Four-Dimensionalism: An Ontology of Persistence and Time

Theodore Sider

https://doi.org/10.1093/019924443X.001.0001
Published: 2001 Online ISBN: 9780191598425 Print ISBN: 9780199244430

CHAPTER

chapter 4 In Favor of Four‐Dimensionalism Part 1 

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

Theodore Sider

https://doi.org/10.1093/019924443X.003.0004 Pages 74–139

Published: October 2001

Abstract
Some traditional arguments for four‐dimensionalism are weak: denying four‐dimensionalism does not
prohibit the application of modern logic to natural language, does not imply the A‐theory of time and is
consistent with special relativity. Others have some force but are inconclusive: the argument from
analogies between time and space and Lewis's argument from temporary intrinsics. Some new
arguments fare better. (1) Only four‐dimensionalists can admit certain (admittedly exotic) possibilities
involving timeless objects and time travel into one's own past. (2) Either substantivalism or
relationalism about space‐time is true. Given substantivalism (and a sensible, exible theory of de re
modal predication), one might as well identify continuants with regions of space‐time, which have
temporal parts. Alternatively, one could identify continuants with instantaneous slices of space‐time
and employ temporal counterpart theory; either way, we have a four‐dimensionalist metaphysics of
continuants. On the other hand, relationalism about space‐time cannot be made to work without
temporal parts. So either way, we have an argument for four‐dimensionalism. (3) It can never be vague
how many objects exist; if temporal parts do not exist then a restrictive account of which lled regions
of space‐time contain objects must be given; but no such account can be given that is plausible and
non‐vague.

Keywords: composition, continuants, counterpart theory, endurance, Four‐dimensionalism, Minkowski

space‐time, parthood, perdurance, relationalism, special relativity, substantivalism, temporal parts, three‐
dimensionalism, time‐travel, vagueness
Subject: Metaphysics
Collection: Oxford Scholarship Online

. . . The theory of the manifold is the very paradigm of philosophic understanding. This is so with
respect to its content, since it grasps with a strong but delicate logic the most crucial and richest
facts.

Donald C. Williams, ‘The Myth of Passage’, 471–2

 It is easy to feel, with Williams, an intellectual joy in contemplating a theory so elegant and beautiful as
four‐dimensionalism, and it is tempting to accept the theory simply on this basis, utilizing arguments to
rationalize more than justify. Despite this temptation it is to arguments that I now turn. One will occupy the
whole of the next chapter: temporal parts are needed to solve puzzles in which distinct objects appear to
1
share the same parts and spatial location. In the present chapter I discuss the rest. The rst few are
familiar, while the arguments in Sections 7, 8, and 9 are, as far as I know, new. The arguments are not
uniformly strong. While some are compelling, others have only limited persuasive power; still others I atly
reject. I regard none as knock‐down; that would be too good to be true. In each case my goal will be to set out
the strengths and weaknesses of the argument as accurately as possible. The reader must then judge for him
or herself. I proceed roughly in order of increasing plausibility‐to‐me. Note that some of the arguments

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

presuppose the B‐theory of time, or at least the falsity of presentism, and thus rest on Chapter 2.

p. 75
1. Russell's Argument from Parsimony

Early defenders of temporal parts include C. D. Broad, Jonathan Edwards, David Hume, Hermann Lotze, J. M.
2
E. McTaggart, Bertrand Russell, and A. N. Whitehead. Their arguments were sometimes familiar from a
contemporary perspective, but sometimes not. Russell's defense of temporal parts, for example, is tied up
with the program of constructing the world of physical objects from sense‐data. Sense‐data have temporal
parts; physical objects, being sequences or sums of sense‐data, therefore themselves have temporal parts
(Russell 1914: 109 .).

I will leave a detailed historical survey of these gures to those more able. I will, however, mention one
other Russellian argument, since it remains tempting to this day but is, as I see it, unsuccessful. The
argument is that the postulation of anything more than temporal parts would be empirically unjusti ed. In
perception, all that is immediately given is a sequence of objects. Belief in a single enduring object
3
accounting for the sequence is, according to Russell, unwarranted metaphysical speculation. This comes in
two versions, depending on whether the objects of perception are sense‐data or physical things. In the rst
case, the metaphysical speculation would be twofold: that there exists anything other than sense‐data, and
that there is a single such thing. Set the rst aside; of the second Russell says:

. . . the physical object, as inferred from perception, is a group of events arranged about a centre.
There may be a substance in the centre, but there can be no reason to think so, since the group of
events will produce exactly the same percepts; therefore the substance at the centre, if there is one,
is irrelevant to science, and belongs to the realm of mere abstract possibility . . . (1927: 244).

Suppose I infer, from a sequence of perceptions, a sequence of objects o1, o2, . . . . Grant Russell for the sake
p. 76 of argument that it goes beyond the evidence to claim that the observations are of a single thing, that o1 =
o2 = o3 . . . . It would likewise go beyond the evidence to claim that o1≠ o2 . . . . At best, the conclusion can be
that we should be neutral, so far as observation and science are concerned, about whether objects endure or
perdure; the question must be resolved on philosophical grounds. (It is a separate question whether the
observed objects must be distinct because they di er qualitatively; see Sect. 6 below. Russell's argument is
not based on change, for he applies it even in the case of electrons (ibid. 246–7).)
 2. The Argument from Logic

In Word and Object (1960: 170–6), Quine points out that tensed sentences of English may be symbolized
using tenseless quanti ers and predicates if one utilizes quanti ers over times and relativizes temporary
predicates to times. Thus, ‘A mammal was once trampled by a dinosaur’ becomes ‘∃ x ∃ y ∃ t (t < n & D x & M
y & T xyt)’ (‘n’ is a name for the current time, now; ‘t < n’ means that time t is earlier than time n; ‘T xyt’
means that x trampled y at t). Elsewhere he claims that this is the only workable treatment of tense (1953:
442–3):

The only tenable attitude toward quanti ers and other notations of modern logic is to construe

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

them always, in all contexts, as timeless. This does not mean that the values of ‘x’ may not
themselves be thing‐events, four‐dimensional beings in space‐time; it means only that date is to be
treated on a par with location, colour, speci c gravity, etc.—hence not as a quali cation on ‘∃’, but
merely as one of sundry attributes of the thing‐events which are values of ‘x’.

The four‐dimensional view of space‐time is part and parcel of the use of modern formal logic, and
in particular the use of quanti cation theory, in application to temporal a airs.

. . . I do not see how, failing to appreciate the tenselessness of quanti cation over temporal entities,
one could reasonably take modern logic very seriously.

To partake of the wonders of modern logic one must accept the ‘four‐dimensional view of spacetime’.

What Quine seems to be calling the four‐dimensional view of spacetime actually has two components. There
is rst the B‐theory of time. (The B‐theory, recall, also has two components: eternalism—the thesis that
p. 77 past and future entities are real—and the thesis of the reducibility of tense.) Secondly, there is the thesis
that things perdure, are composed of temporal parts. Call the combination of these two views the manifold
theory.

To some extent Quine's argument is simply outdated, for there subsequently occurred the invention of tense
4
logic by Arthur Prior and others. Prior gives a rigorous treatment of tense that is very di erent from
Quine's, and which does not presuppose the manifold theory. But the real problem is that Quine's argument
at best supports the B‐theory. The second component of the manifold theory, namely perdurance, is by no
means required for the logical systems developed by Quine himself and others, for example Carnap (1967)
5
and J. H. Woodger (1937, 1952). Only the B‐theory is needed. Even if past dinosaurs endure, a B‐theorist can
still quantify over them and relativize their temporary predicates to times. Quine (1960: 172–4) later
extended the above treatment of tense by introducing the phrase ‘x at t’ (which denotes x's segment during
time interval t) and interpreting tensed locutions as concerning temporal segments. But the temporal parts
here could be ersatz, not genuine (see Ch. 3, Sect. 2). Assertions about ersatz temporal parts may be
translated as assertions about enduring entities as follows:

x‐at‐t ⇒ < x,t > (where x exists at t)

< x,t > is F ⇒ x is F at t

R(< x,t > , < x',t >) ⇒ R(x,x') at t

R(< x,t > , < x',t' >) ⇒ R(x,x') at t,t'

The nal two cases give the translations for synchronic and diachronic ascriptions of relations between
temporal parts, respectively. The phrase ‘R (x, x′) at t, t′’ is to be read ‘x, as it is at t, bears R to x′, as it is at
t′’. Similar remarks apply to other uses of temporal parts by these authors. Woodger, for example, is
primarily interested in axiomatizing biology to clarify the logical relations between di erent biological
 claims. No use is made of the fact that temporal parts are supposed to be literally parts; their only use is in
an account of the logical form of biological sentences. This use is preserved if further analysis of talk of
temporal parts is subsequently given, which treats the temporal parts as ersatz and which ultimately rests
on a foundation of endurance.

p. 78 Related arguments, if we can call them that, are based on the fact that temporal parts make possible a neat
statement of this theory or that. In any such case we must look to see whether the temporal parts are really
indispensable. Often they are not.

3. The A‐Theory of Time Is Incoherent

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

Another argument from around the same time was that endurance must be rejected because it is part of an
untenable philosophy of time. In chapter 7 of Philosophy and Scienti c Realism, J. J. C. Smart argues against
what he calls an ‘anthropocentric’ account of time, according to which ‘the notions of past, present, and
future apply objectively to the universe’ (1963: 132). This is the ‘A‐theory’ of time. Against this account,
Smart puts forward a view which is in essence what I am calling the manifold theory. A similar argument is
contained in D. C. Williams's classic paper ‘The Myth of Passage’.

Smart has three objections to the A‐theory. First, he claims but does not argue that the manifold theory ts
6
better with contemporary science. Presumably Smart is thinking about special relativity, and he may well
be correct that the B‐theory half of the manifold theory may be thus supported (see Sect. 4 of Ch. 2). But, as
will be argued in Section 4 of this chapter, it is less clear that special relativity supports temporal parts. Like
Quine, Smart (and Williams as well) seems to lump the two components of the manifold theory together and
assume that considerations in favor of one must invariably favor the other. It is vital to distinguish these
components. I suspect that temporal parts have gained undeserved credibility by tagging along in this way
with the B‐theory.

Matters are similar with Smart's second and third objections to the A‐theory, which are that (1) the A‐theory
is anthropocentric and (2) that it introduces ‘unnecessary mysti cation’ (1963: 140). Argument (1):
distinctions of past, present, and future are those that human thinkers make from particular perspectives,
and should not be exalted into objective metaphysical distinctions in the world. Argument (2): the mysteries
p. 79 resulting from the A‐account stem from its claim that time ows. How fast does time ow? Must there be
a hyper‐time in which the ow takes place? (Williams says in this connection that ‘Most of the e ect of the
prophets of passage . . . is to melt back into the primitive magma of confusion and plurality the best and
sharpest instruments which the mind has forged’ (1951: 472).) Moreover, there is McTaggart's famous
‘proof’ of the unreality of time, in which the incoherence of the A‐theory is allegedly demonstrated. Against
7
such claims A‐theorists have responded, but this debate is not my concern here, for it is evident that the
arguments at best support the B‐theory; they do not support the temporal parts component of the manifold
theory. (Of course, if one could argue that the combination of the B‐theory and endurance is untenable then
this gap in the arguments of Smart and Williams would be bridged; see Sects. 6, 7, and 8.)
 4. Four‐Dimensionalism and Special Relativity8

It is often claimed with little argument that the special theory of relativity requires perdurance. Russell
(1927: 286), for example, says ‘The old notion of substance had a certain appropriateness so long as we
could believe in one cosmic time and one cosmic space; but it does not t in so easily when we adopt the
9
four‐dimensional space‐time framework’. In fact, however, the support for temporal parts here is weak;
one can accept special relativity and endurance alike. (At any rate, provided one is not a presentist;
presentism itself is in tension with special relativity (Ch. 2, Sect. 4).)

There is no doubt that the language of Minkowski spacetime at least suggests four‐dimensionalism. The

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

p. 80 world is typically described as a world of events; point particles are construed as the sum total of events
along their worldlines; extended objects are described by worm‐like sections of spacetime, which have time‐
slices; and so on. Moreover, many descriptions of endurance are couched in terms hostile to relativity.
Friends of endurance say that an object is wholly present at every moment of time at which it exists, and
they speak of properties instantiated relative to times. This talk of times appears to con ict with the denial
in special relativity of a privileged separation of spacetime into space and time. But endurantists may just be
ignoring relativistic considerations. (We four‐dimensionalists commit the same sin when speaking of
temporal parts.) The question is whether the endurance theory can be stated in a relativistically acceptable
way. In this section I will show that this can indeed be done. (I ignore general relativity, since it appears to
10
present no new obstacles for three‐dimensionalism.)

Begin by assuming a four‐dimensional Minkowski spacetime, as described in Chapter 2, Section 2. Reality

thus contains a four‐dimensional manifold of points of spacetime. In addition to the spacetime, let us also
suppose that there exist continuants—fundamental particles, tables and chairs, people, and so on—which
11
inhabit the spacetime. This ‘inhabiting’ may be characterized by means of a binary relation of occupation
holding between continuants and points of the spacetime.

The three‐dimensionalist wants to say that these continuants endure, that they have no temporal parts. This
is easiest to characterize in the case of a fundamental particle. Such a particle would have two crucial
features. First, it would have no proper parts whatsoever (no parts, that is, with respect to any points of
spacetime—see below). Secondly, the occupation relation between it and points of spacetime would be one‐
many: it would occupy all the points that comprise its worldline.

This way of viewing things may be contrasted with that of the four ‐dimensionalist. The four‐dimensionalist
would agree that the occupation relation between the particle and points of spacetime is one‐many, but she
p. 81 would deny that the particle is mereologically simple. Rather, at each point along its worldline, the
particle has a spatiotemporal point‐part—an object that (1) is part of the particle, (2) is located at that point,
and (3) is not located at any other point. For the four‐dimensionalist, the occupation relation between
mereologically simple things and points of spacetime can never be one‐many.

We also want to know how to talk about the properties and relations of fundamental particles. Let us rst
discuss spatiotemporal relations. Here the endurantist should say that spatiotemporal relations hold
primarily between points of the spacetime. Fundamental particles then stand in spatiotemporal relations
derivatively by occupying points of the spacetime that stand in those relations. Let us rst get clear about
the spatiotemporal relations that hold between points in Minkowski spacetime. Just as observer‐
independent simultaneity is not a well‐de ned notion in Minkowski spacetime, neither are observer‐
independent spatial or temporal distance. However, relative to any frame of reference—that is,
unaccelerated path through spacetime—not only are simultaneity relations well de ned, but temporal and
spatial distance relations between points are also well de ned. Relative to a frame of reference, F, any two
points x and y are separated by some de nite temporal and spatial distance. These frame‐relative distances
 ow from facts about the intrinsic geometry of the spacetime (they may be derived from information about
the ‘interval’, which is an absolute (frame‐invariant) quantity de ned for any two points in spacetime).

We therefore proceed as follows. Since simultaneity is well de ned relative to any frame of reference, F,
relative to F we can divide the points of spacetime into equivalence classes of simultaneous points. These
equivalence classes may be thought of as times, so long as we remember that the division of points into
times is valid only relative to the chosen frame, F. So now, relative to F, consider any time, t, and any two
fundamental particles x and y. We may rst say that one of these particles exists at t i it occupies some point
12
of spacetime that is contained in t. And given that x and y both exist at t, the spatial distance between x and
y at t (relative to F) may be de ned as the spatial distance relative to F between the points of spacetime in t

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

that are occupied by x and y, respectively.

The defender of endurance must account for the qualities of fundamental particles as well as their
p. 82 spatiotemporal relations. When an endurantist is not thinking of relativity she typically construes
property instantiation as being in some way relative to time, to account for change (see Sect. 6 below). Given
13
relativity, instantiation must be indexed instead to points of spacetime. Thus, an electron will have unit
negative charge, spin 1/2, and so on, relative to the points of spacetime that it occupies. (The four‐
dimensionalist will regard point‐relative property instantiation as being de nable from property
instantiation simpliciter as follows: a particle x has property P at point p i the spatiotemporal point‐part of x
that occupies p has P simpliciter. The three‐dimensionalist will reject this de nition since it presupposes
spatiotemporal point‐parts.) We can then de ne a frame‐relative notion of instantiation at a time: relative to
any chosen reference frame F, and given any time t in the partition of spacetime for frame F, x has P at t i x
has P at the spacetime point in t that x occupies.

