Oxford Studies in Metaphysics Volume 7

OXFORD STUDIES IN METAPHYSICS
OXFORD STUDIES IN METAPHYSICS
Editorial Advisory Board

David Chalmers (Australasian National University)
Andrew Cortens (Boise State University)
Tamar Szabó Gendler (Yale University)
Sally Haslanger (MIT)
John Hawthorne (Oxford University)
Mark Heller (Syracuse University)
Hud Hudson (Western Washington University)
Kathrin Koslicki (University of Colorado, Boulder)
E. J. Lowe (University of Durham)
Brian McLaughlin (Rutgers University)
Trenton Merricks (University of Virginia)
Kevin Mulligan (Université de Genève)
Theodore Sider (Cornell University)
Timothy Williamson (Oxford University)
Managing Editor
Matthew Benton (Rutgers University)
OXFORD STUDIES
IN METAPHYSICS
Volume 7
Edited by
Karen Bennett
and
Dean W. Zimmerman
1
1
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© The several contributors 2012
The moral rights of the authors have been asserted
First Edition published in 2012
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Data available
ISBN 978–0–19–965908–1 (Hbk.)
978–0–19–965907–4 (Pbk.)
Printed in Great Britain by
MPG Books Group, Bodmin and King’s Lynn
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
PREFACE
Oxford Studies in Metaphysics is dedicated to the timely publication

of new work in metaphysics, broadly construed. The subject is
taken to include not only perennially central topics (e.g. modality,
ontology, and mereology) but also metaphysical questions that
emerge within other subfields (e.g. philosophy of mind, philosophy
of science, and philosophy of religion). Each volume also contains
an essay by the winner of the Oxford Studies in Metaphysics Younger
Scholar Prize, an annual award described within.
D.W.Z.
New Brunswick, NJ
This page intentionally left blank
CONTENTS
Oxford Studies in Metaphysics: Younger Scholar Prize ix
I. RELATIVE SAMENESS
1. Possibility Relative to a Sortal 3

Delia Graff Fara
2. Reflections on Counterpart Theory 41
Allen Hazen
II. ABSOLUTE GENERALITY
3. All Things Must Pass Away 67

Joshua Spencer
4. Absolute Generality Reconsidered 93
Agustín Rayo
III. HUMEANISM AND LAWS OF NATURE
5. Goodbye, Humean Supervenience 129

Troy Cross
6. “There sweep great general principles which all the laws
seem to follow” 154
Marc Lange
IV. CONTINGENT OBJECTS AND COINCIDENT OBJECTS
7. Relativized Metaphysical Modality 189

Adam Murray and Jessica Wilson
8. Coincidence Through Thick and Thin 227
Sydney Shoemaker
viii | Contents
V. THE OPEN FUTURE
9. The Real Truth about the Unreal Future 257

Rachael Briggs and Graeme A. Forbes
10. Presentism and Distributional Properties 305
Jonathan Tallant and David Ingram
Name Index 315

THE OXFORD STUDIES IN METAPHYSICS
YOUNGER SCHOLAR PRIZE
Sponsored by the Ammonius Foundation* and administered by the

editorial board of Oxford Studies in Metaphysics, this annual essay
competition is open to scholars who are within ten years of receiv-
ing a Ph.D. or students who are currently enrolled in a graduate
program. (Independent scholars should enquire of the editor to
determine eligibility.) The award is $8,000. Winning essays will
appear in Oxford Studies in Metaphysics, so submissions must not be
under review elsewhere.
Essays should generally be no longer than 10,000 words; longer
essays may be considered, but authors must seek prior approval by
providing the editor with an abstract and word count by 1 Novem-
ber. To be eligible for next year’s prize, submissions must be elec-
tronically submitted by 31 January (paper submissions are no longer
accepted). Refereeing will be blind; authors should omit remarks
and references that might disclose their identities. Receipt of sub-
missions will be acknowledged by e-mail. The winner is determined
by a committee of members of the editorial board of Oxford Studies
in Metaphysics, and will be announced in early March. At the
author’s request, the board will simultaneously consider entries in
the prize competition as submissions for Oxford Studies in Metaphys-
ics, independently of the prize.
Previous winners of the Younger Scholar Prize are:

Thomas Hofweber, “Inexpressible Properties and Proposi-
tions”, Vol. 2;
Matthew McGrath, “Four-Dimensionalism and the Puzzles of
Coincidence”, Vol. 3;
* The Ammonius Foundation is a non-profit organization dedicated to the revival

of systematic philosophy and traditional metaphysics. Information about the Foun-
dation’s other initiatives may be found at <http://www.ammonius.org/>.
x | Younger Scholar Prize
Cody Gilmore, “Time Travel, Coinciding Objects, and Persist-

ence”, Vol. 3;
Stephan Leuenberger, “Ceteris Absentibus Physicalism”, Vol. 4;
Jeffrey Sanford Russell, “The Structure of Gunk: Adventures
in the Ontology of Space”, Vol. 4;
Bradford Skow, “Extrinsic Temporal Metrics”, Vol. 5;
Jason Turner, “Ontological Nihilism”, Vol. 6;
Rachael Briggs and Graeme A. Forbes, “The Real Truth About
the Unreal Future”, Vol. 7;
Shamik Dasgupta, “Absolutism vs Comparativism about
Quantities”, forthcoming, Vol. 8.
Enquiries should be addressed to:
dwzimmer@rci.rutgers.edu
I
RELATIVE SAMENESS
1. Possibility Relative to a Sortal
Delia Graff Fara
1. THE LEADING IDEA
There is a distinction between identity and sameness. There is only

one identity relation but there are many sameness relations. As
many have said, identity is that relation which everything bears to
itself but to no other thing. But something may be the same, in a
number of different respects, as something other than itself. Same-
ness relations fall under at least two categories. First, there are those
that are relativized to a quality—e.g., same color or same height. Sec-
ond, there are those that are relativized to a sort of thing—e.g., same
person or same boat.
Philosophers make the distinction between “qualitative identity”
and “numerical identity.” Numerical identity is identity. Qualita-
tive identity can be thought of as a special case of sameness relativ-
ized to a quality: sameness with respect to every quality. (Or perhaps
the notion is usually relativized, in a context, to every salient or
relevant quality.) Sameness with respect to a given quality obvi-
ously does not require numerical identity. For example, distinct
things can be the same height as each other. But it is generally
assumed that sameness with respect to a sort does require numeri-
cal identity. To ask whether the ship we are on now is the same ship
as the one we were on last week is to ask, it is assumed, whether
this ship and that one are ships that are numerically identical.
I reject that assumption. The ship we were on last week might be
the same ship as the one we are on now without being numerically
identical to it. In my view, to say that they are the same ship is to say
that if we were to count how many ships we have been on in the
past week, the correct answer could be “one” even if the ship we
were on last week is not numerically identical to the ship we are on
now.
Some people, when they are first introduced to the problem of
“personal identity” say that they are not the same person as the one
4 | Delia Graff Fara
they were when they were a child because their interests and values
have changed drastically. In my view, such people are not thereby
making a category mistake. They have not thereby confused qualita-
tive identity with numerical identity. They have not made the mistake
of thinking that significant qualitative change over time precludes
numerical identity over time. Rather, they have made a mistake
about what conditions are required for a particular sort-relativized
sameness relation to hold over time. They mistakenly think that sig-
nificant change with respect to interests and values precludes same-
ness of person over time. That latter relation, sameness of person, is
not, on my view, the relation of numerical identity—even when
restricted just to persons.
In addition to the distinction between sameness relativized to a
quality and sameness relativized to a sort, there is a distinction to be
made even within the category of sort-relative sameness relations.
There is sort-relative sameness of types and also sort-relative same-
ness of tokens. If you and I are both wearing blue platform wedges
then there is a sense in which we are wearing the same shoes even
though the ones on my feet are not on your feet. We are wearing the
same type of shoes, but not the same individual (“token”) shoes
Peter Geach would have thought that in the case of our shoes, not
only are you and I wearing the same type of shoes, my shoes and
your shoes are the same relative to a certain sort, namely, the sort
shoe-type (Geach, 1962). Of course Geach did think that my shoes and
your shoes were not the same shoe-tokens. I disagree with the former
claim. We do not have a case here of things being the same relative to one
sort of thing while not being the same relative to another sort of thing.
Our shoes are not the same shoe-type while being different shoe-
tokens. Although we are wearing the same type of shoes on our feet,
neither of us is wearing a shoe-type on our feet. In other words,
although we are wearing the same type of shoes, there is nothing on
my feet that is the same, relative to any sort, as anything on your feet.
On this point, I agree with William Alston and Jonathan Bennett
(1984) and with John Perry (1970), as against Geach. From here
onwards, I will ignore sort-relative sameness of types.
I will also from here onwards ignore relations of sameness rela-
tivized to a quality. So as shorthand I will use the terms “relative
sameness” and “sort-relative sameness”—even sometime just
“sameness”—to mean sort-relative sameness of tokens.
Possibility Relative to a Sortal | 5
Some have used the term “relative identity” to mean what I am

calling relative sameness. Some proponents of “relative identity”
have denied that there is any relation of absolute identity, instead
only same-F-as relations.1 For this reason it is worth emphasizing
that in admitting the various relative-sameness relations, I in no
way reject the absoluteness of identity.
I am, however, driving a wedge between relative sameness and
absolute identity—as concerns what was, what will be, or what
might have been—that makes for a greater gap than philosophers
typically allow space for.2 For example, I think that the boat that
Michael just embarked on is the same boat as the one he will later
row ashore (should all go as planned), even though the boat that he
embarked on is not absolutely identical to the boat that he will row
ashore. Like Theseus’s ship, Michael’s boat will have exchanged or
lost some of its parts over the course of its long river crossing. This
suffices to make the embarked-upon boat numerically distinct from
the rowed-ashore boat. But it does not suffice to make them differ-
ent boats.3 Modal variant: if Michael does not in the end row his boat
ashore because it capsizes, then the boat that he would have rowed
ashore—if only his boat had not capsized—would be the same boat
as the one he in fact embarked on, even though the boat that he
would have rowed ashore in that event would not be absolutely
identical to the boat he did in fact embark on.
I am not, however, driving so much of a wedge between same-
ness and identity that I would qualify as a “relative identity” theorist.
1
The view is most associated with Geach (1962).
2
A number of philosophers do allow for the space: Peter Geach (1962, 1967, 1972);
Roderick Chisholm (1973, 1976); and Anil Gupta (1980) are examples.
3
With respect to identity over time, I here agree with Roderick Chisholm (1976),
who thought that this ship here now could be the same ship as that one there then
even though they are not numerically identical. Like Chisholm, I also think that the
same-ship relation, rather than the numerically-identical relation is the one that is rele-
vant for counting ships. Following Bishop Butler, though, Chisholm accepted that it
is only in a “loose and popular sense” that identity may hold between numerically
distinct ships. On that view, for two objects to be the same ship they must each be a
ship and must be identical in the loose and popular sense. But for the two objects to
be numerically the same thing is for them to be identical in the “strict and philosophical”
sense. We can speak of identity “in the loose and popular sense” if we like, but there
is really no need to once we recognize that being the same F over time does not
require numerical identity (identity in the strict and philosophical sense) over time.
Although I think that there is a distinction between sameness and

identity, I do not think there is any world-time point at which dis-
tinct things can be the same with respect to any sort. Nor do I think
that there is any world-time point at which identical things can be
different with respect to any sort. I think rather that “transworld” or
cross-time sameness can come apart from transworld or cross-time
identity. A boat a at world-time point <w,t> may be the same boat as
the boat b at world-time point <w’,t’> even though a ≠ b, as long
as <w,t> ≠ <w’,t’>. The boat that I am actually on might have been
a red boat even though it is in fact a blue boat. This is to say not that
there’s a possible world in which there’s a red boat that is identical
to the boat that I’m actually on (though that may be true). Rather, it
is to say that there’s a possible world in which there’s a red boat that
is the same boat as the blue one that I’m actually on. I hold that
those two conditions can come apart. Correspondingly, I do not
think that two things which are both F and both G can be the same
F while not being the same G. This is to say that I reject another
central thesis of the sort of relative-identity theory associated with
Geach, though not with Roderick Chisholm.
Since I do not think that there is any kind of identity other than
absolute identity, I am happy to dispense with the qualification
“absolutely.” The boat we’re on now is the very same boat as the
one we were on last week. But they are not identical, for identity is
“too unyielding a relation” (to borrow Robert Stalnaker ’s words) to
hold of entities composed of different matter.
(By the way, the use of the plural “they” here should not be taken
to suggest that two different boats are being counted any more than
two different planets are being counted when I use a plural verb form
to say that Hesperus and Phosphorus were discovered to be the same
planet. The plural is appropriate because two different noun phrases
were used, not because more than one thing was mentioned.)
I propose that transworld sort-relative sameness is the important
relation for analyses of de re possibility—for analyses, that is, of
what we mean when we say what a certain thing might have been
or was or will be like.4
4
Such a relative-sameness counterpart theory is discussed briefly by Theodore
Sider (1999), who attributes it to Michael Jubien (1993), although Jubien would disa-
vow it. One finds a kernel of the same theory in Jubien (2001).
For example, I propose that a certain woman might have worn

green shoes today, even though she is in fact wearing blue ones,
because there is a possible world in which someone who is the same
person as her did wear green shoes today.
There are two salient competing proposals. The first is associ-
ated with Saul Kripke (1963, 1972): transworld identity is the
important relation for analyses of de re possibility. This is the view
that it is true that a certain woman might have worn green shoes
today just in case there is a possible world in which something that
is identical to that woman did wear green shoes today. The meta-
physical question of whether that thing itself need be a person may
be left open.
The second is associated with David Lewis (1968, 1971, 1986):
transworld superlative qualitative similarity is the important relation
for analyses of de re possibility. This is the view that it is true that a
certain woman might have worn green shoes today just in case there
is a possible world in which a thing wearing green shoes today is
qualitatively similar to the woman and is no less qualitatively simi-
lar to her than is anything else in its world. Again, it is left open in
principle whether that thing is a person. (I’ve used the term ‘super-
lative’ so that it allows for ties. Think of it as synonymous with the
term ‘unbeaten’.)
My proposal comes apart from the identity proposal since trans-
world sort-relative sameness does not require transworld identity.
We might have been on the same ship now as the one we were actu-
ally on last week even if the ship we would have been on now, in
that event, were not identical to the ship that we were actually on
last week.
My proposal comes apart from the qualitative-similarity proposal
since transworld sort-relative sameness does not require trans-
world superlative qualitative similarity. The boat we are actually
on might have (i) been a different color even while (ii) there were
some different boat of the same color as the one we are actually on
and qualitatively just like it—even in very extrinsic ways, such as
having our qualitatively identical twins on it. We are actually on
boat a. We might have been on the same boat b even while there
were some different boat c that were qualitatively just like a—even
in very extrinsic ways.
Granting this much, what do we say about the truth of the fol-
lowing modal and temporal claims? What conditions must be met
in order for them to be true?
(1) Angel, Michael’s boat, might have been rowed ashore.
(2) Angel will be rowed ashore.
The less popular answer, derived from Lewis (1968, 1971, 1986) for
the modal case and from Katherine Hawley (2001) and Theodore
Sider (1996, 2001) for the temporal case, is that these sentences are
true just in case there is a possible world (in the case of (1)), or a
future time (in the case of (2)), at which something that is a counter-
part of Angel is rowed ashore. For the moment, let us leave behind
the question of just what this counterpart relation is. The important
point for us now is that it may hold between things that are not
identical to each other.
The more popular answer, derived from Kripke (1963, 1972), is
that these sentences are true just in case:
(A) There is a possible world, or future time, at which some-
thing identical to Angel is rowed ashore.
Do we accept the popular answer, so phrased? For Kripke, and
those who follow him on this question, there is no distinction
between that condition and another:
(B) There is a possible world, or future time, at which some-
thing that is the same boat as Angel is rowed ashore.
But for us, the two conditions come apart. The leading idea for us is
that (A) and (B) are different views. We may accordingly say that
(B), the second phrasing of the popular view, is the correct one;
while (A), the first phrasing, is incorrect. A de re modal claim “a
might have been Φ” is true when there is something in some pos-
sible world to which a bears a certain relative-sameness relation
and which is Φ in that world—whether or not there is something
identical to a that is Φ in that world. Likewise, a de re future claim “a
will be Φ” is true when there is something in the future to which a
bears a certain relative-sameness relation and which is Φ at that
future time—whether or not there is something identical to a that is
Φ at that time. Which sameness relation is involved (same person, same
body, same boat, et cetera) varies with the context.5 We will have to
revise and sharpen this up in a bit. But before we do, I want to point
out that this view about the truth conditions for de re predications
(temporal or modal) has something in common with both the popu-
lar view and the Lewisian view.
It shares with the popular view the idea that future and coun-
terfactual possibilities for a thing a are determined by the ways
that a thing b is at a future time, or possible world, whenever a
bears a certain relative-sameness relation to b. Because the the-
ory has this feature, I call it a relative-sameness theory of de re
predication.
It shares with the Lewisian view the idea that future and counter-
factual possibilities for a can be determined by the ways that a thing
b is at a future time, or possible world, even if a is not identical to b.
Because the theory has this feature, I call it a counterpart theory of
de re predication.
Let me emphasize two things before moving on. One, I intend for
relative-sameness counterpart theory to cover temporal predication
(de re) as well as modal predication (de re). It simplifies matters,
however, if I focus on just one of these at any given time. For the
most part I focus on modal predication, but it should be somewhat
clear how the view extends to temporal predication.
Two, I use the term ‘counterpart theory’ liberally. To be a counter-
part theorist is not to have any particular view about what it takes
for an individual in one possible world to be a counterpart of some
individual in some different possible world. Any theory of de re
modality is a counterpart theory in my sense if it takes the trans-
world relation that is important for the analysis of de re possibility
to be a relation other than identity. For example, someone who held
the ridiculous view that it is true that a certain woman might have
worn green shoes today just in case there’s a possible world in
which she has a sister who is wearing green shoes today would be a
counterpart theorist in my sense.
5
This appeal to contextual variation can be found in Lewis (1971) and Allan
Gibbard (1975). Gibbard’s views on contingent identity inspired me to develop the
view presented here. I differ from both Lewis and Gibbard on a number of important
issues—among them, the question whether objects are extended in time.
2. MOTIVATING THE IDEA
Why develop a relative-sameness counterpart theory of de re modal-

ity? My aim is to provide a plausible account of possibility and identity
claims while preserving the attractive metaphysical view that ordinary
material objects are identical to the matter that constitutes them even
though they might not have been constituted by that matter. The boat
and the wood that it is made up of are one and the same thing—they
are identical. But they might not have been. They are identical since if
we were to count the things on the river we would have already
counted at least one too many if we added the boat to our count and
separately added the wood to our count. (This is not to say that we
could, even in principle, count all of the things on the river.) Never-
theless, it might have been that that same boat was not constituted by
that same wood. The boat would not have been constituted by that
same wood if some of the boat’s planks had been exchanged during
some previous river crossing. It might have been, therefore, that that
same boat and that same wood were not one and the same thing. It also
follows, moreover, that they might not have been identical.
The boat’s name is ‘Angel’. ‘Dusty’ is the name of some wood
that is a proper part of the wood that makes up Angel. We accept
the following triad.
(3) Angel and the wood that it is made up of are identical.
(4) Angel might not have had Dusty as a part.
(5) The wood that Angel is made up of cannot but have had
Dusty as a part.
One could reason for a contradiction from these three claims as fol-
lows. Appealing to Leibniz’s law and the identity statement in (3),
substitute ‘Angel’ for ‘the wood Angel is made up of’ in (5) to yield
(4†), which contradicts (4).
(4†) Angel cannot but have had Dusty as a part.
Alternatively, substitute ‘the wood Angel is made up of’ for ‘Angel’
in (4) to yield (5†), which contradicts (5).
(5†) The wood that Angel is made up of might not have had
Dusty as a part.
I will show how this reasoning may be resisted: (3)–(5) do not jointly
yield a contradiction.
One could lamely reject the reasoning by noting that identity

statements like (3)—those which have a definite description on one
side of the identity sign—do not license substitution within the
scope of a modal operator when the description involved is not nec-
essarily true of the thing it is actually true of. If Barack Obama is
identical to the president of the United States, and it is not possible
for the president of the United States to have never been elected
to a political office, it does not follow that it is not possible for
Barack Obama to have never been elected to a political office. Why?
Because the description ‘the president of the United States’ does not
necessarily apply to Obama.
But I do not want to reject the reasoning on those lame grounds.
I want to promote the stronger claim that Angel, the boat, is identi-
cal to the wood that she is actually made up of—she is identical to
that wood, to it. Give that wood the name ‘Lumber’. I want to pro-
mote the compatibility of this stronger triad of claims. ‘Dusty’ is the
name of some wood that is a proper part of Lumber.
(6) Angel and Lumber are identical.
(4) Angel might not have had Dusty as a part.
(7) Lumber cannot but have had Dusty as a part.
One could now use the substitutivity reasoning made above to
argue for conclusions that contradict those in the above triad.
(4†) Angel cannot but have had Dusty as a part.
This, again, contradicts (4).
(7†) Lumber might not have had Dusty as a part.
This contradicts (7).
Generalizing from the claims in (3)–(7), I want to defend the fol-
lowing theses.
(IT) The Identity Thesis: Ordinary material objects are iden-
tical to the matter that makes them up.
(ACT) The Accidental Constitution Thesis: Ordinary material
objects could be made up by matter that is different
from the matter that actually makes them up.
(NMT) The Necessary Matter Thesis: The matter that makes up
an ordinary material object could not be made up by
matter that is different from the matter that actually

makes it up.
I call the combination of these three theses The Sensible View. The
person who opposes the identity thesis has an argument for her
view, namely, the one that proceeds by appeal to Leibniz’s Law.
Rather than give a countervailing argument for my views, let me
give some sense of my reasons for individually accepting the three
theses. I just want to give some sense of why I hold them. I then go
on to provide an account of de re possibility that blocks the Leibniz’s
Law argument against The Sensible View.
Some thoughts behind the Identity Thesis: If you were to tally the number
of things on my desk, for example, you would have counted at least
one thing too many if you counted the matter that makes up one of the
books and separately counted the book itself. This is what David Lewis
meant when he described this as “double counting” (1986, 252).6
Some thoughts behind the Accidental Constitution Thesis: Every ordi-
nary material object has some material part, however small, that
could be removed without the object itself either becoming a scat-
tered object or ceasing to exist. If I bite off a small piece of my paper
coffee cup and swallow it, nothing then in my stomach is then a
part of the coffee cup, even though there is something then in my
stomach that used to be part of the coffee cup. And what’s more
obvious, I do not thereby destroy the coffee cup.
Some thoughts behind the Necessary Matter Thesis: This is true by stipu-
lation. I hereby stipulate that I am using the terms “constitutes” and
“matter” so that whenever X is made up of some matter Y, and Y is
made up of some matter Z, then Y is necessarily made up of Z.7 (I’m
using the expression “makes up” interchangeably with “constitutes”.
6
Mark Johnston has argued that there is no plausible argument for the identity
thesis (Johnston 1992). Harold Noonan (1993) argued, in response to Johnston, that
not all of Johnston’s dismissals of arguments for the identity thesis are justified.
That’s as may be. I do not take myself to be providing a direct argument for the iden-
tity thesis, nor even a rebuttal to Johnston’s arguments that there is no plausible such
argument. Rather I take myself in what follows, to be giving a defense of the identity
thesis—since it’s a thesis that I find attractive—by explaining how it does not lead to
a contradiction.
7
I appreciate comments from Dean Zimmerman that led to this formulation of the
stipulation.
I use one or the other phrase according to what sounds more natural
to my ear.)
The leading idea of this paper enables us to coherently maintain
The Sensible View. The leading idea is that sort-relative sameness
does not require identity. This boat here now might be the same
boat as that one there later even though they are not identical.
Clearly, anyone who thinks this—that ordinary material objects,
like boats, are identical to the matter that they are made up of even
though the objects, unlike the matter itself, could be made up of dif-
ferent matter—must accept some statements of contingent identity.
Angel is identical to Lumber, but she might not have been.
The leading idea allows us to make sense of this (merely) seem-
ing contingency of identity. It allows us to say that it is true that
Angel might not have been identical to a thing x when there’s a
possible world in which there’s some boat, Angel*, that is the same
boat as Angel but which is not identical to x. Let us introduce a piece
of terminology. We will call Angel* Angel’s “boat-mate” in that
world.
Since Angel is not identical to her otherworldly boat-mate, there
is no contradiction in her being identical to Lumber even though
Lumber is not identical to the boat-mate. The following are
compatible:
(8) Angel = Lumber;
(9) Angel*, in the counterfactual world β, is the same boat as
Angel;
(10) Lumber ≠ Angel*.
These are compatible since being a boat-mate does not require iden-
tity. The joint truth of (8)–(10) is what on my view renders the fol-
lowing true:
(11) Angel is identical to Lumber, but Angel might not have
been identical to Lumber.
None of this really commits us to the contingency of identity, how-
ever. Identity is a relation that holds necessarily (and perma-
nently). By this we mean that there are no things a and b which are
identical in one possible world but not identical in some other pos-
sible world. (Similarly, there are no things a and b which are identi-
cal at one time but not identical at some different time.) Sort-relative
sameness, in contrast, is not a relation that holds necessarily (or
permanently). Sort-relative sameness does not coincide with iden-
tity, but it, rather than identity, is the important relation for analy-
ses of de re modal (and temporal) claims—even when these are
modal (or temporal) identity claims. That is why the truth of par-
ticular modal statements of contingent identity does not commit us
to the contingency of identity.8 This thought will be developed
and sharpened in what follows.
3. DEVELOPING THE IDEA
Symmetry and transitivity are logical properties of relative same-

ness as well as identity. This is obviously true when the sameness is
relativized to a quality: if I’m the same height as you then you are
the same height as me (symmetry), and if you are also the same
height as Maria then I’m also the same height as Maria (transitivity).
It is also true when the sameness is relativized to a sort: if Rusty is
the same boat as Scarab and Scarab is the same boat as Lancha,
then Lancha is the same boat as Rusty (symmetry and transitivity).
Unlike identity, however, relative-sameness relations are not quite
equivalence relations, since they are not all reflexive. Since Barack
Obama is not a boat, he is not the same boat as anything, not even
himself. The only way for you to be the same boat as something is if
you yourself are a boat. Only a boat can bear this relation to some-
thing. And, of course, every boat does bear this relation to itself.
Since only boats bear the relation to anything, and all boats bear the
relation to themselves, we may say that the same-boat relation is
weakly reflexive. A relation is weakly reflexive when anything that
bears the relation to at least one thing bears the relation to itself.
Weak Reflexivity: ∀x(∃yRxy → Rxx).
We will say that a relation is a weak equivalence relation if it is sym-

metric, transitive, and weakly reflexive. Weak equivalence is a
logical property of all relative-sameness relations.
8
This interpretation of what acceptance of contingent identity amounts to can be
attributed to Gibbard (1975).
Although I call my theory a counterpart theory, I do not accept all

of the postulates that were originally stipulated by Lewis (1968) to
constitute counterpart theory. In particular, I deny Lewis’s postu-
late P2, the stipulation that individuals are “world-bound”—that
none exists in more than one possible world:
P2: ∀x∀y∀z((x is in possible world y ∧ x is in possible world
z) → y = z).
I reject P2. That is, I accept that there are transworld identities. Indi-
viduals are not world bound: an individual in one possible world
may be identical to some individual in a distinct possible world.
I also accept, remember, the following theses:
(IT) The Identity Thesis: Ordinary material objects are iden-
tical to the matter that makes them up.
(ACT) The Accidental Constitution Thesis: Ordinary material
objects and their constituting matter might have been
different things, however.
(NMT) The Necessary Matter Thesis: The matter that makes up
an ordinary material object could not be made up by
matter that is different from the matter that actually
makes it up.
Along with these, I accept for now the following preliminary state-
ment of an example of possibility relativized to a sortal (PRS).
(PRS) An ordinary material object, such as a boat, might have had
properties that are different from the ones it in fact has only
because it might have been that that very boat (something
that is the same boat) did have the different properties.
The view as I’ve developed it so far might at this point seem bizarre.
Here is one illustrative way to depict it. We have the actual world,
α, and a counterfactual world, β.
α (actual) β (counterfactual) be is the boat embarked upon in α;
d is the wood constituting be in α;
d d
br is the boat rowed ashore in β;
be = d br ≠ d d does not constitute br in β;
be and br are the same boat.
In the actual world, be is the boat that Michael embarked upon and
it is constituted by some wood d. Where we have the boat and the
wood that constitutes it we have just one thing, so in α, be = d. In β,
br is the boat that Michael rows ashore. It is not constituted by the
wood d. In β, d does not constitute any boat at all. So in β, d ≠ br. In
the actual world, α, Michael does not row the boat ashore, since the
boat he embarked on capsizes. Nevertheless, Michael might have
rowed the boat ashore. In the part of modal space that I’m describ-
ing, that is because be , the boat embarked-upon in α, and br, the boat
rowed ashore in β, are the same boat. We can see that be and br are
nevertheless not identical, since be = d but d ≠ br.
Now is anything wrong with this picture? We can see that it allows
for transworld identities, which were proscribed by Lewis’s P2, since d
exists in the actual world as well as in the counterfactual world β. But
does be , the embarked-upon boat, exist in the counterfactual world β as
well in the actual world α? Since it is identical with d it has to exist in β,
since d exists there. And so it does. Identity is a necessary relation. So
since the wood d, which in fact constitutes the embarked upon boat be ,
is not identical to the boat rowed ashore in β, be ≠ br. Although be and br
are the same boat, they are not identical—not even in β.
In view of the above, it might seem that we have incurred the fol-
lowing commitments, which cannot all be right:
(12) be = d.
(13) be is the same boat as br in β.
(14) d is not a boat in β and therefore not the same boat as
anything in β.
These cannot all be right since it follows from these that be is the
same boat as br in β but also not the same boat as anything in β. As
we sharpen the idea, we will see that we are not in fact committed
to this inconsistent triad.
4. SHARPENING THE IDEA
Evidently, the relative-sameness relation same-boat must be relativ-

ized to worlds. Whether or not be in the actual world is the same
boat as br in the counterfactual world depends not only on what br
is like in the counterfactual world but also on what be is like in the
actual world. At the very least, be must be a boat in the actual world
and br must be a boat in the counterfactual world. Only boats can be

the same boat as anything. Of course, there are stronger conditions
in addition to that one.
Switch for the moment to sameness over time. Here’s a plausible
condition on cross-time sameness of boat. If I own a boat now and
you own a boat later, then they are not the same boat unless I relin-
quished my right to it or was legitimately evicted from it.9 The
transworld analog is not pretty to state: if I in fact owned a boat for
the whole of last month but you might have owned that same boat
by the end of the month, that can only be because I might have
legitimately relinquished my right to it or have been legitimately
evicted from it by the end of the month. Put another way, trans-
world boat-sameness requires any counterfactual transfer of own-
ership to have been done legitimately. This is not to say that you
could not steal my boat. You might have stolen the very boat that I
in fact own, but you would not in that case own the same boat that
I in fact own.
I will not make any further attempts to say what the necessary
conditions are for transworld sameness of boat. For now, it is suffi-
ciently illustrative to say something further about what necessary
conditions there are not. Since the boat that Michael actually
embarked on might have changed color during his river crossing, it
is not required that the boat he rowed ashore in the counterfactual
world be the same color in that world as the actually embarked-
upon boat is in the actual world. Likewise, since the boat that
Michael embarked on might have been manned by Gabriel by the
time it was rowed ashore, it is not required that the boat rowed
ashore in the counterfactual world be manned by the same person
in that world that manned the embarked-upon boat in the actual
world. And so on for any other accidental property of the boat that
was actually embarked upon.
I presume that there are necessary conditions for something in a
counterfactual world to be the same boat as something in the actual
world. This is just to presume that boats have essential properties.
9
Jubien makes roughly this same point (Jubien 2001). He, however, makes it as a
psychological point about how we are inclined to think about the cross-time same-
boat relation rather than as a metaphysical point about the cross-time same-boat rela-
tion itself.
What those conditions are is not per se part of the theory presented
here.10 I am interested primarily in the formal properties of relative
sameness and in applying these to certain metaphysical views
about constitution that I happen to hold. I want to develop a rela-
tive-sameness counterpart theory whose formal properties allow
for the combination of the Identity Thesis, the Accidental Constitution
Thesis, and the Necessary Matter Thesis while providing a plausible
analysis of de re modal (and temporal) predication. Someone with
different metaphysical views about essence and accident—whether
that be as it relates to material constitution or to something else,
such as personal “identity”—might well find the formal analysis
here congenial for her purposes. That would not be to accept a dif-
ferent view but rather to combine the view here with a different
metaphysics.11
This is like saying that someone could accept Kripke’s identity
analysis of de re possibility but reject his view that sameness of ori-
gin is required for transworld identity. Similarly, it would be like
saying that someone could accept Lewis’s counterpart theory of de
re possibility but think that similarity of origin is more important
for transworld similarity than is similarity in any other respect.
4.1 World-Relativized Relative Sameness

As we have said, whether or not x is the same boat as y will depend
on what x is like and on what y is like. But, like all ordinary objects,
boats might have been different from how they actually are. So if
the transworld sameness of x and y is to depend on what they are
like, it must depend on what they are like in a world. We therefore
need to further relativize relative-sameness relations to worlds. The
relative-sameness relations appealed to in our theory of de re modal
predication will officially be understood as relations holding
between thing-world pairs. In other words, whether <x, w> bears a
10
Sider (1999, 289–90) makes essentially the same point in partial defense of
Jubien’s (1993) version of counterpart theory.
11
Chisholm offered us criteria for cross-time sameness of table (Chisholm 1976,
218–21). But he need not have. I’m pointing out that we can present a theory of the
kind presented here while having no obligation to analyze what’s required for cross-
time or transworld sameness for each sort of thing that there is.
relative-sameness relation to <y, w’> depends on what x is like at w

and what y is like at w’.
Of the many ways that x and y are in a possible world, which
ones count will depend on which sortal the sameness is being rela-
tivized to. For instance, if x in w is to be the same boat as y in w’,
then x and y will have to be boats in w and w’, respectively. Whether
x and y, in their respective worlds, have all the same matter parts as
each other will not be important. When the same-matter relation is
at issue, however, sameness of matter parts is of the utmost impor-
tance. Since there are many different sorts to which a sameness rela-
tion may be relativized, we can indicate this by providing subscripts.
Where F stands for a sort of thing, RF is to be that sort-relative same-
ness relation that is relativized to the sort F. For example, Rboat is the
same-boat relation.
Our relative-sameness counterpart theory of de re possibility can
now be provisionally stated as follows:
(RSC*◊) Relative to a sortal F, ‘◊Φx’ is true at a world w just in
case there is a thing y and possible world w’ such that
<x, w> RF <y, w’ > and such that ‘Φy’ is true at w’.
Our instance of (RSC*◊) is this:
Relative to the sortal boat, ‘Angel might have been rowed ashore’ is
true at a world w just in case there is an object y and a world w’ such
that Angel is the same boat as y, given the ways that Angel is in w
and that y is in w’, and such that y was rowed ashore in w’.
This allows us to rewrite the inconsistent (12–14) as the consistent
(12*–14*):
(12*) be = d;
(13*) be in α is the same boat as br in β;
(14*) d is not a boat in β and therefore, in β, d is not the same
boat as anything in β.
4.2 Formalizing World-Relativization

We have now relativized the (sort-relativized) sameness of objects
to worlds. This is not to admit funny objects—things-at-a-world,
that are somehow world “slices” of (somehow) modally extended
objects. We are regarding these relations as relations on genuine

pairs. The first member of each pair is a ordinary object. Ordinary
objects do not have some of their parts in one possible world, some
other parts in some other worlds. Rather, in each possible world in
which an ordinary objects exists, there are ways that that object is.
This is reflected in our taking relativized sameness to be a relation
on thing-world pairs. These sameness relations are symmetric, tran-
sitive, and weakly reflexive as relations on pairs.
With regard to symmetry, this is to say that if x is the same boat as
y relative to the ways that x is in w and that y is in w’, then y is the
same boat as x relative to the ways that y is in w’ and that x is in w.
For example, if the boat that Michael embarked on in the actual
word is the same boat as the one that Gabriel rowed ashore in world
β, given the ways that these boats are in their respective worlds,
then the boat that Gabriel rowed ashore in world β is the same boat
as the one that Michael embarked on in the actual word, given the
ways that these boats are in the respective worlds. In symbols:
Symmetry: <x,w> RF <y, w’> → <y, w‘> RF <x,w>.
Transitivity and weak reflexivity are now given corresponding
world-relative formalizations:
Transitivity: (<x,w> RF <y,w’> ∧ <y,w’> RF <z,w’’>) → <x,w> RF <z,w’’>;
Weak Reflexivity: ∃y∃w’ (<x, w> RF <y, w’ >) → <x,w> RF <x,w>.
4.3 Relative Sameness and Identity

Despite the need for these relativizations to possible worlds, we
do not deny that there are normal relative-sameness relations that
hold among things simpliciter, ones that are not further relativ-
ized to a world. I am the same person as myself, the same mother
as myself, the same philosopher as myself, . . . simpliciter. It cannot
be that there is something distinct from me that is the same
woman as myself simpliciter, since no woman has properties that
are incompatible with others of her properties. No woman is ever
in two places at once, for example. We can put this point formally
as follows. Simpliciter relative-sameness relations are not only
weak equivalence relations, they are subsets of the identity rela-
tion. They will be proper subsets of the identity relation whenever
they are relativized to a nonvacuous sort of thing—that is, to a

sort of thing that some things are not. For example, I am not the
same boat as myself since I am not a boat. Therefore, the simplic-
iter same-boat relation is a proper subset of the identity relation.
The simpliciter same-boat relation is that subset of the identity
relation which contains those pairs <d, d> such that d is a boat.
Vacuous sorts, if we admit them, might include thing or entity.
The fact that simpliciter relative-sameness relations are always
subsets of the identity relation can be represented formally as a
kind of functionality condition on relative-sameness relations as
they hold among thing-world pairs:
Thing Functionality: <x,w> RF <y,w> → x = y.
There is no possible world in which (or time at which) distinct things
bear a relative-sameness relation to each other. This condition guar-
antees that we are not the same person (woman, philosopher, et
cetera) as anything other than ourselves. So sameness really is a kind
of identity after all—distinct things are never the same F, simpliciter,
as each other. Relative sameness (simpliciter) does on our view, as on
the standard view, require identity. We differ from the standard view
in holding that transworld (or cross-time) relative sameness does
not require transworld (or cross-time) identity. In saying this, we do
not thereby reject the notion of transworld identity. I am happy to
admit this relation, and without it could not state the difference
between my theory of de re predication and the popular view.
Just as there is only one thing in the actual world that Angel is the
same boat as (viz., herself), she is also not the same boat as more than
one thing in any counterfactual world. This is entailed by thing func-
tionality together with symmetry and transitivity. For if there is a coun-
terfactual world in which Angel is the same boat as both Cherub and
Seraph (given the ways that they are in the counterfactual world and
the way that Angel is in the actual world), then by symmetry, Cherub
is the same boat as Angel. Then by transitivity, Cherub is the same
boat as Seraph. But the distinctness of Cherub and Seraph is ruled
out by thing functionality. We can call this stronger requirement
strong thing functionality and formalize it as follows:
Strong Thing Functionality: <x, w> RF <y, w’> ∧ <x, w> RF <z,
w’> → y = z.
Unlike Lewis’s qualitative-similarity counterpart relation, the rela-

tive-sameness counterpart relation does not allow for a single indi-
vidual to have more than one counterpart in some one possible
world. In other words, unlike Lewis’s counterpart relation, our rela-
tion does not allow for multiple counterparts in some one other
world.
4.4 Relative Sameness and Superlative Qualitative Similarity

We have said that on our proposal, relative-sameness relations can
be construed as a kind of counterpart relation given that (i) trans-
world relative sameness is the important relation for analyses of de
re possibility, and that (ii) transworld sameness does not require
transworld identity. But while relative-sameness counterpart rela-
tions are strongly thing functional, qualitative-similarity counter-
part relations are not.
Even once we fold in Lewis’s later (1971) relativization of the
counterpart relation to salient respects of similarity, which vary
from context to context, the reason that superlative qualitative simi-
larity still allows for multiple counterparts in a world, relativized to
a single respect of similarity, is exemplified by this example: geneti-
cally identical twins in one possible world may tie for most person-
similar with a single thing in some one other possible world. There
might have been genetically identical twins, each qualitatively just
like me, one of whom was born two minutes later than my actual
birth-time, the other two minutes before my actual birth-time. So
there is a possible world in which there are genetically identical
twins who would tie for most person-similar to me. They would
both be my qualitative-similarity personal counterparts.
Because transworld relative sameness, when relativized to a sin-
gle sortal F, does not allow for multiple counterparts in a world, we
may speak of such a thing as the same-F counterpart of x in world
w, if it has one. Since this is a functional expression, we may equiva-
lently speak of x’s counterpart in w (if it has one). This uniqueness
is relativized to a sortal, of course. When relativized to the sortal
boat, for example, we called x’s counterpart in a world w (if it has
one) x’s boat-mate in w. In the abstract, we call x’s counterpart in a
world w (if it has one) x’s “F-mate” in w, when relativized to sortal F.
(So there are different instances of F-mates: boat-mates; person-

mates; matter-mates; etc.)
But whether an F-mate relation holds between objects x and y
depends not only on what sortal F is, but also on the ways that x
and y are in their respective possible worlds. Any “F-mate” rela-
tion is functional when doubly relativized to two worlds: given
the way that x is in w, we may speak of such a thing as the F-mate
of x in w’. In terms of our earlier notation, this would be the <y, w’>
such that <x, w> RF <y, w’>. So as to have a convenient, if cumber-
some, functional expression, we use the functional expression
‘rFw, w’ (x)’:
F-mate notation: ‘rFw,w′(x)’ denotes the <y,w’> such that
<x,w> RF <y,w’>;
In other words: ‘rFw, w′(x)’ denotes the y such that x is the same
F as y, given the ways that x is in w and that
y is in w’.
The expression ‘rFw, w′(x)’denotes x’s F-mate in w’, given the way that
x is in w. We can pronounce this functional expression as x’s Fw,w′
-mate—or just as x’s F-mate in w’, when the relativization of x to w
may be left out without causing confusion; or simply as x’s F-mate,
when the double relativization to both w and w’ can be thus left
out.
Given our entitlement to the use of this F-mate functional expres-
sion, we can restate our provisional version of relative-sameness
counterpart theory of de re possibility using that functional expres-
sion as follows:
(RSC◊) Relative to a sortal F, ‘◊Φ’ is true at a world w just in
case there is a possible world w’ such that ‘Φ(rFw,w’(x))’
is true at w’.12
For the Lewisian counterpart theorist, lack of uniqueness within
a world for superlative qualitative similarity means that the
Lewisian counterpart theorist cannot state his theory in the analo-
gous way, replacing ‘x’s F-mate’ with ‘x’s F-ish counterpart’. On
12
We understand ‘Φ(rFw,w’ (x))’ to be the result of replacing each occurrence of ‘x’
in ‘Φ’ with ‘rFw,w’ (x)’. We assume that Φ has no unnecessarily repeated uses of ‘x’.
Lewis’s view, we cannot speak of the F-ish counterpart of x in w’,

since x may have more than one counterpart there (x might have
had an identical twin). So how must he state his theory? Lewis
discussed two salient choices (Lewis, 1968).
First: an individual might have been Φ just in case there is a possible
world in which some counterpart of that individual is Φ.
Second: an individual might have been Φ just in case there is a
possible world in which every counterpart of that individual is Φ.
The difference is whether in order to be possibly-Φ there just
needs to be some possible world in which one of your counterparts
is Φ, or whether there needs to be some possible world in which all
of your counterparts are Φ.
Lewis chose the first (Lewis 1968, 118f). Given his later (1971)
relativization of the counterpart relation to salient respects of simi-
larity—which could vary from context to context—we can give the
following statement of his qualitative-similarity counterpart theory
of de re possibility. It helps to use a variable-binding predicate-
abstraction operator and to assume that Φ contains no free variables
other than x.
QSC◊: ‘xˆ (◊Φ)’ is true of a thing, relative to respects of simi-
larity F, just in case there is a possible world in which
that thing has some F-ish counterpart of which ‘xˆ Φ’
is true.
5. APPLYING THE IDEA
So how, after all this, do we do what we wanted? What we wanted

was to preserve the coherence of the following:
(15) Angel = Lumber;
(16) Possibly, Angel ≠ Lumber.
Angel and Lumber are one and the same thing since Angel is con-
stituted entirely by Lumber. But they might not have been one and
the same thing since Angel might not have been constituted entirely
by Lumber.
The view of de re modality offered here is that Angel might have
been distinct from Lumber since there is a possible world in which
Angel has a boat-mate and Lumber a wood-mate that are distinct

from each other. The leading idea allows that this is compatible
with the actual identity of Angel and Lumber since the transworld
boat-mate of Angel and the transworld wood-mate of Lumber
need not be identical to Angel.
Once Lewis incorporated multiple counterpart relations into
his semantics, he could say (and did say) just the same thing:
Angel and Lumber are actually identical, but they might have
been distinct since there is a possible world in which there is a
boatish counterpart of Angel that is distinct from some woody coun-
terpart of Lumber (Lewis 1971). It needn’t be that either of these
things is identical to Angel/Lumber (in fact, they couldn’t be,
given Lewis’s world-bound modal metaphysics), so there was no
contradiction.
But we have not yet made it explicit how relative-sameness coun-
terpart theory allows for a single modal operator to be associated
with more than one relative-sameness relation. Analogously, we
have also not yet said how Lewis’s counterpart theory allows for a
single modal operator to be associated with more than one qualita-
tive-similarity relation. In each case, what we need is for distinct
names within the scope of a single modal operator to each be asso-
ciated with their own sortal relativization (in our case) or their own
respect of similarity (in Lewis’s case). Within the scope of a single
possibility operator, we need ‘Angel’ to be associated with the sor-
tal boat and for ‘Lumber’ to be associated with the sortal wood. And
this is so not just for names, but for variables as well. Just as we can
say that Angel and Lumber are identical even though they might
not have been, we can also say that there is a boat x and some wood
y such that x and y are identical but might not have been. If we
explicitly apply RSC◊ to the case of two variables, then we get the
following statement of relative-sameness counterpart theory:
RSC◊2 : Relative to sortals Fx and Fy, ‘◊Φ(x,y)’ is true at a
world w just in case there is a possible world w’ such that
‘Φ(rFxw,w’ (x), rFyw,w’ (y))’ is true at w’.13
13
We understand ‘Φ(rFxw,w’(x),rFyw,w’ (y))’ to be the result of replacing each occur-
rence of ‘x’ in ‘Φ’ with ‘rFxw,w’ (x)’ and each occurrence of ‘y’ in ‘Φ’ with ‘rFyw,w’ (y)’.
We assume that Φ contains no unnecessarily repeated uses of ‘x’ or ‘y’.
The idea here is that Fx might be the sortal boat while Fy is the sortal
wood. This application to two variables can be straightforwardly
extended to explicitly cover indefinitely many variables.
The matter of how to extend this sort of generalization to qualita-
tive-similarity counterpart theory is not quite so straightforward.
(If there is a world in which I and my mother both have multiple
personal counterparts, do we have our various counterparts inde-
pendently? Or do we have our counterparts only as a pair? When
Lewis originally presented his counterpart theory, he chose the
former.) But we will come to this soon enough.
The relevant instance of (RSC◊2) for us right now is this:
Instance: ‘◊ Angel ≠ Lumber’ is true just in case there is a
possible world w’ in which Angel has a boat-mate a
in w’, Lumber has a wood-mate l in w’, and ‘a ≠ l’ is
true at w’.
This condition is witnessed in any world in which Angel has a boat-
mate that is not made up of some wood-mate of Lumber.
6. DEFENDING THE IDEA
The main theoretical advantage of counterpart theory (in my liberal

sense of that term) is that it allows for the attractive combination of
the Identity Thesis, the Accidental Constitution Thesis, and the Necessary
Matter Thesis. The particular instances I have focused on are these:
(IT) Angel, Michael’s boat, is identical to Lumber, the
wood that makes it up;
(ACT) Angel might not have been made up of Lumber, the
wood that actually makes it up;
(NMT) Lumber cannot but have been made up of the wood
that actually makes up Angel.
These jointly committed us to the following statement of contingent
identity:
(CI) Angel = Lumber, but ◊ Angel ≠ Lumber.
The claim of possible distinctness (ʹ◊ Angel ≠ Lumber’) is true on
our view not because there is some possible world in which iden-
tical objects are distinct, but rather because the context depend-
ency of modal predicates prevents co-referential names from being
intersubstitutable in modal contexts—no matter whether they
flank the identity symbol or not. Despite this, we can coherently
affirm the necessity of identity by giving it a nonmodal, exten-
sional expression.
(NI) x = y ⇒ there is no possible world in which x ≠ y.
While both of the counterpart theories we have discussed have the
advantage of preserving (it), (act), and (nmt) there are nonetheless
important differences between them. We will proceed, then, with
our defense of relative-sameness counterpart theory by explaining
its advantages over qualitative-similarity counterpart theory. We
will subsequently turn to a discussion of Leibniz’s Law and explain
why, although one version of Leibniz’s Law is not validated by our
theory, there is a superior version of Leibniz’s Law that is. The paper
concludes with a summary and in the end suggests why relative-
sameness counterpart theory provides for a more satisfactory
response than qualitative-similarity counterpart theory does to
Kripke’s famous Humphrey objection.
6.1 Comparison with Superlative Qualitative Similarity

Given the equivalence between ‘¬◊¬Φ’ and ‘◽Φ’, relative-sameness
counterpart theory and qualitative-similarity theories of de re neces-
sity may be stated as follows.
RSC◽: Relative to sortal F, ‘◽Φ’ is true at a world w just in
case every possible world w’ is one in which the fol-
lowing is true: ‘Φ(x’s Fw,w′ -mate)’.14
QSC◽: Relative to respect of similarity F, ‘xˆ◽Φ’ is true of a
thing just in case every possible world is one in which
xˆΦ is true of all of that thing’s F-ish counterparts.
Lewis remarked that his counterpart theory had the consequence
that everything exists necessarily (Lewis 1968, 119). To exist neces-
14
For simplicity of exposition, we are temporarily prescinding from the need for
distinct variables to have distinct sortal-relativizations.
sarily according to QSC◽ is for every possible world to be one in

which all of your counterparts exist. In possible worlds where you
have no personal counterparts, for example—such as those contain-
ing only barren deserts not populated by life of any kind—you will
have no personal counterparts that do not exist. Even though there
might have been no life at all, you exist necessarily.
What is perhaps worse, since there is no possible world in which
you have a personal counterpart that was never alive, you necessar-
ily were alive at some point in time even though it might have been
that nothing was ever alive. Since you are a person that was once
alive in every possible world in which you exist (I may assume for
the sake of argument) while there are possible worlds in which
nothing was ever alive, the following is true according to qualita-
tive-similarity counterpart theory:
(17) ◊¬∃x x was once alive ∧ ◽ you were once alive.
Why does our version of counterpart theory allow for the falsity of
(17), for things to have lived merely contingently? Worlds in which
nothing is the same person as you are not worlds in which the fol-
lowing is true: the same person as you was once alive. In other
words, if the functional expression ‘your person-mate’ does not
denote anything that exists in some world w, then it does not denote
anything that is in the extension of the predicate ‘was once alive’ in
w. According to relative-sameness counterpart theory, you have
been alive but it is not necessary that you were ever once alive. This,
of course, is as it should be.
The problem of necessary existence reflects a more general prob-
lem. As Allen Hazen in essence pointed out, Lewis’s theory makes
no room for the difference between a property’s being among your
essential properties and its being necessary that you have that prop-
erty (Hazen 1979, 317–29). But there is indeed a difference. The con-
trast is displayed by the following schematic statements, where ‘E’
is the existence predicate.
Essentiality: ◽(Ex → Φx): ‘x is essentially Φ’.
Necessity: ◽Φx: ‘x is necessarily Φ’.
The difference is clear: I have a property essentially when I could not
have existed without having the property. It is necessary that I have
a property when it can only have been that I have the property.
Take the property of being born a human. I was essentially born

human. If a thing wasn’t born human, it most definitely is not me.
Any counterpart theorist analyzes this as follows—whether her
counterpart relation be sameness based or similarity based.
There is no possible world in which I have a counterpart that wasn’t born
human.
For the similarity-based counterpart theorist, who allows for multiple

counterparts, this is equivalent to his most plausible analysis of what
it is for it to be necessary that I was born human. According to him, its
being necessary that I was born human is analyzed as follows.
Every possible world is one in which all of my counterparts there were
born human.
But the similarity-based counterpart theorist is clearly wrong on

this score. I was essentially born human—if a thing wasn’t born
human it definitely is not me. It is not possible that I was born a
frog or a bird or any nonhuman animal. But it is not necessary that I
was born human. I might never have been born, so it might have
been that I was never around to be human at all. There might only
ever have been barren deserts, in which case it would never have
been that I was born human.
The sameness-based counterpart theorist, whose counterpart
relation does not allow for multiple counterparts, makes room for
the distinction since the above conditions are not equivalent to the
following gloss on her most plausible analysis of its being necessary
that I was born human:
Every possible world is one in which I have a counterpart that was born
human.
As we said, similarity-based counterpart theory never makes room

for the distinction between your having a property essentially and
it is being necessary that you have it. In what may be the worst case,
this includes existence. According to the similarity-based counter-
part theorist, our having existence among our essential properties is
equivalent to its being necessary that we exist.
Let me now develop another of Hazen’s objections to superlative-
qualitative-similarity counterpart theory. Before I begin I must say
something about the analysis of polyadic predicates in the scope of
modal operators. We have not yet been explicit about how the simi-
larity-based counterpart theorist analyzes modal claims with poly-

adic predicates. In the case of modal claims involving only the
monadic predicate ‘is a frog’, we said that the similarity-based coun-
terpart theorist analyzes ‘◽(a is a frog)’ as being true just in case
every possible world is one in which all of a’s counterparts there are
frogs, and ‘◊(a is a frog)’ as being true just in case some possible
world is one in which one of a’s counterparts is a frog. With rela-
tional expressions, such as ‘being friends with’, a similarity-based
counterpart theorist might want to say that “◽(a is friends with b)”
is to be analyzed as being true just in case every possible world is
one in which every counterpart of a is friends with every counter-
part of b. The problem with this analysis is that it renders necessity
claims about relations too easy to falsify. There is some possible
world in which Abel has two counterparts, Abel1 and Abel2, who tie
for most similar with Abel and in which Adam has two counter-
parts, Adam1 and Adam2, who each tie for most similar to him.
There is some such possible world in which Adam1 is father of Abel1
but not of Abel2 and in which Adam2 is father of Abel2 but not of
Abel1. Given the aforementioned analysis, the existence of this pos-
sible world suffices to render it not essential to Abel that he have
Adam as his father. But that should not suffice. We should consider
rather the counterparts of Adam and Abel taken as a pair. Although
Adam has both Adam1 and Adam2 as counterparts and likewise for
Abel, the pair Adam–Abel has only the pairs Adam1–Abel1 and
Adam2–Abel2 as counterparts—not the two other mix-and-matches
of these.
For this simple case, then, the similarity-based counterpart theo-
rist will say that ‘xˆ yˆ◽Rxy’ is true of a pair <a, b> just in case in every
possible world, every counterpart <a’, b’> of the pair <a, b> is one of
which ‘x̂yˆ◽Rxy’ is true. As before, each argument place may be
associated with its own respect of similarity. It might be that <a’, b’>
is a counterpart of the pair <a, b> only if a’ is a bodily counterpart of
a, and b’ is a personal counterpart of b.
Now Hazen remarked that the following inference form is not
valid on Lewis’s theory, even when Rab is an atomic sentence other
than identity:
(18) ◽Rab,
therefore
(19) ◽∃xRax.
Let us consider a concrete example.
(20) ◽ Adam is Abel’s father, therefore
(21) ◽ Adam is someone’s father.
Lewis analyses the first as true just in case every possible world is
one in which every counterpart <Adam’, Abel’> of <Adam, Abel> is
such that Adam’ is father of Abel’. This is to say that there is no pos-
sible world in which the pair of Adam and Abel have a counterpart
pair with the first member of the pair not being father of the second.
But there are possible worlds in which Adam has a counterpart but
Abel does not, because there are possible worlds in which Adam has
no children at all. But these possible worlds are effectively ignored
for the purposes of evaluating (20). These are possible worlds in
which the pair of Adam and Abel have no pair as counterpart. So
these are possible worlds in which it is vacuously true that every
counterpart of the pair of Adam and Abel have a first member that
is father of the second member. But for the purposes of evaluating
(21), we do not ignore such possible worlds. ‘Adam is someone’s
father’ is analyzed as true on Lewis’s analysis just in case every pos-
sible world is one in which all of Adam’s counterparts have some
child or other. This is false on Lewis’s analysis, since there are worlds
in which Adam has a counterpart who has no child at all.
But the inference should hold good, at least in the case of logi-
cally simple relations, as in (18)–(21) above.15 If it is necessary that a
bear a certain logically simple relation to b, then surely it is neces-
sary that a bear that relation to something or other. The failure of
this inference to hold good within similarity-based counterpart the-
ory stems at bottom from its failure to distinguish essence and
necessity. Since (21) is false and the inference from (20) to (21) is a
good one, (20) must be false. Here’s why it is false. There are possi-
15
Given that the atomic predication ‘F( f (x))’ is not true—discounting identity—
whenever ‘f (x)’ has no denotation, the negation ‘¬F( f (x))’ is true whenever f (x) has
no denotation. For this reason, the truth of ‘◽¬R(x, y)’ does not guarantee the truth of
‘◽∃y¬R(x, y)’. This is as it should be. It is necessary that I am not related to you, but it
does not follow that it is necessary that there is someone to whom I'm not related.
There are possible worlds in which I and my ancestors are the only humans that have
ever existed.
ble worlds in which Abel was never born. So, contra (20), there are
possible worlds that don’t contain both counterparts of Adam and
Abel with the first being father of the second.
It is, though, essential to Abel that he have Adam as a father.
Nothing could be Abel if it didn’t have Adam as a father. For the
similarity-based counterpart theorist, this is equivalent to its being
necessary that Abel have Adam as a father. This in turn is equiva-
lent to its being essential to Adam that he be father of Abel. But
because there are possible worlds in which Adam has a counterpart
who has no children at all, it is not necessary that Adam be father to
some child or other.
We want to be able to make the distinctions—and we can—
between the pairs of sentences (22)–(23) through (26)–(27).
(22) ◽(Abel exists → Abel’s father is Adam), (true)
therefore
(23) ◽(Abel exists → ∃x Abel’s father is x). (true)
This is the sound argument from ‘Abel essentially has Adam as a
father’ to ‘Abel essentially has someone as a father’. (22) is true on
our theory because there’s no possible world in which there’s some-
one who is the same person as Abel but who does not have some-
one who’s the same person as Adam for a father. (23) is then
guaranteed to be true on our theory since it is then guaranteed that
there is no possible world in which there’s someone that is the same
person as Abel but who does not have (in that world) anyone for a
father.16
(24) ◽(Abel’s father is Adam), (false)
therefore,
(25) ◽(∃x Abel’s father is x). (false)
This is the valid but unsound argument from ‘It is necessary that
Abel’s father be Adam’ to ‘It is necessary that Abel have someone as
a father’. The latter sentence, (25), is false since there are possible
16
We have not given truth clauses for all of the logical symbols. The details are
slightly complex once spelled out fully. I refer readers to Fara (2008) for details.
worlds in which nothing exists that is the same person as Adam or

as Abel.
(26) ◽(Adam exists → Adam is father of Abel), (false)
therefore
(27) ◽(Adam exists → ∃x Adam is father of x). (false)
This is the valid but unsound argument from ‘Adam is essen-
tially Abel’s father’ to ‘Adam is essentially the father of someone’.
(27) is false on our theory because there is a possible world in
which someone who is the same person as Adam fathers no-one at
all. (26) is then false on our theory because such possible worlds
are ones in which there’s someone that is the same person as Adam
but who, given that Adam is not the father of anyone, is not the
father of anyone who is the same person as Abel. Adam might
have fathered only Cain. (There is a possible world in which some-
one who is the same person as Adam fathers only one son, and
that son is the same person as Cain, and therefore not the same
person as Abel.)
The difference in truth value between and the pairs (22) and (24),
(23) and (25), reflects the fact that sameness-based counterpart the-
ory, unlike its qualitative-similarity rival, makes a distinction
between essence and necessity. Moreover, each of these pairs repre-
sents one of two argument forms that are valid according to same-
ness-based counterpart theory. Those are the forms:
(28) ◽(Ea → Rab) |= ◽(Ea → ∃xRax);
and
(29) ◽Rab |= ◽∃xRax.
6.2 Substitution Failures? No Problem

One could complain, against the theory presented here, that it does
not validate Leibniz’s Law. Once the complaint is accurately stated,
however, it should not worry us too much. There are two different
versions of Leibniz’s Law: (I) the sentence-schema version, in which
a
‘Φ’ is a sentence-schema to be replaced by a sentence, and ‘Φb’ is
any sentence that results if ‘b’ is put in for ‘a’ in any or all occur-
rences of ‘a’; and (II) the property version.
a
Leibniz Law I (a = b → (Φ → Φb)),
Leibniz Law II ∀x∀y(x = y → ∀z(z is a property → (x has z →
y has z))).
These principles can be easily restated to cover relations of
indefinitely large adicity.
The first version of Leibniz’s Law is properly called “the principle
of Substitutivity of Identicals” rather than Leibniz’s Law. The
two may come apart since it may be that ‘a’ does not occur in a ref-
erential position in ‘Φa’. If we say “a is Φ,” we may not be attribut-
ing a genuine property to the thing named by ‘a’. Substitutivity of
Identicals simply is not a valid principle. A classic counterexample
is Quine’s: ‘Giorgione was so-called because of his size’. Whether a
sentence ‘____was so-called because of his size’ is true once we put
in a name for the blank depends not only on what the referent of
that name is like, but also on which name we have used to refer to
it. That relative-sameness counterpart theory invalidates the substi-
tution principle is therefore not per se a problem with it. Moreover,
we have an explanation for why it fails in cases of attributions using
modal predicates. Our case is ‘____might have been constituted by
different wood’. Whether this sentence is true when we put in
a name for the thing that is Michael’s boat—e.g., ‘Angel’ or
‘Lumber’—depends on which sortal is associated with the name
that is used to refer to the thing in question.
What does it mean to say that different names for a thing can be
associated with different sortals? Let me give an example. We have
a woman, Larissa Seiger, who is both a ballerina and an elementary-
school teacher. Sometimes people ask, “What is Mrs. Seiger?” In
those situations it is typically appropriate to answer by saying that
she’s a teacher (my former teacher, my daughter’s teacher, et cetera).
Sometimes people ask “What is Madame Larissa?” In those situa-
tions it is typically more appropriate to answer by saying that she’s
a ballerina (the principle ballerina in company X, Odette in Swan
Lake, et cetera). This is to say that different ones of her names are
associated with different sortals—teacher or ballerina.
So why is it that ‘Angel’ cannot always be substituted with ‘Lum-
ber’ in the modal context in question? Why does this failure not
falsify the property version of Leibniz’s law? One could say, with
Allan Gibbard, that ‘____might have been constituted by differ-
ent wood’ does not express a property (Gibbard 1975, 201). Or one
could say instead, with Harold Noonan, that the predicate
‘____might have been constituted by different wood’ does express
a property, but it is context dependent. Which property it expresses
depends on which name is substituted for the blank (Noonan 1991,
188). I am inclined to side with Noonan, but his view would have to
be modified so as to account for substitution failures with variables
rather than with names. The satisfiability of the following illustrates
this point:
(30) There is a boat x and some wood y such that x = y and x,
but not y, is possibly not made up of z.
7. CONCLUSION
The leading idea here was that relative sameness does not require
identity. This was subsequently qualified by saying that transworld
relative sameness does not require transworld identity. Michael’s
boat actually had one third of its planks replaced by the time it was
rowed ashore, but if one half of its planks had been replaced instead,
it would still have been the same boat that was rowed ashore in that
event, even though the boat rowed ashore in the counterfactual
world is not (absolutely) identical to the boat rowed ashore in the
actual world. Nonetheless, intraworld sameness—equivalent to
what I called sameness simpliciter—does require identity. Whenever
x is the same boat as y, x and y are identical.
This was not to say that we never have identity between things in
different possible worlds. In other words, it was not to say that indi-
viduals are “world bound.” Much less was it to deny that there is
such a relation as transworld identity. In that sense, I rejected radical
modal realism since I think that things in one possible world typi-
cally are identical to something in some other possible world. The
very thing that is actually Michael’s boat, and also some wood,
exists in many possible worlds in which it is not a boat at all.
Given the distinction between transworld sameness and trans-
world identity, the following analyses of de re possibility come
apart.
(A) Relative to a sortal F, x is possibly Φ just in case there is a

possible world in which something identical to x is Φ.
(B) Relative to a sortal F, x is possibly Φ just in case there is a
possible world in which something that is the same F as
x is Φ.
The second of these analyses is the one that I favor. Whether it is
true to say that x is possibly Φ depends on which sortal we are asso-
ciating it with. For that reason we said that ‘being possibly Φ’ does
not stand univocally for some one property any more than ‘being
appropriately called ‘Mrs. Seiger’ ‘ does. Whether or not Mrs. Seiger
is appropriately called “Mrs. Seiger” depends on the situation she
is in, and on which of the many sortals that she falls under is most
relevant for how she is referred to or addressed in that situation. At
the dinner table, her daughter should address her as “mommy,”
definitely not as “Mrs. Seiger.”
One who accepts (B) but rejects (A) is a counterpart theorist in the
sense that she thinks that the transworld relation that is involved in
the analysis of de re modal claims is something other than identity.
This does not mean that her counterpart theory is like Lewis’s in every
respect. Like Lewis, I think that whether or not x in world w is a coun-
terpart of y in world w’ depends on what x is like in w and on what y
is like in w’. But there are many important respects of disagreement.
First, Lewis thought—but I do not think—that if x is in world w and
y is in a distinct world w’, then x and y are never identical. Lewis there-
fore could—but I cannot—think of the counterpart relation as not need-
ing relativization to possible worlds. He said that x is an F-ish counterpart
of y (simpliciter) just in case x and y are such-and-such a way simpliciter.
I must rather say that x in w is an F-mate of y in w’ just in case x and y
are such-and-such a way, relative to w and w’ respectively.
Second, Lewis also thought—but I do not think—that whether x in
w is a counterpart of y in w’ depends only on comparative similari-
ties. For him, in order for x to be a counterpart of y it just has to be that
x is similar to y and no less similar to y than anything else in x’s world,
w, is. Moreover, Lewis thought that the relevant respects of similarity
were purely qualitative—they never involved any other particular
individuals. I think rather that whether x in w is a counterpart of y in
w’ depends on sort-relative sameness. Sort-relative sameness is not a
comparative relation. There is no sense in saying that x is more
the same horse as y than z is. Furthermore, sameness relations are not
purely qualitative. The boat that Michael rows ashore in the actual
world might be the same boat as the one he rows ashore in some
counterfactual world even if in that counterfactual world the boat
that Gabriel rows ashore is qualitatively just like the one that Michael
actually rows ashore—and even if Michael is, in that counterfactual
world, qualitatively just like Gabriel is in the actual one.
But like Lewis, I am able to maintain the sensible view that
Michael’s boat and the matter that composes it are one, even though
they might not have been, since the boat might have been com-
posed of different matter, while the matter could not have been
composed of different matter. In extensional possible-worlds talk in
terms of relativized counterpart relations, Michael’s boat and the
matter it is composed of are identical but they may have something
in another world as a boat-counterpart (a “boat-mate”) without
having that thing as a matter-counterpart (a “matter-mate”).
We saw that a counterpart theorist gets into trouble if his counter-
part relations allow for more than one counterpart in some one other
possible world. This rendered relative-sameness counterpart theory
far more attractive than superlative-qualitative-similarity counter-
part theory. Unlike superlative qualitative similarity, sameness is a
weak equivalence relation. Moreover, it is functional, albeit partial.
Making explicit the relativization to sortal F and double world rela-
tivizations to w and w’, we expressed this latter fact by saying that
x has at most one Fw,w’-mate. This allowed us to sharpen up (B)
above as follows.
RSC◊: When relativized to a sortal F, ‘◊Φ(x)’ is true at a
world w just in case there is a possible world w’ in
which ‘Φ(x’s Fw, w′-mate)’ is true.
Treating the functional expression on the right-hand side of this
truth clause as a singular term gave us this statement of truth condi-
tions for claims of de re necessity.
RSC□: When relativized to a sortal F, ‘□Φ(x)’ is true at a
world w just in case every possible world w’ is one in
which ‘Φ(x’s Fw,w’-mate)’ is true.
A counterpart theorist without a functional counterpart relation is
not entitled to the above analyses of de re modal claims. Instead,
he must quantify over a thing’s counterparts in a possible world.

