2.1. Deprivationism and the timing problem

Epicurus famously posed a challenge to deprivationism.

Philosophers interpret this passage as implying the following argument:

A

1. (A1) Anything that is bad for a person is bad for her at a time.

2. (A2) There is no time at which death is bad for the person who dies.

3. (A3) Hence, death is not bad for the person who dies.Footnote ²

(A1) is plausible because it is not clear how something could be bad for someone if it is never bad for her. Epicurus did not explicitly distinguish between intrinsic and extrinsic evils, but I will understand his argument as applying to the claim that death is an extrinsic evil.Footnote ³ (Extrinsically) bad events appear to be bad for people at particular times. If I break my leg, this is (extrinsically) bad for me at times after the event occurs when I am in pain or unable to do the things I would like to do. It stops being (extrinsically) bad for me once I am no longer in pain and have recovered my mobility. Extrinsically bad events that do not cause any unpleasant experiences seem to be bad at times too. Suppose that being deceived is intrinsically bad for a person. If a colleague causes me to be deceived about my own aptitudes by offering false praise for my work, this is (extrinsically) bad for me at all times when I have false beliefs about my aptitudes that I would not have had had my colleague not praised my work. A2 is plausible if we suppose that death coincides with its victim’s ceasing to exist. Because a person does not exist after her death, and hence fails to possess a level of well-being after her death, death cannot affect a person’s well-being after it occurs.Footnote ⁴ Because an event cannot affect a person’s well-being before it occurs, if death does not have an effect on a person’s well-being after it occurs, it is unclear when it can be bad for a person. Call this ‘the timing problem.’ It is worth clarifying a few points about the timing problem.

First, I am interested in defending a modest version of deprivationism in response to this problem. According to deprivationism, if death is bad, it is at least extrinsically bad because of the good life that it deprives a person of. It is natural to think that death is bad for someone, when it is bad for them, because they otherwise would have enjoyed more intrinsic goods. I will assume that the contemporary formulation of the Epicurean argument is an attempt to refute the claim that death is extrinsically bad for the person who dies, qua deprivation.Footnote ⁵

Second, I will assume with the proponent of the Epicurean argument what is sometimes called ‘the termination thesis’: that a person ceases to exist when she dies.Footnote ⁶ She does not continue to persist as her body. This view is incompatible with some theories of personal identity, but Epicurus assumes it, and it is generally taken for granted by participants in the debate about the Epicurean argument.

Third, the Epicurean argument does not entail that dying cannot be bad for the person who dies. Dying is a process that culminates in death. A person exists while she is in the process of dying. Thus, dying might be bad because it involves lots of suffering and prevents the person from enjoying goods that she would have enjoyed had she not been in the process of dying. There is no Epicurean problem here. I am also going to assume that the Epicurean argument applies to particular death events at particular times but not to the fact that someone dies at all. When answering the question whether death is bad because of what it deprives a person of, the relevant comparison is not with a life that goes on forever but with a life that goes on for a bit longer and then ends.Footnote ⁷

Fourth, the question ‘when is death bad for the person who dies?’ should not be confused with two distinct questions. One question is ‘when is the proposition that death is bad for the person who dies true?’ The other question is ‘when does the harmful event of death occur?’ The first question asks when a certain proposition is true, not when an event (death) is bad for a person. On one plausible view, propositions have their truth values at all times, so the proposition that death is bad for the person who dies might be true at all times even though death is bad for the person who dies at some times but not others. The second question with which the Epicurean question might be confused asks when death occurs. The answer to this question need not concern us, because identifying a time at which death occurs would be unlikely to move the Epicurean from the position that death is not bad. What concerns the Epicurean seems to be that, whenever death occurs, its occurrence coincides with the beginning of its victim’s non-existence. I will assume that the fundamental question about the timing of death’s badness, the one that leads the Epicurean to deny that death is bad, is the following:

In responding to (Q), priorists, concurrentists, subsequentists, and eternalists agree that there is some time at which S’s death in w makes S is intrinsically worse off in w than S is in w*, where w* is a world in which S’s death in w does not occur. Priorism is the view that death is bad before the person dies, say, by frustrating certain desires the person has while alive. Concurrentism is the view that death is bad at the time of its occurrence. Subsequentism is the view that death is bad for a person at times after it occurs. Eternalism is the view that death is bad for the person at all times or eternally.Footnote ⁸ Atemporalists and Epicureans, on the other hand, agree that there is no time at which death is bad, but they disagree about what this implies for the badness of death. Atemporalists maintain that death is bad for the person who dies even though it is not bad for her at any time.Footnote ⁹ I discuss the advantages of fusionism over these views in Section 4.