So far we have formulated a theory of enduring fundamental particles in which properties of and
spatiotemporal relations between these particles may be characterized. The relativistic physics of the small,
at any rate, is consistent with endurance.

But a relativistic theory of macro‐entities must be formulated as well, and here we encounter new
di culties. One problem is due to Yuri Balashov (1999). One and the same macroscopic object will have
di erent shapes relative to observers traveling at di erent velocities, even as observed from the same point
in spacetime (allow as an idealization that the observers coincide spatiotemporally at the point of
observation). What explains the fact that one and the same object has these di erent shapes for di erent
observers? Balashov puts the point as follows (p. 653):

The same object has di erent 3D shapes. There are strong reasons to believe that the di erence is
due neither to intrinsic change (shapes are observed at the same time) nor to the variation of a
merely spatial perspective (shapes are observed from the same place; furthermore, even if they
were not, it is hard to see how a variation of a merely spatial perspective could a ect the three‐
dimensional shape of an object) . . .

[The three‐dimensionalist] will have a hard time explaining how ‘separate and loose’ 3D shapes
p. 83 come together in a remarkable unity, by lending themselves to an arrangement in a compact
and smooth 4D volume. Where the four‐dimensionalist has a ready and natural explanation of this
fact: di erent 3D shapes are cross‐sections of a single 4D entity, the three‐dimensionalist must
regard it as a brute fact, indeed, as a complete mystery.

According to Balashov, only the four‐dimensionalist has a ready answer of why the object has di erent
shapes for di erent observers, and why those shapes ll up a compact and smooth 4D volume. The three‐
dimensionalist allegedly cannot provide a unifying explanation of the di erent shapes, but must rather
 postulate a brute fact that a plenitude of ‘perspective‐indexed’ 3D shapes all belong to a single enduring
thing, and t into a compact and smooth 4D volume.

But assuming the account of enduring fundamental particles given above, the three‐dimensionalist can
reply as follows. To have a shape at a time is just to have parts that are spatially related in a certain way at
that time. As we have seen, relative to a chosen reference frame we can account for spatial relations between
fundamental particles at times. Provided the three‐dimensionalist can make sense of the part‐whole relation
in a relativistic context, then, she can account for the shapes of macroscopic objects in various reference
14
frames. All of the perspective‐indexed shapes are the result of a single set of facts about the enduring
object, which include (1) the holding of the part‐whole relation, and (2) the holding of the occupation
15

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

relation between fundamental particles and points of spacetime. This also explains also why the shapes t
into the 4D volume that they do. The volume is generated as the sum of all the points occupied by the parts
of the object; the shapes are slices of this volume.

p. 84 This presupposes a relativistically acceptable mereology. Relativistic mereology for the four‐dimensionalist
is easy. Since the four‐dimensionalist's part‐whole relation is atemporal, the theory of that relation is
una ected by relativistic considerations. (Of course, which parts of a spacetime worm count as temporal
parts is a matter a ected by relativity, but that is di erent.) As we saw in Chapter 3, Section 2, the eternalist
three‐dimensionalist rejects atemporal mereology. The strategy discussed there of indexing the part‐whole
relation to times can no longer be pursued without modi cation in the relativistic case since there are no
such things as times. The natural course is to relativize the fundamental mereological locutions to points of
spacetime instead; frame‐relative facts about parts at times may then be derived.

On behalf of the three‐dimensionalist, let us take as primitive a three‐place relation of overlap at a point: one
continuant overlaps another at a point of spacetime. The idea may be made vivid by thinking of the histories
of the relata: the picture is that objects overlap one another at p when their histories both include p (though
this should not be thought of as de nitional since a three‐dimensionalist might not want to rule out, a
priori, the possibility of two mereologically disjoint objects sharing the same spacetime location). The
primitive of overlap should be assumed to obey the law that if objects overlap at a point then they both
occupy that point (though the converse should not be assumed if spatiotemporal co‐location is not to be
ruled out a priori.)

(Notice that the four‐dimensionalist would de ne overlap‐at‐a‐point in terms of two‐place mereological

locutions and occupation: x overlaps y at p i some z is part (simpliciter) of both x and y, and z occupies p. For
the three‐dimensionalist overlap‐at‐a‐point remains primitive.)

A frame‐relative notion of overlap at a time can now be introduced:

x and y overlap at time t (relative to some chosen frame, F) i there is some point in t at which x and
y overlap.

We can then use familiar methods from the Calculus of Individuals to de ne other temporally quali ed
mereological notions, for example parthood‐at‐t:

x is part of y at time t (relative to some chosen frame, F) i x and y both exist at t, and everything
that overlaps x at t also overlaps y at t.

p. 85 In sum, the three‐dimensionalist may appeal to the following sorts of facts holding within Minkowski
spacetime:

• facts about overlap‐at‐spacetime‐points for enduring objects;

• facts about the occupation relation between enduring objects and spacetime points;
 • facts about properties of enduring fundamental particles at spacetime points;

• facts about the intrinsic geometry of spacetime (for example facts about the interval).

The three‐dimensionalist could plausibly go on to claim that all of what we say in science and everyday life
supervenes on these facts. In this sense, then, endurance can be reconciled with special relativity.

There remains the issue of how to formulate the central three‐dimensionalist thesis itself, that objects are
wholly present whenever they exist, within this framework. As we saw in Chapter 3, Section 3 above, even
leaving aside relativity there are serious problems in arriving at an acceptable formulation of three‐

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

dimensionalism. We can give a relativistic analog of the de nition of being strongly wholly present
throughout a temporal interval:

De nition: Given a choice of frame, F, in the partition of spacetime relative to F, x is strongly wholly
present throughout region T i anything that is at any time in T part of x is, at every time in T,
part of x.

But what will be the central claim of three‐dimensionalism? We do not want to claim that, with respect to
every (or even any) reference frames, every object is strongly wholly present at every moment at which it
exists, because this would rule out mereologically inconstant objects. As we saw, the proposition that it is
possible for there to be an object strongly wholly present throughout an extended interval will be accepted by all
three‐dimensionalists; and most three‐dimensionalists will at least conjecture that every fundamental
particle is strongly wholly present throughout its career. This much we already knew from Chapter 3, Section 3.
Relativity now raises the additional question of which reference frames the claim and conjecture will be
made from. The natural course seems to be to relativize to all (inertial) reference frames. The claim would
p. 86 then be that it is possible for there to be an object that is, with respect to every reference frame, strongly
wholly present throughout some extended interval with respect to that reference frame. The conjecture
would be that, with respect to every reference frame, every fundamental particle is strongly wholly present
throughout its career. I see no obstacle to the coherence of these things. It therefore appears that relativistic
three‐dimensionalism is a coherent position.

One nal worry must be faced by three‐dimensionalists who combine their rejection of temporal parts with
the acceptance of ‘arbitrary spatial parts’. Might something count as a spatial part with respect to one
16
reference frame and a temporal part with respect to another? In fact this cannot happen. Consider, for
example, a fundamental particle e that is strongly wholly present throughout its career with respect to some
reference frame, and let p1 and p2 be two points along its worldline. The worry now involves a second
reference frame with respect to which p1 and p2 are simultaneous. Since e is strongly wholly present
throughout its career in the rst frame, we would appear to have a violation of the doctrine of arbitrary
spatial parts in the second frame, for there is an object—e—which occupies two points in space—p1 and p2
—without containing parts that are con ned (at that time) to those points in space. The error in this
argument is, of course, the assumption that there is any reference frame in which p1 and p2 are
simultaneous: since the points are each on the worldline of a particle, one must be in the absolute future of
the other.

The relativistic account of endurance presented in this section has assumed substantivalism about
spacetime. As argued in Section 8 below, there are di culties in combining relationalism about classical
spacetime with endurance; relationalism about Minkowski spacetime would presumably inherit the same
di culties. Moreover, published developments of relationalism about Minkowski spacetime of which I am
17
aware have presupposed perdurance. But a detailed investigation of combining endurance with
relationalism about Minkowski spacetime is beyond my technical competence.
 The account also assumes two kinds of atomism, rst that spacetime is made of points, and secondly that
continuants are ultimately made up of simples. Though it is beyond my technical competence to explore
p. 87 dropping the rst assumption, dropping the second assumption by allowing ‘gunky’ inhabitants of
spacetime may require only a little modi cation. Assuming spacetime itself is not gunky, we can still utilize
overlap‐at‐a‐point as our fundamental mereological relation, and we can still take occupation to be a
relation between things and points of spacetime. (Any gunky thing will of course occupy many points.)
Property instantiation can presumably also remain relative to points of spacetime, although it is di cult to
anticipate obstacles since I have never seen any detailed theory of the properties of gunk constructed, never
mind a relativistic one.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

5. Space and Time Are Analogous

A common argument in favor of temporal parts is based on the analogy between space and time. We observe
an analogy between these dimensions in many respects. All physical objects are represented in our
experience as located in both space and time. There is a common topological and metrical structure between
any given spatial dimension and the temporal dimension (at least relative to any given reference frame).
Objects move in both space and time. Other times are as real as the present, just as other places are as real as
here. (The argument from analogy here depends on eternalism.) The more this analogy holds, the more
entitled we are to expect it to hold in new areas. We thereby should expect the part–whole relation to behave
with respect to time as it does with respect to space. But objects are spread out in space; few think that
everyday particulars are ‘wholly present’ at more than one place. By analogy, therefore, objects are spread
18
out in time as well: four‐dimensionalism is true.

One way of challenging the argument is to point out disanalogies between time and space that even the four‐
dimensionalist will accept. Unlike time, space has three dimensions and lacks a distinguished direction;
unlike space, time seems to be specially connected with causation. A second challenge would be this: why
should similarity in one respect, for example, metricality, persuade us of similarity in a quite di erent
p. 88 respect, namely, parthood? This weakness surfaces in the literature on this argument in various ways.
Chisholm (1971: 15–16; and 1976 app. A) for example, criticizes the argument by alleging the following
disanalogy between time and space:

Possibly, there exists an x such that every part x ever has exists at more than one time at the same
place.

NOT: possibly, there exists an x such that every part x ever has exists at more than one place at the
same time.

The former statement will be denied by a four‐dimensionalist: existing at multiple times requires distinct
parts con ned to those times. The analogy would then be reinstated. Chisholm realizes this, but claims that
it would be circular to base the argument on this sort of consideration, premised as it is on temporal parts.
No more circular, I think, than Chisholm's argument that time and space are not analogous, since
Chisholm's opponent accepts temporal parts. Nevertheless, the argument for temporal parts is dialectically
weak. The three‐dimensionalist need not feel awkward in claiming that three‐dimensionalism itself
constitutes a breakdown of an analogy that was limited to begin with. Similarly, in his classic article ‘Spatial
and Temporal Analogies and the Concept of Identity’ (1955), Richard Taylor argues for a pervasive analogy
19
between space and time, but many of the arguments simply assume the doctrine of temporal parts. This
sort of dialectical impasse is typical.

The traditional argument from analogy, then, is weak. Some improvement may be possible, however. Why
do we think that a desk, say, has spatial parts? Why do we not think that the desk is mereologically simple,
 ‘wholly present’ at multiple places? We do take the desk to have parts, justi ably I think. But various
arguments for spatial parts have plausible temporal analogs in favor of temporal parts.

One argument is that the desk is spatially extended. But if spatial extension implies the existence of spatial
parts, would not temporal extension imply the existence of temporal parts? It is true that we are dealing
with di erent sorts of extension in the di erent cases, but (assuming the B‐theory, anyway) there seems to
be nothing distinctive about spatial extension that makes the argument succeed only in the spatial case.

Another argument appeals to qualitative variation within objects. If the desk is smooth at one place and
p. 89 rough at another, by Leibniz's Law we naturally conclude that the thing that is smooth is distinct from
the thing that is rough, and thus that the desk has proper parts. (The argument of the previous paragraph

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

may be regarded as an instance of this argument: the di ering properties are properties of location. But
perhaps the argument is more plausible when the selected properties are intrinsic.) This argument, too, has
a temporal analog: the well‐known argument from temporary intrinsics, to be discussed in Section 6 below.
When I change from sitting to standing, Leibniz's Law seems to imply that the thing that is sitting is
distinct from the thing that is standing. These distinct things would be my temporal parts. It is common to
resist the argument by indexing properties to times—one and the same enduring thing is alleged to be
sitting at one time while standing at another time. This is indeed consistent with Leibniz's Law; but an
analogous move could render spatially wholly present objects consistent with Leibniz's Law. The desk might
be asserted to be smooth at one place and rough at another.

A third reason for postulating spatial parts is modal. Call the left half of the region of space occupied by the
desk ‘R’. R and its material contents might have been, intrinsically, exactly as they actually are even if the
rest of the world had been eliminated. In that case an object exactly occupying R would have existed. But
then we should postulate an object in actuality that exactly occupies R. For surely the elimination of the rest
of the world outside of region R would not bring a new object into existence; but what other actual object
could this object be, other than a part of the desk that actually occupies region R? (Some—for example Peter
van Inwagen; see Ch. 5, Sect. 6—would deny the premise that an object exactly occupying R would exist in
the world where everything else was eliminated. But even van Inwagen might accept a version of the
argument when R is required to be a point‐sized region of space.)

It is clear that the modal argument for spatial parts has a temporal analog. Note, though, that the
plausibility of either the spatial or temporal version of the argument may depend on the particular region, R,
chosen. In the spatial case, if R is chosen to be a very large part of the entire region occupied by the desk it is
tempting to say that the resulting isolated object is the original desk itself, in truncated form. This would
undermine the argument's nal step, that the isolated object can only be identi ed with an actual proper
part of the desk. In the temporal analog of this argument, if the isolated region is chosen as an initial
p. 90 segment of the life of a thing (even a short one), it is tempting to identify the isolated object with the
original object itself, similarly undermining the argument. Nevertheless, certain cases of each argument
fare better, cases in which we are not tempted to make these identi cations. When a small spatial region of
the desk is selected, the resulting isolated object seems not to be the original desk, and when a small nal
temporal segment of a thing's life is isolated, it is much less tempting to identify the result with the original
thing. And perhaps if the modal argument succeeds in generating parts of a thing in certain of the regions of
space or time that it occupies, perhaps considerations of symmetry would justify postulating parts in the
rest of those regions, regions for which the modal argument is implausible.

A related argument involves division. Suppose the desk is divided into halves. (Followers of van Inwagen
may substitute division of a living thing into point‐sized parts.) Since the halves are spatially separated, they
are distinct objects. These objects were presumably not created by division, so they must have existed before
20
division. But surely it does not matter whether the division ever actually occurs, so anything potentially
divisible must actually have parts.
 One might think the division argument has no temporal analog, until Taylor‐style spatial and temporal
analogies are recalled. Begin by examining the spatial division argument more closely, in terms of the total
region of spacetime involved. At time t1 the desk is intact; by time t2 it has been divided in two halves, which
occupy spatial regions R1 and R2, as in Figure 4.1. The argument is that since R1 and R2 are spatially
separated, they must contain distinct objects, although notice that R1 and R2 are spatiotemporally
connected, via the ‘V‐shaped’ path.

Fig. 4.1.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

Division into spatial parts

But now consider Figure 4.2, in which the temporal and spatial axes are reversed (assume, for simplicity,
that the regions R1 and R2 are point‐sized). We now have a case in which an object is ‘separated into its
temporal parts’, only now the dimension of separation between the temporal parts is temporal, and the
‘dynamic’ dimension along which the separation takes place is spatial. Less cryptically, the gure now
represents a point particle that begins at place p2 at time t1, moves to place p1 at time t2, and then moves
p. 91 back to place p2 at time t3. The temporal analog of the argument for parts now runs as follows. Regions R1
and R2 are separated in time (though connected by a ‘V‐shaped’ spatiotemporal path). Thus, we must posit
distinct objects occupying those regions. These distinct objects can only be temporal parts. But then, the
temporal parts ought to be posited even when the spacetime path does not take such a V‐shape.

Fig. 4.2.

