Philosophical Issues, 15, Normativity, 2005

THE BASIS OF MORAL LIABILITY TO DEFENSIVE KILLING

Jeff McMahan
Rutgers University

There may be circumstances in which it is morally justifiable intentionally

to kill a person who is morally innocent, threatens no one, rationally wishes
not to die, and does not consent to be killed. Although the killing would
wrong the victim, it might be justified by the necessity of averting some
disaster that would otherwise occur. In other instances of permissible killing,
however, the justification appeals to more than consequences. It may appeal
to the claim that the person to be killed has acted in such a way that to kill
him would neither wrong him nor violate his rights, even if he has not
consented to be killed or to be subjected to the risk of being killed. In these
cases, I will say that the person is liable to be killed. Although I borrow the
notion of liability from legal theory, and although much of what I say will be
informed by the literature on liability in both tort law and criminal law, my
concern in this article is with moral rather than legal liability.
Liability is different from desert. The claim that someone deserves to be
killed implies that there is a reason to kill her even if it is possible for no one
to be killed; but the claim that someone is liable to be killed has no such
implication. Liability is the broader notion: desert implies liability but
liability does not imply desert. Thus, if a person deserves to be killed, it
follows that he is liable to be killed, but he can be liable to be killed without
deserving to be killed. My focus will be on forms of liability that do not
involve desert; I will not consider punitive or retributive killing. My focus
will instead be primarily on liability to defensive killing, though I will also
consider whether there can be liability to killing that preserves life but is not
strictly defensive because the person to be killed is not the cause of the
threat to be averted. Liability, of course, also extends to forms of harmful
treatment other than killing, but for simplicity of exposition I will focus on
moral liability to be killed. Much of what I will say, however, can be
generalized to other forms of harming.
 The Basis of Moral Liability to Defensive Killing 387

The question I will address is: ‘‘What is the basis of liability to killing?’’
Or, more precisely: ‘‘What must a person who does not deserve to die have
done to make it the case that he would not be wronged by being killed?’’ I
will examine two widely held accounts of the basis of liability to defensive
killing before sketching the outlines of what I think is the most promising
account.

The Rights-Based Account

The Rights-Based Account holds that what makes a person liable to

defensive killing is that he will otherwise violate a right that is sufficiently
stringent for killing to be a proportionate means of prevention.1 According
to its most original and influential exponent, Judith Jarvis Thomson, this
account offers a justification for self-defensive killing in cases that pose
problems for other accounts. It explains, for example, how it can be per-
missible to kill an ‘‘innocent threat,’’ a moral agent who through voluntary
action poses an unjust threat but is not culpable for doing so. Here is an
example.

The identical twin brother of a notorious murderer is driving during a stormy

night in a remote area when his car breaks down. Unaware that his brother has
recently escaped from prison and is thought to be hiding in this same area, he
knocks on the door of the nearest house, seeking to phone for help. On opening
the door, the armed and frightened resident mistakes the harmless twin for the
murderer and lunges at him with a knife.

Assuming that this is the only way to save his own life, may the twin kill the
resident? Most people believe that he would be justified in doing so, even if
the resident is acting reasonably in the circumstances and thus is fully
excused for the threat he poses to the twin. And the Rights-Based
Account provides an explanation of why the twin would be justified:
because the resident would otherwise violate the twin’s right not to be killed,
the resident lacks a right not to be killed by him. In short, the resident
makes himself liable to be killed.
Another type of case that is even more problematic for many accounts
of self-defense involves a ‘‘nonresponsible threat,’’ a person who poses an
unjust threat without being in any way responsible for doing so. Here is
Thomson’s example.

A fat man is enjoying a picnic on a cliff directly above the deck on which you
are lying with your leg in traction. Suddenly a villain pushes him off the cliff. If
he lands on you he will kill you; but he will survive because you will cushion his
fall. You cannot move aside but can save yourself by hoisting your sun umbrella
and impaling him on it.2
388 Jeff McMahan

Again, most people believe that you would be justified in killing the fat man.
And again the Rights-Based Account offers an explanation, which is that,
because the fat man will otherwise violate your right not to be killed, he
lacks a right not to be killed by you. He too is liable.
Some have argued that it is not true that the fat man will violate your
right if you do not kill him, since only a morally responsible agent can
violate a right. Even on Thomson’s own account, for a person to have a
right is just for others to be morally constrained in certain ways.3 And just
as a falling boulder or a charging tiger cannot be morally constrained and
therefore cannot violate a right, so also a person cannot be morally con-
strained except in the exercise of responsible agency. Since the fat man
cannot be morally constrained from having his body used by someone else
against his will, he does not threaten your right when his body is hurled at
you.
This does not, of course, show that the Rights-Based Account is wrong.
It if is correct that the fat man would not violate your right, all this shows is
that the Rights-Based Account does not, after all, have the intuitively
appealing implication that nonresponsible threats may be liable to defensive
killing.
Here is another case that challenges the Rights-Based Account. It is
familiar from discussions of the Doctrine of Double Effect.

A tactical bomber fighting in a just war has been ordered to bomb a military
facility located on the border of the enemy country. He knows that if he bombs
the factory, the explosion will kill innocent civilians living just across the border
in a neutral country. But this would be a side effect of his action and would be
proportionate to the contribution that the destruction of the facility would
make to the achievement of the just cause. As he approaches, the civilians
learn of his mission. They cannot flee in time but they have access to an anti-
aircraft gun.

