Are animals – and persons – morally “replaceable”?

Over the years, Peter Singer has frequently been vilified for his claim – initially advanced in the first edition of his Practical Ethics in 1979 – that sentient individuals that are wholly non-self-conscious are “replaceable” (Singer, 1979, pp. 99–103). By this he means that it can be permissible to kill a non-self-conscious individual painlessly, provided that one would also cause another individual to exist whose life would contain at least as much of what is intrinsically good for individuals as the remaining life of the other individual would contain were that individual not to be killed. While Singer initially made this claim about non-self-conscious animals, he had come at least by 1985 to realize, as he says explicitly in the third edition of Practical Ethics in 2011, “that if replaceability holds for animals, it must hold for humans at a similar mental level” (Singer 2011, pp. 106–107). ⁽¹⁾ Thus, what caused him to be publicly vilified was not his initial claim that non-self-conscious animals are replaceable but his later claim that human fetuses and newborn infants are as well.

I believe that there is much to be said in support of what we can call Singer’s “replaceability thesis”, though the version of this thesis that seems to me most defensible is weaker than Singer’s and has a somewhat different basis in moral theory. This means that not only Singer but I too must provide an explanation of why it is that persons – that is, individuals who are self-conscious and to some degree rational and autonomous – are not similarly replaceable. In this essay I will explain and examine the arguments that Singer advances both for the replaceability of non-self-conscious individuals and for the non-replaceability of persons. I will also explain and defend my own views about why some individuals are in some circumstances replaceable while others are not.

Because many of the arguments that follow will be critical of Singer’s arguments, I want to stress how privileged I feel to be able to contribute to this special issue in his honour. I first encountered Singer’s work soon after I began my study of philosophy as a student at Oxford, and the effect was revolutionary. I immediately saw in his work the ideal model of all I wanted the effect was revolutionary. I immediately saw in his work the ideal model of all I wanted to try to accomplish in philosophy. Along with my two subsequent graduate supervisors at Oxford – Jonathan Glover and Derek Parfit – Singer, whom I met only many years later, has done more than any other moral philosopher to inspire and, through his work, forge my own thinking. His influence on my life and work, both morally and philosophically, has been both enormous and invaluable. He is, to me, the very paradigm of a moral person and a moral philosopher.

When Singer published the first, second, and third editions of Practical Ethics, he was an avowed adherent of the form of utilitarianism – “preference utilitarianism” – that was developed and defended by his Oxford mentor, R. M. Hare (Hare, 1981). Singer succinctly states this theory’s explanation of the wrongness of killing as follows:

According to preference utilitarianism, an action contrary to the preference of any being is wrong, unless this preference is outweighed by contrary preferences. Killing a person who prefers to continue to live is therefore wrong, other things being equal

A wholly non-self-conscious individual, however, has no conception of itself as having a future life in prospect, and cannot, therefore, desire to continue to live. According to pure preference utilitarianism, then, there is no moral objection to killing non-self-conscious animals – or human fetuses or newborn infants, as they too are not self-conscious and cannot desire to continue to live.

This is intuitively highly implausible. Singer’s response is to recognize an alternative explanation of the wrongness of killing non-self-conscious animals, fetuses, and infants. This is that such killings are impersonally bad insofar as they prevent the occurrence of experiences of pleasure or happiness that the individuals would otherwise have had. These experiences are instances of what Singer refers to as preference-independent value. Singer thus defends a pluralist moral view that encompasses both preference-dependent value and preference-independent value (Singer, 2011, p. 117). The recognition that there are both these forms of value enables Singer to produce an argument for the claim that, while non-self-conscious individuals are replaceable, persons are not.

Non-self-conscious individuals are replaceable because the value that is lost when one of them is killed is preference-independent and impersonal. It makes no difference which individual’s life this value occurs in. Hence, if one prevents a certain amount of preference-independent value from coming about by killing a non-self-conscious individual but also creates at least as much of the same sort of value by causing another individual to exist, there is no net loss of value. And there is, Singer contends, a moral reason to cause the other individual to exist. “We do seem”, he writes, “to do something bad if we knowingly bring a miserable being into existence, and if this is so, it is difficult to explain why we do not do something good when we knowingly bring a happy being into existence” (Singer, 2011, p. 108). ⁽²⁾ It is this “something good” that morally counterbalances or offsets the loss of future happiness or pleasure when a non-self-conscious individual is “replaced”.

