Can the Self Divide?

Full text

Can the Self Divide?1
By John Perry
[Can%The%Self%Divide?%Journal(of(Philosophy%69%(1972):%463;88.%%Reprinted%in%
Identity,(Personal(Identity(and(the(Self.%Indianapolis:%Hackett%Publishing,%
2002.]%
%
BROWN, Jones, and Smith enter the hospital for brain-rejuvenations. (In a brain
rejuvenation, one's brain is removed, its circuitry is analyzed by a fabulous
machine, and a new brain is put back in one's skull, just like the old one in all
relevant respects, but built of healthier gray matter. After a brain rejuvenation
one feels better, and may think and remember more clearly, but the memories
and beliefs are not changed in content.) Their brains are removed and placed on
the brain cart. The nurse accidentally overturns the cart; the brains of Brown and
Smith are ruined. To conceal his tragic blunder, the nurse puts Jones's brain
through the fabulous machine three times, and delivers the duplicates back to
the operating room. Two of these are put in the skulls that formerly belonged to
Brown and Smith. Jones's old heart has failed and, for a time, he is taken for
dead.
In a few hours, however, two individuals wake up, each claiming to be
Jones, each happy to be finally rid of his headaches, but somewhat upset at the
drastic changes that seem to have taken place in his body. We shall call these
persons "Smith-Jones" and "Brown-, Jones." The question is, who are they?2
1. A problem for the mentalist?
One thing is clear: they are not each other. Smith-Jones is lying down, Brown-
Jones is sitting up; Brown-Jones is thinking of his nurse, Smith-Jones is thinking
of Jones's companion (they both think of the nurse while being nursed, the
companion otherwise, but right now the nurse is in Brown-Jones's room). So all
sorts of things are true of the one but not true of the other. Perhaps we could sort

these things out in some way consistent with the single-person hypothesis: a
certain person is sitting-in-room-102- I -and-lying-in-room- 104, etc. But there is
no motivation for such maneuvering, for there is no unity of consciousness.
Brown-Jones cannot tell by introspection what Smith-Jones is seeing, thinking,
wishing, etc., and vice-versa.
Smith-Jones and Brown-Jones each claim to be Jones. Certain
philosophers--John Locke3 Anthony Quinton,4 and H. P. Grice,5 to mention just
three-hold theories of personal identity that seem to commit them to agreeing
with both Smith-Jones and Brown-Jones. They analyze the identity of persons in
terms of memory or "continuity of consciousness," or memory and potential
memory. Each would surely want to say in a simpler case of apparent bodily
transfer (such a case as we would have if either only Smith-Jones or only Brown-
Jones survived) that the resultant person is who he remembers being. These
analyses, when applied to the case at hand, give us the result that both Smith-
Jones and Brown-Jones were Jones. Each did all the things Jones did. They used
to be the same person.
My intuitions agree. I do want to say of this case, that Brown-Jones and
Smith-Jones did all the things they seem to remember doing, that they both were
Jones, and so were one another.
But certain philosophers maintain that it is at least almost as clear that we
should not say that Smith-Jones and Brown-Jones are who they claim to be, as
that we should not say that they are a single person--the former, in fact,
following directly from the latter. And, according to these philosophers, what
Locke, Grice, and Quinton are committed to saying simply shows that their
theories of personal identity are wrong. For consider. We agreed that (1) was
clearly true:
(1) Smith-Jones is not the same person as Brown-Jones.

But the claims of Smith-Jones and Brown-Jones, and the results of
applying the analyses of personal identity mentioned, seem to come to (2) and
(3):
(2) Smith-Jones is the same person as Jones.
(3) Brown-Jones is the same person as Jones.
But from (2) and (3), by the symmetry and transitivity of identity, we obtain (4):
(4) Smith-Jones is the same person as Brown-Jones.
Thus (2) and (3) lead us to (4), which is known to be false as surely as (1) is
known to be true. The theories of personal identity that led us there must then be
wrong (Adapted from Williams, 1957, p. 8ff.; see also Flew, 1951, p. 67.)
In this paper, I defend the theories of personal identity in question against
this argument. I shall not say anything about the respective merits of Grice's and
Quinton's analyses, or others for which this case appears to pose a problem,
being content to defend these various plausible analyses against this particular
argument. I shall refer to such analyses as "mentalist analyses of personal
identity," and shall speak of a defender of such as "the mentalist." Of course,
there are analyses of personal identity that might be called "mentalistic" besides
Quinton's and Grice's, and some of these do not commit us to saying, of this case,
that Brown-Jones and Smith-Jones used to be Jones.6 They seem to me to be
wrong for that reason. I agree with Quinton that mentalist analyses of personal
identity need not be incompatible with behaviorist or materialist theories of
mind; the problems of mind and body and personal identity, though related,
should not be conflated.
2. Idea for a solution.
To defend the mentalist, we need to become dearer about the nature of a
philosophical theory of personal identity and about the nature of the objection
being considered.

Let me begin with a simpler problem, that of table identity. Suppose Alf
has a limited understanding of what we mean by 'table'. If we point to a certain
part of a given table, and ask him what color the table is at that spot, he will give
the correct answer. Thus, if we point first to one leg of a brown table, and ask, "'Is
this table brown here?" and then to another spot on the same table, and ask the
same question, Alf will answer "yes" both times. But if we then ask him, "Is there
a single table that is brown here and also brown here?" (pointing successively to
the same legs we pointed to before), he will shrug his shoulders. What does Alf
lack? He does not know what counts as a single table. We might want to express
this by saying, "Alf doesn't know what relation must obtain between this table
and that table (pointing twice to the same table) for them to be the same." But this
is wrong, for Alf might know quite well that the relation in question is identity.
His problem is with the concept of a table, not with the concept of identity. Alf
doesn't know what relation must obtain between a number of table parts, for
there to be a single table of which they all are parts. The rest of us do know what
that relation is, although of course articulating it in a nontrivial way would be a
philosophical exercise of some difficulty. I shall call the relation that obtains
between two table parts, if and only if there is a table of which they are both
parts, the spatial unity relation for tables. When Alf learns what counts as a
single table, he learns to recognize when this relation obtains.
Now take a somewhat different case. Suppose Alf can now very well say
whether this table part and that table part are parts of a single table. But suppose
we point to a table and ask him, "Is that table brown?" He answers, "Yes." We
then move the table to a different room, paint it green, and ask, ("Is that table
green?" He gives the right answer in both cases. We then ask him, "Is there a
single table that was brown and now is green?" Alf shrugs his shoulders, and
cannot answer.
Alf still lacks mastery of the concept of a table. He doesn't know what
counts as a single table, or the same table, through time. Again, he knows what

