Do you really exist over time?

Parfit in seven parts: Part V

Parfit argues that we do not endure, or exist through time, in quite the way that we ordinarily suppose. He further argues that “identity is not what matters,” as one might split into multiple future “selves” (in the relevant sense) without any of them truly being one and the same person as oneself. Such claims initially sound incredible. But, as we will see, they are supported by reasoning that is difficult to deny.

Reductionism about Identity

Suppose ten people start their own club. It eventually lapses, but years later one of them decides to reinvent the club with a new group of friends. Is there any fact of the matter as to whether this later club is one and the same as the old one? Parfit uses examples like these to support a kind of conventionalism—or what he calls “reductionism”—about identity.

There is a strict sense in which everything is identical to itself, and numerically distinct from every other object (including those that they exactly resemble, which we sometimes call “identical” in a different, qualitative sense: two qualitatively identical tennis balls are nonetheless numerically or strictly distinct objects—there are two of them, not just one). Common sense holds that objects can endure through change: a tree has fewer leaves by winter than it did in mid-summer, but is still (one and) the same tree. Such thoughts can provoke philosophical reflection: what does it take to be the same object at different times? If the tree falls over in a storm, is it still the same tree? What if it is logged and turned into a raft?

These questions can provoke very different reactions in people. Some take them very seriously, as concerning deep truths about the structure of reality. Others take the questions to be superficial, merely verbal, concerning how we choose to talk and categorize the world. Conventionalism seems especially natural when talking about the identity conditions for non-conscious entities such as trees or clubs. (It would seem incredible to claim that if two people had different ideas about the identity conditions for social clubs, one of them must thereby be making some deep mistake about the true structure of reality—as though such entities came with invisible name tags built-in, and we might be wrong about what name was invisibly written on such a tag. More plausibly, disagreeing people just accept different conventions, and nothing in reality forces us to favour either convention over the other.) But it is much harder to believe that the identity of conscious beings, such as ourselves, could be merely conventional in this way.

When I believe that some future person will be me, I anticipate having their experiences in future, and I have a special—prudential—kind of concern for their (my) wellbeing, seemingly different in kind from the other-regarding concern I might have for other people. Or consider a test-case, like Teletransportation: a machine scans my brain and body (down to the last atom), disintegrates it, sends the information via radio signal to another machine on Mars, which then reconstructs the exact configuration of my brain and body (out of all new atoms). The person stepping out of the teletransporter on Mars will have all my memories, beliefs, desires, and personality. They will self-identify as me. But is it really me who will step out on Mars, or have I been replaced with a perfect replica?

This seems like a genuine—and important—question. We may mock the idea of metaphysical name-tags for groups and merely physical objects, but things seem different when it comes to conscious minds: surely there’s a fact of the matter whether the person who steps out of the teletransporter is really me or just a replica.

We can know all the descriptive, qualitative facts about what will happen in the scenario—what the emerging person will think and experience, and how the contents of those thoughts relate to those of the person who entered the teletransporter on Earth—but still wonder whether these thoughts, however similar in content, are being had by the same thinker as before. This way of thinking seems implicitly to presuppose the Featureless Cartesian View that an immaterial mind or soul contains our thoughts/experiences and grounds our identity. We can imagine losing all our memories and turning into an ordinary cat, for example, which is different from imagining that we are (destroyed and) replaced with a cat. The imagined difference presumably lies in whether the subsequent cat-experiences are imagined to be contained within the same conscious mind that used to be ours, however different the contents of this mental container may now be.

What, though, would the identity conditions for such a featureless mental container be? It cannot be given by the contents, since, ex hypothesi, the container may persist through even the most radical changes in its contents (and conversely, containers may be replaced while their contents remain unchanged, as in the case of a replica). But that leaves the identity of the container unmoored. As Parfit objects (following Locke), such a view implies the possibility of brute identity swaps:

[W]hile you are reading this page of text, you might suddenly cease to exist, and your body be taken over by some new person who is merely exactly like you. If this happened, no one would notice any difference. There would never be any evidence, public or private, showing whether or not this happens, and if so, how often. We therefore cannot even claim that it is unlikely to happen.

Could our identities all swap around, to no discernible effect, every minute or even every second? It’s far from clear that such a scenario is coherently intelligible. And this may prompt us to reject the container view of personal identity. Even if immaterial souls existed, they would not be what grounds our identity.¹ What matters is the qualitative content, not the container. But then there is no deep distinction to be drawn between replication and ordinary survival after all. The question whether you survive teletransportation turns out to be no more substantial than our earlier question about the identity conditions for social clubs.

Some thought experiments

Reductionism has striking implications in Parfit’s Branch-Line Case, in which the scanner on Earth doesn’t disintegrate you, but instead damages your heart so that you will die within a couple of hours (while your new Replica on Mars lives on). This is roughly equivalent, Parfit suggests, to taking sleeping pills that cause retrograde amnesia: “if I take such a pill, I shall remain awake for an hour, but after my night’s sleep I shall have no memories of the second half of this hour.” Parfit explains:

Suppose that I took such a pill nearly an hour ago. The person who wakes up in my bed tomorrow will not be psychologically continuous with me as I am now. He will be psychologically continuous with me as I was half an hour ago. I am now on a psychological branch-line, which will end soon when I fall asleep.

