Why the handicapped child case is hard

Why the handicapped child case is hard.∗ Josh Parsons September 5, 2006 Abstract This paper discusses the handicapped child case and some other variants of Derek Parfit’s non-identity problem. (Parfit 1984, s. 122) The case is widely held to show that there is harmless wrongdoing, and that a moral system which tries to reduce wrongdoing directly to harm (“person-affecting morality”) is inadequate. I show that the argument for this does not depend (as some have implied it does) on Kripkean necessity of origin. I distinguish the case from other variants (“wrongful life cases”) of the non-identity problem which do not bear directly on person-affecting morality as I understand it. And finally, I describe a respect in which the handicapped child case is puzzling and counter-intuitive, even on the supposition that it is a case of harmless wrongdoing. I conclude that the case is “hard”: it will take more than the rejection of person-affecting morality to remove its puzzling character. 1 The Case The Handicapped Child Case: A woman, Elizabeth, desires to become pregnant. She knows that if she were to become pregnant immediately, her child would have some mild genetic defect. However, she also knows that if she were to wait a year, her child would not suffer from this defect. Despite this knowledge, Elizabeth becomes pregnant without waiting and subsequently gives birth to a child, Anne, with the foreseen mild defect. Anne’s life is well worth living, but she suffers from this seemingly contingent affliction caused by the circumstances of her conception. For most people, this case evokes a powerful intuition that Elizabeth has done something wrong. Let’s call this the wrongdoing intuition. Usually, people who haven’t heard this argument before also have the intuition that Elizabeth has acted wrongly in a way that makes her accountable to Anne. That is to say, Elizabeth’s wrongdoing is specially connected with Anne. If Anne feels resentment toward Elizabeth, for example, that resentment is justified (and more justified than any resentment you or I might feel toward Elizabeth). ∗ Thanks to Elizabeth Ashford, Daniel Cohen, participants in the University of Dundee Philosophy Department research seminar, and an anonymous referee for Philosophical Studies. 1 This accountability intuition is difficult to satisfy. Suppose Elizabeth had waited a year before becoming pregnant. The child who would then have been born — call her Zoe — would not be Anne. For it is possible that Elizabeth have both children: Zoe could be born a year after the birth of Anne, and then they would be sisters. Possibly, Zoe and Anne will both exist, and so, by the necessity of identity, they can’t be the same person. Since, if Elizabeth had acted rightly — had waited before becoming pregnant — Anne would not have existed, Anne is no worse off than she would have been, had Elizabeth acted rightly. Anne’s life is not so bad that she is better off dead, or better off never having existed. She should be glad that she exists at all. Certainly Elizabeth could not have acted in such a way that Anne would be better off. If ought implies can, then Elizabeth had no obligation to make Anne any better off than she actually did. So Anne’s resentment is unjustified. The received view of this case takes it to show that there is harmless wrongdoing. People can do what is morally wrong without harming anyone — and in this case, that is what Elizabeth has done. If this is right, then the case also refutes that is often called “the person-affecting view” — roughly, that right and wrong reduce to benefit and harm to particular persons.1 It seems to me that this is all basically correct. However, standard discussions of the case often confuse matters in various ways: by making spurious connections with the alleged necessity of origin; by being insufficiently clear about what the person affecting view amounts to; or by confusing the case with another type of case, which I will label “wrongful life”. In addition, I think that there is something troubling about the “received view” — the moral that this is a case of harmless wrongdoing. Regarding Elizabeth’s actions as harmless seems to lead to other problems, which have not been widely commented on. Accordingly, in this paper, I try to give a clear characterisation of the argument that handicapped child cases present for harmless wrongdoing and against person affecting morality. I explain why they do not depend on any essentialist premise, and how they connect to issues in the metaphysics of modality (section 2). I give a detailed statement of how the argument against person affecting morality should proceed (section 3), and explain how the handicapped child case is distinct from “wrongful life” cases (section 4). Finally, I argue that there is something puzzling about these cases, even if they are examples of harmless wrongdoing (section 5). 