relation the table that was brown must have to the table that is green for the right
answer to the question to be "yes." The relation of course is identity.
But Alf does not know what relation must obtain between temporal parts
of a table for them to be temporal parts of a single table. It may be objected to this
that we have no notion of a temporal part of a table; what I glance at when I
glance at a table is a whole table, and not just a part of it. But we do have the
notion of the history of an object--a sequence of events in which it is, in some
sense, a main participant. When we glance at a table, we see the whole table, but
we witness only a portion of its history. Alf's problem, then, is that he doesn't
know what relation must obtain between two portions of table histories for them
to be portions of the history of a single table. And now we can simply introduce
the notion of a temporal part by saying that a is a temporal part of b if and only if
a is a part (in the ordinary sense) of the history of b. Alf doesn't know what
relation must obtain between two temporal parts of a table for them to be
temporal parts of a single table. This relation I call the temporal-unity relation for
tables. To analyze it in a non-trivial way is the problem of table identity.
Notice that I am not merely imagining Alf to be in a poor position to re-
identify tables he hasn't seen for a while. It's not just that he's unclear about what
would be good evidence for table identity. Rather, he's unclear about what this
state of affairs amounts to.
Now we all know, in a sense, what the temporal-unity relation for persons
is. But the philosophical problem is, I take it, to articulate this knowledge in some
nontrivial way, to say what the relation is that obtains between temporal parts
(or, as I shall call them, person-stages) of a single person. This relation is what
Grice and Quinton give explicit analyses of, and what Locke suggests an analysis
of.7
It is extremely important not to confuse the unity relation for an object
with the relation of identity. Of course the two are connected in an important
way. If a and b are (temporal or spatial) parts of an object of certain kind K, and

RK is the (temporal or spatial) unity relation for Ks, then, if the K of which a is a
part is identical with the K of which b is a part, a must have RK to b. But,
nevertheless, RK is not the relation of identity, and must not be confused with it.
The logical properties of identity are well known: identity is necessarily
transitive, symmetrical, and reflexive. Now our example shows that the relations
suggested by some philosophers as an analysis of the temporal-unity relation for
persons are not transitive.8
For let j be a person-stage of Jones that occurs before the operation, and let
b-j and s-j be temporal parts of Brown-Jones and Smith-Jones, respectively, that
occur after the operation, and let R be the relation suggested by Quinton or
Grice. Then j has R to b-j, and s-j has R to j. But s-j does not have R to b-j. Now, if
we confuse identity with the unity relation, it will seem clear that R is an
incorrect analysis. Once we have made the distinction, however, it seems a
legitimate question whether R must necessarily be transitive.
The answer, however, may still seem quite obvious. A simple argument
seems to show that, since identity is a necessarily transitive relation, so with any
unity relation. Suppose a, b, and c are K-parts, and RK the unity relation for Ks.
Then if we have a counterinstance to the transitivity of RK:
a has R to b
b has R to c
not-(a has R to c)
it seems to follow that
The K of which a is a part is identical with the K of which b is a part.
The K of which b is a part is identical with the K of which c is a part.
Not-(the K of which a is a part is identical with the K of which c is a part).
But since this consequence is absurd, so must be the supposition.

This argument is, however, mistaken. To shed some initial doubt on the
dogma that a unity relation must be transitive, consider the following case.
Suppose there to be Siamese twins joined at the thumb. Now consider the three
thumbs, a, b, and c (b is the shared thumb). There is a single body of which both
a and b are thumbs. That is, the (spatial) unity relation for human bodies holds
between a and b. And similarly there is a single human body of which both b and
c are parts. But there is not a single human body of which both a and c are parts.
So, if RB is the spatial-unity relation for human bodies, a has RB to b, b has RB to c,
but not-(a has RB to c). Thus the spatial-unity relation for human bodies is not
transitive.
Why does this not lead to a breakdown of the transitivity of identity? The
reason is simple and instructive. It seems that we should be able to infer:
The body of which a is a part is identical with the body of which b is a part.
The body of which b is a part is identical with the body of which c is a part.
Not-(the body of which a is a part is identical with the body of which c is a part).
which violates the transitivity of identity (given its symmetry).
But given the case in question, the referring expression 'the body of which
b is a part' is, of course, improper. There is no unique body of which b is a part.
Thus, given any reasonable theory of definite descriptions, the first two sentences
of our inconsistent triad are not true, and the transitivity of identity is saved.
This point contains the essential insight that seems to me to lead to a
satisfactory reply to the objection in question. But, as we shall see, its application
to the more complicated case of identity through time is not a simple and
straightforward matter.
3. The Branch Language
To apply the point made in section ii to the case of the apparently dividing self,
the mentalist. might argue as follows. The objection is based on my alleged
commitment to (1), (2), and (3):

(1) Smith-Jones is not the same person as Brown-Jones.
(2) Smith-Jones is the same person as Jones.
(3) Brown-Jones is the same person as Jones.
I seem to be committed to these because I analyze personal identity in
terms of a relation, R, which does obtain between the person-stages j and b-j, and
the person-stages j and s-j. But in fact this does not commit me to (2) and (3).
These sentences contain a proper name, 'Jones.' But this proper name turns out,
contrary to what everyone thought, never to have been assigned to a person at
all. The person-stages of Jones (as we say) that occurred before the operation
were stages of both Smith-Jones and Brown-Jones. Thus any attempt to name a
person by identifying one of these stages would have miscarried: although we
had identified a single person-stage, we would not have identified a single
person and so would not be in a position to assign the name. Thus (2) and (3)
turn out on my analysis to be untrue. I can say that both Smith-Jones and Brown-
Jones did all the things done (as we say) by Jones; they did the things they
remember doing. But there is no single person Jones they both were. And there
was no single person doing the things they did, just as no one person would
have pushed a button if that button were pushed by the shared thumb of the
Siamese twins mentioned in the last section. Thus, although my analysis of the
relation that holds between person-stages when they are stages of a single person
is not logically transitive, this does, not commit me to the absurd denial of the
transitivity of identity.
Before considering the merits of this response, I must make a
methodological digression. The mentalist's problem is that he seems to be
committed to an inconsistent set of sentences. But sentences can be judged
inconsistent only in the framework of a theory about their truth conditions. The
sentences in question state that persons had in the past or will in the future have
certain properties. Any solution to the mentalist's problem will then be a theory
about the truth conditions of such sentences, which shows either that the

particular sentences in question are consistent or that the mentalist is not
committed to them.
The method I shall use, to analyze alternatives, is to state the truth
conditions of sentences about past and future properties of persons, in terms of
statements about the properties of person-stages and the temporal-unity relation
for persons, which I shall refer to as "R." It turns out that, even if we agree upon
the analysis, of the unity relation and upon the properties of the various person-
stages in our example, there are still alternatives as to the account we give of a
person having a property at a time. It is my intention to consider these
alternatives, and to argue that one of them solves the mentalist's problem.
I need first to explain, however, under what conditions a person-stage has
a property. To do this, I must first distinguish between basic and nonbasic
properties. A person's basic properties, at any time, are those properties which he
has in virtue of events that occur at that time. His nonbasic properties are those
which he has wholly or partly in virtue of events that occur at other times. If a
person is now in room 100, but in a few minutes will be in room 102, then he has
both the properties being in room 100 and being about to be in room 102. The first is
basic, the second nonbasic.
Let 'P' designate a basic property. Then a person-stage x, which occurs at
time t, satisfies the conditions for having P if and only if every person of which x
is a stage satisfies the conditions for having P at t. Alternatively, x has P if and
only if "This person has P," uttered while pointing to x, is true.
This method of assigning properties to person-stages does not presuppose
a prior understanding of personal identity-of what a person is. The conditions
under which an ostensively identified person has a basic property may be known
or stated without knowing or stating the conditions under which a person will
have or has had that property. This is the point that was made in the discussion
of Alf and table identity. Alf knew that every table before him was green,
although he did not know whether the table that was brown a moment before

