Does death harm those who die? If so, when? Recent literature on the topic suggests several ways in which death could be harmful for those who die, and several theories on when the harm occurs. Ben Bradley argues that death harms those who die during those times in which the dead would have been living valuable lives had they not died when they did. Bradley’s view, like others, developed partly in response to Epicurus. My purpose here is to outline Bradley’s view, taking account of its features that developed in response to Epicurus, and ultimately argue against his position.

            In order to trace the development of Bradley’s view, one ought to begin with Epicurus. An excerpt from Epicurus’ “Letter to Menoeceus”:

                …death, the most terrifying of ills, is nothing to us, since so long as we exist, death is not with us; but when death comes, then we do not exist. It does not then concern either the living or the dead, since the former is not, and the latter are no more. [1]

Bradley interprets Epicurus as follows:

A) anything bad for someone must be bad for that person at some time, and

B) a person’s death is not bad for that person at any time, so that

C) a person’s death is not bad for that person.[2]

Bradley objects that premise B holds only if we accept two further claims—(1) that intrinsic goods and harms are the only relevant ones, and (2) that a person must exist at a particular moment to be harmed at that moment. We ought not, he thinks, accept either of the two further claims. Bradley’s argument against (1) revolves around intrinsic and extrinsic goods. Intrinsic goods are things desirable for their own sakes; intrinsic evils those that are undesirable in themselves. Extrinsic goods are those things that tend to lead to intrinsic goods or prevent intrinsic harms, and extrinsic evils those things that tend to prevent intrinsic goods or bring about intrinsic harms.[3] Though death may not be intrinsically harmful, it is almost always extrinsically harmful; a person’s death generally deprives her of a further series of intrinsically valuable moments. So as Bradley reads him, Epicurus’ argument does not account for all salient types of harms (i.e. extrinsic harms). Yet one might think death does not deprive one of intrinsic goods. Part of the reason a person normally objects to being deprived of intrinsic goods is that, as a result, she exists in a deprived state. In the case where death deprives her of goods, she does not exist in a deprived state. If she does not exist to be deprived, we might think, then she has not been deprived at all. But Bradley thinks deprivation harms—extrinsic harms which deprive us of intrinsic goods—do not require us to exist in deprived states in order to harm us.[4] The simple fact that our lives contain less net pleasure than they otherwise would is harm enough. Bradley needs more than this to reject premise B, however; though he has delivered the type of harm that death may be (deprivation harm), he has yet to account for when that harm visits its victim. For this, he tackles (2).

Against (2), Bradley wishes to say that death harms those who die during those times after their deaths in which they would have been living valuable lives had they not died. In addition, he would like to be able to account for the specific amount of harm incurred by the deceased at each and every instant. Thus his timing scheme, if it is to do what he expects of it, must at least (a) be able to identify deprivation harms, and (b) allocate those harms over discrete instants. In order to accomplish both (a) and (b), he must show it is possible to compare a subject’s condition after death to that of her hypothetical living counterpart.[5] This is evident for the following reasons. The account must be able to dub some events deprivation harms. In order to be a deprivation harm, an event must cause a subject to have welfare lower than otherwise. To determine whether this is so, we must construct a hypothetical situation in which the event does not occur, and observe the hypothetical counterpart’s level of welfare. Only if the counterpart’s welfare exceeds the actual subject’s welfare is the event a deprivation harm. Thus a comparison satisfies (a); however, in order to account for (b), the account must break up the comparison into momentary parts. Using the account, one must be able to determine whether the hypothetical counterpart’s welfare exceeds the actual persons’ welfare at any given instant to determine whether the subject suffers the deprivation harm at that instant.

