Oxford Studies in Metaphysics Karen

Bennett
Visit to download the full and correct content document:
https://ebookmass.com/product/oxford-studies-in-metaphysics-karen-bennett/
OXFORD STUDIES IN METAPHYSICS OXFORD STUDIES
Editorial Adviso1y Board IN METAPHYSICS
Elizabeth Barnes (University of Virginia)
Ross Cameron (University of Virginia)
David Chalmers (New York University and Australasian National University) Volume 11
Andrew Cortens (Boise State University)
Tamar Szabo Gendler (Yale University)
Sally Haslanger (Min
John Hawthorne (University of Southern California)
Mark Heller (Syracuse University)
Edited by
Hud Hudson (Western \Vashington University)
Kathrin Koslicki (University of Alberta) Karen Bennett
Kris McDaniel (Syracuse University) and
Trenton Merricks (University of Virginia) Dean W. Zimmerman
Kevin Mulligan (Universite de Geneve)
Laurie Paul (Yale University)
Jonathan Schaffer (Rutgers University)
Theodore Sider (Rutgers University)
Jason Turner (University of Arizona)
Timothy Williamson (Oxford University)

Managing Editor
Christopher Hauser (Rutgers University)

OXFORD
UNIVERSITY PRESS
 OXFORD
UNIVERSITY PRESS PREFACE
Grear Clarendon Street, Oxford, OX2 6DP.
United Kingdom
Oxford Unin~rsity Press is a dcpamnem of the University of Oxford. Oxford Studies in Metaphysics is dedicated to the timely publication of new
It furthers the University's objective of excel!ence in research, scholarship. work in metaphysics, broadly construed. The subject is taken to include not
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain mher countries
only perennially central topics (e.g. modality, ontology, and mereology) but
© the several conrribmors 2018
also metaphysical questions that emerge within other subfields (e.g. phil-
osophy of mind, philosophy of science, and philosophy of religion). Each
The moral rights of the authors have been asserted
volume also contains an essay by the winner of the Sanders Prize in
First Edition published in 20 l 8
Impression: 1
Metaphysics, an biennial award described within.
All rights reserved. No part of this publication may be reproduced. stored in K. B. & D. W. Z.
a retrieval system, or transmitted, in any form or by any means. without the
prior permission in wriring of Oxford University Press, or as expressly permitted New Brunswick, NJ
by !aw, 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 acquircr

Published in the United States of America by Oxford University Press

198 Madison Avenue, New York, NY 10016. United States of America
Brirish Libraty Cataloguing in Publication Dara
Data available
Library of Congress Control Number: 20 l 8940688
ISBN 978-0-19-882819-8 (hbk.)
978--0-19-882820-4 (pbk.)
Printed and bound by
CPI Group (UK) Ltd, Croydon, CRO 4YY
Links to third party websites are provided by Oxford in good fairh and
for information only. Oxford disclaims any responsibility for the materials
contained in any third parry website referenced in this work.
CONTENTS

The St1nders Prize in /Vfett1physfo· ix

PART I. RELATIONALISM AND SUBSTANTIVALISM

;\ New Approach to the Relational-Substantival Debate 3

Jill North
2 Relative Locations 44
Andrew Bacon

PART II. TIME AND CHANGE

3 A Passage Theory of Time 95

Martin A. Lipman
4 Fragmenting the Wave Function 123
Jonathan Simon

PART III. RECOMBINATION, RELATIONS,

AND SUPERVENIENCE

5 Possible Patterns 149

Jeffrey Sanford Russell and john Hawthorne
6 Plural Slot Theory 193
T. Scott Dixon
7 Local Qualities 224
Elizabeth Miller

PART IV. VAGUENESS

8 Vague Naturalness as Ersatz Metaphysical Vagueness 243

Rohan Sud
9 Against 'Against "Against Vague Existence"' 278
Roberto Loss

Author Index 289

THE SANDERS PRIZE IN METAPHYSICS

Sponsored by the Marc Sanders Foundation* and administered by the

editorial board of Oxford Studies in Metaphysics, this essay competition is
open to scholars who are within fifteen years of receiving a PhD and students
who are currently enrolled in a graduate program. (Independent scholars
should enquire of the editors to determine eligibility.) The award is
$10,000, and the competition is now biennial. Winning essays will appear
in Oxford Studies in lvfetaphysics, 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 November. To be eligible for
the next prize, submissions must be electronically submitted by 31 January
2020. Refereeing will be blind; authors should omit remarks and references
that might disclose their identities. Receipt of submissions will be acknow-
ledged 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 Metaphysics, independently of the prize.
Previous winners of the Sanders Prize are:
Thomas Hofweber, "Inexpressible Properties and Propositions", Vol. 2;
Matthew McGrath, "Four-Dimensionalism and the Puzzles of Coincidence",
Vol. 3;
Cody Gilmore, 'Time Travel, Coinciding Objects, and Persistence", 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", Vol. 8;
Louis deRosset, "Analyticity and Ontology", Vol 9;

' The Marc Sanders Foundation is a non-profit organization dedicated to the revival of
systematic philosophy and traditional metaphysics. Information about the Foundation's
other initiatives may be found at <11ttp://www.marcsandersfoundation.com/>.
x The Sanders Prize in Metaphysics

Nicholas K. Jones, "Multiple Constitution", Vol. 9;

Nick Kroll, "Teleological Dispositions", Vol. lO;
Jon Litland, "Grounding Grounding", Vol. 10;
Andrew Bacon, "Relative Locations", Vol. 11;
T. Scott Dixon, "Plural Slot Theory", Vol. 11. PART I
Enquiries should he addressed to Dean Zimmerman at:
dwzimmer@rci.rutgers.edu.
RELATIONALISM AND
SUB ST ANTIVALISM
 1
A New Approach to the
Relational-Substantival Debate
Jill North

1. INTRODUCTION

The traditional relational-substantival debate is about whether space-in

modern terms, spacetime-exists. The substantivalist says that it does. The
relationalist says that it doesn't. According to the relationalist, all that exists,
in the physical world, are material bodies related to one another spatiotem-
porally; there is no further thing in which these bodies are located.
This is a debate with a long histo1y. Yet there is still surprisingly little
agreement not only on what is the right answer, but also on how to
understand the very question at issue and the potential answers to it-and
even on whether there is any genuine dispute here. For example, we can try
to formulate the debate in a way that harkens back to the traditional
Leibniz-Newton dispute, as the question of whether space exists as a
substantial entity. But then what it means to call something a substantial
entity is disputed, so that it may start to seem like the two sides are simply
talking past each other.
Some people have concluded that the debate is not substantive. Perhaps it
is merely a verbal dispute about which things to call 'space' versus 'matter',
with no objectively correct answer to be had (Rynasiewicz, 1996). Others
have thought that the dispute has stagnated or become divorced from
physics. 1 A review of the historical dispute and its central examples (Newton's
bucket and globes, Leibniz's shifts, Kant's glove, as well as the more recent

1 Claims that the traditional debate is non-substantive, unclear, or removed from

physics, either in certain contexts or in general, can be found in Stein (1970; I 977b);
Malament (1976); Horwich (1978); Friedman (1983, 221-3); Earman (1989); DiSalle
(1994); Leeds (1995); Rynasiewicz (1996; 2000); Dorato (2000; 2008); Belot and
Earman (2001, sec. 10.7); Pooley (2013, sec. 6.1, 7); Curiel (2016); Slowik (2016).
Earman (1989) advocates the need for a te11ium quid.
4 Jill North
hole argument, all of which live on in today's discussions) may reasonably
suggest a stagnated debate. Each of these aims to show that the opposing side
recognizes either too few or too many spatiotemporal facts for the physics; but
there are various maneuvers, well hashed-out in the literature, allowing each
side to escape the charge. Relatedly, given the variety of different understand-
ings of the dispute, you might think that there is no overarching, well-posed
question in the vicinity (Curiel, 2016). David Malament is not alone in
wondering whether there is any clear-cut dispute between the two sides:
"Both positions as they are usually characterized ... are terribly obscure. After
they are qualified so as to seem intelligible ;md not too implausible, it is hard to
retain a firm grasp on what divides them" ( 197 6, 317). Certainly all of this hints
at "the fragile health of the substantival-relational debate" (Belot, 1999, 38).
These are reasonable concerns when leveled at traditional conceptions of
the dispute. Nonetheless, I believe that there is a debate that is substantive,
not stagnant, and relevant to physics. The debate that I will present is not
exactly the traditional one. But it is close enough in spirit that I think it is
the best way of understanding that dispute, updated to take into account
more recent developments in physics and philosophy. And once we frame
the debate in this way, we unearth a novel argument for substantivalism,
given current physics. At the same time, that conclusion could be overrid-
den by future physics. A seemingly subtle shift yields surprising progress on
a longstanding issue that many people feel has stagnated.
In Section 2, I discuss an idea that will play a central role: structure in
general, and spatiotemporal structure in particular. I will argue that, regard-
less of whether you are a relationalist or substantivalist, you should think
that there are objective, determinate spatiotemporal facts about a world: you
should be a realist about spatiotemporal structure in my sense. This follows
from a general principle we rely on in physics. (The traditional debate was
about the existence of space and time separately. I discuss the question of
spacetime, or spatiotemporal structure, updating things to the terms of
modern physics.) In Section 3, I will argue that, regardless of whether you
are a relationalist or a substantivalist, you can be a realist about spatiotem-
poral structure. I do this by framing the debate in terms of fundamentality
and ground, notions that have gotten lots of press recently in metaphysics.
I show that this way of putting things captures traditional conceptions of
the dispute, while allowing us to formulate the most plausible-if not
entirely traditional-versions of the two main positions on it. (Although
I put things in terms of ground, what's most important is that we make use
of some notion of relative fundamentality.) Finally (Sections 4 and 5), I put
all the pieces together to show that there is a powerful argument for
substantivalism, or at least a powerful challenge to relationalism, given much
of current physics.
A New Approach to the Relational-Substantival Debate 5
At the end, I briefly discuss how the conclusion in favor of substantivalism
may change witl1 future developments in physics. Yet however the physics
turns out, the question of relationalism versus substantivalism should be
settled by means of the new type of argument offered here. Hence, if I am
right, the substantivity of the debate is secured regardless of future develop-
ments in physics, while the conclusion in favor of one view or the other will
ultimately be decided by the physics.
2. SPATIOTEMPORAL STRUCTURE AND
THE MATCHING PRINCIPLE
I'll begin by arguing that both the relationalist and the substantivalist should
posit enough, and not too many, spatiotemporal facts for the physics. As
I will put it, they both should countenance the spatiotemporal structure that
is needed for the physics. (In Section 3, I turn to whether they both can
do this.) I argue that there is a certain methodological principle we are used
to relying on in physics, even if it is not usually mentioned. This principle
guides our inferences from the mathematical formulation of a theoty to the
nature of the world according to the theory. I show by example that we do
generally, and successfully, rely on this principle. The conclusion about
spatiotemporal structure will follow from it.
Consider classical Newtonian mechanics. What does this theory tell us
about the world? Newton thought it tells us that absolute space, a space that
persists through time, exists. He argued that phenomena involving inertial
(unaccelerated) and non-inertial (accelerated, in particular rotated) motion
reveal this. (Think of his bucket experiment and the spinning globes
example.) Although we nowadays agree that the phenomena indicate a
real distinction between inertial and non-inertial motion, we think that
Newton was wrong about what's required to account for this distinction.
In today's terms, Newton was arguing for substantivalism about what is
ofren called Aristotelian, or Newtonian, spacetime. 2 This spacetime has the
structure to support Newton's idea of absolute space, for it has structure that
identifies spatial locations over time. But we now know (as Newton did not)
that Galilean, or neo-Newtonian, spacetime also supports the distinction
between accelerated and unaccelerated motion, without absolute space.
Spelling this out, Aristotelian spacetime has all the structure of Galilean
spacetime, but it also has absolute space, or an absolute standard of rest or
2
Not to be confused with the spacetime that Earman (1989, sec. 2.6) calls 'Aristorel-
ian.' I follow Geroch's (1978) use of the 'Aristotelian' and 'Galilean' labels.
6 Jill North
preferred rest frame. To remind you of what this means, think of an observer
on a platform and another observer on a train moving with constant velocity
relative to the platform. Each observer feels that he or she is at rest and that
the other is moving. Galilean spacetime says that neither one is "correct" or
at rest in any absolute, observer-independent sense. Each is simply in
motion relative to the other, and at rest in her own frame of reference.
(Think of a reference frame as a coordinate system attached to an observer,
representing her own point of view.) According to an Aristotelian spatio-
temporal structure, there is an observer- or frame-independent fact, from
among all the observers in constant relative motion, about which one is at
rest in an absolute, frame-independent sense-namely, the one at rest in
absolute space. For there is a frame-independent fact about whether a given
spatial location is the same location over time, so that an object located there
is at absolute rest. In other words, there is a preferred rest frame: the one
that's at rest with respect to absolute space.
Intuitively, an Aristotelian spatiotemporal structure has more structure than
a Galilean one. It has all the same structure, plus an additional absolute-space,
or absolute-velocity, structure. It recognizes all the same spatiotemporal facts,
but it also says that there are facts about how fast an object is moving with
respect to absolute space.
It turns out that these additional facts are not needed for, or recognized
by, the physics here. Newton's laws are the same in any inertial frame-they
are invariant under changes in inertial frame-which means that they can be
formulated without mentioning or presupposing a preferred frame. Since a
preferred frame isn't needed in the mathematical formulation of the laws, we
infer that it doesn't correspond to anything physical in the world. An
absolute standard of rest isn't part of the theo1y's, or world's, spatiotemporal
structure. The physics does not recognize objective, frame-independent facts
about what velocity an object has. Conclusion: Aristotelian spacetime has
excess, superfluous structure, as far as Newton's laws are concerned. It
recognizes more spatiotemporal facts than the laws do.
These laws do recognize facts about objects' accelerations (as Newton
argued). Think of Newton's first law: an object travels with uniform velocity
unless acted on by a net external force. This law assumes that there is a
distinction between accelerated and unaccelerated motion, since it tells
things to behave differently depending on whether they are accelerating or
not. In terms of spatiotemporal geometry, the law assumes a distinction
between straight and curved trajectories or paths through spacetime, with
the straight ones corresponding to inertial motion, the curved ones to non-
inertial motion. And Galilean spacetime has the structure to support this
distinction. It has an affine connection, or inertial structure, which provides
a standard of straightness for these trajectories. We might put it like this: this
A New Approach to the Relational-Substantival Debate 7
spatiotemporal structure supports a notion or quantity of absolute acceleration
but not of absolute velocity-"absolute" not in Newton's sense, which assumes
the existence of absolute space, but in the sense of being invariant or frame-
independent. 3
All of this suggests that a Galilean spatiotemporal structure is the right
structure for Newton's physics. This is the structure that's required for, or
presupposed by, the dynamical laws; the structure that recognizes the spatio-
temporal facts that the laws do. 4 Newton was wrong to think that a classical
world must contain absolute space and a concomitant quantity of absolute
velocity: the physics doesn't require it. (If the laws were not invariant under
changes in inertial frame, then we would infer that extra structure. Such laws
would implicitly refer to a preferred frame.) Notice that we reached this
conclusion about the structure needed for the laws independently of the
relational-substantival debate, an idea that I will return to soon. 5
First let me say a bit about "structure." On my understanding (and as it is
often used in physics and mathematical physics), 6 structure has to do with
the invariant features or quantities, which are the same in all allowable
reference frames or coordinate systems. Inertial structure, for example, is
part of a classical spatiotemporal structure: there is an absolute, frame-
independent notion of accelerated versus unaccelerated motion. But there
is no "absolute-velocity structure." An object's velocity depends on the
inertial frame we use to describe it. Since Newton's laws are invariant
under changes in inertial frame, we infer that the choice of frame is an
arbitrary choice in description, and that any quantity depending on that
choice, like velocity, is merely frame-dependent, not out there in the world
apart from that choice.
Similarly, we think that a choice of origin is just an arbitrary choice in
description, not corresponding to genuine structure in the world. Choose a
coordinate system with a different origin, and the laws always remain the
same. Since the laws are invariant under changes in origin-they "say the
3 I believe that this sense evades Rynasiewicz' s (2000) arguments against the clarity of
any absolute/relative distinction.
4
Although the inference to a Galilean structure is now relatively standard (Earman
(1970); Stein (1970); Huggett (1999, 194-5); Maudlin (2012, ch. 3)), there is room for
debate. Saunders (2013) and Knox (2014), in different ways, argue that Newtonian
physics requires a different structure. I continue as though the above inference is correct. It is
in any case agreed that absolute space is not needed, and whatever structure is required, the
example illustrates our reliance on the upcoming principle.
5 A similar point is made by Stein (1970, 271-2), although he goes on to say that, "the
question whether ... this structure of space-time also 'really exists', surely seems to be super-
erogatory" (277). In a way I agree, but I also think that there remains a substantive dispute.
6
More is in North (2009).
8 Jill North
same thing" 7 regardless-we infer that this choice is merely a conventional
or arbitrary choice in description. There is no preferred-location structure
in the world, no coordinate-independent fact about whether a given point
is "really" the origin. By contrast, the laws of Aristotle's physics are not
invariant in this way. According to them, there is a preferred-location struc-
ture in the world-a location toward which certain elements naturally fall
and away from which others naturally rise-and preferred coordinate systems
for describing this structure, namely those with an origin at that location.
We likewise think that different choices of unit of measure are conven-
tional or arbitrary choices in description. Change from feet to meters or
some other unit for measuring distances, for instance, and the physics always
remains the same. Since the physics says the same thing regardless, we infer
that there is no "preferred-unit-of-measure structure" in the world.
As I see it, structure corresponds to the intrinsic, genuine, objective
features or quantities, which don't depend on arbitraty or conventional
choices in description. By contrast, frame-, coordinate-, or unit-dependent
quantities depend to some extent on our arbitrary or conventional choices in
description-arbitrary, since according to the physics any choice is equally
legitimate. Such quantities aren't wholly about the world as it is in itself, but
are in part about our descriptions of the world, whereas structural features
are agreed upon by all the allowable descriptions, and so correspond to
genuine features of the world apart from any of those descriptions. No
matter which description you use, after all, you get the same result.
Spatiotemporal structure in particular concerns the intrinsic, genuine,
objective spatiotemporal features of a world, which don't depend on arbi-
trary or conventional choices-that two objects are separated by some
amount under a Euclidean metric, say, or that a particle's trajectory is
straight according to a given inertial structure. Notice that this idea of
structure is neutral between substantivalism and relationalism. Both of
these views can recognize that there is a distinction between spatiotemporal
facts that are more objective, and those that are frame-, observer-, unit-, or
coordinate-relative.
We are still working up to the general principle. Here's an idea that we
have reached so far, which will motivate the principle. As we can see from
the inference to a Galilean structure for Newton's laws, any physical theory
will constrain, or help dictate, a world's spatiotemporal structure. We infer
the structure from the physics in this way. This is because any theory will
require or presuppose a certain spatiotemporal structure. In particular, it
will require the structure needed to support the laws, in that the laws cannot
7 9 Compare Maudlin (2012, 9-12); Pooley (2013, sec. 3).
Brading and Castellani (2007) discuss different ways of spelling out this idea.
A New Approach to the Relational-Substantival Debate 9
be srated or formulated without assuming it-they wouldn't make sense
without it. 8
Two examples illustrate this. Recall Newton's first law, which tells objects
to behave differently depending on whether they are traveling inertially,
with uniform velociry, or not. This law would not make sense if there
weren't a distinction between uniform and accelerated motion: it presup-
poses it. So the world must be such that there is this distinction. The world's
spatiotemporal structure should distinguish between inertial and non-
inertial trajectories. Assuming that the laws are about the objective nature
of the world, there must be objective facts about whether objects are
traveling inertially or not. 9
Consider a different example that I'll return to later. If the laws are not
time reversal invariant-if they "look different" when we flip the direction
of time, swapping past and future-then this suggests a structural, physical
distinction in the world between the two temporal directions. Newton's
laws are symmetric in this sense: any behavior allowed by the theo1y can also
happen backward in time. The film of any Newtonian process (a ball
thrown in the air, billiard balls colliding) run backward also depicts a
process that evolves with the laws. These laws don't distinguish past versus
future: they say the same thing regardless of the direction of time. By
contrast, the second law of thermodynamics says that entropy increases to
the future, not the past: gases expand, ice melts, not the reverse. A reverse-
running film shows something disallowed by the law. Non-time reversal
invariant laws like this mention or presuppose the distinction between past
and future, telling things to behave differently depending on the direction of
time. Such laws would not make sense if there weren't a past-future
distinction in the world, corresponding to an asymmetric temporal struc-
ture, or objective facts about past versus future: they presuppose it. (If you
are worried about this conclusion in the case of the second law, stay tuned:
I return to it later in this chapter.)
Finally, the principle. The above examples are familiar instances of how
we draw certain conclusions about the physical world from the laws that
govern it. These examples all suggest that we rely on a certain methodo-
logical principle, which says to posit in the world the structure that's
presupposed by the laws. We generally posit physical structure in the
world corresponding to the mathematical structure needed to formulate
the laws-such as a Galilean spatiotemporal structure for Newton's laws, an
asymmetric temporal structure for non-time reversal invariant laws, or a
8
Consider Earman' s statement that "laws of motion cannot be written on thin air
alone but require the support of various space-time structures" (1989, 46).
10 Jill North
preferred-location structure for Aristotle's laws. We infer to the world
whatever the laws presuppose, whatever there must be in the world for the
laws to make sense and be true of it. There should be a match in structure
between the laws and the world. Theories obeying what I will call the
matching principle are "well-tuned," to borrow a phrase that John Earman
(1989, ch. 3) uses for a somewhat different idea. 10 (I rake it this is motivated
by a kind of realism. I won't argue for realism here.)
As with any guiding methodological principle, this principle won't yield
conclusive inferences, yet it is still a reasonable guide. We cannot be certain
that there is no absolute space in a Newtonian world, bur it is reasonable to
infer that there isn't. Or rake special relativity. The marching principle lies
behind the thought that there is no preferred simultaneity frame. Since the
laws are invariant under changes in Lorentz frame, we infer that there is
no absolute, frame-independent simultaneity relation. We can't be certain
about this, and some people argue that we have other reasons to posit this
structure (for presentism or for certain theories of quantum mechanics, for
example). Still, we do generally, and reasonably, rely on this principle. We
take it to be successful. A~ the case of special relativity shows, we need an
extra reason to disobey it. To put it another way: all other things being
equal, we should infer a match in structure between laws and world. Those
who believe in a mismatch are saying that other things are not equal, and
must argue as much. 11
It is sometimes said that the reason to posit a Galilean rather than
Aristotelian structure in a Newtonian world is that the latter would yield
in-principle undetectable physical facts. 12 Since Newton's laws are invariant
under changes in inertial frame, no experiment could ever detect which is
the preferred frame. Choose any frame in which to run your experiment,
and the laws always predict the same results. That's right. But I think that
there is a deeper reason for the inference to a Galilean structure, which is the
match between the mathematical structure of the theory and the physical
structure of the world. This match is part of our evidence that we have
inferred the correct structure of the world. This is a more fundamental
reason for the inference than the verificationist-sounding principle to avoid
undetectable physical facts.
10 Earman suggests that there should be a match between the symmetries of the laws
and of the spacetime, as a condition of adequacy on theories.
11 Those who argue from quantum mechanics aren't proposing a mismatch, but that
the laws of quantum mechanics trump special relativity when it comes to inferring this
structure.
12 Mentioned, with varying support, in Earman (1989, ch. 3); Ismael and van Fraassen
(2003); Roberts (2008); Dasgupta (2009); Maudlin (2012, ch. 3); Pooley (2013, secs. 3-4).
A New Approach to the Relational-Substantival Debate 11
I have argued that the matching principle is a core methodological
principle we use to guide our inferences from a physical theo1y to the nature
of rhe world according to that theory. Now we can see that this principle
rells us to posit, or countenance, or somehow be able to talk about,
spariotemporal structure. For the laws generally talk about, they mention
or presuppose, a particular spatiotemporal structure. We should counten-
ance the particular spatiotemporal structure or facts required for the laws;
ipso facto, we should countenance spatiotemporal structure or facts in
general. In other words, the matching principle says that we should be
realists about spatiotemporal structure, since the laws presuppose such a
rhing, and we should generally posit in the world the structure that's
presupposed by the laws.
Importantly, this conclusion is independent of the relational-substantival
debate. Regardless of your position on that debate, the matching principle
rells you to believe that there are objective facts about the spatiotemporal
structure of a world; to recognize the spatiotemporal facts that are recog-
nized by the laws. You should believe that a Newtonian world has a
Galilean spatiotemporal structure, for example (although this claim may
be understood differently by the relationalist and substantivalist, as I discuss
below). Who would reject the principle? The conventionalist, for one, like
Reichenbach or Poincare, who denies that there is an objective fact about
the "right" spatiotemporal structure of a world: there are no objective
spatiotemporal facts. Against such a view, the matching principle suggests
that spatiotemporal structure is out there in the world. It is not conventional
or arbitrarily chosen, as is an inertial frame or origin or unit of measure. 13
This structure exists; it is part of reality. There is an objective, determinate
fact about what spatioremporal structure a world has, evidenced by its laws.
The marching principle is nor Quine's criterion for ontological commit-
ment. Quine says that we are ontologically committed to what the variables
of our theories must range over in order for those theories to be true. This has
to do with ontology, with what entities exist. The marching principle is about
what structure we should posit. It says to align physical structure in the world
with the mathematical structure required to formulate the laws. This has to do
with what spatiotemporal facts we should recognize, which is not simply a
matter of ontology. To see that these come apart, notice first that a given
spatiotemporal structure, say a Galilean one, can be understood by different
13
We can agree with Reichenbach and Poincare that those things are arbitrary, since
the laws indicate that different choices are equally legitimate. Spatiotemporal structure is
different. We cannot arbitrarily alter the metric, for instance, and keep the laws the same,
not without major compensating changes elsewhere.
12 fill North
people as involving different entities: by a certain substantivalist 14 as involving
points of spacetime and a relationalist as involving material bodies. (As Tim
Maudlin (2015) puts it, to attribute "a mathematical structure to physical
items" is to say that those items "have some physical features that make them
amenable to precise mathematical description in some respects". In particular,
it is not yet to say what the items must be.) Second, two people might agree on
what entities exist-say, points of spacetime-but disagree on the spatiotem-
poral structure, for instance on whether the points are arranged in a Galilean
or Aristotelian way. This will become clearer as we proceed.
Question: How should we formulate the laws? It seems as though
different formulations can presuppose different structures. If so, then in
order to adhere to the matching principle, we will first need to know how to
formulate the laws, which is a big question. Trust me for now that we can
make progress in advance of answering this question. I will return to it at the
end of this chapter.
Some have argued for a third view, neither substantivalist nor relational-
ist, called 'structural spacetime realism.' 15 Since that view emphasizes real-
ism about spacetime structure, you might think that it is what I am
advocating. I don't have space to address the alternative in detail, 16 but
I will note that, despite superficial similarities, it is importantly different
from my overall approach. First, I claim that both the relationalist and the
substantivalist should (and can: below) be realists about spatiotemporal
structure, whereas spacetime structural realism aims to be distinct from
either of those views. Second, I understand the idea of spatiotemporal
structure differently, to encompass any objective, intrinsic spatiotemporal
fact about a world. In particular, countenancing spatiotemporal structure in
my sense does not mean eschewing fundamental physical objects (alterna-
tively, intrinsic properties) altogether, nor the possibility of our knowing
about such things, as the structural spacetime realist often seems to do. That
said, below we will see one way in which my account mirrors certain claims
of the spacetime structural realist.
3. A DISAGREEMENT ABOUT GROUND
In order to say that the relationalist and substantivalist both should counten-
ance spatiotemporal structure, I must be able to say that they both can do this.
14 See Section 3.3.
15 Different versions are in Dorato (2000; 2008); Slowik (2005); Bain (2006); Esfeld
and Lam (2008); Ladyman and Ross (2009).
6
1 See Greaves (2011).
A New Approach to the Relational-Substantival Debate 13
You might wonder: How can the relationalist believe in spatiotemporal
structure? Isn't this the very sort of thing the relationalist rejects? On the
other hand, if the relationalist can believe in spatiotemporal structure, you
might then wonder what could be left for the two views to disagree about.
I'll now suggest that the notion of ground gives the sense in which the
relationalist as well as the substantivalist can countenance spatiotemporal
structure, and that this yields a real disagreement that's relevant to physics.
The basic idea will be this. Both views can countenance, or believe in the
existence of, spatiotemporal structure. (Whether each one is able to recog-
nize the particular structure needed for the laws is a question that I will be
sidestepping here, for reasons to come.) The views differ on what underlies
this structure. Essentially, the substantivalist says that spatiotemporal struc-
ture is fundamental to the physical world, whereas the relationalist says that
it arises from the relations between and properties of material bodies.
Putting this in terms of ground. A grounding relation is an explanatory
relation that captures the way in which one thing depends on or holds in
virtue of another, without implying that the dependent thing doesn't exist.
Ground captures a "metaphysical because" in answer to questions about
why something exists or some fact holds. (I use the general idea, without
entering into debates over its metaphysics. I won't take a stand on whether
ground is properly a relation between facts or objects, but deliberately use
both ways of talking. It is generally thought that the grounding relation is
transitive and irreflexive, and that the grounds metaphysically necessitate
the grounded. None of these assumptions have gone uncontested, but
I assume them here. 17)
Using the notion of ground, the relationalist and substantivalist can
each say that spatiotemporal structure exists, that there are objective spatio-
temporal facts about a world. They disagree on what the spatiotemporal
structure holds in virtue of; what metaphysically explains the spatiotemporal
facts. The relationalist says that a world's spatiotemporal structure is
grounded in the features and behaviors of material bodies. All the spatio-
temporal facts are grounded in the facts about material bodies. The sub-
stantivalist says that spatiotemporal structure isn't grounded in anything else
more fundamental to the physical world; in particular, it is not grounded
in material bodies. There are fundamental spatiotemporal facts that are not
grounded in facts about material bodies. Both views can countenance spatio-
temporal structure or facts; they disagree on what, if anything, grounds this
structure or those facts.
17
Different accounts are in Fine (2001); Schaffer (2009). Rosen (2010) defends
the idea.
 14 Jill North
I spell out the two views more in a moment. First, a few notes on the use
of ground in this context. Jonathan Schaffer (2009, 363) and Shamik
Dasgupta (20 l l) also suggest that we can understand this debate in terms
of ground, but they put things a little differently. They say that the
relationalist and substantivalist both believe that spacetime exists, while
differing on what grounds the existence of spacetime. I say that both (can
and should) believe that spatiotemporal strncture exists, while differing on
w?at grounds the ~xisten~e of that structure. I prefer this way of putting
thmgs because, we II see, it allows us to flesh out the competing views in
different ways, all the while maintaining a genuine dispute that the physics
will weigh in on.
It may seem unexciting to exchange a debate about the existence of
spacetime for one about the fundamentality of spatiotemporal strncture.
~here has been mucl.1 discussion in metaphysics oflate about doing a similar
kmd of exchange with other existence debates (as in Schaffer (2009)), so
that this instance may feel like old hat. There have been some related
thoughts about the spacetime debate in recent philosophy of physics as
well. Thus Carl Hoefer ( 1998) frames the question in terms of fundamen-
tality, ,~s that of how "to understand the basic ontology of the physical
wo~ld, alt~ough he formulates aspects of the dispute more traditionally,
saymg for mstance that substantivalism is committed to the existence of
"a substantial, quasi-absolute entity." 18 Gordon Belot (1999; 2000; 2011)
says that the relationalist, like the substantivalist, can be a realist in the sense
of''.attribut[ing] to reality a determinate spatial structure," while disagreeing
on 'the nat~re oft.he existence of space" (2011, 1). 19 This is close to my own
way o.f putting thmgs, although his account is not spelled out in the same
way (it does not use notions like ground or my conception of spatiotem-
poral structure, and it focuses on certain traditional examples), nor does
he dra""'. the sam~ conclusions. The more prevalent attitude in philosophy
of phys:cs, es~ec1ally among those who complain about the substantivity
of the dispute, 1s that the debate concerns the existence question. So although
my proposed way of understanding of the dispute is not without prece-
d~nt, e:en then there are differences, and it is anyway not the prevalent
v1ewpomt. If you disagree with that assessment, though, it will soon be
clear that novel avenues of argument open up once we are completely
explicit about this shift.
. is Hoefer simi!arly argues that this is a substantive dispute, which is likely to remain so
wit.h furu:e ph~sics, and that general relativity supports substantivalism. Yet he puts
vanous thmgs differently from how I do, drawing these conclusions for different reasons.
i 9 Be!ot also says that his formulation, while unorthodox, yields a debate that is
substannve, relevant to physics, and reminiscent of the traditional dispute.
11 New Approach to the Relational-Substtmtival Debate 15
3.1. Relationalism in terms of ground
The relationalist says that certain material bodies, and various of their
properties and :-eta~ions, a:·e fundamental, and a world's spatiotemporal
structure holds m vtrtue of them. All spatiotemporal structure or facts are
grounded in (facts about) material bodies. In saying that "certain material
bodies are fundamental," this means whichever material objects turn out to
be most fundamental: certain particles, say. (I assume the fundamental
relations can include spatiotemporal ones, 20 although the relationalist
might want a different kind of relation to be fundamental, causal ones
being a familiar candidate. I leave this open here. The upcoming argument
rakes aim at all these versions of relationalism equally. 21 )
So, for example, the fact that a world has a Euclidean spatial structure is
grounded in, holds in virtue of, the fact that its particles are, and can be,
arranged in various ways, with various distance relations between them.
(I return to this "can be" phrase soon.) The world has a Euclidean structure
because (in the metaphysical sense) its particles are, and can be, arranged in
those ways; this is what the spatial structure consists in. Similarly, the fact
that a Newtonian world has a Galilean spatiotemporal structure is grounded
in the fact that its particles do, and can, behave in various ways, with various
spatiotemporal relations between them. The fact that a world has a particu-
lar spatiotemporal structure is made true by the facts about material bodies.
A world has the spatiotemporal structure it does because material bodies
(can) behave in certain ways.
Three notes on this use of ground. First, a grounding explanation is
importantly different from a causal explanation. In Kit Fine's words, ground
yields "a distinctive kind of metaphysical explanation," in which the objects
or facts are connected by "some constitutive form of determination" (2012, 37).
Particle behaviors don't cause a Euclidean spatial structure. This is rather
what the spatial structure consists in or depends on, in a metaphysical
sense. Compare this to more familiar cases, such as the grounding of
facts about the macroscopic world in facts about subatomic particles, or
the grounding of mental facts in non-mental facts, or moral facts in non-
moral facts. Ground captures this metaphysical "in virtue of" explanation. 22
As I understand it, when we say that "the fact that x grounds the fact that y,"
2
° Contra Nerlich (1994a, ch. 1).
2
i I also assume that t~e objects and relations are equally fundamental, though there may
be a view with only one rundamental "ontological category" in the sense of Paul (2013).
22 Loewer (2001) discusses the relevant sense of "in virtue of."
16 Jill North
this just means that "the fact that y holds in virtue of the fact that x"; i.e. that
the holding of the grounded fact consists in nothing more than the holding of
the grounding fact.
Second, ground aims to give a "looser" connection between the facts or
objects involved than that given by a definition. An analogy: I am thinking
of ground in such a way that it can articulate the view that the biological
facts are nothing over and above the facts about these systems' particles.
(You may not hold such a view, but ground can specify what it amounts to.)
The history of failed attempts in twentieth-century philosophy of science to
spell out a "tighter" connection between the reduced and reducing facts by
means of correspondence rules that define the biological quantities in terms
of physical ones suggests that this won't work. Yet there is still a way of
capturing the sense in which the biological facts "are nothing but" the
physical facts, which is to say that the biological facts are grounded in the
physical ones. In an analogous way, the relationalist can say that the facts
about spatiotemporal structure are "nothing but"-are grounded in-the
facts about material bodies, even if she can't explicitly define the spatiotem-
poral structure in terms of the relations between material bodies. A grounding
relation can hold even in the absence of a definitional connection. (This is one
reason the notion of ground can help the relationalist, since finding such
explicit definitions is notoriously difficult. Of course, it is not easy to give an
account of the grounding of spatiotemporal structure in material bodies
either, but replacing the definitional requirement with the looser constraints
of ground can ease some of the burden.)
Third, there must be some account of how the facts that the relationalist
takes to be fundamental manage to ground all the spatiotemporal facts
needed for the physics. (For instance, there can't be two worlds with the
same fundamental relationalist facts but different spatiotemporal structures,
since the fundamental facts necessitate the grounded facts.) Simply being a
realist about spatiotemporal structure does not guarantee the ability to
generate the particular structure required by the laws as the matching
principle demands. You might be skeptical that the relationalist can do
this. Much of the literature is taken up with this question of how, and
whether, the relationalist's more meager ontology can recognize all the
spatiotemporal facts we want. 2 3
This is a big question, but I won't try to answer it here. I won't try to tell
you exactly how the relationalist grounds all the spatiotemporal facts in facts
23
A repeated complaint against the varieties of relationalism surveyed by Pooley
(2013) is that the relationalist's resources are too thin to yield predictions of the
phenomena.
A New Approach to the Relational-Substantival Debate 17
about material bodies. 24 As we'll see, I think there is an argument for
subsranrivalism that goes through even if we grant the relationalist the
ability to ground all the relevant facts in ones she takes to be fundamental.
So for the purposes of that argument, I am going to grant the relationalist
that ability.
It is worth mentioning one thing that I do think will be required to
ground that structure, which is some version of "modal relationalism."
I suspect that the relationalist will have to countenance facts not only
about the actual features and behaviors of material bodies, but about their
possible ones as well-facts about what spatiotemporal relations can hold, in
some sense. This is because the actually instantiated relations won't in
general suffice to fix the full spatiotemporal structure required for the
physics. (As long as the relationalist can embed the actual relations uniquely
into a certain structure, it seems as though she can talk of the spatiotemporal
structure of a world. The problem is that the actual relations may not
uniquely fix the structure (up to isomorphism) needed for making predic-
tions about material bodies. 25 ) In order to adhere to the matching principle,
the relationalist will have to go modal. I refer you to Carolyn Brighouse
(1999) and Belot (2011) for discussion of ways the relationalist might do
this and what sort of modality may be involved. 26
(Modal relationalism arguably allows the view to countenance vacuum
worlds, which seem possible according to both classical and relativistic
physics. Such worlds contain no material bodies and yet can have a spatio-
temporal structure. Now, it is open to the relationalist to deny that vacuum
models correspond to physically possible worlds. Nonetheless, the modal
relationalist should be able to allow for these possibilities. All the facts about
spatiotemporal structure will still be grounded in facts about material
bodies-in facts about how these bodies would behave, if there were any.
Such a relationalist can arguably even countenance different spatiotemporal
structures in different vacuum worlds, as general relativity seems to allow
for. This is not to say exactly how the relationalist can do this, just as I haven't
24 From this perspective, those such as Manders (1982); Mundy (1983; 1992);
Huggett (2006); Belot (2011) can be seen as giving accounts of how this grounding
project might go.
25 Examples are in Mundy (1986); Maudlin (1993, 193-4, 199-200); Nerlich
(1994a); Belot (2000; 2011, ch. 2). Field (1984) argues that the modal view is necessary
for the relationalist to solve the problem of quantity. An alternative is conventionalism
(Earman, 1989, sec. 8.6).
26
The view may sound newfungled, bur even Leibniz, according to many, held it: Belot
(2011, Appendix D). The liberalized relationalism of Teller (1991) is a precursor to more
recent versions. See also Sklar (1974, IILB2); Horwich (1978); Mundy (1986). Objections
are in Malamcnt (1976); Field (1984); Earman (1989, sec. 6.12); Nerlich (1994a).
18 Jill North A New Approach to the Relational-Substantiva! Debate 19

