FEINBERG ON VOLUNTARINESS

Let me briefly set out Feinberg's basic view of apotemnophilia and then raise some questions about it. We have to do a bit of extrapolating, because such cases represent something unusual. They are unusual not only statistically or in that they represent aberrant psychologies but also in that Feinberg, a master taxonomist, failed to find a place for them in his account of self-mayhem. He divides these cases into three kinds: those done for fraudulent purposes (say, to cheat an insurance company), which obviously fall under the harm principle; those done by madmen whose basic irrationality undermines the voluntariness of their actions; and those of religious fanatics seeking purification or atonement. Imposing a legal restriction on these latter would be not only “cruelly inappropriate,”⁴

but also potentially self-defeating, since the suffering of incarceration might secure the same end as the intended self-mutilation.

Still, though Feinberg does not consider such cases as the ones I have described, I do not think it too difficult to identify the main lines of his analysis. As is well known, Feinberg rejects what he terms hard paternalism, the view that it is always a legitimate reason in favor of some criminal sanction that it is probably necessary to prevent self-regarding harm. He allows for the plausibility of so-called soft paternalistic statutes, those that restrict a person's liberty against his will and for his own good in cases in which his choice to undertake the action that guarantees or threatens self-harm is nonvoluntary or involuntary. So for liberals who consider cases of self-mutilation, the permissibility of such restraint depends entirely on the voluntariness and self-regarding nature of the choice. In his rejection of hard paternalism, Feinberg argues that these two conditions are jointly sufficient for a moral immunity from state interference. If the people seeking amputations are harming only themselves and doing so voluntarily, then a hands-off policy is in order.

Feinberg tells us that it is not the aim of the soft paternalist to reduce nonvoluntary behavior as such.⁵

Nor is it his aim simply to prevent people from harming themselves.⁶ Rather, it is to “prevent people from suffering harm that they have not truly chosen to suffer or to risk suffering.”⁷ So the authorities, when temporarily placing a would-be amputee in custody, are meant to discover whether her life- or limb-threatening choices are truly her own—whether they are truly voluntary.

Feinberg rejects the view that voluntariness is an either-or concept, applicable or not without degrees. Choices can be more or less voluntary, and given such a view, it is natural to ask: How far less than fully voluntary must a choice be before the law may intervene? Feinberg wisely declines to do anything other than give general guidelines. The greater the threatened self-harm, the higher the probability of such harm eventuating, the more irrevocable the results of a choice, the greater the voluntariness required for legal immunity. How to balance these criteria against one another if they conflict? Well, there is no substitute for sound practical judgment, since no algorithm is possible for rendering such calculations. A critic might take issue with such a stance and demand a greater degree of precision in setting the voluntariness threshold whose passage secures an agent from legal interference with her choices. But I see Feinberg's method here as a plus rather than a minus. He certainly endorsed Aristotle's dictum to the effect that we not demand greater precision of our subject matter than its nature allows. Though this seems to me the path of wisdom, it may, as we shall see toward the end of this essay, cause problems for the liberal.

Let us investigate the matter of whether the choices of apotemnophiliacs are less than fully voluntary. We might be tempted to argue for such a verdict by claiming that such choices are so extremely unreasonable as to necessarily undermine their voluntariness. But as Feinberg understands unreasonableness—namely, as the lack of wisdom and worthwhileness of a given choice—it is completely irrelevant to voluntariness.⁸

A fully competent person can make a choice that is perfectly expressive of her settled preferences with full knowledge of what she is doing and in the absence of coercion and yet choose in a way that reveals attachments to things that we think are relatively valueless. Because choices that meet this description are voluntary ones, the state (says the liberal) has no business interfering with them so long as they issue in actions that wrong no others.

Now, it is true that some choices, such as that to amputate a healthy limb, may raise legitimate doubts as to whether the person is in touch with reality. The pity or revulsion that many feel when contemplating such an act might incline us to a verdict of incompetence and nonvoluntariness. Yet might that not be a rather patronizing attitude, an insult to one whose highly unconventional choices are as voluntary as yours or mine?

If the person is truly irrational in Feinberg's sense, then he is incapable of voluntary choice, and we may well limit his liberty for his own good. In Feinberg's lexicon, irrationality represents derangement,⁹

a wholesale cognitive failure that undermines warranted attributions of moral responsibility.¹⁰ But the unreasonableness of one's choices is compatible with perfect rationality. One's choices may be extremely ill-advised and represent very poor judgment as to what really merits pursuit, and yet one may well know how to locate information about securing one's ends and be very efficient in doing so. Though a judgment of irrationality would indeed undermine the voluntariness of the self-mutilator, we cannot justify such a diagnosis simply on the basis of the unreasonableness of her choices.

There is currently a good deal of debate about whether such people have some kind of mental illness or disorder. But from my admittedly cursory readings on the subject,¹¹

it seems that those tempted by a medical diagnosis are thus inclined only because of the desire's object. Apotemnophiliacs are in other ways perfectly in touch with reality, can have highly coherent sets of desires, and can exemplify as much instrumental rationality as the next person. Though I do not pretend to have a good criterion for distinguishing mental illness from unconventional preference and behavior, I think that most cases are ones in which the apotemnophiliac has a highly unusual basic desire but is in all other respects as sane, rational, and reasonable as the rest of us.

