Killing in Self-Defense

Killing in Self-Defense* Jonathan Quong Can we ever permissibly kill others in self-defense, and if we can, what is the moral principle that permits such lethal acts? Consider a Villainous Aggressor who wants to kill me simply because he hates me, and I see him coming at me with an axe. Can I permissibly kill him if this is the only way to save my own life? What about the case of an Innocent Aggressor? This is someone who has the intention to kill me, but they are not morally responsible for forming this intention: perhaps a villain slipped them a drug which temporarily caused them to have this intention. Finally consider an Innocent Threat—someone who threatens my life even though they have formed no intention to kill me and exercise no agency at all. An example would be someone who has been pushed off a cliff and will land on me and kill me unless I vaporize him with my trusty ray gun first (if I do not shoot he will survive the fall). Like Judith Jarvis Thomson, I think these are all cases where it is permissible to kill one person in order to save my own life.1 Jeff McMahan, Michael Otsuka, and several others, however, have offered a powerful argument against the permissibility of killing Inno* Earlier versions of this article were presented at Manchester Centre for Political Theory and Philosophy events at the University of Manchester, the Nuffield Political Theory Workshop in Oxford, and the Warwick Research Seminar in Political Theory. I would like to thank all the participants at those events for their questions. For comments on the article and helpful discussions about it I am very grateful to Kim Brownlee, Richard Child, Matthew Clayton, Cecile Fabre, Joe Horton, Seth Lazar, Dan McDermott, Jeff McMahan, David Miller, John O’Neill, Michael Otsuka, Katie Portwin, Mark Reiff, Dave Sant, Ben Saunders, Hillel Steiner, Zofia Stemplowska, Rebecca Stone, Victor Tadros, Steve de Wijze, Andrew Williams, and an anonymous referee. Cecile Fabre, Jeff McMahan, and Michael Otsuka each deserve special thanks for their generous help. 1. Judith Jarvis Thomson, “Self-Defense,” Philosophy & Public Affairs 20 (1991): 283–310. My descriptions of Villainous Aggressor, Innocent Aggressor, Innocent Threat, and Bystander are all drawn from Thomson. For an argument which has many parallels with Thomson’s, see Suzanne Uniacke, Permissible Killing: The Self-Defence Justification of Homicide (Cambridge: Cambridge University Press, 1994). Fiona Leverick also endorses the general approach to self-defense developed by Thomson and Uniacke. See Fiona Leverick, Killing in Self-Defence (Oxford: Oxford University Press, 2006), chap. 3. Ethics 119 (April 2009): 507–537 䉷 2009 by The University of Chicago. All rights reserved. 0014-1704/2009/119030001$10.00 507 508 Ethics April 2009 cent Aggressors and Innocent Threats in self-defense.2 They focus on the widely accepted view that we may not kill Bystanders in order to save our own life. Bystanders are those people who have no significant causal involvement in whatever it is that threatens my life and have done nothing that would normally make them liable to be killed.3 I cannot, for example, permissibly push an innocent passer-by in front of a flying javelin that is about to kill me in order to save my own life. But, McMahan and Otsuka argue, there is nothing of moral relevance to distinguish Bystanders from Innocent Threats and Innocent Aggressors. All are equally innocent—the Innocent Aggressor and Innocent Threat have done nothing that would normally make them liable to be killed in the way that Villainous Aggressor clearly has (innocent persons are thus always morally nonresponsible for a threat, and I will use these terms interchangeably). Otsuka and McMahan thus argue that such Innocent Aggressors and Threats must have an equivalent status to Bystanders. If we accept the premise that it is impermissible to kill Bystanders, we must therefore conclude that it is similarly impermissible to kill Innocent Aggressors or Innocent Threats.4 I call this the Moral Responsibility Argument. In Section I of the article I show that despite its intuitive appeal, the moral responsibility argument does not establish the impermissibility of killing Innocent Aggressors or Threats. The most the argument can do is remind us of the presumption against killing innocent persons. This presumption, however, can sometimes be overridden given a sufficiently compelling argument. Sections II and III are devoted to developing such an argument. I claim that Thomson’s distinction between Bystanders on the one hand, and Aggressors and Threats on the other, 2. Jeff McMahan, “Self-Defense and the Problem of the Innocent Attacker,” Ethics 104 (1994): 252–90, The Ethics of Killing: Problems at the Margins of Life (New York: Oxford University Press, 2002), 398–411, “The Basis of Moral Liability to Defensive Killing,” Philosophical Issues 15 (2005): 386–405, and Killing in War (Oxford: Clarendon, 2009), sec. 4.1.5; Michael Otsuka, “Killing the Innocent in Self-Defense,” Philosophy & Public Affairs 23 (1994): 74–94, reprinted in his Libertarianism without Inequality (Oxford: Oxford University Press, 2003), chap. 4. For similar views, also see Kimberly Kessler Ferzan, “Justifying Self-Defence,” Law and Philosophy 24 (2005): 711–49, 733–39; David Rodin, War and SelfDefense (Oxford: Clarendon, 2002), 79–89; and Noam Zohar, “Collective War and Individualistic Ethics: Against the Conscription of ‘Self-Defense’,” Political Theory 21 (1993): 606–22. 3. There is a subcategory of persons who are sometimes included as Bystanders who have had a causal involvement in the lethal threat: Justified or Innocent Causes. I address such cases in sec. B of the appendix. 4. Technically, McMahan believes we can permissibly kill some people he labels Innocent Threats, but that is because he draws the distinction somewhat differently than I do here. McMahan does not believe we can permissibly kill Innocent Threats as I have defined them. See, e.g., McMahan, “The Basis of Moral Liability,” 393–94. Quong Killing in Self-Defense 509 is roughly the right one for thinking about permissible defensive killings but that this distinction requires a different rationale than the one offered by Thomson. My argument for distinguishing Bystanders from Aggressors and Threats turns on a particular account of what it is to use someone as a mere means to save your own life. Before proceeding any further it is worth clarifying a few points about the arguments that are to follow. First, I am only considering cases of self-defense that involve lethal threats; nonlethal threats are not the topic here, except insofar as they may help illuminate some aspect of killing in defense against lethal threats. Second, this essay does not address the interesting issue of when, if ever, we should be excused (due to duress, panic, or lack of knowledge) for an impermissible instance of killing. I am only interested in the question of when, if ever, we are morally permitted to kill a single person in self-defense. The essay thus does not concern itself with those cases of self-defense where one or both parties have mistaken beliefs regarding matters of fact—I assume all parties always have full and accurate knowledge of the situation. Finally, the essay will be limited to the consideration of only those cases of self-defense where the choice is between one life and another. Cases of self-defense where multiple lives are at stake, or where killing (or not killing) one person will have fatal consequences for many other lives, are only considered insofar as they prove helpful in understanding some aspect of the one versus one cases. I. THE MORAL RESPONSIBILITY ARGUMENT Proponents of the moral responsibility argument deny that it is permissible to kill Innocent Threats and Aggressors in self-defense. Like Bystanders, Innocent Threats and Aggressors have done nothing to make themselves liable to be killed. Since it is impermissible to kill Bystanders, it should likewise be impermissible to kill Innocent Threats and Aggressors. In this essay I set aside the various debates and questions surrounding the notion of moral responsibility, for example, the debate between compatibilism and incompatibilism or the distinction between attributive and substantive responsibility.5 Even if we assume the definition of moral responsibility to be unproblematic, my aim in this section is to show that the most plausible version of the moral responsibility argument does not establish that it must be impermissible to kill Innocent Threats and Aggressors in self-defense because it is impermissible to kill 5. For this latter distinction, see T. M. Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University Press, 1998), chap. 6. 510 Ethics April 2009 Bystanders in self-defense.6 The most the moral responsibility argument does is remind us of the general presumption that it is wrong to kill innocent people. The arguments in this section are thus purely negative: they only establish that we cannot move unproblematically from the premise that it is impermissible to kill Bystanders to the conclusion that it must therefore be impermissible to kill Innocent Threats and Aggressors. This section does not establish that it is in fact permissible to kill Innocent Aggressors or Threats. In order to justify that conclusion, some positive argument needs to be provided, and this task is left until Sections II and III. With that clarification out of the way, let us consider the strongest version of the moral responsibility argument:7 1. 2. 3. 4. 5. 6. Each person has a right not to be killed by others unless they have done something to waive or forfeit that right. The only way to waive or forfeit your right not to be killed is either to consent that someone may kill you or by being morally responsible for causing a threat to someone else’s life who has not waived their right not to be killed by you.8 Like Bystanders, Innocent Threats and Aggressors are not morally responsible for what threatens your life, nor have they consented to being killed by you. Innocent Threats and Aggressors therefore retain their right not to be killed by you. It is impermissible to kill someone if they retain their right not to be killed. Therefore it is impermissible to kill Innocent Threats and Aggressors in self-defense. The immediate difficulty, however, is that premise 5 conflicts with what most people believe. Consider, for instance: Trolley: There is a runaway trolley whose brakes have failed headed down a track where five people are trapped and will be killed unless the trolley is diverted. Fortunately there is a side track onto which 6. For two other recent critiques of what I call the moral responsibility argument, see Helen Frowe, “Equating Innocent Threats and Bystanders,” Journal of Applied Philosophy 25 (2008): 277–90; and Seth Lazar, “The Right to Kill? A Critique of Jeff McMahan’s Theory of Liability to Defensive Killing” (unpublished manuscript, University of Oxford). Frowe’s independently developed argument overlaps in some ways with the argument in this article, and I note this below where appropriate. 7. There are several other formulations of the moral responsibility argument, but I believe they are all clearly unsuccessful. 