said how the relationalist can ground any particular structure in material
bodies. Yet once we grant the (modal) relationalist the ability to ground all 3.2. Substantivalism in terms of ground
the spatiotemporal facts in facts about material bodies, there needn't be a
The substantivalist denies that ,all spatio temporal facts hol~ in virtue of fac.ts
special problem for vacuum worlds.)
Keep in mind that the relationalist might not deny the fundamentality of abou t nnterial
, bodies . A worlds spatiotemporal
, structure Is not grounded . m
c r·es and behaviors of material bodies. fhe fact that a world has a given
any spatiotemporal fact or structure. Depending on the version of the view reatu . ld .
iotemporal structure is a fundamental fact about the physical wor ; m
(see the beginning of this subsection), the fundamental facts may include
spat·cular it is not grounded in facts about material bodies. (Clarifications
ones such as that two particles are separated by some distance, or that one part! ' d
below.) The facts about a world's spatio~emporal structure, .in turn.' groun
particle lies between two others. 27 What's important is that the relationalist
the facts about the spatiotemporal relations b.etween material bod1e:. (The
only allows certain kinds of spatiotemporal facts (if any) to be fundamental,
c mer may only partially ground the latter, smce the grounds may mclude
namely those that essentially involve material bodies and their relations- ror b . .
occupation relations that ma~erial bodie: ear to sp~cet1me pomts or
facts that the substantivalist takes to be nonfundamental. The fact that
·ons depending on the version of the view-see Sect10n 3.3.)
a world has a given spatiotemporal structure is grounded in the facts reg l ' . .
For example, the fact that two particles are some distance apart is
about material bodies, even though these latter facts may include certain
grounded in, made true by, the fact that they are separated b~ th~t ~mount
spatiotemporal ones. More exactly, there is no fundamental spatiotem-
according to the fundamental metric structure (where the metric w11l 1tselfbe
poral fact or structure apart from the structure of, or facts about, material
understood in different ways by different substantivalists-see Section 3.3-
bodies. For ease of exposition, I put this as the claim that all spatiotem-
but will in any case not be grounded in features of material bodies). The fact
poral facts are grounded in facts about material bodies. All spatiotemporal
that a particle is traveling inertially in a Newtonian world is likewise gro~nde.d
structure is grounded in the relations between and properties of material
in facts about the fundamental spatiotemporal structure: the particle is
bodies.
following a straight trajectory because (in the metaphysical sense) its path is
So, using the notion of ground, the relationalist can say that there are
straight according to the world's Galilean structure. (The substantivalist then
facts about a world's spatiotemporal structure, which are distinct from
recognizes nonfundamental spatiotemporal facts or structure of a sort, ab?ut
the facts about material bodies and their relations, but are also nothing
the spatiotemporal relations between material bodies. More exactly, the view
over and above those facts about material bodies-just as one might say
holds that there are fundamental spatiotemporal facts or structure not
that there are real facts about macroscopic systems, which are distinct
grounded in (facts about) material b.odies. Notice t.hat. ce:tain fact~ abo~t
from the facts about their particles, but are also nothing over and above
material bodies, for instance about thell' fundamental 111tr1ns1c properties, wt!!
the facts about the particles.
be fundamental. What's not fundamental are the spatiotemporal facts about
This is a non-standard (if not wholly unprecedented) way of formulating
them.) By contrast, for the relationalist, a world's spatiotemporal structure is
relationalism, which captures traditional thoughts about the view, for
Galilean because the particles behave in certain ways. On that view, the facts
instance that spacetime doesn't "really exist": "spacetime" is nothing but
about material bodies metaphysically explain the fact that a world has the
various features of material bodies; certain material bodies are fundamental,
given structure.
and any spatiotemporal talk or fact is really about them. At the same time,
For the substantivalist, facts about the spatiotemporal relations between
this formulation allows the relationalist to say that spatiotemporal structure
material bodies are nothing over and above facts about how these objects are
exists, that there are objective truths about what spatiotemporal structure a
arranged according to a given spatiotemporal structure. Facts about. a
world has, as the matching principle says we should do. It's just that these
world's spatiotemporal structure, on the other hand, are not grounded 111
things all hold in virtue of what's true about material bodies.
facts about material bodies, and in that way are "over and above" any facts
about material bodies. This captures the traditional conception of the view
as holding that spacetime exists "independently of" material bodies: there is
spatiotemporal structure that is not metaphysically due to material bodies.
2
7 Which of these depends on whether the relationalist thinks that fundamental You may worry that this conception of substantivalism is already discon-
relations can be quantitative. firmed by our current best theory of spacetime. According to general relativity,
20 Jill North
the presence of matter affects the local spatiotemporal geometty, which in
turn affects the behavior of matter; whereas on my conception of substan-
tivalism, there is spatiotemporal structure that is independent of matter.
This worry is evaded by noticing that the interdependence between spa-
tiotemporal structure and material bodies in general relativity is of a
different, causal or nomological, kind from that given by ground. Although
the substantivalist says that there is spatiotemporal structure that is inde-
pendent of material bodies in not being grounded in them-these facts
about spatiotemporal structure are "metaphysically over and above" the
facts about material bodies-she can still allow that the behavior of material
bodies causes a certain spatiotemporal structure in accord with the physical
laws. Compare: although the dualist says that mental events are not grounded
in physical events-mental events are "metaphysically over and above" physical
ones-she can still allow that physical events cause mental events in accord
with the scientific laws.
Substantivalism and relationalism, as I understand them, disagree about
the fundamental nature ofthe physical world. They both countenance spatio-
temporal structure or facts, but disagree on whether all such structure or
facts hold in virtue of material bodies. Both views can recognize the fact that
two particles are separated by some distance under a Euclidean metric, for
instance, or that a world has a Euclidean metric structure. But they will
disagree on whether the metric is itself fundamental or grounded in the
behavior of material bodies. To borrow a phrase that Helen Beebee uses for a
different debate, these views "have completely opposite conceptions of what
provides the metaphysical basis for what" (2000, 580). The substantivalist
sees a world's spatiotemporal structure as the metaphysical basis for the
spatiotemporal relations between material bodies. The relationalist sees
material bodies and their relations as the metaphysical basis for a world's
spatiotemporal structure. If we ask, of a Newtonian world, "why (in the
metaphysical sense) does it have a Galilean spatiotemporal structure?" the
relationalist will answer: "because the particles (can) behave thus and so."
The substantivalist will have no answer (or if there is any answer, it won't
reference material bodies: see below). This is a substantive debate about
what makes it the case that the spatiotemporal structure needed for the
physics holds.
3.3. Further clarifications
The substantivalist might not take a world's spatiotemporal structure to be
absolutely fundamental. Newton held that absolute space is a necessary
consequence of God's existence, so that the facts about the world's spatial
A New Approach to the Relational-Substantival Debate 21
structure are not fundamental but grounded in facts about God. Yet
Newton is still a substantivalist, on my understanding, since the facts
about the spatial structure are more fundamental than the facts about
bodies' spatial relations. 28 To put it another way: the facts about the spatial
structure are fundamental to the physical realm. Analogously, the relationalist
will say that all spatiotemporal facts are grounded in facts about material
bodies, regardless of her other metaphysical views, such as whether there is
something yet-more-fundamental that lies outside the physical realm. The
views still disagree over whether spatiotemporal structure apart from mater-
ial bodies is fundamental to the physical world. For ease of presentation,
I continue to put the dispute as the question of whether spatiotemporal
structure is fundamental (to the physical world).
What if there is no fundamental physical level? In that case, the views
might still be distinguished by means of the relative fundamentality of the
behaviors of material bodies and a world's spatiotemporal structure, depend-
ing on the details. This may suggest that the debate should be framed in
terms of relative fundamentality. Substantivalism would then be the view
that the facts about a world's spatiotemporal structure are more fundamental
than the spatiotemporal facts about material bodies, and relationalism
would be the view that the facts about material bodies are more fundamental
than the facts about spatiotemporal structure. But I don't want to put it this
way. That way of putting things would imply that either relationalism or
substantivalism is bound to be true, regardless of future physics, so long as
the two kinds of facts are not equally fundamental. Yet intuitively, if nothing
like either spatiotemporal structure or material bodies turns out to be funda-
mental to the physical world, then neither view has been vindicated. You
could insist that substantivalism would still be correct so long as the facts
about the world's spatiotemporal structure are more fundamental than the
spatiotemporal facts about material bodies, and contrariwise for relationalism.
This strikes me as too far removed from the original views. More generally,
I don't think that one of these views must be correct regardless of future
physics, and it will depend on the details of that future physics whether one or
the other, or neither, is correct.
There is another way to put the difference between the views, which
I want to be careful with. The substantivalist says that there exists a
fundamental physical space(time); the relationalist denies this. Similarly,
the relationalist denies, whereas the substantivalist accepts, the existence
of spacetime points (or regions) as fundamental physical objects. This way
of putting things is familiar and in keeping with traditional conceptions
28
Some argue that Newton wasn't a substantivalist: Stein (1970); DiSalle (2002).
22 fill North
of the dispute. 29 The problem is that it is not entirely clear what it means to
say that a physical space-this "peculiar entity" (Belot and Earman, 2001,
227)-does, or doesn't, exist; relatedly, whether spacetime points or regions
exist as concrete entities. I suspect that this is an underlying reason for the
unclarity of the debate in many people's minds, especially in the philosophy
of physics community. Some philosophers of physics have worried about
taking spacetime points to be concrete physical entities in particular. As
Malament says, in the context of discussing whether spacetime points are
nominalist-friendly, "They certainly are not concrete physical objects in any
straight-forward sense. They do not have a mass-energy content .... They do
not suffer change. It is not even clear in what sense they exist in space and
time" (1982, 532). Others have worried more generally that this kind of
ontological dispute-a dispute that is just about what things exist-is non-
substantive or merely verbai. 30 Howard Stein, in discussing the spacetime
debate, says that, "For me, the word 'ontological' itself presents seriously
problematic aspects"; in particular, "Quine's usage [is) not a very useful one
for philosophy of physics" (1977a, 375).
As I see it, the debate is about the fundamentality of spatiotemporal
structure, in particular about whether there is any spatiotemporal structure
(fact) not grounded in the structure of (facts about) material bodies, where
the substantivalist says that there is and the relationalist says that there isn't.
Within this framework, there is some flexibility as to how exactly to put the
dispute. Neither the matching principle nor my conception of spatiotem-
poral structure says how we must construe the nature of spatiotemporal
structure; and I have not taken a stand on whether ground is primarily a
relation between objects or facts. As a result, although we can put the
disagreement as being about whether there exists a fundamental physical
spacetime or fundamental spacetime points, we do not have to. Anyone
squeamish about putting things in ontological terms can still see the debate
as being about the fundamentality of spatiotemporal structure, understand-
ing this as being not about whether there exist certain objects (over and
above material bodies), but about whether there are certain facts (over and
above the facts about material bodies): the relationalist says that the fact that
a world has a certain spatiotemporal structure holds in virtue of the fact
that material bodies behave thus and so; the substantivalist denies this,
29 See Field (1980, ch. 4); Mundy (1983); Earman (1989, 12); Brighouse (1994).
30 This seems the spirit behind Stein (1970; 1977a); Curiel (2016); perhaps Belot
(2011) and some others in note l; in a different way Wallace (2012). There have been
similar thoughts in metaphysics, for example in Hirsch (2011), but it's not clear that this
is exactly the same idea.
A New Approach to the Relationa!-Substantiva! Debate 23
seeing it as a fundamental fact about the physical world. This allows us to
discuss the dispute, and to evaluate the evidence for either side, while
remaining neutral on how the substantivalist wants to understand the
instantiation of that structure or the ontology behind this fact.
This dovetails with an idea in spacetime structural realism. Jonathan Bain
(2006) argues that classical field theory (this includes general relativity),
standardly given in terms of a tensor formalism, can be formulated in ways
that do not presuppose a differential manifold of points. He describes three
alternative formalisms one could use (twistor theory, Einstein algebras, and
geometric algebra), none of which treat points as fundamental. My under-
standing leaves it open for the substantivalist to spell out the spatiotemporal
structure in any of these ways, or even to refuse to choose among them, as
Bain himself proposes. (Bain argues that we should be realists about space-
time structure and not any particular instantiation of it. He sees this as a
third view, since according to him the substantivalist is committed to
spacetime points, but it counts as substantivalist by my lights.)
To be explicit, there are four different kinds of view that my conception
of substantivalism is meant to encompass, each of which holds that there are
spatiotemporal facts or structure not grounded in material bodies. First is
what we might call Bainianism, on which one is a realist about spatiotem-
poral structure but not about any particular instantiation of it, i.e. not about
any of the (non-material) objects that could be said to instantiate it. On this
view, the different possible descriptions or formulations or instantiations of
spatiotemporal structure do not really differ from one another: one is an
anti-realist about those. Second is what we might call uncommitted sub-
stantivalism, on which one is a realist about a particular instantiation of
spatiotemporal structure-there is a single best way of describing or formu-
lating the spatiotemporal-structure facts, in terms of a certain kind of non-
material object-but one doesn't know what that instantiation or best
formulation is; hence we cannot state the view as propounding one or
another such formulation. Third is what we might call committed substan-
tivalism, on which one is a realist about a particular instantiation of
spatiotemporal structure, one thinks that there is a best formulation of it,
and one does claim to know what it is; e.g. it might be the one in terms of
points (in which case the view approaches traditional substantivalism).
Fourth is the "qualitativist" substantivalism of Dasgupta (2009; 2011), on
which the fundamental spatiotemporal facts are purely qualitative, not
mentioning any entities at all; spacetime is not an entity but a "purely
qualitative structure." One of the things I am claiming is that, when it comes
to the relational-substantival debate, we needn't choose among these ver-
sions of substantivalism. The argument in Section 4 will support each of
them in the same way.
24 Jill North
3.4. Something old, something new
There are too many different notions of "relational," "substantival," and
related concepts in the literature to survey them all here and compare them
t? my own a~count. 31 ~t should be clear that this is a non-standard concep-
t'.on of the dispute, which captures core ideas behind more familiar concep-
tions, both contemporary and traditional. For example, my understanding
ca~tures the thought that the substantivalist believes in "the independent
e.x1sten:e a~? structure of space and time" (Sklar, 197 4, 163)-that space-
t1me exists mdependently of material things ... and is properly described as
ha~ing its own properties, over and above the properties of any material
thmgs that may occupy parts of it" (Hoefer, 1996, 5)-so that "space is
something as real as matter and whose existence does not require matter, but
whi:h is not the same stuff as matter" (Huggett, 1999, 129). It encompasses
the idea that for the substantivalist, "space-time points (and/or space-time
:egions) ~re ~nt~ties that ~xist in their own right" (Field, 1980, 34); "[s]pace
ts an entity m its own nght-a real live thing in our ontology" (Nerlich,
1994a, 3), a "genuine entity of a fundamental kind" (Pooley, 2013, 526).
These ideas are captured by the claim that spatiotemporal structure is
~undamental to the physical world. There is spatiotemporal structure that
ts not grounded in, and is in that way independent of, any material bodies.
My conception also captures the thought that the relationalist "denies
that space, or spacetime, is a basic entity, ontologically on a par with matter"
(B'.·own and Pooley, 2002, 183, n.1), so that "the universe consists solely of
objects and events exemplifying various properties and relations" (Horwich,
1978, 397); "all that exists is material bodies" (Arntzenius, 2012, 153). As a
result, "all our talk of space and time can be reconstructed out of talk about
spatial re.lations between objects" (Brighouse, 1999, 60), and we "regard the
use physical theory makes of space-time and its geometrical structure merely
a~ a convenie~t way of saying something about the spatio-temporal proper-
ties and relations of concrete physical objects" (Friedman, 1983, 216).
These statements are captured by the claim that spatiotemporal structure
apart from material bodies is nonfundamental; whereas certain material
objects, and certain of their properties and relations, are fundamental.
At the same time, this is a non-standard, non-traditional take on things
which allows us to sidestep many of the reasons people feel that the usuai
dispute has stagnated or become non-substantive. Most importantly, it
leaves room for future physics to provide an answer, so that this dispute
cannot be "merely verbal" or "purely metaphysical." We think that there
31
See the many notions listed in Horwich (1978); Friedman (1983); Earman (1989).
A New Approach to the Relational-Substantival Debate 25
is a real difference between a world in which spatiotemporal structure is
fllndamental, and one in which it arises from some pre-spatiotemporal
structure, for instance. Physicists treat these as genuinely different possi-
bilities, governed by different theories. This is evidence of a genuine
difference between the views as I see them.
Against tradition, I claim that the relationalist as much as the substantiv-
alist can recognize "absolute" or frame-independent facts about-quantities
of, structures that support-objects' motions. 32 In particular, it needn't be
the case that "all motion is relative" for the relationalist, since there can be
objective facts about objects' motions even in a world devoid of other
material bodies. 33 The traditional question about the relativity of motion,
then, is not of primary concern. 34 In addition, we needn't distinguish the
two views by means of how they count possibilities, contrary to tradition as
well as s~me recen.t accounts. 35 Further, against some other understandings
of the dispute, this one allows for both sides to believe in, to be realists
about, spatiotemporal structure. 36 (I have argued that they both should do
this, in orde.r to r.espect o~r usual inferences in physics.) I even leave it open
for the relat1onahst to posit the same spatiotemporal structure to a world as
the substantivalist, whereas some have taken the dispute to be over the
relevant structure. 37
My conception also avoids having to draw some of the distinctions that
people have been skeptical of. It does not require that we definitively
distinguish between container and contained, substance and non-substance,
absolute and relative, to name a few. 38 There are three distinctions presup-
posed by my understanding of the dispute, but they are not as unclear as
those required by more traditional conceptions. First, there is the distinction
between the fundamental and the nonfundamental. This is a distinction
that we have a reasonably clear pre-theoretic grasp of, clear enough to be
~seful he~e even without sfelling it out in more detail. Second, my concep-
tion requires that we can identify what structure counts as spatiotemporal.
32 Hoefer notes that traditional relationalism "is connected essentially to the denial of
absolute motion" (1998, 460). '
33• Huggett and Hoefer (2009) note other relationalist views denying the relativity of
motion.
34 T.his ~igns :vith a similar shift away from that question in recent literature,
exemplified 1ll Stem (1970; l 977b); Sklar (1974); Friedman (1983); Earman (1989);
Belot (1999; 2000; 2011); DiSalle (2006).
3
~ Huggett (1999, ch. 8) discusses the traditional arguments. More recent examples
are 111 Earman and Norton (1987); Belot (2000).
36
St~tem~nts intimating that the relationalist cannot believe in spatiotemporal struc-
tur~7 are 1ll Field (1984, 34); Nerlich (1994a); Pooley (2013, 542); Maudlin (2012, 66).
Earman (1989) suggests this at points.
38
Rynasiewicz (1996; 2000) worries about the clarity of all these (and other) distinctions.
26 Jill North
This is something that the physical laws give us a handle on, in ways
discussed earlier, though I admit that there is more that could be said.
Perhaps there is nothing else that makes some fact or structure spatiotem-
poral; perhaps there is. 39 Either way, I take the idea to be relatively familiar
from physics. At least we have some clear cases of spatiotemporal structures,
such as those discussed here.
Third, my conception requires a distinction between material bodies and
other things in the world. Although people have worried about the clarity of
this distinction, 40 I think that it is clear enough for our purposes. At the
least, I suggest that we understand the debate in this way, on the assumption
that we will be able to locate such a distinction. For now I follow Earman,
who says that, "It is a delicate and difficult task to separate the object fields
into those that characterize the space-time structure and those that charac-
terize its physical contents," while also noting that "the vagaries of this
general problem need not detain us here, since there are clear enough cases
for our purposes" (1989, 155-6). For those wanting argument that the
distinction can generally be made, I refer you to Carl Hoefer (1998) and also
David Baker (2005).
One will find, in contemporary discussions, the thought that the rela-
tionalist can believe in the existence of spacetime, understanding this as
being (somehow) constructed out of material bodies and their features. So it
may seem like even the traditional dispute (and contemporary versions of it)
was never about the existence of spacetime but its fundamentality, and my
own formulation may seem like just a new label for an old dispute. This
however is something of an anachronism. Traditional participants, like
Newton and Leibniz, weren't focused on questions of fundamenta!ity:
they were not thinking explicitly in those terms. Neither, of course, were
they thinking in spatiotemporal terms. Atthe same time, to the extent that
we can understand what they were saying in these terms, this shows that my
understanding is, as I claim, an updating of the traditional dispute, using
more recent developments in physics (involving spacetime and its struc-
tures) and philosophy (fundamentality and ground).
4. AN ARGUMENT FOR SUBSTANTIVALISM
I now suggest that if we do understand the debate in this way, then there is a
powerful argument for substantivalism, given much of current physics.
39 Belot (2011) and Brighouse (2014) are two different accounts.
40 See especially Rynasiewicz (1996).
A New Approach to the Relational-Substantival Debate 27
Above I argued that the relationalist should go parrway41 toward adhering
to the matching principle by countenancing spatiotemporal structure, and
that she can do this by understanding all the facts about spatiotemporal
structure as being grounded in facts about material bodies. I am now
going to argue that really the relationalist can't adhere to this principle,
properly understood. The argument differs from the more familiar charge
that the relationalist cannot countenance a particular spatiotemporal fact
or structure.
Recall that the matching principle says to posit in the world the structure
presupposed by the laws; that is, to posit physical structure in the world
corresponding to the mathematical structure needed to state the laws.
Now here is something else about the principle I haven't yet mentioned.
It applies, in the first instance, to the fundamental laws. (By saying "in the
first instance," I mean to indicate that the principle applies at least to the
fundamental laws, and that this is where we begin constructing our picture
of the world from physics, in that we build a world "from the bottom up."
I leave it open whether an analogous idea holds for nonfundamental laws.)
Given the fundamental laws, we should posit in the world the structure
they presuppose. This is clear from our usual inferences about spatiotem-
poral structure. Assuming that Newton's laws are fundamental, we infer a
Galilean structure to the world. From different fundamental laws, we infer
a different spatiotemporal structure-such as a Minkowskian structure for
special relativity, a preferred-location spatial structure for Aristotle's physics,
or a variety of different spatiotemporal structures for general relativity.
The matching principle also tells us to posit, in the fundamental level of
the physical world, whatever those laws presuppose. The fundamental laws,
after all, are about what's fundamental. They don't "care about" or "know
about" or mention the nonfundamental. I take it this is part of what we
mean when we say that they are fundamental. I also take it that this is a
familiar thought. (Michael Townsen Hicks and Jonathan Schaffer (2017)
call it orthodoxy. 42 ) For example, it lies behind our dislike of quantum laws
that mention things like "measurement" or "the observer." This isn't to
deny that fundamental laws have consequences for nonfundamental things.
These laws yield predictions for nonfundamental phenomena when we plug
in initial conditions and use various bridge principles. On their own, though,
41
Partway, since I haven't shown that the relationalist can ground the particular
structure needed.
42
They argue against the idea, concluding that fundamental laws can, and do,
mention nonfundamental properties. I agree that an alternative formulation can be useful
in practice, bur I think that the best formulation won't mention such things.
28 ]ill North
fundamental laws only mention or presuppose or know about things at the
fundamental level. 43
Another way to see this comes from the idea of "the structure presup-
posed by the laws." The sense in which the laws presuppose or require some
structure is akin to an idea familiar from mathematics. In mathematics, we
can define different levels of structure by starting with a lowest level, such as
a set of points, and then defining other objects that add more structure.
These levels of structure form a hierarchy. The ones "higher up" assume or
presuppose or constrain levels lower down, in that the higher-level objects
cannot be defined until the lower-level ones have been assumed or defined.
For example, think of adding differential structure to a topological space.
This structure indicates, from among the continuous curves specified by
the topology, which ones are smooth to varying degrees. In this way the
differential structure assumes or presupposes a topology: it cannot be defined,
it doesn't make sense, absent a topology. Higher-level structure is not simi-
larly constrained by levels lower down-as different metrics, or none at all,
can be added to a differential manifold. In other words, a given level of
structure only "knows"-requires, constrains, presupposes, assumes-things
about that level and below. 44
Analogously for the structure required by the physical laws. This structure
is presupposed by the laws in that it must be assumed in order for the laws to
be formulated or make sense. The laws don't similarly know about-
require, constrain, presuppose, assume-higher-level structure. For funda-
mental laws, the result is that they only know about fundamental structure.
Note that the fundamental laws may constrain things higher up in a
different, metaphysical sense: given the fundamental laws and ontology,
everything else may be "fixed" in some sense. This is a different sense of
constraining from the mathematical notion, which concerns what is needed
for something to make sense or be defined. The other sense is a metaphysical
notion that requires additional metaphysical principles concerning the
relation between different levels of realiry.
An example illustrates and motivates the primary reading of the matching
principle. Recall the discussion of non-time reversal invariant laws. Earlier
43 This is different from Sider' s (2011, ch. 7) purity principle. Purity is a very general
principle about what the fundamental facts or truths can mention. (It says that they
cannot mention nonfundamental concepts.) The above is specific to the physical laws and
what they presuppose and therefore tell us about the physical world.
44 In mathematics one also talks of a higher-level structure "inducing" a lower-level
one (e.g. "the topology induced by the metric"). This makes it sound as though the
higher-level structure is defined first and it then constrains the lower, but in fact it
amounts to the above idea (e.g. once we have defined a metric, there must already be
implicitly a topology).
A New Approach to the Relational-Substantival Debate 29
I said that if the laws are asymmetric in this way, then we infer an
asymmetric temporal structure in the world. The idea is that such laws
presuppose this structure, for they mention or presuppose a distinction
between past and future, by telling things to behave differently depending
on the direction of time. But there is more to the stoty. Take the second law
of thermodynamics. This law is not time reversal invariant, so it may seem to
indicate an asymmetric temporal structure. However, the second law of
thermodynamics is not a fundamental law. It doesn't mention a system's
particles or other fundamental constituents. It is formulated in terms of
higher-level macroscopic quantities like entropy. Whether to infer an
objective past-future distinction in the world then really depends on what
fundamental theo1y accounts for the second law, and whether that theory's
laws are symmetric in time. (It is natural to think that if a past hypothesis
account of thermodynamics is correct, then there is no asymmetric tem-
poral structure; whereas if a non-time reversal invariant theory like GRW
quantum mechanics is true (and able to account for thermodynamics) then
there is. 45 ) The nonfundamental law on its own does not tell us about
fundamental temporal structure: it is too far removed from the fundamen-
tal level to do that. Only a fundamental law can tell us about this.
In other words, we posit fundamental structure in the world needed for
the fundamental laws. We recognize as fundamental the facts that are
recognized by the fundamental laws. The matching principle applies, in
the first instance, to the fundamental laws and fundamental level of physical
reality. The matching principle as discussed in Section 2 says that the world
should "look like" or "fit" its laws. The primary reading of the principle says
that the fundamental level of the world should look like or fit its funda-
mental laws.
Now to the argument for substantivalism. First notice that the kinds of
fundamental laws we are most familiar with are formulated to presuppose
spatiotemporal facts apart from material bodies. These laws mention or
presuppose a spatiotemporal structure in addition to material bodies and
their features. Newton's laws presuppose a Galilean spatiotemporal structure
in addition to the existence of massive particles. These laws assume or
require that the world has this structure, just as the laws of special relativity
assume or require a Minkowskian structure. The laws of Aristotle's physics
mention a preferred-location spatial structure in addition to the elements
that move toward their natural places. Similarly for the laws of general
relativity, even though they allow for different spatiotemporal structures.
45
Albert (2000) discusses these two accounts. See North (2008) on why these
conclusions about temporal structure are natural.
30 Jill North
Think of the usual way of understanding the field equations, as saying how
the distribution of matter and energy relates to the spatiotemporal geometty,
which in turn affects the behavior of matter. These equations are formulated
directly in terms of-they mention or talk about-a spatiotemporal struc-
ture apart from material bodies, coded up in the metric tensor, distinct
from the stress-energy tensor. (See Hoefer (1996; 1998) for arguments
that the metric is most naturally seen as characterizing a spatiotemporal
structure that is not the structure of a material field. This is not uncon-
troversial, but is assumed in standard presentations.) The fundamental
laws that we are familiar with make reference to material bodies, but they
also presuppose or make reference to a spatiotemporal structure apart from
those bodies. 46
Given that the fundamental laws are typically like this, a problem arises
for the relationalist. The problem is not that the relationalist doesn't
recognize enough spatiotemporal facts for the physics, a concern lying at
the root of classic arguments like Newton's, as well as many contempora1y
ones (see notes 23 and 25). Grant the relationalist enough stuff to ground
those facts and make the relevant predictions, and there is still a problem.
According to the core of the view, all the facts about spatiotemporal
structure are grounded in more fundamental facts about material bodies.
The kinds of fundamental laws we are used to, though, presuppose or
mention spatiotemporal facts apart from material bodies-facts that, for
the relationalist, are nonfundamental. This violates the principle that the
fundamental level of the physical world should contain whatever is needed
for or presupposed by the fundamental laws.
So the argument is this. First premise: the fundamental laws are about
what's fundamental to the physical world; they refer to or presuppose
things about the fundamental physical level. Second premise: these laws
are about, they presuppose or refer to, a spatiotemporal structure, or
spatiotemporal facts, apart from material bodies. Third premise: for the
relationalist, this kind of structure or fact exists at a nonfundamental level,
46 There is a difference between the laws mentioning and presupposing something.
That a law explicitly mentions something implies that the law presupposes it, but not vice
versa. The laws of general relativity explicitly reference both material bodies and spatio-
temporal structure. The usual Newtonian laws explicitly mention the former yet only
presuppose the latter. (Hence a difference from Quine's prescription (see pp. 11-12):
Newton's laws, as usually formulated, presuppose a Galilean spatiotemporal structure;
they don't explicitly mention or quantify over that structure, which the matching
principle tells us to posit.) This difference does not matter here. We use the matching
principle to infer structure in the world regardless of whether it is explicitly mentioned or
presupposed. Either way, the laws require it.
A New Approach to the Re!ational-Substantival Debate 31
above that of material bodies. Fourth premise: the prima1y reading of the
