2. Does blameworthiness entail fault?

The first way of denying the moralist’s argument accepts the blameworthiness of breach, BOB, whilst denying BEF, the claim that blameworthiness entails fault. Applied to the gentleman in Stevenson’s anecdote, this view first uses BOB to interpret him as aptly blaming himself for his inability to adequately support his mother through her illness, and then, by denying BEF, allows his blame to remain apt in the absence of fault. That is how this form of anti-moralism avoids pathologising his feelings of responsibility, without being forced to redescribe what he feels responsible for as somehow his own fault.

If, like Stevenson, we found it pernicious how moralism can lead us to find fault in the gentleman’s conduct, then we should find this way of opposing it no less pernicious. After all, they both involve treating the gentleman unduly harshly. But whilst the fault-finding tendency in moralism is unduly harsh in how it evaluates the gentleman’s conduct, the anti-moralism at issue is unduly harsh in the consequences it metes out.Footnote ⁵ So whilst it might not seek to find fault where there is none, it finds punitive feelings of blame apt nonetheless.Footnote ⁶ In this way, the denial of BEF results in too ‘hard’ an anti-moralism, by the anti-moralist’s own lights. We should want to reject hard anti-moralism if at all possible.

Before we consider BEF in more detail, we should note that responses to situations like that of the gentleman in Stevenson’s anecdote have often expressed a hard anti-moralism. An early example is Ruth Barcan Marcus (Reference Marcus1980: 130), who takes moralism ‘to claim that it would be mistaken to feel guilt or remorse about having failed to act according to that obligation … since [the] obligation was vitiated and [our] feelings are inappropriate’, a claim which, she continues, ‘is false to the facts’. Having thus endorsed BOB, she goes on to claim that there are at least some obligations that may be breached ‘however holy our wills or rational our strategies’ (Reference Marcus1980: 135), that is, however faultless we may be.Footnote ⁷ Another example is Martha Nussbaum, who asks if the Mycenaean king Agamemnon’s sacrifice of his innocent daughter Iphigenia is a case of ‘actual blameworthy wrongdoing’ (Reference Nussbaum2001: 27), given that the brutal deed, on Aeschylus’s telling, was in effect necessitated by the gods. Nussbaum finds that Agamemnon’s subsequent downfall speaks ‘to our intuitive demand that even the constrained killer should come to regard himself as a killer and should suffer, in his own person, for his deed’ (Reference Nussbaum2001: 41), a demand reflecting ‘a deep ethical response that would be intelligible in the absence of the divine’ (ibid.). More recently, Amy Sepinwall (Reference Sepinwall2017: 524–5) nods approvingly to aspects of ‘the Ancients’ responsibility practice’, seeking to recover its rationale, within which ‘one can be blameworthy even if one is not at fault’. In a similar proposal, Elinor Mason (Reference Mason and Shoemaker2019: 248–9) considers someone who inadvertently loses her friend’s sentimental necklace, finding that ‘it is plausible that she should feel something in the region of remorse’, and that she ‘should feel really bad about what she did, even though there was no bad will’.

In the hard anti-moralist literature, BEF is frequently denied solely on the strength of the phenomenology in cases like the gentleman in Stevenson’s anecdote.Footnote ⁸ But is there any way of arguing against BEF on more principled grounds? No doubt, the intuitive connection between blameworthiness and fault is extremely tight, and BEF does indeed seem to be thoroughly entrenched in our blaming practices in both their interpersonal and institutionalised forms.Footnote ⁹ One way we might drive a wedge between them is to suggest that though fault may be sufficient for blameworthiness, it is not in fact necessary. For instance, we can suggest that for someone to be blameworthy, all that is needed is a genuine moral reason to blame them, which may but need not be a fault-based reason.

This should sound familiar from consequentialist accounts of blame, which all say, typically hedged in various ways, that it is necessary for the justification of (a system of) blame that it leads to the best outcomes.Footnote ¹⁰ Such accounts are often considered unconvincing because we can accept that blaming someone blameless may on occasion lead to the best outcomes, without conceding that this somehow makes it a genuinely apt or fitting response.Footnote ¹¹ And for someone to be blameworthy for something just is for it to be genuinely apt to blame them for it.