2 Essentialism It is sometimes said of these cases that they depend on on a kind of essentialism made popular by Kripke (1972, pp. 110–115) — necessity of origin. This is just the idea that things have their origin essentially: in the case of human beings, their circumstances of birth or conception. Usually, this is said to give an air of Kripkean authority to the argument. Sometimes the cases are 1 As always with philosophical examples, one has to abstract away a little. In calling this a case of harmless wrongdoing, we are ignoring any minor harms Elizabeth inflicts on herself or others by bringing Anne into existence. For example, if Anne’s affliction is myopia, perhaps Elizabeth will have to go out of her way later on to obtain corrective lenses for her. That might count as a minor self-harm by Elizabeth, so her actions were not entirely harmless. The point, however, is that the harm done does not exhaust the wrongness of Elizabeth’s act, as the person-affecting view would have it. 2 even made to depend on genetic essentialism: the idea that organisms have their genetic makeup essentially (though this is rarer).2 If the specification of the case depends on either of these essentialisms, then it is open to a person affecting moralist to defend herself by denying necessity of origin; or by denying genetic essentialism; or by denying that any kind of essentialism is true. I have a certain amount of sympathy for these objections, but I think that they are misplaced. It’s not necessary to appeal to essentialism in setting up the case, and I don’t think that I did. I did appeal to the following modal premise: Counterfactual premise: If Elizabeth were to wait a year before becoming pregnant, she would not give birth to Anne. There is nothing about this premise that requires necessity of origin (or other forms of essentialism). In particular, I have not appealed to this premise: Essentialist premise: It is impossible that Elizabeth wait a year before becoming pregnant, and give birth to Anne. In logical terms, the essentialist premise employs a strict conditional where the counterfactual premise employs a counterfactual conditional. We can see the difference between them by looking at the standard semantic accounts that are given of them: The counterfactual says only that the “nearest” possible world (the most similar to the actual world, that is) in which Elizabeth waits is a world in which the child she gives birth to is not Anne. The strict conditional says that every possible world in which Elizabeth waits is a world where the child she gives birth to is not Anne. The strict conditional entails the counterfactual, but not the other way around. Nor is this simply an artifact of the possible world semantics. Take an analogous case: the counterfactual conditional “If I were outside right now, I would be cold” clearly does not entail the corresponding strict conditional “It is impossible that I be outside right now and warm”. It would be reasonable to object to the Essentialist Premise on the grounds that it embodies some kind of essentialism. But I have not affirmed it. If the objection is that I’m involved in essentialism by asserting the Counterfactual Premise, then that objection rests on a modal fallacy. Perhaps a further objection is that there is no reason to believe the Counterfactual Premise apart from appeals to essentialism. But I think I have given such a reason: namely that the child who would be born if Elizabeth were to wait, Zoe, could have been born as Anne’s sister. There is room for quibbling about whether this is true in certain cases, but I don’t see any appeal to essentialism here. 2 Thus, Jeff McMahan says, in the context of a version of the handicapped child case, “The view that alterations to genetic materials may be identity-determining with respect to the person who eventually develops from them seems compatible with the widely accepted doctrine of necessity of origin.” (McMahan 1998, pp. 209–210) Gregory Kavka says that, in order to set up his version, he will “assume that sameness of genetic structure is, for practical purposes, a necessary condition of personal identity.” (Kavka 1982, p. 93) And Derek Parfit appeals to “The Origin View” of personal identity of across worlds. (Parfit 1984, s. 119) 3 3 Person affecting morality I would like to explore a bit further exactly how the handicapped child case makes trouble for person affecting morality, and also what exactly the latter is. I’ll start with a principle that a) is clearly some type of person affecting principle; b) looks as though it is refuted by the handicapped child case; and c) apart from this might be independently plausible. Harm principle: An outcome is good only if it benefits someone; bad only if it harms someone.3 This way of stating person affecting morality has two drawbacks. First, it is ambiguous in a way that sometimes leads to misunderstanding of person affecting morality; and second, it entails several weaker doctrines, some of which have a better right to be called “the person affecting view” than it does. “Harm” and “benefit” are contentious terms in their own right, and the