In support of the idea that the painless killing of one non-self-conscious animal can be morally offset by causing another to exist, Singer cites Derek Parfit’s Non-Identity Problem, illustrated by the following comparison. In one case, person P₁ could, without personal cost, correct a genetic defect in a fetus, thereby enabling the person who will develop from the fetus to have a life of substantially higher quality and longer duration; but P₁ is indifferent and allows the defect to remain uncorrected. In another case, person P₂ could prevent a person from coming into existence whose life would be of low quality and short duration and cause a different person to come into existence instead whose life would be of substantially higher quality and longer duration; but P₂ is indifferent and allows the less well-off person to come into existence. Parfit claims, and Singer agrees, that P₁’s and P₂’s acts are equally wrong, and that this shows that good and bad effects brought about by causing individuals to exist can be as important morally as effects for better or worse on existing individuals – though Singer would add the qualification, if there is no preference-dependent value at stake. This, Singer suggests, supports the view that the bad effect on an existing non-self-conscious individual of being painlessly killed can be fully offset morally by causing an equally well-off individual to exist.

In Singer’s view, one of the most important advantages of preference utilitarianism over hedonistic utilitarianism is that, while hedonistic utilitarianism implies that both non-self-conscious individuals and self-conscious persons are replaceable, all other considerations being equal, preference utilitarianism, while accepting that non-self-conscious individuals are replaceable, does not imply that persons are replaceable. Again, he writes of hedonistic utilitarianism that “if killing those leading pleasant lives is bad because of the loss of pleasure, then it would seem to be good to increase the number of those leading pleasant lives” (Singer, 2011, p. 87). Thus the loss of pleasure in killing one person can be outweighed by the gain in pleasure from causing a better-off replacement to exist. But persons typically have a strong desire to continue to live. When this desire is frustrated, there is a loss of preference-dependent value, which is morally more significant than a loss of an equivalent amount of preference-independent value. According to Singer, “the fact that one may desire to continue living means that death inflicts a loss for which the birth of another is insufficient compensation” (Singer 2011, p. 112). So perhaps it is “the capacity to see oneself as existing over time, and thus to aspire to a longer life … [that] is the characteristic that marks out those beings who cannot be considered replaceable” (Singer, 2011, p. 111).

Singer’s pluralistic view seems to be an instance of what Derek Parfit calls a “Two-Tier View” of moral reasons (Parfit, 2011, p. 219ff). There are both reasons to satisfy or not to frustrate preferences and reasons to create and preserve what is impersonally good, or valuable, and not to create or preserve what is impersonally bad. But reasons of the first sort – preference-dependent reasons – are stronger. For example, the reason to satisfy an individual’s desire for a certain pleasure is stronger than the reason to create the same pleasure in the life of a non-self-conscious individual that cannot desire that pleasure. It also may seem, though Singer does not explicitly say this, that the two types of reason are additive, in the sense that there are two independent reasons to provide pleasure for an individual who desires it: a preference-dependent reason and a preference-independent reason, each of which leads to the production of value when it is successfully acted on.

In any case, it seems that for this sort of Two-Tier View to avoid the implication that persons are replaceable, it must accept that not only that preference-dependent reasons are stronger than corresponding preference-independent reasons but also that preference-dependent value has something like lexical priority over preference-independent value. This is because, if killing a non-self-conscious animal results in a loss of impersonal value, and causing another to exist results in a corresponding gain, then causing a person to exist must involve an even greater gain in impersonal, preference-independent value, since the lives of persons typically contain substantially more pleasure or happiness than the lives of non-self-conscious animals. Thus, if preference-independent value were to weigh against and potentially outweigh preference-dependent value, there would then have to be some number of persons one might cause to exist whose total impersonal, preference-independent value would outweigh the frustration of an existing person’s desire to continue to live. Persons would then be replaceable, perhaps not on a one-for-one basis, but at least on a many-for-one basis.

