Jy

Parfit took the A Train:

Psychological criteria of personal identity and what matters in survival

Parfit described a "reductionist" approach to personal identity through time that he called "Psychological Criterion." The psychological approach, as Parfit describes it, focusses on mental events and the relations between them ("psychological continuity") in its account of personal identity through time. According to this some versions of this approach (the "wide" approach), it is coherent to suppose that a person who exists now might continue to exist even after his or her current body had been completely destroyed by death and decomposition in the normal course of events or, in more exotic scenarios, by means of destruction in a teletransporter device that destroys one’s body as it extracts relevant information or by virtue of a sudden replacement of body parts by functionally identical duplicate parts. One’s continued existence would be possible on the psychological account, even after one’s current body had been destroyed in one of these ways, because later on there could be mental states or events standing in the "right sorts" of relations to one’s current psychology --that is, there could be psychological continuity between one’s current states and the later ones. Other versions of the psychological approach ("narrow" approaches) also give priority to psychological continuity, but hold that the physical continuity of realizers of the mental states and event also is required for identity. The psychological approach to personal identity has been widely endorsed in one form or another by many philosophers, including Locke, Grice, Quinton, Shoemaker, Lewis, Nozick, Martin, Rovane, and many others.

My goal here is to formulate a precise version of the Psychological Criterion that can be used in discussion of Parfit’s general theory of identity as well as in evaluation of his arguments for the practical ramifications of acceptance of any form of reductionism about personal identity. I first will consider a reconstruction of the Psychological Criterion by Judith Thomson which I think is not successful. Then I will spell out a form of Psychological Reductionism that uses the concepts that Parfit introduced and that gives the results concerning personal identity that Parfit argues are the "best descriptions" of a number of thought experiments.

Parfit himself does not officially endorse the Psychological Criterion of identity. (At least this is the final upshot of his discussion, although in places he seems inclined to defend it.) He says that reductionists about personal identity should not try to decide between competing reductionist accounts of personal identity since personal identity is not "what matters;" that is, that while it may seem to us that it is identity that matters to us when we wish for continued survival, in fact it is not identity that matters, but rather what matters are the relations that constitute identity under normal circumstances (psychological connectedness and continuity). Since what matters, then, is to be understood in terms of the psychological relations used in the Psychological Criterion of personal identity, Parfit endorses what we may call a Psychological Criterion about what matters in survival. Indeed, what matters are simply those relations that constitute identity in the absence of branching of these relations--and, Parfit claims, this is true even if some reductionist alternative to the Psychological Criterion of identity is true. For Parfit, this is the "most important Reductionist claim," and indeed it is chief among the practical implications of reductionism. He also uses the stories in which continuity branches to explain how many of our significant beliefs about morality and rationality are wrong or at least unsupported (Ethics volume). For this reason, the Psychological account (and what is involved in "branching" of the key psychological relations) is important, according to Parfit, even if one rejects the Psychological Criterion as an account of personal identity through time.

The significance of the conceivability of fission is that it highlights our not being "essentially unified," as Johnston has pointed out. In this paper I repeatedly will suggest that the conceivability of time travel also illustrates this central point of the reductionist picture: just as fission illustrates that we are not esentially unified dichronically, time travel shows that we are not essentially unified synchronically.

It will interesting and provocative to see that Parfit’s Psychological Criterion contrasts with Grice’s version of the psychological account as reconstructed by John Perry, according to which branching of the relevant psychological relations turns out not to be possible. For this reason, Grice’s theory will give results quite different from Parfit’s in some central cases, including "fission" cases. I will try to explain why the Perry-Gricean alternative is well-motivated and presents a difficulty for Parfit, similar to that presented by David Lewis’s alternative interpretation of fission, insofar as Parfit wants to establish practical conclusions that should be accepted by all reductionists, since his main arguments for conclusions about the practical ramifications of reductionism depend on a "no survival" interpretation of the fission thought experiments. I will try to offer a general diagnosis of the prospects for this type of argument in light of the Perry-Gricean and Lewisian alternatives. Another fission-based argument that identity does not matter can be based on David Velleman’s conception of "self"-relatedness; I also will examine this further argument and try to show why it also does not succeed. My general conclusion can be summarized as follows: "Earth to Parfit: Forget about fission!"

More precisely: forget about the fission-based arguments. I do not think it is likely any of them will work to show that reductionism entails practical conclusions. But having formulated the Psychological Criterion in a precise way, the field is left open, it seems to me, for Parfit to re-formulate his arguments for the practical implications without depending on fission-based arguments. And the fission thought experiment, along with time travel and the Combined Spectrum thought experiments, are useful in making explicit what it means not to be necessarily determinate or essentially unified either synchronically or diachronically. This is why relianace upon thought experiments like fission have been, and still remain, useful. And it seems plausible to me that there would be practical consequences, given that the intuitions of necessary determinacy and essential unity are "very strong in us" (as Thomson says is an intuition of determinacy). If we see why the fission-based arguments fail, it might help us see how to develop Parfit’s insight that reductionism should yield practical consequences.

It seems to me that Parfit’s over-reliance on fission arises from his wider approach to the practical consequences. Parfit says that reflection on reductionism led him to a sense of increased separation between the different parts of his own life and a sense of decreased separation between this own life and the lives of others. There are two distinct trains of thought here, and it seems to me that in his arguments about the normative ramifications of reductionism Parfit focusses on (A) potentialincreased separation within the life of a single person rather than (B)decreased separation between distinct people. That is why I say that Parfit took the A train.

*

Thomson’s reconstruction of the Psychological Criterion.

Thomson begins with Parfit’s formulation of the Psychological Criterion in Reasons and Persons:

Thomson realizes that Parfit’s discussion was limited and imprecise, and she suggests a number of revisions and simplifications that, for present purposes, I will accept. First, clause (5) suggests the "if" right before "(2)" should be "if, and only if" ("iff"). Secondly, she asks why clause (3) is present: doesn’t Parfit say any cause will do? Well, yes, at least in the "Widest" version; of course, Parfit does not endorse the Widest version. For the moment, though, let’s not go into this. Thirdly, it would be best to eliminate the self-reference in clause (5). So Thomson gets

Thomson like Parfit understands the phrase "psychologically continuous" in terms of a notion of "strong connectedness":

Strongly-connected-with(X,t,Y,t*)

This says that X at t is strongly connected with Y at t*. This relation holds between X at t and Y at t* when there are "sufficiently many" direct psychological connections between X at t and Y at t*. "Direct psychological connections," in turn, exist between X at t and Y at t* when X at t remembers some experience Y had at t*, or believes or wants many of the things Y believed and wanted at t*, and so on. Thomson says she will let this notion remain as vague as Parfit did.

To make clauses (2) and (4) more precise, Thomson keeps a close eye on a story that apparently involves branching of the relation of psychological continuity. The relevant stories in which continuity branches include the Star Trek episode in which the inimitable Kirk simultaneously got teletransported twice to two different planets. According to Parfit’s "best description" of this case, and according to the Psychological Criterion as he develops it, Kirk would not survive the double teletransportation, even though before the double teletransportation (his fission) he would have had stages that would be psychologically continuous with stages after the double teletransportation (more down to earth versions of fission involve transplanting each hemisphere of Kirk’s brain into numerically distinct bodies, where it is assumed that each side of the brain would guarantee psychological continuity). Continuity does not guarantee identity with the earlier person, Parfit holds, because if there were two distinct people equally continuous, as is possible, then neither can be identified with the earlier person on pain of contradiction, assuming that transitivity is a property of the relation of personal identity. Transitivity of personal identity means if A=B and B=C, then A=C (abbreviating "is the same person as" as "="). Here I use "A is the same person as B" or "A=B" to mean "is numerically identical with" so that "A=B" says that A is numerically one and the same person as B. The same phrase "is the same person as" also can be used, of course, to talk about a person’s qualities, how they would identify themselves, etc. --such as in "She is not the same person as she was before receiving the Nobel Prize," meaning she has changed a lot since receiving the award, etc. But in the Kirk example we are talking only about the numerical sense of identity --just the conditions under which one would be numerically one and the same person through time.

To illustrate the branching idea, and Parfit’s use of it, consider the following diagram.

Let the terms "ZK" and "YK" refer to the people constituted respectively be the stages z and y (for more precision about "constitution," see below). The question then is whether the original person Kirk (constituted by x) is the same person as ZK and YK? Another way to put the question is to ask where stages x and z are "co-personal." If YK=Kirk and Kirk=ZK then the transitivity of personal identity means ZK=YK. The objection to this idea is that ZK and YK might be on different planets. How could be they be one and the same person? Assuming then that they are distinct, and neither has a better claim than the other to identity with Kirk, Parfit says the best description of the case is that Kirk does not survive the double teletransportation insofar as he is identical with neither ZK nor YK (nor anyone else) after that event. Parfit’s Psychological Criterion will entail that Kirk does not survive, and he uses this result to argue that it is continuity (or continuity and connectedness), and not numerical identity, that matters to us when we wish for continued survival. Kirk does not survive fission (or, on a later version of Parfit’s argument, it is not determinate that Kirk survives) but still he has "what matters" insofar he is continuous with both guys later on.

The Kirk example introduced above can be used to explain Thomson’s proposals. Kirk is psychologically continuous with both XK and YK. In light of such cases, Thomson proposes the following truth conditions for clause (2):

Now this strikes me as an odd way to formulate the the truth conditions for (2), given Parfit’s clause (1), which says there is continuity when there are "overlapping chains" of strong connectedness. Thomson just seems to ignore (1), and in any case, her truth condition does not incorporate anything like the "overlapping chains" idea. I will return to this point below.

As for Parfit’s clause (4), Thomson notes that its presence makes the criterion circular. To spell out the "no-branching" proviso, Thomson again attends to branching cases, and about the Kirk case would say that the significant feature is that there are times t at which XK will have memories or wants or beliefs ... that YK does not have at t, so that XK is not psychologically the same as YK. [Footnote: see her quote about the Replica case, 222] And Thomson defines the predicate "psychologically-the-same-as" as follows:

She suggests that the "non-branching" idea that Parfit’s clause (4) is intended to express can be represented simply by psychologically-the-same-as(X,Y). Thus, we get

(2) psychologically-continuous-with(X,Y), and

(4’) psychologically-same-as(X,Y).

Let’s see if it works. Is Kirk=ZK true? If so, then psychologically-same-as(Kirk, ZK). Now if so, on the supposition that YK is psychologically continuous with Kirk if ZK is, by parity of reasoning, we get psychologically-same-as(Kirk, YK). But since psychologically-same-as is an equivalence relation, it would follow that psychologically-same-as(YK,ZK) which is false, given Thomson’s supposition that ZK and YK are distinct persons, with ZK having beliefs at certain times that YK does not have, and so forth. So not Kirk=ZK and not Kirk=YK. Thomson’s use of the "psychologically-same-as" relation to represent the non-branching requirement seems to work fine--if we assume ab initio that ZK and YK are distinct.

On the other hand, were we to assume ab initio that ZK=YK, then of course any psychological property of ZK would be regarded as a psychological property of YK as well. So (4’) would hold. What about clause (2)? Well, again, if we assume Kirk=ZK then there are times (before the fission) at which ZK exists and is strongly connected with YK afterwards. So Thomson’s criterion permits the truth of clause (2), psychologically-continuous-with(ZX,YK), as well as (4’) psychological-same-as(ZK,YK) and hence the truth of ZK=YK. Surely this is contrary to what Parfit had in mind in formulating his psychological criterion in terms of non-branching continuity: the criterion should be designed to preclude the truth of ZK=YK in what, after all, is a paradigm sort of "branching" case. The non-branching proviso should entail that in this case that not ZK=YK, that is, it should yield the result that ZK and YK are not identical rather than presuppose it. Thomson’s no-branching clause is just as circular as Parfit’s, even if not so obviously so.

One might attempt to refine Thomson’s criterion by explaining why the concept "psychological property" is such that it cannot be the case that ZK and YK are "psychologically the same" when they live on different planets and are, perhaps, eating different breakfasts at a certain time t*. Perhaps one could appeal to the fact that there are at t* two distinct unified sets of conscious events associated with ("constituting") ZK and YK, that is, the "stages" z and y (so to speak; I use the term ‘stages’ loosely here--for more precision, see below), in order to explain why the following could not be true:

For example, we can imagine that stage z that constitutes ZK at t* includes an experience of tasting cantaloupe but not blueberries, and a belief about eating cantaloupe, whereas y that constitutes YK at t* includes an experience of tasting blueberries, but not cantaloupe, as well as beliefs about blueberries, etc. So we might want to say that ZK has the property of tasting cantaloupe at t* (but not the property of tasting blueberries) and conversely for YK. It is not quite obvious (to me), however, how to make this idea work to build the no-branching clause. For one who posits that ZK=YK will say that in fact ZK indeed has the property of tasting blueberries at t*. (And of course, it was a short-sighted mistake to assume that ZK was not tasting blueberries simply because there happened to be no event of tasting blueberries in the stage z).

The fact that two stages overlap temporally signifies little with respect to branching of persons, as is shown by the coherence of time-travel stories in which two distinct stages of a single person overlap at a single moment, even though they are not simultaneous in "personal" time (see Lewis): in such a case the temporal overlap would not constitute branching. For instance, one might recall Michael J. Fox’s character at the end of the first Back to the Future: his later self (so to speak) watches his earlier self in the parking lot. There are two distinct unified "stages" and the Psychological criterion should be formulated so that it is coherent to speak of Fox’s character as having all of the psychological properties associated with both stages. Since there can be temporal overlap of stages of a person without branching ----I assume this should be at least conceivable for any psychological (and probably for any physical) form of reductionism about perons --I think it would be difficult to fix Thomson’s no-branching clause in terms of psychological properties the person has "at a time" because it will be unable to distinguish the temporal overlap in a time travel from that in the fission case.

A second difficulty with the "no temporal overlapping" approach to the no-branching clause is that apparently there can be significant branching, on some psychological accounts, even when there is no temporal overlap of stages. Suppose for instance, that YK is not produced until some time after ZK has finished his life and died. This is "branching with no temporal overlap"

Parfit mentions a hypothetical example of this type, in which one "branch" lives a life as a philosopher and the other branch lives the life of a psychologist, where every stage of the psychologist (after the branch point) occurs later than any stage of the first branch philosopher; also see Unger and Johnston about this type of case. If we assume the "fission point" is 2001 we can assume the philosopher lives until 2025, and the psychologist takes up in 2026 being psychologically continuous with the person before the branch point. In this type of example there will be no time when distinct competing stages co-exist, but surely any Psychological criterion that treats branching of R as possible will have to treat this sort of case as involving a genuine branch. It would at least be odd to hold that there is no branching here simply because there is no temporal overlap of stages. Even though there is no time t at which both YK and ZK exist (at least according to Parfit’s favored description of the case), nonetheless their lives are constituted by distinct streams of interconnected events each of which is continuous with the life of the original person.

I conclude that Thomson has failed to elucidate non-branching in a non-circular way. The Psychological Criterion need not and should not regard the temporal overlap of stages in time travel as a branching of continuity; on the other hand, whether or not there is temporal overlap in the Kirk cases, they both clearly involve the sort of branching of continuity that Parfit wanted his no-branching clause to preclude. Earlier I mentioned that in her formulation of clause (2), Thomson seemed to ignore Parfit’s straightforward definition of psychological continuity in terms of strong connectedness. So instead of trying to refine Thomson’s approach, I will prefer to attempt to formulate a criterion that is closer in spirit to Parfit’s.

*

Towards a noncircular psychological criterion.

Parfit sugggests that a person’s existence just consists (according to the Psychological approach) in the occurrence of a series of interrelated mental events. And Parfit is sympathetic to the idea that all the facts about a person could be given an "impersonal description," which means that all the facts could be described without using the concept "person" or assuming the existence of any particular person. The "impersonal description" thesis is controversial even among Psychological theorists who agree with Parfit on many other issues. [Shoe footnote] Some philosophers hold that the thesis cannot be made coherent. [McDowell] With this in mind I will try to develop a framework in which the coherence of the "impersonal description" thesis can be defended and challenged in precise ways. To do this, I first will try to represent the key notions in Parfit’s account in terms of the occurrence of mental events (without comment here on the plausibility of the assumptions that are needed).

The view I want to characterize says that at any moment at which a person exists there will be some mental events taking place at that moment, and the fact that a particular person exists at that time consists in facts about those events taking place at that time. Of course these facts may include facts about the relations between events taking place at that time and events taking place at other times. Since events typically have duration, a person’s existence at time t consists in the occurring at t of certain mental events with some, perhaps all, of these events beginning before t and with some, perhaps all, continuing after t. Of course some events occurring at t are not constitutive of the person, so the relevant events taking place at t need to be unified in some way at t ("unified synchronically") to distinguish those that are constitutive of a certain person at t from those that are not.

Parfit could hardly have said less about synchronic unity, but he assumes that there are certain states or events of awareness that stand in a special sort of relation, "awareness of," let us say, with events. Parfit says "states of awareness" (250) but the term "events" conforms better with a project of "impersonal description" (see 211). Let A represent this relation so that Axy says "x is an awareness of y." Various simultaneously occurring mental events are partitioned by virtue of events of awareness which (apparently by their nature) unify distinct mental events into a single entity. For lack of a better term, I will say these diverse events are unified into "incidents." Grice used the term "total temporary state" for all the experiences any one person is having at any given time, p. 86. I want to avoid an account of synchronic unity that is tied to temporal moments, however, for reasons given below, and for that reason also avoid "person stage" or "person slice" because both of these notions, like Grice’s, usually are tied to moments or brief periods of time. (Lewis, cf. Maddy re "person intervals").

Let an incident be a set E of mental events such that there is an event a in E such that for all e in E either a=e or Aae. More precisely, an incident is a relational structure <E,a,A> where E is a set of mental events, a is a member of E, and A is the awareness relation such that for all e in E, either e=a or Aae. The disjunction is included so as not to require the possibility of Aaa. In what follows I will speak of incidents as sets of events with this more complicated structure taken for granted. A person’s existence in a particular situation consists in the occurring, in that situation, of the events in at least one incident.

Shoemaker’s functionalist Psychological approach provides him with an account of synchronic unity of the mental states of a person that is somewhat different from this approach. Since the relevant states are ontologically dependent upon the existence of a person, including the various relations in which the states stand to each other over time, synchronic unity is derivative from diachronic unity. Also Shoemaker can, but this account apparently cannot, countenance unconscious mental events as constitutive of the person. This point is not as damaging to Parfit’s approach as might be thought (for a criticism see White), for it can treat such states, like unperceived states of one’s body, as causally relevant to the person even if not constitutive. And any Psychological account will treat the body upon which the relevant psychology is supervenient as causally relevant in a central way. Indeed, an alternative account of synchronic unity in terms of "co-conscious" mental events might be formulated in terms of mental events supervening at a certain time on a given body.

Assume that typically mental events have temporal duration and that, typically, all the events in a incident overlap temporally. That is, for any incident x there typically is some time t such that for each e in x, e is occurring at t. I suppose there may be no reason why Aae could not hold for some a and e even though there is no single time at which both a and e are occurring, but such strange goings on will not enter into the discussion here. It may be, of course, that the events in an incident x overlap not only with other events in x but also with events not in x, as for instances does e3 in this diagram:

----e1 ----e2

------a1

time ->

I assume this is what Parfit has in mind when he speaks of connections between mental events overlapping like "strands in a rope" (222). If x ={e1,a1} and y = {e2,a1} then x and y are distinct incidents even though they both contain e3. Because the events in x ={e1,e3} and y = {e2,e3} overlap in time it is reasonable to say that incidents x and y also overlap temporally.

The rope metaphor is a good one, and it can be used to illustrate why synchronic unity is derivative for Parfit from diachronic unity even though Parfit does not assume a functionalist account of mental events. The reason Parfit was so schematic about synchronic unity (referring vaguely to "awareness") is that he was mainly interested in the causal and intentional diachronic relations between mental events that he called direct connections. These relations include normal memory, intention, and continuity of beliefs, desires, and projects. Planning to make a loud bang at a certain time may be directly connected to a later mental event of remembering to make (and then making) the loud bang. Anticipating a future experience may be directly connected with later enjoying the experience; identifying with the subject of a humiliating or embarrassing or delightful experience may regarded as a form of direct psychological connection between the "identifying" thought and the experience. To see why synchronic unity is derivative from the diachronic unity grounded in direct connections, consider this diagram

----e1 ----e2 ----7

------a1

Suppose that e7 is a memory of e1; or suppose that e1 is a decision and e7 an action based on e1. Either way, there is a relation between e1 and e7, and it is this type of relation between mental events that is the basis for Parfit’s Psychological account of personal identity. Awareness can play a secondary unifiying role since if a1 is an awareness of both e1 and e2, then it functions to connect them even if there are no other direct connections between them.

----e1 ----e2 ----7

------a1

For this reason, we can assume that the A-relation, "awareness of," is simply one of the direct connections that needs to be included in a complete list of the significant psychological relations such as memory, planning, and so forth. Whether or not a state or event of awareness connects simultaneous events or not is secondary (in this example just given, I assumed e1 and e2 are not co-occurring). Of course it may play a special role insofar as it helps to unify co-occurring events as well; but other types of psychological relations may also be able to do that. (Grice said that he would treat two experiences E and E’ as belonging to the same "total temporary state" just in case "E and E’ would, given certain conditions, be known by memory or introspection to be simultaneous" (where being "simultaneous" meant "whatever would be meant in ordinary speech by ‘occurring at the same time’"), 88-89.) My seeing x now plays a causal role in my touching x now--is there really an event of awareness that is unifying the seeing and touching? More plausibly, these events simply are causally interdependent and co-conscious because each is a conscious event and because of the nature of their causal interdependence.

My main point here is that Parfit does not need to offer a special account of synchronic unity in addition to what is available given an ontology with mental events and relations of the type "directly connected". The notion "incident" is useful simply to abbreviate talk about how events are unified due to direct connections overlapping like strands in a rope. A cotton rope is made up of small threads of cotton twisted together. The metaphor is useful to illustrate the idea that causally interacting mental events create a structure in which synchronic unity is derivative from diachronic unity. The unity of a rope is not derived from discrete "rope segments" that are connected in some way. A rope is in this respect different from a train, with its discrete cars. If we are measuring a rope for some purpose, we might find it useful to talk about ‘the rope at point x’ ("Put the knot at x!"). In Parfit’s intended picture, what I am calling ‘incidents’ is analogous a phrase like ‘the rope at point x’ -- and it is disanalogous to the discrete cars of a train (which is why I introduced a new term instead of using the more familiar term "person stages," which tends to be used on analogy with the cars of a train). The notion "incidents" is an intermediary notion that may help us speak clearly about relations between "persons at times" (as does, for instance, Parfit’s Psychological criterion) even though a person’s existence at a time is constituted by nothing more than events and relations between those events and other events, past and future.

With these points in mind, I can proceed to give a characterization of some of the formal aspects of the key elements in the Psychological Criterion. The core idea for each is that mental events stand in various relations to each other by virtue of which they are copersonal. I will examine several different versions of the psychological criterion, where the alternatives will differ in their conception of copersonality. A non-reductionist might claim that copersonality cannot be analyzed in any terms that does not presuppose the concept of person or facts about personal identity (McDowell). The criteria that I will outline here are reductionist in the sense that the criteria do not explicitly make these presuppositions (another reason for avoiding the term "person stage" is the appearance at least that the person is conceptually primitive, and the "stages" derivative: by using the term "incident" I hope to defuse any confusion about this.) Non-reductionists who believe that no intelligible criterion of identity can be formulated in "impersonal" terms might use the framework I develop to make their case, by showing wherein lies the incoherence. Some reductionists like Shoemaker will accept a reductionist criterion as an acceptable explication of copersonality, even while denying that "impersonal descriptions" of all the facts about persons can be given (for Shoemaker the mental events themselves are not ontologically independent of the existence of persons).

It is not my purpose here to discuss the list of types of direction connections between mental events; instead, I hope to sketch the general framework of Parfit’s theory, which requires primarily keeping an eye on what formal properties should characterize the relations that are introduced to express the Psychological criterion. As we will see, relatively small differences in assumptions about the formal properties of the relations (especially concerning transitivity and symmetry) easily ramify into larger differences in overall theory; and these differences make a difference in the evaluation of Parfit’s arguments for the practical ramifications. Since the decisions about the properties at times can appear arbitrary, it will be important to take note of these points.

Let us consider, then, whether the relation "directly connected," which holds between mental events, should generally be regarded as transitive or symmetric. As for transitivity, it might seem not, since it might seem that there can be events e, f, and g such that e is directly connected with f, and f with g, but e is not directly connected with g. Thinking about making a bang might be directly connected with remembering to make it, and remembering to make it might be directly connected with later remembering having made it, but the prior thinking may not be directly connected with later remembering having made it. This point counts against the transitivity of the directly-connected relation. On the other hand, there may be a significant causal relationship between the prior thinking and the later remembering, which would count in favor of transitivity. Whether or not the relation is treated as transitive depends on how the relation is understood substantively-- for instance, whether in terms of causal relations (which favors transitivity) or in terms of similarity of phenomenological content (which opposes transitivity, since my memory of it may include information about my making it without including information about my earlier thinking about making it). As it is not my goal now to provide a substantive account of the relation, I will leave open the question of transitivity.

On the other hand, it seems clear that, like a vector, being "directly connected with" has a "direction". If so, this is grounds for saying the relation is not symmetric. It may be that event e is directly connected with f, but f is not directly connected with e. The denial of symmetry is more plausible given the causal account of the relation. An experience may cause a later memory, but not conversely, which provides grounds for saying the experience is directly connected with the memory, but not conversely. On the other hand, focussing on similarity of content would count in favor of symmetry.

