1. Introduction
The locus classicus for discussions of the autoinfanticide paradox is Kadri Vihvelin's (Reference Vihvelin1996) ‘What Time Travelers Cannot Do’:
Suppose that the time traveler, unhappy with her life, decides that it would have been better never to have lived it. She packs a gun and travels back through time, determined to kill her infant self. She picks a time when she knows that the baby will be alone. She checks carefully to make sure her gun is loaded. She fires. (Reference Vihvelin1996: 315)
The paradox lies in the fact that, despite the time traveler's seemingly surefire position, it also seems that her efforts must all fail, for the baby's ‘survival is what makes possible the journey of the time traveler’ (Vihvelin Reference Vihvelin1996: 315). Here I discuss the metaphysical and physical possibility of autoinfanticide.
It is not the aim of this paper to resolve the autoinfanticide paradox, nor is to address Vihvelin's conclusion that the time traveler to the past can't kill her younger self. To resolve the paradox, one would need to explain the enticing but seemingly contradictory judgments to the effect that the time traveler both can and cannot commit autoinfanticide. In this regard, I have elsewhere (2010) defended a version of David Lewis's (Reference Lewis1976a) metalinguistic resolution of the grandfather paradox and have also used this Lewisian resolution to support aspects of Vihvelin's treatment of autoinfanticide against criticisms raised by Ted Sider (Reference Sider2002). As I see it, a resolution of the paradox must focus on what is expressed by natural language modal terms, such as ‘can’ and ‘can't’.
Just so, despite some of the recent literature, the autoinfanticide paradox is at its core not about what is metaphysically possible; it is also not about what is physically possible. Although these are two useful and important concepts, they are semitechnical notions of philosophy. They do not directly help one to understand the seemingly contradictory judgments about what the time traveler can and can't do. Still, there are interesting arguments in the literature that include judgments of autoinfanticide's physical and metaphysical possibility. It is these judgments that are the focus of this paper.
Caution is required with judgments about the physical possibility of time travel because this is in large part an issue of theoretical physics. There is agreement among theoretical physicists that time travel to the past is consistent with the general theory of relativity. There is also similar agreement that the answer to the question about physical possibility awaits a theory that fully unifies quantum mechanics with the general theory (see Smeenk and Wüthrich [Reference Smeenk, Wüthrich and Callender2011] and Thorne [Reference Thorne2009]). To sidestep the difficult physics, I address the physical possibility of autoinfanticide simply assuming that time travel to the past is physically possible. If this assumption turns out to be wrong, then obviously it would be physically impossible for anyone to travel in time to the past and to kill his or her younger self.
Early claims to the effect that autoinfanticide is impossible include:
Suppose I meet my former self . . . I draw a pistol. Can I shoot my old self? It seems not, for to do so would entail my having died before I did so. And this is absurd. (Gorovitz Reference Gorovitz1964: 366–67)
To be sure, if I could go back and kill my infant self, some sort of contradiction would arise. (Malament Reference Malament1984: 98)
Thus we concede that autoinfanticide is impossible but deny that its impossibility follows from the possibility of time travel. (Horwich Reference Horwich1987: 141)
Much more recently, Vihvelin (Reference Vihvelin1996) holds that autoinfanticide is logically possible, but physically impossible. Peter Vranas (Reference Vranas2009) thinks that autoinfanticide is metaphysically possible, but physically impossible partly based on his belief that resurrection is physically impossible. David Horacek (Reference Horacek2005) believes that autoinfanticide has a nonzero chance of occurring and so is physically possible, but he also thinks that autoinfanticide is metaphysically impossible.
To help sort this all out, I present here six different ways to commit autoinfanticide. Following this introductory section, I argue in section 2 that five of these ways are metaphysically and physically possible. In section 3, I provide an example from Horacek that requires an extended discussion of chance that touches on the modal issues at the core of the autoinfanticide paradox. Horacek's example provides that sixth way to commit autoinfanticide, but for interesting reasons it turns out to be only a derivative way of doing the dirty deed. The concluding section, section 4, details some reasonable concerns one might have about my understanding of autoinfanticide, but it also makes clear that true claims to the effect that autoinfanticide is impossible are bound to rely on substantive assumptions, ones that are often not made explicit.
2. The First Five Ways
I use the two names, ‘Suzy’ and ‘Baby Suzy’, which Vihvelin chose for the time traveler. These names allow the reader to keep better track of this person at two different spatiotemporal locations, with ‘Baby Suzy’ calling attention to this person as she is at the spatiotemporal locations at which she has not aged beyond infancy. ‘Suzy’ calls attention to her as she is at the spatiotemporal locations at which she has.