In his classic fashion, Feinberg provided us with a chart that identifies the features of a fully voluntary choice.¹²

Such a choice is made by a competent chooser free from coercion and more subtle manipulation whose choices are not caused by ignorance or mistaken belief and are not made in circumstances that are temporarily distorting. Now, certainly many choices to amputate one's own limb reflect failures to measure up to these standards. But judging from the quick literature survey I undertook, many such choices are close approximations of the fully voluntary. Many apotemnophiliacs know precisely what they are doing, choose their actions from a steady preference for what they are about to undertake, and are not succumbing to duress in making the choices they do. If we set the voluntariness threshold in such a way as to allow most of the choices that the liberal wants to allow, we will be forced to concede that the self-regarding choices of many apotemnophiliacs pass the bar.

TWO LIBERAL ARGUMENTS AGAINST CRIMINALIZING SELF-MUTILATION

That is not enough, however, to warrant a noninterventionist policy on the part of the state. And that is because some self-regarding actions may be so harmful as to create a strong presumption of nonvoluntariness on the part of the actor, thus licensing state intervention. The liberal principle restricting state action tells us that voluntary self-regarding conduct is beyond state scrutiny. But because of evidentiary considerations, policy may not always match principle. It may be that a person is fully voluntarily choosing to slice off a limb. If we do not know this person, however, and we see what she is about to do, we (private citizens as well as state officials) are warranted in stopping her and questioning her to make sure of her condition. In cases where the threatened harm is very great, very likely to occur, and impossible or difficult to remedy, the state is even permitted to suspend the person's liberty and commit her.¹³

Such commitment, however, cannot permissibly constitute a punishment. It is an interference with the person's liberty for purposes of fact-finding rather than the expression of reprobation and the infliction of hard treatment.

Might we go further and seek to criminalize either self-mutilation itself or, in the case of second parties, soliciting, abetting, or performing such mutilation? Given the possibility that those in either single- or two-party cases can voluntarily choose to undertake their actions, it seems that the liberal is bound to regard such criminal sanctions as morally impermissible.

Here we might recall Feinberg's very influential view of legal punishment, which sees the hard treatment that it metes out as conceptually linked with the conventional expression of reprobation. And this view can provide us with the basis of a potentially satisfying argument that shows why any criminal sanctions on either self-mutilation or the mutilation of others with their consent would be unjustified. For ease of exposition, I will run this argument (and the next) as if it applied only to the case of self-mutilation, but the same things could be said, mutatis mutandis, for the second-party case. Let us call this the expressive function argument:

1. A kind of conduct is properly subject to legal punishment only if it is appropriately resented and reproved.
2. Conduct is appropriately resented and reproved only if it is immoral.
3. Self-mutilation is not immoral.
4. Therefore self-mutilation is not properly subject to legal punishment.

This looks pretty good. The first premise is, according to Feinberg, a conceptual truth. It helps us to explain a number of features of legal punishment, such as its ability to vindicate existing law, authoritatively disavow certain actions of state agents, absolve those who might be suspected of having committed a crime, and express symbolic rejection of hitherto accepted conduct.¹⁴

The second premise is also highly plausible. Feinberg relies on it when arguing against the justice and even the coherence of strict criminal liability.¹⁵

There is something fundamentally unfair and perhaps even senseless in knowingly resenting conduct that one regards as morally aboveboard. If there is no moral blemish, there is no ground for reprobation.

And although Feinberg never in fact pronounces on the morality of self-mutilation, premise 3 is not without its attractions. We could, of course, straightaway defend the premise by providing a compelling account of the contours of morality, according to which self-regarding behavior is invariably morally permissible. Though such a view is nowadays fairly widespread, there is unfortunately no account of the scope of morality's domain that is both widely agreed on and such that it makes all self-regarding behavior morally innocuous. Indeed, the major contemporary theoretical contenders—virtue theory, Kantianism, and utilitarianism—all allow room for immoral self-regarding behavior. This may be one of the points where theory and contemporary common sense part company, but in such cases, we cannot just uncritically rely on the prevailing wisdom (if that is what it is) to deliver the final word on this crucial premise.

We can, however, look to Feinberg's own claims about consent to do some work for us. He endorses the volenti maxim¹⁶

and records its place within all modern legal systems and common sense.¹⁷ Consent does not protect against harm, really, but against harms that are also wrongs, that is, violations of moral rights. Provided, as seems plausible, that (1) one's moral rights are not violated by conduct that one has truly consented to; and (2) voluntary self-regarding conduct is conduct one has consented to, then voluntary self-regarding behavior involves no violations of moral rights. This is not enough to show that all such behavior is morally permissible, since some immoralities may not be rights violations. But since the paradigmatic immoralities do involve such violations, there is a significant argumentative burden on those who would argue for the immorality of self-mutilation.

Perhaps this burden could be met, but I am not optimistic and will not pursue such a line here. So, granting this third premise, it seems that we have a sound argument for the liberal conclusion. Yet there is a problem. For Feinberg allows that some morally permissible activities may be properly subject to criminal sanctions and so, presumably, to legal punishment. Consensually entering a duel or a gladiatorial combat, or eating one's feces on a crowded bus—if Feinberg is right, these actions may each be permissibly criminalized. Yet it is not at all obvious that they are immoral. Since that is so, we cannot be sure that self-mutilation, even if morally permissible, is immune from criminal sanctions.