8. Or by being morally responsible for causing someone else to reasonably believe they face a lethal threat. Quong Killing in Self-Defense 511 the trolley can be diverted, but there is one person trapped on this side track, and this person will be killed if we divert the trolley.9 Is it permissible to divert the trolley? Most people believe, and I agree, that it is permissible. Even those who insist that it is impermissible to turn the trolley will likely believe that there are some circumstances under which it is permissible to kill an innocent person (e.g., proportionate and unintended civilian casualties caused in the pursuit of a just war). It therefore seems that the moral responsibility argument fails because it implausibly assumes that people who retain their right not to be killed can never be permissibly killed. There is, however, an obvious reply available to the proponents of the moral responsibility argument. They can reply that rights may be permissibly infringed when there is a sufficiently important reason for doing so. Saving five innocent people’s lives in Trolley seems to be a sufficiently important reason, but saving a single life, even if it is your own, is not sufficiently important to justify infringing a person’s right not to be killed. Provided the number of lives at stake are equal, it is impermissible to kill some in order to save others. So the appeal to Trolley fails to undermine what we can call The Narrow Thesis: It is always impermissible to kill one (or more) innocent person(s) in order to save another single person (or equivalent number of people). If the narrow thesis is true, then the moral responsibility argument appears perfectly valid. In fact, the narrow thesis makes most of the steps in the argument above redundant—it establishes the conclusion in a single step. But notice what this means—the narrow thesis establishes the truth of the view that it is impermissible to kill Innocent Aggressors and Threats in self-defense, but it does so by stipulating the very point which is disputed, namely, whether or not it is always impermissible to kill one innocent person to save another person. This is what critics of the moral responsibility argument deny, and in the face of such disagreement one cannot simply assert the truth of the narrow thesis. If it is to do more than establish the conclusion by stipulation, the narrow thesis must be supported by arguments which show that it is plausible and that it is not vulnerable to objections that might be pressed against it. Of course the narrow thesis does have a great deal of plausibility, primarily because it tracks the widely accepted intuition that killing is 9. For the classic presentation of Trolley and different variations of it, see Judith Jarvis Thomson, Rights, Restitution, and Risk (Cambridge, MA: Harvard University Press, 1986), chaps. 6, 7. 512 Ethics April 2009 worse than letting die.10 But even if killing is worse than letting die, this does not entail that the narrow thesis must be correct. There might be other nonconsequentialist considerations that sometimes override or nullify the fact that killing is worse than letting die, even when the number of lives at stake is equivalent. Indeed, there seem to be examples where this is exactly what happens. Consider the following pair of cases: Man on the Track : There is a runaway trolley whose brakes have failed headed down a track where your child is trapped and will be killed by the trolley. Fortunately there is a side track onto which the trolley can be diverted, but there is one man trapped on this side track, and he will be killed if you divert the trolley. Man on the Overpass: There is a runaway trolley whose brakes have failed headed down a track where your child is trapped and will be killed by the trolley. Fortunately you are standing on an overpass under which the trolley must pass before it reaches your child. Next to you stands a large man. This large man’s weight (but not your own) would be sufficient to stop the trolley. If you pushed him off the overpass, he would land on the tracks and stop the trolley before it reaches your child, though he will be killed if you do so. I assume that many people will believe that considerations of partiality make it permissible to act in the first case, but that considerations of partiality do not suffice to make it permissible to act in the second case. If the first half of this judgment is right, then the narrow thesis is false. But even those who do not share this view regarding the permissibility of acting in Man on the Track may still believe there is some moral difference between these two cases. This difference (even if one does not believe it changes what is permissible) is presumably explained by the fact that you would be using or intending the death of the large man in a way you do not use or intend the death of the man in Man on the Track. Provided it is more clearly impermissible to act in Man on the Overpass than it is in Man on the Track, this should cast some doubt on the claim that the narrow thesis is obviously or unproblematically true. The narrow thesis implies that there can never be countervailing considerations in one versus one cases which trump the prohibition on killing an innocent person. It implies that nonconsequentialist considerations—considerations of partiality, the distinction between intending and foreseeing, or the idea of using someone as a mere means—can 10. The narrow thesis might also be thought to track another nonconsequentialist distinction: the difference between intending and merely foreseeing. I am less certain that this supports the plausibility of the narrow thesis, but I omit further discussion of this point for the sake of brevity. Quong Killing in Self-Defense 513 never justify killing one innocent person to save another. But if these kinds of nonconsequentialist reasons can make a difference in distinguishing cases like Man on the Track versus Man on the Overpass, why assume, without further argument, that such considerations cannot be used to distinguish Bystanders from Innocent Aggressors and Threats in cases of self-defense? Of course, after careful reflection we may decide that there is no sound nonconsequentialist reason for treating Bystanders differently from Innocent Aggressors and Threats, but it seems right to say that we must consider the various arguments that might support this asymmetric treatment before we can confidently endorse the narrow thesis. What we cannot do is assert the truth of the narrow thesis as a way of denying that any nonconsequentialist method for distinguishing Bystanders from Innocent Aggressors and Threats is even possible. If this is right, then the moral responsibility argument cannot establish the conclusion that it is impermissible to kill Innocent Aggressors and Threats in self-defense. The most the argument does is point us to the thesis that would need to be true (the narrow thesis) in order for this conclusion to be true. Whether or not the narrow thesis is true will turn on whether any successful argument can be mounted justifying the differential treatment of Bystanders as opposed to Innocent Aggressors and Threats. In Sections II and III, I provide such an argument, and so if the argument I present in those sections is sound, the narrow thesis is false. One final point before moving on. Throughout, I assume that general rights, such as the right not to be killed by others, do exist, but can sometimes be permissibly infringed. I distinguish between a rights infringement (permissible act) and a rights violation (impermissible act).11 I will not provide a complete explanation of why I think this view of rights is superior to a view of rights-as-absolutely-inviolable, but here is one reason for thinking so. Regardless of which view of rights you hold, you are likely to believe that in cases such as Trolley where we may permissibly harm or kill an innocent person, something is owed to the innocent person (or their beneficiaries) who has been harmed or killed. On the infringing picture of rights it is quite easy to explain why something is owed to this person: something was done to them or taken from them that they had a right not to have done or taken, and so they are entitled to compensation. If you believe that rights are absolute, and thus believe there is no rights violation or infringement committed in cases such as Trolley (i.e., you believe that the right not to be killed is qualified in a way that does not give the one person a right not be killed in cases like Trolley), then it is more puzzling to explain why compensation is owed in cases such as Trolley. Since nothing was done to that person that they had a right not to have done to them, why is 11. For a version of this distinction, see Thomson, Rights, Restitution, and Risk, 40–42. 514 Ethics April 2009 there a further moral requirement to compensate them? It will not be very convincing to reply that they have a right to compensation when they are harmed or killed in such and such circumstances—what we are asking is why they have this right. The most plausible answer is because something that was theirs was taken from them. The view of general rights which are sometimes permissibly infringed fits nicely with this sensible answer, whereas the conception of rights as absolutely inviolable seems to beg the further question as to why we compensate people when we do things to them that are permissible.12 But, to be clear, I think the arguments I present in the rest of the essay (with certain modifications) could also be accepted by those people who hold the view that rights cannot be permissibly infringed. Of course the burden of justification must be borne by the side arguing in favor of the permissibility of some rights infringement. The default position is that the right protects the right holder, and any infringement of the right thus requires a special sort of justification. But if we believe that premise 5 is false—that rights are not absolute— then it remains open to provide a positive account which explains why killing Bystanders would constitute a rights violation, whereas killing Innocent Threats and Aggressors would be merely a rights infringement. This is what I try to do in the final section. II. A PRINCIPLE OF DEFENSIVE KILLING Having examined the moral responsibility argument, it will be helpful briefly to consider Thomson’s alternative position which does not focus on the moral responsibility of Threats and Aggressors.13 After considering various formulations of a principle for killing in self-defense, Thomson settles on the following (my paraphrase): T1: If X’s killing you will violate your right not to be killed, then X loses their right not to be killed by you, and so you may permissibly kill X in self-defense.14 12. For a more comprehensive defense of the view that rights can be permissibly infringed, see Thomson, The Realm of Rights (Cambridge, MA: Harvard University Press, 1990), or Rights, Restitution, and Risk, chap. 5. The alternative position is presented in Hillel Steiner, An Essay on Rights (Oxford: Blackwell, 1994), particularly chap. 3. Also see Steiner, “Self-Ownership and Conscription,” in The Egalitarian Conscience: Essays in Honour of G. A. Cohen, ed. Christine Sypnowich (Oxford: Oxford University Press, 2006), 88–101. 