While the question of symmetry, like that of transitivity, need not be fully settled at this point, we should notice that for Parfit it would be counterintuitive to treat the relation as both transitive and symmetrical. In branching cases like that of Kirk, we may have events e, f, and g such that e is directly connected with both f and g, where f is an event in the incident z on one branch, and g is an event in the incident y on the other branch.

f

e

g

Given symmetry of "directly connected" it would follow that f is directly connected with e, and by transitivity (given f is directly connected with e and e with g) that f is directly connected with g. There is little doubt but that Parfit did not intend for connectedness to hold between f and g in this sort of case, so for Parfit the relation should not be regarded as both transitive and symmetric. (Parfit will reject symmetry, as I show below.)

Let us turn now to three relations between incidents that can be fully characterized in terms of the direct connectedness of the mental events that constitute incidents. (1) Psychological connectedness is characterized completely as follows:

(Parfit speaks of persons at times being connected; we’ll get to that in a moment.) The following diagram represents the relation between direct connectedness between events and connectedness between incidents.

e f

x y

Parfit speaks of connections between mental events overlapping like "strands in a rope" (222) Connectedness between incidents is transitive if and only if direct-connectedness between events is transitive (likewise for symmetry).

Connectedness comes in degrees. The degree of connectedness between two incidents x and y depends on the number and significance of the direct psychological connections between x and y. (Parfit seems to treat connectedness as a quantitative notion, but probably it is qualitative as well, with some connections counting for more than others in determining degree of connectedness between incidents. This might be understood on analogy with the varying "weight" of norms within a normative system. But I cannot go into the details here without getting too far off track.)

(2) The idea of strong psychological connectedness between incidents is central to Parfit’s account. For incidents 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 x and y.

About the relevant threshold, Parfit says there is enough connectedness for strong connectedness if there areat least half of the mental connections that occur in one day for a normal adult, but it is difficult to believe that he intended this formula to be taken very seriously, since it is so clearly arbitrary to treat the connections over one day as the relevant standard. (Why is one day the right interval to use? Why not two days? Why not use one hour intervals as the right period of time? Or five second intervals? etc) As noted with degree of connectedness, it is equally difficult to treat strong connectedness as a purely quantitative notion. Could one count the connections? And aren't some more important than others? More important than the formula is the fact that because "connectedness is a matter of degree, we cannot define precisely what counts as enough" (206). 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 (in Parfit’s way or some other way), there are going to be possible cases in which there incidents for which it will not be determinate where or not there is strong connectedness between those incidents.

We should notice that the claim that it can be indeterminate whether or not there is strong connectedness between two incidents is not the claim that strong connectedness comes in degrees. To see this, suppose it is determinate that there is enough connectedness for strong connectedness between incidents 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 (even if connectedness is 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.

O O O

x y z

(3) Finally we get topsychological continuity which is the ancestral of R*, that is, letting ‘Rxy’ abbreviate ‘there is continuity from x to y’:

That is, R holds between two incidents just when there is a chain of strong psychological connectedness linking them. This, I think, is what Parfit has in mind in clause (1) of his Criterion 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 incidents 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 this case we do have Rxz by virtue of Rxy and Ryz.

O O O

x y z

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. As for R, 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). In any case, just as it was noted that "direct-connectedness" between events cannot plausibly be regarded by Parfit as both transitive and symmetric, a similar and decisive argument can be given here about the R relation between incidents. In the "branching" cases like the Kirk fission case discussed earlier, we may have incidents x,y, and z such Rxy and Rxz:

Given symmetry of R, we would have Ryx, and given transitivity, Ryz. Now I do not see that saying Ryz is incoherent (indeed see below for discussion of Perry’s version of Grice’s theory, which has the consequence in this type of case that Ryz), but in any case it is contrary to what is intended by Parfit.

It will be important later to have noted why Parfit is interested in guaranteeing that R is not symmetric. That continuity is nonsymmetric is what he means when he says while continuity is transitive, we would not treat it as transitive "if we allow it to take both directions in a single argument" (302). Because Parfit’s phrasing is awkward one can understand why critics have been confused--e.g. Brueckner argues that Parfit would be committed to R holding between incidents y and z. In reply to Brueckner, Parfit says "psychological continuity is only transitive when considered in one direction in time" (24). What he means here, more simply put, is that continuity is not symmetric. Again this is another awkward way to say that since R is nonsymmetric we cannot use the transitivity of R together with Rxy and Rxz to infer Ryz. (Notice, by the way, nonsymmetry is not asymmetry, which would mean we never could have Rxy and Ryx.) But in the fission cases, it is clear that not Ryz.

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). I will return below to this matter of the symmetry of R in discussing Grice’s, Lewis’s and other theories.

Continuity and identity of persons.

The relations R* (strong connectedness) and R (continuity) have sets of incidents 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 speaks this way. Even though a person’s existence at a time just consists in the occurrence of the events in a certain incident (so to speak), Parfit envisions a form of psychological reductionism that would not identify the person with the series of interrelated incidents in which his or her existence consists (see 211). 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 incident in which A’s existence at t consists.

A is constituted by x at t iff F(A,t)=x.

We can use this idea to characterize the following:

Moreover we also may want to talk about continuity of persons (without any mention of times):

This is how statements about the continuity of persons can be constructed out of the "impersonal descriptions" about mental events and their relations-- assuming of course that the mental events and relations themselves can be given "impersonal descriptions (see below).

To leap ahead, we a can state the General Psychological Criterion at this point as well. Let Ixy say "x is copersonal with y."

The General Psychological Criterion (GPC) of personal identity:

F(A,t)=x and F(B,t*)=y and Ixy.

I skipped a step, of course, since we have not yet said what it means for Ixy. But this general Criterion is useful because we can use it to identify the differences in different forms of the Psychological Criterion, which will characterize the "copersonality" relation in different ways.

Parfit’s basic idea, of course, is that copersonality of incidents corresponds to "non-branching" psychological continuity between them, that is, incident x is copersonal with y (Ixy) if Rxy and there are "no significant branches" in the R-chain from x to y. Suggesting that the next section can be skipped without loss of continuity would be in the spirit of what Parfit says is his most important claim (241): that is relation R that matters in survival; and that "personal identity is not what matters" (241). The main argument that Parfit gives for the claim that identity does not matter (261ff.) depends on the possibility that R branches in the fission cases (although other arguments that Parfit has given for this claim depend on weaker claims, for instance, that it is "indeterminate" whether R can branch; I will discuss these differences below). Therefore, it may be important to know how the details about "nonbranching continuity" can be explicated. It may not be necessary to follow the tortorous details themselves, however, in order to follow the later discussions about "what matters."

What then aboutnon-branching continuity? Given continuity "from x to y," Rxy, I first will offer a non-branching proviso based on the idea that there is no branch from x to y (Nxy) if there is a linear ordering of all incidents "between" x and y. To begin, let us look once again at the Kirk story.

In this case we have both Rxy and Rxz. Assuming both that we ultimately want the criterion for personal identity in terms of non-branching R to entail that not ZK=YK (where as before ZK is constituted by z and YK by y) and that we want our criterion for identity just to be non-branching R, we know that we want an analysis according to which there is branching in this case. Why is there branching here? Well, obviously, because of z, where Rxz but not Rzy. This suggests that, generally, there is "no branch from x to y" (Nxy) only if there is no z like this one; that is (given Rxy):

Nxy only if for all z, if Rxz then Rzy.

This isn’t sufficient for Nxy, however, since there could be some w such that Rxw and not Rwy but nonetheless Ryw as in the following diagram.

R

x –– y –– w

R R

In this simple case there is an "overlapping chain of connectedness" (since both Rxw and Ryw are true, but not Ryw) but not a branch--at least, it obviously is not the sort of R branch that should defeat copersonality of the three incidents mentioned in this simplest of cases. To take care of this case, we can add a disjunct as follows:

However this proviso is too weak. For example, as in the following diagram, each z where Rxz is such that Rzy, yet there is a branch in R from x to y.

The problem is that Nxy does not yet guarantee that R determines a linear ordering of incidents from x to y. So add a second condition (ii) in case Rzy;

This guarantees that when Rxy, R determines a linear ordering of all the incidents z on any R-path from x to y. However, the condition still is not sufficient for Nxy since there could be other incidents z that may be relevant where not Rxz: the proviso is still too weak to guarantee there is no branch from x to y since there can be "fusions" as well as fissions that represent branches from x to y, as represented here:

/

z

Now we can preclude these "fusion" frays by requiring that when Rzy either Rzx or Rxz, as follows.

I claim, finally, this condition is necessary and sufficient for Nxy, the "strict" nonbranching proviso when Rxy.

There is non-branching continuity from x to y just in case Rxy and Nxy.

This, of course, is the notion we sought. We can say that

Incidents x,y are copersonal (that is, Ixy) iff (a) Rxy & Nxy or (b) Ryx & Nyx.

Call this the Strict Parfitian (Strict P) Criterion of copersonality. It is "strict" because of how it contrasts with the more lenient Parfitian criterion developed later. Given the General Psychological Criterion stated earlier it follows that A=B just in case there are times t, and t* such that F(A,t)=x and F(B,t*)=y and (a) Rxy & Nxy or (b) Ryx & Nyx.

A no "significant" branching proviso.

This isn’t quite the Criterion Parfit had in mind. This criterion is based on the intuitive idea that R determines a linear ordering of all copersonal incidents. However, Parfit suggests (and it is plausible to suppose) that the psychological criterion should be more lenient than that (even if not as lenient as PG which does not recognize bother with branching insofar as branching is impossible). There are some cases that have the forms which we have just considered, where strictly speaking R branches, but the branching is not significant enough to preclude copersonality. The cases in which Parfit says the "best description" will violate the strict no-branching clause include the Sleeping Pill and Physics Exam cases.

In the Sleeping Pill case (287-288) one takes a sleeping pill after which one will fall asleep in about an hour, but later will have no memories of the thirty minutes before one fell asleep. This case and other cases of retrograde amnesia have the form of the cases in the the Daydream family of cases that worried Grice.

The branch from v to z represents the thirty minute branch. A less dramatic version of Sleeping Pill would be a case like one in which I am knocked out for a few seconds after getting hit in the head and afterwards cannot recall anything about the five minutes prior to getting hit. On the other hand, an even less dramatic case than getting momentarily knocked out would be one ("Daydream") in which after being lost in daydreams for a few minutes I have no idea what happened to me during that period of time even though I may vaguely recollect being conscious. If we assume that there is no psychological continuity between incidents during that period (moments represented by z in the diagram) and those right afterwards, as might be plausible in some such cases, then the Criterion as it now stands would have to deny that one survives such bouts of daydreaming.

If our existence is constituted by connections between mental events that overlap like strands in a rope, these "Daydream" type of cases suggest at least that the rope can be frayed. Indeed, even in the Branchline case in which there is temporal overlap of incidents, because a replica is fashioned on Mars while one still seems to exist on earth (but only for a few minutes) Parfit agrees with Nozick about such cases that brief overlaps "cannot be rationally thought to have much significance" (see Parfit, 289): "I need not assume that my Replica on Mars is someone else" (288). However reflection on these types of cases do not lead Parfit simply to abandon symmetry of R, for he claims, "if the overlap was large, this would make a difference" (289). In the Branchline case, I need not assume non-identity because the temporal overlap is brief and the disunity modest (otherwise it would be a branch undermining identity). Presumably he would say the same thing about cases involving no temporal overlap but where the branch is quite long: I have heard of actual cases in which this period of time was as long as two years. There are going to be incidents on the branch that are not "closely enough" connected to the main line branch to be copersonal. (But perhaps it is the overlap that matters? Or perhaps just that there is overlap of a future competitor?) Of course Parfit will want to argue that all of these details are secondary to the fact that identity itself does not matter; whatever we say about the replicas in Branchline cases we can know that one would have "what matters" even if it is false that, or indeterminate whether, one survives. As we will see, however, the details about these issues will be important in the evaluation of his arguments--this, at any rate, is my reason for going into this technical wilderness. (Before I am done, the complexity of the account may give as a reason simply to embrace the Perry-Gricean account which I describe later and which just embraces symmetry of R to deal with these cases rather than engage in the present sort of ad hoc gerrymandering! --But I digress.)

As for the Physics Exam case, which also is relevant here, Parfit imagines one can "divide" one’s mind at will so as to work on two distinct sets of problems at once.

\ /

z’....

Parfit says that "given the brief and modest nature of this disunity, it is not plausible to claim that this case involves more than a single person" (248). A person’s mental history is like a river, "occasionally having separate streams" (247). But in this case R does not order all the relevant incidents, so the Strict Parfit-Shoemaker Criterion rules out copersonality. It is a genuine branch and according to the strict Criterion, the person goes out of existence; now we want to build in more leniency.

To handle all these cases for theories with a nonsymmetrical R, such as Parfit’s and Shoemaker’s, I will revise the characterization of copersonality by replacing the the "no branching" proviso with a "no significant branching" proviso.

First let Qxz represent the idea that the degree of connectedness between incidents x and y passes some threshold so that x is not too weakly connected with y to be copersonal. Perhaps Q is just the same relation as R* (strong connectedness), but I will not assume so. That is, I will not assume that the threshold for determining whether Rxz is a "significant" branch or an "insignificant" fray is just the same threshold as for strong connectedness. Let Q be symmetric but it definitely is not transitive.

We formulate the idea that "there is no significant branching from x to y" (N*xy) by building qualifications into the characterization of Nxy (the strict no branching criterion). First add to our strict no branching proviso an escape clause using Q, as follows. Assuming Rxy:

This additional handles some Daydream cases, since if Qxz, z automatically is regarded as close enough to x to treat Rxz as insignificant even though neither Rzy nor Ryz.

But this won’t quite do, since Qxz may be false in the case as represented in the Daydream diagram (x and z are not "close enough"), and yet if Qvz is true, for some v on the R line between x and y (as depicted in the diagram) then the branch off of the R chain connecting x to y is not significant. So we replace the (c) disjunct Qxz with the more general statement that there is some v such that Rxv and Rvy and Qvz (the more general statement is true if Qxz, given reflexivity on R); that is, given Rxy,

Of course "fusion" frays also need to be considered for they also may not be significant:

/

z

These can be handled by a similar addition to the clause, clause (ii), that governs fusions:

So much, then, for insignificant branching in the Daydream cases. What about the Physics Exam? Well, any incident occurring on either of the separate streams can be regarded as a fission fray off of x or a fusion fray into y. So if I have dealt with fission and fusion properly, and if the streams themselves in the Physics Exam are taken to be insignificant, any incident within either stream will be close enough to either x or y so as to be treated as insignificant by the characterization of N*xy. Being separated "too much" is represented here by the idea that there is some incident on one of the branch streams that is not strongly enough connected to any incident on the mainline.

With N*xy then representing the more lenient notion "no significant branching" we can say

There is no significant branching from x to y (Bxy) just in case Rxy and N*xy.

I now will use this idea to construct a characterization of copersonality that is more lenient than the "strict" one.

Unfortunately we cannot define copersonality simply as continuity with no significant branching, that is, we cannot define copersonality (Ixy) directly in terms of Bxy (such as, as one might have hoped: Ixy iff either Bxy or Byx). For reconsider:

and focus on the relation between x and y here. We have neither Rxy nor Ryx because of our nonsymmetric R. The problem is that we wanted x and z to turn out copersonal (indeed that is why we trying to make the criterion more lenient). Despite the fact that Bzy (there is no significant branching from z to y, since Rzx is not "significant") we have neither Rxy nor Ryx in the case as depicted here.

To be able to treat x and y as copersonal in this type of case, let us first record first that P*xy holds if there is no significant branching between x and y or vice versa:

P*xy if either Bxy or Byx.

Now we want P*xy also to hold if as in this case there is some z such that Bzy and Qzx. Equally with fusion we want P*xy in the cases depicted like this:

/Q

y

That is, we want P*xy if for some z, Bxz and Qyz. The idea here is that if Bxz (so x and z are going to end up copersonal), then "close enough" incidents off of z should be copersonal with x and y. Summing up so far we have:

P*xy if Bxy or Byx or there is a z such that (Bzy & Qzx) or (Bxz & Qyz).

Now to get a symmetrical P* (so that P*xy iff P*yx) we add two conditions as counterparts to the last two disjunctions which guarantee symmetry and make the condition necessary and sufficient for the desired copersonality relation:

The relation P* holds between each of the incidents in the following diagram:

P* is both transitive and symmetric, as can be proven. As for reflexivity, P* here is reflexive if the "no significant branching" relation is; and it is reflexive if we assume the reflexivity of R. So copersonality is an equivalence relation and we can be sure, at least, that it meets this minimal condition for copersonality (as normally conceived; but see below re Lewis and the nonsymmetric criterion) even while allowing for Gricean frays in the rope of R-interconnectedness.

The Lenient Parfitian (Lenient P) Criterion of copersonality:

Incidents x,y are copersonal (Ixy) iff P*xy.

And given the general Psychological Criterion stated earlier it follows (for the lenient P criterion) that A=B just in case there are times t, and t* such that F(A,t)=x and F(B,t*)=y and P*xy.

*

Copersonality in Grice’s theory.

One of the influential earlier Lockean accounts of the co-personal relation between (what I am calling) incidents posited the symmetry of a relation very much like strong connectedness, and as a result, this theory denies the possibility of "branching" of the relation that plays the role of R. In Grice’s theory (as reconstructed by John Perry) copersonality turns out to be necessarily coextensive with R, and it will be useful to examine that theory before turning to Parfit’s arguments that identity does not matter.

Grice did not distinguish between what Parfit calls connectedness and strong connectedness, and the only type of direct psychological connections he considers is memory. These differences are not important for present purposes, I think, and I will assume that Parfit’s additions are important improvements in Grice’s theory. The memory relation plays the role for Grice, roughly, that strong connectedness plays for Parfit, with the ancestral of the memory relation corresponding to Parfit’s R (which, recall, we defined as the ancestral of R*). Suppose there is a memory connection between incidents x and y, that is, y contains or could contain a memory of an experience in x. Obviously this "memory connection" as such is not symmetrical, and neither is its ancestral. Grice worried that the ancestral could not be used in the criterion for identity because it was not symmetrical. He was concerned about the possibility of there occurring experiences that could not for some reason be subsequently remembered, in which case the memory connection (the only relevant psychological connection for Grice) cannot hold between them and later experiences.

John Perry dramatizes Grice’s worry in the "senile general" case in which a general can remember experiences as a boy, as can the young officer, but the general can remember nothing about his experiences as a young officer. (Notice that this thought experiment is different from Reid’s original senile general story in which the young officer could remember being a boy, and the general could remember being the young officer but not the boy -- a problem that is solved by defining R to be transitive since it is is the ancestral of strong connectedness, or in Grice’s case, the ancestral of the memory relation.) The description of the senile general case is similar to that of some actual cases of Alzheimer’s Disease (AD) in which people report they seem to retain long term memory but seem to have lost more recent memories. They can remember childhood experiences but are unable even to recognize their adult children. Another type of case involves repressed memories of traumatic events (Grice expresses his general worry in terms of the possibility of an experience that "even a prolonged process of psycho-analysis would not bring about the occurrence of a memory" of which (87-88).) Indeed the cases that worried Grice are just those "Daydream" types of cases which motivated the move from the strict or lenient Parfitian no-branching proviso. A more mundane example is the situation of someone who is knocked unconscious and loses all memories of what happened during a period of time before getting knocked out. A yet even more mundane example that arguably is of the type that worried Grice involves short periods of time when someone is idly daydreaming and later has no memories of that period, and later performs no actions due to intentions formed during that period, and so forth. Each of these cases has this structure:

(Here I use "RG*" to represent Grice’s memory connection since this relation in his theory corresponds to Parfit’s R*, strong connectedness.) There is an important problem for the Psychological theorist if one thinks, with Grice, that the person constituted by y may (at least in some cases of this type) be same person as the person constituted by z. In y there are memory connections to x, and in z to x; but there are none either way between y and z.

Grice’s solution then is to define a relation which is like memory except that it is symmetrical (see 88; also see Perry, his relation disjoins the memory relation with its converse, allowing us "in effect, to change directions and thus ignore time order" when linking incidents)-- that is, Grice imposes symmetry on the relation RG* (and this, then, officially is his relation that corresponds to R*). Let us use "RG" to refer to the ancestral of RG*; then RG, which corresponds to Parfit’s R, also is symmetrical. It may be that no later incidents whatsoever will or could contain memories of experiences in y in the cases represented in the above diagram, but y nonetheless can enter into a symmetrical RG* relation because it does or could contain memories of experiences in other incidents (like x); and since GriceR*xy Grice has RG*yx because of the symmetry of RG*. Where RG is the ancestral of RG*, we also have RGyx; and since RGxz and since RG by definition is transitive, RGyz as well.

Perry's formulation Grice's criterion for copersonality of incidents is such that any two incidents are co-personal just in case both are in some "Grice set" where a "Grice set" is a set of incidents (again, to use my term for what Grice called a "total temporary state") which can be put into a sequence (not excluding repetitions and not necessarily ordered temporally) so that each member of the sequence is RG* related to the next member and to which no more incidents can be added in the sense that there is no incident that both is not in the Grice set and yet is RG*-related to some member of the set. This means that any two incidents will be co-personal just in case RGxy (where RG, recall, is the ancestral of RG*). On Perry’s interpretation of Grice, the person A constituted by y in the Daydream type case may indeed be one and the same person as B, constituted by z (that is, A=B). For there is this RG*-related sequence

y-x-z

by virtue of which {x,y,z} constitutes a Grice set. (Notice that the sequence

z-x-y

also could be used to show that {x,y,z} is a Grice set.) Not only do we get A=B in the Daydream family of cases. We also get the result that ZK=YK in the Kirk fission case, since RG and copersonality are necessarily coextensive. Indeed "branching" of RG is not possible on this account!

"Fission" cases are the most plausible examples of branching but for Perry-Grice’s symmetrical RG we can always get a sequence of the necessary sort to get copersonality of all relevant incidents.

y

/

w --- x

\

z

Since RG* and RG are symmetrical we can always get a sequence of the necessary sort for copersonality in these cases, for instance

z-x-w-x-y

so that {w,x,y,z} counts as a Grice set and all the incidents are copersonal. ZKirk and YKirk would be R-related in the Kirk story, according to this theory, and therefore copersonal. And so would all the incidents of Aulisio’s friend (who "lost" two years after the basketball game), including those in the two year period that later were not remembered. R just is the copersonality relation for Perry’s Grice; that is, using the apparatus introduced above, we can state the

Perry-Gricean (PG) Psychological Criterion of copersonality:

Incidents x and y are copersonal just in case RGxy.

Notice that RG can be defined in terms of Parfit’s R as follows:

RGxy iff Rxy or Ryx.

So this Psychological account does not need a "non-branching" clause in its criterion of identity, since it treats RG itself as copersonality. And given the General Psychological Criterion stated above, the Perry-Gricean theory holds simply enough that

A=B just in case A is continuous with B.

This account is, I believe, a viable and well-motivated Psychological criterion.

People do not always balk at the idea that one could survive (using ‘survive’ to entail personal identity) with "two bodies and a divided mind" in the fission cases. Even Parfit grants this could happen in some cases, as we have seen.

First, the transitivity of = can be denied. Arthur Prior suggested giving up transitivity on personal identity to handle this type of thought experiment. John Perry also explored its denial, David Lewis presents a theory that entails its denial, and Michael Hand argues for it nowadays. [Also Dainton?? In the time travel article?] I will return to discuss this view later. For the moment, however, notice that one need not deny transitive on the personal identity relation in order to dispute the no-survival interpretation of fission. One can say that say that despite appearances, the two guys after the double beaming both are Kirk, and they are identical to each other, as according to the PG criterion. Although it looks like there are two distinct people, there is just one person, Kirk, but he has two bodies afterwards-- one on planet Z and one on planet Y. It is true that one person having two bodies and a divided mind would be quite weird, but Parfit goes too far in claiming it would require a great distortion of our concept of a person (256; see Johnston). For instance, after some years, the two people might meet again and end up playing tennis together, and having aged, maybe a face lift or two, not even know that the other woman she’s playing tennis with just is herself!

However, the time travel cases show that idea of one person with two bodies and a divided mind is at least coherent (even when the disunity is not "brief"). In time travel you could have one person with two bodies playing tennis with herself without realizing it. So the idea at least makes sense. Strictly speaking nobody need be regarded as going anywhere (as might be pictured if one imagines the person as a "separately existing entity"). As Lewis explained this, one would need backwards causation, so that events in "external" time can be said to cause other events earlier in external time. So you now could "remember" something that will happen an hour from now. Given backwards causation (events at t causing earlier events at t-1) there could be two incidents of one and the same person coexisting at the same moment in external time, even though one of the incidents is later than the other in the person’s "personal" time, as in the Back to the Future scenario

McFly’s birth_______z__x

y______________________z*________

<- external time ->

The key is that even though event x is later than event y in external time, both (a) x is a cause of y (so there is where we need causation backward in external time), and (b) x is earlier than y in McFly’s "personal time" (so y could for example contain a memory of x). And, of course, z and z* can represent the two incidents of McFly overlapping at one moment in the parking lot near the end of the first Back to the Future.

One can object that the use of this type of story is irrelevant to interpretations of fission since in this story all of McFly’s incidents can be lined up in a linear ordering within his "personal" time, whereas in the fission story, neither ZKirk nor YKirk at any given time would be earlier or later than the other since they are regarded as being on totally separate streams. This is true, but the relevance of time travel is to show only that Parfit’s specific objection to the survival interpretation of fission fails (based as it is on the alleged incoherence of playing tennis with oneself).