In the second and third editions of Practical Ethics, Singer notes that, in a review of the first edition, Herbert Hart argued that preference utilitarianism is in fact no different from other forms of utilitarianism in implying that persons as well as non-self-conscious individuals are replaceable. All forms of utilitarianism require maximizing some measure of what is good, and thus, “if preferences, even the desire to live, may be outweighed by the preferences of others, why cannot they be outweighed by new preferences created to take their place?” (Singer, 2011, p. 113). In short, preference utilitarianism, understood as an impersonal maximizing moral theory, implies that it can be permissible to kill a person with a strong desire to continue to live if that is necessary to cause another person to exist who would have an at least equally strong desire to live that would be at least equally likely to be satisfied, if all other relevant considerations, such as side effects, would be equal.

In response to this challenge, Singer argues that the moral effect of the satisfaction of a desire that exists or will exist independently of one’s action is entirely different from that of the creation of a new desire that will be satisfied. He contends that if we ask ourselves whether we would welcome the acquisition of a new desire that would be satisfied, we will welcome the new desire only if it “is a means of achieving something that” we already desire. Otherwise, the acquisition of a new desire – one that is wholly unconnected with our preexisting desires – is unwelcome, as it comes as a “privation” that requires remedying. Thus, we do not, he observes, “make ourselves thirsty because there will be plenty of water on hand” to satisfy our desire for it. According to this “debit view of preferences”, as Singer calls it, “the creation of an unsatisfied preference” is merely the creation of “a debit in a kind of moral ledger” that is cancelled by the satisfaction of the preference. There is, therefore, no preference-dependent value in causing individuals to exist, and “the thwarting of existing preferences” is not “something that can be outweighed by creating new preferences – whether in existing beings or in beings we bring into existence – that we will then satisfy” (Singer, 2011, pp. 113–114). While there may well be a moral reason to bring a new individual into existence, it must be a preference-independent reason concerned with impersonal value rather than a preference-dependent reason. As such, it cannot outweigh the preference utilitarian reason not to frustrate a person’s desire to continue to live.

It is thus this debit view of preferences, together with the priority of preference-dependent value over preference-independent value, that excludes the replaceability of persons. Indeed, Singer explicitly acknowledges that an “objection to th[is] debit view” is also “an objection to my attempt to argue that persons are not replaceable” (Singer, 2011, pp. 115–116).

There are, I believe, decisive objections to the debit view of preferences. This view, it must be conceded, does apply to some desires – namely, desires that Bernard Williams refers to as “conditional”. The satisfaction of such a desire matters only on condition that the individual who has it is going to remain alive. Singer’s example of the desire to drink water is often of this sort (though, as a long-time squash and tennis player, I can testify that if one is dehydrated enough, the satisfaction of a desire to drink water can be a source of intense pleasure). Conditional desires contrast with what Williams calls “categorical” desires, which give those who have them reasons to continue to exist in order to be able to satisfy them (Williams, 1973, pp. 85–88). A plausible form of preference utilitarianism must, I believe, accept that the creation and satisfaction of categorical desires has preference-dependent value.

While the satisfaction of a purely conditional desire contributes only to the avoidance of ill-being, the acquisition and satisfaction of categorical desires must contribute to our well-being on any view that recognizes preference-satisfaction as at least one dimension of well-being. After all, we begin to exist with no desires at all, and initially acquire only conditional desires. Only after we begin to mature psychologically do we begin to develop and satisfy categorical desires. If this did not contribute to our well-being, then on a preference-dependent conception of value, we would have no well-being. If to acquire a new desire were to enter a state of privation, so that to satisfy the desire would be nothing more than a return to a neutral state, it seems that there could be no such thing as preference-dependent value. There would be nothing positively good for us in the satisfaction of a desire. Indeed, in the absence of preference-independent value in our lives, there could, on the debit view, be no well-being at all, only ill-being and neutral-being. The development and satisfaction of a desire would be only the development of a state of privation and the subsequent elimination of that bad state. But the development and frustration of a desire would be the development of a state of permanent privation (unless, perhaps, ceasing to have a frustrated desire would eliminate the privation). No one’s life could be worth living. It seems, therefore, that any plausible version of preference utilitarianism must accept that the development (or creation) and satisfaction of a desire is more than just the development and elimination of a state of privation. When a desire develops or is created in an individual and is then satisfied, that must be a positive good for that individual. And if that is true of the creation and satisfaction of a desire in an existing person, then it should also be true of the creation and satisfaction of a desire by causing the individual who will develop and satisfy the desire to exist.