More recently, hard anti-moralists have attempted to deny BEF by allowing, like the consequentialists, that it is sufficient for someone to be blameworthy that there is simply a genuine moral reason to blame them, not necessarily a fault-based one. But unlike the consequentialists, the fault-independent moral reasons are based, not in the outcomes, but in our close relationships. To be precise, the relevant reasons either are or are explained by the fact that a close relationship obtains between the putatively blameworthy person and the one they let down by their conduct, such as between the gentleman in Stevenson’s anecdote and his mother. Relationship-based arguments against BEF on this pattern are less obviously unconvincing than simple consequentialist ones. Of the hard anti-moralists, it is perhaps Sepinwall who has gone furthest in advancing them. I now consider and rebut three such lines of thought adapted from Sepinwall (Reference Sepinwall2017: 546–552).

The first line of thought is the argument from higher standards.Footnote ¹² If we are lucky, our close relations are deeply important to us in ways that demand that we assess our own behaviour towards them by higher standards. So, if we let down our family, say, this can make it reasonable – at least from our own perspective – to believe we were at fault in so doing, even if in fact we were not. For instance, if the gentleman in Stevenson’s anecdote is unable to adequately support his mother through her illness, it may be perfectly understandable for him, out of love for his mother, to believe that he is at fault in letting her down. The hard anti-moralist suggestion here is that if the gentleman’s tears express feelings of self-blame (e.g., guilt or remorse), then that may be a genuinely apt response to his belief that he is at fault. The close relationship is said to make this doxastic fact a good moral reason for him to blame himself, in virtue of which he is blameworthy despite not being at fault.

The problem, I submit, is that this argument falls foul of the plausible principle that apt feelings are in general veridical. To see this, consider, first of all, that if our feelings admit of assessment as apt, that is because they have representational content. For instance, fear represents its object as dangerous, and that is why it can be apt or not, depending on whether its object is dangerous or not. To generalise: for any given feeling to be genuinely apt, its object has to genuinely be as that feeling represents it as being. This is why, if the object is not in fact as the feeling represents it as being, then merely believing otherwise is not enough to secure the feeling’s aptness. Imagine your immature friend finds an unfunny joke funny. There is no contextual feature, no unintended meaning, nothing edgy about it that your friend is responding to – the joke is straightforwardly unfunny, yet they believe it to be funny. This doxastic fact may make their laughter intelligible, but it cannot make it apt. By the same token, even if Stevenson’s gentleman believes he is at fault in letting his mother down, that cannot make his self-blame apt, no matter how reasonable his belief might be in light of the higher standards his love demands. The relationship may, at best, turn this doxastic fact into a fault-independent moral reason for him to blame himself, but he will not be made blameworthy thereby.

The second line of thought is the argument from degrees of blameworthiness. This argument trades on the following observation. When one person injures another, their degree of blameworthiness depends on the type of relationship they stand in with the victim, even when we hold fixed: (a) the injury itself, and (b) the degree of fault. To see this, imagine two equally negligent drivers hit and injure a child. If one is the child’s parent, and the other merely a stranger, we may think it apt for the former to blame themselves more than the latter. Similarly, if the gentleman in Stevenson’s anecdote is unable to pay for his mother’s medical treatment, we may think it apt for him to blame himself more than does a random stranger for their inability to do the same. The hard anti-moralist infers that his close relationship to his mother must therefore furnish him with fault-independent moral reasons that bear in the right kind of way on his self-blame.