A more effective objection to the survival interpretation of fission, it might be suggested, is that after the fission there are two distinct centers of consciousness and action. Recall from that what unifies a person at a time ("synchronically") are simply relations between various events co-occurring at that time. While there is no substantial, determinate "center" of consciousness, Parfit treats "awareness" as the fundamental unifier at a time. Of course, synchronic unity of incidents might be achieved by means of the idea of "supervening at a moment on a particular body"--given the assumption that mental event’s supervene on bodily events, the body upon which one’s mental events supervene at a certain time might be used as the principle of synchronic unity, even for a Psychological account. But whatever the details about synchronic utnity, the time travel example show that Parfit would have to abandon the idea that synchronic unity of incidents is what unifies a person at a time -- and more generally, we must abandon the idea that what normally unifies a person at a time is always decisive as to the extent of a person’s existence at at time (even if, for example, the body plays a role in synchronic unity). Whether incidents are unified by means of awareness or by means of the body upon which mental events supervene or by some other means, the time travel examples show that the unity of incidents cannot be the full story about the unity of a person at a time, since in these stories we assume that a single person is constituted by two temporally overlapping incidents. After a time travel, there might be only one person on the tennis court at a certain time, yet a match is being played. There are two unified sets of mental and physical events. The Psychological Criterion says awareness events are doing the unifying in some way on each side of the net yet that cannot be the bottom line basis for unification of the person, since there are two distinct awareness unifying events (or two distinct human bodies playing the unifying role) but there is only one person. What the time travel example shows is that the unification across times ("diachronic" unity, that is, R or non-branching R) can take priority for personal identity over whatever unifies a person synchronically.

In deciding what to say about fission while one can give priority to a certain conception of synchronic personal unity that treats the synchronic disunity of the fissionees as decisive. But one need not do that--and the time travel case shows that one probably better not do it in all cases. We can give priority to the diachronic unifiers, namely, R-- which is what we have to do to make sense of the coherence of time travel. And if we do that, if we say diachronic unity takes priority when we make sense of the possibility of a time travel case, then it is at least coherent to do something similar in thinking about fission. Whereas awareness unifies each of the two incidents, it is R that unifies the person. And perhaps (as for Perry’s Grice) R should be defined so that the two incidents x and y in the above diagtram turn out copersonal. Even though there certainly is synchronic disunity for Kirk after the fission --there is no single unity of all the events that constitute him at various times after the fission --nonetheless one can maintain that ZKirk = YKirk = Kirk. It seems to me that one who would argue against survival of fission has to assume that unity at a time is more important than R, and for the reductionist without a simple, indivisible substance in the picture it is not clear why one should assume that.

There are some Psychological theories, of course, for which synchronic unity is determined by the diachronic unifiers. Shoemaker achieves unity of the person, both synchronically and diachronically, by virtue of the functional integration of the person through time. He has a holistic view, deriving from functionalism about mental states, in which memories and intentions and so forth exist only by virtue of their functional roles in the person.

Shoemaker’s account raises a problem for the claim that the time travel cases are relevant to our thinking about fission. A normal form of functional integration would include the connection between one’s deliberation and decision-making with what happens, with what one does, and so forth--a functional integration that would be lost after the two Kirk streams diverge--for the deliberation by ZKirk would not play its normal causal role in what YKirk does. In the time travel case, the synchronic disunity of the person is not really a problem for Shoemaker; there are two distinct incidents coexisting at certain times, but since one is later than the other in personal time (even though they are simultaneous in external time), he can consistently posit all the normal functional relationships between plans and actions, and so on. So the planning in one of the incidents might actually be having effects simultaneously in the other incident. And thiis, it can be argued, is the deeper point underlying the importance of the linear ordering (and the refusal to countenance copersonality given branching).

However, serious problems with Shoemaker’s theory, and with any form of Psychological reductionism like Parfit’s that countenances the possibility of branching. For there are possible and actual cases in which intuitively we probably will wish to grant survival but which on Shoemaker’s and Parfit’s theory cannot be regarded that way even after we have engaged in the gerrymandering in which we did engage to construct the "no significant branching" clause. Indeed the problematic cases for Parfit and Shoemaker are precisely the cases that worried Grice. As we saw, Grice had a psychological theory much like Parfit’s Psych. Criterion, a purely psychological-based theory, and he was worried that some experiences (in the Daydream/Alzheimer family of cases) couldn't be regarded as mine because at no later point were they retrievable in any way. They might be R-linked to earlier experiences but he imagined they’re not R-linked to any later ones.

Diagram. Aulisio/Alzheimer’s cases

One case of this type may involve a two year branch, say, after a guy got hit on the head playing basketball. Now even if close examination of cases of this type reveal there is enough psychological connectedness of some sort so as to yield psychology continuity in this type of case, we can imagine hypothetical cases that are very similar to the actual ones in which there is no connectedness. Given the possibility of branching, we certainly have a genuine branch in such cases.

Yet it is difficult to believe that Shoemaker or Parfit would be happy to agree that Aulisio’s friend did not survive the basketball injury, or to speak more precisely (and even more problematically) that he did not survive the moment two years before the basketball game at which the branching took place. As noted, Parfit grants that some brief branches may be possible, which is why ultimately his criterion is not exactly "continuity with no branching" but rather "continuity with no significant branching". But if the two year branch in the Aulisio case is treated as "insignificant"--that is, if all the incidents on the branch are regarded as "close enough" so as get Qvy in this case -- then it seems to me that the "no branching" clause is simply idle; two years is not "brief". (Or if it is, just change the example: see Nozick about this.) So I assume that Shoemaker and Parfit are committed to significant branching in the two-year Aulisio case --it is an actual fission case.

Of course, even if identity is lost --there is no survival--Parfit and Shoemaker might say this is where the "identity does not matter" thesis comes to life. It is another case that shows that it is not identity that matters to us, insofar as we want to survive. Aulisio’s friend actually went out of existence two years before the basketball game, due to the R-branch back to that point, and of course that is especially surprising since it may very well be that nothing particularly unusual happened at that time. But nonetheless he still had "what matters" insofar as later on he had two R-related streams.

The Aulisio case is just the sort of case that worried Grice. The reason that the Perry-Gricean alternative to Parfit’s criterion is important, it seems to me, is that Grice set out to construct a theory of personal identity which is very much in the spirit of Parfit’s Psychological account. Along the way he found reasons based on cases like Aulisio’s to treat the central relation (what I am calling "RG") as symmetrical. Whatever may be the overall merits of the symmetrical RG as opposed to the nonsymmetrical R that Parfit presupposes, it is only this technical difference in the two approaches that leads to the radically differing accounts of those fission cases about which they disagree. (PG’s disagreement with Shoemaker is more substantive, mirroring Parfit’s disagreement about the possibility of impersonal description of all the facts about persons.) The fission cases figure prominently in Parfit’s discussions of the practical implications of reductionism about personal identity, including his arguments that identity does not matter.

*

Parfit’s first argument that identity does not matter

His argument has taken two different forms. Both forms of the argument depend upon describing circumstances in which fission is used to attempt to develop what Penelope Maddy calls a "logical wedge" (p.151) between identity and the relation that matters (or should matter) insofar as we wish to survive into the future.

The first form of the argument depends upon the premise that one would not survive fission (here I put the argument in terms of the Kirk fission example):

(1) Kirk’s relation to zKirk contains what matters.

(2) Kirk’s relation to zKirk is not identity.

Therefore,

(3) Identity is not what matters.

Of course the PG criterion regards premise 2 as false. Since Kirk is continuous with zKirk, Kirk=zKirk according to the PG criterion and the non-identity premise 2 is false. Insofar as Parfit wanted to establish conclusions that hold for any form of reductionism, this form of the argument fails.

Parfit has acknowledged in a paper subsequent to his book that this form of the argument cannot succeed. He grants that he should not appeal to premise 2, since in a fission case "there is no special further fact which would make it determinately true" that Kirk would be neither of the two resulting people:

Here I believe that Parfit acknowledges that his argument should not utilize a premise like 2 which simply asserts that the original person fails to survive.

The other form of Parfit’s argument that identity is not what matters is not so easily dismissed, since it attempts to utilize the idea that there is no fact of the matter according to which competing criteria of identity can be factually mistaken. Premises 1 and the conclusion 3 are as before, and 2 is replaced by 2*:

In order to evaluate Parfit’s second argument, it will be useful first to sketch two additional forms of the Psychological criterion, both of which agree with PG that one could survive fission. With these sketches before us, we will be in a good position to understand what Parfit means by saying it is an "empty question" whether Kirk’s relation to zKirk is identity.

*

Copersonality in Lewis’s theory

David Lewis like PG holds that copersonality and the central "R" are necessarily coextensive. However Lewis’s "R"-relation is symmetrical as is his copersonality relation (the "I-relation"); but neither is transitive (see Lewis, 23-25)-- which means his "R" differs from both PG’s and Parfit’s. I will refer to Lewis’s "R" relation as ‘RL’ to make sure it is distinguished from R as defined above. PG and Lewis differ in their conceptions of copersonality. To explain this, we should notice first that Lewis uses ‘R’ simply to mean "the relation that matters in survival" (20), by which he has in mind "mental continuity and connectedness" understood in terms of bonds of similarity and lawful causal dependence (17). While Lewis refrains from going into details, he says enough so that we can distinguish RL from R. An attractive simple suggestion is that Lewis’s phrase ‘mental continuity and connectedness’ means exactly the same thing as Parfit’s phrase ‘psychological continuity and connectedness,’ in which case Lewis’s R-relation (RL) is related to Parfit’s R in the following way: RL is continuity (R) plus connectedness. This is not exactly correct, however, since Lewis explicitly formulates his R so that it is symmetric (whereas recall that for Parfit neither R nor "connected" are symmetric). The correct formulation of RL in terms of R, then, is as follows:

This definition of RL not only mirrors Lewis’s phrasing in a fortuitous way, but also corresponds to his use of ‘R’ insofar as it explains why he should deny transitivity on RL: although R is transitive, RL (as just defined) is not transitive. To see this suppose we have both Rxy and Ryz (and hence Rxz) and suppose there is some direct connection between events in x and y (henced connectedness), as well as between y and z. Yet it is consistent to suppose that there are no direct connections between x and z. Therefore RL is not transitive. The role of connectedness surfaces again for Lewis insofar as RL can be a matter of degree (32). Since Lewis is committed to treating RL and the I-relation (copersonality) as necessarily coextensive (and this is the basis for his response to Parfit’s argument that identity is not what matters), clearly Lewis requires some degree of connectedness, in addition to continuity, for copersonality. This is where Lewis differs from PG (since PG requires only continuity for copersonality). Given the translation from Lewisian into Parfitian terms (for instance, referring to incidents rather than person stages and accepting, at least arguendo, the Parfitian architecture about connectedness, strong connectedness, and continuity --I take it these are among the details Lewis set aside), here is how to express the conception of the I-relation (copersonality) that Lewis presupposes:

where RLxy is as defined above.

On this reconstruction, Lewis establishes the necessary coextension of RL and copersonality (his "I-relation": for Lewis x and y are copersonal just in case there is at least one person of whom both x and y are stages). Moreover, Parfit accepts that RL is what matters (footnote: a least in some passages: 301 "both relations [continuity and connectedness] matter",.. Other places 292 he says what matters is "relation R, that is continuity and/or connectedness" which is yet a 3d use of ‘R’. While this corresponds to his explicit defintions of R.. p. It is just a slip. Since the and/or reading implies there is R when there is even just one connection, which surely is not what he meant. The equivocation between R as continuity or R as continuity-and-connectedness is more serious since it underlies the confusion between his account and Lewis that I am trying to clarify.) So Parfit even can accept that Lewis’ I-relation is what matters. The conflict--or at least, the appearance of conflict-- is, of course, their divergent conceptions of copersonality: for Lewis it is R plus connectedness, and for Parfit it is non-branching R.

Lewis’s conception of copersonality has the consequence that personal identity is intransitive, given the General Psychological Criterion introduced earlier. [I think it would be a good idea just to drop reference to "identity" or "personal identity" for accounts like Lewis’s which do not treat personal identity as an equivalence relation: but Lewis himself uses the phrase "personal identity" even while acknowledging it is not a transitive relation: see quote below.] To see this, notice first Lewis rejects transitivity on his I-relation (copersonality) because of the role he gives to connectedness; and the intransitivity of I entails intransitivity in judgments about personal identity. Proof. Using ‘A=B’ as before to abbreviate "A is the same person as B," A=B just in case there are times t, and t* such that F(A,t)=x and F(B,t*)=y and x, y are copersonal. For Lewis, x and y are copersonal just in case RLxy. Because of the role Lewis gives connectedness, we saw that RL is not transitive. This is illustrated in the Methuselah case:

O O O

x y z

This means for Lewis there can be A,B,C and times t,t*, and t** such that F(A,t)=x, F(B,t*)=y, and F(C,t**)=z where RLxy, RLyz, but not RLxz; and this means A=B and B=C but not A=C; that is, A can survive as B, and B survive as C, but A fail to survive as C. (See Lewis, 31: I do not see that Lewis explicitly notes that personal identity can fail to be transitive, but it clearly is a consequence of his theory (even if it is reconstructed in terms other than those of the GPC--spell this out in his terms: and explain how tensed identity figures in here; if it does).

I do not see that Lewis explicity acknowledges that "survival as" is not "identity". Nonetheless Lewis explicitly distinguishes personal identity from identity in discussing how connectedness makes both I and R scalar:

Lewis is committed to a similar claim about transitivity: personal identity need not be transitive, even though identity is. (See Michael Hand who argues for that position.)

The debate between Lewis and Parfit is impossible to follow unless one realizes that Parfit is arguing that fission shows that what matters and identity can diverge (as we saw in his first argument), whereas Lewis argues that what matters is survival, even in fission. [Note: I believe that Lewis claims it is identity that matters--get the quote-- but this is not the case insofar as the relation that matters cannot be identity, since it is not transitive.] Parfit is talking about identity and Lewis is talking about survival. But strictly speaking, for Lewis survival does not require identity, since strictly speaking, identity is transitive whereas his conception of survival is not transitive. It is startling to realize that, despite the apperance of disagreement, Parfit and Lewis really are not in disagreement about the central issue whether identity is what matters. Moreover, Lewis presents himself as the conservative in their debate (even protesting that he’d have to reject philosophy if it conflicted with common sense). That Lewis manages to write as if he is disagreeing with Parfit’s claim that identity is not what matters, and as if he is doing so on behalf of common sense, when in fact his position is based upon the assumption that identity should just be ignored --this can be described only as amazing and beautiful philosophical magic: distraction plus sleight of hand. In any case, since personal identity is intransitive for Lewis, it is not an equivalence relation and is not identity; whereas Parfit’s claim that identity does not matter presupposes that identity is an equivalence relation.

The rejection of transitivity on personal identity is essential for Lewis’s analysis of the fission case, since if he were not to reject transitivity on I and his R (and hence on personal identity) then in the fission case he would be committed to the copersonality of y and z, hence to the personal identity of zKirk and yKirk. (His theory collapses into PG.)

Lewis and Parfit both disagree with PG, for whom identity and what matters are necessarily coextensive. As I have just shown, Lewis agrees with Parfit from the outset that identity does not matter insofar as he holds that it is survival, when understood in terms of his I-relation, that matters; and on his conception of survival, X survives as Y does not entail that X is identical with Y. Even though they agree that identity is not what matters, they still disagree about whether personal identity (or survival) matters: Lewis holds that survival is necessarily coextensive with what matters, whereas Parfit disagrees insofar as he holds that survival entails identity (and not merely personal identity), and that identity does not matter. More on this below.

Insofar as Lewis intended to be disagreeing with Parfit’s claim that identity is not what matters insofar as we want to survive, Parfit wins the debate with Lewis, who simply abdicated from the word go. Of course, he abdicated in an interesting and provocative way; as I will argue below, Parfit should do something similar. At any rate, Lewis’s theory cannot be used to challenge the first argument, considered above, that identity does not matter, since Lewis like Parfit is committed to the truth of premises (1) and (2), which entail the conclusion. (This does not mean Parfit’s argument works, however, as we saw above. The challenge from PG is different from the challenge from Lewis since PG holds that it really is identity that matters). To summarize, then, identity is not what matters for Lewis insofar as what matters is survival understood in terms of the I-relation. Survival is not identity because of the fact that survival is intransitive because of the role of connectedness in survival, as Lewis conceives it.

Even though degree of connectedness plays no role in fission, nonetheless it is an essential aspect of Lewis’ theory, when the theory is applied to fission, because the intransitivity of I (and RL) is what enables him to deny that yK is zK in fission since the stages y and z are not RL-related.

Since Lewis holds that it may be that F(M,t)=x and F(N,t)=x for distinct persons M and N, he can hold that zKirk and yKirk share the incident x prior to the fission. This means that Lewis’ view has the consequence that the term ‘Kirk’ is ambiguous when used prior to the fission since two persons, zKirk and yKirk share the stage x --and we make a mistaken presupposition when we talk about the person, Kirk, constituted by x.

Lewis successfully can explain, of course, why fission does not undermine the necessary coextension of RL and copersonality (his I-relation). Since RL is not transitive, we can posit both RLxy and RLxz in this fission case, and even though we have both RLyx and RLzx (because Lewis guarantees symmetry of RL) neither RLzy nor RLyz is entailed here since RL is not transitive. Lewis consistently can deny both RLzy and RLyz. But of course RL is not intransitive either, so it might seem mysterious why one should assume that transitivity fails in this case. That is, while we realize that RL can fail to be transitive, due to the failure of connectedness such as over long spans such as in the Methuselah case, why should we assume that connectedness would not hold between y and z in this case? There is, after all, not a long span here between z and y: it can be as short as one pleases. "Direction of causation" reasons are not available here to Lewis, since he abstracted away from those considerations when he explicitly made RL symmetric; indeed, Lewis imposed symmetry on RL explicitly in order to be able to defend the necessary coextension of his R-relation with his I-relation (copersonality) even though he acknowledges that intuitively R "has a direction":

What then about the combinations of forward and backward R-relatedness as in the relation between y and z in the fission example? Lewis does not explicitly consider forward/backward combinations, but clearly he needs to deny RL between y and z, and of course it is arguable that there may not be the appropriate bonds of similarity and causal dependence either way between y and z; nonetheless, it is not clear to me exactly why that should be assumed. Couldn’t there be some connections between y and z in terms of a conception of direct connectedness like Lewis’s that encompasses both similarity and causal dependence? We can imagine there being some connection or quasi-connection between incidents z and y.

Diagram. Kirk fission

z

Suppose for example that a memory (or quasi-memory) of an experience in z* is implanted in y by means of a brain-trace operation (see Parfit 220). Would this make for Ryz? It looks like it should, since after all, there is both continuity and connectedness between y and z. But surely Lewis wishes to deny Ryz here. But why? There is connectedness between y and z* (due to the implanted quasi-memory) and because Rxz* we have Rz*x (because R is symmetric), and of course Rxy. Given Rz*x and Rxy, plus the connectedness of z* with y, it is not clear why transitivity should fail in this case (this is not a case like Methuselah’s in which there is a long span of R-connectedness).

In any case, Lewis denies it; and the failure of transitivity of R and I (and personal identity) technically can provide for Kirk’s survival. But the claim that transitivity fails in fission must appear ad hoc simply to get the desired result in fission. Martin asks the transitivity-deniers whether they can give a principled reason for when transitivity of personal identity fails ("a persuasive argument that [personal] identity should not be regarded as transitive would have to specify the circumstances under which we are and are not entitled to infer from the fact that X is identical with Y, and Y to Z, the conclusion that X is identical to Z" (293, Kolak and Martin, eds). How could Lewis answer this? First, Lewis wants to treat cases in which there is not connectedness as such cases, as in the Methuselah case. (Also see Parfit’s recommendation for immortals, 304): so the failure of connectedness surely is a reason. Perhaps a reason also can be provided for the failure of transitivity in fission. Had Lewis rejected symmetry on RL in addition to (or rather than) his rejection of transitivity, then it would be easier to accept the assumption that y and z are not RL related in the fission case. Lewis may be correct that survival is not transitive (but it is clear that survival is not symmetric: A survives as B does not entail that B survives as A). But of course Lewis cannot reject symmetry on RL (given his assumption that the I-relation is symmetrical) unless he gives up the necessary coextension of I and RL.

What about fusion for Lewis? Here we have Rzx and Ryx.

Diagram. Kirk fusion

looks like the same problem as in fission (but nothing really new).

How can Lewis handle the Daydream case? He has no problem surviving Daydreams (the problem that worried Grice). However, one cost of his view is that he has to posit new people all along. Strictly speaking, nobody ever has a brief random idle daydream -- either it is just the end of one’s life (on the brief daydreaming branch) or it is somebody else having it (somebody else with whom one shared stages for a long time they branched briefly and died). That, surely, is a another cost of this view-- and indeed it is a cost that later should be compared with the costs of the NS view described below.

In any case, Lewis’s treatment of fission rests upon the assumption that y and z are not RL related, and it is not my purpose here to make a full-scale evaluation of the proposal. Other theorists also have considered abandoning transitivity on personal identity (Prior, Perry, Johnston?, Hand). Even if the rejection of transitivity is ad hoc, this may not be a very strong objection to Lewis if reflection on fission shows that some ad hoc conceptual extension or revision is going to be necessary to deal with the thought experiment (unless, per PG, one is willing to treat the fissionees as identical persons). Overall Lewis’ maneuvre need not be regarded as any more ad hoc than the tortured "non branching" proviso required to make sense of Parfit and Shoemaker. In any case, insofar as the transitivity-rejecting theorists are motivated simply to save people like Kirk in the fission case, they might have done as well or better simply to deny symmetry on copersonality and personal identity, as in a proposal I will make later. For the moment let us return to Parfit’s second argument.

The squabbling family of psychological criteria about identity and about what matters in survival

[remove NS from this section unless NS moved back here]

My purpose in discussing Lewis and NS was to prepare to discuss Parfit’s second argument that identity does not matter. The relevant differences between the criteria discussed so far concern the characterization of copersonality:

Perry’s Grice (PG): IGxy iff RGxy

(where RGxy iff Rxy or Ryx;

IG is an equivalence relation)

Lewis: ILxy iff RLxy

(where RLxy iff Rxy plus connectedness of x with y, or Ryx plus connectedness of y with x;

IL is reflexive, symmetric but not transitive)

Parfit (lenient or strict): IPxy iff non-branching Rxy or Ryx

(lenient/strict versions depend on non- branching proviso; either way, IP is an equivalence relation)

Non-Symmetric (NS): INxy iff Rxy

(IN is reflexive, transitive, not symmetric)

These are different conceptions of copersonality.

Although I personally have become enamored by the NS, mainly because I thought of it, [but footnote Dainton??] I do not know whether NS is overall a better strategy for thinking about fission than the denial of transitivity (as in Lewis or in Hand) or PG’s denial of neither transitivity nor symmetry (while biting the bullet about the personal identity of zK and yK), or Parfit’s non-branching gerrymandering. As for relative costs, notice that all of these alternative except Parfit’s meet both of the requirements that Parfit claimed no criterion could meet (whether a future person will be me depends only on intrinsic features, and is not dependent upon trivial facts, 267). As for Johnston’s three principles (violated by all the answers about fission that Johnston considered), namely, (1) Parfit’s intrinsic features requirement, (2) no person constituted at one time by two or more separately living human bodies, and (3) no person spatially separated from himself in the manner of an instantiated property, 161): clearly PG violates 2, as probably does NS (since Kirk is said to be the same person as both yK and zK; but whether NS really violates 2 could be argued since neither zK nor yK is the same person as the other or the same person as Kirk). But I don’t see that Lewis violates any of these principles. (But so what?) Perhaps a large-scale argument could be mounted for one or another of these proposals, but I am not going to attempt to do that, since I want to consider whether the simple fact that there are competing proposals about fission can ground Parfit’s second argument that identity does not matter.

These alternatives differ in their treatment of fission primarily because of assumptions about symmetry of their diachronic unifier (to have a general technical term for this: let us say the diachronic unifier is the glue).

PG’s glue is: RGxy (Rxy or Ryx) (equivalence relation)

Lewis’s glue is: RLxy (Rxy plus connectedness of x with y,

or Ryx plus connectedness of y with x)

(reflexive, symm but not trans)

Parfit’s glue: Rxy (reflexive, transitive but not symmetric)

NS’s glue: Rxy (reflexive, transitive but not symmetric)

The glue and copersonality are necessarily coextensive for all but Parfit. Assuming that the glue is "what matters," PG, Lewis, and NC can agree that "what matters" and copersonality are necessarily coextensive (even though they would disagree with each other, of course, about what matters, with PG saying it is R or its converse, and NS saying it is R, and Lewis saying it is R plus connectedness or the converse of R plus connectedness). But the important point is this:

Even though Lewis and NS disagree both about whether the glue and copersonality are symmetric (Lewis saying yes and NS saying no) they agree (along with PG) that the glue is symmetric just in case copersonality is. Only Parfit permits divergence here: copersonality is symmetric (it is an equivalence relation) whereas the glue is not symmetric.

It is the divergence between the glue and copersonality that is the basis for Parfit’s arguments that identity does not matter. Yet we have seen that technical assumptions about symmetry were made in each case in order to formulate a theory that entailed desired results about the fission cases or related ones. It seems to me that in this case we will not go far wrong in speculating about the motives of our theorists, based on their actual comments about symmetry of the glue. Parfit insisted upon the failure of symmetry in his discussion of fission in order to guarantee that R did not obtain between incidents on the post-fission branches. On the other hand, Grice imposed symmetry on Grice-R precisely to ensure Grice-R-relatedness of incidents in the Daydream cases. Lewis imposes symmetry on Lewis-R to make sure that it corresponds with the I-relation, which in the context of his discussion obviously is motivated by his insight how to guarantee that copersonality and what matters do not diverge in the analysis of fission. The NS theory I just made up (and here I can speak with rather more definiteness about the motivation of the theorist) was designed to put another coherent alternative on the table.

The apparently arbitrary and relatively minor differences between competing criteria about symmetry (on the glue and on copersonality) give rise to different judgments about identity, and then to different judgments about whether identity necessarily corresponds with what matters. These differences sometimes are said to give rise to "indeterminacy" about personal identity (Parfit, for instance, says this about fission cases). It is this indeterminacy that gives rise to the "empty question" that figures in Parfit’s second argument that identity does not matter. Before turning to that argument, then, it will be worthwhile to distinguish two different types of indeterminacy about personal identity that are possible in terms of the criteria under consideration.

*

Vagueness and plasticity

A question is said to be "empty," in Parfit’s work, just in case one could know all the facts there are to know without having an answer to the question. There are two distinct ways in which a criterion of identity might treat identity as indeterminate or treat a question of identity as empty.