Singer is aware that the debit view is vulnerable to an objection of this sort, as he notes that the view may seem to entail antinatalism – the view that it is wrong to cause individuals to exist. It is in response to this objection that he invokes the notion of impersonal, preference-independent value, arguing that it is permissible to cause people to exist because their lives will contain pleasure and happiness, and perhaps also because they will contain other types of impersonal value, such as love, friendship, meaningful work, knowledge, and so on. He says that this pluralist view, which combines preference utilitarianism, including the debit view of preferences, with the claim that there is also a utilitarian reason to produce impersonal, preference-independent value, entails neither antinatalism nor the view that “all sentient beings, even those who are self-aware, are replaceable” (Singer, 2011, pp. 114–117).

Although this is the view he adopts, he confesses that it “sacrifices one of the great advantages of any form of utilitarianism that is based on just one value, which is that there is no need to explain how different values are to be traded off against one another” (Singer, 2011, p. 119). But this acknowledgement that preference-dependent values have to be traded off against preference-independent values in order for it to be permissible to cause individuals to exist prevents the pluralist view from being able to deny the replaceability of persons. If it is permissible to cause a person to exist when the impersonal, preference-independent value in the person’s life would outweigh the unavoidable frustration of many of the person’s desires, then the frustration of one person’s desire to continue to live must be able to be outweighed by the impersonal, preference-independent value in the life or lives of one or more persons created as replacements. In short, only if preference-dependent value has lexical priority over preference-independent value, so that one person’s desire to continue to live cannot be outweighed by some increase in preference-independent value, can the pluralist view avoid implying the replaceability of persons. But if the pluralist view gives preference-dependent value lexical priority, then it entails antinatalism.

It is perhaps worth mentioning still further reasons for rejecting the debit view. One is that, even as we are presently constituted, we are constantly satisfying categorical desires we have had, so that it is essential to our continuing well-being over time that we develop and satisfy new ones. One might argue that, to contribute to our well-being, our development and satisfaction of new desires must always be, as Singer suggests, a means to the satisfaction of preexisting, more encompassing desires, such as the desire to continue to have a good life, or the desire to develop and satisfy new desires. But the appeal to second-order desires of this sort simply empties the preference-based conception of what is good for individuals of its content and implicitly replaces it with an alternative conception. For what, on the debit view, could continuing to have a good life consist in other than continuing to satisfy one’s preexisting first-order desires? And what reason would there be to desire to develop new desires if their satisfaction would not be good for us and developing them would risk their being frustrated?

One might add that it is not just the satisfaction of our categorical desires that contributes to our well-being but also, at least in many instances, the process of striving to satisfy them. Bertrand Russell once wrote, in a passage that might have been addressed to Schopenhauer, whose views Singer mentions in connection with his defense of the debit view, that

The human animal, like others, is adapted to a certain amount of struggle for life, and when by means of great wealth homo sapiens can gratify all his whims without effort, the mere absence of effort from his life removes an essential ingredient of happiness. … If he is of a philosophic disposition, he concludes that human life is essentially wretched, since the man who has all he wants is still unhappy. He forgets that to be without some of the things you want is an indispensable part of happiness

That the acquisition of new desires is normally necessary to make continued striving to fulfill one’s goals possible is another reason for thinking that there is a preference-dependent reason to create new desires that can be satisfied. Russell’s observation also directly challenges the claim that having unsatisfied desires is just a condition of privation.

A caveat is necessary here, which is that, for the acquisition and satisfaction of new categorical desires to make a person’s life better, this process must be compatible with the preservation of a high degree of psychological continuity in the person’s life – otherwise it would undermine the basis of the person’s rational prudential concern about the future (Parfit, 2011, Part Three). Obviously, however, this caveat is unnecessary in the case of acts that increase preference-dependent value through causing individuals to exist who will then develop and satisfy categorical desires.