The traditional question is how the tactical bomber can be justified in

bombing the facility when it would not be justifiable for a terror bomber
to drop a bomb in the same spot, producing the same effects, with the
intention of killing the civilians. My question is different. Assuming that the
tactical bomber would be morally justified in dropping his bomb, would the
civilians be justified in shooting him down in self-defense?
To explain why this case challenges the Rights-Based Account, I need to
distinguish two ways of acting against a right. When one is morally justified
in doing what another has a right that one not do, one infringes her right.
When one acts without justification in doing what another has a right that
one not do, one violates her right.4 Because the tactical bomber acts with
justification, he will merely infringe the civilians’ rights. The usual assump-
tion is that one does not lose one’s rights by virtue of morally justified
 The Basis of Moral Liability to Defensive Killing 389

action. If that is right, the tactical bomber retains his right not to be killed
even though he will otherwise infringe the civilians’ rights. According to the
Rights-Based Account, therefore, the civilians may not kill him in self-
defense.
But this is hard to believe—unless, perhaps, his mission is so important
that they are morally required to sacrifice themselves for the sake of its
success. Can the Rights-Based Account accommodate the common intuition
that the civilians may kill the tactical bomber in self-defense? It cannot be
claimed that his right is overridden by morally weightier considerations; for
the stipulation that his act would be proportionate entails that the failure of
his mission, which would be a consequence of their killing him, would be
worse from an impartial perspective than their being killed. Perhaps then,
contrary to the common assumption, one may lose rights by threatening to
infringe rights. If so, then the fact that the tactical bomber will otherwise
infringe their right not to be killed means that he lacks a right not to be
killed by them—that is, he makes himself liable to be killed by them.
But this too is hard to believe. For if he has lost his right not to be killed
by them, it seems that he can have no right of self-defense against them. Yet
intuitively it seems that, just as it is permissible for the civilians to kill the
tactical bomber in self-defense, it is also permissible for the tactical bomber
preemptively to kill the civilians in self-defense. It is, however, hard to see
how the Rights-Based Account can explain this apparent symmetry. How
can each party have a right to kill the other while at the same time lacking a
right not to be killed by the other? How can one be justified in killing
someone to prevent him from doing what he has a right to do?
The case of the tactical bomber is a problem not only for the Rights-
Based Account but for other accounts as well. I will return to it at the end of
the paper. For the moment I will consider another widely held account of
the basis of liability to defensive killing.

The Culpability Account

According to Thomson, the violation of a right is a matter of what one

person causes to happen to another. It does not require culpability, respon-
sibility, or even agency on the part of the violator. Thus both innocent and
nonresponsible threats may violate rights. Although they may be fully
excused, they are nevertheless liable to necessary and proportionate defen-
sive action. Yet Thomson acknowledges that some people believe that
culpability on the part of the agent is necessary for the violation of a
right, and that even more believe that responsible agency is necessary. The
latter believe that nonresponsible threats cannot violate rights and thus are
not liable to defensive killing, while the former believe that this is true of
both nonresponsible and innocent threats.
390 Jeff McMahan

Some theorists have developed accounts of liability to defensive killing

that focus on issues of responsibility and culpability for unjust threats. An
unjust threat is a threat of harm that would wrong the victim—harm to
which the victim is not morally liable. The Culpability Account of liability
holds that the criterion of liability is culpability for an unjust threat.5 There
is a narrow version of this account that assigns liability only for culpably
posing an unjust threat, while a broader version assigns liability for culpable
responsibility for an unjust threat. According to the broader version, a
person may be liable to killing now if that is necessary to avert an unjust
threat for which he is responsible through past culpable action, even if he is
now no part of the threat. I will assume that the Culpability Account takes
this broader form.
It is possible that the Culpability Account is equivalent to a version of
the Rights-Based Account that assumes that culpability is necessary for the
violation of a right. Like the Rights-Based Account, this account has
trouble with the case of the tactical bomber, since it implies that, because
the tactical bomber does not act culpably, he does not make himself liable
and thus, presumably, may not be attacked in self-defense by the civilians.
But I will put these issues aside in order to focus on a different problem for
the idea that culpability is the basis of liability. Consider a case of attempted
murder.6

A villain, fully intending to kill you, points his gun at you and slowly begins to
squeeze the trigger. You and he both reasonably believe that the gun is loaded
but in fact it is not. You have just enough time to kill him with your concealed
derringer before he pulls the trigger.

Most people believe that you are justified in killing him, for he is culpably
attempting to kill you and you reasonably believe that he will succeed unless
you kill him first. It is not, moreover, obvious that you will wrong him if
you kill him. Yet he in fact poses no threat to you. He is not culpably
responsible for a threat to your life. He does not, therefore, satisfy the
Culpability Account’s criterion of liability.
There are several options. One is to retain the Culpability Account’s
criterion of liability—culpability for an unjust threat—and conjoin it with
an objective account of justification. On this view, because objectively the
villain poses no threat to you, he is not liable and you cannot claim necessity
as a justification. Your killing him would therefore be objectively unjusti-
fied, though it would be fully excused because it would be reasonable given
your epistemic limitations.
A second option is to retain the Culpability Account’s criterion of
liability but conjoin it with a subjective account of justification, according
to which justification depends not the objective nature of the relation
between you and the villain but on what you believe or reasonably believe
 The Basis of Moral Liability to Defensive Killing 391

about that relation. On this view, the villain is not liable but because you
believe, or reasonably believe, that he is liable and that killing him is
necessary for self-defense, your killing him would be justified.
A third option is to modify the criterion of liability. Intuitively this may
seem to be a case in which the villain is liable to defensive killing. Because he
is culpably attempting to kill you, it does not seem that you would wrong him
if you were to kill him, given that you reasonably believe that killing him is
necessary for self-defense and that he is culpably responsible for putting you
in a situation in which you have to act on that belief. It seems, in short, that
the villain’s culpability for the attempt on your life is sufficient for liability
even though there is in fact no threat for which he is responsible.7
Suppose we accept that a person can be liable to be killed by you in self-
defense either if he is culpably responsible for an unjust threat to your life or
if he is culpably making an unjust attempt on your life. We may still believe
that your killing the villain in the case of the culpable attempt is unjustified
(though excused). For the villain’s liability is not a sufficient justification for
killing him if killing him would be pointless because he is in fact harmless. It
would, after all, clearly be unjustified for you to kill him if you knew that his
gun was unloaded, and it is not clear how your lack of morally relevant
knowledge can establish a justification where otherwise there would be
none. (This is one of the oddities of the subjective account of justification:
that it makes mistake of fact a ground of moral justification rather than
excuse.)
It seems, however, that there can be cases in which a person can be
liable for a culpable attempt on your life and in which your killing him
could be justified even when you know that he poses no threat to you. These
are cases in which the kind of threat the person is attempting to pose arises
from a different source and can be averted only by killing the person who is
making the attempt. Here is an example.

