory. If my donations fail to affect my well-being, or affect it only trivially, making these donations is morally required.

A note on But for Sacrifice. S is a proper subset of all supererogatory actions. Some – like myself – hold that most, even all, supererogatory actions will involve some sacrifice on the part of the agent. Others believe that not all supererogatory actions would be morally required were it the case that there were no agential sacrifice.Footnote ⁴ In the remainder of the article, I will limit my discussion to cases that fall within S (or, at least, that seem to do so given my considered judgements). These cases generate serious problems with the traditional view, even if the relative size of S is very small.Footnote ⁵

The traditional view might be supplemented by a number of additional principles concerning the structure of supererogatory action. For instance, someFootnote ⁶ might argue that the idea of the supererogatory should be limited to acts that are in some way beneficial, or are particularly praiseworthy. But Permissible not Required, Morally Good, and But for Sacrifice seem to be the beating heart of traditional analyses of supererogatory actions, whether further principles are required or not.

II. A CHALLENGE FOR THE TRADITIONAL VIEWFootnote ⁷

James Dreier notices a problem for the traditional view:

One way to state this challenge more precisely is to introduce a modicum of terminology. First, consider the distinction between a supererogatory action, and a merely erogatory action. Though both supererogatory and merely erogatory actions are permissible, supererogatory action goes ‘beyond’ one's duty. Merely erogatory action does not. Consider the following case. Imagine that you can react in one of three ways to a person down on her luck. You can assist her by going out of your way to buy her a nutritious meal. Second, you can offer her one dollar. Finally, you can do nothing. Assume that the third option is morally disallowed and that the first is supererogatory. One might say, plausibly, that the second is ‘merely erogatory’; it counts as the fulfilment of one's moral obligations, but not in a way that is particularly morally special. The supererogatory action is morally special in comparison to it.

Second, consider the idea, mentioned by Dreier, of a moral reason. A moral reason to ϕ is a fact that, morally speaking, counts in favour of ϕ-ing. Moral reasons come in different strengths. In comparison to some moral reason s, a moral reason r might support ϕ-ing in a weaker, stronger or equivalently strong way. Reasons can also add up to support actions in stronger or weaker ways: it could be that the collection of reasons in favour of ϕ-ing is sufficient to render it the case that there is a stronger balance of moral reasons in favour of ϕ-ing rather than some other act ψ, for which there is also moral reason.

Given the traditional view, if ψ -ing is merely erogatory, and ϕ-ing is supererogatory, ϕ-ing must be supported by moral reasons (or collections of moral reasons) that are stronger than the moral reasons that support ψ -ing. The reasoning for this runs as follows. If ψ -ing is both merely erogatory and supported by stronger moral reasons than ϕ-ing, it would appear that ϕ-ing is certainly not morally meritorious; one does not behave in a particularly morally good way if one behaves in a way for which there is weaker moral reason than the act that is merely erogatory. For the same reason, it cannot be the case that ϕ is supported by reasons that are equivalent in strength to ψ . This would entail that ϕ and ψ are of equivalent moral importance or quality. But this is incompatible with Morally Good. Hence it must be the case that supererogatory actions are supported by stronger moral reasons (or a stronger collection of moral reasons) than merely erogatory actions. With this terminology in mind, the puzzle is easy to see: given Morally Good, supererogatory actions will be supported by stronger moral reasons than merely erogatory actions. But given Permissible not Required, supererogatory actions cannot be morally required: merely erogatory actions remain permissible. But, in any collection of potential actions a person might perform, it seems right to say that this person ought to perform the action that is supported by the strongest balance of moral reasons. And hence either the supererogatory action will be required (violating Permissible not Required) or it will not be supported by stronger moral reasons (violating Morally Good).

Of course, there is an easy way out. The purported problem seems to assume that intra-moral rationality – put roughly, the relationship between moral reasons and moral requirements – is of a particular kind: one is morally required to perform the action for which there is strongest moral reason. Call this ‘the standard view’ of intra-moral rationality. If the standard view is correct, the traditional view of the supererogatory cannot be accommodated. But if we reject the claim that we are morally required to perform the action for which there is strongest moral reason (or strongest balance of reasons), we can say it is permissible to ψ even if there is stronger moral reason to ϕ. However, though this is clearly the right response for partisans of the traditional view, it is not enough simply to say that one can be permitted to refrain from performing actions for which there is strongest moral reason. To solve the puzzle one has to adopt an alternative account of the relationship between moral reasons and moral requirements such that the peculiar character of supererogatory action can be accommodated.