In order to compare a subject’s condition after death and her hypothetical condition had she not died, one must (a) have a method of comparison, and (b) evaluate the subject’s actual and hypothetical conditions via the chosen method. Bradley chose to explain welfare in terms of a single number representing the amount of intrinsic value for the subject at the particular moment in question. Each number is generated by aggregating the intrinsic goods to the subject at that moment, and subtracting from it the intrinsic harms to the subject at that moment. Higher numbers represent net greater amounts of intrinsic value accrued to the subject at the time in question, lower numbers less, zero none, and negative numbers net intrinsic evil. Having chosen a method, it remains only to attribute some set of intrinsic values to each moment in the life and post-mortem times of the subject and hypothetical counterpart in question.

All would agree that positive, negative, and likely zero value can be assigned to a subject for those moments in which she lives. Controversy arises over what values are permissible with regard to post-mortem moments. Recall that in evaluating the deprivation harm of death, the counterpart will be living in some moments for which the actual person is deceased, since the hypothetical situation conjectures that the harming event , in this case death, did not occur. There are three general guidelines one could follow for value attributions to the deceased, each of them predicated on a different view of the kind of things which are capable of having intrinsic value and evil accrued to them.  View 1 (the Any View): Subjects can have intrinsic value after death (in the way that some state of affairs can be valuable for a natural landscape or for objects of human craft). Any positive, negative, or zero value is in principle appropriate for assignation. View 2 (the Zero View): Subjects cannot have positive or negative assignations after death, because only persons or living animals accrue intrinsic value or evil. Only zero is an appropriate assignation. View 3 (the Undefined View): Subjects cannot have positive, negative, or zero value assignations after death. Assigning any value results either in a false claim or in an unacceptable conclusion. Bradley’s theory makes use of the Zero View, and he mentions two reasons for adopting it over the Undefined View. Bradley does not mention the Any View, but we can see why he rejects it regardless. In order to adopt the second view over the first, we must believe one of two things. We may believe that personhood is essential to intrinsic value bearing (at least for humans, the only subjects Bradley mentions), and that the dead have ceased to be persons. Alternatively, we could believe that we are annihilated upon death, so that even if personhood is not essential, whatever the essential element is does not survive our deaths. Regardless, the Any View will not serve; the value assignment is supposed to act as a stand-in for welfare levels. Even if there are states or situations more valuable to crafts or landscapes than others, the resultant values are not indicators of the welfare level of those items or areas. It remains to be seen what reasons Bradley advances for the Zero View over the Undefined View.[6]

            The first of Bradley’s arguments for the Zero View is that it leads to intuitively plausible results.[7] By this, he means that it enables us to treat death like other harms; it is simpler and more elegant. Of course, conclusions are not accepted on the grounds that they are elegant; a view may be correct despite being complicated or awkward. The Zero View allows us to treat death like other harms under his framework, but one ought to fit frameworks to plausible assumptions, not vice versa. Hence, we should reject Bradley’s first argument.

            Bradley’s second argument asserts that some states of living persons contain zero intrinsic value.[8] He says there are some times a subject does not have pleasure and is not in pain. In these cases, the subject is in a state with zero intrinsic value.[9] The analogy: in death, we are neither in pleasure nor in pain. If this state had zero value when the subject was alive, it should have zero value when she is dead. The states are identical with respect to value, and we should treat them as such. This argument has merit, and is discussed below.

We have discussed Bradley’s attempts to reject premise B of Epicurus’ argument. First, he argued that death may deprive us of intrinsic goods, and may thus be bad for us. We saw that for us to name death a deprivation harm at a particular instant, we must determine that the subject’s actual levels of intrinsic value were lower than they would have been had the death not occurred (at that time). For this to be possible, the actual person must have an amount of value that can be compared to the hypothetical amount of value they would have had. This requirement led into Bradley’s second argument, that we should assign zero value to the subject at all times after death that fall under the scope of the comparison. With the groundwork laid, we will look at the particular method Bradley chooses to evaluate this harm.