Without any further ado, here are my first five ways for a person to travel in time and kill his or her younger self.
2.1 Option 1: Ordinary Suicide
Suzy takes that trip to the past. If Suzy commits ordinary suicide (blows her adult head off, say), she kills Baby Suzy because she kills herself and she is Baby Suzy. I am not resorting to endurantism; the endurantist and the perdurantist both agree that Baby Suzy is Suzy. They only disagree on whether a person is a perduring thing (i.e., a temporally extended spatiotemporal worm with temporal parts) or an enduring thing (i.e., a thing that may wholly exist at each of two times without having temporal parts).
2.2 Option 2: Slow-Acting Poison
Not satisfied? OK, let's have Suzy bring some very slow, slow-acting poison back with her, and let's have her kill Baby Suzy with the poison. It's so slow-acting that Baby Suzy does not die until she has reached adulthood, travels back in time, and has given her younger self the slow-acting poison.
2.3 Option 3: Resurrection/Resuscitation
Is resurrection metaphysically possible? If resurrection requires an act of God, then, if it is not metaphysically possible that God exists, resurrection must also not be metaphysically possible. If it is metaphysically possible that God exists, then the path is at least open for resurrection to be metaphysically possible. In that case, it would be metaphysically possible for Suzy to kill Baby Suzy and then for Baby Suzy to be resurrected to grow up to be the adult who committed the autoinfanticide. Parallel points could be made about the physical possibility of Suzy's killing Baby Suzy conditional on its being physically possible that God exists. (For that matter, if God should actually exist, then there is even that storied case of resurrection in the actual world to be taken into consideration.)
Vihvelin (Reference Vihvelin1996: 321–23, 327) says resurrection is logically possible and nomologically (i.e., physically) impossible and that it only takes place in worlds very different from the actual world. Vranas (Reference Vranas2009: 524, 529, footnote 7) claims that resurrection is metaphysically possible, but physically impossible. Not much is said by either author in support of these claims. To avoid the religious aspect of all this, I prefer to consider (mere) resuscitation from death. In other words, no act of God is required. This manner of ‘resurrection’ is both metaphysically and physically possible. Resuscitation from clinical death (i.e., cardiac arrest) is metaphysically and physically possible, because it is actual. Though resuscitation from brain death is not actual, it is almost certainly metaphysically and physically possible. Such a resuscitation requires only a chemical and biological rearrangement that returns certain cellular structures to their normal state. As difficult a practical matter as that is, it would not require a violation of a law of nature.Footnote 1 If so, Suzy could go back in time and cause the clinical or the brain death of Baby Suzy, who could then be resuscitated to grow up, travel to the past, and kill her infant self. Thus, especially regarding clinical death, it is hard see how Vihvelin might motivate the claim that Suzy's killing of Baby Suzy occurs only in worlds very different from the actual world. More to the point of this paper, Vranas and Vihvelin both make a mistake in asserting that autoinfanticide is physically impossible.
2.4 Option 4: Fissioned Twin Paradox
Arguably, Baby Suzy could be bilocated through a suitable brain-fission (cf., Ehring Reference Ehring1987; Miller Reference Miller2006: esp. 331–32; and Wright Reference Wright2006). For example, with a partial brain transplant into two new infant bodies, Baby Suzy would be in two places at one time. For convenience, I introduce the names ‘Stay-at-Home Baby Suzy’ and ‘Astronaut Baby Suzy’ to allow the reader to keep better track of this person along her two different paths through space-time. Just so, one could ‘twin paradox’ Baby Suzy by sending Astronaut Baby Suzy on a rocket trip traveling close to the speed of light relative to the frame of reference of Stay-at-Home Baby Suzy. When Astronaut Baby Suzy returns home, Stay-at-Home Baby Suzy could be an adult though Astronaut Baby Suzy is still an infant. Even without a time travel trip to the past, Stay-at-Home Suzy could fire a gun and kill Astronaut Baby Suzy. In other words, Suzy could kill Baby Suzy. Both the fissioning and the high speed space trip are extremely impractical, but there is no reason to deny that they are physically possible.