I agree that fault-independent moral reasons can indeed determine how much blame is apt: considerations such as proportionality, the relationship between the blamer and the injurer (including their relative ‘standing’), and whether and to what extent the injurer is independently remorseful. I also agree that the type of relationship that obtains between injurer and victim should be counted amongst these considerations. Quite clearly, the ethical significance of letting down our own family members differs from letting down a stranger’s, and this supports genuine moral reasons that bear on how much we should blame ourselves, in a way that reasons of political expediency (say) do not. But that is not really to the hard anti-moralist’s point. What they need is to provide fault-independent moral reasons that tell us not how much blame is apt, but that blame is apt in the first place. And the considerations at issue do not plausibly play this role. So, whilst I agree that how much we blame (say) a negligent injurer may aptly vary with their relationship to the victim, whether it is apt to blame them at all is surely not something that varies with the relationship. It is blameworthy to injure negligently, no matter the relationship between injurer and victim. By the same token, if the gentleman in Stevenson’s anecdote is not obviously to blame for being faultlessly unable to meet someone’s medical costs, I do not see what difference it makes to point out that that someone is his mother.

The third line of thought is the argument from causal involvement. Consider that our relationships are usually important to us in ways that rightly make demands, not just on how we are attitudinally disposed to one another – and how that finds expression in our agency – but on how our agency impacts the other independently of the attitudes behind them. In such cases, the hard anti-moralist suggests, the relationship can turn our mere causal involvement in injuring the other party into a fact that makes self-blame fitting. So, for example, the sheer fact that the gentleman in Stevenson’s anecdote is causally involved in his mother’s inability to pay for medical treatment – even if only by omission – may come to make it fitting for him to blame himself, given his relationship to his mother. Since we can be involved in causing outcomes without being at fault, the sheer fact of the gentleman’s causal involvement can thus be said to make him blameworthy without fault.

An initial problem with this line of thought is that the feeling usually taken to befit such a fact is what Williams (Reference Williams1976: 123) famously introduced under the heading of ‘agent-regret’. And though Williams and his followers would say that our gentleman is apt to feel agent-regret despite the situation being ‘no fault of his’ (Reference Williams1976: 124), they would also say, on phenomenological grounds, that ‘there is something wrong’ with identifying it with ‘remorse’ (Reference Williams1976: 126) and other such forms of blame.Footnote ¹³ Thus, any attempt to do so only makes blame fault-independently fitting by denying a widely accepted and phenomenologically robust distinction.

On a more sophisticated variant of this line of thought, the fact of causal involvement does not bear directly on the fittingness of self-blame, so much as indirectly, by the following route. Imagine that, unable to help his mother, our gentleman becomes preoccupied with proving to his mother (or perhaps to himself) that it was not his fault, turning inward to the complexities of his own predicament instead of outward to how he can still be there for his mother. Such a focus on exculpation, on his own moral purity, seems not to do right by his mother – whether or not his inability to help her was in fact his fault.Footnote ¹⁴ The idea is to say that, given their relationship, the fact of his causal involvement bears indirectly on his blaming himself by providing a fault-independent moral reason not to try to exculpate himself. Call this the argument from inappropriate exculpation.

I am inclined to agree that our relationships can indeed give us such reasons. Even so, I believe the argument only tends to show that deflecting blame would be unfitting, not that the blame thereby not deflected is itself fitting. To see this, imagine that your hopelessly immature friend has just been broken up with by the fifth partner this year. They are still perilously fragile from the breakup, and now is not the right moment for brutal honesty – it would be unfitting not to lie to them about their qualities. But what you say, in lying to them, does not thereby become true. In general, we can accept that the circumstances sometimes make it unfitting not to dissimulate in our responses, without deluding ourselves that those responses become fitting thereby. So, even on the argument from inappropriate exculpation, the appeal to the sheer fact of our causal involvement does not explain why self-blame is fault-independently fitting, even if it explains why deflecting it would be unfitting. And it is the former, not the latter, that the hard anti-moralist needs.

3. Is breach of obligation blameworthy?

In the previous section, I argued that the denial of BEF is unsustainable, and with it, hard anti-moralism itself. On its own, of course, that is grist to the moralist’s mill, which is the view I am concerned to deny in this paper. Fortunately, as we already saw, the moralist’s argument allows us to oppose it another way, by denying the blameworthiness of breach, or BOB, instead of BEF. I call the resulting view ‘soft’ anti-moralism.

