Theories of self-defense are numerous and quite varied. Most take as their master hypothetical an Attacker and a Victim and then ask when the Victim is justified in preventing the Attacker’s threatened attack by employing force. (From now on I shall assume that the Victim is employing deadly force, and that, unless otherwise noted, the harm Victim fears at the hands of Attacker is death or serious bodily injury.) Usually, theorists only vary the master hypothetical by varying the extent to which Attacker is culpable for the threatened attack.
I believe that most of the extant theories of self-defense are deficient and are so because the master hypothetical is far too simple to represent the complexity of selfdefense. For example, there is considerable disagreement over whether Victim’s mistaken but “reasonable” perception of an imminent attack renders his defensive use of force justifiable or only excused. But his mistake can only be exposed by reference to a different perspective on the events—God’s, a trial court’s or jury’s, or a Third Party’s. And note that the perspective of the trial court, jury or Third Party can itself be mistaken relative to that of God.
I believe that by focusing on the permissibility of Victim’s response, these theories overlook the value of beginning the analysis from the perspective not of Victim, but of a Third Party who is willing to intervene to prevent the attack. Moreover, I believe that instead of seeking to pigeonhole responses as either justified or excused, we should ask whether a response is culpable or nonculpable. I hope that the utility of emphasizing the Third-Party perspective and culpability will become apparent.
So let us assume Third Party perceives what he believes is likely an impending attack on Victim (with deadly force) by Attacker. Third Party believes he can prevent the attack by using deadly force against Attacker. Will Third Party—who is not a member of Victim’s family and has no duty to defend Victim—be culpable for using deadly force against Attacker?
To answer that question requires a theory that tells us when such a use of defense-ofothers force is culpable or nonculpable. What elements should go into such a theory?
The Array of Possible Attackers
First, the theory should account for the different kinds of Attackers that Third Party might perceive. Who are they?
(1)The Culpable Attacker (CA)
CAs come in different stripes. There is (i) the CA who intends to kill Victim. There is (ii) the CA who intends to impose a risk of death on Victim (for example, by playing involuntary Russian roulette on Victim). There is (iii) the CA who is intending to do something that he knows will impose a risk of death on Victim, though he might wish it did not (for example, the CA intends to detonate explosives in the vicinity of Victim, or to drive 100 miles per hour within inches of Victim).
(2)The Innocent Attacker (IA)
Like CAs, IAs come in different stripes. There is (i) the IA who mistakenly believes Victim is a CA out to kill him or out to kill someone else who is not culpably threatening Victim. There are (ii) IAs who are not morally responsible agents—the young, the insane and the senile. And there is (iii) the IA who is nonculpably unaware that his conduct threatens Victim (for example, an IA who does not realize that the use of his cell phone will detonate a bomb in Victim’s house).
(3)The Anticipated Culpable Aggressor
Some persons are not yet CAs in that they are not contemplating attacking Victim. However, Third Party may predict that one who is not yet a CA will become one in the future, when it may be impossible to protect Victim. So suppose Third Party sees Anticipated Culpable Attacker (ACA) about to enter ACA’s office. Third Party knows that Victim is in ACA’s office, awaiting ACA for a business conference. Third Party knows that ACA carries a loaded gun; knows that ACA is murderously jealous; knows the ACA’s wife has left a voicemail message on ACA’s office answering machine confessing an affair with Victim; and knows that ACA always listens to his voicemail as soon as he enters his office, even when there is someone else present waiting to speak with him. So Third Party anticipates that once in his office, ACA will become a CA before Victim can leave the office. The question a theory of self-defense must answer is whether Third Party may nonculpably use deadly force against an ACA who is not yet a CA. (Assume Third Party does not believe he has a nonlethal method available that would be equally as effective as deadly force for averting the threat to Victim—though whether Third Party must use nonlethal force when it is available and effective is an issue I will raise when I take up proportionality, necessity and retreat.)
(4)The Anticipated Innocent Aggressor
Some persons are not yet contemplating attacking Victim. However, as with ACAs, Third Party may predict that such persons will become aggressors in the future when it may be impossible to protect Victim. However, unlike ACAs, such persons will be IAs. For example, suppose CIA agents have positioned themselves just outside Enemy Nation’s missile silo. They have been informed that Enemy Nation’s president has just ordered a missile launch against America. They also have been informed that the order is inside the silo on the control computer, awaiting the new shift of soldiers who man the silo. Finally, they know that the soldiers are quite likely to carry out such an order when they read it, although because they believe in the rectitude of their president and the propaganda to which they have been subjected, they will be nonculpable for doing so. And once they are inside the silo and read the order, the CIA agents will be powerless to stop the launch. They believe that to avert the missile launch, they must kill the approaching shift of soldiers before they get inside the silo and read the order—that is, before they form the intent to aggress. (Assume delaying the launch will give America time to take actions that will avert future launches.) Those soldiers are Anticipated Innocent Aggressors or AIAs.