13. Here I refer exclusively to Thomson’s most recent position presented in “SelfDefense.” In an earlier paper Thomson makes one comment which is consistent with the position I defend in this article. See Rights, Restitution, and Risk, 41. 14. Thomson, “Self-Defense,” 302. Uniacke presents a similar principle, arguing that we possess a right to act in proportionate self-defense when we are subject to an unjust harm or threat of harm, where unjust means roughly ‘would violate one or more of our rights’. See Uniacke, Permissible Killing, chaps. 5, 6. For her agreement with Thomson regarding the definition of unjust threats, see Permissible Killing, 175. Quong Killing in Self-Defense 515 Why does Thomson rely on the idea of a rights violation in her principle of defensive killing? Thomson asks us to consider an alternative principle of permissible defensive killing that does not rely on the idea of a rights violation: T2: You can permissibly kill X if X will otherwise kill you.15 The problem with T2, as Thomson shows, is that it can be used by a Villain to justify killing an Innocent Victim in self-defense who was only attacking the Villain because the Villain was about to kill them. T1 avoids this perverse conclusion provided we accept Thomson’s further claims that (a) the person who poses the initial lethal threat in any situation is threatening to violate someone’s right not to be killed, and so (b) the initial lethal threat loses their right not to be killed by their potential victim. Villains thus forfeit their right not to be killed by their Innocent Victims when they initiate an attack, but Innocent Victims do not forfeit their right not to be killed when they act in self-defense since the Villain has already forfeited their right not to be killed, and thus the Victims’ act of self-defense involves no rights violation. Thomson applies this argument in exactly the same way to all Aggressors and Threats regardless of their moral responsibility. Thomson maintains that all Aggressors and Threats are, by virtue of being the initial deadly threat, going to violate someone’s rights. This means Aggressors and Threats forfeit their right not to be killed, and so it is permissible to kill all Aggressors and Threats in self-defense without thereby losing one’s own immunity from being killed. There is, however, a serious problem with Thomson’s T1. As Otsuka and McMahan rightly point out, it doesn’t seem plausible to suppose, as Thomson does, that Innocent Aggressors or Threats are guilty of violating anyone’s rights.16 This is because we tend to believe that, in order for one person to violate another’s rights, there must be responsible agency. For example, although it would make sense to say that you were killed by a boulder which fell from the top of a cliff as the result of an earthquake, it would not make sense to say the boulder violated your right not to be killed. The boulder has no agency, thus cannot be subject to any moral duties, and so cannot violate (or even infringe) your rights. The same is surely true of a large baby falling from a cliff.17 Since the large falling baby is an Innocent Threat, this seems to show, 15. Thomson, “Self-Defense,” 289. 16. Otsuka, “Killing the Innocent in Self-Defense,” 79–84; McMahan, “Self-Defense and the Problem of the Innocent Attacker,” 276. 17. The example of a falling stone or boulder is used both in Otsuka, “Killing the Innocent in Self-Defense,” 80; and in McMahan, “Self-Defense and the Problem of the Innocent Attacker,” 276. 516 Ethics April 2009 contra Thomson, that T1 does not justify the killing of Innocent Threats or Innocent Aggressors in self-defense since neither has exercised responsible agency and thus neither can be shown to violate someone’s right not to be killed.18 The difficulty we face is that in order to distinguish between Villains and Innocent Victims, it looks as if we need to make use of the concept of a rights violation and the notion of moral responsibility on which this concept relies. But if we make this move to moral responsibility, we will then be unable to distinguish Bystanders from Innocent Aggressors and Threats as Thomson would like to do. The central problem with views of self-defense like Thomson’s, and the moral responsibility argument, is that they wrongly focus our conceptual attention on the possible targets of the defensive killing and ask whether the possible target is going to violate someone’s rights or is responsible for the lethal threat. Instead, I believe we should begin by focusing on the agent acting in self-defense. On this view the permission to act in self-defense is justified by appeal to the agent-relative value that each person’s life has for them. By that I simply mean that your life is of particular importance to you. One implication of this fact is that most deontological theories do not require people to make great sacrifices on behalf of others unless they have voluntarily incurred an obligation to do so.19 You would be a hero if you risked your life by rushing out into choppy and dangerous waters to rescue someone who was drowning, but morality does not demand that you risk your life in this way. Your life is your own, and so morality does not require that you give it up or put it at significant risk for another person or even several other people.20 By appealing to the agent-relative value each person’s life has for them, I therefore mean nothing more than the following: that each person is understood to have a powerful agent-relative per- 18. The same problem holds true for Uniacke’s account of the right to act in selfdefense against “unjust” threats, since this notion also relies on the idea that an unjust threat is someone who is going to violate your rights. For a version of this critique applied to Uniacke, see Tziporah Kasachkoff, “Killing in Self-Defense: An Unquestionable or Problematic Defense?” Law and Philosophy 17 (1998): 509–31, 518–19. 19. This claim is clearly consistent with the view that you may sometimes be required to save others when you can do so at reasonable cost to yourself. For an argument that this duty of rescue can also include the duty to kill attackers in defense of third parties, see Cecile Fabre, “Mandatory Rescue Killings,” Journal of Political Philosophy 15 (2007): 363–84. 20. Some people believe that if the numbers are significant enough (a hundred, a thousand, a million . . .) morality can in fact require you to sacrifice your life for others. In this essay I remain agnostic about this claim. I simply hold that morality does not require you to risk your life for another single person or even several other people. Quong Killing in Self-Defense 517 mission to avoid sacrificing or significantly risking their own life for the sake of others (absent any obligations voluntarily incurred).21 This appeal to agent-relative value is the source of a moral permission and not merely the grounds for an excuse. We are excused when we act in a way that we should not have, but mitigating circumstances entail that it would have been exceedingly difficult or even impossible for any normal or reasonable person to act as morality requires. Skeptics of the agent-relative view might believe this is what happens when someone kills an Innocent Threat or Aggressor in selfdefense: they should not have killed under those circumstances, but panic or self-interest made it very difficult to act in accordance with morality’s requirements. I believe, however, that agent-relative considerations can do more than provide the grounds for excuse in life and death cases: they can alter the boundaries of what is permissible. Consider cases of rescue. If a child is drowning and X can rescue the child at the cost of muddying their trousers, most will agree X is required to save the child. But suppose instead X can only save the child at the cost of becoming a paraplegic. Here I think many would agree X is no longer required by morality to save the child. Since the death of the child is worse than the cost of becoming a paraplegic, the only explanation is that agent-relative considerations have altered what morality permits. While it is true that this is a case of rescue rather than killing, I think it usefully highlights that agent-relative considerations can dramatically alter what morality permits us to do, even when someone else’s life is at stake. I believe it is this agent-relative view about what morality can reasonably require of us that is at the heart of the permission to act in selfdefense.22 Just as you cannot be expected to sacrifice or incur a significant risk to your life to rescue another person unless you have agreed to do so, I do not believe (subject to certain constraints) that morality can demand that you allow yourself to be killed if you have not consented to this killing, or otherwise engaged in some voluntary act which has the same effect. Morality cannot require you to sacrifice your life for 21. This agent-relative value is objectively, rather than subjectively, specified. That is, I do not hold that a person may permissibly attach any weight they like to their own life or projects. Rather, I claim that there is an objective sense in which each person may permissibly attach much greater weight to their own life in comparison to the lives of others. I do not claim that this agent-relative permission attaches to other things that a person may value, for example, their religious beliefs. 22. In this respect I agree with what Nancy Davis and Susan Levine have said about the basic grounds for self-defense. See Nancy Davis, “Abortion and Self-Defense,” Philosophy & Public Affairs 13 (1984): 175–207, 192–93; and Susan Levine, “The Moral Permissibility of Killing a Material Aggressor in Self-Defence,” Philosophical Studies 45 (1984): 69–78, 73–74. 518 Ethics April 2009 another single person when you rightfully possess the means to save yourself, and this is what provides the basic grounds for a permission to kill in self-defense.23 There are, however, a number of difficulties with even a limited agent-relative permission to kill in defense of one’s own life.24 First, there seems to be a great deal of difference between the claim that morality cannot require you to engage in some life-threatening activity to rescue someone else and the claim that morality permits you to kill someone else in defense of your own life. Another difficulty with the agent-centered approach to self-defense is that it does not appear to draw any distinction between Threats and Aggressors on the one hand and Bystanders on the other. If the reason why you are permitted to kill in self-defense is explicable in terms of the value your life has for you, then surely it should be irrelevant whether it is a Threat, an Aggressor, or a Bystander that you need to kill to save yourself? This, of course, is deeply counterintuitive, and it is not a conclusion that I wish to endorse. I will address these challenges in Section III, but first let me introduce the agent-centered Principle of Defensive Killing (PDK) I will defend against them: PDK: You can permissibly kill X if X will otherwise kill you,25 provided the following things are true: (i) killing X is the only reasonable course of action that can save your life, and (ii) you have not waived or forfeited your permission to act in self-defense. Now consider the clauses attached to PDK. The first clause should be relatively uncontroversial, even if the idea of reasonable necessity cannot be rendered precise. You cannot permissibly kill X if you could 23. Again, for different expressions of this agent-relative view of the permission to act in self-defense, see Davis, “Abortion and Self-Defense”; and Levine, “The Moral Permissibility of Killing a Material Aggressor in Self-Defence.” 