Division into temporal parts

Those who wish to resist temporal parts face a double challenge. It is not enough to show why the temporal
arguments just discussed fail, for the opponents of temporal parts typically accept spatial parts and so must
show why their objections to the temporal arguments do not carry over, invariably, to their spatial
p. 92 counterparts. Of course, it can be admitted that some of the spatial arguments fail as well. But at least one
must succeed if the belief in spatial parts is to be justi ed, and then we can ask why its temporal analog
fails.
 The response will presumably be a particularly stubborn version of the response to the original argument
from analogy. Some spatial argument succeeds where its temporal counterpart fails because it involves
space, not time. The rst argument assumes that spatial extension implies spatial parts, the second that
properties should not be indexed to places, the third that spatial isolation cannot ‘create’ a new entity, and
the fourth that spatially disjoint (though spatiotemporally connected) regions must be occupied by distinct
things. Perhaps the analogous principles assumed in the temporal arguments are less plausible: that
temporal extension implies temporal parts, that properties should not be indexed to times, that temporal
isolation cannot create a new entity, and that temporally disjoint (though spatiotemporally connected)
regions must be occupied by distinct things. Perhaps. In a sense, we have returned to our starting‐point: the
friend of temporal parts upholds a spacetime analogy that the opponent resists. But now there is more

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

pressure on the opponent of temporal parts, for she must positively commit to an argument for parts in the
spatial case, and so has the burden of explaining why that argument does not carry over to the temporal
case.

6. The Problem of Temporary Intrinsics

The traditional problem of change is that a changing object appears to exemplify incompatible properties.
Right now I am bent, since I am sitting. When I stand I will be straight. Straightness and bentness seem to be
incompatible, and yet I instantiate each. The formal contradiction can, of course, be resolved by pointing
out that these incompatible properties are had at di erent times, but some are not satis ed with this glib
response. Hermann Lotze dismissed similar glib responses to the problem of change as follows:

In what sense can that at di erent moments remain identical with itself, which yet in one of these
moments is not identically like itself as it was in the other? It is scarcely necessary to remark how
entirely unpro table the answers are which in the ordinary course of thought are commonly given
to this question; such as, The essence always remains the same with itself, only the phenomenon
p. 93 changes; the matter remains the same, the form alters; essential properties persist, but many
unessential ones come and go; the Thing itself abides, only its states are variable. All these
expressions presuppose what we want to know. (Lotze, 1887: 62–3)

My guess is that he would have as little patience for the present response as for the ones he considered.

The glib response is not entirely useless. The problem of change is sometimes put forth by a skeptic as
showing that no object persists through change. The glib reply is then that being bent now, and being straight
in 30 minutes are not incompatible. This will not convince the skeptic that I do persist over time, since it
presupposes that I have the property being straight in 30 minutes, and therefore that I persist. But it is
familiar that convincing the skeptic and remaining unconvinced by the skeptic are two very di erent things.

It is likewise familiar, however, that the skeptic plays a valuable role even if we remain sure she is wrong,
for discovering exactly where her arguments fail may teach us something valuable about the concepts
21
contained therein. The glib response to the problem of change is unsatisfying, for the skeptic about
persistence will ask: what is it for objects to have properties at di erent times? A more satisfying answer
requires constructing a metaphysics of change.

The temporal parts account of change is that incompatible properties are had by di erent objects, di erent
temporal parts of the whole. Change is therefore no more remarkable than the variation of a road with some
bumpy stretches and some smooth stretches.

The question then is whether endurantist accounts of change are possible. David Lewis's discussion of ‘the
problem of temporary intrinsics’ (1986a: 202–4) consists primarily of a negative answer to this question.
 Lewis presents three models of change and argues that the only acceptable one invokes temporal parts. The
three models are these:

(1) An eternalist‐endurantist model according to which so‐called properties are in fact relations to times.
This model in e ect elevates what I called the glib response into serious metaphysics. I bear the bent‐
22
at relation to the present time, and the straight‐at relation to another.

p. 94 (2) A presentist model. On this model, the only properties a changing object has are its present
properties, because only the present time is real. All facts are facts about what objects are like now.
On this view, the apparent contradiction between incompatible properties vanishes with the
banished past and future. Since I am presently bent, I am bent simpliciter, and in no way straight,

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

since there is no future in which to locate my straightness.

(3) A temporal parts model, according to which the incompatible properties involved in change are
really had, not by the persisting object itself, but rather by its temporal parts. Change is
heterogeneity of temporal parts.

Lewis provides reasons to reject models (1) and (2), leaving us with model (3) and temporal parts.

Given Chapter 2, the presentist model may be set aside. Note, however, that there is a worry here that
Lewis's trilemma is not exhaustive. The attraction of the presentist's account of change is that it allows us
to claim that I am bent simpliciter, rather than bent with respect to some time. But this virtue might be
thought to ow just from taking tense seriously, without requiring the additional claim that only the
present is real. As mentioned in Chapter 2, it is possible to combine eternalism with irreducible tense. The
defender of this combination might claim that objects have temporary intrinsics simpliciter, which change
over time in the sense that irreducibly tensed propositions about the instantiation simpliciter of temporary
intrinsics are true. An object that exists currently has the property of being bent, objects located in the past
have the property of being formerly bent, and so on. This view might appear to have the virtues of
presentism while lacking its defects.

Suppose someone accepts the existence of past and future objects while denying that tenseless facts about
these objects exhaust the facts about reality. Of any such person we must ask what exactly is missing from a
tenseless description of reality. One answer is that of the moving spotlight view discussed in Chapter 2: the
B‐theoretic description of the world is completely adequate except that it leaves out information about
which time is present. But information about the properties of objects at various times is part of the
tenseless information regarded as unproblematic by the moving spotlight theorist; the moving spotlight
theorist, therefore, cannot deliver an alternative to the three models of change listed above.

p. 95 For a distinctive solution to the problem of change we would need to go further and claim that more is
wrong with a B‐theoretic description of reality beyond its omission of facts about presentness. Irreducibly
tensed facts must include facts about property instantiation in addition to facts about presentness. All that
can be accurately reported from a tenseless point of view is which objects exist from the past, present, and
future; to report the features of these objects—their properties and relations—requires irreducible tense.
But now we have a very strange combination of views. What properties does Socrates have? The standard B‐
theorist's answer is that all talk of property instantiation is fundamentally tenseless. One can then say that
Socrates has the property of sitting at various times at which he existed, has the property of standing at
other times at which he existed, and so on. Once tenseless property attribution is rejected in favor of taking
tense seriously, any properties attributed to Socrates would need to be properties he has now. This turns
Socrates into a very shadowy gure: he has properties like formerly sitting, formerly being wise, and so on, but
for no ‘manifest’ or categorical property, F, can we say that Socrates is F. We got ourselves into this sorry
position by accepting an element of the B‐theory—atemporal quanti cation—while rejecting too much else
 of the picture, namely tenseless property exempli cation. This fourth model of change, therefore, is
23
unattractive. It seems, therefore, not to be a viable alternative to Lewis's three models of change.

The most important part of Lewis's argument is his argument against model (1). According to this model,
what we ordinarily think of as properties are in fact relations to times. But this is an implausible view,
according to Lewis, given that change sometimes involves intrinsic properties. Being bent is intrinsic; its
instantiation by me cannot, therefore, involve my relations to other things, not even times. Surely I am just
plain bent, not bent with respect to something else.

It is worth distinguishing two objections Lewis gives to this view, though I think that the bottom line is the
same. The rst is that the eternalist‐endurantist model would obliterate the distinction between intrinsic

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

and relational properties; all properties would turn out relational. As Sally Haslanger points out (1989a:
p. 96 123–4), the eternalist‐endurantist could respond with an understanding of ‘intrinsic’ that would allow
for a distinction to be drawn, despite the fact that instantiation of properties is temporally relative:

P is intrinsic i for every time at which an object has P, it has it then solely in virtue of the way it is
then.

This is in contrast to Lewis's gloss on intrinsic properties as those instantiated by an object in virtue of the
way that object is, considered in itself (1983b: 197). Lewis anticipates such a response:

[a defender of properties had at times can draw a distinction] he will call the distinction between
matters of one's own intrinsic character and matters of one's relationships . . . [but] his account
reveals that really he treats shape, no less than unclehood, as a matter of relations. In this account,
nothing just has a shape simpliciter. (1988a: 65)

Lewis is insisting on the following as a datum: some things have shapes simpliciter.

The second version of the objection is this: ‘If we know what shape is, we know that it is a property, not a
relation’ (Lewis 1986a: 204). Some have responded by defending theories of property instantiation that
allow shapes to be properties but which relativize the instantiation relation to times in some way. Property
instantiation for van Inwagen (1990b) and Mark Johnston (1987) is a three‐place relation between a thing,
property, and time; for Sally Haslanger (1985, 1989a), things and one‐place properties combine to form
propositions, which are then true or false relative to times. On all of these views, shapes are properties but
are nevertheless had relative to times; they are ‘relativized properties’. I agree with Lewis (1988a: 65–6
note 1) and Mark Hinchli that such circumvention accomplishes little. Hinchli puts this well: ‘As
appealing as the relativization strategy is, it fails to accommodate our intuition that shapes are properties.
Relativized properties are not properties, because a thing cannot just have them. So what are they? They are
nothing new; they are relations in disguise’ (1996: 122). We have arrived at the same place as above: Lewis's
argument against model (1) rests on the alleged datum that some objects are ‘just plain straight’, that
shapes can be ascribed to objects without quali cation or relativity of any kind.

Can any progress be made here, or must the datum be baldly asserted as intuition? As discussed in Section 5
p. 97 above, one might press the defender of eternalist endurance on the analogy between space and time. If
the argument from temporary intrinsics for temporal parts is to be resisted by indexing to times, why not
also resist the analogous argument for spatial parts, index properties to places, and claim that objects are
wholly present across space as well as time? Of course, this move inherits the weakness of the other
arguments from analogy considered above: the eternalist can admit some analogies between time and space
while denying a general analogy.

The argument, then, seems only as strong as the claim on which it is based, that some objects are ‘just plain
straight’. It is unclear how strong this basis is. There is indeed something odd about relativizing shapes:
 ‘One and the same enduring thing may bear the bent‐shape relation to some times, and the straight‐shape
relation to others. In itself, considered apart from its relations to other things, it has no shape at all’ (Lewis
1986a: 204). But, it might be asked, when we are clear that time is the source of the relativity, what is wrong
with denying that objects are ever just plain straight? We ordinarily speak as if objects are just plain
straight, but present tense is normally implicit. It is therefore easy to forget that ‘x is straight’ means ‘x is
straight now’. Once this is made clear, Lewis faces E. J. Lowe's (1988a: 73) question: ‘why can't we just
happily deny that changeable, persisting physical objects can ever correctly be described as “just plain being
straight”?’ (As Lowe immediately observes, Lewis would say we can deny this, since it is not changeable
persisting objects—spacetime worms—that are just plain straight, but rather temporal parts. But what
Lowe seems to be saying in the following sentences is that we can happily deny that anything is just plain

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

24
straight.) Nothing is simply straight or bent, since things change.

At this point it is hard to evaluate the argument. The main premise of the argument, that some things are
just plain straight, is accepted by the argument's proponents, and denied by its targets. The argument does,
I think, favor four‐dimensionalism, but not strongly so. We do have some initial attachment to the main
premise, but the rejection of that premise is certainly not rationally unacceptable.

It is worth noting that the version of four‐dimensionalism that is best supported by the argument from
p. 98 temporary intrinsics is not actually the view that ordinary continuants are spacetime worms, but rather the
view (to be defended in Ch. 5, Sect. 8) that ordinary continuants are to be identi ed with instantaneous
stages. As noted, Lewis himself cannot claim that ordinary continuants are just plain straight since for him
continuants are sums of stages, some straight, others not. Lewis's objection to model (1) must be carefully
quali ed: it is that model (1) implies that no objects are just plain straight. But here Lewis's critics have
pounced. How can Lewis lay claim to ordinary intuitions if his claim does not involve continuants—the
objects of our everyday ontology? Moreover, Lewis's theory of change has its own counterintuitive
consequence: it violates a plausible principle about change, that an object that changes shape must itself
25
have a shape simpliciter. As Hinchli (1996: 120) puts it, discussing a changing candle, ‘If the candle never
has the shapes itself, it cannot change its shape’. However, this rejoinder to Lewis only applies to the worm
view. According to the stage view, I myself have the property being straight, for I am a stage, not a spacetime
worm. Thus, the stage view allows both that temporary intrinsics are instantiated simpliciter and that they
26
are instantiated by ordinary continuants such as persons and candles.

7. Arguments from Exotica

Assume the B‐theory of time (see Ch. 2), that past and future objects exist in addition to presently existing
ones, and that tensed notions reduce to tenseless ones. (Or, for that matter, assume either the moving‐
spotlight or the growing block universe theory of time; in this section it will not matter.) Then either the
27
subjects of momentary property instantiation are stages, or we must index instantiation to times. So any
argument against indexing is an argument for temporal parts. Lewis's argument against indexing has been
discussed; in this section I consider two others. The rst is based on the possibility of a timeless world.
Some may consider this possibility too bizarre to be useful in argumentation. Fortunately, while the second
p. 99 argument is also based on a rather exotic possibility (time travel), there are powerful reasons to take that
possibility seriously—that argument cannot be so lightly dismissed.

7.1. The Argument from Timeless Worlds

If being straight‐shaped is a relation to times, nothing would remain straight‐shaped if you cut away all the
times from a world. And yet surely objects in a timeless world could be straight‐shaped.
 More carefully, consider two worlds: rst, a non‐relativistic world with a Euclidean space and separable time
dimension, and secondly, a world without any time dimension at all. It seems plausible that a persisting
object in the rst world could be straight‐shaped in exactly the same way as a timeless object in the second.
Since time and space are separable, shape seems to be a purely spatial property, and thus should remain if
the time dimension is removed. Given four‐dimensionalism this is straightforward: straightness is a property
that can be instantiated by instantaneous stages and alien timeless objects alike. But in one way or another,
the three‐dimensionalist will need to treat the straightnesses of the timeless and temporal objects
di erently. If straightness is a relation to times, there is no time in the timeless world for the object to be
related to. At best, the timeless object could instantiate some closely related property. But what would this
property have to do with the straightness relation? Similar but more complicated remarks apply to endurance

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

theories that postulate three‐place instantiation relations, or states of a airs true at times: the timeless
object cannot instantiate straightness in the same way as the temporal object since a di erent relation of
property instantiation or truth of states of a airs will be involved in the di erent cases.

The claim that timeless and temporal objects could be straight in the same way is not so plausible when
applied to a relativistic world. There, shapes are not purely spatial since no dimension is purely spatial.
However, an analogous argument will hold for non‐spatial intrinsic properties: it should be possible for an
electron to have unit negative charge in a purely spatial world without time in the same way that a
persisting electron in the actual, relativistic, spatiotemporal world has unit negative charge. A three‐
dimensionalist must say that actual charge is a relation to points in spacetime, which cannot be instantiated
p. 100 in a timeless world. It might be responded that electrons have their charge permanently, and so charge is
had absolutely, not relative to points in spacetime. However, it is unlikely that the three‐dimensionalist
would want to rest the case on this; for all we know, we might discover tomorrow that some subatomic
particles have their charges temporarily.

It might be argued that a single abstract entity, straightness, is a relation to times in the temporal world and
a property in the timeless world. Thus the temporal and timeless objects could each instantiate straightness.
This reply strains our conception of the modal properties of abstract entities. But even if we grant the
objector that straightness is involved in both worlds, the timeless object still is not straight in exactly the
same way as is the persisting object since straightness is in one case a property and in the other case a
relation.

Timeless worlds are admittedly quite distant from ordinary concerns; denying my claims about them, or
denying their possibility, would not count too terribly heavily against the three‐dimensionalist. My own
feeling is that we do have some opinions about such exotic possibilities; hence the argument retains some
weight.

Furthermore, a more mundane version of the argument can be advanced against certain versions of three‐
28
dimensionalism. Three‐dimensionalists deny that continuants perdure, but some—call them dualists—
29
believe that in addition to continuants there exist events, or facts, which do perdure. My life history, for
example, is an event that is in many ways exactly like me, except that it, being an event, has temporal parts.
It is present everywhere that I am present, is straight‐shaped when I am, and bent when I am. However, its
being straight is importantly di erent from my being straight, for when I am straight I bear a relation to a
time, whereas it is straight by having a temporal part at the time that is straight simpliciter. This relationship
between straightness‐at‐a‐time and straightness simpliciter is mysterious—why does each count as a kind of
straightness? Moreover, why is it a necessary truth (as it must be) that I bear the straightness relation to a
time i the temporal part of my life at that time instantiates the straightness property? Relatedly, the
dualists say that my life history has parts; but these parts would be parts simpliciter, and what does the two‐
place relation of parthood simpliciter have to do with the three‐place relation of having a part at a time? (Yet
p. 101 another related argument: some three‐dimensionalists want to apply parthood in other cases where
temporal relativization is inappropriate: to abstract entities, or to sets, or to intervals of time. But parthood
 could not hold between such entities in the same way it holds between continuants, if parthood for the latter
must always be temporally relativized.)

