Parfit and Lewis on Survival of Fission and What Matters in Survival: Why Parfit won the Debate Draft (summer, 2002) mbelzer@bgnet.bgsu.edu Imagine that your body is cloned twice today. Imagine that two resulting organisms will be grown and kept alive and reasonably healthy. Now imagine that in 20 years, your own brain will be removed from your body (which will die at that point) and your brain will be divided, with each half of that brain being installed into one of the two healthy clone bodies (whose own brains have been removed). While this would be immoral to do (since it involves killing the two clones), in this thought experiment we imagine your upcoming "fission" -- if we make the assumption that half of your brain by itself could support all of your mental life. In 20 years, after the operations, there will be two apparently distinct persons in the world with bodies much like the one you had when you were 20 years old, each of whom will have the psychology you will have in 20 years--each will have the memories, projects, and so forth that you will have at that time. This is one way to imagine your personal fission. Another way, not so immoral even if more fantastic, is to imagine in twenty years you happen to go into a malfunctioning Star Trek teletransporter, which beams you twice, resulting in Good-you and Bad-you (perhaps), although of course the diverging qualities might emerge only well after the fission, since the two you's (so to speak) initially come out pretty much the same. Or, without worrying about technical details at all, we can imagine the poet Robert Frost out taking a walk, when
and then suddenly Frost does find himself just walking down each path. (Or rather: an observer in a helicopter hovering overhead sees a person on each path after the fork.--What Frost himself in such a story can be said to find afterwards, if anything, is the main topic of this paper.) There are apparently serious interpretations of quantum mechanics that posit that such fissions are taking place all the time (see Albert and Loewer 1988 --this is how they interpret the "collapse of the wave function"). Derek Parfit says in his 1984 book, Reasons and Persons, that he decided to study philosophy "almost entirely because [he] was enthralled" (518) by the possibility of personal fission as described by David Wiggins. In 1971 he published an article, "Personal Identity," in which he sketched the themes that later he developed in Part III of his Reasons and Persons (RP), and which he continues to develop today. One theme is the "reductionist" thesis that personal identity through time is constituted by relations between mental and physical events in the absence of anything like a necessarily determinate, indivisible soul. The second general theme is that given reductionism, some of our commensensical beliefs about rationality and morality need to be revised. The fission thought experiment has played a central role for Parfit in the arguments for the practical ramifications of reductionism, including his argument that identity is not what matters in survival. Parfit's argument has been challenged by many philosophers (including Sosa, Unger, Johnston). In this paper I will focus on David Lewis's challenge to the `71 article. Lewis presents a reductionist theory which is in general agreement with Parfit's metaphysics. But Lewis claims there is no need to posit a revisionary logical wedge between identity and what matters in survival. In Lewis's interpretation of fission, two or more persons share stages prior to fission, and it is primarily this feature of his theory that has drawn fire insofar as Lewis purported to be defending common sense against Parfit. But I will not be much concerned here with the stage-sharing aspect of Lewis's theory. My goals are to show (1) why Parfit wins the debate with Lewis about the question whether their shared metaphysical views entail conflict with common sense, since Lewis simply assumes that identity (more precisely: copersonality) does not matter. More importantly, I also will show (2) why, if one accepts Lewis' presupposition that his I-relation need not be an equivalence relation, an alternative account (which treats the I-relation as not symmetric) provides a superior treatment of fission --and, indeed, this superior treatment already is implicit in Parfit's original article. (3) I would encourage Parfit now to emulate Lewis: instead of trying to develop the general argument that identity does not matter (which has dubious prospects, at best) Parfit should return to his own `71 advice and simply presuppose that identity does not matter (as Lewis did) and then spell out potential practical and normative ramifications in terms of the account that he began to develop there but later abandoned in Reasons and Persons. Parfit's theory Even though there is general agreement in the metaphysical accounts of persons presupposed by Lewis and Parfit, there also are some technical differences between them that have significant consequences. I will try to reconstruct their accounts in an accurate way so as to minimize the differences while also being able to focus on the relatively few formal differences that are significant. Parfit's version of the Psychological account of personal identity through time does not change from `71 to R&P. Parfit sugggests that a person's existence just consists (according to the Psychological approach) in the occurrence of a series of interrelated mental events, where such events can be "directly connected" with each other. Relevant direct connections include memory, intention, and continuity of belief and desire. Assume that person stages are collections of such states and events (see Lewis SI, 20; and "Postcript," 76-77; Parfit does not use this notion, but his views can be reconstructed accurately by borrowing it from Lewis. For a detailed discussion, see "Psychological criterion" in "References" to this paper.) The relations of "psychological connectedness" and "psychological continuity" between stages provide a basis for the Psychological account of personal identity through time. These relations can be understood generally and schematically as follows. (1) Psychological connectedness is characterized completely as follows, for person stages x and y:
(Parfit speaks of persons at times being connected; we'll get to that in a moment.) Let us abbreviate `x is psychologically connected with y' as `Cxy'. Relation C ("connectedness") is not transitive, Parfit says, since the mental relations that constitute connectedness are not transitive. Let us leave open for the moment whether or not C is symmetric (it seems arbitrary, but this will turn out to be an important question). C comes in degrees. The degree of connectedness between two stages x and y depends on the number and significance of the direct psychological connections between x and y. (2) The idea of strong psychological connectedness between stages is central to Parfit's account. For stages x and y, I will let `R*xy' abbreviate `x is strongly psychologically connected with y':
That is, R*xy just in case there are enough direct psychological connections between the events in x and y. Any specific formula for the threshold-- for what is "enough"-- is going to be somewhat arbitrary, so there will be borderline cases. This means that, however the threshold might be characterized, there are going to be possible cases in which there stages for which it will not be determinate where or not there is strong connectedness between those stages. We should notice that the claim that it can be indeterminate whether or not there is strong connectedness between two stages is not the claim that R* comes in degrees. To see this, suppose it is determinate that there is enough connectedness for strong connectedness between stages x and y as well as between x' and y'. But suppose as well that there is more connectedness in the first case, between x and y, than there is in the second, between x' and y'. It is important to see that this is not to say that the first case has "more" strong connectedness than the second. The two cases are equal with respect to strong connectedness insofar as the threshold is exceeded in both cases. So R* does not come in degrees. As for the formal properties, R* is symmetric if connectedness is; but clearly R* is not transitive since obviously there can be cases in which there is enough connectedness (for R*) between x and y, and enough between y and z, but not enough between x and z, as is in this diagram (where the arrow represents R*).