The key features of soft anti-moralism are exhibited in how it represents the gentleman from Stevenson’s anecdote. Its ‘softness’ lies in how it can accept that the gentleman is faultless in breaching his obligation to his mother without the hard anti-moralist implication that he is nonetheless blameworthy, which it avoids by denying BOB. But it remains an ‘anti-moralism’ because it can accept the faultlessness of our gentleman – it avoids seeking to find fault – without being forced to pathologize his feelings of responsibility. We may wonder how it achieves this if it denies BOB. The trick is to reinterpret the phenomenology: soft anti-moralists say that though the gentleman is perfectly apt to feel what he feels, what he feels is what Williams calls agent-regret. And as we saw previously, agent-regret can be fitting without fault and is not itself a form of blame. The appeal to agent-regret, as we shall see in Section 4, is not without its challenges. But for now, suffice it to note that soft anti-moralism is only available if the denial of BOB can be sustained. That is the question I turn to in this section.

Before we discuss whether breach of obligation is indeed blameworthy, let me note that ‘obligation’, as it figures in that claim, admits of what Jonathan Dancy (Reference Dancy2004: 15–16) would call contributory and overall readings. An ‘obligation’ in the contributory sense is defeasible as a reason for action, whereas an ‘obligation’ in the overall sense is undefeated as a reason for action. This will become important as the argument progresses.

My case against BOB proceeds by a series of counterexamples. I will start by considering cases where breach of obligation is adequately excused. Most accounts of excuses accept the following two criteria: (a) they presuppose a failure to satisfy some reason or other (which may but need not be limited to obligations), and (b) they are exculpatory when adequate, reducing blameworthiness or eliminating it entirely.Footnote ¹⁵ Whatever we may think of how different accounts go on to explain these criteria, or how excuses function in human life more broadly, the intuitiveness of the criteria is, I believe, both bedrock and pervasive.

For instance, someone who misses a bidding deadline for a rare item they wanted on eBay might beat themselves up over it (‘How could you miss the deadline? You really wanted that thing!’), but if they missed it because the Wi-Fi went down, it would be familiar and intelligible that they cut themselves some slack, and we would understand this using the concept of an excuse. Even such trivial cases exhibit the need for the two criteria. Had they relaxed their self-disapproval in response to something they did that was in no way regrettable (i.e. were (a) absent but (b) present), then this behaviour would not be intelligible under the concept of an excuse. Similarly, had they responded to missing the eBay deadline in any way such that they still beat themselves up over it (i.e. were (a) present but (b) absent), then they could not be understood as excusing themselves either. Recognisably the same concept (with the same criteria) applies to less trivial examples from everyday life, where the relevant failures strike us as breaches of obligation: for example, if we replace missing the eBay deadline with the breaking of an important promise. And again, recognisably the same concept is institutionalised in the criminal law of both common and civil law jurisdictions, through the practice of excusatory defences. Since only pleas that presuppose legal responsibility for an offence count as defences, excusatory defences satisfy criterion (a).Footnote ¹⁶ And since they would not count as defences if they did not get defendants off the hook of criminal liability when successful, they also satisfy criterion (b).

I take these examples to show that we not only frequently take ourselves to encounter cases of excused breach of obligation in ordinary life, but that we are also not wrong to think of such cases in these terms. This makes it hard to deny the possibility of excused breach of obligation. Given that we are blameless for breaches of obligations for which we have an adequate excuse, it is just as hard to deny the possibility of cases of breach without blameworthiness. For now, we can treat these counterexamples to BOB as dispositive, though we shall return to them later on in this section.

The proponent of BOB might respond by seeking refuge in the distinction between the contributory and the overall. To see how, consider that excuses operate at the overall level. In general, when someone does something, it is only if there is undefeated reason for them not to have done it that they stand in need of an excuse. The same applies if what they did was in breach of obligation: it is only if the breached obligation happens to be undefeated – if it lies at the overall level – that the question of excuses can sensibly arise. If the breached obligation lies only at the contributory level, so that it is defeasible by other considerations, then its breach is not automatically the sort of thing to which excuses can apply. When an obligation is defeated by competing considerations, so that breach is justified overall, no excuse is necessary: we do not need an excuse for what we were justified in doing. The proponent of BOB may thus attempt to evade the counterexample by saying that ‘obligation,’ as it figures in the premiss, is meant in the contributory rather than the overall sense.