was before him, and did not even know under what conditions that would be
true.
Without the distinction between basic and nonbasic properties, the
method of assigning properties to person-stages would be circular. In order to
assign a nonbasic property to a person at a given time, one would have to know
under what circumstances certain things happened to that same person at other
times.
I assume that the nonbasic properties that a person has at a given time are
a function of the basic properties he has at that and other times. Having made
this assumption, I feel free to ignore nonbasic properties in the sequel. The
project is to examine accounts of the truth conditions of sentences of the form 'N
has F at t, where `N' names a person, `F' designates a basic property, and `t'
designates a time.
The defense suggested in this section amounts to one theory of this sort. It
asserts that we speak what I shall call the branch language. Let us say that a set of
person-stages is a branch if and only if all the members of the set have R to one
another, and no stage that has R to all the members of the set is not a member.
Given a mentalist analysis of R, all the person-stages thought to be of Jones, plus
all the post-operative stages of Smith-Jones form a branch, and all the person-
stages thought to be of Jones plus all the post-operative stages of Brown-Jones
form another. The set containing all the person-stages in both of these branches is
not itself a branch. The view suggested is that the history of a person forms a
branch; there is a one-to-one correspondence between persons and branches, the
branch of each person containing just his person-stages. To say that a person has
a certain property at a certain time is Just to say -that there is a person-stage
belonging to that person's branch, which occurs at that time and has that
property. The view needn't be that persons are branches, but for the stake of
simplicity we shall suppose that that identification is made.

Now suppose that persons acquire names in the following way. Names
are assigned to person-stages---say at baptism. The name names the person
(branch) of which that person-stage is a member. A sentence
N has F at t
is true if and only if the branch named by N, that is, the branch containing the
person-stage to which N is assigned, contains a person-stage that occurs at t and
has property F. If there is no such branch-as there would not be if the person-
stage to which N is assigned is a member of two branches-the sentence is false. A
sentence of the form
N is identical with M
is true if and only if the branch containing the person-stage to which N is
assigned is identical with the branch containing the person-stage to which M is
assigned. If there is no unique branch for N or M, the sentence is false. The
language for which this sketch is correct I shall call the branch language. In the
response above, the mentalist was supposing that English is the branch
language.9 What are the merits of this view? Is our concept of a person the
concept embodied in the branch language? I think not. The mentalist, in adopting
this solution, would be leaving the ordinary man far behind; for the ordinary
man is not willing to admit that there was not a single person, Jones, before the
operation, doing all the things Smith-Jones and Brown-Jones seem to remember
doing. Metaphysical considerations also seem to weigh against this view.
Consider the possible world just like the world described in section i1, except
that Jones dies the day before the operation. In that possible world there is a
single person Jones before the operation, even according to the branch language.
But before Jones's death there is, by hypothesis, no difference between that
possible world and the world described in section 1. Whatever the merits of this
last argument, it seems clear that we are reluctant to abandon the principle that
each person-stage identifies a person, so that if we assign a name to a person-
stage, we cannot but have named a person.

Nevertheless, I think this response suggests a more promising line. The
next three sections are devoted to its development.
4. Another Strategy
Can the mentalist use the impropriety of 'Jones' to save himself from self-
contradiction, without giving up the view that Jones was a single person before
the operation? It seems that he might if he can give sense to the view that 'Jones'
was proper before the operation, but improper after. He could then reject (2) and
(3):
(2) Smith-Jones is the same person as Jones.
(3) Brown-Jones is the same person as Jones.
on ground of the impropriety of 'Jones', but assert nevertheless (2') and (3'):
(2') Before the operation, Smith-Jones was Jones.
(3') Before the operation, Brown-Jones was Jones.
(2') and (3') answer the reasonable question, "Which of the persons who existed
before the operation were these two persons?" They were both the person, Jones.
(21) and (31) do not lead directly to the objectionable (4):
(4) Smith-Jones is the same person as Brown-Jones.
They lead only to
(4') Before the operation, Smith-Jones was the same person as Brown-
Jones.
But can the mentalist assert (2'), and (3') and (4'), without contradicting
himself, given his commitment to (1)?
(1) Smith-Jones is not the same person as Brown-Jones.
At first sight, the prospects for this seem slim. A variety of arguments can
be given to show that (1), (2'), and (3') lead, along with certain other things the
mentalist wants to say, as surely to self-contradiction as do (1), (2), and (3). The

essential reasoning behind any of these arguments will be something like this.
(2') and (3') say that Smith-Jones and Brown-Jones were the same person before
the operation. That means that, uttered before the operation, (4) would have
expressed a truth: 10
(4) Smith-Jones is the same person as Brown-Jones.
(Of course, only someone who knew what was going to happen would
have bothered to say it.) But from (4) it would follow that everything that was
true of Smith-Jones was true of Brown-Jones. But now suppose that after the
operation Smith-Jones is in room 102 and Brown-Jones in room 104. Then the
mentalist surely wants to say that (5) and (6) expressed truths before the
operation:
(5) After the operation, Smith-Jones will be in room 102.
(6) Not-(After the operation Brown-Jones will be in room 102).
But then something, namely, the open sentence 'After the operation, __will
be in room 102' was true of Smith-Jones but not true of Brown-Jones. So (4)
cannot have been true.
In the rest of this section, I discuss the moves the mentalist must make if
he is to evade this argument; these moves are in fact simply consequences of the
view that 'Jones' can be proper at one time, improper at another. In the next two
sections, I consider whether the mentalist can give an account of what a person
is---that is, an alternative to the branch language---that justifies making these
moves.
First let us consider what we should say in a case the mentalist might find
similar. The definite description, 'the senator from California' is occasionally
proper, although usually it is not. Suppose Murphy resigns as senator Saturday,
and Tunney is not sworn in as his successor until Tuesday. In the interim, 'the
senator from California' is proper, and denotes Cranston. Now, if on Monday we