2.2. Desire satisfaction theories and the problem of temporal well-being

Desire satisfaction theories are one family of views about the basic constituents of well-being or ‘value atoms.’ A value atom is any state of affairs that is intrinsically good (for someone) in the most fundamental way. Value atoms are non-derivatively intrinsically good (for someone). That is, their intrinsic value is not solely a function of the intrinsic value of their proper parts.Footnote ¹⁰ For example, hedonists would say that my being happy today and my being happy tomorrow is a derivatively intrinsically good state of affairs for me. It is good for me only in virtue of the intrinsic value of the value atoms that are its proper parts. Taken together, value atoms determine the overall intrinsic value of a world (for a person). Desire satisfaction theorists typically include restrictions on the sort of desire satisfaction that counts as a value atom. Chris Heathwood (Reference Heathwood2007), for example, argues that the satisfaction of a desire is a value atom only if the person desires the object intrinsically and de re. I will not take a position on the sort of restrictions a desire satisfaction theorist should adopt, because, with one exception, any desire satisfaction theory must confront the following time-related problem.

Well-being seems to be a largely temporal phenomenon. Just as things can go well or poorly for someone, considering her life as a whole, things can go well or poorly for someone at particular times during her life. Desire satisfaction theories seem ill-equipped to account for temporal well-being. Suppose that I desire in 2008 that I climb Mt. Everest in 2010. Suppose also that I climb Everest in 2010, but I no longer desire that I climb it. Or suppose that I strongly desire today that I not have made a fool of myself last night while out for drinks with my friends. Last night, I did not desire that I not make a fool of myself, when I was too tipsy to care. If I benefit from the satisfaction of my desire that I not have made a fool of myself last night or my desire that I climb Everest in 2010, when does the satisfaction benefit me? When I have the desire? When the object of my desire obtains? Since neither the desire nor the object of the desire, considered on its own, is sufficient to secure a benefit, there is no reason to say that I benefit when one, but not the other, obtains. Moreover, because the desire and its object never coincide in time, this is some reason to think that the benefit of the desire satisfaction is itself atemporal. If the desire satisfaction theory cannot temporally locate the benefit of future or past-directed desire satisfaction, the desire satisfaction view has a problem accounting for temporal well-being. Call this ‘the problem of temporal well-being.’

There is one desire satisfaction theory that does not face the problem of temporal well-being. Concurrentism about desire satisfaction is the view that a desire counts as satisfied only if the desire and its object coincide in time (Heathwood Reference Heathwood2005, 490). This view precludes the possibility of future or past-directed desire satisfaction. Therefore, it precludes the possibility that future or past-directed desire satisfaction is prudentially intrinsically valuable for a person. Because it does not allow for future or past-directed desire satisfaction, concurrentism does not face the problem of temporal well-being.

I think we should resist concurrentism for two familiar reasons: first, concurrentism entails that future and past-directed desires cannot affect one’s well-being, and this is an unattractive feature of the view. If I desire now that I not have made a fool of myself last night, it seems that I am better off if I did not in fact make a fool of myself last night. Many of our desires are directed toward the past or future in this way, and it would be appealing if the desire satisfaction view could account for their effect on our well-being.Footnote ¹¹ Second, part of the appeal of desire satisfaction theories over simple hedonism is that they do not limit benefits to experienced events. Suppose I desire that my friends respect me but that they do not. I may never come to experience the fact that my friends do not respect me, yet this seems to be bad for me. It is a virtue of desire satisfaction theories that they can explain why. A concurrentist can of course accommodate these cases if they deny the requirement that only experienced events can benefit or harm a person. But, as Dorsey (Reference Dorsey2013, 157–158) points out, echoing Parfit (Reference Parfit1984, 495), once it is granted that the experience of desire satisfaction is not necessary to receive a benefit, if concurrentists accept that spatial distance between the experience of the desirer and the object of her desire is irrelevant to whether her desire counts as satisfied, then why should temporal distance matter? So I think there are reasons to be unhappy with concurrentism, but one can understand my defense of fusionism about temporal well-being as conditional on concurrentism being false.Footnote ¹²