The problem with this manoeuvre is it faces counterexamples of its own, this time from justified breach of contributory obligation. We already leaned on the intuitive possibility of such cases in the last paragraph, and concrete examples are not too far to seek. Consider promises that are broken because something comparatively important arises, as when a firefighter on call breaks a promise to attend her boy’s birthday party due to an emergency response request citing immediate threat to life. Or consider the institutionalised practice of justificatory defence in criminal law procedure, the logic of which begins, as with all legal defences, by conceding the commission of an offence, and proceeds to argue that this was in fact justified under the circumstances.Footnote ¹⁷

Once we accept that such cases are possible, it is a short step to seeing that they are counterexamples to this reading of BOB. Sometimes, as we said, we have an undefeated reason to breach a contributory obligation. When we do so for that undefeated reason, we are justified in so doing, and that seems to get us off the blame hook.Footnote ¹⁸ Pressed to say why, the answer, roughly, is that blame criticises how we fare as responders to reasons: it criticises that and how we failed to satisfy the balance of reasons we were faced with. So, if we acted in accordance with and for an undefeated reason, there is nothing to critique in this regard: we are not to blame. This makes justified breach of contributory obligation a straightforward counterexample to BOB on this reading.

Does the proponent of BOB have any other options left at this stage? One final strategy returns us, as promised, to a version of the overall reading. The challenge for any such view, we saw, is how to deal with cases of excused breach. I believe the proposal is one that many tacitly endorse, though David Owens makes it explicit when he says, ‘Suppose the agent breaches a binding obligation. Then ceteris paribus we would expect the agent to feel guilty and others to blame them for the breach’ (Reference Owens2016: 5), and then includes amongst the ceteris such qualifications as that ‘the [obligor] has an excuse’ (Reference Owens2016: 5, n. 9). Owens is asking us to read a ceteris paribus clause into BOB, such that breach is blameworthy only other things equal, where other things are not equal when (for example) the breach is excused. Thus modified, BOB is no longer threatened by cases of excused breach, because it no longer claims unqualified entailment from breach to blameworthiness. After all, it no longer denies that where breach is excused this inference cannot be drawn.

An immediate question is how exactly the ceteris paribus clause is supposed to make the inference from breach to blameworthiness defeasible. There are two ways of understanding the defeasibility at issue. The first is to understand it consistently with what logicians would call the ‘monotonicity’ of entailment. On this approach, the ceteris paribus clause makes the inference defeasible because it expresses a suppressed conjunct in BOB’s antecedent, which if false prevents the inference from being drawn in the usual way. In this case, the suppressed conjunct comes out true only if there is no excuse. To spell it out, with the ceteris paribus clause included, BOB now reads: if someone is in breach of obligation and unexcused, then they are blameworthy. Hence, for someone who is in breach but has an excuse, the modified BOB straightforwardly does not license the conclusion that they are blameworthy.

The second way to understand the defeasibility introduced by the ceteris paribus clause does away with the monotonicity of entailment. On this approach, the inference from breach to blameworthiness is defeasible by an excuse not because the antecedent harbours a suppressed conjunct, but because the inference itself may be retracted depending on the informational context. To illustrate the idea, imagine a list comprised of two premisses: that breach entails blameworthiness (i.e. the unmodified BOB), and that someone is in breach. From this list it can be inferred that that person is blameworthy. But whilst a monotonic logic allows the inference to be drawn no matter what premisses we add to the list, in a ‘nonmonotonic’ logic the inference is retractable depending on the premisses we add. For example, if we add that the person also has an excuse, then in a nonmonotonic logic this new information may allow for the original inference to be retracted, such that we may no longer conclude that the person is blameworthy. On the nonmonotonic approach, the ceteris paribus clause makes the inference from breach to blameworthiness defeasible by encoding this information-dependent retractability.