contention extends to the interpretation of the harm principle. “Harm” is naturally used to mean what I will call “temporal harm”: the event of a person’s welfare decreasing from one time to the next. Analogously, “benefit” can mean temporal benefit, the event of a person’s welfare increasing from one time to the next. I am temporally harmed at a time t iff my welfare before t is greater than my welfare after t; temporally benefited at t iff my welfare before t is less than my welfare after t. However, that is not the type of harm and benefit that is relevant here. Suppose I am unemployed, and am applying for an academic job. You are one of my referees, and you have a choice of writing me a favourable reference or an unfavourable one. My other references are favourable enough that either way I will get the job. If you write a favourable reference I will be paid $40,000 p/a, if you write an unfavourable one, $39,000 p/a. In either outcome, then, I am temporally benefited — being on a salary of either $39,000 or $40,000 is better than being unemployed. Either way I am better off than I was. But in the relevant sense it is only a favourable reference that will benefit me, and only an unfavourable one that will harm me. It is only if you write a favourable reference that I am better off than I would be. The type of harm that is relevant for us is not temporal harm, but “global” harm: Definition of global harm: S is harmed in outcome A iff A is worse for S than rival outcomes (or, for short, iff A is bad for S). S is benefited in outcome A iff A is better for S than rival outcomes (for short, iff A is good for S).4 3I have chosen to formulate this principle in axiological terms — in terms of goodness of outcomes — in order to keep exposition simple. Henceforth I will be assuming that, according to both person affecting morality and according to its principal foes, right action is justified by its bringing about a good outcome. This outcome talk does cut me off from one type of person affecting morality. According to the view I call “quasideontic person affecting morality”, right actions are justified not in terms of the good, but directly in terms of their benefit to moral patients. A quasi-deontist does not like to say anything at all about the goodness of outcomes, only about right, wrong, benefit, and harm. It is still consequentialist — benefits and harms are consequences of actions — but it will not fit into the framework of ranking outcomes according to goodness that I will be using. If you are tempted by the quasi-deontic view, it should be possible to translate everything I say into its terms. 4 A problem with this conception of harm and benefit is that it might be over-demanding — it might require you to be constantly maximising my welfare, on pain of harming me. That doesn’t follow from the definition above, 4 Outcomes, here, are thought of as complete ways the would might be (as possible worlds, in metaphysical parlance), and persons as complete lives lived. This account asks us to compare S’s whole life as it would be if A came about to S’s whole life as it would be if A did not come about.5 To avoid the ambiguity introduced by different ways of understanding “harm” and “benefit”, we could fold the definition of global harm into the harm principle to produce the following: Person affecting constraint: An outcome A is good only if there is someone for whom A is good. The person affecting constraint is, I claim, the common thread to all person affecting moral systems. It is appealed to indirectly by many accounts of person affecting morality, including those that endorse the harm principle. Some people will want to say that this constraint is too weak to properly capture person-affecting morality — in a certain sense I agree (and this will be discussed further below). However, I think that it is important to isolate it because there are multiple ways of strengthening it, each of which leads to a different version of person-affecting morality. The way of strengthening it that I prefer will be described in the remainder of this section. I also discuss a rival way in section 4. I am now in a position to explain how the handicapped child case constitutes a counterexample to this constraint. It will be helpful to represent the case in a type of diagram whose form I have borrowed from Broome (1999): Elizabeth Anne Zoe A (we , ... 