Deprivation Harm and OVT

            Bradley formulates his theory as follows:

                OVT: The overall value of p for s at <w,t> = the intrinsic value of t for s at w minus the intrinsic value of t for s at the nearest world to w at which p does not obtain.[10]

            The variables above are p for some event or state of affairs, w for possible world, t for time, and s for subject. Bradley intends OVT to determine the timing and magnitude of the value of any event or state of affairs p for s. The t in OVT may be either a particular moment (e.g. t₁), or a period of time (e.g. t₁-t₁₂). To operate OVT, we simply aggregate the intrinsic value of the subject in the world in which p obtains for all moments in the time set, and subtract from it the aggregate of intrinsic value of the subject in the closest not-p world for all moments in the time set. At any moment we care to designate, then, we can discover the value of p for s.

Of particular interest is OVT’s timing mechanism. An example will make this clear. Suppose we wish to know how harmful Tim’s stubbed toe was for him. We notice he stubbed his toe at time t₁, and the toe hurt until time t₄. Because the toe hurt, Tim missed his chance to go on his afternoon jog, which he would have found pleasurable. Let us conjecture that Tim’s intrinsic values in the actual world and the hypothetical world where Tim does not stub his toe and goes jogging as normal were as follows:

Using OVT, we can judge the harm of Tim’s toe stubbing to Tim at any particular moment t₁-t₄, or we can judge the harm for any set of moments t₁-t₄. At t₁, the harm of the toe stubbing for Tim is -250. For times t₁-t₄, the harm of the toe stubbing for Tim is -725. When we determine an overall value for t₁ or for t₁-t₄, we are determining a value for some particular time-slice of Tim. We could rephrase our earlier statement to “Tim-at-t₁ suffers harm of magnitude -250”. In breaking down harms into a moment-by-moment analysis, Bradley gives us the intrinsic values for particular time-slices of subjects. The harm of death computes just like any other harm, with the reminder that we must input zero intrinsic value for all post-mortem moments.

One important task remains for our explanation of Bradley’s view. When using OVT to calculate the harm of death, an interpretive issue arises. Two incompatible readings both seem plausible interpretations; in the following section, I will outline the interpretations, ultimately selecting the superior interpretation and attributing the associated view to Bradley.

Interpreting Bradley

An example will help to set the scene. Suppose Pat dies at t₀. If Pat had not died at t₀, Pat would have died instead at t₂. According to OVT, the overall value of Pat’s death for Pat at t₁ in this world is equal to the intrinsic value of t₁ for Pat in the actual world minus the intrinsic value of t₁ for Pat at the nearest world to the actual world in which Pat’s death at t₀ does not obtain. The way in which Bradley must reference Pat depends on our interpretation of “for Pat at t₁.” Bradley may wish to determine the value of a particular time-slice of Pat, that at t₁, or he may wish to determine the computed value of Pat’s death at t₁ for Pat as Pat existed at some other time (e.g. t_-1). Let us take each interpretation in turn and see where it leads.

         Suppose Bradley means to say that the calculation above is the intrinsic value of t₁ for a non-t₁ Pat-slice. If we take this route, it is plain that OVT does not describe the intrinsic value of Pat at t₁ at all. It is simply reporting the intrinsic value to an earlier time-slice of the intrinsic value to a later one. We may struggle with what mechanism Bradley might intend for Pat to achieve this feat of inter-temporal value bearing, but we need not look far to see that Bradley would not endorse this reading. He intends OVT to provide consistent results across all permissible combinations of p, s, and t.[11] In cases where Pat time-slices exist at t, he clearly means for Pat-at-t to bear the value of p at t. He certainly would not want to exclude this possibility. If a past Pat time-slice could bear the value for Pat at t as well as Pat-at-t, we would have a problem with double counting. To illustrate this, let us refer back to the table above. If Tim-at-t₁ could bear the value for both Tim-at-t₁ and Tim-at-t₂, then Tim-at-t₁’s intrinsic value would be -450. Since Tim-at-t₂’s intrinsic value would still be -200 (or less, if Tim-at-t₂ could also bear value for future Tim time-slices), Tim-at-t₂’s intrinsic values would be double counted. More importantly, the system requires the subject to bear the intrinsic value of an event before the occurrence of the event; the appears to constitute backwards causation. We might be able to alleviate the charge of backwards causation and double counting if the value that Pat-at-t bears for all non-t time-slices is zero.  This seems hard to justify.  It appears to be a linguistic trick to allow OVT to treat both harms equally while not requiring any actual value bearing. 