First, as in the middle cases of the Combined Spectrum, no Psychological criterion will yield necessarily determinate results because of the ineliminable vagueness inherent in the notion of strong connectedness. There is no prospect for refining a Criterion to make it precise enough--indeed, a Criterion that was precise enough to give necessarily determinate results in the middle cases of the CS would be objectionable for that very reason (which, recall, is why Parfit’s own precise formula for strong connectedness was objectionable). It seems to me that Parfit’s CS-based argument for indeterminacy establishes this point. That questions about identity can be empty, due to the vagueness of strong connectedness, is something that any acceptable Psychological criterion should entail. What makes for personal identity is not necessarily "all or nothing." The facts about persons entail the possible indeterminacy of persons, as well as the possibility of questions that are empty and for which it would be wrongheaded to attempt to supply answers. To have a term for it, let us say this is "factual vagueness" (or simply, "vagueness").

The second type of empty question, as in the fission cases, differs from the first in the respect that it is not in itself an objection to a given criterion that it offers a determinate answer to such questions. Insofar as reflection on fission reveals indeterminacy, it is indeterminacy in our concept of (or our ordinary beliefs about) personal identity.

There are various ways extend our concept (or beliefs) to handle unusual situations, and it would be difficult to argue that one of the ways is uniquely required, given the facts. (Johnston, shoe re this] And while it may be futile to argue about survival in fission cases (and wrongheaded for that reason) but it is not at any rate part of Parfit’s reductionist picture that is wrongheaded to believe one way or the other about it. One might even argue, as Parfit himself did, that some conclusions about survival are better descriptions of the case than others. Concepts can be extended or refined so that within a given conceptual framework, determinate conclusions not inappropriately are held, but these conclusions are not required by the factual basis. The facts about persons underdetermine certain features (such as our capacity to endure the dematerialization involved in teletransportation) so there is, let us say, "conceptual indeterminacy" in our concepts--or rather, to have a term to distinguish this type of indeterminacy from the factual indeterminacy in the Combined Spectrum, let us say there is "conceptual plasticity" ("plasticity"). [Putnam re cats.]

The best illustration of conceptual plasticity is to be found when two criteria differ only because of more or less arbitrary assumptions about the formal character of connectedness, strong connectedness and continuity, even when they agree on more substantive issues of "width" (Parfit’s narrow, wide, and widest versions that differ in terms of requirements about normal causes of connectedness) and "strength" (the number or type of connections that make for strong connectedness). Of course, differences about width and strength also can be classified as merely "conceptual" [footnote: also of course contrast between any of our psych criteria and others incl physical criteria and criteria that lead to eliminativism] but the relevant sort of conceptual plasticity, for present purposes, is exhibited most strikingly when two criteria agree about width and strength and differ only about symmetry of connectedness --especially when the differences appear to be due largely to the desire to formulate a theory that gives certain outcomes in cases like fission. This is why the contrast between PG and the lenient Parfit strikes me as significant. We can imagine, for example, a PG theorist and a lenient Parfitian theorist who agree about width (they are "wide" ones, let us suppose, so both countenance the possibility of surviving teletransportation, whereas their "narrower" colleagues deny it) and they accept the same formula for determination of strong connectedness. They also can agree about whether or not there is copersonality in a large number of puzzling thought experiments, getting the same results about copersonality even in the branching Daydreams and Physics Exams and Sleeping Pill stories, when the branching is not "significant." But they disagree about what to say about copersonality when the branching involves incidents not "closely enough" connected: and the difference hinges only on what they say about symmetry of connectedness. While the reflexivist has theoretical reasons to deny symmetry here, the Parfitian who is not a reflexivist may deny symmetry simply to get the result that there is no survival in fission (as, it seems to me, was the case for Parfit himself when he wrote the book).

Vagueness and plasticity yield two types of indeterminacy and two types of "empty questions." An factually empty question is one where we do not have (and would not want) an answer even given an extended criterion. Parfit claims plausibly that any theory or criterion of personal identity without "separately existing entities" will yield factual indeterminacy. (It is interesting and important that even non-reductionists like McDowell acknowledge the possibility of indeterminacy in this sense.) The vagueness of any acceptble criterion of identity means that some questions can be empty, and there are cases in which any acceptable account of identity will yield situations in which the question of identity is empty.

A conceptually empty question, on the other hand, is one in which there is no uniquely correct answer because of the plasticity in our concepts: there is no uniquely correct extension of our conceptual scheme in order to understand a thought experiment that involves unusual situations. This is plasticity about how to extend our concepts in order to deploy them in thinking about bizarre cases. (See Johnston 161). Questions about identity are, let us say, "controversial" because the concept itself is "essentially contested." There are acceptable criteria such that situations are possible about which the criteria will give divergent answers to questions about identity.

The simple claim that identity is (or can be) indeterminate is ambiguous between factual indeterminacy and conceptual plasticity. Likewise "empty questions" about identity can be empty due to either indeterminacy or plasticity. This distinction will be important in the analysis of Parfit’s second argument that identity does not matter.

*

Parfit’s second argument that identity does not matter.

The second argument is a sort of meta-argument about the various interpretations of fission that can be made, as is illustrated in the list of criteria we have discussed so far. The argument is as follows:

(1) Kirk’s relation to ZK contains what matters.

Therefore,

(3) Identity is not what matters.

(See page 30 in reply to Brueckner). The "empty question" mentioned in premise 2* clearly is due to what I have called conceptual "plasticity", [footnote Paul Churchland re "plasticity" ??] and it is not due to factual "vagueness" as I just distinguished these, for in fission we do not confront the sort of empty question that we confront in the middle cases of the Combined Spectrum; rather the question can be empty because of the diversity of acceptable (or at least not obviously unacceptable) theories or criteria that can be deployed without contravening facts; and it is no objection to a criterion that it entails a determinate result about survival in fission. So clearly the empty question of premise 2 is due to conceptual plasticity, which means the second premise should be read then as:

But given the interpretation of 2 as 2*c in terms of plasticity, the step from 1 and 2*c to 3 obviously becomes problematic. Isn’t it reasonable to expect that the plasticity about identity is going to be relevant to a position about "what matters" too?

Consider, for example, the PG theorist who thinks that there is identity in the Kirk case, and so (retaining the background commonsense belief that identity is what matters) she continues to believe that identity is what matters. She accepts premise 1, of course, but does so only because she believes on the basis of her own PG theory that Kirk=ZK. Having refined her concepts of personal identity to accommodate the fission cases, she has a theory that entails 1; and she may very well accept 2*c insofar as she realizes that other theories just like hers, except for symmetry on R, make Kirk=ZK false; and yet she still has no reason to infer 3 from 1 and 2*c. Because she holds that R and copersonality are necessarily coextensive, she need not agree that 2*c and 1 entail 3 just as she rejected the first argument because of its second premise (the determinate non-identity of Kirk and ZK).

Parfit might reply that given 2*c, the PG theorist is cheating when she uses her background theory’s entailment of Kirk=ZK as the basis for her acceptance of 1. Ok; but without that assumption, she has no reason to agree that 1 is true, because she believes that R and copersonality (identity) are necessarily coextensive. It seems to me that Parfit simply cannot use the conceptual plasticity of criteria of identity to construct the "logical wedge" he needs to construct between identity and what matters.

To approach this from another angle, consider a theorist who accepts the Lenient PS Criterion and believes identity is lost. But (being relatively humble) he grants 2* because he recognizes it is not a factual mistake to say K=ZK (in Johnston’s terms, he realizes that is one extension of our concepts that could acceptably have been made, just as the PG theorist makes it). What about the first premise 1 then? Given the background belief that identity is what matters, should he grant this premise? It seems not (at least not necessarily), since he can consistently believe that the acceptability of 1 would be dependent upon an extension according to which Kirk=ZK; but since he believes Kirk is not ZK, he denies (or at least consistently can deny) premise 1. The point here is that, like the PG theorist, this theorist can presuppose the background necessary coextension of identity with what matters. But since he believes there is nonidentity, unlike the PG theorist, the background belief will lead him to deny premise 1. Either way, Parfit’s argument fails.

These points suggest that given the conceptual plasticity presupposed in 2* the most that can be claimed, generally, along the lines of 1, is a much weaker first premise that also introduces plasticity there as well.

But 1*c and 2*c do not entail 3. Nor do 1*c and 2*c together even entail that it is an empty question whether identity matters, since it is consistent to accept that both 1*c and 2*c are true while continuing to endorse the background commonsense belief that Parfit is attacking, namely, that identity and what matters are necessarily coextensive. The PG theorist, for example, recognizing the variety of views even amongst those who accept the general Psychological Criterion, accepts 1*c and 2*c.

Parfit’s argument would be more promising if his second premise could be reconstructed as presupposing factual indeterminacy rather than conceptual plasticity, because this might block the maneuvre by which plasticity about identity is assumed to accompany plasticity about what matters. In the Combined Spectrum there is not merely conceptual plasticity in the middle cases, there is factual indeterminacy.

But factual indeterminacy is irrelevant to the fission cases, as these are normally understood -- it is taken for granted that the relation between Kirk and each successor is determinately sufficient for R-relatedness. So the relation between Kirk and ZK is much different from the relation between Parfit and, say, Pargarbo in one of the middle stories of the Combined Spectrum, where ‘Pargarbo’ refers to the person who emerges from the operation and where what is in question (and must remain in question, for any plausible criterion) is whether there is a sufficient degree of connectedness to make for strong connectedness and R. The indeterminacy of Parfit=Pargarbo differs from the controversity about whether Kirk=ZK (and, of course, this illustrates the difference between vagueness and plasticity). And notice that nobody would try to use the Combined Spectrum F indeterminacy like this

The problem, of course, is that it is quite obvious that nobody who accepts the second premise need accept the first one.

To summarize, Parfit recognizes that he needs to exhibit a case in which identity and what matters come apart (the "logical wedge"; see Harris, ed 33). The Combined Spectrum does not show that, and of course Parfit never claims it does. But he does claim that fission shows it. Since he realizes that he cannot claim that identity does not obtain in fission without taking a position that excludes reasonable theories (like PG), in his reply to Brueckner he revised the arguments from RP and Ethics by weakening the second premise to 2*, the "empty question" claim. (That is, he abandons the first argument with premise 2 that we considered above.) If indeed what matters holds but identity did not hold, then his argument would work. It is of course not inconsistent to take that position, but the problem is that he has not given a reason why a theorist cannot make the same response as she would make if the combined spectrum had been offered as a basis for the claim that identity does not matter, namely: if it is indeterminate whether Kirk=ZK, then it is indeterminate whether Kirk has what matters. All this of course is based on the background commonsense assumption that Parfit is trying to challenge (that Kirk=ZK iff Kirk’s relation to ZK contains what matters), but making that assumption is fair enough until one is shown what is wrong with making it. Parfit is trying to show this, but so far it seems to me his argument begs the question.

Parfit’s second argument trades on the ambiguity between (what I am calling) vagueness and plasticity when he speaks simply of "indeterminacy" and "empty questions". For instance, he summarizes his second argument as follows:

Parfit’s question seems reasonable, and his second argument initially may appear attractive. But it can work only if the empty question is due to factual indeterminacy --for the premises entail the conclusion only relative to factual indeterminacy; and yet relative to that interpretation the first premise clearly is false if the second one is true. Relative to conceptual plasticity, on the other hand, the premises do not entail the conclusion. To summarize: the problem here is that the "empty question" to which this sentence refers needs to due to factual indeterminacy for the argument to work; but there is only conceptual plasticity (and not factual indeterminacy) in fission cases.

Corresponding to the difference between indeterminacy and plasticity are two very different types of "refinements" that can be involved in anwering questions that "are not covered" by our concept of a person. For example, if our concept of "person" is simply "locus of mental activity" (Johnston) and our concept of "personal identity through time" is simply, say, copersonality as expressed above in the General Psychological criterion, then of course many questions are left unanswered. At this point one’s theory leaves open questions about how to deal with the cases in the middle of the CS as well as what to say about teletransportation and fission. But then suppose one accepts PG so one does have an answer to the question about teletransportation; yet questions about the middle CS cases are still empty, and they should remain empty no matter what specific refined criterion one may accept.

Parfit may wish to hold that the differences are not relevant to his argument, and that the plastiticity in interpretations of fission is enough for his argument. He may wish to hold that opting for PG (and trying to defend it) is as irrational and arbitrary as drawing a sharp line, say, right in the middle of the CS. After all, even if there is some difference between indeterminacy and plasticity, nonetheless in both cases one could know all the facts there are to know without having an answer to the question "would I survive teletransportation?" One’s opting for one theory or another to answer the question does not change anything about the factual situation.

But I don’t think this response will work. Even if the differences between indeterminacy and plasticity are ignored, the obvious problems that would undermine a CS based argument involving Pargarbo also undermine the fission argument. In the Pargarbo argument it is obvious that indeterminacy about identity is coextensive in each case with indeterminacy about what matters, so the truth of one premise undercuts the truth of the other one. Parfit needs an argument based on a case in which it is determinate that the relation between Kirk and ZK is not identity yet it is uncontroversial that the relation between Kirk and ZK has what matters. The controversity about identity is not sufficient here, because the question of identity need not be taken as independent of the question about whether or not Kirk has what matters (as in the middle CS cases).

How could Parfit show that identity is indeterminate but Kirk’s having what matters is not indeterminate? One way he could do this is by showing that it is R that matters (in the way we tend to think survival matters) or that it is R plus some degree of connectedness or some related thesis-- whether or not identity is determinate. Since we know that Kirk is R-related to ZK, then we would have a basis for the first premise. Perhaps this can be shown: that R matters whether or not identity is determinate. But Parfit has not yet shown this. Without having shown it, it will be open to an opponent such as Johnston to claim that the intuition that Kirk’s relation to ZK contains what matters already is tainted by an implicit extension of our concepts of identity and attribution of appropriate self-referential attitudes such as self-concern (we regard as justifiable an extension of self concern, by Kirk, to include ZK) -- and that those refinements can be made without abandoning the assumption of the necessary coextensiveness of identity and what matters.

Since Johnston would claim that it is determinate that Kirk is not ZK in the fission case (see his comment on teletransportation), his objection would be to premise 1: if Kirk is not ZK, that is grounds for denying that Kirk has in ZK what matters. One cannot say whether what Kirk has in ZK contains what matters unless one knows whether or not it is identity between Kirk and ZK (or at least so one can reply to Parfit, given the commonsense background assumption that identity is what matters).

Johnston’s reply to Parfit about this type of argument is best framed in terms of the brain division version of fission since Johnston focusses exclusively on body-based reductionist criteria. In fission one may very well extend his concepts to countenance survival (even if he wouldn’t do so in teletransportation cases ). And while one reasonably could extend one’s self-concern in a brain-based fission case so that self-concern is extended to the fission products, this doesn’t show that what matters (or should matter) is not identity, because we can distinguish between what matters (or should matter), on the one hand, and what would matter given the extension of our concepts and self-referential attitudes in an unusual situation (See Johnston 170). It seems to me that Johnston’s "quarantining" strategy has to be respected as a coherent conservative conceptual move in response to plasticity. I don’t see how Parfit reasonably can object to this reply, given his own reliance on plastiticity in the argument as reconstructed, until he establishes independently that one can have what matters, by virtue of R, even in a case in which identity is indeterminate.

Can Parfit do it? It would be best to do this in terms of the all or nothing F-indeterminacy like that in the Combined Spectrum, since the following argument is valid:

Kirk has what matters because of R-relatedness to ZK.

It is (factually) indeterminate whether Kirk=ZK.

Therefore, identity is not what matters.

So: what about such a case? We want factual indeterminacy about identity but it is determinate that R holds. But that is a contradiction according to the Psychological criterion! (The argument is "valid" only because the premises are inconsistent.) If Rxy then there will be continuity between x and y, meaning there is chain of connections wherein each connection contains enough direct connections for strong connectedness. And if so, it is not an empty question, anywhere in the chain, whether there are enough connections--which means there is no F indeterminacy.

What about the 100 operations (Harris ed) argument? A brain-based psych criteria may posit survival for Kirk if there are 100 small operations, but no survival if only one operation (replacement of brain). (See p 38). Consider the one operation situation; and this argument

Kirk has what matters because of R-relatedness to ZK.

It is (factually) indeterminate whether K=ZK.

Therefore, identity is not what matters.

Pretty clearly J’s quaranting strategy works here as well. Even though this thought experiment is not as bizarre as fission, nonetheless it is an unusual situation. Since our conceptions of self and what matters presuppose determinacy, this case like fission requires an extension (refinement) of our concepts. Not just any old answer will do, if we are motivated to come up with one, but nonetheless given the plasticity of our concepts, one not-irrational option is to presuppose the necessary coextensiveness of identity and what matters, and to reply that the second premise is true if, and only if, the first premise is not true.

[? By the way, clearly Parfit has the basis for an argument with a weaker conclusion, namely, personal identity does not necessarily matter. --Isn’t this what the next argument does?]

Another meta-argument that identity does not matter

What about this one? Lewis and NS agree that identity does not matter in survival --it is personal identity that matters. Parfit doesn’t need to convince them that identity does not matter.

(1) What matters cannot depend upon arbitrary or trivial differences in the interpretation of fission.

(2) Survival of fission (identity) depends upon assumptions about the symmetry or nonsymmetry of the glue and copersonality: one survives fission according to criterion X iff X treats the glue and copersonal alike relative to symmetry.

(3)The assumption about symmetry of the glue and copersonality is an arbitrary or trivial matter: it is largely a matter of theoretical taste or whim (such as, for example, the desire to put another alternative on the table).

Therefore, survival (identity) is not what matters.

Can this version of the argument evade Johnston’s quarantining strategy?

A conception of copersonality as asymmetric: "the asymmetry of personal identity through time"

So far we seem to have learned that personal identity is indeterminate in two senses, one of which is conceptual plasticity. The fission cases force us to accept this conclusion unless like Chisholm we insist on determinacy about such cases. As thomson pointed out, the intution that Chisholm articulates is "very strong in us" (Chisholm said it was "clear and distinct" to him that he would survive as one or the other or neithcr??--of the fissioness). But in fact this is not clear and distinct, and if there is no soul in the picture than the intuition that is very strong in us is wrong. It needs revision. Parfit realized that our commonsense view of ourselves needs revision, and attempted to articulate this realization in his claim that identity does not matter. The arguments for this claim do not succeed generally, however, since the fissions arguments simply do not work, and in my opinion cannot be made to work. As we have seen, conceptual revisions can be made so that personal identity and what matters are necessarily coextensive (as in PG), or so that at least survival and what matters are necessarily coextensive (as in Lewis). However, PG nonetheless pays elsewhere, insofar as it allows for the possibility of surviving radical disintegration.

I think we can say we have learned that there is room for creativity here. The evaluation of alternative extensions of our conceptual schemes isn’t going to be "factual"-based. Recall Parfit’s premise in the second argument: What standards of evaluation should we use? Aesthetic and ethical, perhaps. A conception of how to deal with fission is a work of art, or a normative proposal.What Parfit should do is to quit trying to argue for the conclusion that identity does not matter, and instead follow Lewis in formulating an account of copersonality that presupposes that identity does not matter; and in spelling out the alternative account, explain its virtues.

It is difficult to believe that the world needs another criterion of personal identity. All the same, I would like to propose a new conception of copersonality.

If one is willing to give up the assumption that personal identity is an equivalence relation in order to handle the fission cases, another alternative is to abandon symmetry on copersonality and personal identity (and keep transitivity); one can simply treat copersonality as necessarily coextensive with Parfit’s nonsymmetric R.

Incident x is copersonal with y (INxy) just in case Rxy.

This gives the result that copersonality is not symmetric, since to consider Kirk’s fission once again:

we have Rxy but not Ryx; hence INxy but not INyx. Given the General Psychological Criterion, we then would have Kirk=yKirk since PNxy, but not yKirk=Kirk because not INyx. On this view the relation IN ("copersonal") has a direction -- indeed it is just the direction of R (since on this view copersonality just is Parfit’s R). Which direction? Parfit’s R goes from past to future in personal time. We can say Kirk=yKirk and Kirk=zKirk, but since neither Rzy nor Ryz, we have neither zKirk=yKirk nor yKirk=zKirk; and this is consistent because while = is transitive on this view, it is not symmetric. (So from Kirk=yKirk we do not derive yKirk=Kirk; if we could get that, then from Kirk=zKirk and transitivity we’d get yKirk=zKirk).

This proposal is not as odd as it at first is likely to seem. Suppose we know that

(1) A survives as B.

From this it follows that A is the same person as B, since generally

(s1) A survives as B only if A is the same person as B,

but (1) does not entail

(2) B survives as A.

Claim (2) does not follow from (1), given ordinary usage, because "survives as" is not symmetric; the logic of survival encodes the direction of becoming (so to speak) --that is, "survival" encodes the asymmetric relation by which some personal states or events are successors of others insofar as some arise out of others (see Nozick for this phrase)--indeed, it can be regarded as just the asymmetry of Parfit’s R. In any case, survival is asymmetric, which is why (1) does not entail (2).

Normally we assume a tight link between survival and personal identity. It might even be natural to assume that

(s2) A survives as B if and only if A is the same person as B

but of course this fails if personal identity is symmetric but survival is not symmetric. But rejecting symmetry on personal identity (which I will abbreviate as ‘=’ unless explicitly noted otherwise), one can accept s2. So NX establishes a closer link between survival and personal identity than can be established if personal identity is taken to be symmetric. Just as 1 does not entail 2, so also on this proposal A=B does not entail B=A.

Some philosophers inspired by Parfit’s claim that identity is not what matters, have suggested it is survival, not identity, that interests us-- and of course this is in the spirit of Parfit’s claim that identity is not what matters in survival (alternatively, it is not what matters or should matter to us insofar as we want to survive). As Georges Rey says (using ‘=’ to refer to identity proper):

The link between survival and identity can be severed at two different points: (a) personal identity is not the identity relation, or (b) survival is not (and does not entail) personal identity. We saw Lewis deny transitivity on personal identity at least implicity, committing him to (a), insofar as personal identity lacks a property (transitivity) that identity generally has. I regard this proposal to be a form of philosophical magic (using the phrase "personal identity" here strikes me as confusing--WAIT! It can be confusing, but also I think it can be done without confusion if (unlike Lewis) one is explicit about it --see below, where I think I can do it!!. But Lewis did not separate survival from personal identity; and of course Lewis’ purpose was to argue that personal identity and survival (and what matters in survival) need not be regarded as diverging under any possible circumstances.

While Parfit treats the logic of personal identity as the logic of identity (see eg 298: "fusion does not fit the logic of identity" and other passages), in his early work distinguished survival from personal identity. Other philosophers also have accepted (b), rejecting even s1; that is, they use the term ‘survive’ so as explicitly to sever the entailment from ‘A survives as B’ to ‘A is the same person as B’. (Even if we assume that Kirk is not the same person as zK, still one can say that Kirk survives as zK. For this usage, See early Parfit, Martin). But rejection of s1 may be more likely to obscure than to clarify since in ordinary use ‘survives’ entails personal identity. (I regard acceptance of (a) and rejection of s1 equally to be willful misuses of English. Lewis’s and Martin’s insights need, in my opinion, to be expressed in other ways: for Lewis, he simply assumes neither identity nor personal identity matter; for Martin..... WAIT: see note above: I think the best strategy now is ot use the pharse "personal identity" and "is the same person as" so that neither entails identity.. So follow Lewis on that in developing NX view: but make it explicit that if this is too jarring, the view can be developed in other terms --and either way it involves an acceptance of the Parfitian insight that Rey expressed above: identity is not quite to the point relative to our concern with survival.

On the other hand, Parfit in RP does not use ‘survive’ so as to sever the entailment to personal identity; and, unless I am mistaken, he does not de-link survival from personal identity per (b). That is, he does not reject s1 in his book, where he argues instead that identity is not what matters in survival--a claim that equally can be expressed by saying that it is not survival that matters. Kirk does not survive fission, but he has what matters in survival (that is, he has what matters insofar as he and we wish to survive into the future).

The NX theorist who denies symmetry on personal identity follows Lewis here, distinguishing personal identity from identity as in (a)--alternatively, simply rejects talk about identity?????? --See brief sermon above about misues of English!!! --Wait! But I will show how my proposal actually squares with English statements we are drawn to make --re blurring the numerical/qualitative identity distinction.... Whereas proposal (b) severs the logical link between survival and personal identity by denying (s1), the NX strengthens the link since, as noted, it supports not only (s1) but also (s2).

I survive insofar as my earlier self is the same person as my later one (so to speak) --and what do I care about the converse relation? Of course it is natural to assume I will be the same person as X later iff X later will be the same person as me--but that is because it is natural to assume that personal identity just is identity. Consider, for example, the two questions in the following passage from Lewis’s article:

According to NX, the two questions here have different answers. If I survive until a later time, then I who am reading these words will still exist at the later time; there will be someone X such that I now am one and the same person as X by virtue of being R-related to that person. But, nonetheless, none existing later will be me, since none is (will be) R-related to me. None later will survive as me. The NX proposal is to treat the logic of personal identity as the logic of survival (not the logic of identity simpliciter); in so doing, it can agree with Lewis also on the points that what matters in survival is survival, and what matters is personal identity--for survival and personal identity are necessarily coextensive. (It seems to me NX is superior to Lewis at this point because of the asymmetry of both R and survival: whereas for Lewis’R is symmetrical. This is related to the critical questions posed earlier re his interpretation of fission.)

As Rey pointed out, our concern is with survival. Indeed it is tempting to say that just as I now do not survive as me yesterday, in fact I now am not the same person as I was yesterday. Normally philosophers regard this sort of statement, of course, as incorporating an ambiguity between numerical and qualitative identity. (see Parfit re ‘After his accident, he is no longer the same person’, 201). Yet perhaps, on the contrary, NX offers a conception of numerical personal identity that is closer to statements about qualitative personal identity in an appropriate way insofar as the NX conception is necessarily coextensive with the survival relation. As just noted, ‘A survives as B’ does not entail ‘B survives as A’. Consider ‘I am not the same person I used to be’. True, I survived the accident, in the sense that the person who I was is identical with me now, but I am not the same person I was before it. (The use of ‘I’ is tricky here --see Velleman.)