matching principle. Conclusion: relationalism is incorrect. Substantival-
ism posits the spatiotemporal structure or facts needed for the laws at the
fundamental level.
General relativity provides an example. This theory establishes a nomo-
logical connection between material bodies and a spatiotemporal structure
apart from them. On their own, the laws do not say whether material bodies
and spatiotemporal structure are at the same level of physical reality, nor
which is more fundamental if not. Without some further principle, both
relationalism and substantivalism seem satisfactory: both recognize facts
about material bodies as well as a world's spatiotemporal structure. Enter
the matching principle. The substantivalist does, the relationalist does not,
adhere to it.
You may wonder why the spatiotemporal structure presupposed by the
laws is apart from material bodies, as premise two claims. After all, the
relationalist, in my view, can countenance this structure, but will say that it
has to do with the (actual and perhaps possible) spatiotemporal relations
between material bodies. In what way do the laws presuppose a spatiotem-
poral structure that is in addition to material bodies? The answer comes
from the way that the fundamental laws are usually formulated. (I turn to
potential reformulations in Section 5.) These laws are typically formulated
to directly mention material bodies, with a term that directly refers to
them-such as the mass term of Newton's dynamics, or the mass density
of some formulations of Newtonian gravitation, or the elements mentioned
in Aristotle's laws, or the stress-energy tensor of general relativity. 47 At the
same time, these laws also presuppose that the world has a spatiotemporal
structure apart from those bodies-apart in that it is presupposed by the
laws in the mathematical sense given above, or else is directly mentioned by
or coded up in a distinct term.
Recall that the matching principle tells us to infer that a special relativistic
world lacks an absolute simultaneity structure. The laws don't require this
mathematical structure, which suggests that the world doesn't have the
corresponding physical structure. To fail to adhere to the matching principle
is to fail to heed this evidence from the laws about what the world is like.
The relationalist fails to adhere to the primary reading of the principle in the
same way. The fundamental laws are giving us evidence that spatiotemporal
structure is fundamental to the physical world, which the relationalist fails
to heed. The relationalist may respond that there are good reasons to
47
In the context of this debate, both views take certain material objects to exist at the
fundamental level. (Supersubstantivalism would then deny this.)
32 Jill North
disregard this apparent evidence from the laws. The burden is then on the
relationalist to show this, just as the burden falls on the proponent of
absolute simultaneity.
You might think that there are two distinct notions, that of what's
physically fundamental versus metaphysically fundamental; that the matching
principle governs the first whereas substantivalism and relationalism are
views about the second; and conclude that the argument from the matching
principle doesn't make contact with those views. In particular, you might
think it open for the relationalist to say that spatiotemporal structure is
metaphysically nonfundamental, in accord with relationalism, yet physically
fundamental, in accord with the matching principle-that a world's spatio-
temporal structure is less metaphysically fundamental than, but more phys-
ically fundamental than, the spatiotemporal relations between material
bodies. I suppose that such a view is possible, but it seems implausible on
its face. Imagine an analogous reductionist who says that macroscopic
systems (boxes of gas) are metaphysically nonfundamental, grounded in
more fundamental microscopic objects (their particles), yet physically fun-
damental. This is a puzzling view. Surely the thought that microscopic
objects are metaphysically fundamental goes hand in hand with evidence
from physics suggesting that they are physically fundamental. Relative
physical and metaphysical fundamentality cannot plausibly go in opposite
directions. More generally, I'm inclined to reject the idea that there are two
distinct notions of fundamentality here.
Suppose that what I have been calling "spatiotemporal structure"
involves, at least in part, facts that must be stated using universal general-
izations. On a standard axiomatic approach to geometry, for instance, a
given spatiotemporal structure will be defined via a universal generalization
over a domain of points. Suppose further that generalizations are not
fundamental but grounded in their instances, in accord with a familiar
way of thinking about grounding. Then it may seem as though the sub-
stantivalist doesn't adhere to the matching principle either, simply because
spatiotemporal structure, qua generalizations, cannot be fundamental.
However, the substantivalist will avoid the wony, for one of the following
reasons. First, one might for independent reasons think that generalizations
are fundamental, a not-unprecedented (to my mind, not implausible) view,
even among grounding proponents. Second, even if spatiotemporal-
structure-qua-generalizations is not absolutely fundamental, it is very close
to being fundamental, so that the fundamental structure of the world almost
directly matches the structure for the fundamental laws. The only "gap"
there is between spatiotemporal structure and the fundamental level is the
one created by the gap between generalizations and their instances. This is
an intuitively smaller gap than that between a world's spatiotemporal
A New Approach to the Relational-Substantival Debate 33
structure and features of material bodies. The former is just a "gap in logical
form"-the "size" of the separation between a generalization and the col-
lection of particular claims that grounds it-whereas the latter is a larger,
physical gap. The substantivalist then adheres to the matching principle
more than the relationalist does. Finally, notice that even if the generaliza-
tions that axiomatize a given structure are not absolutely fundamental, the
various facts about the points still can be, and these facts are included in my
conception of spatiotemporal structure; in which case there are still funda-
mental spatiotemporal facts or structure apart from material bodies.
(The worry would also seem to go too far. It would force us to say that no
particular collection of fundamental facts is to be preferred to any other on
the basis of the physical laws, simply because any structure required for those
laws takes the form of a generalization, and no generalization is fundamental.
But surely a matching-type argument can sometimes work-as when we
want to say that Berkeleyan idealism posits a world that radically fails to
match the structure indicated by the laws. It seems we might reject that view
for the reason that the fundamental nature of the world does not match the
structure for the laws-even though that structure is given by generaliza-
tions, and even if generalizations are not fundamental but grounded in
their instances.)
Notice that the argument for substantivalism is independent of one's view
on the metaphysics of laws. The question of what makes a statement a law is
distinct from the injunction to posit, assuming that a certain statement is a
law, the requisite structure in the world. Even the Humean, who denies that
laws of nature are metaphysically fundamental, can agree to posit, in the
fundamental physical level of the world, the structure presupposed by
the fundamental physical laws. To put it another way, the content of the
law claim, the proposition p of the statement "it is a law that p," is what
indicates structure in the world. It is irrelevant whether what makes it the case
that p is a law is itself metaphysically fundamental. Whatever your account of
laws of nature, you can, and should, adhere to the matching principle.
Current physics therefore gives us reason to believe that substantivalism
is correct. Nonetheless, it is open for future physics to turn the tide. If
a quantum theory of gravity or some other future fundamental theory
contains laws that only presuppose things about material bodies and their
relations, which in turn give rise to the spatiotemporal structure presup-
posed by current theories, we can conclude that relationalism is correct.
Future laws might even suggest a view that doesn't look like either
relationalism or substantivalism, presupposing facts about neither mater-
ial bodies nor spatiotemporal structure but something else. (A causal set
theory approach to quantum gravity, for example, might support rela-
tionalism, depending on the particulars, or it could be a case on which
34 Jill North
neither view is correct. 48 ) In this way the debate will remain relevant to,
and continue to be informed by, future developments in physics.
5. A CHALLENGE FOR RELATIONALISM
Finally, let me turn to the question raised at the end of Section 2. I have
been assuming that the fundamental laws we currently have are formulated
to presuppose a spatiotemporal structure apart from material bodies. This
reveals one other way for the tide to turn: the relationalist could try to
reformulate these laws to only presuppose things about material bodies. If
such a reformulation is possible, then the argument will turn on how we
should generally formulate the laws, which is a big question that I can't fully
answer here. Even so, the argument poses a significant challenge to any
relationalist attempt to reformulate the laws.
Consider an illustrative example: the relationalist reformulation of New-
tonian mechanics initially suggested by Bas van Fraassen (1970, sec. 4.1)
and filled out in one way by Nick Huggett (2006). According to their idea,
we can reformulate Newtonian mechanics to include the statement that,
"Newton's Laws hold in some frames," where these will be the inertial
frames. (There is also a force law, and on Huggett's account a law about
the spatial geometry.) These laws then pick out a standard of inertia or
straightness of trajectories-they recognize a quantiry of, or facts about,
acceleration-without assuming that spacetime exists. In my terms, they
only presuppose spatiotemporal facts about material bodies. This is because,
according to Huggett, the facts about inertial frames-indeed, all the
spatiotemporal facts-themselves supervene on facts about the history of
relations between material bodies. (Huggett rejects modal relationalism.)
This is a genuinely relationalist formulation, on my construal, which
respects the primaty reading of the matching principle. The truth of the
laws in certain frames effectively substitutes for an inertial structure, so that
the laws themselves do not have to mention or presuppose this structure.
The problem is that this is a worse formulation of the laws, for a couple of
reasons. First, this formulation does not respect the idea that fundamental
laws only mention fundamental things. These laws are given in terms of
facts about inertial frames, which for Huggett are not fundamental but
grounded in facts about the relations between material bodies.
48
See Huggett and Wiithrich (2013) and the other papers in that journal issue on the
emergence of spacetime in quantum gravity.
A New Approach to the Relational-Substantival Debate 35
Second, this formulation is given in terms of reference frames. Why is this
wors. el I take
' it that fundamental physical laws are .best formulated
, in terms
.
f rhinas about the world itself, and reference frames don t fit the bill.
o ord~1 g to Newton's laws, inertial frames are like units of measure or
Acc c b ! . .
coordinate systems, in that a choice of 1rame is an ar itrary c 101ce m
description. Now, Huggett's formulation does not mentio~ any particular
c "'e nor does it directly mention inertial frames. Instead It says that there
rra .. , ' , ! 49 B h c
e frames you can choose such that Newton s aws are true. ut t e ract
arhat a choice of inertial frame is ar b itrary suggests t !iat .mertta. 1 rrames
c ·
111
t articular, and reference frames in general-these objects as a group or kind
pf thing-are merely descriptive or labeling devices we use, not inherent in
0
hysical systems themselves; 50 hence they should not, other things equal, be
P entioned in the fundamental physical laws. I gather that this is what
:nderlies the general feeling in foundational discussions that formulating
the laws in geometric, coordinate-free terms is desirable. (Consider formu-
lations of classical mechanics in terms of so-called generalized coordinates,
which do not mention any particular coordinate system. Even this reference
to coordinates is seen as ideally replaceable by geometric objects with no
mention of coordinates.)
An idea from Hartry Field bolsters the thought that such a formulation is
worse in this way. Field draws a distinction between 'intrinsic' and 'extrinsic'
explanations. The former "explain what is going on without appeal to
extraneous" entities, things "extrinsic to the process to be explained" (1980, 43).
As a result, intrinsic explanations are better, more "illuminating" (1980, 43) or
"satisfying" (1989, 18). He says,
[E]xtrinsic explanations are often quite useful. But it seems to me that whenever one
has an extrinsic explanation, one wants an intrinsic explanation that underlies it: one
wants to be able to explain the behaviour of the physical system in tenns of the
intrinsic features of that system, without invoking extrinsic entities ... whose proper-
ties are irrelevant to the behaviour of the system being explained. If one cannot do
this, then it seems rather like magic that the extrinsic explanation works. (1989, 193;
original italics)
The best explanations cite intrinsic features relevant to the system's
behavior.
By analogy to Field's idea, call formulations of the laws in terms of
reference frames or coordinate systems or the like "extrinsic formulations."
49 See Dorr (2010) for argument that "existential quantification as such is a distinctive
source of badness" (166; original italics).
°
5 Compare Einstein on a coordinate system, which is "only a means ofdescription and
in itself has nothing to do with the objects to be described" (2002, 203; original italics).
36 Jill North
Extrinsic formulations are then worse for the same reasons Field says that
extrinsic explanations are worse: they reference things outside the system
or world itself, whose properties aren't directly relevant to the system's
behavior. 51 This makes the success of the formulation seem like magic. All
things equal, it is better to have an intrinsic formulation-or what I prefer to
call a direct formulation, since extrinsic entities, like coordinate labels, can
tell us about the system in question; only they do so in an indirect, and
therefore less preferable, way. It's analogous to characterizing the geometry
of the Euclidean plane by saying that, "there are coordinate systems in which
the distance formula takes the usual Pythagorean form," rather than by
giving the metric tensor (or, for that matter, Euclid's axioms). That charac-
terization gives the structure of the plane, but in a needlessly indirect way,
by means of the kinds of coordinate systems we can lay down on top of it.
Better to have a formulation of the laws that more directly reflects realiry.
(It is not uncommon for physics books to state the laws in terms of reference
frames or coordinate systems. The claim is that this is not the best formulation.)
Of course, direct formulations may seem preferable only if you are a
realist to begin with-only if you think that it is the job of a physical theory
to tell us what the world is like. An instrumentalist may be unbothered
by indirect formulations and extrinsic explanations. (The instrumentalist
should be used to the charge that the success of science seems like magic.)
Since it is not my aim to argue for realism here, I leave it to the anti-realist to
parry the objection that such formulations are worse. Let me note, though,
that indirect formulations seem particularly problematic for fundamental
laws, since the elements that feature in them, like reference frames or
coordinate systems, don't seem the sorts of things that can be truly funda-
mental or explanatory.
There are other relationalist reformulations to consider in more detail
than I have space to do here. However, the above strikes me as indicative of
the kinds of problems that any such reformulation will face. In order for
relationalism to be victorious, the proffered reformulation must be genuinely
relationalist, presupposing facts only about material bodies; it should be
direct; and it should respect the primary reading of the matching principle.
A brief look at three more examples further suggests that a relationalist
reformulation meeting these constraints will be hard to come by. (1) Julian
Barbour's relationalist mechanics (Barbour (1982; 2000; 2001); Barbour
51
Consider Field's reason that a scientific explanation citing direct relations between
physical objects and numbers is extrinsic and therefore worse: "[T]he role of the numbers
is simply to serve as labels for some of the features of the physical system: there is no
pretense that the properties of the numbers influence the physical system whose behav-
iour is being explained" (1989, 192-3). The role of reference frames in physics is similar.
A New Approach to the Relational-Substantival Debate 37
and Bertotti (1982)), which eschews any fundamental temporal structure,
arguably presupposes a spatial structure above that of material bodies, 52 in
which case the theory is substantivalist, on my understanding. Setting that
aside, the theo1y is not formulated directly. 53 (2) David Albert (1996)
suggests that in classical mechanics, the Hamiltonian energy function
gives rise to a three-dimensional spatial structure. Since the Hamiltonian
is defined in terms of particle features, this may count as a relationalist
rheory, on my construal. (Albert is not arguing for relationalism.) Yet there
is also a case to be made that the mathematical formulation presupposes a
spatial structure apart from material bodies (in partirnlar for the kinetic
energy term), in which case it would either count as substantivalist, or fail to
respect the primary reading of the matching principle. (3) Huggett men-
tions another law of his reformulation of Newtonian mechanics: "'There is
an embedding of the relational histo1y into G', for some specific Riemann-
ian geometry G" (2006, 53), where for him the privileged embedding
supervenes on the history of relations between material bodies. Facts
about the embedding geometty (spatial structure) are not fundamental but
grounded in facts about material bodies. This makes the law relationalist.
The problem is that it, too, explicitly mentions nonfundamental things, and
is formulated indirectly, in terms of a structure into which the relations can
be embedded. (A similar charge applies to Albert's (2012) suggestion for a
relationalist Newtonian mechanics that says: "The physically possible his-
tories of inter-particle distances are those which can be embedded in a full
substantivalist Newtonian space, or imagined as taking place in such a space,
in such a way as to satisfy F = ma.")
This does not prove that no relationalist reformulation can succeed, and
more work must be done to fully evaluate the various proposals on offer in
these terms. 54 But it does suggest that it won't be easy to find a relationalist
52 See the presentation in Earman (1989, secs. 2.1, 5.2). Arntzenius (2012, sec. 5.11);
Pooley (2013, sec. 6.2) suggest this for Barbour's reformulation of general relativity in
particular.
53
The indirectness enters in recovering the topological temporal structure and the
inertial structure: Arntzenius (2012, chs. l, 5).
54
A few more examples. On the dynamical approach of Brown (2005); Brown and
Pooley (2006), a world's spatiotemporal structure holds in virtue of the behavior of
material bodies via the laws and their symmetries. This seems relationalist, on my
conception (in particular if the laws are grounded in facts about material bodies). They
presumably reject my idea that the laws presuppose a certain structure in order to be
formulated. Another relationalist theory is that of Belot (1999; 2000), which seems
indirectly formulated (cf. Brown and Pooley (2002, 192-3); it also presupposes a
temporal structure apart from material bodies: Brown and Pooley (2002, 194)). Another
is that of Albert (2017), on which there is no fundamental, pre-dynamical spatiotemporal
structure: all spatiotemporal facts are grounded in facts about the behaviors of material
bodies. Albert reformulates the laws in an indirect way.
38 Jill North
reformulation that has the features we want of fundamental laws. Current
laws are generally formulated to presuppose a spatiotemporal structure apart
from material bodies. The problem is that typical relationalist substitutes for
that kind of structure-facts about things like reference frames or coordin-
ate systems or embedding geometries-are not candidates for direct formu-
lations of the laws. Future laws, however, may be different.
6. CONCLUSION
Many people have thought that the arguments for relationalism or sub-
stantivalism will have to resort to considerations like simplicity, ontological
parsimony, or explanatory power. 55 Some have said that the relationalist's
ontology is more parsimonious, and therefore favored by Occam's razor. 56
Others have said that the substantivalist's theory is simpler, and therefore
favored by ordinary criteria of theory choice. 57 Some have argued that
the relationalist's theory is more explanatory. Others have claimed that the
substantivalist's is. 5B
You might conclude that the debate is hopelessly vague, since the criteria
of simplicity, parsimony, and explanatory power needed to adjudicate it are
themselves vague; nor is it clear which to favor when these virtues com-
pete.59 I don't object to relying on such considerations even so, but it is
worth noting that the argument from the matching principle is different.
The matching principle doesn't say to refuse to posit unnecessary entities or
to go with the simplest or most explanatory theory. It says to posit in the
world the structure presupposed by the laws. The argument based on this
principle escapes those particular worries about the status of the debate.
The matching principle is a familiar and successful guiding principle. It
applies, in the first instance, to the fundamental laws and fundamental level
55 Dasgupta (2015) discusses the effects of these criteria on the spacetimc debate for
classical physics.
56 Huggett (2006); Huggett and Hoefer (2009); Pooley (2013).
57 Huggett (2006); Arntzenius (2012, ch. 5).
58 Earman (1989) suggests that the relationalist' s theory will be worse; Brown and
Pooley (2002) argue against this. Maudlin (1993, 196) says that the substantivalist's
theory is more explanatory in some ways; Nerlich (1994a; 1994b) argues that it is more
explanatory in general.
59 See Horwich (1978); Earman (1989, sec. 3.3); Huggett (2006, 70) for this kind of
complaint. Sklar (1974, 231) notes a tradeoffbetween the substantivalist's explanatory
power and relationalist's parsimony; Mundy (1983, 207) notes one between the relation-
alist's parsimony and substantivalist's simplicity. Belot (2011) suggests that parsimony in
fact favors substantivalism (while arguing against using simplicity considerations to draw
metaphysical conclusions).
A New Approach to the Relational-Substantival Debate 39
of physical reality. The substantivalist and relationalist, as I see them,
disagree about the fundamental physical level, which is why the matching
principle can distinguish between them. This is a substantive debate about
rhe fundamental nature of the world according to physics; a debate
about what makes it the case that the spatiotemporal structme required
by the physics holds.
The traditional debate centered on whether we need to posit an inde-
pendently existing space in order to account for objects' motions. The
debate that I have presented is a natural descendant: a debate about whether
we need to posit a spatiotemporal structure apart from material bodies to
support the theory that best accounts for objects' motions. This is a
substantive debate, which we currently have reason to believe the substan-
rivalist is winning. 60
Rutgers University
REFERENCES
Albert, David (1996). "Elementary Quantum Metaphysics." In J. T. Cushing,
A. Fine, and S. Goldstein, eds., Bohmian Mechanics and Quantum Theory: An
Appraisal, pp. 277-84. Dordrecht: Kluwer Academic.
Albert, David (2000). Time and Chance. Cambridge, MA: Harvard University Press.
Albert, David (2012). "Philosophy of Physics." In Encyclopaedia Britannica Online
Academic Edition. Encyclopaedia Britannica Inc.: https://academic.eb.com/.
Albert, David (2017). "On the Emergence of Space and Time." Unpublished
manuscript.
Arntzenius, Frank (2012). Space, Time, and Stuff Oxford: Oxford University Press.
Bain, Jonathan (2006). "Spacetime Structuralism." In Dennis Dicks, ed., The
Ontology of Spacetime, Volume I, pp. 37-65. Amsterdam: Elsevier.
Baker, David John (2005). "Spacetime Substantivalism and Einstein's Cosmological
Constant." Philosophy of Science (Proceedings) 72(5): 1299-311.
Barbour, Julian ( 1982). "Relational Concepts of Space and Time." British Journalfor
the Philosophy of Science 33(3): 251-74.
Barbour, Julian (2000). The End of Time. New York: Oxford University Press.
60
For comments and discussion, I am grateful to Ori Belkind, Karen Bennett, Jim
Binkoski, Carolyn Brighouse, Andrew Chignell, Ted Sider, and audiences at Rutgers
University, the University of Illinois-Champaign, the University of Western Ontario, the
University of Wisconsin-Milwaukee, Brown University, the Philosophy of Science Asso-
ciation, the University of North Carolina-Chapel Hill, Tel Aviv University, the Hebrew
University of Jerusalem, the University of Massachusetts-Amherst, and the University of
Oxford. Many thanks also to the anonymous reviewers for, and the edirors of, this
volume. This research was funded in large part by the National Science Foundation
STS Scholars Award No. 1430435.
40 Jill North
Barbour, Julian (2001). The Discovery ofDynamics. New York: Oxford University Press.
Barbour, Julian and Bruno Bertotti (1982). "Mach's Principle and the Structure of
Dynamical Theories." Proceedings of the Royal Society A 382(1783): 295-306.
Beebee, Helen (2000). "The Non-Governing Conception of Laws of Nature."
Philosophy and Phenomenological Research 61: 571-94.
Belot, Gordon (1999). "Rehabilitating Relationalism." International Studies in the
Philosophy a/Science 13(1): 35-52.
Bclot, Gordon (2000). "Geometry and Motion." British journal far the Philosophy of
Science 51: 561-95.
Bclot, Gordon (2011). Geometric Possibility. Oxford: Oxford University Press.
Belot, Gordon and John Earman (2001). "Pre-Socratic Quantum Gravity." In Craig
Callender and Nick Huggett, eds., Physics Meets Philosophy at the Planck Scale,
pp. 213-55. Cambridge: Cambridge University Press.
Brading, Katherine and Elena Castellani (2007). "Symmetries and Invariances in
Classical Physics." In Jeremy Butterfield and John Earman, eds., Handbook of the
Philosophy o/Science: Phiwsophy ofPhysics, Part B, pp. 1331-67. Amsterdam: Elsevier.
Brighouse, Carolyn (1994). "Spacetime and Holes." In David Hull, Mickey Forbes,
and Richard Burian, eds., PSA: Proceedings ofthe Biennial Meeting ofthe Philosophy
of Science Association, 1994, Volume I: Contributed Papers, pp. 117-25. East
Lansing, MI: Philosophy of Science Association.
Brighouse, Carolyn (1999). "Incongruent Counterparts and Modal Relationism."
International Studies in the Philosophy of Science 13(1): 53-68.
Brighouse, Carolyn (2014). "Geometric Possibility-An Argument from Dimen-
sion." European journal for Philosophy of Science 4( 1): 31-54.
Brown, Harvey R. (2005). Physical Relativity: Space-Time Stn1cture ftom a Dynamical
Perspective. Oxford: Oxford University Press.
Brown, Harvey R. and Oliver Pooley (2002). "Relationalism Rehabilitated? I:
Classical Mechanics." British journal far the Philosophy of Science 53: 183-204.
Brown, Harvey R. and Oliver Pooley (2006). "Minkowski Space-Time: A Glorious
Non-Entity." In Dennis Dicks, ed., The Ontology of Spacetime, pp. 67-89.
Amsterdam: Elsevier.
Curiel, Erik (2016). "On the Existence ofSpacetime Structure." Britishjournalfar
the Philosophy of Science. doi: 10.1093/bjps/axw014.
Dasgupta, Shamik (2009). "Individuals: An Essay in Revisionary Metaphysics."
Philosophical Studies 145: 35-67.
Dasgupta, Shamik (2011). "The Bare Necessities." Philosophical Perspectives
25: 115-60.
Dasgupta, Shamik (2015). "Substantivalism vs Relationalism about Space in Clas-
sical Physics." Philosophy Compass 10(9): 601-24.
DiSalle, Robert (1994). "On Dynamics, Indiscernibility, and Spacetime Ontology."
British journal far the Philosophy of Science 45( 1): 265-87.
DiSalle, Robert (2002). "Newron's Philosophical Analysis of Space and Time." In
I. Bernard Cohen and George E. Smith, eds., The Cambridge Companion to Newton,
pp. 33-56. Cambridge: Cambridge University Press.
DiSalle, Robert (2006). Understanding Space-Time: The Philosophical Development of
Physics ftom Newton to Einstein. Cambridge: Cambridge University Press.
A New Approach to the Relational-Substantival Debate 41
Dorat<), Mauro (2000). "Substantivalism, Relationism, and Structural Space-time
Realism." Foundations ofPhysics 30(10): 1605-28.
Dorato, Mauro (2008). "Is Structural Spacetime Realism Relationism in Disguise?
The Supererogatoty Nature of the Substantivalism/Relationism Debate." In
Dennis Dicks, ed., The Ontology o/Spacetime, Volume II, pp. 17-37. Amsterdam:
Elsevier.
Dorr, Cian (2010). "Of Numbers and Electrons." Proceedings of the Aristotelian
Society CX: 133-81.
Earman, John (1970). "Who's Afraid of Absolute Space?" Australasian journal of
Philosophy 48(3): 287-319.
Earman, John (1989). World Enough and Space-Time. Cambridge, MA: MIT Press.
Earman, John and John Norton (1987). "What Price Spacetime Substantivalism?
The Hole Story." British journal far the Philosophy of Science 38: 515-25.
Einstein, Albert (2002). The Collected Papers of Albert Einstein, Volume 7, trans.
Alfred Engel. Princeton, NJ: Princeton University Press.
Esfeld, Michael and Vincent Lam (2008). "Moderate Structural Realism about
Space-Time." Synthese 160: 27-46.
Field, Hartty (1980). Science Without Numbers. Oxford: Blackwell.
Field, Hartry (1984). "Can We Dispense With Spacetime?" In P. Asquith and
P. Kitcher, eds., PSA 1984: Proceedings of the 1984 Biennial Meeting of the
Philosophy a/Science Association, Volume 2, pp. 33-90. East Lansing, MI: Michigan
State University Press.
Field, Hartty (1989). Realism, Mathematics and Modality. Oxford: Blackwell.
Fine, Kit (2001). "The Question of Realism." Philosopher's Imprint 1: 1-30.
Fine, Kit (2012). "Guide to Ground." In Fabrice Correia and Benjamin Schnieder,
eds., Metaphysical Grounding: Understanding the Structure of Reality, pp. 37-80.
Cambridge: Cambridge University Press.
Friedman, Michael (1983). Foundations of Space-Time Theories. Princeton, NJ:
Princeton University Press.
Geroch, Robert (1981). General Relativity from A to B. Chicago, IL: University of
Chicago Press.
Greaves, Hilary (2011). "In Search of (Spacetime) Structuralism." Philosophical
Pmpectives 25: 189-204.
Hicks, Michael Townsen and Jonathan Schaffer (2017). "Derivative Properties in
Fundamental Laws." British journal far the Philosophy of Science 68(2): 411-50.
Hirsch, Eli (2011). Quantifier Variance and Realism: Essays in Metaontowgy.
New York: Oxford University Press.
Hoefer, Carl (1996). "The Metaphysics of Space-Time Substantivalism." The journal
ofPhilosophy 93(1): 5-27.
Hoefer, Carl (1998). "Absolute versus Relational Spacetime: For Better or
Worse, the Debate Goes On." British journal for the Philosophy of Science
49: 451-67.
Horwich, Paul (1978). "On the Existence ofTime, Space and Space-Time." Nml.s 12(4):
397-419.
Huggett, Nick (1999). Space fiwn Zeno to Einstein: Classic Readings with a Contem-
pormy Commentaiy. Cambridge, MA: MIT Press.
42 Jill North
Huggett, Nick (2006). 'The Regularity Account of Relational Spacetime." Mind
115: 41-73.
Huggett, Nick and Carl Hoefer (2009). "Absolute and Relational Theories of Space
and Motion." Stanford Encyclopedia of Philosophy (Fall 2009 Edition): https://
stanford. library.sydney .edu. au/ archives/ fall2009 I en tries/ spacetime-theories/.
Huggett, Nick and Christian Wi.ithrich (2013). "The Emergence of Spacetime in
Quantum Theories of Gravity." Studies in Hist01y and Philosophy ofModern Physics
44: 273-5.
Ismael, Jenann and Bas C. van Fraassen (2003). "Symmetry as a Guide to Super-
fluous Theoretical Structure." In Katherine Brading and Elena Castellani, eds.,
Symmetries in Physics: Philosophical Reflections, pp. 371-92. Cambridge:
Cambridge University Press.
Knox, Eleanor (2014). "Newtonian Spacctime Structure in Light of the Equivalence
Principle." British jounzal for the Philosophy of Science 65: 863-80.
Ladyman, James and Don Ross (2009). Eveiy Thing Must Go: Metaphysics Natural-
ized. Oxford: Oxford University Press.
Leeds, Stephen (1995). "Holes and Determinism: Another Look." Philosophy of
Science 62: 425-37.
Loewer, Bany (2001). "From Physics to Physicalism." In Carl Gillet and Barry Loewer,
eds., Physicalism and Its Discontents, pp. 37-56. Cambridge: Cambridge University
Press.
Malament, David (1976). "Review of Lawrence Sklar, Space, Time, and Space-
time." journal ofPhilosophy 73(11): 306-23.
Malament, David (1982). "Review of Hartry Field, Science Without Numbers:
A Defense ofNominalism." journal of Philosophy 79(9): 523-34.
Manders, Kenneth L. (1982). "On the Space-Time Ontology of Physical Theories."
Philosophy of Science 49(4): 575-90.
Maudlin, Tim (1993). "Buckets of Water and Waves of Space: Why Spacetime is
Probably a Substance." Philosophy of Science 60(2): 183-203.
Maudlin, Tim (2012). Philosophy ofPhysics: Space and Time. Princeton, NJ: Princeton
University Press.
Maudlin, Tim (2015). "How Mathematics Meets the World": https://fqxi.org/
data/ essay-contest-files/Maudlin_How_Mathematics_Mee. pdf.
Mundy, Brent (1983). "Relational Theories of Euclidean Space and Minkowski
Space-Time." Philosophy of Science 50: 205-26.
Mundy, Brent (1986). "Embedding and Uniqueness in Relational Theories of
Space." Synthese 67(3): 383-90.
Mundy, Brent (1992). "Space-Time and Isomorphism." Philosophy of Science
(Proceeding;) 1: 515-27.
Nerlich, Graham (1994a). The Shape of Space. 2nd edn. Cambridge: Cambridge
University Press.
Nerlich, Graham (l 994b). What Spacetime Explains: Metaphysical Essays on Space and
Time. Cambridge: Cambridge University Press.
North, Jill (2008). "Two Views on Time Reversal." Philosophy ofScience 75(2): 201-23.
North, Jill (2009). "The 'Structure' of Physics: A Case Study." journal ofPhilosophy
106(2): 57-88.
A Nezu Approach to the Relational-Substantival Debate 43
l A (2013). "Categorical Priority and Categorical Collapse." Aristotelian
Pau'1 " ·
Society Supplementmy Volume 87: 89-113.
1 y Oliver (2013). "Substantivalist and Relationalist Approaches to Spacetime."
Po~ne Robert Batterman, ed., The Oxford Handbook of Philosophy of Physics,
522-86. Oxford: Oxford University Press.
R
0
k~ts, John T. (2008). "A Puzzle about Laws, Symmetries and Measurability."
British journal far the Philosophy of Science 59: 143-68.
Rosen, Gideon (2010). "Metaphysical Dependence: Grounding and Reduction." In
Bob Hale and Aviv Hoffmann, eels., l\llodality: Metaphysics, Logic, and Epistemology,
PP· 109-35. Oxford: Oxfor~, University Press. .
R nasiewicz, Robert (1996). Absolute Versus Relational Space-Time: An Out-
)'rnoclecl Debate?" journal of Philosophy 93(6): 279-306.
Rynasiewicz, Robert (2000). "On the Distinction Berween Absolute and Relative
Motion." Philosophy of Science 67: 70-93.
Saunders, Simon (2013). "Rethinking Newton's Principia." Philosophy of Science
80(1): 22-48.
Schaffer, Jonathan (2009). "On What Grounds What." In David Chalmers, David
Manley, and Ryan Wasserman, eds., Jvfetarnetaphysics, pp. 347-83. Oxford:
Oxford University Press.
Sider, Theodore (2011). Writing the Book of the World. Oxford and New York:
Oxford University Press.
Sklar, Lawrence (1974). Space, Time, and Spacetirne. Berkeley, CA: University of
California Press.
Slowik, Edward (2005). "Spacetime, Ontology, and Structural Realism." Inter-
national Studies in the Philosophy of Science 19(2): 147-66.
Slowik, Edward (2016). The Deep Metaphysics of Space: An Alternative Histo1y and
Ontology Beyond Substantivalism and Relationism. Switzerland: Springer.
Stein, Howard (1970). "Newtonian Space-Time." In Robert Palter, ed., The Annus
Mirabalis of Sir Isaac Newton, 1666-1966, pp. 258-84. Cambridge, MA: MIT
Press. (Originally published in Texas Quarterly (1967) 10: 174- 200.)
Stein, Howard (l 977a). "On Space-Time and Ontology: Extracts from a Letter to
Adolf Gri.inbaum." In John S. Earman, Clark N. Glymour, and John J. Stachel,
eds., Foundations of Space-Time Theories: Minnesota Studies in the Philosophy of
Science, Volume VIII, pp. 374-402. Minneapolis, MN: University of Minnesota
Press.
Stein, Howard (l 977b). "Some Philosophical Prehistory of General Relativity."
In John S. Earman, Clark N. Glymour, and John J. Stachel, eds., Foundations of
Space-Time Theories: Minnesota Studies in the Philosophy ofScience, Volume VIII,
pp. 3-49. Minneapolis, MN: University of Minnesota Press.
Teller, Paul (1991). "Substance, Relations, and Arguments about the Nature of
Space-Time." The Philosophical Review 100(3): 363-97.
van Fraassen, Bas C. (1970). An Introduction to the Philosophy of Time and Space.
New York: Columbia University Press.
Wallace, David (2012). The Emergent Multivme: Quantum The01y According to the
Everett Interpretation. Oxford: Oxford University Press.