One might argue that the activities just mentioned are indeed immoral, at least once they have been criminally proscribed. But that will not do. Intrinsically morally permissible conduct, such as driving without a seat belt, does not automatically become immoral upon being criminalized. Even if there is a prima facie moral duty to obey the law, it surely is not so strong as to make all actual criminal conduct immoral. Further, Feinberg himself rejects the idea that there is even a prima facie moral duty to obey the law.¹⁸

So what we have in these examples are cases of morally permissible conduct that is nevertheless properly subject to criminal sanction. Thus even if self-mutilation is morally permissible, we cannot rely on the present argument to secure it from criminalization. We have to look elsewhere.¹⁹

I agree with Feinberg that it would be pointless to criminalize the person who has managed to or is about to physically mutilate himself. In drawing parallels between the cases of suicide and those of self-mutilation, Feinberg admits that “there is something moot about the philosophical debate over the propriety of suicide laws when a person with the resources and opportunities can kill himself if he is so determined, whatever the law says.”²⁰

Indeed, this parallel offers Feinberg the opportunity to construct an argument that could at one blow undercut any justification for making self-mutilation illegal. Call this the analogical argument:

1. If criminal prohibitions against suicide are unjustified, then so too are criminal prohibitions of self-mutilation.
2. Criminal prohibitions against suicide are unjustified.
3. Therefore criminal prohibitions on self-mutilation are also unjustified.

We might think this argument a quick clincher for the liberal case against restrictive laws on self-mayhem. Premise 1 seems beyond scrutiny. And premise 2 also seems strong. For, without begging any questions on behalf of Feinberg's liberalism, we could point to the fact that antisuicide laws would in fact be ineffective in preventing suicide and would be unduly harsh in requiring the punishment of those who have tried but failed to take their own lives. We can vindicate the second premise with three quite plausible assumptions: (i) that such laws are bound to be largely ineffective in achieving their aim (that of curtailing suicide); (ii) that their enforcement will result in significant misery; and (iii) that any potential law known in advance to be highly ineffective in achieving its aim and resulting in such misery is unjustified.

The problem is that even if we accept this, the argument is questionable because the first premise is not, in fact, beyond scrutiny. The natural thought behind the premise is this: if even the greatest possible self-harm does not warrant legal restriction, then less severe self-harm cannot do so either. And that does sound very plausible. But there is a bit of a puzzle here. Most of us would regard death as a greater harm than enslavement. By the reasoning that supports premise 1, then, the de jure liberty to commit suicide should imply a legal power to sell oneself into slavery. But it does not. Feinberg rejects antisuicide laws but thinks that those prohibiting slavery can be justified. So perhaps we need to rethink the first premise.

I suggest that we devote a bit more attention to Feinberg's argument against the legality of voluntarily selling oneself into slavery, with an eye to seeing whether such an argument can be brought to bear on the matter of self-mutilation. The basis for disabling citizens from making enforceable slavery contracts is a complicated one. At first blush, it seems that such a course is straightforwardly paternalistic. But, as with many laws that seem both intuitively plausible and illiberal, Feinberg seeks to find a way to accommodate such legal strictures within his framework. Perhaps we can do the same with laws preventing self-mutilation.

A LIBERAL ARGUMENT FOR CRIMINALIZING SELF-MUTILATION

In order to press the slavery arguments into best service, I think that we should concentrate on the two-party mutilation case, that is, the sort of case in which one enlists the aid of another to amputate a healthy limb. Focusing our attention in this way will have two benefits. First, where antimutilation laws might have little or no deterrent effect in single-party cases, they might have greater effect when a second party is involved. Second, the slavery cases are essentially two-party cases, and so if we are to look to them for advice on the mutilation issues, it would be best to have the analogy be as close as possible. Further, it seems that we can draw direct parallels from two-party to single-party cases (and vice versa) provided we are considering cases in which agents act voluntarily. As Feinberg tells us, “A person's directly self-affecting actions and the consented-to behavior of others that affects him are united and placed in the same moral category.”²¹

Where the arguments we have considered so far are efforts to substantiate a clearly liberal conclusion about self-regarding conduct, the argument we are now to consider tries to justify the legal prohibition of apparently voluntary conduct. Feinberg does not want to see a society in which slavery is allowed, even in cases in which all parties have voluntarily consented to the transaction. But I am not sure that I would like to see a society that allows people to set up shop and advertise their mutilation services, much less practice them. And if slavery is relevantly analogous to mutilation, then society might be warranted in outlawing mutilation by others as it is in outlawing slavery, even in those cases where citizens voluntarily sign up.