There is, I should note, one way in which Singer could accept that it is good for us to develop and satisfy new categorical desires that are not merely means of satisfying preexisting desires, while rejecting the claim that there is a preference-dependent reason to create and satisfy new categorical desires by causing people to exist. This would be to accept that preference utilitarianism must take what he calls a “prior existence” form, or what I would call a “narrow individual-affecting” form. According to a view of this sort, there is a moral reason to create a desire that would be satisfied, or to prevent the existence of a desire that would be frustrated, only if doing so would be better or worse for an individual who does or will exist independently of the act that creates the desire. But, as Singer rightly observes, preference utilitarianism cannot plausibly take such a form, as it would then imply that there is no preference-dependent reason not to cause an individual to exist whose every desire would unavoidably be frustrated and whose life would therefore be utterly miserable.

I am forced to conclude that the debit view of preferences is untenable and, therefore, that Singer’s defense against Hart’s charge that preference utilitarianism implies that persons are replaceable is unsuccessful.

In all three editions of Practical Ethics, Singer suggests, as I noted earlier, that one important reason that preference utilitarianism is superior to hedonistic utilitarianism is that it, but not hedonistic utilitarianism, can explain why persons are not replaceable. More recently, however, Singer has converted from preference utilitarianism to hedonistic utilitarianism, and for some time I wondered where that has left him on the issue of replaceability. In a book published three years after the third edition of Practical Ethics, he and his co-author, Katarzyna De Lazari-Radek, consider the problem of replaceability from a hedonistic utilitarian perspective, but their discussion is brief and inconclusive (De Lazari-Radek & Singer, 2014, pp. 346–348). Because of the objections to preference utilitarianism that I have advanced above, it seems to me now that Singer’s conversion to hedonistic utilitarianism makes no difference to the view he can consistently take on the issue of the replaceability of persons. After all, a “prior existence” or narrow individual-affecting version of hedonistic utilitarianism would, like a narrow individual-affecting version of preference utilitarianism, reject the replaceability of persons, as it would deny that there is a reason to produce pleasure by causing individuals to exist. What seems to make the difference to whether a form of utilitarianism entails the replaceability of persons is not its conception of well-being or the good but what form – impersonal or narrow individual-affecting – it asserts that its utilitarian reasons must take. It may just be that Singer assumes, as most others do, that hedonistic utilitarianism reasons take an impersonal form.

It is perhaps worth noting that there is also a problem with part of Singer’s defense of the claim that non-self-conscious individuals are replaceable. This is his appeal to the Non-Identity Problem to show that the reason to cause a better-off individual to exist can be just as strong as the reason to make a preexisting individual better off. Suppose we agree with Parfit and Singer that in the two choices Singer cites, the reason to cause a better-off individual to exist rather than allow a different, less well-off individual to exist is just as strong as the reason to enable a preexisting individual to have a life as good as that of the better-off individual rather than allow that same individual to have a life that would be only as good as that of the less well-off individual. The problem is that that judgment has no direct bearing on choices concerning replacement. This is because the choices that Singer considers are what Parfit calls “Same Number Choices” – choices in which the same number of people exist in all the possible outcomes. But a choice between allowing an individual to continue to live and replacing that individual with another individual are “Different Number Choices”, as the choice of replacement involves causing an individual to exist rather than not causing anyone to exist. And, as Parfit recognized, there may be highly important differences between Same Number Choices and relevantly similar Different Number Choices.

According to Parfit, the reason to cause a person to exist who would live to 80 rather than allow a different person to come into existence who would live only to 60 is just as strong as the reason to save the life of a 60-year-old, thereby enabling that person to live to 80. This is a Same Number Choice. A relevantly similar Different Number Choice might be between saving the life of a 20-year-old, thereby enabling that person to live to 80, and causing a person to exist who would live to 60. While some may agree with Parfit’s judgment in the Same Number Choice, it is difficult to believe that in the second choice, the reason to cause a person to exist is just as strong as the reason to save a person’s life – even though the number of years of good life one would enable someone to have would be the same in both choices. It is notable that Parfit never made such a claim or, indeed, inferred from his view about Same Number Choices that there is a moral reason to cause a well-off individual to exist rather than not cause anyone to exist. While Parfit’s arguments about the Non-Identity Problem do, as Singer recognizes, challenge “prior existence” forms of utilitarianism, they do not support the replaceability even of non-self-conscious individuals.

