Supererogation

Why is the heroic act merely praised rather than required, if more good is always better?

A soldier throws himself onto a live grenade and dies so that four others live. We call it heroism. What we do not say is that the four who hesitated failed in their duty, or that the man beside him is a wrongdoer for not having jumped first.

Hold those two judgements together and something is strange. If the act was the best thing anyone could have done, and morality is in the business of telling us to do the best thing, why is it not required? And if it was required, why does no one blame those who did not do it? J. O. Urmson pressed exactly this in a 1958 paper called "Saints and Heroes," arguing that our inherited three-way sorting of acts — obligatory, permitted, forbidden — has no shelf for the act in question.

Try the simplest fix first: deny the category. Perhaps morality really does require the grenade and our reluctance to blame is squeamishness. Shelly Kagan argued something close to this, that the supposed permission not to maximise is unearned. The position is coherent, and its cost is visible: every waking choice becomes a case of falling short, since there is nearly always something better available.

So look at what distinguishes the two verdicts. "Wrong" is not simply the negative of "good." Wrongness is bound up with what may be demanded: it licenses blame, reproach, the sense that the person owes an account. Mill made this explicit — to call an act wrong is to say it is fit to be punished, by law, by opinion, or by the person's own conscience. Read that way, asking whether the grenade is obligatory is asking whether you could stand at the funeral and reproach the survivors. Could you?

That reframes the puzzle. The question is no longer "which act produces most good" but "what may one person demand of another." And the shape of the answer appears: the cost of the heroic act falls entirely on the agent, and it is total. A morality that could compel that leaves no room for a person having a life of their own — projects, attachments, a self not wholly at the disposal of the aggregate. Samuel Scheffler called the permission that protects this an agent-centred prerogative: a licence to weight your own interests above what an impartial calculation would allow.

A second thread is harder to make precise but hard to dismiss. Some of the act's value seems to depend on its not having been owed; a gift you were legally obliged to make is not a gift. If heroism were conscripted we would still want the four men saved, but the thing we admire would be gone — so a morality requiring everything good would abolish a good it wanted.

Two honest complications. First, none of this tells you where the line sits, and the line is disputed. Peter Singer's 1972 argument about famine relief attacks precisely that gap: if you can prevent something very bad at small cost, you must — and once that is granted, much of what we file under optional generosity moves into the required column without any hero being asked to die. Second, traditions disagree about whether the category exists. Act-utilitarianism struggles to make room for it, and Kantian ethics gives duties of beneficence latitude about when and how much, with scholars still arguing whether that latitude generates real supererogation or conceals it inside imperfect duty.

The refinement that makes the rest hang together is dropping the assumption Urmson exposed: that better always implies more required. Call it the good-ought tie-up. Once "obligatory" is understood as demandable rather than highest-scoring, the gap between them is not an anomaly but the expected result of two different questions — how much good does this do, and how much may be extracted from this person.

That also fixes the relation to its neighbours here. Doing versus allowing asks about the causal route to a harm, holding the outcome fixed; supererogation holds the causal route fixed and asks about the deontic status line — required or merely admirable. Anonymous giving asks what the giver's motive was; supererogation is indifferent to motive, since a grudging rescue at great cost is still beyond duty. Cost to the agent does the work here, and it is the fixed point of difference in both comparisons.

One consequence is that the gap should have a mirror image, and it does. If there are acts good but not required, there should be acts bad but not forbidden — the small meanness you may permissibly commit and be thought worse for. Julia Driver named these suberogatory in 1992. That the same seam opens on both sides is evidence the three-way scheme is genuinely incomplete, rather than that we are being soft about heroes.

The disagreement is conceptual, and worth labelling as such. What people praise and blame can be measured, and moral psychology does measure it — but a finding that most people decline to blame the non-hero does not show they are right to. Progress would look like a defensible account of what makes a demand demandable, or a debunking showing the praise-blame asymmetry tracks the enforcement costs of our practices rather than anything moral.

• The demandingness objection to consequentialism, and whether agent-centred prerogatives can be justified rather than merely wanted.
• Moral saints: Susan Wolf's argument that a life of maximal moral devotion is not one to want.

Every term the collection defines is gathered in the glossary.