2 ,eight in what follows on the arguments from ordinary language. In my
Relative Locations
Andrew Bacon
Substantivalism is the view that locations exist independently of the objects
that they are locations of. Thus, for example, the moon's present. location
would have existed even if it had not been located there and, reciprocally,
there are regions that could have been the location of some material object,
but aren't (unoccupied regions).
Substantivalism can have a number of different motivations. In ordinary
English we frequently talk about, and quantify over, locations. For example,
I can talk about where Emily went on vacation, or where I left my keys; I can
in some cases talk about unoccupied locations, such as the place I would
have been had I traveled halfWay to the moon. The substantivalist has a
ready answer for what we are doing when we quantify in this way:. we. are
quantifying over regions of space-time. If this were the central m~t1vat1on,
however, it would suggest that the dispute is primarily about the existence of
locations. This is not quite right: even those who oppose substantivalism-
the relationists-are often happy to accept ordinary talk about locations
provided that that talk can be recovered from a more acceptable fundamen-
tal ontology. (For these theorists, locations are a bit like shadows and holes:
they are, metaphysically speaking, 'second-rate' entities in some sense ..We
can readily quantify over them, but their existence depends on other kmds
of things: a location on the thing that occupies it, 1 much like a shadow on
the thing that casts it, and a hole on the thing it penetrates.)
Substantivalists arguably should not accept this reduction of our ordinary
talk of locations to quantification over regions of space-time either. My
pocket is not a region of space-time, and nor is Pari~-for ex~mple, my
pocket changes its shape as it goes through the washmg machme, ~nd a
region of space-time cannot change its shape. But my pocket and Pans are
1 Or perhaps, if the region is empty, the possible objects that could have occupied it.
See Forbes (1993).
Relative Locations 45
rhe kinds of things I might be asking ab?ut when I ask where my keys are or
hen I ask where Emily went on vacanon. 2
w Since I find the above sorts of objections compelling, I shall place little
w. \V , the ·strongest motivations for substantivalism-the motivations that
vie
lay a central role in this paper-are slightly more theoretical. Space-time
points tend to appear in the formulation of many of our simplest physical
~heories and this gives us good reason to take their existence seriously.
Unfortunately the most straightforward versions of substantivalism and
relationism suffer from having a pair of undesirable, and arguably related,
consequences. Substantivalism predicts the existence of physically indistin-
uishable worlds that differ solely concerning where things are located. For
~ccording to substantivalism there could be two worlds that differ from each
other only in that each object's location in one world has been displaced
relative to the other by some fixed distance in some particular direction (see
Leibniz and Clarke 2007).
By refusing to take talk of locations and regions of space-time at face value
the relationist does not face this problem in its most acute form. However,
relationists are subject to a somewhat similar concern. The most straight-
forward versions of relationism take geometrical properties and relations to
be a matter of fundamental relations holding between material objects and
abstract objects, such as numbers to represent distances between objects,
vectors to represent forces, and so on. However much like the substantival-
ist's location relation, I shall argue, one can have physically indistinguishable
worlds in which the relations between the material objects and the
appropriate abstracta have undergone a similar kind of displacement.
In this paper I shall in response to these worries be advocating for a novel
version of substantivalism underwritten by a non-standard account of
the relation between material objects and the regions they occupy. As
I cautioned against earlier, the theory is not intended to model our ordinary
way of speaking about locations. However, the theory does allow one to give
a story about how the shapes of objects, the distances between them, and
other such facts can emerge from the relations they stand in to a sufficiently
structured space-time manifold.
Here is an outline of the paper. In section l, I review Leibniz's shift
argument against substantivalism and argue that it provides us with
reasons to look for alternatives to the orthodox version of substantivalism.
In section 2.1, I consider the analogous problem involving abstracta for
2
Of course some people-the supersubstantivalists-do identify my pocket with a
region; however, some fairly elaborate maneuvers are needed to make sense of the idea
that ordinary things could have had different shapes.
 46 Andrew Bacon
(versions of) relationism, and consider the options for a relationist who
wishes to give an account of geometrical properties without reference to
abstract objects. To this effect I outline in section 2.2 a result that says
that any first- or second-order language that can express, relative to some
class of models, all possible configurations of a system of n particles either
must have an infinite set of fundamental primitives or must admit a mode(
with an infinite ontology (whether that be an ontology of space-time
points, abstract objects, or something else). In section 3.1, I develop a
substantivalist theory according to which an object is located at a region of
space-time if and only if it is located at every transformation of that
region. Thus unlike orthodox substantivalism, objects are multiply located
in a fairly far-reaching way. Finally, I consider two farther technical issues
in the appendices: the treatment of space-time fields and the first-order
theo1y governing the interaction of parthood and location.
1. LEIBNIZ'S SHIFT ARGUMENT
Substantivalism is subject to an old objection that traces all the way back to
Leibniz (Leibniz and Clarke 2007). Pick some direction and distance, and
imagine a world at which the location of every object has been translated
uniformly in that direction by that distance, but in which every other
property of the object has been kept the same. According to the substantival-
ist, worlds related by such operations are genuinely different because they
disagree about facts concerning which particular regions of space-time each
object is located at. Substantivalism thus falls afoul of the following seem-
ingly attractive thesis:
No SHIHS: There are no differences between shifted worlds.
A pair of worlds are 'shifts' of one another if they are related by a translation
of every object's location of the sort described above. The crucial feature of
shifts to highlight here is that they preserve all the laws of physics, and
arguably all observable properties and relations between objects. 3 Very
similar arguments could be made by appealing to other spatial transform-
ations operations: rotating about some axis or reflection about some plane.
3
One might object here on the grounds that one can have singular thoughts about
particular regions of space-time, and thus knowledge, that isn't preserved under trans-
formations. (I can know that I'm located right here, at this particular region of space-rime
for example-I clearly wouldn't know this in a world where I wasn't located right here).
I'm going to set this issue to one side for the time being, but see Maudlin (1993) for more
discussion.
Relative Locations 47
d just as we can consider spatial transformations, we may also consider
Anperations involving time: time reversal (a kind of temporal reflection),
0
poral shifts (moving every event fo1wards or backwards a fixed amount
te(;ime) and boosts (uniform shifts of velocity), and of course, arbitrary
0
mbinations of any of these operations. We shall call a combination of
cohese operations
. a c·,ia1·1 c .
I ean trans1ormatton.
4
t The principle No SHIFTS has been given a number of distinct motivations
ver the years, some less convincing than others. While some have overtly
~heistic or verificationist premises, No SHUTS has remained central to the
debate about the existence of space-time, even though attempts to motivate
it from more general principles have changed. Most contemporary phil-
osophers take the best justification for No SHIFTS to be a defeasible one:
theories that postulate undetectable structure that play no role in explaining
the observable world are not to be preferred over theories that do not
postulate this structure but are otherwise just as simple and explanatoty
(see, for example, Russell 2014 and Pooley 2013). As a silly example, one
could imagine a theory which postulates the existence of an absolute origin
from which every point can be assigned an absolute distance. Although a
particular origin could well make a nominal appearance when the dynamical
laws are formulated in terms of a particular coordinate system, a world in
which an alternative point were the absolute origin would be identical to
our own in all physical respects. This kind of additional structure is
completely undetectable and is also unnecessary to the formulation of the
dynamical laws. A less silly example is Newton's original formulation of
classical mechanics, in which there was a distinguished velocity called
'absolute rest': objects traveling at that velocity counted as being at absolute
rest, and objects moving relative to it had absolute motion. The move from
Newtonian space-time to Galilean space-time, in which this redundant
structure is eliminated, is now universally accepted.
The main problem for this line of reasoning is that unlike the case of the
absolute origin and absolute rest, we don't have a clear alternative theory
in which the redundant structure is eliminated. Thus, while many phil-
osophers take this version of Leibniz's challenge to be a serious one, ultimately
they reject it on the grounds that the most prominent alternative to
4
Note on terminology: sometimes reflections of various sorts are excluded from the
definition of a Galilean transformations-I always intend them to be included in what
follows. Under the operation of composition these transformations form a group called
the Galilean group. Note, however, that I do not include enlargements in the space of
transformations: these are not in fuct symmetries of the laws of Newtonian physics, since
the rate at which two particles of equal mass will accelerate towards one another depends
on the distance they arc separated.
48 Andrew Bacon
substantivalism-relationism-doesn't share the theoretical virtues of the
substantivalist theory (see, for example, Maudlin 2012: 65-6 and Pooley
2013). 5 This strikes me as sufficient reason to explore alternative theories
which don't posit this sort of redundant invisible structure. 6
2. RELATIONISM
The best argument for No SHIITS rests on general considerations of theory
choice, and is thus successful only if there is a simple alternative to the
standard version of substantivalism that respects it. It is often assumed that
relationism is this alternative. Unfortunately relationism faces a somewhat
similar set of issues.
2.1. Mathematical relationism
According to relationism, the motions and geometrical properties of mater-
ial objects are not grounded by the relations they stand in to a background
space-time manifold. Thus the relationist is in need of some other kind of
theory of these sorts of properties and relations. To keep things simple we
shall restrict our attention to a relatively sparse world consisting of three
point particles, arranged in a certain configuration, traveling with certain
momenta as governed by the laws of Newtonian gravitation. Since it is clear
that there is a difference between worlds in which the particles are arranged
in, say, a regular triangular shape, and those in which they are colinear, or
worlds in which the masses of the particles, and thus the forces between
them, are greater or smaller, the relationist owes us an account of what these
differences consist in.
According to the most straightforward versions of relationism, geomet-
rical properties like these consist in fundamental relations holding, in certain
configurations, between the material objects, numbers, and other kinds of
abstract objects. Thus, for example, distances can be represented by a
fundamental three-place relation, Dxyz, taking two particles and a real
5Although there have been several attempts to develop a 'Machian' theory of this sort,
it's unclear how successful they are. See Barbour and Bertotti (1977; 1982). See Pooley
(2013) for an overview.
6 This general sort of project has been attempted by Dasgupta (2009) and Russell
(2014), but both sorts of theories involve contentious metaphysical ideology such as a
many-many grounding relation or a primitive factuality operaror. The approach I will be
recommending here is straightforwardly intelligible to anyone who has the concepts
presupposed in the original relationism/substantivalism debate-specifically, anyone
willing to theorize in terms of the location relation.
Relative Locatiom 49
ber, satisfying the axioms of a metric space. (In mathematical contexts,
1
'.u:U standard to equivalently describe such a relation as a function d(x,y)
1t !S
ping .x- and y to t he d'!stance between them.) Thus three pamcles .
ma~iged in a regular triangle are such that each pair of them is related to
a~:same number. The gravitational force exerted by one particle on another
t n similarly be represented by a relation, Fxyz, taking two particles and a
c:ctor from some suitable normed three-dimensional vector space, and mass
van be represented by a bina1y relation between particles and real numbers
M.xy (equivalently a function/(x,y) mapping pairs of particles.to vectors and
a function m(x) mapping particles to real numbers).7 Given this kind of
background theo1y, one can begin to attempt to reconstruct a relationistically
acceptable version of Newtonian physics: for example, we might subject F to
the restriction that if F holds between x, y, and v it holds between y, x, and -v
according to Newton's third law. 8 Although there is room for much variation
in the details let us call this general kind of approach, in which geometric
and physical properties are grounded in relations between the physical and
the platonic realms, mathematical relationism. (For a prominent theoty of this
sort, which freely employs relations to mathematical objects, see Barbour and
Bertotti 19 8 2.)
The framework just outlined seems to be susceptible to an objection
similar to the one afflicting substantivalism: it seems that one can describe
rwo different configurations of these relations between objects and numbers
that correspond to the same physical scenario. For consider any legal
arrangement of our three-particle system: this will consist in the relations
D,F, and M holding between our particles, numbers, and vectors in some
configuration. In particular, each pair of particles, x and y, will be related by
F to some vector v in an abstract vector space V, representing the force that x
exerts on y. Now consider another configuration which agrees with the
original regarding which things are related via M and D, but in which the
vector related to each pair of particles by F has been uniformly switched, at
every time, for the result of rotating that vector by 90 degrees about some
fixed axis in the abstract vector space it belongs to. Notice that this operation
7
I have set aside the issue of time here. A proper treatment may involve introducing
another argument place to these predicates for a time, or giving them a tense-logical
treatment (the latter seems natural for those relationists, such as Arthur Prior, who reject
time as well as space).
8
One might wonder if one could eliminate forces just by appealing to distances and
masses (or perhaps masses in favor of distances and forces). This is far from clear: for
example, a pair of equally massive particles orbiting one another in a circular motion will
ha~e constant masses, and be at a constant distance from one another, although the force
acting between them is constantly changing, so that forces can't be recovered from mass
and distance alone.
 50 Andrew Bacon
will preserve eve1ything of importance for the modeling offorces-cruciaJ[
it will preserve the inner product which represents the lengths of the vecto~f
and the angles between them. ~i
This operation bears many of the marks of Leibniz's shift argument: the~
operation of rotating the vector that represents the force between any pa· ~.
of particles appears to be a symmetry of our relationist theo1y. Although th~!'.
is not on its own a sufficient reason to reject this theory, it does mean it~
lacks the advertised advantages over substantivalism. We must, however, be;,
careful to distinguish the symmetry just described from a different oper.!
ation. For example, one can imagine a possible world that is not legal by the~
standards of the Newtonian theory of gravitation, in which the force that thef
particle y exerts on x i~ not parallel; ~ut orthogonaJ to the line ~assingl
between them-see the Illegal world diagram below. (Note: each d1agralll'
is supposed to represent three different states all at the same time t.)
Legal world 1 Legal world 2 Illegal world
v y
x ____,.. u y
x ____,..
x y
This is not the operation I am describing: if the force y exerts on x really is
acting perpendicular to the line between them, then at later times x would
accelerate in a direction perpendicular to that line, so that the distance
between x and y ought to increase for a period of time (absent other forces).9
In the original example, however, I stipulated that the distance facts are
the same at both possibilities at all times, so if the original case is legal, the
distance between x and y will decrease for a period in the variant world. The
'shifted' world is therefore a world in which the physical force in fact points
from x toy (since that is the direction that x will move in), but in which its
so pointing consists in x and y standing, via F, to a different abstract vector
(u instead of v, as in 'Legal world l' and 'Legal world 2').
Of course, one could insist that one of these configurations is not rea.lly
a metaphysical possibility. Perhaps, in that configuration, only v but not 11
could possibly represent the force exerted by yon x. Similar moves can aJso
be made in the substantival case: perhaps no two worlds that agree about
the non-locational profile of the material objects can disagree about their
locations in space-time. But such responses do little to assuage the feeling
of arbitrariness-why, for example, is it necessary that when the particles
9
We are assuming here that the world is legal in at least the sense that particlei
accelerate in the direction of the forces acting on them, even though it is illegal in the
sense that the force exerted by a particle does not point towards the particle.
Relative Locations 51
f are air.. anged this particular way, they are always located at this region of
ce-wne.
I
spNote also that the vector representing the force y exerts on x can differ
, en times, even when those times agree about the mass and distance
betwe For cons1'der a system o f two equa IIy massive· pomt · Ies, x an d y,
· parttc
facbt~'. g one another in a circular motion. The distance between the
0 r inn
. Jes and their masses, remain fixed over time. However, the vector
Parncsenting
'
the force between x and y is rotating at a constant rate. 10 Thus,
:ep~ect the states depicted in 'Legal world l' and 'Legal world 2' could
111
r:s:nt the state of the same world at two different times. So the thought
r~t it's impossible for v or impossible for u to represent the force acting
~e~veen x and y when the distance and mass facts are in such a state seems to
le out certain kinds of physically possible scenarios. 11
ru It may strike the reader that the above problem is specific to the
articular form of mathematical relationism I have outlined above, which
;elied on the use of vectors. However, the problem outlined above is but an
instance of a more general problem for relationism outlined by Hartry
Field (1980). For example, Field notes that we can describe the world
equally well using different units. Thus a configuration of M, D, and Fin
which M relates each particle to its mass in kilograms, and the configur-
ation in which M relates each particle to its mass in pounds but is
otherwise the same, correspond to the same physical scenario. The 'shift'
that we are performing in this case is that of multiplication by a factor of a
scalar quantity, rather than a rotation of a vector quantity. 12 There is
something ve1y suspect about a fundamental metaphysics in which one
10 This is part of the reason why this version of relationism can avoid Newton's bucket.
11 One could try to make this argument more rigorous by appeal to the principle that
whatever sometimes happens could happen (see Dorr and Goodman Forthcoming), for in
the two-particle world described, x and y sometimes bear F to v and sometimes to u, but
always bear M and D to fixed numbers. (Although I do not myself subscribe to this
principle, there is something very compelling about the intuition in this case.)
12 In the case of scale dependence there's a fairly straightforward fix: instead of having a
primitive binary relation Mxy relating each object to a number, have a ternary relation
Mxyz relating two physical objects to a number that simply tells us what the ratio of the
mass of x and y is. Whatever units we choose, the ratio of two masses will remain the
same. However, this doesn't really speak to the underlying problem that when abstract
numbers appear in physics there are usually other abstract objects that would do the job
just as well; usually the choice of mathematical objects to use is made based on conveni-
ence. A slightly contrived version of this problem applies even to ratios of physical
quantities-there are many abstract objects that could do the job of ratios equally well.
Ratios are real numbers, and set theoretically we can construct these entities in several
equally natural ways: we can represent them by certain sets of rational numbers, called
Dedekind cuts, or by certain kinds of converging sequences of rational numbers
called Cauchy sequences. Even if real numbers are sui gm eris entities, and are not identical
to either of these constructions out of rational numbers, that just means that we have one
more isomorphic mathematical structure to choose between.
52 Andrew Bacon
has to choose between kilograms and pounds, quite aside from its the fact
that it generates invisible distinctions.
The issue to do with units above appears to be part of a broader set of
problems affiicting theories that postulate relations to mathematical objects
in order to account for the fundamental properties of physical objects. In
general there is a lot of arbitrariness in our choices when we represent the
world using abstracta: not just in the choice of units, but sometimes in more
far-reaching ways. For example, Newtonian mechanics is simple enough to
be formulated in terms of affine spaces-in which pairs of space-time points
are related to objects in a mathematical vector space-but it could also be
formulated in the language of differential geometty in which case each
space-time point is instead related to an equivalence class of real-valued
smooth functions on a neighborhood of that point. Similar points extend to
most uses of abstracta by physicists, whether it concerns the choice of origin
and orientation of a coordinate system, the choice of units, or sometimes the
very choice of mathematical formalism itself. 13
Thus some philosophers, such as Field, have maintained on this basis that
'relations between physical things and numbers are conventional relations
that are derivative from more basic relations that hold among physical things
alone' ( 1984: 46). Field has in mind relations of magnitude in particular and,
although he is presumably at least partly being motivated by his nominalism,
also mentions considerations such as those about units mentioned above.
I am not a nominalist, but I think the idea is compelling nonetheless. Even if
one rejects nominalism, it would be puzzling if the fundamental laws of
motion, for example, depended for their truth on the existence of abstract
objects. One could dramatize this intuition by imagining that all numbers
were to disappear tomorrow: insofar as we can entertain such a hypothesis, it
doesn't seem likely that there would be any serious consequences for the
non-mathematical universe-it's not like the earth would stop orbiting the
sun, or that planes would start falling out of the sky. Gravity, for example,
seems to be a physical force whose existence does not depend on the numbers
we use to represent it: surely concrete things could have moved about in the
way that physics demands even if there hadn't been any numbers.
2.2. The prospects for alternative versions of relationism
It is best, I think, for a relationist to maintain that material objects have their
topological, metrical, and geometrical structure intrimical!y: structure that is
13 For example, some physicists prefer to theorize in the language of category theory,
which often leads to certain kinds of more familiar abstracta being replaced with
categories.
Relative Locatiom 53
. ternal to the domain of material objects and is not merely inherited by their
1
:lacion to a platonic realm of numbers. There are a couple of ways we could
: about this project, depending on whether we tal<e there to be an infinite
0
7iurnber of metaphysically primitive geometrical properties or a finite number:
(1) GEOMETRIC PRIMITIVISM: Take each geometrical property and relation
to be metaphysically primitive. On this picture all geometrical properties
are equally fundamental.
This theoty will have an infinite number of primitives: for each possible
value of a there is a primitive relation between particles of being a meters
apart; for each possible shape, there is a primitive fundamental property
of having that shape; and so on.
(2) GEOMETRIC R.EDUcrIONISM: Attempt to fix the geometrical structure
by a smaller finite set of primitives.
Perhaps to the notion of an object being an open sphere (Tarski 1983), or
to the notion of two pairs of particles being congruent to another, and a
particle being between two others (Hilbert 1899; Tarski 1959). 14
If one thought that our fundamental properties and relations are governed
by simple laws, the first option should strike us a deeply unsatisfactoty.
There are, for example, general geometrical laws relating collections of
particles with certain shapes, and moreover physical laws governing how
those shapes should evolve as the particles are attracted to and repel one
another. In the first kind of theoty we have no hope of writing these sorts of
laws down: to achieve quantification over distances (which we need to do in
order to talk about rates of change, for example) would require one to
employ large infinitary conjunctions and disjunctions. 15
The latter sort of approach to geometty has been developed extensively by
Hilbert and Tarski, and has most prominently been championed in the
philosophical literature by Field (1980) (see also Casati and Varzi 1999,
Maudlin 2010, and Arntzenius and Dorr 2012 for some similar approaches
to geometrical and related structure). To illustrate let us focus on the work
of Hilbert (1899) and Tarski (1959). The usual way to represent distances
between points would be to introduce a metric-a function from pairs of
space-time points to real numbers-telling us how far apart the points are.
14 If we are interested in properties relating to the smoothness of the manifold, the
notion of a converging sequence of points (see Arntzenius and Dorr 2012). Or if we are
interested in topological properties, the notion of two things touching (see Casati and
Varzi 1999) or a closed line (see Maudlin 2010).
15 For a comprehensive discussion of related issues, see Skow (2007); see also the
further discussion in Kleinschmidt (2015).
54 Andrew Bacon
The mark of the Hilbert-Tarski program, by contrast, is that it employs a
small number of geometrical properties and relations whose relata are
concreta. These geometrical properties can be understood in both a sub.
stantivalist and a relationist setting by interpreting the primitives as applying
to either regions of space-time or material objects respectively. In Tarski's
geometry of solids, for example, we have the following primitives:
(1) A binary relation x :Sy whose intended interpretation says that xis a
mereological part of y.
(2) A unary predicate, Sx, stating that xis an open sphere. 16
Hilbert's axiomatization, also later refined by Tarski, instead only invokes a
pair of relations whose arguments are point-like objects:
(1) A three-place betweenness relation, Bxyz, whose intended interpret-
ation states that x lies on the straight line segment between y and z.
(2) A four-place spatial congruence relation, Cxyzw, whose intended
interpretation states that the distance between x and y is the same as
that between z and w. (For short: xy is congruent to zw.)
It should be noted that in this setting the quantifiers are understood as
ranging over point-like objects (space-time points or point-like material
objects); if one wanted to make that restriction explicit one could introduce
another primitive applying to mereological atoms, or one could simply
stipulatively understand congruence so that it applies to no complex objects
and define atomicity as standing in congruence relations to some things.
The general approach is not limited to quantities representing distances
either. Field has shown that the latter sort of theory can be extended to other
quantities. If I want to talk about the numerical value of a field at a given
point-for example, the mass density field or the gravitational potential-I
can employ a similar trick. One can introduce a congruence relation stating
that the difference in gravitational potential, for example, at x and y is the
same as the difference of the gravitational potential at z and w, and a
betweenness relation saying the potential at x is between the potential at y
and z respectively (see Field 1980). 17
16
This is what modern mathematicians would call an open ball; the word 'sphere' is
now typically reserved for the two dimensional surface of a ball; however, this is not what
Tarski meant by a sphere.
17
In order for this to work there must be a rich enough variety of field comparisons for
us to be able reconstruct the field values (up to a scale) from the comparisons. This is
guaranteed if we make the assumption, typical in physics, that physical fields are always
continuous, so that a field is either constant everywhere or inhabits an open interval of the
space of possible field values.
Relative Locations 55
One might hope that a relationist could employ a small set of geometrical
. itives such as those above, and attempt to recover geometric structure
prtm '
chat way.
Unfortunately, as Hartry Field has noted in (Field 1984), some of the
b ve theories are simply not available to the relationist. In particular
ah; Hilbert-style theo1y, employing congruence and betweenness, fixes the
t [evant geometric structure only if there is a sufficiently large number of
r:ometric objects hanging around. This can be illustrated with a simple
~xample: consider a situation in which t~e dis~ance between x and y is two
imes the distance between z and w. While this may on the surface look as
:bough we have a relation between four points and the number two, in the
Hilbertian setting reference to the number two can be eliminated. This
relation can be stated instead as follows:
There is some point u between x and y such that both xu and uy are
congruent to zw.
If we wanted to say the distance between x and y was three times the
distance between zw we'd say there was a u 1 and u2 such that xu1, u1 u2,
and u2f were each congruent to zw. One can see, without much trouble,
how to paraphrase away talk of arbitra1y rational ratios of distances in this
fashion. Note that once you have pinned down the rational distances
between points, all remaining distances between points are fixed. So,
given the existence of enough point-like objects, betweenness and congru-
ence facts are enough to pin down all metric structure: no two worlds can
agree about the betweenness and congruence facts and disagree about the
geometrical facts.
The above statement captures the notion of a pair of particles being
twice as far apart as another pair perfectly well in a substantivalist setting,
since according to that theory, for any two space-time points there is
another space-time point between them. However, Field notes that this
is not so for point-sized material objects: if I take the closest point to the
earth on the edge of the moon, p, and take the closest point to the moon
on the edge of the earth, q (suppose for a moment that both have definite
boundaries) then, at least by the relationist' s lights, there are no entities
between p and q, not even a space-time point. Thus, even ifI were twice as
far away from the earth as the moon is, I wouldn't count as such by the
lights of Hilbert's analysis given a relationist ontology. Indeed, it is easy to
see that there are distinct arrangements of me, the moon, and the earth,
that agree about all betweenness and congruence facts but are nonetheless
very different geometrically.
The problem Field has identified here seems to be much more general.
For example, Tarski defines the topological notion of two spheres touching
56 Andrew Bacon
(i.e. touching but not overlapping) in terms of the notion of a sphere as
follows:
x touches y if and only if, x is disjoint from y, and any two spheres
containing x and disjoint from y are such that one is part of the other.
Neither this definition nor any other will work in a relationist setting.
Imagine two worlds containing two perfectly spherical balls, x and y, in an
otherwise empty space. Suppose also that in the first they are touching, and
in the second they are not. Both worlds agree with one another concerning
which things are spheres and which things are parts of what, but they disagree
about which things are touching. Thus, given a relationist ontology, the
touching facts are not fixed by the sphere and parthood facts (in particular,
because the only sphere that contains xis x itself, according to the relationist
ontology, x and y count as 'touching' in both worlds by Tarski' s definition).
It follows that in order to pin down the geometrical structure the
relationist needs to introduce twice the distance and touching as new primi-
tives. It is natural to wonder whether this can be done by adding only finitely
many primitives; thus evading the undesirable aspects of PRIMITIVISM.
To address this question let us focus on a simple example world consisting
of only three point particles, arranged in some shape. There are an uncount-
able infinity of arrangements like this that differ regarding the ratios of the
distances between the three particles.
In this setting the primitivist strategy involves introducing, by brute
force, an uncountable infinity of primitive ternary relations, so that for
each possible arrangement a of the particles x,y,z there is a fundamental
relation Ra that holds between x,y,z (in any order) iff those particles are in
that arrangement.
It is natural to wonder whether we can do better. Can we specify a theo1y
of three point particles with finitely many primitives without expanding our
ontology to include space-time points or numbers? In order to answer this
question we need to state it a bit more precisely. An arrangement of three
particles can be represented by the three particles x,y,z and the ratios of
the distances between any pair of them. Formally, the set of possible
arrangements of three particles, A, is the set of metric spaces (M, d) such
that M={x,y,z} quotiented out by scale. 18 Someone hoping to write
down a theory capturing the geometry of three point particles must choose
a language £-given by specifying some set of non-logical primitives-and
present a theory which can be either specified axiomatically, or by a class of
18 That is to say, an arrangement is an equivalence class of such metrics where (lvf, d)
and (M', d') are equivalent iff d = a.d' for some a E !R.
Relative Locations 57
intended models, C. For example, in the Hilbert-style theory the primitives
four language are the congruence and betweenness relations, three con-
~tants denoting the three particles, and a location relation for stating the
location of each particle. The class of intended models consists of three-
dirnensional Euclidean spaces with the three particles x,y, and z located at
three points of that space. Each model of the theory ought to correspond to
some <trrangement of the particles: there ought to be a surjective function
Arr : C---> A. In the Hilbert-style theory Arr is easy to specify-each model
easily determines an arrangement because the underlying metric of the
Euclidean space tells us what the distances between the three distinguished
points are.
The Hilbert-style theory has a nice feature. For any two models Mand M'
in Cof the theo1y in which the arrangement of x,y, and z differ (i.e. Arr(M) ::/=
Arr(MJ, one can find a sentence in the language of congruence and between-
ness, rp, such that M I= r/J and M' Jr: r/J. 19 In this way any two arrangements
can be distinguished by some sentence of the language.
The primitivist relationist theo1y also satisfies this constraint. The language
of this theo1y has a primitive relation Raxyz for each possible arrangement
a E A. The models of this theory have a minimal relationist ontology: the
domain of each model contains only the three particles x, y, z. Moreover, for
any arrangement a E A there's some model in which the objects are arranged
that way: a model in which R" applies to x,y, and z in any order, but in which
Rb doesn't apply, in any order, for any b distinct from a. Each model is
therefore associated with a unique arrangement Arr(M) =a where a is the
arrangement such that the extension of R" is non-empty in M. As with the
Hilbert theory, whenever Mand M' correspond to different arrangements-
i.e. when Arr(M) ::/= Arr M' -there is a sentence that is true in M but not
M', namely Raxyz where a=Arr(M).
The principle we have appealed to in each case is the following principle.
Suppose that £ is a theory with a class of intended models C, that includes
models representing each possible arrangement of the three particles (that is
to say, the function Arr associating each model with an arrangement is
surjective). Then if the primitives of£ express a physically complete set of
19
Proof: let the distances between x andy in a model M be denoted dM(x,y). If the
arrangement of particles in M and M' differ then for some a, b, c, d E { x, y, z} the ratio
~M(a, b) / di11 (c, d) and d1>r (a, b) / di'>f'(c, d) differ (if all these ratios agreed, they would be
m the same arrangement). That means there are a pair of rational numbers, q and q' such
that q<d,11(a, b)/dM(c, d) <q' but doesn't hold when Mis substituted for M'. As we
indicated earlier, it is quite easy to express facts about rational distance ratios using a single
sentence.
58 Andrew Bacon
fundamental relations, any two models representing different arrangements
of the particles ought to be distinguishable by some sentence of£:
DISTINGUISHABILITY: If Mand 1'11 correspond to different arrangements of
the particles (i.e. Arr(M) =F Arr(M')) then there is some closed sentence
¢ E £., such that M I= ¢ and M' JC ¢.