This way to commit autoinfanticide is not as straightforward a case as the previous ones. There are philosophical challenges to the claim that fissioning preserves identity. I am assuming that the fission is done properly. What I mean by ‘properly’ is that the ordinary factors that are important to survival of the person—the criteria of identity-over-time—are met, whatever those criteria are. That is, things like memory, psychological connectedness and continuity, physical connectedness and continuity, being part of the same life, or whatever the core criteria are, they need to be met. I do need to deny that the bilocation itself (or the appearance that Stay-at-Home Suzy and Astronaut Suzy do not have all the same properties) threatens ‘their’ identity. Some (e.g., Unger Reference Unger1990: 255–58) hold that the post-fission persons are not identical to each other, and as a result neither is identical to the pre-fission person.
There is something of a consensus in the time-travel literature on self-visitation that shows that the bilocation itself (or some other apparent violation of Leibniz's Law) is a bad reason for denying the post-fission identity. There is something of a consensus that if Bill travels back in time to visit his younger self as he was yesterday at noon and remains standing while his younger self is sitting, then it is one person who is both sitting and standing at noon. There are a few different ways that metaphysicians ground this claim. Perdurantists (e.g., Lewis Reference Lewis1976a: 147; and Sider Reference Sider2001: 101–2) hold that Bill is both sitting and standing yesterday at noon in virtue of there being a sitting spatial part and a standing spatial part of Bill's temporal part at noon yesterday.Footnote 2 Endurantists sometimes relativize the sitting and the standing to spatial and temporal location (Miller Reference Miller2006: 315–16). I am sympathetic to another endurantist approach (Carroll Reference Carroll2011: 366–69). It too holds that Bill is sitting and standing at noon yesterday, but with no further explanation needed. No one should claim that Bill is sitting and that it is not the case that he is sitting. When self-visitation or fission is involved, that a person is standing does not imply that the person is not sitting.
2.5 Option 5: Branching Time
There are a variety of models of time that seem to permit a time traveler to change the past (e.g., Meiland Reference Meiland1974; Goddu Reference Goddu2003; and van Inwagen Reference van Inwagen and Zimmerman2010). Suzy killing Baby Suzy fits as well with these models as do any other proposed changes to the past. Though these models differ on some details, all of them are based on the rough idea that the arrival of the time traveler to the past is on a branch that is not the branch from which he or she departed.
Along the arrival branch, the past proceeds differently than along the departure branch. Baby Suzy could be born sometime between t 0 and t 1, grow up normally, and then, at t4, depart on a trip to the past, arriving on the arrival branch just after t 1. She leads the rest of her adult life on this branch, along which she kills Baby Suzy. David Deutsch and Michael Lockwood (Reference Deutsch and Lockwood1994) take a similar approach to time-travel paradoxes by invoking the many-worlds interpretation of quantum mechanics, letting the many worlds play the role of the branches. This interpretation of quantum mechanics is one Deutsch (Reference Deutsch1991, Reference Deutsch1997) champions. Given my working assumption that time travel to the past is physically possible, I expect Deutsch would be prepared to endorse that Suzy's killing Baby Suzy is too.
Some have argued that branching time travel is not really time travel to the genuine past. They deny that the real Baby Suzy exists along the arrival branch. For example, Lewis says of views like Jack Meiland's:
When the traveler revisits the days of his childhood, will his playmates be there to meet him? No; he has not reached the part of the plane of time where they are. He is no longer separated from them along one of the two dimensions of time, but he is still separated from them along the other. (1976a: 145; also see Le Poidevin Reference Le Poidevin2005: 341)
Instead, I treat this issue in the same way I treat fission in section 2.4: So long as the conditions for identity over time are in place between the newborn Baby Suzy and Arrival-Branch Baby Suzy, then it will be the real Baby Suzy that is along the arrival branch as Adult Suzy arrives. These branching models hold that events prior to the start of the arrival branch are part of the causal history of the events along both the departure branch and the arrival branch. One standard condition of identity-over-time, however, does not hold between Baby Suzy along the departure branch at t 2 and Baby Suzy along the arrival branch at t 2. A causal consideration does not hold; neither Baby Suzy is causally ‘downstream’ (cf. Wright Reference Wright2006: 138) of the other. Still, I see no reason to conclude that they are not the same individual given that, by standard criteria, both are the same person as, say, the newborn Baby Suzy before t 1. (All the same points apply to Astronaut Baby Suzy and Stay-at-Home Suzy at times after the fissioning.)