The following two sections consider three potential alternatives. None, I argue, is successful. The first two (each a version of a satisficing approach to intra-moral rationality) cannot accommodate But for Sacrifice. The third can accommodate the principles of the traditional view, but ends up with very implausible results in so doing. I conclude that we must reject the traditional view in favour of an alternative analysis of the supererogatory.

III. SATISFICING

One alternative account of intra-moral rationality is a satisficing view. According to Dreier, ‘Ethical satisficing is structurally similar to rational satisficing. Ethical satisficing theory says that it can be morally right to choose an alternative that is good enough . . . even though there is a better alternative available.’Footnote ⁹ On a scheme of moral satisficing, one violates a moral obligation only if one acts in a way that is not morally good enough. The content of satisficing views will differ markedly; in this section, I consider the simplest form of satisficing, one that sets an absolute target (call this t), and holds that one need only perform actions that meet this target threshold. Call this ‘simple satisficing’.

Simple satisficing accommodates Permissible not Required and Morally Good. If we accept that one is morally required only to perform an action that is at least as morally good as the target threshold, this can leave open the possibility that actions that are morally better than the threshold are permissible, but not morally required. Furthermore, moral satisficing can explain why supererogatory action is plausibly morally heroic or ‘beyond’ the call of one's duty: if one's duty is only to conform to the relevant threshold, any action that is morally better than the threshold is, plausibly, especially morally good in comparison to action that barely passes the threshold.

Despite its virtues, simple satisficing cannot accommodate the traditional view. To embrace moral satisficing, one must reject But for Sacrifice. Assuming that ψ -ing is ‘morally good enough’ (i.e. just barely passes the threshold), satisficing permits ψ -ing whether or not a morally better action involves any sacrifice at all. And hence there can be no supererogatory action that would have been required but for sacrifice, as it were.

But simple satisficing is not the only form of satisficing. Dreier advocates a form of intra-moral rationality that may be more promising. In considering the purportedly supererogatory action of helping to return a stranger's hat that has blown away in the wind, Dreier argues that there are, in fact, two moral points of view:

Dreier's account is interesting, and much more can and should be said about it than I can or will say here. For Dreier's view, intra-moral rationality is a mix of two different moral points of view. The point of view of ‘justice’ determines the moral permissibility and impermissibility of actions. For Dreier, however, there is no reason of justice to fetch the stranger's hat. If so, refraining from so doing is permissible, not required. The second point of view, the point of view of ‘beneficence’, determines the relative goodness or badness of actions, but does not say anything about the permissibility or impermissibility of these actions. Reasons of beneficence rank-order actions, but that ϕ-ing is supported by greater reasons of beneficence can never count in favour of a requirement to ϕ. Putting these points of view together, one performs a supererogatory action if one performs an action that is better, as determined by the point of view of beneficence, than the merely erogatory action (which one might interpret as the morally worst action that is compatible with the requirements of justice).

But Dreier's view inherits the problems of simple satisficing. It also fails to accommodate But for Sacrifice. Take Urmson's case. Dreier, to claim successfully that saving one's comrades at a sacrifice of one's own life is supererogatory, must say that, from the point of view of justice, ‘there isn't anything to be said in favour’ of saving one's fellow soldiers. Of course, there is something to be said in favour of rescuing one's comrades from the point of view of beneficence, and this is what renders one's action supererogatory. But Dreier's view cannot accept the further claim that were there no sacrifice involved, one is morally required to save one's comrades. If ‘there isn't anything to be said in favour’ of so doing from the point of view of justice, and justice is the point of view that determines ‘judgements of wrongness’, it seems difficult to see how Dreier might deliver the verdict that it is wrong not to save one's comrades under any circumstances at all. It would seem that any reason to save one's friends is a reason of beneficence not justice, and hence this reason cannot influence the extent to which it would be wrong not to save one's friends when it is of no sacrifice.

Of course, Dreier claims that the domain of justice is the domain of ‘right and wrong’. Hence we might be tempted to suggest that my requirement to save my comrades if it is of no sacrifice is a requirement of the point of view of justice. Doing so in this case does ‘make me more just’. But one cannot make this claim and maintain the integrity of Dreier's proposal. To allow Dreier to claim that, e.g., saving one's comrades by sacrificing one's life is supererogatory, he must say that the reason to save one's comrades is a reason of beneficence: it is something that makes the act ‘better’, but not morally required. But to say that one is required to save one's comrades when there is no sacrifice involved would seem to collapse the distinction between reasons of beneficence and reasons of justice. Such a position seems to require reasons of betterness to play a role in determining one's moral obligations, after all. But if this is the case, it is difficult to see how such reasons wouldn't play a role in the case in which saving one's comrades would entail sacrifice. If so, we must now search for an alternative form of ‘justice rationality’ that would allow one to perform a suboptimal act from the point of view of justice when sacrifice is involved. But this just is the challenge for the traditional view.