Lewis chose to existentially quantify in the case of possibility and
therefore to universally quantify in the case of necessity:
QSC◊: Relative to respects of similarity F⃗, ‘◊Φ(x⃗)’ is true just
in case some possible world is one in some F⃗-ish coun-
terpart of x⃗ is Φ.
We use ‘x⃗’ to stand for the n-tuple of constants and variables in ‘Φ’
and we use ‘F⃗’ to stand for the series of the various respects of
similarity associated, respectively, with each of these constants
and variables.17
QSC□: Relative to respect of similarity F⃗, ‘□Φ(x⃗)’ is true just
in case every possible world is one in which every F⃗-
ish counterpart of x⃗ is Φ.18
The difference between the two counterpart relations enables the
relative-sameness counterpart theorist to make a distinction between
essence and necessity where the qualitative-similarity counterpart
theorist cannot. There are other criticisms of qualitative-similarity
counterpart theory that also turn on its allowance of multiple coun-
terparts within a single world. I have in mind in particular those that
criticize it for being unable to give a plausible semantics for a lan-
guage enriched with the modal operator ‘actually’.19
And what of another famous objection to counterpart theory?
Kripke complained about Lewis’s counterpart theory that according
to it, whether or not Hubert Humphrey might have won the election
does not depend on whether Humphrey himself won the election in
some other possible world, but rather on whether someone else, a
“counterpart,” won the election in some other possible world (Kripke
1980, nt. 13, p. 41). But while Humphrey does care whether he might
have won the election, he has no concern whatsoever whether some
different person—the counterpart—won the election. The relative-
17
The qualitative-similarity counterpart theorist who gives up world-boundedness
needs to relativize his counterpart relation to pairs of worlds, just as we have done.
18
Murali Ramachandran (1998) has formalized Lewis’s version of sortal-relative
counterpart theory by devising a “translation scheme” of the kind we find in Lewis
(1968).
19
On this see Hazen (1979, 330) and also Michael Fara and Timothy Williamson
(Fara and Williamson 2005).
sameness counterpart theorist need not say anything as radical as this,

however. The relative-sameness counterpart theorist says that win-
ning the election is a possibility for Humphrey since there is a possible
world in which he, Humphrey himself, that man, does win the elec-
tion. For by this we mean that there is a possible world in which some-
one who’s the same man as Humphrey does win the election. Whether
that man be identical to Humphrey or not, though, he will not be
someone else. To be someone else would be to be a different man.
Princeton University
I intend for this paper to be an informal companion of my ’Relative-Sameness

Counterpart Theory’ (Fara 2008). I’m grateful for discussion to Michael Graff
Fara, Gilbert Harman, Ned Markosian, Theodore Sider, Robert Stalnaker,
Jason Stanley, Judith Jarvis Thomson, and Timothy Williamson. I’m also
grateful for discussion to participants of the New York Corridor Workshop as
well as audiences at Rutgers University, the 2007 BIRS Mathematical Methods
in Philosophy Conference, and the 2007 Arizona Ontology Conference, as
well as to Berit Brogaard, my commentator on that last occasion. I’m also
indebted to Allan Gibbard, whose (1975) work greatly influenced me and first
got me to see that there could be any merit to the idea of contingent identity.
REFERENCES
Alston, William and Bennett, Jonathan (1984), ‘Identity and Cardinality:
Geach and Frege’, Philosophical Review 93(4): 553–68.
Butler, Joseph, D.C.L. (1844), The Analogy of Religion, Samuel Hallifax, edi-
tor, Robert Carter & Brothers, New York.
Chisholm, Roderick M. (1973), ‘Parts as Essential to their Wholes’, Review
of Metaphysics 26: 581–603.
—— (1976), Person and Object, Allen and Unwin, London.
Fara, Delia Graff (2008), ‘Relative-Sameness Counterpart Theory’, Review
of Symbolic Logic 1: 167–89.
Fara, Michael and Williamson, Timothy (2005), ‘Counterparts and Actu-
ality’, Mind 114: 1–30.
Geach, Peter (1962), Reference and Generality, Cornell University Press, Ithaca.
——, ‘Identity’, Review of Metaphysics 21: 3–12. Reprinted in Geach 1972,
pp. 238–47.
——, Peter (1972), Logic Matters, Blackwell, Oxford.
Gibbard, Allan (1975), ‘Contingent Identity’, Journal of Philosophical Logic
4: 187–221.
Gupta, Anil (1980), The Logic of Common Nouns, Yale University Press,
New Haven.
Hawley, Katherine (2001), How Things Persist, Clarendon, Oxford.
Hazen, Allen (1979), ‘Counterpart-Theoretic Semantics for Modal Logic’,
Journal of Philosophy 76(6): 319–38.
Johnston, Mark (1992), ‘Constitution is Not Identity’, Mind 101(401): 89–106.
Jubien, Michael (1993), Ontology, Modality, and the Fallacy of Reference,
Cambridge University Press, Cambridge.
—— (2001), ‘Thinking About Things’, Philosophical Perspectives 15: Meta-
physics pp. 1–15.
Kripke, Saul (1963), ‘Semantical Considerations on Modal Logic’, in Pro-
ceedings of a Colloquium on Modal and Many-Valued Logics, Acta
Philosophica Fennica, Helsinki, pp. 83–94.
—— (1972), ‘Naming and Necessity’, in D. Davidson and G. Harman,
eds., Semantics of Natural Language, Reidel, Dordrecht, pp. 253–355.
Reprinted as Kripke (1980).
—— (1980), Naming and Necessity, Harvard University Press, Cambridge,
MA. Reprinted, with an added preface, from Kripke (1972).
Lewis, David (1968), ‘Counterpart Theory and Quantified Modal Logic’,
Journal of Philosophy 65(5): 113–26.
—— (1971), ‘Counterparts of Persons and Their Bodies’, Journal of Philoso-
phy 68(7): 203–11.
—— (1986), On the Plurality of Worlds, Blackwell.
Noonan, Harold (1991), ‘Indeterminate Identity, Contingent Identity and
Abelardian Predicates’, The Philosophical Quarterly 41(163): 183–193.
—— (1993), ‘Constitution is Identity’, Mind 102: 133–46.
Perry, John (1970), ‘The Same F’, Philosophical Review 79: 181–200.
Ramachandran, Murali (1998), ‘Sortal Modal Logic and Counterpart The-
ory’, Australasian Journal of Philosophy 76: 553–65.
Sider Theodore (1996), ‘All the World’s a Stage’, Australasian Journal of
Philosophy 74(3): 433–53.
—— (1999), ‘Critical Study of Michael Jubien’s Ontology, Modality and the
Fallacy of Reference’, NOÛS 33: 284–94.
——(2001), Four-Dimensionalism: An Ontology of Persistence and Time,
Clarenden Press, Oxford.
Stalnaker, Robert (1986), ‘Counterparts and Identity’, Midwest Studies in
Philosophy 11: 121–40. Page references are to reprinted version (Stal-
naker 2003).
—— (2003), Ways a World Might Be, Oxford University Press, Oxford.
2. Reflections on Counterpart Theory
Allen Hazen
David Lewis’s project of analyzing modality de re in terms of a rela-

tionship of counterparthood between entities from different worlds
has given rise to many applications, some more closely related to
Lewis’s original intentions than others. At one extreme the addi-
tional degree of freedom allowed by a counterpart-theoretic inter-
pretation of quantified intensional languages can be exploited
for purely technical ends: Stalnaker 1995 proves the independence
of a certain schema from certain axioms by use of a counterpart-
theoretic model satisfying formal constraints which have no
Lewisian philosophical motivation. Other applications are closer to
Lewis’s ideas.
Delia Graff Fara has recently developed the suggestions in
Lewis 1971 to give a general account of a variety of philosophical
puzzles about identity. On this proposal there would be more
than one counterpart relation, and terms (constants and varia-
bles) would each have a counterpart relation associated with
them, with no requirement that terms assigned the same object
as designatum have the same associated counterpart relation.
The truth values of modalized formulas like ◊Fa would depend
on the properties of those objects, in various possible worlds,
related, by the counterpart relation associated with a, to the
object assigned to a. Such a semantics would allow the satisfi-
ability of formulas like a=b & ~◻a=b: the first conjunct would
require that the two terms, a and b, be assigned the same object
as designatum, but if the associated counterpart relations were
different this single object could be related to distinct objects in
another world.
In Lewis’s original formulation, the counterpart relation was not
required to be symmetric or transitive, an object in one world was
42 | Allen Hazen
allowed to have multiple counterparts in some other world, and

distinct objects in one world were allowed to share a counterpart in
another. This is made plausible by his characterization of the coun-
terpart relation in terms of comparative similarity. Many writers
have felt, however, that it generates an implausibly weak modal
logic, failing to validate a variety of intuitively valid formulas and
inferences. (This is, of course, an objection that presupposes a choice
of interpretation for the “modal” operators. The logic generated by
Lewis’s counterpart theory is too weak as a logic of metaphysical or
logical necessity and possibility: one can consistently allow this
while claiming that the full generality of the counterpart relation is
desirable for other applications. It seems to me at least initially
plausible that a semantic analysis of fiction might say that Charles
Martel and Charlemagne, though distinct entities in the actual
world, have a common counterpart in a world of Medieval French
chanson de geste.) To obtain the “standard” quantified modal logic
for the logical or metaphysical modalities—a logic whose simplest
model theory involves worlds with overlapping domains—on a
counterpart-theoretic basis, we have to add non-Lewisian condi-
tions on the counterpart relation: we require it to be an equivalence
relation (reflexive, symmetric, transitive), with each of its cells hav-
ing at most one member in the domain of a possible world.
Fara explicitly disavows an analysis of her counterpart relations
in terms of similarity, and imposes the equivalence and single rep-
resentative per world restrictions on each counterpart relation. The
resulting semantics is very similar to that of Gupta 1980: the exten-
sion of one of her counterpart relations and the corresponding sort
in Gupta’s account can be defined from each other by elementary
set theoretic means. It seems to capture many intuitions about iden-
tity in modal or temporal contexts.
In the remainder of this essay I would like to look at modal
semantics using a similarity-based counterpart relation. I do not
mean thereby to criticize Fara’s work: there are different analytic
tasks, and I think the issues about identity that she (and Gupta)
address are in a sense orthogonal to those which motivate the
project of defining counterparthood in terms of (something like
some sort of) similarity. (Fara and Gupta’s project is descriptive,
aimed at capturing and systematizing our “native speakers” intui-
tions. Lewis’s is more foundational or metaphysical, and so eschews
Reflections on Counterpart Theory | 43
metaphysical assumptions he found dubious, even if the validity

concept for a quantified modal language derived from such assump-
tions does not jar our intuitions.)
II
Let us consider some very general kind of alethic modality: perhaps

what some philosophers have called metaphysical necessity and pos-
sibility, perhaps (if this is not the same thing) logical necessity and
possibility. There is a certain amount of vagueness here: I doubt that
the philosophers who have used and discussed these notions have
said enough to single out a unique, precise, sense for their modal
operators. I believe, however, that there are meaningful notions in
the neighborhood, that at least some of the arguments that have
been formulated in terms of metaphysical or logical modality make
sense, even if the precise sense of their premises and conclusion
may need a bit of further specification. Moreover, vague as it is, the
sense that modal reasoning makes is one that claims truth values.
Modal reasoning doesn’t have the status of a comprehensible fairy
tale, one that can be rewritten with different endings, with the
choice of ending responsive solely to aesthetic concerns: rather,
some modal statements are right and others wrong (even if others
are neither because they are in need of further clarification). And
the truth values of modal statements are not isolated: they can have
implications for other theoretical or practical questions. (A baseball
fan would be a fool to bet that a team will win the pennant when the
distribution of wins and losses in the league has made it “mathe-
matically impossible” for the team to finish first: the modal claim
supports a practically relevant prediction!) There may be other
ways of explaining the status modal statements have, but the obvi-
ous one is realism: modal sentences are among those which have a
real subject matter, and are true or false in virtue of the facts about
this subject matter.
Modal statements, in other words, have ontological commit-
ments. Now, one of the useful techniques Quine has taught us is
that clarity in the evaluation of ontological commitment is often
best served by paraphrasing other kinds of discourse in the lan-
guage of quantifiers. (I think Quine would be happy to see his
44 | Allen Hazen
“criterion of ontological commitment” construed pragmatically in

this manner. I am by no means certain that his chosen yardstick—
classical first-order logic—is always the best one to use: questions
about constructivity in the philosophy of mathematics, for exam-
ple, may depend on a subtler analysis than the simple determination
of what domain the quantified variables range over. Still, examina-
tion of explicit first-order formulations often seems illuminating.)
So I think we should take talk of possible worlds seriously.
There is, I think, compelling reason to take modal talk to be
“about” possible worlds. As a logician, I am most impressed by the
parallels between the rules of logical inference with modal opera-
tors and those for quantifiers: modal logic (particularly alethic
modal logic, particularly S5) has a very simple and convenient nat-
ural deduction formulation, and the formal rules for the introduc-
tion and elimination of the necessity (resp. possibility) operator are,
formally, simplified analogues of those for the universal (resp. exis-
tential) quantifier. (Simplified by doing without the machinery of
variables.) If it walks like a quantifier and quacks like a quanti-
fier . . . I find it very hard not to think of necessity and possibility
operators as (lightly disguised) quantifiers. The logical analogy is
not perfect: if we compare a formal language of quantified modal
logic with a first-order language with quantified variables for pos-
sible worlds, interpreting them in the same models, we find that the
modal language is expressively weaker, in the sense that every sen-
tence in the modal language can be translated into the possible-
worlds language, but some sentences about possible worlds are not
(equivalent to) the translations of any modal sentences. Lewis and
others, however, have pointed out many examples of informal, nat-
ural-language, modal locutions that also do not translate into the
standard modal language with quantified variables for individuals,
but can be expressed in the worlds language: ordinary modal dis-
course has a richness that may come closer to that of a formal lan-
guage with quantified variables ranging over worlds! Indeed, the
English expression of a piece of modal reasoning can contain locu-
tions like “in that case” (or “then” in a non-temporal sense) func-
tioning in ways similar to the English pronouns that corresponding
to formal variables.
Over the years there has been a fair amount of argle-bargle among
philosophers as to the nature of possible worlds: what kind of entity
are they? Lewis’s own view (which he himself dubbed “extreme”

modal realism) was that the actual world—the world we find our-
selves in, that we explore—was a typical example of a possible
world, and that there was no reason to think that other worlds dif-
fered from it in fundamental metaphysical ways: in particular, no
reason to think of them as abstract in any sense in which our own
surroundings are concrete. Others preferred one or another concep-
tion of non-actual worlds as abstract entities: ersatz worlds in the
terminology of Lewis 1986. I would argue that modal thinking fun-
damentally involves pretty much the same sort of abstraction as
mathematical thinking: constructing a scenario to show that some
proposition is not necessarily true seems fundamentally the same
sort of intellectual activity as defining a weird space to refute a con-
jecture in topology. Epistemology and metaphysics ought to have
something to do with one another, and it is tempting to think that
subject matters we find out about in similar ways are of similar
nature. My own preferred view, then, would be a variant of what
Lewis 1986 calls linguistic ersatzism, with certain set-theoretic objects
(viz, models in the sense of standard model theory) taking the place
of his concrete possible worlds. In the bigger philosophical picture,
I think such a view would represent progress in that, though onto-
logical and epistemological problems about mathematics would
remain, we would be rid of the separate ontological and epistemo-
logical perplexities faced by Lewis’s modal concretism. This is, how-
ever, an issue independent of the problem I wish to discuss here: I
think what follows is equally applicable on my view or on Lewis’s
as to the nature of non-actual worlds.
In one respect, however, Lewis was an ersatzist. Our modal rea-
soning is often de re, and the simplest way to accommodate this in a
possible-worlds framework is to allow the very same people (and
other entities) who have one set of characteristics in one world to
exist with different characteristics in other worlds. There is an obvi-
ous analogy, here, with temporal language: we make de re past and
future tense statements, and their truth conditions rest on having
people and things continue to exist, but with changing properties,
over time. (The analogy, of course, has been successful and fruitful
for semantic theory: David Lewis and Delia Graff Fara are only two
of the many writers who have given parallel accounts, using the
same or at least closely analogous technical concepts, of modal and
46 | Allen Hazen
temporal semantics.) The criteria of personal or material object

identity over time—memory, spatio-temporal continuity, assorted
causal relations—just don’t apply from one possible world to
another. Whether identity over time can be fully analysed or defined
in terms of spatio-temporal continuity, etc. is of course a controver-
sial issue. What should not be controversial is that there is a close
conceptual connection: assertions that an object at one time is iden-
tical to one at another would be pointless if identity wasn’t related
to these other things. Our understanding of identity through time
depends on these connections; if we didn’t understand them we
wouldn’t have the concepts of persons or material objects that we
do. A theory of possible worlds simply postulating the literal iden-
tity of inhabitants of different worlds while admitting that this
identity has no connections with “criteria” of identity would make
identity (in the transworld case) into a completely “opaque” primi-
tive, and leave it utterly mysterious how we could ever grasp the
concept. So something else must take the place of the familiar crite-
ria of identity. Similarity relations, in a suitably broad and abstract
sense of similarity, are the obvious candidate. The entity in some
other world that I take, in my modal thinking, for you will have to
resemble you at least to the extent of sharing your essential proper-
ties. The obvious proposal for defining transworld identity is thus
simply that, in the absence of anything else that looks hopeful, we
should take this necessary condition as also sufficient. The problem
is that this doesn’t work. Identity is an equivalence relation, and
it is necessarily true (so: true in every possible world) that each
object is identical to itself and not to anything else. But, however
carefully we specify the sort of similarity between objects in differ-
ent possible worlds we want to have be criterial of transworld iden-
tity, there is bound to be an object in (say) the actual world which
bears this sort of similarity to two distinct objects in some other pos-
sible world, and this would force us to say that the object in the
actual world was identical to two objects not identical to each other:
a manifest repugnancy. (One could, I suppose, try saying that x, in
world w, is identical to y, in world v, just in case they bear the appro-
priate sort of similarity to each other but x does not bear that sort of
similarity to anything in v other than y and y does not bear that sort
of similarity to anything in w other than x. This is an unattractive
option, however: at least some worlds are big places, containing
many similar objects, with the result that anything resembling one
of them in some specified way would resemble many. And, indeed,
we want to allow an object to be transworld identical to an object
that closely resembles some other object in its world: I might have
had an identical twin is intuitively plausible.)
The logical point can be made without assuming “world-bound”
individuals. Suppose, for the sake of argument, that individuals
literally exist in multiple worlds. We can still define a domain of
world-bound pseudo-individuals as, say, ordered pairs of a world
and an individual existing at it. Identity of individuals induces a
pseudo-identity relation over these pseudo-individuals, and this
relation is an equivalence with no cell containing two distinct
ordered pairs with the same world as a component. We can also
define pseudo-similarity relations over them: relations hold between
pairs <w,i> and <v,j> according to the manner and degree of simi-
larity and difference between what i is like at w and what j is like at
v. (Note that whether a pseudo-similarity holds between <w,i> and
<v,j> is independent of whether i=j: our ability to say that someone
“has changed,” or that he “looks just like his father at that age,”
depends in effect on our understanding of the analogous intertem-
poral similarity relations as independent of the identity or distinct-
ness of the terms.) Stated in this framework, the problem is that no
relation between world-individual pairs defined in terms of pseudo-
similarity relations will coincide with the relation induced by iden-
tity of individuals. All this should be acceptable even by the
staunchest defender of transworld identity. For Lewis it was a start-
ing point, providing (what must have been for him and certainly is
for me conclusive) grounds for scepticism about literal transworld
identity.
So Lewis went ersatz. Instead of postulating the sort of multi-
verse our intuitive modal thinking seems to be about—one in which
the same object often inhabits, and has different characteristics in,
multiple worlds—he postulated a replacement multiverse in which
no object exists in more than one world. (Those of us who prefer to
think of worlds as set-theoretic entities can enforce the disjointness
of worlds by stipulation: let non-actual worlds be models of the sort
usual in model theory, but take the “objects” existing in them to be,
not the entities in the domains of the models, but ordered pairs of
these with the models themselves. Assorted ordinal numbers, for
48 | Allen Hazen
example, can then be in the domains of many models, but when

thought of as objects existing in worlds they are distinguished by
using the models themselves as tags.) By itself, making the sets of
objects inhabiting different worlds disjoint would trivialize de re
modal locutions: since no object exists in any other than its “home”
world, it would follow that if an object has some characteristic, it is
a necessary truth that it has it. (Leibniz, of course, was tempted to
say precisely this when speaking strictly, but most of us want a
semantics for our vulgar modal language, a semantics that allows
non-trivial de re modality.)
Lewis’s counterpart theory was an attempt to provide a non-triv-
ial semantics for the de re without assuming that individual objects
could genuinely inhabit (and possess different properties in) multi-
ple possible worlds. De re modal statements were explained in
terms of a sort of resemblance—counterparthood, to give it its tech-
nical name—between objects in different worlds, but without tak-
ing it to define identity of objects. Unfortunately, as I argued in the
critical part of my dissertation (basically reproduced in Hazen
1979), Lewis’s theory gave the wrong truth conditions, and even the
wrong formal modal logic, for de re sentences. The (fundamental)
problems stemmed from the fact that, on Lewis’s theory, an object
might have multiple counterparts in some worlds. There are some
minor problems—such as those of what I called “There but for the
grace of God” examples: “I might have led a life like the one you
have actually and vice versa”—that can be solved by fine-tuning the
account of what sorts and degrees of resemblance make for coun-
terparthood, but solving them if anything makes the main prob-
lems worse. The biggest problem, insoluble I think without some
fundamental change to Lewis’s framework, has to do with rela-
tional de re statements. Intuitively we want to say that it is essential
to some individuals that they be related in specific ways to specific
other individuals: it is essential to Queen Elizabeth II, perhaps, that
King George VI and no one else was her father. Consider, however,
a possible world with two England counterparts (on planets in gal-
axies far, far apart), each with a pair of a royal father and royal
daughter resembling the actual George VI and Elizabeth II enough
to be their counterparts. On Lewis’s original formulation of the
truth conditions of de re statements, such a world would prevent
Elizabeth from essentially having George as her father: she has a
counterpart in that world who is not the daughter of all the counter-
parts of George in that world. One can tinker with the truth condi-
tions (substituting, say, existential for universal quantifiers, etc.),
but nothing seems to work in general.
What is needed, then, is something which yields what Fara, in
her paper, has assumed: some way of picking out one of the per-
haps many counterparts an object has in some world, and allowing
the chosen counterpart, but not the others, to figure in our semantic
account of de re modalities.
III
When I wrote my dissertation, and for many years afterward, I

thought I had solved the problem with what I called stipulational
semantics. This took worlds as the basic truth-givers for modal locu-
tions, like Lewis’s theory, but replaced the counterpart relation with
a family of (partial) functions from the domains of worlds into the
domains of other worlds. (As a first approximation, we can think of
one of these functions as mapping an object in one world that has at
least one counterpart in a second world to one of these counterparts.
As a second approximation, note that, in order to give an account of
essential relational predication, we may want the choice of a coun-
terpart for one object to depend on which counterpart in the second
world is chosen for another object, and this can be done by allowing
the family of “counterpart functions” from one world to a second to
be a proper subset of the set of all functions mapping objects in the
first to counterparts in the second: Lewis, in the afterword appended
to Lewis 1968 when it was reprinted in his Philosophical Papers, refor-
mulates this idea in terms of a counterpart relation holding, not just
between simple objects, but between fusions of objects.) Or rather,
that is what it would have been for a simple modal language with-
out nesting of modal operators. (This simplified modal language
and its semantics are outlined in the appendix to Hazen 1979; much,
though of course not all, intuitive modal reasoning can be expressed
in this fragment of quantified modal logic, and I still think my
semantics is perfectly adequate for it.)
For the full language of standard first-order modal logic my
semantics had more complicated set-theoretic machinery. What
50 | Allen Hazen
played the semantic role of a possible world in my semantics—the

entity “at” which a formula, on an assignment to its variables, is
evaluated—was not the underlying world but an ordered pair
<S,I>, where S is a finite sequence of worlds (the sequence starting
with the actual world), and I (an “identification”) is a partition of
the union of the domains of the worlds in the sequence, with each
cell of the partition containing at most one object in the domain of
any particular world. (The partition I is obtained from an equi-
valence relation which itself is the union of one counterpart func-
tion for each ordered pair of worlds in S, but not all such unions will
do: the different counterpart functions have to be “compatible.”)
This I, obviously, takes the semantic place of the counterpart rela-
tion: if d is an object in the domain of one world of S, the truth val-
ues of modal formulas about it are determined by the properties in
other worlds of the members (if any), in those worlds, of its I-cell. In
a fully developed semantics (my discussion in my thesis didn’t go
much beyond providing a model theory) conditions, formulated in
terms of some sort of similarity between objects, would be placed
on the class of admissible I, in analogy to Lewis’s proposal in (3)
that a counterpart relation be defined in terms of degrees of similar-
ity between objects.
The “shortest” such pair—adequate for formulas containing no
modal operators—would simply consist of the actual world (con-
strued as a sequence of length + 1) and the discrete partition on its
domain. For any pair <S,I> with S of length = n, define its direct
extensions to be the pairs <S+,I+>, where S+ is the sequence (of
length = n+1) obtained from S by adding one more world at the
end, and I+ is an admissible “identification” partition of the union
of the domains of worlds in S+ which agrees with I when restricted
to the union of the domains of worlds in S.
Where S is a finite sequence of worlds, let end(S) be the last world
in S. (In particular, if S is the actual world, @, construed as a one-
world sequence, end(@) = @.) The basic idea of my semantics was
that a formula embedded within the nested scopes of n modal oper-
ators should be evaluated at end(S) for a sequence of length = n+1.
The quantificational logic the semantics was designed to validate
was a version of Universally Free Logic with an existence predicate,
so constants were allowed to be vacuous. Non-vacuous constants
had designata in the domain of @. (It wasn’t clear to me then, and
still isn’t, how to extend the semantics to allow constants to act as

names of non-existent objects that exist in other worlds. I took this
to be a desirable feature of the semantics from an analytic point of
view: it provided an explanation (in some sense) for Kripke’s thesis
that Sherlock Holmes couldn’t possibly have existed.) Variables were
handled by assignments, which were partial functions from the set
of variables into the union of the domains of the worlds. (Allowing
assignments to be partial amounts to allowing free variables to be
“non-denoting singular terms.” This was a departure from the lit-
erature on Free Logic I was familiar with, but seemed convenient in
the context of modal logic.)
For any assignment, a, and term, t, we define a(t), the denotation
of t on a, by the clauses
I. if t is a vacuous constant, a(t) is undefined,
II. if t is a non-vacuous constant, a(t) is the designatum of t,
III. if t is the variable x and a is undefined for x, a(t) is unde-
fined, and
IV. if t is the variable x and a is defined for x, a(t) is the value
given by a for x.
Given a pair <S,I>, an assignment a, and a (Lewis-)world w, we
define, for an expression d of the form a(t), rep(d,I,w) (“the representa-
tive, under I, of the object d in the world w”) to be the unique object
in the domain of w which is in the same cell of I as d if there is such an
object, and undefined otherwise. Note that rep(d,I,w) is undefined if
and only if d is undefined, or w is not in S (since in that case I isn’t
defined over the domain of w), or d is defined but stands for an object
whose I-cell does not contain a member of the domain of w.
Convention: an expression <d,e,f, . . .> for an n-tuple of objects is
taken to be undefined just in case one or more of the expressions
d,e,f, . . . is undefined.
Finally, we define by recursion what it is for a formula of the lan-
guage of first-order modal logic to be true at a pair <S,I> and an
S-assignment a:
i. an atomic formula, F(t,u,v, . . .), with an n-ary predicate F and

n terms t,u,v, . . . , is True(<S,I>,a) if and only if the n-tuple
<rep(a(t),I,end(S)), rep(a(u),I,end(S)), rep(a(v),I,end(S)), . . .>
exists and is in the extension in end(S) of the predicate F,
52 | Allen Hazen
ii. a conjunction, (A&B), is True(<S,I>,a) if and only if both of

its conjuncts, a and B, are True(<S,I>,a) (and similarly for
other truth-functional compounds),
iii. a universal quantification, ∀xA, is True(<S,I>,a) if and only
if A is True(<S,I>,a*) for every S-assignment a* which
assigns x an object in the domain of end(S) and is otherwise
like a (and dually for existential quantifications), and
iv. a necessitation, NecA, is True(<S,I>,a) if and only if A is
True(<S+,I+>,a) for every direct extension <S+,I+> of
<S,I> (and dually for possibility formulas).
(Note that clause (i) of the truth definition adopts the
“falsity convention,” that atomic predicates are never true
of non-existents. This seemed to me (and still seems to
me) plausible for “ordinary” predicates: those expressing
physical properties, for example. The analysis of “inten-
sional” predicates (e.g. worships) seemed to me a topic for
another thesis.)
Note also that the truth definition would still make
sense if we allowed the identification, I, to be defined
over the domains of the worlds in a longer sequence
than S: this makes possible a fairly simple extension to
cover the Actuality operator (for which I could see no
adequate treatment in Lewis’s original semantics).
Abusing the notation a bit,
v. an “actualitation,” AA, is True(<S,I>,a) if and only if A is
“True(<@,I>,a)”: true, not in end(S), but in the actual world
@, with the denotations of terms, however, taken as
rep(d,I,@), where I is the identification from <S,I>.
IV
Before going on to discuss the possible inadequacy of even this

complicated variant of Lewis’s semantics, let me say something
about why I called it stipulational semantics.
Many philosophers have been convinced by Kripke’s arguments
in Naming and Necessity that “the problem of transworld identifica-
tion” rested on conceptual confusion from the start: we don’t first
discover objects in other possible worlds and then, after further
investigation, discover which ones are to be identified with which

actual objects: as Kripke put it, we stipulate possible worlds, and as
part of this stipulate the identities of their inhabitants. There is no
need, they may therefore be tempted to argue, for any special
machinery of transworld identification, either in Lewis’s original
counterpart relation or my more elaborate version.
I find such a response to the problem unsatisfying. Kripke’s use
of “stipulate” is metaphorical: in the literal use of the word we
speak of stipulating deadlines or preconditions or conventions or
definitions, not worlds. (The kind of stipulation most familiar in
logical and philosophical discourse is, of course, stipulative defini-
tion. It seems to me that a rough characterization of the ordinary use
of the verb, inspired by this case but capable of more or less neat
extension to others would be: stipulation is a performative act—an
exercitive in the terms of Austin’s (1960) classification—which
establishes at least a temporary linguistic convention.) The analytic
problem of transworld identification that remains, after the confu-
sion is cleared away, amounts to that of saying what the literal con-
tent of Kripke’s metaphor is. We can’t just stipulate possible worlds
by means of arbitrary descriptions: we cannot, for example, stipu-
late a world containing round squares. It is up to us to decide what
sense of possibility—historical, physical, metaphysical, etc—we
wish to express, but once we have made that decision it is a matter
for investigation rather than for further stipulation what sorts of
worlds are possible in that sense, and round squares are not histori-
cally or physically or metaphysically possible. The idea that there is
a class of worlds (or perhaps, after further epistemological and
ontological consideration, of suitable ersatzes) which is not subject
to stipulation and which determines what the de dicto possibilities
are is simply a recognition of this truism. The usual formal lan-
guages of quantified modal logic, however, allow the formulation
of de re modal statements, and, as Kripke’s own discussion dramati-
cally emphasized, it seems intuitively that many de re statements
have non-trivial truth values: thus a full semantics for modal lan-
guage needs something beyond the class of worlds.
Now, what the essence of an object is, and so what de re statements
about it are true, seems to depend on what sort of thing it is: on its
intrinsic and relational properties. (Kripke does not disagree: it is,
after all, because Queen Elizabeth II is a human person and has in fact
54 | Allen Hazen
the relational property of being the daughter of George VI and Eliza-

beth the Queen Mother that she has this property essentially.) This,
it seems to me, is an important truth, and Lewis’s proposal that the
counterpart relation should be determined by similarities in respect
of intrinsic and relational properties was an attempt to give a precise
statement of its significance. Any adequate semantic account of de re
modality will recognize this as part of the story, and as a part which
constrains anything plausibly described as stipulation. Once we have
decided to mean metaphysical possibility, in Kripke’s sense, by “pos-
sible,” we are not at liberty to “stipulate” a possible world in which
Queen Elizabeth II has other parents. As Lewis pointed out (and as
was implicit in other discussions of the so-called problem of trans-
world identification in the late 1960s) the definition of a counterpart
relation based on similarity does not amount to a characterization of
transworld identity in terms of similarity: if nothing else, an object can
have many counterparts in a single world. And, as I argued in criti-
cizing his proposal, the counterpart relation by itself does not suffice
for an intuitively plausible semantics of de re modality.
It is at this point, and only at this point, that something like
stipulation enters the picture. When we say, for example, that
Queen Elizabeth II might not have had any sons (in some chosen
sense of “might”), this is true because there are possible worlds
containing objects enough like Queen Elizabeth II in their intrinsic
and relational properties to be admissible representatives of her—
in a word, counterparts of Queen Elizabeth II—who have no sons.
If, however, as seems overwhelmingly likely, some possible uni-
verse contains more than one such counterpart, there is no need to
try to define which one is “really her” in that possible world. It is
as if we could arbitrarily choose, or stipulate, one counterpart to
represent her. But of course we don’t literally do this! We don’t
spread out a possible world like a giant map and point to the
counterpart we choose. A rigorous semantic theory (which is not
what Kripke was trying to expound in the lectures published in
Naming and Necessity) must not depend on such metaphors. I tried
to outline such a semantics: I proposed that modal statements had
the same truth conditions as certain general statements about
abstract objects (the counterpart functions, the I components in
the pairs <S,I>) which can be thought of as, well, counterparts of
these imaginary stipulations.
The semantic machinery described in Section III is designed to cope

with sentences containing nested modal operators. It now seems to
me that it may fall down if the inner modal operators do not have
nested scopes, but are allowed to occur “side by side.” For the sake
of example, assume the correctness of Kripke’s thesis that human
persons have their parentage essentially: if parents x and y have
child z, then necessarily (if z exists, then x and y exist and have z). It
follows that siblinghood is an essential relation: if x and y are sib-
lings, then necessarily (if x and y both exist, they are siblings). On the
other hand, it seems plausible that it is not essential to one of a set of
siblings that the other(s) should exist. (One’s intuitions about this
may be strained by monozygotic twins. Further, anyone who thinks
the whole of history up to the time a contingent being comes to be is
essential to that being—a view which has been suggested by some,
but which seems extreme to me—would think that the existence of
one’s older siblings is essential to one. For the sake of having an
example, let us assume that neither of these worries is relevant.)
Now, I am an only child, but I might have had siblings, siblings
who might have existed without me. The following statement (of
modal degree = 2 but with two inner modalities which are not
nested) seems reasonable (if perhaps unreasonably complex!) on
the assumptions of our example:
(*) It holds necessarily of any two people x and y, that if it is
possible that x and I both exist and are siblings and also
possible that y and I both exist and are siblings, then x and
y are siblings.
Or, in symbols, with ◻ and ◊ for modal operators, S for the sibling-
hood relation, a for me, and leaving out the clauses about existence
as made redundant by the falsehood convention on S:
(*) ◻(∀x∀y((◊(S(ax)) & ◊(S(ay))) => S(xy))).
On the semantics of my dissertation, however, (*) comes out false
(or rather, more accurately, comes out dangerously close to coming
out false).
Evaluate at the actual world. Since it is a necessitation, what fol-
lows its initial operator must come out true at every <S,I> of length =
56 | Allen Hazen
2, <<@,w>,I>, if (*) is going to be true at @. But suppose (as seems

likely) that there is a world, w1, in which I have no counterpart but in
which there are two objects, x and y, which are not siblings (perhaps
they live in different galaxies within w1) but which each have the
appropriate (intrinsic and relational) properties to be siblings of me.
Since the class of admissible identifications (like Lewis’s original
counterpart relation) is supposed to be determined by the (intrinsic
and relational) similarities between objects in different worlds, each
will be a counterpart of some sibling of mine. Now the fact that the
two possibility clauses in (*) are not nested becomes a problem.
◊(S(ax)) is true at <<@,w1>,I> because there is some world, w2, con-
taining a pair of siblings s and t such that there is an identification I’
defined over the union of the domains of the worlds in the sequence
<@,w1,w2> whose restriction to <@,w1> is just I and which identi-
fies me with s and x with t. ◊(S(ay)), on the other hand, is true at
<<@,w1>,I> because there is some world, w3, containing a pair of
siblings u and v such that there is an identification I’’ defined over
the union of the domains of the worlds in the sequence <@,w1,w3>
whose restriction to <@,w1> is just I and which identifies me with u
and y with v. (Perhaps, indeed, w3 = w2, and even s = u and t = v: all
that is required is that I’’ not be identical to I’.) Which would make
(*) false.
Now, in fact the counterexample as described doesn’t quite work.
The essentiality of the siblinghood relation, recall, is derivative from
the essentiality of parentage. For I’ to identify object t with a sibling
of an object, s, that it identifies with me, the domain of w2 must
contain parents it identifies with my parents, and these parents
must have both s and t. By the essentiality of parentage, any other
world—in particular, w1—containing an object, like x, that I’ identi-
fies with t must also contain parents I’ identifies with the parents of
t, and which, therefore, it identifies with my parents. Mutatis
mutandis for I’’, w3, and u. But, ex hypothesi, x and y are not siblings,
so I (which agrees with both I’ and I’’ when restricted to the domains
of @ and w1) cannot identify my parents both with the parents of x
and with the parents of y.
But let’s leave my parents out of this! Suppose there were a rela-
tion with the intuitive properties of siblinghood (viz. that it holds
essentially between any objects it relates, and if it possibly relates
an object to a second and possibly relates the second to a third,
then necessarily, if the first and third both exist, it relates the first
to the third: crossworld transitivity, to give it a name) but which
was not derivative from another relation to further objects. I am
not at all confident that there really are such relations (though, if
your biological species is part of your essence, being conspecific
with might be one), but I also don’t feel that I could, with any con-
fidence, declare that there are none. So it seems to me that the
modal semantics of (1) is, at least, exposed to a counterexample of
the structure described here.
Since one of my main criticisms, in both (1) and (5), of Lewis’s
original semantics (and thus one of the main motivations for my
complicated modification of it) was that it didn’t handle examples
involving essential relations, it is acutely embarrassing to me that
my own semantics may have problems with more complicated
examples involving essential relations.
VI
In Sections II and IV above, I argued that we need some account of

the semantics of de re modality that, like Lewis’s counterpart theory,
avoids an appeal to “brute” presence of the same objects in different
worlds. Lewis’s original account is unsatisfactory, and we have just
seen that the variant of it that I devised to repair its inadequacies
may not work either. What to do?
One obvious thought is to use “identifications” defined over the
domains of bigger systems of worlds. The problem in the previous
section was that I’ was an admissible “stipulation” of identities over
the domains of worlds in the sequence <@,w1,w2> and I’’ was
admissible over <@,w1,w3> but that they were incompatible. Why
not require that an identifier be admissible over the full set of worlds
relevant to the evaluation of the sentence: {@,w1,w2,w3}? There are
minor and major technical difficulties, and a conceptual issue. In
my proposed semantics, there was no need to specify which world
in a set was relevant to the evaluation of a given subformula: evalu-
ation is always at the last world in a sequence whose length is deter-
mined by the depth of modal embedding. When we go from
sequences to sets we will have to index the worlds, and the identifi-
cations will have to be partitions, not of the simple union of their
58 | Allen Hazen
domains but of their disjoint union. (Recall, from the discussion in

Section III, that w2 and w3 can be the same world, with t (= v) identi-
fied with different objects in w1 for purposes of evaluating the two
possibility subformulas.) This is ugly, but doable. More seriously, it
is not at all obvious to me how to formulate the truth definition:
somehow the definition will have to provide that the set of possible
worlds is big enough, that adding a world (and extending the “stip-
ulation” of identities to its domain) wouldn’t alter the truth value of
the sentence. And—I think this leads to a serious conceptual issue—
the set of worlds involved will generally have to be quite large. The
discussion in Section V mentioned only four worlds in toto, but
managed this by ignoring the fact that the embedded modal subfor-
mulas contained variables bound by quantifiers outside them. In
the general case, quantifying in might force us to include a new
world for each object in the domain of a given world.
One way (but one I do not want to take) of not worrying about
just what worlds have to be included would be to consider identifi-
cations defined over the union of the domains of all the worlds.
(Dan Marshall, of Monash University, seemed to be proposing
something like this a couple of years ago in work in progress.) My
reason for not wanting to do this (both when I wrote my thesis and
now) is ontological. My “working philosophy of mathematics,” for
these purposes is that Zermelo–Fraenkel set theory (with urele-
ments) is a true (though of course very far from complete!) descrip-
tion of the whole universe of abstract entities, and I wanted every
entity mentioned in my semantic theory to be identifiable with
something in the intended interpretation of ZF(U). (So in particular,
my preferred candidates for ersatz worlds are particular sets.) On
the other hand, I accept as a thesis of modal metaphysics that, for
every cardinal number, it is (metaphysically) possible for there to be
that number of concrete individuals. Thus the worlds themselves
would not form a set. An identification defined over the union of
the domains of all the worlds would have to be a proper class, and
as a good ZF-ist I am willing to countenance proper classes only as
a façon de parler, and only when they can be explicitly defined. (My
view of proper classes is close to that presented in Parsons 1974.)
Perhaps the issue can be avoided. I am at least somewhat attracted
to the idea that it is essential to any concrete object in a given world
that it inhabit a world whose fundamental physics is the same as
that of the given world. If the fundamental physics of a world deter-

mines the structure of its space-time (or whatever takes the place of
space-time in it), then this would put a limit on the number of con-
crete objects that could coexist with a given object: it will exist only
in worlds with domains of some limited cardinality. And this, in
turn, might allow us to restrict our attention to identifications
defined over the unions of the domains of sets of worlds. I am not
confident of the intuition here, however, and even if it is valid for
some conception of essence, there may be some understandable
and useful modalities requiring objects to inhabit . . . unusual . . .
worlds. After all, there are non-trivial counterfactuals whose ante-
cedents require an object to lack properties ordinarily thought
essential to it: if Queen Elizabeth II had been the daughter of Rabbi
and Mrs Kripke of Omaha. . . .
VII
To summarize the conclusions of the last section, semantics would

be easier if we could postulate an identifier (an equivalence rela-
tion, relating each object to at most one object in the domain of any
possible world, associating unique “counterparts” in other worlds
to objects) defined over the (union of the) domains of absolutely all
possible worlds. Such a relation would have to be a proper class,
not a set, but for proper classes, esse est definiri. And it was already
clear to Lewis and others in the late 1960s that it was not possible to
define a way of identifying objects with unique otherworldly coun-
terparts on the basis of similarity (broadly construed).
At this point it is tempting to try theft instead of honest toil: why
not be Fictionalists about transworld identity? Interpret our modal
discourse, that is, as based on the assumption that there is a proper-
class identifier, but don’t claim that the assumption is true. The
problem here is common to many forms of fictionalism. Modal dis-
course is serious: modal reasoning can form part of our route to
non-modal theoretical or practical conclusions, and a satisfying
philosophical account of it will provide some assurance that the
conclusions reached will be sound. A non-fictionalist semantic
theory—based either on “extreme modal realism” or on some version
of ersatzism that specifies particular abstracta as playing the role of
60 | Allen Hazen
worlds—has a simple answer here: modal reasoning may be about

a special domain, but it is nonetheless about a domain of actually
existing entities, and the specific assertions it assumes about these
entities are true. It may be, in the end, that a fictionalist account of
transworld identity is the best we can hope for, but even fiction, if it
is to play a role in serious reasoning, needs some sort of justifica-
tion. The situation is analogous to one that arose in the foundations
of mathematics early in the twentieth century.
Modern mathematics makes strong assumptions about infinite
sets, and uses them in deriving conclusions within “applicable”
mathematics. Now, a naïve Platonism (such as the ZF-ism I described
earlier as my working philosophy of mathematics!) will not find
this disturbing: it holds that the infinities assumed actually exist
and that the mathematical axioms appealed to in modern proofs
are simply true. Such an attitude, however, was not convincing for
critical minds, and did not even yield any conviction that the
infinitistic axioms of set theory were consistent! Hilbert, in par-
ticular, argued at some length that there was no reason to believe
in the actual reality of infinite sets. His view on infinitistic set the-
ory can, I think, with some plausibility be described as fictionalis-
tic: he did not hold out the hope of finding a semantic interpretation
on which the axioms of set theory would be seen to be true. But—
and this seems to me evidence that his views were deeper and
more sophisticated than most of the fictionalisms proposed in
the recent philosophical literature—he made a concrete proposal
about how to justify the continued employment of “ideal” or fic-
tional theories. Hilbert’s program of consistency proofs was meant
to find proofs that the infinitistic theories of modern mathematics
would have true consequences in applicable mathematics, proofs
that did not depend on finding a true semantic interpretation for
those theories.
In Section III above, I argued that a certain restricted modal lan-
guage—essentially the fragment of quantified modal logic in which
no necessity or possibility operator stands within the scope of
another—did have an adequate semantics in terms of “counterpart
functions.” It is not just a formal language: it is an interpreted
language, one whose sentences have (often unknown) truth values.
We might take this fragment as the analogue, for modal discourse,
of the applicable mathematics of which Hilbert wanted “ideal”
mathematics to be a consistent extension. A fictionalist analysis of

full quantified modal logic (or of some richer language with explicit
quantification over possible worlds) proposes that this fragment be
embedded in a theory postulating the existence of a “universal
identifier”: an undefinable proper class of ordered pairs of objects
in different worlds which, for each pair of worlds, coincides with
some counterpart function between the domains of those worlds.
This theory is not true, but we would like some assurance that it is
consistent with all the truths of the restricted modal fragment, so
that reasoning making use of its fictitious postulate will not lead to
false conclusions in the fully interpreted part of its language.
I think there is some reason to hope that a proof of this consist-
ency or conservativity claim is possible. The usual axiom of choice
for Zermelo-Fraenkel set theory (“Local Choice” in the jargon of the
trade) guarantees that, for every set of non-empty sets, there is a
selection function: a mapping that takes each set belonging to the
given set to one of its members. Some formulations of set theory,
however, have a stronger choice principle: Global Choice. This can
be expressed in an extension of the usual language of ZF(U): add a
function symbol σ (the selection operator or selector) with an axiom
saying that if x is a nonempty set σ(x) is one of its members. Since
the selection operator is taken to apply to all sets, σ, on a classical
semantic interpretation of this stronger system, would have to
denote a class function: a proper class of ordered pairs, but one
which, restricted to any particular set of non-empty sets, would
coincide with one of the selection functions for that set postulated
by the Local axiom of choice. Now, the ZF-ist philosophy of math-
ematics I am presupposing here is thoroughly realist about ordinary
set theory: set theory as expressed in the usual language of ZF(U).
Even the most fanatically Platonistic upholder of this philosophy,
however, should be a fictionalist about the extended version of set
theory with a selection operator. Proper classes are façons de parler,
and should only be asserted to exist when they can be defined, and
there is no reason to suppose that a class suitable to interpret σ can
be defined. Global Choice, however, is a justifiable fiction: it can be
proven that Zermelo-Fraenkel set theory with Global Choice is a
conservative extension of ordinary ZF(U)C.
The easiest proof is by a forcing construction. Cohen’s original appli-
cation of forcing used finite sets, thought of as finite approximations
62 | Allen Hazen
to an indefinable infinite set. The “finite” part of that isn’t essential:

forcing constructions are possible in which some other limitation of
size is imposed on the “small” approximations. A forcing construction
for showing that ZFC + a selection operator is conservative over ZFC
would use, as forcing “conditions,” selection operators defined over
sets: thus, for example, we could allow as a condition any function,
defined over sets of rank less than some ordinal, mapping each non-
empty set in its domain of definition to one of its own members (and
the null set to some throwaway value).
So perhaps we can do something like this. Let a condition be
an identification scheme defined over a set of worlds: an equiva-
lence relation over the union of their domains, each cell of which
has at most unit overlap with the domain of each of the worlds
in the set. One condition will extend another if it operates over a
superset of the other’s set of worlds and the two coincide over
the worlds the smaller one operates on. We’ll want to add some
further constraints:
(i) For some cardinal number it is the case that for every
world with a domain of that or smaller cardinality there is
an isomorphic world in the set the condition operates on.
There’s a metaphysical presupposition here: that the “signature” of
a possible world—the length of the sequence of its “natural proper-
ties,” in terms of which its isomorphism type is defined—is bounded
by some function of the cardinality of its domain. I think I’m will-
ing, tentatively, to buy into this. Perhaps it makes sense to speak of
a world in which very many distinct natural properties are co-
extensive, but in that case we can take a more abstract view of
things: allow the Armstrongian universals existing at a world to be
part of its domain, and limit the “signature” to properties of being
an individual, being a universal, and instantiation.
(ii) For any worlds W1 and W2 in the set on which a condi-
tion operates and any counterpart function F from the
domain of W1 to the domain of W2, there is a world W3
in the set which is isomorphic to W2, and the restriction
of the transworld identification relation to Dom(W1) X
Dom(W2) is the composition of F with an isomorphism
from W2 to W3.
(Note that it is a consequence of the idea that counterparthood is

somehow a matter of similarity that composition of a counterpart
function with an isomorphism between worlds will always be a
counterpart function.)
This is very far from being a fully worked-out proof. The breeze
you just felt was produced by hands waving.
University of Melbourne
REFERENCES
Austin, J.L., How to Do Things with Words, (Oxford: Oxford University
Press, 1960).
Cohen, P.J., Set Theory and the Continuum Hypothesis, (New York: W.A. Ben-
jamin, 1966).
Gupta, A.K., The Logic of Common Nouns: an investigation of quantified modal
logic (New Haven: Yale University Press, 1980).
Hazen, A.P., “Counterpart theoretic semantics for modal logic,” Journal of
Philosophy 76 (1979), pp. 319–38.
Kripke, S.A., “Naming and necessity,” in G. Harman and D. Davidson,
eds., Semantics of Natural Language (Dordrecht: Reidel, 1972); reprinted
with new preface as S.A. Kripke, Naming and Necessity (Cambridge:
Harvard University press, 1980).
Lewis, D.K, “Counterparts of persons and their bodies,” Journal of Phi-
losophy 68 (1971), pp. 203–11; repr. In D.K. Lewis, Philosophical Papers,
vol. I, (New York: Oxford University Press, 1983).
—— On the Plurality of Worlds (Oxford: Basil Blackwell, 1986).
Parsons, C.,”Sets and classes,” Nous 8 (1974), pp. 1–12; repr. in his Math-
ematics in Philosophy (Ithaca: Cornell University Press, 1983).
Stalnaker, R., “The interaction of modality with quantification and iden-
tity,” in W. Sinnott-Armstrong et al., eds., Modality, Morality and Belief:
essays in honor of Ruth Barcan Marcus (Cambridge: Cambridge Univer-
sity Press, 1995), pp. 12–28.
II
ABSOLUTE GENERALITY
3. All Things Must Pass Away
Joshua Spencer
0. INTRODUCTION
The notion of all things must pass away from philosophical theori-
zing. By this, I do not mean that we must reject the set of all things.
Rather, we must reject the notion that some things are such that any
things are amongst them. This is a claim made in a purely plural
language. It may be formulated as follows:
(AT) (∃xs)(∀ys)(ys are amongst xs)
Unfortunately, (AT) is false. The argument against (AT) is quite sim-
ple. It involves three premises:
(1) There are two or more things.
(2) For any things, there is a unique thing that corresponds to
those things.
(3) For any two or more things, there are fewer of them than
there are pluralities of them.
Given (3), if there are some things that are all things and there are
two or more things, then there are fewer things altogether than plu-
ralities of them. But, given (2), there is a unique thing that corre-
sponds to each plurality of things. So, there are at least as many
things as there are pluralities of things. So, either there are no things
that are all things or there are fewer than two things. But, according
to (1), there are two or more things. It follows that there are no
things that are all things. That is (AT) is false.
Before I defend the argument above, I would like to indicate just
a bit of what is and is not at stake in a denial of (AT). Many people
might worry, for example, that one consequence of my thesis is
that unrestricted singular quantification is impossible. In section 1,
I indicate that this worry is, to some extent, legitimate. Once we
acknowledge that there are no things that are all things, we must
68 | Joshua Spencer
recognize a limitation on usefulness of plurals in logic. In section 2,

I show that there is a surprising metaphysical ramification as well.
It seems that one formulation of Unrestricted Composition entails
that everything is a proper part of something further. The remain-
der of this paper is devoted to defending the argument outlined
above. In section 3, I argue that a certain cardinality thesis involving
pluralities is true. That is, I argue for premise (3). In section 4, after
indicating several metaphysical views that seem to entail both (1)
and (2), I show that a commitment to propositions supports both
of these premises. Finally, I consider several objections to my argu-
ment. First, in section 5, I consider an objection that attempts to
undercut the support for (2) by endorsing coarse-grained views
about entities like propositions. In sections 6 and 7, I consider and
respond to an objection according to which my view leads to
paradox.
1. SPEAKING OF ALL THINGS
Some people might think a denial of (AT) entails that unrestricted

singular quantification is impossible. After all, if there are no things
that are all things, then there are no things that every single thing is
amongst. But, one might think, in order for unrestricted quantifica-
tion to be possible, it must be that there are some things that every
single thing is amongst.1
At the very least, though, a denial of (AT) is consistent with a
denial of the possibility of unrestricted quantification. That is, a
denial of (AT) does not logically entail that our singular quantifiers
are indefinitely extensible. Consider any universal statement of the
following form:
(US) Everything is either identical to α or identical to β or . . .
Someone who believes that our singular quantifiers are indefinitely
extensible is committed to saying that for any context in which a
statement that has the form of (US) is true, there is another con-
1
This thought may simply be an instance of what Uzquiano calls The All-in-Many
Principle. According to the All-in-Many Principle, “quantification over objects satis-
fying a certain condition presupposes that there are some objects which are all and
only those objects that satisfy the condition” (2009, 312).
All Things Must Pass Away | 69
text in which that same statement is false. One might say that the
domain of any putatively unrestricted singular quantifier is always
extensible.2
On the other hand, consider someone who denies (AT). That is,
consider someone who accepts the following:
(~AT) ~(∃xs)(∀ys)(ys are amongst xs)
This person might consistently hold that there is a sentence of the
form (US) which expresses a truth in every context. Here is an
example to show that this is consistent. Suppose we restrict our
attention to things that are finite in number and each one of which
is a positive integer. Now, it is clear that the following three claims
are consistent:
(i) There are no integers that are finite in number and are all
integers.
(ii) For every integer, either it is identical to 1 or identical to 2
or . . .
(iii) There is no context in which the quantifier of (ii) is
expanded to make (ii) false.
Notice that (i) involves the plural quantifier ‘there are no integers’
and expresses a restricted variant of (~AT). Thus, under an appro-
priate interpretation of our plural language, (~AT) expresses (i).
Moreover, (ii) involves the singular quantifier ‘for every integer’
and expresses a restricted variant of (US). Thus, under an appropri-
ate interpretation of our plural language, (US) expresses (ii). Finally,
(iii) is just a denial of the indefinite extensibility of the singular
quantifiers in (ii). But, since (i)–(ii) are consistent and are appropri-
ate restricted interpretations of (~AT) and (US), and since (iii) is a
2
This, though, may be mistaken. On one view, there are certain unrestricted quan-
tifiers that fail to have a domain. Consider someone who believes that domains are
sets of things and yet still believes that unrestricted singular quantification is possi-
ble. On this view, when an unrestricted singular quantifier is employed, it has no
domain. This is because domains are sets and there is no set of all things. Moreover,
since there is no set of all things, domains are indefinitely extensible even though the
unrestricted quantifier is not. So, on this view, the claim that unrestricted quantifiers
are indefinitely extensible comes apart from the view that domains are indefinitely
extensible. The same may be true if domains are pluralities rather than sets and there
are no things that are all things.
70 | Joshua Spencer
denial of indefinite extensibility under that interpretation, we must

conclude that the impossibility of unrestricted quantification is not
a consequence of (~AT). The denial of indefinitely extensible quan-
tifiers and the denial of the existence of some things which are all
things, is a coherent position.3
One lesson we might draw from this is that although we may
quantify over absolutely everything, there need not be some things
that are all things. However, this lesson may make us worry about
the prospects of giving a model theoretic account of logical truth
and logical consequence for various languages using plurals instead
of sets. After all, if there is no modal theoretic interpretation of an
unrestricted quantified sentence that appropriately corresponds to
the intended meaning, then perhaps we cannot be sure that the sen-
tence is true even if it is true in all models.
In the case of first-order logic, this worry is misplaced. Complete-
ness results for first-order logic show that if a sentence is true in all
models, then it is provable and, on the assumption that the axioms
of that language are true and the inferences truth preserving, we get
that all provable sentences are true. Hence, any sentence of such a
language which is true in all models is simply true. Of course, the
worry is not misplaced when one considers second-order lan-
guages. This is because completeness results have not been obtained
for various second-order languages. Hence, one might legitimately
worry that the prospects for an account of logical truth in a second-
order language may be encumbered by the fact that there are no
things that are all things.4
3
Here is another way to make the same point. The way that I understand a plural-
ity, a single thing is merely a very sparse plurality. Given this fact, we can simply
introduce singular quantification as a restriction on plural quantification. Singular
quantification is merely plural quantification restricted to those things that are one in
number (McKay 2006). But the argument only shows that there are no things that are
all things. As long as there is more than one thing, the argument doesn’t show any-
thing about things that are one in number. Moreover, since the argument relied on the
premise that there are two or more things, it doesn’t show anything about things that
are one in number even if there is only one thing.
4
For more details see Rayo and Uzquiano (1999) and Uzquiano (2009). Also see
Williamson (2009) for related problems for unrestricted quantification. The limita-
tions placed on plural languages by a denial of (AT) may impede certain attempts to
solve the problems posed in Williamson (2009) using plurals.
2. UNRESTRICTED COMPOSITION AND JUNK
Unrestricted Composition is the thesis that for any things there is

something that composes them. This thesis is supposed to be incom-
patible with the claim that everything is a proper part of something
else (that the world is junky).5 That is to say that these two theses are
supposed to be incompatible with one another:
(UC) For any xx, there is a y such that y is composed of xx
(Junk) For any x, there is a y such that x is a proper part of y.
The argument seems to be straightforward. If (UC) is true, then any
things compose something. But, that means that all things compose
something, namely U. If (Junk) is true, then the thing that all things
compose must be a proper part of something else. That is, there
must be an object, O, that U is a proper part of. But, since U is com-
posed of all things, it follows that O is a part of U. It is clear, though,
that nothing can be a proper part of one of its parts.6 So, either (UC)
is false, or (Junk) is false. However, this argument clearly relies on
the premise that there are some things that all things are amongst.
If there are no such things, then the argument above is unsound.
In fact Unrestricted Composition and (~AT) together entail (Junk).
I have two arguments for this claim. The first is as follows: assume
(UC) and (~AT). Now arbitrarily choose an object, hereby named
‘Galileo’, and consider some xs amongst which can be found every
part of Galileo. Given that (AT) is false, there must be some zs that
those xs are amongst and that are not amongst those xs. That is, there
must be some zs that those xs are properly amongst. Again, by (UC),
there is something composed of those zs. Call that thing ‘Jupiter’. But,
the following principle seems pretty plausible, if some xs are properly
amongst some ys, then anything every part of which is amongst those
5
Jonathan Schaffer (2010), for instance, says “Classical mereology—with its axiom
of unrestricted composition—guarantees the existence of a unique fusion of all con-
crete objects. Thus there are gunky models of classical mereology, but no junky mod-
els. Indeed, a mereologically maximal element is the only individual that classical
mereology guarantees on every model.” Einar Duenger Bohn, in both (2009a) and
(2009b), has made similar remarks.
6
Some people might reject this premise if they think it’s possible for something to
be shrunk down and sent back in time to become a part of itself. Although I am a fan
of the possibility of such science fiction examples, I will not be discussing the implica-
tions of such fantastical possibilities here.
72 | Joshua Spencer
xs is a proper part of anything those ys compose.7 It follows that Gali-

leo is a proper part of Jupiter. But, since Galileo was arbitrarily chosen,
we may conclude that everything is a proper part of some further
thing. That is, (Junk) is true. So, (UC) and (~AT) entail (Junk).
Some people might be reluctant to accept the principle that if some
xs are properly amongst some ys, then anything every part of which
is amongst those xs is a proper part of anything those ys compose.
Someone might hold that Galileo is not a part of Jupiter, but rather
just shares some parts with Jupiter. Perhaps this kind of view would
allow someone to hold (UC) without (Junk). This seems like a plau-
sible position, but my next argument shows that it is mistaken.
The second argument is as follows: Consider some arbitrary thing.
Call that thing ‘Chunk’ and call all of its parts ‘Bits’. Now, if (AT) is
false, there are some things that Chunk’s parts are properly amongst.
Call those things ‘Bits+Pieces’. Now, there are some things amongst Bits+
Pieces that are not parts of Chunk. Those things that are amongst
Bits+Pieces which are not parts of Chunk are Pieces. According to
(UC) there is something composed of Chunk and Pieces, namely Big
Chunk. Chunk is a part of Big Chunk. Moreover, since Big Chunk has
parts that are not parts of Chunk, namely Pieces, it follows that Chunk
is a proper part of Big Chunk. So, there is something that Chunk is a
proper part of. But, since Chunk was arbitrarily chosen, it follows
that for anything whatsoever, there is something that it is a proper
part of. So, (Junk) is true. Again, (UC) and (~AT) entail (Junk).
Both of these arguments rely on the assumption that for any com-
posite object, there are some things that are all of its parts. One may,
of course, deny this assumption and hold that (UC) and (~AT) are
both true and yet (Junk) is false. But, of course, the denial of this
assumption is, in itself, an interesting mereological result.
3. “THE MORE YOU APPROACH INFINITY, THE DEEPER

YOU PENETRATE TERROR” 8
I would like to start by considering the third premise of my argu-

ment first: (3) for any two or more things there are fewer of them
than there are pluralities of them. This is just as true of pluralities as
7
This principle is entailed by Classical Extensional Mereology.
8
Gustave Flaubert (or so I've heard).
it is of sets. However, unlike sets, there really aren’t any entities

which are the pluralities. Admittedly, I seemed to reify pluralities to
state this cardinality thesis.9 But, we can write out, individually,
each claim if we have to. For example, if we have three things, A, B,
and C, then in addition to the three singulars (which are merely
very sparse pluralities) there are the things: A and B; the things: A
and C; the things: B and C; and the things: A, and, B and C. Here, A,
B, and C are fewer in number than the items on a list of things
amongst which just A, B, and C are individually found; there are
only three things, whereas there are seven items on the list. If we
help ourselves to a perplural language, we can say that there are
three things and there are seven thingses total.10 In fact, we can
make our general claim in a language of perplurals: For any two or
more things, there are fewer of those things than there are thingses
amongst which just those things are found. It may sound weird to
talk about thingses, but I think it makes the cardinality thesis rather
clear.11 Formally, the cardinality thesis can be stated in a second-
order language with plurals as follows: There is no relation R, such
that for any xs there is a y such that all and only those xs stand in R
to y. However, in the remainder of this section, I will speak infor-
mally of lists that pair-up pluralities with individuals.
9
Literally speaking, something is a cardinal number only if it numbers the mem-
bers of some set. I am going to use the notion of a cardinal number in an extended
sense. On the notion I will be employing, something is a cardinal number if it num-
bers some things, whether or not they form a set.
10
Hazen (1993) introduced a language of perplurals. Hazen, however, tried to use
the sensibility of such a language to argue against plural quantification. However, I
think we can understand the language and need not accept Hazen’s rejection of plu-
ral quantification. Perhaps, we can understand the language through a convenient
fiction according to which pluralities are real entities over which perplural quantifi-
ers range. The fictional interpretation of the language need not lead to inconsistencies
as long as we stipulate that the plural quantifiers of such a language range only over
the non-fictional entities of the universe. The plural quantifiers will have their ordi-
nary meaning whereas the perplurals will have a fictional meaning that merely helps
us in our counting.
11
Something like it follows from a plurals version of Cantor’s theorem. Stewart
Shapiro (1991) has a proof of a second-order version of Cantor’s theorem and George
Boolos (1984) has famously suggested that we can reinterpret second-order state-
ments in a language of plurals. If Boolos’s method of doing so is adequate, then Sha-
piro’s proof and statement of Cantor’s theorem should be reinterpretable in a
language of plurals. Something like the cardinality claim immediately follows. In the
remainder of this section, I present a Cantorian argument for the cardinality princi-
ple. See also Rayo (2002).
74 | Joshua Spencer
It will be worthwhile to present an argument for the cardinality

claim. The argument has two steps. We begin by assuming that
there are at least two things and attempting to make a list that con-
tains in one column all the things that are one in number, all the
single things, and contains in the second column all the things that
are any number whatsoever. If the cardinality thesis is false, then
there should be such a list, each individual uniquely paired with a
plurality and no plurality left unpaired. Our first step is to show
that some individual on that list must be paired with a plurality
that that very individual is not amongst.
Assume that there are at least two things and that we can uni-
quely pair-up each individual with a plurality, leaving no plurality
unpaired. Arbitrarily choose a very sparse plurality consisting of
one thing and name it ‘Tom’. Either Tom is paired with itself or it is
not. If it is not, then there is at least one individual that is paired with
a plurality that that individual is not amongst (namely the individ-
ual paired with Tom). So, assume that Tom is paired with itself. Now
arbitrarily choose another sparse plurality and name it ‘Dick’. Again,
either Dick is paired with itself or it is not. If it is not, then there is at
least one individual paired with a plurality that it is not amongst. So,
assume Dick is paired with itself. Now, consider Tom and Dick. They
cannot be paired with either Tom or Dick, since Tom is already
paired with Tom and Dick is already paired with Dick. So, they must
be paired with some new object, ‘Harry’, which is not amongst Tom
and Dick. Hence, if there are more than two things and there is a
pairing of individuals to pluralities, then there is at least one indi-
vidual that is paired with a plurality that it is not amongst.12
The second step in our argument is to show that if there are more
than two things, then there cannot be a list that pairs every plurality
with an individual and leaves no plurality unpaired. We know from
the first step that if there is such a list, then there must be at least
one individual paired with some things that it is not amongst.
Suppose there is such a list and let those individuals that are not
amongst the pluralities they are paired with be called ‘sinful indi-
viduals’. Consider all and only the individuals that are not amongst
the pluralities they are paired with; all and only the sinful individu-
als. Call that plurality ‘The Fallen’. The Fallen must also be paired
12
Thanks to Karen Bennett for this argument.
with an individual, call that individual, ‘Eve’. Now we may ask: Is

Eve amongst The Fallen or not?
If Eve is amongst The Fallen, then Eve must be a sinful individ-
ual. This is because The Fallen are a plurality of only the sinful indi-
viduals. But, if Eve is a sinful individual, then Eve is not amongst
The Fallen. Because sinful individuals (by definition) are not
amongst the pluralities they are paired with. So, if Eve is amongst
The Fallen, then Eve is not amongst The Fallen.
On the other hand, if Eve is not amongst The Fallen, then Eve is a
sinful individual. This is because Eve is not amongst the plurality it is
paired with. But, if Eve is a sinful individual, then Eve is amongst The
Fallen. After all, The Fallen are a plurality of all the sinful individuals.
So, if Eve is not amongst The Fallen, then Eve is amongst The Fallen.
It clearly follows that Eve is amongst The Fallen iff Eve is not
amongst The Fallen. But, of course, either Eve is amongst The Fallen
or Eve is not amongst The Fallen. If the first disjunct is true, then
Eve both is and is not amongst The Fallen. If the second disjunct is
true, then, again, Eve both is and is not amongst The Fallen. Either
way, we are led to a contradiction. So, it must be that our original
supposition is false. But, our original supposition was that there is
a list that pairs every plurality with a unique individual. So, there
cannot be such a list.
As I mentioned before, if the cardinality thesis is false, then there
must be a list that pairs every plurality with an individual and leaves
no plurality unpaired. Since we have just shown that there cannot be
any such list, we must conclude that the cardinality thesis is correct.
So, for any things, there are fewer of them than there are pluralities of
them. Hence, premise (3) of the argument against all things is true.
4. SOMETHING FOR EVERYONESES 13
Certain metaphysical views commit us to the existence of two or

more things and to a unique thing for each plurality of things. That
is, if certain metaphysical views are true, then premises (1) and (2)
13
The argument of this section is similar to, but distinct from, arguments I presented
in my (2006). However, I am indebted to Greg Fowler for many helpful discussions
about some puzzling aspects of my previous argument. These discussions helped me to
see some of my underlying assumptions and develop the argument that appears here.
76 | Joshua Spencer
are true as well. Which metaphysical views have such consequence?

