Double-effect and comparing intentions
There are two main ways one might seek to argue that Finnis’s further constraint on killing does not lead to the impermissibility of practically all abortion if the fetus is a person. One is to show that many or most abortions involve only the indirect intention of fetal death. The other is to try to collapse the moral distinction Finnis draws between directly and indirectly intended killing in abortion.
With regard to the first strategy, the violinist scenario can be brought into play again. Finnis sees it as a relevant point of disanalogy between unplugging the violinist and almost all abortion that whereas the violinist’s death is only indirectly intended, abortion aims directly at fetal death. When you unplug yourself from the violinist, you foresee that his death is inevitable, but you do not will it or aim at it as part of your enterprise. If he happened to survive, your purpose would not have been frustrated. In contrast, abortion often does seem to aim at fetal death either as an end in itself or as a necessary means of detaching the fetus from the woman, not merely a by-product of doing so.
It is important to note at this point that the violinist scenario will mirror most abortions in failing to meet the strict proportionality requirements for necessity killing. Only if both violinist and host are doomed to die imminently in any event, and only if unplugging the violinist can save the host but nothing can be done to save the violinist, will Thomson’s scenario be a candidate for a ‘lesser of two evils’ justification for homicide. So let us amend the violinist case to assume that these conditions are met. The scenario therefore reflects the therapeutic abortion where nothing can be done to save the fetus. We can reframe Finnis’s argument as stating that when the proportionality condition is equally met in both cases, the difference between directly and indirectly intentional killing would still show us that unplugging the violinist is permissible whereas therapeutic abortion is not.
One apparent objection to this is that in therapeutic extraction abortions at least, the fetus’s death is no more directly chosen than is the violinist’s death in Thomson’s analogy. In neither case is the victim attacked directly, we may say, but rather just separated from a source of life-support. In both cases, therefore, the death of the victim might be understood as a foreseen consequence but not an aim of that separation. Should either the fetus or the violinist live, the aim, which is to discontinue bodily support, would still be achieved. Finally, in neither case is the death of the victim even an interim aim of the separation process: neither fetus nor violinist needs to be killed in order to be detached. Thomson consequently argues that that death is no more intended in an extraction abortion than when unplugging the violinist.
With regard to most abortions, however, whichever method is used, the death of the fetus is specifically intended by the pregnant woman and her doctors. Indeed, I think it fairly transparent that the main intention most women have in procuring abortions is not to end their pregnancy by extracting the fetus, but to avoid motherhood or the emotional turmoil of adoption by ending the life of the fetus through the abortion. This is a serious problem for the claim that the death of the fetus is no more directly intended in extraction abortion than in the violinist’s unplugging.
David Boonin responds to this objection by claiming that if an abortion is otherwise justified, say, to spare the woman the physical burdens of pregnancy, that justification will not be invalidated merely because those performing it also happen to desire the fetus’s death. Suppose, he imagines, that you wanted to unplug the violinist and win back your freedom, thus killing him intentionally, and are permitted to do so. However, suppose that the violinist also happened to be your chief rival in a symphony, and that you independently wished him dead. Boonin argues that so long as one acts on a ‘permissible’ intention (to unplug oneself), an additional, ‘impermissible’ intention (to kill the violinist) will not render the conduct impermissible. Just so with abortion and fetal death.
If Boonin’s answer is sound, it will still, at this point, only act as a defence of abortion by extraction where the other necessity conditions obtain: where the abortion is carried out to save the woman and fetal death is proportionate (it will save more life than if nothing is done). Consequently, it can be argued that equating the violinist’s unplugging with extraction abortion in terms of the directness of the intention to kill does not provide a defence for killing the fetus where it could be safely extracted alive in the alternative. Were the fetus capable of being extracted without being killed, there would be no way to claim, by analogy with the violinist, that it is equally permissible to abort in a way that ensures its death. This suggests that the necessity doctrine can therefore only offer a defence to the therapeutic extraction of the fetus.