Aside from concurrentism about desire satisfaction, two answers to the problem of temporal well-being have been offered in the recent literature. Dale Dorsey (Reference Dorsey2013) defends ‘the time-of-desire-view.’ According to this view, a person benefits from the object of a future or past-directed desire at the time she has the desire but not when the object obtains. Eden Lin (Reference Lin and Timmonsforthcoming) defends what he calls ‘asymmetrism,’ the view that satisfaction of future-directed desires benefits a person at the time the desired object obtains and that the satisfaction of past-directed desires benefits a person at the time of the desire. I will criticize the time-of-desire view and asymmetrism later in Section 3 in order to showcase the comparative advantages of fusionism.

3.1. Fusionism and internalism

If Internalism is true, then fusionism is the best solution to the problem of temporal well-being. Fusionism says that P2 benefits me at the fusion of t ₁ and t ₃. Formally:

Fusionism about temporal well-being entails that I benefit from P2 at the fusion of t ₁ and t _3, but I do not benefit at t ₁ or t ₃ considered individually. The fusion of t ₁ and t ₃ is the smallest fusion at which P2 benefits me in the sense that it is the fusion with the fewest temporal parts, but it is not the only fusion at which I benefit from P2. I also benefit from P2 at any larger temporal fusion that has the fusion of t ₁ and t ₃ as a proper part (e.g. the fusion of t ₁, t _3, and t ₄).Footnote ¹⁷ Perhaps we should say that I am benefitted only in a derivative sense at any of these larger fusions. I benefit at these larger fusions only because they include the fusion of t ₁ and t ₃. If I am very happy on Monday, then I am intrinsically better off at the fusion of Monday and Tuesday, but this is true only derivatively in virtue of what happens on Monday.

Fusionism denies that I benefit at either t ₁ or t ₃, so fusionism respects Internalism. The value for me of the fusion of t ₁ and t ₃ is determined solely by features that are intrinsic to the fusion. The complex state of affairs my desiring in 2008 that I climb Everest in 2010 and my climbing Everest in 2010 is intrinsic to the fusion of 2008 and 2010. Fusionism thus avoids the implication that the value of a time for someone depends on features that are extrinsic to that time.

It may seem odd that a state of affairs could benefit someone at the fusion of times at which it obtains but not at either of the times that compose the fusion. This should not seem odd to desire satisfaction theorists, however, given what they must say about the structure of value atoms. For desire satisfaction theorists, value atoms are complex states of affairs. These complex states of affairs are composed of more basic states of affairs that are not themselves value atoms. They are thus already committed to the view that complex entities can have intrinsic value for individuals when their individual parts do not.

Fusionism solves the problem of temporal well-being only if fusions of times are temporally located. I have been assuming that fusions of times are times. I will now defend that assumption. Some fusions of times seem clearly to be times. A day is a time but also a fusion of a bunch of times that make it up. A day is a continuous fusion of each moment between its beginning and end. The satisfaction of a past or future-directed desire will, on the other hand, sometimes obtain at a discontinuous fusion of times. P2 obtains at a discontinuous fusion of times; there are times between t ₁ and t ₃ at which P2 has no parts.

Is a discontinuous fusion of times a time? It seems so. Ordinary events occur at discontinuous fusions of times. For example, baseball games may involve rain delays during which the game is stopped and later resumed. Some cricket matches take place intermittently over the course of an entire week. The match is occurring at some times during that week but not others. The match is not taking place when the bowler is at home, asleep in bed. It is better to say that the match has stopped and will resume again in the morning. Plays stop during intermissions and then resume for the next act.Footnote ¹⁸ Academic courses might take place Mondays, Wednesdays, and Fridays, but not Tuesdays or Thursdays. In these cases, one and the same event is spread out across discontinuous times. But these events are not timeless. There are analogous cases involving spatially located things. The United States of America is located at the fusion of its territories and states, which are spread out discontinuously in space. There are spaces in between its parts that are not a part of the U.S. Nevertheless, the U.S. territory is spatially located. Moreover, events occur at the fusion of discontinuous spatial locations. U.S. holidays occur at the fusion of its territories and states. If time is analogous to space, then, if we permit discontinuously spatially located events in our ontology – as we should – we should not be squeamish about discontinuously temporally located events.