(3) Finally we get to psychological continuity which can be defined as the ancestral of R*, that is, letting `Rxy' abbreviate `there is continuity from x to y':
That is, R holds between two stages just when there is a chain of strong psychological connectedness linking them. This, I think, is what Parfit has in mind when he says there is continuity when there are "overlapping chains of strong connectedness" --although in my explication one chain will do (there may not be "overlapping chains," but they are possible). There can be continuity between two stages even if they are not strongly connected or perhaps even if there are not any direct connections at all between them. For example, consider the diagram presented above to illustrate that R* is not transitive. In that case we have Rxz by virtue of R*xy and R*yz. R is transitive by definition. As for symmetry, R is symmetrical if R* is. We left open the question whether or not R* is symmetric. Now a reason for treating R as not symmetric goes back to the directed nature of direct connections (perhaps because of the causal nature of these relations: later person stages arise out of earlier ones but normally not conversely). In any case, Parfit consistently assumes R is not symmetric in order to avoid R-relatedness of post-fission stages. In the fission thought experiment as diagrammed below, both y and z arise out of x; but x does not arise out of either y or z. We have stages x,y, and z such Rxy and Rxz:
Given symmetry of R, we would have Ryx, and given transitivity, we would have Ryz. Now I do not see that saying Ryz is incoherent (Perry reconstructs Grice's account so that it has this consequence--see "Psyc Crit." paper), but in any case it is contrary to what is intended by Parfit. Given that Parfit's R definitely should not hold between z and y, and since his R equally definitely is transitive, we have a reason for treating R as not symmetric (and so also, then, for going back down the line and treating all the relations R*, connectedness, and directly-connected also as not symmetric). In both 1971 and 1984 (R&P) Parfit can be understood as treating copersonality as R-relatedness among stages in the absence of "branching":
Copersonality is an equivalence relation (symmetric and transitive). Finally, "what matters," he argues in R&P, is continuity plus connectedness, which I will abbreviate as RC so that RCxy says "Rxy and Cxy". RC is neither symmetric nor transitive: it is not symmetric because R is not symmetric and RC is not transitive because of the role C (and degrees of C) in RC. Parfit's argument that identity does not matter is grounded in the fact that P (copersonality) is an equivalence relation whereas RC is neither symmetric nor transitive, as we will see. Continuity and identity of persons. The relations C (connectedness), R* (strong connectedness), R (continuity), and RC (continuity plus connectedness) have sets of person- stages as their domain and ranges. But Parfit often speaks as if persons also are in the domains and ranges of these relations --indeed his Psychological Criterion, for example, speaks this way when it says person X is continuous with Y (R&P, 207). Even though a person's existence at a time just consists in the occurrence of the events in a certain stage (so to speak), Parfit envisions a form of psychological reductionism that would not simply identify the person with the series of interrelated stages in which his or her existence consists (R&P, 211); rather personal existence is constituted by the series of interrelated stages. We formally can accommodate the sort of reductionism that Parfit envisions by supposing that there is a function F that maps any person A and time t on the person stage in which A's existence at t consists.