It is perhaps also worth mentioning, in concluding this discussion of the relevance of the Non-Identity Problem to replaceability, that Parfit’s view in Same Number Choices is inconsistent with the view that Singer defends. Suppose that one could either confer a desired benefit of magnitude x on an existing person or prevent one person from coming into existence and cause a different person to exist whose life would be better by amount x. On Parfit’s view, the reason to benefit the existing person is no stronger than the reason to cause the better-off individual to exist, other things being equal. But on the pluralist view that Singer sought to defend in Practical Ethics by appealing to the Non-Identity Problem, the reason to benefit the existing person is a preference-dependent reason and thus is stronger than the impersonal reason to cause the better-off individual to exist. This too challenges Singer’s use of Parfit’s claims about the Non-Identity Problem in support of his replaceability thesis.

As I noted in the Introduction, I believe that there are important insights in Singer’s notion of replaceability. Many of us, for example, believe that when a fetus is discovered to have an incurable condition that would unavoidably make the life of the person who would develop from it of low quality and short duration, there is a moral reason for the pregnant woman to have an abortion and conceive a different fetus that would not have the disadvantageous condition. If one accepts this claim, one accepts that some human beings are replaceable. And if one accepts this claim in its application to a fetus late in the third trimester, then one ought, in consistency, to accept it in its application to a chronologically younger and thus less well developed premature infant.

I accept these claims but would seek to justify them differently from the way that Singer does. In my view, a non-self-conscious individual can have only very weak interests in its own future life and, in particular, can have only a very weak interest in continuing to live. This is because I believe that the strength of an individual’s present interest in having something intrinsically good in the future is a function of two variables: (1) the degree to which the future benefit would be good when it occurred and (2) the strength of the psychological connections between the individual now and itself in the future when the benefit would occur. In the case of a conscious fetus, there is really only one psychological relation binding it to itself in the future, and that is continuity of consciousness, or sameness of mind. That this relation is a basis for egoistic concern about the future, or third-party concern about an individual’s future for that individual’s own sake now, is shown by a well-known thought experiment of Bernard Williams’s. Suppose that one knows that, over the next hour, the neurological bases of every element of one’s mental life – all of one’s beliefs, desires, memories, character traits, and so on – will be destroyed, so that, while one’s brain will remain conscious, all of its particular potentially conscious contents will have been erased. If one also knows that in an hour, at the end of this process, one’s body will be horribly tortured, it seems clearly rational for one now to fear that torture (Williams, 1973, p. 52).

A newborn infant is normally slightly more strongly psychologically continuous with itself in the future than a conscious fetus is with its future self, but even in the infant’s case the degree of psychological relatedness is extremely weak. Hence, in my view, the painless killing of a conscious fetus or a newborn infant can frustrate at most only a very weak interest. That killing it would frustrate this interest is a moral reason not to kill it that is not a wholly impersonal reason. It is a reason of the same type as one of the strong reasons we have not to kill persons.

Like Singer, I believe not only that there is a moral reason to cause a better-off individual to exist rather than a different, less well-off individual but also that there is a moral reason to cause a well-off individual to exist rather than not cause any individual to exist. ⁽³⁾ This latter reason cannot be what Parfit would call “narrow person-affecting”, or what I would call “narrow individual-affecting” (so that the term covers effects for better or worse on animals and human non-persons as well as persons). According to Singer, the reason is impersonal. I think it is better understood as what Parfit would call a “wide person-affecting reason”, or what I would call a “wide individual-affecting reason”. But for present purposes, I will follow Singer in saying that the reason to cause a well-off individual to exist rather than not cause any individual to exist is an impersonal reason. In my view, the reason to satisfy an individual’s interest in having a benefit is stronger than the impersonal reason to produce the same benefit by causing an individual to exist. Thus, on my view, the reason not to kill a fetus or newborn infant is stronger than it is on Singer’s view. Still, because the interest in continuing to live of a fetus or newborn infant is very weak, it can be overridden by the reason to cause a better-off individual to exist when it is infeasible for both to live, particularly when the future life of the fetus or infant would be of unusually lesser quality or duration than most people’s lives because of a disadvantageous congenital condition.