Aware that a villain plans to kill you, you begin to carry a gun. On one occasion
you have the opportunity to empty the bullets from his gun and you do so.
Immediately thereafter, he confronts you in an alley and tries to fire. As he
continues to pull the trigger in frustration, you see that a second villain is
preparing to shoot you from behind a narrow basement window (it is a tough
neighborhood). Unable to flee in time and also unable to fire with accuracy
through the tiny window, you can save yourself only by shooting the first
villain, causing him to slump in front of the window, thereby blocking the
second villain’s line of fire.

According to the Culpability Account’s revised criterion of liability—cul-

pability for an unjust threat or culpability for an unjust attempt—the first
villain is liable to be killed by you because he is making a culpable attempt
on your life. You know that this villain is harmless but your killing him
392 Jeff McMahan

would not be pointless, for it is necessary to prevent yourself from being

killed by the second villain. My intuition, which is not shared by everyone, is
that you would be justified in killing the first villain. Because he is culpably
attempting to pose the same threat you in fact face from another agent, it is
difficult to accept that you would wrong him by killing him in order to avert
that threat.
Perhaps another case may push us even further.

All of the details are the same as in the previous case except that the second
villain is preparing to shoot not you but another innocent person who has
appeared in the alley. You can prevent the second villain from killing this
innocent person only by killing the first villain, thereby causing him to drop
to the ground and block the window.

Again, my intuition is that you would be justified in killing the first villain.
My sense that he would not be wronged by being killed by you is slightly
weaker in this case than in the previous case, but even in this case I doubt
that he has any persuasive ground for complaint.
In short, I suspect that liability to defensive killing can extend beyond
culpability for an unjust threat. But once liability becomes detached from
threat, various questions arise. One is: ‘‘Culpability for what?’’ These cases
suggest, at least to many people, that we should accept an extended criterion
of liability to defensive killing that recognizes liability for culpable attempts.
But why stop there? It seems arbitrary to draw the line at attempts. Suppose
that in the initial case involving the second villain, the first villain follows
you into the alley with the intention of killing you but decides to check his
gun before firing. Just as the second villain appears at the window, the first
villain discovers that his gun is empty. As in the original case, the only way
you can prevent the second villain from killing you is to kill the first villain,
who not only poses no threat but also now makes no attempt to kill you—
though he would kill you if he could. My sense is that his merely discovering
that an attempt would be futile cannot make the difference between his
being liable and his not being liable.
But is it plausible to suppose that a person can be liable to be killed
simply because he would unjustly kill an innocent person if he could? And if
we accept this as a basis of liability, should we stop here or go further? What
about someone who now intends to kill an innocent person in the future?
Does it make a difference whether he is actively preparing for the killing?
(These questions are crucial to understanding the permissibility of preven-
tive defense.) What about someone who intends to kill an innocent person if
certain conditions obtain? Does it matter how likely it is that the conditions
will obtain?
Other possibilities are suggested by further variations of the case invol-
ving the threatening villain behind the basement window. What if the person
 The Basis of Moral Liability to Defensive Killing 393

who might be killed in order to block this villain’s fire is not now attempting
to harm you but made an attempt on your life earlier in the day? Or a month
ago? What if, instead, he made an attempt on someone else’s life a year ago?
What if he is not now attempting to kill anyone but is a murderer who was
never punished? Or a murderer who has been punished but remains unrepen-
tant? What if, though harmless, he is simply a thoroughly nasty person?
Clearly we must draw a line before we reach this final possibility. I find
it plausible that the person who is culpably but futilely attempting to kill
you would not, in our first three cases, be wronged if you were to kill him.
But once liability is divorced from culpability for an actual threat, I do not
know where to draw a principled line at which culpability ceases to be a
basis of liability.
There is another problem. If, as the Culpability Account originally
asserted, liability to defensive killing arises only from culpability for an
unjust threat, that implicitly limits the purpose for which the liable person
may be killed to the elimination of the threat. But if liability is divorced
from responsibility for an actual threat, it is an open question for what
purposes a person who is liable to be killed may justifiably be killed. This
problem is illustrated by the case in which the purpose of killing the first
villain is not to defend the person he is attempting to kill but to defend a
different person from a different threat. But there are other possibilities. We
might, for example, kill those who are liable for culpable attempts in order
to use their organs for transplantation. It is unclear, however, whether there
is a stable, principled line between this unappealing suggestion and the view
that it is permissible to kill the first villain to prevent an innocent bystander
from being killed by the second villain.
The cases I have considered thus far pose the question whether certain
forms of culpability are sufficient for liability. But one point on which the
Culpability Account is firm is that culpability is necessary for liability. It
thus denies that either innocent or nonresponsible threats can be liable to
defensive killing. Many people will regard this as a sufficient basis for
rejecting the account. For most people agree with Thomson that it can be
permissible to kill innocent and nonresponsible threats in self-defense.
I disagree. I think that there is no moral basis for holding nonrespon-
sible threats liable.8 And it is possible that there is no basis for holding
certain innocent threats liable. But some innocent threats are liable. In other
words, culpability is not necessary for liability. Here is an example of
liability without culpability.