The following point is critical. It could very well be that some supererogatory action would be supererogatory even if no sacrifice were involved. However, even if there is one action that falls within S, one and only one action of which it can rightly be said that without sacrifice it would be required, satisficing views fail.

IV. THE PORTMORE/RAZ VIEW

The failures of satisficing approaches trace to the same source: neither could allow that a lack of agential sacrifice would, of itself, render a previously supererogatory action morally required. To accommodate But for Sacrifice, then, one must allow that agential sacrifice (or lack thereof) can influence the content of moral obligations.

But the three tenets of the traditional view entail that agential sacrifice must play this role in a special, and perhaps non-straightforward, way. By But for Sacrifice, any supererogatory action ϕ, falling within S, must be supported by reasons that would be sufficient to require ϕ, were it not for the sacrifice to the agent involved. But, by Permissible not Required, ϕ cannot be required as it stands. Hence the sacrifice involved in ϕ-ing must itself must have some sort of force in determining our moral requirements. But this force is either the product of per se moral reasons or not. If this force is the product of per se moral reasons, the sacrifice, which justifies a merely erogatory act ψ in comparison to ϕ, would itself have moral significance or per se moral weight, strong enough morally to justify in comparison to the reasons in favour of ϕ-ing. And if this justificatory strength is the result of moral factors themselves, it cannot be that ϕ-ing is morally special in comparison to ψ -ing; both are justified by significant moral concerns.Footnote ¹¹ And if this is right, by Morally Good, ϕ-ing is not supererogatory. Hence the force of agential sacrifice to justify merely erogatory action (and render supererogatory action unrequired) must, at least in part, be the product of reasons that themselves lack per se moral weight. Call any view that accepts the possibility that the non-moral significance of, e.g., prudential sacrifice can influence the moral permissibility of actions a version of the ‘Portmore/Raz view’.

Two versions of the Portmore/Raz view are offered, not coincidentally, by Douglas Portmore and Joseph Raz. According to Raz, the supererogatory is the outcome of a set of reasons known as ‘exclusionary permissions’. Raz explains this idea as follows: ‘The permission to refrain from performing an act [of] supererogation is an exclusionary permission, a permission not to act on certain reasons. An act is a supererogatory act only if it is an act which one ought to do on the balance of reasons and yet one is permitted not to act on the balance of reasons.’Footnote ¹²

For Raz, an exclusionary permission is a not a per se moral reason (it is not part of the ‘balance of [moral] reasons’), but is rather a ‘second-order’ reason, viz., a reason that allows a person to ignore the balance of moral reasons, or to refuse to grant them weight in one's moral deliberation. Surely there are strong moral reasons to save one's comrades by jumping on a grenade. But the soldier is permitted, given his exclusionary permission, to exclude these reasons in his moral deliberation. And hence, on this view, refraining from so doing is morally permissible given that the soldier maintains an exclusionary permission to ignore the moral reasons in question.

This proposal appears to accommodate the traditional view: Permissible not Required is accommodated, given that (in the case of the soldier) jumping on the grenade is perfectly permissible: the balance of first-order moral reasons favours doing so. But, given this exclusionary permission, one can fail to take these reasons into account, rendering the supererogatory action unrequired. Morally Good is also accommodated. The balance of first-order moral reasons, after all, favours the supererogatory action. Though the soldier has ‘exclusionary permission’ to ignore the moral reasons in question, the fact that there are such reasons allows Raz to claim that following through on the sacrifice is morally superior: it is an act that conforms to the balance of first-order moral reasons one nevertheless has permission to exclude.

Raz's view can also be interpreted to accommodate But for Sacrifice. Its ability to do so depends on what triggers an exclusionary permission. One thought might be that the exclusionary permission is present in any case in which the ‘first-order’ reasons are reasons of beneficence. But this interpretation fails to accommodate But for Sacrifice, because this exclusionary permission would be present whether or not the beneficent action requires agential sacrifice, and hence Raz's view would be no less problematic than Dreier's. But one could say that the exclusionary permission is triggered only in cases in which the sacrifice to the agent in question is strong enough. If we do this, Raz's view successfully accommodates But for Sacrifice. For supererogatory actions falling within S, the exclusionary permission is triggered by the sacrifice involved; no sacrifice, no permission. On this interpretation, Raz's view does exactly what I argued must be done to accommodate the traditional view: it treats agential sacrifice as lacking ‘first-order’ moral importance, but nevertheless justifies merely erogatory action (by triggering an exclusionary permission to ignore the moral reasons that favour supererogatory action).