So let us turn the tables for a bit and see whether we can come up with a liberal case for outlawing behavior that, in some instances, is genuinely voluntary. That is what Feinberg tries to do for slavery, and we might apply the argument to mutilation. When considering slavery contracts, he argues that one good reason to forbid them is the likelihood that they'll be abused. Feinberg does not deny that selling oneself into slavery can be voluntary. But given the extreme nature of such conduct, “the state might be justified simply in presuming nonvoluntariness as the least risky course.”²²

This is the most efficient way of preventing harm that one has not truly chosen for oneself. Two principles support such a position:

Perhaps we can say the very same thing about apotemnophiliacs. Though some undertake their actions voluntarily, such actions are so risky and so irrevocable in the damage they render that we might be best advised just to presume nonvoluntariness. It is better that many be prevented from undertaking these actions, even if they have voluntarily chosen them, than that a few be indulged in preferences that are not truly their own.

We might have some qualms about this argument, since the analogy between slavery and mutilation is not perfect. Slavery is an essentially two-party relation, whereas mutilation often does but need not involve a second party who is doing the surgery. Benefits to this second party are likely far greater in the case of slavery than in mutilation, and so the prospects for exploitation and undue advantage are greater in the slavery case. Further, we might suppose that there is a better chance that self-mutilators exercise voluntary choice and that we are better equipped to detect their voluntariness when we encounter it.

Despite these differences, the analogy between slavery and mutilation may be close enough to get the argument off the ground. Suppose it is. Suppose, further, that we grant the analogue to claim 2 and so say that legally allowing mutilation will result in many wrongfully performed amputations. Still, we might ask about the underlying principle expressed by claim 1.

We need a way to refer to this claim easily—I suggest that we call it the comparative claim. On its face it seems very plausible. It is very like another, deeply plausible principle—one that forbids state executions because it is better that one hundred murderers be allowed to remain alive than that one person be wrongly executed. I think this last principle is true, but I have never known how to adequately defend it. Now Feinberg never explicitly offers a defense of the comparative claim. But we might try to reconstruct an argument on its behalf, and this seems the most natural defense:

1. If we allow conduct C, then there is a near-certainty that some people will suffer, against their will, an extremely grave harm.
2. If we forbid conduct C, there is a near-certainty that a much greater number of people will be restricted from engaging in conduct that they want to undertake (namely, C).
3. The prevention of extremely grave harm to a few is morally more important than the exercise of freedom for many others.
4. Therefore, we should prevent such harm and so forbid the conduct that will, with near-certainty, generate it.

The problem with this argument, of course, is that the truth of the third premise would undermine Feinberg's antipaternalism. On the assumption that the law ought to be directed by the morally most compelling considerations, an endorsement of premise 3 would allow us to legally forbid all sorts of extremely risky activities even if they were voluntarily undertaken. The liberal needs an alternative defense of the comparative claim.

I think that we can get to such a defense by considering some of the remarks that Feinberg makes against the enforcement of voluntary slavery contracts. He asks us to consider what enforcing such contracts would really look like and provides the following answer:

A few pages later, Feinberg concludes his line of thinking:

As I say, Feinberg never explicitly defends the comparative claim. But we might see these remarks as constituting such a defense. If successful in this capacity, we will have a solid basis for outlawing voluntary, self-regarding behavior that is both highly risky and, when the risks are realized, irrevocably and significantly harmful. Though in principle beyond the scope of state scrutiny, such behavior may be outlawed because on closer inspection it is really a kind of harmful other-regarding behavior. We can subsume it under the harm principle and so, with a clear liberal conscience, outlaw it after all.

I think that liberals should balk at this appeal to the psychic costs of third parties. That an activity will demoralize those who witness it, make them uncomfortable or dispirited, or even make them question the moral authority of the state that allows such conduct is not, on liberal grounds, sufficient to render it criminalizable. We might concede that these unhappy witnesses are in some way harmed by being faced with the choice of inuring themselves to suffering or paying out of pocket to assist a desperate stranger. Yet harm is not enough to justify criminal prohibition. We also need a rights violation, and it is not clear that the rights of these third parties are being violated.

The central situation faced by the third parties is this: either refuse to aid another in immediate need or provide assistance that one is legally and perhaps morally free to withhold. I think there are least two important points to make here. First, we have not received an argument to the effect that one has a moral right not to be placed in such situations. Without a compelling argument to that conclusion, there is no liberal case for outlawing conduct that forces such a choice on others. Second, if there were such an argument, it would allow the state to limit our liberty in a way that seems highly illiberal. Any quite risky activity is one such that its practitioners might die or be greatly harmed “right in front of our eyes.” And so, apparently, all such activities may rightly be prohibited on grounds that allowing them violates the rights of third parties.

That obviously will not do and is obviously not what Feinberg had intended. Though he does not offer this consideration, Feinberg might have said that the chances of facing such difficult choices are far greater with enforceable slavery contracts than with (say) allowing private alcohol consumption or mountaineering, two activities that can land people in life-threatening situations in which they must rely on the help of others. Not every activity that forces the aforementioned choice is criminalizable; just those that pose a very substantial threat of doing so.

Before we can assess this point, we need to distinguish between two interpretations of what might constitute a substantial threat of imposing the relevant third-party harms. The first understands conduct to impose a substantial threat if most instances of it cause significant harm. The second understands conduct to impose a substantial threat if there is a large number of instances in which the significant harm eventuates, even if these instances are realized in only a relatively small percentage of cases.