There are lots of them. Some people, for example, believe in states
of affairs. If there are states of affairs, then there are at least two of
them, one that obtains and one that does not. Moreover, if there are
states of affairs, then for any things, there is a state of affairs of just
those things existing. Other people believe that there is an omnis-
cient God. If there is an omniscient God, then there are at least two
things, God and God’s first person existential thought. Moreover,
if there is an omniscient God, then for any things, God has some
thought about just those things (the thought that they exist, for
example). There are even those who believe in possible and impos-
sible worlds, or ways for the world to be. If there are possible and
impossible worlds, then there are at least two worlds, one possible
and one impossible. Moreover, if there are such worlds, then for
any things, there is a world where just those things exist.14 Views
like these commit one to the thesis that there are no things that are
all things.15
I’d like to consider, in more detail, one view that seems to
commit us to (1) and (2). Consider the view that there are truths
and falsehoods; the view, in other words, that there are proposi-
tions. If there are propositions, then there are at least two things.
First, there is the proposition that some propositions are true. Sec-
ond, there is the proposition that every proposition is false. These
are two distinct propositions because they must have different
truth values; the first must be true whereas the second must be
false. So, on the view that there are propositions, there are at least
two propositions.16 So, on the view that there are propositions,
statement (1) is true.
14
Many of these worlds will be impossible worlds since many things cannot exist
without other things also existing. For example, I cannot exist without the number
two also existing.
15
One more view that commits us to two or more things and a unique thing associ-
ated with any things is a robust view of properties. According to a robust view of
properties, there are many properties and for any things, there is a property that they
and only they have in common.
16
Moreover, there is also at least one more proposition. Not every proposition has
its truth value necessarily. Some propositions are contingent. But, since each of the
two propositions above have their truth values necessarily, it follows that there is at
least one more proposition. So, on the view that there are propositions, there are at
least three things.
Now, for any things, there is a proposition just about those things.
I think it is fairly clear what it means to say of some things that a
proposition is about them. The proposition that Nicholas and Brie
both exist is, for example, about Nicholas and Brie. Some philoso-
phers might use the word ‘about’ in an extended sense. They might
say that, in addition to being about Nicholas and Brie, the proposi-
tion that Nicholas and Brie both exist is also about the property of
existence. But, I do not intend to use the word in this extended sense.
The way I am using ‘about’, the proposition that Nicholas and Brie
both exist is about Nicholas and Brie and is not at all about the prop-
erty of existence. There are, of course, propositions that are about
properties (or at least I think there are). The proposition that exist-
ence is monadic is about the property existence. But, it should be
clear, now, that it is not also about the property of being monadic.
It is obvious that for any things there are some propositions about
them. After all, for any things, it is obvious that there are some truths
about them; that they exist, perhaps, or have some kind of being.
But, I want to make a slightly stronger claim. I believe that for any
things there is a proposition about them. Moreover, I believe that for
any things, there is a proposition just about them. That is to say that
for any things there is at least one proposition about those things and
there are no further things that that proposition is also about. We can
state this Propositions Thesis a bit more formally as follows:
(PT) for any xs, (i) there is a proposition about those xs and
(ii) for any ys which are not the same as the xs, there is a
different proposition about those ys.
If (PT) is true, then statement (2) is true as well.
One way to support (PT) is to indicate a certain type of proposition
that is such that for any things whatsoever, some proposition of that
type is just about those things. That is, indicate some propositions
that correspond, in the appropriate way, to all the various pluralities
in the world. It seems to me that existential propositions are good
candidates for the appropriate correspondence. If that is correct, then
the following Existential Propositions Thesis might be true.
(EPT) For any xs, (i) there is the proposition that those xs exist
and (ii) for any ys which are not the same as the xs, the
proposition that those ys exist is distinct from the prop-
osition that those xs exist.
78 | Joshua Spencer
If (EPT) is true, then for any things there is a proposition, an existential

proposition, just about those things.17 Hence, (PT) is true and so is (2).
In addition to existential proposition, there are also composi-
tional propositions. That is, propositions that say of some things
that they compose a further thing.18 Of course, some people believe
that some things cannot compose a further thing. However, even if
there are some things which cannot compose a further thing, there
is still a compositional proposition just about them. After all, if it is
necessarily false that they compose some further thing, then some
proposition about them is necessarily false. So, perhaps the follow-
ing Compositional Propositions Thesis is true:
(CPT) For any xs, (i) there is the proposition that those xs
compose something and (ii) for any ys which are not
the same as the xs, the proposition that those ys com-
pose something is distinct from the proposition that
those xs compose something.
If (CPT) is true, then (again) for any things, there is a proposition,
this time a compositional proposition, just about those things.
Hence, (PT) is true and so is (2).
Finally, there are doxastic propositions, propositions that say of
some things that someone or other is thinking about them. Again, it
seems intuitively correct that for any things there is a proposition
that someone is thinking about those things. Moreover, this seems
right even if it turns out that most such propositions are false. So, it
seems that the following Doxastic Propositions Thesis is true:
(DPT) For any xs, (i) there is the proposition that those xs are
thought about by someone and (ii) for any ys which
17
Admittedly, there are some who deny the first tenet of (EPT). Some people think
that existence is not a first order property of either individuals or pluralities. Rather,
existence is a higher order property of properties. Thus, we cannot, for any individ-
ual x, say that x exists. Rather, we can only say, for some property F, that there are Fs.
Furthermore, some people might claim that for all we know, there may be some
things that share no properties. Thus, we cannot say of them that they exist because
we cannot say of some property they all share that some things have that property.
However, this worry should not be too worrisome since any things have the prop-
erty of being just those things. This is a property that they have and any other things
lack. So, even if existence is a second-order property, we can say of any things that
existence applies to the property of being just those things.
18
Even if existence is not a first-order property of plurals, surely the property of
composing something is such a property.
are not the same as the xs, the proposition that those
ys are thought about by someone is distinct from the
proposition that those xs are thought about by someone.
If (DPT) is true, then for any things there is a proposition, a doxastic
proposition, just about them.19 Hence, (PT) is true and so is (2).
So, if any of (EPT), (CPT), or (DPT) are true, then (PT) is true as
well. That is, each of the three theses above supports the claim that
for any things there is a proposition just about those things. Of
course, if (PT) is true, then so is (2). But, we need not appeal to the
existence of particular types of propositions to support (PT).
It is easy to show that for any things, there is a proposition just
about them; that is, it is easy to show that (PT) is true. Suppose, for
reductio, that there are some things such that there is no proposition
just about them. Let’s say that a1 . . . an are some such things. Then, it
is true that a1 . . . an are such that there is no proposition just about
them. But, since every truth is a proposition, it is clear that there is
at least one proposition about them, namely the proposition that
a1 . . . an are such that there is no proposition just about them. Moreo-
ver, this proposition is just about them.20 So, there is at least one
proposition just about them. But, this contradicts our claim that
19
Notice that the denial of (DPT) entails that there can be no omniscient being. For,
if there could be an omniscient being (even a non-actual one), then for any things
whatsoever, that being would believe that they possibly exist.
20
One might wonder whether this proposition really is just about a1 . . . an. I have to
admit that I have no criteria for determining whether or not a proposition is just
about some things. However, the following seems intuitively correct to me. P is about
some xs if those xs are some of the things that make P true and they are the only such
things that are such that P logically implies that they have some feature or other. For
example, Nicholas and the property of being tall both help to make the proposition
that Nicholas is tall true. But, the proposition that Nicholas is tall logically implies
that Nicholas has some feature or other. It does not logically imply that the property
of being tall has some feature or other. Moreover, there are no other things that help
to make that proposition true. So, the proposition is about Nicholas and not tallness.
Similarly, being blue and being a color are the only things that help to make it true
that blue is a color. However, the proposition that blue is a color logically implies that
blue has some feature or other but does not logically imply that being a color has
some feature or other. So, that proposition is about the property of being blue and not
about the property of being a color. Now consider the proposition that a1 . . . an are
such that there are no propositions about just them. It seems that a1 . . . an are things
that make that proposition true and that proposition logically implies that a1 . . . an
have some property or other (namely being such that there are no propositions about
them). Moreover, there are no other things that help to make that proposition true
which are also such that that proposition logically implies they have some property
or other.
80 | Joshua Spencer
there is no proposition just about them. Hence, the supposition

that there are some things of which there is no proposition just
about them is false. So, for any things, there is a proposition just about
them. That is, (PT) is true.
So, it seems clear that the thesis that there are propositions com-
mits us to rejecting the claim that there are some things that are
all things. For, if there are propositions, then there are two or more
of them. Hence, statement (1) is true. But, if there are propositions,
then (PT) is true. Moreover, if (PT) is true, then for any things, there
is a unique thing that corresponds to just them. Hence, statement
(2) is true. Moreover, the cardinality thesis I argued for in the last
section says that for any things there are fewer of them than plurali-
ties of them. Hence, statement (3) is true. But, as I have shown, from
(1)–(3) it follows that there are no things that are all things. That is,
it follows from (1)–(3) that (AT) is false. Hence, (AT) is false.
The remainder of this paper is devoted to defending this argu-
ment from objections. In section 5, I consider a popular view about
propositions that might undermine the support for premise (2) in
the argument. In section 6, I consider an objection that says premise
(2) leads to paradox. Finally, in section 7, I consider an objection
according to which my response to the paradox of section 6 under-
mines my argument.
5. ON THE PROPOSITION THAT THERE IS

A PLURALITY OF WORLDS
In addition to the claim that for any things there is a proposition

about them, I also endorse the claim that for any things there is a
proposition just about them. That is to say that for any things there
is a proposition about those things and there are no further things
that that proposition is also about. Some people, though, believe in
a sparse view of propositions according to which propositions are
merely sets of possible worlds.21 A proposition, P, is true at a world,
w, just in case P is identical to a set of possible worlds and w is a
21
A better view might be that propositional attitudes (and other properties and
relations that seem to take propositions as objects) are really irreducibly plural. The
fundamental belief relation, on this view, is irreducibly plural in its second place and,
hence, it will have the form ‘S believes those xs’ where the plural variable is satisfied
member of that set. On this view, for any propositions P and Q, if it

is necessary that P is true iff Q is true, then P=Q. But, now consider
the proposition that the number 2 exists and the proposition that
the number 7 exists. Since numbers are necessarily existing entities,
it is impossible for one of those propositions to be true without the
other proposition also being true. But, then it follows, on this view,
that the proposition that the number 2 exists is the same as the
proposition that the number 7 exists. Hence, the proposition that
the number 2 exists is not just about the number 2, it is also about the
number 7. On this view, it might turn out that although it is true that
for any things there is a proposition about them, it is false that there
is a proposition just about them. Hence, if this sparse view is true,
then (PT) may be false and part of the support for (2) may be
undermined.
There are two things that I would like to say in response to this
objection. First, the best sparse theory of propositions is one accord-
ing to which propositions are sets of worlds. But, it seems to me
that this view of propositions is mistaken. In what follows, I will
say a bit about why I think this sparse view of propositions is mis-
taken. I recognize, however, that there are some who have very
by worlds. Consider the sentence ‘grass is green’. On this view, that sentence picks
out, plurally, all those worlds where grass is green. Moreover, if someone believes the
content of that sentence, then she stands in the belief relation to those worlds. This
view might be favored by those who believe that some worlds are too numerous to
form a set. If propositions are sets, then there is no proposition that corresponds to
some worlds that are too numerous to form a set. However, if (speaking vulgarly)
propositions are pluralities, then those worlds that are too numerous to form a set are
still propositions and might still be expressed and believed.
One downside of this view is the following. There are some truths about proposi-
tions that seem to be irreducibly plural. For example, when I say that Nicholas’s
beliefs are consistent, I seem to be saying something irreducibly plural about propo-
sitions. However, if propositions are pluralities, then I must be saying something
irreducibly perplural. As I mentioned before, some people think that there’s no way
to make sense of a perplural language without reifying pluralities (which is exactly
what we want to avoid). The only hope for a defender of this view is to find a way to
paraphrase away those claims that seem to be irreducibly plural; like the claim that
Nicholas’s beliefs are consistent. Gabriel Uzquiano (2004) has a good discussion of
the prospects for paraphrasing seemingly irreducible plural claims in another con-
text. Much of what Uzquiano says will apply in this circumstance as well. Luckily,
nothing that I say in the remainder of this section depends on whether propositions
are sets of worlds or pluralities of worlds. So, I will simply focus on the former
view.
82 | Joshua Spencer
reasonable defenses of the sparse view of propositions. So, I will

close this section by noting that there are many metaphysical views
that support premise (2) of my main argument against all things.
Endorsing a sparse view of propositions in order to avoid com-
mitment to (2) is merely the first step on a dark path to a generally
sparse metaphysics.
Let’s start by considering a sparse view of propositions. Let’s
consider a bare-bones view according to which (i) for any proposi-
tion P, P is possibly true iff there is a possible world, w, such that P
is true-at-w, and (ii) for any possible world, w, and any proposition,
P, of the form there are Fs, P is true-at-w iff P is identical to a set of
F-worlds and w is a member of that set. At the very least, if the
sparse view of propositions is correct, then (i) and (ii) must both be
true.
Unfortunately, (i) and (ii) are not both true. Let’s suppose that
necessarily, for any proposition, P, P is possibly true just in case
there is a possible world, w, and P is true-at-w. But, it is necessary
that there are merely possible truths; claims that are not true but
are possibly true. It follows from these last two claims that neces-
sarily, there is a merely possible world; worlds, distinct from the
actual world, at which the merely possible truths are true.22 Since,
necessarily, there is a world that is not merely possible, namely the
actual world, it follows that necessarily there is more than one pos-
sible world. That is, the following claim turns out to be necessarily
true:
(PW) There is (quantifiers wide open) a plurality of possible
worlds.
So, if (i) and (ii) are both true, then (PW) is necessarily true.
But, if (i) and (ii) are both true and (PW) is necessarily true, then
(PW) is identical to the set of all worlds. After all, given that a neces-
sary proposition is one the negation of which is not possibly true
22
Strictly speaking this does not follow from the last two claims. On Lewis’s view,
a merely possible de re modal truth about me may be true because there is some other
actual individual who is my counterpart. However, it is necessary that there are
merely possible de dicto truths. The claim that there is a world distinct from the actual
world follows from the claim that there are merely possible de dicto truths and the
claim that necessarily, for any proposition, P, P is possibly true just in case there is a
possible world, w, and P is true-at-w.
(that is, given the duality of necessity and possibility), it follows

from (i) that a necessary truth is true at all worlds. But, according to
(ii), if (PW) is true at all worlds, then (PW) is identical to the set of
all worlds. If (PW) is identical to the set of all worlds, then (PW) is
identical to any other proposition that is true at all worlds; in
particular, (PW) is identical to the proposition that arithmetic is
incomplete. (PW), however, is not identical to the proposition that
arithmetic is incomplete. Many of us believe the latter but disbe-
lieve the former and, moreover, Gödel proved the latter without
also proving the former. So, (PW) is not the set of all worlds. So,
either (i) or (ii) is false. Since the sparse view of propositions is true
only if both (i) and (ii) are true, it follows that the sparse view is
mistaken.
As I mentioned before, although I take the above objection to be
sound, I recognize that there are some who have very reasonable
defenses of the sparse view of propositions. Many of the defenders
have plausible things to say about objections like the one above.
But, there is something that I find more troubling about this style of
response to my argument against all things. This style of response
seems to be merely one step down a dark path to a generally sparse
metaphysical view. As I noted at the beginning of section 4, there
are several metaphysical views that seem to support premise (2) of
my argument. I chose to focus on the view that for any things there
is a proposition just about them. However, I could have easily
focused on the view that for any things, there is a state of affairs of
just those things existing; or the view that for any things there is a
property of being just them; or the view that for any things there is
a possible or impossible world were just they exist. Any of these
metaphysical views supports premise (2) in my argument against
all things. So, anyone who wishes to avoid the conclusion of my
argument must accept not only a sparse view of propositions, but
also a sparse view of states of affairs, properties, and worlds; one
must accept a generally sparse metaphysics.
Moreover, if one accepts a generally sparse metaphysics in
response to my argument, then one must also hold that how these
entities are sparsely distributed over individuals must match up.
That is, one must reject the view that for any things there is either a
proposition just about them or a state of affairs of just those things
existing or a property that just they have, etc. In other words, in
84 | Joshua Spencer
order to undermine all hope of support for premise (2), one must
accept a generally sparse and rather radical metaphysics.23
6. WE MUST PASS OVER IN SILENCE
In the last section, I considered a sparse view of propositions

which some might have taken to be both true and inconsistent
with (PT). If such a view were true, then the primary support for
premise (2) would be undermined. However, it seems that the
sparse view is false. Moreover, regardless of whether or not it’s
false, one would have to accept a generally sparse and radical
metaphysics to undermine all potential support for premise (2). In
this section, I consider what I take to be the strongest objection to
(PT). According to this objection (PT) entails a contradiction all by
itself, and hence must be false. However, if (PT) is false, then our
primary support for (2) is undermined. Moreover, the form of this
objection can be applied to any robust metaphysical view. So, if
this objection were successful, it would indicate a path in support
of a generally sparse metaphysics.
The objection to (PT) is fairly clear.24 First, assume that (PT) is
true; that is, assume that for any things there is a proposition just
about them. Now, it is clear that some propositions are not about
23
To make the darkness of this path more acute, note that one who accepts that
propositions are merely sets of possible worlds should say exactly what a possible
world is. But, the standard views of possible worlds cannot be held given a sparse
metaphysical view. We cannot, for example, say that a possible world is a complex
state of affairs or a complex property or even a complex sentence in a lagadonian
language (where everything is its own name and every property is a predicate that
expresses itself). For if any of these resources are rich enough to use in our construc-
tion of possible worlds, then they are also rich enough to generate support for premise
(2). I see three options available to the defender of a sparse metaphysics. First, one
could, of course, simply endorse Lewis’s unorthodox view according to which pos-
sible worlds are concrete things composed of individuals which are appropriately
related. Second, one could accept Magical Ersatzism according to which there is no
true account of how possible worlds represent ways the world could be (i.e. they just
do). Finally, one could accept a poor world-making language and allow most repre-
sentation to be implicit. However, I think few of us are willing to accept the counter-
intuitive costs of Lewis’s concrete modal realism or Magical Ersatzism (though see
van Inwagen (1986) for a defense of the latter). So, the third option seems most
plausible.
24
A version of this argument is presented by Rayo and McGee (2000).
themselves. For example, the proposition that Parmenides was born

in Elea is not about itself (at most it is about Parmenides and Elea).
Let all and only those propositions that are not about themselves be
called ‘humble propositions’. Since there are some humble prop-
ositions, by (PT), there must be a proposition that is just about the
humble propositions (perhaps the proposition that they exist or the
proposition that they are humble). Arbitrarily choose any propo-
sition that is just about the humble propositions and call it
‘Confusion’.
Given the reference-fixing description used to name Confusion
and the definition of ‘humble proposition’, the following two claims
are true.
A. For any proposition P, Confusion is about P iff P is
humble.
B. For any proposition P, P is humble iff P is not about P.
But, by universal instantiation on (A) and (B) respectively, we get
(C) and (D).
C. Confusion is about Confusion iff Confusion is humble.
D. Confusion is humble iff Confusion is not about
Confusion.
And, of course, (E) follows from (C) and (D).
E. Confusion is about Confusion iff Confusion is not about
Confusion.
But, (F) is a logical truth and (G), (H), and (I) all follow by classical
logic.
F. Either Confusion is about Confusion or Confusion is not
about Confusion.
G. If Confusion is about Confusion, then Confusion both is
and is not about Confusion.
H. If Confusion is not about Confusion, then Confusion both
is and is not about Confusion.
I. So, Confusion both is and is not about Confusion.
This contradiction followed from the supposition that for any things
there is a proposition just about those things. So, there must be some
things that no proposition is about. That is, (PT) must be false. But,
86 | Joshua Spencer
since (PT) was our primary support for premise (2), it looks like our
support for (2) has been undermined.
This is a very powerful argument against (PT). However, I do not
believe the argument is sound. I do not accept (F).25 This may seem
shocking, given that (F) seems like a logical truth. But, it turns out
that propositions of the form P or not-P are not logical truths. This is
one lesson we should take away from the Liar Paradox. Consider
the English sentence, hereby named ‘(LIAR)’: “(LIAR) does not
express a truth”. If we accept that either (LIAR) expresses a truth or
it does not, then contradiction immediately follows. So, we must
not accept that either (LIAR) expresses a truth or it does not. But,
that means that there is at least one proposition of the form p or not-
p which is unacceptable and hence, not a logical truth.
Now, I said that I do not accept (F). But, I do not accept that Con-
fusion is neither about itself nor not about itself either. To move
from not accepting (F) to accepting the negation of (F) is to make the
same mistake as before. Such a move presupposes that (F) or not-(F)
is true. Rather, we must remain silent about whether (F) is true and
we must remain silent about whether Confusion is about itself.26
It turns out that ‘is not about itself’ and ‘does not express a truth’
express partial properties. Let’s call those things that are instances of
a property the ‘metaphysical extension’ of that property.27 Partial
properties are properties that have a definite metaphysical extension
and a definite metaphysical anti-extension. An atomic proposition
of the form a is F is true if a lies within the metaphysical extension of
the property F. The negation of an atomic proposition of the form a
is F is true if a lies within the metaphysical anti-extension of the
property F. However, there are some things that lie outside both the
metaphysical extension and the metaphysical anti-extension of par-
tial properties. If a lies outside both the metaphysical extension and
25
The view that I present is very similar to the view presented by Field (2008) and
is inspired by Soames’s (1999) view of truth.
26
If we were to model a language that behaves this way, then we would assign to
each predicate an extension and an anti-extension. However, some predicates, such
as ‘is not true’ and ‘is not about itself’, will be such that the union of their extension
and anti-extension fails to include the entire universe.
27
Here I am following Salmon (1981 pp. 46) in distinguishing a metaphysical
extension from a semantic extension. Properties have metaphysical extensions
whereas predicates have semantic extensions.
the metaphysical anti-extension of F, then we must remain silent

about whether a is F; that is, a is F is unacceptable.
There are lots of partial properties and it is very easy to introduce
a predicate that expresses such a property into a language. Suppose
I say “being a humanoid who is over 10 meters tall is sufficient for
being a scholossal and being a humanoid under 200 centimeters tall
sufficient for not being a scholossal. Furthermore, no non-human-
oids are scholossals.” Suppose Breetai comes from a species of
humanoids who are between 10 and 15 meters tall. If Breetai is over
10 meters tall, then clearly he is a scholossal. If, on the other hand,
Max is an ordinary human who is under 200 centimeters tall, then
clearly he is not a scholossal. What, however, about someone who is
210 centimeters tall? Such a human falls outside the metaphysical
extension of the property of being scholossal. In this case, we must
remain silent about whether such a human is a scholossal.
The properties expressed by ‘is not about itself’ and ‘does not express
a truth’ are partial properties. It is clear that some propositions fall
within the metaphysical extensions of these properties and some prop-
ositions fall inside the metaphysical anti-extensions of them as well.
However, there are a few propositions that are neither in the meta-
physical extension nor in the metaphysical anti-extensions of these
properties. (LIAR) is a sentence that lies outside both the metaphysical
extension and the metaphysical anti-extension of the property
expressed by ‘does not express a truth’.28 Similarly, any propositions
that are about all and only the humble propositions, including Confu-
sion, lie outside of the metaphysical extension and metaphysical anti-
extension of the property expressed by ‘is about itself’.
7. THE BOUNDS OF SILENCE
Some may have noticed that the argument presented in the last sec-
tion against (PT) is similar in structure to one part of the argument
28
Moreover, given the truth conditions for negation, (LIAR) lies outside of the
metaphysical extension and anti-extension of the property expressed by ‘is true’ as
well. This means that the proposition, hereby named (TRUTH), that this proposition
is true is also outside the metaphysical extension and anti-extension of the property
expressed by ‘is true’. Hence, we must remain silent about whether the truth teller is
true. See Soames (1999) for a discussion of this consequence.
88 | Joshua Spencer
used to defend the cardinality thesis in section 1. Remember that I

called the plurality of all and only those individuals that are not
amongst the pluralities they are paired with ‘The Fallen’. I also
called the individual paired with The Fallen ‘Eve’. I then argued for
the claim that Eve is amongst The Fallen iff Eve is not amongst The
Fallen. Finally, I claimed that either Eve is amongst The Fallen or
Eve is not amongst The Fallen and derived a contradiction. At that
point, I concluded that the supposition that some things are such
that there are as many of them as there are pluralities of them is
false. That is, I concluded that the cardinality thesis is correct. How-
ever, when I considered an argument of the same form in section 6,
I refused to accept premise (F), the disjunctive premise that either
Confusion is about itself or it is not. Why, one might legitimately
ask, did I decide to remain silent about whether Confusion is about
itself but I did not decide to remain silent about whether Eve is
amongst The Fallen? How could the premises of the first argument
be acceptable yet the premise of the second not?
The latter question is difficult to answer. I do know that some
arguments of the form we are considering are sound. For example,
suppose I say “let ‘Mark Barber’ name the barber of Syracuse who
shaves all and only those who don’t shave themselves.” The propo-
sition that Mark Barber shaves himself iff he does not shave himself
is inconsistent with the proposition that either Mark Barber shaves
himself or he does not shave himself. However, I do not remain
silent over the claim that either Mark Barber shaves himself or he
doesn’t shave himself. Rather, I accept that premise, reject the bicon-
ditional and conclude that Mark Barber does not exist.
I also know that some arguments of this form are not sound.
Obviously, (LIAR) expresses a truth iff it does not express a truth.
However, I will not accept that either (LIAR) expresses a truth or it
does not. Moreover, I cannot accept that premise lest contradiction
ensues. The difference between the case of Mark Barber and the
case of (LIAR) is that we must accept the existence of (LIAR). Here
it is before us on this very page!
(LIAR): “(LIAR) does not express a truth.”
I can think of no clearer proof that (LIAR) exists than its very pres-
ence before my eyes. Mark Barber, on the other hand, is not present-
ing himself so clearly.
Eve is like Mark Barber. Eve’s existence is not foisted upon us and
we are free to avoid contradiction by denying that Eve exists. This
route leads us to the cardinality principle. Confusion, on the other
hand, is like (LIAR). We know that Confusion exists because we
know that there are some humble propositions. Moreover, I am con-
vinced by the discussion of section 2 that there must be a proposi-
tion just about the humble propositions. The existence of Confusion
forces us into silence. That is, we are forced to remain silent about
whether Confusion is about itself.
There is also an independent reason for avoiding silence about
the claim that either Eve is amongst The Fallen or Eve is not. Sup-
pose the claim that either Eve is amongst The Fallen or Eve is not
amongst The Fallen is unacceptable. That is, suppose that we must
remain silent about that disjunction. If that is the case, then we must
also remain silent about each of the disjuncts. So, the claim that Eve
is amongst The Fallen is a claim that we must remain silent about.
But, we know that the amongst relation is transitive. But, that means
that if Eve is amongst any things that are amongst The Fallen, then
Eve is amongst The Fallen as well. So, if we must remain silent
about whether Eve is amongst The Fallen, then we must remain
silent about whether Eve is amongst any things that are amongst
The Fallen. Some things that are amongst The Fallen are individuals
and individuals are simply pluralities that are one in number. So,
for any things that are one in number and amongst The Fallen, we
must remain silent about whether Eve is amongst those things. One
interesting discovery of the logic of plurals is that there is a relation-
ship between the amongst relation and the identity relation. In par-
ticular, the following Identity Principle is true:
(IP) (∀xs)(∀ys) (xs=ys ↔ (xs are amongst ys & ys are amongst
xs & xs are one in number))
That is, one plurality is identical to a second plurality iff each is
amongst the other and neither is more than one in number.29 But,
supposedly we must remain silent about whether Eve is amongst
any things that are amongst The Fallen (including individuals).
Moreover, Eve is definitely one in number. It follows from these two
claims that for any individual amongst the fallen, we must remain
29
See, for example, McKay (2006, 129).
90 | Joshua Spencer
silent about whether it is identical to Eve. This suggests that there

are some identity claims that we should remain silent about and the
root of that silence is in the identity relation. But, if we must remain
silent about any identity claim, the root of that silence is not in the
identity relation. So, our supposition must be mistaken. That is,
the claim that either Eve is amongst the Fallen or Eve is not amongst
the Fallen is perfectly acceptable. But, if that is right, then the proof
of the cardinality principle is sound.
8. CONCLUSION
Although the claim that there are some things that any things what-
soever are amongst seems intuitively plausible, I believe this thesis
must be rejected. Those who disagree must accept widespread met-
aphysical limitations, not only with respect to propositions, but also
with respect to states of affairs, properties, divine thoughts, and
impossible worlds (to name a few). Moreover, they must accept that
there are some things that are such that there are is no proposition
just about them and there is no state of affairs involving just them
and there is no property had by just them etc. On the other hand,
those who reject all things, must face certain limitations on the use-
fulness of plurals and perhaps accept certain surprising metaphysi-
cal theses. It seems to me that the costs of rejecting all things are less
drastic than the costs of sparse metaphysics.30
University of Wisconsin, Milwaukee
BIBLIOGRAPHY
Bohn, Einar Duenger (2009a) “An Argument Against the Necessity of
Unrestricted composition,” Analysis Vol. 69: 27–31.
—— (2009b) “Must there be a Top Level?” The Philosophical Quarterly
Vol. 59: 193–201.
Boolos, George (1984) “To Be is to be the Value of a Variable (or to be Some
Values of Some Variables),” The Journal of Philosophy Vol. 81: 430–49.
30
Thanks to Karen Bennett, Gregory Fowler, Mark Heller, Hud Hudson, Kris
McDaniel, Tom McKay, Chris Tillman, and Gabriel Uzquiano, for discussing these
issues and reading earlier drafts of this paper. Thanks also to audiences at The Uni-
versity of Manitoba and Syracuse University for helpful comments and discussion.
Cartwright, Richard (1994) “Speaking of Everything,” Nous Vol 28: 1–20.

Field, Hartry (2008) Saving Truth from Paradox (Oxford: Oxford University
Press).
Grim, Patrick (2000) “The Being that Knew Too Much,” International Jour-
nal of Philosophy of Religion Vol 47: 141–54.
—— (1991) The Incomplete Universe (Cambridge, MA: MIT Press).
—— (1984) “There is No Set of All Truths,” Analysis Vol. 44: 206–8.
—— (1983) “Some Neglected Problems of Omniscience,” American Philo-
sophical Quarterly Vol. 20: 265–76.
Leonard, Henry and Goodman, Nelson (1940) “The Calculus of Individu-
als and its Uses,” Journal of Symbolic Logic Vol. 5: 45–55.
Lewis, David (1991) Parts of Classes (Oxford: Basil Blackwell).
—— (1986) On the Plurality of Worlds (Oxford: Basil Blackwell).
McKay, Thomas (2006) Plural Predication (Oxford: Clarendon Press).
Plantinga, Alvin and Grim, Patrick (1993) “Truth, Omniscience, and
Cantorian Arguments: An Exchange,” Philosophical Studies Vol. 71:
267–306.
Rayo, Agustín (2002) “Word and Objects,” Nous Vol. 36: 436–64.
—— and McGee, Van (2000) “A Puzzle about De Rebus Belief,” Analysis
Vol. 60: 297–9.
—— and Uzquiano, Gabriel (1999) “Toward a Theory of Second-
Order Consequence,” Notre Dame Journal of Formal Logic Vol. 40:
315–25.
Rosen, Gideon (1995) “Armstrong on Classes as States of Affairs,”
Australasian Journal of Philosophy Vol. 73: 613–25.
Salmon, Nathan (1981) Reference and Essence (Princeton: Princeton Uni-
versity Press).
Schaffer, Jonathan (2010) “Monism: The Priority of the Whole,” The Philo-
sophical Review Vol. 119: 31–76.
Shapiro, Stewart (1991) Foundations Without Foundationalism: A Case for
Second-Order Logic (Oxford: Clarendon Press).
Simons, Peter (1987) Parts: A Study In Ontology (Oxford: Clarendon
Press).
Soames, Scott (1999) Understanding Truth (Oxford: Oxford University
Press).
Spencer, Joshua (2006) “Two Mereological Arguments against the Possi-
bility of an Omniscient Being,” Philo Vol. 9: 62–71.
Tarski, Alfred (1935) “On the Foundations of the Boolean Algebra,”
reprinted in Tarski, Alfred (1983) Logic, Semantics, Metamathematics
(Indianapolis, IN: Hackett Publishing Company).
Uzquiano, Gabriel (2004) “Plurals and Simples,” The Monist Vol. 87:
429–51.
92 | Joshua Spencer
Uzquiano, Gabriel (2009) “Quantification Without a Domain,” New Waves

in Philosophy of Mathematics edited by Otávio Bueno and Øystein Lin-
nebo (New York: Palgrave Macmillan).
van Inwagen, Peter (1986) “Two Concepts of Possible Worlds” Midwest
Studies in Philosophy Vol. 9: 185–213.
Varzi, Achille (2004) “Mereology,” The Stanford Encyclopedia of Philosophy
(Fall 2004 Edition), Edward N. Zalta (ed.), URL = <http://plato.stanford.
edu/archives/fall2004/entries/mereology/>.
4. Absolute Generality Reconsidered
Agustín Rayo
Years ago, when I was young and reckless, I believed that there was
such a thing as an all-inclusive domain.1 Now I have come to see the
error of my ways.
The source of my mistake was a view that might be labeled ‘Trac-
tarianism’. Tractarians believe that language is subject to a meta-
physical constraint. In order for an atomic sentence to be true, there
needs to be a certain kind of correspondence between the semantic
structure of the sentence and the ‘metaphysical structure’ of reality.
The purpose of this paper is to explain why I think Tractarianism is
mistaken, and what I think an anti-Tractarian should say about
absolutely general quantification.
1. THE PLAN
‘Just is’-statements will be important in what follows, so let me start

by giving you some examples:
1. Sibling
For Susan to be a sibling just is for her to share a parent
with someone.
2. Water
For the glass to be filled with water just is for it to be filled
with H2O.
3. Physicalism
For such-and-such a mental state to be instantiated just is
for thus-and-such brain-state to be instantiated (and for the
environment to be thus-and-so).
4. Properties
For Susan to instantiate the property of running just is for
Susan to run.
5. Death
For a death to take place just is for someone to die.
1
Relevant texts include Rayo (2002), Rayo (2003), and Rayo and Williamson (2003).
94 | Agustín Rayo
6. Tables
For there to be a table just is for there to be some things
arranged tablewise.
7. Dinosaurs
For the number of the dinosaurs to be Zero just is for there
to be no dinosaurs.
Statement 1 is utterly uncontroversial. Statement 2 should be pretty
uncontroversial too, at least if we ignore certain complications (such
as the possibility of impurities). Statement 3 is not totally uncontro-
versial (Chalmers (1996)), but it seems to be the dominant view
amongst philosophers.
Statements 4–7, on the other hand, are all highly controversial met-
aphysical theses. My own view is that they are all true, but I won’t try
to convince you of that here. The aim of this paper is to argue that
they shouldn’t be rejected on general linguistic or metaphysical
grounds. I will proceed by defending a conception of language I call
compositionalism, and showing that it makes room for Statements 4–7.
I will then argue that a compositionalist who accepts Statements 4–7
is left with an attractive metaphysical picture of the world.
The plan for the paper is as follows. I will start by explaining
how I think the ‘just is’-operator should be understood (section 2).
I will then introduce my foil: Tractarianism. I will explain why I
think Tractarianism is bad philosophy of language (section 3), and
develop compositionalism as an alternative to Tractarianism (sec-
tion 4). Attention will then turn to metaphysics. I will argue that
compositionalism does not lead to untoward metaphyical conse-
quences, even if one accepts ‘just is’ statements such as 4–7 (section
5). I will conclude by addressing the problem of absolute generality
from the perspective of the compositionalist (section 6).
2. THE ‘JUST IS’-OPERATOR
Before mounting my defense of compositionalism, it will be use-

ful to say something about how I will be understanding the
‘just is’ -operator, as it occurs in Statements 1–7.
Consider Sibling as an example. What it takes for Sibling to be
true is for there to be no difference between Susan’s having a sibling
and Susan’s sharing a parent. If Susan is a sibling it is thereby the
Absolute Generality Reconsidered | 95
case that she shares a parent, and if she shares a parent it is thereby
the case that she is a sibling. More colorfully: when God created the
world, and made it the case that Susan shared a parent, there was
nothing extra She had to do, or refrain from doing, in order to ensure
that Susan was a sibling. She was already done. And when God cre-
ated the world, and made it the case that Susan was a sibling, there
was nothing extra She had to do, or refrain from doing, in order to
ensure that Susan shared a parent. She was already done.
In the special case in which Susan is, in fact, a sibling, there is an
additional way of clarifying the meaning of Sibling. For Sibling to
be true is for ‘Susan is a sibling’ and ‘Susan shares a parent’ to be
full and accurate descriptions of the same feature of reality.
Other ‘just is’-statements should be understood in the same sort
of way. For Death to be true is for there to be no difference between
someone’s dying and a death’s taking place. When someone dies it
is thereby the case that a death takes place, and when a death takes
place it is thereby the case that someone dies. The feature of reality
that is fully and accurately described by saying ‘A death took place’
is also fully and accurately described by saying ‘Someone died’.
It is useful to compare Sibling and Death with an identity state-
ment such as ‘Hesperus is Phosphorus’. If you accept ‘Hesperus is
Phosphorus’, you believe that there is no difference between
traveling to Hesperus and traveling to Phosphorus. Someone who
travels to Hesperus has thereby traveled to Phosphorus, and some-
one who travels to Phosphorus has thereby travelled to Hesperus.
The feature of reality that is fully and accurately described by say-
ing ‘A Soviet spaceship traveled to Hesperus’ is also fully and accu-
rately described by saying ‘A Soviet spaceship traveled to
Phosphorus’.
Since ‘just is’-statements are treated as equivalent to the corre-
sponding ‘no difference’ statements, the ‘just is’-operator is treated
as symmetric. There is a different reading of ‘just is’ on which it fails
to be symmetric. One could suggest, for example, that a ‘just is’-
statement should only be counted as true if the right-hand-side
‘explains’ the right-hand-side, or if it is in some sense ‘more funda-
mental’. This is not the reading that will be relevant for present pur-
poses. If you find the asymmetric reading more natural than the
symmetric reading, please substitute a suitable ‘no difference’-
statement for each ‘just is’-statement in the text.
96 | Agustín Rayo
There is a lot more to be said about ‘just is’-statements.2 But we

had better plunge ahead. Otherwise we’ll never get to absolute
generality.
3. TRACTARIANISM
In this section I will introduce my foil: a view that will be referred

to as Tractarianism.3 It makes no difference for present purposes
whether there are any actual Tractarians. The point of introducing
Tractarianism is that it’ll make it easier to explain what composi-
tionalism amounts to, and why it is an attractive thesis.
A Tractarian believes that in order for an atomic sentence to be true,
there needs to be a certain kind of correspondence between the seman-
tic structure of the sentence and the ‘metaphysical structure’ of reality.
Consider ‘Susan runs’ as an example. Let us agree that in order for this
sentence to be true, it must supply a full and accurate description of
some feature of reality. As long as one is suitably deflationary about
fact-talk, one might think of the relevant feature of reality as a fact: the
fact that Susan runs. What is distinctive about Tractarianism is the
claim that ‘Susan runs’ can only supply an accurate description of the
fact that Susan runs if the semantic structure of ‘Susan runs’ is in sync
with the particular way in which the world’s metaphysical structure
carves up the fact that Susan runs. In particular: metaphysical struc-
ture must have carved up the fact into an object and a property such
that ‘Susan’ refers to the object and ‘runs’ expresses the property.
More generally, Tractarianism is the view that in order for an
atomic sentence ⌜F(t1, . . . ,tn )⌝ to constitute a full and accurate
description of a given feature of reality, the following three condi-
tions must hold:
1. The world’s metaphysical structure carves up the relevant
feature of reality into the objects a1, . . . ,an and the property P.
2. For each i, the singular term ⌜ti⌝ refers to ai.
3. The predicate F expresses P.
2
See Rayo (forthcoming) "Neo-Fregeanism reconsidered". For a different way of
thinking about ‘just is’-statements, see Bennett (2009).
3
For further discussion of Tractarian conceptions of language, see Heil (2003). For
criticism, see Eklund (2009).
Tractarianism is a substantial view. Notice, in particular, that Trac-

tarians are immediately barred from accepting certain ‘just is’-state-
ments. Consider ‘for Socrates’s death to take place just is for Socrates
to die’. The embedded sentences are both atomic, but they have dif-
ferent semantic structures. Suppose for reductio that they both
describe the same feature of reality, as the ‘just is’ statement would
have it. How does the relevant feature of reality get carved up by
the world’s metaphysical structure? At most one of the following
can be true:
• It gets carved up into Socrates and the property of dying.
• It gets carved up into Socrates’s death and the property of an
event’s taking place.
If the former is true, then the relevant feature of reality can be accu-
rately described by ‘Socrates is dying’, but not by ‘Socrates’s death
is taking place’. If the latter is true, then the relevant feature of real-
ity can be accurately described by ‘Socrates’s death is taking place’,
but not by ‘Socrates is dying’. Either way, the ‘just is’ statement
turns out to be false.
Here I am taking for granted that the logical form of a sentence
can be read off more or less straightforwardly from the sentence’s
surface grammatical structure. This is a non-trivial assumption. Say
you believe that proper logical analysis of ‘Socrates is dying’ reveals
it to have the same logical form as ‘Socrates’s death is taking place’.
Then you should think that the Tractarian could accept ‘for Socra-
tes’s death to take place just is for Socrates to die’ after all. Contem-
porary linguistics does suggest that there are certain cases in which
there is a real mismatch between surface structure and semantically
operative lower-level syntactic structure. But, as far as I can tell, it is
not the sort of mismatch that would offer much comfort to the Trac-
tarian. (See, for instance, the treatment of semantics in Heim and
Kratzer (1998).) If this is right, then the assumption that logical form
can be read off more or less straightforwardly from grammatical
structure is a harmless simplification in the present context.
There are certain ‘just is’-statements that the Tractarian is in a
position to accept. She is free to accept ‘for Susan to be a sibling just
is for Susan to share a parent’, for example. For, as long as she is
happy to identify the property of being a sibling with the property
of sharing a parent, she will be in a position to claim that the seman-
98 | Agustín Rayo
tic structures of both ‘Susan is a sibling’ and ‘Susan shares a parent’

are in sync with the metaphysical structure of the fact that Susan is
a sibling. For similar reasons, the Tractarian is free to accept Water
and Physicalism, from section 1.
The Tractarian is, however, barred from accepting Properties.
And, on reasonable assumptions about the treatment of non-atomic
sentences, she is also barred from accepting Dinosaurs and Tables.
As I noted above, these are all controversial metaphysical theses.
What is striking about Tractarianism is that it rules them out merely
on the basis of syntactic considerations.
Metaphysical Structure
Tractarianism is a hybrid of linguistic and metaphysical theses. It
deploys a metaphysical assumption—the existence of metaphysi-
cal structure—to impose a constraint on linguistic theorizing.
I suspect that the notion of metaphysical structure is not in good
order, and I would like to make a few remarks about why I think
this is so.
Let me start by talking about objectivism. Most of us are objectiv-
ists about truth. We believe that it makes sense to speak of what is
objectively the case, as something over and above what is true
according to one person or another. Many philosophers, but not all,
are objectivists about morality. They believe that it makes sense to
speak of what is objectively good, as something over and above
what would be good with respect to some value system or other. Few
philosophers, if any, are objectivists about fashion. You may think
that ascots are fashionable. But it would be preposterous to suggest
that they are objectively fashionable: fashionable over and above the
tastes of some community or other.
Objectivism comes at a cost. An objectivist about fashion, for
example, would be faced with the awkward task of elucidating a
non-trivial connection between what is objectively fashionable and
what various communities take to be fashionable. She would also
have to choose between coming up with an explanation of what it
takes for something to be objectively fashionable and burdening
her picture of the world with the view that there are brute facts
about objective fashion. And the rewards for her efforts would be
decidedly meager. For it is not clear what theoretical advantages

fashion objectivism could bring. As far as I can tell, interesting theo-
retical questions concerning fashion can all be addressed by using a
community-relative notion of fashionability.
When the price of objectivism is not worth paying, one should do
more than simply deny that the relevant objectivist notion has any
instances. One should deny that the notion makes sense. Someone
who claims to understand the notion of objective fashionability
faces the burden of elucidating the connection between objective
and community-relative fashionability, whether or not she thinks
the world happens to contain any instances of objectively fashion-
able outfits. For what gives rise to the explanatory burden is the
concept of objective fashionability, not the assumption that it has
any instances.
In some cases, of course, the price of objectivism is worth paying.
The notion of objective truth is fruitful enough that few would feel
unduly burdened by the need to explain the connection between
objective truth and truth according to an agent, or by a picture of
the world whereby there are brute facts about what is objectively
true and what is not.
Metaphysics is filled with objectivist views. There are metaphysi-
cians who believe that it makes sense to speak of objective similar-
ity, as something over and above what might strike an agent as
similar.4 There are metaphysicians who believe it makes sense to
speak of objectively fundamental vocabulary, as something over
and above the role a piece of vocabulary plays in some scientific
theory or other.5 And—most relevantly for present purposes—a
metaphysician might think that it makes sense to speak of the objec-
tively correct way of carving up reality into objects, as something
over and above the syntactic properties of the various representa-
tions one might use to describe the world.
Before embracing a form of metaphysical objectivism, it is impor-
tant to be mindful of the costs. My own view is that when it comes
to metaphysical structure, the price is not worth paying. For I sus-
pect that many of the most interesting metaphysical questions can
be addressed without having to appeal to the notion of metaphysi-
4
See Lewis (1983) and Lewis (1984).
5
See Fine (2001), Schaffer (2009), and Sider (typescript).
100 | Agustín Rayo
cal structure. Because of this, the need to elucidate the connection

between an objectively correct way of carving up reality and the
ways in which reality gets carved up by our representations strikes
me as too high a price to pay for the resulting theoretical benefits.
(At the same time, I don’t think it would be irrational to think
otherwise.6)
Bad Philosophy of Language

Even though I suspect that the notion of metaphysical structure
makes no sense, I will not be relying on this assumption anywhere
in the paper. My argument against Tractarianism will be based on
the claim that Tractarianism is bad philosophy of language.
As the name suggests, Tractarianism is a close cousin of the ‘pic-
ture theory’ that Wittgenstein advocated in the Tractatus.7 And it
ought to be rejected for just the reason Wittgenstein rejected the pic-
ture theory in his later writings. Namely: if one looks at the way
language is actually used, one finds that usage is not beholden to
the constraint that an atomic sentence can only be true if its seman-
tic structure is in suitable correspondence with the metaphysical
structure of the world.
It is simply not the case that ordinary speakers are interested in
conveying information about metaphysical structure. The sentences
‘a death took place’ and ‘someone died’, for example, are used more
or less interchangeably in non-philosophical contexts. An ordinary
speaker might choose to assert one rather than the other on the
basis of stylistic considerations, or in order to achieve the right
emphasis. But it would be tendentious to suggest that her choice
turns on her views about metaphysical structure. It is not as if an
ordinary speaker would only be prepared to assert ‘a death took
place’ instead of ‘someone died’ if she has a certain metaphysical
view about events: that they are amongst the entities carved out by
the world’s metaphysical structure. Think about how inappropriate it
6
See, for instance, Schaffer (2009) and Sider (typescript). For a critique of certain
forms of metaphysical objectivism, see Hofweber (2009).
7
Here I have in mind a traditionalist interpretation of the Tractatus, as in Hacker
(1986) and Pears (1987). See, however, Goldfarb (1997).
would be to respond to an assertion of ‘a death took place’ in a non-

philosophical context by saying ‘I am certainly prepared to grant
that someone died, but I just don’t think that the world contains
events amongst its ultimate furniture.’ One’s interlocutor would
think that one has missed the point of her assertion, and gone off to
a different topic.
If ordinary assertions of ‘a death took place’ are not intended to
limn the metaphysical structure of the world, what could be the
motivation for thinking that the truth-conditions of the sentence
asserted play this role? As far as I can tell, it is nothing over and
above the idea that semantic structure ought to correspond to meta-
physical structure. Remove this idea and there is no motivation left.
To buy into Tractarianism is to start out with a preconception of the
way language ought to work, and impose it on our linguistic theo-
rizing from the outside—from beyond what is motivated by the
project of making sense of our linguistic practice.
Moderate Tractarianism
There is a moderate form of Tractarianism according to which the
constraint that there be a correspondence between semantic struc-
ture and metaphysical structure applies only to assertions made by
philosophers in the ‘ontology room’. When I use the term ‘Tractari-
anism’ here, the view I have in mind is always non-moderate Trac-
tarianism. My arguments for the claim that Tractarianism is bad
philosophy of language do not apply to moderate Tractarianism.
For all I know, there is a special convention governing discourse
in the ontology room, which demands correspondence between
semantic and metaphysical structure. If you are sympathetic
towards moderate Tractarianism, that’s fine. Just make sure you
don’t interpret me as a moderate Tractarian.
A moderate Tractarian is free to accept a ‘just is’-statement such
as ‘for a death to take place just is for someone to die’. All she needs
to do is insist that at most one side of the ‘just is’-statement is taken
in an ontology-room spirit. To avoid confusion, moderate Tractari-
ans might consider introducing a syntactic marker for ontology-
room discourse, as in Fine (2001). They could then say:
What it really is for a death to take place is for someone to die
102 | Agustín Rayo
or:
What it is, in fundamental terms, for a death to take place is for
someone to die
to indicate that the feature of reality described by ‘a death takes
place’ gets carved by the world’s metaphysical structure in a way
that corresponds to the semantic structure of ‘someone dies’.
Just to be clear: this is not what I intend when I use ‘just is’-state-
ments here.
4. COMPOSITIONALISM
I will now defend an alternative to Tractarianism: the view I shall

refer to as compositionalism.
Suppose you introduce the verb ‘to tableize’ into your language,
and accept ‘for it to tableize just is for there to be a table’ (where the
‘it’ in ‘it tableizes’ is assumed to play the same dummy role as the
‘it’ in ‘it is raining’). Then you will think that what would be required
of the world in order for the truth-conditions of ‘it tableizes’ to be
satisfied is precisely what would be required of the world in order
for the truth-conditions of ‘there is a table’ to be satisfied. In both
cases, what would be required is that there be a table (equivalently:
that it tableize). So you will think that—for the purposes of stating
that there is a table—object-talk is optional. One can state that there
is a table by employing a quantifier that binds singular term
positions—as in ‘there is a table’—but also by employing an essen-
tially different syntactic structure—as in ‘it tableizes’.
If object-talk is optional, what is the point of giving it a place in
our language? The right answer, it seems to me, is ‘compositionality’.
A language involving object-talk—that is, a language including
singular terms and quantifiers binding singular-term positions—is
attractive because it enables one to give a recursive specification of
truth-conditions for a class of sentences rich in expressive power.
But there is not much more to be said on its behalf. In setting forth
a language, we want the ability to express a suitably rich range of
truth-conditions. If we happen to carry out this aim by bringing in
singular terms, it is because they supply a convenient way of speci-
fying the right range of truth-conditions, not because they have
some further virtue.
A proponent of this sort of view will disagree with the Tractarian

about what it takes for a singular term to succeed in referring.
Whereas the Tractarian would insist that a singular term can only
succeed in referring if it is paired with one of the objects carved out
by the world’s metaphysical structure, proponents of the present
view will claim that all it takes is a suitable specification of truth-
conditions for sentences involving the term.
More specifically, we shall let compositionalism be the view is that
all it takes for a singular term t to refer is for the following three
conditions to obtain:
1. Truth-conditions have been specified for every sentence
involving t that one wishes to make available for use.
2. The assignment of truth-conditions respects composition-
ality, in the following sense:
If ϕ is a syntactic consequence of ψ, then the truth-condi-
tions assigned to ψ impose at least as strong a requirement
on the world as the truth-conditions assigned to ϕ.
3. The world is such as to satisfy the truth-conditions that
have been associated with the sentence ‘∃x(x = t)’.
Compositionalism is a substantial view. The best way to see this is to
imagine the introduction of a new family of singular terms ⌜the direc-
tion* of a⌝, where a names a line. The only atomic sentences involving
direction*-terms one treats as well-formed are those of the form ‘the
direction* of a = the direction* of b’, but well-formed formulas are
closed under negation, conjunction, and existential quantification. A
sentence ϕ is said to have the same truth-conditions as its nominali-
zation [ϕ]N, where nominalizations are defined as follows:8
• [⌜the direction* of a = the direction* of b⌝]N = ⌜a is parallel to b⌝.
• [⌜xi = the direction* of a⌝]N = ⌜z i is parallel to a⌝.
• [⌜xi = xj⌝]N = ⌜zi is parallel to zj⌝.
• [⌜∃xi(ϕ)⌝]N = ⌜∃zi([ϕ]N)⌝.
• [⌜ϕ ∧ ψ⌝ ]N = the conjunction of [ϕ]N and [ψ]N.
• [⌜_ϕ⌝]N = the negation of [ϕ]N.
8
It is worth noting that the nominalizations of open formulas turn out to be open
formulas, and therefore lack truth-conditions. Fortunately, all that is required for
present purposes is an assignment of truth-conditions to well-formed sentences.
104 | Agustín Rayo
It is easy to verify that every condition on the compositionalist’s list

is satisfied. Notice, in particular, that since [‘∃x(x = the direction*
of a)’]N is ‘∃z(z is parallel to a)’, and since every line is parallel to
itself, all that is required for the truth-conditions of ‘∃x(x = the direc-
tion* of a)’ to be satisfied is that a exist.
Accordingly, the compositionalist will claim that the existence of
a is enough to guarantee that the singular term ‘the direction* of a’
has a referent. Moreover, by employing the newly introduced
vocabulary in the metalanguage, the compositionalist will claim
that the referent of ‘the direction* of a’ is the direction* of a, and
therefore that the existence of a is enough to guarantee the existence
of the direction* of a.9
How is this possible? How could a linguistic stipulation, together
with the existence of lines, guarantee the existence of directions*?
There is nothing deep or mysterious going on. As a result of the
linguistic stipulation, the fact that is fully and accurately described
by saying ‘a is parallel to b’ can now also be fully and accurately
described by saying ‘the direction* of a = the direction* of b’. Simi-
larly, the fact that is fully and accurately described by saying ‘a is
parallel to a’ can now also be fully and accurately described by say-
ing ‘the direction* of a = the direction* of b’. So of course the exist-
ence of lines is enough to guarantee the existence of directions*: for
the direction* of a to be self-identical (equivalently: for the direc-
tion* of a to exist) just is for a to be self-parallel (equivalently: for a
to exist).
Needless to say, the Tractarian would insist that such a linguistic
stipulation is inadmissible. She would insist, in particular, that ‘a is
parallel to b’ and ‘the direction* of a = the direction* of b’ cannot be
full and accurate descriptions of the same fact. For they are atomic
sentences with distinct semantic structures. So they cannot both be
in sync with the way in which the fact that a is parallel to b gets
9
For closely related views, see Frege (1884), Wright (1983), and Rosen (1993). Dis-
cussion of compositionalism amongst contemporary metaphysicians in the United
States has tended to focus on the work of Eli Hirsch, who draws on earlier work by
Hilary Putnam. (See, for instance, Putnam (1987) and Hirsch (2002); for criticism, see
Eklund (2008) and Bennett (2009).) It is worth keeping in mind, however, that some
of Hirsch’s linguistic theses go significantly beyond anything defended here, and
that his general attitude towards metaphysics is profoundly different from my own
(see, in particular, the discussion in section 4.1 of the present text).
carved up by the world’s metaphysical structure. According to the

Tractarian, there can’t be directions* unless some fact gets carved
up into directions* by the world’s metaphysical structure. And even
if the existence of lines is taken for granted, no stipulation can tell
us that directions* are amongst the objects carved out by the world’s
metaphysical structure.
The compositionalist, on the other hand, believes that an atomic
sentence can be true even if there is no correspondence between
semantic structure and metaphysical structure. So there is no imme-
diate obstacle for atomic sentences with different semantic struc-
tures to deliver full and accurate descriptions of the same fact. We
needn’t check whether directions* are amongst the objects carved
out by the world’s metaphysical structure in order to determine
whether there are any directions*. It is enough to observe that ‘a is
parallel to a’ and ‘the direction* of a = the direction* of a’ are full and
accurate descriptions of the same fact, and that a is indeed parallel
to itself. (Notice, incidentally, that it is no part of compositionalism
that there is no such thing as metaphysical structure. The point is
simply that the notion of metaphysical structure does not figure in
a proper account of the reference of singular terms.10)
It is important to be clear that compositionalism does not entail
any interesting ‘just is’-statements unless it is supplemented with
further claims. Notice, in particular, that compositionalism does not
entail that all it takes for the direction of a (as opposed to the direc-
tion* of a) to exist is for a to exist. In order to get that conclusion we
would need a substantial hypothesis about the truth-conditions of
sentences involving the ordinary word ‘direction’. In particular, we
would need to help ourselves to the claim that the ordinary sen-
tence ‘the direction of a exists’ has the same truth-conditions as ‘a
exists’. And this is non-trivial assumption. We would, in effect, be
assuming that for the direction of a to be identical to the direction of
b just is for a to be parallel to b, which is a controversial metaphysi-
10
It is also worth noting that compositionalism is not in tension with the view—
first suggested in Lewis (1983) and Lewis (1984)—that problems of referential inde-
terminacy can sometimes be resolved by attending to metaphysical naturalness.
Compositionalism is a view about what it takes for a singular term to be in good
order, not about the sorts of considerations that might be relevant to fixing the refer-
ence of singular terms. There is room for thinking that Lewis himself was a composi-
tionalist: see Lewis (1980).
106 | Agustín Rayo
cal claim. It is true that we made the analogous assumption in the

case of directions*. But back then we were introducing a new term,
and were therefore free to introduce truth-conditions by
stipulation.
Although compositionalism does not commit one to the accept-
ance of any interesting ‘just is’ statements, it does eliminate an obsta-
cle for the acceptance of ‘just is’-statements. One is no longer barred
from accepting a ‘just is’-statement merely on the basis of syntactic
considerations. But there is substantial work to be done before one
can make a case for accepting any particular statement.
4.1 ‘Just is’-Statements in Metaphysics

Some of the ‘just is’-statements a compositionalist is in a position to
accept constitute interesting metaphysical theses: for the number of
the dinosaurs to be Zero just is for there to be no dinosaurs; for
Susan to instantiate the property of running just is for Susan to run.
Let me say something about the sorts of considerations that might
be relevant to deciding whether to accept ‘just is’-statements such as
these. (My discussion is very much indebted to Block and Stalnaker
(1999) and Block (2002).)
When one accepts a ‘just is’-statement one closes a theoretical gap.
Suppose you think that for a gas to be hot just is for it to have high mean
kinetic energy.11 Then you should think there is no need to answer the
following question: ‘I can see that the gas is hot. But why does it also
have high mean kinetic energy?’ You should think, in particular, that
the question rests on a false presupposition. It presupposes that there is
a gap between the gas’s being hot and its having high kinetic energy—a
gap that should be plugged with a bit of theory. But to accept the
‘just is’-statement is to think that the gap is illusory. There is no need to
explain how the gas’s being hot might be correlated with its having
high mean kinetic energy because there is no difference between the
two: for a gas to be hot just is for it to have high mean kinetic energy.
The decision whether to treat the gap as closed is partly a termino-
logical issue. (How should we use the word ‘heat’?) But in interesting
11
This is a badly inaccurate statement of the thermodynamic theory of heat. Fortu-
nately, the inaccuracies are harmless in the present context.
cases the terminological issue is tied up with substantial theoretical

issues. (Is the thermodynamic theory of heat superior to the caloric
theory of heat?) And it isn’t always easy to separate the two.
Rejecting a ‘just is’-statement comes at a cost, since it increases
the range of questions that are regarded as rightfully demanding
answers (why does this hot gas have high kinetic energy?), and
therefore the scenarios one treats as intelligible (there is a hot gas
with low mean kinetic energy). But having extra scenarios to work
with can also prove advantageous, since it makes room for addi-
tional theoretical positions, some of which could deliver fruitful
theorizing. (A proponent of the caloric theory of heat, for example,
would want to make room for a scenario in which a substance is hot
because it contains high quantities of caloric fluid, even though it is
not made up of particles with high kinetic energy.) Disagreement
about whether to accept a ‘just is’-statement is best thought of as
disagreement about whether the additional theoretical space would
be fruitful enough to justify paying the price of having to answer a
new range of potentially problematic questions.12
We have been focusing on an example from the natural sciences,
but our conclusions carry over to ‘just is’ -statements in metaphys-
ics. One has to balance the cost of rejecting the relevant statement—
an increase in the range of questions that are regarded as rightfully
demanding answers—with the cost of accepting the statement—a
decrease in the range of theoretical resources one has at one’s
disposal.
There is no quick-and-easy criterion for determining whether the
extra theoretical space is fruitful enough to justify paying the price
of having to answer a new range of potentially problematic ques-
tions. The only reasonable way to proceed is by rolling up one’s
sleeves and doing metaphysics.13
Suppose we are considering whether to accept ‘for a time to be
present just is for it to have a certain relational property’. By accepting
the ‘just is’-statement one would eliminate the need to answer an
awkward question: what does it take for a time to be present simplic-
iter, as opposed to present relative to some time or other? But there is
12
For a more detailed discussion of these matters, see Rayo (forthcoming) "The
Contruction of Logical Space"
13
Here I am indebted to Andrew Graham’s PhD thesis.
108 | Agustín Rayo
a price to be paid, because it is not immediately obvious that one will

have the theoretical resources to explain the feeling that there is some-
thing special about the present. By rejecting the ‘just is’-statement, on
the other hand, one would be left with a gap to fill—one needs to
explain what it is for a time to be present simpliciter as something over
and above being present relative to some time or other. One could try
to fill the gap by saying something like ‘to be present simpliciter is to
be at the edge of objective becoming’, and thereby introduce a new
theoretical resource. It is not immediately obvious, however, that such
a move would lead to fruitful theorizing, or be especially effective in
explaining the feeling that there is something special about the present.
The decision whether to accept the ‘just is’-statement is a decision
about how to best negotiate these competing theoretical pressures.
Here is a second example. Suppose we are considering whether
to accept ‘to experience the sensation of seeing red just is to be in a
certain brain state’. What sorts of considerations might be used to
advance the issue in an interesting way? Jackson’s Knowledge
Argument immediately suggests itself:14
Mary is confined to a black-and-white room, is educated through black-
and-white books and through lectures relayed on black-and-white televi-
sion. In this way she learns everything there is to know about the physical
nature of the world . . . If physicalism is true, she knows all there is to know.
For to suppose otherwise is to suppose that there is more to know than
every physical fact, and that is just what physicalism denies . . . It seems,
however, that Mary does not know all there is to know. For when she is let
out of the black-and-white room or given a color television, she will learn
what it is like to see something red, say. (Jackson (1986))
What Jackson’s argument brings out is that physicalists face a chal-

lenge. They must somehow accommodate the fact that it seems like
Mary acquires information about the world—information she did
not already have—when she first experiences the sensation of see-
ing red, even though physicalism appears to entail that she does
not. My own view is that the challenge can be met.15 But someone
who thinks that the challenge cannot be met might see the argu-
ment as motivating the introduction of possibilities that a physical-
14
See Jackson (1982) and Jackson (1986); for a review of more recent literature, see
Byrne (2006).
15
See Rayo (forthcoming) “Neo-Fregeanism reconsidered”, ch. 4.
ist would regard as unintelligible. According to the physicalist, to

experience the sensation of seeing red just is to be in a certain brain
state. So it makes no sense to consider a scenario in which someone
is in the brain state but lacks the sensation. If, however, one were to
give up physicalism and countenance the intelligibility of such a
scenario, one might be able to relieve some of the pressure gener-
ated by Jackson’s argument. For one could claim that, even though
Mary knew all along that she would be in the relevant brain state
when she was first shown a ripe tomato, she did not yet know if she
would also experience the relevant sensation. It is only after she is
actually shown the tomato, and experiences the relevant sensation,
that she is in a position to rule out a scenario in which she is in the
brain state without having the sensation. And this ruling out of sce-
narios substantiates the claim that Mary does indeed acquire infor-
mation about the world when she is first shown the tomato.
I think there are good reasons for resisting this way of addressing
the puzzle. (See, for instance, Lewis (1988).) But suppose one takes
it to work. Suppose one thinks that by creating a gap between being
in the relevant brain state and experiencing the relevant sensation—
and thereby making room for the possibility of being in the brain
state without having the sensation—one can adequately account for
a case like Mary’s. Then one will be motivated to give up the ‘just
is’- statement that keeps the gap closed (‘to experience the sensation
of seeing red just is to be in a certain brain state’). But doing so
comes at a cost because it opens up space for awkward questions.
For instance: ‘I can see that Mary is in the relevant brain state. What
I want to know is whether she is also experiencing the relevant sen-
sation. I would like to understand, moreover, how one could ever
be justified in taking a stand on this issue, given that we would find
Mary completely indistinguishable from her zombie counterpart,
or from someone with “inverted” sensations.’
4.2 Avoiding the Tractarian Legacy

The Tractarian can be expected to reject Statements 4–7 from section 1.
But it seems to me that they are all cases in which the advantages of
accepting the ‘just is’-statement far outweigh the disadvantages. Con-
sider Tables. By accepting the claim that for there to be a table just is
110 | Agustín Rayo
for there to be some things arranged tablewise, one eliminates the need
to address an awkward question: what would it take for a region that
is occupied by some things arranged tablewise to also be occupied by
a table? It is true that one also loses access to a certain amount of theo-
retical space, since one is no longer in a position to work with scenarios
in which there are things arranged tablewise but no tables. It seems to
me, however, that this is not much of a price to pay, since the availabil-
ity of such scenarios is not very likely to lead to fruitful theorizing. (Not
everyone would agree; see, for instance, van Inwagen (1990).)
For similar reasons, it seems to me that Properties, Death, and
Dinosaurs are all eminently sensible ‘just is’-statements. Again,
not everyone will agree. But I hope to have convinced you that
these statements shouldn’t be rejected merely on the basis of syntac-
tic considerations. They should be rejected only if one thinks that
the resulting theoretical space leads to theorizing that is fruitful
enough to pay the price of answering awkward questions.
And the relevant questions can be very awkward indeed. By
rejecting Dinosaurs, for example, one is forced to concede that the
following is a legitimate line of inquiry:
I can see that there are no dinosaurs. What I want to know is whether it is
also true that the number of the dinosaurs is Zero. And I would like to
understand how one could ever be justified in taking a stand on the issue,
given that we have no causal access to the purported realm of abstract
objects.16
If, on the other hand, you accept Dinosaurs you will think that
such queries rest on a false presupposition. They presuppose that
there is a gap between the non-existence of dinosaurs and dino-
saurs’ having Zero as a number—a gap that needs to be plugged
with a philosophical account of mathematical objects. Dinosaurs
entails that the gap is illusory. There is no need to explain how the
non-existence of dinosaurs might be correlated with dinosaurs’
having Zero as a number because there is no difference between the
two: for the number of the dinosaurs to be Zero just is for there to
be no dinosaurs.17
16
That this is a legitimate line of inquiry is famously presupposed by Benacerraf
(1973).
17
For an account of mathematics along these lines, see Rayo (2009) and Rayo
(forthcoming) The Construction of Logical Space.
Of course, you won’t see the closing of this theoretical gap as a

real benefit unless you think that the resulting theory is consistent
with a sensible metaphysical picture of the world, and unless you
think that it gives rise to a sensible philosophy of mathematics. The
remainder of this paper will be devoted to addressing the first of
these two challenges: I will argue that a compositionalist incurs no
untoward metaphysical commitments by accepting a ‘just is’-state-
ment like Dinosaurs.18
5. LIFE AS AN ANTI-TRACTARIAN
Let an anti-Tractarian be a compositionalist who accepts some meta-

physically contentious ‘just is’-statements. (Like Tractarianism, but
unlike compositionalism, anti-Tractarianism is a hybrid of linguis-
tic and metaphysical theses.19)
Anti-Tractarianism has a distinguished provenance. When Frege
claims that the sentence ‘there is at least one square root of 4’
expresses the same thought as ‘the concept square root of 4 is real-
ized’, and adds that ‘a thought can be split up in many ways, so that
now one thing, now another, appears as subject or predicate’ (Frege
(1892) p. 199), it is natural to interpret him as embracing the ‘just
is’-statement:
For the concept square root of 4 to be realized just is for there to be at least one
square root of 4.
And when he claims, in Grundlagen §64, that in treating the judge-

ment ‘line a is parallel to line b’ as an identity, so as to obtain ‘the
direction of line a is identical to the direction of line b’, we ‘carve up
the content in a way different from the original way’, it is natural to
interpret him as embracing the ‘just is’-statement:
18
I address the second challenge in Rayo (forthcoming) The Construction of Logical
Space, where I develop a semantics for mathematical discourse and an account of
mathematical knowledge.
19
Tractarianism and anti-Tractarianism are incompatible with each other, but they
are not contradictories. A compositionalist who accepts no metaphysically conten-
tious ‘just is’-statements would reject them both. So would someone whose views of
reference fall somewhere between Tractarianism and compositionalism.
112 | Agustín Rayo
For the direction of line a to equal the direction of line b just is for a and b to
be parallel.
More recent texts with broadly anti-Tractarian sympathies include

Parsons (1974), Wright (1983), Rosen (1993), Stalnaker (1996), and
Burgess (2005).20 My impression is that many contemporary meta-
physicians are nonetheless suspicious of anti-Tractarianism. The
purpose of this section is to get clear about what the view entails,
and what it does not.
Realism
A Tractarian might be tempted to complain that if anti-Tractarian-
ism were correct, there would fail to be a definite fact of the matter
about how the world is. I have sometimes heard arguments such as
the following:
Say you believe that for the number of the dinosaurs to be Zero just is for
there to be no dinosaurs. You believe, in other words, that a single fact can
be described fully and accurately by asserting ‘the number of the dinosaurs
is Zero’ and by asserting ‘there are no dinosaurs’. This presupposes that a
single fact can get carved up into objects and properties in different ways.
When the fact is described by asserting ‘the number of the dinosaurs is
Zero’, it gets carved up into an individual (the number Zero), a first-order
property (the property of being a dinosaur), and a second-order function
(the function taking first-order properties to their numbers); when it is
described as ‘there are no dinosaurs’, it gets carved out into a first-order
property (the property of being a dinosaur) and a second-order property
(non-existence).
But if this is so, there can’t be an objective, language-independent fact of
the matter about whether there are numbers. It all depends on how we
choose to describe the world.
I am happy to grant everything in the first paragraph of this argu-

ment (as long as the metaphor of fact-carving is spelled out prop-
erly; see section 6). The argument’s second paragraph, on the other
hand, strikes me as deeply misguided.
20
Hirsch is a compositionalist (see Hirsch (2002)). But it is not clear to me that he
is also an anti-Tractarian.
The anti-Tractarian is certainly committed to the view that a sin-

gle feature of reality can be fully and accurately described in differ-
ent ways. But this does not entail that there is no fact of the matter
about how the world is. On the contrary: it is strictly and literally
true that the number of the dinosaurs is Zero, and therefore that
there are numbers. And this is so independently of which sentences
are used to describe the world—or, indeed, of whether there is any-
one around to describe it. The point is simply that the relevant fea-
ture of the world could also be fully and accurately described in
another way: by asserting ‘there are no dinosaurs’.
Moral: If realism is the view that there is a definite, subject-
independent fact of the matter about how the world is, then anti-
Tractarianism is no less of a realist position than Tractarianism.
The World as a Structureless Blob

‘Wait a minute!’—you might be tempted to reply—‘Isn’t the anti-
Tractarian still committed to the view that the world is a structure-
less blob?’
Absolutely not. The anti-Tractarian believes that it is strictly and
literally true that there are tables, that a death took place, that the
number of the dinosaurs is Zero, and so forth. So if the strict and
literal existence of tables, deaths, and numbers is enough for the
world not to be a structureless blob, then it is no part of anti-Tractar-
ianism that the world is a structureless blob.
Perhaps what you mean when you say that the world might be
a structureless blob is that the world might fail to be endowed with
metaphysical structure. In that case, you should think that anti-
Tractarianism is neutral with respect to the question of whether the
world is a structureless blob. It is compatible with anti-Tractarian-
ism that reality be carved up by the world’s metaphysical structure.
The point is simply that such a carving is not presupposed by ordi-
nary language.
A brief aside: I don’t really understand what Putnam has in
mind when he talks about Internal Realism. But perhaps one
could interpret some of what he says as an endorsement of an
anti-Tractarian form of realism. (See, for instance, Putnam (1987)
pp. 18–19.)
114 | Agustín Rayo
Comprehensivism
Comprehensivism is that view that it is in principle possible to give a
comprehensive description of the world—a description such that: (1)
there is precisely one way for the world to be that would satisfy the
description, and (2) the world, as it actually is, satisfies the description.
A critic might be tempted to think that anti-Tractarianism is incompat-
ible with comprehensivism. ‘According to anti-Tractarianism’—the critic
might argue—‘the same fact can be described in many different ways.
One can say that there is a table, or that some things are arranged table-
wise, or that the world tableizes, or that tablehood is instantiated, or that
two half-tables are put together in the right sort of way, and so forth,
with no natural end. But one hasn’t given an exhaustive description of
the world until one has described it in all these ways. So the anti-Tractar-
ian could never give a comprehensive description of the world.’
To see where the critic goes wrong, it is useful to consider an
example. Suppose I hand you a box and ask you to give me a com-
prehensive description of its contents. You examine it and say:
‘There is a hydrogen-1 atom in such-and-such a state, and nothing
else.’ It would be inappropriate for me to respond by complaining
that your answer is incomplete on the grounds that it failed to men-
tion at least two objects: a proton and an electron. Such a response
would be guilty of double-counting. Part of what it is for there to be
a (non-ionized) hydrogen-1 atom is for there to be a proton and an
electron. So when you mentioned that there was a hydrogen-1 atom,
the presence of protons and electrons was already included in the
information you gave me. It is true that you never mentioned pro-
tons and electrons explicitly. But that was not required for your
description to be comprehensive. All that comprehensiveness
requires is that there be precisely one way for the contents of the
box to be such that it would satisfy your description.
Moral: Anti-Tractarianism does not entail that comprehensivism
is false. What it entails is that there could be more than one way of
giving a fully comprehensive description of the world.
Paraphrase
It is tempting to think that in accepting a ‘just is’-statement one com-
mits oneself to the availability of a paraphrase-method for translating
vocabulary that appears on one side of the statement into vocabulary

that appears on the other. Consider, for example, an anti-Tractarian
who accepts every instance of the following schema:
Numbers
For the number of the Fs to be n just is for there to be exactly n Fs.
It is tempting to think that she should also be committed to the

claim that arbitrary arithmetical statements can be paraphrased as
statements containing no mathematical vocabulary.
It seems to me that this would be a mistake. The availability of a
suitable paraphrase-function depends on the expressive richness of
one’s non-mathematical vocabulary. And the decision to accept
Numbers should be based on a cost–benefit analysis of the sort sug-
gested above, not on whether one has access to a powerful enough
stock of non-mathematical linguistic resources.
It is easy to overestimate the importance of paraphrase when one
sees things from the perspective of a nominalist: someone who
thinks that numbers don’t exist. For a nominalist might think that
non-mathematical paraphrases are needed to give an adequate
statement of our best scientific theories, and of the real content of
our mathematical accomplishments. But a friend of Numbers is no
nominalist, and would see little advantage in stating our scientific
theories or mathematical accomplishments in a non-mathematical
language. Suppose, for example, that ‘there is an even number of
stars’ can be paraphrased as a non-mathematical statement, ϕ. In all
likelihood, ϕ will be significantly more cumbersome than its math-
ematical counterpart. And a friend of Numbers will think that its
truth-conditions impose no less of a demand on the world, since
she will think that for ϕ to be the case just is for there to be an even
number of stars.21 She will therefore see little point in reformulating
her scientific or mathematical theorizing in terms of ϕ.
The question of whether it is possible to paraphrase arbitrary
mathematical statements as statements containing no mathematical
vocabulary is an interesting one, and I take it up in Rayo (forthcom-
ing) The Construction of Logical Space. The present point is simply
that one should not confuse Numbers with the view that a suitable
paraphrase-function exists. Accepting a ‘just is’-statement is one
21
For a related point, see Alston (1957).
116 | Agustín Rayo
thing; committing oneself to the availability of paraphrase-func-

tions relating vocabulary on either side of the statement is another.
6. ABSOLUTE GENERALITY
In this section I will consider the question of whether an anti-Trac-

tarian should think that there is such a thing as an all-inclusive
domain.
The first thing to note is that there are several different ways of
cashing out the claim that there is such a thing as an all-inclusive
domain:
• First Reading [Realism + Comprehensivism]
There is a definite fact of the matter about how the world is,
and it is in principle possible to give a fully comprehensive
description of its contents.
• Second Reading [Metaphysical Absolutism]
There is a ‘fundamental domain’—a domain consisting of the enti-
ties that are carved out by the world’s metaphysical structure.
• Third Reading [Recarving-Absolutism]
There is a ‘maxi-domain’—a domain consisting of the entities
that result from every possible way of carving up the world
into objects.
What should the anti-Tractarian say about the existence of an all-
inclusive domain, on each of these readings?
We have seen that anti-Tractarianism is compatible with both
Realism and Comprehensivism. So, on the first of the three read-
ings, there is no tension between anti-Tractarianism and the exist-
ence of an all-inclusive domain.
What about the second reading? Anti-Tractarianism is neutral with
respect to the existence of a ‘fundamental domain’. To address the
issue of a fundamental domain would require deploying the notion of
metaphysical structure, and anti-Tractarianism does no such thing.
Let us therefore turn our attention to the third reading. The anti-
Tractarian believes that there are tables. So a ‘maxi-domain’ would
have to include tables. But according to the anti-Tractarian, the fact
that there are tables could also be described by saying that there are
half-tables put together in the right sort of way, or that the property
of tablehood is instantiated, or that some mereological simples are

arranged tablewise, or that the set of tables is non-empty, or that the
number of tables is greater than Zero. So the maxi-domain would
also have to include half-tables and instantiated properties of table-
hood and mereological simples arranged tablewise and non-empty
sets and numbers greater than Zero and Zero itself, and so forth.
Could such a list ever be completed? It seems to me that anti-
Tractarians should be skeptical about the claim that it could. It is not
that an anti-Tractarian should think that the world is somehow
incomplete. The problem is that there is no reason to think that our
concept of ‘carving the world into objects’ is determinate enough to
allow for a final answer to the question of how it might be possible
to carve up reality into objects. Let me explain.
Unpacking the ‘Carving’ Metaphor

As I understand it, a ‘carving’ of the world is nothing more than a
compositional system of representation for describing the world.
In the most familiar case, a carving is a compositional system of
linguistic representation: a language in which the truth-conditions
of sentences are generated recursively from the semantic values of
a restricted set of basic lexical items. To say that a subject carves
the world into objects is simply to say that she represents the world
using a language that contains singular terms, or variables that
take singular term positions. Similarly, to say that a subject carves
the world into properties is simply to say that she represents the
world using a language that contains predicates, or variables that
take predicate positions.
Carving up the world is not like carving up a turkey. For the purposes
of spelling out the carving-metaphor, one is not to think of the world
as a big object—the mereological fusion of everything there is—and
of a carving as a way of subdividing the world into smaller parts.
The world, for these purposes, is to be thought of as ‘the totality of
facts, not of things’, and a carving is to be thought of as a composi-
tional system for describing these facts.22
22
Compare Eklund (2008).
118 | Agustín Rayo
When the carving metaphor is spelled out in this way, the

existence of a maxi-domain would require a final answer to the
question of what counts as a possible system of compositional
representation. And I see no prima facie reason to think that our
notion of representation (and our notion of linguistic representa-
tion, in particular) are constrained enough for this question to
have a definite answer. From the perspective of Tractarianism,
the range of admissible compositional languages is restricted by
metaphysical structure, since only languages whose semantic
structure is in correspondence with the metaphysical structure
of the world are potential vehicles for truth. From the perspec-
tive of anti-Tractarianism, on the other hand, the only constraint
on semantic structure is that it deliver an assignment of truth-
conditions to sentences from the semantic values of basic lexical
items. So it is hard to say in advance what would count as a pos-
sible compositional language. Whenever we dream up a new
mechanism for representing reality, the potential for a new com-
positional language—and hence for a new way of carving up the
world—will be in place.
An Analogy
An analogy might be helpful. Suppose you are told that the ordinals
are built up in stages. One starts with a ‘base’ ordinal, and at each
stage one gets a new ordinal by pooling together all the ordinals that
have been constructed so far. The process is to be carried out
indefinitely.
In the absence of further constraints, your understanding of
‘ordinal’ will be hopelessly incomplete. It will be consistent
with taking the ordinals to be isomorphic with the natural
numbers. But it will also be consistent with taking the ordinals
to be isomorphic with the natural numbers followed by an addi-
tional copy of the natural numbers—or two additional copies, or
three, or as many copies of the natural numbers as there are nat-
ural numbers. In fact, one’s understanding of ‘ordinal’ will be
consistent with taking the ordinals to be isomorphic with any
limit von Neumann ordinal.
Notice, moreover, that assuming that there is a definite plurality

of von Neumann ordinals wouldn’t bring a natural end to the
process. For although your understanding of ‘ordinal’ is consist-
ent with taking the ordinals to be isomorphic with the von Neu-
mann ordinals, it is also consistent with taking the ordinals to be
isomorphic with the von Neumann ordinals, followed by an
ω-sequence of additional objects—or two ω-sequences of addi-
tional objects, or an additional ω-sequence for each von Neumann
ordinal. And so on.
If you give me a definite characterization of ‘ordinal’, I can use
it to supply a significantly more generous one. (I can say, for
instance, ‘the ordinals are isomorphic to the structure you just
articulated followed by a copy of the structure you just articulated
for every point in the structure you just articulated’.) And, cru-
cially, I am not able to say anything definite about what it would
mean to continue this sort of process ‘all the way up’—anything
significantly more illuminating than the original invitation to
carry on the process ‘indefinitely’. The upshot is that there is no
sense to be made of an absolutely general ordinal-quantifier. But
this is not because of some dark metaphysical thesis about the
nature of ordinals, and it is not because of some mysterious limit
on our referential abilities: it is simply that the notion of ordinal
is not well enough defined.23
I think something similar holds for our general notion of lin-
guistic representation. If you give me a definite characterization
of ‘linguistic representation’, I suspect I’ll be able to use it to sup-
ply a significantly more generous one. If you give me a first-order
language, I can give you a second-order language. If you give me
an α-level language for some ordinal α, I can give you an α +
1-level language.24 But I’m not able to say anything very informa-
tive about what it would mean to iterate this process ‘all the way
up’—anything significantly more illuminating than the vague
idea that it could be carried out ‘indefinitely’. And, of course, the
order of the quantifiers is not the only dimension along which the
expressive power of a language might be expanded. If you give
23
For more on this sort of picture, see Parsons (1974).
24
For more on languages of transfinite order, see Linnebo and Rayo (typescript).
120 | Agustín Rayo
me a definite system of linguistic representation, there may be

other ways in which I can use it to supply a significantly more
generous one.
A Light-weight Conception of Objecthood?