A person keeps his car well maintained and always drives cautiously and alertly.
On one occasion, however, freak circumstances cause the car to go out of
control. It has veered in the direction of a pedestrian whom it will kill unless
she blows it up by using one of the explosive devices with which pedestrians in
philosophical examples are typically equipped.
394 Jeff McMahan

This conscientious driver is an example of one type of innocent threat,

which I call an ‘‘inadvertent threat.’’ The pedestrian has done nothing to
make herself liable to be killed but the driver will kill her unless he is himself
killed first. Yet he is in no way culpable. The unjust threat he poses is not
the result of wrongful intent, recklessness, or negligence; it is the result of
sheer bad luck. On what basis, then, may he be said to be liable?
What makes him liable is that, as a morally responsible agent, he
voluntarily chose to set a couple of tons of steel rolling as a means of
pursuing his ends, knowing that this would involve a tiny risk that he
would lose control of this dangerous object that he had set in motion,
thereby imperiling the lives of the innocent. It is important to notice that
the claim here is not that he is liable because he alone had the chance to
avoid the tragic conflict between himself and the pedestrian. For the pedes-
trian could have avoided it as well by simply staying at home. The claim is,
rather, that the driver is liable because he voluntarily engaged in a risk-
imposing activity and is responsible for the consequences when the risks he
imposed eventuate in harms.
Admittedly, what I am citing as the relevant difference between the
driver and the pedestrian is of comparatively slight moral significance. But
given that the costs of the driver’s voluntarily engaging in a risk-imposing
activity cannot be divided, it is fair that he should suffer them all rather than
impose them all on the pedestrian. Yet because he is not culpable for the
threat of unjust harm he poses, it might be reasonable to demand that the
pedestrian share the costs with him if that were possible. If, for example, the
pedestrian could defend her life in a way that would inflict nonlethal harm
on the driver but also allow his action to inflict a nonlethal injury on her,
she might, particularly if the harm she would suffer would not be severe, be
obliged to suffer this harm rather than kill the driver. To kill him when this
alternative is available might indeed be to wrong him. (Note that this
assumes that liability and proportionality in self-defense are interrelated.)

The Responsibility Account

The case of the conscientious driver suggests an alternative account of

the basis of liability. According to what I will call the Responsibility
Account, the criterion of liability to defensive killing is moral responsibility,
through action that lacks objective justification, for a threat of unjust harm
to others, where a harm is unjust if it is one to which the victim is not liable
and to which she has not consented. For convenience, I will sometimes
abbreviate this to ‘‘responsibility for an unjust threat.’’ The conscientious
driver satisfies this criterion because he is a morally responsible agent who
has acted voluntarily in a way that foreseeably imposed risks on others and
now threatens the pedestrian with unjust harm. And he acts without
 The Basis of Moral Liability to Defensive Killing 395

objective justification because, although he may act reasonably given his

epistemic limitations, it is not justifiable to drive a car that is in fact going to
go out of control. Yet he is in no way culpable for the threat he poses. The
epistemic constraints under which he acts are fully exculpating.
The Responsibility Account treats liability as a matter of degree.
Responsibility for an unjust threat can vary in degree, and liability varies
concomitantly. A person’s liability is therefore greater when his action is
culpable, and the degree of his liability varies with the degree of his cul-
pability. The degree of a person’s liability is manifest in the severity of what
may be done to him without wronging him—that is, in the stringency of the
proportionality requirement.
It may be instructive to compare liability to defensive action to liability in
tort law. Just as we may think of liability in torts as a matter of corrective
justice, or justice in the distribution of harm ex post, so we may think of
liability to defensive action as a matter of preventive justice, or justice in the
distribution of harm ex ante. According to the Culpability Account, liability
to defensive action corresponds to fault liability in torts. But the
Responsibility Account treats liability to defensive action as strict, as in the
case of the conscientious driver. It says to all people who engage in socially
permitted risk-imposing activities such as driving: ‘‘You know that if you
drive you impose a very small risk on other innocent people. If you choose to
drive, the consequences are your responsibility unless others also contribute
to the outcome through their own risk-imposing activities. You will be liable
to defensive action even if you satisfy the relevant standards of due care.’’
It is, perhaps, significant that there is no account of liability in torts that
corresponds to the Rights-Based Account as it is usually understood, for
example by Thomson. Suppose the fat man who is pushed off a cliff actually
lands on you but merely breaks your leg rather than killing you. There is no
conception of strict liability that is strict enough to hold him liable for
damages.
It is also true, however, that if the conscientious driver were merely to
injure the pedestrian rather than kill her, he too would escape liability in
tort law. His act would not be among those that tort law subjects to a
standard of strict liability. More generally, the range of injurious acts for
which tort law holds people strictly liable is narrower than the range of
threatening acts for which the Responsibility Account holds people liable to
defensive action. I believe, however, that tort law would correspond more
closely to our sense of corrective justice if it were to hold the conscientious
driver liable. Yet the provisions of tort law have to be formulated to take
account not only of considerations of justice but also of consequences. If
people like the conscientious driver were held strictly liable ex post, that
might help to deter people from driving for frivolous reasons, which would
be desirable, but it might also have undesirable effects as well. I cannot
pursue these issues here.9
396 Jeff McMahan