A three‐dimensionalist might accept the existence of events but construct them in such a way as to render
this argument ine ective. The event x 's being F at t might be identi ed with the ordered triple < x, F, t>;
such an event might be said to instantiate property F simpliciter i x instantiates F at t. Property
instantiation by events would thus be analyzed in terms of continuants bearing relations to times, and the
argument would not apply. It is towards those dualists like D. H. Mellor (1981, 1998), according to whom
events are sui generis entities that genuinely perdure, that the argument is directed.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

7.2. The Argument from Time Travel
A second argument from exotica, to my mind more compelling, may be based on time travel. Three‐
dimensionalism, I will argue, is inconsistent with the possibility of traveling back in time to meet one's
former self. Since there are philosophical and physical reasons to believe in the possibility of time travel,
this in exibility of three‐dimensionalism is unwelcome.

Suppose I travel back in time and stand in a room with my sitting 10‐year‐old self. I seem to be both sitting
and standing, but how can that be? The four‐dimensionalist's answer is that there are two distinct person
30
stages, one standing, the other sitting. (Given the Chapter 3 de nition of a temporal part, the fusion of
these two stages counts as my temporal part at the time in question, so let us understand ‘person stage’ to
refer to ‘person‐like’ parts of temporal parts. Ordinarily my temporal part at any time is a person stage, but
not in cases of time travel.) If three‐dimensionalism is true, on the other hand, the case involves only a
single ‘wholly present’ person, which seems to be both sitting and standing.

p. 102 This might be denied. Perhaps the object that sits and the object that stands are two distinct spatial parts of
me at the time. These spatial parts cannot both be identical to me; otherwise they would be identical to each
other. Indeed, the three‐dimensionalist will probably want to say that neither is identical to me; rather, I
have two person‐like proper parts at the time. (Anyway, remarks similar to those I am about to make would
apply if one of the spatial parts is me.)

Where did these spatial parts come from? Presumably they popped into existence upon arrival of the time
machine; there seem to be no future or past objects with which they could be identi ed. But then ‘a meeting
of my former self and me’ no longer seems an apt description of the event. Rather, it seems to be an event in
which two wholly new persons meet each other. The later self who looks at what he takes to be himself and
thinks ‘I wonder what I was thinking at this moment’ is not me, and is not in fact looking at himself. He is a
person who just came into existence and mistakenly thinks he is a time traveler. It might be argued that it is
really the composite of the two spatial parts who is looking and thinking, though of course it is doing so
through one of its spatial parts. If this is correct, then since I am the composite, I really would be looking at
myself and thinking about myself. But this is a distortion of the concept of a person. The correct description
is quite clearly that the spatial parts are the persons, and that the case is not one of a person meeting his
former self.

It seems, then, that to genuinely accommodate time travel, the three‐dimensionalist must describe the case
as involving only a single wholly present person. This leads to trouble. The most immediate oddity is that
the person appears to have incompatible properties: standing‐at‐t and sitting‐at‐t, where t is the time in
question. (Similar remarks apply to the sophisticated variants of the relations‐to‐times theory, for example
that instantiation is three‐place.)

The three‐dimensionalist may be willing to live with this oddity, that something can stand and sit at a single
time. After all, it is not as if the thing is standing and also not standing. Standing and sitting turn out not to
 be incompatible in the bizarre case of time travel.

There is a further objection that cuts deeper. In the story I told, I am standing while my former self is sitting.
But our roles might have been reversed—I might have sat where he sits while he stood where I actually
p. 103 stand. We have here what appear to be two distinct possibilities. When I meet my former self I think: ‘I
am standing and he is sitting, but it could have been the case that I sat and he stood.’ The problem is that the
three‐dimensionalist cannot distinguish these possibilities. The four‐dimensionalist can easily distinguish
them. Each case involves two person stages, T1 and T2. In one case T1 stands and T2 sits, whereas in the other
case T2 stands and T1 sits. But the three‐dimensionalist can only speak of what properties I have at t. In the
rst possibility, I am standing at t, and also sitting at t. Exactly the same is true in the second possibility.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

(Relatedly, the three‐dimensionalist will have trouble distinguishing the case where ‘each’ sits from the
31
case in which ‘just one’ sits. )

The cases could be distinguished after all if I and my former self were qualitatively di erent. If I were taller
than my former self, for example, then in the case where I stand a certain point in space would be occupied
by me—namely, the point occupied by the top of my head—but would not be occupied by me in the case
where it is my shorter former self that is standing. So let us stipulate that I and my former self do not di er
in ways that would thus distinguish the cases.

The three‐dimensionalist might try to distinguish the possible worlds in which these cases occur by
di erences at times other than the time of the meeting. To get around this sort of problem, let us stipulate
that my time machine transports me ‘instantaneously’ into the room with my earlier self, and that my later
self is then immediately annihilated after the meeting. But might not whether my later self sits or stands be
caused by whether I am sitting or standing while entering the time machine? Given any such causal
dependence of the later self's state on my state while entering the time machine, the cases might be
distinguishable after all. I cannot get around this problem by stipulating that my emergence in the past is
entirely causally unconnected to my entering the time machine since I accept that causation is a prerequisite
of personal identity. What I need to do instead is, roughly, stipulate that my entry into the time machine
causally determines that I emerge at the time of the meeting, but does not cause whether I am sitting or
standing then. More carefully, I stipulate the following. First, in each case, I am standing when I enter the
time machine. Secondly, in each case I am standing shortly before the time at which my former and later
p. 104 selves meet. Thirdly, the laws of nature in the two cases are identical, and are indeterministic. As applied
to my entry into the time machine, they determine that I will emerge at a certain spatial location in the past,
but they leave it open whether I am sitting or standing at the time. As applied to me immediately before the
meeting of my former and later selves, the laws determine that I will remain in a certain spatial location, but
leave it open whether I will be sitting or standing at that time. Given these stipulations, it can be argued that
the cases cannot be distinguished by causal means. Given stipulations one and two, the cases do not di er in
non‐causal ways at times other than the time of the meeting. And given three‐dimensionalism, the two cases
do not di er in non‐causal ways at the time of the meeting (in each case I am sitting at the time and
standing at the time). Given stipulation three, the laws of nature are common to the two cases. But any
di erences in causation between the cases would need to be due to di erences in the laws of nature or non‐
causal di erences. Thus, the worlds do not di er causally.

(‘Singularists’ about causation reject the principle that di erences in causation must be due to di erences
in the laws of nature or non‐causal di erences. But even they may be willing to allow me to stipulate that the
causal relation holds between my entering the time machine and my emergence at a certain place at the time of
the meeting, and between my being located at a certain place at times before the meeting and my being located at
that place at the time of the meeting, but that the causal relation does not hold between any event after or
before the meeting and any event specifying whether I sit or stand at the time of the meeting.)
 Suppose my ‘later self’ is located at place P, whereas my ‘earlier self’ is located at place P′. We might try to
distinguish the possibilities by saying that in the rst case I have the property standing (at t ) at P, but not
the property standing (at t ) at P′, whereas in the second case I have the latter property but not the former.

This response appears unsuccessful. What is it to have a property at a certain place (at a given time)? One of
two things, I would have thought: (1) to be both located at the place (at the time) and to have the property (at
the time), as when we say that I have the property of typing in my o ce at 3.15 p.m., EST, on 5 November
2000; (2) to have a part (at the time) that has the property (at the time) and is located at the place (at the
time), as when we say that a road is bumpy at a place. Neither helps the defender of endurance. The second
reading may be set aside, since the three‐dimensionalist denies I have temporal parts. As for the rst

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

p. 105 reading, in each possibility I am located at P (at t), and also at P′ (at t). And in each case I am standing (at
t). Thus, in each case, I have both standing (at t ) at P and standing (at t ) at P′. The possibilities still have not
been distinguished.

The defender of endurance might claim that the relativity of instantiation to places is irreducible. For that
matter, as we saw in Section 4 above, special relativity requires relativizing property instantiation to
spacetime points anyway. The latter di culty may be circumvented by stipulating that the case in question
is to take place in a non‐relativistic world. Still, the endurance theorist might insist that even in non‐
relativistic worlds, property instantiation is to be irreducibly relative to both places and times. But there is
no independent motivation for this claim. Moreover, if all properties are relativized to places, one important
argument for spatial parts goes away, namely the problem of ‘local’ intrinsics (Sect. 5); why not then claim
32
that objects are wholly present across space as well as time?

I have granted relativization to times for the sake of argument and then complained about relativization to
places. Interestingly, something like this complaint can be made even under special relativity: one can insist
that point‐particles, anyway, can di er with respect to a property at a pair of points only when the points are
time‐like related. This is probably just what the defender of endurance wants, since it allows intrinsic
change yet preserves the argument from local intrinsics for spatial parts. But now imagine a time‐traveling
electron meeting its former self, where the former self is F and the later self is F′. Since the spacetime points
at which F and F′ are instantiated are space‐like related, the time travel argument goes through.

I note nally that the argument could appeal to time‐traveling objects that lack spatial location, in which
case the relativization response is unavailable. A desperate reply would be to appeal to di erences in which
properties are co‐instantiated with which. The reply is desperate because it depends on rejecting what is
surely a correct account of co‐instantiation, namely that F and G are co‐instantiated by x at t i x instantiates
p. 106 F at t and x instantiates G at t. For the reply to work, property instantiation would need to be irreducibly
relative, not only to time and to place, but also to other properties.

One might try to distinguish the possibilities by relativizing property instantiation to ‘personal time’ (as in
Lewis 1976) rather than external time. Personal time is time experienced by the time traveler, whereas
external time is time simpliciter, time according to the public ordering of events. If I travel back to the time
of the dinosaurs, my arrival is located before my departure in external time, but after it in my personal time.
Now imagine I travel back in time and meet myself in 1977. Though I am both sitting and standing at a
single moment of external time, there are two moments of my personal time involved, call them pt1 and pt2.
The cases may then be distinguished by claiming that in one case I stand at pt1, whereas in the other I stand
at pt2.

The problem is that personal time, as construed by Lewis anyway, is not an additional fundamental physical
element of the world, but is rather a de ned quantity. Roughly, experiencing one minute of personal time is
de ned as undergoing the amount of change that would normally occur to a person during one minute of
external time. Since personal time is a de ned quantity, all facts about personal time must be capable of
 being stated in vocabulary that just mentions external time. But it has already been argued that the
possibilities in question cannot be distinguished in this vocabulary by the defender of endurance. One might
consider instead cases of time travel in which personal time is not a de ned quantity, but rather a second
physically real temporal dimension. I do not deny that two‐dimensional time is one way to make sense of
time travel, but I also think that time travel is intelligible in worlds with a single temporal dimension, and I
am free to stipulate that the case of time travel I am considering is in such a world.

The argument from time travel is only as powerful as its assumption that time travel is possible. Some may
wish to deny that claim. Some might even regard inconsistency with time travel as an advantage of three‐
dimensionalism, as a vindication of a prior belief that time travel is impossible! I see no merit in these
33

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

claims. Arguments that time travel involves some sort of inconsistency fail. And there are two good
reasons to think that time travel is possible. First, the fact that physicists take the physical possibility of
p. 107 time travel seriously suggests that time travel may well be physically possible. This will be discussed
below. Secondly, time travel is conceivable or imaginable, which creates a presumption in favor of its
possibility. The presumption is defeasible. It would certainly be defeated if time travel could be shown to
involve some hidden inconsistency, but as I have said, arguments attempting to demonstrate such an
inconsistency do not succeed. Presumptions of possibility can also be defeated if the possibility of the
proposition in question would con ict with other things reasonably believed. Whether this is the case for
time travel depends on an assessment of all the issues discussed in this book. But so long as time travel is
presumptively possible, the argument from time travel provides a prima‐facie reason to reject three‐
dimensionalism, and can take its place in the cumulative case I am amassing in favor of four‐
dimensionalism.

A priori arguments about possibility are sometimes thought to be threatened by the phenomenon of the
necessary a posteriori (Putnam 1975; Kripke 1972). Once it was perfectly conceivable that water was not H2O
and that heat was not the motion of molecules; nevertheless, it is argued, these things have turned out to be
metaphysically impossible. Perhaps time travel and timeless objects are also conceivable but impossible.

My arguments do not directly assume that conceivability entails possibility. So long as conceivability
provides a defeasible reason in favor of possibility, the arguments have some force. I admit, though, that if
the analogy with familiar cases of the necessary a posteriori is correct then my arguments would be
undermined. They would be like arguing against the identity of Hesperus and Phosphorus because of the
conceivability of discovering its falsity.

Fortunately the analogy is bad. Take a typical case of the necessary a posteriori: that water is H2O. It is
crucial to the standard Kripkean explanation of the divergence between conceivability and possibility in this
case that there is a possible world in which our conventional activity concerning the term ‘water’ is just the
same as it actually is, but in which ‘water is H2O’ expresses a falsehood. Such a world might be one in which
XYZ, rather than H2O, is in rivers and oceans. We think that water might not have been H2O because we see
that a certain scenario is indeed possible—a scenario in which everything is ‘qualitatively’ exactly as it
actually is but in which ‘water is H2O’ expresses a falsehood—and we then confuse the possibility of this
p. 108 scenario with the possibility of water's failing to be H2O. The world in which XYZ lls our rivers and
oceans is not a world in which water fails to be H2O. Why not? Why doesn't the existence of the XYZ world
imply the truth of ‘possibly, water is not H2O’? Because it is part of the semantics of ‘water’ that an
utterance of ‘water’ in a world w rigidly refers to whatever stu plays the ‘water role’ in w. That is,
‘possibly, water is F’ is true, as uttered in w, i for some world, v, the stu that plays the water role in w is F
in v; ‘necessarily, water is F’ is true, as uttered in w, i in every world, v, the stu that plays the water role in
w is F in v. So the inhabitants of the XYZ world could say, truly, ‘water is not H2O’ (and indeed, ‘necessarily,
water is not H2O’). Nevertheless, we speak truly when we say ‘necessarily, water is H2O’.
 But in the cases of time travel and timeless worlds, there is no term involved in the description of the
relevant possibilities to play the role ‘water’ plays in the case just considered. ‘Water’ is a natural kind term
that rigidly refers to di erent substances depending on what substance happens to play the ‘water‐role’ in
the world of utterance. In David Chalmers's (1996: 56–71) terminology, ‘water’ is two‐dimensionally
inconstant, in that it has di erent secondary intensions relative to di erent worlds of utterance. That is
how there can be a world in which ‘water is not H2O’ could be truly uttered (while remaining governed by
the same conventions that actually govern that sentence) without rendering actual utterances of ‘possibly,
water is not H2O’ true. But persistence words presumably have constant two‐dimensional intensions.
Relative to any world of utterance, in any world of evaluation, persistence words pick out whatever plays the
‘persistence‐role’ in that world. For example, if our world is non‐relativistic, relativistic worlds should count

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

as worlds in which things persist; persistence words would not rigidly designate actual, non‐relativistic
persistence. Similarly for four‐dimensional persistence: if there are four‐dimensional worlds in which
inhabitants using our same linguistic conventions could say truly ‘time travel has occurred’, then the
34
sentence ‘it is possible that time travel occur’ would be actually true.

The phenomenon of the necessary a posteriori, then, gives us no reason to distrust our modal intuitions
p. 109 about time travel and timeless worlds. There might be other reasons to distrust those intuitions. One
might, for example, be some sort of modal skeptic (see e.g. van Inwagen 1998). Or one might have a view of
modality on which the facts of modality, even facts expressible in two‐dimensionally constant terms,
depend crucially on what happens to be actual. The a priori part of the defense of the possibility of time
travel is therefore dependent on a modal epistemology on which intuitions about possibility supply genuine,
even if defeasible, evidence.

A three‐dimensionalist might grant that time travel is possible, and that its possibility requires temporal
parts, and thus concede that temporal parts are possible, while denying that they are actual. Three‐
dimensionalism would then be a contingent thesis. I note that three‐dimensionalists seem not to have this
attitude. They tend to regard three‐dimensionalism as a necessary truth, perhaps because they take it to be a
conceptual truth that perdurance would not be genuine persistence, that four‐dimensional change would
not be genuine change, and so on. Moreover, even if a three‐dimensionalist were to admit that four‐
dimensionalism is metaphysically possible, she would presumably want to claim that it is impossible in some
sense. Four‐dimensional worlds would be very distant possibilities, worlds unlike our own in fundamental
ways. At the very least, the three‐dimensionalist will not want to admit that four‐dimensional worlds are
physically possible.