Thomson is in fact willing to grant that a pregnant woman has no right to demand fetal death in abortion as an end in itself where a viable fetus can be safely extracted alive.49 However, she nevertheless disputes Finnis’s central claim that abortions which directly aim at fetal death as a means or an end are prohibited in every eventuality. In effect, she denies that a hysterectomy or a craniotomy for the same reason really can be distinguished solely in terms of the directness of the intention to kill. Along similar lines, Boonin contends that directly intended killing is plainly permissible in the same circumstances where it is permissible to kill as a side- effect if choosing the direct form of killing involves less risk of harm to the agent or the beneficiary. If the craniotomy is safer for the woman than the abortifacient medicine, and the outcome is the same for the fetus, surely it is unreasonable to prohibit that procedure specifically. Boonin’s argument is, in short, that killing directly rather than indirectly (through extraction) does not manifest greater disrespect for the value of the fetus’s life if it avoids a more ‘dangerous and invasive’ procedure for the woman. He explains:
[I] t is extremely difficult to see how causing a person to die by one means, rather than causing him to die by another, in cases where it is permissible to cause death by the first means and the cause of death makes no difference to him, is using him as a means, or failing to respect him as an end.so
Boonin’s denial of the moral difference between killing directly and indirectly where it makes no difference to the victim might cut against our moral reactions to amended violinist scenarios. If unplugging the violinist could only safely be performed by first ensuring that he is dead, could we be so confident that a third party is justified in killing the violinist in order to detach him from you? Would the Court of Appeal in Re A have consented to separating Mary from Jodie if this could only be achieved by first poisoning, stabbing, or suffocating her? I think it doubtful that the Court would have consented to such a thing even if it were the only means of detaching Mary, let alone only a safer means for Jodie.
But the deeper problem with Boonin’s argument here bears some similarity to the pitfalls of his argument (considered in the previous chapter) that the difference between killing and letting die is of no moral consequence where there is justification enough for letting die, but where killing is safer for the agent. Boonin expressly grants that there is a substantial moral distinction between indirectly and directly intended killings.51 He only seeks to show that the distinction does not affect the permissibility of killing directly where indirect killing would be permissible, but directly intended killing has some distinct advantage for the agent or beneficiary. Yet again, however, one might answer Boonin by pointing out that if there is such a moral distinction, as he grants, it will only ever make a difference when everything else is equal but when killing directly has some added advantage. If the moral distinction does not affect one’s permissions in these cases—when all else is equal but killing directly is better for the agent—then it surely does not exist at all. It simply follows from there being a moral distinction between directly and indirectly intended killing that the difference must matter when it is the only difference there is.
Rather than denying that the indirect—direct distinction matters for morally permissible killing when all other things are equal, a defender of Thomson might instead draw on the DDE to question whether therapeutic abortion by direct attack on the fetus really intends fetal death any more or less directly than extraction abortion by hysterectomy. Finnis claims that when a malignant womb is removed to save the life of a pregnant woman, DDE can be deployed to show that the overriding intention of the operation is that of saving life, not killing, notwithstanding the fact that the fetus’s death is inevitable. But for his conclusions to be correct, we must be confident that the same cannot be said of therapeutic abortion by craniotomy. Can a doctor who performs an abortion by craniotomy not say, in the end, that the fetus’s death was, here too, neither a means nor an end for her—not an end because it was not directly desired, and not a means because the fetus’s death is not logically necessary for ending the pregnancy, the true aim, but only practically inevitable? If the fetus happened to survive its skull being crushed, the purpose of the abortion, to end the pregnancy, could still be achieved. We can also ask whether the fetus’s death in this scenario is really any more aimed at than the violinists death in Thomson’s case.
An important part of Finnis’s direct intention test looks to whether there is logical room to hope that a victim of killing will live by asking if ‘the same action would have been chosen if the victim had not been present’. 52 This is true of removing the cancerous womb, but not of the therapeutic craniotomy, or of any extraction abortion other than the hysterectomy. Where a hysterotomy (removal of a pre-viable fetus by caesarean section) or a medical abortion (inducement of miscarriage using drugs) is concerned, it cannot be said that the same action would have been chosen completely apart from the fetus’s presence, even if the reason for the abortion is that the pregnancy poses a risk to the pregnant woman’s life. But as Thomson points out, this condition does not hold for the violinist either: his presence is a sine qua non of your decision to unplug. Thus the following question is raised. If the death of the ‘victim’ is logically severable from the agents’ intentions in the violinist scenario, why is the same not true of all life-saving therapeutic abortions by extraction, which do not directly attack the body of the fetus, and 
5i ibid 213.
whose purpose is not to kill the fetus but only to end the bodily imposition of the fetus upon the woman?