I believe that some version of the Responsibility Account will turn out
to be the best account of liability to defensive killing. Like the other
accounts, however, the Responsibility Account, at least in the form in
which I have stated it, faces a battery of objections. In the remainder of
this article I will review some of these objections and suggest some initial
responses.
Suppose that the only way you can prevent yourself from being killed
by a culpable attacker is to kill his mother. If you do kill her, can you then
claim that she was liable to be killed because, as a morally responsible agent,
she voluntarily chose to engage in an activity (having a child) that had a tiny
probability of resulting in an unjust threat and that this made her respon-
sible for the threat you faced from her son?10 Obviously not. But it is less
obvious what the right explanation is of why the mother is not liable.
One explanation is that the woman’s act of having a child is causally too
remote for her to be responsible for the threat he poses. It might be argued
that just as tort law insists that a person must have been the ‘‘proximate’’
cause of harm to another in order to be liable to repair that harm, so a
person must be the proximate cause of an unjust threat to another in order
to be liable to defensive killing. This seems, however, to be too strong a
requirement. Suppose that a wholly innocent person is about to be tortured
as a terrorist suspect under newly adopted rules that wrongfully permit
certain forms of torture during interrogation. She cannot defend herself
against the person who will actually conduct the torture but, through a
remarkable mechanism known only to those who devise philosophical
examples, she can prevent the torture by inflicting a proportionate harm
on an official who provided sophistical arguments for the legal permissi-
bility of the new rules, thereby making it possible to implement them. This
official seems responsible for the practice of torture in a way that makes him
liable to defensive action. He would not be wronged if this woman were to
inflict a proportionate harm on him as a necessary means of avoiding
torture that she would not have faced were it not for his action. Yet the
action that makes him liable is causally remote from the unjust threat the
woman faces. The torturer, not the official, is the proximate cause of that
threat. Still, it seems that the causal connections are of the right sort for the
transmission of moral responsibility. I will not try to determine what sorts
of causal connection are the right sorts. But it seems that the official’s causal
role in implementing the policy of torture is part of the cause of the
torturer’s action and that this makes him responsible to a significant degree
for what the torturer does. By contrast, the mother’s action in causing her
child to exist does not in anything like the same way cause his subsequent
acts and does not make her to any degree responsible for them.
Not only are there some who, like the official, are liable for an unjust
threat without being its proximate cause; there are also some who are the
proximate cause of an unjust threat but are not liable. This is true, for
 The Basis of Moral Liability to Defensive Killing 397

example, of the falling fat man. But it also seems true of some innocent
threats who might seem to be liable under the Responsibility Account as I
have stated it. In earlier work, I assumed that the only instances in which a
person can be altogether lacking in responsibility for harm he causes are
those in which he causes harm independently of his responsible agency. I
therefore assumed that whenever a morally responsible agent causes harm
through his own voluntary action, he is to some degree responsible for it,
even if he is fully excused for causing it because of some excusing condition
or simple bad luck.11 But I now believe that this was a mistake. Here is an
example that shows why.

A villain has secretly tampered with a man’s cell phone in such a way that if the
man presses the ‘‘send’’ button, he will detonate a bomb to which you have been
tied by this same villain. The villain has trussed you up so that you cannot
escape or alert the man with the cell phone to your plight. But the villain has
given you a weapon with which you can kill the man with the cell phone.

Suppose the cell phone operator has pressed in a number and is about to
press the send button. He now inadvertently poses an unjust threat to you—
that is, a threat to which you are in no way liable. He is, moreover, a
morally responsible agent whose pressing the button would be a free,
voluntary act. But, even though he is the proximate cause of the threat to
you, he seems in no way responsible for that threat. Consequently, he seems
no more liable than the falling fat man. It is possible that you could be
justified in killing him, but you could not plausibly claim that he had made
himself liable to be killed.
In our earlier example, the killer’s mother could reasonably have foreseen
that there was a very small probability that if she had a child he would
become a murderer. But in that case there was not the right sort of causal
connection between her having a child and that child’s later posing an unjust
threat to transmit responsibility for that threat back to her. In the example of
the cell phone operator, there is certainly the right sort of causal connection
between the man’s action and the unjust threat you face. The necessary
ingredient of responsibility that is missing in this case is foreseeability.
There is, of course, a problem about how probable a threat or harm
must be in order to count as foreseeable. I cannot address that problem
here; for present purposes I will sidestep it by simply invoking the notion of
a ‘‘risk-imposing activity’’—that is, an activity of a type that is known or can
be seen to impose a risk of significant harm on others, even if the probability
is comparatively low, as it is in the case of driving. I will assume that it is a
condition of responsibility for an unjust threat that the action that gave rise
to the threat either was of a risk-imposing type or was such that in the
circumstances the agent ought to have foreseen that it carried a non-negli-
gible risk of causing a significant unjust harm.
398 Jeff McMahan

While the conscientious driver and the cell phone operator are both the
inadvertent but proximate cause of an unjust threat, only the driver is
engaged in a risk-imposing activity. Pressing a button on a cell phone is
not the kind of act that typically carries any risk to others, however slight.
This is why the cell phone operator cannot be held responsible or therefore
liable for the harm his pressing the button would cause. (Having a child is
also not a risk-imposing activity, but this is not because it is not a type of
activity that leads ultimately to risks to others. It is because any subsequent
risks or harms cannot be traced in the right way back to the activity itself.)
In short, the cell phone operator is not liable to defensive killing
according to the Responsibility Account because he fails to meet one of
the conditions of moral responsibility—the foreseeability condition—for the
unjust threat he poses. Although he is a morally responsible agent who acts
voluntarily and without objective justification (since his pressing the send
button is objectively unjustified), he is nevertheless, like the falling fat man,
a nonresponsible threat.
I find it a virtue of the Responsibility Account that it identifies a basis
for attributing liability to the conscientious driver. Intuitively, however, it
seems to matter what his reason is for driving. On considering the example,
we naturally assume that this person is driving for one of the self-interested
reasons for which people typically drive: to go to the store, to get to work,
or whatever. But some people, such as ambulance drivers, drive for more
important reasons—reasons for which other people reasonably want them
to drive despite the risks they impose. If we imagine a variant of the case of
the conscientious driver involving an ambulance driver on the way to the
scene of an accident, we are likely to question whether his voluntarily
engaging in this risk-imposing activity can make him liable to defensive
killing by the pedestrian when the ambulance goes out of control.
It is tempting to suppose that the ambulance driver’s action is justified
in a more robust sense than the action of the conscientious person who
drives for reasons of self-interest and that this blocks an attribution to him
of liability. But objectively neither driver is justified. For in each case,
however faultless or even meritorious the driver may be, there is an objective
description of what he will do unless he is killed—killing an innocent
pedestrian—under which it is clear that his action is unjustified. (I am
assuming that getting to the accident site is not sufficiently important to
make it justifiable for the ambulance driver knowingly to kill a pedestrian if
that were necessary to get there.)
This, of course, presupposes an objective account of justification. But
even a subjective account of justification is unlikely to be able to distinguish
between the two drivers in a way that supports an attribution of liability to
the self-interested driver but denies that attribution to the ambulance driver.
For any plausible subjective account that implies that the ambulance dri-
ver’s action is justified will have the same implication for the conscientious
 The Basis of Moral Liability to Defensive Killing 399