Just to be clear and fill in some details: Suzy's time travel trip to the past could be through a wormhole to the past during which all the usual criteria of identity-over-time are met. She exits the wormhole on the arrival branch just after t 1. Maybe she arrives at her one-month birthday party. This party began prior to t 1 and continues along the departure branch with no sign of Adult Suzy, but it also continues along the arrival branch with Adult Suzy present and ready to carry out her miserable plan. That Baby Suzy exists on the arrival branch at time t 1 + Δ and also exists on the departure branch at t 1 + Δ need not be seen as a threat to their identity any more than the different spatiotemporal locations threaten the identity of the post-fission Astronaut Baby Suzy and Stay-at-Home Suzy.
Some would deny that killing your younger self along a branch would amount to changing the past even though the time travel trip and the killing would cause events along the arrival branch to go differently after t 1 than they do along the departure branch after t 1. For example, Frank Arntzenius and Tim Maudlin say about Deutsch's approach:
Now whatever one thinks of the merits of many worlds interpretations . . . in the end one does not obtain genuine time travel in Deutsch's account. The systems in question travel from one time in one world to another time in another world, but no system travels to an earlier time in the same world. (Reference Arntzenius, Maudlin and Zalta2013)
Whether this amounts to changing the past need not concern me. What matters is that this is a metaphysically possible way for Suzy to commit autoinfanticide. If Deutsch is correct about the many-worlds interpretation of quantum mechanics and about how time travel to the past finds a welcome home in that interpretation, then there would be a good case to be made that this is also a physically possible way to commit autoinfanticide.
3. Option 6: Schrödinger's Baby
If Suzy puts Baby Suzy in a Schrödinger's cat situation with a 50 percent chance that she would thereby kill Baby Suzy and an equal chance that Baby Suzy would live, then surely it is possible—even physically possible—that Suzy kills Baby Suzy. There was a 50 percent chance that she would even though—as it turns out—she does not. This example is from Horacek (Reference Horacek2005: 423–24) and requires more discussion than any of the first five options.
Surprisingly, Horacek argues that, although the automurderer has a 50 percent chance of succeeding and the success itself is a physical possibility, the success itself is not a metaphysical possibility. That is surprising because, for any proposition P, if P is metaphysically impossible, then P is physically impossible, and the chance of P is zero. Metaphysical impossibility entails physical impossibility, since physical possibility is defined in terms of what is metaphysically compossible with the laws of nature. Regarding chance, plausible principles governing chance have been proposed that entail that if P is metaphysically impossible, then the chance of P is zero (cf., Bigelow, Collins, and Pargetter Reference Bigelow, Collins and Pargetter1993: 459; Schaffer Reference Schaffer2007: 124; and Eagle Reference Eagle and Zalta2014). That Horacek judges autoinfanticide to have a nonzero chance and that it is also metaphysically impossible suggests that he has made some kind of a mistake.
Here is what Horacek says about why he thinks autoinfanticide is metaphysically impossible.
There is no possible world in which someone travels back in time and kills herself as a baby. All possible worlds are internally consistent; the one just described would not be. For this reason, autoinfanticide is metaphysically impossible. (2005: 432)
In other words, he thinks that the proposition that someone travels back in time and kills herself as a baby is metaphysically impossible. For the moment, I will go along with Horacek's conclusion that autoinfanticide is metaphysically impossible.
How does Horacek manage to argue that the autoinfanticide has a nonzero chance of taking place? Here is the key juncture:
Does an autoinfanticide have a chance to succeed? There are two ways to disambiguate this question:
(1) Given (Lw & Ctw), what is the Chancetw (baby dies & killer = baby)?
(2) Given (Lw & Ctw & killer=baby), what is the Chancetw (baby dies)?
I think (2) is the correct way to understand the question. It does not ask about the chance of a certain conjunction being true; it is a question about whether a certain special baby has a chance to die, given certain facts about the world. (2005: 425)
‘Lw’ stands for the laws of nature of world w. ‘Ctw’ stands for conditions relative to time t in world w, which for Horacek include the causal ancestry of the proposition that the event in question occurs, including the backward causes that occur after t (Horacek Reference Horacek2005: 424 and see 423, Definition D). It is not part of standard characterizations of chance that propositions about the future backward causes be included in Ctw. So, for now, as does Horacek in his initial discussion of the example, I will set that nonstandard characterization aside. (He does return to the issue. I will too.)
What is important for now is that in reporting the chance of autoinfanticide, Horacek is prepared to disambiguate ‘the chance of autoinfanticide’ in such a way that what he is really reporting is that if Ltw, Ctw, and killer = baby, the chance baby dies is nonzero. So the proposition within the chance operator is just that baby dies, which is clearly not metaphysically impossible. Yet, when the question at issue is the possibility of autoinfanticide, the proposition in question is one that Horacek thinks is metaphysically impossible. It is the proposition that killer = baby and baby dies.