(5)Culpable Persons (who are not CAs)
Another category of persons against whom Third Party might be tempted to use deadly force are persons who are in some way culpable but who are not presently attacking Victim. For example, suppose two people attack Victim with deadly force—in this case, guns—but CA1 has run out of ammunition. CA2 is continuing the attack on Victim, but Third Party’s ability to protect Victim by using deadly force against CA2 is thwarted by the fact that CA1—now just a culpable person (CP)—is in the way of a clean shot. (He currently has his foot caught in something; he is not intentionally shielding CA2, an act that would make him a CA through complicity.) May Third Party nonculpably use deadly force against CA1, the CP, in order to then use deadly force against CA2 and save Victim? Notice that CPs, whose presence is unintentionally preventing defense of Victim, can, like CAs, come in a variety of types. There are CPs (i) who, as in the example above, have just finished attacking Victim; (ii) who attacked Victim in the more distant past (but were not punished for this); (iii) who attacked, not Victim, but someone else on a prior occasion (and were not punished); or (iv) who were culpable to a lesser extent than a deadly CA (and were not punished). A good theory of self-defense should tell us whether Third Party may nonculpably use deadly force against any or all of these CPs to protect Victim.
Finally, there are CPs who are accessories to attacks by CAs. Some of these accessories are actually CAs, as their present participation is instrumental to the success of CA’s attack. But others are persons whose participation in CA’s attack is over—the crime boss who ordered CA’s “hit,” for example, or the accessory who gave CA the gun. If for some reason Third Party can avert CA’s attack by using deadly force against these accessories—perhaps the hit man CA will call off his attack if he learns that his crime boss has been killed—may Third Party use deadly force against the accessory CP rather than against the CA?
There is one category of culpable actors against whom the use of force by Third Party will generally be culpable. Suppose someone fakes a deadly attack on Victim as a (bad) practical joke. We can assume that person has acted culpably in frightening Victim. He is a Culpable Faker (CF). If Third Party knows CF is in fact a CF, then Third Party cannot use deadly force against CF, for that would be unnecessary for Victim’s defense. (Victim needs no defending from a fake attack.) On the other hand, Victim may be acting nonculpably in using deadly force against CF. (Victim will be an IA vis-vis CF by virtue of Victim’s innocent mistake.) What should Third Party do in such a case? It should be clear that he should not use deadly force against Victim, despite Victim’s mistake. If anyone should bear the risk of death here, it should be CF. (The same is also true of someone who was at one time a CA and who has changed his mind but cannot signal Victim of this change of mind; he is not a CF, but like a CF, he is culpable for creating Victim’s belief that he is a CA.) So Third Party should not intervene to protect CF if that requires him to use deadly force against Victim. And if CF, who is now a potential victim of a deadly attack from an IA (Victim), seeks to defend himself against Victim with a real gun, Third Party should either not intervene or intervene on the side of Victim, the IA. CF, though not initially an attacker, is culpable for having created the incident. His life should not be preferred to that of Victim.
Such is the array of possible persons against whom Third Party might be tempted to employ deadly force to protect Victim. My own view is that, other factors to be discussed put aside, Third Party may nonculpably use deadly force against all categories of CA and against CPs who are unintentionally shielding CAs and have culpably risked Victim’s or others’ deaths and have not been punished. ACAs are a more difficult case because they are not yet culpable. They raise the question of just how far preemptive strikes can go without being culpable. (All defensive force is preemptive, a point discussed more fully below; however, ACAs raise the issue of preempting not just an attack but the attacker’s culpability.)
IAs, on the other hand, cannot ordinarily be attacked nonculpably. There is no reason why Third Party should prefer Victim’s life to an IA’s. Both are innocent. There are various accounts of self-defense that suggest Victim’s life should be preferred to IA’s, but these are unpersuasive because they are usually advanced from Victim’s rather than Third Party’s impartial perspective. From the latter perspective, IA and Victim are equally innocent, and their confrontation is equally tragic no matter who survives it. The fact that IA initiates the attack does not alter this verdict. For if Victim may employ deadly force to protect himself from IA, he is as much a threat to IA as IA is to him.
On the other hand, if the numbers count, or if there are other reasons to intervene (for example, Victim—or IA—is on the verge of finding a cancer cure), then Third Party may nonculpably use deadly force on the preferred side of the conflict. There may be two or more IAs and only one Victim. Likewise, there may be more than one Victim and only one IA. If two deaths are worse than one, then Third Party may nonculpably use deadly force to save the greater number, even if those against whom the force is used are innocent. (The situation is similar to the famous trolley hypothetical, where one can save five trapped workers by diverting a runaway trolley from the track on which the five are trapped to a siding on which one is trapped. The numbers favor the diversion, and the body, labor or talents of the one are not necessary to bring about the saving of the five (the one is not “used”). The same is true in a Victim—IA confrontation where the numbers support intervening on one side or the other.)