This constraint is quite important and is effectively a way of saying, in mode[
theoretic terms, that the arrangement of the particles supervenes on the facts
expressible in £.,. Without it one could not formulate an adequate physical
theory: the forces particles exert on each other, for example, depend on the
arrangement of those particles. In particular, particles arranged differently
can behave differently. If the fundamental primitives do not distinguish
between two possible arrangements of the particles then the behavior of the
particles will not be determined by kinds of facts we are taking to be
fundamental. Neither the arrangements nor the motions of the particles
will supervene on the distribution of the fundamental properties and rela-
tions; particles could be arranged differently in two worlds even when the
two worlds agree about all the fundamental facts as stated in our fundamen-
tal language, £.,,
We can now prove the following !imitative theorem.
Theorem 2. Let[, be a first or second order language, C a collection of models ofL, and
Arr : C-> A a sU1jective association ofarrangements to models. Suppose that our class of
models also satisfies the distinguishability constraint. Then one of the following is true:
(I) [, has infinitely many non-logical primitives.
(2) C contains an infinite model.
This theorem is not particularly deep from a mathematical perspective. 20
However, it does clarify the situation: the two most salient responses to the
theorem are (i) to adopt infinitely many primitives, as with PRIMITIVISM, or
(ii) to adopt an infinite ontology, as with the substantivalist theory or the
theory that postulates mathematical objects to account for the relations
between the three particles. Note, however, that our theorem is quite
general-we have allowed, for example, that which 'extra' objects exist can
20 If£ has a finite signature, then for each finite cardinality n, there are only finitely
many different isomorphism classes of models of £ of that cardinality (for an k place
relation there are only 2"' possible relations over a domain of cardinality n, so there are
21" ' + ··· + 11"' .n1 possible models in a language with I constants and predicates of arity
k1, ... , k,,,). It follows that if each model in C is finite then there are at most countably
many isomorphism classes of models of M. But since there are uncountably many
arrangements, Arr must map two models in the same isomorphism class to the same
arrangement, contradicting D1;mNGVISHABILITY.
Relative Locations 59
d end on the arrangement of the particles. Perhaps if the particles are
~~ly and symmetrically arranged one can characterize that arrangement
nt·~h a finite set of primitives without postulating extra objects. What the
~eorem shows is that whatever your theo1y, if it has finitely many non-
~ ical primitives, there will always be some arrangements for which one
ogds infinitely many objects present to distinguish those arrangement from
nee
ther such arrangements.
0
Note that one could also question the constraints placed on the kind of
language employed in our theorem: we allowed ourselves first-order and
second-order resources, but one might think it possible to do better with
modal resources. A particularly natural strategy would be to combine modal
resources with second-order resources. With a modest finitary principle of
recombination for possible particles, guaranteeing that the outer domain of
quantification is infinite, one could force the domain of the second-order
quantifiers to range over an infinite collection of intensions even whilst
keeping the inner domain of the first order quantifiers finite at each world.
This could in principle allow one to simulate the kinds of things one would
normally do with space-time or mathematical objects, whilst technically
keeping our first-order ontology finite. 21 Even if one grants that second-
order quantification does not in itself carry a commitment to abstracta (as
I am inclined to think myself), such proposals are not without their own
difficulties, although it is beyond the scope of this paper to survey them in full
(but see the discussion in section 2 (on using properties to simulate space-
time) and (on modal approaches) in Field 1984, section 9). (See also Mundy
1987 and Eddon 2013 for property theoretic account of quantities, and Belot
2011 for a fairly sophisticated example of a modal treatment of geometry.) 22
21 Here is an example of this kind of strategy, where we employ modal operators and
higher-order quantification into sentence position: one could adopt a primitive A(x, y, P)
taking two terms and a sentence, roughly meaning P is an arrangement of x and y: in a
possible worlds style model, for some distance d, P is the set of worlds at which the
distance between x and y is d. From this one can define what it means for P to describe
the arrangement of any finite collection of particles, and helping oneself to propositional
quantification and a necessity operator, one can recover the congruence and betweenness
relations. Thanks to Jeremy Goodman for discussion here.
22
According to the Mundy-Eddon view, there are infinitely many properties corres-
ponding to each quantity (e.g. one property for each possible mass) and primitive higher-
order relations between these properties that determine their quantitative structure. It is
natural to view the Mundy-Eddon view as falling under the 'infinite ontology' branch of
our dilemma, as particles are inheriting their quantitative structure by standing in
relations to an infinite collection of mass properties. However, one could also formalize
it in third-order logic with primitive third-order relations over second-order entities, in
which case the theory is not within the remit of our theorem, and the first-order ontology
could well be finite (even though the higher-order domains are infinite).
60 Andrew Bacon
One could respond to these arguments not by introducing relations to
space-time points or numbers, but by introducing some other infinite
physical structure that, unlike space-time, is invariant under Leibnizian
symmetries. Although not relationism as traditionally conceived, one
might hope to find an alternative to space-time from which distances and
other quantities can be recovered without relinquishing No SHIFTS.
There are, of course, lots of ways of going about this strategy, and the
project is too broad to say anything too conclusive about its prospects.
A particularly natural way to go about this is to take a group of displacements
of the universe of material objects as a primitive physical entity in its own
right. According to this theoty, particles do not get locations in this space,
but rather pairs of particles get assigned 'locations'-i.e. vectors in our space
of parallel displacements representing the displacement between them.23
Crucially, this 'location' remains the same under Leibniz shifts. 24
Note, however, that these projects all involve accepting a form of sub-
stantivalism, even if not the familiar sort. Rather than evaluating the
prospects for theories like this-a worthwhile project for another time-I
want to focus on the possibility of carrying out this sort of idea within the
confines of orthodox substantivalism.
3. MULTIPLE LOCATION
These sorts of considerations strike me as a good prima fade reason to be a
substantivalist: by appealing to relations between material objects and space-
time points we can recover enough structure to represent the distances and
other quantitative properties and relations, without placing an undue sig-
nificance on any particular mathematical representation of these facts.
23
Nore that the order of the pair matters: if vis the displacement between p and q, -v
is the displacement between q and p.
24
For simplicity we might focus on a physical structure that is invariant under
translations in space-the group of parallel displacements-however, mathematically
it is straightforward ro extend this idea to rotations using the notion of an angular
displacement, and boosts and other continuous symmetries using other similar notions.
A parallel displacement is just a vector quantity that points from one point of space-time
to another, telling us the displacement between the two points; this quantity is invariant
under translations in the sense that if one pair of particles is a translation of another the
displacement vector between the first pair is the same as the second pair. The group of
parallel displacements of Galilean space-time is itself a four-dimensional manifold, just
like Galilean space-time, but it has more structure: it is a normed vector space, and thus
unlike Galilean space, has a special point that is distinguished from the others (the 0
vector). The manifold is rich enough that the mathematical structure can be recovered in a
Hilbert-Tarski-Field-style setting.
Relative Locations 61
A. relationist, on the other hand, must either take the role of abstracta in
h sics more seriously than seems wise or attempt to emulate substantival-
f ~using certain kinds of higher-order and modal resources.
isr for someone following this line of reasoning, it would not be unreason-
ble to take these considerations to also reflect negatively on the No SHIFTS
a rinciple. After all, substan tivalism is often taken to entail that there are
~ifferences between shifted worlds. Note, however, that the role substantival-
isrn plays in fixing the geometrical structure of material objects is quite
different from the role that regions of space-time play in the shift argument,
which make some specific assumptions about the location relation.
To address the issue of geometric structure one postulates a manifold with its
own intrinsic geometric structure. Geometric relations between material objects
then exist only in a derivative sense: in virtue of material objects standing in some
relation-call this the 'location relation'-to space-time regions that have the
geometric properties in question. Nothing we have said so far, however, rules out
the possibility that the location-like relations material objects bear to space-time
regions are invariant under Leibnizian transformations.
Note that as I have introduced it above, the location relation is defined
implicitly by its role as that relation between objects and space-time fi'om which
the geometrical properties of objects can be recovered. As such, one shouldn't
assume that it corresponds to the pretheoretic of notion of location we are
used to employing when we are not doing fundamental metaphysics. The
ordinary notion of location usually relates us to places-like Paris, the
moon, and so on-and not regions of space-time. 25 With this caveat in
mind, we can ask what it would mean for the world to remain unchanged by
a uniform shift of the locations of each material object. Ignoring time for a
moment, and restricting ourselves to a three-dimensional universe it would
require the following:
SHIFT INVARIANCE: If xis located at y and y' is a Euclidean transform of y
(a combination of shifts, rotations, and reflections) then x is located at y'.
Later we will see that SHIFT INVARIANCE can be formulated vety simply
without appeal to Euclidean transforms (see section 3.4).
SHIFT INVARIANCE arguably follows from a certain conception of the
location relation. As mentioned above, we are introducing the location
relation by the job it plays, regardless of the distance from our pretheoretic
notion. Assuming that an object's geometric properties are entirely deter-
mined by its shape, the job to be satisfied can be summarized by the idea
25
More importantly, objects are usually thought to be located at no more than one
place on this conception.
 62 Andrew Bacon
that the shape of an object is given by the shape of its location. 26 Thus the
location relation must satisfy the constraint that if it holds between a
material object 0 and a region R, then 0 has the shape of R. The simplest
relation satisfying this constraint is the relation '0 has the same shape as R',
and on this interpretation SHIFT INVARIANCE is true. 27 (Of course, there are
also one-one relations that satisfy the job, relating each object to a unique
region: but there are infinitely many of them that are equally good, and if
the job description is our only criteria for choosing, to pick one of thern
seems arbitrary.)
Of course SHIFT INVARIANCE requires material objects to be multiply
located in a fairly radical way. The crucial point is that if we are at a world
in which objects are multiply located in this sort of way, then uniformly
subjecting every object to a Euclidean transformation will leave each object
with the locations it had before the transformation. Thus we do get to
maintain the principle No SHIFTS.
This idea generalizes to other kinds of transformations. For example, if
you thought that embiggenings don't generate genuine differences then
perhaps a world in which the displacement between some particular point,
p, and the location of each particle had been doubled (an embiggening
around p) also results in a state of affairs no different from the one you
started out with. (In this case, however, it is less clear that the transformation
is a symmetty of the underlying physics. 28 ) In the special theory of relativity,
the relevant transformations are Lorentz transformations. And in the con-
text of general relativity, the transformations that seem natural are diffeo-
morphisms. In this latter setting the metric structure of the manifold seems
like a contingent and changeable feature of the world, and thus transform-
ations which preserve the differential structure but not the metric properties
seem like the natural transformations to use. (Although, perhaps surpris-
ingly, this means that properties like shape are not independent features of
26
More generally, we can read off geometrical relations between more than one object
from the shape of their fusion.
27
Note that although one can identify the location relation with the relation of having
the same shape, that needn't be the order of reduction: in particular, it's consistent to
assume, as we have been, that space-time has its geometric structure intrinsically and that
the shape of a material is given by its location(s).
28
In a world in which the distances between two point particles of mass m has been
doubled, they will accelerate towards each other at different rates, because the forces
between them are inversely proportional to the square of the distance between them.
(Note that talk of'doubling the distance between two points' must be treated with some
care. One can of course change between units in a way that doesn't change the physics,
but if we keep the scale the same and move each particle so that the distance between pairs
of particles is doubled we won't in general keep the physics the same, unless we also
change other properties like their masses to compensate.)
Relative Locations 63
bjects location in GR, and only emerge in the presence ofa gravitational
~l~). Thus although classical physics, with its Galilean symmetry group,
e been our focus the fimdamental idea can be generalized ro other theories
has
in natural ways. . h f, ll , 'd . 1 c . f'
SHIFT INVARIANCE 1s not enoug : or a we ve sat , a smg e rns1on o
rtides could be located at a table-shaped region of space-time and a chair-
p~aped region of space-time, so long as it is located at eve1y region of space-
~irne that has that table or chair shape. To rule this out we want to require
that any two locations of an object are Euclidean transforms of one another:
SHIH EQUIVALENCE: If x is located at y and x is located at y' then there's a
Euclidean transformation taking y toy'.
SHWI' EQUIVALENCE is required if we are to cany out the project described as
GEOMETRIC RrnucnoNISM: if the shape of a material object, for example, is
determined by the shapes of its locations, but SHIFT EQUIVALENCE failed,
objects simply wouldn't have well-defined shapes. 29 If we want to satisfy No
SHIFI'S and GEOMETRIC RrnucnoNISM at once, we had therefore better
accept SHIFT INVARIANCE and SHIFT EQUIVALENCE. (Note, on the other
hand, that we could in principle attempt to take geometric properties as
primitive properties of material objects and regions, and attempt to reduce
the location relation to them: an object is located at a region if and only if
they both have the same shape. This would, of course, guarantee SHIFI'
INVARIANCE and SH!l'T EQUIVALENCE, but would require us to be a geometric
primitivist. 30)
Finally, it is natural to require that the locations of mereological simples
be themselves mereological simples (space-time points):
SIMPLE LOCATIONS: If xis mereologically simple then its locations are too.
This rules out extended simples.
With these three principles in place we are in a position to see, at least in
outline, that the present view is in as good a position as the relationist is. For
the relationist the arrangement of a collection of particles is given by the
distances between each pair of particles. SIMPLE LOCATIONS tells us that the
locations of a mereological simple are themselves mereologically simple; they
are space-time points. Given a natural principle governing the interaction
of parthood and location, to be discussed in the section 3.3, it follows that
the location of the fusion of two mereological atoms is the fusion of two
29
Note that if we weakened Euclidean transformations to diffeomorphisms, then
objects won't have well-defined shapes. This is perhaps not unexpected given the conse-
quences of general relativity for the naive notion of shape.
30
Thanks to Jeff Russell and Shieva Kleinschmidt for discussion here.
64 Andrew Bctcon
space-time points. (For convenience we shall call the fosion of two mereologicaj
atoms a 'diatom'.) By SHIH EQUIVALENCE we know that all the locations of a
diatom must be congruent to one another, and thus all consist of a fosion of
two space-time points that are the same distance from one another. Thus the
distance between two mereological atoms is always uniquely determined froni
the locations of the diatom they fose, so that all distance facts between point
particles (and thus the geometric properties of their fosions) can be recovered
from their relation to the space-time manifold.
3.1. Locations in ordinary language
Now one might wonder how the radical proliferance of locations posited by
this kind of view is consistent with our experiences. After all, ifl am looking
at the General Sherman I see a solitary tree. If I were looking at a bilocated
tree-a single tree with two exact locations-one expects to see two tree-like
shapes belonging to the same tree. This is at least how philosophers rypically
describe paradigm cases involving multiple-location. By this reasoning, if
the tree were multiply located at each of its Euclidean transforms we should
expect to see a tree smeared out over all of space (if that is even possible to
visualize), but this is emphatically not what we see.
One crucial difference between the bilocated tree and the present case is
that in the former the observer is not herself bilocated. Indeed, in the
former case the distance between the observer and the tree is not obviously
well-defined: there is the distance between the observer's location and the
tree's first location, and the distance between the observer's location and
the tree's second location. In such a case SHIFT EQUIVALENCE could fail,
since the locations of the fusion of the observer and the tree might not be
congruent to one another if the observer is not symmetrically positioned
between the two locations of the tree (this result can be demonstrated more
rigorously in the theo1y of part and location to be developed in section 3.2).
As we have just seen, however, in the present setting in which we have
S111n EQUIVALENCE, the distance between the observer's eyes and the tree is
completely well-defined: it simply falls out of a property that the locations of
the eye-tree fusion share-roughly, being a disconnected eye-tree-shaped
region whose connected parts are separated by a certain distance. Of course
it would be a tall order to give a complete theory of perception in more
fundamental terms, but surely whatever the correct theory is, the position of
the tree in our visual field will depend only on the relative distances between
our eyes, the trees, and other background objects, and will not depend on
which particular regions of space-time the tree is located at. (One way to
convince oneself that there is no conflict with experience, perhaps, is to
Relative Locations 65
al to the fact discussed in section 2.2. The pattern of locations on this
a~pe re determines all the facts about relative distances that the relationist
. has the means to account f'or perceive
P1ctu Thus the present view · d re Iat1ve
·
nee ds. . .
d' ances if the relat10mst does.)
is~imilar puzzles can be warded off in analogous ways. One might think
h tit is possible to uniquely specify the General Sherman's location simply
~ a esturing towards a particular location. I could point at the General
she;man and say something like 'the General Sherm~n is located at that
e ion of space-time over there, and not anywhere else . But on the present
r g I haven ' t rea11y succeeded 111
view · s111g
· 1·111g out a umque
· · o f space-time;
reg10n ·
since my hand is multiply located the gesture I made is simultaneously
related to every region of space-time with that shape.
All of this admittedly sounds a little wild at first. To put it in perspective,
it might be worth recalling our opening remarks in which we distinguished
between two kinds of motivations for substantivalism. The reasons I have
been discussing so far have been rather theoretical-the simplest way to
formulate a physical theory of distances and other physical quantities
without giving particular mathematical objects undue physical significance
appeals to space-time points. However, one might have much more direct
motivations for being a substantivalist: perhaps you think that regions of
space-time play a more explicit role in our lives than I have been acknow-
ledging. Perhaps when I wonder where my keys are I am implicitly won-
dering which region my keys occupy, for example, or when I learn where
Jones went on holiday I learn something about a particular region. On this
picture facts about particular regions of space-time can be revealed simply by
observing the objects that occupy those locations, and the theory that best
explains our observations is the orthodox theory in which every object has
no more than one location.
I suspect that locations in the everyday sense exist and are indeed the
subject of our ordinary talk of'locations'. But I also suspect that locations in
this sense aren't fundamental entities. Perhaps they are places: 'my pocket'
and 'Paris' both refer to places, and seem like reasonable answers to the
question of where my keys are and where Jones went on holiday, respect-
ively. Countries and pockets, like other material objects, will be multiply
located in the more fundamental sense. Or perhaps places in the colloquial
sense are material objects, but are ontologically 'lightweight' objects like
holes and shadows-certain kinds of non-fundamental entities whose exist-
ence supervenes on the properties of other more fundamental objects. (Holes
and shadows, then, are also multiply located in the fundamental sense.)
Relationists are often perfectly happy to engage in talk of places and
locations in the way understood above. In this way the kind of substantival-
ist I have been describing can also accept this sort of non-fundamental
 66 Andrew Bacon
talk of locations. But like the relationist, I deny that this way of talking is a.
perspicuous description of the fondamental metaphysics. Indeed, even the
ordinary substantivalist should think twice before attempting to identify ordin.
ary locations, like my pocket and Paris, with regions of space-time, since the
former have much more interesting modal profiles than the latter.31
To a substantivalist motivated by the more direct sort of reasons men.
tioned above, the view that people are multiply located in this radica.1
mann~r might se~m particularly bad. But the fact is that the methodology
of taking our ordmary use of language at face value is a notoriously bad way
to do fundamental metaphysics. One can see the present view as what one
gets from taking a relationist picture and then expanding the ontology with
a richly structured physical entiry that plays the role in a theo1y of quantities
that a mathematical object would otherwise have played. Counterintuitive
results arise when one attempts to identify that physical entiry with the non.
fundamental way of talking about places and locations; but I think this kind
of identification is ill-advised in the first place.
3.2. The combined theory of part and location
According to orthodox substantivalism there is a fairly simple way to
determine the location of a complex object. If you know the locations
of its parts-in particular, if you know the locations of all its atomic
parts-then the location of the whole is just the fusion of those parts. In
short: the locations of the atoms determine the location of the whole.
This is consistent with the broader Humean thesis that the properties of
and relations between point-sized objects determine the properties of and
relations between all objects. The fact alluded to, however, does not come
for free-it is delivered by a plausible theo1y governing the interaction of
parthood and location. One can concoct formal models in which the
location of a whole is not determined by its atomic parts and which thus
violate this theory. A pair of point particles, both located at space-time
points, has a fusion that is located at the fusion of those two points.
However, it is simple enough to construct formal examples where the
fusion is located elsewhere: on one not so far-fetched model the fusion
might be located at a larger extended region containing the original two
points-on this picture ordinary objects like tables and chairs can be
located at extended regions (with non-zero volume) even if they are finite
31
Of course some people do make these identifications-see footnote 2-but they are
not completely pain free (see e.g. Sider 2001: §4.8).
Relative Locations 67
. ns of zero-volume point particles. 32 But there are models that deny
fustoconnection between the locanons · · parts an d t he fius10n:
o f tI1e atomic ·
~~y]ogically consistent that the location of the fusion of two particles be
its, region whatsoever, whether containing the locations of the individual
an} tides or not. However, insofar as these kinds of situations are deemed
pa:hological they should be ruled out by our theory of part and location.
pa!he situation for the multiple location theory is not so simple. It seems
clea r that whatever kind of theory one adopts connecting parthood and
I cation it cannot be true that the locations of a complex object are
;etermined by the locations of its atomic parts. To see this, note that by
SHIH INVARIANCE, SHIFT EQUIVALENCE, and SIMPLE LOCATIONS, a mereo-
logical atom has an exact location at each space-time point and is located
only at space-time points. It follows that any two mereological atoms have
exactly the same locations. Yet a complex object can have much more
interesting locations. Consider two different objects each composed of three
atomic parts: one could have locations that are all equilateral triangles, whereas
the other might have locations that are all the shape of some particular
irregular triangle. These two composite objects have different locations, yet
as we have seen their atomic parts have exactly the same locations. In general,
knowing the locations of the particles is not enough to determine the locations
of the things they compose.
Indeed it might at first seem that on this picture one must completely
relinquish the principle that the location of a complex object is determined
from the locations of its smaller parts. As it turns out things aren't this bad:
the location of a whole is determined by the locations of its diatomic parts
(objects with exactly two proper parts)-but in order to see this we must
develop the theory of parthood and location a little more. 33
To that end let us begin with the pure theory of location and part: the
theory one gets by looking only at the relation between the parthood and
location relations, ignoring any geometrical constraints such as SHIFT
EQUIVALENCE and SHIH INVARIANCE. Since, to my knowledge, no one has
given a thorough analysis of the relation between locations and parts among
multiply located objects we shall need to develop a little bit of theory.
-" The fosion of the locations of the particles composing a table is an extremely
disconnected object-the gaps between the particles is significantly greater than the
sizes of the particles-so this intuition holds even if we do not assume that particles are
point-sized. On this view the table itself is a solid, connected object merely having a
location that contains the fusion of locations of the particles (see Fine 2003).
33
In addition to the above, there are also quite general problems surrounding the
interaction of mereology and location when multiple location is permitted (see
Kleinschmidt 2011 ). The following is an attempt to formulate a simple logic of mereology
and location that is not subject to these sorts of worries.
68 Andrew Bacon
A natural constraint to impose on our theory is that it should have what
I'll call the classical theory ofpart and location as a special case: it should entail
that theory given the assumption that evety object has a unique location.
The classical theoty effectively says that the location relation determines an
isomorphism between the parts of a thing and the parts of its location (this
theory is sometimes called 'mereological harmony'; see Uzquiano 2011 and
Saucedo 2011). For simpliciry I shall assume classical extensional mere-
ology. I shall write x + y to denote the fusion of x and y and x \y to denote
the relative complement ofy from x, which exists whenever x is not a part of
y (formally, it is the fusion of x's parts disjoint fromy).
(1) The domain of objects that have locations is closed under fusions and
parthood.
(2) The location relation, L, is functional (on the domain of objects that
have locations). We will write l(x), to denote the unique y such that
Lxy.
(3) I preserves atoms: if a is mereologically atomic then l(a) is mereologi-
cally atomic.
(4) I preserves relative complements l(x\y) = l(x) \f (y ).
(5) l preserves fusions: for any x and y, l(x+y)=l(x)+l(y). (More
generally, the location of the fusion of some things is the fusion of the
locations of those things.)3 4
Regarding the theory of parthood as it applies to both material objects and
regions, the classical theory consists of classical extensional mereology. On
the intended interpretation both regions and material objects individually
and jointly form a complete Boolean algebra with the bottom element
removed.
The classical theory rules out many of the interesting possibilities that
have traditionally preoccupied philosophers interested in location. (5), for
example, says that an object is located at the sum of its parts locations; thus
ruling out the pathological example we opened with. (1) ensures that fusions
and parts of located objects have locations, (2) rules out the possibility of
multiple location, (3) rules out extended simples and (4) effectively rules out
colocation and partial colocation (when mereologically disjoint objects are
located at overlapping regions), and (3), (4), and (5) together rule out
unextended complexes.
34 The parenthetical part of (5) is stated using plural quantification. If a first order
axiomatization of this theory is sought, one should replace this axiom with a schema,
saying: if x fuses the </Js then x's location is the fusion of the locations of the </Js. Note that
without the parenthetical these principles do not entail that I preserves arbitrary fusions.
Relative Locations 69
These principles jointly entail that the location function preserves other
. portant mereological properties: for example that l preserves parthood
imd disjointness, and that no two things can have the same location. 35
an Although many of the possibilities suggested above are worthy of philo-
ophical attention, our purpose at present is to explore the consequences of
~ultiple location, and the best way to do this is to screen off these other
kinds of non-standard behavior. Thus the theory we will consider is the
minimal generalization of the classical theory that allows for multiple
location.
If f is a function that satisfies these conditions we shall call it a location
function. The intended models of the classical theory oflocation and part are
therefore models in which the location relation is a location function. (More
formally the intended models of the classical theory of location and part
consist of a tuple (D, :::; , o, l) where (D, :;) is a standard model of classical
extensional mereology, o is an element of D whose improper parts represent
the objects that have locations, and l is a location function whose domain
consists of the improper parts of o and whose range is disjoint from o.) In
order to develop the multiple location theory further, we need to start by
specifying the intended models of that theory.
By way of motivating our choice of model, let us start with a simple
example. Consider a simple Lego construction consisting of three Lego
bricks, Bl-B3, that have been put together in a pyramid shape, with two
at the bottom and one brick on top holding them together. Now suppose
further that this Lego pyramid is bilocated. By examining our intuitions
about this simple case we shall attempt to tease out some general principles
connecting the location and parthood relations.
To keep our intuitions clean we shall make a few simplifying assump-
tions. (i) The three bricks themselves might be composed of smaller parts-
indeed assuming they are not extended simples they must-but for simplicity
we shall pretend that these three bricks and their fusions are the only parts
of the pyramid. (ii) We shall assume that the two locations of the Lego
pyramid are both congruent to one another (in accordance with SHIFT
EQurvALENCE). This geometric constraint is not required by the pure theory
35
(5) entails that I preserves parthood: if x is a part of y then x + y = y and thus
l(x) + l(y) l(y) by (5), which means that l(x) is a part of l(y). Without this principle
one could be located in one ciry even when one's arms, legs, torso, and head are located
in another. By contrast (4) ensures that l preserves disjointness: if xis disjoint for y then
x\y=x and so l(x)\l(y)=l(x) which means that l(x) is disjoint from l(y). It also
ensures that no two things can have the same location, for if x is distinct from y
then either x\y or y\x exists, but if l(x)=l(y) then neither l(x)\l(y) nor !(y)\f(x)
would exist.
70 Andrew Bacon
of part and location-the different locations of the pyramid could in
principle have been any other shape or size (although it would then becorne
unclear what grounds we have for calling it a Lego pyramid, if some of its
locations did not have that shape). (iii) We shall assume that the bricks Bl-
B3 that compose the pyramid also have two locations each, and that these
two locations are congruent. (iv) We shall also assume the three bricks are
arranged symmetrically at each location-one could in principle have had
B 1 be the 'top' brick relative to one location and B2 on top at the other.
Again this is completely consistent with the pure theory of part and location
we are exploring in this section. (v) Lastly we shall assume for simplicity that
the two locations of the whole pyramid do not overlap. To make things
vivid we'll suppose one of the pyramid's locations is entirely contained
within a box, and the other within a jar. 36 Our theory ultimately won't
require this, and of course any constraint like this will have to be dropped if
we are to accommodate SHIH INVARIANCE.
The interaction between parthood and location here is relatively clear:
there are seven (i.e. 23 - 1) parts of the pyramid: B 1, B2, B3, B 1 + B2,
B2 + B3, B1 + B3, and B 1 + B2 + B3. Each of these parts, including the
pyramid itself, has two locations-a box location and a jar location.
Insofar as we are generalizing the classical theory, then the box location
of B 1 + B2 should just be the fusion of the box location of B 1 and the box
location of B2, and similarly for the other three composite parts of the
pyramid. Since we have only finitely many objects this in effect gets us the
classical theory of location and part when we restrict quantification over
locations to locations within the box. None of the deviant behavior ruled
out by (1)-(5) ought to arise when restricting ourselves to box locations
(ignoring, for the moment, that the bricks are extended simples): thus
there ought to be a location function-a function satisfying (1)-(5)-
mapping parts of the pyramid to regions within the box. Parallel reasoning
suggests that there should be a location function mapping the pyramid
and its parts to regions inside the jar. These location functions, of course,
are also defined on material objects disjoint from the pyramid; however,
assuming there is no other multiple-location going on, we can assume that
the two location functions agree about the locations of every material
object disjoint from the pyramid. An arbitrary object is thus located at a
region iff one (or both) of these two location functions maps the object to
36
Note if the two locations of the pyramid did overlap, that would be completely
consistent with our ban on colocation. That ban rules out two objects being entirely located
at the same region, or entirely located at overlapping regions: but this is a case where only one
object is entirely located at two overlapping regions, and so is not subject to this ban.
Relative Locations 71
h t region: formally this means that the location relation is just the union
~f~hese two location ~unctions (recalling that both functions and relations
sets of ordered pairs).
areThis example is hopefully instructive enough as to make the following
rnodel of a location relation seem particularly natural:
A proper location relation, L, is a union of location functions, where
each location function is a mapping of material objects to regions of
space-time satisfying ( 1)-(5).
Thus any location relation, L, can be decomposed into a set of location
functions roughly telling us how objects and their parts are located relative
to certain locations. The intended models therefore consist of all tuples
(D, ::; , 0 , L) where (D, ::;) is a standard model of classical extensional
rnereology, o is an element of D whose improper parts represent the objects
that have locations, and L is a proper location relation whose domain
consists of the improper parts of o and whose range is disjoint from o.
It should be stressed that the decomposition of L into location functions
is not unique: one can have two sets oflocation functions that have the same
union. Here is a very simple example of this phenomenon. Suppose that two
point particles a and b are both triply located at each of three space-time
points, x, y, z, arranged in an equilateral triangle. Suppose further that the
fusion a+ b is multiply located at the three locations x + y, y + z, x + z
respectively (note that this stipulation is consistent with the constraint that
L be a union of location functions).
This situation can be represented by three location functions as follows:
Here each box represents how a and b are located in space relative to each
location function.
~~II
t_j~~
Note that it can also be represented by the following three location
functions:
~~II
t_j~~
72 Andrew Bacon Relative Locations 73