Various statements about persons can be constructed out of the descriptions of stages and their relations. If F(A,t)=x and F(B,t*)=y:
Moreover we also may want to talk about continuity of persons (without any mention of times):
The "Psychological Criterion" of personal identity through time (R&P, 207) can be expressed as follows in terms of "copersonality" (P) as defined above (where P is non-branching R). If F(A,t)=x and F(B,t*)=y,
Finally, Parfit says in R&P that his most important claim is that "personal identity is not what matters" (241), and that it is continuity plus connectedness that matters in survival. This can be reconstructed as follows, given F(A,t)=x and F(B,t*)=y:
The argument that identity does not matter The main argument that Parfit gives for the claim that identity does not matter (RP, 261ff.) depends on the possibility that R (and RC) branch in the fission cases. The argument is as follows: in fission one would have what matters (at least twice); but one would be identical with no one living afterwards--in anticipating a fission, one's current stage will not be copersonal with any post-fission stages.. Therefore, identity (copersonality) is not what matters in survival. For example, Frost has in ZFrost what matters in fission since RCxz is true (if `Frost' refers to the person constituted by x in the earlier "Frost fission" diagram, and `ZFrost' refers to the person constituted by z). But Frost is not the same person as ZFrost, since Pxz is not true--for while R holds between x and z, there is NOT non-branching R between x and z (and we have to deny both Pxz and Pxy since otherwise the transitivity and symmetry of P will guarantee the counterintuitive result that Pzy). Therefore, Parfit concludes, personal identity is not what matters in Frost's (or anyone's) survival. Survival The argument that identity does not matter can be found in both the `71 article and in R&P but there is a significant shift in Parfit's assumption about the relation between survival and personal identity. I believe the only important difference (for present purposes) between Parfit's views in `71 and RP concerns this shift in his conception of "survival as." The `71 position would define survival simply as RC, so that, supposing F(A,t)=x and F(B,t*)=y:
Both continuity and connectedness are necessary for survival, and it is the inclusion of connectedness that makes survival a "matter of degree" (PI, 211). Parfit in `71 also introduced the notion of a "future self" and this idea can be linked to survival. Given F(A,t)=x and F(B,t*)=y:
Notice that on this reconstruction, "survival as" and "future selfhood" are necessarily coextensive with each other, as well as with "what matters in survival":
and
On this account, neither survival nor "future selfhold" entails personal identity, and indeed Parfit suggests in `71 that the concept of "survival" should be prized apart from "identity" so that "I will survive" does not entail "there will be someone alive who is the same person as me." This would be a revision in our normal concept of "survival as" --if "survival as" entails identity. It is natural to assume the entailment, but of course `survival as' is not a commonly used locution. On this revisionary interpretation (if it is revisionary), one would survive fission even though afterwards there would be no one alive with one whom is identical. Parfit did not develop this idea in his book, R&P, however, where he does not say much about survival at all; but when he does talk about survival in R&P, his use of the concept is such that survival entails personal identity (see, for example, R&P, 261). Possibly following Shoemaker (in Personal Identity), Parfit in R&P argues not that survival should be prized apart from identity, but that what matters in survival should be prized apart from survival (and identity): one can have what matters in survival, and one would have it in a fission case, even if one would not survive. Even though Parfit shifted his position about whether we should prize survival apart from identity, in both cases Parfit argues (as in the argument outlined earlier) that we can have what matters in survival without personal identity. In both cases, that is, he holds that the best interpretation of fission will establish what Penelope Maddy called a "logical wedge" between what matters in survival and personal identity. The difference between the `71 and RP positions has to do with where the wedge is to be placed. In the `71 article, the wedge is placed between survival and personal identity (and he does not envision a wedge between survival and what matters in survival), whereas in R&P it is placed between survival and what matters in survival (and he does not discuss the earlier wedge between survival and identity). In the following diagram let > represent entailment:
The `84 wedge is between (1) and (2); that is, it denies that
entails
On the other hand, the `71 wedge denies that (2) entails
Given the reconstruction so far, we have the logical wedges in the right place for Parfit given the two divergent conceptions of "survival as": it is transitive for R&P but not for `71, but symmetrical for neither account. There are two differences between the `71 and R&P treatments of survival. The first is that R&P requires copersonality for survival whereas `71 does not. The second difference is that R&P does not require any degree of C for survival (since copersonality does not require any degree of C): the bottom-line difference in the two treatments of survival is that `71 links survival with continuity and connectedness (RC) whereas R&P links survival with copersonality (P). But, as noted, these two differences are irrelevant to Parfit's general argument that identity does not matter. For in both cases, Parfit places a logical wedge between what matters (RC) and copersonality (P), and it is this feature that grounds his argument that identity does not matter. No matter what your form of reductionism, he holds, you should conclude that identity does not matter. Insofar as Parfit wanted his argument to convince any reductionist that identity does not matter, the argument is challenged by reductionist theories, like Lewis's, which claim to drive no wedges at all between what matters in survival survival identity. Lewis's reply: the non-transitive I-relation Lewis claims that the metaphysical picture of persons that he shares with Parfit is consistent with the commonsensical assumption that there is no Parfitian wedge between what matters and identity:
The appearance of conflict is because "we make an unequal and inept comparison" (as Johnston puts it, FF 385) between (a) what matters conceived as a relation between stages and (b) identity as a relation between continuant persons. Lewis notes that it is "pointless" to compare these relations as such (21) since the relata of the two relations are different: in the one case stages, in the other continuant persons. To mark the distinction Lewis has in mind let us distinguish "stage-level" statements such as those about the relations between person stages (Cxy, Rxy, RCxy, Pxy, and so forth) from the "person-level" statements about relations between continuant persons (A at t is continuous with B at t*, A is continuous with B, A at t is the same person as B at t*, A at t survives as B at t*, and so forth--where it is important to remember that `A at t' refers to a continuant person, not a person stage). Even though the distinction is important, there nonetheless can be relationships between the two levels insofar as persons are constituted by stages and the relations between stages. (This general claim is true even if I have not succeeded in properly explicating Parfit's views; or even if I have succeeded but Parfit's views are incoherent; etc.) And, of course, Lewis does not deny that relations between the two levels can be described; for example, he freely combines them in his key expression of the commensensical position that what matters in survival is identity: "you have what matters in survival if and only if your present stage is I-related to future stages" (22). Lewis agrees with Parfit that what matters --expressed in person-stage level terms-- is "mental continuity and connectedness." There is in Lewis, however, a significant formal difference from Parfit about this relation; that is, Lewis's "continuity and connectedness" does not correspond exactly to RC so I introduce `RL' for Lewis's version of "continuity and connectedness," which can be understood as follows. Since Lewis seeks an account in which his RL will be necessarily coextensive with the relation "that holds between the several stages of a single continuant person" (21) --that is, the I-relation--and since Lewis assumes that the I-relation must be symmetric (23), he also defines RL so that it turns out symmetric. He does so by abstracting away from the "direction" of continuity and connectedness, even while acknowledging that it "has a direction" (23; note: in the following passage Lewis uses simply `R' wherever `RC' or `RL' appears; I make the replacements for coherence):
Formulated in the terms I used earlier to explicate Parfit, the forward relation is simply RC and the backward relation is RC-1, the converse of RC. Lewis is saying that he will guarantee that his version of "mental continuity and connectedness" is a symmetrical relation by defining it so that
which means
which, again, means
RL is symmetric by definition, but RL is not transitive because of the role of C: as noted earlier, there can be stages x,y,z such that Cxy (there are connections from x to y) and Cyz (connections from y to z) but not Cxz (no connections from x to z). With RL defined in this way, and then positing the necessary coextension of RL and I, it turns out the I-relation is symmetric but not transitive. The non-transitivity of I=RL plays a crucial role in Lewis's interpretation of the fission thought experiment, according to which there are two continuant persons. (Something is a continuant person if and only if it is a maximal I-interrelated aggregate of person-stages (see 22).) Lewis grants RLxz and RLxy (which entails RLzx and RLyx, given symmetry of RL); but neither RLzy nor RLyz can be derived because RL is not transitive.
In particular, we cannot reason: "RLxz, so RLzx by symmetry [that step is ok]. And [here is the mistake] RLzx and RLxy together entail RLzy." This is a mistake because it assumes that RL is transitive. So Lewis is committed to neither RLzy nor RLyz --which means of course neither Izy nor Iyz, since RL and I are necessarily coextensive; that is, there is no continuant person of whom z and y are stages. And Lewis consistently can hold that I is the relation between stages that matters in survival --assuming, for course, that two distinct continuant persons share the pre-fission stage x. On this account, there is no "wedge" between what matters (RL) and relation-I since they are defined, as above, to be necessarily coextensive. Nonetheless one might be dubious about this account as a reconciliation of the reductionist metaphysics with common sense, even bracketing worries about stage-sharing), because Lewis's I-relation is not copersonality. Copersonality, after all, is an equivalence relation (both symmetric and transitive) but relation-I is not an equivalence relation, since I is not transitive. Lewis pointed out rightly that what matters as a relation between stages should not be compared directly with identity as a relation between continuant persons, but nonetheless what matters coherently can be compared with copersonality, a relation between stages. And were common sense to consider copersonality, it would certainly (I assume) treat it as an equivalence relation: for this is just the stage-level correlate to the person-level statement that identity is an equivalence relation. In any case, the conclusion of Parfit's revisionist argument that fission shows that identity does not matter can be expressed equally well by saying that fission shows that copersonality does not matter: stages x and y are not copersonal, but since RCxy, the relation between x and y has what matters. The copersonality relation is one-one, but RC is not one-one; therefore copersonality is not what matters. Put this way, it becomes stunningly clear that rather than opposing Parfit's revisionist position, Lewis simply assumes it! That is, Lewis presupposes a logical wedge between the I-relation (that is, RL) and copersonality, P:
Since I is necessarily coextensive with what matters, Lewis presupposes a logical wedge between what matters and copersonality just as surely as did Parfit. Insofar as he uses `survival' so as to entail I-relatedness but not P-relatedness between relevant stages, the wedge that Lewis presupposes is simply Parfit's `71 wedge between survival and copersonality. That is, the relation between stages by virtue of which there is survival for Lewis, the I-relation, can hold even when the copersonality relation does not hold between those stages (since, to repeat, copersonality is transitive but the I-relation is not). This point is illustrated, for example, when Lewis says
Since RL=I, the account of survival presupposed in this statement is as follows, assuming F(A,t)=x and F(B,t*=y):