1, Ω) B (we , ... Ω, 10) A and B are two outcomes available to Elizabeth in deciding when to have her child. Between the brackets are listed the levels of welfare that each person has in the outcome whose name is written to the left. The ellipsis (...) represents the fact that many more people exist in each outcome than Elizabeth, Anne, and Zoe. I will be assuming that these persons’ welfare is identical in each outcome. Finally, an omega (Ω) indicates that a person does not exist at that outcome. So, in both A and B, Elizabeth has some level of welfare we . In A, Anne has welfare 1, and Zoe does not exist. In B, Zoe has welfare 10, and Anne does not exist. The assignment of numbers to levels of welfare is nearly arbitrary. I chose a bigger number for Zoe in B than for Anne in A, because Zoe is supposed to be better off in B than Anne is in A. The order is significant but the scale is not. though: we also need a further principle to tell us what the “rival outcomes” are in the context of a particular action or decision. For example, is the outcome in which you give me all you own and become my slave a rival to the outcome in which you write a favourable reference and I earn $39,000? If so, perhaps even writing a good reference would harm me. But, plausibly, making yourself my slave is not a rival outcome in that context, because it’s not an outcome of a decision about what kind of reference to write. 5 The distinction between global and temporal harm is connected with two distinct types of welfare. (Bigelow et al. 1990, pp. 120–123) When I described temporal harm as a decrease in welfare, I had in mind a level of welfare as a state that is had by a person at a time, or, perhaps, by a temporal part of a person. In thinking of whether an outcome is worse for a person than another, we are assessing the total welfare of a person’s whole life. These two types of welfare are different properties: the former, temporal, welfare is had only by person stages (or persons-at-times), the latter, global welfare, is had only by whole persons. 5 A represents the outcome in which Elizabeth has a child now, B the outcome in which she waits. Intuitively, B is definitely better than A. Elizabeth ought to wait — that is, she ought to bring about B, rather than A. The person affecting constraint, however, will require that B is no better than A, contrary to our moral intuitions. It won’t quite do this on its own, however. What it does is zeroes in on Anne and Zoe, who are the only people (according to the description of the case) whose welfare differs from outcome to outcome. They are the only people, therefore, who could be harmed by Elizabeth’s choice of A over B. Anne is not harmed by Elizabeth’s choice, because she’s not any worse off in A than she is in B. But we do need another premise to ensure that Zoe is not worse off in A by virtue of not having been created. This seems very plausible. Zoe is merely one among an infinite number of merely possible persons. If all of them were harmed by not being brought into existence, then person affecting morality would instruct us refrain from bringing about this harm by creating as many people as possible, and that would be absurd. The missing principle is a doctrine to the effect that merely possible persons, or indeed merely possible patients of any kind, are not worthy of moral consideration — a point urged vigorously by Jan Narveson: [T]he sole ground of duty is the effects of our action on other people, and from this it follows that whenever one has a duty, it must be possible to say on whose account the duty arises — i.e. whose happiness is in question. [I]f a person is not born, he does not exist... And as we all know, non-existent people are not just a special kind of people (Narveson 1967, pp. 63–64) The principle suggested by Narveson, as I understand him, is this one:6 Actualism about “good for”: Outcomes can only be good for actually existing moral patients. The point here is that we can only properly evaluate Elizabeth’s action in the light of who is actual — who exists, has existed, or will exist. If we are evaluating Elizabeth’s actions in a world where she has chosen A, then Zoe is a merely possible person, and the potential benefit to her of her own existence does not contribute to the goodness of outcome B. Now we can show how the counterexample works. The intuition provoked by the description of the case is that B is better than A — Elizabeth has done wrong in bringing about A rather than B. But B is no better than A for Elizabeth, or any of the other people who exist in both — all of them, by hypothesis, have the same level of welfare in both. Nor is B any better for Zoe, who is not an actually existing moral patient. Nor is B any better for Anne, because, by hypothesis, Anne’s life is worth living, and she would not be better off if she did not exist. Hence, by the person affecting constraint, B is no better than A. But intuitively it is, therefore either the person affecting constraint or actualism about “good for” must be false. Of the two, the most plausible principle to reject is the person affecting constraint. 6 For a defence of this exegetical point about Narveson (and further discussion of actualism about “good for”), see my (2002). 