You may be worried that my way of cashing out the carving meta-
phor is too lightweight. ‘If the only relevant difference between
asserting “there are tables” and asserting “some things are arranged
tablewise” is to do with the system of compositional representation
one chooses to employ’—you might be tempted to complain—‘then
someone who asserts “there are tables” hasn’t really committed her-
self to the existence of tables. For what she says could be true even
if there are really no tables.’
As far as I’m concerned, all it takes for there to really [table
thump!] be tables is for an English sentence like ‘there are tables’
to be strictly and literally true. And all it takes for ‘there are
tables’ to be strictly and literally true is that there be some things
arranged tablewise (equivalently: that the property of being a
table be instantiated; equivalently: that there be two half-tables
put together in the right sort of way; equivalently: that there be
tables).
Perhaps you mean something different by ‘really’. Perhaps what
you have in mind is that in order for something to really exist, it
must figure in a ‘fundamental’ description of the world. It must, in
other words, be carved out by the world’s metaphysical structure.
In this sense of real existence, the view defended in this paper is
neutral on the issue of whether there is anything that exists but
doesn’t ‘really’ exist.
A Language-infused World?
‘Wait a minute!’—you might be tempted to complain—‘Are you
setting forth a view according to which the existence of objects is
somehow constituted by language?’
Absolutely not. What is ‘constituted by language’ is the use of
singular terms. If we had no singular terms (or variables taking
singular-term positions) we wouldn’t be able to describe the
world in a way that made the existence of objects explicit. But

there would be objects just the same. Speakers of a language with
no singular terms can say things like ‘Lo, tableization!’. But for it
to tableize just is for there to be a table. So even without singular
terms, they would be in a position to convey information about
tables.
For the anti-Tractarian, the existence of tables depends entirely
on how the non-linguistic world is. If there are things arranged
tablewise (equivalently: if it tableizes; equivalently: if there are
tables), then there are indeed tables. If no things are arranged
tablewise (equivalently: if it fails to tableize; equivalently, if
there are no tables), then it is not the case that there are tables.
The Tractarians’ mistake is to conflate form and content. They
think there is a difference in content (i.e. truth-conditions)
between ‘there are tables’ and ‘some things are arranged table-
wise’, when in fact there is only a difference in form (i.e. seman-
tic structure).
7. CONCLUSION
I have argued that a ‘just is’-statement like Tables or Dinosaurs

shouldn’t be rejected on general linguistic or metaphysical grounds.
I argued, first, that Tractarianism is bad philosophy of language, and
suggested compositionalism as a promising alternative. Composition-
alists are in a position to accept a number of interesting ‘just is’-state-
ments. The possibility of accepting such statements can be extremely
valuable, because by accepting a ‘just is’-statement one eliminates the
need to address a certain kind of awkward question. (By accepting
Dinosaurs, for instance, one eliminates the need to answer a question
that many philosophers have found troubling: ‘I can see that there are
no dinosaurs, but why is it also true that the number of dinosaurs is
Zero’?)
I then argued that anti-Tractarians—compositionalists who
accept metaphysically contentious ‘just is’-statements—are not
saddled with an unattractive metaphysics. They are not commit-
ted to anti-realism, or to the view that the world is a structure-
less blob, or to the view that the existence of objects is constituted
122 | Agustín Rayo
by language. I noted, finally, that anti-Tractarians have reasons

to resist the claim that there is such a thing as a maxi-domain—a
domain consisting of the entities that result from every possible
way of carving up the world into objects.25
Massachusetts Institute of Technology
REFERENCES
Alston, W. (1957) “Ontological Commitments,” Philosophical Studies
8–17.
Beall, J., ed. (2003) Liars and Heaps, Oxford University Press, Oxford.
Beaney, M., ed. (1997) The Frege Reader, Blackwell, Oxford.
Benacerraf, P. (1973) “Mathematical Truth,” Journal of Philosophy 70,
661–79. Reprinted in Benacerraf and Putnam (1983).
—— and H. Putnam, eds. (1983) Philosophy of Mathematics, Cambridge
University Press, Cambridge, second edition.
Bennett, K. (2009) “Composition, Colocation and Metaontology.” In
Chalmers et al. (2009).
Block, N. (2002) “The Harder Problem of Consciousness,” The Journal of
Philosophy 99, 391–425.
—— and R. Stalnaker (1999) “Conceptual Analysis, Dualism, and the
Explanatory Gap,” Philosophical Review 108, 1–46.
Bueno, O. and Ø. Linnebo (2009) New Waves in Philosophy of Mathematics,
Palgrave Macmillan.
Burgess, J. (2005) “Being Explained Away,” Harvard Review of Philosophy
13, 41–56.
25
For their many helpful comments, I would like to thank Karen Bennett, Ross
Cameron, Matti Eklund, Andrew Graham, Caspar Hare, Jeremy Hartman, Øystein
Linnebo, Ricardo Mena, David Nicolas, Damien Rochford, Brad Skow, Bob Stalnaker,
Pedro Stepanenko, Eduardo Villanueva, Steve Yablo, and Dean Zimmerman. I would
also like to thank seminar participants at MIT, and audiences at the Paris Absolute
Generality Workshop, the MIT Work in Progress Seminar, the First Annual Meeting
of the Latin American Association for Analytic Philosophy, the Institut Jean-Nicod,
the University of Leeds, the University of Connecticut, and Institute for Philosophi-
cal Research at the National Autonomous University of Mexico. Portions of this
paper were written during the tenure of an ACLS Fellowship, for which I am
extremely grateful.
Byrne, A. (2006) “Review of There’s Something about Mary,” Notre Dame

Philosophical Reviews 2006.01.20. Available at <http://ndpr.nd.edu/
review.cfm?id=5561>.
Chalmers, D. (1996) The Conscious Mind: In Search of a Fundamental Theory,
Oxford University Press, New York.
—— D. Manley, and R. Wasserman, eds. (2009) Metametaphysics, Oxford
University Press, New York.
Eklund, M. (2008) “The Picture of Reality as an Amorphous Lump,” In
Sider et al. (2008).
—— (2009) “On Some Recent Criticisms of the ‘Linguistic’ Approach to
Ontology,” Dialectica 63, 313–23.
Fine, K. (2001) “The Question of Realism,” Philosophers’ Imprint 1, 1–30.
Frege, G. (1884) Die Grundlagen der Arithmetik. English Translation by
J.L. Austin, The Foundations of Arithmetic, Northwestern University
Press, Evanston, IL, 1980.
—— (1892) “On Concept and Object,” Vierteljahrsschrift für wissenschaftli-
che Philosophie 16, 192–205. The English translation by Peter Geach is
reprinted in Beaney (1997).
Goldfarb, W. (1997) “Metaphysics and Nonsense: On Cora Diamond’s
The Realistic Spirit,” Journal of Philosophical Research 22, 57–73.
Hacker, P. (1986) Insight and Illusion: Themes in the Philosophy of Wittgen-
stein, Oxford University Press, Oxford.
Heil, J. (2003) From an Ontological Point of View, Clarendon Press, Oxford.
Heim, I. and A. Kratzer (1998) Semantics in Generative Grammar, Blackwell
Textbooks in Linguistics, Blackwell, Oxford.
Hirsch, E. (2002) “Quantifier Variance and Realism,” Philosophical Issues
12, 51–73.
Hofweber, T. (2009) “Ambitious, Yet Modest, Metaphysics,” in Chalmers
et al. (2009).
Jackson, F. (1982) “Epiphenomenal Qualia,” Philosophical Quarterly 32,
127–36.
—— (1986) “What Mary Didn’t Know,” The Journal of Philosophy 83,
291–5.
Kanger, S. and S. Öhman, eds. (1980) Philosophy and Grammar, Reidel Pub-
lishing Company, Dordrecht.
Lewis, D. (1980) “Index, Context, and Content,” in Kanger and Öhman
(1980) pp. 79–100. Reprinted in Lewis (1998).
—— (1983) “New Work for a Theory of Universals,” Australasian Journal
of Philosophy 61, 343–77. Reprinted in Lewis (1999).
—— (1984) “Putnam’s Paradox,” The Australasian Journal of Philosophy 62,
221–36. Reprinted in Lewis (1999).
124 | Agustín Rayo
—— (1988) “What Experience Teaches,” Proceedings of the Russellian Soci-

ety 13, 29–57. Reprinted in Lewis (1999).
—— (1998) Papers in Philosophical Logic, Cambridge University Press,
Cambridge.
—— (1999) Papers in Metaphysics and Epistemology, Cambridge University
Press, Cambridge.
Linnebo, Ø. and A. Rayo (typescript) “Hierarchies Ontological and
Ideological.”
Morton, A. and S. Stich, eds. (1996) Benacerraf and his Critics, Basil Blackwell,
Oxford.
Parsons, C. (1974) “Sets and Classes,” Noûs 8, 1–12. Reprinted in Parsons
(1983), pp. 209–20.
—— (1983) Mathematics in Philosophy, Cornell University Press, Ithaca,
NY.
Pears, D. (1987) The False Prison: A Study of the Development of Wittgen-
stein’s Philosophy, volume 1, Oxford University Press, Oxford.
Putnam, H. (1987) The Many Faces of Realism, Open Court, La Salle, IL.
Rayo, A. (2002) “Word and Objects,” Noûs 36, 436–64.
—— (2003) “When does ‘Everything’ Mean Everything?” Analysis 63,
100–6.
—— (2009) “Towards a Trivialist Account of Mathematics,” In Bueno, O.
and Ø. Linnebo (2009) New Waves in Philosophy of Mathematics, Palgrave
Macmillan.
—— (forthcoming) “Neo-Fregeanism reconsidered,” in Abstractionism,
edited with Philip A. Ebert. Oxford University Press.
—— (forthcoming) The Construction of Logical Space, Oxford University
Press.
—— and T. Williamson (2003) “A Completeness Theorem for Unrestricted
First-order Languages,” In Beall (2003), pp. 331–56.
Rosen, G. (1993) “The Refutation of Nominalism (?),” Philosophical Topics
21, 149–86.
Schaffer, J. (2009) “On What Grounds What,” in Chalmers et al. (2009).
Sider, T. (typescript) Writing the Book of the World.
—— J. Hawthorne, and D. Zimmerman, eds. (2008) Contemporary Debates
in Metaphysics, Blackwell, Cambridge, MA.
Stalnaker, R.C. (1996) “On What Possible Worlds Could Not Be,” in
Morton and Stich (1996), 103–19. Reprinted in Stalnaker (2003),
40–54.
—— (2003) Ways a World Might Be: Metaphysical and Anti-Metaphysical
Essays, Clarendon Press, Oxford.
van Inwagen, P. (1990) Material Beings, Cornell University Press, Ithaca,

NY.
Wittgenstein, L. (1921) Tractatus Logico-Philosophicus, London. Published
as “Logisch-Philosophische Abhandlung,” in Annalen der Naturphiloso-
phische Vol. XIV, 3/4, 1921, pp. 184–262.
Wright, C. (1983) Frege’s Conception of Numbers as Objects, Aberdeen
University Press, Aberdeen.
III
HUMEANISM AND LAWS

OF NATURE
5. Goodbye, Humean Supervenience
Troy Cross
THE MARCH OF THE CAUSAL
The 1960s witnessed a remarkable string of causal theories. Func-

tionalism—in essence a causal theory of mental states—was first on
the scene (Putnam 1960) followed by a causal theory of perception
(Grice 1961), a causal theory of action (Davidson 1963), a causal the-
ory of knowing (Goldman 1967), a causal theory of events (Davidson
1969), and finally in 1970, a causal theory of reference (Kripke 1980).
Though the most famous causal theories were already in play by
the end of the decade, the trend had not run its course. In 1978,
David Armstrong defended the Eleatic Principle, which might be
characterized as a causal theory of being and followed it two years
later with the somewhat less dramatic causal theory of object persist-
ence (1978; 1980). It was in that same volume—Time and Cause: Essays
Presented to Richard Taylor—that Sydney Shoemaker published his
causal theory of properties, which is the focus of the present essay
(1980).
Why the causation fixation? Each of these causal theories, of
course, had roots in its local dialectical history. Goldman was
responding to the Gettier problem (1963), Kripke to the failures of
descriptivism, and so on. But there’s also no denying that causation
was in the air like a fine mist. It was the default analysans, followed
closely in popularity by related notions like counterfactual depend-
ence, chance, and nomic necessitation. Thus, Nozick rejected the
causal theory of knowing only to try a counterfactual theory of
knowledge instead, and Armstrong did likewise with nomic neces-
sitation (Nozick 1981; D.M. Armstrong 1973). But causation, coun-
terfactuals, laws, and chance all belong to the same tightly knit
family. The question, suitably qualified, remains: why were these
causal notions so popular?
One reason was the rise of naturalism in the period. Naturalism
attempts to shoehorn all meaningful talk of action, knowledge,
130 | Troy Cross
mind, reference, and the like into a non-primitively-mentalistic

world. It’s a tight fit. Most of the philosophical subjects up for natu-
ralistic treatment involve relations between persons and their envi-
ronments. To ground such relations appropriately, theorists had to
find topic-neutral substitutes, relations that could apply univocally
across the categories of mind and body, and of these there are pre-
cious few with any prospect of success: causation, counterfactual
dependence, and nomic necessitation nearly fill out the roster. So,
the ubiquity of causal analyses in philosophy during the rise of
naturalism is no surprise. The surprise is how brilliantly philosophers
managed (and still manage) with such crude implements.
In any case, naturalism did not halt its advance upon reaching
causal theories of mind, knowledge, action, reference, and the like,
but turned upon its own devices, the very ideas like causation, law,
chance, dispositions, and counterfactual dependence, which were
themselves marked as spooky theological vestiges in need of some
kind of reductive or eliminative treatment (Goodman 1983, 31–40).
The modest strategy here was to pick the most central of these
notions and use it to explain the others. But the more popular tradi-
tion, tracing back, perhaps, to David Hume, and culminating in
David Lewis, aimed for more. Throughout the 1970s, Lewis offered
ingenious reductions for the entire family of causal notions to facts
about the distribution of what he called “the perfectly natural prop-
erties”. Lewis reduced the laws of nature to maximally simple and
informative facts about the distribution of these properties (2001a,
73), reduced counterfactuals to complicated similarity relations
(1979), and reduced causation to chains of counterfactual depend-
ence (1973). Here is his purist vision in canonical form:
Humean supervenience is named in honor of the greater denier of necessary con-
nections. It is the doctrine that all there is to the world is a vast mosaic of local
matters of particular fact, just one little thing and then another. (But it is no part
of the thesis that these local matters are mental.) We have a geometry: a system of
external relations of spatiotemporal distance between points. Maybe points of
spacetime itself, maybe point-sized bits of matter or aether or fields, maybe both.
And at those points we have local qualities: perfectly natural intrinsic properties
which need nothing bigger than a point at which to be instantiated. For short: we
have an arrangement of qualities. And that is all. There is no difference without
difference in the arrangement of qualities. All else supervenes on that. (Lewis
1987, ix–x)
Goodbye, Humean Supervenience | 131
Lewis’s conjecture holds undeniable “desert landscape” appeal: the

distribution of point-sized qualities is a strikingly sparse ontologi-
cal base. I think Shoemaker’s causal theory of properties, however,
was no less ambitious. Rather than trying to reduce away all of the
causal notions to something else, Shoemaker instead anchored to
the idea of a causal power itself. So far, this sounds like the modest
strategy mentioned above, but Shoemaker also used causal powers
to regiment something lying outside the small causal circle, viz., the
notion of a natural property, which Lewis himself could only regard
as an untamed primitive. In this way, Shoemaker claimed equal
rights, alongside Lewis, to the coveted title of “most naturalistic”,
and—perhaps along with Armstrong’s Eleatic pronouncements—
brought the sixties-era celebration of the causal to its climax.
I’m going to discuss Shoemaker’s paper in this essay, but my goal
is not intellectual history or critical exegesis. My goal, rather, is to
offer a new argument for Shoemaker’s theory. Like Shoemaker, I
will argue that the intrinsic properties of concrete objects are each
necessarily correlated with a unique causal power, i.e., that proper-
ties have dispositional shadows. But unlike Shoemaker, I will make
no use of (even slightly) controversial premises in my argument.
Furthermore, I will note that if we take necessarily co-extensive
properties to be identical then it follows that the intrinsic properties
of concrete objects are identical to powers, i.e., that properties are
nothing distinct from their dispositional shadows, i.e., that proper-
ties are dispositions.
Though my argument is perfectly general, it’s most instructive to
consider it as a direct attack on Lewis’s reductive project. If I am
right, then though Lewis never saw this implication, his Humean
accounts of the dispositions, laws, chances, and the like, together
with his theory of property identity, commit him to saying that each
of his beloved qualities—his purely intrinsic, categorical, local
properties—is in fact a disposition. The famous “Humean super-
venience” excerpt above would represent Lewis’s metaphysics with
equal veracity if it were to substitute “dispositions” for “qualities”.
(The horror!)
In order to resist my conclusion, Lewis must either abandon his
reductionism about the dispositional/causal/nomic/subjunctive in
favor of full-bore eliminativism or else abandon his modal criteria
of property individuation and take refuge in hyperintensional
132 | Troy Cross
distinctions, i.e., of cases where a disposition and a quality are nec-

essarily co-extensive, Lewis would have to say those dispositions
and qualities are nevertheless wholly distinct, and that only the
qualities are perfectly natural. Either strategy would represent not
only a departure from the letter of Lewis’s metaphysics, but also a
betrayal of its spirit.
SHOEMAKER’S THEORY
Shoemaker’s thesis in “Causality and Properties” is that the intrin-

sic properties of concrete objects are uniquely and essentially cor-
related with the powers those properties bestow (1980). The view
was not entirely novel, prefigured as it was in the work of Achin-
stein (1974), Harre and Madden (1975), Kneale (1952), Putnam
(1970), Sellars (1948), and, indeed, perhaps in nearly every philoso-
pher’s work before the early occasionalists, al-Ghazali and Gabriel
Biel, wrested powers out of ordinary things and granted them to
God alone (Freddoso 1986). But Shoemaker improves upon his
predecessors in two noteworthy respects. First, he defines the view
with a new measure of precision, identifying powers by appeal to a
function from circumstances to effects, <ci,ei> such that objects pos-
sessing the power and in ci “have a certain effect”: ei, then claiming
that every property necessarily and uniquely correlates with a func-
tion from other properties to these powers:
Just as powers can be thought of as functions from circumstances to causal
effects, so the properties on which powers depend can be thought of as
functions from properties to powers or, better, as functions from sets of
properties to sets of powers. (Shoemaker 1980)
Since the function by which Shoemaker individuates properties itself

takes properties as arguments, and the circumstances and effects are a
matter of which properties are instantiated where, he is forced to
acknowledge, and embrace, the non-well-foundedness of his pro-
posal. Still, just as the individuation of sets by their elements and the
individuation of ur-elements by the sets to which they belong is
informative, though ultimately circular, Shoemaker’s theory promises
us a non-trivial, substantive constraint on the identity of properties.
(If we’re willing to allow that circumstances are not merely extrin-
sic, but can include intrinsic components as well, we may simplify
Shoemaker’s formulation, neatly folding the “other properties”

into the circumstances that identify a power. For example, if F is
identified by the function from another property, G, to the power to
bring about E in circumstances C, we can call the conjunction of C’s
holding, together with the object in question’s having G, “C*”. Now,
F can be identified by appeal to the function from C* to E.)
The second notable advance in “Causality and Properties” is the
bevy of arguments, both epistemic and semantic, that Shoemaker
marshaled in support of the newly clarified view. If properties were
untethered from the powers of objects, he claimed, we could neither
know much of anything nor successfully refer to the properties of
things. Since we do have knowledge of the world, however, and
also manage to refer to the properties of objects, properties are, in
fact, tied to the causal powers they bestow in just the way his theory
describes.
Though much discussed in the three decades since the publica-
tion of “Causality and Properties,” Shoemaker’s epistemic and
semantic considerations have not fared as well as his articulation of
the view itself. To the contemporary philosophical ear they mani-
fest a peculiar, twentieth-century anxiety about individuation,
knowledge, and reference. Thus, although the theory itself boasts
many fans among contemporary metaphysicians, it is rare indeed
to find an approving citation of Shoemaker’s arguments for it (Bird
2007, 263).
I shall briefly discuss the arguments, and in the predictably dis-
approving manner. But recall that my purpose in so doing is not
refutation. Rather, it is to expose a new and powerful line of reason-
ing for Shoemaker’s view.
EPISTEMIC ARGUMENTS AND REPLIES
Shoemaker’s central argument in “Causality and Properties” is a

sort of reductio. He first asks us to assume that properties are not tied
to powers in the manner he proposes and then points out that all
sorts of new skeptical scenarios are possible.
For instance, suppose there could be causally inert properties.
Then two very similar-appearing objects—say, two ballpoint pens
from the same package—may in fact be vastly dissimilar, because
134 | Troy Cross
one is stacked with multitudes of inert properties that the other

lacks. This may be the case despite the pens having exactly the same
causal powers throughout their lifetimes. Likewise, two very dis-
similar-appearing objects—say, a pen and a person—may in fact be
highly similar, by virtue of sharing a multitude of inert properties.
Further, if properties can endow different powers at different times,
it may be that an object appearing to change—say, undergoing a
pen-to-person transformation—actually isn’t changing its intrinsic
properties at all; the intrinsic properties are remaining fixed, while
the property/power relation shifts. And finally, it may be that what
appears to be unchanging is, in fact, undergoing radical change in
its intrinsic properties while the relation between properties and
powers evolves in a disguising, compensating fashion.
According to Shoemaker, if such skeptical scenarios were pos-
sible we could not have any knowledge of similarity and change.
But we do have such knowledge. Therefore, the scenarios are
impossible.
It’s an odd way of doing modal metaphysics. After all, it’s never
argued that the envisioned scenarios embed any hidden conceptual
confusion. So it’s not clear how ruling them out metaphysically is
supposed to do any epistemic work. Let me put the problem as a
dilemma. Upon first considering the scenarios, we either know
we’re not in them or we don’t know we’re not in them. In the former
case, we needn’t worry about whether they’re metaphysically pos-
sible, so long as they’re not actual (or nearby). But in the latter case,
it seems we also don’t know Shoemaker’s metaphysics is correct.
That is to say, either Shoemaker’s ad hoc modal surgery is unneces-
sary or else we lack the nerve to perform it.
Fortunately, standard epistemology itself saves us from system-
atic ignorance, and without a modal metaphysics driven by wish
fulfillment. As many have pointed out (Cross 2004; Hawthorne
2001; Schaffer 2004), standard anti-skeptical theories of knowl-
edge, or of “knowledge”, of the traditional fallibilist, reliabilist,
subject-sensitive invariantist, contextualist, and relativist varie-
ties, all address Shoemaker’s skeptical challenge with exactly the
same methods—and just as successfully (or unsuccessfully)—as
they address Descartes’ evil genius scenario. Epistemically speak-
ing, there is nothing to see here, nothing new. Move along,
everyone!
Although the foregoing replies deflect Shoemaker’s argument,

the most satisfying is a tu quoque. Analogous skeptical worries about
similarity and change arise even if properties are linked to powers
in exactly the way that Shoemaker suggests. To see how this is so,
first observe that one of the central motivations for dispositionalism
is the thought that dispositional differences between objects need
never be manifest in the actual world. Moreover, these dispositions
might contain wholly novel manifestations, alien to the actual
world. For example, two fundamental particles might never actu-
ally meet, but if they did, their collision would result in a brand
new kind of fundamental particle, an “alien” property (C.B. Martin
1993; C.B. Martin 1997). Of course, this alien property itself would
have powers, perhaps powers that would be manifest only upon
reacting with other fundamental kinds of particles in the actual
world, perhaps powers to produce further fundamental alien prop-
erties, and so on.
Thus, actual properties may be sensitive to aliens; the actual
world may contain any number of properties with potential effects
that would be manifest only in the company of an alien. Call such
properties “alien-sensitive powers”, and call their triggers “alien
catalysts”. Now we’re in a position to mirror Shoemaker’s skeptical
scenarios.
Suppose a property is sensitive only to an alien catalyst. While
not strictly speaking inert, it is for all practical purposes undetect-
able, requiring, as it does, an alien trigger. Call such properties
“quasi-inert”. Now we can load up one ballpoint pen, but not the
other, with multitudes of these quasi-inert properties, each of
which has different alien sensitivities. The two pens will seem,
under any actual test, to be exactly alike, while in fact being vastly
dissimilar. Quasi-inert properties can also make a pen and person
more alike than a pen and another pen, by the same means, mutatis
mutandis.
We can also mimic property/power swapping, as follows. Let F
and F’ differ only with respect to (actual) catalysts C and C’. In the
presence of C, F endows power P but F’ does not. In the presence of
C’, F’ endows power P but F does not. Imagine that at time t, every-
thing F switches to F’ and everything C switches to C’. Though
everything will change, no change will be detected. Likewise for
the case of objects seeming to change (pen-to-person, say) when
136 | Troy Cross
they are in fact static, which could be the result of the introduction
of new catalysts, rather than an actual intrinsic change.
Though Shoemaker’s envisioned skeptical scenarios and the ana-
logues that arise from within dispositionalism itself are subtly dif-
ferent, standard epistemologies will not draw a bright line between
them. Shoemaker may have called our attention to a new and dev-
astating species of skeptical argument, but he has failed to make a
convincing case for dispositionalism.
THE ARGUMENT THAT LEWIS’S LOCAL QUALITIES

SATISFY SHOEMAKER’S CONDITIONS
The purpose of the previous section was not primarily to review

Shoemaker’s epistemic arguments, but to introduce and motivate
the notion of alien sensitivity, the idea that dispositions may have
radically non-actual activation conditions. Once we see that Shoe-
maker’s criterion distinguishes properties whose dispositional dif-
ferences are only revealed under alien circumstances, we are very
close to recognizing that even paradigmatically categorical proper-
ties are powers.
Begin by asking yourself whether Lewis thinks there are possible
conditions such that, for arbitrary perfectly natural property F, F
endows some power to objects in those conditions.
The answer, quite clearly, is yes. Lewis’s view is not that perfectly
natural properties cannot endow powers. Rather, it is that they
endow different powers in different circumstances. Exactly how the
powers granted by F vary with circumstances is a complicated mat-
ter. A nomic theory of powers would say something like the follow-
ing: in worlds where it follows from the laws of nature that Fs
become Gs, F endows the power to become a G. And in general, in
worlds where it is nomically necessary that Fs do something, then F
endows the power to do that thing. There is reason to think Lewis
does not hold the nomic theory of powers, but it is nevertheless
very close to his view, and helpful to bear in mind.
But we needn’t worry about precisely which conditions are
required for F to endow the power to become G. What matters is
that there are possible conditions in which F disposes things to
become Gs. These conditions may be states of the whole world, as
on the nomic theory of powers, or states of subregions of a world,

if, for instance, F disposes things to become Gs in one part of the
world but not in another. The important thing is that the regions of
modal space at which F disposes things to become G are not ran-
dom, but a matter of how powers generally supervene on the distri-
bution of Humean facts.
Let “C” name the set of conditions under which F endows the
disposition to become G, (or the conditions under which F-ness dis-
poses things to become G). Lewis should accept the following with-
out objection:
(1) In C, F disposes things to become Gs.
Now to the substantive claim. Having said (1), I think Lewis must
also say that:
(2) F disposes things to become G in C.
The step from (1) to (2) seems to add metaphysical weight, because
(1) contingently correlates F with a disposition, while (2) pushes
that contingency into the very specification of the disposition itself,
yielding a necessary connection between F and a disposition. While,
according to (1) there are certain conditions under which F disposes
things to become G, according to (2), F, everywhere in modal space,
disposes things in a certain way: to become G under conditions C.
The transition is hard to resist. When someone says, “In the Sun,
ice is disposed to melt”, we paraphrase them seamlessly as saying
that ice is disposed to melt in the Sun. The same can be said for the
habitual, e.g., ice melts in the Sun. Here it is even clearer. One does
not hesitate in the inference from “In the Sun, ice melts,” to “Ice
melts in the Sun”. Yet, on at least some theories of dispositions, the
habituals are true only if being in the Sun ice is disposed to melt,
and ice is disposed to melt in the Sun, respectively (Fara 2005).
Compare also: (a) This joke is funny in England; and (b) In Eng-
land, this joke is funny. It is very hard to see the difference. Suppose
(a) and (b) are uttered in the US. (a) directly attributes a power to
the joke: it has the power to make people laugh in England. (b) does
not directly attribute a power, but says that the joke has a power
under other conditions: being in England. Yet, it would be absurd to
halt the inference from (a) to (b) or vice versa. “Yes, of course, if the
joke is told in England, then it has the power to amuse audiences.
138 | Troy Cross
But don’t go thinking that means that here in the US, the joke has
any power to amuse audiences in England!” Bizarre.
I think this is just a general feature of power and disposition talk.
F disposes things to become G in C iff in C, F disposes things to become G.
And it is easiest to see for habituals: Fs G in C iff in C, Fs G.
It is worth asking why the analogous inference is invalid for sub-
junctive conditionals. At least according to the standard semantics
for subjunctives, it does not follow from (x)(Fx >cf (Gx > cf Hx)) that (x)
(Gx > cf (Fx > cf Hx)). It may be true that all soluble things are such that
if they were immersed in water, they would dissolve, but false that all
things actually immersed in water would dissolve if they were solu-
ble. Suppose the only soluble things are not in water, and that noth-
ing would prevent their dissolving if put in water, but that objects in
the water are such that if they were to become soluble, they would be
prevented from dissolving. Perhaps a sorcerer guards over all of the
immersed objects, and would cast a spell instantly removing them
from water, or else in some way interfering with the dissolution proc-
ess in some way at the microphysical level, should they become solu-
ble. This sorcerer, however, may be perfectly content to let soluble
objects that are not actually immersed dissolve, should they be put in
water. He watches over only the immersed objects.
So in general, it would be a bad idea to infer from the fact that if
something were F, then if it were G, it would be H to the fact that
if something were G, then if it were F, it would be H. Given the
undeniably close relation between disposition attributions and sub-
junctive conditionals, we must take care to avoid that fallacious
inference in another guise. Yet it seems undeniable that if in water,
soluble objects are disposed to dissolve, then soluble objects are dis-
posed to dissolve in water, and also that if in water, soluble objects
dissolve, then soluble objects dissolve in water.
Why is this so? I think the cases where the relevant inference fails
for subjunctives are precisely those cases where disposition ascrip-
tions succeed and counterfactual or subjunctive analyses appear to
falter. So-called “finkish” dispositions are dispositions that disap-
pear in their stimulus conditions (C.B. Martin 1994). Lewis imag-
ines a sorcerer guarding a fragile chalice, poised to cast a spell
making it not fragile if it is going to be dropped (1997). The chalice’s
disposition is finkish, vanishing exactly when called upon. It’s frag-
ile, intuitively, but would not break if dropped. Relatedly, Bird notes
that “antidotes” are external factors that disrupt and prevent the
manifestation of dispositions in the stimulus conditions, e.g., the
antidote to a poisonous venom (1998). Here the venom has a
disposition to kill when ingested, even though it may be ingested
without causing death because of the interference of the antidote.
Finally, we can generate the reverse phenomenon, where a certain
subjunctive conditional holds without the associated disposition,
by appealing to what Mark Johnston calls “mimics” (1992). Imagine
a non-fragile piece of clay attached to a motion-sensing machine
that, if it detects jarring or dropping of the clay, will instantly stream
liquid nitrogen over the clay, causing it to shatter.
This is my claim: all counterexamples to the pattern of subjunctive
inference above are also cases of finks, antidotes, or mimics. That is why
the analogous form of reasoning is permissible in the case of pow-
ers, dispositions, and habituals.
I do not have a proof that all of the cases in which the relevant
inference pattern for subjunctive conditionals fails are cases of finks,
antidotes, or mimics, but you’re invited to submit counterexamples.
My only goal is to explain away the threat that swapping the
conditions in and out of the scope of the power operator inherits the
danger of the analogous operation for subjunctives. To dismiss this
threat, it’s enough to have strong initial intuitions about the legiti-
macy of the inference and a plausible explanation of the difference
between the good case (dispositions) and the bad case (subjunc-
tives), and that much, I think we have.
Even on a pure subjunctive conditionals account of dispositions,
Lewis is in trouble. While, in general, the form of inference with
subjunctives noted above is invalid, it may still be valid for a subset
of cases. For instance, suppose something is an F-detector iff it is
disposed to register “1” if in a world with an F. Now, we can make
this disposition to register “1” if in a world with an F as strong as
we like. So let’s make it a full-strength, deterministic, unconditional
power. Lewis might argue that there couldn’t be any such power.
But I don’t see why he would say that. For him, it would just be a
highly non-natural property. Now, isn’t every F such that, if it were
in a world with an F detector, the F detector would register “1”? If so,
then F underwrites a non-trivial subjunctive conditional. So, on a
subjunctive conditional account of dispositions, F underwrites a
disposition.
140 | Troy Cross
But perhaps we can deny that F entails the conditional. Perhaps

some Fs are guarded over by F-detector-averse sorcerers waiting to
cast spells should an F-detector threaten. If this supposition is
coherent, what we have here is a paradox equivalent to the exist-
ence of two omnipotent beings—the F-detector and the F-detector-
averting-sorcerer: they aren’t jointly possible. But now consider the
conditions of being alone in a world with an F-detector. If an F were in
those conditions, it wouldn’t have its protective sorcerer with it,
and the F detector would register “1”.
Let’s recap. If there are possible conditions, C, under which a
“categorical” property F disposes things to become G, then through-
out modal space, F disposes things to become G in C. The condi-
tions may be strange. They may involve non-actual laws. But
dispositions may have alien triggers, as we have seen, so this is
nothing distinctive. Some parts of the pluriverse are poised to send
the property into action. By the same token, the property, wherever
instantiated, is poised to go into action in those parts of the pluriv-
erse, and not in others. In this way, categorical properties, far from
the inert, modally innocent creatures they purport to be, are in some
sense modal monads, representing the full range of possible condi-
tions (but unlike monads, causally interacting as well).
That was the first question: do properties endow powers? It seems
they do. In fact, it is difficult to see how one could offer any theory of
how powers supervene on properties without giving us a recipe for
such endowment. Now comes the second question: is there a unique
power endowed by each natural property? After the last question, this
should be relatively easy. Suppose we start with two arbitrary proper-
ties, F and F*. Would Lewis think there are any conditions under which
F and F* would endow different powers? Of course. Again, the nomic
account of powers would say that in worlds where it’s a law that Fs
become Gs and F*s do not, F endows the power to become G and F*
does not. Lewis’s theory may be a bit subtler than that, but it doesn’t
matter. Given the wealth of the pluriverse, he will have such distin-
guishing conditions, for any two properties. Thus, we can find a single
power that is both necessarily correlated with F and unique to it by tak-
ing the union of each of the powers (sets of pairs of circumstances and
effects) that distinguishes F from F*, F from F**, and so on.
For rhetorical purposes, my discussion has meandered through
some of the specifics of Lewisian reductive metaphysics, but it must
be noted that my central claim in this section—that natural proper-

ties can be paired one-to-one with powers—depends on none of the
implementational detail. It depends only on each perfectly natural
property making a potential difference to the powers of things
instantiating that property, and on no two perfectly natural prop-
erties making all the same potential differences. I simply note a
general feature of powers, viz., that whatever the potential differ-
ences in powers that a natural property contributes to its bearers,
those potential differences are also actual differences in potential. This
is simply what powers do: they code potential differences as
actual ones.
Since Lewis aims to explain, rather than eliminate, causal pow-
ers, it should be uncontroversial that for him, every natural prop-
erty potentially contributes to the powers of things, and that no two
perfectly natural properties make all the same potential contribu-
tions. If one simply attends to what powers are, one can see that
every perfectly natural property is shadowed by a unique power,
i.e., Shoemaker’s thesis from “Causality and Properties” is true.
ANTI-HYPERINTENSIONALITY
Observe that for Lewis, the perfectly natural properties, like all
other properties, are sets of possibilia: “. . . the property of being
donkey comes out as the set of all donkeys, the donkeys of other
worlds along with the donkeys of ours” (Lewis 2001b). The natural
properties may be specially marked by perfectly resembling tropes,
universals, or simply primitive naturalness, but properties, whether
natural are not, are sets, and sets are, by the axiom of extensionality,
individuated by their membership. Thus if a perfectly natural prop-
erty P and a power P’, which is endowed by P, are necessarily co-
extensive, then P=P’.
Given the argument above for Shoemaker’s thesis, it follows that
for Lewis the perfectly natural properties—all of them—are pow-
ers. And if the perfectly natural properties, the local qualities of
which Lewis speaks, are all powers, then Humean Supervenience
isn’t worthy of its name. Lewis might as well have said that every-
thing supervenes on powers, or that there is just one little disposi-
tion and then another.
142 | Troy Cross
NON-NOMIC, NON-GLOBAL CONDITIONS
The heart of my strategy is to take global states of the world such as

possible laws of nature and treat them as circumstances which,
together with the original property, would trigger an effect. Just as
placing salt in water or dropping a glass reveals its disposition to
dissolve, or to break, so the laws of nature reveal the causal powers
inherent in so-called categorical properties.
One might think that this is cheating. It is certainly unortho-
dox. One might think that laws of nature are of an entirely differ-
ent category from standard activation conditions such as being
put in water or being dropped. Laws concern the relations between
ordinary properties, and being in a world with such-and-such laws is
itself a property only in a highly attenuated, highly derivative
sense. (So goes the objection.) Standard categoricalist opposition
to Shoemaker’s individuation and essence claims rested on
everyone tacitly ignoring my strategy—which certainly would
explain why no one has called attention to it—and so a repair
should be easy enough. All we have to do is make that tacit
assumption explicit.
Very well. Let’s admit that Shoemaker is correct if we allow that
the triggering circumstance in the specification of a power can be
any sort of state, but contend that his view is only interesting if we
restrict the relevant circumstances to exclude nomic facts, facts
about what the laws are. The proposal, then, is that for Shoemaker,
but not Lewis, property possession entails having a power to bring
about certain effects under certain non-nomic conditions. And
remember: this is supposed to be no substantial departure from
Shoemaker’s or Lewis’s intentions, but rather a matter of noting a
previously tacit restriction.
To my mind, the proposed revision is grossly ad hoc. The power
to bring about a G in a world where it’s a law that Gs follow Fs is
still a power, and it’s a power that characterizes F in particular, and
not properties generally. But I won’t presume my audience shares
that immediate judgment of arbitrariness. The “non-nomic condi-
tions” restriction has a more blatant flaw. Since Shoemaker thinks
the laws of nature are necessary, every circumstance is in a sense
nomic: it entails the laws of nature, simply because everything does.
So Shoemaker, on this proposal, would end up agreeing with Lewis
that properties do not necessarily endow powers with non-nomic

activation conditions, because on Shoemaker’s view, there aren’t
any non-nomic conditions.
Nor is it useful to stipulate that the circumstances identifying the
relevant powers cannot “mention” the laws, even if they entail
them. This is a debate about properties and not predicates. The cir-
cumstances are a matter of which things have which properties.
They don’t “mention” anything. And remember that we are work-
ing in an anti-hyperintensional environment. Any attempt to get
fine-grained enough to block the entailment of the laws by Shoe-
maker-envisioned circumstances will likely require distinctions too
medieval for Lewis. If Lewis cannot distinguish his view of the
nature of properties from Shoemaker’s without recourse to hyper-
intensionality, the game is already over.
A related strategy would be to forbid the use of triggering cir-
cumstances that are temporally later than the effect in the specifica-
tion of powers. Since the Humean conditions for F’s endowing the
power to become G will undoubtedly involve large swathes of the
future, this would rule out the troublemaking powers. But this is
overly restrictive, particularly for someone of Lewis’s methodologi-
cal bent. The recent theory that the Large Hadron Collider’s many
failures and delays are due to time travel and the universe’s “hatred
of the Higgs” should be ruled out on physical, not metaphysical
grounds (Nielsen and Ninomiya 2008).
DLEWIS
Before we try in earnest to engineer a better restriction on the condi-

tions, or to find some way to block the inference pattern, let’s look
ahead to the end game. The problem is not with formulating a
proper version of the restriction, or with blocking a pattern of infer-
ence for that matter. The problem is deeper.
Consider the following fictional philosopher: David K. Dlewis.
Dlewis is much like Lewis. He thinks that all of the possible worlds
exist as spatiotemporally disconnected concreta. He thinks proper-
ties are sets of possibilia. He endorses counterpart theory with
table-banging glee. He strokes his beard. He likes trains. He inspires
students. He writes beautifully. And he’s really really smart.
144 | Troy Cross
Dlewis agrees with Lewis about which properties are the perfectly
natural ones. That is, when Lewis and Dlewis survey the pluriverse
from their God-like pedestals, there is never an occasion on which
Lewis says some set of possibilia, S, is natural and Dlewis does not,
or vice versa. They also agree on principles of recombination. They
even agree on counterfactuals.
There are notable differences in their metaphysics, however.
Dlewis says that all of the natural properties are dispositions to have
certain effects, conditional on global states of the world. Which glo-
bal states? Precisely the ones Lewis says are the conditions in which
the perfectly natural properties endow the powers to bring about
those effects. In other words, Dlewis does not fight the general
inference I was trying to draw earlier, but embraces it.
Lewis’s actual views are a bit more complex, but if we think about
the (very nearby) nomic theory of powers discussed earlier, we can
simplify and say that F endows the power to become G just in the
circumstances in which it is nomically necessary that Fs become Gs.
Thus, for Dlewis, F-ness is (in part) the power to become G if in the
following condition: the global state of the world is such that the
simplest and most informative axiomatic description of that world
entails that Fs become Gs (i.e., it’s a Lewis-law that Fs become Gs).
For Dlewis, such global states of the world do not constitute laws.
He thinks, like a typical dispositionalist, that laws reflect the dispo-
sitional essences of properties. He thinks laws are necessary truths
and that (deterministic) dispositions metaphysically necessitate
their effects. Yet, he agrees with Lewis that there are worlds where
Fs do not become Gs. As Dlewis sees it, those are worlds where the
crucial (global) activation conditions, e.g., the supervenience basis
for its being a “Lewis law” that Fs become Gs, are absent, and thus
fail to trigger the disposition into action.
What Lewis sees as contingency in which powers are granted by F,
Dlewis sees as contingency in the existence of the activation condi-
tions for various F-involving dispositions. For Dlewis, that contin-
gency in no way counts against the metaphysical necessity of the
relation between a disposition, its activation conditions, and its
manifestation. And Lewis agrees that each connection Dlewis thinks
is necessary is indeed necessary.
Suppose Lewis somehow resists my earlier argument. Dlewis’s
metaphysics still looks coherent, and, when it comes to the charac-
terization of natural properties, only nominally different from

Lewis’s own. The differences between Lewis and Dlewis are not dif-
ferences in the modal behavior of the fundamental properties. So
it’s hard to see what Lewis’s insistence that his properties are “gen-
uinely” non-dispositional, his disavowals of Dlewisian metaphys-
ics, amount to.
Is Dlewis’s view somehow conceptually confused? How so? The
position isn’t terribly attractive, but it does seem coherent. The inco-
herent-seeming position is Lewis’s. It is just very hard to imagine,
of all people, David Lewis, insisting that although he and Dlewis
agree about the modal extension of all the natural properties, all of
the necessities, and all of the counterfactuals, that in fact, their views
about the nature of fundamental reality are radically different, that
Humeanism amounts to refusing the label “dispositional” to some-
thing that behaves, modally, just like a disposition.
Intuitively, what Lewis should say to Dlewis is this:
You’re right that the perfectly natural properties, in some sense, correspond
to these funny powers with highly non-local activation conditions. But the
perfectly natural properties are not themselves powers. Rather, they are
perfectly qualitative, intrinsic features of point-sized things. Any powers to
which they correspond are, in the end, a matter of how these purely quali-
tative intrinsic features are globally distributed. Powers are not primitive.
They are highly unnatural, and exist only in virtue of the distributions of
qualities, which are wholly non-dispositional in character. The purely qual-
itative properties do all of the deep explanatory work. They lie at the bot-
tom of things. And they do not have dispositional essences.
Well and good. Except that for Lewis to say such a thing, he would
also have to say that properties can be non-identical even if they’re
necessarily co-extensive. That would mean that properties are not
sets of possibilia, that naturalness is neither a primitive feature of
sets, nor a matter of whether sets have members instantiating a sin-
gle universal nor perfectly resembling tropes (Lewis 1983).
Hyperintensionalism, then, is a perfectly acceptable counter to
my argument, and I would be content if it were widely recognized
as a presupposition of categoricalism. But there is reason to assume
that Lewis himself would not abandon his set-based theory of
properties. One reason, though certainly not the only one, is that
the concession would undermine part of Lewis’s argument for
modal realism, namely, his objection to building a semantics for modal
146 | Troy Cross
talk, a la Robert Stalnaker, on an ontology of properties rather than

on concrete, spatiotemporally disconnected things (Lewis 2001b,
3.4; Stalnaker 1976). If properties are individuated primitively,
rather than by their transworld extensions, they could achieve the
theoretical payoff of possible worlds, without a plentitude of spa-
tio-temporally disconnected concreta.
If hyperintensionalism is not an option, though, how can Lewis
meet the challenge of my argument? Only, so far as I can see, to
trade his reductivism for expressivism or eliminativism about pow-
ers. Suppose Lewis were to say that there is only really the Humean
base—just a spread of qualities—and that all talk of powers, dispo-
sitions, potentialities, chances, and the like, while perhaps permis-
sible or impermissible, felicitous or infelicitous, is either false or
non-truth-evaluable (Ward 2002). Then he could deny the central
premise of my argument. He could say that different properties do
not make different potential contributions to the powers of objects,
because objects are all—whatever their qualities—impotent, speak-
ing strictly. However, this would trade his somewhat plausible
supervenience claim for a far more controversial eliminativism.
OBJECTIONS AND REPLIES
1. Significance
Some philosophers have granted the central argument of this paper,
but charged it with triviality. According to this objection, the
Humean should simply accept my conclusion—that natural prop-
erties are all powers—without batting an eyelash. After all, the
powers identified with Lewisian natural properties are highly unu-
sual, and easily screened off from the powers endorsed by mystery
mongering dispositionalists: the Lewisian is committed only to per-
fectly local powers, instantiated by point-sized objects, and sensi-
tive only to global states of the world. What’s wrong—so goes the
objection—with saying that qualities are, in fact, identical to these
kinds of powers? Yes, ontologically speaking, Lewis is Dlewis, but
so what?
We can resolve the matter by careful attention to the disposi-
tionalist/categoricalist dialectic. The expressions “quality” or
“categorical property” are typically used by categoricalists to mark
something that is not a disposition, and typically in order to analyze

away the latter in terms of the former. In particular, Lewis’s iconic
statement of Humean supervenience, quoted above, makes no
sense if we allow that qualities can be dispositions. Lewis seems to
be claiming that everything supervenes on non-dispositional prop-
erties, i.e., that the dispositional, nomic, chancy, subjunctive, and
the like, are all fixed by the non-dispositional, non-nomic, non-
chancy, non-subjunctive, and the like. But according to the present
objection, Lewis would happily gloss his doctrine as, in part, saying
merely that some dispositions fix others, which is hardly visionary.
The problem is not that Lewis’s ontology is filled with disposi-
tions. That is hardly disturbing, so long as the dispositions all
supervene on a purely Humean base. The problem is rather that
we’ve shown that there are no non-dispositional grounds for the
supervenient dispositions; it’s dispositions “all the way down”.
Even though, strictly speaking, if we read “quality” as compatible
with “disposition”, dispositions will supervene on qualities (and
vice versa), the lack of a non-dispositional base should be troubling
to committed Humeans.
In any case, suppose we take the non-standard reading on which
“quality” is taken to be compatible with “dispositional”. Lewis’s
view then looks to be in broad agreement with C.B. Martin and John
Heil’s (Martin 1997; Heil 2005; Martin and Heil 1998), on which the
natural properties are all both dispositional and categorical. Martin
and Heil would only take issue with Lewis about which dispositions
are perfectly natural. This is, at the very least, news.
Moreover, once the disagreement is recast as about which disposi-
tions are perfectly natural, the Martin Heil view seems to have a
clear advantage over Lewis’s. The powers that shadow—and are in
fact identical to—Lewis’s qualities are an odd choice for fundamen-
talia. They are, if you recall, globally sensitive, dispositions that
specify how an object possessing the property would be poised to
act under every possible circumstance, even the counter-nomic
ones. Why would a metaphysician start here, rather than with the
fundamental properties of physics, understood as dispositions to
resist acceleration, warp spacetime, alter electro-magnetic fields,
and such, with no reference to other possible laws? Note that Lewis
cannot answer this question by appealing to the spookiness of
dispositions generally, or the need to reduce the dispositional to the
148 | Troy Cross
non-dispositional. Once we have admitted that the world is, indeed,

composed of dispositions all the way down, it’s unclear why those
dispositions would have only point-sized instances, and why they
would all be globally sensitive, and never purely locally-sensitive,
why they would always vary so radically when embedded in non-
actual global patterns in the way that Lewis prefers. Given that the
global distribution of (non-dispositional) qualities does not explain
the powers of objects, Lewis’s view looks highly idiosyncratic and
lacks its standard atomistic, anti-modalist motivations.
2. Symmetry
Another objection charges me with favoritism. The Lewis/Dlewis
scenario, as depicted, has no obvious asymmetries, so if Lewis is in
trouble, then so, it seems, is Dlewis. In particular, Dlewis, as a fel-
low anti-hyperintensionalist, is just as incapable of distinguishing
himself from Lewis as Lewis is incapable of distinguishing himself
from Dlewis. So, why shouldn’t we conclude by extension that dis-
positionalists generally, at least if they are anti-hyperintensionalists,
are in the same boat as categoricalists?
The objection has a grain of truth: anti-hyperintensionalism can
create a problem for categoricalists and dispositionalists alike.
Dlewis is a case in point. However, actual dispositionalist philoso-
phers differ from Dlewis in at least three respects.
First, they tend not to be anti-hyperintensionalist, at least not
explicitly. Perhaps this is because anti-hyperintensionalism is a
remnant of nominalism and dispositionalists tend to be less squeam-
ish about property realism, but the literature simply does not hold
any loudly anti-hyperintensionalist dispositionalists.
Second, dispositionalists do not attempt to reduce or explain the
categorical in terms of a purely dispositional, non-categorical base.
Rather, they are often friendly to the idea that dispositions are them-
selves categorical. It is categoricalists, historically, who have insisted
on the non-categoricity of dispositions, in an effort to banish them
from ontology along with the rest of the merely possible. If any-
thing, dispositionalists tend to applaud C.B. Martin’s insistence
that dispositions are as actual, as categorical, as anything else (1993,
75; 1996). Though there is a literature on the plausibility of pure
dispositionalism (Bird 2007, 6; Holton 1999; McKitrick 2003), it is

difficult to find actual advocates of the view. Shoemaker, to whom
the pure dispositionalist theory is often attributed, originally held
that the dispositional/categorical distinction applies to predicates,
and not to properties, thus leaving it open that a dispositional pred-
icate and a categorical predicate might share the very same seman-
tic value (1984).
Third, Dlewis’s dispositional properties are tailored precisely to
fit Lewis’s pure qualities and his reductive account of powers. Even
setting aside the two disanalogies listed above, to force Dlewis’s
problem on actual dispositionalists we would have to find a cate-
goricalist who has a unique and necessarily correlated purely non-
dispositional quality for each of the dispositionalist’s fundamental
dispositions. It is not clear to me that this can be done. Say we are
searching for a categorical property to match the modal profile of
the disposition to resist acceleration, where that disposition is not
taken to be conditional on the laws or the global state of the world.
What is the categorical shadow of this disposition? How does it not
violate categoricalist principles of recombination? (Remember: this
power is not conditional on laws or global regularities of any kind.)
As such, how can it be a candidate for a perfectly natural categorical
property?
At the very least, the search for a categoricalist shadow for every
disposition is a different project from the one going in the other
direction, and it is not obvious that it will be successful. This should
not be surprising. Lewis’s account of dispositions in terms of the
distribution of qualities allows us to easily locate the dispositional
shadows of his properties and conjure up a philosopher who takes
just those dispositional shadows to be fundamental: Dlewis. But
dispositionalists do not give us the analogous conjuring recipe.
As I said, the symmetry objection has some genuine import.
Though Dlewis thinks of his view as a radically different ontology
from Lewis’s, he is mistaken, by his own lights. Just like Lewis. But
Dlewis is a very peculiar dispositionalist, purpose-built as a foil. He
thinks dispositions cannot also be categorical, he is anti-hyperinten-
sionalist, and his fundamental powers, oddly enough, all share
modal profiles with Lewis’s local qualities. Actual dispositionalist
theories lack some or all of these features, and are therefore immune
from the parallel criticism.
150 | Troy Cross
CONCLUSION
When a property potentially makes a difference to its bearers, it

makes an actual difference to the potential of its bearers. This sim-
ple observation dooms Lewis’s grand vision of resting all disposi-
tions upon a foundation of pure qualities. For it is easy enough to
see, given Lewis’s own reductive theories, that for arbitrary quality
F, being an F makes a difference to the powers of objects across the
modal pluriverse. These potential differences in powers are, in fact,
powers that F actually endows, and when the powers are consid-
ered collectively, powers that only F endows. That is enough to
establish Shoemaker’s thesis in “Causality and Properties”. If, like
Lewis, we then identify necessarily co-extensive properties, it fol-
lows that Lewis’s qualities are, one and all, dispositions.
The small story here is that Shoemaker, not Lewis, turns out to
be the hero of the naturalistic mini-epic recounted in the introduc-
tion of this essay. But there is a larger story too. My argument is
easiest to process as a direct attack on Lewis, but it poses a per-
fectly general dilemma for realists about dispositions: either the
world is dispositional at its fundamental level or else it is
hyperintensional.
Reed College
WORKS CITED
Achinstein, Peter. 1974. The Identity Conditions of Properties. American
Philosophical Quarterly 11: 257–75.
Armstrong, D. M. 1973. Belief, Truth and Knowledge. Cambridge University
Press.
—— . 1978. A Theory of Universals: Volume 2: Universals and Scientific Real-
ism. Cambridge University Press.
—— . 1980. Identity Through Time. In Time and Cause: Essays Presented to
Richard Taylor, ed., P. van Inwangen, 67–78, Dordrecht: D. Reidel.
Armstrong, D. M., C. B. Martin, U. T. Place, and Tim Crane. 1996. Disposi-
tions: A Debate. Psychology Press.
Ashwell, Lauren. 2010. Superficial Dispositionalism. Australasian Journal
of Philosophy 88: 635–53.
Bird, Alexander. 1998. Dispositions and Antidotes. Philosophical Quarterly
48: 227–34.
—— . 2007. Nature’s Metaphysics. New York: Oxford University Press.
Bird, Alexander and Toby Handfield. 2008. Dispositions, Rules and Finks.
Philosophical Studies 140: 285–98.
Carroll, John. 1994. Laws of Nature. New York: Cambridge University
Press.
Choi, Sungho. 2008. Dispositional Properties and Counterfactual Condi-
tionals. Mind 117: 795–841.
—— . Intrinsic Finks and the Dispositional/Categorical Distinction. Noûs,
forthcoming.
Clarke, Randolph. 2010. Opposing Powers. Philosophical Studies 149: 153–60.
Cross, Troy. 2004. The Nature of Fundamental Properties. Dissertation,
Rutgers, the State University of New Jersey.
—— . 2005. What is a Disposition? Synthese 144: 321–41.
Davidson, Donald. 1963. Actions, Reasons, and Causes. Journal of Philoso-
phy 60: 685–700.
—— . 1969. The Individuation of Events. In Essays in Honor of Carl
G. Hempel, ed. Nicholas Rescher, 216–34.
Davis, Wayne. 1984. A Causal Theory of Intending. American Philosophical
Quarterly 21: 43–54.
Eagle, Antony. 2009. Causal Structuralism, Dispositional Actualism, and
Counterfactual Conditionals. In Dispositions and Causes, 65–99. Oxford
University Press.
Evans, Gareth. 1973. The Causal Theory of Names. Proceedings of the Aris-
totelian Society 47: 187–208.
Everett, Anthony. 2009. Intrinsic Finks, Masks, and Mimics. Erkenntnis 71:
191–203.
Fara, Michael. 2005. Dispositions and Habituals. Noûs 39: 43–82.
Freddoso, Alfred J. 1986. The Nature of Necessity. Midwest Studies in Phi-
losophy 11: 215–42.
Gettier, Edmund L. 1963. Is Justified True Belief Knowledge? Analysis 23:
121–3.
Goldman, Alvin I. 1967. A Causal Theory of Knowing. The Journal of Phi-
losophy 64: 357–72.
Goodman, Nelson. 1983. Fact, Fiction, and Forecast. 4th ed. Harvard Uni-
versity Press.
Grice, H. P. 1961. Symposium: The Causal Theory of Perception. Proceed-
ings of the Aristotelian Society 35: 121–68.
Harre, R., and E. H. Madden. 1975. Causal Powers: A Theory of Natural
Necessity. Oxford: Blackwell.
Hawthorne, John. 2001. Causal Structuralism. Philosophical Perspectives:
361–78.
Heil, John. 2005. From an Ontological Point of View. Oxford University
Press, USA.
152 | Troy Cross
Holton, Richard. 1999. Dispositions all the way round. Analysis 59: 9–14.
Johns, Richard. 2002. A Theory of Physical Probability. University of Toronto
Press, September 28.
Johnston, Mark. 1992. How to Speak of the Colors. Philosophical Studies
68: 221–63.
Kneale, William. 1952. Probability and Induction. Oxford University Press.
Kripke, Saul A. 1980. Naming and Necessity. Harvard University Press.
Lewis, David. 1973. Causation. Journal of Philosophy 70: 556–67.
—— . 1979. Counterfactual Dependence and Time’s Arrow. Noûs 13:
455–76.
Lewis, David K. 1983. New Work for a Theory of Universals. Australasian
Journal of Philosophy 61: 343–77.
—— . 1987. Philosophical Papers: Volume II. Oxford University Press.
—— . 1997. Finkish Dispositions. Philosophical Quarterly 47: 143–58.
—— . 2001a. Counterfactuals. 2nd ed. Wiley-Blackwell.
—— . 2001b. On the Plurality of Worlds. Wiley-Blackwell.
Manley, David, and Ryan Wasserman. 2008. On Linking Dispositions and
Conditionals. Mind 117: 59–84.
Martin, C. B, and John Heil. 1998. Rules and powers. Philosophical Perspec-
tives 12: 283–312.
Martin, C. B. 1993. Power for Realists. In Ontology, Causality and Mind:
Essays in Honour of D M Armstrong, ed. John Bacon. New York: Cam-
bridge University Press.
—— . 1994. Dispositions and Conditionals. Philosophical Quarterly 44: 1–8.
—— . 1997. On the Need for Properties: The Road to Pythagoreanism and
Back. Synthese 112: 193–231.
Maudlin, Tim. 2007. The Metaphysics Within Physics. Oxford University
Press.
McKitrick, Jennifer. 2003. The Bare Metaphysical Possibility of Bare Dis-
positions. Philosophy and Phenomenological Research 66: 349–69.
Nielsen, H. B., and M. Ninomiya. 2008. Search for Effect of Influence from
Future in Large Hadron Collider. International Journal of Modern Physics
A (IJMPA) 23: 919–32.
Nozick, Robert. 1981. Philosophical Explanations. Harvard University
Press.
Prior, Elizabeth W., Robert Pargetter, and Frank Jackson. 1982. Three The-
ses about Dispositions. American Philosophical Quarterly 19: 251–7.
Putnam, Hilary. 1960. Minds and Machines. In Dimensions of Mind : A
Symposium, ed. Sydney Hook: 362–85. New York: New York University
Press.
—— . 1970. On Properties. In Essays in Honor of Carl G. Hempel, ed. Nicho-
las Rescher. Dordrecht: D. Reidel.
Schaffer, Jonathan. 2004. Quiddistic Knowledge. In Lewisian Themes, ed.

F. Jackson and G. Priest, 210–30. Oxford University Press.
Sellars, Wilfrid. 1948. Concepts as Involving Laws and Inconceivable
Without Them. Philosophy of Science 15: 287–315.
Shoemaker, Sydney. 1980. Causality and Properties. In Time and Cause:
Essays Presented to Richard Taylor, ed. Peter van Inwagen, 109–35.
Dordrecht: D. Reidel.
—— . 1998. Causal and Metaphysical Necessity. Pacific Philosophical Quar-
terly 79: 59–77.
Stalnaker, Robert C. 1976. Possible Worlds. Noûs 10, 65–75.
Tye, Michael. 1982. A Causal Analysis of Seeing. Philosophy and Phenome-
nological Research 42: 311–25.
Ward, Barry. 2002. Humeanism without Humean Supervenience: A Pro-
jectivist Account of Laws and Possibilities. Philosophical Studies 107:
191–218.
6. “There sweep great general principles
which all the laws seem to follow”
Marc Lange
My title is taken from a passage in Richard Feynman’s classic book,

The Character of Physical Law:
When learning about the laws of physics you find that there are a large
number of complicated and detailed laws, laws of gravitation, of electricity
and magnetism, nuclear interactions, and so on, but across the variety of
these detailed laws there sweep great general principles which all the
laws seem to follow. Examples of these are the principles of conserva-
tion . . . (Feynman 1967, 59)
My aim in this paper is to understand this conception of the conser-

vation laws and to examine what it reveals about the character of
physical law generally—in particular, about the laws’ modality,
their explanatory role, and the adequacy of the dispositional essen-
tialist conception of laws as metaphysical necessities arising from
the causal powers essential to the sparse fundamental properties of
physics.
My main positive result is that science recognizes an important
distinction: between conservation laws as constraints on the funda-
mental forces there could be, on the one hand, and conservation
laws as coincidences of the fundamental forces there happen to be,
on the other hand. In the above passage, Feynman characterizes
conservation laws as constraints. I do not argue that they are con-
straints; on my view, this is a matter for science, not metaphysics, to
decide. However, I argue that some conservation laws have some-
times been taken (with good reason) to be constraints, that their
status as constraints would make an important difference to their
role in scientific explanations, and that the distinction between con-
straints and coincidences applies to other laws besides the conser-
vation laws. I ultimately cash out the distinction between constraints
and coincidences in terms of the truth of various counterfactual
"There sweep great general principles" | 155
conditionals, and I briefly sketch how this way of elaborating the

distinction relates to my broader account of natural law (Lange
2009).
My main negative result is that metaphysical pictures along the
lines defended recently by Alexander Bird (2007), Brian Ellis (2001,
2002), and Stephen Mumford (2004)—for which, despite some disa-
greements among these authors, I will use the catch-all term “dis-
positional essentialism”—cannot accommodate the distinction
between constraints and coincidences. Such a picture must portray
all conservation laws as coincidences. It thus forecloses options that
science has (with good reason) taken seriously. This failure is a
weighty count against dispositional essentialism.
1. CONSTRAINTS VERSUS COINCIDENCES
Consider the law of energy conservation. (I could just as well have

chosen any of the other conservation laws that have been proposed
in the history of physics, such as the conservation of linear momen-
tum, angular momentum, electric charge, mass, parity, baryon
number, or lepton number.) As Feynman emphasizes, though the
various kinds of fundamental interaction differ in a host of ways (in
their range, their strength, the kinds of bodies that participate in
them, and so forth), they are all alike in conserving energy. As con-
venient examples of kinds of fundamental interactions, I shall fol-
low Feynman and take gravitational and electric interactions as
described in classical physics by Newton’s gravitational-force law
and (in the static case) Coulomb’s law, respectively. Despite their
differences, these two types of interactions are alike in both con-
serving energy. Of course, gravity is not in fact a force at all accord-
ing to general relativity, and electric and magnetic forces are not
actually distinct kinds of force according to special relativity. But
none of this matters to my argument. I shall be appealing to these
two forces only to illustrate my claim that physical theory recog-
nizes an important distinction between two different ways in which
a law like energy conservation could hold: as a constraint or as a
coincidence. I contend that any metaphysical account of natural
law should leave room for both of these possibilities. The same dis-
tinction must be drawn whatever the fundamental forces actually
156 | Marc Lange
are—indeed, even if there is in fact only a single kind of fundamen-

tal force (a “grand unified field”).
Why are gravitational and electric interactions alike in conserv-
ing energy? Here are two possible explanations.
1. Gravitational interactions conserve energy because the
gravitational force law requires them to. Electric interac-
tions conserve energy because the electric force law requires
them to.1 The two interactions are therefore alike in con-
serving energy—but for separate reasons.
2. Both kinds of interaction conserve energy for the same
reason: because the law of energy conservation requires
them to.
On the first option, it is just a coincidence that these two different
forces conserve energy, since there is no common explanation of
their doing so. Just as it would be a coincidence for two friends both
to be in Chicago on the same day if there was no important com-
mon reason for their both being there then (e.g., they had made no
plans to meet there, they were not both attending the same conven-
tion), so likewise it is a coincidence for various distinct forces all to
conserve energy if there is no important common “cause”, but
rather each does so for a substantially separate reason.2 On the sec-
ond option, in contrast, the law of energy conservation is not a coin-
cidence. Rather, the various fundamental kinds of interaction all
have a common reason for conserving energy: the conservation law.
1
The force law alone is not enough to entail that the interaction will conserve
energy. The explanation must also appeal to the fundamental dynamical law: the law
relating forces to the motions they cause (in classical physics: Newton’s second law
of motion).
2
I say “important” and “substantially” in order to acknowledge that two compo-
nents of a coincidence may have some explainers in common—as long as they are
beside the point in the context in which an explanation of the two components is
being demanded. For instance, suppose that the two friends both happened to travel
to Chicago on the same airplane flight. Then there would be some common explain-
ers of their both being there (e.g., the flight, the natural laws governing jet engines).
But these are not the sorts of explainers that we would (ordinarily) be asking for in
asking “What brings you to Chicago?” Likewise, although the fundamental dynami-
cal law (see note 1) is common to the explanation that gravitational interactions con-
serve energy and to the explanation that electric interactions conserve energy, it is
incidental; the force laws involved would (typically) be the focus of our explanatory
demand. Hence, it is a coincidence that both forces conserve energy.
It is a constraint on the forces. That is, the law of energy conserva-

tion limits the kinds of forces there could have been. The only kinds
of force there could have been are forces that conserve energy, and
that is why every kind of force there actually is conserves energy.
The difference between constraint and coincidence is a difference
in what is explanatorily prior to what. If energy conservation is a
coincidence, then the various force laws are explanatorily prior to
the law of energy conservation. On the other hand, if energy conser-
vation constrains the force laws, then the conservation law is
explanatorily prior to them. It does not entail the particular force
laws there are, but it explains why they each exhibit a certain fea-
ture. These two options (constraint or coincidence), then, are mutu-
ally exclusive.3
However, these two options are alike in one important respect:
whichever option holds, the law of energy conservation is physi-
cally necessary—a law rather than an accident. As a coincidence,
the conservation law is physically necessary in virtue of following
exclusively from laws, such as the gravitational-force law, the elec-
tric-force law, and the law that all fundamental forces are gravi-
tational or electric or . . . (a “closure law”). As a constraint on the
force laws, the conservation law transcends the grubby, pedestrian
details of the various particular force laws. It is a higher-order law,
as Feynman suggests. It does not depend on the kinds of forces
there actually happen to be. It limits the possible kinds of forces.
Since the difference between constraint and coincidence is a dif-
ference in explanatory priority, the conservation law’s status as con-
3
They are not collectively exhaustive. Rather, they are the extremes; there are
intermediate cases. For instance, suppose that some fundamental kinds of interac-
tions have a certain feature (e.g., are capable of both attraction and repulsion)
whereas others (namely, interactions A, B, and C) do not have this feature. Suppose
it is a law that every kind of interaction with that feature conserves energy, and sup-
pose that law is a constraint. Then the fact that every kind of interaction conserves
energy might be explained by this constraint together with A’s force law, B’s force
law, and C’s force law (along with the fundamental dynamical law). In that case,
energy conservation is neither a complete coincidence nor a constraint. As another
kind of intermediate case, the law of energy conservation might follow from exactly
two separate constraints (e.g., that all of the forces capable of both attraction and
repulsion must conserve energy, and that all of the forces capable only of attraction
or only of repulsion must conserve energy). For the sake of simplicity, I shall not
return to these intermediate possibilities, but I believe that it is clear how my remarks
apply to them.
158 | Marc Lange
straint or coincidence makes a difference to whether certain

arguments carry explanatory power. It makes a difference to the
success of many putative explanations well beyond whether the
conservation law explains why gravitational and electric interac-
tions both conserve energy. Consider the fact that an ideally incom-
pressible, nonviscous fluid in a container at rest in a uniform
downward gravitational field is not undergoing any circulation;
none of its parcels at the top feels an unbalanced force pulling it
downward, nor do any bottom parcels feel unbalanced forces push-
ing them upward. Why is that?4 If energy conservation is a con-
straint, then it explains why. A force arising from no outside agency
that would make the fluid parcels begin to circulate from rest would
violate energy conservation: in beginning to circulate, the parcels’
kinetic energy would increase but their total potential energy would
be unchanged. (As ascending parcels gain gravitational potential
energy, descending parcels lose an equal quantity of it.) Energy con-
servation as a constraint rules out any circulation-inducing force.
However, as a coincidence, energy conservation cannot supply
this explanation. If energy conservation is a coincidence, then the
reason why the fluid undergoes no circulation is that electric forces
fail to induce circulation (because of the electric-force law), gravita-
tional forces fail to induce circulation (because of the gravitational-
force law), and so forth for all of the actual kinds of forces experi-
enced by the fluid parcels. This is a “bottom-up”, causal/mechani-
cal explanation. As a coincidence, the general principle of energy
conservation cannot figure in such an explanation. It cannot explain
why various kinds of fundamental force are alike in failing to induce
fluid circulation, since as a coincidence rather than a constraint, it is
not explanatorily prior to the force laws. The reason why electric
forces fail to induce circulation (and the reason why they conserve
energy) is not the coincidence that all forces conserve energy; it is
the electric-force law.
Suppose that, instead of trying to take the comprehensive con-
servation law and slot it into the explanation somewhere explanato-
rily prior to the force laws, we try to place it somewhere explanatorily
4
The explanandum is a scientifically significant fact; it is not a fact that only a
philosopher would inquire into. (Don’t pretend that you don’t know what I mean!)
For example, it is central to the reason why Archimedes’ Principle holds.
posterior to the force laws. Then we encounter a different problem.