And since the right side here does not entail Pxy, survival does not entail copersonality. The commonsensical platitude that what matters in survival is identity is not in fact expressed by Lewis's claim that you have what matters in survival if and only if your present stage is I-related to future stages (22)-- this sounds good only if one assumes that the I-relation is copersonality. The platitude, properly expressed is that you have what matters in survival just in case your present stage is copersonal with (that is, P-related to) future stages. The Lewis wedge between what matters and copersonality does not reveal itself in person-level statements because if we use `Frost' or `Frost at t' to refer to the pre-fission person constituted by a certain stage x, we make ambiguous reference (since two continuant persons share stage x, there is no unique continuant person constituted by x). So even though we have both Ixy and ~Pxy in the fission case, Lewis's theory never generates person-level conflict between competing statements like the Ixy-based `Frost is the same person as ZFrost' (where `ZFrost' refers to the person constituted by z, assuming there is only one), and the ~Pxy-based `Frost is not the same person as ZFrost'. Why not? Because `Frost' is an improper description in this case. (That is, if t is the time of x, F(Frost,t) can make no assignment because `Frost' as used in that situation is ambiguous.) For these reasons, I think it must be granted that Lewis drives no wedge between what matters in survival and identity. But it turns out that this is only because all the relevant person-level identity statements that correspond to I-relation statements are ill-formed whenever there is failure of transitivity of I, for those are just the cases when continuant persons overlap (share stages). And this explains why Parfit himself failed to explain Lewis's magic. In his reply to Lewis, Parfit imagined his own forthcoming fission, and asserted the following (here Parfit uses `R' not distinguishing between `RC' and `RL'; given either reading it will be problematic by Lewis's lights, as is explained below):
The problem with this objection to Lewis is that even first-personal reflexive thoughts and statements about "me" will, for Lewis, be subject to ambiguity if I am about to undergo fission. Parfit's objection may sound plausible, but pre-fission thoughts like `I hope I have future stages' or `I am about to undergo fission' do not succeed in referring to a unique continuant person, since two (or more--the number depends on the number of future fissions) continuant persons share "my" current stage when I think `I am about to undergo fission.' [See Lewis in Postscript, page 75: a pre-fission desire expressed by "let me survive" cannot be satisfied, since "it rests on the false presupposition that they [the two people sharing the stage] are a single person." Of course Lewis's analysis here undermines his claim in the original article that:
Since these questions rest upon the false presupposition that `I' and `me' refer to a single person, Lewis is committed to saying either the questions are meaningless or should be answered no (depending on he treats questions that make false presuppositions) but either answer to these questions is inconsistent with the claim that because of RL-relatedness one will have what matters even in fission. (Lewis needs to be able to answer them yes, but he cannot do so given the ambiguity of reference.) Given his later discussion in "Postscripts," Lewis should simply withdraw the comments just quoted from the original article.' In any case, my point in this section is that I do not think that person-level objections to Lewis's account will succeed; or at least, they will not succeed conclusively since Lewis suggests plausible ways to talk coherently about pre-fission stage-sharing. Nonetheless, as I have shown, Lewis has to posit a logical wedge at the stage-level between what matters and copersonality. It is widely accepted that if one is willing to go along with pre-fission stage-sharing, Lewis succeeded in sketching a theory in which what matters and identity do not come apart in fission. While this widely held consensus is partly right, it more importantly is mostly wrong. It is partly right because indeed person-level judgments about what matters and identity do not diverge for Lewis, for the reasons explained above. Whenever transitivity of I fails there will be overlap of continuant persons, in which case referring terms are ambiguous. As for being mostly wrong, on the other hand, I have shown that for Lewis what matters can diverge from copersonality. The weird thing about this debate, then, is that Lewis already is one step ahead of the adventurist Parfit, already theorizing, that is, within a framework that assumes that identity (copersonality) does not matter, even while he (Lewis) is presenting himself as the commonsensical conservative in the debate. Later I will suggest that, in light of what I take to be the failure of Parfit's general argument that identity does not matter, and the failure of his related arguments based on fission, Parfit should emulate Lewis here and follow his own `71 advice: develop a theory in which survival is prized apart from identity (and copersonality) and exhibit practical ramifications of that sort of conceptual shift. The virtues of Parfit's original `71 account will emerge even more fully in the next section, where I explain why Lewis does not actually succeed in offering a satisfactory treatment of fission cases in terms of his symmetrical non-transitive RL-relation. Why Lewis's account of fission is unsatisfactory The non-transitivity of RL is essential for the use to which Lewis puts it in replying to Parfit about fission, as was explained. But recall that RL fails to be a transitive relation for the same reason that RC is not transitive, that is, because connectedness (C) is a component of those relations. I claim that even though Lewis consistently can deny that RLyz in the fission as diagrammed above, he lacks an explanation why RL should fail to hold between the post-fission stages y and z on the separate branches. To see this, notice that we coherently can imagine a fission case in which there is some degree of connectedness between y and z (due perhaps to an implanted memory or q-memory; see Parfit's "Venetian memories," R&P, 220). I claim that given the symmetry of RL for Lewis, there is no reason to deny RLyz in this case: after all, we have RLxz (hence by symmetry RLzx) and we have RLxy. Now we cannot simply infer RLyz, since the RL-relation generally can fail to be transitive, as noted, because of the failure of connectedness. But we just imagined that in this case we do have Czy. So Lewis cannot appeal to the role of C to explain why ~RLzy in the case. In fact, in his account, there is no reason why ~RLzy. But certainly supposing that there is