6 4 Wrongful Life and Epicureanism The handicapped child case is sometimes presented in a different form, which, it seems to me, is a weaker argument against the person affecting constraint. In this section, I want to discuss this other form, and explain why I think it is a bad argument against person affecting morality. The case is this: The Wrongful Life Case: A woman, Elizabeth, desires to become pregnant. She knows that if she were to become pregnant immediately, her child would have some severe genetic defect. However, she also knows that if she were to wait a year, her child would not suffer from this defect. Despite this knowledge, Elizabeth becomes pregnant without waiting and subsequently gives birth to a child, Anne, with the foreseen severe defect. Anne’s life is not worth living, and she wishes Elizabeth had not brought her into existence. To clear up the terminology a little, I will henceforth refer to cases like this one as “wrongful life cases”, and not as “handicapped child cases”. It seems to me that the distinction between them is not always as widely realised as it ought to be. The difference between the two cases is that in the handicapped child case, it is stipulated that Anne’s handicap does not prevent her from living a worthwhile life. In the wrongful life case, on the other hand, Anne’s handicap is as severe as you like, and her life as miserable as you like. Using the type of diagram introduced above, we might show the wrongful life case as: Elizabeth Anne Zoe C (we , ... −10, Ω) D (we , ... Ω, 2) Now let us run through the argument given at the end of the previous section using this case instead of the handicapped child case. The intuition is that D is better than C. D is not better for Elizabeth, or for the other people who exist at both. Nor is D better for Zoe because she doesn’t actually exist. However, in this case, it’s much more tempting to say that D is better than C for Anne. Of course this will be a contentious point. It depends on whether we can make sense of the idea of someone’s being so badly off that they would be better off not existing at all. Notice, by the way, this is not quite the same as being better off dead. It is one thing to be dead, and another never to have been born. The temptation, then, in this case, is to say that Anne is so miserable that she should never have been born. D is, therefore, better for her than C. If this is right, then the wrongful life case does not refute the conjunction of the person affecting constraint with actualism about “good for”. The person affecting moralist can point to Elizabeth’s creation of Anne in a harmed state as the harm that her actions have caused, and the ground of their wrongness. The wrongful life case, therefore, is a red herring from the point of view of arguments about person affecting morality. However, things are not so simple. Some person affecting moralists would also like to believe this doctrine: 7 Global epicureanism: Outcomes can only be good (or bad) for moral patients who exist at those outcomes.7 If global epicureanism is true, D is not better for Anne than C, because Anne does not exist at D. The global epicurean will say that I have cheated by representing Anne’s welfare in C as a negative number. Perhaps there is no non-arbitrary zero point on the scale of welfare? If there is one, why suppose that it coincides with the point at which lives worth living give way to lives not worth living? These are all good questions. I don’t wish to argue against global epicureanism here. I simply want to point out that it is independent of the person affecting constraint and actualism about “good for”. I will also try to answer a common argument that is given in favour of global epicureanism. The argument runs like this: 1. Existing is a pre-requisite for having a level of welfare at all. 2. Therefore, you can only have a level of welfare in an outcome at which you exist. 3. Therefore, outcomes can only be good (or bad) for moral patients who exist at those outcomes. To answer this argument, I will need to work by alternatives. I am now going to temporarily drop my neutrality about different accounts of welfare. I will show that, assuming a certain account of welfare, the argument does not work; and then that, assuming any other account of welfare, it still doesn’t work (though for different reasons). 4.1 Preference satisfaction and epicureanism The account of welfare in question is the preference satisfaction theory. According to this theory, facts about a moral patient’s welfare reduce to facts about which of their preferences are satisfied. (“Satisfaction” here is used in its dry logical sense — a patient’s preference is satisfied when the world matches the way the patient would like it to be, regardless of how the patient feels about the matter). We might think of a patient’s preferences as an ordinal ranking of possible states of affairs from better to worse. It is this ranking that we are talking about when we describe a state of affairs as being good for a patient, or bad for a patient. The feature of this account that is important now is that the welfare a patient has in a given outcome is had in virtue of that outcome’s position in the ranking. “Good for” is prior in explanation to welfare. The preference satisfaction theory has a feature that the global epicurean should like: because the facts about welfare are exhausted by a merely ordinal ranking of states of affairs, there is no non-arbitrary zero point in the scale of welfare. There is no structural feature of an ordered list of states of affairs that tells you where to put the zero point. 