This does not seem so bad. Even if your closest future continuant is only truly continuous with your recently-past self, rather than with your current self, you could reasonably regard this as roughly as good as ordinary survival. So even in the Branch-Line Case where you are about to die while your Replica lives on, this too should probably be regarded as roughly as good as ordinary survival. It might be psychologically difficult to accept this. Even if we accept Parfit’s reductionism on an intellectual level, our intuitions may fail to fall into line, much as we may feel fear in a glass-floored elevator despite knowing, intellectually, that we are perfectly safe.

My earlier container/content argument was loosely inspired by Parfitian themes. Parfit’s own argument for reductionism invokes a different thought experiment, the Combined Spectrum:

At the near end of this spectrum is the normal case in which a future person would be fully continuous with me as I am now, both physically and psychologically.… At the far end of this spectrum the resulting person would have no continuity with me as I am now, either physically or psychologically. In this case, the scientists would destroy my brain and body, and then create, out of new organic matter, a perfect Replica of… Greta Garbo.

At intermediate points along the spectrum, varying proportions of the cells in Parfit’s brain and body are replaced in ways that make the resulting person more and more like Greta Garbo. But the spectrum lacks drastic discontinuities. We can imagine a full spectrum of possible people, starting with pure Parfit, then Parfit with a hint of Garbo, through various mixes of the two, until we reach Garbo with a hint of Parfit, and finally pure Garbo. Parfit uses this thought experiment to argue that personal identity is vague, rather than being an all-or-nothing phenomenon. For early in the spectrum, it’s clear that Parfit survives. By later in the spectrum, it’s clear that Parfit is destroyed. But could there really be a sharp borderline between two adjacent points on the spectrum, where removing just one more cell makes all the difference to whether the resulting person is still Parfit? What could make it true that Parbo #721 was still Parfit, while the practically indistinguishable Parbo #722 was not? As Parfit asks, “What would the difference consist in?”

The only plausible response to the Combined Spectrum, Parfit suggests, is to embrace reductionism. We can then respond to the spectrum cases as follows:

The resulting person would be me in the first few cases. In the last case [she] would not be me. In many of the intervening cases, neither answer would be true. I can always ask, ‘Am I about to die? Will there be some person living who will be me?’ But, in the cases in the middle of this Spectrum, there is no answer to this question. Though there is no answer to this question, I could know exactly what will happen. This question is, here, empty. In each of these cases I could know to what degree I would be psychologically [and physically] connected with the resulting person. And I could know which particular connections would or would not hold. If I knew these facts, I would know everything.

Many ordinary language terms are vague, admitting of borderline cases. They can still communicate useful information. But their boundaries are a matter of linguistic convention, and so cannot carry great normative weight. Often they serve to track some underlying scalar property that matters more. (Imagine ordering a “heap” of sand, and then disputing whether the delivered quantity was sufficient. It would seem more productive to discuss the sand’s mass or volume.) The Combined Spectrum seeks to establish this result for personal identity. It is not all-or-nothing, and so it makes no sense to have an all-or-nothing attitude of prudential concern. Our prudential concern should instead track the underlying relations of physical and/or psychological continuity, which come in degrees. On this view, the further along the Combined Spectrum you go, the less of a prudential interest you (the original subject) should have in the resulting person.

This makes an important difference to how concerned the subject should feel about the various outcomes along the Spectrum. Whereas the all-or-nothing view assigns 100% significance to the border between identity and non-identity (and faces puzzles about where the borderline lies), Parfitian reductionists may assign only trivial significance to the difference between any two adjacent points on the Spectrum. On the other hand, they may regard larger differences as being highly significant even if they don’t alter whether or not the resulting person qualifies as “you”: going from 100% to 90% you, say, or from 10% to 0%. And this pattern of concern seems much more rational. Faced with such a spectrum, it makes sense to feel less and less attached to the resulting person, the less of the original “you” they will contain.

We now have our first argument for why identity is not what matters: identity is all-or-nothing, whereas the Combined Spectrum shows that what matters in survival is instead something that comes in degrees—whatever relations of connectedness and continuity underlie our attributions of personal identity.² Parfit himself uses the Combined Spectrum thought experiment in a more limited way, just to argue for his reductionist view of identity. The further argumentative steps we’re now considering are a natural addition, however. (Parfit agrees with the conclusion, that identity is not what matters. His argument instead draws on fission cases, which you can read about in sec. 6.2 of Parfit’s Ethics, if you’re interested.)

Does anything matter in survival?

