We begin our reply to Huddle’s response to our critique of his original paper by reminding readers that this exchange should be considered a debate among intellectual allies. We share with Huddle the view that the distinction between killing and allowing-to-die (K/ATD) is logically valid and ethically meaningful. We differ, however, in how to parse and defend that distinction, which leads to significant differences in judgments about marginal cases. But neither Huddle nor we dismiss the distinction altogether. We and Huddle agree that there is an extremely important moral difference between discontinuing life support and injecting a poison. Common sense and the law both readily grasp this distinction, but providing a philosophical understanding of this moral intuition is complicated. Our explication of the difference would allow the discontinuation of a somewhat broader range of medical interventions than would Huddle’s. Our ethical conclusion is a consequence of a clearer understanding of the philosophy of mind, notions of causation, language, and philosophy of medicine at play in making sense of the distinction.

An overriding philosophical issue at the core of our disagreement with Huddle concerns the relationships governing ordinary language, moral intuition, and ethical analysis. While these three usually align, they can sometimes diverge, and philosophers differ in how they treat such conflicts. Like Huddle, we accept the intuition that killing and allowing-to-die are different, yet we recognize that ordinary language introduces a wrinkle. ‘Killing’ tends to be used as a moral term in ordinary language, referring to cases in which someone dies as a result of another person’s actions, in a manner that is deemed morally wrong. Yet, as Rachels pointed out decades ago, there are cases that ordinary language would deem ‘allowing-to-die’ that are also morally wrong.^{Footnote 1} That led Rachels, and many other moral philosophers, to the conclusion that the distinction is confused. Huddle, by contrast, tries to save the intuition that the K/ATD distinction is ethically significant while clinging to the ordinary language meaning of ‘killing’ as a term encompassing all those acts that lead to death and are morally wrong. We do not believe that it is possible to hold both. Huddle’s attempt to do so leads to ontological, linguistic, and ethical confusion. Our solution is to uphold the ethical significance of the K/ATD distinction while correcting the imprecision of ordinary language. One of our crucial moves is to argue that, to be consistent, ‘killing’ ought to refer only to one class of morally-wrong acts that lead to death. All medical acts that can be classified as killing, as we define it, are morally wrong.^{Footnote 2} Yet some acts of ATD are also morally wrong, even if ATD is very often justifiable in clinical settings, especially at the end of life. ‘Killing,’ on this view, does not exhaust the class of morally wrong acts that result in death. Our contention is that only by accepting this correction of ordinary language can one properly align moral intuition, ethical analysis, and speech. That Huddle will not even grant this minor correction for the sake of argument, nor explain why one might oppose this correction, lies at the heart of most of his objections to our view.

We will consider the four questions Huddle presented in his response to our critique of his paper and explain why these points do not challenge Sulmasy’s account of the ethical justifiability of at least some cases of cardiac implantable electrical device (CIED) deactivation.

First, Huddle contends that our “doing/allowing” distinction is exclusively a matter of causal relationships. In so doing, he misunderstands our version of the K/ATD distinction. One must first take note of the slippage in his language. Our formulation of the K/ATD distinction is not a “doing/allowing” distinction. We do not use that language, and Sulmasy explicitly rejects this formulation.^{Footnote 3} Killing is defined as “…an act in which an agent creates a new, lethal pathophysiological state with the specific intention in acting of thereby causing a person’s death.”^{Footnote 4} ATD is defined as “…an act in which an agent either performs an action to remove an intervention that forestalls or ameliorates a preexisting fatal condition or refrains from action that would forestall or ameliorate a preexisting fatal condition, either with the specific intention in acting that this person should die by way of that act or not so intending.”^{Footnote 5} Philosophically, human acts are taken to include both motor actions and refrainings from motor action. All killings are motor actions. Yet some ATDs are also motor actions (such as turning off a switch), even though many ATDs are refrainings. Huddle is just wrong to call our formulation a doing / allowing distinction. Further, Huddle maintains that Sulmasy’s K/ATD distinction involves a separate consideration of causal relationships and factors such as intention. This is also inaccurate. Simply put, intentional action is causal. When an action occurs intentionally, the agent’s intention helps to explain how the resulting event came about. This is not a “conceptually separate consideration.”^{Footnote 6} The determination of K versus ATD requires consideration of intention as integral. This is why the determination of “intention” is included in the definitions of K and ATD that Sulmasy provides.