The extent to which persons will be psychologically connected to themselves in the future is normally vastly greater than the extent to which non-self-conscious animals, fetuses, newborn infants, and even higher animals will be psychologically connected to their future selves. Also, the future lives of persons will normally contain much more that is intrinsically good for them than the future lives of non-self-conscious animals and animals who are to some degree self-conscious. For these reasons, persons normally have a vastly stronger interest in continuing to live than animals, fetuses, or newborn infants. And if the moral reason to satisfy an interest in having certain benefits is substantially stronger than the impersonal reason to produce the same benefits by causing an individual to exist, then the reason not to kill a person will almost always outweigh the reason to cause a person to exist, thereby making it impermissible to “replace” a person with a different, better-off person.

This is insufficient, however, to show that, while fetuses, newborn infants, and non-self-conscious animals are in principle replaceable, persons are not. There remains the problem I noted earlier: numbers. Even if the interest-based reason not to kill a person is substantially stronger than the impersonal reason to cause a person to exist whose life would contain more of what is intrinsically good for persons, it seems that the interest-based reason could be outweighed if killing the one person were necessary to enable a very large number of persons to exist whose lives would be better than that of the person they would replace. It is for this reason, among other related reasons, that I cannot accept any form of utilitarianism. It seems that the only form of utilitarianism that can avoid implying the replaceability of persons is the “prior existence” or wholly narrow individual-affecting version that holds that the only effects on well-being that matter morally are effects on individuals for better or worse. ⁽⁴⁾ This version implies that no sentient being is replaceable in Singer’s sense, as there is never a reason to cause an individual to exist solely because of the effect on that individual. This is because coming into existence cannot be better or worse for someone than never existing, since there is no one for whom never existing can be worse or better than existing. But even though this form of utilitarianism implies that individuals are irreplaceable, it also, as Singer observes, implies that there is no moral reason, apart from effects on others, not to cause an individual to exist whose life would be wholly intrinsically bad. For this and other reasons, this form of utilitarianism is unacceptable.

The best way to be able to reject the idea that persons are replaceable, it seems to me, is to accept that there are deontological constraints on action, and in particular that there are constraints against harming and killing individuals who have not made themselves liable to be harmed or killed. This, along with the priority of interest-based reasons over impersonal reasons, ensures that persons are never in practice replaceable. It may be true that, if no constraints are absolute, so that all are in principle overridable (as I believe they are), then the replacement of a person could be in principle permissible in extreme conditions. But it is perhaps sufficient if the replacement of a person is never in fact permissible in practice.

The idea that deontological constraints prevent persons from being replaceable is compatible with the replaceability of non-self-conscious animals, other non-human animals, fetuses, and newborn infants. In my view, deontological constraints apply only to acts that affect individuals with a certain moral status. It is possible, on this view, that there is no deontological constraint against killing a fetus (even though there are constraints against affecting fetuses in ways that would adversely affect their interests later when they had achieved a certain moral status). But I believe that an individual’s moral status is determined by that individual’s psychological capacities and is also a matter of degree – and that the strength of deontological constraints on action varies with the moral status of the individual that would be affected by an act. Thus, for example, the constraint against inflicting a certain harm on a person is stronger than the constraint against inflicting the same harm on a non-self-conscious animal (McMahan, 2021). If this is right, then non-self-conscious animals, fetuses, and newborn infants can be replaceable even if they have interests and moral status, as their interests are weak and their moral status is low.

I note in concluding that, if Singer were still receptive to accepting a pluralist utilitarianism, one component of it could be the recognition of a narrow individual-affecting preference-dependent reason to satisfy existing preferences (and preferences the existence of which is independent of one’s choice of action). This view could reject the debit view and recognize the satisfaction of a preference as a positive good. Singer could then combine this view with his acceptance of impersonal reasons to produce preference-independent value, while assigning priority to preference-dependent reasons over preference-independent reasons. The resulting pluralist utilitarianism would be similar to the view I have sketched in this section. It would imply that there is no preference-dependent reason to cause people to exist because they would develop desires that would be satisfied. It would, like the view I have sketched, imply that conditions are unlikely to arise in practice in which it would be morally required, or even permissible, to kill a person if this were necessary to cause a better-off person, or even a number of better-off people, to exist. It would not, moreover, entail antinatalism, and it could appeal to impersonal reasons to explain the wrongness of causing miserable individuals to exist. But if it is purely utilitarian and does not recognize deontological constraints, it must accept that it would be permissible to kill an innocent person who strongly desires to continue to live if that were necessary to enable some number of well-off individuals – even some large number of well-off animals – to come into existence.