Whichever way we understand the defeasibility introduced by the ceteris paribus clause, the overall reading of BOB that results can handle cases of excused breach. But equally, however, it will be unable to handle another counterexample, this time from justified breach of overall obligation. Consider, for example, that we sometimes face dilemmas between incomparable and mutually unsatisfiable obligations. A famous case is owed to Sartre (Reference Sartre, Macomber and Kulka2007: 30), or at least a student of his during World War II, who faced ‘the choice of going to England to join the Free French Forces – which would mean abandoning his mother – or remaining by her side to help her go on with her life’. This is clearly a dilemma between incomparable and mutually unsatisfiable obligations. By their incomparability neither obligation defeats the other, and so both are undefeated in the circumstances, making them considerations at the overall level. This means it would be justifiable for the student to act in satisfaction of either obligation, so that either obligation is justifiably breached in the course of satisfying the other. Thus, if he acts in satisfaction of either obligation for that reason – if he is justified in breaching the other obligation – then he cannot, as we said earlier, be blameworthy for that. We therefore have a straightforward counterexample to any ceteris paribus version of BOB on its overall reading.

Perhaps the proponent of BOB will reply that the ceteris paribus approach can accommodate the new counterexample, by making other things unequal not only when breach is excused but also when justified. Maybe so. But there is a deeper problem with any ceteris paribus proposal, whether it makes the blameworthiness of breach defeasible (monotonically or not) by excuses, justifications, or indeed both. To see the problem, let us return to the gentleman in Stevenson’s anecdote. The gentleman, by hypothesis, is faultless, meaning that his inability to help his mother is justified, or at least excused, under the circumstances. Since ceteris paribus modifications to BOB will therefore make his breach blameless, those who appeal to them will have a hard time explaining exactly what apt response it is that the gentleman’s tears express. It cannot be any form of self-blame, after all, since the ceteris paribus clause tells us that that would be pathological. The problem, then, is that this approach renders proponents of BOB unable to take the gentleman’s feelings of responsibility seriously. My point is that whether or not they in fact care to do so (as hard anti-moralists do, but some moralists do not), they should care to do so, though any ceteris paribus approach will rule out this possibility.

This concludes my case against BOB. What follows from this in the context of our dialectic? Firstly, it follows that we should reject the moralist’s argument, and with it, moralism about obligation. Secondly, in tandem with the section 2 defence of BEF, it follows, against the hard anti-moralists, that we should soften our anti-moralism.

4. Interpreting feelings of responsibility

What distinguishes anti-moralists as such is their ability to accept the faultlessness of the gentleman in Stevenson’s anecdote without pathologising his powerful feelings of responsibility. Hard anti-moralists do this by first using BOB to interpret him as blameworthy, and then, by denying BEF, allowing that to stand in the absence of fault. Their softer counterparts instead deny BOB, interpreting the gentleman’s feelings as apt, but as apt agent-regret rather than any form of blame. I have argued that we should reject moralism, but that our anti-moralism should be soft rather than hard. In this final section, I address a major challenge for the soft anti-moralist appeal to agent-regret – a challenge that, if left unanswered, leaves us with no way to represent the gentleman in Stevenson’s anecdote felicitously.

Agent-regret has several features that make it especially suitable for interpreting the phenomenology in cases like our gentleman.Footnote ¹⁹ First of all, it is, like blame, a feeling of responsibility, which is to say, a negative retrospective feeling about the undesirable results of our agency as such.Footnote ²⁰ It captures how the gentleman’s despair tracks his inability to cover his mother’s medical costs – to live up to that important reason for action, an obligation.Footnote ²¹ In this, agent-regret is more suitable than what Lucy McDonald (Reference McDonald2021: 143) would call ‘non-agential shaming’, because such shaming, in her analysis, does not involve holding people responsible for anything at all. As mentioned already, agent-regret is also unlike blame in being capable of aptness in the absence of fault, as appears to be the case with the feelings of our gentleman. This also makes it more suitable than what McDonald would call ‘agential shaming’, since, if we accept her analysis, ‘all agential shaming counts as blaming’ (ibid.). But even if we do not, she is surely correct to hold that shaming, in general, ‘must invoke the judgment of an audience, real or imagined’ (Reference McDonald2021: 142). And that is not obviously needed to understand what our gentleman feels: his despair is rooted in his love of his mother, independently of what anyone else (real or imagined) thinks about his inability to live up to it. So, even if agential shaming is not itself a form of blame, it is less suited to interpreting the gentleman’s feelings of responsibility than is agent-regret.