One might attempt to explain away each example by denying that each is a single event that occurs at a discontinuous fusion of times. A baseball game isn’t itself an event that occurs at a discontinuous fusion of times, but a series of distinct events (the first inning, the second inning, and so on) each of which occurs at continuous fusions of times. Whether we call the baseball game an ‘event’ or a ‘series’ is not of great importance – though I find it very strange to deny that a baseball game is an event – unless we were also to hold the view that a series of events is itself not located at any time. But this last view is implausible. There is something that one of the teams involved in the game wins and that game occurs. Whether it is an event or a series of events, it is not timeless. Finally, even if discontinuous fusions of times were not times, this would call at most for a minor modification of fusionism as stated above. The phrase ‘any fusion of times’ could be replaced with ‘any continuous fusion of times.’ Clearly, there are many such fusions, since any set of times that together constitute a discontinuous fusion will also be part of a continuous fusion.

3.2. Advantages of fusionism over other views

I have argued that fusionism offers a way to temporally locate the benefit of future and past-directed desire satisfaction while respecting Internalism. Existing answers to the problem of temporal well-being violate Internalism. According to Dorsey’s time-of-desire view, a person benefits from the object of a future or past-directed desire at the time she has the desire (Dorsey Reference Dorsey2013).Footnote ¹⁹ On this view, P2 benefits me at t ₁, when I have the desire but before the desired object obtains. It is important to emphasize that, according to Dorsey, it is the desired object that is the value atom (Dorsey Reference Dorsey2013, 152). What is good for me, in the most fundamental way, is the object of my desire. More on this below. The time-of-desire view violates Internalism because it entails that the intrinsic value of a time (t ₁) for a person can depend on value atoms (my climbing of Everest) that are not intrinsic to that time. Whether I benefit at t ₁ – how good t ₁ is for me – depends on value atoms that obtain at t ₃. According to Lin’s asymmetrism, the satisfaction of future-directed desires benefits a person at the time the desired object obtains, and the satisfaction of past-directed desires benefits a person at the time of the desire. On asymmetrism, P2 benefits me at t ₃, when B obtains but when I no longer have the desire. Asymmetrism also violates Internalism, but the way it does so depends on what the value atom is supposed to be. If the value atom is the complex state of affairs my desiring in 2008 that I climb Everest in 2010 and my climbing Everest in 2010, then asymmetrism violates Internalism because it entails that you benefit intrinsically from the satisfaction of your future-directed desire to climb Everest at t ₃, the time the desire’s object obtains. So the intrinsic value for me of t ₃ is determined by a value atom that does not obtain at t ₃ (by Parts and Atoms) in violation of Internalism. For similar reasons, if the value atom is my desire, then asymmetrism will violate Internalism in cases involving future-directed desire satisfaction. If he says that the value atom is the object of my desire, then asymmetrism will violate Internalism in cases involving past-directed desire satisfaction. The only way for asymmetrism to avoid violating Internalism is to hold that desires are value atoms when they are directed at past objects but that the objects of desire are value atoms when desires are directed at future objects. This view seems implausible on its face.

Dorsey argues that the time-of-desire view is in fact compatible with Internalism. He says,

The success of Dorsey’s argument depends on his ‘Hobbesian’ view that the object of one’s desire – my climbing Everest – is a value atom (Dorsey (Reference Dorsey2013, 152). This requires rejecting the ‘Moorean’ view I have been assuming: that the value atoms posited by desire satisfaction theories are complex, consisting of a desire and its object. Following Bradley (Reference Bradley2009, 19), I think we should resist the Hobbesian view and hence the claim that the state of affairs my climbing Mount Everest is a value atom. If my climbing Everest is a value atom, then it involves the instantiation of a fundamentally good-making property. But the desire satisfaction view says that my climbing Everest in 2010 is good for me only because I desire it in 2008 (or at some other time). This means that whether my climbing Mount Everest is good for me depends on a state of affairs that is not a proper part of my climbing Mount Everest. Thus, my climbing Everest fails to be a value atom, because it fails to involve the instantiation of a fundamentally good-making property. The value atom must consist of my climbing Everest and my desiring to climb Everest.Footnote ²⁰ If we think that value atoms must involve the instantiation of fundamentally good-making properties, Dorsey’s argument for the claim that the time-of-desire view is compatible with Internalism fails; it depends on the false claim that 2010 is (derivatively) intrinsically good for me in virtue of having a value atom as one of its parts. If the time-of-desire view is incompatible with Internalism, then so is asymmetrism. Asymmetrism locates the benefit of future or past-directed desire satisfaction at either the time of the desire or the time its object obtains (but not at the fusion of those times). Asymmetrism will always violate internalism by making the intrinsic value of a time depend on value atoms that are extrinsic to that time.