7 The lower-case ‘e’ is to indicate that I am not putting forward any claims about the views of Epicurus or his followers here. 8 But it has another feature that the global epicurean won’t like: there is nothing in the structure of a ordered list of states of affairs that prevents it from containing states of affairs in which the orderer does not exist. Some people, however unreasonably, like the thought of playing golf. If I like the thought of playing golf, then I put states of affairs in which I play golf high up in my preference ranking. In virtue of that, I get to have a high level of welfare at those states of affairs where I play golf. Some people, however unreasonably, like the thought of never having been born. If they like that thought, then they put states of affairs in which they were never born high up in their preference rankings. In virtue of that, they get to have a level of welfare at states of affairs where they do not exist. If all that is needed a for patient to have a level of welfare at a state of affairs is that they place that state of affairs in their preference ranking, then there is no reason that a patient cannot have a level of welfare at a state of affairs at which they do not exist. If the preference satisfaction account of welfare is correct, then existence is not a pre-requisite of having a level of welfare, and the premise of the Global epicurean’s argument is false. 4.2 Non-Preferencism and epicureanism I will call the denial of the preference satisfaction theory “non-preferencism”. According to nonpreferencism, a patient’s having a certain level of welfare at an outcome does not consist in that patient’s ranking that outcome in a preference ordering. Rather, the welfare of a patient at an outcome is some ordinary property which the patient has according to that outcome. For example, hedonism is a form of non-preferencism, on this usage. According to hedonism, the welfare of a patient at an outcome is the happiness (or unhappiness) of that patient at that outcome. Happiness and unhappiness are ordinary properties of moral patients, and are not constituted by patients’ preferences. A patient is neither happy nor unhappy if she does not exist. Such a patient has, in consequence, no level of welfare. So according to hedonism, the first premise of the argument for global epicureanism is true. A similar type of argument will apply for any non-preference account of welfare. Suppose, for example, that welfare consists in certain primary goods including, perhaps, health, vigour, intelligence, imagination, rights, liberties, powers, opportunities, income and wealth. (Rawls’s list) (1971, p. 62) As was the case for hedonism, a patient has none of these if she does not exist. Like the preference theory, non-preferencism has a feature that the global epicurean should like: it is indeed impossible to have welfare without existing, if welfare is an ordinary property such as happiness. However, the relationship between the welfare has at a state of affairs and that state of affairs’s goodness for that person might be quite complex. For example, consider a theory of value that identifies welfare with pleasurable sensations. It needn’t follow from such a theory that someone who never has any pleasurable sensations — and thus, has no welfare — is no worse off than someone with welfare. Indeed, such a theory would surely say that the state of having no welfare is worse for the person in that state than any state of having welfare. Similarly, it needn’t follow that someone who never has any health, wealth, etc, is no worse off than someone who has these 9 things. So it doesn’t in general follow, according to non-preferencism, that a state of affairs being good or bad for an patient requires that the patient have a level of welfare at that state of affairs. If you are going to identity welfare with an ordinary property (such has happiness) you will want to say that to completely lack that property is bad for you. So, in this case, the conclusion of the little argument for global epicureanism does not follow from the intermediate conclusion that you can only have a level of welfare in an outcome at which you exist.8 Regardless of whether the preference theory or some form of non-preferencism is correct then, this argument gives no reason to believe global epicureanism. 