The uni~ns of these two sets of location functions are identical, and locate
a and b m ~xactly the way we described above. 37 It should also be obvious 3.3. The theory of part and location
t~1at the un10n of all .six location functions gives us yet another representa-
Now that we have specified the intended models of the theory, let us turn to
tion of the same relation. However, this latter representation is special in the
axiomatizing it. A major choice-point is whether to pursue a first- or second-
sense that it is the largest set of location functions whose union is L. Thus
order axiomatization. In what follows I shall present both; however, there
although there is no unique decomposition of a location relation int
location functions, there is always a unique maximal decomposition: th~ are several substantial technical questions concerning the first-order theories
that remain open. I outline these in further detail in Appendix B. These
set of location functions that are subsets of the location relation. When we
mainly concern completeness (is eve1y sentence true in all intended models
talk about the decomposition of a location relation we shall always mean the
provable in our system), and the status of representation theorems (can
maximal decomposition.
every relation that satisfies the axioms of the theo1y be expressed as a union
T~is observation suggests that while location functions are a useful way to
of location functions).
specify .the .structu'.·e of the location relation they don't have any independ-
The plural language augments first-order logic with plural variables,
ent real1~; mdee~ 111 th.e next se~tions we shall look at ways of characterizing
written xx, yy, zz, and so on, a plural quantifier Vxx and a singular-plural
the locat10n relation directly without appealing to location functions. It is
relation x-<.xx meaning that 'xis one of the xx'. In both the first- and second-
worth comparing our theory with prima fade similar responses to the shift
order cases, the non-logical vocabulary consists of a bina1y relation, L, a
argument that place more importance on the role oflocation functions. For
unary predicate 0, and a binary function symbol +. L represents the
example, Jeff Russell (2014) has recently defended the view that material
location relation, 0 applies to material objects, and + is a fusion operation.
objects have unique locations but it is a completely indeterminate or non-
We assume, for simplicity, that every object is either a material object or a
factual matter which exact locations they have. Each 'precisification' of the
region of space-time so that regions of space-time can be defined as •Ox.
location relation, on this view, is a location function, telling us where each
We have chosen to take the binary fusion operator as primitive for conveni-
object is located according to that precisification. It is interesting to note
ence, but note that many of the more familiar mereological relations can easily
that if one were to apply this model to the above example there would be a
be defined from it. Partl1ood is defined by x:::; y = df x + y = y, overlap by
genuine difference between a world where it is indeterminate whether the
xo y=dJ 3z(z:S x /\ z:Sy), and atomhood At(x) =df Vy(y :S X-Tx=y).
locations of a, b and a+ b are given by the first, second, or third location
We write xx <;;;Y.Y as short for 'iz(z--< xx -7 z--< Y.Y ). Finally, if x and y overlap,
functions, and another world where it is indeterminate whether the loca-
it is also convenient to have a term x n y-the product of x and y-that
tions of a, b and a + b are given by the fourth, fifth, or sixth (so that the
denotes the fusion of things that are parts of both x and y, and a term
admissible precisifications of'located at' are given by the first three locations
Fus( xx) which, whenever there is at least one xx, denotes the fusion of
functions in the first world, and the second three in the second world). On
the xx.
the picture where they're multiply located, rather than indeterminately
We may now start laying out the axioms of this theory. We shall start
located, there would be no difference between these two possibilities because
with the plural version, which results from adding to a standard axiomatiza-
they determine the same location relations.
tion of plural logic, the following (self-explanato1y) principles:
Ultimately ~hese distinctions get washed out in Russell's theory: he
accepts a prmc1ple analogous to SHwr INVARIANCE that effectively forces us The axioms of classical mereology (in terms of+ ).3 8
to choose the maximal set of precisifications. But this demonstrates that our
theo1y of multiple location is at least conceptually very different; it does not
~nake the kind~ of invisible dis~in.ctions that would have to be acknowledged 38
.F;om the .fi.rs~ three axioms and our definition of parthood one can prove the
1f we. started :"1th a se~ of a~m1ss.1ble location functions instead of the single reflex1v1ty, trans1t1v1ry, and anti-symmetry of the parthood relation. (For example, to
location relat10n that 1s their umon. show anti-symmetry suppose that x :::'. y and y :::'. x. Thus by definition x +y = y and
Y.+:~=x. But by the second axiom x+y=y+x so x=y. The other principles are
s1m1larly ~trn1ghtforward.) Thus every principle of classical mereology is provable given
our defi~1t1~n of parthood (the fusion principle needs no modification). Conversely, the
below pn~c1pl~s can be proven in classical mereology if we add the linking principle that
37 We also constructed it so that it satisfied SHrFT EQUIVALENCE. x+ y=z tff z is the fusion of x andy: x+ y=z <-> Vu(u oz<-> (u o xVu o y)).
74 Andrew Bacon
x+x=.x:
x+y=y+x
x+(y+z)=(x+y)+z
\ixx3x\iy(xoy <---+ 3z(z-<xx!\yoz))
All and only material objects are located somewhere.
Ox<---+ 3yLxy
Locations aren't material objects.
Lxy-+ -iOy
Atoms are located at space-time points.
At(x) !\Lxy-+At(y).
In the classical theory of part and location, we had an axiom that said
that the location relation preserved fusions: I (x + y) I (x) + I (y). Our
approach to axiomatizing the theory of part and location will be to
produce a suitable analog of this axiom. However, one must be careful:
it is not true, for example, that the locations of x +y are all possible
fusions of the locations of x and the locations ofy. In the case of the Lego
pyramid, for example, the fusion of Bl's box location and B2's jar
location is neither wholly located in the box or the jar, and is thus not a
location of B 1 + B2.
Here is something that is true, however. If the pyramid (the fusion
Bl + B2 + B3) is located at a region R, then there is a way of cutting R
up into three disjoint pieces Rl, R2, and R3 such that (i) Bl is located at Rl,
B2 at R2, and B3 at R3, and (ii) Bl + B2 is located at Rl + R2, Bl + B3 at
RI+ R3, B2 + B3 at R2 + R3 ... and (iii) Bl+ B2 + B3 is located at
RI +R2+R3.
This idea can be generalized in the following way. Suppose that an
arbitrary object x is located at a region y. Then for any way of cutting x
up into smaller disjoint pieces, (x;) iEI (indexed by some set I), there is a way
of cutting up y into an equal number of little pieces (y;) iEI such that
whenever j cl, Fus( {x1 Ij E J}) is located at Fus( {y1 Ij E J}). Thus by
letting j range over singletons, we can see that each part of x, x; say, is
located at the corresponding part of y, y;. Moreover, each fusion of these
parts of x (say, x; + x1 + Xk) are located at the fusion of the corresponding
parts of y, (that is, y; +Yi+ Jk). (And similarly for infinite collections of
these parts.)
This brings us to the final and most important axiom of our theory,
which effectively encodes this thought in plural logic. A partition
Relative Locations 75
of an object, x, are some things xx which are pairwise disjoint
Vx (x-<xx !\y-< xx)-> (x o y-+ x= y)) but fuse to x (Fus(xx) =x).
i :rtidon xx divides x if and only if x is a fusion of some of the xx.
Ftn, l!vJ write mat(x) for the
~1
. material part of x (i.e. x n Fus( (x : Ox)))
and reg(x) for the space-ttme
•
part of x (.i.e. x n Fus ( ( x .• -, 0 x·))) .·>9 Tl1e
final axiom says:
ARBITRARY PARTITIONING: Suppose that x is located at y and that the xx are
a partition of the material objects that divides x. Then there are some
things, the zz, that are a partition and that divides x + y, such that:
(a) If z and ware one of the zz, and mat(z) = mat(w) or reg(z) = reg(w)
then z= w.
(b) If the ww <;;;; z.z then mat (Fus( ww)) is located at reg (Fus( ww)).
(c) If z-< zz and mat(z) is atomic, reg(z) is atomic.
Our formalization of the intuitive thought, which uses indexing sets, is a
little less direct, since we are restricted to a plural language. The zz effect-
ively encode a pairing between the given partition of x and some partition of
y, and a little reflection should reveal that ARBITRARY PARTITIONING correctly
captures the thought explained above.
The above plural axiomatization has an important properry: it accurately
captures the intended models of the pure theory of part with multiple
location. Any full model of the axioms is one in which L can be represented
as a union of location functions. 40
In order to state a first-order version of this theory we must first replace
the mereological composition axiom above with a schema:
3zrp-+ 3x\iy(xoy<-+ 3z(ijJ!\yoz))
39 Here (x : ¢x) is a plural expression for the ¢s.
40
A model of a plural language is fit!! if the plural quantifiers range over every subset of
the domain. We show that L is the union of the location functions that are subsets of L
(here we use L to denote the interpretation of the location function in a model). In
particular this means showing that if Lxy then some location function that is a subset of L
maps x toy (we show this under the assumption of atomism; a more intricate argument is
needed without the assumption). Suppose that xis located at y and consider the partition
that consists of all the material atoms. Arbitrary partitioning guarantees that there is a
bijection f between these atoms and the atoms of a region of space-time containing y,
whose graph is determined by material and space-time parts of the elements of the
partition zz. This bijection will be such that (i) the fusion of any set of those atoms, X,
is lornted at the fusion of the image of X underf,f(X). If we let !(a) be the fusion of the
image of as atoms under f then l is a location function that is a subset of L. Moreover,
since zz divides x + y, l maps x toy as required.
76 Andrew Bacon
We must also find a suitable first-order replacement for ARBITRARY
PARTITIONING. The most straightforward substitute is the following fini-
taty version:
PARTITIONING: Ifx is located at y, and x 1, •.. , x,, partition x, then there are
y 1, •.• , y,, such that each x; is located at y;, each x; + x1 is located at y; + y1,
each x; + x1 + Xk is located at y; + YJ + Jk (and so on).
Unlike ARBITRARY PARTITIONING, PARTITIONING is a schema, with a different
instance for each choice of n. That this schema really is expressible in first-
order logic is shown in Appendix B.
Here the logical issues are not as clean as with the second-order case. In
any model of the theory in which the extension of 0 is finite, the extension
of L is a union oflocation functions. However, there are unintended models
when the extension of 0 is infinite. Two natural questions then present
themselves: are there any first-order axiomatizations that rule out unin-
tended models, and are there any first-order axiomatizations that are
sound and complete for the class of intended models (even if such axioma-
tizations admit unintended models)? Both these questions are explored
further in Appendix B.
3.4. Combining the theory with shift invariance
We can eliminate reference to the mathematical notion of a Euclidean
transformation in our statements of SHIFT INVARIANCE and SHIH EQUIVA-
LENCE, so that our entire theory can be stated in the internal language of
congruence, betweenness, location, and part.
In fact the statement of both principles takes a particularly simple form,
and turns out to be a restriction of both principles to diatoms. For example,
'diatomic shift invariance' says that if a diatom x is located at p + q (a fusion
of two space-time points) then it is located at every Euclidean transform of
p + q. Note, however, that p' + q' is a Euclidean transform of p + q iff p and
q are congruent to p' and q', which is something we can state in our
fundamental vocabulary consisting of congruence and betweenness rela-
tions. Diatomic shift equivalence similarly becomes the claim that if x is
located at p + q and p' + q' then p and q are congruent top' and cJ. Putting
both together we get:
CONGRUENT DIATOMS: Lx(p+q)-> (Lx(p' +c/) ,_... Cpqp'q') where
p, q, p'
and cJ range over space-time points.
To see that this principle is adequate, we need to know that if the diatoms
are located at all and only the Euclidean shifts of their locations then every
Relative Locations 77
object is located at all and only the Euclidean shifts of their locations. In other
words, we need to know that CONGRUENT DIATOMS entails SHIH INVARIANCE
and SHIFT EQUIVALENCE. In fact this follows from a fact we mentioned earlier
and which we are now in a position to prove: the locations of a whole are
determined by the locations of its diatomic parts (even though, as we saw
earlier, they are not determined by the locations of their atomic parts).
Theorem 3.1 (The Locations of the Diatoms Determines the Locations of Every-
thing) Let L be a proper location relation (a union of location fimctions), between some
domain ofmaterial objects and Euclidean space E 3 . Suppose that for eve1y diatom there is
a pair ofspace-time points such that the diatoms locations are given by the collection ofall
pairs of space-time points congruent to that pair. Then for any material object, there is
some region such that the objects locations are given by the collection of all regions
congruent to that region.
Proof Suppose that Lis a union of a set oflocations functions, F, and that dis
some metric that coheres with the congruence and betweenness structure on
£3. We want to show that if Lxy and Lxz then y is congruent to z: there is some
Euclidean transformation that maps y to z. It is a standard result that y and z are
Euclidean transformations of one another if and only if there is a distance
preserving bijection (an 'isometry') between the points in y and the points in z.
From the fact that Lxy and Lxz it follows that there are location functions
f, g E F such that/ (x) = y and g (x) = z. I claim that the mapping t =go
1- 1
restricted to points is an isometry between points iny to points in z. One
can see by inspection that it maps y to z. Suppose that p and q are two
points in y and that t(p)=p' and t(q)=q'. Now let a=/- 1 (p) and
b=r 1
(q). It follows that g(a)=p' and g(b)=cJ. Note also that a
and b must be mereological atoms because they are mapped to atoms
3
in £ by a location function. Thus a+ b is a diatom, and it is located at p + q
by f (note that/(a + b) =/(a)+ /(b) = p + q) and located at p' + q' by g
(since g (a+ b) = g (a)+ g (b) = p' + q'). By CONGRUENT DIATOMS
it follows that p + q is congruent to p' + q', which means that
d(p, q) = d(p', q') d( i(p ),/( t(q)) as required of an isometry.
A small modification to the argument must be made to extend to the
four-dimensional case: regions are transforms of each other in the relevant
sense iff there is an isometry that also preserves simultaneiry and non-
simultaneity of each pair of points.
4. CONCLUSION
w~ have considered two sorts of theories of the geometry of material
objects: theories in which they inherit their structure by their relation to a
78 Andrew Bacon
space-time manifold which has its geometric structure internally, and the-
ories in which they inherit their structure by their relation to a platonic
realm of abstracta. We have seen that the most straightforward versions of
both these views give rise to certain kinds of objectionable invisible struc-
ture. That does not mean we should reject these approaches, for they may
have other theoretical virtues that could not be achieved othe1wise. How-
ever, I have outlined a theory that can provide an adequate account of the
geometry of physical objects that does without these invisible differences,
and is otherwise quite simple. To be sure, the view does not describe (nor
does it purport to describe) our untutored ideas about the nature of space-
time as it relates to our ordina1y talk of locations. But it strikes me that this
talk is just a convenient shorthand for describing facts about the relative
locations of different objects (much like the physicists use of coordinate
systems) and should not be taken seriously as a guide to how things are
fundamentally.
5. APPENDIX A: FIELDS
Fields, as they are ordinarily conceived, are properties of space-time: a field
determines a field value for each point in space-time, which specifies the
strength (and possibly also direction) of the field at that point. On this view
the strength of the field values at a particle's location determines the
behavior of that particle. This picture, however, is fundamentally at odds
with the present view, since each particle has many locations, and so no
unique field value can be associated with a particle. This obstacle is not fatal
to the present approach: Newtonian mechanics can be formulated as an
'action at a distance' theory-a theory that works directly with the forces
acting on each particle by other particles without invoking fields to mediate
these forces. However, this approach is not in the spirit of (if not incom-
patible with) the idea that forces and other quantities usually represented by
mathematical objects are fundamentally reducible to more basic relations
between concrete entities like space-time points. For example, the action at a
distance theory assigns a number to each ordered pair of particles telling us
the force one exerts on the other. A problem similar to that posed for
relationism then arises: unless there are a continuum of particles between
each pair of particles, relations encoding the forces between the particles
alone will be insufficient to specify all the possible kinds of forces a particle
can exert on another. If one could consider the forces exerted at a con-
tinuum of points between the two particles (as one might in some sort of
field theory) we could employ the methods discussed in section 2.2 to
Relative Locations 79
eliminate numbers. At any rate, the use of space-time fields are so pervasive
in modern physics it would be remiss not to treat them.
In what follows I'm going to focus on strategies, in the spirit of those
proposed by Field (19~0), .for demathematizing scalar and vector fields by
reducing them to certam kmds of congruence and betweenness facts. This
strategy comes in two parts, the first of which is to note that most vector
fields of physical interest can be equivalently described by a scalar field, so
that our problem can in effect be reduced to the problem of giving an
account of scalar fields. The second is to see that a scalar field can be
represented, up to an affine transformation, by a pair of congruence and
betweenness relations.
If a vector field is 'conservative'-a property most physical fields share-
then it can be represented as the gradient of a scalar field. 41 To get a feel for
this result it helps to visualize the special case of a scalar field on two-
dimensional space: one can picture this as a kind of hilly terrain imposed
over the surface where the height of the field above the surface represents the
magnitude of the scalar field. The slope of the terrain at a point-the vector
pointing downhill with a magnitude proportional to the steepness-will be
a vector field of the requisite kind. This idea easily generalizes to higher
dimensions. When a vector field is generated by a scalar field in this way, we
call the scalar field a 'potential' for the vector field.
It should be noted that many different scalar fields generate the same
vector field-one can uniformly raise or lower the height of a terrain
without changing the direction or magnitude of the slope at each point.
For this reason people normally don't consider the potential scalar field to
be fundamental. When the laws of physics depend only on the values of a
vector field, and thus do not discriminate between different scalar fields
that generate that vector field, it is extremely natural to think that the
only physically real distinctions are those that give rise to different
assignments of vectors to points. It thus seems prima facie inadvisable
to attempt to reduce vector fields to scalar fields. However, our strategy is
to reduce both scalar and vector fields to congruence and betweenness
relations, so, for example, the potential betweenness relation would say
11
' A vector field is conservative if the 'work-done' to get from one point to another
does .not depend on the path you take. Most fields that have a chance of representing real
physical fields are conservative so our restriction won't cost us much in the way of
generality. Note, though, that an arbitrary vector field on an n-dimensional manifold
can be decomposed into n scalar fields representing its n-components (although this
decomposition is highly non-unique). So, if needed, a more general reduction procedure
is available.
80 Andrew Bacon
'the potential at xis between the potential at y and z'. Interestingly, in this
setting the correspondence between vector fields and the potential congruence
and betweenness facts becomes one-to-one. In particular, one e<m change the
height of a potential without either changing the vector field it generates or the
congruence and betweenness facts it generates. Thus the strategy of reducing
conservative vector fields to facts about potential congruence and betweenness
at space-time points seems like a promising place to start.
However, once we adopt SHIFT INVARIANCE we quickly encounter a
completely independent difficulty that has nothing to do with the project
of demathematizing fields. To illustrate the problem, consider a single
particle with external forces acting on it. We would normally represent the
gravitational potential by a distribution of quantitative properties over
space-time points and determine the particle's motion by the distribution
of these properties in a neighborhood of the particle's location. If the
particle is multiply located at evety space-time point this method breaks
down: the gradient of the potential is different at each point of space, so
we can't tell anything about the force acting on a particle just by looking
at its locations. It is natural to think that this means that fields have to be
multiply located as well: if a point p has a certain gravitational potential,
every Galilean transformation of p must have that potential as well. But
this is not sufficient either, for it would entail that every field value that's
had anywhere is had by every space-time point at once (since every space-
time point is a Galilean transform of every other space-time point). The
location of the field has to be somehow correlated with the locations of the
material objects for this to work. 42 The following is an exploration of one
way of meeting this challenge.
5.1. Demathematizing fields in orthodox Newtonian physics
We shall start by getting a clearer handle on the representation of forces in
the ordina1y substantivalist setting, in which each particle has a single
42 Let me note in passing that one way to do this would be to expand our ontology of
material objects to include fields as well, consisting of point-sized parts. Field values can
then be construed not as properties of space-time points, but as properties of the point-
like parts of the field. Thus the field values are multiply located in a way that correlates
with locations of the field itself; and any other material objects we might want to include.
This strategy is not without contention, however. Since each material object has a unique
'location' within a given field, this maneuver potentially reinstates the problems we
originally had with space-time. For example, Amtzenius (2012) points out (in the context
of a discussion of relationism) that you can shift the field values at each point of a field to
some other point on the field in a uniform fashion, in a way completely analogous ro a
Leibniz shift on space-time.
Relative Locations 81
I cation. Let us begin by considering a very simplified example consisting of
;.., point particles, a and b, constrained to one dimension with no external
0
forces acting on them. Both a and b generate a force field defined over space:
the value of the field generated by a at the location of b determines how
uickly and in what direction b will accelerate. Such forces can be equiva-
~ntly represented by a scalar potential (the potential energy): in this case it
will be the values of the potential in a neighborhood of a and of b that will
determine their respective motions. 43 Now we can imagine what would
happen if we were to uniformly translate the positions of the two particles
in some direction by some fixed amount without also shifting the fields.
The result would not be legal: each force field has a source-a point from
which the field emanates-and these sources would not match the loca-
tions of the particles. In particular the force acting on a and b respectively
would depend on their distances from the sources, and not on the distance
from each other as it ought to. In order to preserve the laws we must
translate a, b, and the gravitational field as well. Similarly, when the two
particles move in a lawlike manner, we must also provide laws telling
us how the field values change, so that the field sources 'keep up' with
the particles.
A metaphysically distinct (although mathematically equivalent) way to
represent our two particle example is to imagine the system having a single
location in a higher-dimensional space called configuration space, where each
point in configuration space represents all the particles' locations in ordinary
space. Recall that both particles in our example reside in a one-dimensional
space, and there are only two of them, so in this case configuration space is
two-dimensional. (More generally, when we are considering all three dimen-
sions, and there are N particles, configuration space will have 3N dimen-
sions.) The gravitational forces can similarly be represented by a vector field
on this two-dimensional space, with the first component of the vector at a
point representing the force a exerts on b and the second component
representing the force b exerts on a when a and b are located according to
that point in configuration space. As before, this vector field can be equiva-
lently represented by a scalar potential. A uniform shift of the locations of
a and b in configuration space corresponds to moving the location of the
system diagonally in configuration space (north-east or south-west, as it
43
Sometimes it is natural to consider a single vector field defined over space-the
gravitational field-telling us the force per unit mass that would be exerted at each point
of space-time due to every other particle. I have chosen to illustrate things with two
distinct fields, rather than merging them into one, because it makes for a more straight-
forward comparison with views considered later.
Another random document with
no related content on Scribd:
described; and the value of his work has been very differently
appraised. Cuvier had small opinion of it. Camper and Saint-Hilaire
considered the author the greatest naturalist of modern times, the
French Aristotle. Topinard (1885, p. 33) thus describes the opinion of
the public: “Le public, lui, n’hésita pas; dans l’Histoire naturelle des
animaux il sentit un souffle nouveau, vit un pressentiment de l’avenir.
La libre pensée était dans l’air, 89 approchait; l’œuvre de Buffon,
comme l’Encyclopédie, Voltaire, Rousseau et Bougainville, contribua
à la Révolution française.”
The genius of Linnæus lay in classification. Order and method
were with him a passion. In his Systema Naturæ he fixed the place
of Man in Nature, arranging Homo sapiens as a distinct species in
the order Primates,[14] together with the apes, the lemurs, and the
bats. He went further and classified the varieties of man,
distinguishing them by skin colour and other characters into four
groups—a classification which holds an honourable place at the
present day.
14. The tenth edition, 1758, is the first in which the order Primates occurs. Earlier
editions have the order Anthropomorpha. See Bendyshe, p. 424.