Asymmetrism and the time-of-desire view violate a second attractive principle:

Both asymmetrism and the time-of-desire view are incompatible with Temporal Independence. Because of this, the time-of-desire view in particular must say some strange things about well-being at a time. The time-of-desire view must say, ‘of a person raised in adversity, that his youth was not so bad after all, simply because his childhood hopes were eventually fulfilled later in life’ (Velleman 1991, 340).Footnote ²¹ Or suppose that our lives in 2008 are identical, but you have one of your 2008 desires satisfied in 2010 when you climb Everest. I would have no reason to be envious of your life in 2008.Footnote ²² But if the time-of-desire view is true, then I would have reason to be envious in 2008, because things are going better for you then than they are for me.

So fusionism is compatible with Internalism and Temporal Independence. The time-of-desire view and asymmetrism are not compatible with these principles. I will briefly discuss three further advantages of fusionism. The first is that fusionism locates the benefit a person receives from the satisfaction of her desire at the time (i.e. the fusion of times) when desire satisfaction occurs. It is highly attractive to say that intrinsically valuable events benefit a person precisely when they happen. Take an experience of pleasure for instance. If an instance of pleasure benefits a person intrinsically, it benefits her precisely the time at which it occurs and no other time. A desire satisfaction theorist should be able to say that having your desire satisfied benefits you when the desire satisfaction occurs. Concurrentism is attractive partly because it allows us to say this by restricting desire satisfaction to cases where the desire and its object occur at the same moment. Fusionism also allows us to restrict the timing of the benefit of desire satisfaction to the time of its occurrence, but the time of occurrence is a fusion of moments rather than a discrete moment. Fusionism says that a person benefits from the satisfaction of her desire at the fusion of times at which the desire and its object obtain. This allows a desire satisfaction theorist to say, attractively, that the satisfaction of a future or past-directed desire benefits a person intrinsically at precisely the time (i.e. the fusion) at which it occurs. The time-of-desire view and asymmetrism cannot say this.

Fusionism is supported by a further attractive principle. Lin argues that one reason to prefer asymmetrism over the time-of-desire view is that asymmetrism, but not the time-of-desire view, is consistent with the following principle.

Fusionism is compatible with the First Principle. Suppose that a condition has been met at a time t if and only if it either is met at t or was met at some time prior to t. According to desire satisfaction theories, the necessary conditions on receiving a benefit are that you desire an object and that the object obtains. Both of the necessary conditions are met on your receiving a benefit any fusion at which the desire and its object occur. The necessary conditions have been met at the fusion because they both obtain at the fusion.

Finally, fusionism is compatible with ‘the resonance constraint’ on personal good, which is taken up by both Lin (Reference Lin and Timmonsforthcoming) and Dorsey (Reference Dorsey2013). Dorsey quotes Peter Railton’s clear articulation of this constraint.

The resonance constraint is a significant motivation for desire satisfaction theories. They are attractive in part because they respect the thought that one must possess a favorable attitude toward that which is good for her. It is plausible that, if we accept the general version of the resonance constraint, we should accept a more specific version (Dorsey Reference Dorsey2013, 156–157). Lin calls this the ‘Synchronic Resonance Constraint.’

If the Synchronic Resonance Constraint is true, it poses a problem for asymmetrism, because asymmetrism entails that you can benefit from climbing Everest at a time when you do not desire it.Footnote ²³ The Synchronic Resonance Constraint does not pose a problem for fusionism. According to fusionism, climbing Everest benefits you – derivatively, in virtue of being one part of a complex value atom – only at the fusion of times at which you (i) climb Everest and (ii) desire that you climb Everest. Any fusion of times at which the object of a desire benefits a person is one at which both the desire and its object obtain, as required by the Synchronic Resonance Constraint.