The fluid parcels feel gravitational forces, so the fact that gravita-
tional forces conserve energy may help to explain why there are no
circulation-inducing forces. But if the fluid parcels feel no magnetic
forces, then the fact that magnetic forces conserve energy does not
help to explain why there are no circulation-inducing forces. If the
conservation law is just a coincidence, then it is effectively the fact
that gravitational forces conserve energy, magnetic forces conserve
energy, and so forth for all of the actual kinds of forces. But if some
of these forces are not experienced by the fluid parcels, then the fact
that they conserve energy is not explanatorily relevant, and so nei-
ther is the general principle of energy conservation. For a coinci-
dence to be explanatorily relevant to an outcome, all of its
components must be relevant. For instance, the reason why you
and I ran into each other at the mall this afternoon might be the
coincidence that you and I both chose this day to go shopping there—
but the coincidence that you, I, and Frank all chose this day to shop
there does not explain why you and I encountered each other there.
As a coincidence of the various kinds of fundamental forces, the
conservation law explains only if all of those forces are explanato-
rily relevant.
Suppose, then, that the fluid parcels feel every species of funda-
mental force so that every component of the energy-conservation
coincidence is explanatorily relevant. Then the resulting explana-
tion from energy conservation would still not be a top-down expla-
nation. Rather, it would have to include the fact that each of these
forces is actually felt by the fluid parcels. The top-down explanation
does not specify which kinds of fundamental force the fluid parcels
experience. Its point is that the outcome does not depend on what
possible forces are actually at work; no matter which possible forces
were operating on the fluid parcels, the fluid would inevitably still
fail to circulate.
If energy conservation is a coincidence, then the reason why
the fluid undergoes no circulation is that electric forces fail to
induce circulation (because of the electric-force law), gravita-
tional forces fail to induce circulation (because of the gravita-
tional-force law), and so forth for all of the kinds of forces actually
experienced by the fluid parcels. In contrast, if energy conserva-
tion is a constraint, then this bottom-up argument (though, of
160 | Marc Lange
course, still sound) does not explain why the fluid is not circulat-
ing because it would then inaccurately depict the explanandum
as depending on the particular kinds of forces that happen to be
acting on fluid parcels. The only explanation is a top-down expla-
nation: that any circulation-inducing force would violate energy
conservation, which is impossible. Here is an analogy. Consider
Jones and Smith, each convicted in separate trials before separate
judges of possessing (independently) 100 kilograms of marijuana.
Why did each of them receive a sentence exceeding five years?
The reason is not that Smith’s judge passed this sentence because
he believed that Smith’s crime rose to a certain level of serious-
ness because of various factors including . . . and Jones’s judge
passed this sentence because he believed that Jones’s crime rose to
a certain level of seriousness because . . . if the two judges were
constrained by a mandatory minimum sentencing law to pass
sentences of at least five years for the possession of 100 kilograms
of marijuana. If there is such a law, then it is no coincidence that
the two judges handed down sentences that are alike in this
respect. Rather, the law is a common explainer—and any account
is mistaken if it depicts the two sentences as the products of inde-
pendent judicial decisions that weighed the particulars of the
individual cases.
The success of various proposed top-down scientific explana-
tions, then, depends upon the status of energy conservation as a
constraint. Even if energy conservation is a coincidence, the law
that gravitational forces conserve energy could still be used to help
explain why the fluid does not circulate. But this explanation would
simply be a bottom-up account that portrays the fact that the fluid
does not circulate as arising from the coincidence that each of the
particular kinds of forces acting on the fluid conserves energy. In
contrast, if energy conservation is a constraint, then the fact that the
fluid fails to circulate does not depend on the particular kinds of
forces at work on it.
Here is another way to bring out this contrast. Consider a wooden
block (of any shape) sitting on top of a post, and suppose that across
the upper surface of the block is laid part of a uniform loop of rope
(or chain), while the rest of the loop hangs below the block, experi-
encing uniform downward gravity. Why does the rope loop, having
been laid across the block, not spontaneously begin to turn round
and round the block?5 If energy conservation is a constraint, then it

explains why no force puts the rope loop into circulation: any such
force is ruled out by energy conservation for exactly the same rea-
son as it precludes a force inducing spontaneous fluid circulation.
However, if energy conservation is a coincidence, then this explan-
ation is unavailable. Energy conservation does not help to explain
features of the force laws. Instead, the explanation is that one kind
of force felt by the rope fails to induce circulation, another also (for
independent reasons) fails to do so, and so forth for all of the kinds
of forces at work on the rope. This explanation may involve differ-
ent kinds of fundamental forces from the corresponding explan-
ation of the fluid’s behavior; different forces may be at work on
ropes and fluids. Therefore, the bottom-up explanations do not
unify these two cases. In contrast, the top-down explanations not
only unify these two phenomena under the same explainer (the law
of energy conservation), but also unify them further by giving them
explanations of the very same form.
My point is that whether energy conservation is a constraint or a
coincidence makes a big difference to features of the world that sci-
ence cares greatly about: to the kinds of explanations that there are
and to the unifications that those explanations bring.6 These are
matters for empirical work to discover. I am not arguing that if
every single fundamental kind of force conserves energy, then this
“conspiracy” is unlikely to be a coincidence – that it probably has a
5
The explanandum is a very important fact. For example, it is central to Simon
Stevin’s 1586 clootcrans explanation of the law of the inclined plane (Stevin, 1955,
Vol. 1, 178).
6
If we believe that energy conservation is a constraint if it is true, then we are
prepared to confirm the hypothesis that energy is conserved very differently than if
we believe it is a coincidence if it is true. Roughly speaking, if we believe that energy
conservation is a coincidence if it is true, then we regard the fact that one fundamen-
tal kind of interaction conserves energy as no evidence that another kind does (just
as we take my being in Chicago as no evidence that you are there, too, if we believe
that our both being there would be coincidental). However, if we believe that energy
conservation might be a constraint, then we may take the fact that one fundamental
kind of interaction conserves energy as some evidence that another kind also does.
Feynman (1967, 76) says that we are “confident that, because we have checked the
energy conservation here, when we get a new phenomenon we can say it has to
satisfy the law of conservation of energy.” A good example of such a new phenom-
enon was radioactive decay which physicists believed to conserve energy before
they had any significant confidence in any theories regarding the particular force(s)
involved.
162 | Marc Lange
common “cause”. I insist only that the hypothesis positing such a

constraint is sometimes a reasonable one to entertain, that science
has frequently taken such hypotheses seriously, and that therefore
metaphysics should not foreclose such hypotheses.
A conservation law need not be a brute fact in order for it to be a
constraint. It may have an explanation. In fact, one way for a con-
servation law to be a constraint is for it to arise from a symmetry
principle, since if it so arises, then each of the actual forces con-
serves the relevant quantity for the same reason: because of the
symmetry principle. As is well known, various classical conserva-
tion laws follow from various space-time symmetries within a
Hamiltonian dynamical framework: energy conservation follows
from the laws’ invariance under arbitrary temporal displacement,
linear momentum conservation from their invariance under arbi-
trary spatial displacement, and so forth. If these derivations explain
why the conservation laws hold (as they are often said to do), then
the conservation laws are constraints, not coincidences. As Eugene
Wigner says:
[F]or those [conservation laws] which derive from the geometrical princi-
ples of invariance it is clear that their validity transcends that of any special
theory—gravitational, electromagnetic, etc.—which are only loosely con-
nected. (Wigner 1972, 13)
In other words, Wigner contends that those symmetries are not

coincidences of the particular kinds of forces there happen to be,
and so the associated conservation laws transcend the idiosyncra-
sies of the force laws figuring in bottom-up explanations.
2. OTHER POSSIBLE KINDS OF CONSTRAINTS BESIDES

CONSERVATION LAWS
Conservation laws are not the only “great general principles” that
have sometimes been reasonably thought to “sweep” across the
various force laws, explaining why all of those laws share certain
features. One candidate proposed by Heinrich Hertz may not turn
out to succeed. But an adequate metaphysical account of natural
law must at least leave room for explanations of the kind Hertz
proposed.
Newton’s gravitational-force law is an inverse-square law. So is

Coulomb’s law for the electric force between two point charges at
rest. So is Ampere’s law for the magnetic force between two electric-
current elements. In his 1884 lectures delivered at Kiel, Hertz said
that (as far as science has been able to discover) all fundamental
force laws are inverse-square—and that this regularity has never
been thought coincidental [zufällig] (Hertz 1999, 68).
What could Hertz have meant by this? Presumably, he did not
mean that no one has thought this regularity to be physically unnec-
essary—since although this is true, it is a trivial remark: obviously,
no regularity among the force laws could be accidental. The laws
alone must suffice to logically entail any such regularity. Rather,
I suggest, Hertz meant that the inverse-square character of all of the
fundamental forces has always been considered to be a constraint,
not a coincidence. In other words, Hertz meant that there is (accord-
ing to widespread consensus) a common explanation for each force’s
being inverse-square; they are not independently inverse-square.
This interpretation of Hertz’s remark is confirmed by his charac-
terizing this regularity among the various fundamental forces as
too remarkable for its instances not to have a common explainer: “Is
it not marvelous [wunderbar] that all long-range forces follow [an
inverse-square] law?” (Hertz 1999, 68). Indeed, Hertz immediately
suggests one possible common explainer: “Kant and many others
before and after him have tried to relate this feature [the inverse-
square character of the force laws] to the three-dimensional nature
of space.” But whereas Kant offers the inverse-square character of
forces as explaining why space is three-dimensional (see Callendar
2005), Hertz proposes that explanatory priority runs in the opposite
direction.
Hertz’s proposed explanation begins with another regularity
among the fundamental forces that he takes to be a constraint on
any force there might have been rather than a coincidence of the
various forces there actually are: that every fundamental force acts
by contact—that is, by a field acting at the same point in spacetime
as the force that it causes, so that the field causally mediates between
the two, perhaps spatiotemporally widely separated bodies thereby
interacting (see Lange 2002). In other words, Hertz’s explanation
begins with the premise that none of the fundamental kinds of
164 | Marc Lange
interaction constitutes action at a spatiotemporal distance. In fact,

Hertz presents himself as arguing for this premise by inference to
the best explanation: the most plausible explanation of the “marve-
lous” fact that all fundamental forces are inverse-square is that all
fundamental forces must operate by contact action.
In his Kiel lectures, Hertz does not say anything about why, in
turn, all fundamental forces must operate by contact action. But to
fund his explanation of the inverse-square character of all funda-
mental forces, this contact-action regularity must also be a con-
straint rather than a coincidence. For if it were a coincidence, then it
could not be a common reason why every force is inverse-square. At
best, the electric-force law would be inverse-square because electric
charges interact by contact (i.e., through the electric field at each
charge’s location), the gravitational-force law would be inverse-
square because gravity acts by contact, and so forth. In that case, it
would be a coincidence that all of the fundamental forces are inverse-
square, contrary to Hertz’s view (following a broad consensus,
Hertz says) that this regularity is no coincidence.
How is the constraint that all forces be inverse-square supposed
to be explained by the constraint that all forces act by contact (in
three-dimensional space)? Consider a configuration of bodies and
any imaginary surface enclosing them. If a given sort of influence
operates by contact action, then the influence of those bodies on
any body outside of the surface must pass through the intervening
surface (rather than hop “over” it). Therefore, the field at all points
on the surface must fix the influence of the enclosed bodies on any
body outside of the surface. Hence, any two configurations with
the same field at all points on the surface must have the same field
everywhere outside of the surface. The existence of such a “unique-
ness theorem” (as it is commonly called today) imposes strict lim-
its on the form that the force law can take. As Hertz (1999, 68)
rightly notes, the requirement that there be a uniqueness theorem
rules out a force that declines linearly with distance or with the
cube of the distance. Indeed, though Hertz does not mention this
result explicitly, it is a mathematical theorem that for a 1/rn force,
a uniqueness theorem is possible (in three-dimensional space)
only for n = 2 (Bartlett and Su 1994). That is why (according to
Hertz) all of the various fundamental forces are inverse-square
forces.
On Hertz’s view, at least two constraints are not conservation

laws: that all fundamental forces are inverse-square and that all
fundamental forces act by contact. Notice once again that a con-
straint need not be a brute fact; on Hertz’s view, the inverse-square
constraint is explained by the contact-action constraint. It seems to
me that Hertz’s explanation cannot be correct as it stands since an
inverse-square force is not quite the only kind of central force (with
a force law consisting of an analytic function that is real-valued
except, perhaps, at isolated singularities) that permits a uniqueness
theorem.7 Rather, a uniqueness theorem holds for such a force if and
only if it is proportional to 1/(ekr r2) for some real k. This is called a
“Yukawa force law” (or a force with a “Yukawa potential”). An
inverse-square force is the special case where k = 0.
More about the various types of forces is known today than in
Hertz’s day. Not all of the forces that physicists today look upon as
perhaps fundamental are inverse-square. However, if all actual fun-
damental forces are Yukawa forces, then perhaps an argument like
Hertz’s explains why this is so. A Yukawa force was famously pos-
ited by (can you guess?) Yukawa in 1935 as the strong nuclear force
(i.e., the force holding protons and neutrons together in atomic
nuclei). However, even if Hertz’s proposed explanation fails because
not all actual fundamental forces are governed by Yukawa force
laws, my point still stands. An adequate metaphysics must not fore-
close explanations of the sort Hertz proposes on pain of failing to
do justice to the fact that science has rightly taken such proposals
seriously. Many sorts of regularities among the various forces could
be constraints rather than coincidences.
3. CONSTRAINTS AS MODALLY MORE EXALTED

THAN THE FORCE LAWS THEY CONSTRAIN
I have suggested if top-down explanations appealing to conserva-

tion laws succeed, then they work because those laws constrain the
lower-level laws figuring in bottom-up explanations—namely, by
limiting the kinds of forces and force laws there could possibly be.
7
A “central force” is a force directed along the line joining the body exerting it and
the body feeling it.
166 | Marc Lange
In that case, the kinds of forces and force laws there could have
been go beyond the kinds there actually are. For gravitational forces
to exist and to diminish with the inverse-cube of the distance, for
example, is physically impossible (i.e., is logically inconsistent with
some laws of nature—in this case, the gravitational-force law) but
nevertheless possesses some broader species of possibility in being
logically consistent with the various constraints on the force laws.
In contrast, if energy conservation is a constraint, then energy’s con-
servation is not just physically necessary, but also possesses an even
stronger kind of necessity (that is, one that applies to some but not
all of the physical necessities).
On this view, a top-down explanation may proceed by expressly
considering hypothetical states of affairs that the lower-level laws
rule out (and that are not even approximations to or idealizations of
some physical possibility). A top-down explanation may succeed
even if it appeals to a physical impossibility, as long as that hypo-
thetical state of affairs is not ruled out by the constraints on the
lower-level laws. The top-down explanation exploits this broader
species of possibility since it works by showing the explanandum
to possess the corresponding species of necessity (stronger than
physical necessity).
Here is an example of such an explanation: the standard textbook
explanation (originating with J. Willard Gibbs) of the entropy of a
mixture of two non-interacting ideal gases. The explanation uses
energy conservation to account for the expression for ΔS: the differ-
ence between the mixture’s entropy and the entropy of the gases
when separated. Suppose NA molecules of gas A occupy volume VA
(the left side of a container) and NB molecules of gas B occupy vol-
ume VB (the right side); the container is isolated and the two gases
have the same pressure P and temperature T. Suppose gas A is con-
fined behind a freely moveable membrane permeable to B but not
to A, and gas B is similarly confined behind a membrane permeable
to A but not to B. Initially, the two membranes divide the container
along the same plane, so the gases are entirely separated. Then the
membranes are allowed to move slowly, each gas expanding quasi-
statically, so that ultimately the two membranes reach opposite
ends of the container and both gases fill the entire container
(volume VA + VB ). Each gas’s expansion is a reversible isothermal
process. Let W be the total work done on the system:
W = − ∫ VA PdV − ∫ VB P dV = − ∫ VA
VA +VB VA +VB VA +VB dV
N A kT
V
− ∫ VBA
VA +VB dV VA VB
N B kT = N A kT ln + N B kT ln
V +
VA VB VA + VB
By energy conservation (i.e., the first law of thermodynamics), the
change ΔU in internal energy and the heat Q absorbed are related
by
ΔU = Q + W
Since the gases expand isothermically, ΔU = 0 , so Q = −W. Thus
VA + VB V + VB
Q = N A kT ln + N B kT ln A
VA VB
Then since ΔS = Q/T,

VA + VB V + VB
ΔS = N A k In + N B k In A
VA VB
which is the explanandum: the formula for the entropy of a mixture
of two non-interacting ideal gases.
Crucially, this explanation does not presuppose that the lower-
level laws make it possible for there to exist a pair of membranes,
one permeable to A but not to B, the other permeable to B but not to
A. Whether there are any possible materials that could constitute
such membranes depends on the particular gases involved. Gener-
ally, such membranes are impossible. For instance, if molecules of
A are small and uncharged whereas molecules of B are large and
charged, then typically there is nothing that could form a mem-
brane permeable to B but not to A according to the lower-level laws
(which specify the molecular constitution of A and B, as well as the
behavior of physically possible membrane materials).
But remarkably, the thermodynamic explanation is not thereby
undermined. That is because it proceeds entirely from constraints
on possible lower-level laws. As far as those constraints are con-
cerned, such membranes are possible for any molecular species. As
Max Planck said in 1891 in commenting on Gibbs’ derivation of
this equation:
The enormous generalization that Gibbs has given to this tenet and which
must, in and of itself, appear irresponsibly daring, rests clearly on the
168 | Marc Lange
self-evident thought that the validity of so fundamental a tenet as that of

the entropy of a mixed ideal gas, cannot depend on the arbitrary circum-
stance of whether we really have available in each individual case a suitable
semi-permeable membrane. (translated in Seth 2010, 108)
It is not entirely an “arbitrary circumstance” since, after all, it is a

matter of physical necessity. Yet it is “arbitrary” as far as the con-
straints are concerned, since whether any such membranes are
possible for a particular pair of gases is a matter of what the fun-
damental force laws happen to be. Because the top-down explana-
tion shows the explanandum to depend on thermodynamics
alone, the explanation can afford to posit membranes that are
impossible according to lower-level laws.8 The laws of thermody-
namics transcend the laws concerning various particular physi-
cally possible kinds of gas and kinds of materials out of which
membranes could be constructed. That is because the laws of ther-
modynamics (and any explanandum they entail) would still have
held, whether or not the lower-level laws allow a suitable pair of
membranes for a particular pair of gases—an “arbitrary circum-
stance”, as Planck says.
Similarly, to explain various laws concerning dilute solutions,
Planck in 1887 (cf. Fermi 1956, 115) considered what would happen
were the temperature so high and the pressure so slight that the
solute and solvent vaporized into a mixture of ideal gases. After
explaining the equations in this way, Planck wrote:
[I]ncidentally, it is completely inconsequential [gleichgultig] if the given
state can really be arrived at experimentally, and certainly whether it repre-
sents a stable state of equilibrium or not; because these expressions [the
explanandum] are completely independent of this [question]. (translated in
Seth 2010, 102)
Later he elaborated:
8
Many textbooks dance lightly over the fact that these membranes are generally
physically impossible, characterizing the two gases as separated “conceptually”
(Yourgrau, Van der Merwe, and Raw 2002, 235) or “hypothetically” (Annamalai and
Puri 2002, 145) without elaborating any further. Fermi (1956, 101) is admirably forth-
right: “We should notice . . . that in reality no ideal semipermeable membranes exist.
The best approximation of such a membrane is a hot palladium foil, which behaves
like a semipermeable membrane for hydrogen.”
In reality, such a process [vaporizing a dilute solution into a mixture of

ideal gases] will admittedly often not be realizable, because in many cases,
at high temperatures, as are necessary here, chemical transformations occur,
and the molecules are thereby altered. (translated in Seth 2010, 108)
Seth (2010, 102) comments:

[Planck’s point], introduced so casually, was, in fact, anything but inciden-
tal. Planck had clearly seen an apparent objection, and an obvious one at
that. If one considers the case described (a dilute solution, say, of NaCl in
water), lowering the pressure and increasing the temperature does not
automatically produce the result required [a mixture of ideal gases, one of
the solute and one of the solvent]. In most common cases, the water will
vaporize, leaving a solid salt. For Planck’s process, however, one requires
both the salt and the water to vaporize and to maintain their molecular
integrity as compounds. Whether it was at all possible to carry out such a
procedure cannot have been clear to Planck. . . . The argument, however,
was a thermodynamic one and the details of the process, including the very
possibility of its experimental realization, did not matter for Planck. It was
thermodynamically possible and hence the result followed.
“Thermodynamic possibility”, as Seth nicely terms it, is broader

than “physical possibility” because the laws of thermodynamics
constrain the lower-level laws; like logical possibility, thermody-
namic possibility includes more than just the physical possibilities.
Planck’s explanation succeeds, despite trafficking in physical
impossibilities, because it works by showing the explanandum to
be thermodynamically necessary, not merely physically necessary.
No bottom-up explanation could explain the explanandum by
showing it to be inevitable in just this respect because no bottom-up
explanation, rooted in the various particular lower-level laws, could
show that the explanandum would still have held, had the lower-
level laws been different. The lower-level laws do not entail what
the lower-level laws would have been like, had they been different.
I shall now focus on such subjunctive conditionals.
4. MY ACCOUNT OF THE DIFFERENCE BETWEEN

CONSTRAINTS AND COINCIDENCES
Let’s now try to be more precise about what it would take to make
a conservation law into a constraint rather than a coincidence. As
170 | Marc Lange
I hinted at the close of the previous section, I suggest that this dis-
tinction be elaborated in terms of subjunctive conditionals. Energy
conservation constrains the possible force laws exactly when energy
would still have been conserved even if there had been an additional
kind of force (that is, a force that is not electric or gravitational or any
of the other actual kinds) acting together with the various actual
kinds—that is, even if there had been an additional kind of interac-
tion experienced by some of the same entities undergoing some of
the actual kinds of interaction. (If the additional kinds of force were
uninstantiated, then they would obviously pose no threat to energy
conservation. If forces of the additional kinds were not influencing
any of the actual sorts of entities, then they would pose no threat to
the conservation of quantities possessed exclusively by those enti-
ties.) The subjunctive fact associated with energy conservation as a
constraint is supposed to be roughly that energy’s conservation is
resilient: that energy would still have been conserved even if there
had been additional kinds of force threatening to undermine its con-
servation. On the other hand, to say that energy conservation is a
coincidence of the actual force laws is to say that it is not the case that
energy would still have been conserved, had there been additional
kinds of force. Rather, energy is conserved because as it happens,
each of the actual kinds of force conserves energy as a result of its
own particular force law. So had there been additional kinds of force,
energy might still have been conserved, but then again, it might not
have been, depending upon the force laws of the additional forces.
This means of distinguishing constraints from coincidences por-
trays constraints as like “higher-order” laws. The lawhood of Cou-
lomb’s law is traditionally thought to be associated with the fact
that Coulomb’s law would still have held, had there been additional
charged bodies. Similarly, the accidental character of the fact that
each of the families on my block has exactly two children is associ-
ated with the fact that it is not the case that had there been an addi-
tional family on my block, it would still have been true that each of
the families on my block has exactly two children. My account
draws the same sort of distinction at a “higher order”: energy con-
servation is a constraint exactly when energy would still have been
conserved, had there been additional kinds of forces.
This means of distinguishing constraints from coincidences fits
nicely into my more general account of natural law. I have presented
this account in detail elsewhere (Lange 2009). Within the confines of

this paper, I am obviously not able to offer much in the way of argu-
ment for my general account. Rather, my main aim is to argue that
any account of natural law must recognize the distinction between
constraints and coincidences. This cannot easily be done by the
accounts of natural law currently on the market (as I will illustrate
in the following section on dispositional essentialism). Here I offer
my account simply as an example of how it is possible for an analy-
sis of lawhood to leave a natural place for the distinction between
constraints and coincidences (as I have just drawn it) and thereby to
recognize the important role that this distinction plays in science (as
I have suggested in the preceding sections).
As I just mentioned, laws of nature have traditionally been
thought to differ from accidents in having greater perseverance
under counterfactual suppositions. For instance, since it is a law
that no body is accelerated from rest to beyond the speed of light,
this cosmic speed-limit would not have been broken even if the
Stanford Linear Accelerator had now been cranked up to full power.
On the other hand, if it is just an accident that all gold cubes are
smaller than a cubic meter, then had Bill Gates wanted a gold cube
larger than a cubic meter, I dare say there would have been one.
Of course, laws are unable to persist under counterfactual sup-
positions with which they are logically inconsistent. This suggests
the following proposal:
m is a law if and only if in any conversational context, under any counter-
factual supposition p that is logically consistent with all of the laws, m
would still have held (i.e., p □→ m).
In this proposal and until further notice, I reserve letters like “m”
for “sub-nomic” claims, i.e., for claims such as “The emerald at spa-
tiotemporal location . . . is 5 grams” or “All emeralds are green” as
contrasted with “nomic” claims such as “It is an accident that the
emerald at spatiotemporal location . . . is 5 grams” or “It is a law that
all emeralds are green”. (On my view, a claim is “sub-nomic” exactly
when in any possible world, what makes the claim hold (or fail to
hold) is not that a given fact in that world is a law or that a given
fact in that world is an accident.) Let me also note that the account
of laws I am sketching here presupposes that every logical conse-
quence of laws qualifies as a law and that every broadly logical
172 | Marc Lange
truth (e.g., every truth holding with narrowly logical necessity,

metaphysical necessity, mathematical necessity, moral necessity,
etc.) is by courtesy a natural law. (These convenient simplifications
could be dropped, however.)
Although the above proposal captures an important difference
between laws and accidents in their behavior in counterfactuals,
this proposal uses the laws themselves to pick out the relevant
range of counterfactual suppositions. This is problematic since if
there is no prior, independent reason why this particular range of
counterfactual suppositions is special, then the laws’ invariance
under these particular suppositions fails to make the laws special.
They merely have a certain range of invariance (just as a given acci-
dent has some range of invariance).
This problem can be avoided. Let’s start by characterizing what I
shall call “sub-nomic stability”:
Consider a non-empty set Γ of sub-nomic truths containing every sub-
nomic logical consequence of its members. Γ possesses sub-nomic stability if
and only if for each member m of Γ and for any p where Γ∪{p} is logically
consistent (and in every conversational context), it is not the case that ~m
might have held, had p held (i.e., ~ (p ◊→ ~ m)).9
Notice that ~ (p ◊→ ~ m) logically entails p □→ m. Therefore, a set

of truths is sub-nomically stable exactly when its members would
all still have held (indeed, not one of their negations might have
held) under any counterfactual supposition with which they are all
logically consistent. So in contrast to the earlier proposal, stability
does not use the laws to pick out the relevant range of counterfac-
tual suppositions. Rather, each set picks out for itself the range
under which it must be invariant in order for it to be stable.
This suggests my proposal for distinguishing laws from accidents:
that the set Λ of all sub-nomic truths m where it is a law that m is sub-
nomically stable, whereas no set containing an accident is sub-nomi-
cally stable (except perhaps for the set of all sub-nomic truths,
considering that the range of counterfactual suppositions under which
this “maximal” set must be preserved in order to qualify as stable
does not include any false suppositions since no falsehood is logically
9
For the sake of simplicity, this definition of “sub-nomic stability” omits some
details from my (2009) that will not make any difference here.
consistent with all of this set’s members). For instance, the set spanned
by the fact that all gold cubes are smaller than a cubic meter is unsta-
ble because this set’s members are all logically consistent with Bill
Gates wanting a gold cube larger than a cubic meter, yet the set’s
members are not all invariant under this counterfactual supposition.
It is a law that m, then, exactly when m belongs to a (non-maxi-
mal) sub-nomically stable set. Now let’s show that this account
leaves a natural place for the distinction between constraints and
coincidences. Are there any other non-maximal sub-nomically sta-
ble sets besides Λ? The sub-nomic broadly logical truths form a sub-
nomically stable set. I’ll now show that for any two sub-nomically
stable sets, one must be a proper subset of the other:
1. Suppose (for reductio) that Γ and Σ are sub-nomically stable,
t is a member of Γ but not of Σ, and s is a member of Σ but
not of Γ.
2. Then (~s or ~t) is logically consistent with Γ.
3. Since Γ is sub-nomically stable, every member of Γ would
still have been true, had (~s or ~t) been the case.
4. In particular, t would still have been true, had (~s or ~t)
been the case. That is, (~s or ~t) □→ t.
5. So t & (~s or ~t) would have held, had (~s or ~t). Hence, (~s
or ~t) □→ ~s.
6. Since (~s or ~t) is logically consistent with Σ, and Σ is sub-
nomically stable, no member of Σ would have been false
had (~s or ~t) been the case.
7. In particular, s would not have been false, had (~s or ~t)
been the case. That is, ~((~s or ~t) □→ ~s).
8. Contradiction from 5 and 7.
Thus, the sub-nomically stable sets must form a nested hierarchy.
Since no non-maximal superset of Λ is stable (since it would
include an accident), any other stable sets must be among Λ’s proper
subsets. Many of them are clearly unstable. For instance, the set
spanned by a restriction of Coulomb’s law to the past is unstable
since had Coulomb’s law been violated sometime in the future,
then (with Coulomb’s law “out of the way”) it might have been
violated sometime in the past.
However, some of Λ’s proper subsets may be stable, and I sug-
gest that any constraint must belong to at least one such set. Other
174 | Marc Lange
members of such a set plausibly include the fundamental dynami-

cal law, the law of the parallelogram of forces, the space-time
transformations, and other laws that “transcend” the particular
kinds of forces there happen to be; the various force laws and the
“closure law” specifying the actual kinds of forces are excluded
from such a set. In other words, if a conservation law belongs to
no non-maximal stable set besides Λ, then it is a coincidence. If
energy conservation belongs to a stable proper subset of Λ that
omits the various force laws and the closure law, then the set’s
stability requires the subjunctive fact that (I proposed) distin-
guishes constraints from coincidences (that energy would still
have been conserved, had there been additional kinds of forces)
since the supposition that there are additional kinds of forces is
logically consistent with each of the set’s members. Energy con-
servation’s status as a constraint is then associated with its invari-
ance under a certain range of counterfactual antecedents, and that
range consists of those antecedents that are logically consistent
with every member of a stable subset of Λ to which energy conser-
vation belongs. For instance, if the various particular force laws
are all omitted from that set, then in connection with its status as
a constraint, energy conservation would still have held, had grav-
ity not been an inverse-square force.
This approach leaves room for multiple levels of constraints on
the force laws, each one associated with a stable set that occupies
a spot in the nested hierarchy of stable sets somewhere between
the set of broadly logical truths and Λ. Moreover, this approach
accounts for the role of constraints as higher-order laws—that is,
laws that are modally more exalted than the force laws they con-
strain, and so able to explain why all of the forces share certain
features. It accounts for the way in which the constraints carve out
a species of possibility that is more inclusive than physical possi-
bility (as we saw in connection with “thermodynamic possibility”,
which embraces some physical impossibilities). The members of a
stable set would all still have held under any counterfactual sup-
position with which they are all logically consistent—that is,
under which they could (i.e., without contradiction) all still have
held. In other words, a stable set’s members are collectively as
resilient under counterfactual suppositions as they could collec-
tively be. They are maximally resilient—that is to say, necessary. Accord-
ingly, I suggest that a sub-nomic truth has a species of necessity

exactly when it belongs to a non-maximal sub-nomically stable
set, and that for each of these sets, there is a distinct species of
necessity that is possessed by exactly its members. On this view,
then, whereas constraints and coincidences are both physically
necessary, a constraint also possesses a species of necessity (a
stronger cousin of physical necessity) that the coincidences and
force laws lack. Thus, we are entitled to say that a constraint limits
the kinds of forces there could have been, whereas a coincidence
merely reflects the kinds of forces there happen to be. The actual
inventory of forces is a matter of physical necessity and yet also a
matter of happenstance in that it lacks the stronger necessity pos-
sessed by a constraint.
Finally, this view explains why any conservation law that follows
from a symmetry principle within a Hamiltonian dynamical frame-
work constitutes a constraint rather than a coincidence and so (as
Wigner says) “transcends” the various force laws. A symmetry
principle, such as the fact that the laws are invariant under arbitrary
temporal displacement, is not expressed by a sub-nomic claim.
Rather, a symmetry principle is made true by which facts are laws.
It is expressed by a “nomic” claim, i.e., a claim that purports to
describe which truths expressed by sub-nomic claims are (or are
not) matters of law. So to characterize the invariance that is charac-
teristic of symmetry principles as “meta-laws”, we need an ana-
logue of sub-nomic stability that applies to sets of claims that are
either sub-nomic or nomic. Here it is (now allowing letters like “p”
to stand for claims that are either sub-nomic or nomic):
Consider a non-empty set Γ of truths that are nomic or sub-nomic contain-
ing every nomic or sub-nomic logical consequence of its members. Γ pos-
sesses nomic stability if and only if for each member m of Γ (and in every
conversational context), ~ (p ◊→ ~m) for any p where Γ∪{p} is logically
consistent.
The symmetry principles are meta-laws in that they form a nomi-

cally stable set more exclusive than the set spanned by all truths
about which sub-nomic claims are laws and which are not. Moreo-
ver, if the conservation laws are logically entailed by symmetry
meta-laws within a Hamiltonian dynamical framework, then they
belong to a sub-nomically stable set that is more exclusive than Λ
176 | Marc Lange
and therefore are constraints. That is because for any nomically sta-
ble set, its sub-nomic members must form a sub-nomically stable
set. Here is the proof:
1. If p (a sub-nomic claim) is logically inconsistent with a
nomically stable set Γ, then Γ must entail ~p (also sub-
nomic), and so p is logically inconsistent with the set Σ con-
taining exactly Γ’s sub-nomic logical consequences.
2. Conversely, if p is logically inconsistent with Σ, then obvi-
ously p is logically inconsistent with Γ.
3. By Γ’s nomic stability, Σ is preserved under every sub-
nomic antecedent p that is logically consistent with
Γ—which (we have just shown) are exactly those sub-
nomic antecedents that are logically consistent with Σ.
Hence, Σ is sub-nomically stable.
Therefore, if the symmetry meta-laws (forming a nomically stable
set) entail that a given conservation law holds under the Hamilto-
nian dynamical framework, then the conservation law’s holding if
the Hamiltonian dynamical law holds belongs to a sub-nomically
stable set that is more exclusive than Λ (since presumably, not all of
Λ’s members follow from the symmetry meta-laws’ nomically stable
set). Hence, if the Hamiltonian dynamical law is not a member of
that set but (in transcending the various particular force laws)
belongs to another sub-nomically stable set that does not include the
force laws, then (since the sub-nomically stable sets form a nested
hierarchy) the fact that the conservation law holds under the Hamil-
tonian framework must also belong to that set, and hence (by the
set’s logical closure) the conservation law must belong, too. So it
constitutes a constraint. In other words, that the conservation law
would still have held, even if the force laws had been different, fol-
lows from the fact that not only the fundamental dynamical law, but
also the symmetry meta-law would still have held had the force laws
been different.
Of course, I cannot do more here than sketch the relevant parts of
this conception of natural law. But it is worth seeing how an account
of lawhood can incorporate the constraint/coincidence distinction
in a natural way. Now I shall conclude by turning to an approach to
natural law that is ill-equipped to have the same success.
5. DISPOSITIONAL ESSENTIALISM RULES OUT

CONSTRAINTS
According to Alexander Bird (2007), every sparse fundamental

property of physics is constituted by one or more dispositions. On
Bird’s view, the association between a fundamental property of
physics and some disposition is a matter of metaphysical necessity;
moreover, the identity of a given fundamental property of physics
is exhausted by its dispositional character. Therefore, it is meta-
physically necessary that any entity possessing a certain sparse fun-
damental property of physics exhibit certain further properties if
suitably stimulated. These regularities, or the corresponding rela-
tions among properties, are the laws of nature. Although meta-
physically necessary, the laws do not perform the explanatory
heavy-lifting. The motor and cement of the universe are the dispo-
sitional essences of the fundamental properties of physics. Views
along roughly similar lines have been proposed by Brian Ellis (2001;
2002) and Stephen Mumford (2004), among others. These views dif-
fer in some details from Bird’s: for example, Ellis takes the disposi-
tional essences responsible for laws to be the essences of the natural
kinds rather than of the sparse fundamental properties of physics,
whereas Mumford holds that since any fundamental property of
physics is constituted by a cluster of causal roles and other connec-
tions to other properties (such as its excluding certain properties
and being compatible with certain others), there are no laws because
nothing governs property instances in the manner traditionally
ascribed to laws. However, these differences will make little differ-
ence here.
A conservation law does not say that any entity possessing a cer-
tain sparse fundamental property of physics exhibits certain further
properties if suitably stimulated. Unlike (for instance) the occur-
rence of a certain force, energy’s remaining conserved is not the
manifestation of a particular disposition. Therefore, it is difficult for
views like Bird’s to accommodate conservation laws, as Bird (2007,
211) himself notes.
It is worth distinguishing three forms that this objection can take.
The first form is that energy’s remaining conserved under any inter-
action of a given kind is simply not the manifestation of any dispo-
sition constituting any of the sparse fundamental properties of
178 | Marc Lange
physics manifested in such interactions. The dispositions associ-

ated with electric charge, for instance, may manifest themselves in
various accelerations or various contributions to the electric field.
Thus, the laws specifying that charges manifest themselves in cer-
tain ways under certain conditions do not include among these
manifestations anything about energy being conserved. The law
that electric interactions conserve energy therefore seems to have
been left out of the account. However, it seems to me that the law
has not been neglected. From the laws concerning electric interac-
tions, it follows logically that such interactions conserve energy. So
on this account, energy’s conservation in electric interactions is a
metaphysical necessity arising from the dispositions essential to
various fundamental properties of physics, including charge. There
is no problem yet for Bird’s view.
The second form of the objection is that the law of energy conser-
vation does not follow simply from the law that energy is conserved
in electric interactions, the law that energy is conserved in gravita-
tional interactions, and so forth. There must be a further premise,
since these force laws taken together do not preclude the existence
of another kind of process that fails to conserve energy. One premise
that would close the gap is a “closure law”: that electric interac-
tions, gravitational interactions, and so forth are all of the kinds of
interactions there are, i.e., that every fundamental natural process
belongs to one of these kinds. The objection, then, is that this “clo-
sure law” does not reflect any property’s dispositional essence.
Hence, the overall conservation law is not the reflection of the dis-
positional essences of the fundamental properties of physics.
Ellis’s view purports to be invulnerable to this form of the
objection. Ellis takes the laws to be grounded not in the disposi-
tional essences of various properties, but rather in the essences of
the various natural kinds. He suggests (following Bigelow, Ellis,
and Lierse 1992) that the world is the only member of a certain
natural kind and that the essence of this kind includes various
quantities being conserved (Ellis 2001, 205 and 250). Its essence
also includes its being populated by exactly certain sorts of parti-
cles and fields undergoing exactly certain sorts of interactions,
thereby accounting for the “closure laws”. I have replied else-
where that this move “seems like a desperate attempt to find
something the essence of which could be responsible for various
laws. Even if gravity and the electron have essences, it is not obvi-
ous that ‘the world’—reality—does” (Lange 2009, 83). Bird also
finds this move “somewhat ad hoc” (2007, 213). However, I shall
not pursue this concern here.
This form of the objection may turn out not to pose a terribly
severe challenge for Bird’s view. One option that Bird might explore
is to say simply that although the conservation “law” is not a meta-
physical necessity, the fact that energy is conserved is no accident
either. Rather, it is grounded in the sparse fundamental properties
of physics, i.e., in the world’s repertoire of fundamental causal pow-
ers. But this option would require Bird to allow different repertoires
in different possible worlds, which Bird is reluctant to do (since in
that case the actual laws, though still true in all worlds, are not laws
in some of them). Another option for elaborating Bird’s view accords
metaphysical necessity to the conservation law and allows every
world to have exactly the same laws. As I just explained, energy’s
conservation in electric interactions is a metaphysical necessity aris-
ing from the dispositions essential to various fundamental proper-
ties of physics. So likewise is energy’s conservation in gravitational
interactions, and similarly for any other actual type of interaction.
On Bird’s view, the various fundamental properties of physics
would be different properties if they bestowed additional powers
and susceptibilities—for instance, the power to exert and the sus-
ceptibility to feel some alien force that fails to conserve energy. Pre-
sumably, no metaphysical necessity precludes the instantiation of
alien fundamental properties of physics. But “electric charge” (or
any other non-alien property) would be a different property if it
bestowed susceptibility not just to the influence of other electric
charges, but also to some other alien influence. What if we go fur-
ther and suppose that all of the non-alien properties (or at least
enough of them that all non-alien kinds of entities must possess
one) have as part of their essences that they bestow immunity to
being influenced by any alien property (and so their possession by
an entity precludes its also possessing any alien property that would
bestow susceptibility to being so influenced)? Then the possession
of any alien properties (by entities not possessing any of the non-
alien properties) could not influence the behavior of any entities
possessing non-alien properties and so could not disturb the con-
servation of the quantities that are in fact conserved under each of
180 | Marc Lange
the various actual types of interaction, as long as those quantities

are possessed only by entities possessing the non-alien properties.
This is obviously the case for electric charge and, by the disposi-
tional essentialist’s lights, arguably the case for energy and momen-
tum, since these properties (the essentialist might say) are constituted
by various combinations of non-alien fundamental properties such
as mass, velocity, charge, separation, and so forth. In that case,
although these quantities are not instantiated in every possible
world, they are conserved in any world where they are instantiated,
since the non-alien properties constituting these quantities bestow
immunity to any other influences besides the various actual types
of forces. Energy conservation is then metaphysically necessary
even if the “closure law” is not and even though the conservation
law is not the manifestation of any single fundamental property’s
dispositional essence.
However, let’s now consider a third form of the original objection
that views like Bird’s cannot properly accommodate conservation
laws. This form of the objection will prove more difficult for views
like Bird’s to answer. On Bird’s picture, even if energy conservation
is metaphysically necessary, it is the product of the various particu-
lar types of interactions there are: that gravitational forces conserve
energy, electric forces conserve energy, and so forth. The various
dispositions essential to various fundamental natural properties are
responsible for energy’s conservation. So although energy conser-
vation is metaphysically necessary, it is a coincidence rather than a
constraint.
To appreciate this, recall the subjunctive fact that (I have sug-
gested) is what it is for energy conservation to be a constraint: that
energy would still have been conserved even if there had been an
additional kind of interaction experienced by some of the entities
undergoing some of the actual kinds of interaction. This condition-
al’s antecedent is a countermetaphysical according to the view
I have just offered Bird in response to the previous form of the objec-
tion. Even without that response, on Bird’s view there is no disposi-
tion (or collection of dispositions) essential to some sparse
fundamental property of physics (or collection of such properties)
that is available to underwrite this conditional’s truth. On Bird’s
view, the causal powers constituting those properties do the meta-
physical heavy lifting, but these causal powers cannot underwrite
this conditional’s truth; none of them has as its manifestation that

certain alien fundamental natural properties are instantiated or has
as its manifestation-eliciting stimulus that some alien kind of inter-
action occurs. The conditional associated with being a constraint
does not concern what the actual causal powers would produce
under suitable stimulation, but rather concerns what causal powers
there would be if there were more besides the actual ones—so the
actual causal powers do not sustain this conditional’s truth. (The
same applies to other subjunctive conditionals whose truth would
reflect a conservation law’s status as a constraint, on my view of
constraints as belonging to stable proper subsets of Λ. For example,
the fact that energy would still have been conserved, had gravity
not been an inverse-square force, posits a countermetaphysical and
is underwritten by no causal powers, on Bird’s view.) Ellis’s view
faces a similar problem: there is no natural kind whose essence can
step in to sustain the conditional needed to make energy conserva-
tion a constraint. Even the world’s essence cannot do the job since
the conditional’s antecedent is inconsistent with that essence. On
Ellis’s view, had there been additional kinds of interactions, there
would have been a different kind of world; it might have been a
world where energy is conserved, but then again, it might not.
Energy conservation is therefore a coincidence, not a constraint.
On a view like Bird’s, a conservation law cannot be a constraint
because a conservation law cannot be explanatorily prior to the
force laws—or, rather, to the various dispositions responsible for
the force laws and essential to the various sparse fundamental
properties of physics. Gravitational interactions conserve energy
not because the law of energy conservation requires them to, but
because of the causal powers involved in gravitational interac-
tions (which the gravitational-force law reflects). The conserva-
tion laws cannot explain why (e.g.) no fluid circulates; rather, the
explainers must be the particular powers involved. Even if a
given power’s conserving energy when it is manifested suffices
to entail the explanandum, energy conservation as a comprehen-
sive law covering all of the actual powers is explanatorily irrel-
evant; all that matters to the outcome is the fact that the particular
power at work in this case conserves energy. This is how all
scientific explanations must work, according to Ellis—from the
bottom up:
182 | Marc Lange
Essentialists seek to expose the underlying causes of things, and to explain

why things are as they are, or behave as they do, by reference to these under-
lying causal factors. Consequently, explanations of the sort that essentialists
are seeking must always have two parts. They must contain hypotheses
about the underlying structures or causal powers of things, and hypotheses
about how things having these structures and powers must behave in the
specific circumstances in which they exist. (Ellis 2002, 159–60)
This approach to scientific explanation leaves no room for the top-

down explanations that constraints supply. Why are two forces
alike in conserving energy? Why do no kinds of interaction put flu-
ids spontaneously into circulation? That energy conservation con-
strains the kinds of forces there could have been is not the kind of
explanation Ellis allows. On Ellis’s picture, the explananda are
merely upshots of the various particular kinds of interaction built
into the world’s essence, all of which conserve energy.
Bird likewise seems to find top-down explanations from conser-
vation laws problematic:
[T]here is something mysterious about conservation laws. They seem to
require explanation . . . How does a system know that energy should be con-
served? . . . It is not clear how these could be fundamental laws—they seem
to stand in need of a deeper explanation. (2007, 213)
The deeper explanation Bird has in mind would be in terms of the

particular causal powers at work in the system. It is these powers
that allow a system to “know that energy should be conserved”.
This is the order of explanatory priority that is characteristic of
conservation laws as coincidences: the various force laws (or, for
the essentialist, the various particular causal powers responsible
for the force laws) come before the conservation law. But as we saw
earlier in this paper, conservation laws as constraints provide
explanations that are in some respects “deeper” than any explana-
tions appealing to the particular powers involved. By constituting
common explainers, conservation laws as constraints unify the
non-circulation of ropes and fluids, for example, and likewise unify
various otherwise unrelated forces as all inverse-square (or all sub-
ject to uniqueness theorems) or all conserving energy for the same
reason. These explanations reveal that the fact to be explained
would still have held even if different fundamental forces had been
at work.
That conservation laws must be coincidences rather than con-

straints on his account seems to be roughly what Bird has in mind
when he acknowledges that his account cannot accommodate con-
servation laws:
Conservation and symmetry laws tell us that interactions are constrained
by the requirement of preserving, e.g., mass-energy or momentum . . . [T]he
dispositional essentialist holds that the laws are necessary. If that is correct
there is no room for further constraints. Properties are already constrained
by their own essences and so there is neither need nor opportunity for
higher-order properties to direct which relations they can engage in. (Bird
2007, 211 and 214)
Bird’s thought seems to be that as constraints on (the causal powers

responsible for) the various force laws, conservation laws would
have to impose limitations on the possible manifestations of the
powers associated with electric charge, mass, and so forth, limiting
the interaction laws to those that conserve energy. But these interac-
tion laws are already fully determined by the properties involved in
the interactions since those properties are essentially constituted by
various causal powers, which fix their own possible manifestations.
So there is no further constraining for a conservation law to do. It
seems to me that there is more that the conservation law could con-
strain: what powers there would be, if additional fundamental nat-
ural properties were instantiated. But this would require that the
conservation law do some metaphysical heavy lifting—a job that,
on Bird’s picture, is reserved for the causal powers.
In short, a conservation law as a constraint is a higher-order law;
it has greater necessity than any of the motley collection of various
particular force laws and thus explains why those laws are all alike
in exhibiting certain features. But on views like Bird’s and Ellis’s,
the particular force laws hold as matters of metaphysical necessity.
No greater necessity is left for a constraint to possess; necessity has
already been maxed out in the force laws. So there is no genuine
“thermodynamic possibility”—no sense in which various actual
properties could have been associated with certain alien powers but
not with others (namely, with only those alien powers that, in being
manifested, would conserve energy, momentum, and so forth).
Therefore, conservation laws cannot be constraints; they can only
be coincidences.
184 | Marc Lange
Bird seems content with this result:

The dispositional essentialist ought to regard symmetry principles as
pseudo-laws. . . . So it may be that symmetry principles and conservation
laws will be eliminated as being features of our form of representation
rather than features of the world requiring to be accommodated within our
metaphysics. (Bird 2007, 214)
But I have argued (in the opening sections of this paper) that the
price of adopting this view is too high. It precludes top-down expla-
nations of a kind that science has reasonably taken seriously—
indeed, on which science has often placed great importance. Views
like Bird’s rule out scientifically respectable theories. Perhaps Bird’s
prognostication regarding future science will be proved right; per-
haps symmetry principles and conservation laws, along with the
alleged top-down explanations they supply, will ultimately be
eliminated from physics—and their place not be taken by other
top-down explanations (like Hertz’s) employing other sorts of
constraints. Personally, I doubt it. But more importantly, a meta-
physics that cannot do justice to top-down explanations and the
constraints they require is at a serious disadvantage even if as a
matter of fact, there turn out to be no such explanations. Room
should still be left for them; it should be up to science rather than
metaphysics to foreclose them.
Consider, for instance, the conservation of baryon number in
contemporary physics. By energy conservation, an isolated proton
can decay only into particles that have less rest-mass than it does,
and the proton is the lightest baryon (i.e., the lightest particle with
non-zero baryon number). The conservation of baryon number thus
entails that the proton is stable (radioactively, I mean—not “sub-
nomically!). Does baryon-number conservation explain why the
proton is stable? This remains controversial.
It is no explanation if the conservation of baryon number is a so-
called “accidental symmetry” (a term that was introduced by Ste-
ven Weinberg; see Weinberg 1995, 529). An accidental symmetry
reflects merely the particular forces in action at lower-energy
regimes rather than some deeper “symmetry of the underlying the-
ory” (Weinberg 1995, 529). Accordingly, if baryon-number conser-
vation is an accidental symmetry, then it may not even hold at
higher energies (and so the proton may turn out not to be stable, but
rather to have an extremely long half-life). But even if an accidental

symmetry is unbroken, it would still be a coincidence of the partic-
ular kinds of interactions written “by hand” into the underlying
theory, and so it would fail to explain. (The stability of the lightest
baryon would then help to explain why the baryon number turns
out to be conserved, not vice versa.)
On the other hand, baryon-number conservation may turn out to
be a consequence of a more fundamental symmetry, in which case it
would help to explain the proton’s stability. Thus, whereas some
physicists cite baryon-number conservation as explaining why the
proton is stable (e.g., Davies 1986, 159; Duffin 1980, 82), other physi-
cists put scare-quotes around “explain” (Lederman and Teresi 1993,
303) or say that the jury is still out (Ne’eman and Kirsh 1996, 150–1).
The uncertainty regarding the conservation law’s explanatory
power is matched by the uncertainty regarding the law’s status as
constraint or coincidence. But this live scientific controversy would
be settled outright by views like Bird’s. I do not think that meta-
physics should prejudge the outcome.
If the laws are the upshot of the inventory of powers, and hence
(according to dispositional essentialism) of the sparse fundamental
properties of physics, then any regularity in those powers (even if
metaphysically necessary) is coincidental; no law transcends the
various powers, constraining those there could have been. How-
ever, an adequate metaphysics should allow the “great general
principles which all the laws seem to follow” to be constraints; it
should not require them to be coincidences.
University of North Carolina at Chapel Hill
REFERENCES
Annamalai, K. and Puri, I. (2002), Advanced Thermodynamic Engineering.
Boca Raton, CRC Press.
Bartlett, D.F. and Su, Y. (1994), “What Potentials Permit a Uniqueness
Theorem,” American Journal of Physics 62: 683–6.
Bigelow, J., Ellis, B., and Lierse, C. (1992), “The World as One of a Kind:
Natural Necessity and Laws of Nature,” British Journal for the Philoso-
phy of Science 43: 371–88.
Bird, A. (2007), Nature’s Metaphysics. Oxford: Oxford University Press.
186 | Marc Lange
Callendar, C. (2005), “Answers in Search of a Question: ‘Proofs’ of the

Tri-dimensionality of Space,” Studies in History and Philosophy of Modern
Physics 36: 113–36.
Davies, P. (1986), The Forces of Nature. Cambridge: Cambridge University
Press.
Duffin, W.J. (1980), Electricity and Magnetism, 3rd ed. London: McGraw-Hill.
Ellis, B. (2001), Scientific Essentialism. Cambridge: Cambridge University
Press.
—— (2002), The Philosophy of Nature: A Guide to the New Essentialism. Mon-
treal & Kingston: McGill-Queen’s University Press.
Fermi, E. (1956). Thermodynamics. New York: Dover.
Feynman, R. (1967), The Character of Physical Law. Cambridge: MIT Press.
Hertz, H. (1999), Die Constitution der Materie, ed. Albracht Fölsing. Berlin:
Springer-Verlag.
Lange, M. (2002), An Introduction to the Philosophy of Physics: Locality,
Fields, Energy, and Mass. Malden, MA: Blackwell.
—— (2009), Laws and Lawmakers. New York: Oxford University Press.
Lederman, L. and Teresi, D. (1993), The God Particle. New York: Dell.
Mumford, S. (2004), Laws in Nature. London and New York: Routledge.
Ne’eman, Y. and Kirsh, Y. (1996), The Particle Hunters. Cambridge:
Cambridge University Press.
Seth, S. (2010), Crafting the Quantum. Cambridge: MIT Press.
Stevin, S. (1955), The Principal Works of Simon Stevin. Volume 1. Amster-
dam: C.V. Swets & Zeitlinger.
Weinberg, S. (1995), The Quantum Theory of Fields, volume 1. Cambridge:
Wigner, E. (1972), “Events, Laws of Nature, and Invariance Principles,” in
Wigner, Collected Papers, Part A, volume 3. Berlin: Springer, 185–96.
Yourgrau, W., Van der Merwe, A., and Raw, G. (2002), Treatise on Irrevers-
ible and Statistical Thermophysics. New York: Dover.
IV
CONTINGENT OBJECTS
AND COINCIDENT
OBJECTS
7. Relativized Metaphysical Modality*
Adam Murray and Jessica Wilson†
INTRODUCTION
Metaphysical necessity and possibility are commonly supposed to be

necessity and possibility in the broadest, not merely syntactically logi-
cal, sense. Hence it is that metaphysical modality is often contrasted
with other, restricted forms of modality, as when Burgess (2009) says:
[W]e may distinguish the species of physical necessity, or what could not have
been otherwise so long as the laws of nature remained the same, from meta-
physical necessity, what could not have been otherwise no matter what. (46)
In quantificational terms, the supposition is that a single domain of

possible worlds enters into metaphysical modal evaluation; a claim
is metaphysically necessary just in case it is true in every possible
world in the domain, and metaphysically possible just in case it is
true in some possible world in the domain. We argue here that the
standard understanding is strictly incorrect; rather, whether a given
claim is metaphysically necessary or possible depends on which
world is, as we put it, ‘indicatively actual’. In brief: metaphysical
necessities and possibilities are relativized to indicative actualities.
The proper understanding of metaphysical modality thus takes
modal space to have a complex, relativized structure. The sense in
which the standard view is correct concerns its coinciding with
metaphysical modality when relativized to our very own indica-
tively actual world; the sense in which the standard view is incorrect
concerns its failing to be sensitive to the more complex relativized
structure of metaphysical modality.
* Special thanks to Karen Bennett and Dean Zimmerman for extensive comments
which greatly improved this paper. Thanks also to Benj Hellie, Phil Kremer, Dan
Rabinoff, Chris Tillman, and audience members at the University of North Carolina
and the University of Miami, for helpful comments and questions, and to Brent
Cromwell for assistance in constructing the figures.
†
Department of Philosophy, University of Toronto; adam.murray@utoronto.ca,
jessica.m.wilson@utoronto.ca
190 | Adam Murray and Jessica Wilson
We motivate the alternative proposal by attention to discussions in

Salmon (1989) and Fine (2005). In each discussion, the author canvasses
data which he takes to support a certain thesis—in Salmon’s case, that
the transitivity of the accessibility relation between possible worlds,
and associated systems of modal logic S4 and S5, should be rejected as
characteristic of metaphysical modality; in Fine’s case, that nomological
and metaphysical modality should be taken to be distinct and equally
basic (as per “modal pluralism”). We argue that the data in each case
can be accommodated, compatible with transitive accessibility and
with modal monism, if metaphysical necessities and possibilities are
relativized to indicative actualities; and we offer two ways of imple-
menting the relativized conception within a possible-worlds semantics
for metaphysical modal logic. We also note, for heuristic purposes, a
formal analogy between the relativized conception and a thoroughly
metaphysical interpretation of the ‘secondary’ or ‘horizontal’ intensions
associated with the two-dimensional semantic framework, which inten-
sions may be seen as representing what is counterfactually possible
relative to each indicatively actual world. We close by observing the
neutrality of our conception as regards the actualist/possibilist and
trans-world identity/counterpart theory distinctions.1
1 SALMON’S ‘WOODY’ CASE
1.1 Possible Worlds Semantics and Transitive Accessibility

It is intuitively natural and historically familiar (following Leibniz
1686) to characterize modal claims in quantificational terms, where
the evaluation of such claims reflects the spectrum of truths across
a given range of possible worlds. Such a characterization is for-
mally vindicated in possible worlds semantics for modal logics
(Kripke 1963). A modal logic extends the usual propositional
(or predicate; here we follow Salmon in focusing on the simpler
case) logics by introducing symbols ‘□’ and ‘◊’, along with certain
1
Our project here bears some similarity to but is in key respects different from
projects of the sort at issue in Crossley and Humberstone (1977) and Davies and
Humberstone (1979), in which the standard modal logic(s) are supplemented with an
actuality operator (‘A’) and associated modal operators (‘fixedly’, ‘fixedly actually’).
In the earlier paper, the motivation for an actuality operator is that scope interactions
between quantifiers and standard modal operators fail to allow expression of claims
like ‘It is possible that everything that is in fact red is shiny’, and the additional modal
Relativized Metaphysical Modality | 191
rules or axioms supposed to govern expressions containing these

symbols on any of a wide range of interpretations, which include
the necessitation rule (if p is a theorem of the logic, infer □p), and
the distribution axiom (K) (□(p → q) → (□p → □q)).2 Creating a
model for such a logic then involves two steps. The first step
involves specifying a frame: a set W of possible worlds, along with
a relation R between worlds; the desired features of the relation are
encoded in certain axioms, to be discussed shortly. The second step
involves specifying a valuation function v setting up the basic non-
modal facts in each world; truth clauses for expressions prefaced
with one or other modal symbol are then added to the usual truth
clauses in such a way that the truth of all basic and non-basic claims
in the model is determined. The relevant modal clause (also deter-
mining, given that necessity and possibility are duals, the clause
for claims involving the possibility operator) is then schematically
as follows:
v(□p, w) = T iff for every world w’ in W such that wRw’, v(p, w’)
= T.
operators are briefly introduced in order to respond to concerns about the validity of
axioms of the form Aα→□ Aα on the semantics offered for ‘A’, that ‘“actually a” need
not have been true because another world might have been actual’ (1979, 2); in the
later paper, these new modal operators are applied in service of accommodating
putative cases of the contingent a priori and necessary a posteriori, with the assist-
ance of certain theses about names and natural kind terms. Our motivations and
target applications are different. We aim to show that certain puzzles, having to do
with natures or essences as opposed to scope or semantics (or epistemology), ulti-
mately arise from a failure of standard metaphysical modal logics to incorporate rela-
tivization to indicatively actual worlds, and to argue that proper incorporation of a
relativized structure makes for better resolutions of these puzzles than those cur-
rently on offer. Our suggested ways of making sense of such relativization within a
possible worlds semantics for modal logic do not involve any additional operators,
and though in the course of explicating our view we heuristically appeal to properly
metaphysical interpretations of the notions, familiar from 2-D semantics, of consider-
ing worlds ‘as actual’ or ‘as counterfactual’, unlike Davies and Humberstone we are
officially neutral on both the semantics and epistemology of names and natural kind
terms. These differences aside, in arguing that the traditional modal operators should
be relativized to indicatively actual worlds we are on the same side as these other
authors; our contribution here is, first, to offer distinctively metaphysical reasons for
incorporating such relativization, and second, to show that this can be done in ways
minimally departing from standard modal logic(s).
2
More precisely, modal logics that include the distribution axiom (□(p → q) → (□p
→ □q)) are known as normal modal logics. We assume in what follows that the modal
logics under discussion are normal.
As above, different systems of modal logic impose different con-

straints on the relation R at issue, which are encoded, either explic-
itly or indirectly, in certain axioms. Typically, the relation R is
understood as an ‘accessibility’ relation; in the case of metaphysical
modality, the features of this relation are intended to ensure—again
reflecting the standard understanding of metaphysical necessity as
necessity in the broadest sense—that the facts holding at any and all
possible worlds are relevant to metaphysical modal deliberation.
Such an accessibility relation is standardly supposed to be reflexive,
such that any world is accessible to ‘(can “see”)’ itself; this require-
ment is encoded in axiom (T):
(T): □p → p (for any necessarily true proposition p, the propo-
sition that p is true).
The resulting system (i.e., the system imposing no further con-
straints on R) is system T. The accessibility relation is also stand-
ardly supposed to be transitive, which requirement is satisfied by
adding axiom (4) to system T:
(4): □p → □□p (for any necessarily true proposition p, the
proposition that p is necessarily true is itself necessarily
true).
The resulting system is system S4. Finally, the accessibility relation
is standardly supposed to be symmetric, which requirement is sat-
isfied by adding axiom (5) to system T:
(5): ◊p → □◊p (for any possibly true proposition p, the propo-
sition that p is possibly true is itself necessarily true).
The resulting system is system S5. Since in S5 the accessibility rela-
tion R is also reflexive and transitive, R in S5 is an equivalence rela-
tion. It is commonly assumed that S5 is the correct logic for
metaphysical modality (see Sider 2010).3
3
This assumption reflects, in part, that the theorems of S5 coincide with those on
a modal logic where the accessibility relation is ‘total’, such that (as per the standard
conception) every world is accessible to every other. To prefigure a bit: one way of
implementing our transitive relativized conception of metaphysical modality takes
advantage of the fact that, notwithstanding the coincidence of theorems, S5 is com-
patible with modal space being partitioned into non-overlapping equivalence classes
(see S2.2).
1.2 Salmon’s Rejection of Transitive Accessibility

Salmon (1981, 1984, and 1989) argues (following Chandler 1976) that
this is a mistake: axiom (4) has false instances, and so S4 and the stronger
system S5 are fallacious logics of metaphysical modality. Salmon takes
this result to follow from consideration of a case where a table (‘Woody’)
could have originated from matter m’ slightly different from the matter
m it actually originated from, but could not have originated from some
matter m” very different from its actual originating matter:4
Wherever one may choose to draw the line between what matter Woody
might have originated from and what matter Woody could not have origi-
nated from, it would seem that [. . .] we may select some [. . .] matter m” such
that, although Woody could not have originated from m”, m” is close
enough to being a possibility for Woody that if Woody had originated from
certain matter m’ that is in fact possible for Woody—matter differing in as
many molecules from the actual original matter m as possible, and sharing
as many molecules with m” as possible, while remaining a possibility for
Woody—then it would have been possible for Woody to have originated
from m”, even though it is not actually possible. [As such] the conditional
claim (which is an axiom of S4) that if Woody necessarily does not originate
from m”, then it is necessary that Woody necessarily does not so originate
fails. [. . .] S4 modal logic is fallacious. (1989, 5)
Somewhat more formally, Salmon’s argument is as follows:

1. Woody originates from matter m.
2. It is possible that Woody originates from matter m’.
3. It is not possible that Woody originates from matter m”.
4. If Woody had originated from matter m’, then it would
have been possible for Woody to originate from matter m”.
5. It is possible that it is possible that Woody originates from
matter m”. (2, 4)
6. It is not possible that Woody originates from m”, but it is
possible that it is possible that Woody originates from mat-
ter m”. (3, 5)
7. It is necessary that Woody does not originate from matter
m”, but it is not necessary that it is necessary that Woody
does not originate from matter m”. (6)
4
For simplicity we have altered the indexing on the hunks of matter at issue, and
will later do so for associated worlds.
The last claim expresses that a certain claim is necessary, but not
necessarily necessary, contra axiom (4). Salmon thus concludes that
neither S4 nor S5 is the correct logic for metaphysical modality.5
We find the data that Salmon canvasses in the Woody case to be
intuitively compelling. But do the data really establish, as Salmon
maintains, that axiom (4) should be rejected as a general constraint
on metaphysical modality?
1.3 The Woody Data, Specified

In fact, the data do not clearly establish this. As we now show, one
of the premises of Salmon’s argument is under-specified, and the
argument must be reformulated accordingly. This discussion serves
two purposes. First and most importantly, it makes explicit that
accommodating the data of the Woody case requires that metaphys-
ical possibilities and necessities be relativized to indicative actuali-
ties; as we will see, this is a claim with which Salmon arguably
agrees. Second, the discussion reveals two objections to Salmon’s
argument for the rejection of transitivity; these objections will make
room for our preferred treatment of the data, and an associated
‘S4-friendly’ relativized conception of metaphysical modality.
(Before continuing, a small caveat. We present our concern with
Salmon’s argument in terms of premise 4’s having two ‘readings’; but
by this we do not mean to imply that this premise or any of its consti-
tuting bits of language are ambiguous (perhaps the premise is not
ambiguous, but its assessment is sensitive to certain presuppositions—
namely, concerning which world is held fixed as indicatively actual).
What we are mainly concerned to do is make explicit, via the compara-
tively coarse-grained means of a difference in truth-value of the two
readings of premise 4, the role that relativization to indicatively actual
worlds plays in appropriately accommodating the Woody data. Simi-
lar remarks will apply to our reformulation of Salmon’s argument.)
To start, note that there are two readings of premise 4. The first
reading follows premises 2 and 3 in presupposing (or as we’ll put it,
5
Salmon’s discussion focuses on the normal modal logical systems S4 and S5.
However, insofar as Salmon’s target is the transitivity of the accessibility relation, as
characterized by axiom (4), his conclusion plausibly extends to simpler modal logics
such as K4, which also include (4). Thanks to Phil Kremer for discussion of this point.
‘holding fixed’) that Woody actually originates from matter m. As a

first pass, the first reading of premise 4 might be expressed as follows:
Holding fixed that Woody actually originates from matter m: if Woody had
originated from matter m’, then it would have been possible for Woody to
originate from matter m”.
The first pass is not yet sufficiently specified, however, since it fails
to express the sense in which Woody’s actual origins in matter m are
‘held fixed’, notwithstanding that in evaluating the antecedent of
the conditional, it is supposed that Woody actually originates in
some different matter. What is needed is the distinction between a
given state of affairs (or whatever) being indicatively vs. its being
counterfactually actual. To prefigure our heuristic analogy: such a
distinction is operative when we allow that, holding fixed that
‘water is H2O’ is (‘indicatively’) true in our very own actual world,
‘water is H2O’ would remain true were a world where the watery
stuff is XYZ to be (‘counterfactually’) actual. The first reading of
premise 4 should mark this distinction, as follows:
Holding fixed that Woody indicatively actually originates from matter m: if
Woody had originated from matter m’, then it would have been possible for
Woody to originate from matter m”.
The second reading of premise 4 does not follow premises 2 and 3

in holding fixed that Woody indicatively actually originates from
matter m. Rather, on this reading the premise is read as presuppos-
ing that which world is held fixed as indicatively actual is one where
Woody originates from matter m’. Making this explicit, we might
express premise 4 as follows:
Holding fixed that Woody indicatively actually originates from matter m’:
if Woody had originated from matter m’, then it would have been possible
for Woody to originate from matter m”.
These two readings of premise 4 are not equivalent, of course. On

the first reading, premise 4 is false. Here, that Woody indicatively
actually originates from matter m is held fixed; hence even if the
antecedent of the embedded conditional is counterfactually true (as
it might be, as per premise 2) the consequent of this conditional will
be false, since (as per premise 3) the fact that Woody (‘indicatively
actually’) originates from matter m places constraints on the mate-
rial Woody could have (‘counterfactually actually’) originated from

instead.6 On the second reading, however, premise 4 is (under the
relevant assumptions) true. Here, that Woody indicatively actually
originates from matter m is not held fixed; rather, in evaluating the
embedded conditional the indicatively (as opposed to merely coun-
terfactually) actual world is taken to be one where Woody origi-
nates from matter m’. Whether the conditional is true will then
depend on whether Woody’s (counterfactually actually) originat-
ing from matter m″ is possible given that Woody (indicatively actu-
ally) originates from m’. And as Salmon points out, for properly
chosen m’ and m″, the conditional will indeed be true.
How does the fact that premise 4 has distinct readings, only one
of which makes sense of the Woody data, bear on Salmon’s argu-
ment against axiom (4)? This fact indicates, at a minimum, that the
truth-values of modal claims are, somehow or other, sensitive to
which world is held fixed as indicatively actual. As such, we need
to rewrite Salmon’s argument in a way that respects this sensitivity,
which we will do by appending subscripts to the modal operators
indicating which world is held fixed as indicatively actual when the
possibility or necessity at issue is evaluated. (Again, in presenting
this reformulation, we do not mean to commit ourselves or Salmon
to any particular semantics of the modal operators; we append sub-
scripts to operators here merely to make explicit the need for rela-
tivization of modal claims to indicative actualities. Later, we’ll offer
two ways of accommodating the needed relativization within a
possible worlds semantics for modal logic.) So, for example, premise
2 should be rewritten to indicate that the possibility at issue is eval-
uated given that w1 is held fixed as indicatively actual:
2’. It is possiblew1 that Woody originates from matter m’.
And premise 4, if it is to be true, should be rewritten to reflect that,
given the background suppositions noted above, the possibility at
issue is evaluated holding w2 fixed as indicatively actual:
6
Note that the claim here is not that premise 4 is false because the consequent of
the embedded conditional is actually false; the claim is rather that, holding fixed that
Woody indicatively actually originates in matter m, the consequent would be false in
a world where the antecedent is (counterfactually) true. In other words: the con-
straints imposed by Woody’s indicatively actual origin are in force even in contexts
(e.g., w2) where Woody counterfactually originates from different matter.
4’. If Woody had originated from matter m’, then it would have
been possiblew2 for Woody to originate from matter m”.
Correspondingly, premise 5 should now reflect that the possibilities
at issue are evaluated with respect to different indicatively actual
worlds:
5’. It is possiblew1 that it is possiblew2 that Woody originates
from matter m”. (2’, 4’)
Properly specified, then, Salmon’s argument is as follows:
1’. Woody originates from matter m.
2’. It is possiblew1 that Woody originates from matter m’.
3’. It is not possiblew1 that Woody originates from matter m”.
4’. If Woody had originated from matter m’, then it would have
been possiblew2 for Woody to originate from matter m”.
5’. It is possiblew1 that it is possiblew2 that Woody originates
from matter m”. (2’, 4’)
6’. It is not possiblew1 that Woody originates from matter m”,
but it is possiblew1 that it is possiblew2 that Woody origi-
nates from matter m”. (3’, 5’)
7’. It is necessaryw1 that Woody does not originate from mat-
ter m”, but it is not necessaryw1 that it is necessaryw2 that
Woody does not originate from matter m”. (6’)
Note that the last claim is no longer a clear counter-instance to axiom
(4). As per the standard assumption that metaphysical modality is
modality in the broadest (non-syntactically logical) sense, previous
discussions of the axiom have not incorporated the need for relativi-
zation to an indicatively actual world. How the axiom should be
understood in light of the need for relativization—and in particular,
whether it should be understood to apply to modal claims involving
iterated or (as we’ll put it) ‘in situ’ shifts in which world is held fixed
as indicatively actual (of the sort occurring in 5’–7’) remains to be
seen. This result constitutes our first objection to Salmon’s argument
that the Woody case motivates the rejection of transitive accessibility:
if we appropriately attend to which worlds are held fixed as indica-
tively actual in this case, no clear violation of axiom (4) results.
More generally, the need to incorporate facts about which world is
held fixed as indicatively actual in order to appropriately express the
data indicates that the standard conception of metaphysical modality

is strictly incorrect. Hence though we disagree with Salmon’s diag-
nosis and treatment of the Woody case, we agree with him that the
data here motivates a revision of the standard conception in the direc-
tion of relativization. (The need for relativization is an underappreci-
ated insight of Salmon’s discussion, which has been, we speculate,
obscured by his rejection of transitivity and endorsement of so-called
‘impossible worlds’.) More specifically, a proper understanding of
the data concerning Woody indicates that metaphysical modal rea-
soning and any associated modal logics must be able, first, to make
room for different worlds to be indicatively actual; and second, to
keep track of which world is held fixed as indicatively actual.
That said—and here we raise our second objection to Salmon’s
argument against transitive accessibility—we do not think that the
best way to implement the needed relativization is to make sense,
one way or another, of metaphysical modal reasoning that involves
in situ shifts in which world is held fixed as indicatively actual, of
the sort which explicitly occurs in the disambiguated premise 5’ of
Salmon’s argument. On the contrary, we are inclined to see some-
thing defective in such claims. To again prefigure, compare (what we
will later argue is) the formally analogous epistemic interpretation
of the two-dimensional (2D) semantic framework, endorsed by,
e.g., Jackson and Chalmers (2001). Epistemic two-dimensionalism
distinguishes between ‘considering as actual’ and ‘considering as
counterfactual’; it makes room for our being able to ‘consider as
actual’ either a world where water is H2O, or a world where water
is XYZ, and to go on to ‘consider as counterfactual’ other worlds
against the assumption that one or other world has been considered
as actual. Yet within this framework there is little motivation to
accommodate the following sort of claim:
Considering as actual a world where water is H2O: considering as actual a
world where water is XYZ, then necessarily, water is XYZ.
We similarly do not see any motivation for a revision of metaphysi-

cal modal logic or associated semantics on which claims such as
Salmon’s disambiguated premise 5’ are accommodated (indirectly,
on Salmon’s view, by relaxing constraints on accessibility, or directly,
by revising rules of modal logical inference so as to explicitly incor-
porate, e.g., double indexing). As we see it, such claims illegiti-
mately shift indicatively actual horses in modal mid-stream.
Our view is rather as follows: metaphysical modal claims and asso-

ciated reasoning need to be appropriately sensitive to which world is
held fixed as indicatively actual, primarily in order to avoid such ille-
gitimate in situ shifts in which world is held fixed as indicatively
actual. From this perspective, Salmon’s argument against transitive
accessibility is problematic not just in that no clear counter-instance to
axiom (4) follows from its (properly specified) premises, but also in
that the argument, once specified, relies on the ill-formed premise 5’.
In response to our objections, Salmon might maintain that, though
no counter-instance of axiom (4) is explicitly entailed by the specified
data, the axiom’s rejection, and associated acceptance of claims
involving in situ shifts in which world is held fixed as indicatively
actual, are required to accommodate the data concerning Woody.
As we’ll argue in S2, such maneuvers are unnecessary. First, though,
we present Salmon’s treatment of the data; this will serve to high-
light how our relativized conception of metaphysical modality dif-
fers from Salmon’s, and to flag certain general concerns with his
treatment which our preferred conception avoids.
1.4 Salmon’s Intransitive Relativized Conception

Salmon understands possible worlds as maximal abstract ways for
goings-on to be, and endorses ‘the standard identification of neces-
sity with truth in every possible world and possibility as truth in at
least one possible world’ (1989, 5). He maintains that these commit-
ments are compatible with the data concerning Woody, if one
accepts impossible worlds—‘total ways things cannot be’—and
allows that a world that is impossible ‘relative’ to one world may be
possible relative to another. So, for example, a world where Woody
originates in matter m” is such a world. That world is not possible
relative to the actual world, but it is possible relative to a world in
which Woody originates from matter m’. Relative to the actual
world, it is merely possibly possible. Salmon suggests that ‘other
impossible worlds may not be even possibly possible, but only pos-
sibly possibly possible, and so on; hence the binary relation between
(possible or impossible) worlds of relative possibility—the modal
relation of accessibility—is not transitive’. (1989, 7).
Beyond the relativization of what is possible and necessary to
which world ‘obtains’, and the rejection of transitive accessibility,
Salmon notes two related ways in which his treatment departs from
the standard understanding of metaphysical modality. First, ‘[i]f
worlds include ways things metaphysically cannot be in addition
to ways things metaphysically might have been, then the idea
that metaphysical necessity corresponds to truth in every world
whatsoever is flatly mistaken’ (1989, 15). Second, given that ‘[a] pos-
sible world is a total way for things to be that conforms to metaphysi-
cal constraints concerning what might have been [. . .] metaphysical
modality is definitely not an unrestricted limiting case’ (1989, 12–13).
The conception of metaphysical modal space that is in the first
instance suggested by Salmon’s treatment of the Woody case is of a
single space of (‘maximal’, abstract) worlds, whose status as possi-
ble or impossible is relative to whatever world is supposed to
obtain,7 and which are connected by an intransitive accessibility
relation. In pictorial terms: the standard conception of metaphysical
modal space—the conception that Salmon, and we, reject—has the
following structure:
w2
w1 w3
Figure 1. The Standard Conception: Unrelativized Transitive

Accessibility.
(Here, and in the figures to follow, arrows point towards worlds

accessible to the origin world.) And the conception of metaphysical
modal space suggested by Salmon’s treatment of the Woody case is
as follows:
7
We uniformly interpret Salmon’s talk of metaphysical possibilities and necessi-
ties as relative to which world ‘obtains’ (in our terms: is indicatively actual), such
that, e.g., w3 is possible relative to w2 when the latter obtains in just the same way that
our very own actual world obtains. Though Salmon’s emphasis on impossible worlds
might be thought to suggest that he sees a substantive difference between whatever
world in fact obtains and other worlds (e.g., w2) that merely hypothetically obtain,
Salmon seems to reject such a privileging of our world when disparaging what he
calls the ‘ostrich approach to modality, with its consequent misconstrual of “necessar-
ily” as meaning actual necessity and “possibly” as meaning actual possibility”’ (1989,
29). In any case, a restricted understanding would still have the structure to follow.
w1 w2 w3
Figure 2. Salmon’s Intransitive Relativized Conception.