By contrast, Huddle’s definition of killing neglects the role of intention altogether, thereby obscuring the most ethically significant aspect of the act: “If a death-causing act intervenes upon homeostasis, it is killing,” he writes.^{Footnote 7} To claim that intention is a secondary consideration or is not innately causal is neither representative of Sulmasy’s view nor helpful in differentiating K and ATD.

To conclude our reply to Huddle’s first question, we maintain that his use of McMahan’s Burning Building II case is irrelevant to deactivating CIEDs. It is impossible to engage Huddle about the fire rescue analogy if he cannot agree to accept, at least for the purposes of argument, our definition of killing and our view that while all killing is wrong, some ATD is also wrong. For example, Huddle writes that we (Sulmasy and Courtois) “agree, after all, that removal of the net ‘without good reason’ would be ‘an unjustified instance of allowing-to-die’ (I think most would call it killing).”^{Footnote 8} Here one sees a clear example of his insistence on clinging to the use of ‘killing’ as an ordinary language judgment that the act is wrong. Moreover, Burning Building II does not even provide a good analogy for his definition of killing. Putting a net under someone about to jump from a building is hardly a “completed” (or even an initiated) saving of a life; it is not interference with a homeostatic state; it is not even an “ongoing” intervention. It is an intervention that has been intended provisionally, but the life-saving intervention has not yet taken place. Moving a net is not analogous to discontinuing a ventilator. In media res, during a fire or other disaster, intentions change and the potentially life-saving intervention that one intended provisionally to use for one victim may be redirected to maximize the number of lives saved. A firefighter might move her net from under one person to place it under two persons. Similarly, a physician running with an oxygen tank toward a lone trapped disaster victim might spy two disaster victims trapped together 20 yards away. She might then change direction and supply oxygen to the two from the one tank. The underlying “good reasons” of either the firefighter or the physician fit into the same ethical framework and answer to the same guiding principles of benevolence, which remains consistent across practices. This is why the Burning Building II example is analogous to triage, not treatment, and certainly not to deactivating a CIED.

A second question posed by Huddle concerns his view that a distinction between ongoing (O) and completed (C) therapies better describes whether the discontinuation of the therapy is a case of killing rather than a case of ATD than our distinction between replacement (R) and substitutive (S) therapies. Huddle seeks a bright line where none can be drawn. He calls our R/S distinction “arbitrary.”^{Footnote 9} Yet we would argue that our distinction between treatments—that substitute (S) for a disordered function or organ and those that have replaced it (R)—is more faithful to the ambiguities of clinical reality. There is no escaping judgment. The fact of twilight does not mean that there is no difference between night and day. No single one of the considerations we give in distinguishing R from S is decisive. In particular, “organic unity” is but one of many features (such as need for external monitoring and control, need for nonphysiological energy supply, etc.) that must be considered in making the distinction. Sulmasy’s distinction remains the more serviceable and compelling, and it is capable of evaluating future developments in medical science. Advances in technology may progress to the point that implanted electronic devices are so physically and physiologically integrated into the body that they are best described as replacement parts of the body rather than as therapies that substitute for pathological deficiencies. When we arrive at the judgment that an implanted device has become a replacement, then we would agree that discontinuing that device’s normal function would be an act of killing. We do not believe that a pacemaker fits the bill.

Moreover, almost all life-sustaining interventions contribute to homeostasis, whether completed or ongoing. Huddle’s reply to us is not clear on this point, but if he is changing his criteria to suggest that interference with a contribution to homeostasis suffices to render the judgment that stopping the treatment is killing, then the K/ATD distinction collapses. There would be no ATD because stopping any treatment would be killing. As a defender of the distinction, this cannot be Huddle’s aim. Yet it would follow from accepting contribution to homeostasis as criterion to mandate the continuation of therapy.

Huddle seems to suggest that we rigged our analysis in order to achieve a particular outcome, stating that, “A possible motivation” for our formulation of the K/ATD distinction was to allow a broader class of patients to avoid “discomfort and suffering” through the discontinuation of life-sustaining interventions.^{Footnote 10} This is not only uncharitable; it is false. Sulmasy first presented this analysis in 1995,^{Footnote 11} well before left ventricular assist devices and implantable cardioverter-defibrillators were widely used and certainly before there was a debate about the ethics of their deactivation. The theory is driving the result, not the other way around. Moreover, one way to test the adequacy of a theory is to see what unanticipated cases it can address (i.e. consilience). The fact that Sulmasy’s theory can be applied to these newer cases with such sensible results further supports its validity.