To see the problem, consider an obviously inconsistent conjunction of propositions about a situation that does not involve time travel: the conjunction that Joe sneezes at noon and that it is not the case that Joe sneezes at noon. The chance at 11:00 a.m. (and at all other times) of this conjunction being true is zero, because this conjunction is inconsistent, and so it is also metaphysically and physically impossible. But with all that granted, it may still (a) be true at 11:00 a.m. that it is not the case that Joe sneezes at noon, and it may (b) be metaphysically and physically possible at 11:00 a.m. that Joe does sneeze at noon.
What goes wrong with the example is that when Horacek judges the chance of autoinfanticide, he is judging the chance that baby dies, but when he judges the possibility of autoinfanticide, there is more to the proposition in question; it is that baby dies and killer = baby. It really does not matter what is part of the background or what is within the scope of the possibility operator and the chance operator; pick the same proposition to be within the scope of the two operators without changing the background, and you'll never have a case in which a metaphysical impossibility also has a nonzero chance. Nothing warrants ‘scoping out’ part of the inconsistency regarding the evaluation of the chance of autoinfanticide while not ‘scoping out’ the same parts of the inconsistency regarding the evaluation of the metaphysical possibility of autoinfanticide.
That addresses the mistake that leads Horacek to his surprising view about chance and possibility. Nevertheless, there remains a question about what is the proper way of reporting the chance of autoinfanticide and the metaphysical possibility of autoinfanticide. As I have argued, the proposition within the scope of the probability operator and the scope of the possibility operator should be the same proposition. But what proposition should it be: that baby dies or that baby dies and baby = killer? If the concern really is with autoinfanticide, then the proposition should be the latter. It is not enough to show only that there is a chance that Baby Suzy dies or only that there is a possible world where Baby Suzy dies.
The issue remains, however, whether putting Baby Suzy in a Schrödinger's box is genuinely a way for Suzy to commit autoinfanticide, a way that is both metaphysically and physically possible. Ultimately, my judgment is that it does not provide a way to commit autoinfanticide that is either metaphysically or physically possible without riding the coattails of some other more basic way. For example, I acknowledge that if Baby Suzy is dead when the box is opened and then is resuscitated (as in option 3), then Suzy would have killed Baby Suzy by putting her in the Schrödinger's box. Well, yes, that is a way to commit autoinfanticide that is physically and metaphysically possible, but the Schrödinger setup is incidental—it is more like a bizarre choice of a weapon than a way to dodge a contradiction. Indeed, parallel points could be made using option 4—with Stay-at-Home Suzy putting Astronaut Baby Suzy in a Schrödinger's box—or using option 5—with Adult Suzy doing the same to Arrival-Branch Baby Suzy along the arrival branch. In a straightforward application of the Schrödinger's method, however, should the chancy mechanism kill Baby Suzy and there is no resuscitation (or prior fission or branching time), it seems that it could not have been Suzy who killed Baby Suzy. Should Baby Suzy die, it is hard to see how the criteria of personal identity could have been met such that the baby is the murderous adult. How exactly could it have been Suzy who killed Baby Suzy?
Regarding the chance of autoinfanticide, I still need to consider what belongs in Ctw in a case where backward causation is involved. As pointed out above, Horacek calls for including the entire causal ancestry of the event in question in Ctw. This seems a natural suggestion in that it parallels including the past history of the event, including the causal history, as is usually done with Ctw in situations where the possibility of backward causation is not on the table. The idea is to conditionalize on facts that affect the chance at t of the event. This proposal certainly is worth investigating and developing, but as Melissa Schumacher has pointed out in an e-mail exchange, it appears that this only makes more trouble for Horacek's position. Since one of the backward causes of Suzy's placing Baby Suzy in the box would be that Baby Suzy survives the ordeal, Baby Suzy's surviving would be a member of Ctw. This would entail that her chance of dying is zero, which is good news for Baby Suzy but no help to Horacek or to Suzy.
Issues remain. It seems possible that Suzy emerges from a wormhole, collects Baby Suzy, and drops her in a Schrödinger's box. It also seems that it is possible that things go well for Baby Suzy and she reaches adulthood. It is looking doubtful, however, that there is a metaphysically possible world in which the surviving Suzy kills Baby Suzy in this manner. That is really puzzling given that a Schrödinger's box is supposed to be genuinely chancy. Isn't there some sense in which Suzy could have killed Baby Suzy?