5 The Resentment Problem I have now stated what I take to be the prima facie case presented by the handicapped child case against person affecting morality. The case presents a counterexample to what is predicted by the person affecting constraint conjoined with actualism about “good for”. The most natural way to reply to this counterexample is to deny the person affecting constraint. If the person affecting constraint is false, then there is no reason to deny that Elizabeth has done wrong. The wrong Elizabeth has done, however, does not consist in making Anne (or anyone else, for that matter) worse off than she might be. Thus, by rejecting person affecting morality, we are able to retain the wrongdoing intuition. Tempting as it is, however, there is something unsatisfactory about the argument just outlined. The trouble is with the admission that Elizabeth’s wrongdoing is not connected with Anne’s welfare. That is, the proposed way of understanding the case does not seem to satisfy the accountability intuition. Here is a way of making the problem more explicit: we might imagine that Anne feels a strong and lasting resentment toward Elizabeth. Pre-theoretically, this resentment seems only reasonable under the circumstances. But if Elizabeth’s wrong has nothing to do with Anne’s welfare, Anne’s resentment is inappropriate. According to the standard account, resentment is only justified toward someone who has negligently “offended or injured” us (Strawson 1962, p. 7). A way to bring this point out is to consider the contrast between resentment and indignation. Indignation is an emotion appropriate for me to feel toward someone who is responsible for a moral wrong, but not offended or injured me in particular. Resentment, on the other hand is a much more personal matter. According to the argument from the handicapped child case against the person affecting constraint, Elizabeth has not offended or injured Anne. If there was a way of salvaging some kind of offence or injury to Anne, then that offence or injury could count as a respect in which Anne’s welfare is less than it would be if the offence or injury had not occurred. If Anne’s 8 What a non-preference theorist probably will say is that the goodness or badness of a state of affairs for a patient must supervene on the patient’s welfare at that state of affairs. That is to say, no difference in goodness without a difference in welfare. But global epicureanism doesn’t follow from this either. It only follows that if we rank states of affairs where I do not exist as good or bad for me, we must do so consistently — ranking all such states of affairs equally. To do otherwise would be to have a difference in goodness or badness for me without a corresponding difference in welfare. 10 resentment is justified, then the argument against the person affecting constraint cannot work. For this reason, I think that a person affecting account of the handicapped child case would be desirable. Alternatively, a non-person affecting account that explained why it is reasonable for Anne to feel resentment toward Elizabeth would suffice. But it is extremely difficult to give either. It is hard to do the former for the reasons given in section 3, and hard to do the latter for the reasons just given above: Anne’s resentment is justified only if Elizabeth has harmed her. That is not to say that either way of proceeding is impossible. We could start to make progress in the person affecting direction by weakening the definition of global harm. We might say that S is harmed in A iff A is worse for S than rival outcomes are for whomever is “in S’s shoes” is those rival outcomes. We would need to spell out “in S’s shoes”, and we would need to reply to criticisms along the lines of “Why should S care about what happens to other persons in her shoes at various outcomes?” Or we could start to make progress in the non-person-affecting direction by abandoning the Strawsonian account of resentment endorsed above. We would need either an account that makes Anne’s resentment justified without requiring that she be “offended or injured”, or at least an account that explains why her resentment would persist even on coming to fully understand the facts of the case. Either way heavy-duty theoretical work is needed. The handicapped child case does not straight-forwardly rule out person affecting morality. It makes it hard — but the case is hard for everyone, not just for the person affecting moralist. References Bigelow, J., J. Campbell, and R. Pargetter (1990). Death and well-being. Pacific Philosophical Quarterly 71, 119–140. Broome, J. (1999). The value of a person. In Ethics out of Economics, pp. 228–242. Cambridge: Cambridge University Press. Kavka, G. (1982). The paradox of future individuals. Philosophy and Public Affairs 11(2), 93–112. Kripke, S. A. (1972). Naming and Necessity. Cambridge, MA: Harvard University Press. McMahan, J. (1998). Wrongful life: Paradoxes in the morality of causing people to exist. In J. Coleman and C. Morris (Eds.), Rational Commitment and Social Justice: Essays for Gregory Kavka. Cambridge: Cambridge University Press. Narveson, J. (1967). Utilitarianism and new generations. Mind 76(301), 62–72. Parfit, D. (1984). Reasons and Persons. Oxford: Clarendon Press. Parsons, J. (2002). Axiological actualism. Australasian Journal of Philosophy 80(2), 137– 147. Rawls, J. (1971). A Theory of Justice. Cambridge, MA: Harvard University Press. 11 Strawson, P. F. (1962). Freedom and resentment. In Freedom and Resentment and Other Essays, pp. 1–25. London: Methuen. 12