This brings us to the second reason for believing that time travel is possible: the fact that physicists take its
possibility seriously. This is not just a fringe movement: time travel is currently discussed in major physics
35
journals. It would be rash to rule out, a priori, the possibility of a physical hypothesis that is taken
seriously by physicists. Even defenders of contingent three‐dimensionalism, then, should be moved by the
time travel argument. So should skeptics who doubt conceivability arguments for possibility. So should
anyone who doubts my defense against the challenge of the necessary a posteriori. Anyone reluctant to
tread on the toes of physicists should reject metaphysical theories like three‐dimensionalism that preclude
time travel.
p. 110
8. The Argument from Spacetime

Given eternalism, which I continue to presuppose, spatiotemporal reality apparently consists of a spacetime
in which past, present, and future objects are located. Concerning spacetime itself one can be either a
substantivalist or a relationalist. The substantivalist takes the physicist's talk of spacetime at face value:
points and regions of spacetime are genuine entities. The relationalist, on the other hand, rejects the
genuine existence of points or regions of spacetime and reduces all talk of spacetime to talk of
spatiotemporal relations between entities she is willing to accept. Either substantivalism or relationalism is
true; I will argue in each case that the four‐dimensional ontology is favored. In brief: if we are
substantivalists then we should identify objects with regions of spacetime, which perdure; on the other

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

hand, relationalism is tenable only if the relata of spatiotemporal relations perdure; either way, the result is
a metaphysics of perduring objects.

First, assume that substantivalism is true, that there are such things as points and regions of spacetime.
There is then the question of whether there is anything else, whether spatiotemporal objects occupy, but are
distinct from, regions of spacetime, or whether they simply are regions of spacetime.

There is considerable pressure to give the latter answer, for otherwise we seem to gratuitously add a
36
category of entities to our ontology. All the properties apparently had by an occupant of spacetime can be
understood as being instantiated by the region of spacetime itself. The identi cation of spatiotemporal
objects with the regions is just crying out to be made. Given the identi cation, perdurance follows, since
37
spacetime perdures.

‘A region of spacetime bounded out the door and barked at the mailman’—it sure sounds strange to say!
p. 111 Indeed, it sounds like a ‘category mistake’. But this is not a good reason to resist the identi cation, any
more than the strangeness of saying that pain is located in the brain is a good reason to reject the identity
theory of mental states. As Paul Churchland (1988: 29–31) points out, Copernicus's claim that the earth
moves may well have seemed like a category mistake at the time. Our language may well have a prohibition
against saying that regions of spacetime bark. Perhaps this is due to ignorance of the metaphysical nature of
dogs: our progenitors simply did not know that dogs are regions of spacetime. Perhaps if we became
convinced of the identi cation of things with regions of spacetime we would reform our language. Or
perhaps not. We might accept the identi cation and yet choose to treat thing‐language and spacetime‐
language di erently. We might regard the predicate ‘barks’ as inappropriate to apply to an object when the
object is thought of as a region of spacetime, even if we acknowledge that the object is indeed a region of
spacetime. Appropriateness of a predicate would be determined not only by its object but also by how that
object is conceptualized. This sort of restriction on the appropriateness of predicate application emerges in
unrelated cases: those who regard persons as identical with aggregates of subatomic particles can
acknowledge the strangeness of saying ‘I was just kissed by a swarm of quarks and electrons’.

It may be objected that the very fact that spacetime perdures is reason to resist the identi cation:
spatiotemporal objects do not perdure and thus cannot be identi ed with anything that does. But what is the
evidence for this claim? It is perhaps not part of ordinary belief that spatiotemporal objects perdure, but
that should not stand in the way of a theoretical identi cation, any more than the fact that it is not part of
the everyday concept of heat that heat is molecular motion. Philosophical arguments against perdurance
will be taken up and rejected in Chapter 6; those do not stand in the way of the identi cation. Any remaining
resistance should be weighed against the ontological extravagance of a redundant ontological category.

Some will claim that regions of spacetime di er from the occupants of spacetime by having their
spatiotemporal size and location essentially. But given an appropriately exible understanding of de re
modality, the worry vanishes. On the counterpart‐theoretic account I favor, the claim that a is essentially F
means that all of a's counterparts are F. The counterpart relation is a similarity relation: a's counterparts
p. 112 are objects in other possible worlds that are similar to a in certain respects. But in di erent
conversational contexts, di erent respects of similarity, and thus di erent counterpart relations, are
relevant to the evaluation of the claim. In the case of modal claims about regions of spacetime there are two
di erent counterpart relations one could utilize. One counterpart relation stresses spatiotemporal
similarity; the other stresses similarity of a sort relevant to ordinary objects. The thought that
spatiotemporal facts are essential to regions of spacetime is a result of thinking in terms of the former
counterpart relation, whereas ordinary modal judgments are based on the latter counterpart relation. Thus,
the fact that we make apparently di erent modal judgments about spatiotemporal objects and regions of
spacetime does not preclude their identi cation: the judgments are judgments about the same objects under
38
di erent counterpart relations.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

As with other cases of ‘counterpart theory to the rescue’ in this book, there will be doubters. But it strikes
me as being fundamentally wrong‐headed to allow conservatism about de re modality to dictate one's
ontology in the way envisioned by the doubters. We have available to us a compelling reduction. It would be
terrible methodology to forgo this reduction because of worries about de re modality. Moreover, most of the
reasons people have for rejecting counterpart theory are fallacious, as I will argue in Chapter 5, Section 8.

It is worth stressing a point made by Harold Noonan (1991: 190; see also Lewis 1986a, sect. 4.5) that
counterpart theory is not strictly needed to rebut modal arguments of the sort considered here. The modal
argument against identifying things with regions of spacetime can be rejected by appeal to any account of de
re modality according to which the content of a de re modal judgment depends on how its subject is
conceptualized. (In Noonan's phrase, we need the claim that modal predicates are ‘Abelardian’.)
Counterpart theory is one way, but not the only way, of making good on this idea. And even if the truth of de
re modal judg‐ments is not held to depend on how their subjects are conceptualized, an analogous view
about modal intuitions may yet be correct. A four‐dimensionalist could grant that spatiotemporal objects
have their spatiotemporal size and location essentially but go on to give an error‐theory of ordinary modal
p. 113 intuitions. Intuitions about the modal properties of an object, on this view, depend on whether we think
39
of it as an ordinary object or as a region of spacetime. I think counterpart theory is the best of this lot, but
these alternate views would also defuse the modal objection to the reduction.

In exible accounts of de re modality lead to trouble anyway, for example in the modal versions of familiar
paradoxes of identity over time. A modal version of the statue/lump example, put forward by Alan Gibbard
(1975), runs as follows. A lump of clay, which we may name Lump1, is created already in statue form. It
composes a certain statue, which we may name Goliath; we then later destroy Lump1 and Goliath
simultaneously by annihilating the clay completely while it is still in statue form. Surely Lump1 and Goliath
are identical, for they coincide at all times. And yet modal di erences appear to stand in the way of this
identi cation: Lump1 appears to have, while Goliath appears to lack, the properties possibly having survived
attening, possibly failing to be a statue, and so on. Surely, this apparent di erence should be resisted; surely
it is due in some way to a shift in our conceptualization of a single object, rather than a di erence between
two objects; surely, adopting a exible account of de re modal predication, or a exible error‐theory, is a
40
more sensible alternative than multiplying entities corresponding to the apparent modal di erences. The
counterpart‐theoretic version of this is familiar. To say that Lump1 might have survived attening is to say
that Lump1 has lump counterparts that survive attening; to say that Goliath could not have survived
attening is to say that Goliath has no statue counterparts that survive attening. Nevertheless, Lump1 is
Goliath.

So far the argument has been that substantivalists should identify spatiotemporal objects with regions of
spacetime and must therefore admit that spatiotemporal objects perdure. Let us now turn to the other horn
of the dilemma. Can relationalism be made to work without accepting perdurance? As mentioned in Section
4 above, recent published versions of relationalism have utilized perduring entities; but is this inevitable?
p. 114 Though I have no knockdown objection, I will argue that the combination of spacetime relationalism and
 endurance has unattractive consequences. My discussion will ignore relativity; I presume this begs no
important questions.

Let us rst examine what a perdurance‐based relationalism looks like. The relationalist wants to capture
temporal facts using temporal relations between objects existing in time. Given perdurance, the objects
standing in these relations may be taken to be instantaneous stages, and the temporal relations will include
the binary relations of simultaneity, being temporally before, being one minute before, being two minutes before,
and so on. What object stages exist and how they stand in these relations determines the totality of temporal
facts. Moreover, facts about intrinsic change involve the instantiation of monadic properties by these object
stages. For example, when one of my object stages has the property sitting, and another has the property

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

standing, and the second stage is after the rst, then I have changed from sitting to standing.

This simple picture is made possible by the assumption of perdurance. Once that assumption is dropped, the
statement of temporal relationalism gets much more complicated.

For one thing, the temporal relations from which all temporal facts are supposed to ow can no longer be
the same binary relations used by the perdurantist. If two objects to be temporally compared are both
instantaneous, then either they are simultaneous, or one is some de nite temporal distance before the other
(I continue to ignore special relativity). But now the objects to be compared exist throughout stretches of
time; moreover, these stretches of time may overlap. Suppose I create a statue and then destroy it; the
statue therefore starts existing after I do and stops existing before I do. What kinds of temporal relations
relate me and the statue? Nothing quite so simple as the binary relations of the perdurantist.

Another feature of the relationalist picture complicated by enduring objects is the account of change.
Change is no longer the instantiation of di erent monadic properties by di erent temporal parts. It is
common for defenders of endurance to account for change by indexing property instantiation to time in
some way (Sect. 6), but indexing makes best sense on a substantival picture of time. Otherwise what are
properties indexed to?

The defender of the endurance‐based relationalist theory of time should, it seems to me, pursue something
p. 115 like the following strategy. The strategy overcomes both obstacles at once: it allows the defender of
relationalism to make temporal comparisons between enduring things, and simultaneously yields an
account of change that does not require indexing to substantival times. The endurantist relationalist should
say that all facts about persisting objects are captured by statements of the form: ⌈x is F n units of time after
y is G⌉, which is o cially primitive but explained informally as meaning that at some pair of times n units
41
apart, x is F at the earlier time and y is G at the later. Figure 4.3 gives the idea, indulging in the ction of
substantivalism for clarity of presentation. We have two objects: x, which exists between times t 0 and t 3
and is F throughout, and y, which exists between times t 1 and t 2 and is G throughout. In this situation the
following facts hold: x is F one unit after x is F (informally, this holds in virtue of t 0 and t 1, for example, as
well as t 1 and t 2, and t 2 and t 3); x is F two units after x is F (in virtue of t 0 and t 2, or t 1 and t 3); x is F three
units after x is F; y is G one unit after y is G; y is G one unit after x is F (in virtue of e.g. x at t0 and y at t1); y is G
two units after x is F; x is F one unit after y is G (in virtue of e.g. x at t2 and y at t1); and so on; it is not the case
that x is F four units after x is F; and so on.
 Fig. 4.3.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

Relational predication

p. 116 On this conception of property instantiation, change is easily described. An object x's being F for one unit of
time followed by its being G for one unit of time, for example, would be (partially) characterized by the
following facts: x is F (0.1, 0.2, . . . 1) units of time after x is F; x is G (0.1, 0.2, . . . 1) units of time after x is G; x is G
(0.1, 0.2, . . . 2) units of time after x is F.

Moreover, temporal relations between objects are given by the totality of relative property ascriptions
involving those objects. On some views, a property can be had by an object only at times at which it located;
the totality of relative property ascriptions thus speci es the relative temporal locations of continuants. But
even if some properties can be instantiated at times at which an object does not exist (perhaps the property
being famous), others are existence‐entailing (for example, the property of having 50g mass). Facts about the
relative instantiation of these existence‐entailing properties would then x the temporal locations of
objects.

This is an apparently consistent view. However, it seems to me to have two unattractive consequences. First,
it collapses certain possibilities that seem, pre‐analytically, to be distinct. Imagine, for example, worlds
consisting of multiple time lines laid end to end, which may be intuitively described as follows: (1) each time
line is in nite; (2) each moment of each time line is temporally after each moment of all the earlier time
lines; (3) no two moments from di erent time lines are any nite temporal distance from each other; and
(4) each time line contains the same objects, with the same intrinsic character, as the rest. One would have
thought that di erent worlds of this sort containing di erent numbers of distinct time lines are possible,
but the endurantist‐relationalist view does not allow this, for whether the world has one, two, or twenty
such time lines, the endurantist‐relationalist's description is the same. Similarly, the view cannot
distinguish a world with an in nite linear time line containing only a single unchanging electron from an
otherwise similar world in which time is circular. Call the electron e, and let C be the property having unit
negative charge. Presumably, in the circular world, for any n, e is C n units after e is C, since one can begin at
any point and traverse the circle repeatedly until n units have elapsed. Since it is also true in the in nite
linear world that e is C n units after e is C, for all n, the endurantist‐relationalist cannot distinguish these
p. 117 possibilities. The endurantist might deny that in the circular world, for any n, e is C n units after e is C, and
hold instead that e is C n units after e is C only if n is less than or equal to the temporal circumference of the
circle. But then the circular world cannot be distinguished from an otherwise similar world with a nite time
line of temporal distance n. All of these possibilities can be distinguished if one gives up either endurance or
relationalism, by appealing to the di ering pattern of temporal relations among temporal parts or
42
spacetime points.

The possibilities envisioned in the preceding objection are admittedly distant from common opinion; the
second objection must therefore bear a good deal of weight. It is this: the endurantist‐relationalist view
 makes all predication objectionably relational. Recall Lewis's objection to turning temporary intrinsic
properties into relations to times. As noted, one could perhaps stomach indexing if the indices were times,
but now objects have shapes relative to other objects. What is more, my having a given shape is relativized to
every other object in the universe. Very little remains of the intuition that an object can be just plain
straight.

This last objection could be avoided by altering the fundamental predicational form from ⌈x is F n units after
y is G⌉ to ⌈x is F n units after x is G⌉. This fundamental form only relates an object's properties at a time to its
own properties at later times. The proposal amounts to turning every pair of ordinary property attributions
to a single thing into an attribution of a single property. These properties are examples of what Josh Parsons

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

(2000) calls ‘distributional’ properties. Examples include being hot ten minutes after one is cold, being straight
one second after one is bent, and so on. The advantage of this approach is that predication is no longer relative
to other objects. One still cannot say that anything is just plain straight, so a hard‐line defender of the
problem of temporary intrinsics (Sect. 6) may remain unsatis ed. But at least being straight is no longer a
relation to other objects or times.

Just as there are temporary intrinsic properties, there are also temporary intrinsic relations. The present
theory must account for these as well. Presumably it will do so using sentences of the form: ⌈x bears R 1 to y
n units of time after x bears R 2 to y⌉. Such predications bring in nothing other than x and y, and so appear to
preserve the intrinsicality of R 1 and R 2.

p. 118 However valuable this progress on the temporary intrinsics front may be, the proposed account is
unacceptable as a version of relationalism. (Not that Parsons thinks otherwise. He only introduces
distributional properties to respond to the problem of temporary intrinsics, not in an attempt to develop a
version of spacetime relationalism.) Suppose x is at some time hot and y is also at some time hot. Given only
sentences like ‘x is hot n units after x is hot’ and ‘y is hot m units after y is hot’, one cannot specify which of x
or y is hot rst, or whether they are hot at the same time. The whole point in moving to the new view was to
remove the objectionably relational aspect of the original theory, which made all predication to objects
relative to predications to other objects. But by ruling out relational predication the new theory loses the
ability to capture certain temporal facts.

The defender of the new theory might try a divide‐and‐conquer strategy. First, account for intrinsic change
using the proposed locution for property instantiation (⌈x is F n units after x is G⌉). Secondly, ‘temporally line
up’ objects by introducing a new predicate ‘x begins to exist n units after y begins to exist’. The hope is that
once we specify the temporal interval between the beginnings of x's and y's existences, and further specify
all the distributional properties of x and y, then the temporal intervals between x's and y's instantiating
various properties will thereby be xed. But this works only in the special case where x and y have temporal
beginnings. Suppose two things have always existed (and, for good measure, always will exist). If each is hot
at some point in time, the current theory still cannot say which is hot rst, or whether they are hot
simultaneously. The divide‐and‐conquer strategy fails.