Suppose that Bradley means to say that the calculation above is the intrinsic value of Pat-at-t₁, as our second interpretation would suggest. Bradley would then require there to be such a thing as Pat-at-t₁, else the entire calculation would be meaningless. Yet we see that Pat goes out of existence at t₀. There is no such thing as Pat-at-t₁. There is a possible Pat-at-t₁, but we require an actual Pat time-slice.[12] So under the second interpretation, OVT provides meaningless results when the subject does not exist at t.

It appears that no other readings fare better. Either we have a non-existent Pat-slice, or we have double counting. In order to protect the integrity of OVT, Bradley must move with the second reading, and argue that it is permissible and intelligible to reference, and attribute value states to, persons at time slices in which they do not exist. That Bradley must argue this way is a direct result of his adoption of the Zero View.

Chief among the arguments against the Zero View are those that support the Undefined View, as described above. The following section illustrates some reasons why one ought to accept the Undefined View rather than the Zero View.

No Subject Problem

            The “no subject problem” revolves around Russell’s observation that sentences of the form xRy (where R may stand for any predicate linking subject x with property y) assert that (1) there is an x and (2) xRy.[13] Using this interpretation, one sees that any statement xRy where x does not exist is false. The application here is that Bradley’s view may require statements of this sort; namely statements that attribute properties to non-existent subjects.

Whether Bradley requires these statements or not depends on his view of what happens to subjects when they die. He may believe either that death annihilates subjects, such that they exist before death and do not afterwards (this view has been called the Termination Thesis). Bradley hints that he accepts the Termination Thesis. He admits that we must accept an eternalist outlook if deprivation accounts are viable.[14] Eternalism, in Bradley’s words, is the view that:

 past and future object and events coexist with present ones in a four-dimensional manifold. Other times are treated relevantly like other places; objects and events that exist in the past or future, bud don’t exist now, nevertheless exist, just in the way that objects that don’t exist here nevertheless exist if they exist in a faraway place.[15]

            If corpses were identical to subjects, one would not need to accept an eternalist account in order to accept a deprivation theory. The subject would still exist after death, albeit as a corpse. That Bradley plainly thinks eternalism is essential to deprivation accounts indicates that he thinks the subject stops existing at death. For present purposes, I will assume this interpretation is correct and that Bradley endorses the Termination Thesis, though I will revisit this question later.

            The no subject problem plagues Bradley because of his endorsement of the Zero View.  Recall that the conclusion of the Undefined View was that one cannot assign welfare values to non-existent subjects. By claiming both that the subject does not exist after death, and that in these circumstances the Zero View is preferable, he directly contradicts the Undefined View. Bradley makes comments to this effect in support of his view:

   So the question becomes this: are there any states that obtain during t that have s as a subject, or are “directly about” s, if s does not exist? If there are, I presume that their basic intrinsic values will all be zero…”[16]

It will help to clarify the manner in which people make use of posthumous reference. Surely, there are states that obtain that have S as a subject when S does not exist. “Socrates was dead in 2000 A.D.” is one such statement, when t is 2000A.D. When we reference Socrates, we reference a four-dimensional being, a being that existed in three spatial dimensions and a temporal dimension. We can refer to the entirety of this four dimensional being, or parts of it, depending on our purposes. In a statement like the above, we refer to the entire set of spatial and temporal parts. But we might make the statement, “Socrates argued with Antiphon”. If the works of Plato are believable, this statement is true. This statement does not assert that the entire four-dimensional Socrates argued with Antiphon; there were temporal parts of Socrates that did not argue with Antiphon. Instead, it asserts that there was at least one temporal part of Socrates that argued with Antiphon. Suppose this temporal part of Socrates existed from t_-1000 to t_-950. One could truthfully say, “Socrates argued with Antiphon at t_-975”. Let us call some temporal portion of Socrates that does not range over all the temporal portions of Socrates a “Socrates time-slice”. The statement, “Socrates argued with Antiphon at t_-975” is equivalent to the statement, “The Socrates time-slice at t_-975 argued with Antiphon.” We can abbreviate “the Socrates time-slice at t_-975” as “Socrates-at-t_-975”. We can then translate the common usage sentence, “Socrates argued with Antiphon” to the more specific but roughly equivalent statement “Socrates-at-t_-1000à-950 argued with Antiphon”.