24. Both objections to the agent-relative view in this paragraph are raised by Kasachkoff and Leverick, and the latter objection is also raised by McMahan. See Kasachkoff, “Killing in Self-Defense,” 524–26; Leverick, Killing in Self-Defence, 52; and McMahan, “SelfDefense and the Problem of the Innocent Attacker,” 270–71. 25. I set aside two complications. First, I set aside complications that arise when we consider cases where there is some uncertainty as to whether X will kill you if you do not kill them. I will not try to settle the question regarding how likely a lethal attack must be before acting in self-defense becomes permissible. Second, I set aside complications that arise when we consider cases where an agent’s beliefs do not match the true state of the world. That is, cases where the agent may believe, reasonably or not, that X is going to kill them when X is not in fact going to kill them. I will only say that someone may act without fault or blame, even if the agent’s action is impermissible given all the relevant information. Quong Killing in Self-Defense 519 also save yourself by pinching X on the nose. On the other hand, you can permissibly kill X if the only other way to save yourself would render you a quadriplegic. Between these extremes there will be difficult cases where it is unclear whether you have a reasonable option other than using lethal force, but again, I do not believe the existence of this gray area renders the clause untrue or unnecessary.26 Second, I suppose that you have waived or forfeited your permission to act in self-defense if you have voluntarily done any one of a number of things that would also constitute waiving or forfeiting your right not to be killed. These actions include but are not necessarily limited to consenting that X may kill you, attempting to kill or gravely injure X (or some other innocent person) without his consent when X poses no threat to you, intentionally causing X (or some other innocent person) to believe that you are attempting such an act, or otherwise being culpably responsible for initiating a threat to X’s life. You do not waive or forfeit your permission to act in self-defense when you act in self-defense against someone who is the initial Threat or Aggressor, or if you act in defense of someone else who is being attacked by a Villainous Aggressor. Your permission to kill in self-defense is thus closely tied to your right not to be killed, in the sense that so long as you retain the latter right you normally also retain the former permission. However, and this is one of the ways PDK differs from Thomson’s T1, what justifies or grounds your permission to act in self-defense is not the fact that X will otherwise violate your right not to be killed. The permission to kill in self-defense is grounded rather in the agent-relative value your life has for you. While this is what grounds the permission to act in self-defense, the permission is also (as with all moral permissions) conditional on your not having waived or forfeited the permission. It makes sense to suppose that, in general, whenever you waive or forfeit your right not to be killed, you have also forfeited your permission to act in self-defense. There could, however, be cases where you waive your right not to be killed while retaining your permission to act in self-defense (e.g., this 26. This reasonable necessity clause applies differently to morally guilty as opposed to morally Innocent Aggressors and Threats. That is, the Victim will be required to bear a greater burden when facing Innocent Aggressors and Threats. For example, we might expect the Victim to endure five minutes of extreme pain if this was necessary to avoid killing an Innocent Aggressor, but this might not be morally required in the case of a Villainous Aggressor. This asymmetry fits nicely with the view that we infringe an Innocent Aggressor’s right if we kill them, but we infringe no such right when we kill a Villainous Aggressor. It seems plausible to suppose that we have a moral duty to bear certain costs to avoid harming or killing someone when this would also wrong them, but that either we do not have a duty to bear those costs when the harming or killing would not wrong the target, or at least the costs we must bear in the latter case are substantially less. This asymmetry is suggested in McMahan, “Self-Defense and the Problem of the Innocent Attacker,” 265–66. 520 Ethics April 2009 is a common view regarding combatants in war). For simplicity, I will assume that whenever you waive or forfeit your right not to be killed, you have also lost your permission to act in self-defense. I believe that PDK does, roughly, everything that a principle of permissible defensive killing should do.27 To begin, PDK successfully distinguishes between Villains and Innocent Victims. A Villainous Aggressor coming at me with an axe cannot use PDK as the justification for killing me in self-defense when I try to shoot him first. He has forfeited his permission to act in self-defense, but I have done nothing similar, and so if it is necessary, I may kill him in self-defense. PDK also picks out Threats and Aggressors, but not Bystanders, as permissible targets of defensive killing. Of course it does so merely by stipulation— by focusing on those people who would otherwise kill you, and so this stipulation will require a justification. The task of providing this justification will be taken up in Section III. Finally, notice that PDK makes it permissible for Innocent Threats and Innocent Aggressors to ‘fight back’, that is, to kill someone in selfdefense who is about to kill them in self-defense. This is a major difference between PDK and Thomson’s principle T1.28 PDK allows anyone to act in self-defense provided they have not done anything to waive or forfeit their right not to be killed. This means that if you are about to zap a falling Innocent Threat with your trusty ray gun, the Innocent Threat may also permissibly use his gun, should he have one handy, to shoot you to prevent you from shooting him. Since neither of you has done anything to waive or forfeit your right not to be killed by the other, you each have a permission to kill the other in self-defense.29 There are two worries about this conclusion. First, is it coherent? It is. There need be nothing problematic about two people each having a permission to act in ways that are incompossible. In a race, for example, both athletes are permitted to try to win, but only one of them can. Advocates of the view that rights must be compossible and inviolable may object that races are different since each party has a liberty-right to try and win the race, and this means that no party has a claim-right against any other not to try to win the race.30 But cases involving In27. I say ‘roughly’ since, though I believe PDK correctly deals with all standard cases of self-defense, I also believe that there are further cases, not covered by PDK, where it is permissible to kill someone to save yourself even when that person will not otherwise kill you. I discuss some of these cases in Secs. III and IV. 28. It also differs from Uniacke’s position in this way. 29. Again, in this respect I follow Davis’s position as presented in “Abortion and SelfDefense,” 193–94. 30. Here I draw on the Hohfeldian categorization of rights. For clear expositions of these categories, see Thomson, The Realm of Rights, chap. 1; or Steiner, An Essay on Rights, chap. 3. Quong Killing in Self-Defense 521 nocent Aggressors or Threats seem to be different. If both the Threat and the Victim are innocent then each retains their right not to be killed, and this should mean that each has a claim-right not to be killed by the other, so it cannot also be true that each party is at liberty to kill the other. This would imply the paradoxical conclusion that each is under a duty not to kill the other, but each is also at liberty to kill the other. Surely this cannot be correct? This worry raises complex questions about the nature of rights, issues I cannot fully engage with here. I will restrict myself to the following remarks. It does not seem to me incoherent to think of claimrights as generating powerful, yet defeasible, prima facie duties to do or forbear from doing certain actions. If this is true, then under certain circumstances the permission we each possess to act in self-defense may outweigh the claim-right others have against us killing them. On this view, acting in self-defense will necessarily involve the infringement of the claim-right Innocent Aggressors and Threats possess against being killed. The permission to act in self-defense would thus be no ordinary liberty-right. It would be much more than the mere absence of a duty: it would be a special permission or exemption from observing other people’s valid claim-rights under certain circumstances.31 By contrast, those who believe that rights cannot conflict in this way are invited, in cases involving Innocent Threats and Aggressors, to assume that each party loses their claim-right against the other and retains only the libertyright to act in self-defense.32 31. Note the following feature regarding the structure of my argument. I claim that Innocent Aggressors and Threats have the same moral status as Bystanders but that we each possess a special agent-relative permission to kill Innocent Aggressors and Threats, but not Bystanders. As I will argue in Sec. III, this agent-relative permission covers the former, but not the latter, group because the mode of agency we exercise in the two cases is different. I thus endorse Warren Quinn’s view that the stringency of one’s right not to be killed depends on the mode of agency that is exercised in the killing. One has more stringent rights not to be killed in certain ways (e.g., when one’s body or property is used as a means) than in other ways (when one’s body or property is not used as a means). See Warren Quinn, “Actions, Intentions, and Consequences: The Doctrine of Double Effect,” Philosophy & Public Affairs 18 (1989): 334–51, 346–47. 32. There is then the further question of whether, in cases involving Innocent Threats or Aggressors, third parties would be prohibited from interfering. Is it, for example, plausible to suppose that the spouse of one of the parties would not be permitted to intervene on behalf of their spouse? A satisfactory resolution of this question as it pertains to agent-relative accounts of self-defense is beyond the scope of the present essay. All I will say here is that in such cases I am tempted to endorse the proposition that some third parties may permissibly intervene for partial reasons. In cases where one party has forfeited their right not to be killed but the other party has not, then it is clearly true that third parties may intervene on behalf of the person who retains their right, but not on behalf of the person who has forfeited their right. For an illuminating discussion of third party rights and duties in defense cases, see Davis, “Abortion and Self-Defense,” esp. 194–96. 