At this point, the dialectic is bumping up against the intuitive core of the autoinfanticide paradox, though the core is disguised a little by the chanciness. Regarding chance and the modal paradoxes of time travel, my sympathies lie with an approach to chance proposed by Anthony Eagle (Reference Eagle2010: 282–91; also see Handfield and Wilson Reference Handfield, Wilson and Wilson2014: 37–40). It treats chance as a relative modality and provides truth conditions of ‘chance’ ascriptions that eschew the standard approach of trying to specify the facts to be conditionalized on in the chance function. Eagle treats ‘chance’ as a relative modal term. He treats ‘chance’ sentences as context sensitive. What is conditionalized on is partly determined by the context. About option 6, Schrödinger's Baby, in line with Lewis's (Reference Lewis1976a) resolution of the grandfather paradox, Eagle's sketch suggests that, relative to the fact that Baby Suzy survives the ordeal and grows up to be an adult, it is true to say, ‘There is no chance that Baby Suzy dies’ even though relative to only prior quantum mechanical facts about the Schrödinger setup—which does not include that Baby Suzy survives the ordeal—it is true to say, ‘There is a 50 percent chance that Baby Suzy dies’.
Even though it seems likely that Eagle's approach has the consequence that some sentences in some contexts that report a positive chance of autoinfanticide via a Schrödinger's box are true, that is not what I was looking for. Traditionally, metaphysical possibility is thought to be absolute. Metaphysical possibility sentences are thought not to be context sensitive. Traditionally, physical possibility is defined in terms of metaphysical possibility and the laws of nature—and there is a kind of relativity to the laws—but, traditionally, physical possibility sentences are not taken to be context sensitive either. Thus, though Eagle's approach is promising regarding puzzles about chance and time travel, it certainly does not provide a basis to judge the method using a Schrödinger's box to be a metaphysically and/or physically possible, nonderivative way for Suzy to commit autoinfanticide.
4. Conclusion: Reflection on the Ways
With option 1, an ordinary suicide, you probably feel that I am not taking the autoinfanticide paradox seriously. This option does not reflect what, in the standard telling of the paradox, is fully intended by Suzy killing her infant self. Though option 1 is a way for Suzy to kill Baby Suzy at a time when Baby Suzy is an infant, I concede it is not a way to commit autoinfanticide as that word has been used in the literature on time travel. Prima facie, autoinfanticide is killing oneself as an infant such that the killing action succeeds in killing Baby Suzy at a location at which she has not aged beyond infancy.
For some philosophers, options 2–5 may also seem not to be in the spirit of the paradox. About option 2, they might say that there was an implicit background assumption that the baby would be killed as an infant and die as infant. About option 3, they might say that there was a background assumption that, once dead, the dead infant would remain dead (Sider [Reference Sider2002: 115] is explicit about understanding ‘death’ as permanent death). Regarding option 4, it is not very likely that philosophers were assuming that surviving fission was impossible, but it may be that they were assuming something like that there was nothing funny going on. Regarding option 5, the case can be made that many philosophers were not prepared to take seriously the possibility of branching time (Kiourti [Reference Kiourti2008: 344, footnote 2] explicitly assumes time does not branch). Of course, there will also be dissatisfaction with the examples stemming from some of the philosophical claims I have made, especially regarding options 4 and 5. That, however, is a quite different matter than the accusation that I am not being fair to the example.
That is enough musing about what philosophers might have been thinking about the case of autoinfanticide. My concern really is not with the historical facts regarding what Vihvelin and others had in mind as the literature on autoinfanticide developed. My concern is to call attention to the need for clarity about what autoinfanticide is and the kind of additional assumptions needed to ground claims of impossibility.
One could explicitly build in lots of assumptions. There surely are ways of setting up the example such that autoinfanticide is not metaphysically possible and therefore is also not physically possible. For example, if it is assumed that (a) to commit autoinfanticide Suzy must kill her younger self in a manner that results in Baby Suzy's permanent death as an infant, that (b) Baby Suzy does grow up to be an adult, that (c) there is never any fissioning, and that (d) time does not branch in any manner, then along the one and only nonbranching timeline there would have to be a time at which Baby Suzy is alive, growing up to be an adult, and is also permanently dead. Understood this way, it may well be that autoinfanticide is metaphysically and physically impossible. Nevertheless, since the contradiction would have been so carefully and painstakingly built in, it is hard to see what the significance of this result would be.