Of course I now am not the same person I was just one second ago either! (Of course I now do not survive as me a second ago either, even though me a second ago survives as me now.) The direction of succession built into NX personal identity follows the asymmetrical R. Notice that our NX-theorist needn’t simply reject the general distinction between numerical and qualitative identity, since in a future oriented statement like "I won’t be the same person after the face lift" is subject to the familiar numerical/qualitative identity distinction: since I will survive the face lift, I will be numerically the same person after the face lift. Where the NX theorist fails to acknowledge the numerical/qualitative distinction is in statements like Parfit’s where a later person (so to speak) is said to be not the same as an earlier one. Here there never is numerical identity (at least never insofar as there never is R-relatedness, which means, I think, never in the absence of causal loops: if there are loops there could be).

Lewis considers the statement, "in later life I will be a different person" (30) and remarks that for shortlived beings like ourselves this statement (interpreted in terms of numerical identity) is an extravagance, although it wouldn’t be an extravagance for a longlived being like Methuselah (because degree of connectedness is relevant to numerical identity for Lewis). On the NX view, though, such remarks are not an extravagance even in daily life from moment to moment when there is no question about degree of connectedness: one can say truly that in a moment I will be a different person, I will become a different person: I will become him, because I will be him (I will be numerically one and the same person as he); yet I will be a different person, because he will not be me. These statements have the ring of truth, sort of, as well as the ring of paradox, and the NX theory strikes me as at least refreshing iNXofar as it explains why: from earlier to later in the series of R-relatedness, we have personal identity; but we don’t have it from later to earlier in R. I am not the same person I was this morning --but he is me; indeed, he became me. I didn’t become him. He survived as me, whereas I will not survive as him. See also the quotes from Proust (in Rey and Parfit, 305).

To make this more precise, consider a person at a time as constituted diachronically at that time by the set of incidents in which his existence consists or ever has consisted: A at t is constituted by the set F*(A,t). (Define a new function F* to do this, where our old F just picks out what constitutes me synchronically.) As time passes, my existence is constituted by monotonically increasing sets of incidents, meaning for t* later than t, F*(me,t) is or is included in F*(me,t*). The set of incidents at earlier t will be included in the set of incidents at later t, but not conversely. From this we might say that the earlier self is included in the later self, on this conception, but not conversely --and this has the ring of truth; and this is the grounds for saying I earlier have become me later, that is, I earlier am the same person as me now (insofar as I earlier "am included" in me now); but I now am not the same person as me earlier (since the later set of incidents is neither identical to nor included in the earlier set). One who simply identified the person with sets of incidents might be saying this; this is something like the view Parfit describes as identifying reductionism (Harris ed). Given the asymmetry in these relations it is not so odd to say that the earlier self A is the same person or self as the later self B (I use ‘self’ and ‘person’ interchangeably) but the later B is not the same person as A, since the earlier self becomes, survives as, the later self whereas the later self is not, does not become, the earlier self.

On this account: A survives as (becomes) B just in case F*(A,t) is included in F*(B,t*) for some t and t*. Wait. This locutions are problematic for the same reasons that Thomson’s criterion gets screwed up: if A=B, then A at t just is B at t*. So be careful with this sort of formulation.

The sort of challenge that Ray Martin presented to the deniers of transitivity can, I think, be met by NX. The challenge in this case would be specify the circumstances under which we are and are not entitled to infer from the fact that X is identical with Y the conclusion that Y is identical with X. Answer: never. While the NX criterion need not treat copersonality as antisymmetric, since we could have both X=Y and Y=X just in case there are incidents x,y and times t,t* where F(Y,t)=y, F(X,t*)=x and both Ryx and Rxy. Given the weight this approach puts on the directedness of R (emphasizing the direction of causation) it would be rare for distinct x,y to have both Rxy and Ryx (requiring a causal loop). In any case, even if it is possible that there are cases in which both X=Y and Y=X, the denial of symmetry is not ad hoc because it is grounded in the direction of causation, in the way in which one’s states arise out of earlier states, with some being successors of others. The direction of R is the direction of becoming and survival. Fission cases offer nothing new: Kirk becomes both zKirk and yKirk; he survives as both zKirk and yKirk; and he is the same person as zKirk and yKirk. Of course, the twist is that neither zKirk nor yKirk is the same person as Kirk (but of course neither zK nor yK survives as Kirk either; and neither becomes him). But equally I now will survive (I hope) in me tomorrow, I will become me tomorrow, and I will be the same person as me tomorrow. But (and here is the twist) I tomorrow will not be the same person as I am now; indeed, as we have seen, some of the usual examples of confusion between numerical and qualitative identity --"I am not the same person I was yesterday"-- become pedestrian truths about personal numerical identity through time. Of course I tomorrow will not become me now and neither will I tomorrow survive as me now (and the plausibility of these points may ameliorate somewhat the cost of being committed to saying that I tomorrow will not be the same person as I now am).

Re qualitative and numerical identity again. I will be a different person tomorrow. This statement like I will not be the same person tomorrow can be used to illustrate the distinction between numerical and qualitative identity (as does Parfit): to describe changes in me, we need to presuppose that there is numerical identity. That is, there is a person, me, who will have different features tomorrow. I will change, become a "different person" in the sense that I will exist but my properties will be different. NX agrees with this. But, on the other accounts, my having the different properties requires that the person tomorrow with those other properties be numerically one and the same person as me now. NX of course does not reject the general distinction between numerical and qualitative identity, but NX denies that the distinction is illustrated by the statement that I will not be the same person tomorrow (or in two seconds either). As changes take place, as I have new experiences or perform new actions, the person I become will not be numerically one and the same person as the person who I am now. And the new experiences and the new actions are what make the difference (if time passes but there are no new relevant mental events, then there are no new incidents, so no basis for the claim that there is a different person). So the qualitative differences are what make for the numerically distinct person. The person tomorrow (or in two seconds) will not be me due to the changes in the series of events that constitute me, and of course, by virtue of which I survive into the future (and of course I do survive since the person who I now am will be numerically one and the same person as a person existing tomorrow, assuming the R-series continues). These sort of statements would amount to nothing more than paradox-mongering but for the facts that (a) the theory is a well-motivated alternative response to the fission puzzle and (b) the statements themselves can be understood in a way that makes sense. I will survive until tomorrow by virtue of being R-related to a person existing tomorrow: I will be that person. Yet the asymmetry of personal identity allows us also to maintain that that person will not be me, due to the qualitative changes. I will be numerically the same person as him (I survive as him; I become him). But he will be numerically distinct person from me. Yet, despite his being numerically distinct from me, I will be (become) that person, which is to say that I will be a different person tomorrow --I will become a different person.

"Asymmetry, identity, and the void": if personal identity through time is asymmetric, then it can be that I now am not the same person as me in the past, even though me in the past is the same person as me now. Likewise I will become a person (me later) in the future, since I now will be one and the same person as me in the future, but I am not yet that person (and indeed it is not yet settled which person will be me, of those who are possible now). [See notes comparing with Lewis’ eternalism and bringing in determinism/indeterminism variation as well: given NX plus indeterminism, from perspective of t: there is no one in the past with whom to identify, and no one yet in the future...] So right now? I am me now but there is no one past or future with whom I can identify; yet I am aware, and I know that as each moment dissolves into the next, I as that person whom I will become will find my situation then, in this respect of not having anyone past or future with whom to identify, to be precisely the same as my situation now. I will become me-in-the-future, yet that future me will not be me ("me" as thought now), nor will there be any person existing in my past, at that time, with whom I will be able to identify; as the unsettled future becomes settled and fixed, definite facts condition my future. Yet none of those facts are such that I now can, or then will be able to, identify with any past person. There is no one in the past with whom I now can identify, and there never will be no matter how long I continue to become. This is the lucid beautiful void.

*

"The asymmetry of personal identity through time: the relation between facticity and existenz"

Facticity: I became who I am, but I am not he who became me. I have beomce the actual subject of the intnetions and the actions resulting from those intentions, where those intentions had a purely notional subject when formulated. "Purely notional," for one thing, because there are infinitely many individual actions that are possible that conform to the intention, even when it is realized; morever, when we intend we also formulate conditional intentions for what to do if things go wrong ("contrary-to-intention intentions") so I can intend to tell neither sally nor Penrose, but to tell both if I tell one. So: int(I not to tell P) but int(I to tell P/I tell Sally).

The asymmetry of personal identity through time parallels direction of fit: for the past (or rather, for what is settled at a certain time, even if not yet realized) the direction of fit is "mind to world" --and this is true even for my own actions. For instance, when it becomes settled that I will tell Sally (even though I intended not to tell her and even if I haven’t yet actually told her), then my action of telling Sally becomes part of the circumstances that should determine what I do or not do re Penrose. Before I tell Sally gets settled, hoever, my not telling her is required (relative to my intentions). The direction of fit is "world to mind": the world should conform to my intention.

I am not describing anything particularly astonishing here. In practical reasoning, I am Janus: facing, first, both the fixed past and (still looking backwards but seeing the future in my peripherial vision) the fixed future insofar as it is fixed by the past (facticity); and, secondly, looking forwards to face the open future insofar as it is open (existenz).

The Kantian challenge to reductionism about persons from the past is that reductionism cannot handle the knowledge about fixed circumstances via memory; the Kantian challenge from the future is that it cannot deal with agency resulting from deliberation, decision, choice, planning, intending, and acting in light of the open future. We consider these challenges later (after Velleman stuff gets into the picture --note also see Velleman’s earlier article on agency).

*

commitment etc

And of course when I do go along with it --well that is part of R-relatedness. Of course, remembering the earlier decision, and decding not to do it--well, there is plenty of connectedness in that process as well.

What about my sense of commitment to ceratin plans and actions for the future? Well, if via earlier self’s plan same point as above. I now am not necessarily merely an intermediary conduit between past and future selves> what about plans and earlier decisions I now endorse? Well, same point: this continuity of endorsed plans just makes for connectedness, and that is the sort of process by virtue of which I become me later.

*

Punishment/praise/blame

Suppose that A is constituted by x and B by y where Rxy, so A=B, but ~Rxy so not B=A. Suppose x includes or part of a series of incidents that include committing a crime during t and y includes punishment for that crime. Given asymmetry of = we need to be careful how we state the retributivist principle concerning punishment that B is justifiably punished for A’s crimes only if B is the same person as A. We should say: only if A=B (since that is true in this case, whereas B=A) is not true. But we can put it that way, and there isn’t necessarily anything radical emerging here. (Eg not Parfit’s extreme claims.) [Note: punishment, blame, as welll as praise, is different from first personal commitment where A=B is not sufficient to "commit" B, since B can reason "I am not A, so why should what A decided determine my actions?" If B says, "I am not A, so how can you punish me for what A did?" the ansswer is: You may not be A, but A is the same person as you, and ..."

Notice that we do not punish A for the (future ) crimes of B. Now of course one reaons for this is that the crimes have not yet occurred and may not be settled. But suppose in a certain situation we did know somehow that the person B whom A will beomce will commit a heinous crime; suppose it is settled now that B will do it and we know that it is settled now. Would this give us grounds to punish A simply because A=B? I don’t think so. So justifiable punishment is asymmetric in just the way that personal identity is asymmetric. "It is justifiable to punish A for the crimes of B if A=B" is false; we also need B=A. [This needs to be reworded but the point is correct: perhaps the point is simply that from the first paragraph: not only is B=A not necessary for punishment, it also is not sufficient; whereas A=B is .....--spell it out better]

of course even re punishment: the fact that not B=A might be relevant, eg with respect to excusing conditions, etc -- but this is secondary probably to connectedness being weak (even if R holds: note this might lead to alternative view where R plus connectedness is the criterion for copersonality)

*

Regret, remorse

One source of regret is a sort of yearning grounded in past unfulfilled hopes and desires. Whose desires? Well the desires I used to have and may still have. --But this is false, if I did not exist in the past, as NX claims. True, a person existed in the past with certain hopes and desires, and indeed that person survives as the person I am now. That person became me. But the person who I am now is not that person. I grant that that person had certain hopes and desires, just as many people did. But why should I have any special concern for the past desires of the person who survived as me? That person is me but I am not him. Special concern for my past is impossible since I do not have a past (and never will). Since I am not the person whose past desires and hopes went unfulfilled --what is that to me? Is it irrational not to have a special concern for one’s past? I say it is not, because NX is an acceptable conception of copersonality and according to NX one does not have a past. Likewise for past injuries and injustices. Of course it may not be irrational to have the special concern, but clearly it is not irrational not to have it. If one adopts the NX frame of mind, it may become easier to let go of regret and remorse and sense of injury.

*

special concern for the future

There is no reason NX should affect concern for one’s future: R provides for personal identity, survival. Special concern for my own future is an aspect of R (see McInerny, Martin, also Whiting, Rovane??). Of course I should say future(s) since there is no reason in principle why I should have just one. [SO REPLYY TO VELLEMAN AS AN ASPECT OF DEFENDING COHERENCE OF NX SOLUTION TO FISSION: Velleman argues that in fission I could not be self-related to each fissionee. So first interpret self-relatedness as an aspect of R, then reply to Velleman’s argument.

*

Intrapersonal compensation

Let A, B as above.. where A=B but not B=A. Since A=B it is ok to deprive A for benefits to B (e.g. The child for the adult person he becomes) but it is not ok to deprive B due to the beneifts to A, since not B=A (for the adult is not the child). So justifiable intrapersonal compensation plausibly is asymmetric in just the way that personal identity is asymmetric per NX.

That is, it at least conforms ot attitudes we do have: we do justify acts in terms of delayed gratification (a parent doing so with the child on behalf of the person the child will become) but we don’t (I think) justify actions in terms of prior gratification --unless the prior gratification violated the rights of others -- in affirmative action, for example, the fact that I recieved certain benefits that were themselves the result of past injustice might be grounds for denial of certain benefits now. But I think that is a completelly different matter.

*

Lewis v. NX view

Lewis takes the eternal point of view: his theory of copersonality derivative from a four-dimensional ... See notes and diagrams on this.

*

Ray re transformation and "identity does not matter"

Ray says we’d give up identity in order to transform into the persons we’d most value being. This undoubtedly is true. Its bearing on the question whether identity matters is unclear since it might be due to a general altruistic approach to life. But I think we have already established that identity does not matter in a good enough way. How? [What about PG?} -- INCLUDE THIS AS A SEPARATE SECTION EARLIER, THAT IS, SPELL OUT A SUCCESSFUL ARGUMENT FOR PARFIT, AS IN THE SECOND META-ARGUMENT]

For Martin if I transform enough, then I now will not be the later person into whom I transform becuase we will not be R-related. (Speaking loosely: see above for precise formulation). So I will not be him just as he will not be me.

Bring in Velleman here: (a) I might be genuinely reflexively related to the person -- this at least can be envisioned in a case like Martin’s, where there is no other sort of continuity --wait! What is Martin’s case? There is no memory at all? The plans and projects are totally different? That doesn’t sound right. So need to look more closely at this.] My point here is that possibly I am R-related to the later guy just insofar as I can imagine and my intetion re "him later" (= "me later") is genuinely reflexive

(b) But even if I am not R-related I must have him as the actual subject of the notional me when I strive or decide etc. to transform.

NX says that we are transforming all the time by virtue of which we give up identity all the time: ;I now will become someone who is not hte same person as me now. [Note again: this is not confusion about the identity of the incidents or stages: rather the person who I am now will beomce a person who is not the same person as the person who I now am. This sort of statement would be paradox mongering of course, if it were not grounded in the asymmetry of personal identity through time, a view which itself is as defensible as the alternatives. Or so I argue.] the difference of Martin’s transformation case, though, is that if there is R-relatedness between me now and person later, then I will be that person (though she or he will not be me).

What about fusion? Suppose Ryx and Rzx in a fusion case. So again zKirk=Kirk and yKirk=Kirk. But we don’t have Kirk=zKirk or Kirk=yKirk so again we don’t end up with zKirk=yKirk. What about daydreams? Fission daydream: I survive them ok. But looking back it wasn’t me (of course, looking back it never is me), so no special problem here. Fusion daydreams: both survive as me later. Again no problem.

Rejecting symmetry strikes me as a more plausible way to deal with fission than rejecting transitivity of personal identity insofar as either move is designed primarily to undercut both Rzy and Ryz in the analysis of fission. (Of course one may also wish to deny transitivity insofar as one requires some degree of connectedness in the glue: that is, one may wish to regard the glue as nonsymmetric R plus some degree of connectedness.). It is enough at this point to establish that it is at least not obviously worse than the other alternatives. Evaluating one as better than the others would be difficult since I’m not sure what the aesthetic and ethical considerations would be, but I will return to this question later on. Now I would like to discuss an objection to NX’s solution to fission, the objection is based on Velleman’s theory.

**

Is NX believable?

NX claims that, if I survive until t, I will survive as a later person who exists at t but who will not be me. I will become him, I will be him, but he will not be me. Is this believable or is it just a philosopher’s game? It is not just a game. The possibility of fission shows, as Johnston pointed out in defending the use of the thought experiment, that as persons we are not essentially unified, even though, as Thomson noted, the intuition that we are necessarily determinate and unified is very strong in us. Perhaps someone will someday explain how that intution can be saved. (Perhaps NX can be used to explain it: see section re "full presence" assumption following from no past, and no future yet.) Meanwhile, let us consider our options insofar as we think seriously about the possibility of fission. PG: we survive fission (again, as always, using ‘survive’ so that ‘A survives as B’ entails ‘A is the same person as B’) because one has two bodies afterwards, with the persons supervenient upon each body identical with each other. Not a grotesque distortion, perhaps--at least no more grotesque than the other options; but pretty weird. Parfit/Shoemaker: one does not survive fission, but one would have survived if one of the two fissionees had lived only a short while. Lewis: one does survive, but only because personal identity is not transitive: you could survive as B, where B survives as C, but you do not survive as C. [Add comment about Nozick: also add a section for him earlier]

Given this sort of competition, it is not so odd to hold that you will survive as the person you will become tomorrow, but he or she will not be you. I have treated personal identity as antisymmetric (except in causes of causal loops), because I do not have a way to explain why personal identity would have to be fail to be symmetrical in fission if it were not generally antisymmetric (recall Martin’s challenge to the deniers of transitivity)--and yet its being genereally antisymmetric conforms to the way in which our states arise out of other states.

So --do you have an account that does not bring in any weirdness? If not, my claim is that the NX theory is prima facie no weirder than the other criteria on offer; and NX has some distinctive advantages (not the least of which is survival of fission without the bizarre consequence of PG that a person afterwards will be bi-located for good; or the bizarre theory from Lewis that there were two people all along sharing stages. I have been looking only at Psychological theories, but resorting to a Bodily criterion is of no help relative to fission (even if overall such a criterion is better): the Bodily criterion also needs to solve the fission problem. [Is it any easier for the Bodily Criterion? No, I don’t think so. There can be survival --though certainly not hte PG type. But either Lewis or NX can be appropriated, as can the Parfit/Shoemaker model of nobranching bodily continuity. [See what Unger says about fission]

Copersonality and the reflexive identity theory

Velleman distinguishes between "self" and "person" (p. 66): "selfhood denotes "the relation borne to me by those whom I can think of first-personally" whereas "personal identity" is the relation "among those who are one and the same person." Velleman’s emphasis on "first personal access" recalls Nozick re role of personal values in determining closeness; Martin re role of anticipation in what matters; Schechtman re the self narrative as constitutive; Rovane also?? and even McDowell’s attention to the Lockean formula re seeing oneself as the same... Neither Schechtman nor McDowell would be happy being classified as reductionists, of course, and it does not matter whether the label sticks. My point here is there is a commonality of emphasis in these divergent discussions that are, I think, articulated by Velleman. Let us call these theorists "reflexivists."

Relevant direct psychological connections between mental events for Velleman may differ from the set of direct psychological connections in Parfit’s theory, since Velleman uses only those special types of "direct connections" that are such that they "function like memory in giving us first-personal access to other points of view" (42). In memory there is a relation between events e and f such that e is a memory of f, and the first personal access is to be explicated in terms of "unselfconscious" (identity-free) reference (which "does not rely on antecedent specification" of the target of the reference: such thoughts "require no other thought about whom they refer to" (60)). Notice that Velleman’s account of the direct psychological connections does not reject (at least, need not reject) Parfit’s starting point; he simply restricts the relevant connections to those with a certain type of content. Presumably direct connections have the usual sort of causal connection. This is important because the causal relation between e and f can allow us to make sense of the notion of genuinely reflexive reference here (without using the concept person from the beginning).

It appears to to me that Parfit’s framework for relations between incidents (connectedness, strong connectedness (R*), and continuity (R)) can be adopted by reflexivists like Velleman. Recall that:

Parfit and Velleman agree that connectedness is not symmetric. But there is an important difference in their background reasons for nonsymmetry. Whereas Parfit’s explicit comments about symmetry of R seem to be motivated primarily by the wish to make sure that branching is possible (and perhaps by the "directed" nature of causal relations, where an experience causes a memory but not vice versa), Velleman says enough about the connects to explain why connectedness must not be regarded as symmetric; it is due to the direction of genuinely reflexive unselfconscious first personal reference.

To see this, suppose incidents x and y contain events e and f, respectively, where the memory f makes genuine first personal reference to the person constituted by y, by virtue of y’s containing e (due to the notional subject of e and the relevant causal chain); but it is not necessarily the case that e makes a similar genuine first personal reference to the person constituted by x. So connectedness between incidents is not necessarily symmetric (just as "being connected" is not necessarily symmetric between events).

Even though Parfit and Velleman agree that "connectedness" is not symmetric, they can get different results in some cases as to the direction in which the relation fails. For example, in fission cases they agree that symmetry on connectedness is not exhibited (and that it is crucial for each that it fails, for their general positions about what matters), but nonetheless they disagree about how it fails.

For a simple causal view (associate with what Parfit says), x is connected to y insofar as events in x cause events in y but not vice versa, whereas on Velleman’s "reflexive content" view of connectedness, y after the fission is connected to x, because events in y have genuine identity-free first person access to events in x, but not vice versa. Alternatively, of course, both PG and Lewis presuppose a "connectedness" relation that is symmetrical. (Another way to put this: Velleman restricts the PG extension for "psychological connected"to those pairs in which there also is, let us say, reflexive connectedness, whereas Parfit restricts the PG extension to those pairs for which there is causal connectedness.) These technical points emerge below as crucial for wider philosophical arguments about what matters.

As for strong connectedness, it may be that Velleman would require only a relatively small degree of connectedness in order to get strong connectedness (this would be in the spirit of giving priority to genuinely reflexive reference; fewer types of connections are relevant but when they obtain they have more significance; see Nozick on this).

Whatever he may wish to say about strong connectedness, however, it is clear that it is not symmetric, for the same reasons that connectedness is not symmetric; and neither is Velleman’s R. In this respect, then, Velleman’s R is like Parfit’s (and unlike RG and RL, which are symmetric). As for transitivity, his R ("reflexive-R") is transitive again by definition; presumably neither connectedness nor R* are transitive (for the same reasons Parfit’s are not). And then, to speak of persons, as for Parfit, we let function F map any person A and time t on the incident in which A’s existence at t consists; and, as before:

A is constituted by x at t iff F(A,t)=x.

As for copersonality, finally, Velleman also can accept the strict or lenient parfitian criteria for copersonality. (Nothing about his theory that I can see indicates otherwise, at any rate.) so, for instance, incidents x and y are copersonal just in case Rxy&Nxy or Ryx&Nyx (as on the strict Parfitian criterion).

I have spelled all this out to make sure that Velleman’s view coheres generally with Parfit’s about personal identity; and it seems to me that it does. They differ only about the substantive nature of connectedness (and Parfit might have good reasons for becoming a reflexivist; one consequence of what I argue here will be that Parfit could accept Velleman’s conception of connectedness with minimal consequences for his overall account).

Now let us return to "selfhood." Well, why can’t this just be understood in terms of the substantive reflexivist conception of psychological connectedness --or perhaps strong connectedness-- or continuity? That is: why not say that

(We also might replace the right side with "x ix strongly connected with y," or with "x is continuous with y.") And then (speaking of persons rather than incidents) A at t is a "self" to B at t* just in case A at t is connected to (or strongly connected or continuous with) B at t*.

As for the alternative readings in terms of R* or R, I think it would be in the spirit of Velleman’s discussion to adopt the weakest reading here (in terms simply of connectedness: one direct connection making for genuine reflexive reference is sufficient for self-relatedness; see Martin re identification: suppose that "in at least one situation one can appeal to [salient dimensions ...that facilitate identification] to explain why it makes sense for that person--without altering his basic beliefs and values--to choose to cease to exist in order that a continuer who he does not believe is himself may begin to exist" (APQ 328); if eg one believes will not survive fission, then even given extreme psychological discontinuity, on this view one might be "connected" --and even "self"-related-- to the successor insofar as one now identifies with the later person, and insofar s one’s so identifying plays a causal role in future states of that person), and this is what I will presuppose in what follows (although I do not anticipate that it will make any difference; but of course Velleman himself might prefer a stronger conception of what is necessary for self-relatedness, even if he were to accept the framework offered).

This interpretation of Velleman is plausible since Velleman’s conception of connectedness (at least as exlicated above) already incorporates first-personal access and unselfconscious reference. In fission cases, Velleman wants to say that incident y after the fission may be connected to x before the fission, because y has first person access "back" to x (or, more precisely, since the reference is to persons not incidents, any person constituted by x), but x does not have first genuine first person access "forward" to y (or, that is, to any person constituted by y). But as noted this already can be built into his notion of connectedness, and it is built in given the definitions proposed above: y is connected to x but not vice versa.

And this interpretation of Velleman strikes me as attractive because of its simplicity and because it ties directly into Parfit’s argument that identity does not matter, which is a connection Velleman wants to make (42): whether we define self-relatedness as connectedness, strong connectedness, or continuity, it is clear that self-relatedness (self identity) can diverge from copersonality (personal identity) because of the nonsymmetry of self-relatedness and the symmetry of copersonality.