522 Ethics April 2009 The second worry is whether it is plausible to suppose that Innocent Threats and Innocent Aggressors should be permitted to fight back once they realize they are about to be killed in self-defense. I believe it is. If you are under attack from an Innocent Threat, you are both innocent in the relevant sense—neither of you has waived or forfeited your permission to defend your life—and so neither of you should be precluded from acting in self-defense. There is no good reason why you should be permitted to do so, but that the equally Innocent Threat to you should not be permitted. Some may object that PDK is problematic because it fails to offer a unitary account of self-defense, in the sense that it treats Villainous Aggressors very differently from Innocent Aggressors and Threats (i.e., no rights violation or infringement occurs when the former are killed, but an infringement does occur when the latter are killed). According to Suzanne Uniacke it is “inappropriate and distracting to posit separate justifications of homicide in self-defence based on a distinction between culpable and blameless aggressors.”33 On this view, each permissible instance of defensive killing must have, at bottom, the same justification. PDK does, however, offer a unitary account of self-defense. The reason why it is permissible to act in self-defense is the same in all cases: each person’s life has special value to them, and this is what grounds the permission to act in defense of our own life. PDK’s dualism only applies to the question of who retains the permission to act in selfdefense. Here PDK does depart from accounts, such as those offered by Uniacke and Thomson, which make no distinction between Villainous as opposed to Innocent Aggressors and Threats. But this is one of the virtues of PDK. Unlike the theories of Uniacke and Thomson, PDK does not implausibly hold that Innocent Aggressors or Threats violate anyone’s rights and thereby forfeit their permission to act in self-defense. Instead PDK accepts that people who exercise no responsible agency cannot plausibly be said to forfeit anything, and so if we threaten to kill them in self-defense, it must be the case that they then have a similar permission to act in defense of their own life. It is a major advantage that PDK recognizes this difference between the different types of permissible defensive killing, and it accords with our wider intuition that when our permissible actions harm or kill innocent people we may be required, where possible, to compensate those people or their legal beneficiaries. In this respect I am in complete agreement with an early comment Thomson made on this topic: “If you are an ‘innocent threat’ to my life . . . and I can save my life only by killing 33. Uniacke, Permissible Killing, 207. Quong Killing in Self-Defense 523 you, and I therefore do kill you, I think I do owe compensation, for I take your life to save mine. If so I infringe a right of yours but do not violate it.”34 III. THE RATIONALE Even if PDK fits with our intuitions, we also want to know why we should endorse a principle of permissible defensive killing that allows us to kill people even when those people are morally innocent. As mentioned earlier, PDK faces two major challenges. First, we must explain why the permission or liberty to act in self-defense should include the permission to harm or kill an innocent person. Even if our life has special agentrelative value, the permission to kill innocent people in self-defense seems much stronger and more controversial than the permission to let innocent people die if rescuing them would be too risky. Second, we must also explain why we can permissibly kill Innocent Aggressors and Threats, but not Bystanders. Since I have claimed that the permission to act in self-defense is fundamentally grounded in the agent-relative value each person’s life has for them, why does PDK distinguish certain targets of killing as permissible and others as impermissible? From the agent’s point of view this distinction seems irrelevant. The difficulty for PDK is not to explain why we can permissibly kill Villainous Aggressors in self-defense—I assume this is a relatively unproblematic case where someone forfeits their right not to be killed. In this final section I argue that when we act in self-defense, we still face certain moral constraints. Your life may have great agent-relative value to you, but this does not mean that morality permits you to do anything necessary in order to stay alive. The exercise of this permission must cohere with other normative beliefs we have, and I suggest that one of these beliefs—the belief that it is wrong to use someone as a mere means—can explain why we are permitted to kill some innocent people in self-defense but not others. Let’s begin with a relatively trivial case: Sleepwalking: You are sleepwalking, and are about to unwittingly trample on your good friend’s prize-winning ant colony. The only way your friend can stop you is by pinching you hard on the arm which will cause you to wake up and stop walking (you will suffer temporary pain, but no lasting damage from the pinch). Is your friend permitted to pinch you? I think that in this case it would be permissible for your friend to infringe your right not to be pinched in order to protect his ant colony. If this explanation is on the right track, then the following principle seems plausible: 34. Thomson, Rights, Restitution, and Risk, 41. 524 Ethics April 2009 Infringing in Defense: You can permissibly infringe someone else’s rights when this is necessary to defend something that belongs to you. Of course this principle is too simple. First, it needs to be restricted by some notion of proportionality. It would not be permissible, for example, for your friend to kill you if this was for some reason necessary to save his ant colony from destruction. So this means we need: Infringing in Defense 2 : You can permissibly infringe someone else’s rights when this is necessary to defend something that belongs to you, provided the infringement is proportionate. How exactly we should settle the matter of proportionality is a difficult question, and one that I cannot properly address here. I will assume that an infringement is proportionate provided that the value of what the right infringed is protecting is no greater than the value of what is being threatened. Some may feel this definition of proportionality allows too much, others too little, but we should be able to agree on the idea that a proportionality clause is necessary.35 But our principle is still incomplete. Suppose that the only way for your friend to wake you up and thus prevent you from trampling his ants is by pinching a third person, his neighbor, hard on the arm (this will cause the neighbor to yelp loudly which will wake you in time). Is this permissible? It is far less obvious that it is. But why should there be a difference between these two versions of the story? I think the difference lies in the fact that your friend will be taking advantage of the neighbor’s presence to do something he could not do without him, whereas this is not true in the case where he pinches your arm. When you threaten to trample your friend’s ants and your friend pinches you to save the ants, your friend does not use you to do something he could not do without you.36 You are threatening to destroy something your friend would safely possess so long as you were not there. In the case where he pinches the neighbor, however, things are different. Here he 35. There remains the further question of whether the reference to value in the proportionality clause is to be understood in objective or subjective terms. My view is that value will have to be understood in an objective sense. To interpret the proportionality clause in subjective or agent-relative terms would mean that, if the agent cared enough about what they were defending, there would be no infringement that was too great to be permissible, and I take this to be an obviously unacceptable conclusion. 36. One might object that your friend does use your pain as a means of manipulating you (i.e., to wake you) and thus he uses you in the same (but much less severe) manner that a torturer uses his victim. Setting the severity of the pain caused aside, there is another significant difference between your friend’s pinch and the torturer. In pinching you, your friend does not use you to do something he could not do without you (save his ants), whereas the torturer does need his victim to achieve his objective (extract information or simply cause suffering). This difference is the one that my example is meant to illuminate. Quong Killing in Self-Defense 525 can only save his ants because the neighbor is present—if the neighbor were not there he would be unable to protect his ants. One way of explaining this distinction is to say that your friend would be using his neighbor if he pinched his arm in a way that he would not be using you if he pinched your arm. X uses Y, in the special sense I am interested in here, when Y’s presence is causally necessary for the successful execution of X’s action. Put counterfactually, without Y’s presence, X cannot perform the required action or achieve the relevant goal. Y is, we could say, a tool which X needs to successfully complete some action or attain some goal. I believe that, other things being equal, it is morally worse to act contrary to someone’s rights when this allows you to do something you could not do without that person than it is to act contrary to someone’s rights as a means of keeping what you would be able to keep without that person. This view, I believe, captures part of the Kantian idea that there is something particularly wrong about treating another person as a mere means to your own ends.37 When your friend pinches the arm of his neighbor he uses his neighbor in a manner that has normative salience, but in the case where he only pinches you he does not. In the latter case he is not taking advantage of your presence—he is preventing you from taking something that he would have were it not for your threat. This distinction is not novel. In particular, Warren Quinn proposes something similar, distinguishing between those cases where the harmed or killed person’s presence presents an opportunity or advantage, as opposed to those cases where the harmed or killed person’s presence is perceived as an obstacle. Quinn suggests that when we harm or kill in the former cases our agency is ‘opportunistic’, whereas in the latter cases our agency is merely ‘eliminative’.38 This same point can be put another way. We tend to believe it is wrong, other things being equal, to profit or gain from someone else’s suffering, but I think this intuition is strongest when the person who has been harmed can complain—“you were just using me!” This complaint makes sense and has moral force when the person who acts could not have successfully executed their action but for the presence of the person harmed, but it is not clear this complaint is plausible when this is not the case. An Innocent Aggressor or Threat cannot plausibly complain, “you were just using me,” if they are harmed or killed by someone acting in self-defense. In these cases there is no sense in which the 37. I do not deny, however, that there are serious objections that have been pressed against the view that there is a coherent and morally salient conception of what it is to use someone as a mere means. See, e.g., Jonathan Bennett, The Act Itself (Oxford: Clarendon, 1995), 212–25. 