It is often assumed that the rejection of the doing–allowing distinction had far-reaching consequences for morality. As Peter Singer famously writes:

Singer, like other consequentialists, is, to put it mildly, skeptical of the moral relevance of the doing–allowing distinction. He does, therefore, accept the claim that one ought to channel all of one's disposable income into assisting the extremely poor, as not to do so would be tantamount to harming and, in many cases, killing the potential recipients of assistance.Footnote ² Obviously, the underlying assumption here is that harming or killing these recipients is morally unacceptable. When the above argument has been criticized, it has usually been on the basis that the doing–allowing distinction (or some other related variant of this distinction) is in fact morally relevant.Footnote ³ It has, therefore, been generally assumed among the critics of the argument that if one gives up the doing–allowing distinction, then the drastic consequences identified by Singer and other consequentialists follow.

What those critics have missed, in my view, is that the argument put forward by Singer and others suffers from a deeper flaw than that they have identified. The flaw can be identified by considering a crucial premise of the argument which is that harming or killing innocent people is morally unacceptable. For one's intuitions about the badness of killing themselves depend on an implicit distinction between killing and letting die. And it is the rejection of that distinction which is supposed to give rise to the consequences identified by Singer and others. The argument that the rejection of the doing–allowing distinction has the radical consequences suggested is thus at root incoherent, for it depends on that distinction. In this article I would like to develop this train of thought.

Most people accept the view that there is a significant moral difference between killing a person and letting her die; killing is usually considered to be much worse than letting die. This, of course, implies that the distinction between killing and letting die is morally relevant. The killing–letting die distinction is an instance of a more general distinction between doing and allowing harm. Those who are skeptical of the moral relevance of the doing–allowing distinction have tried to argue either that (a) there is no workable criterion for telling when a certain harm is being done by a given agent, and when it is merely being allowed by her, or that (b) even if we can come up with a criterion for applying the doing–allowing distinction, there is no reason for taking the distinction in question to have any moral significance.

I do not in this article take a position on the doing–allowing distinction. However, in order to give those who reject the distinction (and draw radical consequences from that rejection) a run for their money, I will assume for the sake of argument that we should reject the moral relevance of the doing–allowing distinction, so that either (a) or (b) is correct. It follows that whether the outcome of a behavior is being done or is merely being allowed is not relevant to the moral assessment of that behavior. Some of the consequences of that rejection are apparently counter-intuitive, a fact accepted by a majority of its proponents. Given, for example, that it is morally forbidden for me to kill a person in order not to lose $10,000, then it would also be forbidden for me not to give up $10,000 needed to save a person's life, if the doing–allowing distinction is considered to be morally irrelevant.Footnote ⁴ It appears then that rejecting the moral relevance of the doing–allowing distinction expands considerably the demands morality places on us. As Jonathan Bennett succinctly puts it:

If one breaks down the argument about the consequences of rejecting the doing–allowing distinction into its basic parts, it takes the following form, when applied in the context of killing and letting die.

Premise 1 of the above argument states that one is never permitted to engage in non-optimal killing. There is no doubt that this premise has a very strong intuitive appeal. It occupies a secure part of our ordinary morality. It is also endorsed by deontological as well as consequentialist moral doctrines.Footnote ⁶ Premise 2 is a direct corollary of rejecting the moral relevance of the doing–allowing distinction. The conclusion of the argument, which clearly follows from 1 and 2, is that one is never morally permitted to engage in non-optimal acts of letting die. This claim, however, has clear counter-intuitive implications, since it leads to over-demanding morality. It implies, for example, that other things being equal I am morally required to sacrifice my life, if doing so is necessary for saving the lives of two other persons, even if I have no connection towards these people.

Given the strong intuitive appeal of premise 1 and the validity of the argument, the most attractive option for those who want to reject its counter-intuitive conclusion is to reject premise 2, i.e. to reject the claim that there is no moral difference between killing a person and letting her die. In what follows, I will propose a different take on the argument. Instead of opting to reject premise 2, I will argue that the argument fails due to an internal inconsistency. More specifically, I will argue that the strong intuitive appeal of premise 1 depends on accepting the moral relevance of the doing–allowing distinction. In other words, I will argue that our confidence in the truth of premise 1 depends on the rejection of premise 2. This does not mean that the critics of premise 2 are wrong; only that even if premise 2 is correct then the argument is still fundamentally flawed.

Assume that premise 2 is true and that the argument is valid. We are then faced with a choice between rejecting the intuitively appealing premise 1 and rejecting the counter-intuitive conclusion on the one hand, and accepting premise 1 and accepting the counter-intuitive conclusion on the other. We must, in other words, either reject the claim that one is never permitted to engage in non-optimal killing or accept the claim that one is never permitted to engage in non-optimal letting dying. Faced with this choice, it is tempting to argue that we should opt for the latter, i.e. opt for accepting the conclusion rather than rejecting premise 1. This choice would be based on the fact that it is definitely more counter-intuitive to reject premise 1 than it is to accept the conclusion of the argument. We are less prepared to accept that one is sometimes permitted to (non-optimally) kill an innocent person, than we are prepared to accept that one is never permitted to (non-optimally) let an innocent person die.