After all of that, then, one might expect that the distinction between person and self would be relevant to Velleman’s discussion of fission. But in fact fission does not drive a wedge between person and self, for Velleman. He treats fission as driving a wedge, rather, between selfhood and "psychological engagement" (p. 74), not between self and person. Personal identity and self identity both fall on the same side of the wedge in fission, as Velleman interprets it (and in any case they surely do for Parfit and my Parfitian reconstruction of Velleman): there is neither personal nor self identity in fission --but there nonetheless is "psychological engagement".

What is important here, I think, is that even though "person" and "self" fall on the same side of the wedge in fission for Velleman, as for Parfit, they ("person" and "self") would NOT fall on the same side of the wedge for PG or Lewis. The incidents x and y in a fission case as diagramed here turn out copersonal for PG and Lewis, as we have seen; but x cannot be self-related to y, if Velleman is right.

Diagram. Branching of "connectedness"

Even though a permissive conception of R (such as RG or RL) treats x and y as copersonal, y cannot be self-related to x if Velleman is correct. So if it can be established that being self-related is what matters, then Velleman will have established a logical wedge between what matters and copersonality as it is conceived by PG and Lewis-- which of course would make for an objection to the PG and Lewis claim that copersonality is necessarily coextensive with what matters.

That, at any rate, is how I will understood Velleman to be following Parfit in arguing that identity does not matter, as he claims to do (42): "What matters most," he suggests,

The point is that self identity is more important than personal identity; and, Velleman argues, it is self identity that matters not personal identity. It seems to me that it helps, to understand Velleman’s fission argument, to focus on theories like those of PG and Lewis because they entail survival of the person in fission. But if Velleman is right there is not, and cannot be, the sort of access that makes for self identity in such cases (even if there were identity, as PG and Lewis say there is). And if self-identity is what matters, one could not have what matters even if one survived. This type of argument for the conclusion that personal identity does not matter differs from Parfit’s fission-based arguments, as we will see: Parfit argues that one would have what matters even without identity, whereas this type of argument says one would not have what matters even if one survived.

**

An argument that identity does not matter based on the "reflexive identity" theory -- note: put this in footnote, or in a parenthetical paragraph: keep the focus on Velleman as an objection to NX’s analysis of fission.

Velleman claims, like Parfit, that personal identity does not matter; rather, it is self identity that matters. Velleman’s work bears on Parfit’s arguments that identity does not matter in two ways. First Parfit holds that in the fission cases, one would have what matters but not identity, and suggested that it is continuity plus connectedness that matters. Velleman argues that it is "self"-relatedness that matters, and he also argues that in fission one could not be self-related to one’s fission successors. Therefore, his argument if successful would show that Parfit and Lewis are wrong to think that it is R plus connectedness that matters (assuming now connectedness does not include self-relatedness) because in fission there could be R plus connectedness but there could not be self-relatedness (which is, or is a part of, what matters). As Velleman suggests, however, Parfit easily can accommodate the insights of the reflexivists by accepting their substantive proposals about how to think about connectedness, and I tried to make this explicit earlier by explaining how it can be done. That is, Parfit could revise his view of connectedness and R (about the substantive natures of which he says little) so that connectedness is coextensive with, or includes, self-relatedness (just as earlier I identified self-relatedness with connectedness in a context where genuine first personal access is essential for connectedness). It is more central to Parfit’s overall work that he be able to show that personal identity does not matter than that he be able to defend a particular answer about what does matter.

Secondly, Velleman’s disagreement with Parfit about what matters actually opens the door for a new argument for the claim that personal identity does not matter. For if Velleman’s argument about the fission case (that one could not be self-related to one’s fission successors) is successful, it can help Parfit with the wedge he needs against PG/Lewis. Recall that PG/Lewis hold that there is no logical wedge between personal identity and what matters, since (they claim) what matters and copersonality are necessarily coextensive. If Velleman is right both that self-relatedness is what matters (or is a component of what matters) and that in fission cases there could not be the the sort of access between me and my fission successors that makes for self identity, then one could not have what matters in fission even if there is identity (as PG and Lewis say there is).

This type of argument for the conclusion that identity does not matter differs from Parfit’s fission-based arguments, since Parfit argues that one would have what matters even without identity, whereas this new type of argument says one would not (and could not) have what matters even if one did survive. This would an effective argument against PG/Lewis, and more generally, for the claim that personal identity does not matter. Of course, the wedge is explicit for theories that say one would survive fission; for the nonbranching R theories that deny survival there is no similar wedge. But perhaps the argument also can be formulated in terms of plasticity: it is controversial whether one could survive fission (due to plasticity), but one could not have what matters in fission; therefore identity is not what matters. But my main interest is in the argument that would be effective against a position that posits survival of fission.

To consider this argument, I will assume arguendo that self-relatedness is, or is a component of, what matters insofar as we want to survive (Martin re anticipation; Schechtman re self-narrative; Nozick re relativity to my own values), and will focus my attention on Velleman’s argument that one could not be self-related to one’s fission successors.

**

The failure of Velleman’s argument against the possibility of being a self to a future fission successor.

To put my cards on the table: I will be trying to show that Velleman’s argument does not work. In the end, I believe, he does not succeed in showing that one could not have what he calls self-relatedness in the fission cases. And this means that we still will not have found a successful argument that copersonality is not necessarily what matters insofar as we want to survive. [NoTE: given the weak interp of "self"-related as simply connectedness: well that provides a different basis: one can say there is what matters with just one connection as long as it involves genuine reflexivity. This shifts the argument: and the PG response can be that if that is enough for self, then it is consistent to suppose it is enough for "person" as well --so one could move to a very weak conception of strong connectedness for use in the criterion for copersonalty. So that type of argument will not work either.]

The crux of the matter, according to Velleman, is that I could not think "first personally" about each of the two later selves in fission. Why not? Why couldn’t I think "first personally" about the future selves in fission? Velleman claims that even though I could be "psychologically engaged" with each (so that each later could have genuinely first person reflexive thoughts about me now), I cannot now have genuinely first personal reflesive thoughts about either of them. More precisely,

To understand his view, let me briefly discuss his use of both "notional subject" and "unselfconsciously."

Velleman distinguishes the "notional subject" of a thought from "actual subject" so that in imagining that I am Napolean surveying a battle scene I am the actual subject of the imagining, whereas Napolean is the notional subject. (53) Intention and anticipation have "a notional subject, whom they present as ‘me’" (70). Intentions "project" themselves into the future in two ways: they represent the world from a future point of view, and they are "sent" into the future, "by depositing them in memory for future retrieval" -- but I don’t have to specify "a person from whose point of view I am trying to frame my intention, because that point of view is fixed by the future causal history of the intention itself" (70), and the referent of "me" in the later context is "simply whoever fills the role of subject within that perspective" (that is, the perspective in the future at which the intention arrives to guide action). Under normal circumstances, my later self

To say that a thought is "unselfconscious" about its reference is to say that it requires "no other thought about whom they refer to": "genuinely reflexive thoughts don’t rely on an antecedent specification of their target: they just point to the subject, at the center of thought" (60). This is Velleman’s way of expressing "identity freedom" of thoughts that refer to a subject even without there being anything like picking out the subject as the thing being referred to.

Being "self"-related per Velleman may offer a way to explicate "identification" (Perry, Martin in "Self-interest and survival" (APQ): Martin seeks an analysis for "identification" that might help explain ...[what exactly? Choice in branchline by earthling to let replica have the good experience where it is assumed the earthling is not the replica --so the question is to explain how earthling’s choice might be "narrowly self interested" even though he would not be choosing for himself: indeed he focusses on the "narrowly self-interested trading of continued identity for other benefits" (as in Shoe’s example with the operation). (Martin notes that for Lewis and others the problem does not arise since there would be identity in all of the cases; but Martin says he himself has argued the problem arises in other cases as well: see Transformation: include a section to discuss this.)

The way Velleman helps Martin here: while there can be explicit first personal judgements of identity ‘that was me..’ ‘that will be me ..’ these thoughts themselves depend upon identity free memory or anticipation. Martin considers some suggestions: Perry’s proposal (Martin calls it "empathy proposal":

But Martin says : still might not be self interested.

Then Rovane: one identifies.. To the extent that she anticipates being psych connected wtih them.

Velleman’s argument that before a fission I could not make either successor person "the notional subject of my anticipations unselfconsciously" is as follows:

This argument does not convince me that fission would indeed deprive me of real future selves; or, more to the point (for my purposes), that it would be incoherent to hold that in fission I do have real future selves each of whom could, in Velleman’s terms, be "‘notional subjects" of my current anticipations in an unselfconscious way.

First, we should notice that in our ordinary lives we do not have "unique future perspectives" and so the fact that there would not be unique future perspectives in fission cannot be what results in there being not being real future selves in such a case. Under ordinary circumstances I am confronted with divergent future perspectives insofar as I now can anticipate and prepare for each of several different future circumstances (being in SF tomorrow, and in NY next week).

+-----------------+-------------------+

x y z

Of course Velleman did not mean to deny anything like that is possible. Nonetheless if at x I try to picture my perspective at y "specified merely as forthcoming" (what Velleman says I cannot do re my fission successors) then it is relevant to point out that if this were an objection against being self-related to the fissionees, it also would be an objection against being self-related to myself at several different points in the future. Likewise, about my anticipation of y at x in fission, Velleman says, "I cannot frame it from a perspective specified merely as that in which it will be remembered" (emphasis added): but again, if this is an objection againsted self-relatedness in fission, it also appears to defeat self-related in ordinary circumstances as well, since of course (in either case) the anticipation of y at x might be remembered at z as well. Right now I do not have a unique future successor in the sense that there is one unique momentary perspective awaiting me (at least, so I hope) --I hope there will be many such perspectives in many different places and times, including places like nice beaches in southern Thailand. My first point then is that indeed we do plan for and anticipate divergent future selves.

Of course Velleman would not wish to deny that. More to Velleman’s point, of course, would be the case in which the two unique future perspectives are perspectives on the same spatiotemporal situation. He says, "I cannot be trying to draw it from both perspectives at once" 75, emphasis added). But is it really inconceivable to do this? Velleman’s objection to the coherence of first personal attitudes to each of the two fissionees turns out to be an objection very much like Parfit’s objection to the dual survival of fission view -- that, for instance, one could end up playing tennis with oneself. As before, I claim it is at least conceivable.

To make this vivid, consider time travel once again the situation in Fox’s character, Marty McFly, is observing himself in the parking lot at the end of Back to the Future.

---------+

z

+-----------------+---

x y

t t*

Even though y and z overlap in a situation t*, I don’t see why x could not have genuinely reflexive thoughts inherited by both y and z. Suppose I were a basketball player, preparing at t for an upcoming game at t*, and anticipating the time travel scenario and knowing the persons constituted by y and z (me in each case) will be on differing teams, and indeed will guard each other, I now practice both defense and offense. I form strategies for both offense and defense, where "strategies" make explicit the highly conditional feature of normal intentions insofar as strategies spell out what to do under various circumstances.

Now suppose I experience a basketball game and then time travel and play it again (so to speak). This time I guard myself. Etc. In each case I can be acting on the strategies I developed in practice. ("When the opponent fakes a shot don’t stupidly just jump, etc..."; "when planning to make a shot, give a head fake first..."). Now of course when I am guarding myself, the better player in that game (and maybe indeed the outcome of the game) will depend on whether I am better at offense or defense. I agree it is weird, but of course it is time travel that is weird (and, of course, don’t forget that the context of the discussion is an argument advanced on the basis of considerations from a fission case). The question here is whether in practice I can make strategies that in the game are implemented by both me’s in the game. I say I can do it.

Other stories could be told to make vivid its conceivability. Well, ok (if you insist): later I time travel then, yes, have a sex change, fall in love with myself, and now I imagine that splendid day when I marry myself. As you know, in a wedding there is a routine for both the bride and the groom. I now sit here vividly practicing both roles over in my mind; or, to be more precise, I practice the role of the bridegroom for awhile, then I practice the brides’ role. Since I can imagine being a groom and doing what a groom is supposed to do; and since I can imagine having a sex change and being a bride and doing what a bride is supposed to do, I can indeed practice doing everything that I need to practice. This is not to say, of course, that I can practice both roles at the same time at any given moment. But of each role, I can practice it now just as easily as I can first go over what I need to do tomorrow when I get to SF, and then next week when I get to NY (even though it would be difficult to go over both plans in my mind at once).

What is unusual adding the time travel twist is that there will be or could be situations in which I am implementing my strategies in two different ways in a single situation.

I claim that imagining what I will do in such a situation given the the time travel twist does not even require new conceptual resources. I rely on my ability to strategize (use conditional intentions) and on the fact that I know I will have future selves in diverse situations. The fact that given the time travel twist my future me’s might interact playing basketball or getting married does not necessarily undermine my ability to do anything that I normally can do.

But before moving on to consider fission, notice that, at the least, Velleman doesn’t describe the problem or coherence correctly (if such there be) : he talks about a "future moment when I have two successors, and he talks about trying to picture that moment as it will appear in an experience specified "merely as forthcoming." The point about time travel is simply that because (a) in ordinary circumstances we do confront a whole series of divergent future perspectives and can and do routinely plan for them, and because (b) it is coherent to suppose that incidents overlap, given the coherence of time travel, it follows that it is coherent to suppose that in a single situation two or more future people are implementing my current strategies, and that I can anticipate them doing that just as they will be able to remember my anticipations. Moreover just as I can now envision what I will doing in SF tomorrow and then in NY next week, so also I can envision interaction of my two future me’s on the basketball court or at a wedding.

As noted earlier, when Velleman says "my anticipation may be remembered in two different perspectives, and so I cannot frame it from a perspective specified merely as that in which it will be remembered" (75) -- we certainly should not read this uncharitably so that it would defeat ordinary planning when I am coordinating what I will do in SF tomorrow and then in NY next week. My planning for each of the two successors (one in SF and one later in NY) is, in each case, conditional on expected perceptions, memories, etc that each will have at the time when action will be appropriate, first in SF, then in NY. My question then, of course, is why couldn’t planning for each of two simultaneous successors work in just the same way? Why can’t conditionalization of plans and so forth work there just as it must be brought in to make sense of how we coordinate more than one future self?

Velleman says, about fission, "In order to specify the perspective from which I’m trying to picture the future, I’ll have to identify it with one of my psychological successors or the other. That is, I’ll have to pick out the person whose perspective is the intended target and destination of my projective thoughts... My anticipation of the future will be nothing more than an act of imagination" (75).

The time travel case shows, it seems to me, that one could plan for two simultaneous successors in just the way that one can plan for two successive successors under ordinary circumstances. The simultaneity of successors in itself does not introduce any complications with which we are not already familiar because of the fact that we can plan for diverse non-simultaneous successors. So it cannot be the fact that, in fission, my successors may overlap in time that defeats my ability to anticipate and plan for each in a genuinely first personal way. And indeed, since fission does not require temporal overlap (Parfit’s philosopher and psychologist story), we can tell stories in which there are no times at which incidents from the two branches will overlap. In those cases, my plans for my fission successors can be very much like my plans for successive selves under ordinary circumstances: intending for each of them is very much like intending for different parts of my normal future when I will be in NY tomorrow, and in SF next week. Of course, in the fission case, I will not be able to assume the later later-me will know things just because the earlier later-me learns them. But in planning now for the later later-me, I can assume the later later-me knows pretty much what I know now (discounting for forgetting, as we normally do; and so --writing things down). And of course the earlier later-me can also can write notes for the later later-me (like the 30 minute branch in Parfit’s "sleeping pill" case does). In any case, there is no conceptual difficulty here.

It seems to me that Velleman has not shown that the fission case differs in relevant ways from ordinary planning when we know that we will have more than one successor self later (one in SF, one later in NY). True, my planning has to distinguish them in the sense that the plans I make are conditional upon my being in one place rather than the other: when I picture seeing Union Square, this picturing certainly will be associated with other plans and expectations, so that it is correct to say that I identifying this image with one my successors, namely, me in NY. But this surely does not mean that I "have to pick out the person" who is the intended target, since Velleman has argued that ordinary intentions do not require that the person be picked out as such.

In ordinary deliberation I may formulate several different courses of action, dependent upon what happens. In so doing I may think at a certain time t* that if p then I will do A at t; but if not p, then I will do B at t, where p or not p is something that will be settled at some time just before t. Before that time when p or ~p gets settled, I have no unconditional plan concerning A or B.

At t* I do not know whether I will do A at t or B at t, but nonetheless I can anticipate doing one or the other, and moreover I can envision doing each of them -- and this is the important point; and then whether I do A at t or B at t, in either case I regard the agent of the envisioned "unselfconsciously ‘as me’" and identical with the notional subject of the thoughts at t* by virtue of which I envision each of the actions. If it turns out the p gets settled, and I do A at t, then in doing A I make it turn out that I am the actual subject referred to in the thought "I will do A at t, if p" when at t* I was making my plans. But equally, if it turns out that ~p gets settled, so I do B at t, then in doing B at t I make it turn out that I am the actual subject of the thought at t*, "I will do B at t, if ~p."

I assume that these points about deliberation are not controversial, and that Velleman would agree with them. When I realize that p has been settled, and then do A at t, as a result of my earlier intending to do A at t, given p, then agent who does A at t is (or will be) me, since I will turn out to "occupy the position of notional subject in my intention," and I will "turn out to be the person of whom I was thinking first-personally" (to borrow phrases from Velleman, p. 72) when I formed the conditional intentions. Of course, what is true of the agent who does A at t, would also have been true of the agent who would have done B at t, had ~p rather than p been settled at t*. On some occasions I may even rehearse what I will do, given p, as well as what I will do, given ~p -- so as to be ready.

Now in a fission case, let us suppose, there are circumstances that can be known beforehand concerning the diverse situations each of the two successors later; and indeed I now make conditional intentions that are conditional upon those differences; for example, one of me will be in NY and the other in SF.

Diagram. Branching of "connectedness"

Let p be "I am in NY at t". I now intend conditionally to do A at t, given p, where p or ~p will be settled at some t* shortly before t. I also conditionally intend to do B at t, given ~p. Just as I can form the conditional intentions in a case in which I assume that I will perform A or B at t, but not both, I also can form the conditional intentions in a situation in which I believe that I will both perform A at t and perform B at t. Suppose, relative to the diagram, that incident y contains the performance of A at t, and suppose z contains the performance of B at t, while the earlier x includes the conditional intentions. Velleman claims that x cannot be a self-related to y. But I claim that if in ordinary deliberation of the sort described above, x can be self-related to y because of the action in y based on the conditional planning in x [the diagrams for deliberation and fission indeed can be drawn in just the same ways; note draw it out that way above] then so also x can be self-related to y in this case as well.

In saying that I believe I will perform A at t and that I will perform B at t, I do not wish to imply that I believe that I will perform A&B at t. Normally, of course, if I intend to A at t and B at t, then by some principle of "composition" there is some sense in which I intend to A&B at t --or someting like that. The details about this composition principle are not relevan here, since I mention it only to point out that, whatever the details, there is no such intention in this case (although there is a sense in which it is true, if indeed I survive fission by virtue of the successors); and no principle of composition like that should be presupposed in characterizing a logic of intention that will be useful in planning for fission (or time travelling successors). And it does not need to be presupposed.

Generally, then, I deny that we have "unique future perspectives" in ordinary experience that differs in any fundamental way from the "unique future perspectives" that we would have in future time travels or fissions.

There are differences, of course; but can the differences be shown to be relevant to the question of first personal relatedness? Consider complications in the time travel stories. one of the future me’s in the basketball game, for example, may be able to remember the game from the other person’s point of view (even may know the outcome). That could be odd (of course the game, like we imagined the wedding, could take place after the time travel loop so that neither me knows how the game or marriage will turn out). I don’t see how these types of possible complications will help Velleman’s argument, since there are stories we can tell without those complications. simply having two overlapping incidents by virtue of time travel is not quite to the point.

Perhaps the "no-linear-ordering" point is more relevant. In time travel we imagine two incidents simultaneous in external time, but nonetheless ordered in "personal time", but we do not imagine this in fission. Is this a relevant difference, so that Velleman might grant that two distinct successive selves is coherent for time travel but not for fission?

I don’t think so. First we can add a linear ordering to a fission story without changing much. Indeed fission is like time travel with the least complications of the sort where one of me knows how our basketball game is going to end.

Consider a time travel with only a one hour overlap.

y___ z*_________>

<_______ _z___x

t t*

this is almost like the "branchline" case, except there is a linear ordering via time travel so that there is no branch, that is, Rxy. (There is overlap in external time but not in personal time.) Suppose that at t I know the time travel will take place as described. It seems to me that I easily can plan at t for two divergent courses of action during the hour, so that at t* I am coordinating with myself. Suppose my job at t when I am constituted by incident z (at t*) is to get some information, whether p or not p for some p; and then, according to my plans at t, to do A at t* given p; and to do B at t* given not p. When I am constituted by z* at t, of course, I will simply need to remember whether p or not p, and act accordingly. All of this can be planned at t as simply as if I were planning to find out today whether p or not p so as to do A or B tomorrow.

Speaking of coordination to get things done, if my house were really dirty and the party is going to begin in an hour I might need at t to time travel repeatedly in order to get enough of me to get the house clean in time (before t**).

y5__________________>

y4 _______x4

y3_______x3

y2______x2

y1______x1

________________x

t t* t**

At t* there are a number of me hard at work (in a coordinated way) to clean the house, where Rxy1, Ry1x1, Rx1y2, Rx2y3, and so forth. To coordinate this, all I would need is a list telling me what jobs should be done in what order; and I intend to do the jobs in order (these are complex conditional intentions). So at t I write the list and do the first job. When I get it done I time travel back and remember having done the first job (only), I do the second one. And so forth. Of course, there may be coordination problems concerning use of the vacuum cleaner, but these can be resolved satisfactorily given good will and general self interest.

Is it any more complicated in a case in which I fission, say, six or seven times in order to get the house clean? Well, yes, it certainly is more complicated in a number of ways. For example, there may not be enough food at the party, given all of the me’s who will expect to stay around and eat something (at t I am already pretty hungry even before doing the chores). Coordination would not limited to the one hour or so (unless suicide is envisioned as part of the plan for all but one of me, after the one hour cleaning period--probably counterproductive, by the way, since it would mess up the house again), so there are definitely going to be more complications in fission than in the time travel case.

One type of complication in coordinating six or seven replicas (as compared with coordinating six or seven overlapping time-travelling me’s) is that I cannot use memory to do it. I have a list of the jobs. The way I coordinated in the time travel case was just to go down the list, using memory to determine what was next to be done. How do I coordinate the fissionees? Well, I can use conditional intentions. As part of my plan I set things up so that it will be as easy for the fissionees to coordinate as it would be for the time travellers. How exactly? I don’t know. It depends on the situation. But what I claim is that it is not only conceivable that I do it (which is all I really want to establish here) but also that in at least some cases it would not be difficult to do it. (One way, for instance: after fissioning I intend to gather in the living room at t. All the fissionees have this intention when they "wake up" (let us suppose fissionees "wake up" in the sense that when I am fissioning I am first put to sleep). Beforehand I prepared a dish with numbered slips of paper and I intend to draw a slip of paper out of the dish (I intend this unconditionally; or rather, conditional upon having gone into the living room) and then to compare numbers with others in the room, thereby coordinating the cleaning of the house. (Or no comparison of numbers may be necessary: I intend to draw a number and do the job associated with the number on the list. That is, to put it conditionally, to do job number n, given that I draw the number n.)

It is not difficult, of course, to imagine fission situations in which beforehand I would be overwhelmed by the number of fissionees or the complexity of coordination so as to be unable to do it very well, if at all. (See Unger re 1000 fissionees and the difficulty of identifying with each 269-275; see Martin APQ, 324: who says Unger says "a person may be able to anticipate the experiences of two fission descendents more fully than she can anticipate the experiences of any of a thousand fission descendents" (Martin, 324). But the questions of complexity and conceivability should be kept distinct: even with one thousand fissionees it as at least coherent that I intend (for example) to go to a certain football stadium and find a seat in the stands, and that each of the fissionees inherits this intention, and acts on it in such a way that I become the actual subject of his mental states and actions as he makes his way to the stadium. I suppose numbers could be distributed at the stadium, or the seats numbered, etc, as part of some plan I might make up so as to coordinate further. If I knew it now was settled it was going to happen I might get really serious about doing it, using committees; but even though I could predict the outcomes of the committees, Unger surely is right that I could not really anticipate very much about each of the future experiences of any of the fissionees insofar as I stayed focussed on a complex plan that really would guide each of them.

My primary interest here is conceivability, of course, not dealing in any detail with the relative complexities of 2- or 7- or 1000-person fissions. I claim that Velleman’s self-relatedness is conceivable even with many fission descendents: that there could be genuinely reflexive first personal thought between me now and each of many future fission successors.

The main point here was to reply to Vellemans claim that self-relatedness in fission is undermined because of the impossibility of first person anticipation and intentions. But I have tried to argue that the fact that we can form first person conditional intentions for diverse future successors (NY tomorrow, in SF next week; or conditional intentions diverse later circumstances) allows us to see how it is possible to have first person intentions for each successor.

It might be claimed that, even if I am right, we could not form unconditional intentions, or have unconditional anticipatory thoughts, by virtue of which we are "self"-related to any successor. Even if this were true, it would not defeat my response to Velleman; but, in any case, I don’t see why even this point should be accepted generally. Suppose I now look forward to playing basketball, without anticipation of fission. I plan unconditionally to play and I am looking forward to it. But my anticipation is in terms of types of events, not token events. If I also now anticipate fission, but know both successors will be playing in the game, it seems to me that my anticipatory thought, occurring before the fission, is such that each of the successors can become the actual subject of the thought in just the normal way. The fact that two or more of them are doing it is irrelevant. Just as Velleman grants that each can be self-related back to me now, so also I now can be self-related to each of them.