Portmore's version of the Portmore/Raz view uses a slightly different mechanism, but with a similar upshot. Following Joshua Gert,Footnote ¹³ Portmore exploits an important ambiguity in the way reasons operate. Note that the common understanding of reasons, i.e. as facts that ‘count in favour of’ particular actions, is vague.Footnote ¹⁴ Reasons may have requiring strength, i.e. may count in favour of a requirement to ϕ, or justifying strength, i.e. may count in favour of the permission to ϕ. According to Portmore, non-moral reasons (such as prudential reasons to avoid first-personal sacrifice) can possess moral justifying strength. For Portmore, non-moral reasons (including prudential reasons) can count in favour of moral justification to ψ rather than ϕ, despite the fact that ϕ-ing is morally better (i.e. supported by stronger moral reasons) than ψ .Footnote ¹⁵ Non-moral reasons (including prudential reasons) are sufficient to morally justify merely erogatory action when, say, the prudential reason not to ϕ is of sufficient weight compared to the moral reason to perform the supererogatory action.

Portmore's view can accommodate Permissible not Required, because the reasons that tell in favour of merely erogatory actions have the power to justify merely erogatory action (and hence render supererogatory action unrequired), but do not have the power to require merely erogatory action, maintaining permission to perform supererogatory action. In addition, it can accommodate Morally Good. The moral justification allotted to a merely erogatory act is a result of non-moral reasons; this allows us to say that a supererogatory act is morally superior (since it is supported by a stronger balance of moral reasons), though not required (given that the merely erogatory action, though morally worse, is morally justified as a result of non-moral reasons). Further, it can accommodate But for Sacrifice. In cases like that of Urmson's soldier, were there no sacrifice, there would have been no relevant non-moral reasons that would justify a morally worse act, and hence only the previously supererogatory act (i.e. saving one's comrades) would be permissible.

The Portmore/Raz view, in either iteration, can accommodate the traditional view. Indeed, as I argued above, only the Portmore/Raz view can adequately accommodate the traditional view. What remains is to determine whether the Portmore/Raz view is plausible. I claim that it is not.

V. THE BOUNDS OF MORAL JUSTIFICATION

The best way to see this is to compare two cases (for the purposes of brevity, I focus on Raz's statement of the view; the problems are the same for any version of the Portmore/Raz view). The first is:

On reflection, it is plausible to say that Gus's donation would be supererogatory. Not only this, it seems plausible to say that were there no cost involved in saving ten people from death, Gus would surely be morally required to do so. Should Gus fail to save these ten from death at no cost to himself, this is surely grounds for moral complaint. Hence we should accept:

1. 1. Gus's potential donation would be supererogatory, falling within S.

So far so good. But before I introduce the second case, let's shift gears slightly. There is surely some reasonably serious harm to which we may morally permissibly subject an individual for the purposes of saving some number of people (perhaps large) from death. Moral theories will differ on the seriousness of harm-to-number of people saved ratio; act-consequentialism, for instance, will claim that one person harmed will justify one person saved, for any harm short of death. Others will hold that there are agent-centred restrictions against harm, and that any reason to harm can only be outweighed by a larger amount of good, a larger number of individuals saved. For any plausible moral theory, however, there is some number of people (n) we could save from death that would morally justify, say, torturing one person, even leaving other things equal (e.g. that the agent is not among those who would be saved). If so, the moral reason to save n from death is of (at least) equivalent strength to the moral reason not to torture. Further, with each harm of decreasing seriousness, the moral reason not to harm someone in that less serious way is of equivalent strength to the reason to save fewer individuals. If we are licensed to torture an arbitrary person (call him ‘Jerry’) to save, say, 1,000 people from death, surely we are morally justified in harming Jerry in a less significant way to save some lesser number of people. With all this in mind, the following seems plausible:

1. 2. Other things (including non-moral reasons) being equal, one is morally justified in beating Jerry up to save ten from death.