Interestingly, Finnis’s condition that the fatal action would be performed even without the presence of the victim does not hold true for the conjoined twins’ separation either. That operation would not have been performed if it were not for Mary’s existence and situation. In Re A, the Court of Appeal in fact considered whether the DDE could be invoked to reach the conclusion that the doctors who performed the operation would not truly intend Mary’s death (and would not be guilty of murder for that reason). Two out of the three judges, Ward LJ and Brooke LJ, thought that it could not beTh Although the operation would save Jodie, it would not benefit Mary in any way, and it was their view that DDE could not apply where the ‘overriding’ benevolent intention concerned one person and the foresight of death another. The judges were surely right in making this determination about DDE. Re A could be contrasted with the typical DDE case of the death-hastening pain relief, where both effects, good and bad, are experienced by the same person. Surely, the doctrine could not similarly be relied upon to displace the intention to kill if, for instance, the patient could be relieved of his pain only by an action which also speeds up the death of another patient in the bed next to him.
Working from the Court of Appeal’s analysis of DDE, it might seem that Finnis is in error to think that the doctrine even applies to the removal of the malignant womb. Just like the conjoined twins’ separation, the hysterectomy in this case does not aim at the victim’s death as a means or an end. But similarly also, the ‘good’ intention (saving life) is not directed at the victim, the fetus, which, I suggested, it must be. If this second condition on DDE is correct, then it will not be possible to distinguish the intention in a life-saving therapeutic hysterectomy from that of every other kind of extraction abortion by using DDE.
In spite of the Court of Appeal’s refusal in Re A to apply DDE to the twins case, however, the causal structure of Mary’s death still seemed to matter a great deal. The majority of judges believed that the doctors performing the separation would indeed intend Mary’s death. But there was greater ambivalence about whether the separation procedure would constitute a direct attack on Mary. They agreed that, by separating the two, the doctors would be committing an assault on the weaker twin. However, both Brooke LJ and Walker LJ proposed that the separation would also confer on Mary a certain kind of bodily integrity that she had been denied through being joined to her sister, restoring her, in a way, to some natural state that she ought to have been in—a state of individual embodiment.54 Furthermore, while the operation certainly interfered with Mary’s body, it might be argued that it did not attack her in any way certain to bring about death if not for her unusual situation of dependence. Notwithstanding their rejection of DDE, the judges seemed influenced by this idea not only that Mary’s death was not the aim of the operation, nor an intermediate aim of the procedure, but more than this, that it would not 
54 ibid 1052; 1066-7.
be the result of a kind of attack on her bodily integrity that would have killed her completely apart from her strange state of dependence.
Like the violinist’s death and the therapeutic hysterectomy, it can be argued that the causal structure of Mary’s death was consistent with it being only indirectly intended, even though the separation procedure would not have taken place but for her existence (a part of Finnis’s indirect intention test). Doubtless, the Court would not have authorized the procedure if it had required a truly direct attack on Mary’s body either prior to or as part of the separation—the sort of interference that no normal baby could survive.
But the therapeutic craniotomy abortion is different from unplugging the violinist, from the therapeutic hysterectomy, and from the conjoined twins’ separation in some meaningful respects. Like the latter three scenarios, it might be argued that its true purpose is to preserve the life of another, the fetus’s death being only a regrettable yet unavoidable consequence. But where the fetus is directly attacked, its death is not merely a consequence of a procedure with a different aim; it is logically inseparable from the nature of the act performed. A big part of why it seems that to aim to crush another person’s skull is to aim directly at his death, no matter what ulterior purposes lie behind the act, is that there is no counter-factual in which that act might not amount to a choice to end his life. More like Dudley and Stephen’s act of cannibalism, and less like Re A, one could argue there is no logical room to wish that the fetus could survive, but only to wish something incoherent: that crushing skulls did not cause death, or that people did not have to be killed to be eaten.