wanted to say of the unique person who is then the senator from California, that
he will be in Washington Tuesday, we might try (7):
(7) The senator from California will be in Washington on Tuesday.
This we could distinguish from (8):
(8) Tuesday, the senator from California will be in Washington.
(7), we might say, requires a person to uniquely fit the description on
Monday, and be in Washington on Tuesday; (8) requires that a person both
uniquely fit the description and be in Washington on Tuesday. This would not be
a report of ordinary usage, but a pardonable regimentation thereof. (7) is true, (8)
false. Thus on Monday we could not infer from the truth, then, of (9):
(9) Tuesday, Cranston will be in Washington.
and of (10):
(10) Not-(Tuesday, the senator from California will be in Washington).
to the falsity of (11):
(11) Cranston is the senator from, California.
That is, we could not at any time infer from (9) and (10) the falsity of (12):
(12) Monday, Cranston was the senator from California.
The point is that temporal adverbs have two roles. In initial position, they
state that the sentence that follows is true at the time indicated With (12) we can
express at any time the proposition that ,we express Monday with (11). Within
the predicate, the temporal adverb indicates at what time the subject has the
property expressed by the predicate. When the time at which the sentence is true
is the time of utterance, no initial adverb is needed; when the time at which the
property predicated is to be possessed is the same as the time indicated by the
initial adverb, or lack of it, no adverb is called for in the predicate.
The fact that 'the senator from California' can be proper at one time,
improper at another, is of course just a special case of the more general fact that

'the senator from California' may denote different objects when used at different
times, or in the scope of different temporal adverbs in initial position. Similarly,
if we can show that it makes sense for a proper name to be proper at one time,
improper at another, that will be a special case of the more general fact that such
names may name different entities at different times, or when in the scope of
different temporal adverbs in initial position. If so, the argument:
(5) After the operation, Smith-Jones will be in room 102.
(6) Not-(After the operation, Brown-Jones will be in room 102).
Therefore,
(13) Not-(Smith-Jones is the same person as Brown-Jones).
is fallacious. The names 'Smith-Jones' and 'Brown-Jones' do not occur in (13) in
the scope of the temporal operator 'After the operation' as they do in (5) and (6).
So they cannot be assumed to name the same entities, and so (5) and (6) cannot
be seen as establishing that something true of one of the entities named in (13) is
not true of the other. All that can be inferred is
(14) After the operation, Smith-Jones is not Brown-Jones. But the
mentalist readily admits this; in fact, he insists upon it.
I believe that, by distinguishing between (2) and (3) and (2') and (3'), and
distinguishing between the two roles of temporal adverbs, the mentalist can say
everything he needs and wants to say about the case in section I without self-
contradiction. The latter maneuver blocks the arguments that derive a
contradiction from (1), (2'), and (3').
This defense, however, will not be very powerful until we have said more
about the relationship between temporal adverbs in initial position and names.
It's fairly clear why 'the senator from California' denotes differently in (8) and
(12). Intuitively, the temporal adverb completes the definite description, it tells
us when the property in question, being a senator from California, is to have

been possessed by the denotation. But it is not clear why a temporal adverb is
needed to "complete" the name 'Jones', nor how exactly this works.
In the next two sections I sketch two alternative accounts of our language,
each of which provides an explanation of the way in which temporal adverbs
and names function, and each of which assigns, to the sentences in question,
truth conditions that do not lead the mentalist to inconsistency. The first account
I reject; the second, I argue, is essentially correct.
5. The Person-stage Language
In discussing the notion of "strict identity" J. J. C. Smart once remarked:
When ... I say the successful general is the same person as the small boy who
stole the apples I mean only that the successful general I see before me is a time
slice of the same four-dimensional object of which the small boy stealing apples
is an earlier time-slice (Smart, 1959, p. 37) 11
The intuition behind the branch language was that persons are enduring
objects in some way composed of person-stages; although we may always
identify one or more person-stages in ostensively identifying a person, the words
'this person' denote not the person-stage occurring at the time, but the larger
whole of which he is in some sense a part. But Smart's remarks suggest a
radically different theory: we really refer, each time we use a personal name, to a
particular person-stage. Persons are just person-stages and not the "four-
dimensional" objects these compose. When I say, "the person you danced with
last night is the person sitting on the sofa," the 'is' does not express identity, but
simply the relation R. The sentence says that this relation obtains between two
distinct persons, the-girl-you-danced-with-last-night and the-girl-sitting-on-the-
sofa. If we use 'is' to express identity, the girl you danced with last night is not
the girl sitting on the sofa. But in such contexts., we would use 'is' not in this
way, but just to express R. In that sense, the one girl is the other. We might object
to this theory by pointing out that we say, for instance, that the girl you danced
with last night is now on the sofa. How could she be doing anything now, if she

was no more than a person-stage who barely survived the night with you? Here
the person-stage theorist can respond that 'the girl I danced with last night is
sitting on the sofa' can be understood as an abbreviated version of 'the girl I
danced with last night is someone sitting on the sofa', where 'is' again just
expresses the relation R. Thus the sentence says that a certain dancing person-
stage has the relation R to a certain sitting person-stage. We might object further
that a single name, 'Hilda', names both girls, and names are presumed to stand
for single objects. But the person-stage theorist denies the presumption. 'Hilda' is
systematically ambiguous; it names different persons at different times, so it is
ambiguous, but the persons it names share the name by virtue of having the
relation R to a certain person(-stage), say, Hilda-being-baptized, and so the name
is systematically and coherently used, in such a way that we are easily misled
into supposing that it names a single entity.
Thus we can sketch the person-stage language. As with the branch
language, each name is assigned to a person-stage. Now, however, instead of
supposing that the name then names the branch of which that person-stage is a
member, we suppose that it ambiguously names all the person-stages that have
R to the assigned stage. But the ambiguity is systematic. At any given time of
utterance, or within the scope of any temporal adverb, the name will name only
those person-stages which occur at the time of utterance, or at the time indicated
by the temporal adverb, and have R to the assigned stage. If and only if there is
exactly one such person-stage at a given time, the name is proper at that time.
A sentence of the form
N has F at t
uttered at time t', is true if and only if the person-stage named by N at t' has R to
some person-stage that occurs at t and has F. A sentence of the form
N is identical with M
is true at t' if and only if the person-stage named by N at t' is identical with the
person-stage named by M at P. A sentence with an initial temporal adverb,

At t', N has F at t
or
At t', N is identical with M
is true if and only if the sentence following the adverb is true when uttered at the
time indicated by the adverb.
Thus consider (15) and (16):
(15) Jones will be in room 102 after the operation.
(16) After the operation, Jones will be in room 102.
The first is true, before the operation, if and only if (15TC) is true:
(15TC) The person(-stage) named by 'Jones' before the operation has R to
some person(-stage) which occurs after the operation and is in
room 102.
(We are assuming that 'before the operation' and 'after the operation' pick out
definite times.) The second is true if and only if (16TC) is true:
(16TC) The person(-stage) named by 'Jones' after the operation is in room
102.
Given our example, (15) is true, (16) false; for "the person(-stage) named
by 'Jones' after the operation" is improper---that is, 'Jones' is improper after the
operation---there being two person-stages at that time, s-j and b-j, which have R
to Jones-being-baptized.
If we speak the person-stage language, the mentalist is in good shape. The
three sentences which the mentalist claims to express truths before the operation
but which seemed to lead him into contradiction, are clearly consistent:
(4) Smith-Jones is the same person as Brown-Jones.
(5) After the operation, Smith-Jones will be in room 102.
(6) Not- (after the operation, Brown-Jones will be in room 102).