But agent-regret prima facie fares worse than all the above alternatives in a potentially non-negotiable respect: it is not obviously a moral emotion. Consider, for example, what Marcus would say about interpreting the gentleman’s feelings in terms of agent-regret. She would say that ‘to describe [his] feelings […] as ‘regret’ seems inadequate’ (Reference Marcus1980: 130, n. 9), and that ‘To insist that ‘regret’ is appropriate rather than ‘guilt’ or ‘remorse’ is false to the facts’ (Reference Marcus1980: 133, n. 12). We find the same point in Marcia Baron, who writes that agent-regret ‘seems out of place … because it is so mild’, and that whatever the relevant feeling, it has to be ‘more than regret’ (Reference Baron1988: 261). In other words, although agent-regret may indeed be a feeling of responsibility capable of aptness in the absence of fault, it is not, or not sufficiently, possessed of a ‘moral’ quality that would do justice to the phenomenology as it strikes us. As far as feelings of responsibility with a special ‘moral’ quality go, forms of blame like guilt or remorse (or perhaps agential shaming) would appear to fit the bill. If that is all the canonical moral psychology has to offer, this leaves the anti-moralist with little choice, to put it crudely, but to go hard or go home. It is incumbent on soft anti-moralists to explain the special ‘moral’ quality of agent-regret in cases like the gentleman in Stevenson’s anecdote.

I agree with Marcus and Baron that, ordinarily, agent-regret is all too quotidian. Other things equal it can befit a failure to satisfy just about any reason for action, very trivial ones included.Footnote ²² It is the kind of feeling that can be appropriate to the bus we just missed (precipitating a twelve-minute wait), or to the unordered tiramisu that now looks so delicious, and so on. We may want to say that a special ‘moral’ quality accrues to agent-regret insofar as it attends the failure to satisfy very weighty reasons. But I do not think that will do. Weight is a quantitative notion, but we are after a qualitative difference. No matter how long the wait for the next bus, or how big or delicious the tiramisu, our agent-regret will intuitively never amount to what the gentleman in Stevenson’s anecdote feels, at least other things being equal. Or at least, if it does, then the explanation will not lie in the weight of the reasons left unsatisfied.

However, this line of thought does, I believe, point towards a more promising explanation. When thinking about practical normativity, the general idea that the force of some reasons for action is qualitatively distinctive in that no quantity of weightiness can ever add up to it – such that they can trump certain other reasons for action, however weighty they may be – has often been explained in terms of obligation, or related deontic categories. So, if we are looking to explain certain agent-regrets as qualitatively distinctive, and we intuit that this cannot be secured by the weight of the reasons left unsatisfied, then this may well indicate that it is not their weight but their obligatory force that is in fact doing the work. Let me explain why I think this is indeed the case.

I believe, as Joseph Raz sometimes suggests (Reference Raz, Hacker and Raz1977: 223–4; Reference Raz1994: 40), that the normative force of obligations as reasons for action is explained by their having two independent features.Footnote ²³ One is that they are normatively structured to exclude certain countervailing considerations, however weighty they may be, from guiding our deliberations about what to do. They are not only reasons to do what they say, but also reasons not to act for certain reasons to do otherwise. This gives them the character of requirements to do what they say, that is, even if what they say to do is unsupported by the balance of all reasons (including the excluded ones). It is their character as requirements that explains why obligations can trump other reasons (i.e. excluded reasons) however weighty they may be.Footnote ²⁴ The second feature of obligations is that their existence, unlike many of our reasons for action, is independent of our conative attitudes: our goals, plans, intentions, desires, and the like. Our obligations are attitude-independent and in this sense inescapable at will, that is, by simply revising our conative attitudes.Footnote ²⁵