Slavery contracts would impose significant harms in the first sense—a great many such contracts would yield the unpalatable situations that Feinberg describes in his excerpted passage. Alcohol imbibing would not impose substantial threats in this sense, since only a small percentage of such drinking leads to situations in which third parties are forced to choose between allowing continued alcoholic suffering or intervening at their own expense to alleviate it. But alcohol imposes substantial threats in the second sense—it is the primary cause of a huge amount of suffering, even if such suffering occurs in only a minority of cases in which people drink.

Clearly, then, the liberal needs to defend this first sense of “substantial threat” over the second. Yet it is far from obvious that there is good reason to prefer the first to the second sense. Self-regarding activities might very reliably endanger their practitioners and yet, if practiced only by a few, will force only a relatively small number of unpalatable choices on third parties. Self-regarding activities that are practiced on a wide scale, however, might generate an even greater number of such unpalatable choices, even if the chance of their doing so is far smaller than activities of the first kind. We need a defense, which we have not received, for the claim that the second sort of activity must be allowed although the first can be criminalized. We also, of course, need a defense of the most important claim—that the rights of third parties are violated by witnessing the extreme vulnerability of the unfortunates whose self-regarding activities did not turn out as planned. Until we have these arguments in hand, the liberal case for nullifying slavery contracts, and so, by analogy, for prohibiting second parties from mutilating those who consent to such treatment, is insecure.

Now the liberal might just bite the bullet here and elsewhere and so say that if truly voluntarily undertaken, one should be legally allowed to sell oneself into slavery, to have all of one's healthy limbs amputated, to engage in private duels, and so on. But if these really are the costs of adhering to liberalism, we might want to reexamine its foundations and ask whether we have received a compelling argument for thinking that actions that are both self-regarding and voluntary must be legally allowed.

TWO ARGUMENTS FOR LIBERALISM

Joel Feinberg was not the sort of philosopher who worked from the bottom up. He did not articulate and defend ultimate ethical principles or fundamental political strictures and then develop a system of derivative principles (say) for the criminal law. Rather, he worked from middle theorems, such as the harm principle, that could find support in much of what we already believed and that did not, at least as Feinberg tells the tale, need a vindication from absolute first principles. So it is no surprise that we do not see him deriving the liberal case from some yet deeper principle. Still, we would like a succinct statement of a master argument against hard paternalism, and it is a surprise to realize how difficult it is to locate one in Feinberg's writings.

That said, I do believe that Feinberg has provided the essential elements of two distinct and mutually supportive arguments on behalf of the liberal view of self-regarding conduct. Though I am quite sympathetic to such a view, I do not believe that either of these arguments is conclusive.

What follows are two reconstructed arguments that, I hope, are faithful both to Feinberg's intentions and to his writings. The first of these can be extracted from the following passage:

Feinberg here deploys a method that does yeoman's work for him throughout his corpus. He is getting us to appreciate what we already believe. He does this by presenting, with adequate detail and a care for relevant distinctions, the salient options in a nuanced way. His is an essentially conservative way of approaching philosophical problems in which the aim is to preserve our fundamental beliefs and have us see that our deepest commitments imply an endorsement of the conclusions that he is arguing for.

What I will call the considered opinion argument reconstructs the line of thinking here:

1. If we carefully reflect on the prospect of legal interference with our own voluntary self-regarding choices, then we will find ourselves opposed to such interference.
2. If we find ourselves opposed to such interference, then we should oppose it not only for ourselves but also for others.
3. Therefore if we carefully reflect on the prospect of legal interference with our own voluntary self-regarding choices, then we should oppose it not only for ourselves but also for others.
4. If a view about personal legal sovereignty is indicative of or an implication of the considered opinions of the citizenry and is such that it would violate no rights if widely implemented, then it should be enshrined into law.
5. Antipaternalism meets these conditions.
6. Therefore antipaternalism should be enshrined into law.

The argument is formally valid and its premises are all plausible.

Feinberg does not defend premise 1. He instead issues an invitation to reflection and expects quite reasonably to receive the answer that is reported in this premise. Premise 2 relies on an unstated principle of universalizability or impartiality, one that requires us to recognize that those who are relevantly alike should be treated similarly. In this context, the way in which we are relevantly like others is just our capacity for voluntary, self-regarding choice. Thus any legal protections for such choices that we seek for ourselves should, in consistency, be granted to others who share this capacity. Premise 3 follows from 1 and 2. Premise 4 receives no defense in the passage I have quoted, but it seems to me a fair summary of Feinberg's overall views. Premise 5 is true on a few plausible assumptions: first, that Feinberg's expectations of our reply to his thought experiment are on target—we really would, upon reflection, resent the suppression of our voluntary, self-regarding choices; second, that this supports a general antipaternalist preference—that the argument from 1 to 3 is sound; and third, that refusing to interfere with self-regarding choices would not violate anyone's rights. The conclusion, 6, follows directly.