driver’s self-interested action, given how low the antecedent subjective

probability of his car’s going out of control was.
Suppose, however, that we were to find a plausible basis for claiming
that, while the conscientious driver is liable for having chosen to engage in a
risk-imposing activity, the ambulance driver’s reason for engaging in that
same activity shields him from liability in the same circumstances, even if his
action is objectively unjustified. We would still face a line-drawing problem.
For people’s reasons for engaging in a risk-imposing activity such as driving
lie along a spectrum from culpable to morally neutral to praiseworthy. If we
claim that the ambulance driver’s reason for driving blocks an attribution of
liability to him, what should we say about a person who drives to take his
child to the doctor? Or to take his sick dog to the vet? Or to present a paper
on liability to defensive killing at a philosophy colloquium? There does not
seem to be any point at which a principled line can be drawn between liable
and not liable on the spectrum of cases between the conscientious driver and
the ambulance driver. (And even if we could find such a point, identifying it
would be largely irrelevant to matters of practice, since in cases of self-
defense, and particularly cases of defense against inadvertent threats, agents
tend to have little or no reliable information about the reasons for which
those who pose an unjust threat have acted.)
It may seem that the ambulance driver is similar to the tactical bomber
in that both unintentionally threaten the innocent as a side effect of enga-
ging in morally desirable action. According to the Responsibility Account,
however, they are quite different, for the tactical bomber acts with objective
justification while the ambulance driver does not. This fact about the
tactical bomber may, indeed, enable the Responsibility Account to deal in
an intuitively satisfactory way with the problem the tactical bomber poses.
In the case of the tactical bomber, the civilians are not liable. The
tactical bomber will therefore wrong them if he drops his bomb. Yet he
acts with objective moral justification. And, according to the Responsibility
Account, justification defeats liability: one is liable to defensive killing by
virtue of responsibility for an unjust threat only if one acts without justifica-
tion. This seems intuitively plausible; for it is hard to see how one’s moral
immunity to being killed could be compromised merely by one’s acting in a
way that is morally justified.
As I noted earlier, however, it seems intuitively that the civilians are
permitted to kill the tactical bomber in self-defense. How can that be if he is
not liable? I suggest that people are permitted a necessary and proportionate
defense against being wronged either by unjustified or justified action.
People are not morally required, or at least not always, simply to submit
to being wronged by another’s morally justified action. They are, at least
sometimes, morally permitted to resist. This is a special form of justification
for killing; it does not appeal either to liability on the part of the victim or to
the overriding moral importance of avoiding disastrous consequences.
400 Jeff McMahan

It is essential to this form of justification that it applies only to

those who would otherwise be wronged by morally justified action.
A person who acts with justification to threaten another with harm to
which the other is morally liable does not threaten to wrong that other
person. There can be no permissible defense against harm to which one is
morally liable.
Because the tactical bomber’s justified action would wrong the civilians,
they are permitted a proportionate defense. And killing him would be
proportionate. But because he is not morally liable to be killed by them,
they will wrong him if they kill him in self-defense. He too, therefore, is
permitted a proportionate defense against their permissible defensive action.
He is justified in killing them in preemptive self-defense.
This reasoning establishes the intuitive moral parity between the tactical
bomber and the civilians. Neither party to the conflict is liable to be killed
by the other, but each is justified in killing the other in self-defense.
This reasoning is also compatible with the plausible view that third
parties are not, other things being equal, permitted to intervene. Although
the fact that each party threatens to wrong the other seems to provide a
ground for third party intervention, that is canceled by the fact that each
party also acts with justification.
This understanding of the case of the tactical bomber can, it seems, also
be stated in the language of rights. The civilians have a right not to be killed.
Because the tactical bomber acts with justification, his killing them would
not violate but would infringe their rights. But people are permitted a
necessary and proportionate defense of their rights against both violation
and infringement (though perhaps what is proportionate to prevent violation
may be disproportionate as a means of preventing infringement). Hence the
civilians are permitted to kill the tactical bomber in self-defense. In doing so,
however, they would infringe his right not to be killed by them. For while it
is true that one may lose rights by threatening to violate rights, one does not
lose rights by threatening to infringe rights. For one cannot lose rights
merely by acting in a way that is morally justified. Thus, because the tactical
bomber would merely infringe the rights of the civilians by killing them, he
retains his own right not to be killed by them. It is therefore permissible for
him to defend his right against infringement by killing them in preemptive
self-defense. In short, he is justified in dropping his bomb, though that
would infringe the rights of the civilians; they are justified in killing him
to prevent him from infringing their rights; but their killing him would
infringe his rights; hence he is also justified in killing them to prevent
them from infringing his right not to be killed. Moreover, disinterested
third parties may not intervene on behalf of either party to the conflict;
for while they may, and sometimes must, prevent the violation of a right,
they may not prevent the infringement of a right, since they must not kill a
person in order to prevent him from engaging in morally justified action.
 The Basis of Moral Liability to Defensive Killing 401