My argument has been against relationalists that make use only of enduring objects. It does not apply to a
relationalist who accepts enduring and perduring entities. The three‐dimensionalist dualist mentioned in
Section 7, for example, might utilize perduring events in her relationalism while holding that ordinary
continuants endure. But dualism faces problems of its own. As noted in Section 7, the dualist has trouble
explaining the relationship between the properties of things and the properties of events. Moreover, the
dualist is subject to the rst horn of the dilemma of the present section. An ontology that contains a (sui
generis) history for each enduring continuant seems to contain a redundant ontological category: why not
reduce the enduring continuants to the events?
p. 119 Before concluding this section a loose end must be tied. The ‘arguments from exotica’ from the previous
section assumed that, given certain philosophies of time, change must be accounted for either by
postulating temporal parts or by indexing property instantiation to times. The relationalist theory of
endurance discussed in this section now provides a third way of accounting for change. Objections have been
given to this theory, but set them aside. How do the arguments from exotica fare against the combination of
endurance and spacetime relationalism?

The argument from timeless worlds carries over with a slight modi cation. I can no longer characterize a
‘timeless world’ as one in which no times exist, since the relationalist says there are no such entities as
times even in a temporal world like ours. A timeless world must be understood instead as one in which no

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

temporal relations are instantiated. Since the actual relations that capture property instantiation, namely
those of the form x is F n units of time after y is G, are temporal relations, no object in the timeless world
could be straight in the same way that actual straight objects are straight. It may be replied that whether an
object, x, is actual or timeless, it could be true that x is straight zero units of time after x is straight. But this
locution is still temporal, and hence inappropriate to a genuinely timeless world. Compare the claim that an
object x is straight zero units after another object, y, is straight. This requires that x's straightness be
simultaneous with y's straightness—a temporal comparison.

The argument from time travel carries over as well. The argument was that anyone who indexes properties
to times cannot distinguish what appear to be two distinct possibilities, one in which I am standing and my
former self is sitting, the other in which our roles are reversed. Neither can the relationalist‐endurantist
distinguish these worlds, since each is a world in which I am sitting zero units of time after I am standing. One
response considered and rejected was indexing property instantiation to space as well as time. The
analogous (and equally unpalatable) response for the relationalist‐endurantist would be to regard all
property instantiation as having the form x is F n units of time and m units of space from where y is G.

p. 120
9. The Argument from Vagueness43

The nal argument of this chapter is, I think, one of the most powerful. In outline, it runs as follows. Under
what conditions do objects come into and go out of existence? As a believer in temporal parts and
unrestricted composition, I say this always occurs. Any lled region of spacetime is the total career of some
object. Others say that objects come into and go out of existence only under certain conditions. When bits of
matter are arranged in certain ways, an object—say, a person—comes into existence; and that thing goes
out of existence when the bits cease to be arranged in the appropriate way. But what sorts of arrangements
are suitable? If one arrangement is suitable, then a very slightly di erent arrangement would seem to be as
well. Iterate this procedure, and we have the conclusion that objects always come into and go out of
existence, no matter how bits of matter are arranged. But this, it will be seen, is tantamount to admitting
that four‐dimensionalism is true. The obvious problem with ‘slippery slope’ arguments of this sort is that
they neglect vagueness. But there cannot be vagueness of the sort needed to block the argument. The
argument will at this point make some assumptions, most notably that vagueness never results from ‘logic’
(i.e. from boolean connectives, quanti cation, or identity). These assumptions could be coherently denied,
but they are very plausible. I also suspect that they are widely held, even among those hostile to temporal
parts. There is, therefore, considerable interest in showing that anyone who accepts the assumptions must
accept four‐dimensionalism.

Those familiar with David Lewis's On the Plurality of Worlds (1986a) will recognize a parallel here with his
argument for the principle of unrestricted mereological composition. In the next subsection I develop
Lewis's argument for unrestricted composition in my own way, and then in the following two subsections
show how it can be modi ed to yield an argument for temporal parts.
p. 121 9.1. Unrestricted Mereological Composition
Here is Lewis's argument (1986a: 212–13):

We are happy enough with mereological sums of things that contrast with their surroundings more
than they do with one another; and that are adjacent, stick together, and act jointly. We are more
reluctant to a rm the existence of mereological sums of things that are disparate and scattered
and go their separate ways . . .

The trouble with restricted composition is as follows . . . To restrict composition in accordance with
our intuitions would require a vague restriction. But if composition obeys a vague restriction, then

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

it must sometimes be a vague matter whether composition takes place or not. And that is
impossible.

The only intelligible account of vagueness locates it in our thought and language. The reason it's
vague where the outback begins is not that there's this thing, the outback, with imprecise borders;
rather there are many things, with di erent borders, and nobody has been fool enough to try to
enforce a choice of one of them as the o cial referent of the word ‘outback’. Vagueness is semantic
indecision. But not all of language is vague. The truth‐functional connectives aren't, for instance.
Nor are the words for identity and di erence, and for the partial identity of overlap. Nor are the
idioms of quanti cation, so long as they are unrestricted. How could any of these be vague? What
would be the alternatives between which we haven't chosen?

The question whether composition takes place in a given case, whether a given class does or does
not have a mereological sum, can be stated in a part of language where nothing is vague. Therefore
it cannot have a vague answer. . . . No restriction on composition can be vague. But unless it is
vague, it cannot t the intuitive desiderata. So no restriction on composition can serve the
intuitions that motivate it. So restriction would be gratuitous. Composition is unrestricted . . .

It may be summarized as follows. (I follow Lewis in speaking of parthood atemporally; I consider temporally
relativized parthood in the next subsection.) If not every class has a fusion then there must be a restriction
on composition. Moreover, the only plausible restrictions on composition would be vague ones. But there
can be no vague restrictions on composition, because that would mean that whether composition occurs is
sometimes vague. Therefore, every class has a fusion.

There is a weakness in this argument. The rst premise of my summary is that if not every class has a fusion
p. 122 then there must exist a ‘restriction on composition’. On a natural reading, a ‘restriction on composition’
is a way of lling in the blank in the following schema:

A class, S, has a fusion if and only if—

such that what goes into the blank is not universally satis ed. That is, a restriction on composition would be
44
an answer to Peter van Inwagen's ‘special composition question’. (For example, one answer might be that
a class has a fusion i its members are ‘in contact’.) But perhaps the special composition question has no
informative answer because whether composition takes place in a given case is a ‘brute fact’ incapable of
45
informative analysis.

There are two senses in which composition might be brute. Composition is brute in a strong sense if it does
not even supervene on causal and qualitative factors. This is extremely implausible; one would need to
admit a pair of cases exactly alike in terms of causal integration, qualitative homogeneity, and so on, but
which di er over whether objects have a sum. But even if supervenience is admitted, composition might be
brute in the weaker sense that there is no natural, nite, humanly statable restriction on composition. Since
 I do not wish to reject weak brute composition out of hand, I will reformulate Lewis's argument so as not to
(directly, anyway) presuppose its falsity.

Let us understand a ‘case of composition’ (‘case’, for short) as a possible situation involving a class of
objects having certain properties and standing in certain relations. We will ask with respect to various cases
whether composition occurs; that is, whether the class in the case would have a fusion. In summary, my
version of Lewis's argument runs as follows. If not every class has a fusion, then we can consider two
possible cases, one in which composition occurs and another in which it does not, which are connected by a
‘continuous series of cases’ selected from di erent possible worlds, each extremely similar to the last. Since
composition can never be vague, there must be a sharp cut‐o in this series where composition abruptly

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

stops occurring. But that is implausible. So composition always occurs.

More carefully. First, consider any case, C 1, in which composition occurs—the case of a certain class of
p. 123 subatomic particles that are part of my body, for example. Now consider a second case, C 2, which occurs
after I die and am cremated, in which my molecules are scattered across the Milky Way. Some would say that
in C 2, composition fails to take place: there is nothing that is made up of these scattered, causally
unconnected particles. Next, let us further imagine a nite series of cases connecting C 1 and C 2, in which
each case in the series is extremely similar to its immediately adjacent cases in all respects that might be
relevant to whether composition occurs: qualitative homogeneity, spatial proximity, unity of action,
comprehensiveness of causal relations, etc. I call such a series a ‘continuous series connecting cases C 1 and
C 2’.

My argument's rst premise is:

P1: If not every class has a fusion, then there must be a pair of cases connected by a continuous series
such that in one, composition occurs, but in the other, composition does not occur.

I can think of only two objections. Given ‘nihilism’, the view that composition never occurs (i.e. that there
are no composite objects): since there are no cases of composition at all there cannot exist a continuous
46
series connecting a case of composition to anything. I argue against nihilism in Chapter 5, Section 6. The
second objection is that not every pair of cases can be connected by a continuous series. No continuous
47
series connects any case with nitely many objects to a case with in nitely many objects, for example.
However, P1 only requires that some pair of cases di ering over composition be connected by a continuous
series, if composition is restricted. No one will want to claim that the jump from nitude to in nity is the
thing that makes the di erence between composition and its absence. So even if no continuous series
connects C 1 with C 2, one can choose another pair of cases C 1′ and C 2′, like C 1 and C 2 with respect to
whether composition occurs, which are connected by a continuous series.

By a ‘sharp cut‐o ’ in a continuous series I mean a pair of adjacent cases in a continuous series such that in
one, composition de nitely occurs, but in the other, composition de nitely does not occur. Surely there are
p. 124 no such things:

P2: In no continuous series is there a sharp cut‐o in whether composition occurs.

Adjacent members in a continuous series are extremely similar in certain respects. By including more and
more members in a continuous series, adjacent members can be made as similar as you like. Accepting a
sharp cut‐o is thus nearly as di cult as rejecting the supervenience of composition on the relevant factors.
It would involve saying, for example, that although certain particles de nitely compose a larger object, if
one of the particles had been 0.0000001 nanometers displaced, those particles would have de nitely failed
to compose any object at all. Of course, sharp cut‐o s in the application of a predicate are not always
implausible—consider the predicate ‘are separated by exactly 3 nanometers’. What I object to is a sharp cut‐
o in a continuous series of cases of composition.
 To postulate such a sharp cut‐o would be to admit that the realm of the macroscopic is in some sense
‘autonomous’ of the microscopic. By ‘autonomous’ I do not mean ‘non‐supervenient’, since accepting a
sharp cut‐o in a continuous series of cases of composition does not threaten supervenience. Rather, I mean
that there would seem to be something ‘metaphysically arbitrary’ about a sharp cut‐o in a continuous
series of cases of composition. Why is the cut‐o here, rather than there? Granted, everyone must admit
some metaphysically ‘brute’ facts, and it is a hard question why one brute fact seems more or less plausible
48
than another. Nevertheless, this brute fact seems particularly hard to stomach.

A possible objection to P2 would be based on precisely statable topological restrictions on the regions of
space that can possibly be occupied by a composite object, perhaps regions in which any two points are

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

connectable by some continuous path con ned to the region. But this would rule out too many objects:
galaxies, solar systems, and so on. More importantly, under the classical physics conception of matter, all
macroscopic objects are discontinuous. This is less clear on a quantum‐mechanical picture, but a classical
49
world should not turn out devoid of macro‐objects.

p. 125 The nal premise of the argument is, I think, the most controversial:

P3: In any case of composition, either composition de nitely occurs, or composition de nitely does not
occur.

P1, P2, and P3 imply the desired conclusion. P1 requires that if composition is not unrestricted, we have a
case of composition connected by a continuous series to a case of non‐composition. By P3, there must be a
sharp cut‐o in this series where composition abruptly ceases to occur; but this contradicts P2. It must be
emphasized that this is not ‘just another Sorites’. The correct solution to traditional Sorites paradoxes will
surely involve a region in which the relevant predicate (‘is a heap’, ‘is bald’, and so on) neither de nitely
applies nor de nitely fails to apply (note: the epistemic theory of vagueness is discussed below). But this is
just what P3 prohibits.

I turn now to the defense of P3. Recall that a ‘case’ was de ned as involving a class of objects, by which I
mean a non‐fuzzy class. Thus understood, classes have precisely de ned membership, and so must be
distinguished from class descriptions, which may well be imprecise. P3 pertains to classes themselves, not
their descriptions. Thus, indeterminacy of truth value in the sentence ‘The class of molecules in the
immediate vicinity of my body has a fusion’ would not be inconsistent with P3. In virtue of its vagueness,
the subject term of this sentence may well fail to refer uniquely to any one class. Also note that P3 is not
concerned with the nature of the resulting fusion, only its existence. It may well be indeterminate whether a
given class of molecules has a fusion that counts as a person. This is not inconsistent with P3, for the class
may de nitely have a fusion which is a borderline case of a person.

50
Lewis's method for establishing P3 appeals to the ‘linguistic theory of vagueness’. This view's slogan is
that ‘vagueness is semantic indecision’. Whenever a sentence is indeterminate in truth value due to
vagueness, this is because there is some term in the sentence that is semantically vague, in that there are
multiple possible meanings for that term, often called ‘precisi cations’, no one of which has been singled
out as the term's unique meaning. There is no vagueness ‘in the world’; all vagueness is due to semantic
indecision. An oversimpli ed example: ‘bald’ is vague because no one has ever decided which of its
p. 126 precisi cations it is to mean, where the precisi cations are properties of the form having no more than n
hairs on one's head, for various integers n in a certain range. (Realistically, baldness depends on more than
how many hairs one has on one's head. Distribution, length, and other factors also matter.)

In virtue of the de nition of ‘fusion’ in terms of parthood, we can formulate the assertion that a given class,
C, has a fusion as follows:

(F) There is some object, x, such that (1) every member of C is part of x, and (2) every part of x shares a
 part in common with some member of C.

If (F) has no determinate truth value relative to some assignment to ‘C’, this must be due to vagueness, for
other potential sources of truth value gaps (such as ambiguity or failed presupposition) are not present.
Given the linguistic theory of vagueness, one of the terms in (F) would need to have multiple
precisi cations. But it is di cult to see what the precisi cations of logical terms, or the predicates ‘is a
member of’ and ‘part of’, might be.

The weakest link here is the rejection of precisi cations for ‘is part of’. Notice that in ruling out ‘part of’ as a
source of vagueness, Lewis is not ruling out all vagueness in ascriptions of parthood, for ascriptions of
parthood may contain singular terms (e.g. ‘the outback’) with multiple precisi cations. (F), however,

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

apparently contains no vague singular terms. Vague ascriptions of parthood are therefore not a good reason
for saying that ‘part’ lacks precisi cations. But what is a good reason against? Lewis's reason is that it is
di cult to see what the precisi cations might be. But perhaps this is because they are not easily statable in
natural language. Some non‐logical terms, ‘is bald’ for example, have prima facie easily statable
precisi cations (namely, properties expressed by predicates of the form ⌈has a head with less than n hairs⌉,
and even then statability is in doubt, given the vagueness of ‘head’ and ‘hair’). But other vague predicates
lack easily statable precisi cations, for example ‘person’, ‘table’, and artifact terms generally. Someone
might argue that ‘part’ has precisi cations corresponding to precisi cations of answers to van Inwagen's
special composition question. I do not say that this response to Lewis can be made to work, but I cannot see
how to show that it cannot be made to work.

Fortunately, P3 may be supported without making any assumptions about parthood, for if it were vague
p. 127 whether a certain class had a fusion then it would be vague how many concrete objects exist. Lewis's
assumptions about vagueness can then be replaced by weaker assumptions that concern only logical
vocabulary.

Let us stipulatively de ne concrete objects as those which do not t into any of the kinds on the following
list:

sets and classes

numbers

properties and relations

universals and tropes

possible worlds and situations

If I have missed any ‘abstract’ entities you believe in, feel free to update the list. Suppose now for reductio
that P3 is false—that is, that it can be vague whether a given class has a fusion. In such a case, imagine
counting all the concrete objects in the world. One would need to include all the objects in the class in
question, but it would be indeterminate whether to include another entity: the fusion of the class. Now
surely if P3 can be violated, then it could be violated in a ‘ nite’ world, a world with only nitely many
concrete objects. That would mean that some numerical sentence—a sentence asserting that there are
exactly n concrete objects, for some nite n—would be indeterminate. But numerical sentences need
contain only logical terms and the predicate ‘C’ for concreteness (a numerical sentence for n = 2 is: ∃ x ∃ y
[Cx & Cy & x≠ y & ∀ z (Cz → [x = z ∨ y = z])]). Mereological terms are not needed to express numerical
sentences, and so need not be assumed to lack precisi cations.