There might be some concern about whether this argument is question-begging. It does not offer anything to those who lack the strong antipaternalist sentiments that Feinberg's thought experiment is intended to evoke. Rather, it presupposes the existence and the authority of such sentiments. So we might ask just how widely shared the antipaternalist response is—whether, say, it is something rather local and confined to liberal philosophers or to citizens raised within liberal political institutions, or whether it is a response we might expect most everyone to have. Further, even if we were to discover that such judgments are very widely shared, we might still ask about their normative force. How much credibility should we assign to intuitive judgments that enjoy broad support? That most of us respond the same way to a thought experiment might be taken, as it is by Feinberg, to be a very powerful argumentative consideration. Or it might be thought that such intuitive judgments lack credibility until vindicated by something else—say, by more general moral principles, by a method or procedure of normative confirmation, or by a comprehensive political theory.²⁷

It is hard to know what to say here. From this armchair, it certainly seems that most everyone would resent the sort of state interference that Feinberg is discussing. True, there are bound to be some extremely self-effacing citizens or some whose confidence in state authority is so great as to quash any resentment at paternalistic interference. Perhaps we are justified in dismissing such attitudes in the way that we do racist or misogynist attitudes that clash with liberal principles. But do we not need an argument to do this, an argument that reveals the error of their ways? It is a very complicated business to try to determine when one does and does not need an argument in ethics or political theory. Part of answering that question is knowing when we can credibly rely on widely shared though not universally held intuitive judgments. I do not have a general answer to that question, but as it turns out, we do not need one here. Let us sidestep some of these central questions of moral epistemology and grant Feinberg both the pervasiveness of these antipaternalist sentiments and his warrant in relying on their presumptive credibility. Still, these concessions do not clearly deliver the verdicts that Feinberg seeks.

Though most of us would indeed resent interference with our voluntary self-regarding choices, most of us would also think it warranted of the state to prohibit voluntary slavery contracts, deny people the right to set up amputation shops that cater to apotemnophiliacs, forbid private duels and gladiator contests, require buckled seat belts, prohibit polygamy, and restrict the sale of heroin. If all of our considered judgments are on one side of an argument, then, absent a very compelling argument to the contrary, we are probably right to favor that side in constructing principles of governance. But here we have conflicting considered judgments and no reason available within the argument for giving priority to the antipaternalist resentment over the paternalistic convictions expressed in our responses to the cases I have just mentioned. What is needed is either an argument that shows how liberalism can accommodate these apparently paternalistic practices or an argument to the effect that the paternalistic intuitions are less credible than the antipaternalistic resentment evoked in Feinberg's thought experiment. But the arguments we have thus far examined, designed to fulfill one or the other of these burdens, have as yet been inconclusive.

So perhaps we'd better have a look at the other antipaternalistic argument. I think, in fact, that the argument to follow is really Feinberg's central argument against hard paternalism, so let us call this the master argument:

1. There must be a sharp, principled boundary that distinguishes the self-regarding choices that are immune from legal interference from those that are not.
2. There are only two plausible candidate boundaries: (i) the state may not intervene to prevent voluntary self-regarding choices; or (ii) the state may intervene to prevent such choices provided that it thereby protects a citizen's true interests and thus her good.
3. (ii) is not a viable option.
4. Therefore self-regarding choices are immune from legal interference if they are voluntary.

The entire discussion about the merits of hard paternalism presupposes the truth of the second premise, so I will do the same here. A great deal obviously hangs on the reasons Feinberg can muster in support of premise 3. As we shall see, attending to these reasons leads naturally to a consideration of premise 1, so I will take these in turn.

As far as I can tell, the basic reason Feinberg endorses 3—and so believes that the hard paternalist strategy is bound to fail—is because of what I will call a theoretical slippery-slope argument. Such an argument has three steps. The first asserts the need for a theoretical demarcation between classes of things. The second claims that the theory or principle under consideration cannot offer the needed demarcation. And the third concludes with the falsity of the relevant theory or principle.²⁸

In the present case, Feinberg's concern is that the hard paternalist lacks any principled way to demarcate those cases of self-harm in which the state is and is not allowed to intervene. The liberal places an absolute prohibition on state interference with voluntary self-regarding behavior, no matter how harmful. That is a very bright line, clearly drawn. The paternalist would blur this line. Feinberg's worry is that once we open the door to intervention in such cases, there is no principled way to identify how much self-regarding harm or what kinds of such harm the state may step in to prevent. We might feel quite confident in the merits of seat belt and helmet laws, or laws that prohibit polygamy or voluntary slavery, yet “the trick is stopping short once we undertake this [hard paternalist] path, unless we wish to ban whiskey, cigarettes and fried foods, which tend to be bad for people, too.”²⁹

Feinberg fully realizes that things are not completely cut-and-dried for the liberal. His position rests on the ability to clarify the distinction between self- and other-regarding behavior, which has never been done with complete success:

This last point is extremely important, as it complements Feinberg's view that the boundaries of personal sovereignty are morally absolute:

These remarks reveal two points at which opponents might dissent. First, they might argue that even though voluntary, self-regarding conduct is within the scope of personal sovereignty, such sovereignty does not, as Feinberg thinks it does, erect a morally absolute barrier against interference. Sovereignty would not be relevantly like virginity but, rather, more like reasonableness or voluntariness—open-textured concepts, instantiated in degrees, with vague boundaries. Or, second, opponents might accept Feinberg's conceptual point and so concede that it is never permissible to infringe one's sovereignty, but then proceed to argue that certain voluntary, self-regarding choices—say, the significantly harmful ones—are outside the scope of such protection. To take such a stand does not entail that the state ought to intervene or is in fact permitted to do so in a given kind of case, but only that the state may legitimately put such harm on the scales in creating its laws concerning self-regarding conduct.