the one random implanted memory connection should not make a difference in the analysis of this type of case: surely Lewis would not want to admit that given the random connection, it turns out after all that y and z would be RL- and I-related (stages of one continuant person). Before pressing my argument, let me admit that there is a flaw in what I have said so far. Even though Lewis cannot appeal to the lack of connectedness to explain ~RLyz, he can reply that RLyz fails to hold, even given the connectedness between y and z, because ~RCyz (there isn't "forward"-continuity and connectedness between y and z); and if ~RCyz (and ~RCzy) then ~RLyz. He can say this because continuity (R) itself is not symmetric: from Rxy we cannot infer Ryx, so we cannot get Ryz in this case. He can say ~RLyz because ~Ryz even though Cyz. I agree that this is a coherent reply for Lewis. But the reply rings false. It is odd for him to appeal to the non-symmetry of R (continuity) in this case after scrambling to guarantee that his RL (continuity and connectedness) would be symmetric. Recall that in introducing these notions to reconstruct Parfit's views, it appeared more or less arbitrary what we said about symmetry on the relations of continuity, strong connectedness and connectedness; but having defined continuity R so that it is transitive, Parfit had a conclusive reason for treating R as not symmetrical precisely in order to avoid R-relatedness of the post-fission stages: Parfit consistently denies symmetry on R (and consequently on Ry) in order to prevent Ryz (and RCyz) in the fission cases. Of course the non-symmetry of continuity can be motivated in terms of the asymmetry of the causal and intentional processes by virtue of which later person stages arise out of earlier ones but normally (in the absence of causal loops requiring backwards causation) not conversely. In the fission example, both y and z arise out of x; but x does not arise out of either y or z. And the point is that it is the non-symmetry of RC, not its non-transitivity, that is relevant to the failure of Pyz (and for Lewis, Iyz) in fission: just as x does not arise out of y, neither does z. Even given some degree of connectedness between y and z, and even given RCxy and RCxz, we need not posit RCyz because there is not the right sort of causal or intentional relatedness between y and z. Lewis's up-front motivation for ensuring the symmetry of RL was to have a relation that corresponded with the symmetrical I, and yet he grants that
If what matters really is symmetric, then y has it relative to x just as surely as x does to why (referring to the Frost-fission diagram again), and since x has it to z as well, it follows that y has it to z unless an inference based on transitivity can be blocked. And I claim that, while it is quite true that generally inferences based on the transitivity of RL are not valid since the relation is not transitive, the failure of transitivity of RL, when it fails, is due to the failure of connectedness. Therefore, in a case where there is connectedness between y and z, as in the imagined case, the claim that inference based on transitivity should be blocked rings false. It seems to me that it is much more plausible to suppose that RLyz fails in fission because the relevant relations are not symmetrical (and nontransitivity becomes relevant only to the denial of RL over long spans of R-relatedness --because there is not any connectedness-- as in Parfit's and Lewis's discussions of long-lived or immortal people like Methuselah). I conclude that it would be better for Lewis to deny symmetry on RL and the I-relation in order handle fission cases properly --and if he denies it for both, then he can maintain their necessary coextension. But I agree this is not yet a knockdown argument. Now for the knockdown argument. Lewis faces an additional not-unrelated problem because of the symmetry of his RL. This problem provides Lewis, it seems to me, with a conclusive reason to reject symmetry on RL and I. Lewis points out that
Call this the "future orientation of survival claim." Let us use `I*' for the relation "being I-related to future stages" (where I*xy iff Ixy and the time of y is later than the time of x). I* is normally asymmetrical: if I*xy so that Ixy and y is future to x, then ~I*yx (since x won't be future to y; I say "normally" since temporal loops are at least possible). The future orientation of survival says that what matters in survival is coextensive with I*. But I* is not RL, for RL, was defined so that RL=I, and I* is not I. As we just saw, I is symmetrical whereas I* (being I-related to future stages) is not a symmetrical relation. If Ixy and y is future relative to x, then it is not the case that x is future relative to y. And yet, of course, what Lewis says--the "future orientation of survival" claim--is correct: what matters in survival, as a relation between stages, is asymmetrical (in the absence of loops). The earlier has it relative to the later but not conversely. A person constituted by x at t survives at t* (and has what matters in survival at t*) by virtue of x being RL-related to a stage y at t* only if t* > t (at least normally, that is, in the absence of time travel; see below.) The point is that having what matters in survival is coextensive only with the "forward" -looking I*-relation. The problem for Lewis is illustrated by noticing that "future" as it is used in his phrase "your present stage is I-related to future stages" needs to be interpreted relative to personal time (not external time; see Lewis's own time travel article for this distinction). The orientation to the future in survival is orientation to one's "personal" future. We coherently can imagine time travel scenarios in which personal time could diverge from "external" future --I now might survive, for example, by virtue of stages occurring in 1856, much earlier in external time. My main point here, then, is that the symmetrical RL offers Lewis no conception of personal time. To see this, suppose there is a stage y that is future, externally, to the one that constitutes me now--that is, it is future in external time (the stage is in, say, 2005); and suppose additionally the stage is prior to me now in personal time, for I have time-"travelled" backwards, and suppose that after a couple of moments there will be no stages future to me now in personal time. In other words, after a few moments in external time, there will be no stages that are RC-related to the one that constitutes me now. Will I now have what matters in survival come 2005? No. My current stage is not properly related to any stages in 2005. True, some stages in 2005 survive now by virtue of my stage now; but I will not survive by virtue of that relation since I now have only the backwards-RC relation to those stages. My present stage will not be forward-RC-related to any stages in 2005. Of course, it now may be nice to be able to remember what will happen then; but that, obviously, is different from surviving! And yet Lewis is committed to my saying yes, I will survive: for on his account, there is a "future" I-related (that is, RL-related) stage--there is at least such a stage that is future in external time. The problem for Lewis is that the symmetrical RL as such does not order stages so as to generate personal time, and my point of course is that Lewis himself needs a nonsymmetrical I-relation to make the relevant distinction between personal and external time --if, that is, I is going to be necessarily coextensive with what matters in survival). We defined the needed I-type relation I* so that