(4) is true if and only if the person-stage named before the operation by 'Smith-
Jones' is the person-stage named before the operation by 'Brown-Jones'; all that
can be inferred from (5) and (6) is that the person-stage named after the
operation by 'Smith-Jones' is not the person-stage named after the operation by
'Brown-Jones'.
On this theory, it seems that (17)-(20) are all true before the operation:
(17) Jones will be in room 102 after the operation.
(18) Jones will be in room 104 after the operation.
(19) Jones will not be in room 102 after the operation.
(20) Jones will not be in room 104 after the operation.
This means only that (19) and (20) must be carefully distinguished from
the negations, of (17) 'and (18), which are false. (19) is true if and only if the
person-stage named by 'Jones' has R to some person-stage that occurs after the
operation and is not in room 102. This complication arises from the complicated
nature of the facts, given the example in section 1 and so is hardly an objection to
the person-stage language.
Similarly, the person-stage language must distinguish (21):
(21) Jones will be in room 102 after the operation and Jones will be in
room 104 after the operation.
from (22):
(22) Jones will be in room 102 and room 104 after the operation.
(21) requires that the person-stage named before the operation by 'Jones' have R
to some person-stage occurring after the operation who is in room 102 and also
have R to some person-stage occurring after the operation who is in room 104.
But (22) requires that Jones have R to a single person-stage occurring after the
operation who is in both room 102 and room 104. There is no such person-stage;
so (22) is false.

The person-stage language allows the mentalist to say just what he wants
about the example of section 1. Nevertheless, I think it would be a serious
mistake to suppose that English is the person-stage language---that our notion of
a person is just that of a person-stage.
The person-stage language and the branch language represent two very
different ways of looking at the function of sortal terms like `person'. According
to the branch language, when we say, "This person will have F at t,'' the word
'person' is a part of the referential apparatus of the sentence. Together with the
demonstrative, it identifies a certain enduring object, which has property F at
time t. Such analyses of sortals I call subject analyses.
The person-stage language suggests what I shall call an adverb analysis of
the function of a sortal. In the sentence, "This person will have F at t,'' we are to
think of the word 'this' as identifying the subject of the sentence---a person-stage-
--and the remainder of the sentence as telling how, in what manner, that person-
stage will have F at t. To be, as we might put it, personally F at t is not to be a
person(-stage) that has F at t, but to have R, the relation of personal identity (now
opposed to "strict identity," rather than a restriction of it to the domain of
persons), to some person(-stage) that has F at t. Being personally F at t is like
being married to a janitor; it's not being the janitor, but having a certain intimate
relation to someone who is.
The adverb analysis of sortals is radically mistaken. The apparatus that
must come with it-that 'is the same as' does not mean is the same as, that the little
boy stealing apples is strictly speaking not identical with the general before me---
seems to be, however consistently it may work out in the end, the progeny of
confusion. Usually the confusion takes this line: we think that the general before
us is big, but the little boy was small; if they were identical---strictly identical---
every thing true of the one would be true of the other. So they are not strictly
identical. Nevertheless, we say that the general is just the same person as the
little boy; so 'is the same person as' must not mean strict identity. This is all

confusion. There is nothing true of the general that is not true of the little boy.
They were both small, and neither was a general at that time. The general and
the little boy both had the property of being small, neither has it now. If we pick
a temporal perspective and stick to it, not ignoring tenses, there is no difficulty. If
we choose a timeless perspective, we must build dates into the properties we
ascribe. We shall find that both the general and the small boy have the property
of (say) being small in 1920. Only if we ignore both tenses and dates do we get
into trouble, and that is mere carelessness.12
It seems, therefore, a mistake for the mentalist to take refuge in the view
that we speak the person-stage language.
6. The Lifetime Language
The mentalist cannot take refuge in the theory that English is the branch
language, for it allows what cannot be: that before the operation, in talking to
Jones, we were not, in a perfectly clear sense, talking to a single person. It
violates our linguistic intuitions-what we want to say about our example.
He cannot take refuge in the person-stage language, for it denies what
clearly is true: that when I say of someone that he will do such and such, I mean
that he will do it. The events in my future are events that will happen to me, and
not merely events that will happen to someone else of the same name. The theory
that English is the person-stage language violates our semantic intuitions; it gives
an unduly complicated account of our language.
Is there any middle ground? Well, what entity is there that meets these
two conditions: (i) there is, in a perfectly clear sense, just one of these entities
identified by an ostension to Jones before the operation; (ii) everything that is in
Jones's future (that is, everything that will happen to Smith-Jones or Brown-
Jones) happens to a person-stage belonging to this entity? Clearly, one entity that
meets these requirements in the "Y-shaped" structure composed of the branches
of both Smith-Jones and Brown-Jones. I shall call any set of person-stages that
meets the following condition a lifetime: there is some member in the set such

that all and only members of the set have R to that person-stage. The Y-shaped
structure, although not a branch is a lifetime; for all person-stages in it have R to
j, the pre-operative stage of Jones. But the two branches that compose the Y-
shaped structure are also each lifetimes.
At first sight the suggestion that persons are lifetimes seems quite
unpromising, leaving the mentalist worse off than the hypothesis that persons
are branches. For the pre-operative stages of Jones being to three lifetimes: the Y-
shaped structure and each of its branches. If it was implausible to suppose that
Jones was two persons all along, surely it is more implausible by at least a half to
suppose that he was three.
But notice that each person-stage does identify a unique lifetime -the
lifetime containing all person-stages that have R to it. Thus the principle that
when we have identified a person-stage we have identified a person, violation of
which made the branch language seem implausible, is not violated by the
lifetime language.
Let us say that a person-stage determines the lifetime it identifies in the
way just described-the lifetime containing all person-stages With R to it. This is
not the only way that a person-stage may identify a lifetime. There is another,
and the different ways may, in bizarre circumstances, lead to different results.
Notice that, although each person-stage determines one and only one lifetime, it
may be a member of several. j, the pre-operative stage of Jones, determines the Y-
shaped lifetime. But in addition, it is a member of both branches of the Y, and
each of the branches is also a lifetime. Neither of these is determined by j, but they
both contain j.
Now, at any given time t, only a certain number of lifetimes will be
determinable, in the sense that they are determined by some person-stage
occurring at that time. Before the operation, neither of the branches are
determinable in this, sense: there is no person-stage occurring which has R to all