However, if we accept (2), a problematic result arises. Consider now the second case:

If we accept (1) and (2), and the account of exclusionary permissions that will successfully accommodate the traditional view, it would seem that Stan is perfectly justified – morally justified, mind you – in beating up Jerry to buy a new car. Here's why. Gus's sacrifice triggers an exclusionary permission to refrain from saving ten. But Stan's potential sacrifice (i.e. not getting his new car) is identical to Gus's. And so one would assume that this sacrifice would also generate an exclusionary permission to beat up Jerry.

The obvious response is to hold that the relevant exclusionary permission is triggered only in Gus's case, not Stan's case. But, given (2), this is not tenable. As we have already seen, it is at least as morally important (that is, as important from the perspective of first-order moral reasons) to save ten as it is not to beat up Jerry. And so one would expect that if an exclusionary permission allows one to exclude the moral reasons in favour of saving ten in Gus's case, it must allow Stan to exclude the moral reasons not to beat up Jerry. Of course, there are differences between the two cases. Importantly, Gus faces moral reasons of beneficence; Stan faces moral reasons of non-maleficence. But this, in itself, cannot make a difference to the possibility of exclusionary permission unless that difference is reflected in the first-order moral significance of the reasons in question. And given (2), it is not: saving ten is at least as morally important as refraining from beating up Jerry.Footnote ¹⁶ The problem is identical for Portmore's view. Given the relative moral significance (as laid out in (2) ) of saving and refraining to harm, any statement of the Portmore/Raz view will have the result that Stan is morally justified in beating up his neighbour given the moral justificatory significance of agential sacrifice. If agential sacrifice (by whatever mechanism) morally justifies pursuit of one's own interests against an action supported by moral reasons of strength s, it should also morally justify pursuit of one's interest against any other action supported by reasons of strength s-or-weaker.

Care is required here. The mere fact that Stan could be morally justified in beating up Jerry to buy a new car shouldn't by itself be regarded as a fatal result of the traditional view. After all, many views – including act-consequentialism, for instance – will hold that such an act can be morally justified, especially if Stan's well-being is furthered by a new car more than Jerry is burdened by the beating. But the problem for the Portmore/Raz view is not this verdict per se, but rather the fact that Stan can be morally justified in beating up Jerry even if his interests are, comparatively, of insignificant moral weight (given that the justificatory power agential sacrifice, according to the Portmore/Raz view, outstrips its per se non-moral significance). This verdict is simply incredible, and must be rejected. If Stan is justified, this justification cannot be moral justification.

One important implication of the above argument is that it doesn't matter in the slightest what the relevant moral reasons are, or what moral factors might favour refusing to beat up Jerry versus what factors favour saving ten from death. One could, in principle, declare (in Kantian fashion) that the reason not to beat Jerry up is derived from an interest in not treating Jerry as a mere means, or that this reason is derived from Jerry's moral rights, or that Stan is under some particularly morally relevant relationship to Jerry, or is accountable to Jerry, in a way that tells against beating him up. Furthermore, one could assume that the reason to save ten is provided by their overall aggregate welfare, respect for persons, or any other reason. Because the reason – whatever it is – not to beat up Jerry is not stronger than the reason to save ten (in so far as one is allowed to beat up Jerry to save ten), and because the relevant agential sacrifice is significant enough to justify refusing to save ten (as seen in Gus's case), it must also be significant enough to justify beating up Jerry. To say otherwise is to accept a logic of exclusionary permission that seems utterly arbitrary.

The partisan of the traditional view could deny either (1) or (2). But (1) seems obvious; indeed, Gus's potential donation seems a paradigmatic instance of the supererogatory; furthermore, if we can save such individuals from death at no cost to ourselves and fail to do so, those whom we failed to save surely could complain on moral grounds. This leaves (2). One might complain that I have no genuine reason to believe (2); after all, that we are morally justified in beating up one to save ten, rather than, say, eleven, seems arbitrary. But the specific number identified by (2) is neither here nor there; the number can change without changing the seriousness of the problem. Imagine that Gus could save 100 people from death as a result of his Oxfam donation. Many would say that even under these conditions, Gus's donation remains supererogatory; his prudential reason is strong enough morally to justify his refusal to save 100 from death. But to deny that one could be morally justified in beating someone up to save 100 people from death is positively absurd. Hence Stan's prudential reason, which is identical to Gus's, must also have the power morally to justify beating up Jerry, in so far as refraining from beating up Jerry is not morally more important than saving 100. One might put this in general terms: there is some number (‘n’) of individuals Gus could save from death at the cost of his new car and new life with his significant other, for which an exclusionary permission is no longer applicable. For the Portmore/Raz view to survive, we must say that we are unjustified in harming someone like Jerry to save any sub-n number of people from death. But this claim is worse than dubious.