Relative to w1, w2 is possible, and w3 is not possible; relative to w2,
w3 is possible.
(Here, and in the figures to follow, dotted lines around a world indicate
that the world is held fixed as indicatively actual. Solid-line ovals around
worlds represent the relativization of accessibility due to shifts in which
world is held fixed as indicatively actual.8) Again, we emphasize that in
our view, Salmon’s conception is on the right track, in recognizing the
need for metaphysical modal deliberation to be sensitive to which world
is supposed to ‘obtain’ (be indicatively actual). Still, Salmon’s approach
to a relativized conception is revisionary, in departing from the usual
assumption of transitive accessibility and associated systems of modal
logic; and many have found the posit of metaphysically impossible
worlds problematic (see Lewis 1986, 7, fn. 3 and 246–8).
2 THE TRANSITIVE RELATIVIZED CONCEPTION

(RELATIVIZED METAPHYSICAL MODALITY)
Our approach to a relativized conception accommodates the Woody

data, compatible with both transitive accessibility and the rejection
of impossible worlds. Schematically, our conception is one natu-
rally seen as involving not a single space of mutually accessible
worlds (as on the standard conception), nor a single space of
intransitively accessible worlds (as on Salmon’s conception), but
rather multiple spaces, each containing one indicatively actual
world, along with whichever worlds are (one might reasonably
8
Our purpose here is merely to sufficiently distinguish Salmon’s intransitive rela-
tivized conception from the standard, transitive conception of metaphysical accessi-
bility (see Figure 1). Plausibly, on Salmon’s intransitive conception, had w2 obtained
(in our terminology, been ‘indicatively actual’) then w1 as well as w3 would have
been accessible/metaphysically possible from w2. In this sense, Figure 2 as it stands
represents the relations of accessibility that obtain between worlds w1, w2, and w3 in
a fairly abbreviated form.
qualify: counterfactually) metaphysically possible relative to that

indicatively actual world. In fact, there are two ways of implement-
ing our conception, reflecting two ways of locating the relativiza-
tion at issue. In presenting these versions and their application to
the Woody case, we will speak freely (as we have been doing all
along) of possible worlds as located in spaces; depending on one’s
further commitments, one may or may not understand such talk as
metaphorical. Either way, the structural distinctions we aim to
identify will be clear enough to get a feel for our proposal(s), and—
most importantly—to see that endorsement of either version of the
relativized conception is compatible with taking the correct logic
of metaphysical modal reasoning to conform to S4 or S5.
2.1 The Overlapping Spaces Interpretation: (Some) Truth Relativized

to an Indicatively Actual World
On the first implementation, the spaces of worlds associated with
different indicatively actual worlds overlap. Here it is assumed that,
prior to identification of any world as indicatively actual, there is a
single space of (what we will call) ‘basically individuated’ possible
worlds, connected, we assume, by a transitive accessibility relation.
The principle of basic individuation of worlds might be primitive; or
it might proceed by way of, e.g., ‘semantically stable’ (Bealer 2000)
or ‘canonical’ (Chalmers 2006) descriptions, or some combinatorial
function of basic elements (as per Lewis 1986 or Armstrong 1989).
While the individuating principle suffices to distinguish worlds,
such individuation, we assume, leaves open the truth values of vari-
ous claims at a given world. For example, in the pre-relativized
space there will be a world, w3, containing a table-shaped hunk of
matter m”. In w3, is it true or false that Woody originates from m”?
In the pre-relativized space, there is no answer to this question, nor
to many other questions, whose answers depend on which world is
(held fixed as: henceforth we take this qualification as understood)
indicatively actual.9 Pictorially, the pre-relativized space of basically
individuated worlds is structured according to the standard concep-
tion of metaphysical modal space discussed above:
9
Hence, on this implementation, worlds are not alethically ‘maximal’ prior to rela-
tivization; accordingly, in the pre-relativized space, v(p) (where p is the proposition
that Woody originates in matter m”) is indeterminate.
w2
w1 w3
Figure 3. A pre-relativized space of basically individuated worlds.

w1 contains a table-shaped hunk of matter m, w2 contains a table-shaped
hunk of matter m’, w3 contains a table-shaped hunk of matter m”
For each world in the pre-relativized space, there is an associated

relativized space, containing the same worlds as in the pre-rela-
tivized space, but with the truth values of at least some previ-
ously undetermined truths—namely, those dependent on which
world is indicatively actual—now determined. In other words, on
the present interpretation (some) truth is relative to which world
is indicatively actual. Implementing this interpretation requires
that the valuation function v assigning truth values to proposi-
tions be sensitive to which world is indicatively actual, for both
non-modal and modal clauses. We might do this, for a language,
by incorporating such a reference in an additional argument place
in these clauses (as we do below: see Figures 3.1 and 3.2), or we
might keep the usual semantic clauses, and distinguish valuation
functions for each indicatively actual world. Now, when the
worlds in the pre-relativized space are relativized to w1, then the
proposition that Woody originates from m” is thereby rendered
false in w3:
w2
w1 w3
Figure 3.1 The space of basically individuated worlds, relativized to

w1 (w1 is indicatively actual).
p = the proposition that woody originates from matter m”
v(p, w3, w1) = F
When the worlds in the pre-relativized space are relativized to w2,

however, the proposition that Woody originates from m” is thereby
rendered true in w3:
w2
w1 w3
Figure 3.2. The space of basically individuated worlds, relativized to

w2 (w2 is indicatively actual).
v(p, w3, w2) = T
Crucially, the need for such relativization is no barrier to the stand-

ard assumption that modal reasoning proceeds as per S4 or S5; we
may continue to assume, as per usual, that the worlds in any post-
relativized space are mutually (totally) accessible. In particular, as
we’ll now show, on our understanding of the data of the Woody
case, no violation of axiom (4) is forthcoming.
Again, we start with a pre-relativized space of possible worlds. Let’s
start by relativizing to w1—that is, holding w1 fixed as indicatively
actual; we want to consider the relevant semantic clauses concerning
the proposition p, that Woody originates from matter m”. Given that
Woody actually originates from matter m in w1, p will be false in w3:
v(p, w3, w1) = F
(Read: the semantic value of p in w3 relative to w1’s being indica-
tively actual is F.) More generally, given that w1 is indicatively actual,
v will assign F to p in every world in the relativized space. Hence v
will assign F to p in every world accessible to w1; hence v will assign
F to the proposition that p is possible in w1; and similarly for w2:
v(◊p, w1, w1) = F
v(◊p, w2, w1) = F
For the same reason, given that w1 is indicatively actual, v will
assign F to the proposition that p is possible in every world accessible
to w1. Consequently, v will assign F to the proposition that p is pos-

sibly possible in w1:
v(◊◊p, w1, w1) = F
When w1 is indicatively actual, no violation of axiom (4) is in the
offing. And similarly if we relativize to w2 (hold w2 fixed as indica-
tively actual).10
We can now be more explicit about how our characterization of
what it is for a proposition p to be possible at a world contrasts with
Salmon’s. Salmon and we agree that propositions such as ◊p, for p as
above, are (to speak in purposely rough terms) only relatively true.
Salmon takes this to mean that ‘◊p’ is true in some worlds but not
others. This suggests, on the standard semantics, that some worlds
can access a p world and others cannot; it follows that not all worlds
can access the same worlds, contra S5. The rejection of transitive
accessibility is thus built into Salmon’s understanding of the rough
thought that some modal claims are only relatively true. We, on the
other hand, do not take the rough thought to mean that ‘◊p’ is true in
some worlds and not in others. Rather, we take it to mean that, though
‘◊p’ is either true in all worlds or true in none, which truth value is
assigned to ‘◊p’ in all worlds is relative to which world is indica-
tively actual. (Similarly for the implementation presented in S2.2.)
Relative to some indicatively actual worlds, ‘◊p’ is true in all worlds;
relative to other indicatively actual worlds, ‘◊p’ is false in all worlds.
Either way, transitive accessibility is maintained.
Two questions remain. The first concerns what justifies thinking
of worlds in differently relativized spaces as ‘the same’, such
that, e.g., we may speak of w3 as existing both in a space relativized
10
In this case it will rather be true at w3 that Woody originates from matter m”:
v(p, w3, w2) = T
Furthermore, given that w2 is indicatively actual (together with the fact that, in w2,
Woody originates in matter m'), v will assign T to the proposition that p is possible in
both w1 and w2:
v(◊p, w1, w2) = T
v(◊p, w2, w2) = T
Moreover, since w2 is accessible to w1, v will assign T to the proposition that p is pos-
sibly possible in w1:
v(◊◊p, w1, w2) = T
Again, no violation of transitivity results.
to w1 and in a space relativized to w2. A fairly straightforward

answer is that worlds in different relativized spaces may be consid-
ered to be ‘the same’ notwithstanding that they differ on the truth
values of certain claims (e.g., that Woody originates in matter m”) in
virtue of being basically the same—that is, strictly the same at the
level of basic individuation, understood in one or other of the prim-
itive or non-primitive ways mentioned earlier.
The second question concerns what justifies thinking of individu-
als in differently relativized spaces as in some sense ‘the same’, so
that, e.g., we may speak of Woody as existing both in a space relativ-
ized to w1 and in a space relativized to w2. The answer here is less
straightforward, and will depend on further details concerning the
metaphysics and individuation of material objects. One way to go
here, putting things in the formal mode, would be to suppose that
expressions denoting material objects (e.g., ‘Woody’) are relevantly
like expressions for natural kinds (e.g., ‘water’), in having something
like a descriptive or reference-fixing sense, whose association with a
referent is dependent upon which world is indicatively actual, and
once fixed, is metaphysically necessary (allowing for some counter-
factual flexibility). Another way to go, putting things in the material
mode, would be to suppose that material objects such as Woody
have a sort of ‘relativized essence’, such that, as a primitive or non-
primitive matter, Woody exists and has certain metaphysically nec-
essary features at some relativized worlds, but either fails to exist or
has different metaphysically necessary features at others.11
2.2 The Non-overlapping Subspaces Interpretation: Domain Relativized

to an Indicatively Actual World
On the second implementation of our relativized conception, the
spaces of worlds associated with different indicatively actual worlds
do not overlap. Rather, the single space of worlds is partitioned into
non-overlapping subspaces of finely individuated worlds, each of
11
What if some claims concerning Woody—e.g., that Woody is a table—are neces-
sary relative to each indicatively actual world? We may define, if we like, a notion of
‘absolute’ metaphysical necessity. But given the desirability of avoiding shifts in
indicatively actual worlds, we should see the absolute notion as grounded in the rela-
tivized notion—as tracking a uniform pattern of variation in what is relatively meta-
physically necessary—as opposed to taking ‘absolute’ metaphysical necessity to be
either prior to or distinct from relativized metaphysical necessity.
which is, unlike the basically individuated worlds of the overlap-

ping subspaces implementation, alethically ‘complete’; and each
subspace corresponds to a space of worlds relativized to a single
indicatively actual world (also contained in the subspace). Again,
the principle of individuation here may be primitive or otherwise.
Within a subspace, worlds are (we may assume) mutually accessi-
ble; but worlds are not accessible across subspaces.12
On this implementation, it is domains that are relativized to indic-
atively actual worlds. Here, the relativization happens at the level
of the frame—that is, in the selection of one subspace of worlds
from among the many subspaces. On the standard conception, a
frame contains a single set of worlds W. On our conception, the
frame contains not a set, but a partition, of worlds:
w2 w2⬘ w2⬘⬘
w1 w3 w1⬘ w3 ⬘ w1⬘⬘ w3 ⬘⬘
S1 S2 S3
Figure 4. Non-Overlapping Subspaces. In subspace S1, w1 is indica-
tively actual; in subspace S2, w2’ is indicatively actual; in subspaces
S3, w3” is indicatively actual. Each ‘w1-type’ world contains a table-
shaped hunk of matter m; each ‘w2-type’ world contains a table-
shaped hunk of matter m’; each ‘w3-type’ world contains a
table-shaped hunk of matter m”.
v(p, w3, w1) = F v(p, w3’, w2’) = T
12
Even though worlds in different subspaces are not identical on this view, worlds
in different partitions may be taken to be ‘basic’ or ‘canonical’ counterparts of each
other, in that (lifting some of the structure of the overlapping spaces interpretation)
worlds in different spaces may be basically alike; such similarity may serve as the
basis for loose (as opposed to strict) identification of worlds across subspaces. So, for
example, distinct subspaces might each contain worlds that are basically, canonically,
or qualitatively similar, in containing, e.g., a table-shaped hunk of matter m”; such
worlds, we might say, are of type w3. In a world of type w3, is it true or false that
Woody originates from m”? That depends. In a subspace where the indicatively
actual world is (of type) w1, this is false; but in a subspace where the indicatively
actual world is (of type) w2, this is true.
Upon relativization, one of the subspaces in the partition is selected

as a basis for subsequent modal and semantic deliberation. So, for
example, one might select the subspace in which w1 is indicatively
actual; or one might select the subspace in which w2 is indicatively
actual. Having so chosen, all the usual logical axioms and semantic
clauses can remain exactly as per the standard conception. In par-
ticular, the valuations of iterated and non-iterated modal clauses
involving the proposition p in a given post-relativized space will, as
in the ‘overlapping spaces’ interpretation, conform to the assump-
tion of transitive accessibility.13
As such, the need for relativization is clearly no barrier to the
standard assumption that modal reasoning proceeds as per S4 or
S5; we may continue to assume that the worlds in any given sub-
space are mutually accessible. Indeed, given that the accessibility
relation on S5 is an equivalence relation, it is well suited to charac-
terize, not just the mutual accessibility of worlds in a single space,
but also the relativized mutual accessibility of worlds partitioned
by such an equivalence relation into subspaces.
Summing up: a conception of metaphysical possibilities and
necessities as relative to indicative actualities can accommodate the
data at issue in the Woody case without incurring the revisionary
costs associated with Salmon’s treatment of the data.
13
Indeed, given the satisfaction of transitive accessibility within a particular sub-
space (either overlapping or non-overlapping), the evaluation of claims of necessity
at a world w in a subspace in which some distinct world w' is indicatively actual is
straightforward (thanks to Phil Kremer for pressing us on this point). Given transitiv-
ity, if some proposition A is necessary at a world w, A is necessarily necessary at w,
and hence necessary at every world in the subspace accessible from w; thus, in par-
ticular we have:
v(□A,w,w') = T iff v(□A, w',w') = T
(A is necessary at w, given that w' is indicatively actual, if and only if A is necessary
at the indicatively actual world w' of the subspace). Necessary truth at the indica-
tively actual world of a subspace can then be given its familiar analysis in terms of
truth at all accessible worlds:
v(□A, w',w') = T iff ∀w[wRw' → v(A, w) = T]
Combining these results yields the truth-clause for necessary propositions evaluated
at worlds accessible from the indicatively actual world of a particular subspace:
v(□A,w,w') = T iff ∀w[wRw' → v(A, w) = T]
2.3 Heuristically Situating the Transitive Relativized Conception: A

Properly Metaphysical Interpretation of the 2-D Semantic Framework
Our proposal is not as unusual as it might appear, for it may be
naturally situated in a thoroughly metaphysical interpretation of
the sort of 2-D semantic framework developed by Kaplan (1979,
1989), Stalnaker (1978), and others. Although our interest here is
metaphysics, not meaning, we take the formal analogy between our
relativized conception of metaphysical modality and (a metaphysi-
cal interpretation of ) the 2-D framework to be heuristically useful.
The so-called ‘epistemic’ interpretation of the 2D semantic frame-
work (see, e.g., Chalmers and Jackson 2001, Chalmers 2006), com-
monly applied to natural kind terms and associated Kripkean
necessities (e.g., ‘water’, ‘water is H2O’), appeals to a distinction
between possible worlds, as either considered as actual or considered
as counterfactual. One starts by constructing a 2-D matrix listing
worlds potentially considered as actual in the far-left column, and
worlds potentially considered as counterfactual along the top row;
conventionally, the first world in each list is our very own actual
world, and the expression whose meaning is at issue appears in the
top left-hand corner. The basic suggestion, then, is that (at least
some especially salient) aspects of meaning may be represented by
intensions, understood as functions from worlds to extensions.
More specifically, the suggestion is that aspects of meaning may be
represented by functions taking as arguments two worlds (hence
‘2-D’)—one considered as actual (drawn from the leftmost column),
one considered as counterfactual (drawn from the topmost row);
different aspects of meaning are then associated with different 2-D
intensions.
On the epistemic interpretation, one salient aspect of meaning
corresponds to metaphysical reference. This aspect is associated with
the function which takes our very own world as its first argument,
and a world considered as counterfactual as its second argument;
this function is sometimes called the ‘secondary’ or ‘horizontal’
intension (we’ll use the latter terminology, as visually more evoca-
tive), and for relevant expressions, is understood to encode (or as
we’ll loosely say, ‘represent’) what is metaphysically necessary. Con-
sider a portion of the 2-D matrix associated with ‘water is H20’,
where an H2O-world is one where the predominant liquid falling
from the sky, found in lakes, etc. (the ‘watery stuff’, for short) is
H2O, and an XYZ-world is one where the watery stuff is XYZ; and
where it is furthermore assumed that our very own actual world is
an H2O-world:
‘water is H2O’ H2O-world XYZ-world

H2O-world T T
The horizontal intension for the expression ‘water is H2O’ here

reflects, among other relevant facts, that the horizontal intension of
‘water’ is sensitive to the way the world actually turns out to be, so
that, given that the actual world is an H2O-world, ‘water is H2O’
will be true in every world considered as counterfactual. As such,
the horizontal intension of ‘water is H2O’ returns ‘T’ for every world
considered as counterfactual (if an XYZ-world were to be counter-
factually actual, then ‘water is H2O’ would have been true, etc.),
consonant with Kripke’s claim that ‘water is H2O’ is metaphysically
necessary.
Other modally implicated intensions may be defined within this
framework. Chalmers, Jackson and others maintain that terms such
as ‘water’ have an aspect of meaning corresponding to epistemic
sense, an aspect of meaning supposed to be a priori accessible, in
being independent of details about which world is considered as
actual, and which is posited as explaining, e.g., intuitions that it
might have turned out that water was not H2O. The associated func-
tion is sometimes called the ‘primary’ or ‘diagonal’ intension (again,
we’ll use the latter terminology), and takes as arguments pairs of
identical worlds <w, w>. Consider a portion of the 2-D matrix asso-
ciated with ‘water is the watery stuff’:
‘water is the watery stuff’ H2O-world XYZ-world

H2O-world T
XYZ-world T
We want to call attention to a third, underappreciated class of inten-

sions associated with the 2-D framework, that is required if the
matrix is to be appropriately ‘filled in’. Note that the diagonal inten-
sion, but not the horizontal intension, takes as input worlds other
than our very own actual world ‘considered as actual’. We may also
define generalized horizontal intensions, where a generalized horizon-
tal intension is a function which takes a world considered as actual
as its first argument, and a world considered as counterfactual as its
second argument; we call such secondary intensions ‘generalized’
in that the world considered as actual need not be our very own
actual world. So, for example, consider the generalized horizontal
intension associated with ‘water is XYZ’, when an XYZ-world is
considered as actual:
‘water is XYZ’ H2O-world XYZ-world

H2O-world
XYZ-world T T
On the usual epistemic interpretation, such a generalized horizon-

tal intension is taken to represent a merely epistemic necessity: fol-
lowing the usual interpretation of Kripke’s results, only the
non-generalized horizontal intension, taking as its first argument
our very own actual world, is capable of representing what is genu-
inely metaphysically necessary (or possible).
But as we see it, there is good reason to interpret the necessities
represented by generalized horizontal intensions as genuine. After
all, the actual world might have turned out to be an XYZ-world, in
which case the horizontal intension associated with ‘water is XYZ’
would have represented a genuine metaphysical necessity. Rather
than obscure this fact by treating the represented necessity as
epistemic, why not treat all horizontal intensions on a par as repre-
senting genuine relativized metaphysical necessities—that is, as repre-
senting (for relevant expressions) what is metaphysically necessary
relative to a given indicatively actual world? Under a properly meta-
physical interpretation of the generalized horizontal intensions, the
2-D framework is well suited for such representation; and more
generally is structurally analogous to the relativized conception, in
encoding what is counterfactually the case, relative to each indica-
tively actual world.
Let’s make the structural analogy explicit. To accommodate the
data concerning Woody, we represent two ways in which the coun-
terfactual possibilities for Woody might depend on Woody’s actual
origin, one of which holds fixed that Woody indicatively actually

originates from matter m, and the other of which holds fixed that
Woody indicatively actually originates from matter m’. We do this
by letting entries in the leftmost column represent which world is
held fixed as indicatively actual; entries along each row then repre-
sent what is the case, as regards Woody’s origin, in worlds that are
counterfactual relative to the associated indicatively actual world
(and where an ‘m-world’ is a world where the salient candidate for
being Woody originates in matter m, etc.).
‘Woody does not m-world m'-world m”-world

originate from m”’
m-world T T T
m'-world T T F
As desired, the structure allows us to represent the dependence of

what is metaphysically possible and necessary concerning Woody
on which world is indicatively actual.
At this point we want to revisit our earlier (S1.3) observation that
attention to the 2-D framework supports thinking that in situ shifts
in which world is held fixed as indicatively actual, of the sort asso-
ciated with Salmon’s disambiguated 5’, are in some sense defective
or ill-formed. As above, one of the salient intensions associated with
the 2-D framework is the diagonal intension, which for each world
w returns the extension of the relevant expression at w when w is
considered as actual, and which on the epistemic interpretation
is taken to represent epistemic sense—an aspect of meaning that is
supposed to be a priori accessible, in being independent of details
about which world is considered as actual, and which allows for
representation of certain modal truths, such as ‘necessarily, water is
the watery stuff’. As an extension of our ‘metaphysical’ interpreta-
tion of generalized horizontal intensions, constant diagonal intensions
might be taken to represent genuine facts—namely, those inde-
pendent of which world is indicatively actual. Such independence,
in turn, might be understood in terms of a notion of ‘absolute’ meta-
physical modality, tracking patterns in what is the case, either non-
modally or relatively modally, when different worlds are held fixed
as indicatively actual. (See note 11.)
Does the fact that (as attention to the diagonal and generalized
horizontal intensions suggests) we can make sense of claims involv-
ing shifts in indicatively actual worlds pose a problem for our think-
ing that some such shifts are illegitimate, from the perspective of
metaphysical modal reasoning? No. Certain claims and associated
reasoning involving shifts in which world is (considered as) indica-
tively actual are, on our view, perfectly legitimate—namely, those
which are appropriately seen as ‘meta-modal’, in tracking patterns
of what is the case, non-modally or relatively modally, when differ-
ent worlds are held fixed as indicatively actual. We see nothing
defective in claims like the following, tracking patterns in what is
non-modally the case across different worlds considered (held
fixed) as indicatively actual:
Considering as (indicatively) actual a world where water is H2O: water is
the watery stuff; and considering as (indicatively) actual a world where
water is XYZ: water is the watery stuff.
Nor do we see anything defective with claims like the following,

tracking patterns in what is modally the case across different worlds
considered (held fixed) as indicatively actual:
Considering as (indicatively) actual a world where water is H2O: it is neces-
sary that water is H2O; and considering as (indicatively) actual a world
where water is XYZ: it is necessary that water is XYZ.
Again, the sort of claims that we maintain are in some sense defec-
tive, from the point of view of metaphysical modal reasoning, are
those involving iterated or ‘in situ’ shifts in which world is indica-
tively actual, of the sort characteristic of Salmon’s disambiguated
premise 5’, or of the following sort of claim:
Considering as (indicatively) actual a world where water is H2O: consider-
ing as (indicatively) actual a world where water is XYZ, then it is necessary
that water is XYZ.
Such in situ shifts are not motivated by diagonal intensions, or the

associated generalized horizontal intensions, whether these are
understood to involve merely epistemic or properly genuine
necessities.
Relatedly, we have no deep complaint against claims, e.g., to the
effect that some world w* different from our very own actual world
‘could have been’ indicatively actual—so long as such claims are

interpreted as meta-modal claims, whose evaluation requires look-
ing at the space of worlds or subspaces ‘from the outside’, as it
were.
3 FINE’S ‘SCHMASS’ CASE
3.1 Fine’s Rejection of Modal Monism

In ‘The Varieties of Necessity’ (2002), Fine notes that there appear to
be different ways in which a claim might be said to be necessary or
possible, reflecting, e.g., logical, conceptual, mathematical, meta-
physical, nomological, or normative necessity or possibility; he then
considers whether any of these can be defined in terms of the oth-
ers, and if so, which are most basic. Fine characterizes metaphysical
necessities as necessities which hold in virtue of the natures and
identities of the entities at issue, and takes it to be plausible that
logical, conceptual, and mathematical necessity may be defined in
terms of metaphysical necessity, with the former varieties of neces-
sity being defined as restrictions on the latter. So, for example, the
logically necessary claims are those that are, first, metaphysically
necessary and second, true in virtue of the nature of logic. Fine does
not think, however, that metaphysical necessity is the only basic
variety, but rather argues that nomological necessity is also basic, in
not being appropriately seen as a restricted form of metaphysical
necessity.
The focus of Fine’s discussion is the view, typically endorsed by
those (Shoemaker 1980, Ellis 2001, and Bird 2007) taking powers or
laws to be essential to properties, according to which nomological
necessities are metaphysical necessities (as Shoemaker puts it, are
‘necessary in the strongest sense’), based in the nature or identity of
laws of nature or natural kinds. While Fine is inclined to agree that
some nomological necessities (e.g., that electrons are negatively
charged) are metaphysically necessary, certain other nomological
necessities, he claims, are such that their denials are metaphysically
possible.
Suppose, for example, that it is a law of nature that massy entities
attract according to an inverse square law. As Fine notes, the neces-
sitarian may plausibly maintain that it is metaphysically necessary

that massy entities so attract, reflecting the nature or identity of the
property of being massy. Still, Fine continues, there is a nomological
necessity in the vicinity that is not metaphysically necessary, to
which necessitarians appear to be committed.
He reasons as follows. Among the necessitarian’s burdens is to
explain away intuitions that massy entities might have entered into
different laws—say, an inverse cube rather than an inverse square
law. One might maintain that the intuition is merely epistemic—the
content is not really ‘imaginable’—and there is no genuine possibil-
ity corresponding to the intuition of contingency at issue; but given
the important role intuitions play in supporting modal claims there
is a case to be made that this line is unprincipled. Hence it is that
Kripke prefers to treat intuitions of the contingency of certain iden-
tities as tracking possibilities that are genuine but misdescribed.
Necessitarians about laws also typically implement such a rede-
scription strategy, as in Shoemaker’s treatment of intuitions that
strychnine might not be fatal to humans:
Let the law be that strychnine in a certain dosage is fatal to human beings.
We can grant it to be imaginable that ingesting vast amounts of what passes
certain tests for being strychnine should fail to be fatal to what passes cer-
tain tests for being a human being, but deny that this amounts to imagining
a human being surviving the ingestion of that much strychnine. (1998, 62)
Applying the redescription strategy to the case at hand would allow

the necessitarian to maintain that massy entities necessarily attract
according to an inverse square law; but on the other hand such an
implementation is puzzling, in seeming to undermine the necessi-
tarian’s core claim that nomological necessities are metaphysically
necessary. As Fine notes, the strategy requires commitment to there
being some property—call it ‘schmass’—which enters into the
redescription of the purported counterexample to the nomological
necessity of the inverse square law. Given that schmass enters into
the redescribed scenario in this way, however, it follows that a
world containing schmass is metaphysically possible. Furthermore,
Fine surmises, the proposition that there is no schmass is nomologi-
cally necessary, given that mass exists in the actual world and the
existence of schmass is nomologically incompatible with it. In that
case, there appear to be some nomological necessities—e.g., ‘There
is no schmass’—that are not metaphysically necessary. Hence, Fine

continues, nomological necessity cannot be seen as a species or
restricted form of metaphysical necessity, contra the usual necessi-
tarian line.
To be sure, the necessitarian has certain options for response:
they may maintain (as Shoemaker does) that intuitions of the meta-
physical possibility of schmass are mistaken; or they may deny that
the non-existence of schmass is nomologically necessary, on grounds
that the incompatibility of mass and schmass is due, e.g., to nomo-
logically contingent initial conditions. Such responses do not seem
fully principled, however. Given that many intuitions of contin-
gency admit of genuine redescription, why not the one concerning
schmass? And given that many claims are nomologically necessary
as a matter of nature, why not the one concerning the non-existence
of schmass?
Indeed, attention to redescriptive strategizing isn’t necessary to
see that it is problematic to suppose that nomological necessity is a
restricted form of metaphysical necessity, when the latter is charac-
terized, as per usual, as involving a single space of mutually acces-
sible possible worlds. After all, necessitarians are typically not
modally nomocentric; as Shoemaker says, ‘Nothing I have said
precludes the possibility of there being worlds in which the causal
laws are different from those that prevail in this world.’ (1980, 248).
Such worlds must involve completely alien properties, but no mat-
ter—such alien worlds can serve as witness to the general claim
that some nomological necessities are not necessary tout court. But
how are such alien possibilities not precluded, one wonders, if
‘nomological necessity is necessity in the strongest sense’? How-
ever one interprets the data concerning ‘schmass’, it remains to
make sense of why necessitarians like Shoemaker, on the one hand,
subsume nomological under metaphysical necessity; yet on the
other, allow that some nomological necessities are not metaphysi-
cally necessary.
Fine interprets the data concerning ‘schmass’ along the follow-
ing lines. Insofar as some nomological necessities (‘There is no
schmass’) are not metaphysically necessary, it follows that nomo-
logical necessity is not a restricted form of, and more generally
cannot be defined in terms of, metaphysical necessity. Meanwhile
(though this step is implicit in his discussion) other forms of neces-
sity (e.g., conceptual necessity) to which nomological necessity

might be reduced are plausibly taken to be restricted forms of
metaphysical necessity, so that nomological necessity cannot be
defined in terms of these other forms of necessity either. Putting
the previous results together, Fine concludes that nomological
necessity is a basic form of necessity—as basic as metaphysical
necessity.
We have three concerns with Fine’s proposal. First, the suppo-
sition that nomological modality is fundamentally distinct from
metaphysical modality fails to sync with the fact that, as Fine
grants, it is natural to see many nomological necessities (e.g., that
massy entities attract as per an inverse square law) as grounded
in the natures and identities of the entities at issue, and hence as
metaphysically necessary.14 Second, if nomological necessities
aren’t grounded in the natures or identities of the entities at issue,
then what are such necessities grounded in? In virtue of what are
they true? As it stands, Fine’s proposal to take nomological
modality as basic is unilluminating. Third, Fine’s proposal fails to
illuminate how it could be—as necessitarians typically allow—
both that nomological necessities are metaphysically necessary
and that some nomological impossibilities are metaphysically
possible. To be sure, as above, necessitarians haven’t explained
how this could be, either. But a more satisfying treatment of the
data concerning ‘schmass’ would show how, when properly
understood, the necessitarian’s seemingly contradictory claims
might jointly make sense.
3.2 An Alternative Treatment of the ‘Schmass’ Case: Relativize

Necessities to Indicative Actualities
Taking metaphysical necessities to be relative to indicative actuali-
ties makes sense of the data concerning schmass and provides the
14
Moreover, qua natural property schmass appears to be on a par with mass: it too
is a property that, in appropriate circumstances, lawfully influences the motion of
entities having the property. Hence considerations rendering it natural to think that
necessities involving mass are grounded in its nature and identity would seem
equally to motivate thinking that necessities involving schmass would be grounded
in its nature and identity.
basis for a consistent necessitarianism, while avoiding concerns

associated with taking nomological necessity to be a basic form of
necessity. (Either implementation of the relativized conception will
do for purposes of treating the ‘schmass’ case.)
According to the relativized conception, what is nomologically
possible or necessary is, like what is metaphysically possible or neces-
sary, relative to which world is indicatively actual. Since different
worlds may be indicatively actual, the necessitarian can avoid being
nomocentric. The necessitarian may rather suppose that relative to
the world which is in fact actual (that is, our very own world), it is
nomologically necessary that there is no schmass; but allow that rela-
tive to another indicatively actual world, it might rather be nomologi-
cally possible that there is schmass. If we are operating with the
domain-relative version of the relativized conception, for example,
then on the necessitarian’s view, there may be multiple subspaces of
possible worlds, associated with different laws of nature. Moreover,
the relativized conception can accommodate the core necessitarian
claim that the laws are metaphysically necessary: here it will be sup-
posed that the laws of nature operative at each indicatively actual
world impose constraints on the laws at all the other worlds in the
associated subspace—namely, that these laws be the same as (or rele-
vantly similar to) the laws operative at the indicatively actual world.
Relativized metaphysical modality thus has the resources to recon-
cile the basic necessitarian claim that what is nomologically necessary
is metaphysically necessary with Fine’s observation (with which Shoe-
maker agrees) that it is metaphysically possible that there be worlds
governed by entirely different laws: the first claim may be accommo-
dated by supposing that, relative to a given indicatively actual world,
every nomological necessity is metaphysically necessary; while the
second claim (like the data concerning schmass) may be accommo-
dated by supposing that indicatively actual worlds may differ with
respect to what is metaphysically, hence nomologically, necessary.
Here again the real culprit giving rise to the seemingly prob-
lematic nature of the data is the insensitivity of the standard con-
ception of metaphysical modality to the need for relativization to
indicatively actual worlds. Insofar as (a live interpretation of ) the
data concerning schmass indicates that some nomological impos-

sibilities are metaphysically possible, there is no way, on the
standard conception, to reconcile the data concerning schmass
with the necessitarian view that nomological necessity is a species
of metaphysical necessity. However, on the supposition that meta-
physical modalities are relative to indicative actualities, this rela-
tivization may be interpreted in necessitarian-friendly fashion as
indicating that relative to an indicatively actual world, every
nomologically necessary claim is metaphysically necessary. As
with the Woody case, the key moral of the schmass case, in the
first instance, is that appropriately accommodating the data
requires that metaphysical modal space have a relativized
structure.
More generally, to return to Fine’s deeper concern, this structure
illustrates how nomological necessity might be, in an appropri-
ately relativized sense, a restricted form of metaphysical necessity.
Every nomological necessity is a metaphysical necessity, relative to
some indicatively actual world. As such, on the relativized concep-
tion nomological necessity need not be seen as a basic form of
necessity, but rather may be seen, in a fashion desirably unified
with the other non-metaphysical forms of necessity, as ultimately
grounded in the natures or identities of the entities at issue in
nomological claims.
4 THE BROAD NEUTRALITY OF RELATIVIZED

METAPHYSICAL MODALITY
In closing, we want to briefly flag the broad neutrality of relativ-

ized metaphysical modality with respect to the actualist/possibil-
ist and transworld identity/counterpart theory distinctions and
associated debates. We can’t do full justice to the options here, but
will try to illustrate the flexibility of the relativized conception,
and note a couple of choice points, by attention to how the con-
ception might accommodate certain standard positions in these
debates.
4.1 The Actualist/Possibilist Debate

Actualists subscribe to the thesis that everything that exists is
actual. Possibilists disagree. According to the possibilist, in addi-
tion to the actual world and actual individuals, there exist other,
merely possible worlds and individuals. Relativized metaphysical
modality is broadly neutral with respect to this debate: actualists
and possibilists alike can in principle help themselves to either the
overlapping spaces or non-overlapping subspaces implementa-
tions of the view. We say ‘in principle’, though, since depending on
how a given version of actualism or possibilism is spelled out, one
or other implementation of a relativized conception might be
thought a better fit.
On a standard actualist treatment, possible worlds are identified
with some sort of actually existing abstract entity—a complex prop-
erty (Stalnaker 1976), a complex state of affairs (Plantinga 1976), or
a set of propositions (Adams 1974); the actual world is distinguished
from merely possible worlds as being the world that is instantiated,
obtains, or is such that the constituent propositions are true, respec-
tively. Can abstractionist actualists endorse worlds of the sort enter-
ing into either implementation of the relativized conception? First,
note that abstractionist actualists typically assume that possible
worlds are ‘maximal’, which assumption might be thought to fit
better with the non-overlapping subspace implementation, on
which worlds, both pre- and post-relativization, are maximally
characterized; on the overlapping spaces implementation, worlds
are incomplete prior to relativization, hence (in abstractionist terms)
represent only the ‘canonical’ or basic truths (e.g., as a non-maximal
set of propositions or complex property). Still, on either implemen-
tation worlds post-relativization will be maximal; and since there is
no in-principle problem with abstract entities’ being non-maximal,
it seems the abstractionist actualist can go either way.
The question remains: does the actualist supposition that every-
thing that exists is actual make good sense in a context where differ-
ent worlds can be indicatively actual? Again, we see no conflict
here. On one reading, the concern is that the relativized conception
can’t make sense of the intuition, sometimes seen as supporting the
actualist view, that the actual world is somehow ‘special’ as com-
pared to other merely possible worlds. As above, for the abstrac-
tionist actualist, the special nature of the actual world is reflected in

one of the worlds being instantiated or obtaining; as such, that some
other world might instead be indicatively actual is no more prob-
lematic than that some other properties than those that are actually
instantiated might instead be instantiated. In any case, one needn’t
insist that making sense, e.g., of the Woody data requires that w2
‘really’ (somehow or other) be instantiated; if no worlds besides our
very own actual world can be so lucky, then one may rather under-
stand the relativized conception as tracking a certain complexity in
our hypothetical deliberations (as involving consideration of not
just counterfactually actual, but also hypothetically indicatively
actual, goings-on). On another reading, the concern is that relativi-
zation to a world different from our very own actual world would
introduce a non-actualist domain. But the relativized conception,
while making room for worlds relevant to modal deliberation to
involve non-actual individuals, does not require any such thing.
Here the action is in the further details of what worlds the modal
theorist thinks exist; given that the actualist constructs merely pos-
sible worlds from actualia, as per the abstractionist and other stand-
ard ‘domain-inclusion’ versions of actualism, then the relativization
to such a world as indicatively actual will not introduce a non-actu-
alist domain.
Indeed, it is worth pointing out that either implementation of
relativized metaphysical modality is broadly consistent with ver-
sions of actualism according to which possible worlds and individ-
uals do not exist at all, but merely could exist, as on the
‘non-domain-inclusion’ actualism recently developed in Bennett
(2005). For example, holding fixed our own world as indicatively
actual, the non-domain-inclusion actualist may consistently hold
both that everything that exists is actual and that each of the other
possible worlds in the post-relativized space does not exist, but
merely could exist, as a matter of fact not grounded in any existing
entity.
The relativized conception can also accommodate standard
accounts of possibilism. Broadly conceived, possibilism allows,
contra the actualist view, that possible worlds and their occupants
may not actually exist. So broadly characterized, possibilism is
compatible with either implementation of a relativized conception;
indeed, one standard way to make out the view is as extending the
sort of abstractionist actualist account to allow that some abstracta

corresponding to possible worlds may advert to alien individuals
or properties (see Menzel 2008). Here again the usual supposition
of the maximality of worlds may be accommodated, post-relativi-
zation, on either implementation; and either implementation may
accommodate the status of worlds as constructed from possibilist-
friendly resources.
On a more specific, and more notorious, approach to possibilism,
this view is combined with the thesis that possible worlds and their
occupants are in some sense ‘concrete’ (see Lewis 1986, Bricker
2008). Supposing, as in Lewis (1986) and McDaniel (2004), that con-
crete worlds are determinate with respect to all matters of particu-
lar fact, the non-abstractionist possibilist will find the overlapping
subspaces implementation unappealing, in requiring a pre-relativ-
ized space of worlds individuated at the level of canonical or
semantically stable descriptions which leave out, as above, many
(perhaps most) truths of matters of particular fact. Possibilists who
take on this additional (but by no means mandatory) metaphysical
constraint on the nature of possible worlds will then presumably
find the non-overlapping spaces implementation more amenable.
4.2 The Transworld Identity/Counterpart Debate

A similarly broad neutrality applies to the issues of transworld
identification and representation de re. Does a given individual ever
literally exist at more than one possible world, in the sense defended
in Kripke (1972), Plantinga (1973), and van Inwagen (1985)? Or do
worlds represent that something is possible or necessary, for an
individual i, in virtue of containing a numerically distinct counterpart
of i which resembles i in certain (typically contextually determined)
respects? Relativized metaphysical modality does not force a choice
here: each implementation of this view is consistent with either lit-
eral transworld identity across worlds or its denial in favor of some
counterpart-theoretic means of de re representation.
Paradigmatic of accounts that reject transworld identity is
Lewis’s treatment according to which representation of an individ-
ual in modal claims involves not (necessarily) that individual itself,
but rather the individual’s counterparts at various possible worlds,
where the notion of a counterpart is defined in terms of overall (see

Lewis 1968) or context-dependent (see Lewis 1971) similarity. So,
for example, given that Woody actually originates from matter m in
world w1, what it would be for a possible world w2 to represent that
Woody originates from some different matter m’ would be for
Woody to have a counterpart in w2 that originates from m’. In the
same vein are accounts on which representation de re is based in
sameness of maximally specific qualitative roles (see McMichael
1983). Here, for distinct worlds w1 and w2 to represent de re facts
concerning Woody is for a certain qualitative role to be exemplified
at both w1 and w2 (exemplification of the role will presumably
make appropriate room for the flexibility of Woody’s origins). This
proposal, like Lewis’s, is typically offered against a background
where individuals are strictly speaking world-bound. The relativ-
ized conception, in either version, has no trouble accommodating
the failure of individuals in different worlds to be strictly identical;
even on the first implementation of the conception, talk of overlap-
ping worlds might be understood as involving type rather than
token identity of basically individuated worlds. To be sure, if the
modal facts are context-dependent in the way that counterpart the-
ory supposes, then this will introduce another degree, so to speak,
of relativization: rather than the metaphysical modal facts being
relative just to which world is indicatively actual, such facts will
also be relative to which counterpart relation is in place. In any case,
nothing in either version of the relativized conception rules out
incorporating further contextual aspects, in line with counterpart
theory.
On the other hand, one might rather treat de re representation in
terms of literal transworld identity. Here again there are options.
One might suppose (following Kripke) that in considering what is
possible or necessary for a given individual, one may stipulate
that it is that very individual that one has in mind—notwithstand-
ing, of course, that we cannot stipulate (assuming the falsity of
modal conventionalism) what is modally the case with the indi-
vidual in question. Alternatively, one might suppose that Woody’s
existence across various possible worlds is grounded in the exem-
plification of a haecceity (roughly, the property of being identical
to Woody), as in Plantinga (1974). Or, as per Spencer (2006), one might
endorse an intermediate position and treat representation de re in
terms of counterpart relations that are restricted so as to model

the formal properties of the identity relation (i.e., symmetry and
transitivity). Each of these options is formally compatible with
either implementation of relativized metaphysical modality, as
developed so far. This conception simply leaves open such fur-
ther details concerning how and why de re representation is to
proceed.
That each implementation of relativized metaphysical modality
is compatible with either transworld identity or counterpart theory
leads to a final moral of the Woody case; namely, that debates over
the viability of transworld identity are largely orthogonal to debates
over essentialism and the sorts of flexibility in material origins
brought out by the data in that case.15 This orthogonality is liable to
be overlooked, given Salmon’s own treatment of the data in terms
of literal transworld identity, and Lewis’s subsequent (1986) reply
and critique, couched entirely in the language of counterpart the-
ory. The choice presented by the Woody case is not, as the Salmon–
Lewis debate suggests, between a view on which transworld
identity is retained by accepting intransitive accessibility (along
with the ‘impossible’ worlds that gave Lewis such pain), and a
view on which transitive accessibility between worlds is retained
by accepting counterpart theory between individuals. Indeed, on
relativized metaphysical modality, the Woody case can be closed
while leaving all these options open.
University of Toronto
REFERENCES
Armstrong, D.M., 1989. A Combinatorial Theory of Possibility. Cambridge:
Bealer, George, 2000. ‘A Theory of the A Priori’. Pacific Philosophical Quar-
terly 81 (1):1–30.
Bennett, Karen, 2005. ‘Two Axes of Actualism’. Philosophical Review
114(3):297–326.
Bird, Alexander, 2007. Nature’s Metaphysics: Laws and Properties. Oxford:
Oxford University Press.
15
Thanks to Laurie Paul for calling this advantage to our attention.
Bricker, Phillip, 2008. ‘Concrete Possible Worlds’. In Contemporary Debates

in Metaphysics, J. Hawthorne, T. Sider, and D. Zimmerman (eds.).
Oxford: Blackwell Publishers.
Burgess, John P., 2009. Philosophical Logic. Princeton: Princeton University
Press.
Chalmers, David J., 2006. ‘The Foundations of Two-dimensional Seman-
tics’. In Two-Dimensional Semantics: Foundations and Applications, Manuel
Garcia-Carpintero and Josep Macia, (eds.). Oxford University Press.
Chalmers, David and Jackson, Frank, 2001. ‘Conceptual Analysis and
Reductive Explanation’. Philosophical Review 110: 315–60.
Chandler, Hugh, 1976. ‘Plantinga and the Contingently Possible’. Analy-
sis, 36:106–109.
Davies, Martin and Humberstone, Lloyd, 1980. ‘Two Notions of Neces-
sity’. Philosophical Studies. 38:1–31.
Ellis, Brian, 2001. Scientific Essentialism. Cambridge: Cambridge Univer-
sity Press.
Fine, Kit, 2005. ‘The Varieties of Necessity’. In Modality and Tense. Oxford
University Press.
Kaplan, David 1979. ‘On the Logic of Demonstratives’. Journal of Philo-
sophical Logic. 8.
—— 1989. ‘Demonstratives’. In Themes from Kaplan, Almog, J., Perry, J.,
Wettstein, H., and Kaplan, D. (eds.). Oxford: Oxford University Press.
481–564.
Kripke, Saul, 1963. ‘Semantical Considerations on Modal Logic’. Acta
Philosophica Fennica, 16:83–94.
—— 1972. Naming and Necessity. Cambridge MA: Harvard University Press.
Leibniz, Gottfried, 1686. Discourse on Metaphysics and the Monadology.
Drier Publications.
Lewis, David, 1968. ‘Counterpart Theory and Quantified Modal Logic’.
Journal of Philosophy 65:113–126.
—— 1971. ‘Counterparts of Persons and their Bodies’. Journal of Philoso-
phy. 68:203–11.
—— 1986. On the Plurality of Worlds. London: Blackwell.
McDaniel, Kris, 2004. ‘Modal Realism with Overlap’. Australasian Journal
of Philosophy 82(1): 137–52.
McMichael, Alan, 1983. ‘A Problem for Actualism about Possible Worlds’.
Philosophical Review, 92(1): 49–66.
Menzel, Christopher, 2008. ‘Actualism’. Stanford Encyclopedia of Philosophy,
Edward N. Zalta (ed.), http://plato.stanford.edu/entries/actualism.
Plantinga, Alvin, 1973. ‘Transworld Identity or Worldbound Individu-
als?’. In Logic and Ontology, Milton Munitz (ed.). New York: New York
University Press.
Plantinga, Alvin, 1976. ‘Actualism and Possible Worlds’. Theoria (42):

139–160.
Salmon, Nathan, 1981. Reference and Essence. Princeton: Princeton Univer-
sity Press.
—— 1984. ‘Impossible Worlds’. Analysis, 44:114–17.
—— 1989. ‘The Logic of What Might Have Been’. Philosophical Review,
98:3–34.
Shoemaker, Sydney, 1980. ‘Causality and Properties’. In Time and Cause,
Peter van Inwagen, (ed.). 109–35. Dordrecht: D. Reidel.
Sider, Theodore, 2010. Logic for Philosophy. Oxford: Oxford University
Press.
Spencer, Cara, 2006. ‘Keeping Track of Objects in Conversation’. In Two-
Dimensional Semantics: Foundations and Applications, Manuel Garcia-
Carpintero and Josep Macia, (eds.). Oxford: Oxford University Press.
Stalnaker, Robert, 1976. ‘Possible Worlds’. Noûs 10 (1): 65–75.
—— 2008. ‘Assertion’. In Formal Semantics: The Essential Readings, P. Portner
and B. H. Partee, (eds.). Blackwell Publishers.
van Inwagen, Peter, 1985. ‘Plantinga on Trans-world Identity’. In Alvin
Plantinga, van Inwagen and Tomberlin, (eds.). Dordrecht Publishers,
101–120.
8. Coincidence Through Thick and Thin
Sydney Shoemaker
Coincident objects are numerically different objects that are com-

posed of the same matter and so occupy the same portion of space.
Supposing this is possible, the coincidence could be either temporary
or permanent; two objects might coincide for a while and then part
company and go their separate ways, or they might coincide through-
out their careers. Whether coincidence is possible is disputed. Some
philosophers think it obviously is possible, and cite as examples a
statue and the piece of clay of which it is composed, and persons and
their bodies. The usual reason for holding this is that such pairs of
entities appear to differ in their historical and modal properties. Oth-
ers think it is not possible, and claim that the view that it is possible
has counterintuitive consequences, raises unanswerable questions,
and involves a severe case of overcounting.
Important issues turn on whether coincidence is possible. Per-
haps the most serious of these concerns the nature of personal
identity. On psychological accounts of personal identity, sometimes
called neo-Lockean accounts, it is possible for persons to change
bodies, via a brain or cerebrum transplant or via some sort of brain
reprogramming.1 This seems to have the consequence that persons
are not identical with their bodies. Yet persons and their bodies are
composed of the same matter. This seems to give us coincident
1
The idea that a person could change bodies by way of a brain transplant (see
Shoemaker (1963)) takes its inspiration from Locke’s example in which the soul of a
prince is transferred to the body of a cobbler. Eric Olson (1997) has claimed that since
the whole brain includes the brainstem which is the control center for the biological
life of a person, the claim that the person would go with the brain in a brain transplant
is compatible with the rejection of the psychological view; it is compatible with per-
sonal identity consisting in biological continuity rather than psychological continuity.
If this is right, the psychological account should say that the transplantation of just the
cerebrum could be person preserving. For a case in which the change of bodies is via
brain reprogramming, see Perry (1972).
228 | Sydney Shoemaker
entities. It also seems that on such accounts persons are not identi-
cal with the biologically individuated animals with which they
share their space and matter. If the body of a person can survive
death, as a corpse, and if both the person and the biological animal
cease to exist at death, then wherever there is a person there are (at
least) three different coincident entities – the person, the person’s
body, and the biological animal that shares the space and the matter
of both.2
Advocates of perdurance (four-dimensionalist) views of the per-
sistence of objects are committed to a kind of coincidence. They are
committed to the coincidence of temporally extended objects with
their temporal parts, and to the coincidence during the period of
overlap of temporally extended objects that share temporal parts. For
some perdurance theorists this is the only sort of coincidence that is
possible. And they see it as a benign form – since they think that in
the first instance it is the momentary temporal parts of objects that
have the properties things have at times, and since these parts are
shared in cases of temporal overlap, they are not committed to what
may seem an obnoxious consequence of coincidence, namely there
being multiple instantiations of the same property at the same time
and place. E.g., they are not committed to there being two instances
of weighing one hundred pounds where a coincident statue and
piece of clay, each weighing one hundred pounds, are located, and
the apparent consequence of this that a scale ought to read two hun-
dred pounds when the statue is placed on it. If the only coincidence
allowed is in cases of temporary overlap, the perdurance theorist will
not allow that there can be pairs of objects that coincide throughout
their careers. However, it appears that a perdurance theorist could
allow this for the same reasons that others do, namely modal rea-
sons – if the careers of the statues and the piece of clay could come
apart, even if in fact they don’t, it would appear that the statue and
the piece of clay must be different.
In any case, I will not be concerned here with the sort of coinci-
dence that all perdurance theorists allow – that which consists in
the sharing of temporal parts. I don’t myself believe that persisting
2
By a “biological animal” I mean an animal with biological persistence conditions.
A neo-Lockean can hold that persons are animals, but not that they are biological
animals.
Coincidence Through Thick and Thin | 229
objects, as opposed to events and processes, have temporal parts.

The sort of coincidence I will be concerned with is a sort that can
occur in objects as conceived by endurance theorists, i.e. by three-
dimensionalists. I shall not here argue for the three-dimensionalist
view. And it may be that some of what I say can be incorporated
into a four-dimensionalist account. My aim in this paper is to give
an account of what coincidence would consist in, and why we
should allow that it is both possible and actual. It is with respect to
a certain kind of objects, or a certain conception of objects, that
I think it is possible; and it is natural to think of this conception in
three-dimensionalist terms.
II
Let me start by sketching the conception of objects I will be working

with here. The conception is one I think is satisfied by the objects
that figure in the commonsense picture of the world, and in a good
part of the scientific picture as well – whether it fits the objects that
figure in microphysics I am not competent to say. I do not claim, in
any case, that all of the objects there are fit this conception. If there
are objects that are arbitrary chunks of spacetime, composed of
whatever matter is contained in those chunks, or are mereological
sums of the members of arbitrary sets of familiar objects, they do
not satisfy the conception. I am not denying (or affirming) that there
are such objects.
It is essential to objects on this conception that they have careers,
which can be thought of as causally unified series of sets of simulta-
neous property instances. The property instances in each of the sets
in the series will be unified by synchronic unity relations, and the sets
in the series will be unified by diachronic unity relations. There being
such a career, a series of sets of property instances standing in appro-
priate unity relations, will be sufficient for there existing an object
having that career. But this should not be understood as amounting
to a reduction of objects to property instances, a theory which makes
objects bundles of property instances. A property instance is the
instantiation of a property in a thing at a time, and the notion of a
property instance involves the notion of objects in which properties
can be instantiated. What we have here is a “package deal” – the
notion of an object and that of a property instance are internally

related. The synchronic unity relation for property instances involves
the instances being so related that they jointly produce certain effects
or bestow certain powers – what effects and powers these are being
determined by the causal profiles that individuate the properties. So,
for example, an instance of a certain mass and one of a certain shape
jointly bestow, when synchronically unified, the power to make a cer-
tain sort of impression on things made of certain materials when
applied to them with appropriate force, and instances of certain
beliefs and certain desires that bestow, when synchronically unified,
a disposition to engage in certain actions. There is a two-way connec-
tion here. If the property instances are synchronically unified, they
jointly produce the relevant powers and effects. But also, if the prop-
erty instances are so related that they jointly produce the relevant
powers and effects, they are synchronically unified.
The diachronic unity relation for property instances involves the
instances being so related, causally, that the instances that occur in
the series at one time generate, in combination with input from the
outside, the instances that occur in it at subsequent times. This does
not require that every instance that occurs in the series at one time
is causally connected to every instance, or even to any instance, that
occurs at some subsequent time. Property instances occurring at
different times can be diachronically unified in virtue of the fact
that each is synchronically unified with a member of a pair of prop-
erty instances whose members are causally related, the earlier one
being a partial cause of the later one.
There is, I believe, an internal relation between the causal profiles of
properties and the persistence conditions of the things that have
them.3 Transtemporal identity requires that the successive stages in
the career of a thing be causally related, where this involves there
being relations of counterfactual dependence of later stages on earlier
ones – often later stages inherit their properties from earlier ones, and
even when they don’t it will be true that if the earlier stages had been
different in certain ways the later stages would have been correspond-
ingly different (a growing tree changes its size and shape, but what its
size and shape are at a later time is partly determined by what they
3
See Shoemaker (1979).
were earlier). These relations of causal and counterfactual depend-

ence hold in virtue of the causal profiles of the properties instantiated
in the course of the object’s career. But the persistence conditions of
the subjects of the properties figure in the specification of the causal
profiles. For in some cases what is built into the causal profile of a
property is not just that instances of it will contribute to the produc-
tion of certain effects but that they will contribute to the production of
those effects in the future career of the thing in which that property instance
occurred. This is clearly the case with many dispositional properties.
For something to be combustible is for it to be such that if it is sub-
jected to a certain degree of heat it, that same thing, will burn. For some-
thing to be elastic is for it to be such that if subjected to certain forces
it, that same thing, will undergo a change of shape and then, when the
force is removed, revert to its (that same thing’s) original shape. And
this will be reflected in the causal profiles of the properties that realize
these dispositional properties. We have here the same sort of two-way
connection we noted earlier in the case of synchronic unity – the hold-
ing of certain causal relations is implied by, but also implies, diachronic
unity and the transtemporal identity it brings with it. This is all part of
the “package deal” mentioned earlier – the internal relation between
the notion of an object and the notion of a property instance.
The objects that have careers of the kind I have characterized
include ourselves. If one thinks that there are objects that are not of
this kind, one might speculate that it is because we ourselves are
objects of this kind that our commonsense ontology is built around
such objects. One might even think that there could be knowers and
conceivers that are not of this sort, and that these would have a
different conception of the objects that surround them, taking these
to be entities like themselves. But on reflection it seems that it is not
possible for there to be conceivers and knowers that are not of this
sort. Maybe there are objects that are simply chunks of spacetime, or
are mereological sums of ordinary objects. But these could not have
the sorts of properties that knowers and conceivers must have.
III
Assuming physicalism, as I will do throughout this paper, the men-

tal properties of persons are realized in physical properties. And as
we will see shortly, all of the physical properties that figure in our
ordinary thought and discourse are realized in other physical
properties.
It is useful to distinguish two different sorts of realization.4 One is
the sort that figured in the preceding paragraph, where the realizer
of a property instance is another property instance. Call this prop-
erty realization. I favor an account of this on which one property
realizes another just in case the forward-looking causal features of
the realized property are a subset of the forward-looking causal fea-
tures of the realizer, and, accordingly, one property instance realizes
another just in case the realized properties are so related. In the
other sort of realization what realizes a property instance is a micro-
physical state of affairs – its being the case that certain micro-entities
are propertied and related in a certain way. Here too the notion of
realization involves the notion of a causal profile; microphysical
states of affairs have causal profiles, consisting in their aptness to
cause or contribute to causing certain effects, and the causal profile
of the microphysical state of affairs that realizes a property instance
must match the causal profile of the instantiated property.5 Call this
microphysical realization. In both cases there will be a multipli-
city of ways in which the instantiation of a given property can be
instantiated – in both there will be “multiple realization” – but in
both the existence of a realizer of a property instance constitutes
that property instance.
Realization of the second of these sorts, microphysical realization,
brings with it a special case of realization of the first sort, property
realization. If a thing’s career embeds a microphysical state of affairs
at a given time, the thing thereby has at that time the property of hav-
ing a career that at that time embeds such a state of affairs. Call this a
microphysical-state-of-affairs-embedding property, or, for short, an
MSE property. There is a one–one relation between types of micro-
physical states of affairs and MSE properties. Since every physical
property instance has a microphysical state of affairs realizer, every
physical property instance has a property realizer that is an instance
of an MSE property. MSE properties are not properties that figure in
ordinary thought and discourse. They are too fine-grained for that.
4
See Shoemaker (2007).
5
For a discussion of this sort of matching, see Shoemaker (forthcoming).
The slightest difference between two microphysical states of affairs,

e.g., a slight difference in the location of an electron, makes for a dif-
ference in MSE properties. So it cannot be known in any ordinary
way whether two things share an MSE property, or that something
has an MSE property it had a moment before (although in both cases
it can be known that such sameness of MSE properties is highly
unlikely).
III
As a final preliminary to presenting my account of coincidence,