Suppose we accept Parfit’s reductionism, and the associated view that relation R (one’s degree of psychological connectedness or continuity across time) is all that really matters in survival. Still, we may wonder to what extent relation R matters. Does it matter as much as we thought personal identity mattered, back when we took our identity to involve a further fact? After all, one way of characterizing Parfit’s reductionism would be as a kind of illusionism or anti-realism about personal identity: you could say that we don’t really persist through time at all—we can just talk as though we do, for convenience.

Here’s a crucial question: is it rational to anticipate experiences that will be felt by some “future self” to whom you are strongly R-related? Or does anticipation implicitly presuppose a non-reductionist view of identity? Parfit does not commit himself either way, suggesting that it “seems defensible both to claim and to deny that Relation R gives us reason for special concern.” Of course, your “future selves” (or R-related continuants) are as closely-related to you as can be, so if we have reason to be partial towards anyone, we presumably have reason to partial towards them. But it would still seem a significant loss if we could no longer think of our future selves as ourselves: if they became mere relatives, however close.

I don’t think such a bleak view is forced on us, however. The distinction between philosophical reduction and elimination is notoriously thorny, and analogous questions arise all over the philosophical map. Consciousness, normativity, and free will are three examples for which it is comparably contentious whether reduction amounts to elimination. Parfit himself took a firm stance on the latter two, viewing reduction as amounting to elimination in these areas. He notoriously insisted that we need non-natural normative properties, or else nothing would truly matter. He also rejected strong compatibilist claims about free will, insisting instead that nobody could truly deserve to suffer, because they were not responsible for their own original character—however vicious it may be. So it’s interesting that he did not straightforwardly view reductionism about personal identity as amounting to elimination, instead suggesting that either answer here could make sense.

I find it tempting to give different answers in different cases. Consciousness and normativity strike me as sui generis phenomena, missing from any account that countenances only things constituted by atoms. For free will and personal identity, by contrast, I’m inclined to think that the “non-reductive” views don’t even make sense (the idea of ultimate sourcehood, or originally choosing the very basis on which you will make all choices—including that first one!—is literally incoherent). Reductive accounts of these latter phenomena can fill their theoretical roles satisfactorily, in my view.

Others may carve up the cases differently. However you do it, my suggestion would be that reductionists can more easily resist eliminativist pressures if they think there is no coherent possibility there to be eliminated. If ultimate sourcehood makes no sense, it would seem unreasonable to treat it as a requirement for anything else, including moral desert. So we might comfortably accept a compatibilist account as sufficing to make one responsible in the strongest sense, as there simply is nothing more that could be required. Perhaps a similar thing could be said of personal identity. If we think that “Further Fact” views are not merely theoretically extravagant, but outright impossible, it might be easier to regard relation R as sufficient to justify anticipation. What more could be required, after all?

This reasoning is not decisive. Eliminativists could insist that anticipation is essentially irrational, presupposing something that could not possibly be. Or they could insist that the Further Fact view is not incoherent, but merely contingently false. Even so, their side too seems to lack decisive arguments. As is so often the case in philosophy, it is up to us to judge what strikes us as the most plausible position, all things considered.

The non-eliminative view is, at least, much less drastically revisionary. (If our future selves are better regarded as entirely new people, there would seem no basis for distinguishing killing from failing to bring into existence. You would have to reconceive of guns as contraceptive agents. Nobody survives the present moment anyway, on this view, so the only effect of lethally shooting someone would be to prevent a new, qualitatively similar person from getting to exist in the next moment. Not so bad!) Though even if Parfit’s reductionism can vindicate ordinary anticipation and self-concern, it certainly calls for some revisions to our normative thought (see sec. 6.4 of Parfit’s Ethics).

Parfit in Seven Parts

1. Curiously, Parfit denies this. He suggests that “a non-reductionist view might have been true,” had it turned out that there was reliable evidence of reincarnation. But this strikes me as a mistake. Even if we had reason to believe in immaterial souls that could preserve psychological content (such as memories and personalities) across human lives, there’s no reason to attribute name-tags to those souls, or to take personal identity to be a deep further fact grounded in such name-tags rather than in content-continuity relations. It would be more consistent for reductionists to hold that, even supposing that Sally is Napoleon reincarnated, this fact (still) consists in nothing more than the relations of psychological continuity and counterfactual dependence that hold between the two of them. Ordinarily, our brains serve as the vehicles that underpin psychological continuity and counterfactual dependence between different “timeslices” or momentary stages of people; in the imagined scenario, immaterial souls can also play this vehicular role. But if we previously thought that personal identity was a matter of content-continuity relations (rather than vehicular name-tags), it’s entirely unclear why adding a second possible vehicle into the mix should suddenly make us rethink this.

2. In Parfit’s terminology, ‘continuity’ involves overlapping chains of memories and other connections. The distinction is important for explaining the transitivity of identity: if A = B, and B = C, then A = C, but C might directly remember being B, who in turn remembered being A, without C remembering being A. Such cases suggest that the criteria for personal identity should be formulated in terms of continuity rather than connectedness. But it leaves open which of the two kinds of relation we should ultimately care about.