The Portmore/Raz view accommodates the traditional view. I have argued, however, that given straightforward assumptions, and straightforward reasoning, this view implausibly expands the bounds of moral justification. This is not to say, of course, that there is no way to avoid this conclusion. But the problem here seems robust enough to warrant the search for an alternative to the traditional view. In what follows, I offer just such an alternative that is (a) satisfying in itself and (b) keeps the boundaries of moral justification right where we want them.

VI. THE SUPEREROGATORY

To begin, consider the notion of a moral requirement. Moral requirements are, well, just that: if I fail to conform to a moral requirement, this entails that I will have behaved immorally, or in a morally unjustified way. But there are many different sorts of requirements – not just moral – that I face. I face legal requirements, prudential requirements, requirements of etiquette, requirements of my neighbourhood association. Sometimes these requirements will conflict. But in cases of conflict, it seems natural to ask ourselves what we ought to do really, or all-things-considered. More generally, in the case of conflicting requirements, how should I live? For the sake of brevity, I will refer to this ‘all-things-considered’ requirement, which is distinct from, e.g., moral, legal or prudential requirements, as the ‘rational’ requirement, or rational ‘ought’.

I take this conceptual territory to be familiar. Furthermore, we have an intuitive grasp on some of the answers to the questions asked above: most would agree that it is more important to conform to a moral norm than a norm of one's neighbourhood association. This is reflective of a standard assumption, viz. that morality is the most important domain, at least in determining how I all-things-considered ought to act. In particular, a standard analysis of the relationship between rational and moral requirements holds that the requirements of morality are rationally supreme: if one is morally required to ϕ, one is thereby rationally required to ϕ.

But just for the sake of argument, let's say that morality is not supreme. Let's say that, in some cases, immorality can be be rationally permitted.Footnote ¹⁷ If we do this, it seems to me a natural – and alternative – way to understand the supererogatory emerges. Take the following case.

How would we describe Rose's action? Ex hypothesi, Rose behaves in accordance with her moral obligations. But also, ex hypothesi, Rose is not rationally required to do so. One might, of course, simply describe Rose's action as an instance of action that conforms to a moral requirement. But that doesn't seem to say it all; given our assumptions, Rose's action is not just morally required, Rose's action is morally required in a way that isn't required of her. Though it is morally required, it is – one could correctly say – ‘beyond the call of duty’.

I want to be very clear about what I take Rose's case to establish.Footnote ¹⁸ I won't assert that her case, by itself, makes for an argument that one can, as a matter of practical rationality, behave immorally. Perhaps the assumptions I've described in the case are plausible; perhaps they are not. What matters at this point in the argument, though, is not the plausibility of these assumptions per se, but rather the way in which we might be tempted to describe Rose's actions under these assumptions. If we believe the case as described, i.e. that Rose is, in fact, morally required to help but rationally permitted not to help, we would in fact reasonably describe Rose's helping as an instance of the supererogatory: an action that is morally better than Rose's duty. It is surely, as Rawls says, ‘an act of benevolence and mercy, of heroism and self-sacrifice’.

Reflection on Rose's case seems to offer an alternative to the traditional view. Rather than holding that supererogatory actions are morally good but not morally required, one might say that, very roughly, supererogatory actions are morally good, but not rationally required.

This proposal requires rethinking of each of the tenets of the traditional view. First, Permissible not Required, on this view, becomes:

Morally Good becomes:

But for Sacrifice becomes:

Though this proposal, like the traditional view, permits of a number of potential additions and supplementary principles, these theses are the heart of what I call ‘the anti-rationalist view’.

The anti-rationalist view is attractive when compared to the traditional view. The key advantage is that the anti-rationalist view has no need of an alternative account of intra-moral rationality. The anti-rationalist view is perfectly free to say that one is morally required to conform to the strongest balance of moral reasons. This is an advantage for two reasons. First, this view is prima facie plausible (as noted by Dreier). If we can both accommodate the supererogatory and accept the standard account of intra-moral rationality, this is a comparative boon. But even aside from its prima facie plausibility, the ability to avoid an alternative account of intra-moral rationality saves the anti-rationalist view from the implausible results of the Portmore/Raz view. Because those wishing to accommodate the traditional view must accept some version or other of the Portmore/Raz view, any problems with the Portmore/Raz view bleed into the traditional view. And problems there are: the Portmore/Raz view implausibly expands the boundaries of moral justification (as explored in section V). But because the anti-rationalist view can accept a standard account of intra-moral rationality, this problem is avoided.