I think Finnis was correct to say, in his response to Thomson, that these counterfactual thoughts ‘remove morally relevant intention too far from common sense intention’.55 By the same analysis, even the sadist who kills for pure satisfaction could argue that she does not directly intend her victims’ deaths since she would be just as satisfied if they came back to life without her knowledge. We cannot say that a killing is not truly ‘aimed at’ only because one can imagine different laws of biology according to which death is not an inevitable consequence of, say, a point blank shooting, a stab to the heart, or holding someone underwater for ten minutes.
It is true that, at points, the distinction can seem extremely fine-grained. Is the death of a fetus in a therapeutic, life-saving vacuum aspiration abortion truly more aimed at than in the life-saving hysterectomy? Again though, Finnis might retort that this is simply the upshot of it ever mattering how a killing is performed, and with what kind of intention, and that if Thomson and Foot wish to dispute his conclusion on the ground that the difference is negligible in terms of outcome, this may call into question their broader acceptance that there are some acts which cannot be given results-based justification.
Still, Foot’s ‘man in the cave’ scenario may give us reason to hesitate before concluding that directly intentional killing is never permissible. Can the trapped people really not blow up the large man to save many lives? One might think that what one is permitted to do when there are many more lives at stake is a different question from what one may do to save only one. Foot’s own condemnation of the simple therapeutic abortion (where the fetus can live if the woman were left to die), given the personhood proviso, was simply that one may not choose to inflict death on one rather than allowing another to die, since the amount of life preserved is equal in both scenarios, and killing is worse than not saving. But Foot still believed that one may inflict death directly in order to maximize life saved in the cave scenario, and in the therapeutic craniotomy where the fetus cannot possibly be saved. This is the only point on which she and Finnis differ. Even, then, if one were to side with Foot and reject the absolute constraint on directly intended killing, the circumstances for permissible abortion if the fetus is a person are still reduced to one: where the fetus cannot be saved at all, but aborting the fetus can save the woman.
There is one remaining problem for Finnis, however, if one believes both that unplugging the violinist kills him only indirectly, and that in the scenario where he is hooked up to your kidneys, you are permitted to unplug. This is that abortion by extraction to spare a woman the burdens of pregnancy seems to be no more direct a killing than unplugging the violinist. In a hysterotomy or a medical abortion using drugs, the fetus is not directly attacked. Unlike the removal of the malignant womb, these abortions do not meet Finnis’s condition that the action would be chosen even without the victim’s presence, but neither does unplugging the violinist. This is a serious problem for Finnis if he wishes to maintain that my amended violinist case (where both parties will die if nothing is done) and the therapeutic extraction abortion, say, by inducing miscarriage, can be distinguished in terms of the directness of the intention to kill. As Thomson argues, if unplugging the violinist is only an indirect killing, then a woman’s taking medication to bring on miscarriage is surely as indirect. Consequently, it could be argued that all extraction abortion should amount to indirectly intended killing if this holds for unplugging the violinist.
This difficulty may well indicate that analysing the act of unplugging the violinist as the failure to save rather than as an indirect killing is more integral to Finnis’s position than might first appear. Ultimately, Finnis still maintains that unplugging the violinist is only, in any event, the denial of life-sustaining aid, whereas almost all abortion, including by extraction methods, positively kills the fetus.56 This ground of distinction was the subject of the last chapter and the core issue for the Good Samaritan Thesis (GST). If Finnis is right that, in the final analysis, unplugging the violinist is still only a failure to save, whereas ejecting the pre-viable fetus is killing it, he may be able to maintain that unplugging the violinist is permissible to avoid the sorts of burdens that would not justify extracting the fetus. However, when the comparison is between extraction abortion and unplugging, he cannot rely on the causal structure of death to argue that the former is killing and the latter failing to save. The causal structure is the same in both, and neither entails a more directly intended death than the other. Finnis would have to find some other reason, therefore, for claiming that only the abortion is an act of killing, if he wishes to defend 
the general impermissibility of extraction abortion while granting that you or others may unplug the violinist.