38. Quinn, “Actions, Intentions, and Consequences,” 344. 526 Ethics April 2009 acting agent takes advantage of, or exploits the presence of, the Innocent Aggressor or Threat for their own gain.39 I thus disagree with Otsuka when he says that killing an Innocent Threat is “analogous to those most deplorable cases in which you kill a Bystander in order to eat her body to prevent yourself from starving or in order to replace your failing vital organs with healthy ones.”40 You may ‘use’ people in all these cases in some loose sense, but there is a morally salient disanalogy since in killing the Innocent Threat you do not exploit their presence as a means of doing something you could not do without them, but this is what you do in the Bystander cases.41 If I am right about this, we need to modify our principle yet again: Infringing in Defense 3 : You can permissibly infringe someone else’s rights when this is necessary to defend something that belongs to you, provided (a) the infringement is proportionate and (b) the presence of the person whose right is infringed is not necessary in order for you to keep what belongs to you.42 But we still face a problem, which can be seen clearly in the following example: Alcove : You are in a tunnel and see a runaway trolley headed straight for you and it will kill you if do not escape. You can only escape the trolley by squeezing into a small alcove in the tunnel. Unfortunately for you, there is already someone in the small alcove. You could pull them out of the alcove and onto the tracks, where they will die, so you may fit in the alcove and save yourself.43 The person already in the alcove is an Innocent Obstructor. Innocent Obstructors are not morally responsible for what threatens your life, but they block your path to safety and can only be moved at the cost of their life. I believe it is clear that in cases such as Alcove it is impermissible to kill Innocent Obstructors. The difficulty, however, is that if you were to kill the Innocent Obstructor in Alcove, you would not be 39. Shelly Kagan considers this method of explaining the normative difference between Aggressors and Threats as opposed to Bystanders and Shields, and while he declares that it is “a promising possibility,” he ultimately rejects it for reasons I consider below. See Shelly Kagan, The Limits of Morality (Oxford: Clarendon, 1989), 140–44. 40. Otsuka, “Killing the Innocent in Self-Defense,” 87. 41. Frowe independently developed a similar objection to Otsuka. See Frowe, “Equating Innocent Threats and Bystanders.” 42. By ‘presence’ I mean their presence in the current circumstances, not their existence more generally. The principle is thus temporally specific, and so I believe it is not vulnerable to a certain form of objection that might be pressed against it. For the form of this objection in a different context, see Kagan, The Limits of Morality, 95–99. The justification for this temporal restriction is discussed in sec. B of the appendix. 43. This example is borrowed from Thomson, “Self-Defense,” 291. Quong Killing in Self-Defense 527 taking advantage of, or exploiting, their presence: you would be safe from the oncoming trolley if they were not present. Cases like this thus appear to pose a serious problem for our proposed principle.44 Consider, for example, the following objection to the argument thus far.45 Even if we grant the distinction I have drawn (or Quinn’s similar distinction) between acts which require the presence of the person who is harmed and acts which do not require the presence of the person who is harmed, and even if we further grant that the former acts are morally worse than the latter acts, this does not establish that the latter acts are permissible. It is still impermissible to kill an innocent person to save your own life, the critic might say, though it may be even worse to do so when the killing is also a ‘using’ in the manner I have specified. In support of this claim, the critic can then point to examples involving Innocent Obstructors. In these cases, like Alcove, it is impermissible to kill the Innocent Obstructor even though doing so would not be a ‘using’. If it is impermissible to kill Innocent Obstructors even though such killings do not involve ‘using’ Obstructors, surely it should also be impermissible to kill Innocent Threats and Aggressors? I believe there is an effective reply to this objection. The objection assumes that it is always impermissible to kill Innocent Obstructors, but this is false. Consider the following example: Meteor: A small meteor is falling toward you and will kill you if it lands on you. The only safe place where you can avoid the meteor is your very tiny one-person car. But there is already someone in your car—this person was placed there without their consent by some third party. You could, however, pull them out of the car, thereby ensuring they will die, so you can get inside to safety. I think it is clear that you may act in this case—it is, after all, your car. It is unfortunate that you and the other person cannot both survive the oncoming meteor, but you have a prior claim to your car and so can permissibly remove the other person from the car to save your life. Here we have a case where it is permissible to kill an Innocent Obstructor to save your own life. (Some might protest that you do not kill the Innocent Obstructor in Meteor but rather you let them die since they only die because you deny them something that belongs to you [your 44. It is partly because of examples such as Alcove that Kagan doubts that the rationale for self-defense I am offering is sound. See Kagan, The Limits of Morality, 140–44. 45. McMahan offers this objection as a means of showing why Quinn’s distinction between eliminative and opportunistic agency cannot ground a permission to kill Innocent Aggressors and Threats in self-defense. See McMahan, Killing in War, sec. 4.1.5. 528 Ethics April 2009 car].46 But the example can be modified. Suppose the car is on the edge of a cliff and the only way to remove them from the car is to throw them out the door and off the edge of the cliff. In this case you certainly kill the Innocent Obstructor, but I think you are still permitted to do so.) Now we must identify the relevant difference between Meteor and Alcove. I suggest the relevant difference is this: in Meteor you have an exclusive prior claim to the space being occupied by the Innocent Obstructor, whereas in Alcove you do not. This, I think, explains why acting in the former case is permissible, whereas acting in the latter case is not. Before elaborating on that thought, let me first defend the claim that the person in the alcove has a prior claim to that space. It makes a great deal of sense to suppose that we each have at least a prima facie claim to the space we occupy, provided we have not entered someone else’s private property without their consent. Even if no harm is caused, it is presumptively wrong to move someone without their permission. Suppose Albert and Betty are in a public park, and Albert desires to get a better view of the lake but can only do so if Betty moves from her present location where she is enjoying a picnic. Albert cannot simply move Betty, even if this causes her no harm, and even if he is somehow able to do this without touching Betty (thus avoiding violating any claims she might have against nonconsensual touching). Betty has a presumptive claim-right to her location even though it is a public park, and even though it is more or less arbitrary that Betty arrived at that particular spot first. Of course, Betty’s rights may sometimes be permissibly overridden if the benefits to some other person or persons are substantially greater than any interest of Betty’s in remaining where she is. But we cannot simply ignore Betty’s rights whenever there might be some small overall good in doing so. The same consideration that supports the view that people have presumptive claim-rights over their body motivates the thought that people must have presumptive claim-rights over the physical space where their body is. Without a right over the space they occupy, people would have almost no control over their own life. Without such rights we would lack the most basic elements of human agency—we would be unable to limit others from using our body for their own purposes. We must, at a minimum, have control over our own body, and the space it occupies, in order to realize an idea of persons 46. For the idea that the main difference between killing and letting die depends on whether the acting agent deprives the dying person only of things (aid or resources) which belong to the acting agent, see F. M. Kamm, Morality, Mortality (Oxford: Oxford University Press, 1996), vol. 2, pt. 1; or McMahan, “Killing, Letting Die, or Withdrawing Aid,” Ethics 103 (1993): 250–79. Quong Killing in Self-Defense 529 as more than mere means to be used for others’ purposes.47 It would be strange to assert rights over the body grounded in a view of persons as self-directing agents rather than mere tools yet insist that individuals lack any claim against others from having their body moved without their consent. If the foregoing remarks about each person’s prima facie claim to the space they occupy are sound, we are now in a position to see how this ought to inform our more general principle of infringing in defense. Suppose I am right, and it is permissible to kill in Meteor but impermissible to kill in Alcove. We have identified the consideration that distinguishes the two cases: in the former case, but not in the latter, the acting agent does not require space to which the Innocent Obstructor has some legitimate claim. I therefore suggest that killing in Alcove is relevantly analogous to using a Bystander’s body as a shield to save yourself from some oncoming lethal projectile. In both cases you cannot save yourself but for something to which the Bystander or Obstructor has a prior claim. To kill in such cases would be to take advantage of, or exploit, the body or the physical space of someone else. In both cases you take advantage of things over which the Bystander or Obstructor has a rightful claim, and in doing so you shift the burden of death from yourself onto them. This, I think, is why killing in such cases would be impermissible. However, in cases involving Threats and Aggressors, as well as nonstandard cases of Innocent Obstructors such as Meteor, the acting agent does not take advantage of, or exploit, the body or the physical space of someone else. We can thus use the following counterfactual test to determine if our killing of an innocent person, X, would be impermissible: if X and all the things currently belonging to X were suddenly to disappear, would your life be saved? 