            Socrates time slices exist only when Socrates is alive; we cannot make a reference to a Socrates time slice which exists outside of Socrates’ life span. Recall that a Socrates time slice picks out some four-dimensional part of Socrates. If Socrates is annihilated at death, there can be no postmortem four-dimensional parts of Socrates. Hence, we cannot legitimately reference Socrates time slices existing before Socrates’ birth or after his death. To do so would be to assert Socrates’ existence outside of his lifespan.

            Yet we see that this is exactly what OVT does. In our example with Tim and his stubbed toe, we made use of time slices; Bradley requires references to person time slices in order to calculate the value of a person at any given moment. Let us observe OVT in action when computing the harm of death. Suppose Johanna dies at t₁, and we wish to compute the magnitude of the harm of her death at t₅. OVT will assign Johanna-at-t₅ zero intrinsic value, and then assign the hypothetically still living Johanna-at-t₅ some other value. Johanna-at-t₅ is not a Johanna time-slice. Johanna-at-t₅ does not refer to an actual Johanna time-slice; Johanna ceased to exist before t₅.

            Since Johanna-at-t₅ does not exist, there are no literally true statements that ascribe intrinsic or relative properties to it. Of course, we can legitimately ascribe qualities to non-existent subjects within a fictional framework. For example, we can say, “Pegasus has wings,” with the understanding that we mean, “there exists some fiction in which the character Pegasus had wings.” But we cannot treat Johanna-at-t₅ as we treat Pegasus, since we do not mean to say that Johanna-at-t₅ is a fictional entity.[17] We should instead interpret the statement, “Johannat-at-t₅ has zero intrinsic value” as “There is one and only one entity Johanna-at-t₅, and Johanna-at-t₅ has zero intrinsic value.”[18] This is false. Bradley requires the attribution of properties to non-existent subjects, which translates effectively into Bradley’s requiring false statements to hold if his theory is to stand. We ought to reject this type of comparison, and Bradley’s view with it.

In Bradley’s discussion of the “no subject problem,” he mentions that arguments of this type prove too much.[19] He reasons as follows: past events can be bad for an existing thing. It follows that the “bad-for” relation does not require two existing things. So the “no-subject objection” is wrong; the “bad-for” relation does not require an existing subject. We might best respond by rejecting Bradley’s interpretation of the objection. He seems right in suggesting that past events can harm persons later, and this does lead to the conclusion that non-existing things can be part of a “bad-for” relation. But suppose we stated the objection as: there must be some subject suffering disvalue or enjoying value at the time that the disvalue or value is suffered or enjoyed. This statement seems to capture the reasoning behind the discussion of the “no subject objection” as presented above. Clearly, Bradley’s response does not present a counterexample; his example includes some agent bearing value. Alternatively, one could rejoin that though the example shows events need not overlap our lives in order to harm us, it does not show that we are harmed before our lives begin or after they end.

            I remarked earlier that one of Bradley’s rejoinders to the Undefined View did have merit. Bradley argued that (1) some states of living persons contain zero intrinsic value; at some times a subject has neither pleasure nor pain. In these cases, the subject is in a state with zero intrinsic value.[20]  The analogy, again: in death, we are neither in pleasure nor in pain. So, (2) if these states had zero value when the subject was alive, then (3) they should have zero value when she is dead. They are identical with respect to value, and we should treat them as such.