I have offered a cumulative case for fusionism about temporal well-being. I showed that fusionism is compatible with Internalism, Temporal Independence, and the attractive thought that an intrinsically valuable event benefits a person at precisely the time at which it occurs. The time-of-desire view and asymmetrism are not compatible with those principles. Fusionism also respects the First Principle, which poses a problem for the time-of-desire view. Finally, fusionism respects the Synchronic Resonance Constraint, which poses a problem for asymmetrism.

4.1. Advantages of fusionism over competing solutions to the timing problem

Fusionism avoids a common objection to subsequentism: that death cannot be bad for someone at times after her death because a person cannot be intrinsically worse off at a time if she doesn’t exist at that time.Footnote ²⁷ Fusionism does not locate the badness of death at times after death, so it is not open to this objection. The person who dies exists at fusions of times before and after her death, even though she fails to exist at some of the times that compose the fusion. Therefore, she can be worse off at those fusions.

Fusionism is also superior to priorism and concurrentism. Priorism holds that death is bad for a person at times before she dies because death frustrates certain projects and desires that she had prior to her death. This seems to have implausible implications about the badness of death in certain cases. If S’s death makes no difference to the satisfaction or frustration of S’s actual desires then the priorist who accepts the Epicurean premise (A1) must say, implausibly, that S’s death is not bad for S. The priorist must say this even if, had S not died at t, then S would have acquired desires after t, which would have been satisfied.Footnote ²⁸ Priorism also presupposes the time-of-desire view, which, I have argued, is implausible.

Concurrentism locates the time of the badness of death at the moment one dies. But there may not be anything bad about the moment of one’s death. Suppose one dies painlessly in her sleep and that she would have been unconscious at that moment even had she not died then. Because she is deprived of nothing at the moment of her death, and because she suffers no pain at the moment of her demise, concurrentism entails that her death is not bad for her at any time. It is plausible that many deaths are like this. For many actual deaths, then, concurrentism will entail either the atemporalist conclusion that death is bad for the person who dies at no time or the Epicurean conclusion that death is not bad for the person at all. I say more about the disadvantages of atemporalism in the next section. Unlike priorism and concurrentism, fusionism does not leave anything out of its explanation of the time of the badness of death. The fusion at which a deceased person is intrinsically worse off includes all of the times in virtue of which things would have gone better for the deceased had they not died. Finally, as pointed out in my initial description of fusionism about the badness of death, if death were bad partly in virtue of frustrating the desires of the living, fusionism could account for this by including the fusion of the time of the desire and the time at which the desired object would have obtained had the person not died among the times at which a person’s death is bad for her.

4.2. The uniformity objection

One might object that locating the timing of the badness of death at a fusion of times makes the timing of death’s badness unlike the timing of any other evil. We must locate death’s badness at fusions of times whereas we can locate the badness of other extrinsically bad events at non-fused times. This ‘uniformity objection’ is sometimes raised against atemporalism, but fusionism seems open to the objection as well.Footnote ²⁹ Here, however, fusionism about temporal well-being helps fusionism about the badness of death. If the benefit of the satisfaction of future and past-directed desires is located at the fusion of the time at which the desire obtains and the time its object obtains, then the extrinsic badness of some non-death-related evils will be located at fusions of times as well. If someone frustrates my 2008 desire to climb Everest in 2010 by preventing me from climbing Everest in 2010, they deprive me of something that would have been intrinsically good for me at the fusion of 2008 and 2010. Depriving someone of something that would have been intrinsically valuable for them at a fusion is extrinsically bad for them at that fusion. Thus, fusionism about temporal well-being allows for non-death events that are extrinsically bad for individuals at fusions. This goes some way toward addressing the uniformity objection to fusionism about the badness of death. This is an advantage of fusionism over atemporalism, because atemporalism cannot address the uniformity objection this way. To make a similar reply to the uniformity objection, the atemporalist would have to argue that the benefit of the satisfaction of future and past-directed desires is not located at any time. Though this reply would address the uniformity objection, it would not move the Epicurean to abandon his position. An Epicurean will reject the claim that a person timelessly benefits from the satisfaction of future and past-directed desires for the same reasons he will reject the claim that death is timelessly bad for the person who dies.Footnote ³⁰

But the uniformity objection suggests a further objection to fusionism. Perhaps an Epicurean will accept not only (A1) but the more precise:

1. (A1*) Anything that is bad for a person is bad for that person at some moment.