I have just argued that "time" here should be understood in terms of personal time, not external time, so it turns out that I* just is the forward-looking subset of RL, that is, RC; so
In other words, the needed I*-relation is necessarily coextensive with Parfit's `71 conceptions of survival and "future selfhood" as these notions were explicated earlier (when expressed as relations between stages). Recall that each of these relations in `71 were treated as necessarily coextensive, as explained earlier, and as distinct from P (copersonality) insofar as the `71 wedge is placed between survival and identity (copersonality). Earlier I argued that Lewis simply presupposes Parfit's `71 wedge between survival and copersonality. Now I have argued that, in order to handle fission properly and to be able to account for the future orientation of survival, Lewis should adjust his account of the I-relation to I*, which turns out to correspond exactly with the "future self" (survival) relation in Parfit's `71. In other words, Lewis needs to presuppose exactly that allegedly counter-commonsensical account to which he was attempting to reply on behalf of common sense! Back to the future If like Lewis one is willing to abandon equivalence on a relation I* that plays the role of I for Lewis, that is, placing a wedge between I* and P,
I have just argued that it is better (relative to fission) to deny symmetry on I* than to posit it, which means I*=RC. In fission a person would survive (have a "future self") and, of course, have what matters in survival on this account by virtue of RC-relatedness, even if afterwards one would not be identical (copersonal) with anybody. There are other advantages of the `71 account over Lewis's. One advantage is that it has a concept of copersonality (P=nonbranching R) whereas Lewis does not bother to formulate one. (Recall, once more: his I-relation is not "common sense's" copersonality!) And placing the wedge between P (copersonality) and RC (continuity and connectedness=survival=future self) makes possible suggestive and more-or-less sane treatments of fission and longevity. Anticipating fission: I will survive, I will have a future self, and I will have what matters (all due to RC-relatedness) in fission; but no one afterwards will be me, since there is not non-branching R between me now and anyone post-fission (this point conforms with what Parfit sensibly argues in R&P is the best interpretation of the case). Unlike Lewis there is no stage-sharing. Unlike Parfit in R&P and Shoemaker we can say that the original person survives: indeed, the original person becomes each of the two post-fission persons; he survives as each these two persons (using `becomes' so that it corresponds with `survives as'.) If I look forward to a fission, I anticipate surviving, for I will become each of two distinct persons. That is, there will be two distinct continuant persons each of whom I will become and as each of whom I will survive. Additionally (and this perhaps is not an advantage, but a cost; but see below) one can say in anticipation of fission: I will become someone else insofar as I will survive by virtue of stages with which my current stage is not copersonal. That is, I will have a future self with whom I am not copersonal. Anticipating longevity: I will continue to exist (there is non-branching R), but after a certain point I will no longer survive, I will have no future selves, and I will not have what matters (since my current stage will not be RC-related to those later stages even though it will be copersonal with them). These seem to me to be pretty good things to be able to say about such cases. Advice for Parfit Pretty good, maybe--but not required. There are other not unacceptable interpretations of fission and longevity that say other things. Reflection on fission shows both that as persons we are not essentially unified, as Johnston says (Dancy, ed., 161), and that there is indeterminacy in our concepts of person (FF, 393). Some not-unreasonable interpretations of fission will even refuse to place a logical wedge anywhere between
such as those that say one would have none of these in fission (Sosa, Johnston) and those that say one would have all of them (unlike Lewis, Perry's reconstruction of Grice results in a theory in which there really is presupposed no relevant logical wedge, see "Psychological Criterion"). Such not unreasonable views will reject at least one of the premises in Parfit's argument that identity does not matter: (a) that you would have what matters or (b) that you would not be identical with a post-fission person; and this means, I believe, that Parfit's general argument is bound to fail. Parfit's other fission-based arguments for practical ramifications of reductionism, in R&P and in Ethics, are problematic for the same reason. Johnston suggests that reflection on fission need have no impact upon ordinary attitudes and practices, since the extensions of our concepts inspired by fission can be "quarantined" so as not to infect our ordinary attitudes and practices. I agree that since the quarantine is one acceptable response to fission, Parfit's general argument that identity does not matter would still fail even if one were for some reason able to exclude, as unacceptable, views that placed no logical wedge (that is, even if all the acceptable alternatives interpretations of fission were committed to a logical wedge somewhere between what matters and identity). While Johnston's quarantine blocks Parfit's general arguments, it seems to me that there is no reason simply to plump for Johnston's conservative minimalism either. Ray Martin distinguishes changes in attitudes and practices that are rationally required by reflection on fission and other problem cases, and those that might be rationally motivated (see 27-29 of Martin's new book). The fact that our concept of person is malleable enough to permit the family of not-unacceptable alternative interpretations of fission means that, under ordinary circumstances, one or more of the alternatives may not only break out in a rationally motivated way, but also might be useful in practice, even though its acceptance would not be required. Parfit suggested impressionistically in R&P that when he began thinking about people in his reductionist way, he found it