and only their members. The Y-shaped lifetime is similarly not determinable after
the operation.
Consider the person-stage b-j---a post-operative stage of Brown-Jones. b-j
determines a lifetime, one of the branches of the Y-shaped lifetime, the b-j branch.
Before the operation, the b-j branch is not determinable. Nevertheless, there is a
lifetime determinable before the operation that contains b-j. There is in fact, one
and only one such lifetime-the Y-shaped lifetime. Thus b-j can be used to identify
the Y-shaped lifetime: it is the unique lifetime, determinable before the operation,
that contains b-j
For any time t and any person-stage s, we can speak of the lifetime identified
by s at t. This description will denote the unique lifetime determinable at t which
contains s, if there is such; otherwise it is improper. In normal circumstances, if
there is a lifetime identified by s at t, it will just be the lifetime determined by s. But
in bizarre circumstances, such as those at issue in this paper, this identity will not
hold. The lifetime identified by b-j before the operation is not the lifetime
determined by b-j, but the lifetime determined by j. In normal circumstances, if s
is contained in any lifetime determinable at t, the lifetime determined by s will be
determinable at t. But, in bizarre cases, this will not be so: the lifetime
determined by i is not determinable after the operation, but j is contained in a
lifetime determinable at that time-as a matter of fact, in two; for both the s-j
branch and the b-j branch are lifetimes.
Given these notions, we can sketch an account of a final and I think
satisfactory refuge for the mentalist, the theory that English is a lifetime language.
The lifetime language embodies a subject analysis of sortals. A person has a
property F at t if and only if his lifetime contains a person-stage that occurs at t
and has F. But the lifetime language also retains a systematic ambiguity of
personal names reminiscent of the person-stage language. Indeed, it assigns
exactly the same truth conditions to the relevant sentences as does the person-

stage language. The lifetime language justifies the line of defense drawn in
section 4.
Again we assume that names are directly assigned to person-stages.
Where u is the person-stage to which N is assigned, the lifetime determined by u
is the primary referent of N. But N will also have a number of secondary referents,
which probably will but may not be identical with its primary referent. The
secondary referent of N at time t is the lifetime identified by u at t. If u does not
identify a lifetime at t---if there is no unique person-stage at t with R to u---then
N has no secondary referent at t, and N is improper at t. Thus, in the ordinary
case, the secondary referents of N and its primary referent will be one. But in
unusual cases, they will not.
A sentence of the form
N has F at t
uttered at time t', is true if and only if the secondary referent of N at t' contains a
person-stage that occurs at t and has F. A sentence of the form
N is identical with M
is true at t' if and only if the secondary referent of N at t' is identical with the
secondary referent of M at t'. A sentence with an initial temporal adverb,
At t', N has F at t.
or
At t' N is identical with M.
is true if and only if the sentence following the temporal adverb is true when
uttered at the time indicated by the temporal adverb.
Thus consider:
(15) Jones will be in room 102 after the operation.
(16) After the operation, Jones will be in room 102.
(15) is true before the operation if and only if (15TC') is true:

(I5TC') The secondary referent of 'Jones' before the operation contains a
person-stage that occurs after the operation and is in room 102.
(16) is true if and only if (16TC') is true:
(16TC') The secondary referent of "Jones" after the operation is in room
102.
(15) is true, (16) false; for 'Jones' has no secondary referent after the operation.
The sentences (4), (5), and (6), to the truth of which before the operation
the mentalist is committed:
(4) Smith-Jones is the same person as Brown-Jones.
(5) After the operation, Smith-Jones will be in room 102.
(6) Not-(After the operation, Brown-Jones will be in room 102).
are consistent. (4) is true if and only if the secondary referents of 'Brown-Jones'
and 'Smith-Jones' before the operation are one; all that can be inferred from (5)
and (6) is that the secondary referents of 'Brown-Jones' and 'Smith-Jones' after the
operation are distinct.
As before, (17)-(20):
(17) Jones will be in room 102 after the operation.
(18) Jones will be in room 104 after the operation
(19) Jones will not be in room 102 after the operation.
(20) Jones will not be in room 104 after the operation.
all come out true before the operation. Again we must distinguish the negations
of (17) and (18), which are false, from (19) and (20). (19) is true if and only if the
secondary referent of 'Jones' before the operation contains a person-stage that
occurs after the operation and is not in room 102.
As before, we must distinguish (21) from (22):

(21) Jones will be in room 102 after the operation and Jones will be in
room 104 after the operation.
(22) Jones will be in room 102 and room 104 after the operation.
(21) requires that the secondary referent of 'Jones' before the operation contain a
person-stage occurring after the operation which is in room 102 and contain a
person-stage occurring after the operation which is in room 104. (21) is true. But
(22) requires that the secondary referent of 'Jones' before the operation contain a
single person-stage occurring after the operation which is both in room 102 and
in room 104. (22) is false.
Now what is the answer to the fair question, "How many persons were
there in Jones's room (room 100) before the operation?"?
On the one hand , "one" seems to be the correct answer; for there was only
a single person, Jones, in the room. On the other hand, Smith-Jones and Brown-
Jones were both there; so "two" seems like the correct answer. But, after all, three
lifetimes (the Smith-Jones branch, the Brown-Jones branch, and the Y-shaped
structure) contain the person-stage in room 100; so the answer would appear to
be "three."
All three answers are correct-but they are answers, to different, and
distinguishable, questions. Consider open sentences of the form
x has F at t
A person satisfies such an open sentence at a time. z satisfies the given open
sentence at time t', if and only if z is identifiable at t' and contains a person-stage
that occurs at t and has F. The open sentence
x is in room 100, before the operation
is satisfied by exactly one person before the operation; so the answer to the
question, "How many persons are in room 100,?" asked before the operation (and
to the question, "Before the operation, how many persons were in room 100?"
asked at any time), is "one." After the operation, two distinct persons satisfy the

open sentence; so the answer to the question asked at that time (and to the
question, "After the operation, how many persons were in room 100, before the
operation?" asked at any time) is "two." There is no one time at which the correct
answer to the question, "How many persons were in room 100 before the
operation?" is "three." But the lifetime language will have to allow us to make
assertions such as
At some time, Brown-Jones was in room 100 before the operation.
which will be true just in case there is some time t such that
At t, Brown-Jones was in room 100 before the operation.
is true. The open sentence
At some time, x was in room 100 before the operation.
will be satisfied by person z if and only if z is identifiable at some time and
contains a person-stage occurring before the operation in room 100. This open
sentence is satisfied by three persons, and so the answer to "At any time, how
many persons were in room 100 before the operation?" is "three."
Can these three persons be identified within the lifetime language? Not
simply by the names 'Smith-Jones', 'Brown-Jones', and 'Jones'; for these identify
persons only in conjunction with temporal adverbs or times of utterance. Not by
definite descriptions, of the form
the x such that x has F at t
for these too may denote different persons at different times or in the scope of
different temporal adverbs. We can identify them, however, by use of definite
descriptions built up from more complicated open sentences. A definite
description of the form
the x such that at t1, x has F at t
denotes the person, if any, who is determinable at t, and contains a person-stage
that occurs at t and has F. And it denotes this person at any time. Thus, the three
characters in our story may be identified as:

the person x, such that, before the operation, x was in room 100
before the operation
the person x, such that, after the operation, x was in room 102 after
the operation
the person x such that, after the operation, x was in room 104 after
the operation
These three persons, are distinct, and never were identical-and nothing I have
said denies that, nor do (2') and (3').
What of fusions? In a convincing case of person fusion, in which a single
person-stage has R to two simultaneous but distinct antecedent person-stages, I
would argue that we should say the survivor was both of his precursors and had
done everything each of them had done. The lifetime language gives this result.
In cases of combined fusion and fission that I have considered, the lifetime
language seems to remain adequate.
Thus the suggestion that persons are lifetimes (or at any rate entities
correlated one-to-one with lifetimes) proves satisfactory. In any normal case, the
lifetimes are just branches. This explains our propensity for making inferences
valid in the branch language but not quite valid in the lifetime language, as when
we infer that, tomorrow Smith will be in Dubuque from the fact that Smith will
be in Dubuque tomorrow. Further, the lifetime language., like the person-stage
language but unlike the branch language, allows us to assign names to persons
with confidence, without fear that future events will present us with the choice of
contradicting ourselves or deeming many statements that seemed to be true false
because of unforeseen improprieties. Whenever we isolate a person-stage we
have isolated a person, namely, the person (lifetime) determined by that person-
stage. Finally, the lifetime language, like the branch language but unlike the
person-stage language, allows us to mean by our words what we think we mean,
to wit, identity by 'is the same as' and so forth. It embodies a subject analysis of
sortals. The lifetime language, then---or, more precisely, the theory that English is

a lifetime language---satisfies both our linguistic and our semantic intuitions.
Moreover, it has a certain naturalness. Who is Jones? The person who did all the
things in Jones's past and will do all the things in his future. Jones's future
includes both Brown-Jones's and Smith-Jones's, for it is true of Jones that he will
do all the things they do. This is what the mentalist wants to say, and the lifetime
language allows him to say it.
As was pointed out, the lifetime language and the person-stage language
do not differ in the truth conditions assigned to sentences, but do differ in the
assignment of entities to the parts of the sentence. In both we have a rather
elaborate system of identification. At each time, an entity is identified by a
person-stage u or a name N assigned to it. In the person-stage language the entity
is the unique person-stage occurring at that time with R to u. In the lifetime
language, the entity is the unique lifetime determinable at that time and
containing u. The difference is that, in the lifetime language, the entities
identified by u at any two times when it identifies anything at all will very
probably be the same; in the person-stage language they will certainly be
different. To the extent that there is no branching, the lifetime language is more
economical than the person-stage language, in having fewer entities in its
domain of discourse.
7. Conclusion
In speaking of a case of a dividing self, Jonathan Bennett has remarked, "the
fission of a mind, if it could happen, would involve the concept of identity in the
same way (whatever that is) as the fission of an ameba (Bennett, 1967, p. 112)."
This remark seems to me to conceal a mistake. It may be that, if selves
divided as often as cells or amoebas divide, we would develop a concept for
dealing with the phenomena our concept of a person now deals with that
resembles, in matters of individuation and identification, the concept of a cell or
an ameba. In a language embodying such a concept, it would apparently be
correct to say that Jones died, and two new "persons" were born, at the time of

the operation (See Carnap (1958) Ch. G). 13Whether we would develop such a
concept is, I suppose, a matter for speculative linguistics. It might be rational for
us to do so in those circumstances. That is a difficult philosophical question,
difficult in part because of the importance of memory in questions about persons.
I have not dealt with either of these questions directly, although what I have said
may have some relevance to them. I have dealt with the question whether the
mentalist account of the concept we have can survive an objection based on what
appears to be logically possible. I claim it can. Whether the concept itself could
survive, would survive, or should survive, if that logical possibility became
commonplace in actuality, is another question.14
Our "choice" of a language reflects certain pervasive empirical facts, and
the transitivity of R is one such. A tribe that spoke the person-stage language
might rationally, for the sake of economy, decide to adopt the lifetime language,
if empirical facts were such (as they are) that in doing so they would, for all
practical purposes, be speaking the branch language. But if dividing (or fusing)
selves were commonplace, not only in the minds of the philosopher members of
the tribe but in reality, there would be little to recommend the lifetime language
over the person-stage language. The comparative economy of the lifetime
language is an empirical matter. How often selves would have to divide, before
its retention became more trouble than it was worth, is a matter on which I shall
not speculate.
Given the nature of our world, the lifetime language shares the
advantages of both the branch language and the person-stage language. Insofar
as R remains transitive, it gives us the same economy as the branch language. But
should there be counterinstances to the transitivity of R, the branch language
would let us down. If we spoke it, we would have to check deep into the past
and future to assign a name with confidence. We would have, as it were, no spot
check for identity. We could not assume that we could tell whether A and B were
identical merely by isolating them (him) at a particular time and conducting an

examination. Here the lifetime language shares the advantages of the person-
stage language. In the person-stage language one cannot go wrong in assigning
names; if you have isolated a person-stage, you have isolated a person. The same
is true of the lifetime language, for every person-stage determines a lifetime.
The evidence that the lifetime language is a correct approximation of that
portion of English in which the mentalist describes the case with which we began
is: (i) R (some mentalist analysis of the relation between person-stages that are
stages of a single person) seems to give the correct analysis of personal identity;
(ii) the case described in section I is conceivable; (iii) speakers of English without
an overdeveloped fear of self-contradiction are perfectly willing to describe this
case by (1), (2') and (3') and unwilling to swallow the story that there were two
persons all along; (iv) a subject analysis of sortals is more natural and economical
than an adverb analysis. At one point Arthur Prior was willing to abandon the
transitivity of identity itself in order to preserve the point of view of (i) through
(iv).15 I believe the 1solution I have outlined is much less drastic. If we want the
economy of the branch language and the nominal security of the person-stage
language we should speak the lifetime language. Is it so surprising that we do?
1 I am heavily indebted to many persons for comments on earlier versions of this paper,
especially David Lewis, John Vickers, David Kaplan, John Bennett, Richard Rodewald, Sydney
Shoemaker, and Jaegwon Kim.
2 I first heard this case described by Sydney Shoemaker in a seminar. Shoemaker
discusses a body-transplant case in his book, Self-knowledge and Self-identity (1963). Body
transplants may some day be medically as well as logically possible. See the remarks of Christian
Barnard quoted in Newsweek, December 23, 1968, p. 46.