            I am now in a position to reply to this objection. The Undefined View does not require that all intrinsic value states of a living subject have non-zero value. It simply requires that for any value state, there is some subject in that value state. Assignment of value at a time is, at base, the assignment of a state or characteristic to a thing existing at that time. Though the situation Bradley describes is analogous in that neither state is of positive value or negative value, it is not analogous in the appropriateness of assigning value predicates. Bradley does not present a rejoinder that meets the no subject problem.

Up front, one ought to recognize that although it may be possible to have a neutral welfare level while living, one cannot have a neutral welfare level while non-existent; questions of welfare simply don’t apply.  In addition to the first consideration, it is clear that if intrinsic value, as Bradley’s measure of welfare, can apply to corpses then it is simply not a good measure. So not only does the Zero View fail, even if it was successful it would be of no benefit to Bradley; it would only mean that intrinsic value is a poor measure of welfare.

            Bradley’s argument falls short, but it is useful in that it reveals a confused argument that may  arise. One may reach (3) via the claim that one cannot have welfare while non-existent, reasoning as follows: if one cannot have welfare while dead, then anything happening after death must have no value, i.e. zero value. This argument is confused because it equivocates on “no value”. The sense of “no value” implied by the claim that one cannot have welfare while non-existent is that of “undefined value,” yet the argument requires the sense of “zero value” in order to reach its conclusion.

            But perhaps Bradley does not think existence ends at death. If he denied the Termination Thesis, he could obviate the no subject problem and his critics who advocate the Undefined View. It seems clear that if S continues to exist after death, then she does so as a corpse. Yet this view is open to the no subject problem in another manner. In some cases, not even a corpse remains after death. A person’s body may be obliterated at the time of death via incineration, vaporization, etc. In these cases, OVT could not yield an amount or timing scheme of death’s harm because of the no subject problem, yet it could in cases where the corpse exists after death. But whether one’s corpse survives after one’s death seems entirely trivial as to how much and when one’s death harms one. Thus OVT will yield saliently different values based on trivial discrepancies; one ought to judge it faulty in that regard.

Some may not find this convincing. For those who think identifying subjects with corpses is appropriate, and that cases in which corpses do not survive death are not problematic, I offer the following section.

Subject Incapable of Welfare Objection

Let us grant, for the moment, that S-at-t exists after S’s death. It does not follow that S-at-t can bear zero value, as OVT requires under this interpretation. We ascribe welfare only to certain subjects—persons, most often. We do not commonly think that rocks, desks, or index cards can have welfare, at least in the sense Bradley requires. If S-at-t is incapable of welfare, then the attribution of zero net intrinsic value is a category mistake. Recall that intrinsic value was adopted as the input for comparison because Bradley required some term to which a numerical value could be assigned as a substitute for welfare levels. If an intrinsic value level is a reliable indicator of welfare, then we ought to refrain from assigning a level of intrinsic value to a subject incapable of welfare. As the dead are incapable of welfare, attributing zero value to them constitutes a category mistake. As such, comparisons that yield the harm of death are impossible under Bradley’s view.

Yet some may think the dead may have welfare levels; if they do, likely they identify a person with a body, which continues to exist as a corpse after death. But claiming welfare levels for corpses is engage in fantasy. If we feel obligate to respect corpses, it is not for the same reason that we respect other persons. Corpses are treated with care out of respect for the living family of the person whose body the corpse was, or because of residual deference to the living person herself. Some may disagree. Consider the following.