There are two trains of thought expressed here about the potential consequences of shifting concepts or beliefs in line with the reductionist insights:
Parfit's general arguments in R&P and in the Ethics volume that identity does not matter, that neither punishment nor intrapersonal compensation can be justified, and so forth are based on the A Train. And his focus on increased intrapersonal distance is related, I suggest, to the R&P/Shoemaker wedge between what matters in survival and survival. How so? One fails to survive in fission, but nonetheless the significant intrinsic relation one has to future post-fission stages (RC) is just the same as it is in ordinary circumstances of life: hence, there is greater distance in ordinary life than we had realized. Parfit uses this insight, for example, to attack rational egoism's "requirement of equal concern" for all the parts of one's life: since the degree of connectedness can vary, it is not irrational to apportion concern in proportion to C. The A-train increased intrapersonal distance also underlies the arguments in Ethics against the justifiability of punishment and intrapersonal compensation. It seems to me that Parfit's impressionistic comments about the ramifications of reductionism are well worth exploring even though his arguments so far have fallen flat. It is not much of an exaggeration, I believe, to say that virtually nobody has accepted as persuasive the Ethics arguments against punishment and intrapersonal compensation. What I would suggest is that a more balanced exploration of the ramifications would be possible were Parfit not stuck with the R&P/Shoemaker wedge between what matters and survival. Parfit's earlier `71 wedge between survival and identity, for example, would tend to inspire B-train insights. How so? This wedge countenances one's survival by virtue of the later existence of a person not identical with oneself. This idea decreases interpersonal distance rather than increasing intrapersonal distance. Insights based on this idea need not necessarily be radical; indeed they might illuminate ordinary aspects of life such as the ways in which people ordinarily do make sacrifices on behalf of others. The A-train runs towards nihilism [see Johnston's comments about this in Dancy, ed.], the B-train towards compassion. Fully to develop his insights about the practical ramifications of reductionism, Parfit should just emulate Lewis's example. That is, instead of trying to reconstruct the general arguments that identity does not matter, etc, Parfit should do like his earlier '71 self and simply presuppose that copersonality does not matter (as Lewis did) and then spell out potential practical and normative ramifications in terms of this account which he later abandoned in R&P. References Albert, David and Barry Loewer. "Interpreting the many-worlds interpretation," Synthese 77 (1988): 195-213. Belzer, Marvin. "Notes on Relation R," Analysis 56(1) (January, 1996): 56-62. ---. "Psychological criteria of personal identity and what matters in survival," ms. ---. Dialogues on Personal Identity and Morality (draft, 2001): see "First Night" for informal discussion of the main points of this paper. Brueckner, Anthony, "Parfit on What Matters in Survival,"Philosophical Studies 70 (1993): 1- 22. Brink, David. "Rational egoism and the separateness of persons," in Jonathan Dancy (ed.) Reading Parfit (Blackwells, 1997): 96-134. ---. "Self-love and altruism," Social Philosophy and Policy 14(1) (Winter, 1997): 122-157 Dainton, Barry. "Time and division," Ratio 5(2) (December, 1992): 102-128. Dancy, Jonathan, ed. Reading Parfit (Blackwells, 1997). Ehringer, Douglas. Time travel article .. Grice, H. P. "Personal identity," reprinted in John Perry, ed. Personal Identity (University of California Press, 1975): 73-95. Hand, Michael. "Personal identity is not transitive," ms. Johnston, Mark. "Fission and the Facts," Philosophical Perspectives 3 (1989): 369-397. ---. "Human Concerns without Superlative Facts," in Dancy, ed.: 149-179. Lewis, David. "Survival and Identity," in A. Rorty, ed.: 17-40. ---. "Postscripts," in Collected Papers ... ---. "Paradoxes of time travel," reprinted in Fred Miller and Nicholas Smith, eds., Thought Probes. Lomasky, Loren. Review of Parfit's Reasons and Persons .. Maddy, Penelope. "Is the importance if identity derivative?" Philosophical Studies 35 (1979): 151-170. Martin, Raymond. Self-concern. (Cambridge, 1998). Miller, Fred and Nicholas Smith (eds.) Thought Probes Nozick, Robert. Philosophical Explanations (Harvard, 1981). Parfit, Derek. "Personal Identity," Philosophical Review 80 (1971): 3-27. (Reprinted in Perry, ed., Personal Identity). ---. "Lewis, Perry, and What Matters," in A. Rorty, ed.: 91-107. ---. Reasons and Persons (Clarendon, 1984). ---. "Comments," Ethics 96(4) (July, 1986): 832-872. ---. "The indeterminacy of identity: A Reply to Brueckner" Philosophical Studies 70 (1993): 23-33. ---. "The unimportance of identity," in Henry Harris (ed.) Identity: Essays based on Herbert Spencer Lectures given in the University of Oxford (Clarendon, 1995): 13-45. Perry, John. "The problem of personal identity," in Perry (ed.) Personal Identity (University of California Press, 1975): 3-30. ---. "Can the Self Divide?" The Journal of Philosophy 69 (1972): 463-488. Rey, Georges. "Survival," in A. Rorty, ed: 41-66. Rorty, Amelie Oksenberg, ed. The Identities of Persons (University of California Press, 1976). Rovane, Carol. The Bounds of Agency (Princeton, 1998). Shoemaker, Sydney. "Persons and their Pasts," American Philosophical Quarterly 7 (1970): 269-285. ---. "Personal Identity: A Materialist's Account," in Shoemaker, Sydney and Richard Swinbure, Personal Identity (Blackwell, 1984). ---. "Parfit on Identity," in Dancy, ed.: 135-148. Sosa, Ernest. "Surviving Matters," Nous 24 (1990): 297-322. Unger, Peter. Identity, Consciousness, and Value (Oxford, 1990). Velleman, David. "Self to Self," Philosophical Review 105 (1996): 39-75. Stephen White, "Metapsychological Relativism and the Self," The Journal of Philosophy 86 (1989). Wiggins, David. Identity and Spatio-Temporal Continuity (Oxford, 1967). Wolf, Susan. "Self-interest and interest in selves," Ethics 96(4) (July, 1986): 704-720. |