I am not sure whether either paternalistic strategy can be made to work. But I do not think that Feinberg has shown that they cannot. And this is a problem, because, as Feinberg admits, the burden is on the liberal to show that a balancing strategy must fail. A balancing strategy is one in which the prevention of the evil of self-harm carries some defeasible weight in determinations about the permissibility of state intervention. Short of such a showing, we have both the admitted evil of the harm itself plus many intuitive convictions that appear to be best explained and justified by hard paternalism as strong presumptions in favor of the balancing strategy.³³

In fact, I think that there are two sets of considerations that we can locate in Feinberg's own writings that favor a balancing strategy and so work against the master argument. Let me set out what I regard as a version of the most plausible such strategy (if any is), and then I will introduce the factors from Feinberg's work that, surprisingly, lend some support to this paternalistic approach.

The most plausible balancing strategy might, first of all, take issue with Feinberg's assumption that the playing field is limited to two extreme positions. These extremes are that of the liberal, who asserts that voluntarily accepted self-harm is never a good reason for state prohibition, and that of the paternalist, who argues that such harm is always a good reason for such measures. What is missing here is a middle course, one that asserts that such harm is only sometimes a good and relevant reason.³⁴

Different paternalists could present different proposals about how to set the threshold that qualifies a kind of self-harm as a good and relevant reason for intervention. Some might argue that any significant harm counts as such a reason. Others might argue that it must be significant and not be associated with activity that the citizen regards as centrally important to her happiness. Certainly there are other possibilities. All, I think, would insist at the least that the harm to be prevented be significant. Though we do not have an absolutely sharp threshold for determining the extension of this concept, we can get by perfectly well without it. Feinberg relies throughout his discussions on a de minimis rider³⁵ and does not seem to think that we need any greater precision in applying it than is available through commonsense reflection on cases. Let us follow him here.

On the assumption, granted here for purposes of argument, that prevention of significant self-harm is sometime a good and relevant reason for prohibition, when would such a reason become a decisive reason? Under what conditions (if any)³⁶

would the state actually be justified in preventing voluntarily undertaken conduct that risks or ensures significant self-harm? The balancing strategy, to the extent that it is at all plausible, would probably resemble the following model, which begins with a concession to the liberal. The concession: protection of an agent's voluntary self-regarding choice is always a good and relevant reason against authorizing state intervention that thwarts it. However, the strength of that reason may be outweighed when all of the following conditions are met (perhaps to a very high degree):

1. On a scale of negligible to extremely grave, the self-regarding harm is much closer to the gravest it can be.
2. On a scale of reparability, the harm is irreparable, or nearly so.
3. The probability of the harm being realized is extremely high.
4. The intervention has an extremely high chance of preventing such harm.
5. The intervention employs measures that interfere with an activity or a goal that the agent does not regard as central to her life's enjoyments or ideals.
6. The intervention employs measures that are only minimally physically invasive; it is better to hold someone's hand forcefully to prevent him jumping off a cliff than it is to strap him down, or club him into insensibility, or shoot him in the leg. The less intrusive a paternalistic intervention is, the less problematic it is.

It is open to paternalists, of course, to add or subtract conditions to this list or to argue that while these conditions exhaust the relevant ones, not every condition must be met, or met to a very high degree, before licensing state intervention. But let us proceed with this model, accepting as a near-certainty that further argument will call for significant revisions.

What can advocates of such a model say in response to Feinberg's master argument? Quite clearly, I think, they will take issue with its first premise and so deny the need for a sharp theoretical criterion that can demarcate those self-harms we are allowed to risk from those we are not. They may well concede the second premise, and our model assumes that they will. They should also concede the third premise, which, interpreted in the context of our discussion, says that the paternalist will be unable to provide a precise theoretical boundary that distinguishes when such self-harm may be interfered with and when it may not be. Once we allow that preventing even voluntarily chosen self-harm is sometimes a relevant reason for legal prohibition, I think we are bound to fail to identify a sharp line that distinguishes when such harm may and may not be prevented.

The first premise of Feinberg's master argument assumes that such failure is fatal to the view that allows for it. Feinberg's defense of the first premise consists of two claims: (i) that personal sovereignty is respected in its entirety or not at all; and (ii) that without a sharp boundary there is likely to be no basis at all for rationally deciding when the state may prevent voluntary self-harms and when it must allow such harm to occur. But the paternalist, enlisting some of Feinberg's other views, may question the first premise by questioning both of these supporting claims.

Consider claim (i), that personal sovereignty is, in effect, morally absolute in the sense that, necessarily, any infringement of it is also a violation: it is always immoral to infringe the boundaries of personal sovereignty. Now, on the plausible assumption that sovereignty is best defined in terms of a constellation of rights, this translates to the claim that these rights are never permissibly infringed. But Feinberg is on record as endorsing the view that there are justifiable infringements of rights. He provides a much-discussed example in which a stranded hiker finds an unoccupied cabin and, his life at risk from exposure and starvation, breaks down the door and eats the owner's provisions.³⁷

He glosses this example with the claim that this right in particular and property rights quite generally are not absolute moral protections, and he raises the question (but does not provide an answer) as to whether there are any absolute moral rights at all. His final verdict on the case (and other similar ones) is that the owner did indeed have a right, but one that in this case was permissibly infringed.