To support P3, then, I must argue that numerical sentences can never be indeterminate in truth value. First,
note that numerical sentences clearly have no syntactic ambiguity. Secondly, note that the concreteness
 predicate, ‘C’, presumably has precise application conditions since it was de ned by a list of predicates for
fundamental ontological kinds that do not admit of borderline cases. And even if one of the members of the
list is ill de ned or vague in some way, the vagueness is presumably of a kind not relevant to my argument:
any way of eliminating the vagueness would su ce for present purposes. So if any numerical sentence is to
be indeterminate in truth value, it must be because one of the logical notions is vague.

p. 128 Accordingly, the argument's crucial assumption about vagueness is that logical words are never a source of
51
vagueness. Any sentence containing only logical expressions, plus perhaps predicates with determinate
application conditions (such as ‘is concrete’), must be either de nitely true or de nitely false. This premise
is extremely compelling. Logical concepts are paradigm cases of precision. At the very least, in no case is

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

there evident indeterminacy as with ‘bald’ and ‘heap’.

I am inclined to regard my assumption that logic does not generate vagueness as owing from two further
theses, rst, the linguistic theory of vagueness outlined above, and secondly, the assumption that logical
terms lack multiple precisi cations. The rst thesis I simply assume; the second I am about to argue for. But
all the argument for four‐dimensionalism requires is that the assumption holds, whether or not it is
justi ed by these two theses.

It is overwhelmingly plausible that the boolean operators lack precisi cations. That leaves the quanti ers
and the identity sign.

Might an unrestricted quanti er have precisi cations? It is important to be clear that there is no problem at
all with restricted quanti ers having precisi cations. The restricted quanti er ‘all persons’ will clearly have
precisi cations because the restricting predicate ‘is a person’ has precisi cations. But this is irrelevant.

If predicates can have precisi cations, why not unrestricted quanti ers? The asymmetry is due to the fact
that predicates have subclasses of the universal domain of all things as their extensions, and the universal
domain has many subclasses. But there seems to be only one ‘everything’ for the restricted quanti er to
range over.

This can be turned into an argument against the possibility of multiple precisi cations for the unrestricted
quanti er. Imagine there were two expressions, ∀1 and ∀2, which allegedly expressed precisi cations of the
unrestricted quanti er. ∀1 and ∀2 will need to di er in extension if they are to make any di erence to the
kinds of sentences under consideration in this section; merely intensional di erence will not do. Thus, there
must be some thing, x, that is in the extension of one, but not the other, of ∀1 and ∀2. But in that case,
p. 129 whichever of ∀1 and ∀2 lacks x in its extension will fail to be an acceptable precisi cation of the
unrestricted quanti er. It quite clearly is a restricted quanti er since there is something—x—that fails to be
in its extension.

This argument is directed only at those who share my assumption of the linguistic theory of vagueness.
Those who believe that objecthood itself is somehow vague might resist the step where I concluded that
some object is in the extension of one but not the other of ∀1 and ∀2 from the fact that ∀1 and ∀2 di er
extensionally. Suppose, for example, that reality de nitely contains objects a and b, but that it is
indeterminate whether reality contains a third object, c. Somehow it is indeterminate whether c exists,
where this is not due to semantic indeterminacy of any kind. The believer in vague objects might then claim
that if ∀1 ranges over a and b, and ∀2 over a, b, and c, then even though ∀1 and ∀2 di er extensionally, it is at
least not de nitely the case that there is something in ∀2 but not ∀1, for it is not de nitely the case that c
exists. I mention this position only to set it aside; as I said above, I simply assume that this theory of
vagueness is not correct.

Someone who shares the linguistic theory of vagueness might still attempt to resist this step of the
argument. Its conclusion is that there is an object, x, in the extension of one of ∀1 and ∀2. But this ‘there is’, it
 might be claimed, must be one or another precisi cation of the unrestricted quanti er. In particular, it will
need to be one that includes x. But then it might be argued that it is illegitimate to conclude that ∀2 is not an
acceptable precisi cation for the unrestricted quanti er just because under one of the other precisi cations
there is an object that is not in ∀2's extension. There is no Archimedean point from which to quantify, and
say that ∀2 is a restricted quanti er. All we have are many precisi cations, each of which is complete by its
own lights.

It is hard to understand what these precisi cations are supposed to be. As mentioned above, precisi cations
for predicates are unproblematic, and should be acceptable to all since everyone admits the existence of the
extensions of those precisi cations. But what are these precisi cations of ‘everything’? It might be claimed

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

that existence and objecthood are somehow relative to ‘conceptual scheme’, and that the precisi cations for
the quanti ers correspond to di erent conceptual schemes, or precisi cations of conceptual schemes. This
comes in tame and wild variants. The tame claim that di erent conceptual schemes involve di erent
restrictions on quanti ers, is surely correct, but is irrelevant, concerning as it does merely restricted
p. 130 quanti cation. The wild claim is that all quanti cation, no matter how unrestricted, is relative to
conceptual scheme. A claim so wild that I will not consider it is that the world is the way it is because we talk
in a certain way. A still wild but nevertheless worthy‐of‐discussion claim is that the world is in a certain way
independent of us, but there is no once‐and‐for‐all correct description of it in terms of quanti ers and
variables. Any description using such a ‘thing‐language’ presupposes some division of reality into things,
but this division may be done in various ways depending on one's conceptual scheme or linguistic
framework. Various ways of doing this division count as precisi cations of the concept of an object, and
hence correspond to precisi cations of the quanti ers. I reject this conception of existence for the reasons
explained in the introduction. This book presupposes that existence of things is univocal, not relative to
conceptual schemes or linguistic frameworks.

So I reject the idea that unrestricted quanti ers have precisi cations. Might the identity predicate have
precisi cations? On the face of it, the answer is no: the nature of identity seems conceptually simple and
clear. Identity sentences can clearly have vague truth conditions when they have singular terms that are
indeterminate in reference: ‘Michael Jordan is identical to the greatest basketball player of all time’, for
example. But the only singular terms in numerical sentences are variables relative to assignments, which
are not indeterminate in reference. There are those who say that even without indeterminate singular
terms, and even without precisi cations for the identity predicate, identity ascriptions can be vague in truth
52
value, despite the Evans (1978)/Salmon (1981) argument to the contrary. I nd this doctrine obscure but
have nothing to add to the extensive literature on this topic; here I must presuppose it false.

In summary, then, the argument for P3 has been as follows. If it could be vague whether composition
occurs, this could happen in a nite world; some numerical sentence would then be indeterminate in truth
value. But aside from the predicate ‘concrete’, which is non‐vague, numerical sentences contain only logical
vocabulary, and logical vocabulary, I say, can never be a source of vagueness.

A loose end must be tied before proceeding to the parallel argument for temporal parts. Defenders of
p. 131 epistemicism claim that vagueness never results in indeterminacy of truth value. Imagine removing the
hairs from a man, one at a time. According to the epistemicist there will be a single hair whose removal
results in the man becoming bald. Even though no one could ever know where it lies, this sharp cut‐o for
53
the predicate ‘bald’ exists. Since epistemicists are already accustomed to accepting sharp cut‐o s for
predicates like ‘heap’ and ‘bald’, one might think they would also be happy with a sharp cut‐o in a
continuous series of cases of composition, thus rejecting P2. The epistemic theory seems to me incredible
despite its current popularity, but set this aside. I will argue that even an epistemicist should accept P2.

As explained in the introduction, I assume a ‘best‐candidate’ theory of reference and meaning, according to
which meaning is determined jointly by ‘use and intrinsic eligibility’. Recall the precisi cations for the
 predicate ‘bald’: having no more than n hairs on one's head, for various positive integers n. Most of us think
there is no fact of the matter as to which of these candidate properties is meant by ‘bald’. The reason, I
think, is twofold. First, the candidates appear equally intrinsically eligible. Secondly, it also appears that use
does not distinguish between the candidates. Despite this, the epistemicist says that ‘bald’ means exactly
one of them. If the epistemicist accepts the best‐candidate theory of meaning (and I think he should), he
must therefore say either (1) one candidate is more intrinsically eligible, carves nature at the joints better
than the rest, thus granting it metaphysical privilege, or (2) one candidate ts use better than the rest, thus
granting it semantic privilege.

54
The epistemicist should surely prefer option (2). Somehow, something about our meaning‐determining

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

behavior singles out one of the many candidate properties to be the meaning of ‘bald’. The epistemicist is
therefore committed to the existence of bridge laws from use to meaning that are more ne‐grained than
one might have expected, but at least he avoids the highly implausible metaphysics of option (1).
Epistemicism per se should not lead us to revisionary metaphysics.

So the epistemicist's sharp cut‐o s would not be ‘metaphysical’. Instead of corresponding to unexpected
joints in reality, they would represent unanticipated powers of humans to draw metaphysically arbitrary
lines. They cannot, therefore, be used to give a plausible objection to premise P2 in my argument. Premise
p. 132 P2 says that there are no sharp cut‐o s in a continuous series of cases of composition. A sharp cut‐o in
whether composition occurs would (in a nite world) result in a sharp cut‐o in the number of objects, and
thus in the truth value of a sentence stated solely in terms of logical terms and the predicate for
concreteness. But, as I have argued, there are no multiple candidates to be meant by these terms. So the
epistemicist's explanation of sharp cut‐o s due to vagueness—as being the result of use selecting among
equally eligible candidate meanings—is unavailable in this case.

If, despite this, the epistemicist were to persist in believing in a sharp cut‐o , he would need to revert to the
metaphysical explanation. The sharp cut‐o would represent a ‘logical joint’ in reality: on the one and only
candidate set of meanings for the logical terms (and ‘concrete’), at some point on a continuous series of
cases of composition there is an abrupt shift in the truth value of a numerical sentence. This sharp cut‐o
would be starkly metaphysical. As I have claimed, it is very hard to believe in this sort of cut‐o —it feels
‘metaphysically arbitrary’. Moreover, at this point epistemicism is no longer playing a role in the objection
to P2, for even a non‐epistemicist will admit that if the metaphysical cut‐o exists, then there is a sharp shift
in the truth value of a numerical sentence. The objection to P2, therefore, is not aided by epistemicism, but
rather rests on its own metaphysical credentials, which are unimpressive.

The argument for restricted composition, we have seen, leans most heavily on P3, which in turn rests on the
view that logic, and in particular unrestricted quanti cation and identity, are non‐vague. This view is
attractive, and I have said some things in its defense, but I doubt I have said enough to convince a
determined opponent. My argument for unrestricted composition, therefore, should be taken as showing
that anyone who accepts that logic is non‐vague must also accept unrestricted composition. In the next two
subsections I show that everyone who shares this assumption about vagueness must also accept four‐
dimensionalism.

9.2. Composition Questions and Temporally Indexed Parthood

The argument of the previous section concerned the question of when a given class has a fusion, where
p. 133 ‘fusion’ was understood atemporally. To avoid begging any questions against my opponents, the argument
for temporal parts will be stated using temporally quali ed mereological terms (see Ch. 3, Sect. 3). When
the relation being a fusion of is indexed to times, various questions of composition must then be
55
distinguished.
 The simplest question is that of when a given class has a fusion at a given time. But we are also interested in
‘diachronic’, or ‘cross‐time’ fusions: things that are fusions of di erent classes at di erent times. These are
objects that gain and lose parts. One concept of cross‐time summation may be introduced as follows. Call an
‘assignment’ any (possibly partial) function that takes one or more times as arguments and assigns non‐
empty classes of objects that exist at those times as values; and let us say that an object x is a diachronic
fusion (‘D‐fusion’, for short) of an assignment f i for every t in f's domain, x is a fusion‐at‐t of f(t). For
example, consider two times at which I exist, and let f be a function with just those two times in its domain
that assigns to each the class of subatomic particles that are part of me then. I am a D‐fusion of f, since at
each of the two times I am a fusion of the corresponding class of subatomic particles.

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

A second question of composition, then, is the question of when a given assignment has a D‐fusion: given
various times and various objects corresponding to each, under what conditions will there be some object
that at the various times is composed by the corresponding objects? A third question would be that of the
conditions under which there would be such an object that existed only at the speci ed times. This is the
question of when a given assignment has a minimal D‐fusion—a D‐fusion of the assignment that exists only
at times in the assignment's domain. I am not a minimal D‐fusion of the assignment f mentioned above
because I exist at times other than the two times in the domain of f. To get an assignment of which I am a
minimal D‐fusion, extend f to assign to any other time at which I exist the class of subatomic particles that
are part of me then.

In an intuitive sense, a minimal D‐fusion of some objects at various times consists of those objects at those
times and nothing more. Though it required some machinery to state, the question of which assignments
have minimal D‐fusions is far from being remote and technical. Indeed, we can restate this question in the
p. 134 following woolly yet satisfying fashion: under what conditions do objects begin and cease to exist? Suppose
we make a model of a park bench from three toy blocks, b1, b2, and b3, by placing one on top of two of the
others at time t1; a few minutes later at t2 we separate the blocks. Is there something that we brought into
existence at the rst time and destroyed at the second? This is the question of whether a certain assignment
has a minimal D‐fusion—namely, the assignment that assigns the class {b1, b2, b3} to every time between t1
and t2.

9.3. The Argument from Vagueness for Four‐Dimensionalism

Under what conditions does a given assignment have a minimal D‐fusion? I say that all assignments have
minimal D‐fusions; my argument is parallel to the argument for unrestricted composition. Restricting when
minimal D‐fusions exist would require a cut‐o in some continuous series of pairwise similar cases. Just as
composition can never be vague, neither can minimal D‐fusion. So the cut‐o would need to be abrupt,
which is implausible:

P1′: If not every assignment has a minimal D‐fusion, then there must be a pair of cases connected by a
‘continuous series’ such that in one, minimal D‐fusion occurs, but in the other, minimal D‐fusion
does not occur.

P2′: In no continuous series is there a sharp cut‐o in whether minimal D‐fusion occurs.

P3′: In any case of minimal D‐fusion, either minimal D‐fusion de nitely occurs, or minimal D‐fusion
de nitely does not occur.

The notion of a ‘case’ must be adjusted in the obvious way. A ‘continuous series of cases’ will now vary in all
respects thought to be relevant to whether a given assignment has a minimal D‐fusion, including spatial
adjacency, qualitative similarity, and causal relations at the various times in the assignment, as well as the
beginning and cessation of these factors at various times of the assignment.
 The justi cation of premise P1′ is like that for P1. Like P1, P1′ can be resisted by a nihilist, who rejects the
existence of all composites. For the nihilist, only mereological simples exist; there are no composite objects.
The only cases of minimal D‐fusion concern the entire lifetime of a single particle; such cases cannot be
56
p. 135 connected continuously with cases in which minimal D‐fusion does not take place. Arguments against
nihilism must wait until Chapter 5, Section 6. As for P2′, an abrupt cut‐o in a continuous series of cases of
minimal D‐fusion—a pair of cases extremely similar in spatial adjacency, causal relations, and so on, but
de nitely di ering in whether minimal D‐fusion occurs—seems initially implausible. There is, however, a
three‐dimensionalist ontology that would secure such a cut‐o : a version of mereological essentialism
according to which, intuitively, nothing exists but mereological sums, which have their parts permanently,
and exist as long as those parts exist. Minimal D‐fusions could be restricted non‐vaguely: an assignment has

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

a minimal D‐fusion, roughly, when and only when it is the temporally longest assignment for a given xed
class of objects. The idea is that mereological fusions of objects ‘automatically’ come into existence when
their parts do, automatically retain those same parts, and automatically go out of existence when any of
those parts go out of existence. (Less roughly: where S1 and S2 are sets of objects that exist at times t1 and t2,
respectively, say that pairs < t1, S1> and < t2, S2> are equivalent i every part‐at‐t1 of any member of S1
overlaps‐at‐t2 some member of S2, and every part‐at‐t2 of any member of S2 overlaps‐at‐t1 some member of
S1. The idea is that S1 and S2 contain, if not exactly the same members, at least the same stu , just divided up
di erently. The non‐vague restriction is that an assignment f has a minimal D‐fusion i f is a maximal
equivalence‐interrelated assignment; that is (construing f as a class of pairs), i (1) every two pairs in f are
equivalent, and (2) if a pair < t, S> is equivalent to some member of f, then some pair < t, S′> (i.e. some pair
with the same time) to which it is equivalent is a member of f.) My argument can therefore be resisted by
57
this sort of mereologist. Other arguments can be given against mereological essentialism; see Chapter 5,
Section 7.

p. 136 Just as topological restrictions on regions of space can provide precise restrictions on composition
(although I nd them unmotivated), topological restrictions on regions of time can provide precise
restrictions on minimal D‐fusion. Some may favor a restriction to continuous intervals (although cf.
Hirsch's 1982: 22 . example of a watch that is taken apart for repairs), or to sums of continuous intervals. I
regard each as unmotivated, but we need not quarrel. Given either restriction, the argument would still
establish a restricted version of four‐dimensionalism according to which there exist continuous temporal
segments of arbitrarily small duration. For most four‐dimensionalists that would be four‐dimensionalism
enough.