Some requirements are escapable at will, and to that extent, they strike us as not obviously obligatory in force. For example, suppose we forge a New Year’s resolution to lose weight. We may, on that basis, have a dietary regime that requires us to make a salad for dinner, no matter how delicious, cheap or convenient it may be to order a takeaway instead – those considerations are excluded from guiding our dinnertime deliberations. But we are reluctant to say we thereby have an obligation to make a salad, and this, I take it, is because the dietary regime would no longer require us to do so if we were to stop caring about the New Year’s resolution. Conversely, some inescapable reasons lack requiring force, and that, equally, seems to preclude their being obligations. For example, if we have reasons to give to charity, then we have those reasons whether or not we are attitudinally disposed to do so. But if we do not give to charity because there are weightier reasons not to do so, better ways to use our efforts and resources, then that is perfectly fine – there is no requirement to do so. After all, if we were so required, we would be obligated to give to charity, and that would go against our understanding of charitable giving as beyond the call of duty. I think that, taken together, such examples lend support to the idea that reasons for action are only obligations when they are at once inescapable and requirements.

To see how this could help explain why agent-regret has a special ‘moral’ quality, when it does, we must first consider what goes on when we leave ordinary reasons for action unsatisfied. Suppose you are at a flea market and you stumble upon a very rare bike component that would make or break a vintage restoration project you have in mind. It is a nice project, but it is not as though you feel the need to break the bank over the component, or go out of your way to find it, or anything like that. But it has fallen into your lap, and at a bargain price. The only thing is, the vendor does not accept card payments, so you need to pop to the cash machine. You go round the corner to withdraw the cash, and come back, only to find that the vendor has mysteriously disappeared, taking the component with him. Kicking yourself, you feel, aptly, a pang of agent-regret corresponding to the reason you have to acquire the component, which you have left unsatisfied.

In a case like this, the normative sway of the reason that makes it apt to feel agent-regret can be counteracted in two ways. Firstly, your reason to acquire the component is attitude-dependent, in that it depends for its existence on your plans for the vintage restoration project. Thus, you can escape the reason by revising your attitudes, perhaps by coming up with a new plan, one that can be realised without that particular component. By escaping the reason you left unsatisfied, you escape what makes your agent-regret apt in a perfectly ordinary way: you are moving on from your misery by letting go of what makes it fitting. Secondly, even if you struggle to let go of your original project – it was a nice project, after all – there may come a point where your still feeling agent-regret is getting in the way of living your life, or enjoying it, such that you have weightier reasons of psychological convenience to repress your apt agent-regret. Such reasons of psychological convenience are not trumped by the reason that makes your agent-regret apt, because that reason is not a requirement. If the reason did not exclude such countervailing considerations before you left it unsatisfied, it presumably remains the same afterwards.

I suspect it will be clear by now what I have in mind.Footnote ²⁶ Obligations, unlike ordinary reasons, are inescapable requirements, and so agent-regret aptly felt as a result of breach is resistant to being dislodged in either familiar way. We cannot simply let go of agent-regret following breach because obligations are attitude-independent reasons, they are inescapable by simply revising our conative attitudes. Moreover, it is hard to justifiably repress agent-regret following breach because obligations are requirements, so that they trump even the weightiest of reasons of psychological convenience, reasons that would otherwise guide us to repress the agent-regret we are apt to feel. As Gardner says (Reference Gardner2005: 105), ‘Wrongful action … leaves us with regrets that are hard to expunge and the repression of which is hard to justify’.

My suggestion, then, is that the agent-regrets we are apt to feel in the wake of breach of obligation exhibit the following ‘moral’ quality: they are such as to plague us, to keep us up at night, because we cannot let go of or justifiably repress them. I believe this does justice to what the gentleman in Stevenson’s anecdote feels, a feeling of responsibility for his faultless inability to help his mother through her illness, one so powerful and painful as to reduce him to tears. As such, the account enables those who oppose moralism about obligation – those who sympathise with Stevenson – to soften their anti-moralism, and thus avoid the hard anti-moralist misrepresentation of the gentleman as blameworthy.