48 If the answer to this question is yes, then you do not 47. This thought, for example, is partly what motivates Quinn’s distinction between eliminative and opportunistic agency. See Quinn, “Actions, Intentions, and Consequences,” 350–51. 48. There are, of course, interesting and difficult questions regarding how to imagine counterfactuals when the space occupied by an Innocent Obstructor/Bystander is removed from the situation. I cannot consider this issue in detail, but I believe the following question, or something like it, is essential in determining the permissibility of actions which would harm or kill Obstructors/Bystanders: if the Obstructor/Bystander’s space was suddenly inaccessible to you (e.g., protected by a barrier), could you still save your life? If the answer is no, then your life-saving action falls foul of the proposed test. So, for example, if your only escape route from a lethal threat requires driving over and killing an Innocent Obstructor, you may not do so. This is true because, in the counterfactual, we imagine the Obstructor’s space is blocked by some barrier, and thus you could not escape to safety. On the other hand, the proposed test would permit you to redirect a runaway trolley which was headed toward you on to a side track, even if there was a Bystander trapped on that side track. This is true because even if the area where the 530 Ethics April 2009 exploit or take advantage of X by killing them. You merely keep what you would have (your life) without them. If the answer is no, however, this shows that you do exploit or take advantage of X in order to save yourself. You cannot survive but for the presence of X or other things over which they have a rightful claim. Since your survival depends on things over which X has a rightful claim, to take them and thereby kill X would be to shift the harm of death from yourself onto X by using X’s entitlements against them.49 This, I believe, would be to use X as a mere means to your own survival, and as such it should be deemed impermissible. Thus we arrive at our final version of the infringing principle: Infringing in Defense 4 : You can permissibly infringe someone else’s rights when this is necessary to defend something that belongs to you provided (a) the infringement is proportionate and (b) you would be able to keep what belongs to you if the other person and all their property were not present. Several things are worth noting about this principle. First, the principle does depend on a theory of justice to tell us what parts of the world people have prior claims to. In other words, whether an infringement will be permissible according to our principle will depend upon our other beliefs about rights or justice, but I do not think this kind of dependence on further views about justice is surprising or troubling. Second, I leave open the possibility that infringements which cause only very small harm in pursuit of a very large benefit may be permissible even if they fall foul of clause b. In this way Infringing in Defense 4 should be read as providing sufficient conditions for a permissible defensive infringement, and not the necessary conditions. Third, Infringing in Defense 4 assumes only a two-person conflict—I do not claim that it can be applied to cases where an unequal number of people are involved. Once the benefits to be had from a proposed infringement are significantly greater than the harm done by the infringement, something that will often be true in multiparty cases, we may require a mod- Bystander is currently trapped was blocked by some barrier, you would still be able to successfully turn the trolley on to the side track. 49. Note that because the relevant counterfactual is whether your life would be safe if the innocent person and all their property were absent, you are permitted to use the property of an Innocent Threat or Aggressor to defend yourself from them. If such a person and their property suddenly vanished, you would be safe, and thus if you do need something that belongs to them (e.g., their gun) to defend yourself from them, you may make use of it since you do not take advantage of the Threat or Aggressor in doing so. Quong Killing in Self-Defense 531 ified, or an entirely different, normative principle.50 Fourth, the principle does not rule out all killing or harming of Bystanders: it will sometimes permit the killing or harming of a Bystander (or an Innocent Obstructor) provided that the person is not used in the way that Bystanders and Innocent Obstructors are normally used when they are harmed or killed. Infringing in Defense 4 therefore covers a broader range of cases than PDK. The latter principle focused only on the subset of cases where X would otherwise kill you, whereas the former principle covers all cases where you might infringe one person’s right in order to defend something that belongs to you. Consider, however, an example that might seem troubling for Infringing in Defense 4: Bridge : You are being pursued by a Villainous Aggressor and in order to escape to safety, you must run across a narrow and wobbly public bridge. Unfortunately, a bird-watcher is currently standing on the bridge, oblivious to your plight. If you were to get onto the bridge you would wobble it such that the bird-watcher would fall off into the ravine below and be killed. It is not possible to communicate with the bird-watcher and get them to leave their location before you are killed by the Villainous Aggressor.51 The bird-watcher in this example is another Innocent Obstructor. But is this case, like Alcove, one where it is impermissible to kill the Innocent Obstructor? It might seem as if the account proposed above cannot explain why killing the bird-watcher would be impermissible. After all, you do not need access to the precise physical space where the birdwatcher stands on the bridge (the bridge is wide enough to accommodate two people), it is simply that in getting on the bridge you will foreseeably topple the bird-watcher. Does this mean that you can proceed? I think not. Because the bridge in this example is so wobbly that it cannot sustain two people at the same time, it seems best to treat the bridge as an amount of physical space that can only feasibly contain one person. To get on one part of the bridge is effectively to seize the whole bridge (including where the bird-watcher stands) since it is the kind of structure that can only hold one person at a time. You thus do need the physical space to which the bird-watcher has a prior claim in 50. So, for example, though some people’s intuitions regarding Thomson’s famous Loop variant on Trolley appear to conflict with the infringing principle as stated, this need not undermine the principle’s application in one versus one cases. For Thomson’s Loop example, see Thomson, Rights, Restitution, and Risk, 101–2. For a very persuasive argument in defense of the view that it is impermissible to kill the one in Loop, see Michael Otsuka, “Double Effect, Triple Effect, and the Trolley Problem: Squaring the Circle in Looping Cases,” Utilitas 20 (2008): 92–110. 51. This example is modified from one provided in Thomson, “Self-Defense,” 290. 532 Ethics April 2009 order to escape the Villainous Aggressor, and this makes it impermissible to topple the bird-watcher.52 The picture of infringing rights in self-defense that I have presented resembles the permission to let other people die in certain circumstances. When it is permissible to let someone die, I believe a major part of what makes it permissible is the fact that the dying person loses only what they would have had by virtue of your efforts, or other things that belong to you.53 In letting someone die in these permissible instances you merely keep what belongs to you in order to avoid a risky or costly action. I claim that for an act of killing in self-defense to be permissible, a related idea should apply. In order to permissibly kill an innocent person, it must be true that saving your life does not require their presence or the presence of anything to which they have a rightful claim. It must be true that saving your life does not depend on acquiring that other person’s body or property. To act when this condition is not met would involve treating that other person as a mere means to your own end. You would be exploiting things over which they have a claim in order to shift the harm of death from yourself onto them. Your use of someone in this manner changes the moral character of your action to such an extent that, I claim, it affects the permissibility of your action. Although they are not exactly analogous, the moral distinction between killing and letting die and the constraint on killing in self-defense share a common deontological structure, one which views each person as an independent agent who cannot be used merely as a means for another’s ends. IV. CONCLUSION Proponents of the moral responsibility argument deny that it is permissible to kill Innocent Threats and Aggressors in self-defense. Like Bystanders, such Innocent Threats and Aggressors have done nothing to make themselves liable to be killed, and so the presumption against killing one innocent person in order to save another should protect Innocent Threats and Aggressors in the same way it protects Bystanders. I have argued that this conclusion is mistaken. Each person has an agentrelative permission to kill in self-defense. This permission extends to the killing of all Threats and Aggressors, but not (most) Bystanders, and not (most) Innocent Obstructors. The reason we must treat these groups differently is grounded in a view regarding what it is to treat someone else as a mere means to our own survival. We treat someone 52. If the bridge was your exclusive property then, as in Meteor, I believe it would be permissible to kill the Innocent Obstructor. 53. Again, see both Kamm, Morality, Mortality, vol. 2, pt. 1; or McMahan, ‘‘Killing, Letting Die, or Withdrawing Aid.” Quong Killing in Self-Defense 533 as a mere means to our own survival if we kill them when we could not survive without them or something else to which they have a rightful claim. There are a number of further points worth noting about the argument I have advanced. First, consider what the rationale I have offered implies about Innocent Shields, a category that is not directly addressed by PDK. Innocent Shields are people who are not themselves a lethal threat, but they are attached or close enough to whatever it is that does pose the lethal threat so that you cannot stop the lethal threat without killing them.54 For example, an Innocent Shield may be trapped inside a runaway trolley that is speeding toward you, and the only way to stop the trolley is by firing a small missile at the trolley which will kill the Shield (if you do not fire the missile they will survive). If you kill an Innocent Shield in cases like this you do not fall foul of Infringing in Defense 4, and I think this explains why many people believe such killings are permissible.55 Second, remember that Infringing in Defense 4 does not apply to people who have waived or forfeited their right not to be killed by you. People such as Villainous Aggressor have no right not to be killed by you, and so the conditions of our infringing principle are not applicable to the killing of Villainous Aggressor. This is as it should be. The normative constraints on your actions should be very different when you are infringing someone’s rights as compared to cases where you commit no infringement at all, and so it should be an unsurprising conclusion that there are things we cannot do to innocent people that we may do to save ourselves from Villainous Aggressors. Third, it is possible that Infringing in Defense 4 is merely an instance of a higher-order principle regarding when we can permissibly infringe other people’s rights generally, not merely in cases where we defend something that belongs to us. Whether or not this is true, and if it is true, what this higher-order principle might look like are questions I hope to pursue in the future. Finally, consider what would be true if we were not permitted to kill Innocent Aggressors or Threats in self-defense. A moral theory which made these acts impermissible would be one where we could each be required to allow other people to kill us even though we have done 54. Note that my definition of an Innocent Shield will include persons who are sometimes referred to as Bystanders. For example, on my view X is an Innocent Shield if he is standing close enough to a runaway trolley such that the only way you can stop the runaway trolley from hitting you is by throwing a grenade at the trolley which will also kill X. 55. There can, of course, be more difficult cases where you must also make use of the Innocent Shield or their property in order to save yourself. In such cases I think the killing would be impermissible for the sorts of reasons I have outlined. 