Suppose the existence of a particular statue composed of meat. Particular meat statues may be similar in many ways to a corpse. Both may be shaped similarly, decompose at similar rates, are composed of once living things, etc. Neither are sentient. If a corpse may have welfare levels, I see no reason to deny them to meat statues. It is logically possible for meat statues to be persons; that is, be sentient things capable of the type of welfare I endorse earlier. One could create hypothetical situations in which meat statues think and move about of their own accord. In some of these hypothetical situations, we might imagine the meat statues doing well, and having positive net intrinsic values. Notice that some set of events and natural laws prevent the meat statue from being sentient in the actual world. If we aggregate this set into p, we can compare the hypothetical situation with the actual one via OVT; this comparison will yield an attribution of overall harm to the meat statue for failing to be sentient. Further, one could decrease this harm by injuring the non-sentient actual meat statue. The actual statue’s welfare would remain at zero, but sentience would no longer be as valuable to the meat statue in its damaged state; it could pursue and attain fewer intrinsic goods were it sentient and damaged than were it sentient and undamaged, so the value of the hypothetical world decreases. As a result of the damage, p is less harmful for the meat statue. Similar results would follow for a corpse. Under this view, then, one can obviate the harm of death simply by mutilating the corpse of the deceased to a sufficient degree. This is absurd. Assigning welfare to corpses at all, even zero, is neither plausible in itself nor leads to plausible results.

Misstatement of Harm Objection

A further problem arises when OVT evaluates subject, even if we concede the above points. The formula accepts only net intrinsic values of states as input variables, yet to yield accurate results it ought to consider how closely related the hypothetical counterpart is to the actual person; this closeness will differ depending on the similarity between the actual world in which p and the hypothetical world in which not-p. In ignoring such differences, OVT conflates deprivation harm and relational harm. To see how this happens and why it causes problems with Bradley’s analysis, one only needs a clear account of the types of events which deprivation and relational harms are.

A deprivation harm is essentially the “harm of going without.” It is the harm suffered when an event ensures that we will not gain some future value. For example: suppose Pat breaks a leg. After breaking the leg, it turns out that Pat must forgo some valuable experiences; long walks in the park are out of the question, as are many other enjoyable and otherwise accessible activities. We would say that the broken leg deprives Pat of these experiences, and harms thereby (likely in addition to causing the intrinsic harm, pain).

Relational harm, by contrast, is that suffered when the subject is in a less-than-optimal situation.  For example: in world A, Pat is a musician, and makes a fair living. In world B, Pat is Autocrat of the Universe, and has every whim indulged by a loving populace. Pat in world A suffers a relational harm, since being a musician does not provide for Pat’s welfare as well as being Autocrat of the Universe. Essentially, in any modal comparison, if one counterpart is worse off than the other is, then that counterpart suffers a relational harm.

Any relational harm may be reformulated as a deprivation harm. For instance, Pat in world A is deprived of being Autocrat of the Universe by the aggregate of many different events. These events include the cultural developments that discourage people from wanting to be universal autocrats, the technological innovations that provide the rest of the population some defense against Pat’s attempts, and so on. Each of these events makes it less likely that Pat will be Autocrat of the Universe, and together they ensure that it will not happen. We can then express this aggregate as p, so that we may calculate it with other deprivation harms via OVT. Since any relational harm may be expressed as a deprivation harm, and any deprivation harm may be expressed relationally, it may appear that they are indistinguishable.

Yet we would do well to regard deprivation harms and relational harms as opposite ends of a continuum of possible events. On the deprivation end, we have those harmful events which are unlikely given the world we inhabit, and on the relational end, we have those harms which are quite likely. To extend example, Pat’s broken leg meant that she could no longer (or at least not as easily) engage in some pleasurable activities. At the same time, one can reasonably expect that one’s leg will be serviceable given normal conditions. As we move along the continuum towards likely events, the harms become more purely relational. Failing to be Autocrat of the Universe is a relational harm that everyone undergoes. One should see that at some point, it becomes a misnomer to call these types of events deprivation harms. Pat has not been deprived in the same sense by the broken leg and the failure to be Autocrat of the Universe. One may have reasonably expected to have a fully functioning leg, but the same is not true for being Autocrat of the Universe. Even if being Autocrat of the Universe would be tremendously beneficial, and having a fully functioning leg much less so, the latter appears to be a harm in the way the former is not.