Only those whose rights would be infringed may do that; others must
remain neutral.
The ease of translation here suggests the possibility of formulating a
variant of the Rights-Based Account that is equivalent, or at least exten-
sionally equivalent, to the Responsibility Account. We may, perhaps, think
of the Rights-Based Account as developed by Thomson as only one way of
filling out a formal structure that may yield a rather different theory when
filled out with substantive assumptions different from those made
by Thomson. Among these assumptions are that the conditions for the
violation or infringement of a right are the same as the conditions of
moral responsibility for an unjust harm and that it is permissible to resist
the infringement as well as the violation of one’s rights (though the propor-
tionality restriction may be more stringent in the former case than in the
latter). But while it is possible that all that I say may be translatable without
loss into the idiom of rights, I will not investigate this possibility further
here.
One worry that may arise is that the explanation I have given of why the
civilians are justified in killing the tactical bomber in self-defense also
applies to the potential victims of the falling fat man and the cell phone
operator. If the civilians may kill the tactical bomber to prevent him from
wronging them, why is it not similarly permissible for you to kill the falling
fat man and the cell phone operator in order to prevent them from wronging
you? Admittedly, there are many civilians and only one tactical bomber,
whereas the exchange of lives in the other two cases would be one for one.
But this consideration is offset by the fact that the tactical bomber, but not
the fat man or the cell phone operator, poses an unjust threat through
action that is objectively justified. In short, while the tactical bomber will
wrong his victims with justification, both the fat man and the cell phone
operator will wrong you without justification. Why may you not kill them?
The answer is that they will not in fact wrong you. Whether you are
wronged is not a matter only of what happens to you. It also depends on
facts about agency and responsibility. A falling boulder or a charging tiger
may harm you but not wrong you. Moral responsibility is a precondition of
wronging, and neither the fat man nor the cell phone operator satisfies the
conditions of moral responsibility for the unjust threat he poses. The threat
posed by the fat man is not a product of his responsible agency, while the
threat posed by the cell phone operator cannot be foreseen by him—that is,
he cannot know that he poses that threat. (Expressed in the language of
rights, the claim here is that neither the fat man nor the cell phone operator
will either violate or infringe your right not to be killed. Both violation and
infringement of rights require moral responsibility. Indeed, infringement, as
I understand it, refers to action that is morally justified. And the fat man,
for example, is not justified in falling on you. Involuntarily falling is neither
justified nor unjustified.)
402 Jeff McMahan

In short, the tactical bomber, the fat man, and the cell phone operator
all fail to meet the Responsibility Account’s conditions of liability. But the
tactical bomber does meet the account’s conditions of responsibility; thus,
his action will wrong the civilians. The reason he is released from liability is
that he acts with justification. The fat man and the cell phone operator, by
contrast, are not released from liability because they act with justification.
The reason they are not liable is that they fail to meet the conditions of
moral responsibility; and this means that though they act without justifica-
tion, they will not wrong you even if they kill you.
Because they are not liable, all three—the tactical bomber, the fat man,
and the cell phone operator—would be wronged by being killed. But at least
in the circumstances as described, only the tactical bomber may permissibly
be wronged. Acting with justification does not, it seems, afford the same
degree of immunity as lack of moral responsibility.
Recall now three other cases we have discussed. The resident in the case
of the identical twin, the conscientious driver, and the ambulance driver all
seem to meet the conditions of responsibility for the unjust threats they
pose. All will violate their victim’s right not to be killed. For none acts with
objective justification. It seems that all three are liable to defensive killing
according to the Responsibility Account. (Perhaps implausibly in the view
of most people, the one who may come closest to avoiding liability, accord-
ing to the Responsibility Account, is the resident. For we can imagine that
story in ways in which it would be virtually impossible for him to know, in
the time in which it is reasonable for him to act, that the threat he poses is
unjust. Even so, we may think it a relevant difference between him and the
cell phone operator that he at least knows that he poses a threat, and thus is
responsible for ensuring that it is not an unjust threat.)
It is, perhaps, tempting to think that these last three cases should be
treated in the way I have suggested that we treat the case of the tactical
bomber. Consider, for example, the conscientious driver. I have argued that
he is liable by virtue of having chosen to engage in a risk-imposing activity.
But this, it might be argued, is too flimsy a basis for a claim of liability. For
many of the normal activities of life are risk-imposing, and it would be
absurd to assign liability to people for simply living their lives in normal
ways. Even just going out in public imposes risks of serious harm on others.
One may, for example, be an asymptomatic carrier of a flu virus, which
might prove fatal to a very young or very elderly person whom one might
expose to it. (This, unlike the possibility of one’s cell phone being secretly
transformed into a detonator, is a possibility of which one is, or ought to be,
aware.)
This suggestion is likely to be urged by a proponent of the Culpability
Account, who claims that fault is necessary for liability. To be liable, on this
view, one must depart in certain specific ways from the norms of ordinary
social life. The defender of this view could argue, moreover, that the denial
 The Basis of Moral Liability to Defensive Killing 403

