Doing, Allowing, and Double Effect

Two patients are dying of the same disease. Both have asked to be allowed to die. In the first case the doctor administers a drug that stops the heart. In the second the doctor withholds the antibiotic that would clear the pneumonia, and the patient dies of that instead, four days later.

The outcome is the same, the motive is the same, the request is the same, and the doctor's state of mind is as close to identical as two states of mind get. Most legal systems treat the first as murder and the second as ordinary practice. Most people's initial reaction agrees with the law and cannot immediately say why.

Whether the reaction survives examination is one of the two questions in this article. The other concerns what an agent aims at as against what they merely expect, and the two questions are usually run together though they are entirely separate.

Two distinctions that are not the same

The doctrine is stated in terms of acts and omissions, and the distinction that actually does the work is between doing harm and allowing it. These come apart, and it is worth seeing how before going further.

A conductor who ruins a concert by not turning up has brought about the ruin, and has done so by an omission. A person who lets a child starve, and spends the relevant afternoon on the telephone to their solicitor, has allowed the death, and has spent the time acting. Whether an agent's body moved is not what the moral question is about. What it is about is whether the agent initiated or sustained the sequence of events that led to the harm, or whether the sequence was already running and the agent failed to intervene.

Philippa Foot's account puts it in those terms, and Warren Quinn's refined it: what matters is whether the agent's most direct contribution to the outcome was something they did or something they did not do. On these accounts the doctrine is not about bodily movement at all, and stating it as a doctrine about acts and omissions invites objections that miss it.

The bare difference argument

James Rachels attacked the distinction in a short paper in the New England Journal of Medicine in 1975, and the method he used has been standard ever since. It works by constructing two cases that differ in the one respect under examination and in nothing else, and then asking whether the moral judgements differ.

His cases are these. Smith stands to gain a large inheritance if his young cousin dies. He waits until the child is in the bath, holds him under, and arranges the scene to look like an accident. Jones stands to gain the same inheritance in the same circumstances, goes to the bathroom with the same plan, and finds that the child has slipped, struck his head, and is face down in the water. Jones watches, ready to push the head back under if necessary, and does nothing. The child drowns.

Smith killed. Jones let die. Rachels's claim is that no one who hears the two cases thinks Jones behaved even slightly better, and that if killing were worse than letting die in itself, Jones's conduct would have to come out better than Smith's by that amount. Since it does not, the difference between killing and letting die carries no moral weight of its own.

He then turned this on medical practice. A doctrine that permits withholding treatment while forbidding a lethal injection sends a patient to a slower and more painful death for the sake of a distinction that does no moral work. Worse, it makes life and death decisions turn on irrelevancies, since whether a condition happens to be treatable determines whether the doctor may let the patient go.

The argument is powerful and it has not been abandoned. What has happened is that its scope has been narrowed.

What the argument establishes and what it does not

The most damaging reply is about the choice of cases. Smith and Jones are both monstrous, both act from greed, both intend the child's death, and both are prepared to kill if they must. Given all that, it is unsurprising that they come out equally condemned. What the comparison shows is that killing and letting die make no difference when everything else is held fixed at that particular setting, and the setting is one in which the agent has already resolved on the death.

That leaves it open that the distinction does work elsewhere. Compare instead a person who kills a stranger to take their organs with a person who fails to send money that would have saved a stranger's life. Here almost everyone judges very differently, and the difference is not obviously an error.

Behind that lies the structural point developed by Foot and Quinn. A duty not to kill is a duty of restraint: it can be discharged by everyone at once, it specifies exactly what each person is to do, and it can be complied with perfectly. A duty to save has no such shape. There is no limit on how many people need saving, saving them costs something, and the duty does not by itself say who is to bear the cost. Rights against interference and rights to assistance therefore behave differently, and a moral system that treated them identically would either demand almost nothing or demand a person's whole life.

Jonathan Bennett pressed the case for Rachels from the other side in The Act Itself in 1995, arguing that the distinction can be drawn in purely descriptive terms, roughly in terms of how many of the agent's possible behaviours would have produced the outcome, and that once drawn that way it is obviously without moral significance. The debate since has largely been about whether any descriptive account of the distinction both captures the cases correctly and explains why it should matter, and no account commands agreement.

Where things stand is worth stating plainly, because the literature is often summarised as though one side had won. Rachels showed that the distinction cannot be doing the work that a blanket rule against active killing assumed. He did not show that it never does any work, and the strongest current positions hold that it makes a difference which can be outweighed rather than one that settles cases.

Aiming at a harm and expecting it

This is the second distinction and it is independent of the first. An agent can do a harm intentionally, do one as a foreseen side effect, allow one intentionally, or allow one as a side effect, and the four combinations behave differently.

The doctrine of double effect is the traditional principle built on it. Its usual statement requires four things: that the act itself be good or at least morally neutral, that the good effect and not the bad effect be what the agent intends, that the good effect not be produced by means of the bad one, and that there be a proportionately serious reason for permitting the bad effect at all.

The lineage runs back to Aquinas, who discussed killing an attacker in self-defence in the Summa Theologiae, holding that one act can have two effects, that the act is characterised by what the agent intends, and that a defender may intend to preserve their own life while the attacker's death is outside that intention. Philippa Foot brought the principle into modern philosophy in a paper of 1967, and the whole subsequent literature on trolley cases descends from that paper.

The pairs the principle is meant to handle are these.

A bomber destroys a munitions factory knowing that civilians living nearby will be killed by the blast. Another bomber kills the same number of civilians deliberately, in order to break the enemy's will to fight. The number of deaths is the same and the second is standardly judged much worse. Double effect explains why: the first foresees the deaths, the second uses them.

A surgeon removes a cancerous uterus, knowing the foetus will not survive. Another surgeon crushes the skull of a foetus in obstructed labour to save the mother's life. Again the outcomes are similar and the traditional verdicts differ, and again double effect gives the reason: in the first case the death is a side effect of removing diseased tissue, in the second it is the means by which the mother is saved.

Two features of the principle deserve emphasis. It is permissive, not prohibitive: it says when a foreseen harm may be permitted, and it never licenses anything on its own, since the proportionality condition still has to be met. And it is a constraint on the agent's reasoning rather than on the world, which is exactly what makes it controversial.

Where the two distinctions meet

The trolley cases, which the ethics tier introduced, exist to separate these two distinctions from each other and from everything else. Their interest is that in one version the death of the one is a side effect of diverting a threat, and in the other it is the means by which the threat is stopped, while in both the outcome is one death rather than five. Someone whose judgements differ between the versions is committed to something, and the whole apparatus of this article is an attempt to say what.

It is worth noticing that no single distinction accounts for all the standard reactions. The doing and allowing distinction does not separate the two trolley versions, because the agent acts in both. Double effect separates them, and does not by itself explain the difference between killing a stranger and failing to save one, since in the second case nothing is intended at all. Anyone who wants a principled account of the ordinary pattern of judgements needs both distinctions, and probably more.

Next

This article has treated the difference between causing a harm and failing to prevent one as a difference in strength between two duties, without saying much about the weaker one. What is owed to a stranger in trouble, whether there is any limit on it, and whether the good things a person does beyond what is required have a category of their own, are the subjects of the last article in this tier.