My argument for P3 was that if it is indeterminate whether composition occurs then it will be indeterminate
how many objects there are, which is impossible. I use a similar argument to establish P3′. Indeterminacy in
minimal D‐fusion might be claimed in several situations. But in each case, I will argue, at some possible
world there would result ‘count indeterminacy’—an indeterminacy in the nite number of concrete objects.
This was argued above to be impossible, assuming that logic is not a possible source of vagueness. (Recall
the distinction between existence‐at and quanti cation. Count indeterminacy is indeterminacy in how many
objects there are, not merely in how many objects exist at some speci ed time. It is the former that my
assumption about logic prohibits, and hence the former that I must argue would result from indeterminacy
in minimal D‐fusion.)

I distinguish four situations in which indeterminacy in minimal D‐fusion might be claimed:

(1) Indeterminacy in whether objects have a fusion at a given time, because (say) they are moderately
scattered at that time. This would result in count indeterminacy. For consider a possible world
containing some nite number of quarks that are greatly scattered at all times except for a single
time, t, at which they are moderately scattered. The quarks would then determinately lack a fusion
except at time t, when it would be indeterminate whether they have a fusion. The result is
indeterminacy in how many objects exist: there is one more object depending on whether the quarks
 have a fusion at t. (Similar remarks would apply if ‘scattered’ in this paragraph were replaced by
p. 137 various other predicates deemed relevant to the question of whether a class has a fusion at a given
time.)

(2) Indeterminacy in whether a fusion at t of certain particles is identical to a fusion at some other time,
t′, of some other particles. This, too, would result in count indeterminacy. Suppose I undergo
amnesia in such a way that we feel indeterminacy in whether ‘Young Man Ted is identical to Old Man
Sider’ is true. Presumably we will want to say the same thing about this case if it occurs in a world
with only nitely many concrete things. But in this world, if it is really indeterminate whether a
certain assignment has a minimal D‐fusion (say, one that assigns to times before and after amnesia

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

all my parts at those times), then there will result indeterminacy in the count of the concrete objects
there, for if the identity holds there will be one less object than if the identity does not hold.

(3) Indeterminacy in when an object begins to exist. Again, this would result in count indeterminacy.
Suppose that in some case, C, it is indeterminate when a certain statue comes into existence.
Consider next a case much like C, but in which (i) only nitely many concrete things exist, and (ii)
the molecules that would make up the statue are all annihilated after the time at which the statue is
alleged to indeterminately exist. Then it will be indeterminate whether the statue exists at all, and
hence indeterminate how many things there are in the world in question.

(4) Indeterminacy in when an object ceases existing. This case is similar to the previous case.

If, then, minimal D‐fusion could be indeterminate, it could be indeterminate what the ( nite) number of
concrete things is. But then there could be a numerical sentence that is neither de nitely true nor de nitely
false. Assuming that no indeterminacy can issue from logic, this is impossible. So P3′ is true: a given
assignment must either de nitely have or de nitely lack a minimal D‐fusion. This is not to say that the
phenomena adduced in (1) to (4) are not genuine; they simply must be understood in some way not implying
indeterminacy in minimal D‐fusion: (1) The indeterminacy is due to indeterminate restrictions on everyday
quanti cation. Typically, we do not quantify over all the objects that there are, only over fusions of objects
p. 138 that are not too scattered. If objects are borderline scattered they still de nitely have a fusion, but we
have a borderline resistance to admitting that fusion into an everyday domain of quanti cation. (2) This is a
case involving three objects. Object 1 begins around the time of my birth and ends at the amnesia, Object 2
begins at amnesia and lasts until my death, and Object 3 lasts throughout this time interval. The name
‘Young Man Ted’ is indeterminate in reference between Objects 1 and 3; the name ‘Old Man Sider’ is
indeterminate between Objects 2 and 3; hence the identity sentence is indeterminate in truth value. (3)
There are many objects involved di ering in when they begin to exist; the term ‘the statue’ is indeterminate
in reference among them; hence the sentence ‘The statue begins to exist at t’ will be indeterminate in truth
value for certain values of ‘t’. Case (4) is similar to (3).

P1′, P2′, and P3′ jointly imply:

(U) every assignment has a minimal D‐fusion.

But (U) is a powerful claim, for it entails four‐dimensionalism! The central four‐dimensionalist claim, recall
from Chapter 3, Section 2, is that every object, x, has a temporal part at every moment, t, at which it exists.
Let A be the assignment with only t in its domain that assigns {x} to t. (U) guarantees the existence of an
object, z, that is a minimal D‐fusion of A. It may now be shown that z is a temporal part of x at t. I do so by
showing that z satis es clauses (1) to (3) of the Chapter 3, Section 2 de nition of a temporal part (see p. 58
for principles and de nitions of temporal mereology):

(1) z is a fusion of {x} at t. It follows from the de nition of ‘fusion at t’ that every part of z at t overlaps x
at t; by (PO) (p. 58), z is part of x at t.
 (2) Since z is a minimal D‐fusion of this assignment, z exists at but only at t.

(3) Let y be any part of x at t. Since z is a fusion of {x} at t, x is part of z at t; thus, y is part of z at t; thus, z
overlaps y at t. So: z overlaps at t every part of x at t.

A few people have objected in conversation that the conclusion of the argument, (U), does not entail four‐
dimensionalism. A thing‐event dualist (Section 7), for example, might say that the objects guaranteed by (U)
are temporal parts of events, not of continuants. These events would spatially coincide with continuants,
p. 139 but the continuants would nevertheless endure. This objection is a mistake. Line (1) of the argument
shows that z is a part of x at t, not merely spatially coincident with x. So x could not be an enduring thing,
given (U)—at any moment it (and every other object as well) would have an instantaneous part that

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

overlaps all its parts.

An interesting feature of the argument from vagueness is that it forces one into taking an extreme position
in the philosophy of persistence. Take on board the claim that minimal D‐fusion can never be vague, and
reject as well the existence of a sharp cut‐o in a continuous series of cases of composition. What is out at
this point are moderate views, such as those of David Wiggins (Ch. 5, Sect. 3) and Michael Burke (Ch. 5, Sect.
4), who admit minimal D‐fusion in cases that match up with ordinary intuition, and even Peter van Inwagen
(Ch. 5, Sect. 6), who admits minimal D‐fusion only in the case of living things. The only views left open seem
to be nihilism, mereological essentialism, and four‐dimensionalism. Each gives a non‐vague answer to the
question of when minimal D‐fusions exist and is thereby una ected by the argument. I discuss each in the
next chapter and argue that four‐dimensionalism is the most attractive of the three.

Notes

8 I thank Phillip Bricker, Yuri Balashov, John Hawthorne, and Dean Zimmerman for helpful comments on this section. See
also Balashov (1999, 2000); Mellor (1981: 128 .); Oderberg (1993, ch. 4); Rea (1998); and Smart (1972) on this topic.
43 An earlier version of this argument was given in Sider (1997). Compare Quine (1981: 10) and Heller (1990, ch. 2, sect. 9).
1 The survey is not intended to be exhaustive. For example, I do not discuss the interesting argument of Lewis (1983a,
postscript B), since I have nothing to add. I do note, however, that while the argument from vagueness of Section 9 shares
some of its assumptions, it does not assume that causal relations supervene on local qualities.
2 See Broad (1923: 54–5, 63, 1933: 141–66, esp. 166), in which he defends the view that we can ʻdispense withʼ things in favor
of processes; Edwards (1758, part 4, ch. 2); Hume (1978, book 1, part 4, sects. 2, 6), though it is perhaps questionable to
place him in the temporal parts camp; Lotze (1887, chs. 1–4); McTaggart (1921: 176–7) (who does not really defend
temporal parts since he does not really believe in time!); Russell (1914: 112 ., 1927: 243 ., 284–9); and Whitehead (1920).
3 Russell (1914: 112). This sort of argument can perhaps also be regarded as Hume's; see the Treatise (1978, book 1, part 4,
sect. 2).
4 See the bibliography in Copeland (1996), and Burgess (1984).
5 For this sort of point see also Wilson (1955), Geach (1972b: 303), and Butterfield (1984).
6 Smart says on p. 130 that ʻa four‐dimensional spacetime frameworkʼ has been ʻforced on physicists by the theory of
relativityʼ; and on p. 140 he says that ʻI advocate my way [of analyzing temporal discourse], because it fits our ordinary
way of talking much more closely to our scientific way of looking at the world . . . ʼ.
7 See Prior (1968a) and Markosian (1993) on the seeming paradoxes of claiming that time flows; Prior (1959) in essence
confronts the anthropocentricity objection. On McTaggart's argument see Broad (1938: 309–17) and Prior (1967: 4–7).
9 See also Armstrong (1980: 74) and Quine (1960: 172) (though it is not completely clear whether Quine has in mind
primarily the B‐theory or perdurance).
10 Although see Mellor (1981: 130–1).
11 The combination of endurance and special relativity I am presenting here faces pressure from my argument in Section 8
below that substantivalists about spacetime should identify continuants with regions of spacetime. In the present section
I am only trying to show that endurance is consistent with special relativity; my overall position is of course that
perdurance is the more attractive view.
12 This notion of exists‐at is a spatiotemporal, not quantificational, notion (Ch. 3, Sect. 2).
13 Or must it? At present we are only presenting an endurance theory for fundamental particles, which apparently have their
intrinsic properties, anyway, permanently; one might take such properties to be instantiated simpliciter. But I do not wish
to base the relativistic theory of endurance on this assumption, which may well only hold contingently.
14 For full generality, the explanation should not presuppose atomism about matter. Suppose a macroscopic object is
ʻgunkyʼ, containing no simples as parts. Facts about its shape in di erent reference frames would be generated by the facts
about the points occupied by the infinitely many objects that overlap it at various points in spacetime.
15 Lowe (1987) made a similar response to Lewis on the problem of temporary intrinsics. In reply, Lewis (1988a) claimed that
since the three‐dimensionalist denies that composition is identity, facts about the spatial relations between the parts of a
thing are not the same as facts about the thing itself; thus, shapes cannot be the same thing as spatial relations between
the parts of a thing. As I see it, Lowe should reply that although composition is not identity, still an object's shape at a time
supervenes on the spatial relations at that time between its parts. Thus an adequate ontology need only explicitly address

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

the latter. Mutatis mutandis for the relativistic case.
16 See also Mellor (1981: 129).
17 See Mundy (1983) and Friedman (1983, ch. VI).
18 See e.g. Armstrong (1980). For critical discussion see Chisholm (1971: 15–16; and 1976, app. A); Rea (1998); and Thomson
(1983, sect. V). For extended discussion of space/time analogies see Schlesinger (1980, ch. 1) and Taylor (1955).
19 Cf. Oderberg (1993: 98 .).
20 In some cases this is not clearly correct. Some say that when an amoeba divides, two new amoebae come into existence.
The division argument for parts should not be applied to such cases.
21 Cf. Pollock (1986, ch. 1, sect. 1).
22 This view is explicitly held by Mellor (1981: 110–14). Variations are defended by Haslanger (1989a); Johnston (1987); Lowe
(1987, 1988a); van Inwagen (1990b); whether these variations di er importantly from Mellor's view is controversial; see
below.
23 Dean Zimmerman (1998b) makes similar points.
24 Might something that happens to be permanently straight be just plain straight? No: surely such a thing would be straight
in exactly the same way as is a temporarily straight thing.
25 Haslanger (1989a: 119–20); Hinchli (1996, 120–1); Lowe (1988a: 73–4).
26 See also Sider (2000c).
27 Assuming substantivalism about time, that is. At the very end of the next section I discuss how the arguments of the
present section apply to a spacetime relationalist.
28 I owe this observation to David Lewis. See the section on Mellor in Lewis (2000).
29 See e.g. Mellor (1998, ch. 8, 1981: 104–7); and Wiggins (1980: 25 n. 12).
30 According to the worm theory, these are two stages of a single person, a spacetime worm that has ʻdoubled back on itselfʼ.
According to the stage view (Ch. 5, Sect. 8), there are actually two persons involved, since each person stage is a person.
One of these can say that he will be the other; the other can say that he was the first, provided the stage‐theoretic truth
condition for ʻx will be Fʼ is modified to read: ʻx has a counterpart in his personal future that is Fʼ. See Lewis (1976) on the
distinction between external and personal time.
31 Thanks to John Hawthorne for this point and also for help with the next paragraph.
32 Relativizing to places would be even more unpalatable to a presentist, who normally would not even relativize to times.
Indeed, this ability to avoid relativization is the basis for a currently popular argument for presentism—see Hinchli (1996)
and Merricks (1994a).
33 See Lewis (1976) for an excellent discussion.
34 George Bealer (1996, sect. 4) gives a similar defense of a priori philosophical reasoning from the challenge of the
necessary a posteriori.
35 For extensive references and an illuminating discussion see Earman (1995).
36 Quine (1976b) makes this identification, but then goes further and gives a construction of the regions in terms of pure sets.
It is hard to go all the way with Quine here; even the most devout reductionist has a strong sense of the di erence in
ontological category between pure sets and blooming buzzing physical things. And even setting this aside, there are
reasons I give in Sider (1996b, esp. n. 22) for resisting this ʻhyper‐Pythagoreanismʼ of Quine's.
37 What of a non‐standard view of enduring spacetime? This would face the objections to the combination of relationalism
and endurance discussed below.
38 See Lewis (1968, 1971).
39 Cf. e.g. the defenses of modal mereological essentialism in Heller (1990), Jubien (1993), as well as the more distantly
analogous Della Rocca (1996); see also Ch. 5, Sect. 7.
40 Cf. Lewis (1986a, sect. 4.5). There are, of course, other responses to the argument, for example rejecting the existence of
the entities involved. I discuss (the temporal analogs of) various positions in Ch. 5.
41 Similarly for cross‐time relations: ⌈x bears R to y n units of time hence/before⌉ means (uno icially) that at some pair of
times n units apart, x as it is at the earlier/later time bears R to y as it is at the later/earlier time.
42 John Hawthorne and I hope to discuss these and related objections in more detail in a future paper.
44 Actually, the special composition question is slightly di erent, since it concerns when fusion takes place at a given time;
see van Inwagen (1990a, ch. 2).
45 Thanks to David Cowles and Ned Markosian here. See Markosian (1998).
46 On the usual terminology, a mereological atom is the fusion of its unit class; let us understand ʻcontinuous series
connecting cases C1 and C2ʼ as excluding ʻcasesʼ involving only one atom.
47 I thank Earl Conee for this observation.
48 See, however, Markosian (1998).
49 I thank John G. Bennett for helpful observations here. Another precise restriction of fusions, to classes that are sets, seems

Downloaded from https://academic.oup.com/book/9493/chapter/156450796 by University of York user on 15 November 2023

unmotivated (and of little consequence even if adopted).
50 See e.g. Dummett (1978: 260) (although see Dummett 1981: 440); Fine (1975); Russell (1923).
51 Cf. also Fine (1975: 267, 274–5).
52 See Lewis (1988b); Parsons (1987); Pelletier (1989); Thomason (1982); van Inwagen (1990a, ch. 18).
53 See e.g. Sorensen (1988: 217–52), and Williamson (1994).
54 Williamson (1994, sect. 7.5) prefers option (2).
55 See Simons (1987: 183 .) and Thomson (1983: 216–17).
56 More carefully, for the nihilist, an assignment A has a minimal D‐fusion i for some simple, x, A's domain is the set of times
at which x exists and A assigns to any such time {x}.
57 A variant of mereological essentialism would also secure precise cut‐o s. According to both mereological essentialism
and this variant, if x is ever part of y then whenever x and y both exist, x must be part of y. The mereological essentialist
adds that x must exist and be part of y whenever y exists, whereas the variant adds instead that y must exist and contain x
as a part whenever x exists. The variant allows a thing to survive the destruction of one of its parts. The criticisms of
mereological essentialism in Ch. 5, Sect. 7 apply to this variant as well.