that the conscientious driver is liable does not entail that the pedestrian
must allow herself to be killed—that is, that she may not kill in self-defense.
For my own discussion of the tactical bomber concedes that there are
justifications for killing that appeal to neither the victim’s liability nor the
necessity of averting a catastrophe. Why cannot we say that the pedestrian
may kill the conscientious driver in self-defense simply because the con-
scientious driver will otherwise wrong her, or violate her rights, and that the
conscientious driver, being free of liability because he is not culpable, is
likewise permitted to kill the pedestrian in preemptive self-defense?
The problem is that this sort of resolution does not seem to be available
in the conflict between the conscientious driver and the pedestrian. The case
of the tactical bomber is special: what absolves him of liability is that he acts
with objective justification. But this is not true of the conscientious driver.
Even if he acts reasonably, he has no positive moral justification for doing
what will kill the pedestrian. Since he will therefore violate the pedestrian’s
right not to be killed, it is unclear how he could be free of liability. Is it
because he fails to meet the conditions of moral responsibility? No, because
he clearly does meet those conditions. There is, moreover, a relevant differ-
ence between him and the pedestrian, which is that he chose to act in a way
that exposed the pedestrian to a risk of serious harm whereas the pedestrian
imposes no comparable risk on him. I believe, therefore, that it would be
unjust if the conscientious driver, having chosen to do what exposed the
pedestrian to risk, were then to kill her preemptively to prevent her from
defending herself from the harm that he would otherwise, and without
justification, cause her.
I continue to believe, therefore, that the Responsibility Account’s impli-
cations for the case of the conscientious driver are more plausible than those
of the Culpability Account. I do not, however, point to this in triumph, for
this is not the kind of case against which we can usefully test a theory’s
implications. It is a case about which most of us have only weak or doubtful
intuitions, and thus is precisely the sort of case for which we need guidance
from a theory.
In criticizing the Culpability Account, I suggested that it seems plausi-
ble to regard those engaged in culpable attempts as liable to defensive
action, even though they would otherwise cause no harm. If this is an
objection to the Culpability Account, it is equally an objection to the
Responsibility Account, which also insists on causation as a condition of
liability. (The problem of attempts arises only in cases in which the
Responsibility Account coincides with the Culpability Account—namely,
cases involving culpable responsibility for an attempt to cause an unjust harm.)
In my earlier discussion of culpable attempts, I left the line-drawing
problem unresolved. I will conclude by offering a brief argument in favor of
retaining causation as a condition of liability. The argument appeals, not to
attempts, but to a variant of the case of the conscientious driver. Suppose
404 Jeff McMahan

that when the driver’s car goes out of control, it threatens another driver
rather than a pedestrian. In this case, the imposition of risk has been
reciprocal. Both drivers made the same choice to expose others to risk.
But only one, through moral bad luck, now threatens unjust harm to the
other. Most people are willing to allow that moral luck may affect liability
and thus believe that only the driver who poses an unjust threat is liable.
Most people, I think, would agree that the driver who is threatened would
be justified in acting in self-defense but that the threatening driver would
not be justified in acting to preempt the other’s defensive action. If that is
right, the mere imposition of risk—or, in the case of attempts, mere cul-
pability—is not sufficient for liability in the absence of causation. Causal
responsibility for an unjust threat is a necessary condition of liability to
defensive killing.12

Notes

1. This view is developed and defended in Judith Jarvis Thomson’s important

article, ‘‘Self-Defense,’’ Philosophy and Public Affairs 20 (1991): 283–310.
Thomson uses the term ‘‘innocent aggressor’’ to refer to what I call an innocent
threat. What she refers to as an innocent threat is what I call a nonresponsible
threat.
2. Thomson, p. 287. In the story as she tells it, you can use your awning to deflect
him past the edge of the deck onto the road below. Since there are accounts of
the distinction between killing and letting die that would classify your deflecting
him as allowing him to die rather than killing him, I have altered the example in
a way that makes your act an uncontroversial instance of killing.
3. Judith Jarvis Thomson, The Realm of Rights (Cambridge, MA: Harvard
University Press, 1990), p. 77.
4. I draw this distinction differently from the way Thomson does. Ibid., p. 122.
5. One question I will not pursue is whether a person may be liable for being
culpably responsible for a threat that is not unjust because the victim is liable to
that threat. There are two relevant types of case. In one, a case of ‘‘unknown
justification,’’ an agent poses a threat to which the potential victim is in fact
liable though the agent acts on a wrongful reason in ignorance of the reasons
that could justify his action. In the other, a case of ‘‘ostensible justification,’’ an
agent poses a threat to which the potential victim is liable but does so for a
wrongful reason on the pretense of acting on the basis of the justifying reason. In
these cases, it is controversial not only whether the agent is justified but also
whether the potential victim has a right of self-defense.
6. Some parts of this section draw on my ‘‘Self-Defense and Culpability,’’ forth-
coming in Law and Philosophy. That paper comments on an account of the right
of self-defense advanced by Kimberly Kessler Ferzan in ‘‘Justifying Self-
Defense,’’ which is also forthcoming in Law and Philosophy and from which I
have learned a great deal.
7. This is the view defended by Ferzan in ‘‘Justifying Self-Defense.’’.
 The Basis of Moral Liability to Defensive Killing 405

8. I argue for this in The Ethics of Killing: Problems at the Margins of Life (New
York: Oxford University Press, 2002), pp. 398–411.
9. For relevant discussion, see Steven Shavell, ‘‘Strict Liability versus Negligence,’’
Journal of Legal Studies 9 (1980): 1–25; and Douglas Husak, ‘‘Vehicles and
Crashes: Why is this Moral Issue Overlooked?,’’ Social Theory and Practice 30
(2004): 351–70.
10. I owe this objection to Kai Draper.
11. See The Ethics of Killing, pp. 401–5.
12. For perceptive written comments on an earlier draft, I am grateful to Shawn
Bayern, Ruth Chang, Kimberly Kessler Ferzan, Peter Vallentyne, and Alec
Walen. I am also indebted to Judith Jarvis Thomson for comments on an earlier
manuscript from which this article draws. For exceptionally stimulating discus-
sion, I am grateful to Larry Alexander, Allen Buchanan, Agnieszka Jaworska,
Sanford Kadish, Shelly Kagan, Frances Kamm, Christopher Kutz, William
McMahan, Eric Rakowski, Larry Temkin, Eric Wampler, and, especially,
Samuel Scheffler.