One might believe that unconditional intention involves straightforwardly intending individual actions. But this claim depends upon an assumption about intending and planning that is false. Just as David Makinson argued that individual actions are seldom obligatory, since there will be infinitely ways to perform any obligatory act (type), so also we seldom if ever intend individual acts. Rather we intend act types. For instance, about obligations, suppose that I am obligated to give you a dollar; there is no individual action that my obligation requires. It requires me to do some individual action or other than is of the type "giving you a dollar". The individual act itself can be performed in infinitely many ways. Likewise for intentions. The individual action as performed was not intended as such, but was intended as a type of action. Since plans and intentions typically involved planned or intended types of actions, not the token actions themselves by virtue of which are plans and intentions are realized, our planning and intending is highly conditional in nature (even when implicitly so). I intend to walk across the street. This means implicitly, if there is a truck coming to wait, then look... etc etc But since intentions are highly conditional in the respect that I have described here, indeed it is not difficult to extend our normal planning to imagine a single situation in which I am at once drawing from two distinct sets of experiences of that situation as that token situation unfolds. So when before fissioning I intend simply to go to the living room after the fission: this looks like an unconditional intention (or it at least can be interpreted that way if any can be). Each of my fissionees inherits the intention, so each thinks when he wakes up, "oh, I should go to the living room now" --and each of me gets up and goes to the living room (except for one or two of me, perhaps, who are struck by weakness of will and fall back asleep). Each performs an individual act due to the intention to do the act-type "going to the living room". I do not see why each of the fissionees in thinking what he thinks as he acts so as to realize the intention cannot thereby become an actual subject of the pre-fission thought by virtue of which I intend to go the living room after fissioning.

Of course, there could be thoughts now which could logically have only one successor subject, such as "I tomorrow will play basketball with only one successor in the game" (what I might be thinking after I decide not to fission). But if I think that but fission anyway due to due to my fission addiction, so there are two of me who show up for the game, then neither successfully may do what I intended to do so as to become the actual subject of the intention. But so what? Obviously fission will change some things; my purpose here is to reply to Velleman’s claim that I now could not be self-related to each of two fission successors.

I conclude that Velleman does not show it. It would at least be coherent, then, for one to adopt Velleman’s account of psychological connectedness, and to accept his claim that self-relatedness is what what matters (or a component of what matters) while also rejecting the claim that identity does not matter; and this is what Lewis and the PG theorist have been saying all along. And so we have not found the wedge that Parfit needs to show that identity does not matter.

**

Two Trains

In most of his arguments about the RAMS, Parfit focusses on (A) potentialincreased separation within the life of a single person rather than (B)decreased separation between distinct people. Let me explain. A and B are alternative "takes" on the shift to reductionism. To make it more precise, consider two relations R1 and S1 as represented in this diagram:

When I shift my beliefs to reductionism, I may come to believe that the relations R1 and S1 are more alike than I had been assuming prior to the shift. (This may not happen, and indeed the question whether it should happen, given the attitudes I tend to have, needs to be investigated.) But it is plausible that it will and should happen insofar as it can be shown that the relations R1 and S1 are more alike in important respects given nondensity than they would be if we were dense. Given density, R1 is simply identity of a dense entity (a simple, indivisible thing, etc.) whereas S1 is a social relation of some type between two distinct dense entities. Given nondensity, while R1 is again identity, there is no simple, indivisible entity, and the psychological connections between me at t and me at t* have to play a significant role (e.g. it is not necessarily all-or-nothing, hence the possibility of indeterminacy, as even non-reductionists like McDowell can and do acknowledge). These psychological relations may be similar to the social relations that hold between me at t and you at t*. If so, R1 is more like S1 given nondensity than it would be given density. [Ftnte: By the way, the reason Liddy talks about the "switch to nondensity" rather than (like Parfit) the "shift to reductionism" is to include nonreductionists like McDowell in the discussion. McD is not a "reductionist" because he denies that the relations that constitute R (such as memory and intention) can be given a non-circular "identity-free" analysis. But nonetheless, insofar as he rejects aspects of density he also can ponder the Rams. ]

The two distinct "takes" on the shift to non-density are views about what should happen, concerning my attitudes about R1 and S1, given that the relations become more alike when we make the shift:

TAKE A: My attitudes about R1 should become more like my antecedent attitudes about S1.

TAKE B: My attitudes about S1 should become more like my antecedent attitudes about R1.

These "takes" can be generalized to include reference to the past and attitudes about the past, and to include reference to social or causal relationships other than friendship. On A, my attitudes about my own future or past should become more like my antecedent attitudes about the futures and pasts of others, whereas on B my attitudes about the futures and pasts of others should become more like my antecedent attitudes toward my own future and past. To see the difference between the two Takes on the shift to nondensity, let us consider an example.

Consider death. There will be a time later when I do not exist. On take A on the shift to nondensity, my attitudes become more like my antecdent attitudes about the deaths of others. Even when I have cared a great deal about them, I still have tended to be less concerned about their deaths than about mine. Shifting to nondensity via Take A, one comes to care less about one’s own death: one may have more equanimity of the the type which Parfit expresses when he says that what my death means is simply that at later times there will be no events or states R-related to mine now. One cares less about one’s own death. The Take B effect is different: instead one comes to care more about the deaths of others insofar as one’s attitudes shift to become more like one’s antecedent attitudes towards one’s own death. Liddy points out that if one adopts both Take A and B, which is possible, one develops more equanimity about one’s own death while also coming to care more about the deaths of others. Similar points can be made about injuries or harms that one or others suffer. On a Take A shift to nondensity, one comes to care less about harms to oneself, whereas on B one comes to care more about harms to others.

**

FBI email tap authorization #3459sZA399

1-22-99

--[Edited for Flynt]--

[Moldur to Flynt: Larry, just ignore Scully. She was having one of her bad days.--Fox]

Date: Thurs, January 22 1999 22:13:56 -0400

Subject: Hey! Come back! Take the B train too!

From: Zo <zoalex@aol.com>

To: parfit@nyu.edu

Mime-version: 1.0

X-priority: 3

Status:

Dear Derek,

Hi again. Ok I’ve got some time now. Let me review. Much of your work has been devoted to showing that shifting from non-reductionism to reductionism about personal identity would have interesting consequences for our views about rationality and morality, and you have made spectacular and melodramatic claims, for instance, that continued survival does not matter, that punishment is not justifiable, utilitarianism is more plausible, given reductionism, and so forth. You knew the shift to reductionism was important and you thought it should make a practical difference.

What I saw this week is that your Fission Fixation is connected directly with your Predilection for Take A on the shift to reductionism. This comes out most clearly in your discussion of egoism especially when its contrasted with Brink’s attempt to give the egoist her run for the money. I’m a bit reluctant to talk about egoism because I don’t think it needs to be refuted, either as a psycholgoical theory or a theory of rationality --it is utterly implausible to me either way -- but looking at how Takes A and B figure into the discussions about egoism might help me explain some more important points. So I’m going to try to do that.

First maybe I should remind you of the difference between Takes A and B. On the A shift to reductionism one increases "separation" within oneself -- more detachment about one’s own future. One’s attitudes about one’s own future become more like one’s antecedent attitudes about others’. On B, on the other hand, there is less separation between oneself and others, and one’s attitudes about others become more like one’s antecedent attitudes about oneself. One example I talked about was re death. There will be a time later when I do not exist. On take A on the shift to nondensity, my attitudes become more like my antecdent attitudes about the deaths of others. Even when I have cared a great deal about them, I still have tended to be less concerned about their deaths than about my own. Shifting via A, one comes to care less about one’s own death: one may have more equanimity of the the type which you express when you say that what my death means is simply that at later times there will be no events or states R-related to mental and physical states right now. One cares less about one’s own death. The Take B effect is different: instead one comes to care more about the deaths of others insofar as one’s attitudes shift to become more like one’s antecedent attitudes towards one’s own death. If one adopts both Take A and Take B, which is possible, one develops more equanimity about one’s own death while also coming to care more about the deaths of others. Similar points can be made about injuries or harms that one suffers oneself or that others suffer. On the Take A shift to nondensity, one comes to care less about harms to oneself, whereas on B one comes to care more about harms to others.

Or look at the case where I survive another’s death (perhaps we were in a serious accident together where I survive but she did not). On Take A there may be no effect at all in making the shift. Of course my joy in surviving may be tempered by her not being around anymore --but only because I may miss her, perhaps feel guilty in some way, and so forth. The shift to nondensity itself via Take A wouldn’t change anything (since it pertains only to my attitudes about me now and me later). But on Take B there is an interesting change (when I shift to nondensity) because my attitude towards her death becomes more like my antecedent attitude towards my own death. My joy is tempered not only because I miss her but because her death becomes to me more like my own death. Similarly, adopting Take B on another’s surviving my death, my attitude towards her survival becomes more like my attitudes toward my own. Her surviving becomes to me more like my own surviving.

If you want interesting and significant ramifications of the shift to reductionism, look no further. If one shifts to reductionism via both Takes A and B then one will be less likely to harm others in situations in which some type of self sacrifice is required in order to avoid harming-- and this means one is more likely to respect the universal precept not to harm. Perhaps the effect would be small. It is a modest claim. basically the rest of what I have to say concerns why this is significant and more abstract arguments relatively are not very important. It probably would be best just to shut up now so as not to obscure this point.

But anyway back to egoism. You use reductionism to argue against egoism’stemporal neutralism which says one is a rational person only if one has equal concern for all the parts of one’s life. You argue that, given the reductionist picture, especially the idea that one is constituted by relations over time, where these relations can have varying degrees of strength --given all this, one rationally at t could give less concern to parts of one’s life that are less strongly connected to oneself at t.

Now one might think that Reductionism also would threaten the agent bias aspect of egoism, and David Brink discusses this, but instead of looking very closely at this possibility, his main interest is in TURNING THE TABLES

[Moldur to Scully: See! I told you! -- Moldur]

[Scully to Moldur: Shit Fox you idiot! Quit interrupting! Just shut up!--Scully]

to show how reductionist considerations can be used to BOLSTER egoism by explaining how self concern can be expanded to generate self-derived concern for others.

--I was wondering why you didn’t look into this question yourself? I mean the question about reductionism threatening the agent bias aspect of egoism --it would have been such a natural question for you to pose, since in a number of passages in your writings you note that reductionism reduces the separation between yourself and others. But so far as I can see, you never apply this idea to the discussion of egoism -- indeed you rarely mention this point in your discussion of any of the ramifications of switching to reductionism except for the discussions of that bizarre 19th-20th century moral theory known as utilitarianism :)

I conjecture you never looked at agent bias because you got stuck in Take A on the shift. Let me explain in a more general way. The contemplated shift in beliefs (from belief in density to belief in nondensity) plays a role in arguments like the ones you gave in Ethics against the justifiability of desert and intrapersonal compensation, for you focus on the question, How should such a shift affect other beliefs? Likewise you use the shift idea when you argue that identity does not matter if reductionism is true. All of these arguments depend on the no-survival use of fission AND YOUR ADOPTING TAKE A ON THE SHIFT IS OF A PIECE WITH YOUR NO-SURVIVAL USE OF FISSION. Even though there is R-relatedness in fission, on the no-survival interpretation there is not identity, and self-concern is not justified. (Nor can ordinary conceptions of desert, intrapersonal compensation, etc be applied, insofar as they presuppose identity. Nonetheless since there is R we have what matters so identity is not what matters.) All this is Take A since in each case you are increasing intrapersonal separation.

But if as per Grice --yes! Grice!! --- you would survive fission, all these arguments fall flat. Since there is both R and identity with each of the people after the fission, you cannot drive a wedge between identity and "what matters" and also there is no basis for rejecting application of the ordinary normative concepts of justice. (Obviously there might be plenty of practical problems but those are beside the point here.) And the survival interpretation is a Take B perspective insofar as it decreases the separation between you and (what might seem to be) two distinct people afterwards. (I remind you I’m not saying the Take B survival interpretation is required; I’m saying only that it is coherent, and insofar as you want to give general arguments about the Rams you shouldn’t cling to the Take A type no-survival interpretation).

Your focus on temporal-neutrality rather than agent-bias when you try to formulate reductionist arguments against egoism also exemplies the Take A pattern. In fact your exclusive adoption of Take A is exemplified most vividly in your argument against egoism (the "self interest theory"). You attack the idea that one rationally-should give equal concern to all the parts of one’s life (the "temporal neutrality" feature of egoism). Since the degree of connectedness between me at t and me at later times can vary, it is not irrational for me to believe I should care more about me at those later times when I am more strongly connected to me at t. Normally this would mean caring less about my distant future since normally the degree of connectedness will be less then. Why should any of this follow? I suggest it seems to follow because you assume Take A (and given Take A it does indeed follow). When relation R1 is such that me at t : me at t* involves a relatively low degree of connectedness (perhaps no direct connections at all) then R1 begins to look like a relation between two distinct people. So (as in Take A) my attitudes toward that later me become more like my attitudes toward other people such as you, with the following effect:

My concern for others generally is proportional to the degree of connectedness with them, so if R1 should become more like S1, as Take A proposes, then indeed it is plausible to suppose that my degree of concern for myself at various time should be proportional to the degree of connectedness. Contrast that with the effect of B:

Your tendency to Take A is illustrated again in the following passage (about both egoism and the question whether identity matters) wherein you are discussing Susan Wolf’s criticisms:

Now I do not want to leapfrog here into a discussion about the claim "identity does not matter" (a claim wholly unsupported except by fission-based arguments--which don’t work, and cannot work, as I hope you realize by now) except merely to point out that the "identity doesn’t matter" theme is a Take A perspective on what matters. On Take B, my identity can matter just as much as ever, but my attitudes towards the pasts and futures of others might become more like my antecedent attitudes about my own past and future. So one who adopts Take B (only) on the shift will never agree we should cease to care about identity, and there is no reason for her to think that relationships with others will become more fragile. On the contrary, one’s separation from others is decreased (--but you seem to have just forgotten about that!).

Secondly, you assume here that natural egoism often is bad for us and others so that its being lessened would have good effects. Again you are revealing a Take A predilection insofar as you are assuming the shift to non-density would lessen my own self-concern--rather than increase my concern for others. Natural egoism may not be so bad. Rather it may be that it is bad only when it is not complemented by enough concern for others (that is, natural altruism). It is density that doesn’t matter (but nondensity is compatible with identity) and it is excessive clinging to density (not self concern) that can be bad for others.

In any case, this gets us back to Brink’s story on behalf of the rational egoist (for whom of course any natural altruism is to be justified, if it can be justified as rational, in terms of egoistic considerations). Whereas you take A and ignore B, I suggest that Brink does just the opposite: he Takes B and ignores A in his development of rational egoism in terms of reductionism, for instead of decreased concern for one’s later self, he emphasizes that reductionism allows us to make sense ofincreased concern for others (his special interest being to show how this concern can be understood as derived from self conern). To see how what I’m calling Take B figures into Brink’s discussion of rational egoism, recall first that there are two central features of rational egoism: agent bias and temporal neutralism. He wants to defend egoism (at least to give it its due, that is, to see how it relates to other-directed moral considerations given the best case we can make for it) but first he notices that reductionism might threaten agent bias. To illustrate the threat he turns immediately (I regret to report) to the fission thought experiment (thereby short-circuiting useful thought). Now you know my opinion about the fission thought experiment which is that it is never going to prove anything because of the plasiticity of our interpretations of it relative to survival vs nonsurvival. Indeed it is the plasiticity of our interpretations that is the important point (one that is echoed by the way in the whole dispute about egoism itself --insofar as the problem itself appears only if one assumes density: when one abandons density the question of the source of reasons --in me or in others -- becomes less pressing and interesting). And it is only with extreme reluctance that I discuss fission here. As I pointed out in my thesis

So that is why you just forget about fission, as I told you before. But now to see where Brink is going we have to look at what Brink does with it. Ok, then: fission. Let’s talk about it.

Blah blah blah I now will be divided so that later there are two bodies blah blah blah each of which support mental life that is psychologically continuous (R-related) with me now like Kirk getting teletransported to both planets X and Y. You guys assume branching continuity entails no survival (whereas I side with the great beautiful H. P. Grice about this: since we define R so it cannot branch, Kirk would survive-! HURRAH!! --but I am firmly resolved not to get sidetracked in that stuff here! :) But since Brink does assume Kirk would NOT survive (and by the way this is not quite coherent with his Take B approach, I would argue) -- and since Kirk would be fully R-related to both X-Kirk and Y-Kirk (let us call them) later, there looms the following problem for egoism:

To solve the problem, Brink posits that people can be related to each other in various ways so that one person’s good can be a component of the other’s. Even though (we are assuming for the sake of the argument) X-Kirk is a different person from Kirk, the idea is that because of their R-relatedness, X-Kirk’s good is a component of Kirk’s. This is a good point: it is a Take B point. He is developing your idea that that reductionism reduces the separation between people. If so, the "iNTERpersonal psychological continuity" found in no-survival fission, Brink says, "extends the agent’s interests in much the way that iNTRApersonal psychological continuity does" (127, caps added). And then he generalizes this to ordinary social relationships such as the relationships between parents and children, between friends, colleagues, and so forth-- and he assumes that there are "psychological relationships" that are part of these social relationships.

Now I feel he moves rather too quickly in assimilating psychological and social relationships--in fact, his discussion is just a blur at this point. But, in any case, I believe he is right: the generalization can be made (and, as noted, this is related to your insight that reductionism decreases the separation between people). The details about R matter here as to whether it can be generalized in an interesting way to social relationships ("S" let us say).

But assuming the details can be worked out connecting R with S, then it is plausible to suppose, as Brink claims:

I agree with Brink that this route promises a defense of egoism insofar as derivative noninstrumental concern for others can be grounded in the psychological relationships that are components of significant social relationships. For instance,

In this way Brink develops what he calls metaphysical egoism wherein nondensity can be used creatively to explain how the R- and S-relatedness of two individuals explains how each is justified in regarding the other as "‘another-self" (Aristotle) or an "alter ego" (T. H. Green), see Brink, "Self-love and Altruism," p. 143. So a picture emerges in which the metaphysical egoist in principle can justify other-regarding concern and conduct. He definitely can make use of nondensity, for instance, to make sense of the Aristotelian idea that the question whether or not I live a happy life can be dependent upon events occurring long after my death (these events being S-related to mine now).

I don’t really intend to get into the plausibility of egoism here. I have some ideas how to extend Brink’s discussion so that the egoist gets a good run for her money (how to characterize the S-relation, explain how it can extend even beyond those with whom we are directly S-connected, etc) but it seems to me very unlikely that the egoist could make the case that it is irrational not to adopt the picture she paints. It is a great picture, possibly inspiring, but not rationally required --which means it is mistaken as a theory of rationality the way it usually is put (one has reasons to do x only if x contributes to one’s own happiness). It seems to me here there is an analogy with our fight about how to interpret fission. If someone has a theory that entails that one or the other interpretation HAS to be adopted, then that theory is in trouble. Likewise about rationality, if one has a theory which says that the interconnected web of R- and S-relations HAS to be understood as one in which reasons for action are derived from self-concern then it seems to she is missing the point about the web itself. For instance, why assume that concern along lines of R (and S) connections should be regarded as derivative upon self concern. Why not the other way around (with concern for self deriving from fundamental concern about R- and S-relations). This is a pretty obvious way simply to set aside any form of rational egoism, since if R- and S-based concern is not derivative upon self concern, then of course agent bias is abandonedis. (Note: this issued related to Parfit v. Johnston re args from "above" or "below"; also see Shoe review re egoist response: shifting selfcentered concern to mirroR.) Even if a coherent response is forthcoming, it is quite implausible that the response would be strong enough to defeat the claim that self concern rationally can be derived from R- and S-based concern. It is this point which leads to the unravelling of any form of rational egoism (given nondensity). To see this better, suppose one is contemplating a shift in beliefs from density to nondensity, as in your arguments. There are two initial points to keep in mind about the shift.

(1) A change in belief is not necessarily a change of heart (motivational structure). There may be interesting consequences of the shift to nondensity that go beyond changes in belief.--For instance, Hume getting depressed. Your becoming more cheerful, worrying less about your death, feeling a glass wall separating you from others has been dissolved, and so on. But for the moment I want to focus only on changes of belief.

(2) The consequences of the shift for beliefs are not logical consequences. A change of belief rarely if ever logically requires other changes in a belief system. Of course a belief change often has a ripple effect throughout one’s system of beliefs. But since the ripple is not a logical ripple, a change in belief by itself will UNDERDetermine the change.

q, ~p, q->p, and ~q

Assume now that you are a psychological egoist, meaning you never do any action x unless you believing doing x will contribute to your own overall welfare; and you also are a rational egoist meaning you believe you never have a reason to do x unless it would contribute to your own welfare. Now suppose you are in a situation in which, re-reading Part III of your book, and being reminded of your arguments for reductionism, you make the shift to reductionism, and you do it adopting BOTH Takes A and B together: your concern for your own future decreases and your concern for the futures of others with whom you are S-related and potentially S-related increases. And suppose, --oops! -- under the circumstances, you go to a certain extreme with Take A and your self-concern plummets all the way to ZERO!! Whoa! Your concern for your future suddenly is like the concern you would have had (or, rather, not havfe had) for someone wholly outside your sphere of concern. In this situation your concern for others cannot be derived from your self concern because (in this story) you don’t have any. If this shift is not irrational, then it seems to me that rational egoism is shown to be false.

To consider a melodramatic example, I plan to sacrifice my life in five minutes to save the life of Erin’s daughter, Zo (its a long story) with whom I am strongly S-connected, and of course I’m strongly S-connected with Erin too. I have shifted to nondensity, and lo! --I overdid it, so that at the moment, contemplating my death in five minutes on behalf of Zo, under the circumstances my self concern has plummeted to zero. I think, "wow. In six minutes there will be no events taking place in heaven or on earth or under the earth that are R-connected to these thoughts right now. So what? It doesn’t matter what happens to me" for I have a great deal of equanimity about my own death, plus my heart is overflowing with joy when I think about little Zo, about whom my attitudes are as if she were carrying on my own life, and how my actions in the next few minutes are going to be causally related to hers (after which I’ll be blown to smithereens, but she will escape and run to safety; --its a long story) and I also will be S-related to those whom she influences and affects, including her children and their children, and there is a magnificent stream in the world of which my actions are a part, --it is very beautiful to behold (and not because I’m fantasizing some otherworldly myth or even some this-worldly myth like N’s brave myth about eternal recurrence, but I’m simply looking facts in the face with a certain frame of mind) -- and I have similar attitudes as well about Zo’s mother, my dearest friend Erin, into whose waiting arms Zo soon will run after I’m blown up to defuse the bomb in order to get us, I mean her, away from the poison gas--

Would the envisioned shift (including the plummet) be IRrational? Since reductionism (or at least the nondensity belief) is true, it certainly isn’t irrational to shift your beliefs accordingly. And recall that the shift in beliefs underdetermines our changes in beliefs or other attitudes, so one objecting to this shift must object to the particular details of this shift.

Now one might object that Take A itself is irrational. One might presuppose a sort of all-or-nothing concern for one’s future (related to temporal neutralism). This is highly implausible, given nondensity, as you show, but even if we grant arguendo that one’s concern for one’s future is temporally neutral, the all-or-nothing considerations do not kick in if one has no concern at all for one’s future. That is why I posited self concern plummeting to zero in this story. No doubt in many circumstances that would be irrational, but I suggest it is not necessarily irrational, especially, when one is contemplating imminent death for good reasons.

But even if one has no concern, might not one be acting against one’s own welfare in a way that is irrational (even if admirable etc)? To see why this is not necessarily what is going on in the story, recall that given nondensity, one’s diachronic unity is grounded in R, that is, relations between various types of events and states of body and mind. These include long range projects which both (i) are central to determining what my own welfare is and (ii) they have a complex logical structure. As for the logical structure of projects, the 3-D structure is a good logic for this with its distinction between ideal and guiding norms. My projects are such that ideally I’ll live to a ripe old age etc etc but they are complex enough to include guidance for non-ideal circumstances, and when I find myself in such circumstances the guiding plan may differ radically from the ideal ones. And (under certain circumstances) the guiding plan may be one in which by design my self concern plummets to zero! It is part of the plan for those nonideal circumstances, entailed by my general projects and by the facts of the situation. The point is that I am not acting against my interests in that situation in which I have no concern for my future. Practical reasoning takes place in a context in which my interests drop out completely. Derek, there are passages in buddhist texts where it seems that the buddha seems to be suggesting that reasoning, thought, life can take place in a context in which thoughts of the form "this is me," "this is mine," etc do not occur.

Of course the metaphysical egoist may respond as follows: you may feel no concern for your life, and let us suppose that your life plan is indeed such that in those nonideal circumstances, all things considered you have no interest in protecting your life, but nonetheless your own interests necessarily are represented by Zo’s and Erin’s and others to whom you are S-related. Your joy in contemplating the long chain of events of which your actions are a part simply is the joy of a life whose good endures and perdures like the ripple in a pond far from the source of the ripple itself. Ok. Yes, I see that point. That is true. However, the particular shift to nondensity that I posited was one in which self-concern plummeted to zero but one saw the value of one’s action and its role in the world and took joy in that but with no concern about oneself, due to the Take A aspect of the shift. The limit of decreasing concern for one’s own future is zero concern, and when this shift is not merely nihilistic but rather accompanied by the Take B increase in concern for others -- there is no reason why that should be described as a situation in which one’s other-directed concern is derivative upon one’s self-concern or as one in which the other’s interests really are one’s own interests by virtue of S-relatedness.