There is, however, a serious problem with the above reasoning for accepting the conclusion of the argument rather than rejecting premise 1. For it relies on the comparative strength of the intuitive appeal of premise 1, as compared with the relatively unappealing conclusion of the argument. However, if the reason why premise 1 enjoys such a strong intuitive appeal is that it is about killing rather than merely letting die, then we cannot appeal to the strength of such an intuition if we reject the moral relevance of the killing–letting die distinction, which is precisely what premise 2 does. It is clear that the strength of the intuition we have in favor of premise 1, i.e. in favor of the claim that one is never morally permitted to non-optimally kill an innocent person, does depend on taking the killing–letting die distinction to be morally significant. This is shown by the mere fact that we do not have the same degree of moral intuition in favor of the claim that a person is never permitted to non-optimally let an innocent person die. For the only way of explaining the difference in our intuitions is by appealing to the killing–letting die distinction. Thus, by relying on the strength of the intuitive appeal of premise 1, advocates of the argument are appealing to the strength of a moral intuition which they are not entitled to appeal to, once they also accept premise 2, i.e. once they reject the moral relevance of the killing–letting die distinction.

The self-contradictory nature of the argument can be illustrated thus. The argument, as it stands, after rejecting the doing–allowing distinction, ‘levels up’ moral stigma attached to allowing to die, so that it is seen as being as bad as killing. If one takes seriously the rejection of the doing–allowing distinction, however, there is no reason why that should be the case. Why not ‘level down’ killing to the same level as allowing to die, so that we adopt as relaxed an attitude towards killing as we normally adopt towards allowing to die? One could with equal consistency take our relatively relaxed pre-theoretical intuitions about letting die, and apply them to killing, so that killing carries as little stigma as allowing to die. Hence, one could argue as follows:

The consequences of rejecting the doing–allowing distinction would, thus, be diametrically opposite to those adopted by Singer and others.

The point is not that advocates of the rejection of the doing–allowing rejection ought to ‘level down’ killing so that we adopt the same attitude towards it that we normally have towards letting die. Leveling down killing would, as much as leveling up allowing to die, depend on our pre-theoretical intuitions about the greater moral stigma attached to killing as opposed to allowing to die. Neither move is permissible if we are truly to reject the doing–allowing distinction. Premise 2 simply says that killing and letting die are morally equivalent, nothing more.

A defender of the argument might claim that, say, our pre-theoretical aversion towards killing (premise 1), or our pre-theoretical disinclination to accept the conclusion that we ought to level up allowing to die to the same level as killing, is based on something other than the killing–letting die distinction. However, this would be a difficult claim to sustain. The very fact that we label a certain type of action as ‘killing’ as opposed to ‘letting die’ inevitably brings into play the distinction between the two categories. If one thinks through the consequences of premise 2, therefore, it is clear that one cannot accept premise 1 as it stands. In order to be consistent with premise 2, premise 1 would have to remove any reference to killing. However, were one to remove from premise 1 the reference to killing and our strong intuitions on killing, one would be removing premise 1 altogether. One would be left with premise 2 on its own, from which nothing would follow.

In ordinary morality, we take certain considerations, besides the overall good that results from an act, to be relevant to the moral assessment of that act. We do treat acts that differ in relation to those considerations differently. On the basis of those considerations, we take certain acts to be morally forbidden and others to be morally permissible. The doing–allowing distinction is an example of such a consideration. Suppose, however, that we came to think that some of these considerations, which we used to take to be morally relevant, are in fact morally irrelevant. The only thing that would follow from the mere rejection of the moral relevance of these considerations is that we are no longer justified in treating differently acts that differ merely with respect to the considerations in question. However, in trying to achieve this equality of treatment, we cannot appeal to moral intuitions that are themselves based on assuming that these considerations are morally relevant.

Applying the above general reasoning to the case of the doing–allowing distinction, I have argued in this article that rejecting the moral relevance of the latter distinction implies only that doing harm is morally equivalent to allowing it, i.e. that both should be treated in the same way. The rejection of the distinction does not have implications as to how this similarity of treatment should be achieved. It remains open whether this similarity of treatment should be achieved by changing the way we usually treat cases of doing harm, by changing the way in which we usually treat cases of allowing harm, or by changing the way we usually treat both. More importantly, I have argued that in trying to opt for one way of achieving similarity of treatment between acts of doing harm and acts of allowing harm, one cannot appeal to intuitions – such as those we have against doing harm – which are themselves grounded on accepting the moral relevance of the killing–letting die distinction.Footnote ⁷