The third question that Huddle poses is whether his theory implies, as we suggest, that interference with any “completed” treatment is always killing. He now contends that in some cases completed treatments can be discontinued justifiably, by invoking the “doctrine of double effect.” For example, he suggests that an infected ventriculo-peritoneal (V-P) shunt could be removed and not be considered killing—even though it is a “completed” therapy—because the intention would be to remove the source of infection, not to stop the completed therapy, which would merely be foreseen but not intended. Huddle’s attempt to shore up his theory this way is not only ad hoc, it rests upon a misunderstanding of double effect, which was never historically used to justify the foregoing of life-sustaining treatments.^{Footnote 12} First, one cannot distinguish what one intends from what one foresees by simply giving the same event two differing definite descriptions. The removal of the infection and the removal of the completed therapy are the same event. There is no “double” effect here. Second, although it is true that a distinction between the foreseen death and the intended discontinuation of the intervention is at play, the intention/foresight distinction is only one of several conditions that must be met for the correct application of the rule of double effect. The role of intention is neither sufficient for double effect nor limited to double effect. Were one to attempt to apply the rule of double effect to these cases correctly, it would quickly become apparent that one condition is not met: one cannot plausibly claim not to intend a bad effect that is the very means by which one attains the good effect of the action.^{Footnote 13} In most cases of withdrawing life-sustaining treatments, the relief of suffering comes about by means of the death one claims not to intend, which violates the rule of double effect. Therefore, it seems that Huddle cannot appeal to the rule of double effect to justify the forgoing of completed therapies. Moreover, astonishingly, Huddle seems to be accepting that there are cases in which a completed therapy, such as the implantation of a V-P shunt, can be discontinued.^{Footnote 14} This is not merely a “complication” for his thesis; it completely undermines his argument that all such instances are killings that ought not be permitted. Accordingly, Huddle’s analysis runs aground because by his own analysis, he would be killing patients, an act he considers immoral. By contrast, Sulmasy’s approach would not lead to classifying the discontinuation of therapies that Huddle considers “completed” (such as pacemakers) as killings but as potentially permissible instances of ATD. Huddle’s analysis cannot save the K/ATD distinction. Our analysis does.

The example Huddle provides of the transplant team that withdraws immunosuppression^{Footnote 15} in fact bolsters Sulmasy’s analysis because the transplanted organ itself is not removed by the team. This is because the transplant is a “replacement,” as Sulmasy describes the term. To extirpate a transplanted organ without a plan to replace it would be killing. Immunosuppression, by contrast, is substitutive. When it becomes more burdensome than beneficial, immunosuppression can be discontinued.

Finally, Huddle’s fourth question concerns the implications of Sulmasy’s account for medical practice. As stated above, we did not rig our analysis to fit a preconceived utilitarian outcome. Rather, we have shown that the outcome of a clear analysis can be helpful to patients. Huddle, we have argued, uses ‘killing’ as a synonym for any intentional action that both results in death and is morally wrong. He seems disinclined, for unspecified reasons, to alter this usage. We contend that a slight adjustment in language and logic, so that the class of morally wrong acts resulting in death includes both killings and illicit instances of ATD (as we have defined these terms) allows one to defend the K/ATD distinction. That simple adjustment, making ordinary language consistent, is extraordinarily helpful in defending what people mean when they say there is a distinction between K and ATD. If he refuses to grant us this linguistic clarification, which we specify immediately in our definitions, then Huddle will continue to talk past us. His attempt to maintain this usage for the word ‘killing’ makes it impossible for him to present a coherent account of the K/ATD distinction, and it leads him to trap too many patients in the jaws of therapies from which they ought to be able to free themselves. A day may come in which a piece of electronic hardware implanted in a patient becomes a full replacement for whatever organ or function the patient has lost, becoming so truly a part of the patient that its deactivation would constitute an act of killing. In the meantime, the deactivation of left ventricular assist devices, implantable cardioverter defibrillators, and pacemakers, as substitutive therapies, can be viewed, in the proper circumstances, as perfectly morally appropriate instances of ATD.