What is more important, I would say, than refuting egoism (or any of the other abstract arguments-- about whether survival matters, or how to think about various aspects of justice) is that we can observe in our own minds how shifting to nondensity increases appreciation for the basic precept not to harm. This is what I said earlier where I probably should have just shut up. Just as you pointed out we care less about our own death when we shift (via Take A) we also can observe increased interest in and respect for others when we shift to nondensity via B. And that really is important! Even when we do care a lot about others and are trying to be helpful, it nonetheless is easy to get stuck in a self-centered orientation of mind. This is obvious within oneself and all over the world, and so much of human culture is dealing with the problem of reducing self centered greed. If we come to care more about others, by making a B shift to nondensity, then we will be more likely not to harm them. That is, we will be more likely to respect them, to respect the moral precept not to harm the neighbor --which Nietzsche himself recognized as universal, explicitly pointing this out in the preface to the Genealogy (of all places!!!? --Makes me wonder if perhaps nobody has really understood him) even as he was gearing up to launch his skeptical attack on conventional morality!

And, Derek, I hope it goes without saying that I am not talking about any leap to some univeralist consequentialist theory here. One just needn’t get locked into a model that contrasts egoism with (i) "presentism" --me-now based reasoning-- OR (ii) "neutralism" which is impartial among all beings -- so that the alternative to egoism when concern is extended to others seems to be some sort of impartial universalism where people are replaced by utility functions. Well, no! That is completely different from simple basic respect for the precept not to harm one’s neighbor (those with whom I am S-related, where S might be spelled out in terms of people in the neighborhood = beings whom I can effect causally in a significant way). We don’t need to universalize the maxim of our action, etc, we just need to try to make sure that the S-ripples are not going to harm (where, I would have to add, not benefitting does not = harming). The shift to nondensity via B makes it easer to see the world this way without having to try to do it. (Contrast a dense self afraid of suffering eternally for not doing it.)

The shift can help open one’s mind to something which has little to do with arguments for abstract views like consequentialism in moral theory. The shift cannot help you or anybody defend utilitarianism or unravel commonsense views about justice (desert, intrapersonal compensation, etc). In your ambitious arguments about the ramifications of reductionism you did just the same sort of thing that the buddhists did when they recognized the potential significance of the shift to nondensity and then leapt to their "no self" jargon. Now you certainly took a balanced and nuanced position about that issue: yes, we do exist, but we’re not dense. In other words, you didn’t go eliminativist about the metaphysics. --But you did tend to go eliminativist (or make an analogous move) about the normative issues!

Perhaps the most surprising result would be that our actual lives and practices come to make more sense, given the shift. Parents do take responsibility for their children (recall Take B on desert). People make sacrifices all the time for others or for some project way out of proportion to what they can expect to get out of it were they calculating in terms of an algorithm that assumes density of self. We obviously are not egoists in real life (or when we are, it is pathological). Yet, even if so, there is something interesting going on, for it does seem you are correct in insisting that we tend to belief that (1) we are dense; yet now I’m suggesting that (2) in real life we not need to "shift" as much as we have been pretending. This is an open question. A conscious shift to nondensity may help us appreciate what we are really like --to "become who we are." For our natural altruism makes more sense when we appreciate we are both not dense and yet S-interconnected. When we shift to nondensity, we understand our own natural altruism better. At least that is one option. (Nihilism is another, related to exclusively going Take A: here Take B looks better, espcially when you say that considerations of prudence may be replaced by morality -- instead of prudential considerations governing respect for my future, it becomes more like a moral interpersonal issue. Great! All we need is more guilt! Notice that for Take B what was moral becomes prudential as for the metaphysical egoist.) When our interconnectedness doesn’t make sense, it is easier to get stuck in a drab boring lifeless separateness. Since it cannot be taken for granted (even if is natural), it is good to have practices to cultivate it in the mind, like those, for instance, like the buddhist practices described by Sharon Salzberg inLovingkindness: The revolutionary art of happiness. And were a mind conditioned in this way, it seems not unrealistic to imagine a mind going beyond good and evil, whether moral or prudential. That really is what I had in mind about abandoning density of self in a way so that the question of egoism collapsed. I realize I haven’t really explained it. --actually I am in way over my head. Sorry if I’ve been wasting your time. I need to talk with Erin about this. She is trying to get at this stuff in her Nietzsche book. I will try again later. The idea is that of course I’d blow myself up to save the girl --if somebody killed the girl to save their own life we would be horrified -- I wouldn’t do that (--at least I hope not!) -- and when we focus on an extreme case we see how the nondensity picture might help us do what we’d be ashamed not to have done. But in less extreme ordinary life the same points apply. What I have in mind is respecting the precepts so that the causal ripples are not coming back in forms that equal remorse and regret etc so the mind develops tranquillity, then one lets go of density of self, then one is way beyond good and evil (and just naturally does not harm). --Which (I would like to argue) is how most people I know do in fact live.

The disconnect between (2) real life altruism and interconnectedness and (1) belief in density of self reminds me of the first time I read Descarte’s Meditations. I was a freshman in college, and I clearly remember sitting in my room when it dawned on me that Descartes was saying that I am a nonphysical being not existing in space and that I could exist independent of my body and independent of anything else in the physical world. I was so amazed! I couldn’t believe somebody was putting forth a view like that! I jumped up immediately and ran across campus to my professor’s office --She wasn’t there so I went outside and looked around. AND THEN LO!! -- IT DAWNED ON ME THAT THIS IS WHAT I ALREADY BELIEVED ABOUT MYSELF!! --For it was part of my religious belief system that I Am A SOUL etc etc --so now I am first of all stunned that somebody like Descartes was arguing something so weird, and then I am even more stunned to realize that this crazy view was exactly what yesterday I would have asserted I believed myself! Of course the double realization dissolved my inclination to think that I was a soul (or at least began the process -- which your book continued--and, possibly, which continues still) but my point here is that until that day I really had very little appreciation for WHAT I THOUGHT I BELIEVED! And yet the whole "soul" idea actually was very central to my thinking about my life until that time even if at the same time how I really lived was such that when I see Descartes make it explicit and try to argue for it I realized it was ridiculous. (Russell: the chief merit of proofs is they engender skepticism about the result proved.). An even better example in contemporary philosophy is where Chisholm posits agent causation in a straightforward analysis of our commonsense intuitions about our role in action, but then later on admits he can’t make sense of it himself. (But I don’t know if he explicitly acknowledged what I’m calling nondensity --do you know?) I hope this makes sense and you see what I meant in the points about the connection of reductionism with real life rather than just theories about life.

**

Martin re transformation and what matters

Martin argues that the transformation argument shows that identity does not matter independent of fission cases.

Bring in his points about anticipation as well. He holds that anticipation has to be a part of what matters. This is much like Velleman holding that first personal access should be a part of what matters.

Does he hold that I could anticipate the transformation?

The PG response: either weaken requiement for strong connectedness so that there is copersonality; or deny that one has what matters in re survival, and that this is a case of altruism. (Notice that altruism might make more sense if we can be self-related to or anticipate or write self narratives that incorporate (Schechtman) states of others. See below). But there is no argument here to show that copersonality and what matters in survival are not necessarily coextensive.

**

The unity reaction.

Just as reflection on fission challenges the assumption of an essentially unified self through time (Johnston), so also reflection on time travel challenges the assumption of an essentially unified self at a time. The thought experiments are valuable if they make vivid what it means not to be essentially unified. Chisholm said it was "clear and distinct" to him that there would be "entirely definite answers" to question a like "will I be Lefty?" where ‘Lefty’ refers to one of the fission successors. Of course Chisholm among the most erudite of contemporary philosophers, would not use the phrase "clear and distinct" lightly--he means to be saying that it is as certain as it can get. And, as Thomson comments, we share the same intuition; the inclination to say what Chisholm says "is very strong in us" (225).

And yet we have before us theories that give differing definite answers (yes v. no) to a question like "will I be Lefty?" --and the differing definite answers are due to differences about an abstract theoretical point, the symmetry of connectedness, which is a question that often as not gets answered in order to get the result that the theorist wants in the fission cases --I claim this is a plausible reconstruction of why Parfit avoids symmetry; whereas Lewis scrambles to ensure that his relation is symmetrical to build an opposing view (quote what he says on p 24).

Parfit has tried to use this sort of indeterminacy (what I called conceptual or C indeterminacy) to construct an argument that identity does not matter; and of course he also has given similar arguments (also based on fission) for "extreme" conclusions about normative matters.

I think it is clear that his arguments that identity does not matter have not worked, just as has no argument for that conclusion that I have considered here (or that I have heard of). I think it is unlikely that such an argument will work, precisely because of C indeterminacy, and the conceptual flexibility that provides the recalcitrant theorist, such as we have imagined the PG theorist to be. So I conclude that we are unlikely to see an argument to show that given reductionist premises, identity does not what matters to us insofar as we wish to survive.

But this doesn’t change the fact that reductionism entails that we are not essentially unified, and as we noted this contradicts an intuition "that is very strong in us"--which means that if we accept reductionism we accept a view that is inconsistent with an intuition that is very strong in us; so if we can make our beliefs coherent with our intuitions, our intuitions are going to change. What the failure of the "identity does not matter" arguments show, I think, is that the changes are not going to be logically required changes. But of course changes in belief rarely are (see Harman in Thought: inference is not deductive), and it would be even less likely that changes generally in attitudes would be logically required (such as changes in attitudes that might be associated in various ways with the belief that identity matters).

But even if not rationally required by acceptance of reductionism, there may be changes in beliefs and attitudes that make sense, given the change. Instead of trying to construct arguments to show how reductionism requires changes in our usual theoretical views about rationality and morality, it is better to show how some changes would make sense given a shift to reductionism, or how certain changes are easier given the change in view

BRING IN ZFs here [see Zfs re easier to sacrifice for others ....]

For example, the difference in past v future self-relatedness makes it easier to be self-related to future states of others than to past states of others. [Spell this out as below in terms of token/type relation....]

Here is where the Two Trains come in.

Two trains

Train B: less separation between me and others.

Martin on anticipating the experiences of others and empathy.

Velleman focuses his article on the example of imagining being Napolean, so even though he’s really more interested in anticipation and other future oriented states, his main example of imagining being another person involves the Napolean example directed to the past. He discusses the future only relative to my own future or the future of a fission successor.

When Martin talks about anticipating the experiences of others his main interest is in the fission like thought experiments in which it is controversial whether the successor would be me, but even if "for theoretical reasons" (as he puts it repeatedly, in a manner that ties in well to my discussion of C indeterminacy) even if for theoreticall reasons I believe that I will not survive, nonetheless I can anticipate the experiences of the successor.

Alternative Story stuff

Why not simple anticipation of the experiences of a person whom I know to be distinct from myself? That is, where there is no theoretical question of identity. I know the experiences will take place, let us suppose, and I imagine being the person experiencing what I know will happen. This can be like imagining being Napolean. But of course that isn’t remembering being Napolean; so why should this be anticipating being Sally, for example, receiving an honor or getting married. But wait: I’m not talking about anticipating BEING Sally. I’m talking about anticipating her experience.

More like recalling the experience of another: this could be done via implanted qusasi memory, even though it would not be remembering BEING Napolean, if I had a genuine q memory of his experience on the battlefield then I could recall the experience.

What about the future experiences of a distinct person? Well for myself there is a difference between remembering and anticipating: what I anticipate is a TYPE!!!!!!!! This is Makinson’s point again.What I recall is a token event or situation, and my recalling it is causally tied to the occurrence of that event in that situation. So the token event itself plays a causal role. But things are very different with respecto anticipation. When I anticipate seeing a movie, there is no token event of seeing the movie that is causing my mental state. Rather I have desires, beliefs etc such that I token "my seeing a movie" in some way or other. But the essential thing here is that it is a type not a token.

And so just as Velleman explains how one can imagine being Napolean where N is the notional subject of the imagining, and I am the actual subject. Similarly I can anticipate your seeing a movie (not only in the third person sense of anticpation, eg anticipating an event like the announcement of the results of the election) but from the first personal point of view where you are the notional subject of my thought, I am the actual subject, and my thought is a first personal "my seeing a movie" type thought.

Suppose we aere going to the movies together. How does my anticipating my seeing the movie differ from my anticipating your seeing the movie in the first personal way? Well it is a horror movie which sort of movie I hate but you love. So when I anticipate your seeing it I anticipate feeling good; but when I anticipate my seeing it I anticipate feeling bad. Another big difference, following Velleman, is that my anticipating your seeing it never is recalled in such a way that I become the actual subject of the notional "I" in the thought (which is part of the experience of my anticipating your seeing the movie) that "I am feeling so happy seeing this movie" whereas my anticipating my seeing it may be recalled so that I do become the actual subject of the thought whose notional "I" in my anticipating my "I am feeling so terrible watching this movie."

the fact that anticipation like all the other genuinely reflexive first personal thoughts is "identity free" in the sense that there is no specification of the subject... This is what makes possible anticipation of others’ experiences. Indeed "identifying" with them indirectly.. Not by explicit identity statements -- which would not be identity free-- but simply putting oneself in their situation, imagining their situation the way Velleman describes imagining being N...

Now earlier I specified a big difference between memory and anticipation: the token event versus type event difference. .. This is direction of fit type stuff , see Searle.

This is of course different from imagining how you feel right now. There you have an access I don’t have. And of course in anticipating how you will feel in a few minutes you have access to a lot of relevant information that I don’t have.

But it is different when we consider how you will feel in two years from today. MAYBE I CAN ANTICIPATE THAT PRETTY MUCH AS WELL AS YOU.

Now move to Zoe’s alternative story: combine the CAUSAL points there with MY CURRENT CAPACITY TO ANTICIPATE THE FUTURE EXPERIENCES OF OTHERS, INCLUDING EXPERIENCES OF PEOPLE WHO DO NOT NOW EXIST BUT WHO WILL EXIST LATER ON... For I can anticpate TYPES of experiences, which is all that is involved in anticipation of distant future experiences (even if not so clearly is it all that is involved in near future experiences, assuming that current experiences are usually quite relevant to anticipation in those cases since the near future experiences will tend to evolve out of current states, so that you have access to that process in a TOKEN specific causal way that I do not have access)

HERE is what to combine with the causal points in Zoe’s alternative story: Given the point about DISTant future: WELL THE SAME POINT HOLDS ABOUT MY ABILITY TO ANTICIPATE MY OWN DISTANT EXPERIENCES. So I can anticipate the experiences of others just as well as those of myself.

Notice this is true of me in this life: it is not only a point about what would matter about survival of death (the way I so far have framed Z’s alternative story insofar as she shifts the "what matters" discussion to "what would matter relative to survival of death"... By "true in this life" I mean it is true now that I can anticipate the futures of others as well as I can anticipate my own future some twenty years from now. .. Note: this might relate to specification of "self relatedness" for Velleman in terms of simply connectedness or strong connectedness within a theory in which first personal access is central to connectedness. FOR I NOW AM NOT GOING TO BE CONNECTED OR STRONGLY CONNECTED TO ME TWENTY YEARS FROM NOW. Or rather if I am it is very difficult to know exactly what aspects of me now will be the ones by virtue of which I now will be connected or strong connected with the incidetns that constitute me at that time.

seeing things as another sees them, feeling how another feels .. Empathy.

Mudita

....

Similarly, just as Parfit’s Combined Spectrum argument challenges the assumption of a necessarily determinate self across time, the reality of multiple personality disorder challenges the assumption of a self necessarily determinate at a time.

.....

What Blackburn calls the "unity" reaction, in contemplating future action, underlies the intuition that two or more fission successors would be not identical. But the coherence of the time travel examples calls that intuition into question; and the unity reaction does not necessarily support nonidentity in fission either.

I conjecture that what underlies the "unity" reaction is the belief that tends to be associated with first personal thought that I am fully present as the thinker of the thought. If in each moment of thought and action, I were fully present as the thinker and agent, then it would be not implausible that (relative to diachronic unity) indeterminacy would be impossible: because fulll presence may seem like an all or nothing sort of thing. No matter what happens, either I will be fully present within some incident or not present at all. On the other hand (relative to synchronic unity) if I am fully present with or within one incident, then there hardly could be a distinct incident which does not overlap the one I am fully present within and which yet also constitutes my existence at that time. (This underlies the initial rejection of one person in two places with distinct perspectives and plans.)

[For "overlap" see initial definitions: incidencts x and y overlap just in case there

The combined spectrum argument shows that the diachronic full presence should be rejected; the time travel argument shows that the synchronic full presence should be rejected. Diachronic indeterminacy is possible, as is synchronic division.

Here is where the diachronic R relation can take precedence over "synchronic" unifiers of incidents. The time travel cases show that what unifies incidents cannot be taken as what unifies the person in external time. The unity reaction is relevant only to unity of incidents. But it is not decisive when it comes to self identity (as Velleman conceives it) or personal identity.

Conclusion: Velleman’s discussion does not provide Parfit with ammunition to reply to the PG theorist who still assumes that what matters and personal identity are necessarily coextensive, since this theorist can incorporate "self identity" as an important component of R [this needs to be spelled out]

** [note this should be included earlier in discussion of Parfitian criterion]

Kantian challenge to Parfit "from the past"

Kantian challenges to Parfit come from two directions --roughly, from the past and from the future. The challenge from the past is very roughly this. I now have knowledge based upon my past experience of the world, and yet such knowledge is inconceivable apart from the assumption that I am a unified subject of the knowledge. The subject of my current apparent memories must be the same as the earlier experiences that are remembered if I am to have knowledge on the basis of my current apparent memories. Moreover, when I make inferences about the world from the first person perspective I rely on memories whose content involves identity-free reference to myself; this identity free self-reference is an ineliminable aspect of first person thought about the past. The challenge to Parfit from these points comes in two forms: (a) the necessity of unified subject, (b) the coherence of "impersonal description". RE (b): parfit himself replies effectively to this point when he says that the concept "I", "me" and so forth can be used in the content of thoughts without this undermining an impersonal description thesis. McDowell seesm to overlook this point in his attack on the intelligibility of the reductionist position (see quote on p. 239)

But this is false. The reductionist position is not rendered unintelligible by the fact that the identity of the remember is represented in the content of the retained knowledge. On the contrary, the reductionist position holds that the relevant connections between thoughts can be characterized without presupposing the identity of a subject; but it need not claim that the thoughts themselves have no content of that type (look specifically at Cassam about this). Rovane in "branching" article has a good brief discussion of this.

But McDowell may grant that point, then, but object that to regard oneself as having knowledge based on the thoughts, one has to assume the identity of subject; and the reductionist then who grants that the person has genuine knowledge based on memory also may be committed to a similar thesis. And it seems to me that the reductionist can grant the thesis --but claim that the identical subject can be supplied within a reductionist picture. (What need not be claimed is that the knowledge itself can be understood as justified prior to understanding what the subject is in reductionist terms.) My point here is that the R-related series of incidents that constitutes a person’s existence through time provides an objective touchstone that can play the role of the "subject" insofar as empirical knowledge requires an identical subject persisting through time. Footnote: see Blackburn, 189).

Even if, as Evans claimed, normally one cannot question whether apparent memories are derived from one’s own experience, it is conceivable that there could be situations in which I coherently question in just that way --I seem to remember such and such, but I also have additional information (about a memory transplant; or that I was or may have been fashioned in teletransportation, or in one of the middle cases of the Combined Spectrum; or that I am ZK, a fission product; and so forth) which leads me to question whether the apparent memory is derived from my own experience. This, of course, is a question that can make sense in reductionist terms. I can ask, am I the person who experienced what I now seem to remember experiencing?

Of course there may be situations, such as when I am created in one of the middle stories of the Combined Spectrum, in which my memories are such a jumble that I cannot form a coherent self conception precisely because there are two distinct R chains combined (a fusion case).

To deepen this section: see Andy Hamilton and Cassam (article and book).

[Include above somewhere in discussion of Velleman; that is return to Kantian objection from the past. If V has an objection against self-reference in fission, then we might expect a similar objection against selfreference re memories in a fusion case. fusion for the past may parallel what Velleman will say about fission for the future. However, I must admit that I don’t even begin to see where the problem is here. Even if I find myself with two radically differing sets of memories, purporting of course to memories of a single life (insofar as there is unselfconscious use of me --of course, my general point is that neither memories, nor intentions nor anticipations do purport to be about a "single life"--THAT is extra content], there is nothing in principle INCONCEIVABLE about my making sense of the jumble. If I necessarily assume that my apparent memories are derived from my own experiences, are there ANY two memories that cannot in principle be rendered coherent?

Well, let us do our worst. I seem to remember being Madonna, making love with Parfit, on March 6, 1996 (I remember seeing the calendar as I was making love with Parfit); and lo! I seem to remember being Parfit making love with Madonna, on March 6, 1996 (I too remember seeing the calendar as I was making love with Madonna). Now this is going to strike me as weird and of course I will grant that forming a reasonable view here will be challenging. (Please keep in mind as we continue that we are in the middle of a story which itself is one of the middle stories of the Combined spectrum -- the basis for an effective argument against the necessary determinacy of identity, a feature of our usual selfconception that we have to abandon. I take there to be a general consensus about thhis point. I.e there is a general consensus that Parfit’s CS arguments work.) ok ok. My point that there is nothing inconceivable about treating all the memories as mine. Indeed there is nothing in general required about my having traced a single spatiotemporal path through the world (see Shoe re this)-- as long as I know that I have time travelled, then had a sex change, etc etc, I have the basis for empirical knowledge. So there is nothing about the kantian challenge from the past that requires that at any given time there be only one of me (so to speak; to be more precise, there is no requirement that I have am constituted by only one incident; even more precisely, no requirement that any incident that constitutes me at a given time shares the same spatial location (the latter description is more precise than the just prior since even in ordinary circumstanyes I am consitututed by more than one incident, given how I spelled this out).

If I can know on the basis of all my memories in the middle CS case, then indeterminacy about my identity need not undermine my knowledge. There is no answer to the question whether I am Madonna or Parfit. (Note on PG’s theory there is an answer to this question in the case in which there is "complete" fusion--that is enough connectedness with each so as to have strong connectedness with each -- in which case both survive; but determinate fusion is very different from the middle cases of the CS where there is indeterminacy due to the vagueness of strong connectedness).

Notice that for PG in a fusion case: if Parfit and Madonna fuse, they become identical. Not quite: there were identical all along! --So PG gets a result that is quite odd. You and I now may be one and the same person due to a future fusion. But this is no odder than that we may be one and the same person due to a past fission. (Here denial of transitivity may seem preferable: you and I are not identical even though each of us endured a certain past fission or will endure a certain future fusion. for Lewis: in the future fusion case you and I now are identical (though his "tensed identity" may help with that) and in the past fission example we are distinct (but of course shared stages before the fission)).

*

kantian challenge from the future

What about the Kantian challenge from the future? This is expressed by Blackburn as follows, in summing up his discussion of the "unity" reaction and its place in a theory about personal identity:

Blackburn says he is sketching "a Kantian opposition to Parfit" and he writes as if the unity reaction presents itself as an objection to Parfitian reductionism. It is not exactly obvious what the objection is. Presumably, as in Korsgaards’ article, it is the opposition inherent in the "unity" that seems to be presupposed in agency as opposed to the disunity that is conceivable given a passive bundle of incidents. But why can’t the unity that is presupposed in agency itself result from the causally interconnected incidents that constitute our existence as persons? After all, the events that constitute incidents (and then persons) need not be "passive"; on the contrary they can be as explosive as a temper tantrum.

Here Velleman’s discussion may be relevant because he presents an explication of how a causal picture needs to be supplemented by an analysis of the role of first personal thought in imagination and intentions (framed with the perspective in mind of oneself in the future). I believe he offers a framework in which a Parfitian reductionist can fashion a reply to Korsgaard (maybe even make sense of the unity reaction). So while Velleman indicates he sees his work connecting with Blackburn (see footnote), I still don’t see that it factors into defense of a "yes" answer to the question, "Has Kant refuted Parfit?"

--So examine Korsgaard, show how Velleman points help make consistent the unity reaction re the future with the Parfitian criterion as developed in terms of incidents not only causally interrelated, but also interrelated via "self identity" as explicated by Velleman]

**

The main thing about psychological reductionism generally is that as persons we are not some simple, essentially unified, necessarily determinate type of entity that exists separate from the mental events that comprise our experiences. This is something that western and buddhist philosophers generally agree about. I’m not saying that everybody agrees, of course, but there is a sort of consensus that we just do not have knowledge of anything like a soul, in any way, shape or form; and when we think about it, we see that it is very difficult to make sense of it in terms of our real experience of life.

There also is widespread acceptance of the possible indeterminacy thesis. There is puzzlement about the sources of the "unity" reaction (see Thomson for instance), and some wonderful insights into how the unity reaction leads to a cartesian invisible substance view (McDowell; see Thomson comment; also Blackburn).

But the "unity" reaction does not provide grounds for refuting Parfit, just as the Kantian challenge from the past did not [spell it out]

So Parfit’s reductionism is coherent; but so also is PG/Lewis reductionism --which means that Parfit cannot establish his identity does not matter thesis.

Moreover most of his RAMS arguments depend upon the logical wedge between identity and what matters --so the PG/Lewis positions again prevent the sorts of moves to practical RAMS that parfit has tried to make.

Are there no practical RAMS?

Well, first notice the move from metaphysics to "psychology of perspectives" (V, 41; see Martin reference). This is, I think, similar to the Buddha’s emphasis on practical liberation of (or from) the self rather than on the ontology of the self (or no self)-- even though popular buddhism tends to focus on the latter.

In recognizing the possibility of indeterminacy, there may be effects on beliefs and attitudes and values that are not nonrational (like getting struck in the head) but also they are not such that the change entails other changes (such as unravelling the believe that identity is what we want and should want when we want to survive; or unravelling the belief that one morally ought to keep one’s promises).

**

RE "psychology of perspectives" (Velleman); "anticipating another’s experiences" (Martin): Tibetan practice of exchanging joys and sorrows: this is a practice in taking the other person’s perspective.

Mudita: sharing in the joy of another, rather than feeling envious. Here V’s emphasis on "self identity" (also see Martin re this) rather than an R-based causal interconnectedness may be important.

Discuss various aspects of "reductionism" [include early on]

versions of "reductionism"

supervenience, dependence (Siderits)

no separate entity

no further fact (Johnston)

McD re circularity objection

Cassam

Shoe’s alternative, grounded in Functionalism

Schechtman self narrative

see martin’s book 97

buddhist stuff: siderits re two truths

death seems to disappear (end of Harris article)