I distinguish two sorts of properties of things. There are proper-
ties that can belong to things of very different kinds. Examples
would be colors, shape properties, mass, and electric charge. I call
these thin properties. Their thinness is with respect to causal fea-
tures in their causal profiles – their causal profiles lack features
that limit their instantiation to things of particular kinds. This
distinguishes them from what I call thick properties, which can
only belong to things of certain kinds. Sortal properties, like being
a tree or being an automobile, are of course thick. But so also are
properties that imply sortal properties, like being in bloom or
being in first gear. Thickness varies in degree. The property of
being an oak tree is thicker than the property of being a tree, and
the property of being a 1974 Chevy is thicker than the property of
being an automobile. But for present purposes I will ignore such
differences in degree of thickness.
I mentioned earlier that the causal profiles of properties include
causal features that are such that the effects of their activation nec-
essarily includes effects on the future career of the thing that has
them. It is such properties whose natures are internally related
to the persistence conditions of the things that have them, and so to
the persistence conditions of particular kinds of things. And it is
especially in the case of thick properties that we find such causal
features. Thick properties are associated with kinds of things, and
kinds of things are individuated, at least in part, by their persist-
ence conditions.
I take mental properties to be thick. Their functional roles give them
causal profiles of the sort just mentioned, ones whose specification
refers to effects on the future careers of their possessors. What beliefs

and desires a person has at a time will be to a considerable extent
determined by what beliefs and desires the person had at earlier times,
although they of course will be partly due to inputs to the person’s
mind subsequent to those earlier times. The dispositions associated
with these states are ones that are largely played out within the careers
of their possessors. On psychological, neo-Lockean, accounts of per-
sonal identity it is relations between states of this kind occurring at
different times, relations that reflect their causal profiles, that consti-
tutes the persistence of persons over time.
IV
It is evident that the thin physical properties true of something do

not realize any of its thick properties, since a thin property can have
only a proper subset of the causal features belonging to a thick
property. It follows from this that the thin physical properties
belonging to a person cannot be property realizers of the person’s
mental properties. If all physical properties were thin, this would
have the consequence that mental properties cannot be realized by
physical properties. So if one is a physicalist, one must hold that
there are thick physical properties, these including the physical
realizers of mental properties.
This relates to an objection sometimes raised against psychologi-
cal accounts of personal identity. Such accounts are committed to
the view that persons are not identical to their bodies. But persons
and their bodies are composed of the same matter. It might seem to
follow that persons and their bodies are physically exactly alike,
and share all of the same physical properties. If this were so, then if
the mental states of persons are realized in physical properties it
must be the case that those mental states are also possessed by the
bodies of persons. They must also be possessed by the biologically
individuated animal that shares its matter with the person and the
body. This gives us “too many minds,” or “too many thinkers.” So
it might seem that the proponent of the psychological view must
choose between rejecting physicalism, i.e., rejecting the claim that
the mental states of persons are physically realized, and accepting
the claim that where we take there to be one person there are two or
three psychologically identical subjects. The latter view would of

course undermine the psychological view – for if the body and the
biological animal share the mental properties of persons they
should themselves count as persons, and as persons that lack the
persistence conditions that the psychological view ascribes to per-
sons. The former view will also be unacceptable to psychological
theorists who are physicalists, as most recent psychological theo-
rists have been.
This objection of course collapses if we allow that the only physi-
cal properties shared by coincident entities are thin properties, that
there are thick physical properties, and that it is only the latter that
realize mental properties. But this leaves us with the task of explain-
ing how coincident objects can differ in some of their physical prop-
erties, i.e., of how one of a pair of coincident objects can have thick
properties the other lacks, this despite the fact that the coincident
objects are composed of the same matter and share the same thin
physical properties. A related task is that of explaining how the
coincidence view can be reconciled with the view, central to physi-
calism, that the distribution of fundamental physical particles in the
world determines the distribution of mental properties. That physi-
calist view implies that the distribution of fundamental physical
particles determines what microphysical states of affairs there are,
and so what microphysical realizers of mental states there are. But,
as we have seen, to each type of microphysical state of affairs there
corresponds a physical property, an MSE property, that is instanti-
ated in a thing just in case its career embeds a microphysical state of
that type. These MSE properties appear to be thin properties (though
we will see later that there are also thick MSE properties). The same
microphysical states of affairs can exist in the careers of objects
of different types, and, more to the point, a microphysical state of
affairs that occurs in the career of a thing will occur in the career of
an object coincident with that thing. Because of this, things of differ-
ent types can share the same MSE properties, and coincident objects
will share the same (thin) MSE properties. All of this suggests that
the distribution of MSE properties in the world determines the dis-
tribution of mental properties in the world, and so that the distribu-
tion of mental properties is determined by the distribution of thin
physical properties. How can this be if mental properties are not
realized in thin physical properties?
There is a difference between some properties being instantiated

in a thing guaranteeing that some other property is instantiated in
that thing, and their guaranteeing simply that that other property
is instantiated in something composed of the matter of which the
thing is made. If the distribution of MSE property instances in the
world determines the distribution of mental property instances,
then for each mental property instance there will be a set of MSE
property instances whose members jointly necessitate its exist-
ence. One such set, of course, would be the set of all MSE-property
instances. But there will be much smaller sets whose members
jointly necessitate its existence, and some of these will be sets
whose members occur in a single thing and where their occurring
in a thing does not entail that that thing has the mental property in
question, but does entail that some related thing has it. To put the
matter in terms of microphysical states of affairs, for any mental
property instance there will be a microphysical state of affairs
whose existence entails the existence of that property instance but
which can be embedded in the career of a thing that does not have
the mental property in question.
I have said that an MSE property is one that a thing has in virtue
of a microphysical state of affairs being embedded in its career. But
we need to distinguish different ways in which a microphysical
state of affairs can be embedded in the career of a thing. One sort of
embedding, call this weak embedding, simply consists in the fact that
the microentities involved in the state of affairs are all ones that are,
at the time in question, among those that make up the object in
whose career they are embedded. A property’s being embedded in
this way amounts to its being a microstructrual property, and any
microphysical state of affairs that is embedded at all is embedded in
at least this sense. Such embedding gives us an MSE property that
is a thin property that anything having it shares with any object
coincident with it.
But the microrealization of a thick property requires a stronger
sort of embedding, call it strong embedding. A microphysical state of
affairs being strongly embedded in a career at a time requires that it
be related in certain ways to states of affairs occurring in that career
at other times. As I said earlier, we can think of the career of a thing
of a certain sort as a series of sets of property instances that are so

related, causally, that the persistence conditions for things of that
sort, together with the causal profiles of the properties instantiated,
make that series the career of a thing of that sort. Assuming physi-
calism, this will be realized in a series of microphysical states of
affairs that realize the property instances. Let’s say that a micro-
physical state of affairs that is a member of such a series at a time is
a momentary stage of the thing at that time. This stage will be made
up of smaller states of affairs that are realizers of property instances
occurring in that thing at that time. What makes the stage made up
of the states of affairs the stage of a particular thing is the fact that
the causal profiles of the states of affairs that make it up are such
that their occurrence contributes to the implementation of the per-
sistence conditions for things of the sort in question. This will
involve some of the states of affairs being realizers of thick proper-
ties that can be instantiated only in careers of things of that sort.
A microphysical state of affairs will be strongly embedded in the
career of a thing of a given sort at a time if it is such a part of
the microphysical state of affairs that is the momentary stage of the
thing at that time, i.e., if it has a causal profile that makes the requi-
site contribution to the implementation of the thing’s persistence
conditions, qua thing of that sort.
In the case of coincident entities, like me and my body, there
will be two different careers that involve, during the period of
coincidence, the same micro-entities. These will have the same
microphysical states of affairs weakly embedded in them. Some of
these states of affairs will be strongly embedded in the career of
one of the coincident entities, and some will be strongly embed-
ded in the career of the other. Which is to say that one of these
objects will have one set of thick MSE properties, and the other
will have a different set of such properties. A thick MSE property
will be a property something has in virtue of its career strongly
embedding a microphysical state of affairs. The things will of
course share a number of thin properties, and the thin properties
will be property realized by thin MSE properties. Here the MSE
properties will be realized by states of affairs that are weakly
embedded in the careers of their possessors. Some of these states
of affairs will also be strongly embedded in the career of one or the
other of the coincident entities.
Consider a microphysical state of affairs that realizes an instance

in me of a mental property, say thinking of Vienna. This state of
affairs occurs in me, and it also occurs in my body. But while it is
strongly embedded in my career, in the sense just explained, it
is not strongly embedded in the career of my body, although it is
weakly embedded in it. And so my body does not have the thick
MSE-property constituted by the strong embedding of that state of
affairs. The occurrence of that state of affairs in my body’s career
does guarantee that something has that mental property. But that
something is me, not my body. What its occurrence in my body
entails is that there is something coincident with my body whose
career strongly embeds it and so has the mental property it realizes.
VI
I should clarify what I mean when I speak of the microphysical

states of affairs that make up a momentary stage contributing to the
implementation of the persistence conditions for things of a certain
sort. Microphysical states of affairs have causal profiles, which they
have in virtue of the laws governing the microentities that enter
into them. So the microphysical states of affairs that make up a
momentary stage will contribute to generating subsequent states of
affairs – they will generate them in conjunction with other micro-
physical states of affairs that lie outside the career of which the stage
is a stage. Normally a subset of these subsequent states of affairs will
form a series of sets of synchronically unified states of affairs that
form thing-stages and together constitute part of a career to which
our initial stage belongs.6 The states of affairs that make up these
subsequent stages will realize property instances that stand in rela-
tions of diachronic unity to property instances realized by states of
affairs occurring in other stages of the series, including those in our
initial stage. These property instances will include instances of thick
properties that can only be instantiated in things that are of a cer-
tain sort and have certain persistence conditions; the persistence
6
I say “normally” because there is the exceptional case in which the stage is the
final stage in the thing’s career, because its propensity to contribute to a future
career is overridden by the causal efficacy of other states of affairs that leads to the
termination of the career.
conditions will be reflected in the relations of diachronic unity that

unite the series. Since the series of stages will be generated in part
by states of affairs external to our initial stage, and these can be dif-
ferent in different possible worlds in which that stage occurs, the
future shape of the career to which that stage belongs will be differ-
ent in those different possible worlds. But in all those worlds there
will be future stages with which it is diachronically unified, unless
there occurs something that prevents that stage from making its
contribution to the future – something that ends the existence of the
thing of which it is a stage. It is in virtue of the causal profiles of the
states of affairs that make it up that the stage contributes to generat-
ing a future career of a thing of a certain sort and having certain
persistence conditions.
Given that there can be coincident objects, the states of affairs in
our stage may contribute to the generation of two or more future
careers. Here it will be the causal profiles of different states of affairs
among those that make up the stage that contribute to the genera-
tion of the different careers. The careers may coincide for a while
and then go their separate ways. But they may coincide as long as
they exist. In the latter case the diachronic unity of the series of
stages making up the career will be overdetermined; one subset of
the property instances realized by the states of affairs in these stages
will be unified by one set of diachronic unity relations, and another
will be unified by a different set. And the same will be true during
the interval in which two careers coincide.
I have spoken of careers as series of stages, and have spoken of
coincident objects as sharing stages. That would suggest that coin-
cident objects have the same career as long as they coincide. But if,
as I said earlier, stages are sets of synchronically unified property
instances, the coincident objects will have different stages and their
careers will be different even when they coincide. This seems the
preferable meaning of “stage,” and of “career.” But when things
coincide there will be at a given time a single set of microphysical
states of affairs whose members realize the property instances that
make up the two stages, and we can allow a secondary sense of
“stage” in which this counts as one. (Note that stages of coincident
objects will overlap in what property instances they contain; they
will contain the same thin property instances.) They will differ only
in their thick property instances, and this will be a difference with
respect to which of the microphysical states of affairs are strongly

embedded in which career.
VII
Corresponding to the distinction between strong and weak embed-

ding of states of affairs is a distinction between two sorts of prop-
erty realization. So far I have taken realization to be a same-subject
relation: the realized property instance and the property instance
that realizes it must belong to the same thing. Call this realization1.
But there seems to be room for a notion of realization that permits a
property instance in one thing to be realized by a property instance
in a numerically different thing. E.g., assuming that I and my body
are numerically distinct, it nevertheless seems reasonable to say
that my mental states are realized in states of my body. This requires
a new realization relation, call it realization2. This has to be explained
in a way that is compatible with the fact that no property of my
body realizes1 any mental property and that the only properties
I share with my body are thin properties.
The motivation for saying that properties of my body realize my
mental properties is the acceptance of physicalism together with
the fact that there is a good sense in which I and my body are physi-
cally identical. Of course, our being physically identical consists in
our sharing the same thin physical properties. So if there is a sense
in which the properties of my body realize my mental properties, it
had better be a sense in which my thin physical properties realize
my mental properties. My thin physical properties do not by them-
selves realize1 my mental properties. But in conjunction with the
fact that I have the persistence conditions of a person – the fact that
I have the sortal property of being a person – they do realize1 them.
That is to say, they are conjuncts of conjunctive properties that do
realize1 them. So we can define realization2 in terms of realization1
as follows: where A and B are coincident objects, the instantiation of
thin property P in A realizes2 the instantiation of thick property Q in
B just in case B has a sortal property S such that the conjunction of
P and S realizes1 Q. Taking everything to be coincident with itself,
this gives the result that the instantiation of my thin properties in
me realizes2 the instantiation of my mental properties. But it also
gives the result that the instantiation of these thin properties in my

body realizes2 the instantiation of my mental properties.
If we like we can introduce different notions of microphysical
realization corresponding to the distinction just made between dif-
ferent notions of property realization. We can say that a micro-
physical state of affairs in the career of a thing realizes1 a thick
property instance in that thing when it has the persistence condi-
tions required for the instantiation of that thick property, and that
a microphysical state of affairs in the career of a thing realizes2 a
thick property instance in the career of a coincident thing if the
latter has the persistence conditions required for the instantiation
of that thick property.
VIII
The case of coincidence I have mainly used as my example is that

between a person and that person’s body. I will now discuss how my
account applies to the case of coincidence most cited in the litera-
ture, that between a statue and the piece of clay that constitutes it.
It is not immediately clear what count as thick properties of stat-
ues and, especially, of pieces of clay. Presumably the sortal proper-
ties being a statue and being a piece of clay count. But what else? In the
case of pieces of clay, I think that the only other thick properties are
modal properties having to do with the object essentially having
such and such microentities among those that compose it. In the
case of statues we can cite such properties as having a slender waist
and being in a sitting position, and perhaps aesthetic properties like
being graceful.
Matters are complicated here by the fact that statues are artifacts,
and both the properties and the persistence conditions of artifacts
essentially involve the social context in which they exist, including
the intentions and interests of those who make them, maintain them
in existence, and use them. Few if any of their thick properties are
intrinsic – typically the thick properties involve relations to the
social context. A property like having a slender waist is partly a
representational property; it is the property of representing, by
resemblance, a person or other animal with a slender waist. The
relational, or partly relational, nature of these properties means that
the microphysical realizers of instances of them extend far beyond

their boundaries, and are highly complex – in the case of statue
properties they may include the realizers of the mental states of
sculptors, collectors, and museum curators.
Still, the account I have given applies. Statues and the pieces of
clay that constitute them share thin properties like size, shape, color,
and mass. They also share relational properties that must count as
thin. It is true of the piece of clay, as it is of the statue, that people in
its presence tend to have feelings of admiration, or of revulsion,
although the admiration or revulsion is directed at the statue rather
than at the piece of clay per se. Assuming physicalism, instances of
these properties are realized in microphysical states of affairs. And
some of these states of affairs are strongly embedded in the career
of the statue and only weakly embedded in the career of the piece of
clay. Their being strongly embedded in the career of the statue inti-
mately involves the relation between the causal profile of the prop-
erty and the persistence conditions of the statue. The kind of interest
people have in statues leads them to count certain changes in them,
e.g., the replacement of some of the matter inside their boundaries,
as ones the statue survives, and others, e.g., the compression of the
same matter into an altogether different shape, as amounting to its
destruction. And here – as is the case generally with artifacts – what
people’s interests lead them to count as persistence is partly consti-
tutive of the persistence.
In the case of the piece of clay, the microphysical states of affairs
that are strongly embedded are of course those whose causal pro-
files are such as to make it a piece of clay and, what goes with this,
to make it have the persistence conditions of a piece of clay. How can
a microphysical state of affairs do this? It does it by having a causal
profile that makes it part of a stage that is a stage in the career of a
piece of clay (other microphysical states of affairs that are parts of
this stage will make it also a stage in the career of a statue). Such a
stage must be apt to generate, or contribute to the generation of,
other stages that stand in the diachronic unity relation to it. Presum-
ably the persistence conditions for a piece of clay require that it be
composed of the same clay (or mostly the same clay) throughout its
career, and at each time during that career this clay fills a continuous
portion of space, and that the career be spatiotemporally continu-
ous. Or something like that. The microphysical state of affairs that
realizes a piece of clay must be such as to guarantee that the object in

which it occurs is made of clay, and it must be such as to generate, or
contribute to the generation of, subsequent thing-stages containing
microphysical states of affairs of the same sort. Of course, the full
story would have to be more complicated than this, and would have
to take into account the fact that the persistence conditions of such
things as pieces of clay are somewhat vague.
IX
It is natural to say that the piece of clay constitutes the statue, and
that the body, or at any rate the biological animal, constitutes the
person. The account I have given of coincidence suggests an account
of constitution. As a first pass, let’s say that object A constitutes
object B if A and B are coincident and the states of affairs that are
strongly embedded in the career of B, and so realize its thick prop-
erties, are only weakly embedded in the career of A.
This faces a difficulty. I have taken it that when coincident objects
are of different kinds they have different persistence conditions,
and that something’s being of a kind and having the persistence
conditions that go with that kind is a thick property of it. This means
that each member of a pair of coincident objects has a thick prop-
erty the other lacks and is realized by a microphysical state of affairs
that is strongly embedded in its career and only weakly embedded
in the career of the other. So on the proposed account of constitution
each member of a pair of coincident objects would constitute the
other. While the piece of clay would constitute the statue, it would
also be the case that the statue constitutes the piece of clay. Simi-
larly, the person would constitute, as well as being constituted by,
the body and the biological animal. Yet we presumably want the
constitution relation to be asymmetrical.
If we could be sure that the objects we want to count as constitut-
ing other objects have no thick properties other than sortal proper-
ties that they don’t share with the objects they constitute, then we
could modify the account so as to give us the asymmetry we want:
A will constitute B just in case A and B are coincident and B has
thick properties not shared by A that are realized by microphysical
states of affairs strongly embedded in B’s career and only weakly
embedded in A’s, and A has no thick properties other than sortal

properties that are not shared by B. But it would seem that some of
the objects we want to count as constituting other objects have thick
properties that are not sortal properties and are not shared by the
things they constitute. I have already mentioned that the modal
compositional properties of a piece of clay count as thick properties
of it. Another example is found in the biological animals that consti-
tute persons. The same biological predicates can be ascribed both to
persons and to biologically individuated animals, but it is arguable
that the properties are somewhat different in the two cases, and that
persons have the biological properties they have in virtue of being
coincident with biological animals having the corresponding bio-
logical properties. On a psychological account of personal identity,
a biological property like being anemic could be lost by a person
who acquires a new body in a cerebrum transplant while the same-
named property would not be lost by the biological animal left
behind as a human vegetable. This implies that the causal profile of
the property of the person and that of the property of the biological
animal are somewhat different, making these different properties. If
it is the case that the person’s being anemic is just a matter of his
being coincident with a biological animal that is anemic, this would
suggest that we have here a case in which a microphysical state of
affairs is strongly embedded in the career of the biological animal
and only weakly embedded in the career of the person. Which on
the proposed account has the undesired result that the biological
animal does not constitute the person.
I have noted that where As constitute Bs it is true both that the
microphysical states of affairs that realize the sortal property of
being a B are weakly embedded in the careers of As and that the
microphysical states of affairs that realize the sortal property of
being an A are weakly embedded in the careers of Bs. But there is
the following asymmetry. Where a particular A constitutes a B, the
B could go out of existence simply in virtue of the A’s career under-
going a change consisting in its ceasing to embed the state of affairs
that is the microphysical realizer of the property of being a B, while
it is not true that the A could go out of existence simply in virtue of
the B’s career undergoing a change consisting in its ceasing to
embed the state of affairs that is the microphysical realizer of the
property of being an A. A person can go out of existence simply in
virtue of the continuing career of its coincident biological animal

ceasing to embed the microphysical realizer of the property of being
a person, and a statue can go out of existence simply in virtue of the
continuing career of its coincident piece of clay ceasing to embed
the microphysical realizer of the property of being a statue. But a
biological animal cannot go out of existence simply in virtue of
the continuing career of its coincident person ceasing to embed the
microphysical realizer of the property of being a biological animal;
and a piece of clay cannot go out of existence simply in virtue of
the continuing career of its coincident statue ceasing to embed the
microphysical realizer of the property of being a piece of clay. This
underlies the truth of Frederick Doepke’s observation that “if x con-
stitutes y at time t, then the fact that y could perish at t is explainable
by describing a change, beginning at t, in which x would be the
substratum,” and that (still in Doepke’s words), “if x constitutes y
then y exists because x accidentally has a certain property.”7
It is central to this account that the microphysical states of affairs

involved in microphysical realization are shared by the coincident
objects. While a microphysical state of affairs may be strongly
embedded in the career of one of a pair of coincident objects and not
in the career of the other, it will be weakly embedded in the careers
of both if it occurs in the career of either. This enables us to reply to
one objection sometimes raised against the notion of coincidence.
Suppose that a statue weighs one hundred pounds. Its coincident
piece of clay, which on the coincidence view is numerically differ-
ent, will also weigh one hundred pounds. So their combined weight
will be two hundred pounds. This seems to generate a false predic-
tion about what will happen when this pair of objects is placed on a
scale, namely that the scale will register two hundred pounds. Simi-
lar difficulties can be raised about electric charge; it would seem
that the combined charge of a pair of objects should be greater than
the charge of either one taken separately, but this not what our
instruments record.
7
Doepke (1982), p. 54 and p. 57.
One reply to the weight objection rests on the claim that the coin-
cident objects are composed of the same matter. What determines
the reading on the scale is the weight of the matter placed on it, and
in the example this is one hundred pounds, not two hundred
pounds. I have no objection to this reply, but I doubt if it can be
applied in the case of electric charge.
I think that we can give a better reply by appealing to the fact that
the microphysical states of affairs in the careers of the objects are
identical. The influence of the piece of clay’s presence on the scale is
due to the causal profile of a microphysical state of affairs it con-
tains; and since the very same microphysical state of affairs occurs
in the statue, the influence of the statue’s presence on the scale is
already included in the influence of the piece of clay’s presence. The
same point applies to electric charge; there is a single microphysical
state of affairs that determines the effects of both coincident objects
on our instruments, and its occurring in two objects shouldn’t be
expected to double its effects.
XI
Perhaps the most frequently raised objection to the view that there
are coincident objects is that given that such objects share the same
matter, arranged in the same way, it is impossible to explain how
they can be of different sorts and have different persistence condi-
tions and different modal properties. Michael Burke asks “Given
the qualitative identity of these objects, what explains their alleged
difference in sort?”8 Addressing Alan Gibbard’s Lumpl and Goli-
ath, a supposedly coincident lump and statue that come into exist-
ence and go out of existence at the same time, Karen Bennett asks
“What grounds the difference between Lumpl and Goliath, given
that they are otherwise so alike? They are the same shape, the
same size, made of the same parts, have the same history and
future, are the same distance from the bagel store, and so on and so
forth. So what exactly makes it the case that they could have differ-
ent shapes and sizes, etc.?”9 And Eric Olson poses what he calls
8
Burke (1992), pp. 14–15.
9
Bennett (2004), pp. 339–40. Gibbard (1974). Bennett is addressing here only the
case of coincident objects that come into existence and go out of existence at the same
time. But she clearly does not think that the problem arises only for such cases.
the “indiscernibility problem” as follows: “By definition materi-

ally coinciding objects are made up of exactly the same particles,
arranged in exactly the same way, in identical surroundings…
How could they have the qualitative differences that constitution-
alists say they have?”10
This objection, or one aspect of it, was briefly addressed in section
IV as the “too many minds” or “too many thinkers” problem. My
response to it there invoked the distinction between thick and thin
properties. Let me expand on that response.
As we have seen, the careers of two coincident objects weakly
embed the same microphysical states of affairs, and in virtue of this
the objects have the same thin MSE properties, and in virtue of
this they have the same thin properties. (We can take it that all thin
properties of physical things are realized by thin MSE properties.)
And the properties they share include the realizers2 of all of the
thick properties possessed by either of them. This is the extent to
which they are alike in virtue of being composed of the same mat-
ter. The microphysical realizers of some of these thick properties,
while weakly embedded in the careers of both coincident objects,
are strongly embedded in only one of them – some in one, some in
the other. It is this that accounts for how the objects differ in their
modal properties, their sortal properties, and their persistence con-
ditions. As I have pointed out (section II), there is an internal rela-
tion between the causal profiles of properties – especially of thick
properties – and the persistence conditions of the things that have
them. The causal features in the causal profile of a property include
the aptness of instances of the property to contribute to the genera-
tion of successor states of certain sorts later in the career of the same
object, and the exercise of such aptnesses is partially constitutive of
the persistence of the object. What we have in a case of coincidence
is the instantiation in the same volume of space, realized by differ-
ent arrangements of the micro-particles located in that volume of
space, of properties whose causal profiles constitutively require
that their possessors have different persistence conditions. The dif-
ference in persistence conditions brings with it a difference in sortal
properties and a difference in modal properties. The instances of
10
Olson (2001), p. 339. Note that Olson takes the constitutionalists to hold that
coincident objects have a qualitative difference while Burke takes it to be agreed that
they are qualitatively identical. I think this is only a terminological difference.
these properties will be synchronically unified with the instances of

the thin properties instantiated in that volume of space, so the pos-
sessors of the thick properties will share the same thin properties,
and will be coincident objects.
At bottom, then, the source of the difference in modal properties,
sortal properties, and persistence conditions lies in differences in
the causal profiles of properties instantiated in the same place.
There can’t be instantiations of these properties without there being
things that have them, and because of these differences their pos-
sessors must be different. It should not be surprising that differ-
ences in causal profiles account for differences in modal properties,
since a causal profile is at the same time a modal profile – what
causal profile a property has is partly a matter of what effects it can
contribute to causing, and what sorts of things it can be caused by.
It should be noted that it is not only “constitutionalists” who
require an explanation of how things have the modal properties,
sortal properties, and persistence conditions they do. It often seems
assumed in the literature that if we reject the possibility of coinci-
dent objects we have a ready explanation of this – somehow the
physical makeup, or qualitative character, of an object determines its
modal and sortal properties and its persistence conditions. But how
does it do so? That is a question any theorist of these matters, not just
constitutionalists, must address. It will have to be agreed that there
are different sorts of things that differ in their modal properties and
their persistence conditions, and differ in what kinds of properties
they can have. Somehow the nature of the properties must enter into
the explanation of this. If the explanation is in terms of the causal
profiles of the properties, we get the sort of account I have offered.
Olson grants that my account of properties, and their relation to
identity conditions, provides an explanation (not one he accepts) of
the difference between persons and human animals, but he says that
even if it is right “it offers no solution to the indiscernibility problem.
Like the others, it must begin with the assumption that Person is a
person (or thinking thing) and not an animal while Animal is an ani-
mal and not a person. Without this starting point the account has
nothing to work on. But it offers no explanation of the difference.
Shoemaker too must deny that properties of and relations among a
thing’s parts, even in conjunction with their surroundings, explain,
entail, or even cause its higher-level properties, such as its ability or
inability to think, or its being an animal or an organism” (p. 352).

(Here “Person” and “Animal” name a pair of coincident things which
a constitutionalist must claim are distinct.) It should be apparent
from what I have said that my account does not begin with the
assumption that Olson says it must begin with. It begins, rather, with
the assumption that where Person and Animal are there is a set of
microphysical states of affairs that realize the properties Person and
Animal have. It is then argued that given the nature of these proper-
ties there must be two different objects there, one having properties
distinctive of persons and one having properties distinctive of bio-
logical animals. Having established that, we can if we want give the
name “Person” to one of them and the name “Animal” to the other.
XII
What I have said so far may have left the impression that on my
view the coincidence of two objects requires that they be of differ-
ent sorts and have different persistence conditions. That is, in fact,
a common view among believers in coincidence; citing Locke they
hold that it is impossible for two objects of the same kind to share
their matter and occupy the same place at the same time. But that
is not my view. I think it is possible for there to be coincident objects
that are of the same kind and have the same persistence conditions.
We need to see how this can be.
Kit Fine has offered one simple example of such intra-kind super-
venience.11 Someone writes a note on a piece of paper. The recipient
turns it over and writes a reply on the other side. There are two
notes, but just one piece of paper. The notes coincide with the piece
of paper, but they also coincide with each other.
In several place I have offered a different example, in which the
coincident objects are both persons.12 The case is similar to that of
Dr. Jekyll and Mr. Hyde. (I got it from Eric Olson, who puts a differ-
ent interpretation on it.13) We are to imagine that two persons
11
Fine (2000).
12
See my 2003 and my 2007.
13
Olson presented his example at a conference on the Self at the University of
Arkansas in September 1999. What I take to be a case of coincident persons he takes
to be a case of one person with a highly unusual history.
alternate in controlling a single body, each having control of it for the

period of a day. While one of them is conscious, and manifesting his
mental states in the behavior of the body and responding to the
effects of its surroundings on its sense organs, the other sleeps. But
both of them have mental states – different mental states – at all times.
The one who is asleep has beliefs, desires, intentions, memories, etc.
in the way a normal person has such states while asleep. We can sup-
pose that their mental states are realized in different hemispheres of
their shared brain, but this is not essential. The states of each form a
career in which the successive states are causally connected in such a
way as to exhibit diachronic unity, but the total series of states occur-
ring in the body over an extended interval does not form such a
career, since there are no relations of “immanent” causal relations
between the states of one of the persons and those of the other.
Here the coincident objects are of the same kind – both are per-
sons – and have the same persistence conditions. What makes it the
case that there are two persons associated with the body is that two
sets of thick properties are realized by microphysical states of affairs
occurring in it, where each of those sets is unified by the synchronic
and diachronic unity relations for persons, while the total set is not
so unified. The microphysical states of affairs that are strongly
embedded in the career of each of the persons are weakly embed-
ded in the careers of both, and in virtue of this the persons share the
same thin properties.
XIII
I suspect that one reason for the skepticism many philosophers feel
about the idea that there are coincident objects is the paucity of
plausible examples of coincidence. The most frequently cited exam-
ples are the two I have discussed – the case of the statue and the
piece of clay, and the case of persons and their coincident bodies
and biological animals. Some believers in coincidence cannot avail
themselves of the latter example because they deny that persons
and their bodies are distinct; thus Judith Thomson, a believer in
coincidence, says ”I take cats and people to be their bodies.”14
14
Thomson (1998), p. 169.
It is natural to say that where there is a statue there is a piece of clay

(or of bronze, etc.) that constitutes it, and that where there is a person
there is a human body that constitutes it. But for most ordinary lan-
guage sortal terms S there is not an idiomatic completion of “Where
there is an S there is a . . . that constitutes it.” There is no ordinary lan-
guage sortal term with which we can complete “Where there is an
automobile, there is a…that constitutes it.” And the same is true if for
“automobile” we substitute “house,” “computer,” “river,” or “tree.”
To be sure, we can complete such statements with the likes of “portion
of matter.” But portions of matter, or quantities of matter, or collec-
tions of molecules are not natural candidates for being kinds of objects.
We would be hard pressed to say what, other than being a portion of
matter, is a thick property of a portion of matter. I suppose that a por-
tion of matter can be said to have the shape, size, and mass of the object
that is composed of it, but speaking this way can seem artificial.
Animals other than human beings can be said to have bodies, and
there are, I think, the same reasons for denying that dogs or horses are
identical with their bodies as there are for denying that persons
are identical with their bodies.15 But we don’t speak of trees, or other
plants, as having bodies. Just why this is so is difficult to say. It must
have to do with the ways these different sorts of entities are related to
our interests, and in particular to the interest we have, in the different
cases, in the distinction between being dead and being alive. It must
be for similar reasons that we have no colloquial counterparts of
“human body” in the case of houses, computers, and rivers.
I think that portions or quantities of matter are entities in good stand-
ing, and that there being these makes coincidence ubiquitous. It is
where it is most interesting that there are the resources in our language
for designating the coincident entities. But it should be noted that even
there we need a term of art, “biological animal” (animal individuated
biologically), to designate one of the entities persons are coincident
with, given that there is a good sense in which persons are animals.
XIV
To sum up, I believe that there are cases of coinciding objects and
that such cases are common. I have restricted my attention to how
15
See Unger (2000).
there can be coincidence in a physicalist world – a world in which all

property instances are physically realized. The key to its possibility
is the distinction between properties that are thin, those that must be
shared by coincident objects, and properties that are thick, those
with respect to which coincident objects can differ. And the key to
how that distinction applies to physicalist worlds is the distinction
between two ways in which a microphysical state of affairs can be
embedded in the career of an object, namely what I have called weak
embedding and strong embedding, together with the role of the
causal profiles of properties in determining whether a state of affairs
realizing a property instance is strongly embedded in the career of
an object or only weakly embedded in it. A state of affairs that is
weakly embedded in an object’s career will be weakly embedded in
the career of any object coincident with that object, while such a state
of affairs can be strongly embedded in the career of an object with-
out being strongly embedded in the career of an object coincident
with it. This difference has to do with the fact that strong embedding
requires that the state of affairs be suited for playing a role in imple-
menting the persistence conditions for a particular sort of object,
while weak embedding leaves it open what the persistence conditions
are for the objects in whose career they are embedded. Thin proper-
ties are realized by weakly embedded microphysical states of affairs,
while thick properties are realized by strongly embedded micro-
physical states of affairs. The account solves the “too many minds”
problem by saying that mental properties are thick, and so can
belong to an object of a certain sort, a person, without belonging to
objects of other sorts, bodies or biologically individuated animals,
that are coincident with it. And it explains how things made of the
same matter, identically arranged, can differ in their modal and sor-
tal properties, and in their persistence conditions.16
Cornell University
REFERENCES
Bennett, K. (2004). “Spatio-Temporal Coincidence and the Grounding
Problem.” Philosophical Studies, 118: 339–71.
16
Thanks to Karen Bennett and Carl Ginet for comments on an earlier draft.
Burke, M. (1992). “Copper Statues and Pieces of Copper: A Challenge to

the Standard Account.” Analysis, 52: 12–17.
Fine, K. (2000). “A Counterexample to Locke’s Thesis.” The Monist, 84:
357–61.
Gibbard, A. (1975). “Contingent Identity.” Journal of Philosophical Logic, 4:
187–221.
Olson, E. (1997). The Human Animal – Personal Identity without Psychology.
Oxford: Oxford University Press.
—— (2001). “Material Coincidence and the Indiscernibility Problem.”
The Philosophical Quarterly, 51: 337–55.
Shoemaker, S. (1963). Self-Knowledge and Self-Identity. Ithaca: Cornell Uni-
versity Press.
—— (1979). “Identity, Properties, and Causality.” Midwest Studies in Phi-
losophy, 4.
—— (1999). “Self, Body, and Coincidence.” Proceedings of the Aristotelian
Society, Supplementary Volume, 73: 287–306.
—— (2003). “Realization, Micro-Realization, and Coincidence.” Philoso-
phy and Phenomenological Research, 67: 1–23.
—— (2007). Physical Realization. Oxford: Oxford University Press.
—— (forthcoming). “Physical Realization Without Preemption.” In a vol-
ume, to be published by Oxford University Press, on the ontology of
the mental causation debate.
Thomson, J. (1998). “The Statue and the Clay.” Nous, 32: 148–73.
Unger, P. (2000). “The Survival of the Sentient.” Nous, 34: 325–48.
V
THE OPEN FUTURE

9. The Real Truth about the
Unreal Future*
Rachael Briggs and Graeme A. Forbes
Recent times have been very much focussed on the future. The elec-
tion of Barack Obama in America was accompanied by a wave of
optimism. The Global Financial Crisis and the challenge of climate
change caused many to descend into pessimism. It is not whether
our glass is half-empty or half-full that we worry about, but whether
it will be empty or full. As philosophers, naturally, we want to illu-
minate such concerns, so that we might understand them better
and subject them to rational scrutiny.
According to the Growing-Block view of time, most famously
put forward by C.D. Broad (1923), the flow of time consists in events
coming into existence, so that past things and events are real, while
future things and events are not. While the Growing-Block view is
often considered intuitively appealing, some are concerned it has
little to say about the future it denies the existence of. We show how
Growing-Block theorists can assign meaningful, mind-independent
truth conditions to sentences about the future.
1. INTRODUCING THE GROWING-BLOCK THEORY
Asymmetries between past and future abound. The past, many of

us think, is fixed and determinate; the future is open and indetermi-
nate. The arrows of time and causation point from past to future,
not from future to past.
The Growing-Block view explains the difference between the
fixed past and the open future in terms of ontological commitment:
* Thanks to Jamin Asay, David Braddon-Mitchell, Mark Colyvan, Alison Fern-

andes, Patrick Greenough, Dominic Hyde, Jenann Ismael, Mark Jago, Rosanna Keefe,
Kristie Miller, Jonathan Payne, Huw Price, Lionel Shapiro, Nicholas J.J. Smith, Agus-
tín Rayo, and Richmond Thomason for their helpful questions and comments.
258 | Rachael Briggs and Graeme A. Forbes
the view is committed to the (tenseless) existence of past objects

and events, but not to the (tenseless) existence of future objects or
events. It treats the present like the past, not like the future:
present things are the last things the Growing-Block theory takes
seriously.
The commitments of the Growing-Block theory change—or more
precisely, increase—as time passes. Though the Growing-Block the-
ory takes only the past and present seriously, it holds that in time,
there will be more to the past and present, because more will have
happened, even though the events that will happen are not ones to
which the Growing-Block view is (tenselessly) committed.
Claims about the future create a puzzle for the Growing-Block
theory. We ordinarily think that there are truths about the future—
for instance, it is true that there will be a lunar eclipse on January 21,
2019. (Astronomers know that there will be a lunar eclipse on Janu-
ary 21, 2019, and what is known is surely true.) But if the future is
unreal, then all claims about the future would seem to be trivially
false.
It will not help to stipulate that all sentences about the future are
trivially true or trivially truth valueless, instead of trivially false.
The proposition that there will be a lunar eclipse on January 21,
2019 is true, and the proposition that the moon will be replaced
with a hunk of green cheese on January 21, 2019 is false. Everyone—
Growing-Block theorists included—should distinguish between
truths about the future and falsehoods about the future.
We claim that although Growing-Block theorists must deny the
reality of the future, they can still countenance non-trivial truths
and falsehoods about future events. In the remainder of part 1, we
develop a metaphysical and semantic framework for discussing
modality in the context of the Growing-Block theory. This stage lays
the groundwork for an account of future possibility, or feasibility,
which can in turn be used to define a concept of future truth.
In part 2, we propose three ways in which Growing-Block theo-
rists might deploy our modal framework to give semantics for sen-
tences about the future. Each proposal is associated with a different
non-classical logic. Rather than choose among these proposals, we
state the advantages and disadvantages of each, and leave the
reader to decide.
The Real Truth about the Unreal Future | 259
In part 3, we respond to two general problems, having to do with

expressive power, that face three all of our semantic proposals.
Finally, in part 4, we explain the advantages of the Growing-Block
theory over similar theories proposed by Craig Bourne, Storrs
McCall, and John MacFarlane.
Before we continue, one word about our starting point is in order.
We will treat propositions as tensed—as having truth values rela-
tive to a time and to a possible world. We will use P and F as past
and future operators on tensed sentences. These are to be read ‘At
some past time’ and ‘At some future time’ respectively. This clarifi-
cation out of the way, we are ready to develop the Growing-Block
theory in more detail.
1.1 Times (Genuine and Ersatz)

Our first task is to develop an ontology of possible worlds suited to
the purposes of the Growing-Block theory. This ontology is a type
of linguistic ersatzism: it presupposes that possible worlds are not
concrete hunks of matter like the actual world, but descriptions of
the world in a logically idealized language.
We follow the approach of Craig Bourne (2006, 52– 4), who builds
ersatz worlds from ersatz times. Just as an ersatz world is a com-
plete representation of a world from a tenseless perspective, an
ersatz time is a complete representation of a single time in a world.
Each ersatz time is a maximal consistent set of tensed propositions
which do not themselves contain any tense operators. (Ultimately,
we will develop a way of assigning truth values to sentences con-
taining tense operators—so the propositions contained in an ersatz
time comprise only a proper subset of the propositions which are
true at that time.) We take propositions to be interpreted sentences
in an idealized logical language, and we will switch freely between
talk of propositions and talk of sentences.
We will make some further assumptions about the structure of
ersatz times. We assume that each ersatz time e is associated with a
vocabulary of names De, which (in a slight abuse of the use–mention
distinction) we will call a ‘domain’. The domains of different ersatz
times may overlap partially, completely, or not at all. We assume
that if e contains a sentence of the form ∃xϕ(x), then it also contains
a sentence of the form ϕ(a) for some a ∈ De. (Both these assumptions
are claims about what it is for e to be a complete representation of a
time: e must specify exactly which things exist; and if it says that
something satisfies a particular open sentence, then it also specifies
which thing satisfies that open sentence.) Finally, we assume a ‘one
name, one object’ rule: each name has at most one referent, and
refers to the same object across times and possible worlds, while
each object has at most one name.
Since they are abstract representations, ersatz times contain no
tables and chairs, donkeys and cats, or quarks and electrons. None-
theless, they can be instantiated by concrete times: maximal simul-
taneous hunks of spatially connected stuff that do include tables,
chairs, donkeys, cats, quarks, and electrons as parts. For a concrete
time t to instantiate an ersatz time e, we claim, is simply for e to be
a complete and true description of t. The concrete present instanti-
ates one ersatz time; various concrete past times have instantiated
other ersatz times. Still other ersatz times are not instantiated, and
never have been.
The Growing-Block theorist should claim that concrete times
(past and present) are ordered from earlier to later. We will assume
that this ordering relation is total, transitive, irreflexive, and asym-
metric, and that it has all four properties necessarily. Following
Bourne (2006, 53–5), we can represent possible timelines as sets of
ersatz times ordered by a total, transitive, irreflexive, and asymmet-
ric relation R.1, 2 (The same ersatz time may appear more than once
in the ordering.)
Having defined timelines, we can now explain what it is for a
timeline to be instantiated. The whole of reality instantiates a time-
1
Perceptive readers will note that we have just ruled out a number of apparent
possibilities, including closed time-like curves and branching concrete universes.
We are not at all convinced these apparent possibilities are genuine possibilities. We
believe that our theory could be expanded to accommodate these scenarios, but to do
so here would require adding to an already long and complex paper.
2
In addition to the ordering relation, one might think that there are meaningful
distance relations between ersatz times. Given two times such that one is later than
the other, it makes sense to ask how much later—an hour, a minute, a year? We will
assume that facts about the distances between times supervene on the facts about
which concrete times there are and how those times are ordered, and do not need to
be added to possible timelines as extra information.
line T just in case there is a function f mapping the ersatz times in T

onto concrete times such that
i. for every concrete time t, there is some e ∈ T such that f(e) = t,
ii. for every e ∈ T, f(e) concretely instantiates e, and
iii. for every e, e’ ∈ T, eRe’ iff f(e) is earlier than f(e’).
Timelines, then, are sequences of ersatz times. They are possible
world histories, or put another way, they are ersatz possible worlds.
The Growing-Block theorist may want to place additional restric-
tions on which timelines represent genuine metaphysical possibili-
ties. For instance, perhaps the properties of objects must change in
a more-or-less continuous fashion, so that no possible timeline can
represent a person as having one head at one time and three heads
at a time immediately afterward. (Note that this continuity con-
straint is available, but not mandatory.) Only a subset of the possi-
ble timelines will be physically possible, biologically possible, and
so forth. Saying exactly which timelines are possible in which ways
is a serious philosophical problem, but it is not a problem peculiar
to the Growing-Block view. We will not address it here, except to
make explicit our assumption that what is possible according to T
supervenes on the ersatz times T contains, together with the order-
ing of those times.
1.2 Truth (Relative and Absolute)

Now that we have developed an ontology of possible worlds, our
next step is to define truth at a possible world. Or rather, since we
are working with tensed propositions, to define truth at a time at a
possible world.
A complete account of truth will require a thorough discussion of
how to treat the nonexistent future, and we will not be in a position
to give such an account until part 2. In the meantime, though, we
can give the beginnings of a definition of truth at a time in a time-
line. These beginnings will help us to clarify the question of how to
treat the nonexistent future, which in turn will help us to further
clarify the concept of truth.
We will write veT(ϕ) for the truth value of ϕ according to e in T.
Defining truth for atomic propositions is straightforward.
Where p is an atomic proposition,

1 if p ∈ e
veT ( p) =
0 if p ∉ e
The definitions of truth-functional connectives are similarly
straightforward.
1 if veT (f ) = 0
veT (¬f ) =
0 if veT (f ) = 1
1 if veT (f ) = 1 and veT (y ) = 1

veT (f ∧y ) =
0 if veT (f ) = 0 or veT (y ) = 0
1 if veT (f ) = 1 and veT (y ) = 1

veT (f ∨y ) =
0 if veT (f ) = 0 or veT (y ) = 0
We assume that the material conditional ϕ ⊃ ψ is equivalent to the

disjunction ¬ϕ ∨ ψ, at least until section 2.3, where we present an
alternative definition of the material conditional. We will assume
that ϕ ≡ ψ is equivalent to the conjunction (ϕ ⊃ ψ) ∧ (ψ ⊃ ϕ).
We can define tensed quantifiers ∃ and ∀, which range (at e in T)
over all and only the objects whose names belong to De.
1 if for some a ∈De , such that veT (f x / a) = 1

veT (∃xf x ) =
0 if for all a ∈De , veT (f x / a) = 0
1 if for all a ∈ De , veT (f x / a) = 1

veT (∀xf x) =
0 if for some a ∈ De , veT (f x / a) = 0
We can also partially define tenseless quantifiers Σ and П, which

range over everything there was, is, or will be.
1 if for some e’ in T , there is an a ∈De ’

veT Σx(f x ) =
such that ve ’T (f x / a) = 1
0 if for some e’ in T , there is an a ∈ De ’

veT Πx(f x ) =
such that ve ’T (f x / a) = 0
We have not fully defined Σ or П, because we have not yet estab-

lished a way of assigning truth values to statements about the
future. Where there is no object x such that ϕx, it may be the case
that there will be an object x such that ϕx (once more times have
become instantiated by the Growing-Block), or it may not. In order
to settle the matter, we will need to say more about the future.
We can also fully define the tense operator P, and partially define
the tense operator F.
1 if T contains an e’ such that e’Re and ve ’T (f ) = 1

veT (Pf ) =
0 else
veT (Ff ) = 1 if T contains an e’ such that eRe’ and ve ’T (f ) = 1
F is only partially defined for the same reason that Σ and П are only
partially defined. We’ve said nothing yet about how to ascertain
veT(Fϕ) when T fails to contain an e’ such that eRe’ and ve’T(ϕ) = 1. We
do not want to say that veT(Fϕ) is invariably 0 under these circum-
stances, because we want to allow for the possibility that ϕ will
become true in the future that does not yet exist.
We can introduce an additional class of tense operators—one for
each temporal distance d. (We will write the distance from e to e’ as
positive where eRe’, and as negative where e’Re.)
1 if T contains an e’ such that e’Re ,

the distance from e’ to e is d ,
veT (Pdf ) =
and ve ’T (f ) = 1
0 else
1 if T contains an e’ such that eRe’,

the distance from e to e’ is d ,
and ve ’T (f ) = 1
veT ( Fdf ) =
0 if T contains an e’ such that eRe’,
the distance from e to e is d,
and ve ’T (f ) = 0
Fd is only partially defined, for the same reason that F was only
partially defined.
Since Σ, П, F, and the Fd operators are only partially defined, the
valuation function as a whole is only partially defined. The fragment
of the language containing only atomic sentences, ∧, ∨, ¬, the quanti-

fiers, and P behaves classically. But the valuation function as a whole
is non-classical. We will discuss three different (non-classical) ways
of extending the valuation function in sections 2.1–2.3.
In addition to the concept of truth at a time in a timeline (which we
have yet to fully define), we can define a number of less relativized
types of truth: truth at a time, truth in a timeline, and truth, full stop.
There is one timeline that represents the world as it really is—that is,
it correctly represents all the times that concretely exist, and it does
not represent the world as containing any times that don’t concretely
exist. This is the actualized timeline. We can say that what is true at a
(concrete) time t is simply what is true according to the ersatz time e
in the actualized timeline (an ordered set of ersatz times) instantiated
by a concrete time t in the actual world (a temporally ordered hunk
of concrete future stuff).
To define truth at a timeline, we will need the concept of an absolute
present. Many theorists of time treat ‘present’ as an indexical, so that
a token of ‘it is raining at present’ is true just in case it is raining at the
time when the sentence is uttered. But the Growing-Block theorist can
also pick out a non-indexical or absolute present: the last time in exist-
ence. Furthermore, each timeline has its own (ersatz) absolute present:
the last time in it. The ersatz absolute present in a timeline T would, if
T were actualized, be instantiated by the concrete absolute present.
What is true at a timeline T is simply what is true at T’s absolute
present. Some complete timelines will have no absolute present,
since they will have no last time. There is nothing that is true accord-
ing to these timelines (though there are still propositions that are
true at times in them).
What is true full stop is what is true at the absolute present in the
actualized timeline. That is to say, what is true full stop is what is
true of the concrete hunks of stuff that exist at the current stage of
the Growing-Block. We have four concepts of truth in all, corre-
sponding to cells in the following table.
Timeline
Relative Absolute
Relative Truth at e in T Truth at t
Time
Absolute Truth in T Truth
The cells on the left-hand side of the table correspond to types of

truth that are wholly relative to representations, while the cells on
the right-hand side correspond to types of truth that are somehow
grounded in concrete existence. The cells in the top row correspond
to tensed types of truth, while the cells in the bottom row corre-
spond to tenseless types of truth. (Notice that we have used a sin-
gle, tensed type of proposition throughout, but that our tensed
propositions admit of both tensed and tenseless truth.)
Given our fourfold definition of truth, how shall we define valid-
ity? A valid sentence is necessarily true, while a valid argument is
one that is necessarily truth-preserving. We need a univocal way of
understanding ‘true’ and ‘truth’ (and also a univocal way of under-
standing ‘necessarily’).
Which sentences and arguments are valid should not depend on
what concretely exists. Rather, questions of validity must be evalu-
ated by considering all possibilities—both those that are concretely
instantiated and those that are not. Possibilities, in our framework,
are timelines. Therefore, we should define a valid argument as one
that preserves one of the timeline-relative types of truth on the left-
hand side of the matrix: either truth at e in T or truth in T. We have
two appealing conceptions of validity:
Validity eT Φ ⊧ Ψ just in case, for every time e and

timeline T, if veT(ϕ) = 1 for all ϕ ∈ Φ, then
veT(ψ) = 1 for at least one ψ ∈ Ψ.
Validity T Φ ⊢T Ψ just in case, for every time e and
timeline T such that e is the absolute present
in T, if veT(ϕ) = 1 for all ϕ ∈ Φ, then veT(ψ) = 1
for at least one ψ ∈ Ψ.
Validity eT is the stronger notion: an argument or sentence that is

valid eT must also be valid T, but the converse does not hold.3
We will take validity to be validity eT, rather than the weaker
validity T.4 Likewise, we will take necessity to be truth at every time
in every timeline—not just truth in every timeline.
3
See Section 3.1.
4
Still other interpretations of validity are possible in the supervaluationist frame-
work we develop in section 2.1; see (Varzi 2007) for a detailed discussion.
The distinctions among the four relativized types of truth can be

brought to bear on a recent objection to the Growing-Block view.
We discuss this objection and our solution to it in the next section.
1.3 How Do We Know it is Now Now?

Craig Bourne (2006) and David Braddon-Mitchell (2004) have inde-
pendently raised the concern that, if there are two interpretations of
‘the present’ on the Growing-Block view (viz. the indexical and the
absolute), then it is possible that, say, Marie Curie might ask if, in
addition to being at the indexical present (i.e. being at the time she
is at) she is at the absolute present (i.e. at the time succeeded by
nothing). If we accept that Marie Curie could ask such a question,
despite not being at the absolute present, then we must explain why
Marie Curie would be wrong to think she is at the absolute present,
while we are right to think we are at the absolute present.
One solution, put forward by Peter Forrest, is to deny the possi-
bility of Marie Curie asking such a question, so long as she is located
in the past. According to him “the past, although real, is lifeless and
(a fortiori?) lacking in sentience” (Forrest 2004, 358). More gener-
ally, “Life and sentience are . . . activities not states. Activities only
occur on the boundary of reality, while states can be in the past”
(Forrest 2004, 359). For Forrest, the distinction between the living
present and the dead past is grounded in their different causal
properties—activity requires a ‘causal frisson’ that can occur only at
the absolute present.
We think that the distinction between the living present and the
dead past is better understood in terms of the distinction between
what is true at a past time (i.e. what is true according to the ersatz time
e that corresponds to t in the actualized timeline), and what is simply
true (i.e. true at the absolute present in the actualized timeline). This
distinction is meant to be applied concerning not the content of
Marie Curie’s beliefs, but our attribution of those beliefs to her. It is
true at some past time (e.g. 1903) that Marie Curie falsely believes
she is (absolutely) present, but it is not true (full stop) that she falsely
believes she is (absolutely) present. This is because it is not true (full
stop) that Marie Curie believes anything at all. In the actual world,
at the absolute present, Curie is dead. Dead people lack beliefs.
Whether the concept of absolute truth aligns with any causal con-
cept is a further question, and one that the authors of this essay
disagree about. One of us thinks Forrest’s concept of a causal fris-
son captures something important about the sense in which the
present is ontologically privileged; the other thinks that it is unlikely
to satisfy any of the useful roles carved out for causation in meta-
physics and the philosophy of science.
1.4 Feasibility
So far, timelines have behaved somewhat like the possible worlds
of Robert Merrihew Adams (1974). We should, however, mark sev-
eral important disanalogies between timelines and traditional pos-
sible worlds. First, unlike traditional possible worlds, timelines are
not sets of propositions, but ordered sets of such sets. Second, dis-
cerning what is true at a time (i.e., a set of tensed propositions) in a
timeline, unlike discerning what is true at a traditional possible
world (i.e., a set of tenseless propositions), is not simply a matter of
checking to see which propositions the relevant set contains. Every
proposition contained in a time in a timeline is true at that time in
that timeline, but a proposition may be true (within a timeline) at
times that do not contain it.
Third, unlike traditional possible worlds, timelines stand in inter-
esting parthood relations. In formal terms, we can say that timeline
T is an initial segment of timeline T’ or that T’ is an extension of T, just
in case
every ersatz time in T also occurs in T’,

if eRe’ in T, then eRe’ in T’, and
if eRe’ in T’, then either eRe’ in T, or for all e* ∈ T, e*Re’ in T’.
(We could define final-segment and middle-segment relations
between timelines in a similar way, but the initial-segment relation
is the one that will turn out to be most theoretically useful.)
With the apparatus set forth so far, the Growing-Block theorist is
equipped to start talking about the fixed past and the open future.
The actual world concretely instantiates a timeline—the actualized
timeline. There are a number of physically possible extensions of the
actualized timeline. (Call these feasible timelines.) Feasible timelines,
like the actualized timeline, correctly represent the denizens of the

actual world. Furthermore, like the actualized timeline, they evolve
in a way compatible with the dynamic physical laws that govern
the universe. So in a sense, they accurately represent the world—
that is, everything in the world, they represent accurately. But unlike
the actualized timeline, they represent some things that never hap-
pened. So in another sense, they fail to accurately represent the
world—they represent it as containing things it does not in fact
contain.
Feasible timelines are possible timelines, in some very strong sense
of ‘possible’. For all that the world determines, any feasible timeline
might be actualized in the future, but the world does not determine
which feasible timeline will be actualized in the future.
Feasibility can be understood as a possibility relation among
timelines. ‘It is feasible that’ behaves as a type of possibility opera-
tor, and its dual as a necessity operator. We will assume that the
feasibility relation is transitive (what is feasibly feasible is feasible)
and reflexive (every timeline is feasible relative to itself). Further-
more, feasibility is antisymmetric (no two distinct timelines are fea-
sible relative to each other). The antisymmetry of the feasibility
relation follows from the antisymmetry of the extension relation,
together with the fact that what is feasible relative to a timeline
must be an extension of it.
A set of feasible timelines for a time can be represented as a tree
structure, like the one in Figure 1.
Notice that Figure 1 depicts a great many timelines, since any
initial segment of a timeline is itself a timeline. Some of those time-
lines, which represent particular kinds of possibilities, will play a
special role in the semantics of sentences about the future. We can
make the following distinctions between three types of timelines.
Some timelines end in a way that (together with the laws of
nature) precludes any future happenings; they contain a final ‘big
crunch’, after which all space, time, and matter disappear, or their
laws of nature permit a certain boundless span of time—which is
filled up with events that carry on indefinitely—and nothing more.
Call these timelines complete. Other timelines seem to end abruptly,
with tigers frozen mid-leap, the planets stopped in their orbits,
and human actors posed in lifelike tableau. The laws of nature
require that these timelines go on into the future; call them incom-
Key:
concrete times
ersatz times
earlier later
Figure 1. The Growing-Block and Ersatz Tree
plete. Finally, there may be timelines that the laws of nature per-
mit, but do not require, to extend into the future. Call these
semi-complete.
We will leave open the question of how to distinguish among the
three types of timelines, the question of whether there are any com-
plete timelines, and the question of whether there are any semi-
complete timelines. We will, however, assume that every timeline
has either a feasible complete extension or a feasible semi-complete
extension—that it is possible to tack on enough time to reach an
acceptable stopping point.
The concepts of completeness and semi-completeness are crucial
because they are linked to the concept of an exhaustive description—
one that details everything that ever has happened, is happening, or

will happen. A complete or semi-complete timeline can (in some
appropriate nomological sense) be exhaustive in this sense, while an
incomplete timeline cannot. When it comes to assigning truth values
to sentences about the future, based on a timeline, it will be impor-
tant to know whether the timeline is exhaustive—whether it cata-
logues everything that will be, in addition to everything that is.
The terms ‘complete’ and ‘incomplete’ have logical connotations:
a classically complete valuation function is one that assigns value 0
or 1 to every sentence in some language, while a classically incom-
plete valuation function fails to assign values to some some sen-
tences in the language. We welcome these connotations: it will turn
out that every complete timeline is associated with a classically com-
plete valuation function, incomplete timelines tend to be associated
with classically incomplete valuation functions, and semi-complete
timelines play two roles—one in which they must be associated with
classically complete valuation functions, and one in which they may
be associated with classically incomplete valuation functions.
2. EVALUATING FUTURE CONTINGENTS
Consider the tensed propositions expressed by the following

sentences.
1. Exactly one day into the future, there will be a sea battle.
(FdS)
2. Exactly one day ago, there was a sea battle. (PdS)
3. Exactly one day into the future, it will be the case that:
either there is a sea battle, or there is not. (Fd(S ∨ ¬S))
4. Either there will be a sea battle exactly one day into the
future, or there will be no sea battle exactly one day into
the future. (FdS ∨ Fd(¬S))
5. Either there will be a sea battle exactly one day into the
future, or it is not the case that there will be a sea battle
exactly one day into the future. (FdS ∨ ¬Fd(S))
6. Someone will win tomorrow’s lottery drawing (though not
necessarily anyone in particular). (Fd(ΣxLx))
7. There is someone (in particular) who will win tomorrow’s
lottery drawing. (Σx(FdLx))
How should one ascertain the truth values of these propositions?

Do all of them have truth values? Do any?
One option for the Growing-Block theorist is to claim that, since
all of the sentences except (2) make apparent reference to a future
that does not, strictly speaking, exist, all except (2) are either neces-
sarily false or necessarily truth valueless. We think this option
should be rejected. (1) looks like be the sort of proposition that can
be known, at least under the right circumstances. If I know that the
ships have already been launched on a collision course, and that
there is no nomologically possible way of preventing the battle,
then I appear to be in a position to know (1). But surely what can be
known can be true.
This argument is not unanswerable. Someone who believes that
all the example sentences but (2) are truth valueless might reply
that when one knows the ships have been launched beyond recall,
one’s attitude toward (1) is properly described not as knowledge,
but as justified belief. (There will have to be some further story
about what justified belief amounts to when the proposition
believed is truth valueless, and knowably truth valueless.) But on
the face of it, there appear to be circumstances in which (1) is true,
which can usefully be contrasted with circumstances in which (1) is
false (e.g., circumstances in which no one has any ships, and it is not
physically possible to make any by tomorrow).
So (1) seems to be true at some times in some timelines, and
false at other times in other timelines. Are there also times when
(1) is neither true nor false? The Growing-Block theorist has good
reasons to say “yes”. Sometimes, the past, the present, and the
laws of nature are not enough to settle whether there will be a sea
battle one day into the future.5 But according to the Growing-
Block theory, there is nothing but the past, the present, and the
laws of nature. If something is not settled by the past, the present,
and the laws of nature, then it is not settled. Where nothing in the
world settles whether a sentence is true or false, that sentence
must not have a truth value—there is no truth without some sort
of truthmaking.
5
Even if the actual world is deterministic, we would not want to rule out the pos-
sibility of indeterministic worlds.
Again, our argument is not unanswerable. Roy Sorensen (2001)

and Patrick Greenough (2008) defend the position that all proposi-
tions are either true or false, that some true propositions are not
made true by anything in the world, and that cases of indetermi-
nacy are cases of truthmaker gaps rather than truth value gaps. The
Sorensen–Greenough view has a notable payoff: its proponents can
keep classical logic, while proponents of truth value gaps must give
classical logic up. Still, contingent truths that are not grounded in
reality seem puzzling and spooky. We think they are so spooky that
on balance, it is best to assume that (1) is sometimes truth valueless
(as well as sometimes true and sometimes false). Hardcore adher-
ents of the Sorensen–Greenough approach, however, can re-inter-
pret our discussion of truth and falsity as a discussion of grounded
truth and grounded falsity.
(2), unlike (1), is about past events. The past—no matter what it
was like—seems to be enough to fix the truth value of (2).6 This
suggests that unlike (1), (2) is always either true or false according
to the Growing-Block theory. The possible indeterminacy of (1),
when taken together with the necessary determinacy of (2), is one
way of cashing out the idea that the future is open while the past is
fixed.
Sentences (3)–(5) all appear to be logical consequences of (1). We
will assume that appearances are right, although they could con-
ceivably be disputed. (Disputation is easiest in the case of (3), since
(4) and (5) follow from (1) by tautological consequence alone.) One
might wonder whether in addition to being entailed by (1), (3)–(5)
are logically valid—that is, whether they are entailed by the empty
set. (3) looks valid at first glance—it is simply an instance of the law
of excluded middle embedded in the scope of a future operator.
And isn’t tomorrow bound to obey the law of excluded middle, just
like yesterday and today? On second thought, though, what if there
is no tomorrow? (3) is arguably false at the last moment of the
universe.
6
Here and throughout, we ignore the complications caused by unrelated phenom-
ena such as vagueness and the Liar. A theory that addressed these complications
might replace our ersatz times—assumed to be classical—with ersatz times that were
complete and consistent by the lights of some suitable non-classical logic.
(4) seems to say something stronger than (3)—not just that the
future obeys the law of excluded middle, but that it is determined
by the past and the present, at least as far as sea battles are con-
cerned. For (4) is a disjunction, and each of its disjuncts entails that
the future is determined, at least as far as sea battles are concerned.
One could object to the inference from the claim that each of (4)’s
disjuncts entails that the future is determined to the claim that (4)
itself entails that the future is determined. The supervaluationist
view we discuss in section 2.1 takes this objection seriously. Still,
there is at least a prima facie case for thinking that (4) might not be
true, even when (3) is.
(5) is a classical tautology. This suggests that (5) is valid. But to all
three of the proposals we consider, there are situations where the
sentences FdS and ¬FdS are both untrue (that is, not true—either
false or indeterminate). If (5) is valid, then it is sometimes true with-
out having a true disjunct—a puzzling state of affairs!
(6) says simply that a prize will be given out tomorrow. (7)
appears to say something stronger—that the lottery is a fix. The
relationship between (6) and (7) is closely analogous to the relation-
ship between (3) and (4). (3) says that a particular disjunction will
be true tomorrow, while (4) says that one of its disjuncts will be true
tomorrow. Likewise, (6) says that a particular existentially quanti-
fied statement will be true (of something that did, does, or will
exist) tomorrow, while (7) says that there did, does, or will exist
something that will witness that existentially quantified statement
tomorrow. (4) and (7) seem to claim that certain questions (whether
there will be a sea battle, who will win the lottery) are settled, while
(3) and (6) seem not to make such claims.
We have chosen to translate both (6) and (7) using the tenseless
quantifier Σ rather than the tensed quantifier ∃. Our translation of
(6) does not require that the winner of tomorrow’s lottery be some-
one who will exist tomorrow. (For all the logic says, the winner may
exist today but be dead by tomorrow’s drawing, or may not be born
until several years into the future.) Likewise, our translation of (7)
does not require that the future winner of tomorrow’s lottery be
someone who exists today. (For all the logic says, the winner may
be dead already, or not yet born.) The aim of this choice is simply to
avoid the complications that would result from changes in the
domain of quantification.
We will consider three proposals for assigning truth values to

sentences like (1)–(7): a supervaluationist proposal, a proposal
inspired by Łukasiewicz’s three-valued logic, and an intuitionist
proposal. Each proposal has different strengths and requires different
tradeoffs. We endorse the intuitionist proposal, but think the other
two proposals are serious contenders worthy of attention.
Proponents of any non-classical logic must reject some intuitively
plausible claims. All of the following claims are highly plausible,
but there are numerous incompatibilities among them, which illus-
trate the tradeoffs available to the Growing-Block theorist.
(a) There are some propositions such that (it is at least possi-
ble that) neither they nor their negations are true.
(b) A disjunction can be true only if one of its disjuncts is.
(c) All classical tautological inferences are valid.
(d) All other things being equal, truth values should be
assigned to as many propositions as possible.
(e) A language for describing the open future should be suf-
ficiently powerful to include an expression for determi-
nate truth.
(f) Reductio ad Absurdum, Contraposition, and Conditional
Proof are valid natural deduction inferences.
(g) Truth is determinate truth: a sentence is true just in case it
is guaranteed by history together with the laws of nature.
(h) Truth is disquotational: if a truth operator Tr is introduced
into the language, every instance of the schema ϕ ≡ Trϕ
should be valid.7
All three of our proposals are committed to (a). They count (1) as
sometimes true, sometimes false, and sometimes indeterminate.
They also agree in counting (2) as always either true or false, and (3)
as true in circumstances where the universe is guaranteed to persist
until tomorrow, false in circumstances where the universe is guar-
anteed not to persist until tomorrow, and indeterminate otherwise.
For reasons we discussed at the beginning of this section, Grow-
ing-Block theorists should accept (a). There are (or could be) propo-
sitions about the future whose truth values are not settled by the
7
Some exceptions will have to be made in cases of self-reference; we ignore these
complications here.
past, the present, and the laws of nature. For Growing-Block theo-
rists, this means that there are (or could be) propositions whose
truth values are not settled at all.
Philosophers who accept (a) must make tradeoffs elsewhere.
First, they must choose between giving up (b) and giving up (c). By
(a), there is some disjunction ϕ ∨ ¬ϕ, neither of whose disjuncts is
true—(5) is such a disjunction under some possible circumstances.
By (b), (5) is sometimes untrue. But since (5) is tautologically valid,
by (c) it is always true. So rejecting either (b) or (c) is mandatory.
The supervaluationist proposal gives up (b) and keeps (c).8 It
treats (3) and (4) as equivalent, and (6) and (7) as equivalent. The
Łukasiewicz and intuitionist proposals, on the other hand, give up
(c) and keep (b). They are able to treat (4) as stronger than (3) and (7)
as stronger than (6).
(d) need not be accepted or denied wholesale, but can be accepted
to a greater or a lesser extent, depending on how important one
thinks it is to assign truth values to as many sentences as possible.
If we treat (d) as a desideratum for theories, the supervaluationist
semantics does better than the intuitionist semantics, which in turn
does better than the Łukasiewicz semantics. All three proposals are
outdone by classical logic, but we think the cost of classical logic for
the Growing-Block theorist—spooky truths without truthmakers—
is excessive.
In addition to the tradeoff between (b) and (c), and the decision
about how seriously to take (d), accepting (a) necessitates a tradeoff
between (e) and (f). In any sufficiently powerful semantics that
allows for truth value gaps, one can formulate propositions for
which Reductio ad Absurdum, Contraposition, and Conditional Proof
fail. (A reminder: these three inferences are as follows:
Reductio ad Absurdum If ϕ ⊢ ⊥, then ⊢ ¬ϕ.

Contraposition If ϕ ⊢ ψ, then ¬ψ ⊢ ¬ϕ.
Conditional Proof If ϕ ⊢ ψ, then ⊢ ϕ ⊃ ψ.
8
At least, it keeps (c) in the sense that it validates all classically valid single-con-
clusion arguments. Classically valid multiple-conclusion arguments are supervalua-
tionistically invalid.
Since entailment is transitive and ¬⊥ follows from the empty set,

Reductio ad Absurdum is the special case of Contraposition where
ψ = ⊥.)
The above inferences fail in three-valued logics because they rely
on the assumption that there is no gap between truth and falsity. If
ϕ entails ψ, then there is no time in any timeline where ϕ is true and
ψ is false. But where there are truth value gaps, ϕ’s entailing ψ is
compatible with there being times in timelines at which ϕ is indeter-
minate and ψ is false. In any sufficiently powerful language, one will
be able to define a ϕ and a ψ such that ϕ entails ψ, and ϕ may be
indeterminate where ψ is false. Possibilities in which ϕ is false and
ϕ is not true can be redescribed as possibilities in which ¬ψ is true
and ϕ is not true. In such possible circumstances, ϕ ⊃ ψ—a condi-
tional with a false consequent and an indeterminate antecedent—
should not count as true. Even if ϕ entails ψ, ¬ψ does not necessarily
entail ¬ϕ, nor is the conditional ϕ ⊃ ψ necessarily valid.
Finally, accepting (a) means trading off between (g) and (h). If (a) is
correct, then there are some sentences that are indeterminate—not
false, but not true. It is false that these sentences are determinate. So,
where ϕ is a proposition with indeterminate truth value, and Δ is a
determinacy operator, Δϕ is false. When a biconditional has an indeter-
minate sentence on one side and a false sentence on the other, the entire
biconditional should not count as true—it is either indeterminate or
false. (This claim could be disputed, but it holds according to all the
interpretations of the biconditional we consider.) Therefore, where ϕ is
indeterminate, Δϕ is false, and so the biconditional ϕ ≡ Δϕ is not true.
2.1 The Supervaluationist Semantics

Richmond Thomason (1970) has proposed a supervaluationist
semantics for future contingents, which can easily be adapted to
suit the Growing-Block theorist’s ontology. Our discussion of Tho-
mason will be somewhat technical, but the gist of his theory is this:
ϕ is true if every possible future course of events will make ϕ true,
false if every possible future course of events will make ϕ false, and
indeterminate otherwise.
More technically, for Thomason sentences receive truth values at
times in model structures. Each of Thomason’s times is associated
with a classical valuation function over sentences without tense

operators—our ersatz times are therefore well suited to play the
role of Thomason’s times. A model structure, for Thomason, is a set
of times partially ordered by a relation R. Thomason assumes that R
is ‘treelike’—meaning that each time has a unique past, or more
formally, that for all times e, e’, and e* in the model structure, if e’Re,
e*Re, and e’ ≠ e*, then either e’Re* or e*Re’.
Truth at a time in a model structure is defined in terms of truth at
a time in a history—a maximal set of times h such that for any two
distinct times in the set, one bears the R relation to the other. (One
can picture histories as branches of the tree stretching all the way
from root to leaf-tip.) A proposition ϕ, says Thomason, is true at e
according to a model structure M if veh(ϕ) = 1 for every h in M such
that e ∈ h, false at e according to M if veh(ϕ) = 0 for every h in M such
that e ∈ h, and indeterminate at e in M otherwise.
The next step, then, is to define truth at a time in a history. Where
e is any time and h is any history such that e ∈ h, e and h are associ-
ated with a classical valuation function veh, built up in a familiar
fashion. For any atomic sentence p,
1 if p is true according to e
veh ( p) =
0 else
Thomason adds the usual clauses for the truth-functional connec-

tives (he does not discuss quantification) and the following clauses
for tense operators P and F.
1 if h contains an e’ such that e’ Re and ve ’ h (f ) = 1

veh (Pf ) =
0 else
1 if h contains an e’ such that eRe’ and ve ’ h (f ) = 1

veh (Ff ) =
0 else
Notice that so far, the definition of veh looks exactly like the defini-
tion of veT, except that the F operator is fully rather than partially
defined. F(ϕ) is false at e in h whenever it is not true at e in h. Before
we go on defining veh for a larger range of operators and quantifiers,
it is worth pausing to think about how Thomason’s system maps
onto ours.
We have already said that Thomason’s times correspond to our

ersatz times. In order to finish adapting Thomason’s semantics to
our ontology, we will need to find objects in our ontology to play
the role of his model structures and histories.
Sets of the form {T’: T’ is a feasible extension of T} are ideally
suited to play the role of model structures. These sets, as we
pointed out in section 1.4, have a tree-like structure. Furthermore,
they are exactly the right sorts of things to relativize truth to. Each
set is uniquely associated with a timeline, and timelines represent
possibilities. Since our ultimate aim is a concept of something like
truth in a possible world, we want to relativize truth to time-
lines (or to entities that admit of a natural one-to-one mapping to
timelines).
Given a timeline and the tree structure consisting of its feasible
extensions, we will need to pick out some entities to serve as its
histories. Histories in Thomason’s system must be something like
exhaustive catalogues of events. This must be the underlying ration-
ale for the semantic rule that treats F(ϕ) as false at e in h whenever h
fails to contain an e’ later than e at which ϕ is true. If h does not
explicitly represent the universe as containing a time at which ϕ,
then h represents the universe as not containing a time at which ϕ.
Complete timelines are well-suited to play the role of histories in
Thomason’s system. They are exhaustive catalogues of events, or to
put it another way, they describe ways the Growing Block might
finish up, if it were allowed to grow until it could grow no more.
What about semi-complete timelines? There seem to be two pos-
sible ways of considering them. On the one hand, one can consider
them under the assumption that they are exhaustive catalogues of
the universe’s events—complete descriptions of the way the Grow-
ing Block will end up. (After all, it is nomologically possible for the
Growing Block to grow until it actualizes a semi-complete timeline,
and then stop growing.) In other words, semi-complete timelines
can be considered as histories. This is the appropriate stance if one
is thinking from the perspective of a particular timeline, consider-
ing all possible exhaustive ways things might go next. The semi-
complete extensions of the original timeline are indeed exhaustive
ways things might go next.
On the other hand, one can take semi-complete timelines at face
value, not assuming that they are exhaustive. (After all, it is nomo-
logically possible that the Growing Block to grow until it actualizes
a semi-complete timeline, and then keep growing.) This is the

appropriate stance to take if one is thinking from the perspective of
a semi-complete timeline, wondering what is true according to it. It
is a way of thinking of the semi-complete timeline in itself.
Depending on how we think of a semi-complete timeline—either
as a history or in itself—different sentences will come out as true at
various times in that timeline. Where e is a time in a semi-complete
timeline T, what is really true at e in T is what is true at e according
to T considered in itself. But for the purposes of divining what is
true at times in other timelines, it may be necessary to consider
what is true at e according to T considered as a history.
One must treat complete timelines as histories, since it is nomo-
logically necessary that they be exhaustive. Unlike the case of semi-
complete timelines, there is no difference between considering
complete timelines as histories and considering them in themselves.
And one may not treat incomplete timelines as histories—they can-
not be exhaustive catalogues of the world’s events. Growing Blocks
that actualize incomplete timelines will have an irresistible impetus
to keep growing, adding more to the catalogue of the world’s events.
Incomplete timelines are not exhaustive, and should only be con-
sidered in themselves.
Since histories are timelines considered as exhaustive representa-
tions of the universe, we can add the following clauses to the defini-
tion of veh.
1 if h contains an e’ such that eRe’,

the distance from e to e’ is d ,
veh (Fdf ) =
and ve ’ h (f ) = 1
0 else
1 if for some e’ in h , there is an a ∈De ’
veh Σx(f x ) = such that ve ’ h (f x / a) = 1
0 else
0 if for some e’ in h, there is an a ∈De ’
veh Πx(f x) = such that ve ’ h (f x / a) = 0
0 else
So truth for sentences whose main operators are F, Fd, Σ, and П is
defined for histories exactly as it was defined for incomplete and
semi-complete timelines considered in themselves. But for histo-
ries, the definition of falsity is more extensive: sentences are false

whenever they are not true. In incomplete timelines, and in semi-
complete timelines considered in themselves, there may be truth
value gaps; in histories there are none.
How do propositions about the future get their truth values in
incomplete timelines, or in semi-complete timelines considered
in themselves? For us, each timeline is associated with a model
structure—the set of its complete or semi-complete extensions.
We can say that ϕ is true at e in T just in case it is true at e in the
model structure associated with T. Now it only remains to inter-
pret the concept of truth in a model structure, according to our
conception of model structures as sets of extensions of individual
timelines.
Every model structure contains histories: complete timelines and
semi-complete timelines which can be considered as histories. So,
mapping our ontology onto Thomason’s semantics yields the fol-
lowing inheritance clause:
Where T is an incomplete timeline, or a semi-complete timeline
considered in itself,
1 iff for every extension h of T such that h is either

a complete timeline or a semi-complete timeline
treated as a history,
veh (f ) = 1
veT (f ) =
0 iff for every extension h of T such that h is either
a complete timeline or a semi-complete timeline
treated as a history,
veh (f ) = 0
If the supervaluationist semantics is to be consistent, the new

inheritance clause cannot assign values to compounds which are
incompatible with the values assigned to those compounds by
the definitions of conjunction, negation, and the tense operators.
Fortunately, the types of sentences introduced so far have a con-
venient heredity property (proved in the Appendix). Any sentence
that receives value 0 or value 1 at e in T must receive the same
value at e in every feasible extension of T. So the inheritance
clause cannot generate any inconsistency in the valuation func-

tion—at least for the types of sentences considered so far. (For a
discussion of some sentences that lack the heredity property, see
section 3.1.)
How does the supervaluationist semantics handle sentences
(1)–(7)? (1) comes out true at e in T whenever each complete or
semi-complete extension of T contains a sea battle exactly one
day after e. In other words, (1) is true whenever a sea battle is
guaranteed true by the absolute past and present together with
the laws of nature. Likewise (1) is false at e in T whenever no
complete or semi-complete extension of T contains a sea battle
exactly one day after e—whenever the absolute past and present,
together with the laws of nature, guarantee that there will be no
sea battle. (1) is indeterminate whenever the absolute past, the
absolute present, and the laws of nature fail to determine whether
there will be a sea battle exactly one day after e. (2), unlike (1),
always comes out as either determinately true or determinately
false, since all feasible timelines agree about what happened
exactly one day before e.
(3) comes out as true when all feasible complete timelines are
ones according to which there is an e’ exactly one day after e—when
there is guaranteed to be a tomorrow. Likewise, (3) comes out as
false when there is guaranteed not to be a tomorrow, and indetermi-
nate when tomorrow’s existence is uncertain.
(4) comes out as true in exactly the same circumstances as (3).
This is somewhat surprising. (4) seems to say that the future is
determined, at least as far as sea battles are concerned. After all, (4)
is a disjunction, each of whose disjuncts says that the future is deter-
mined in a particular way. But (4) can be true at e in T even when
some feasible extensions of T contain a sea battle a day after e and
others contain no sea battle a day after e.
The trick is that on the supervaluationist picture, (4) can be true
without either of its disjuncts being true. Consider the following
example, depicted in Figure 2. e is the absolute present of an
incomplete timeline T, which has two feasible complete exten-
sions and no semi-complete extensions. According to one feasible
history that extends T, FdS is true at e; according to the other, Fd¬S
is true at e.
A Timeline and
Two Extensions T1
T
e⬘
such that
Ve⬘T1(S)=1
e
T2
Figure 2. A Timeline and Two Extentions
Sentence (4), the disjunction FdS ∨ Fd¬S, is true at e according to

both histories that extend T. So by the inheritance clause, (4) is true
at e in T. But neither disjunct of (4) is true at e in both histories that
extend T—different disjuncts are true in different histories. So nei-
ther disjunct of (4) is true at e in T.
(5) is a classical tautology—an instance of the Law of Excluded
Middle—and therefore true at every time in every timeline. Like
(4), (5) can be true without either of its disjuncts being true. (In fact,
the reader can confirm that according to the proposed semantics, (5)
will be true without either of its disjuncts being true whenever (4) is
true without either of its disjuncts being true.) It seems like a point
in favour of supervaluationism that it treats (5) and all other classical
tautologies as valid.
(6) comes out as true at e in T whenever, according to all feasible
timelines, there is a lottery one day after e, and history together with
the laws of nature guarantee that the lottery will be won. (Perhaps
this scenario is farfetched when it comes to real lotteries, which
typically stand some chance of being called off. But we will avoid
tinkering with qualifiers in order to keep the logical point as clear as
possible.)
(7) comes out as true whenever (6) does. This putative equiva-
lence is surprising, just as the putative equivalence between (3)
and (4) was surprising. (7) appears to say something stronger
than (6)—that the lottery is a fix. But (7) is true so long as the
absolute past, the absolute present, and the laws of nature guar-
antee that a prize will be given out. In such circumstances, it is
true according to every history that extends the actualized time-

line that there is someone who will win the lottery. There need
not (tenselessly) exist anyone of whom it is true that they will
win the lottery—who wins the lottery on every history that
extends the actualized timeline. Just as (4) was a disjunction that
could be true without having any true disjunct, so (7) is an exis-
tentially quantified statement that can be true without having
any witness.
2.2 The Łukasiewicz Semantics

Growing-Block theorists who find the supervaluationist treatment
of (4) and (7) unsatisfactory might try a different tradeoff between
(b) and (c). Instead of emphasizing the link between what is true
and what is guaranteed to become true no matter how the future
goes, they might emphasize the importance of preserving the mean-
ings of certain logical connectives—in particular, the truth-func-
tionality of of disjunction and the extensionality of existential
quantification.
Supervaluationists exploited a useful insight about the future:
what can be truly said of future events is roughly what is guaran-
teed to be true by the past, the present, and the laws of nature. But
it was a mistake, opponents of supervaluationism might claim, to
treat this insight as though it applied to all propositions. Proposi-
tions whose main operator is F are true whenever history and the
laws of nature guarantee that they will become true—i.e., a proposi-
tion of the form Fϕ is true at e in T whenever it is true at e in all
feasible extensions of T. But no such claim holds for propositions in
general. (In particular, no such claim holds for disjunctions and
existentials.) Instead, the supervaluationist’s inheritance clause
should be restricted to sentences whose main operators are tense
operators, as follows.
Where ϕ’s main operator is a tense operator,
1 if for every complete or semi- complete

extension T ’ of T , veT ’ = 1
veT (f ) =
0 if for every complete or semi-complete
extension T ’ of T , veT ’ = 0
The other connectives can be defined exactly as before, and veT can
be taken to be the smallest fixed point satisfying the recursive defi-
nition of the value function.
This way of defining the connectives yields an extension of the
logic advocated by Jan Łukasiewicz (1970). (Łukasiewicz outlines
definitions for truth-functional connectives and tense operators,
but does not discuss quantifiers.) Unlike the supervaluationist
semantics, the Łukasiewicz proposal treats the connectives as
truth-functional: the truth of a disjunction (existential) will always
be appropriately grounded in the truth of one or more disjuncts
(witnesses).
The benefits of the Łukasiewicz system are not without their
costs, however. In particular, Łukasiewicz’s logic does much worse
than supervaluationism on criterion (d)—it assigns truth values to
far fewer sentences. As a result, it treats fewer sentences as valid—
every classical tautology has untrue instances in the Łukasiewicz
logic, including the law of excluded middle. For instance, (5) is
indeterminate at e in T whenever some but not all semi-complete
extensions of T contain an e’ exactly one day after e at which there is
a sea battle. The situation is not quite as bad as one might expect
from the observation that Łukasiewicz’s truth-functional logic has
no theorems, however. Sentences without F operators or tenseless
quantifiers behave classically, so that every sentence that is both
valid in first-order logic and free of F operators and tenseless quan-
tifiers is a theorem.
Since the Łukasiewicz semantics agrees with the supervalua-
tionist semantics in its treatment of the operator, it agrees with
the supervaluationist semantics in its treatment of (1) and (2).
(1) is indeterminate at T in whenever T‘s feasible semi-complete
extensions disagree about whether there is a sea battle exactly
one day after, while (2) is determinate at every time in every
timeline.
The Łukasiewicz semantics also agrees with the supervaluation-
ist semantics in its treatment of (3) and (6). (3) is true at e in T pro-
vided there will be a future—i.e., provided all of T‘s feasible
complete or semi-complete extensions contain an e’ exactly one
day after e. And (6) is true at e in T whenever a prize will be given
out for the lottery—i.e., whenever all of T’s feasible complete or
semi-complete extensions contain an e’ exactly one day after e at

which the lottery is won. But in the Łukasiewicz semantics, unlike
the supervaluationist semantics, (4) and (7) are stronger than (3)
and (6). (4) is true only if one of its disjuncts is true—that is, (4) is
true at e in T only if either all of T’s feasible complete or semi-com-
plete extensions, or none of them, contain an e’ exactly one day after
e at which there is a sea battle. So (4) says that the past, the present,
and the laws of nature settle whether there will be a sea battle
tomorrow. Likewise (7) is true only if one if its instances is true—
that is (7) is true at e in T only if the same person wins the lottery a
day after e in all of T’s feasible extensions. So (7) says that the lottery
is a fix.
2.3 The Intuitionist Semantics

Łukasiewicz’s semantics can be strengthened by making slight
changes to the definitions of negations and the tenseless operators,
and by adding a non-truth-functional conditional.
if for all complete or semi- complete

1
extensions T ’ of T ,
veT (¬f ) = veT ’ (f ) ≠ 1
0 if veT (f ) = 1
if for all complete or semi- complete

1
extensions T ’ of T ,
veT (f ⊃ y ) = veT (f ) ≠ 1 or veT (y ) = 1
0 if veT (f ) = 1 and veT (y ) ≠ 1
1 if for some e’ in T , there is an ∈De ’T such that

ve ’T (fx/a) = 1
veT ∑ x(fx ) = 0 if for every extension T ’ of T and every e’ in T ’,
there is no a ∈De ’T ’ such that ve ’T (fx/a) = 1
1 if for every extension T ’ of T and every e’ in T ’,

there is no a ∈De ’T ’ such that ve ’T (fx/a) = 0
veT Πx(fx ) = 0 if for some e’ in T , there is an a ∈De ’T ’ such that
ve ’T (fx/a) = 0
Both changes can be reasonably motivated.