534 Ethics April 2009 nothing to waive our rights. This would be an extraordinarily demanding moral theory. Of course sometimes morality may be very demanding, but a morality which required us to allow ourselves to be killed when we have done nothing to forfeit our rights, and when we already possess the means to save ourselves, blurs the distinction between moral heroism and moral requirement. It would be heroic to allow yourself to be killed in this way, but requiring you to do so diminishes the sense in which your life is your own. PDK, on the other hand, does not require any innocent person to allow themselves to be killed by someone else provided they have the means to save themselves. PDK is thus more in tune with a certain idea that I think is central to deontological morality: that each of us has only one life to live. I think this means we cannot be required by morality to allow ourselves to be killed, provided we can prevent this from happening without treating anyone else as a mere means to our own survival. Appendix In this appendix I confront two further objections that might be pressed against the arguments in the main text. A Consider a luck egalitarian objection.56 This objection begins with the premise that inequalities between persons that are the result of brute luck are unjust, and only inequalities which arise from responsible choices are just. Whenever burdens and benefits are purely a matter of brute luck, the luck egalitarian will recommend that these burdens and benefits be equally distributed. But what happens in cases of self-defense when the Aggressor or Threat is innocent? In such cases we have two morally equivalent parties (in terms of the chance/choice cut) and one indivisible burden: death. In these cases it might seem that the appropriate luck egalitarian strategy would be to flip a coin or use some other fair lottery system for deciding who should live and who should die. If this argument is sound, then PDK is wrong because it does not employ a fair lottery to determine who will live. Fortunately for PDK, this luck egalitarian proposal is not sound. First, as Otsuka points out,57 if you are facing a lethal threat from some object, say a javelin, it would be morally impermissible to toss a coin to decide whether or not to push a Bystander in front of you so that the javelin will kill the Bystander rather than you. An advocate of the luck egalitarian proposal would face the difficult task of explaining why the 56. The main aim of Levine’s paper is to respond to a similar objection. See Levine, “The Moral Permissibility of Killing a Material Aggressor in Self-Defence.” 57. Otsuka, “Killing the Innocent in Self-Defense,” 76–77 n. 9. Quong Killing in Self-Defense 535 coin toss is not appropriate in the case of Bystanders but is appropriate in the case of Innocent Aggressors and Threats. Second, if the luck egalitarian insists that you must toss a coin in all cases involving two innocent parties (including Bystanders), they are not preventing someone from being the victim of brute bad luck, they are simply changing the moment at which someone becomes the victim of brute bad luck. The luck egalitarian wants this moment to be when the coin lands either heads or tails, but why is this moment any different than an earlier moment when, say, an Innocent Threat falls off a cliff toward someone with a ray gun? We could instead choose to say that we all face some ex ante probability of becoming Innocent Threats or Aggressors, and if we all face the same ex ante probability, there is no unfairness in instituting a policy which allows people to kill in selfdefense in such cases. It’s not clear what the luck egalitarian can say in reply. Nothing is being distributed unfairly by PDK—we are all given the same permission (the permission to kill Threats and Aggressors in self-defense), we are all assumed to be equally likely to be the target of this permission, and thus ex ante we are all accorded the same package of burdens and benefits with regard to killing in self-defense.58 B Consider a category that I ignored in the main text: Justified or Innocent Causes.59 Justified or Innocent Causes (hereafter Innocent Causes) are people who are causally responsible but are either not morally responsible, or are fully justified, in initiating a chain of events which now results in a threat to your life, but they themselves do not form part of the direct threat. For example, a Miner who nonnegligently caused a collapse of the mine shaft which now threatens your life because you are now trapped in the mine and will die of dehydration is an Innocent Cause.60 Are such Innocent Causes to be counted as Threats or are they Bystanders? They certainly appear to be Bystanders relative to the immediate threat (they are not threatening you now as you face the risk of dehydration), and it seems wrong to claim that you could kill the innocent Miner who was the cause of the mine shaft collapse if this was somehow necessary to save your own life. But McMahan claims that principles of defensive killing, such as PDK, which try to distinguish between Bystanders as opposed to Aggressors and Threats without making reference to the moral responsibility of the targets, will be unable 58. Assuming there are no special reasons to believe that certain types of people are more likely to end up in one category rather than the other. In cases where some types of people are systematically more likely to be killed, then things will be different if we accept the luck egalitarian premise. 59. I take the terms Justified and Innocent Causes from McMahan, “Self-Defense and the Problem of the Innocent Attacker,” 267. 60. This example is borrowed, with modifications, from ibid., 267. 536 Ethics April 2009 to reach this intuitively plausible conclusion.61 After all, the Innocent Cause did cause what now threatens your life, and so why are they any different from an Innocent Threat? I suggest that we approach Innocent Cause examples in the same way I argued we ought to approach other cases, that is, by asking whether, in killing the Innocent Cause, we would be able to do something we could not do but for the presence of the Innocent Cause. Suppose the Miner who nonnegligently caused the collapse of the mine shaft which now threatens your life is trapped in the mine with you. It will take rescue crews more than five days to dig both of you out. You and the Miner each have one bottle of water, unfortunately you would need to drink two bottles of water in order to avoid dying from dehydration during this time, whereas the Miner only needs his own bottle to survive. Can you take his bottle, thereby killing him but saving yourself? It seems clear to me that you may not do this, and the reason you may not do this (or similar things which would kill the Miner) is because you obtain a benefit you could not obtain but for the Miner’s presence or property. Someone might object that if the Miner had never existed, then you would not need his bottle of water (since the mine shaft wouldn’t have collapsed), and so it is not true that but for the Miner’s presence or property you would not survive. This reply, however, ignores the fact that the counterfactual test which determines whether we use someone in the specified manner is temporally restricted (a point made in n. 42). When we evaluate some action and ask whether our action would constitute using someone as a mere means, we are asking, given the facts as they are now, does our action have a certain moral feature? To know if a particular course of action requires the use of some particular person or their property we must look to the facts as they are, not as they might have been had the person in question never existed. We cannot justly appropriate someone’s property in the present by claiming that if they had never existed, the property would be unowned. The only relevant counterfactual is whether we could save ourselves if the person and their property were suddenly removed from the current situation, but everything else remained as it is. There are some who might object that this temporal restriction looks arbitrary. If the Miner is morally innocent at all points in time (e.g., both when he is about to cause the accident and later when he is trapped with you), then shouldn’t it be equally permissible or impermissible to kill him at all points in time? This worry about arbitrariness, however, ignores a central feature of deontological morality, which is that differences in space and time can make a difference to the moral permissibility of our actions since those differences often entail differences in how we treat the people affected by our actions. Consider the contrast between the two cases described in Section I: Man on the Track versus Man on the Overpass. Both men in this pair of examples are 61. Ibid., 278–81. Quong Killing in Self-Defense 537 equally innocent—what differentiates them is merely where they happen to be located in space. But where they are located in space makes a difference to how we might treat the two men. Because we do not use (or perhaps intend) the death of the man on the track it may be permissible to turn the trolley in this case, but it is clearly not permissible to kill the large man on the overpass because this killing would require that we use (or perhaps intend) his death as a means to our end. Although one’s spatial location does not affect one’s moral status (i.e., whether one is morally innocent or culpable), it can affect the permissibility of actions. Since almost all deontologists accept that such spatial differences between persons can make a difference regarding what is permissible, it should not be puzzling to suppose that deontological ethics may also be sensitive to temporal differences.