All this was abominable in the eyes of Buffon. “Une vérité

humiliante pour l’homme, c’est qu’il doit se ranger lui-même dans la
classe des animaux”; and in another place he exclaims: “Les genres,
les ordres, les classes, n’existent que dans notre imagination.... Ce
ne sont que des idées de convention.... Il n’y a que des individus!”
And again: “La nature ne connait pas nos definitions; elle n’a jamais
rangé ses ouvrages par tas, ni les êtres par genres.”
Nevertheless both rendered incalculable service to the science.
Linnæus “found biology a chaos and left it a cosmos.”
“L’anthropologie,” says Flourens, “surgit d’une grande pensée de
Buffon; jusqu-là l’homme n’avait été étudié que comme individu,
Buffon est le premier qui l’ait envisagé comme espèce.”
But Buffon was no believer in the permanent stability of species.
“Nature is far from subjecting herself to final causes in the formation
of her creatures.” He went so far as to make a carefully veiled hint
(the Sorbonne having eyes on him) of a possible common ancestor
for horse and ass, and of ape and man. At least, he says, so one
should infer from their general resemblance; but, since the Bible
affirms the contrary, “of course the thing cannot be.”[15] In 1751 the
old naturalist was constrained by the Sorbonne to recant his
geological heresies in these words: “I declare that I had no intention
to contradict the text of Scripture; that I believe most firmly all therein
related about the Creation, both as to order of time and matter of
fact.”
15. Quoted from Clodd’s Pioneers of Evolution, 1897, p. 101.
 J. F. Blumenbach