An Epicurean who accepts (A1*) and a suitably modified (A2*) will be unmoved by the fusionist’s response to the initial formulation of the Epicurean argument. Fusionism concedes that there is no moment at which death is bad for its victim.

Adopting (A1*) seems to ask us to address a different concern than the one captured by the initial formulation of the Epicurean argument. That motivation, I take it, was the observation that, whenever death occurs, its occurrence coincides with the beginning of its victim’s non-existence. This seems to entail that there is no time at which a person fares worse than she would have fared had her death not occurred. Fusionism addresses this concern. But suppose the Epicurean’s specific concern is that there must be some moment at which a person’s death is bad for her. In reply a fusionist can contend that, depending on what we take ‘moment’ to refer to in (A1*), either no event is such that it is bad for a person at a moment or (A1*) is consistent with fusionism. If internalism is true then, a value atom v makes a difference to S’s well-being at a moment m only if the value atom obtains (in its entirety) at m. An event e that deprives S of v is extrinsically bad for S at m only if v would have obtained (in its entirety) at m had e not occurred. Hedonism is in the best position to account for well-being at a moment, because it holds that the value atoms are simple states of affairs. Ben Bradley takes this to be a significant advantage of hedonism over desire satisfaction theories of well-being, which, in virtue of the complex character of their value atoms, have difficulty accommodating momentary well-being. Bradley writes, ‘the [moment] a pleasure is good for me is just the [moment] of the pleasure, and pleasures are in principle locatable [at moments]’ (Bradley Reference Bradley2009, 18).Footnote ³¹ Suppose that a moment is an indivisible temporal atom. On this view, moments are infinitesimally brief. It’s not even clear that we can say that moments are temporally extended. But, plausibly, an experience of pleasure is not infinitesimally brief. It must span an interval or period of time. If moments are temporal atoms, then, even if hedonism were true, there would not be any events such that they are (intrinsically or extrinsically) bad for a person at moments. (A1*) would therefore imply that no events were bad for anyone. I take this to be a reductio of (A1*). On the other hand, if moments are ‘nothing but rather short periods,’Footnote ³² then (A1*) is consistent with the view that events can be extrinsically bad for individuals at fusions. Indeed it requires that bad events are bad for individuals at fusions, but it arbitrarily restricts the candidate fusions to very short continuous ones. I see no reason to privilege very short continuous fusions over other sorts of fusions. Therefore, whether moments are indivisible temporal atoms or very brief periods, (A1*) is implausible.

Atemporalism seems to enjoy one advantage compared with fusionism. Consider a possible world in which a person dies at the last moment in time. After that moment, time ceases to exist. It seems that death would be bad for a person who dies at the last moment in time. In this case, either a fusionist must deny that death is bad, because there are no times after death to create a fusion with the times of death’s occurrence or the times before death’s occurrence, or they must accept that death is atemporally bad in this case.

A fusionist can reply to the end of time case several different ways. First, a fusionist might contend that such cases are impossible because time is necessarily infinite. If time has an end, a fusionist can point out that only in ordinary cases does a person’s death seem to be bad for her at a time, just as other events seem to be bad at times. So only in these ordinary cases should try to accommodate this intuition. But in the very unusual ‘end of time’ case, we have no clear intuition that death is temporally bad for the victim. Nor do we have a clear intuition that any deprivation, such as losing one’s best friend, is bad at any time in the ‘end of time’ case.Footnote ³³ So, the fusionist can grant that, in that kind of case, the person’s death – and any other deprivation – is atemporally bad for her. This requires her to deny that (A1) is necessarily true, but that is harmless. (A1) is true only for events that occur at times other than the last moment in time. Finally, a fusionist might deny that person’s death is bad for her in the end of time case. Only in ordinary cases does a person’s death seem to be bad for her, just as other events seem to be bad. So only in these ordinary cases should try to accommodate this intuition. But in the very unusual ‘end of time’ case, we have no clear intuition that death is bad for the victim. Nor do we have a clear intuition that any deprivation, such as losing one’s best friend, is bad in the ‘end of time’ case. Accepting the implication that death is not bad in the end of time case would thus be a small cost for fusionism, for it can account for the badness of death in any cases that concern us in the actual world, cases about which we have firmer intuitions.