Riskiness of the Attack
Third Party, in deciding whether to intervene to defend Victim, must consider not only the status of Attacker—culpable or innocent—but also the risks that Attacker will impose on Victim if Third Party does not intervene. Those risks that Third Party perceives may be different from the risks Attacker perceives or intends. If Third Party but not Attacker knows Attacker’s gun is unloaded, then Third Party will perceive the risk to Victim as far less than the risk Attacker perceives. Or if Third Party but not Attacker knows that Attacker’s cell phone is rigged to set off a bomb in Victim’s house, Third Party will perceive a grave risk to Victim even though Attacker perceives none.
Even in a simple case, such as that of a CA who intends to kill Victim by firing a gun, Third Party’s assessment of the risks of death and injury to Victim will be complex. It will be based, first, on Third Party’s estimate that CA will indeed carry through on his present intention and actually fire the shot (a point I take up below). But even if it is 100 percent certain that CA will fire the shot, the risk of death to Victim will be a function of the condition of CA’s gun (is it loaded? Is it jammed?); CA’s marksmanship; the orientation of Victim’s body; Victim’s apparel (is it bulletproof?); and Victim’s physical condition.
Risks not only vary in degree, depending on the epistemic vantage of the person estimating the risk, but also in kind. A risk of death usually is accompanied by a greater risk of bodily injury. And a risk of bodily injury usually carries with it a somewhat lesser risk of death. And frequently risks of death or injury are accompanied by risks to property and to emotional well-being.
Here, as stated earlier, I shall assume that the risk at issue is the death of Victim. Third Party may perceive that risk as high, moderate, small or nonexistent. But again, Third Party’s assessment may differ radically from Attacker’s.
Likelihood of Attack
Theorists of self-defense frequently overlook its perhaps most significant aspect, namely, that it is preemptive in nature—a “preemptive strike,” if you will. Even the CA who at present intends to kill Victim may change his mind and abort the attack. Indeed, even if CA has already fired a shot at Victim, Third Party’s defensive intervention, as opposed to a punitive or vengeful one, will always be aimed at averting the next shot, and the CA can always decide not to fire the next shot. The same is true of all the CAs and IAs. They may change their minds. Or they may be prevented from carrying out the attack by a change in circumstances between the time at which Third Party considers intervening and the feared attack.
One can perhaps think more clearly about these two risks—the risk that there will be an attack, and the risk that the attack, if it occurs, will produce various harms—by considering the following two scenarios, each of which makes one of these two risks a matter of virtual certainty. In Scenario One, Third Party sees CA about to play involuntary Russian roulette on Victim. CA holds a gun to Victim’s head, a gun into which CA has put one live bullet and then spun the chambers. Third Party is certain CA will pull the trigger if Third Party does not shoot CA first. How many empty chambers must the gun have before we will say Third Party should not be permitted to shoot CA? Five, as with a six-shooter? Nine? Twenty? Fifty?
In Scenario Two, CA is again pointing a gun at Victim’s head, but this time Third Party believes that the gun is fully loaded (and is not jammed or otherwise defective, nor is Victim’s head protected by a bulletproof shield, etc.). In other words, Third Party believes that if CA pulls the trigger, Victim will certainly die. However, Third Party is not at all certain CA will pull the trigger. He could be bluffing (a CF). He could intend to do so but lose his nerve. And so on. In this scenario, how confident must Third Party be that CA will pull the trigger in order for shooting CA to be permissible? Will any chance that CA will pull the trigger be sufficient, given that CA is culpable? Or must that chance be above some threshold, and if so, what? If, on the other hand, the person with the gun is an IA rather than a CA—he’s a child, a lunatic or someone who mistakenly believes he is justified in killing Victim—then if Third Party for whatever reason may favor Victim’s life over IA’s, what probability that IA will pull the trigger will permit Third Party to shoot IA? These are important questions that theorists of self-defense virtually never ask, much less answer.
So there are two “risks” that Third Party must consider. There is the risk that, if he does not intervene, Attacker will indeed attack. And there is the risk that, if Attacker does attack, Victim will be harmed. A theorist may believe that those two risks should be treated together as factors in an overall risk to Victim. On that view, a perceived 90 percent probability of an attack with a perceived 90 percent chance of killing Victim should be considered as a single perceived risk of 81 percent of Victim’s death. On the other hand, because one risk is that of a human choice while the other is one of natural events, a theorist may believe they should be disaggregated and treated separately.
Finally, suppose Third Party has 100 percent confidence in his judgment about the status of Attacker—whether Attacker is a CA, a CP, an ACA, a CF or an IA. But suppose, further, that Third Party does not have 100 percent confidence in his judgment about whether there will be an attack and, if there is, how much harm it will cause Victim. In such a case, should Third Party’s estimate of the likely harm (likelihood of attack times, likelihood of various harms if an attack) that renders his defensive force justifiable vary according to whether he believes Attacker is a CA, CP, ACA, CF or IA? In other words, may Third Party use deadly force against Attacker based on a lower estimate of the risk posed by Attacker if Third Party believes Attacker is a CA, say, rather than an IA?