Under a modal model like OVT, we might express the intuition as follows. Suppose p is a deprivation or relational harm for s. The not-p worlds we might encounter could be similar to the actual world, or distantly related. If not-p worlds are quite similar to the actual world, p is closer to a deprivation harm. If not-p worlds are distantly related to the actual world, p is closer to a relational harm. As these relations between deprivation and relational harms come forward, we may come to question whether the relationship between the actual and hypothetical world may have a part to play in assessing harms, in addition to the quantitative aspects measured by OVT. In particular, we may be inclined to discount relational harms because of the difficulty of escaping them, and maintain the value of deprivation harms on account of the similarity between the actual and hypothetical worlds.  

Take an extended example to reveal the intuition. Consider three possible worlds, each with subject Jess. In world A, Jess plays the lottery and wins the jackpot for $326 million. Upon reaching the disbursement office, however, she finds the ticket has been stolen. She never receives the money. In world B, Jess plays the lottery but fails to win the $326 million dollar jackpot. In world C, God fails to instantiate $326 million dollars in Jess’ living room. In each case, Jess suffers either a deprivation harm or a relational harm. In world A, Jess had a reasonable expectation of receiving the money. In world B, Jess could expect some non-zero chance of receiving the money. In world C, Jess could not reasonably have expected a non-zero chance. Under the classifications above, world A presents a deprivation harm, world C presents a relational harm, and world B presents a case somewhere between the two. OVT calculates the harm of “losing” the money equivalently for all three cases. However, one ought to count world A Jess as suffering a harm greater than in the other worlds, for the reason that Jess could reasonably have expected to receive the money, and yet was denied. If world B and world C present equal and significant harms, then it becomes obvious that every actual person suffers extreme harm at every moment of existence, since no actual person ever attains an optimal moment. Every person also reaps fantastic benefits at every time, since we exist in a world far from the worst possible world. This is absurd. Relational harms, if harms at all, must be discounted to yield plausible results. OVT fails, in part, because it does not account for the differences between deprivation and relational harms.

Conclusion

            I have outlined Bradley’s view, interpreting him with regard to the required reference of S-at-t. Based on that interpretation, which appears to be the strongest of the alternatives, I argued that his view fails to yield the harm of a person’s death to that person. This failure may stem from the combination of Bradley’s acceptance of the Termination Thesis and the no subject problem. If Bradley rejects the Termination Thesis, then the failure results either from the inappropriateness of using intrinsic value as an indicator of welfare or from the reductio ad absurdum offered against assigning welfare to corpses. The arguments here do not touch on other types of deprivation accounts—ones which do not attempt to determine the harm that accrues to a subject at each and every instant.[21] Despite this, the motivation behind the arguments here suggests a particular way to understand the timing of death’s harm.

Ultimately, we have four inconsistent intuitions. First, we tend to think that events can have effects only when or after they occur. Second, we tend to think that persons may only have welfare levels while alive. Third, we tend to think that death is harmful. And finally, we tend to think that one essential feature of harm is that it negatively affects our welfare. Most accounts of death’s harmfulness reject one of these intuitions and accept the rest. Having objected to Bradley’s view, I would like to describe, in broad outline, the view I take to be most plausible.

I propose that one ought to reject the first of the listed intuitions, or at least the interpretation of it which, when paired with the other intuitions, precludes death’s harmfulness. The intuition, as part of the case against death’s harm, equivocates on “effects.” There is a sense in which an ‘effect’ may only come after a cause, yet there are effects that may come before the cause as well; the harm of death may reside in this second sort of effect. An example: Berg composed Wozzeck in the 20^th century. By composing Wozzeck in the 20^th century, he made it true, eternally, that Wozzeck was composed in the 20^th century; that statement was true in the middle ages (assuming that we change the verb tense), and it will be true forevermore. In this way, actions can have effects upon the truth values of statements uttered before the action takes place. It may be that death harms its victims by making it true, before the death occurs, that they will never achieve their goals, experience less pleasure than otherwise, etc. George Pitcher suggested a way to understand death’s harm as occurring before death.[22] Steven Luper has followed suit.[23] Of the intuitions above, it appears only intuition one’s applicability to the situation is predicated on a misunderstanding. If we are to discover the timing of death’s harm, we may serve ourselves best by understanding our intuitions rather than combating them.