3 Locke discusses personal identity in chapter XVII of book ii of his Essay concerning Human
Understanding (1694). He says at one point, "…as far as this consciousness can be extended
backwards to any past action or thought, so far reaches the identity of that person…( p. 39)."
4 In "The Soul," (1962), Quinton argues that persons are "'fundamentally" souls, a soul being a
series of mental states, Roughly, two soul-phases belong to the same soul if they are "connected
by a continuous character and memory path" (p. 59). (Quinton gives a more precise account of
this relation.)
5 Grice gives his analysis in "Personal Identity," (1941), in terms of the notion of a "total temporary
state" which is composed of "'all the experiences any one person is having at a given time" (p. 86).
Grice's analysis (pp. 87-88) is equivalent to the following. Let Sxy be the relation between total
temporary states:
x contains an experience such that, given certain conditions, y would contain a
memory of it.
Then two total temporary states belong to the same person if and only if - they are both members
of a set closed under the relation Sxy v Syx. This analysis gives the result that both Smith-Jones
and Brown-Jones would have had all of Jones's experiences. Unfortunately, it also gives the result
that Smith-JONES and Brown-Jones are one. It could be amended to avoid this unfortunate result
by requiring (as Grice's preliminary analyses seem to) that no two members of the set in question
occur at the same time.
6 One such is that developed by Sydney Shoemaker in (1970a). Shoemaker would deny that either
Brown-Jones or Smith-Jones was Jones (see p. 278n). At least part of Shoemaker's motivation for
denying this is his belief that it involves "'modifying the usual account of the logical features of
identity" (279 n). That I deny; this paper is my argument for that denial. Shoemaker's analysis
builds into every claim of the form "This is the person who did A" the negative-existential claim
that no one else in the entire universe has the criterial relation to the doer of A, and this seems
implausible to me. In (1970b) Shoemaker argues that this objection [which was made by David
Wiggins in (1967) does not apply if the criterial relation requires a causal chain, for 'it can be
established without a survey of the entire universe whether some other person's memories are
connected . . . by the same sort of causal chain"' (543). But it's not the difficulty of the survey that
is the point. Rather, it is that the question whether Smith-Jones brushed his teeth before the
operation shouldn't depend on whether Brown-Jones lives or dies.
7 For Quinton (1962) the relation would be the relation of indirect continuity, with the
understanding that each soul-phase is indirectly continuous with itself. For Grice (1941) as

amended in note 18, it is the relation of co-membership in an appropriate set closed under Sky v
Syx. In both cases, the analysis is stated in a format not precisely like the one I have suggested;
Quinton talks about soul-phases, and Grice about total temporary states, rather than person-
stages. I shall not attempt to discuss the comparative merits of these approaches; the points I
make in this paper could be made in the terminology of either Grice or Quinton.
8 Richard M. Gale (1969) argues persuasively that (as I would put it) the temporal-unity relation
for human bodies is not logically transitive either, and so Williams' objection is not a good
argument in favor of a bodily continuity analysis of personal identity. I tend to agree, but this
does not solve the problem of the dividing self; it merely enlarges the number of philosophers
who should be bothered by it.
9 I have not, of course, given anything like a complete account of the branch language, nor do I of
the person-stage language or the lifetime language, which are described later, but I believe I have
said enough to make the solution advanced in this paper to the problem at hand clear.
Schematic letters such as 'N', 'F', and 't' are used in displayed sentences as metalinguistic variables
for the appropriate classes of object-language expressions; such displayed sentences should be
regarded as in quasi-quotation. Elsewhere these same letters are sometimes used as object-
language variables for the appropriate entities; thus I say "The property F" and "the time t" rather
than "the property expressed by F" and "the time designated by t.''
10 I take it that sentences express propositions at times, and some sentences express different
propositions at different times. When I say a sentence is true at a time or true when uttered at a
time, I mean that the proposition the sentence would express at that time is true.
11 To attribute to Smart exactly the theory embodied in the person-stage language as I develop it
would be unfair; he remarks that he is permitting himself to "speak loosely."
12 For discussions of the relation between identity and the unity relation (as I have called it) or
gen-identity (as Carnap calls a similar notion), see Gottlob Frege (1884/1960), sec. 62ff.; Rudolf
Carnap, (1958), ch. G; and W. V. Quine, (1953), pp. 65ff. For a criticism of the Frege view, see
Peter Geach (1967). I discuss Geach's views in Essay 1.
13 See Carnap, op. cit., chapter H.
14 Derek Parfit, (1971) argues that a single instance of self-division is more than our ordinary
concept of a person can handle. This I have in effect argued against. Parfit argues interestingly
that consideration of such cases, and other even more bizarre possibilities, leads us to see that our
concept of a person is unimportant and should perhaps be replaced with other "'ways of

thinking." I would rather say our concept of a person is important in large part because our world
does not realize such possibilities.
15 Prior discussed this problem in two places. In "Opposite Number" (1957) he advocated
abandoning the principle that 'P' and 'n-moments ago it was true that n-moments from then, P'
are logically equivalent. In "Time, Existence, and Identity," (1956) he pointed out the inadequacy
of his earlier suggestion, and suggested abandoning the transitivity of identity. I hope the reader
will agree without extended argument that a less drastic solution than this is desirable. My
solution has some resemblance to one suggested but summarily rejected by Prior in "Opposite
Number," p. 199.
References
Carnap, Rudolf, 1967. The Logical Structure of the World. Berkeley: University of California Press.
Carnap, Rudolf, 1958. Introduction to Symbolic Logic and Its Applications. New York: Dover.
Flew, Anthony, 1951. Locke and the Problem of Personal Identity. Philosophy (January 1951): 53--
68.
Frege, Gottlob, 1884/1960. Grundlagen der Arithmetik. Breslau, 1884; trans. by J.L. Austin. The
Foundations of Arithmetic, 2nd ed. New York: Northwestern University Press. 1960.
Gale, Richard, 1969. A Note on Personal Identity and Bodily Continuity. Analysis XXIX (1969):
193--195.
Geach, Peter, 1967. Identity. Review of Metaphysics vol. 21, no.1 (1967): 3--12.
Grice, H.P., 1941. Personal Identity. Mind 50 (1941). Reprinted in Perry, 1975: 73--95.
Locke, John, 1694. Of Identity and Diversity. Book II, Chapter 27 of Locke, 1694E. Reprinted in
Perry, 1975: 33--52.
Parfit, Derek, 1971. Personal Identity. Philosophical Review LXXX (1971). Reprinted in Perry, 1975:
199--223.
Parfit, Derek, 1973. Later Selves and Moral Principles. In Alan Montefiore (ed.), Philosophy and
Personal Relations. London: Routledge and Kegan Paul. 1973.

Prior, Arthur, 1966. Time, Existence and Identity. Proceedings of the Aristotelian Society (1965/6):
183--192.
Prior, Arthur, 1957. Opposite Number. Review of Metaphysics (1957): 196--201.
Quine, W.V., 1953. From a Logical Point of View. Cambridge, Mass.: Harvard.
Quinton, A., 1962. The Soul. Journal of Philosophy 59 (1962). Reprinted in Perry, 1975: 53--72.
Shoemaker, Sydney, 1970a. Persons and Their Pasts. American Philosophical Quarterly (1970): 269--
285.
Wiggins, David, 1967. Identity and Spatio-Temporal Continuity. Oxford: Blackwells.
Williams, Bernard, 1957. Personal Identity and Individuation. Proceedings of the Aristotelian Society
(1956-57).