Now this does not entail that personal sovereignty is ever justifiably infringed, because the rights that comprise such sovereignty might not include property rights. But that is an implausible claim. Further, when discussing such cases elsewhere, he makes the claim about rights per se that they may sometimes be justifiably infringed.³⁸

So we have here the makings of a reductio:

1. Personal sovereignty is morally absolute.
2. Personal sovereignty is comprised of a constellation of rights.
3. Therefore those rights are morally absolute, that is, never permissibly infringed.
4. But such rights are sometimes permissibly infringed.
5. Therefore either premise 1 or premise 2 is false.
6. Premise 2 is true.
7. Therefore premise 1 is false, and personal sovereignty is not absolute.

If one wants to retain the first premise, the best move is to deny premise 4 and so say that rights are inviolable. Though Feinberg did not think so, that might be the truth of the matter. In the hiker case, for instance, we might deny that the cabin owner really had a right that was infringed by the hiker. Instead, we would narrow the content of the right so that it incorporated an exceptive clause that limited the owner's sovereign domain in cases where his property was needed by another to survive in an unanticipated emergency. Though it is an interesting question in itself, I do not want to pursue the matter of whether rights are always best viewed as possessed of this narrower scope or whether they are, as we suppose in our everyday talk of them, possessed of a rather general content. We can sidestep this issue here because the paternalist can adopt either alternative.

Suppose that rights are possessed of this more general scope but that we may justifiably infringe them in certain cases. If that is so, then even if voluntary self-regarding choices are securely within the domain of personal sovereignty, it might still be permissible to infringe that domain. Just as the hiker is morally allowed to infringe the cabin owner's rights without permission in order to prevent serious harm, so might the state be permitted to interfere with an autonomous citizen's self-regarding choices, provided that doing so probably prevents serious harm. Of course the analogy is far from perfect, but the paternalist is not relying on it to establish the truth of paternalism. Rather, the paternalist is relying on it to establish the moral permissibility of sometimes interfering with sovereign decisions. Whether the conditions under which such interference is permissible include paternalistic contexts is a matter for further argument. Still, success on this more limited point would undermine one of Feinberg's two considerations in favor of the first premise of his master argument. That voluntary self-regarding choice is within the domain of personal sovereignty would not be sufficient to immunize such choice from state interference.

Alternatively, the paternalist could concede the moral inviolability of sovereign choices but then argue that not every voluntary self-regarding choice is one over which the agent is sovereign. This strategy takes issue with the second of Feinberg's supporting claims for premise 1 of the master argument. That second claim, (ii), says, in effect, that without a sharp criterion to distinguish the extent of personal sovereignty, any determination of its scope will be arbitrary. Since we should not base vital decisions about the extent of such sovereignty on arbitrary grounds, we must identify a sharp boundary. From premises 2 and 3 of the master argument, we get the result that only the liberal can provide such a boundary.

What the paternalist should question is whether determinations of the extent of personal sovereignty really are bound to be arbitrary in the absence of a precise demarcation point. After all, other concepts that are crucial to the determination of personal sovereignty, such as those of voluntariness and rationality, designate properties that are instantiated only in degrees. There are no precise thresholds to determine when a choice is voluntary enough, for instance, to be granted immunity from state interference. Though I have been speaking throughout this paper of voluntary self-regarding choices, such talk is really shorthand for those choices that are sufficiently voluntary. But there is no precise threshold to fix the extension of this concept. Since that is so, we may well wonder whether the absence of such precision is really the threat to the balancing strategy that Feinberg takes it to be.

Whether a person is rational or her choice voluntary are matters that can be decided only by balancing a variety of incommensurable factors. The balancing does not admit of any sort of mathematical formula and is imprecise by its very nature.³⁹

Regarding the key concept of voluntariness, for instance, there are a number of considerations that together combine to make for a perfectly voluntary choice,⁴⁰ and when different constituents are realized to varying degrees, the overall assessment of the extent of voluntariness becomes much more of an art than a science.

So the liberal's bright line is not really so bright, since it protects not voluntary self-regarding choices simpliciter but, rather, only those that are voluntary enough to be granted legal immunity. And whether a choice is voluntary enough depends on a variety of factors, each of which is realizable in degrees and none of which is commensurable with one another. So not only is there the standing worry about whether we can sharply demarcate self- and other-regarding conduct, but there is also this new concern about setting a threshold of sufficient voluntariness.

Now this sort of conceptual blurriness is not, I think, at all devastating. Rather, it is a kind of imprecision that we must learn to live with. But if that is so, then claim (ii), the second support for premise 1, is undercut, because we are not forced to regard imprecision in this area as a symptom of arbitrariness. On the assumption, crucial for liberalism, that we can render nonarbitrary determinations about whether choices are self-regarding and sufficiently voluntary, then we must allow that sound political and moral judgments may be made even in the absence of precise thresholds and the commensurability of factors that are constitutive of the relevant concepts. And if that is so, then this last consideration in favor of premise 1 of the master argument is undercut. That leaves us without compelling reason to accept that premise, and so without compelling reason to regard the master argument as sound. We do not as yet have a decisive argument against hard paternalism.