In the case of negation, it seems sufficient for ¬ϕ’s being true that
ϕ should be guaranteed never to become true—even if ϕ is not
guaranteed to become false. Consider the example of a universe
that is divided into epochs, such that the laws of nature permit it to
end after any finite number of epochs, but forbid its going on for
more than finitely many epochs. Consider the sentence ‘There will
be a last time’ (F¬F⊤), as uttered sometime e in the middle of the
first epoch, in a timeline T where e is the absolute present. This sen-
tence is not false, but it is not true, and can never become true. (After
every epoch there always might be another.) So it seems reasonable
to say that there won’t be a last time—that ¬F¬F⊤ is true at e in T.
Likewise, in the case of the quantifiers, it seems sufficient for
Πx(ϕx)’s truth (Σx(ϕx)’s falsity) that the laws of nature and guaran-
tee that there will be no counterexamples (witnesses). In order to
make a true universal generalization, we needn't guarantee, for all
past, present, and future things, that they support the generaliza-
tion. We need only guarantee that all past, present, and future things
support the generalization, whatever they happen to be.
The new conditional is really a strict conditional: ϕ ⊃ ψ is true at
e in T just in case, at all feasible timelines, ¬ϕ ∨ ψ is true at e. As a
strict conditional, it captures the idea that the antecedent’s truth
necessitates the truth of the consequent, where the kind of necessity
is feasibility. Notice that ¬ϕ is equivalent to ϕ ⊃ ⊥.
The result of adding the strengthened negation and conditional
operators is something very close to the semantics for intuitionism
proposed by Saul Kripke (1963). Our feasibility relation satisfies the
requirements on the accessibility relation in the Kripke semantics
for intuitionism—it is transitive, reflexive, antisymmetric, and (as
proved in the Appendix) hereditary—and our semantics defines
truth for conjunctions, disjunctions, negations, conditionals, and
quantified sentences in the same way as Kripke’s. There are, how-
ever, a number of notable differences between our semantics and
Kripke’s. First, our language contains the F operator, which is unde-
fined in Kripke’s semantics. Second, we give falsity conditions in
addition to truth conditions for our propositions. (Kripke uses the
word ‘false’ for the sentences that we are calling ‘untrue’.)
Third, our incomplete timelines generate a slight complication.
While incomplete timelines are feasible extensions of themselves,
they are not complete or semi-complete feasible extensions of them-

selves. T’ is accessible to T, in the sense required by Kripke’s seman-
tics, just in case T’ is a complete or semi-complete extension of T. So
even though the feasibility relation is reflexive, our analogue of
Kripke’s accessibility relation is not. (It is reflexive only when we
restrict our attention to complete and semi-complete timelines.) This
complication, however, makes little logical difference—we can sim-
ply declare each incomplete timeline to be Kripke-accessible relative
to itself, without changing any logical features of the semantics.
The intuitionist semantics does better than the Łukasiewicz
semantics on criterion (d). It is stronger than the Łukasiewicz
semantics in two notable senses. First, at any time in any timeline,
the set of true sentences according to the Łukasiewicz semantics is
a subset (and sometimes a proper subset) of the set of true sentences
according to the intuitionist semantics. Second, the set of theorems
according to the Łukasiewicz semantics is a proper subset of the set
of theorems according to the intuitionist semantics. (The set of infer-
ences validated by the intuitionist semantics is neither larger nor
smaller than the set of inferences validated by the Łukasiewicz
semantics—they are simply incomparable.)
Our motivation for the proposed intuitionist view differs from
that of traditional intuitionism, and from the intuitionism about the
future (and the past) advocated by Michael Dummett (1978a; 1978b).
Traditional intuitionists treat mathematical truth as essentially con-
strained by evidence, in the form of constructive proof. Dummett
extends the idea of evidential constraints on truth to empirical prop-
ositions, including propositions about times other than the present.
A proposition is true for Dummett insofar as it admits of construc-
tive proof from available evidence, and false for Dummett insofar as
it admits of constructive disproof from available evidence. For Dum-
mett, these evidential constraints on truth spring from considera-
tions about the meanings of logical and non-logical terms—meanings
are truth conditions; hence one knows the meaning of a term only
insofar as one can apply it in various evidential situations.
Our approach emphasizes not evidence and meaning, but truth-
making and grounding. A sentence about the future of the form Fϕ
is true only if its truth is grounded by what exists, i.e., past and
present things and the laws of nature. A sentence with any other
logically complex form is true only if its truth is appropriately
grounded in the truth of simpler statements—disjuncts, conjuncts,

or instances. We see no reason to require that all truths about the
future be knowable, or that they be grounded by knowable truths
about the past and present. Our version of the intuitionist proposal
requires only that they be grounded in past and present reality.
2.4 Taking Stock

So far, we have shown how supervaluationism, Łukasiewicz
logic, and intuitionism—all ways of cashing out (a)—require dif-
ferent tradeoffs between (b), (c), and (d). Supervaluationists
reject (b) for (c), and do better than advocates of Łukasiewicz
logic and intuitionism by the lights of (d). Advocates of
Łukasiewicz logic and intuitionism reject (c) for (b). Intuitionism
outperforms Łukasiewicz logic by the lights of (d), but is outper-
formed in turn by supervaluationism.
Logicians who reject (a) must make additional tradeoffs. The choice
between supervaluationism, Łukasiewicz logic, and intuitionism has
consequences for (f)—the claim that Contraposition, Reductio ad
Absurdum, and Conditional Proof are valid inferences. Choosing
Łukasiewicz logic means rejecting (f). Reductio ad Absurdum and Con-
traposition fail since classical contradictions, such as FS ∧ ¬FS, may
be truth valueless rather than false, and may therefore have untrue
negations. Conditional proof fails for much the same reason: FS ⊧ FS,
but FS ⊃ FS is truth valueless whenever FS is.
So advocates of Łukasiewicz logic must give up (f). Intuitionists
and supervaluationists have not yet been forced to give up (f), but
they are not yet in the clear. Like most theorists who countenance
truth value gaps, they must eventually choose between (f) and (e)—
the claim that a language for the open future should be sufficiently
powerful to include an expression for determinate truth.
Furthermore, advocates of all three logics must choose between
(g), according to which truth is determinate truth, and (h), accord-
ing to which truth is disquotational. The tradeoffs between (e) and
(f) and (g) and (h) will be illustrated in section 3.1, where we define
an operator that captures the concept of determinate truth.
Although we have more logical work to do in part 3, we are ready
to conclude our discussion of the tradeoffs between the supervalu-
ationist, Łukasiewicz, and intuitionist proposals. We believe that all

three proposals are viable responses to the situation that faces the
Growing-Block theorist, and that we have presented the reader
with the materials necessary to make an informed choice. We now
turn our attention to two logical issues that affect all three proposals
equally.
3. GENERAL LOGICAL ISSUES
The two issues to which we now turn are closely bound up with
the concept of expressive power. In section 3.1, we consider how
Growing-Block theorists might enrich their language to express
the idea of determinate truth. Our results shed further light on the
tradeoffs between (e), (f), (g), and (h) discussed in part 2. In sec-
tion 3.2, we consider the prospects for expressing the Growing-
Block theorist’s ontological commitments in a formal language.
3.1 Determinacy
Our metalinguistic discussion of truth at times in timelines has made
extensive use of the distinction between sentences that are true (at
times in timelines) and sentences that are not true (at times in
timelines)—either because they are false or because they are indeter-
minate. How can the Growing-Block theorist express the thought
that a sentence is true, rather than false or indeterminate, using an
object language expression in one of the idealized languages we
have developed? The usual response to this question is to define a
determinacy operator. How exactly should this be done?
Thomason defines a type of determinacy operator—the inevita-
bility operator L—as follows:
1 if veh’ (f ) = 1 for every h’ in the model structure
veh (Lf ) = containing e
0 else
How should this inevitability operator be understood in terms of
truth at times in timelines?
The question is somewhat thorny. For any sentence not contain-
ing the L operator, we could define truth at a history, and more
generally, truth at a time in a timeline. But for a sentence whose

main operator is L, we can only define truth at a time in a history in
a model structure (and derivatively, truth at a time in a timeline in a
model structure). Different choices of model structure give different
truth conditions for sentences of the form Lϕ. In order to translate L
into our formalism, then, we need a method for picking out the
relevant model structure in any given situation.
One option, in keeping with the spirit of Thomason’s proposal, is
to index the relevant model structure (for time e in timeline T ) to the
initial segment of T containing e and all the times before it. Inter-
preting L this way gives us:
1 if veT ’ (f ) = 1
veT (Lf ) =
0 else
where T’ is the initial segment of T containing only e and the e’s

such that e’Re
This interpretation of L makes the model structure drop out of
the semantics again—we can assign a univocal truth value at each
time in each timeline to sentences of the form Lϕ.
Thomason claims that L has the following properties.
L1. ϕ ⊧ Lϕ
L2. Fϕ ⊧ LFϕ
L3. ⊨ ϕ ⊃ Lϕ
L4. ⊭ Fϕ ⊃ LFϕ
L5. ⊭ PFϕ ⊃ PLFϕ
(Notice that if Thomason is right, adding L to the language makes it
possible to generate counterexamples Conditional Proof, since Fϕ ⊧
LFϕ, but ⊭ Fϕ ⊃ LFϕ. Adding L to the language also makes it pos-
sible to generate counterexamples to Contraposition, since ϕ ⊧ Lϕ,
but presumably ¬Lϕ ⊭ ϕ)
According to our semantics for L, L4 and L5 are invalid, just as
Thomason claims. Furthermore, they are invalid regardless of
whether we use the supervaluationist semantics, the Łukasiewicz
semantics, or the intuitionist semantics. We can produce the follow-
ing countermodel for both sentences. Let T be a timeline with abso-
lute present e, and some e* such that e*Re. And suppose T has exactly
two feasible complete extensions—T1 and T2. T1 contains an e’ such
A Timeline and
Two Extensions T1
T
e⬘
such that
Ve⬘T1(S)=1
e
e* T2
Figure 3. A Timeline and Two Extentions
that eRe’ and the atomic sentence S is true at e’; T2 contains no such
e’. Furthermore T has no feasible semi-complete extensions. The
example is depicted in Figure 3.
In this example, veT2 (FS ⊃ LFS) = 0 , so that veT(FS ⊃ LFS) ≠ 1.
Similarly, veT2 (PFS ⊃ PLFS) = 0 , so that veT(PFS ⊃ PLFS) ≠ 1.
But according to our semantics for L, and contrary to Thomason’s
claims, L1–L3 are also invalid. Once again, they are invalid regard-
less of whether we use the intuitionist semantics, the supervalua-
tionist semantics, or the Łukasiewicz semantics. In the example just
considered, veT1 (FS) = 1 , while veT1 (LFS) = 0 ; this constitutes a coun-
terexample to L1 and L2. Similarly, veT1 (FS ⊃ LFS) = 0 ; this consti-
tutes a counterexample to L3.
Although L1–L3 are not valid in general—that is, not valid eT—they
are valid T. Thomason’s model structure runs validity eT and validity
T together. For the fragment of the language without the L operator,
this conflation is harmless: validity eT and validity T coincide. But in a
language containing the L operator, the difference matters.
Despite initial appearances to the contrary, the introduction of L
does not provide counterexamples to Conditional Proof or Contrapo-
sition. Enriching a language with the L operator preserves the hered-
ity property (see Appendix for proof). And in any supervaluationist,
Łukasiewicz, or intuitionist language where all sentences satisfy the
heredity property, Conditional Proof and Contraposition are valid.
Distinguishing between validity T and validity eT also reveals
that L is a poor candidate for a truth operator. Not everything true
at a time is inevitable at that time, as the invalidity of inference L1
shows. What is true full stop coincides with what is inevitable at the
absolute present, but it would be a mistake to confuse truth at a
time with inevitability at that time.
We can define a determinacy operator that does a better job of
capturing the idea of determinate truth by making a different choice
about how to pick out the relevant model structure, given a time
and a timeline:
1 if veT (f ) = 1
veT ( Δf ) =
0 else
While Lϕ expresses the thought that ϕ is determinately true at e
from the perspective of the initial segment of the actual timeline
ending at e, Δϕ expresses the thought that ϕ is true at e from the
perspective of the actual timeline. Determinate truth at e and inevi-
tability at e correspond when e is the absolute present, but come
apart when e is in the absolute past.
This definition of the Δ operator creates some trouble for the
supervaluationist and intuitionist proposals, in which certain con-
nectives are defined intensionally rather than truth-functionally.
Sentences containing Δ violate the heredity condition—a sentence
may have a truth value in one timeline without having that truth
value in all of its extensions.
In supervaluationism, the failure of the heredity condition leads
to inconsistent value assignments. In the example shown in Figure 3,
veT1 ( ΔFS ∨ ΔF¬S) = veT2 ( ΔFS ∨ ΔF¬S) = 1 —and so by the supervalu-
ationist inheritance clause, veT(ΔFS ∨ ΔF¬S) should be 1 as well. But
veT(ΔFS) = veT(ΔF¬S) = 0—and so by the semantic rule for disjunc-
tion, veT(ΔFS ∨ ΔF¬S) should be 0.
In intuitionism, the failure of the heredity condition leads to
highly counterintuitive value assignments. In the example shown
in Figure 3, veT(ΔFS) = 0—it is false at e in T that determinately, there
will be a sea battle. Yet veT(¬ΔFS) ≠ 1. This is because at e in T, there
is some feasible timeline—namely T1—according to which ΔFS is
true. Flat-footedly applying the intuitionist semantics yields coun-
terintuitive results.
The trouble can be easily resolved by making the following
adjustment to both the supervaluationist and the intuitionist seman-
tics. To evaluate sentences containing determinacy operators at a
time in T, first set veT ’ ( Δf ) equal to either 1 (if veT(ϕ) = 1) or 0 (other-
wise) for every sentence ϕ and feasible extension T’ of T. This

ensures that the heredity condition is satisfied. Then, and only then,
evaluate complex sentences in the usual way.
Δ, so defined, violates the T schemas. The example in Figure 3
illustrates why. In this example, veT(FS) ≠ 1, so veT ( ΔFS) = 0. But then
veT1 ( ΔFS) = 1, so that veT(FS ≡ ΔFS) = 0 according to the intuitionist
semantics, and veT(ϕ ≡ ψ) ≠ 1 according to the supervaluationist
semantics and Łukasiewicz semantics.
We claim that Δ behaves like an actuality operator rather than a
disquotational truth operator. The values of sentences containing
the Δ operator can take on different values at the same time in the
same timeline, depending on which world is considered as actual.
This is unsurprising, since Δ denotes a type of truth that is grounded
in concrete existence. One cannot tell what is true at a time simply
by considering the representational features of a possible world;
one needs to know which world is actualized.
No wonder, then, that there is a tradeoff between (g), the determi-
nacy conception of truth, and (h), the disquotational conception of
truth. Determinacy, like actuality, and unlike disquotational truth,
violates the T schema. With determinacy, as with actuality, one
might be misled by the fact that the determinacy operator satisfies
the T rule. For arbitrary ϕ, ϕ ⊧ Δϕ, and Δϕ ⊧ ϕ. But since Conditional
Proof is invalid in a language with a determinacy operator, this
does not allow us to infer instances of the T schema. This would be
just as invalid as inferring the validity of the biconditional ⌜ϕ ≡
actually ϕ⌝ from the fact that ϕ and ⌜actually ϕ⌝ always take the
same truth value at the actual world.
3.2 Quantifier Trouble

The eternalist quantifiers Σ and П create trouble for the Growing-
Block theory. Suppose that in T, the actualized world, e is the abso-
lute present. And let B be the property of being born after e. Suppose
the absolute past and present, together with the laws of nature,
determine that someone will be born after e. What are we to make
of the sentence ΣxBx?
The supervaluationist semantics counts it as determinately true
that ΣxBx. But the Growing-Block theorist is committed to the claim
that only past and present things exist. Things look just as bad for
the intuitionist and the advocate of Łukasiewicz logic if the past

and present, together with the laws of nature, decide who will be
born after e. In these situations, they too will have to affirm that
ΣxBx. Our semantics for the tenseless quantifiers appears to be at
odds with our explicit commitments.
We reply that ontological commitment cannot be read off the
tenseless existential quantifier Σ. Quantifying over everything that
was, is, or will be is like quantifying over all detectives, real and
fictional—a useful logical device, but not a way of stating one’s
ground-level commitments, since it involves quantifying over such
unreal entities as Sherlock Holmes and Hercule Poirot.
How, then, should Growing-Block theorists state their ground-
level ontological commitments? We suggest that a Growing-Block
theorist is committed to everything that exists according to the
actualized timeline. One can quantify over things that exist accord-
ing to T’ in the following way.
1 if for some e’ in T ’, there is an a ∈ De ’

such that veT (fx/a) = 1
veT ∑ T ’ x(fx) =
0 If for every e’ in T ’, there is an a ∈ De ’
1 if for some e’ in T ’, there is an a ∈De ’

veT Π T ’ x(fx ) =
0 if for every e’ in T , there is an a ∈De ’T
such that ve ’T (fx/a) = 1
To say that ΣTx(ϕx) is to claim that some object in T‘s ontology is ϕ

(at the indexical present, in the actual world).
We claim that a Growing-Block theorist is ontologically committed
to everything that (by her lights) actually exists. That is, a Growing-
Block theorist is ontologically committed to the existence of an object a
just in case she accepts that ΣTx(x = a) and T is the actualized timeline.
Is there a way of introducing a formal expression for the concept
of ‘actualized timeline’? We fear not. The formal semantics gives us
a way of determining what is true at a time in a timeline, and a way
determining which timeline is actualized according to each time-
line (each timeline is actualized according to itself), but no way of

determining which timeline is actualized full stop. Being actualized
is not a matter of being actualized according to any timeline.
4. WHY THE GROWING-BLOCK THEORY

OUTPERFORMS SIMILAR RIVALS
So far, we have developed a metaphysical and semantic framework

for explicating modality in the context of the Growing-Block theory,
used this framework to distinguish between true sentences about
the future and false ones, developed three semantic proposals for
handling the indeterminacy about the future that is likely to accom-
pany the Growing-Block theory, and addressed questions that this
indeterminacy raises about the expressive power of the languages
available to Growing-Block theorists. We now turn to some of the
Growing-Block theory’s close relatives: Craig Bourne’s ersatzer
Presentism, Storrs McCall’s Shrinking-Tree view, and John MacFar-
lane’s relativistic semantics for future contingents. These theories
are rivals to the Growing-Block theory, since they share many of its
appealing features. We argue that where the Growing-Block theory
differs from these three proposals, it outperforms them.
4.1 Bourne’s Ersatzer Presentism

Bourne accepts a theory of branching time much like our own, in
which ersatz possible worlds consist of sequences of ersatz times.
Unlike us, however, Bourne thinks that only one time—the present—
is concretely instantiated. Sentences of the form Pp are made true by
the existence of ersatz times which represent p as being true, and which
are appropriately related to the present by the ordering of ersatz time.
So if it is now true that Bill the Brontosaurus ate an oak leaf, then what
makes this true is the existence of an ersatz time, appropriately related
to the ersatz present, which represents Bill as eating an oak leaf.
We believe that Bourne’s theory provides insufficient grounding
for truths about the past. What makes that ersatz time appropri-
ately related to the present? Why isn’t the present related exclu-
sively to ersatz times according to which Bill failed to eat oak
leaves? Bourne thinks there is no issue here. “[A]ccording to ersatzer
presentism,” he writes,
what makes ‘It was the case that p‘ true is an actually R-related ordered
triple, whereas according to the tenseless theory, what makes it true is an
actually earlier than-related concrete fact. Now to ask why these ordered
triples are actually R-related is about as fair as asking why the concrete
facts are actually earlier than-related in the tenseless theory, i.e., not at all—
they just are.9
But this misses the point of the objection. Truth should supervene on
being—on the concrete things that tenselessly exist, the properties
and relations those things instantiate, and the laws of nature. Accord-
ing to Bourne’s theory, however, there could be two possible worlds
exactly alike with respect to the concrete things they contained, the
properties and relations those concrete things instantiated, and the
laws of nature, in which the ersatz present was nonetheless R-related
to different ersatz past times. This, we think, grants the abstract
realm of ersatz times far too much spooky autonomy.
The Growing-Block theory does better. Which timeline is actual-
ized is determined by which concrete things tenselessly exist, the
properties and relations those concrete things instantiate, and the
laws of nature. Which complete timelines are feasible are deter-
mined by the whole of concrete reality. There are no brute facts
about which ersatz times are appropriately related to the present:
where concrete reality and the laws of nature are insufficient to
uniquely determine the future, all candidate futures are possible.
3.2 McCall’s Shrinking-Tree View

If we are concerned that Presentism has too much autonomy from
concrete existence, we might consider Storrs McCall’s view (1994, 3)
known as ‘the Shrinking Tree’. The Shrinking-Tree view, like the
Growing-Block view, posits a moving absolute present and a con-
crete past. Unlike the Growing-Block view, the Shrinking-Tree view
holds that all the future possibilities are concrete, with branches ceas-
ing to (tenselessly) exist as they cease to be feasible possibilities:
The universe, then, has in this model the shape of a tree, with a single
four-dimensional trunk for the past and a densely branching set of four-
9
In the interest of notational uniformity, we have replaced Bourne’s ‘E-related’
with our ‘R-related’.
dimensional manifolds for the future. Each of these manifolds in turn

branches, so that the branching pattern is very complex and the number
of branches very large.
McCall’s tree-like structure is like our tree-like structure generated

by a timeline and the set of its feasible extensions, but McCall’s tree
is made up of space-time, whereas the our tree is made up of ordered
sets of propositions. A view where space-time has many future
branches but only a single past trunk might seem a good way of
capturing the intuitive view that the future is pregnant with possi-
bility and the past is implacably fixed. There is no question, on
McCall’s view, of the truths not being grounded in concrete exist-
ence. McCall’s view is not, then, vulnerable to the sort of objection
that affects Bourne. There are some reasons for concern, however.
Michael Tooley (1997, 239) argues that there are two considera-
tions that count against McCall’s view:
Broad’s model of a dynamic world [the Growing-Block view] seems prefer-
able to McCall’s, for at least two reasons. First, it allows one to make sense
of the idea of dynamic worlds that are deterministic, and this seems desir-
able, since deterministic worlds can, no less than indeterministic ones, be
worlds where states of affairs come into existence. Secondly, given that a
world containing no future states of affairs at all is rather more austere than
one that contains states of affairs corresponding to all future possibilities,
Broad’s model is also to be preferred on grounds of simplicity.
McCall defines the present as the point at which branching first

occurs. So, unless there is branching, there is no absolute present.
This rules out the possibility of there being only one feasible future,
on McCall’s view, which means, as Tooley points out, the Shrink-
ing-Tree view is incompatible with determinism. More than that,
the Shrinking-Tree view cannot cope with a world which is deter-
ministic for five minutes, since time could not pass for those five
minutes, but rather would skip until the first branching-point.
McCall’s view rests on a strong claim, then: The laws of nature are,
and always will be, indeterministic. Such assumptions should be
avoided if we can do the same work without them.
When Tooley describes the Growing Block as being more austere
than the Shrinking Tree he seems to understate the case. It is
not just that McCall is committed to more than a Growing-Block
theorist, but that McCall is committed to any number of as yet
unborn children, who continually drop out of existence when they

cease to become feasible. On both the Growing-Block and the
Shrinking-Tree views, something ceasing to be feasible is a question
of ontological change. On the Growing Block something comes into
existence which rules out the possibility, whereas on the Shrinking
Tree the possibility literally ceases to exist. This lends the Shrinking
Tree a destructive air. It might seem as though we are all guilty of
mass-murder for wiping all the possibilities that did exist out of the
universe. It is not clear that we ought to treat possible future per-
sons as having the same moral status as present persons, nor that
this is a consequence of McCall’s view, but it seems odd that we
should treat our prize-winning first novels, which we have not yet
thought of and may never write as being just as real, as the first
short-stories that we wrote at school. That is the sense in which it
seems the Growing Block is more austere than the Shrinking Tree.
You might worry that we face the same sort of objection from
McCall that we put to Bourne. Do our ersatz times not have the
same sort of spooky autonomy that Bourne’s did? In short, they do
not. Our ersatz times, unlike Bourne’s are grounded in what (tense-
lessly) exists and the laws of nature. To explain what will be true
we need no more than that what has already happened and the
laws of nature. It is then left as a question for science whether the
laws of nature and the history of the universe allow of a number of
possible futures or not. The Growing-Block view does not need as
much, ontologically, as the Shrinking-Tree view, nor does it rest on
tendentious assumptions about the indeterministic nature of the
universe.
Our view also has explanatory advantages over the Shrinking-
Tree view. The inevitability operator generates a convenient way of
talking about lost possibilities, which McCall cannot easily repli-
cate. For McCall the flow of time meant that branches of space-time
dropped out of existence when they were no longer feasible. In our
model, there exist not only those possibilities which, given the state
of the universe and the laws of nature, are feasible, but those which
were feasible and are no longer.
Thus not only optimists and pessimists, but also historians, can
be subjected to rational scrutiny through our account. Was it inevi-
table that the Government forces would win at Culloden? Well, if
the Government forces win according to all feasible extensions of
the timeline the Growing Block instantiates up to the start of the

battle, then yes, it was inevitable. If there were feasible extensions
of the timeline instantiated up to the start of the battle according to
which the Government lose, then it was not inevitable. Unlike
McCall, we need not have a further mechanism for accounting for
possibility in addition to the one we use for feasible futures. We
get an account of this ordinary sort of counterfactual possibility
included in the package.
4.3 MacFarlane’s Relativism

John MacFarlane’s (2003) view, unlike the Shrinking-Tree view or
Presentism, is concerned primarily with semantics rather than
ontology. He argues that the best way to capture the idea that some
sentences about the future are truth valueless is to embrace relativ-
ism about truth. On MacFarlane’s theory, the truth values of tensed
sentences must be relativised not only to the time of utterance, but
to the time of assessment.
MacFarlane asks us to consider a particular utterance of
1. Exactly one day into the future, there will be a sea battle.
(FdS)
We will call the utterer, as he does, Jake. Imagine Jake utters (1) on
a Monday, and the world’s history up to Monday together with the
laws of nature fails to determine whether there will be a sea battle
on Tuesday. MacFarlane thinks that on Monday, we should regard
Jake’s utterance as having an indeterminate truth value. MacFarlane
accepts, as we do, that there can be two possible futures (one on
which there is a sea battle and one on which there isn’t), and no
reason to think that one of these futures is ‘singled out’ as the actual
one. In such a case “symmetry conditions seem to rule out saying
either the utterance is true or that it is false. Thus, it seems, we must
count it as neither true nor false. This is the indeterminacy intui-
tion” (MacFarlane 2003, 323).
Next, imagine it is Tuesday, the day after Jake’s utterance of (1)
on Monday. Imagine also that on this particular Tuesday a sea battle
is raging. MacFarlane argues we should be inclined to agree to the
following:
Jake asserted yesterday that there would be a sea battle today

There is a sea battle today
So Jake’s assertion was true.
When we take this retrospective view, we are driven to assign a determinate
truth value to Jake’s utterance: this is the determinacy intuition. (325)
The indeterminacy intuition is in tension with the determinacy

intuition. MacFarlane argues that we can accommodate both by
making the truth of Jake’s utterance relative to the context of assess-
ment, so that his utterance has a different truth value depending on
when you ask if it is true: it is indeterminate when assessed on
Monday but true when assessed on Tuesday. The idea that Jake’s
utterance is true, false or indeterminate full stop, MacFarlane claims,
should be rejected.
We (the authors) do not have to accept that there is no such thing
as truth full stop: we already have the resources to accommodate
both the determinacy intuition and the indeterminacy intuition. We
needn’t relativize truth to contexts of assessment; we can simply
relativize truth to timelines. (Since timelines are possible worlds,
this amounts to evaluating truth in different possible worlds.)
(1) is indeterminate on Monday in the relevant timeline where
Monday is absolutely present—in that timeline, then, Jake’s utter-
ance of (1) is true. But (1) is false on Monday in the relevant timeline
where Tuesday is absolutely present—in that timeline, Jake’s utter-
ance of (1) is false. The difference in truth values can be put down to
a difference between possible worlds, rather than a difference
between contexts. (Which world is actualized will change, of course,
as time passes. This, on the Growing-Block view, is just what the
passage of time is.) On Tuesday, observers can capture the sense in
which Jake’s utterance of (1) was indeterminate by noting that,
although the proposition Jake asserted then was true (PdΔFdS), it
was not inevitably true (¬PdLΔFdS).
There is no need to abandon truth full stop in favour of truth
relative to context of assessment. What is true at a given time is
dynamic, and can change as the Growing Block grows, because
which world is actualized changes as the Growing Block grows.
The continual change in which world is actualized is what C.D.
Broad (1973, 766) described as the “rock-bottom peculiarity” of
time.
5. CONCLUSION
We have, above, considered what view we should have if we wanted

to talk about the future. We have argued that someone inclined to
the Growing-Block view could give a semantics that involves no
commitment to anything more than what they are ontologically
committed to already, but, nevertheless, can accept that there are
non-trivial future truths, and that some options are feasible while
others are not. Three options have been given about how this
semantics might go; one supervaluationist, one Łukasiewicz, and
one intuitionist. We have also considered some similar rivals, to
show that our theory outperforms them.
We have not argued for the Growing-Block view itself. Equally
we have not tried to deal with all of the common objections to
the Growing-Block view. These topics would extend the scope of
this paper beyond that which can reasonably be dealt with in a
single article. We are, however, optimistic about the fortunes of
what we take to be a plausible position in the philosophy of time,
and believe we have shown that claiming the future does not
exist, rather than rendering future possibility mysterious, allows
fruitful analysis.
University of Sydney
University of Sheffield
REFERENCES
Robert Merrihew Adams. Theories of actuality. Noûs, 8(3): 211–31, 1974.
Craig Bourne. A Future for Presentism. Oxford University Press, Oxford,
UK, 2006.
David Braddon-Mitchell. How do we know it is now now? Analysis,
64(3): 199–203, July 2004.
C.D. Broad. Scientific Thought. Kegan Paul, London, 1923.
—— A reply to my critics. In P.A. Schlipp, editor, The Philosophy of C.D.
Broad, pages 711–830. Tudor, New York, 1959.
Michael Dummett. The philosophical basis of intuitionistic logic. In Truth
and Other Enigmas, pages 215–47. Harvard University Press, 1978a.
—— The reality of the past. In Truth and Other Enigmas, pages 358–74.
Harvard University Press, 1978b.
Peter Forrest. The real but dead past: a reply to Braddon-Mitchell.

Analysis, 64(284): 358–9, 2004.
Patrick Greenough. Indeterminate truth. Midwest Studies in Philosophy,
32: 213–41, 2008.
Saul Kripke. Semantical analysis of intuitionistic logic. In Michael Dum-
mett and J. N. Crossley, (eds.), Formal Systems and Recursive Functions.
pages 92–130. North-Holland, Amsterdam, 1963.
Jan Łukasiewicz. On determinism. In Ludwik Borkowski, editor, Selected
Works. North-Holland, Amsterdam, 1970.
John MacFarlane. Future contingents and relative truth. The Philosophical
Quarterly, 52(212): 321–36, 2003.
Storrs McCall. A Model of the Universe. Clarendon Press, Oxford, UK,
1994.
Roy Sorensen. Vagueness and Contradiction. Oxford University Press,
Oxford, 2001.
Richmond Thomason. Indeterminist time and truth-value gaps. Theoria,
36: 264–81, 1970.
Michael Tooley. Time, Tense, and Causation. Oxford University Press,
Oxford, 1997.
Achille C. Varzi. Supervaluationism and its logics. Mind, 116(463): 633–76,
July 2007.
APPENDIX: PROOF OF THE HEREDITY CONDITION

Heredity Condition: Any sentence that receives value 0 or value 1 at
e in T must receive the same value at e in every feasible extension
T ’ of T.
Base Case: The heredity condition holds for atomic propositions (whose
truth values at e in T are solely a matter of e’s representational
features).
Inductive Hypothesis: The heredity condition holds for propositions ϕ
and ψ.
Inductive Steps: Likewise, the heredity condition holds for ∃xϕ(x/a)
and ∀x(ϕ(x/a), since these quantifiers depend only on the extension of
De (which stays constant between T and T’s extensions) and the truth
values of the ϕ(x/a)'s for various values of a (assumed to be herdia-
tary by the inductive hypothesis).
The heredity condition holds for ϕ ∨ ψ and ϕ ∧ ψ, since ∨ and ∧ are
truth functions. To inherit veT(ϕ) and veT(ψ) is automatically to inherit veT(ϕ
∨ ψ) and veT(ϕ ∧ ψ). Likewise, the heredity condition holds for ∃xϕ(x/a) and
∀x(ϕ(x/a), since these quantifiers depend only on the extension of De (which
stays constant between T and T’s extensions) and the truth values of the
ϕ(x/a)’s for various values of a (assumed to be hereditary by the inductive

hypothesis).
The heredity condition holds for ¬ϕ, according to both the Łukasiewicz
definition and the intuitionist definition of negation. On the Łukasiewicz
definition, ¬ is a truth function, so that to inherit veT(ϕ) is automatically to
inherit veT(¬ϕ). On the intuitionist definition, we have two cases to consider:
the case where veT(¬ϕ) = 0, and the case where veT(¬ϕ) = 1. If veT(¬ϕ) = 0, then
by the semantic rule for negation, veT(ϕ) = 1. By the inductive hypothesis, for
every feasible extension T’ of T, veT ’ (f ) = 1. By the intuitionist semantic rule
for negation, veT(¬ϕ) = 0. If veT(¬ϕ) = 1, then by the semantic rule for nega-
tion, at every feasible extension T’ of T, veT ’ (f ) ≠ 1. Since feasibility is transi-
tive, for every feasible extension T’ of T, for every feasible extension T* of T’,
veT* ≠ 1. By the semantic rule for negation veT ’ (¬f ) = 1.
The heredity condition holds for ϕ ⊃ ψ in the intuitionist semantics.
Whenever veT(ϕ ⊃ ψ) = 1, by the semantic rule for ⊃, every feasible com-
plete or semi-complete extension T* of T is such that veT*(ϕ) ≠ 1 or veT*(ψ)
= 1. Since feasibility is transitive, for every feasible extension T’ of T,
every feasible complete extension T* of T’ is such that either veT*(ϕ) ≠ 1 or
veT*(ψ) = 1. So by the semantic rule for ⊃, veT’(ϕ ⊃ ψ) = 1. Whenever veT(ϕ ⊃
ψ) = 0, by the semantic rule for ⊃, veT(ϕ) = 1 and veT(ψ) = 0. By the inductive
hypothesis, where T’ is any feasible extension of T, veT ’ (f ) = 1 and
veT ’ (y ) = 0. By the semantic rule for ⊃, veT ’ (f ⊃ y ) = 0 .
The heredity condition holds for P(ϕ) and F(ϕ). On all three of our logi-
cal proposals, this follows from the definition of an extension. By clauses
(a) and (b) of the definition, whenever a timeline T contains an e and an e’
such that e’Re and ve ’T (f ) = 1 , so does any extension—and hence, any
feasible extension—of T. Likewise, whenever a timeline T contains an e and
an e’ such that eRe’ and ve ’T (f ) = 1 , so does any (feasible) extension of T.
Therefore, whenever veT assigns value 1 to P(ϕ) or F(ϕ), so must veT ’ for
any feasible extension T’ of T. By clause (c) in the definition of an exten-
sion, whenever a timeline T fails to contain an e and an e’ such that e’Re
and ve ’T (f ) = 1 , so does any (feasible) extension of T. Therefore, when-
ever veTP(ϕ) = 0, veT ’P(f ) = 0 for any feasible extension T’ of T. Finally,
whenever veT (Fϕ) = 0, T has no feasible complete or semi-complete exten-
sion T' containing an e' such that eRe' and ve'T' (ϕ) = 1. Since feasibility is
transitive, no feasible extension of T has a feasible complete or semi-com-
plete extension T' containing an e' such that eRe' and ve'T' (ϕ) = 1. Thus, in
every feasible extension T* of T, veT* (Fϕ) = 0.
The heredity condition holds for Pd and F d. This follows from the
definition of an extension, together with the assumption that the dis-
tance between two ersatz times depends only on the ordering, together
with the representational features, of the times between them. If T’ is
an extension of T, and T contains times e and e’, T’ must contain e and

e’, and will contain all the times between them in exactly the same
order. Since the distance between two times supervenes on their con-
tents and the contents and orderings of the times between them, if the
distance between e and e’ is d according to T, then the distance between
e and e’ is d according to T’. Thus, whenever veT(Pd(ϕ)) = 1 or veT(Pd(ϕ))
= 0, for any extension T’ of T, veT ’ (Pd (f )) = veT (Pd (f )) , and likewise,
whenever T contains an e’ such that eRe', e' is distance d from e, and
v eT( F d( ϕ )) = 1 or v eT( F d( ϕ )) = 0, for any extension T ’ of T ,
veT ’ ( Fd (f )) = veT ( Fd (f )). If T contains no e' such that eRe' and e' is distance
d from e, we can see that for all extensions T' of T, veT' (Fdϕ) = veT (Fdϕ) by
observing that veT (Fdϕ) takes on a value if and only if veT' (Fdϕ) takes on
the same value for all complete or semi-complete extensions T' of T. The
proof is then exactly as for the above case of Fϕ.
The heredity condition holds for Σxϕ and Πxϕ. In the intuitionist and
Kleene semantics as well as the supervaluationist semantics. First, con-
sider only the case where veT(Σxϕ) = 1, and the case where veT(Πxϕ) = 0.
Where veT(Σxϕx) = 1, there is an e’ ∈ T such that for some a ∈ De ’ ,
ve ’T (f x / a) = 1. Any extension T’ of T will also contain e’, so that by the
semantic rule for Σ , veT ’ (Σxf x ) = 1. Likewise, where veT(Πxϕx) = 0, there is
an e’ ∈ T such that for some a ∈ De ’ , ve ’T (f x / a) = 0 . Any extension T’ of T
will also contain e’, so that by the semantic rule for Π, veT ’ (Πxf x ) = 0 .
Next, for the intuitionist semantics, consider the case where veT Σx(ϕx) =
0, and the case where veT Πx(ϕx) = 1. (We needn’t consider this case in the
Kleene semantics, since it never arises.) In both the semantic rule for Σ
guarantees that for all extensions T' of T, e' in T', and De’ such that e' ∈ T',
and all a ∈ De', we have that ve’T’ (ϕx/a) = 0. Once again, since feasibility is
transitive, this suffices to guarantee for every extension T* of x, for all
extensions T' of T*, e' in T', and De’ such that e' ∈ T', and all a ∈ De’, we
have that ve’T’(ϕx/a) = 0. Thus, by the semantic rule for Σ, veT Σx(ϕx/a) = 0.
An exactly parallel argument works for the case where veT Πx(ϕx) =1.
The heredity condition holds for ΔTϕ and TrTϕ. By the semantic rule
for ΔT, veT(ΔT*ϕ) and veT(TrT*ϕ) are functions of veT*(ϕ). Likewise, where T’
is any feasible extension of T, veT ’ ( Δ T * f ) and veT ’ (TrT* f ) are the same
functions of veT*(ϕ). By the transitivity of identity, veT ( Δ T *f ) = veT ’ ( Δ T *f ),
and veT (TrT *f ) = veT ’ (TrT *f ) .
The heredity condition holds for Lϕ. By the semantic rule for L, veT(Lϕ)
depends only on veT*(ϕ), where T* is the initial segment of T containing
only e and the e’ such that eRe’. But for any extension T’ of T, T* is also the
initial segment of T’ containing only e and the e’ such that eRe’. So
veT ’ Lf = veT Lf.
10. Presentism and Distributional
Properties*
Jonathan Tallant and David Ingram
0. INTRODUCTION
Presentism is the thesis that everything that exists, exists now. The
view faces a familiar problem. Most are inclined to think that there
are truths about the past. However, if the past does not exist, then
what is it that makes true our talk about the past? To borrow from
Armstrong:
What truthmaker can be provided for the truth <Caesar existed>? The obvi-
ous truthmaker, at least, is Caesar himself. But to allow Caesar as a truthmaker
seems to allow reality to the past, contrary to [presentism]. (2004: 146)1
There are a variety of suggested solutions to the ‘truth-maker prob-

lem’. We focus upon a particular solution offered by Cameron (2011).
Briefly, Cameron claims that the truth-makers for truths about the
past are distributional properties, instantiated by present entities.
Thus, to say of Barack Obama, <He was once a boy>, is true, because
Barack Obama instantiates a distributional property, a part of which
is being-a-boy-at-t* (where t* is earlier than the utterance). We argue
that Cameron’s proposed solution fails.
1. DESIDERATA
The type of solution offered by Cameron is one that we call a ‘prop-

erties solution’: one that involves the instantiation of properties by
objects in order to do truth-making work. This type of solution is
developed by Bigelow (1996) and termed ‘Lucretianism’. Since
* We are very grateful to Karen Bennett, Ross Cameron, Jonathan Curtis, Suki
Finn, Daniel Nolan, Harold Noonan, and Dean Zimmerman for comments and feed-
back on previous drafts.
1
We follow the convention, as Armstrong does, that ‘<p>’ denotes ‘the proposi-
tion that p’.
306 | Jonathan Tallant and David Ingram
Cameron’s proposed solution is a variation on a Lucretian theme,

we now outline Lucretianism and Cameron’s adaptation.
1.1 Lucretianism
An adequate solution to the truth-maker problem must, according to
Cameron, satisfy four conditions, which are: (1) presentism; (2) truth-
maker theory (that truths about the past are made true by some exist-
ing element of our ontology); (3) realism about the past – simply, that
there are (evidence-transcendent and objective) truths about the past;
and, (4) we should not posit ‘suspicious’ properties, where a property
is suspicious iff it makes no difference to the present intrinsic nature of
its bearer.2 To illustrate how we might apply these conditions, con-
sider the true proposition <Socrates existed>. The Lucretian claims
that there is a property being-such-as-to-have-contained-Socrates and the
state of affairs of the world instantiating it. This makes true <Socrates
existed>. Thus, the world instantiates present, past-directed proper-
ties under Lucretianism, and these necessitate truths about the past.
Lucretianism therefore satisfies each of (1)–(3). The Lucretian endorses
(1) by affirming that only the present exists. Lucretians endorse (2):
what makes true our talk about the past are present properties, and so
the Lucretian can thereby satisfy (3): there are truths about the past.
But the Lucretian property of being-such-as-to-have-contained-Socrates
tells us nothing about the present intrinsic nature of the bearer, i.e. the
world. According to Cameron, the property being-such-as-to-have-
contained-Socrates makes no difference to the present intrinsic nature
of its bearer. Lucretian properties are therefore suspicious, because
they fall foul of condition (4), and should be rejected.
We are prepared to concede, here, that (4) is plausible enough.
The Lucretian is not obviously compelled to agree. Indeed, so far as
we can tell, nothing in Cameron’s paper compels a Lucretian who
does not feel the force of the ‘suspicion’ intuition to concur with
Cameron’s rejection of Lucretianism.3
2
This is based on Cameron’s view that a property is suspicious iff it fails to satisfy
his principle of Intrinsic Determination: ‘for all objects x and properties F and times t,
if x instantiates F at t, then x has the intrinsic nature at t that it has partly in virtue of
instantiating F at t’ (2011: 61).
3
The seriousness of this charge is discussed in Tallant (2009). Our preferred reason
for rejecting Lucretianism is given by Sanson and Caplan (2010) and Merricks (2007:
136–7): what we might crudely call the ‘aboutness’ objection.
Presentism and Distributional Properties | 307
1.2. Distributional Properties

Cameron attempts to adapt Parsons’ (2004) spatial distributional
properties (SDPs), an example of which is as follows. Consider an
object, O, which is white with black polka-dots, and its correspond-
ing property being-polka-dotted. This property tells us how O is
across a region of space. O can be seen as both wholly white at
places, and wholly black at others. While nothing can be both
wholly white and wholly black, instantiating the SDP, being-polka-
dotted, is enough to explain exactly how O is across the region of
space it occupies.4
Cameron argues that, in the same way, the appropriate temporal
distributional properties (TDPs) explain how objects are across
time. We borrow a story from Cameron:
Consider a simple world consisting of just one spatial dimension and one
temporal dimension. There is one entity in this world – Flatty – who starts
off his life at time t as a point, but who as time progresses grows continu-
ously in one direction of the one spatial dimension he occupies. After the
beginning of this life, then, he is no longer a point but a line; and at each
moment he is a longer line than he has ever been previously. Exactly one
year later, at t*, Flatty tragically ceases to be, and the world is empty.
(2011: 63)
On this account Flatty is a point at t and a line at t*.5 Flatty’s spa-

tial dimensions are distributed a certain way across a period of time.
Flatty is a point at one moment in time, and a line at others in
virtue of instantiating the relevant TDP. However, the TDP alone
does not give us enough to explain how Flatty is at a particular
moment in time. In order to explain this, we must appeal to a
further property, Flatty’s age at the moment in question. Given
this account, we can fix how Flatty is, was, and will be, by appeal-
ing to the TDP and Flatty’s age.
TDPs and the age property both satisfy (4). A TDP is a single
property, one that charts the nature of its bearer throughout the
4
Being-polka-dotted is just one example of an SDP, a colour-distributional property.
Other dialectically respectable SDPs include being-hot-at-one-end-and-cold-at-the-other (a
heat-distributional property), having-a-uniform-density-of-1kg/m3-throughout (a density-
distributional property), and so forth.
5
It is more accurate to say that Flatty is a line at the moment immediately preceding
t*, because at t* Flatty ceases to be. However, for simplicity, we will assume that it is
true of Flatty that he was a line at t*, and then ceases.
whole of history. At every moment, the TDP makes a contribution

to the present intrinsic nature of its bearer, and so satisfies (4). The
property of age also clearly makes a difference to how old the object
is now, thereby also satisfying (4). Thus the union of TDPs and ages
ground truths about how their bearers were, providing present
truth-makers for truths about the past, and avoiding charges of
‘suspiciousness’.
2. TENSE VS. TENSELESS
Cameron’s preferred understanding of the TDP is that it is tenseless. By

this Cameron (2011: 67) seems to mean that the TDP is akin to being-F-
then-G-then-H. Contrast this with a TDP that ‘builds in’ tense and is of
the form having-been-F-being-G-and-going-to-be-H. As Cameron notes,
if ‘tense’ were ‘built into’ the TDP, then which TDP were instantiated
by an object would have to vary from moment to moment to reflect
the fact that, given the passage of time, an object may change from
going to be F, to being F, to having been F. Thus, an object, O, bearing a
TDP would have to change from bearing the TDP having-been-not-F-
and-being-not-F-and-going-to-be-F, to bearing the TDP having-been-not-
F-and-being-F-and-going-to-be-not-F, and so on. By contrast, an object
instantiating the TDP being-F-then-G-then-H does not need to change
which TDP it instantiates, over time, to generate the appropriate truth-
values. As we explained above, the ‘tenseless’ TDP and the property
of age are sufficient to ground the tensed truths.
2.1. Changing TDPs

We think it possible that an object instantiating one of Cameron’s
(tenseless) TDPs could change over time in such a way that it instan-
tiates different TDPs at different times and we also think that this is
a problem for Cameron’s view.6 Suppose, to illustrate, that we have
the following natural progression through time, where we allow
that the underlined portions of the TDP are those that are ‘now’:
6
Cameron (2011: 76–7) considers this problem as his ‘Objection 5’.
(A) The rose is red and then the rose is dead.7

This would most naturally be followed by:
(B) The rose is red and then the rose is dead.
The progression described would simply be one in which we move
from one portion of the TDP—the rose is red—being present, to
another portion of the TDP—the rose is dead—being present. So far,
this all looks as it should.
But we maintain that the following sequence is possible. Begin as
before:
(A) The rose is red and then the rose is dead.
And then suppose that this is followed by:
(B*) The rose is green (not red) and then the rose is dead.
This situation is one in which the property bearer—the rose—
changes with respect to which TDP it instantiates; first one TDP is
instantiated, and then another. This second TDP has a different first
portion. We take it that this result would be undesirable.8 The sce-
nario described above is one in which <The rose is red> is true, but
where it is never true, later than this, that <The rose was red>.
After all, given (B*) what will be true about the past is that <The
rose was green>. This, we think, is sufficient to motivate a rejection
of Cameron’s view.
We have spoken as if TDPs have ‘portions’ or ‘parts’. Cameron
does not licence such talk, claiming that such properties cannot be
broken down into simple components. If TDPs cannot be broken
down, then how can we say that TDPs with different parts can be
instantiated at different times?
We are prepared to concede the point, but do not think it pressing
to the aims of the paper. A moment ago, we described a sequence
that involved a transition from (A) to (B*). We think it perfectly
intelligible to think of the transition being generated by the replace-
ment of one TDP (the TDP described in (A)) with another (the TDP
described in (B*)). We therefore maintain that the objection stands.
7
These TDPs are ‘tenseless’ in the way described.
8
This problem is discussed for presentist ersatzism (e.g. Crisp (2007)) by Tallant
(2010).
Talk of ‘portions of TDPs’, though useful, is not essential to the

setting up of the problem.
2.2. Cameron on change and the TDP

Cameron, in part, anticipates our objection:
It makes no sense to speak of an object changing its distributional properties.
Why? Because what change is on the account being offered is to instantiate (at
each moment of your existence) a non-uniform distributional property.
Being red at one time and then orange at some later time, for example, is to
be analysed as instantiating (at all times) the distributional property being
red-then-orange. To speak of an object changing its properties is a loose way
of saying something about the distributional property it has that says how it
is across time; it makes no sense to speak of an object gaining or losing the
property that says how it is across time. (2011: 77)
So far as we can tell, this contradicts Cameron’s earlier treatment of

age properties.
As we noted in 1.2, Cameron’s solution requires the union of a
(tenseless) TDP and the property of age. Clearly, the property of age
that is instantiated by any given object must change over time. We
do not now bear the same age-property that we bore five years, five
minutes, or even five seconds ago. Thus, to ‘speak of an object
changing its properties’ is not merely a loose way of saying some-
thing about the distributional property it has across time.
We present Cameron with the following dilemma: the property
of age must be either a tenseless distributional property or some
other kind of property. Suppose, taking the first horn, that the prop-
erty of age is a tenseless distributional property. If that property is
tenseless, then it will fail to reflect the fact that we are now one of
those particular ages. Rather, the distributional property simply
exists, describing how we are at all times. Suppose, to illustrate, that
Cameron instantiates a TDP and a property of age. The TDP is of
the form being-a-boy-and-then-being-a-man-and-then-being-an-old-
man. To simplify, we suppose that Cameron exists for only three
instants, and that each clause of the TDP makes true our talk about
a distinct instant.
We assume that a TDP age property would be something like the
property of being one-instant-old-then-two-instants-old-and-then-three-
instants-old. But we think it plain that this property, in conjunction

with the TDP being-a-boy-and-then-being-a-man-and-then-being-an-old-
man, does nothing to fix the present intrinsic nature of Cameron’s
person. The two properties simply serve to describe the whole
lifespan of the person and are not sensitive to what is now the case.
This is insufficient; nothing in either of the TDPs permits us to fix
how the person is now. Since, as we explained above, this is essential
to our fixing the truths about the past, so Cameron’s view fails.
We now turn to the second horn: there is change that is not
accounted for merely by the instantiation of a non-uniform TDP. If
age is a property, e.g. the property being one instant old, but is not a
distributional property, then change is not merely a matter of something
bearing a distributional property. An object will have to change from
being-now-five-years-old to being-now-six-years-old, for example; but
this change will be due to an ever-changing, non-distributional prop-
erty of age. Thus, some change would not be accounted for by TDPs.
If that is right, then Cameron loses the leverage to insist that an
object cannot change with respect to which TDP it instantiates.
Cameron’s original line of resistance was that an object cannot
change its TDPs because change just is the instantiation of distribu-
tional properties; to say that ‘distributional properties may change’
simply makes no sense. However, if a change in the property of age
is not accounted for solely via the instantiation of a non-uniform
TDP, then change does not consist solely in the instantiation of a
non-uniform TDP. Clearly, it cannot then fail to make sense to say
that ‘things change in ways other than by instantiating a non-
uniform distributional property’.9 We therefore see no reason to
deny that an object can change which TDP it instantiates, over time.
If the idea is coherent and conceivable (and we think we have dem-
onstrated that it is10) then we think it is possible.11
9
Cameron might reply that it is only a change in an object that requires the instan-
tiation of the TDP, and once again attempt to run his argument. We do not see that
this will work: age, surely, is a property of objects.
10
There is no obvious contradiction in the above representation of the change of
TDPs over time. Recall, Cameron forced the contradiction by insisting that change
was simply variation in the tenseless TDP. Since that is not the case, there is no longer
any contradiction.
11
Elsewhere, Cameron (2010: fn.14) makes it clear that he thinks possibility is the
‘default mode’.
Nor can Cameron abandon the property of age. As Cameron

has it:
Flatty couldn’t have the particular distributional property and age that he
has without being pointy at t and without being that length of line at t*, and
it is because he instantiates this distributional property and age that he is
pointy at t and that length of line at t*. And so the existence of the state of
affairs of Flatty instantiating this distributional property and age at t* neces-
sitates that Flatty was pointy at t, and can suffice as a truthmaker for that
truth about the past. (2011: 65)
We share the thought that although TDPs themselves are sufficient

for all of the tenseless truths, about (e.g.) Flatty, it is only when the
TDP is conjoined with the property of Flatty’s age that we generate
the claim that Flatty is now, and was and will be, any particular way.
Since the tensed truths about the past and the future were the pri-
mary target for grounding, we cannot do very well without the
property of age on the tenseless account of TDPs. Thus, denying the
existence of the property of age, whilst preserving tenseless distri-
butional properties, does not seem viable.
So, instead of denying that there exists an age property, can Cameron
claim that age properties are special; that they are the one property for
which change is not a matter of instantiating a TDP? He might be able
to deny that such a move would be ad hoc. After all, the age property
is basically a surrogate for the ‘now’ of a moving spotlighter.12
We agree that Cameron could take such a position, but we are
less certain as to how this will solve the problem. Cameron’s origi-
nal view was that change just is variation in the TDP. This permitted
him to argue that it would make no sense to say that objects undergo
changes in which TDPs they exemplify. If Cameron then allows that
there can be change that is not due to simply having a single change-
implying TDP throughout one’s career, then this line of argument
cannot be mounted. Once Cameron concedes that there is a type of
change that requires the gaining and loss of a particular property –
the property of age – it becomes unclear why it makes no sense to
describe objects as changing with respect to which TDP they exem-
plify. If objects can gain and lose age properties we see no reason
that they not also gain and lose TDPs.
12
We owe this objection to Karen Bennett.
Cameron might then try to deny that it’s possible for objects to
change with respect to which TDPs they exemplify, but we do not
think Cameron should take this line. We take the view that possibil-
ity is the default status for a proposition that seems conceivable,
and that if Cameron wishes to take the view that some element of
our ontology is necessarily unchanging in some regard, then this
generates particular dialectical requirements. Namely, Cameron
must explain why objects cannot gain or lose TDPs.
To see the explanatory burden more clearly, we think it is worth
considering the claim that objects cannot change with respect to which
TDPs they exemplify, in light of claims typically made about concreta
and abstracta. We naturally think of concreta as capable of undergoing
change. Were we to claim of any concrete object that it is impossible for
it to undergo any change, then we think some explanation must be
forthcoming as to why this is the case. For example, were Cameron to
tell us that there was a concrete object that instantiates the property
‘being green’ and, of necessity, cannot lose this property and/or gain
another colour property, despite the fact that this object can gain and
lose other properties, we should want some explanation of why this is
the case. It would not do to simply assert that this is how things are.
Nor, then, should Cameron merely assert that an object cannot change
which TDP it exemplifies, whilst simultaneously allowing that an
object can change which age property it exemplifies.
Now consider abstracta; putative examples of which include
numbers, sets, and propositions. These entities are often thought to
be necessarily unchanging. Certainly, we do not think that ‘2’ can
undergo any change in its properties over time. However, the con-
trast between objects bearing TDPs and abstracta is marked. Abstracta
are usually thought to exist ‘outside’ time, in some sense or other.
Indeed, they lack any location in time. That abstracta lack temporal
location renders them unchanging. Because abstracta do not reside
at any point along the temporal dimension, we have an explanation
of why it is that they cannot change their properties over time. Some
similar explanation is required if we are to endorse Cameron’s posi-
tion, but it is hard to see how the appeal to abstracta can help Cam-
eron here. For, as already noted, the bearers of TDPs, unlike abstract
objects, have a location in time. As a consequence of these argu-
ments we think that Cameron’s view should be rejected.
University of Nottingham
REFERENCES
Armstrong, David (2004) Truth and Truthmakers (Cambridge: Cambridge
University Press).
Bigelow, John (1996) ‘Presentism and properties’ Philosophical Perspectives
10, 35–52.
Cameron, Ross (2010) ‘From Humean truthmaker theory to priority mon-
ism’ Noûs 44:1, 178–98.
—— (2011) ‘Truthmaking for presentists’ in Karen Bennett and Dean
W. Zimmerman (eds.) Oxford Studies in Metaphysics, vol. 6 (Oxford:
Oxford University Press), 55–100.
Crisp, Thomas (2007) ‘Presentism and the grounding objection’ Noûs
41:1, 90–109.
Merricks, Trenton (2007) Truth and Ontology (Oxford: Clarendon Press).
Parsons, Josh (2004) ‘Distributional properties’ in Frank Jackson and Gra-
ham Priest (eds.) Lewisian Themes (Oxford: Oxford University Press),
173–180.
Sanson, David and Ben Caplan (2010) ‘The way things were’ Philosophy
and Phenomenological Research 81:1, 24–39.
Tallant, Jonathan (2009) ‘Presentism and truth-making’ Erkenntnis 71:3,
407–16.
—— (2010) ‘Time for presence?’ Philosophia 38:2, 271–80.
NAME INDEX
Achinstein, P., 132 Cross, T., 134

Adams, R. M., 220, 267 Crossley, J., 190 n. 1
Adams, R., 220
Al-Ghazali, A., 132 Davidson, D., 129
Alston, W., 4, 115 n. 21 Davies, M., 185, 190 n. 1
Ampere, A., 163 Descartes, R., 134
Annamalai, K., 168 n. 8 Doepke, F., 245
Armstrong, D., 62, 129, 131, 202, Duffin, W. J., 185
305 Dummett, M., 287
Austin, J. L., 53
Eklund, M., 96 n. 3, 104 n. 9,
Bartlett, D., 164 117 n. 22
Bealer, G., 202 Ellis, B., 155, 177, 181–2, 214
Benacerraf, P., 111 n. 16
Bennett, J., 4 Fara, D. G., 32 n. 16, 41–2, 45
Bennett, K., 96 n. 2, 104 n. 9, 221, Fara, M., 39 n. 19, 137
246 Fermi, E., 168
Biel, G., 132 Feynman, R., 154–5, 161
Bigelow, J., 178, 305 Field, H., 86 n. 25
Bird, A., 133, 138–9, 148, 155, Fine, K., 99 n. 5, 101, 190, 214–17,
177–85, 214 249
Block, N., 106 Flaubert, G., 72 n. 8
Bohn, E. D., 71 n. 5 Forrest, P., 266
Boolos, G., 73 n. 11 Freddoso, A., 132
Bourne, C., 259–60, 266, 295–6, 298 Frege, G., 104 n. 9, 111
Braddon-Mitchel, D., 266
Bricker, P., 222 Geach, P., 4–5
Broad, C. D., 257, 300 Gettier, E., 129
Burgess, J., 112, 189 Gibbard, A., 9 n. 5, 14 n. 8, 35,
Burke, M., 246 246
Gibbs, J. W., 166, 167
Callander, C., 163 Gödel, K., 83
Cameron, R., 305–12 Goldfarb, W., 100, n. 7
Caplan, B., 306 n. 3 Goldman, A., 129
Chalmers, D., 94, 198, 202, 209–10 Goodman, N., 130
Chandler, H., 193 Graham, A., 107 n. 13
Chisholm, R., 5–6, 18 n. 11, Greenough, P., 272
Cohen, P., 61 Grice P., 129
Coulomb, C., 155, 163, 169, 173 Gupta, A., 5 n. 2, 42
316 | Name Index
Hacker, P., 100 n. 7 McCall, S., 259, 295–9

Harre, R., 132 McDaniel, K., 222
Hawley, K., 8 McGee, V., 84 n. 25
Hawthorne, J., 134 McKay, T., 70 n. 3, 89 n. 29
Hazen, A., 28–9, 38 n. 19, 48–9, 73 McKitrick, J., 148
n. 10 McMichael, A., 223
Heil, J., 96 n. 3, 147 Menzel, C., 222
Heim, I., 97 Merricks, T., 306 n. 3
Hertz, H., 162–3, 164–5 Mumford, S., 155, 177
Hilbert, D., 60
Hirsch, E., 104 n. 9, 112 n. 20 Ne’eman, Y., 185
Hofweber, T., 100 n. 6 Newton, I., 155, 162, 163
Holton, R., 148 Nielson, H., 143
Humberstone, L., 190 n. 1 Ninomiya, M., 143
Hume, D., 130 Noonan, H., 12, n. 6
Nozick, R., 129
Jackson, F., 108, 198, 209–10
Johnston, M., 12 n. 6, 139 Olsen, E., 227 n. 1, 246, 247
Jubien, M., 6 n. 4, 17 n. 9, 18 n. 10 n. 1, 248
Kaplan, D., 209 Parsons, J., 112, 119 n. 23, 307

Kirsch, Y., 185 Pears, D., 100 n. 7
Kneale, W., 132 Perry, J., 227 n. 1
Kratzer, A., 97 Planck, M., 167–9
Kripke, S., 7, 8, 18, 27, 38, 51–5, 129, Plantinga, A., 220, 222
190, 193, 209, 211, 215, 222, 223, Puri, I., 168 n. 8
286 Putnam, H., 104 n. 9, 113, 129, 132
Lange, M., 155, 163, 171, 179 Quine, W. V. O., 34, 43

Lederman, L., 185
Leibniz, G. W., 48, 190 Ramachandran, M., 38 n. 18
Lewis, D., 7–9, 12, 15, 16, 18, 22–5, Raw, G., 168 n. 8
27, 36, 38, 41, 44–6, 48–57, 59, 82 Rayo, A., 70 n. 4, 73 n. 11, 84 n. 24,
n. 22, 84 n. 23, 99 n. 4, 05 n. 10, 93 n. 1, 96 n. 2, 107 n. 12, 108 n.
109, 130–2, 136–50, 201, 202, 15, 111 n. 17, 115, 119 n. 24
222–4 Rosen, G., 104 n. 9, 112
Lierse, C., 178
Linnebo, Ø., 119 n. 24 Salmon, N., 86 n. 27, 87 n. 28, 190,
Łukasiewicz, J., 274, 275, 284–8, 194–201, 208, 212, 213, 224
292, 293 Sanson, D., 306 n. 3
Schaffer, J., 71 n. 5, 99 n. 5, 100 n. 6,
MacFarlane, J., 259, 299, 300 134
Madden, E., 132 Sellars, W., 132
Marshal, D., 58 Seth, S., 168, 169
Martin, C. B., 134, 138, 147, 148 Shapiro, S., 73 n. 11
Name Index | 317
Shoemaker, S., 129, 131–6, 140, 141, Unger, P., 251 n. 15

143, 149, 150, 214–16, 227 n. 1, 230 Uzquiano, G., 68, 69 n. 4, 81 n. 21
n. 3, 232 n., 248
Sider, T., 6 n. 4, 8, 18 n. 10, 99 n. 5, Van der Merwe, A., 168 n. 8
100 n. 6, 192 Van Inwagen, P., 84 n. 23, 110,
Soames, S., 86 n. 25 222
Sorenson, R., 272 Varzi, A., 265 n. 4
Spencer, C., 223
Stalnaker, R., 6, 41, 106, 112, 146, Ward, B., 146
209, 220 Weinberg, S., 184
Stevin, S., 161 n. 5 Wigner, E., 162
Su, Y., 164 Williamson, T., 39 n. 19, 70 n. 4,
93 n. 1
Tallant, J., 306 n. 3, 309 n. 8 Wittgenstein, L., 100
Teresi, D., 185 Wright, C., 104 n. 9, 112
Thomason, R., 276–8, 280, 289
Thomson, J., 250 Yourgrau, W., 168 n. 8
Tooley, M., 297 Yukawa, H., 165

Oxford Studies in Metaphysics Volume 7

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Oxford Studies in Metaphysics Volume 7

Transféré par

Droits d'auteur :

Formats disponibles

OXFORD STUDIES IN METAPHYSICS

OXFORD STUDIES IN METAPHYSICS

Editorial Advisory Board

Oxford Studies in Metaphysics is dedicated to the timely publication

Oxford Studies in Metaphysics: Younger Scholar Prize ix

1. Possibility Relative to a Sortal 3

II. ABSOLUTE GENERALITY

3. All Things Must Pass Away 67

III. HUMEANISM AND LAWS OF NATURE

5. Goodbye, Humean Supervenience 129

IV. CONTINGENT OBJECTS AND COINCIDENT OBJECTS

7. Relativized Metaphysical Modality 189

V. THE OPEN FUTURE

9. The Real Truth about the Unreal Future 257

Name Index 315

Sponsored by the Ammonius Foundation* and administered by the

Previous winners of the Younger Scholar Prize are:

* The Ammonius Foundation is a non-proﬁt organization dedicated to the revival

Cody Gilmore, “Time Travel, Coinciding Objects, and Persist-

Enquiries should be addressed to:

1. THE LEADING IDEA

There is a distinction between identity and sameness. There is only

Some have used the term “relative identity” to mean what I am

Although I think that there is a distinction between sameness and

For example, I propose that a certain woman might have worn

2. MOTIVATING THE IDEA

Why develop a relative-sameness counterpart theory of de re modal-

One could lamely reject the reasoning by noting that identity

matter that is different from the matter that actually

3. DEVELOPING THE IDEA

Symmetry and transitivity are logical properties of relative same-

We will say that a relation is a weak equivalence relation if it is sym-

Although I call my theory a counterpart theory, I do not accept all

4. SHARPENING THE IDEA

Evidently, the relative-sameness relation same-boat must be relativ-

and br must be a boat in the counterfactual world. Only boats can be

4.1 World-Relativized Relative Sameness

relative-sameness relation to <y, w’> depends on what x is like at w

4.2 Formalizing World-Relativization

objects. We are regarding these relations as relations on genuine

4.3 Relative Sameness and Identity

they are relativized to a nonvacuous sort of thing—that is, to a

Unlike Lewis’s qualitative-similarity counterpart relation, the rela-

4.4 Relative Sameness and Superlative Qualitative Similarity

(So there are different instances of F-mates: boat-mates; person-

Lewis’s view, we cannot speak of the F-ish counterpart of x in w’,

5. APPLYING THE IDEA

So how, after all this, do we do what we wanted? What we wanted

Angel has a boat-mate and Lumber a wood-mate that are distinct

6. DEFENDING THE IDEA

The main theoretical advantage of counterpart theory (in my liberal

6.1 Comparison with Superlative Qualitative Similarity

sarily according to QSC◽ is for every possible world to be one in

Take the property of being born a human. I was essentially born

For the similarity-based counterpart theorist, who allows for multiple

But the similarity-based counterpart theorist is clearly wrong on

As we said, similarity-based counterpart theory never makes room

larity-based counterpart theorist analyzes modal claims with poly-

worlds in which nothing exists that is the same person as Adam or

6.2 Substitution Failures? No Problem

(A) Relative to a sortal F, x is possibly Φ just in case there is a

he must quantify over a thing’s counterparts in a possible world.

sameness counterpart theorist need not say anything as radical as this,

I intend for this paper to be an informal companion of my ’Relative-Sameness