One immediate objection should be discussed. The anti-rationalist view is compatible with a standard account of intra-moral rationality, i.e. with the view that one is morally required to perform the action for which there is strongest moral reason. But, on the anti-rationalist view, supererogatory actions are those which are morally better than is rationally required. But given that one is morally required to perform the morally best actions, this entails that sometimes morally required action will be supererogatory. This might be thought a non-starter.Footnote ¹⁹

Of course, I do allow that morally required action can, on occasion, be supererogatory. But this is a feature of my analysis, not a defect. It would be a defect if, conceptually speaking, supererogatory action cannot be morally required. But this is not the case. The concept of the supererogatory for which we seek an illuminating analysis is expressed by, e.g., the thought that some actions go ‘beyond’ one's duty. The traditional view offers a conception of this concept: that supererogatory actions are morally good, but not morally required. But the traditional view is not conceptually true; it is a theory of the supererogatory that can (and, I argue, should) be denied. Thus it is no objection to the anti-rationalist view that the anti-rationalist view (sometimes) treats morally required action as supererogatory.

Furthermore, the anti-rationalist view is faithful to the concept of the supererogatory. Take Urmson's soldier. Urmson appears to indicate that jumping on a live grenade to save others is morally superior to refraining from so doing, but that it is not a feature of any individual's ‘duty’. But one has a duty to perform only those actions one is rationally required to perform. If conforming to a moral requirement is not rationally required, one has no duty to conform to such a requirement. Hence it is perfectly acceptable to say of this case that one has no duty to conform, and also to say that, e.g., we could not ‘possibly say that they failed in their duty by not trying to be the one who sacrificed himself’, or that ‘If he had not done so’ we would not have ‘said to him “You ought to have thrown yourself on that grenade” ’. Furthermore, because he had sufficient rational permission not to jump on the grenade, no ‘superior’ could ‘have decently ordered him to do it’ (given the strength of the prudential reasons against doing it).

I do not wish to gainsay the popular view that supererogatory action cannot be morally required. Nevertheless, it seems to me that this intuition is misleading. It sounds right to say that supererogatory actions cannot be morally required because we have a tendency to identify our duty with our moral duty: we have a tendency to believe that moral requirements are rationally supreme. As noted here, this is a common presumption in our general thought about the relationship between practical rationality and moral requirements. Under these conditions, supererogatory acts cannot be morally required. But if we reject this connection, we are perfectly licensed to say that sometimes moral requirements go beyond our duty. Sometimes we have no duty to conform to moral requirements any more than we have a duty to conform to, e.g., legal requirements or requirements of etiquette. If so, there is no pressure to declare that supererogatory actions cannot be morally required.

VII. HOW TO ACCOMMODATE THE SUPEROGATORY

My proposal is to treat supererogatory actions not as morally better than is morally required, but as morally better than is rationally required. But there are two potential ways this analysis might go wrong. First, it could be that the anti-rationalist view simply doesn't work; it might be that there is no account of all-things-considered reasons that could accommodate the anti-rationalist view. Second, it could be that, even if there is such an account of the rational ‘ought’, this proposal, like the Portmore/Raz view, is just implausible. Because the second objection awaits a response to the first, I discuss the first objection here, and leave the second until section VIII.

Is the anti-rationalist view possible? Obviously, for this view to work we must be anti-rationalists about morality; we must believe that one can be all-things-considered justified in refusing to perform morally required action. However, anti-rationalism comes in many shapes and sizes, many of which clearly cannot accommodate the existence of supererogatory actions. To see this, consider the various ways one might construe the relationship between moral and rational requirements.Footnote ²⁰ One common thing to say about this relationship runs as follows:

Authority holds that morality generates practical reasons: that ϕ-ing is supported by moral reasons entails that there is some practical reason to conform to it. But Authority is very weak. Authority is compatible with the claim that moral considerations are trumped by other considerations, such as prudential considerations. However, one must obviously accept Authority to accommodate the existence of supererogatory actions on the anti-rationalist view. Without doing so, any practical reason whatsoever not to ϕ renders ϕ-ing practically irrational even if ϕ-ing is morally required. This would be tantamount to the rejection of Permissible not Required II.

However, to accommodate the supererogatory, one must deny a stronger association between morality and rationality. One must deny:

Supremacy holds that moral reasons are rationally overriding: whenever morality holds that one ought to ϕ rather than ψ, it is all-things-considered irrational to ψ rather than ϕ. Given the standard account of intra-moral rationality, Supremacy is equivalent to:

The anti-rationalist view cannot accept Supremacy (and hence must also reject Supremacy*, given its embrace of a standard account of intra-moral rationality). To accept the existence of supererogatory actions (on the anti-rationalist's analysis), it must be the case that one occasionally has rational permission to behave in a morally suboptimal way, which Supremacy denies. On this view immorality – failure to conform to one's moral obligations – can, on occasion, be rationally justified given the possibility of other sorts of reasons, including prudential, aesthetic, legal, etc., reasons. This is not to say that conformity to moral requirements will never be rationally required. But, one might say, in cases in which non-moral (including prudential) considerations are strong enough, one is rationally permitted not to conform to the strongest balance of moral reasons.

But the denial of Supremacy and the acceptance of Authority does not guarantee that supererogatory actions, like Rose's, will be rationally permitted. It must be that people are rationally allowed to perform actions that are morally better than those that are rationally required. Hence, the anti-rationalist view must accept:

Permission holds that all action that is morally better than is rationally permissible is rationally permitted. To accommodate the anti-rationalist view, one must accept Permission but deny Supremacy(*).Footnote ²¹

(One further point. I noted in section I that the comparative size of S relative to the set of all supererogatory actions is controversial. Horgan and Timmons, for instance, hold that at least some supererogatory actions are supported only by ‘favouring’ rather than ‘requiring’ reasons, and hence that one needn't be required to perform them even if they involve no sacrifice (such as taking a stranger on a fun outing). But my view seems to treat moral reasons as always of rational requiring strength, and if so S is coextensive with the set of all supererogatory actions. However, my view is flexible on this point. To capture the possibility of a narrower S, one might add a clause to Authority, holding that any moral favouring reason (if, in fact, such things exist) is also a pro tanto rational favouring reason. If this is right, one might then amend Permission to claim that if a particular act ϕ is morally favoured with respect to a rationally permitted act ψ, ϕ-ing is rationally permitted (though merely morally favoured acts would not be required). This would hold that actions supported by favouring reasons are permissible, not required; they are morally better than is rationally required, and hence they are supererogatory without being within the scope of S.)

This view can accommodate the anti-rationalist view successfully. Take Rose. If we accept the standard account of intra-moral rationality, we should say that Rose (given the strongest balance of moral reasons) is morally required to assist the family. But, if we deny Supremacy and accept Permission, we have the power to accommodate the suggestion that Rose's morally required action is, in fact, supererogatory. Rose has prudential reason to avoid assisting this family. Thus though assisting the family is morally good, indeed morally required, it need not be rationally required in light of Rose's sacrifice in so doing, and hence Rose is rationally permitted not to assist the family. But given that assisting the family is morally better than is rationally required, Rose has rational permission to assist the family as well (given Permission). Hence assisting the family, for Rose, is supererogatory. The same applies in the case of Urmson's soldier. One might say that, given the lives at stake, there is surely greater moral reason for the soldier to throw himself on the grenade than to refrain. But given that this action, quite literally, is an instance of self-sacrifice, we should expect that the denial of Supremacy entails that, at least in this case, the soldier is rationally permitted not to hurl himself on the grenade.

If we accept Permission and reject Supremacy, we can accept that a supererogatory action is morally better than the merely rationally erogatory action (satisfying Morally Good II). This view can also accept that supererogatory acts are rationally permitted, but not required (satisfying Permissible not Required II). And this view can also accept that, were it not for the required sacrifice – that is, the prudential reasons against – supererogatory actions would be rationally required; without prudential sacrifice to rationally justify a morally suboptimal act, one is rationally required to perform the action for which there is overriding moral reason (satisfying But for Sacrifice II). To accept Permission but deny Supremacy in the way just illustrated is necessary and sufficient to accommodate the existence of supererogatory actions on the anti-rationalist view.

VIII. OBJECTIONS

So the anti-rationalist view is possible to accommodate. But that doesn't mean that it's plausible. I respond to three potential reasons to reject it here. First, I argue that the anti-rationalist view does not require an implausible first-order account of supererogatory actions. Second, I argue that though some might object to the denial of Supremacy*, one should not consider the ability to accommodate Supremacy* an advantage of the traditional view. Third, some might argue that this form of anti-rationalism view faces problems – just as the Portmore/Raz view did – with Stan. I argue that this is not so.