Blumenbach. It was fortunate for the nascent science that the

next great name on its roll was that of a man of
very wide reading, endowed with remarkable reasoning powers, and
with an exceptional perspicuity for sifting out the true from the false.
Johann Friedrich Blumenbach (1752-1840) was Professor in the
Faculty of Medicine at Göttingen, and early turned his attention to
the special study of man. He was the first to place anthropology on a
rational basis, and in his De generis humani varietate nativa (1775-
1795) laid the foundations of race classification based on
measurement. He noted the variations in the shape of the skull and
of the face, and may therefore be regarded as the founder of
craniology (see below, p. 28). Besides the services rendered by
Blumenbach to the science of anthropology in classification and in
laying the foundations of craniology, there was a third field in which
his work was perhaps even more valuable to his contemporaries.
Monsters. Every successive age is astonished at the
credulity of its predecessor; but when we
remember the grave difficulties which beset the explorer in the
eighteenth century, and the wild “travellers’ tales” which it was
impossible either to verify or to disprove, it is easy to sympathise
with the credence given to the beliefs in “Anthropophagi, and men
whose heads do grow beneath their shoulders.” Tyson, in his
Philological Essay, gives a list, chiefly derived from classical writers,
of the “monstrous Productions,” belief in which had not altogether
died out in the seventeenth century. In fact, it was not long before
Tyson’s time that a distinguished naturalist had given a serious
description of the mermen who lived in the sea and had their hinder
parts covered with scales.[16] Tyson’s account of “Monstrous sorts of
Men” is taken mainly from Strabo:—
16. v. Cunningham, p. 24.

Such are the Amukteres or Arrhines, that want Noses, and have only
two holes above their Mouth; they eat all things, but they must be raw;
they are short lived; the upper part of their Mouths is very prominent. The
Enotokeitai, whose Ears reach down to their Heels, on which they lye and
sleep. The Astomoi, that have no Mouths—a civil sort of People, that
dwell about the Head of the Ganges; and live upon smelling to boil’d
Meats and the Odours of Fruits and Flowers; they can bear no ill scent,
and therefore can’t live in a Camp. The Monommatoi or Monophthalmoi,
 that have but one Eye, and that in the middle of their Foreheads: they
have Dogs’ Ears; their Hair stands on end, but smooth on the Breasts.
The Sternophthalmoi, that have Eyes in their Breasts. The Panai
sphenokephaloi with Heads like Wedges. The Makrokephaloi, with great
Heads. The Huperboreoi, who live a Thousand years. The Okupodes, so
swift that they will out-run a Horse. The Opisthodaktuloi, that go with their
Heels forward, and their Toes backwards. The Makroskeleis, the
Steganopodes, the Monoskeleis, who have one Leg, but will jump a great
way, and are call’d Sciapodes, because when they lye on their Backs,
with this Leg they can keep the Sun from their bodies.

Wild Men. Linnæus did not include these in his Homo

Monstrosus; but various questionable creatures
are inserted by his pupil Hoppius in the treatise Anthropomorpha of
Linnæus, read in 1760.[17] Such were the Satyr of Vulpius, who,
“when it went to bed, put its head on the pillow, and covered its
shoulders with the counterpane, and lay quite quiet like a
respectable woman”; Lucifer (Homo caudatus), the “dreadful foul
animals—running about like cats,” who rowed in boats, attacked and
killed a boatload of adventurers, cooking and eating their bodies; and
the Troglodyta (Homo nocturnus), who in the East Indies “are caught
and made use of in houses as servants to do the lighter domestic
work—as to carry water, lay the table, and take away the plates.” But
all these were classed among the Simiæ. Within the species Homo
sapiens Linnæus included wild or natural man, Homo sapiens ferus,
whose existence was widely believed in at the time. The most
authentic case was that of “Wild Peter,” the naked brown boy
discovered in 1724 in Hanover. He could not speak, and showed
savage and brutish habits and only a feeble degree of intelligence.
He was sent to London, and, under the charge of Dr. Arbuthnot,
became a noted personage, and the subject of keen discussion
among philosophers and naturalists. One of his admirers, more
enthusiastic than the others, declared that his discovery was more
important than that of Uranus, or the discovery of thirty thousand
new stars.
17. Bendyshe, p. 447.
 Blumenbach alone, apparently, took the trouble to investigate the
origin of Wild Peter, and in the article he wrote on the subject
disposed for all time of the belief in the existence of “natural man.”
He pointed out that when Peter was first met he wore fastened round
his neck the torn fragments of a shirt, and that the whiteness of his
thighs, as compared with the brown of his legs, showed that he had
been wearing breeches and no stockings. He finally proved that
Peter was the dumb child of a widower, who had been thrust out of
his home by a new step-mother.[18]
18. Cunningham, pp. 24-5.
 Chapter II.

THE SYSTEMATISERS OF PHYSICAL ANTHROPOLOGY

Hitherto we have been dealing with the great pioneers in

Anthropology, those who laid the foundations, brought order out of
chaos, and suggested the outlines of future work. Henceforward
Anthropology may claim the name of a science, and the work
developed on definite lines. It will be more convenient to treat these
separately, abandoning a strict chronological method.
The first branch to attract workers was Somatology, the physical
aspect of man, of which we have already noted the inception: not
until the nineteenth century can Archæology, or Prehistoric
Anthropology, be said to have developed into a science; while the
scientific study of Ethnology, or Cultural Anthropology, is barely half a
century old.
Craniology. Somatology had already been foreshadowed by
Norma Verticalis Vesalius, Spigel, and Linnæus; but Blumenbach
of Blumenbach. was the first to strike its keynote by recording the
shape of the skull and of the face. He was the
fortunate possessor of a large number of skulls—large, that is, for his
time, and he published a description of them (1790-1820), Decas
collectionis suae craniorum diversarum gentium illustrata, with 70
plates. He noted particularly the norma verticalis—i.e., the shape of
the skull as seen from above, distinguishing by its means three types
—the square shape of the Mongols, the narrow or “pressed in from
the sides” shape of the Negroes, and the intermediate form which he
recognised in the “Caucasians.” He was the first to popularise
craniology, and “it became the fashion to visit the Blumenbachian
Museum, to have the differences which distinguish the different
cranial types pointed out, and to indulge in sentimental rhapsodies
upon the beauty and symmetry of the young female Georgian skull,
which was considered to represent the highest type of all.”[19] But
Blumenbach does not seem to have taken advantage of his own
discoveries. In choosing the norma verticalis as a racial criterion he
made a valuable contribution to science, but he did not reproduce his
normæ in his plates, nor did he base his classification on them.
Indeed, his typical Caucasian skull is really squarer than his typical
Mongolian.
19. Cunningham, p. 26.

Peter Camper (1722-1789) had already been studying head-form,

though from a totally different standpoint, and his deductions were
not published until after his death.
His contributions to Anthropology were an essay on the Physical
Education of the Child, a lecture on The Origin and Colour of the
Negro, and a treatise on The Orang-outang and some other species
of Apes; but only his work on the facial angle has attained
permanent fame. His early inclinations were towards art, and he was
carefully trained in drawing, painting, and architecture; and it was in
the interests of Art, not of Anthropology, that the researches which
resulted in his determination of the facial angle came to be
undertaken. This he describes in his preface to his lectures:—
At the age of eighteen, my instructor, Charles Moor the younger, to
whose attention and care I am indebted for any subsequent progress I
may have made in this art, set me to paint one of the beautiful pieces of
Van Tempel, in which there was the figure of a negro, that by no means
pleased me. In his colour he was a negro, but his features were those of
a European. As I could neither please myself nor gain any proper
directions, I desisted from the undertaking. By critically examining the
prints taken from Guido Reni, C. Marat, Seb. Ricci, and P. P. Rubens, I
observed that they, in painting the countenances of the Eastern Magi,
had, like Van Tempel, painted black men, but they were not Negroes.
Upper and Side Views of Skulls of Men
belonging to the Neolithic and Bronze Age Races; photographed by
the Author from specimens in the Cambridge Anatomical Museum.

A, Long Barrow, Dinnington, Rotherham. Length, 204 mm.; breadth,

143 mm.; cranial index, 70. 1.
B, Winterbourne Stoke. Length, 177 mm.; breadth, 156 mm.; cranial
index, 88. 1.

Facial Angle of To obtain the necessary facial effects

Camper. distinguishing the Negro from the European,
Camper devised his system of measurements. He drew a line from
the aperture of the ear to the base of the nose, and another from the
line of the junction of the lips (or, in the case of a skull, from the front
of the incisor teeth) to the most prominent part of the forehead. “If,”
he said, “the projecting part of the forehead be made to exceed the
100th degree, the head becomes mis-shapen and assumes the
appearance of hydrocephalus or watery head. It is very surprising
that the artists of Ancient Greece should have chosen precisely the
maximum, while the best Roman artists have limited themselves to
the 95th degree, which is not so pleasing. The angle which the facial
or characteristic line of the face makes,” he continued, “varies from
70 to 80 degrees in the human species. All above is resolved by the
rules of art; all below bears resemblance to that of apes. If I make
the facial line lean forward, I have an antique head; if backward, the
head of a Negro. If I still more incline it, I have the head of an ape;
and if more still, that of a dog, and then that of an idiot.”
Camper’s facial angle may be of service to Art, but since the
points from which the lines are drawn are all variable, owing to the
disturbing influence of other factors, such as an increased length of
face or an unusually prominent brow-ridge, it cannot form an
accurate measurement for Anthropology. It was severely criticised by
Blumenbach, Lawrence, and Prichard, but adopted in France, and by
Morton in America.
Dr. J. Aitken Meigs[20] pointed out that as early as 1553 the
measurement of the head appears to have exercised the ingenuity of
Albert Dürer, who, in his De Symmetriâ Partium in Rectis Formis
Humanorum Corporum, has given such measurements in almost
every view. These, however, are more artistic in their tendency and
scope than scientific. A glance at some of the outline drawings of
Dürer shows incontestably that the facial line and angle were not
wholly unknown to him, and that Camper has rather elaborated than
invented this method of cranial measurement. The artist even seems
to have entertained more philosophical views of cephalometry, or
head measurement, than the professor.
20. North American Med.-Chir. Rev., 1861, p. 840.
Various Early The evolution of craniometrical measurements
Craniologists. is of interest to the physical anthropologist, but
even a brief recital of this progress would weary the non-specialist. A
history of Anthropology would, however, not be complete if it ignored
the general trend of such investigations.
Some of the early workers, such as Daubenton (1716-99) and
Mulder, Walther, Barclay, and Serres in the first half of the nineteenth
century, attempted to express the relation between the brain-case
and the face by some simple measurement or method of comparison
in their endeavour to formulate not only the differences between the
races of mankind, but also those which obtain between men and the
lower animals.
Others during the same period investigated the relations and
proportions of portions of the skull to the whole by means of lines.
Spix (1815) adopted five lines. Herder employed a series of lines
radiating from the atlas (the uppermost bone of the vertebral
column); but, more generally, the meatus auditorius (ear-hole) was
the starting-point (Doornik, 1815).
The internal capacity of the skull first received attention from
Tiedemann (1836), who determined it by filling the skull with millet
seed and then ascertaining the weight of the seed. Morton first used
white pepper seed, which he discarded later for No. 8 shot, while
Volkoff employed water. Modifications in the use of these three
media—seeds, shot, and water—are still employed by craniologists.
The most noteworthy names among the earlier workers in
craniology are those of Retzius and Grattan. Anders Retzius (1796-
1860) correlated the schemes of Blumenbach and Camper, and so
arrived at the methods of craniological measurements which are
almost universally in use at the present day.
Cephalic Index In 1840 he introduced his theory regarding
of Retzius. cranial shapes to the Academy of Science at
Stockholm, and two years later gave a course of lectures on the
same subject. He criticised the results attained by Blumenbach,
showing that his group contains varying types of skull form; and he
invented the cephalic index, or length-breadth index—i.e., the ratio of
the breadth of a skull to its length, expressed as a per-centage. The
narrower skulls he termed dolichocephalic, the broader ones
brachycephalic. By this method Retzius designed rather to arrange
the forms of crania than to classify thereby the races of mankind,
though he tried to group the European peoples more or less
according to their head-form. While thus elaborating the suggestion
of Blumenbach, he also recorded the degree of the projection of the
jaws, demonstrated by Camper, and he added the measurements of
the face, height, and jugular breadth. Thus was Craniology
established on its present lines.
Grattan. John Grattan (1800-1871), the Belfast
apothecary, has never received the recognition
that was his due. Having undertaken to describe the numerous
ancient Irish skulls collected by his friend Edmund Getty, he soon
became impressed by the absence of
that uniformity of method and that numerical precision without which no
scientific investigation requiring the co-operation of numerous observers
can be successfully prosecuted. The mode of procedure hitherto adopted
furnishes to the mind nothing but vague generalities ... until we can
record with something approaching towards accuracy the proportional
development of the great subdivisions of the brain, as indicated by its
bony covering, and by our figures convey to the mind determinate ideas
of the relation they bear towards each other, we shall not be in a position
to do justice to our materials.... No single cranium can per se be taken to
represent the true average characteristics of the variety from which it may
be derived. It is only from a large deduction that the ethnologist can
venture to pronounce with confidence upon the normal type of any race.
[21]

21. J. Grattan, Ulster Journal of Arch., 1858.

Grattan devised a series of radial measurements from the meatus

auditorius, and constructed an ingenious craniometer. As Professor
J. Symington points out, “Grattan’s work was almost
cotemporaneous with that of Anders Retzius, and nearly all of it was
done before the German and French Schools had elaborated their
schemes of skull measurements.”[22] He adopted the most useful of
the measurements then existing, and added new ones of his own
devising.
22. Proc. Belfast Nat. Hist. and Phil. Soc., 1903-4; and Journ. Anat. and Phys.
The distinguished American physician and physiologist Dr. J.
Aitken Meigs laid down the principles that “Cranial measurements to
be of practical use should be both absolute and relative. Absolute
measurements are necessary to demonstrate those anatomical
differences between the crania of different races which assume a
great zoological significance in proportion to their constancy. By
relative measurements of the head we obtain an approximate idea of
the peculiar physiological character of the enclosed brain ... the
craniographer, in fact, becomes the cranioscopist” (1861, p. 857). In
this paper Meigs gives craniometrical directions, some of which were
designed to give measurements for portions of the brain.
Broca. In France the greatest names are those of
Broca, Topinard, and de Quatrefages. Pierre Paul
Broca (1824-1880) was first destined for the army, but when the
death of his sister left him the only child he was unwilling to leave his
parents, and resolved to study medicine and share the work of his
father, an eminent physician. He soon distinguished himself,
especially in surgery, not only in practical work, but also in his
writings. With regard to the latter, Dr. Pozzi, in a memoir, says of him:
“There is hardly one of the subjects in which he did not at the first
stroke make a discovery, great or small; there is not one on which he
has not left the mark of his originality.”[23]
23. J.A.I., x., 1881, p. 243.
Paul Broca.
 In 1847 he was appointed to serve on a Commission to report on
some excavations in the cemetery of the Celestins, and this led him
to study craniology, and thence to ethnology, in which his interest,
once aroused, never flagged. The story of the formation of the
Société d’anthropologie de Paris (1859) and of l’École
d’anthropologie (1876), of both of which Broca was the moving spirit,
affords a curious commentary on the suspicion in which
Anthropology was held. To the success of the School he devoted all
his energies, and during many years of anxiety he met and
overcame all obstacles, surmounted all difficulties, wore down all
opposition, and finally placed it in a secure position. He invented
several instruments for the more accurate study of craniology, such
as the occipital crochet, goniometer, and stereograph, and also
standardised methods; but, dissatisfied with the inconclusiveness of
mere cranial comparisons, he turned towards the end of his life to
the study of the brain. He was an indefatigable worker, and his
sudden death in his fifty-sixth year is attributed to cerebral
exhaustion.
“Broca was a man,” said Dr. Beddoe, “who positively radiated
science and the love of science; no one could associate with him
without catching a portion of the sacred flame. Topinard has been
the Elisha of this Elijah.”[24]
24. Anniversary Address, Anth. Inst., 1891.
Topinard. Paul Topinard, pupil, colleague, and friend of
Broca, made valuable investigations on the living
population of France, besides devoting much time to
anthropometrical studies; but his greatest service has been the
preparation and publication of l’Anthropologie (1876), a guide for
students and a manual of reference for travellers and others, voicing
the idea of Broca and his school, and “elucidating in a single volume
a series of vast dimensions, in process of rapid development.” In
1885 he published his classic Eléments d’anthropologie générale,[25]
which aimed at creating a new atmosphere for the science, breaking
free from the traditions of the monogenists and polygenists, and
incorporating the new ideas spread by Darwin and Haeckel.
 25. General Anthropology, according to Topinard’s classification, is concerned
merely with man as an animal, and deals with anatomy and physiology,
pathology, and psychology.
De Quatrefages. Armand de Quatrefages de Bréau (1810-92)
was not only a distinguished zoologist, occupying
himself mainly with certain groups of marine animals, but also
Professor of Anthropology at the Paris Museum of Natural History,
and undertook several voyages along the coasts of the
Mediterranean and the Atlantic in search of information. In 1867 he
published Rapport sur le progrès de l’Anthropologie, “which reduces
to a complete and intelligible system the abstruse and difficult, and,
to many, the incomprehensible science of anthropology, embracing
during his investigations a wide range of topics, and arranging
disjointed facts in due order, so as at once to evince their bearing
upon the subject.”[26] He published many other works, among them
Les Pygmées (1887), L’espèce humaine (1877), Histoire générale
des races humaines (1889), and, together with E. T. Hamy, the
famous Crania ethnica (1875-79). Professor F. Starr, in the preface
to his translation of The Pygmies (1895), says:—
26. Anth. Rev., 1869, p. 231.

A man of strong convictions and very conservative, de Quatrefages

was ever ready to hear the other side, and ever candid and kindly in
argument. He was one of the first to support the Society of Anthropology.
Those who know the story of the early days of that great association
understand what that means. When the claim for man’s antiquity was
generally derided, de Quatrefages championed the cause. A monogenist
[p. 53], a believer in the extreme antiquity of our race, he was never won
over by any of the proposed theories of evolution.... To the very end of a
long life our author lived happily and busily active among his books and
specimens.

Virchow. In Germany the greatest name is that of

Virchow. Rudolf Ludwig Karl Virchow (1821-1902)
had already gained fame in the medical world, especially with regard
to histology, pathology, and the study of epidemics, and was the
prime leader in the “Medizinische Reform” movement before he
began his valuable contributions to the science of Anthropology.
 His first anthropological writings were some papers on cretinism
(1851 and 1852), and from this date onwards his services to the
science can scarcely be over-estimated. Much of his energy was
also given to somatic anthropology, and in 1866 he started his
investigations into prehistoric archæology, combining scientific
method with spade-work.
In a notice of his work by Oscar Israel[27] (p. 656) we read:—
27. Smithsonian Report, 1902. Translated from the Deutsche Rundschau of
Dec., 1902.

Virchow devoted himself to ethnographic studies no less than to other

branches of anthropology, and here he became a center to which the
material streamed from all sides, and from which went forth suggestion,
criticism, and energetic assistance. This never-idle man did not disdain to
teach travelers schooled in other lines of investigation the anthropometric
methods; and, indeed, he found time for everything, and never left a
piece of work to others that he could possibly do himself. Thus, for
example, for ten years following its inception by him in 1876, he worked
up alone the data recorded in German schools as to the color of the eyes,
the hair, and the skin which has proved of such value for the knowledge
of the different branches of the German race.

Sergi. Professor Sergi at one time proposed to banish

measurements from craniology, and to rely solely
on observational methods. He has later modified his extreme
position, while, as a result of his crusade, he has induced most
anthropologists to pay more attention to the configuration of the
skull, and some of his descriptive terms have come into common
use.
Hagen’s Dr. Hagen relates the extreme specialisation
Criticism of into which craniologists were led:—
Craniometry.
A rage for skull measurements, vast, vigorous, and
heedless, set in on all sides, especially after Lucae had discovered and
perfected a method of accurately representing the irregular form of the
object studied. “More skulls” was henceforth the war-cry; the trunk,
extremities, soft tissues, skin and hair, might all go by the board, being
counted of no scientific value whatever. Anthropologists, or those who
aspired to the title, measured and delineated skulls; museums became
 veritable cities of skulls, and the reputation of a scientific traveller almost
stood or fell with the number of crania which he brought back with him.
After two decades of measuring and collecting ever greater quantities
of material from foreign lands, and from the so-called primitive or
aboriginal races, the inadequacy of Retzius’s method became apparent.
Far too many intermediate forms were met with, which it was found
absolutely impossible to classify by its means. In accordance with the
suggestion of the French anthropologist Broca, and of Welcker, Professor
of Anatomy at Halle, a third type, the so-called Mesocephalic form, was
interposed between the two forms recognised by Retzius. Even this did
not suffice, however. In the face of the infinite variety of form of the crania
now massed together, a variety only comparable to that of leaves in a
forest, this primitively simple scheme, with its four and finally six types,
failed through lack of elasticity. Then began complication extending ever
further and further. Attention was no longer confined to the length and
breadth, but also to the height of the cranium, high and low (or flat) skulls
—i.e., hypsicephalic and chamaecephalic varieties being recognised. The
facial part of the skull was examined not only from the side, with a view to
recording the straightness or obliquity of the profile, but also from the
front; and there were thus distinguished long, medium, and short faces,
and also broad and narrow facial types. The nasal skeleton, the palate,
the orbit, the teeth, and the mandible were investigated in turn, and at last
all the individual bones of the cranium and face, their irregularities of
outline, and their relations to one another, were subjected to the closest
examination and most subtle measurements, with instruments of extreme
delicacy of construction and ingenuity of design, till, finally, the trifling
number of five thousand measurements for every skull found an advocate
in the person of the Hungarian Professor V. Török (whereby the wealth of
detail obscured the main objects of study); while, on the other hand,
observers deviated into scientific jugglery, like that of the Italian Professor
Sergi, who contrived to recognise within the limits of a single small
archipelago, the D’Entrecasteaux group of islets near New Guinea, as
many as eleven cranial varieties, which were all distinguished by high-
sounding descriptive names, such as Lophocephalus
brachyclitometopus, etc.

Macalister’s The misuse of Craniometry is also described by

Criticism of Professor Alexander Macalister[28]:—
Craniometry.
28. Presidential Address to Section H., Brit. Ass., 1892.
 Despite all the labour that has been bestowed on the subject,
craniometric literature is at present as unsatisfactory as it is dull. Hitherto
observations have been concentrated on cranial measurements as
methods for the discrimination of the skulls of different races. Scores of
lines, arcs, chords, and indexes have been devised for this purpose, and
the diagnosis of skulls has been attempted by a process as mechanical
as that whereby we identify certain issues of postage-stamps by counting
the nicks in the margin. But there is underlying all these no unifying
hypothesis; so that when we, in our sesquipedalian jargon, describe an
Australian skull as microcephalic, phænozygous, tapeino-dolichocephalic,
prognathic, platyrhine, hypselopalatine, leptostaphyline, dolichuranic,
chamaeprosopic, and microseme, we are no nearer to the formulation of
any philosophic concept of the general principles which have led to the
assumption of these characters by the cranium in question, and we are
forced to echo the apostrophe of Von Török, “Vanity, thy name is
Craniology.”

It is significant that so many of the earlier craniologists recognised

that the really important problem before them was to gain a
knowledge of the size and relative proportions of the various regions
of the brain, this being a direct result of the phrenological studies
then so much in vogue. When phrenology became discredited, this
aspect of craniometry was largely neglected; but recently it has
exhibited signs of a healthy revival, and the inner surface of the
cranium is now regarded as more instructive than the outer.
Though for a time craniology was hailed as the magic formula by
which alone all ethnological tangles could be unravelled,
measurements of other parts of the body were not ignored by those
who recognised that no one measurement was sufficient to
determine racial affinities.
Anthropometry. Thus Anthropometry began to map out definite
lines of research, and detailed studies were made
of arms and legs, hands and feet, curves and angles, brains and
viscera; while, shorn of its extravagant claims, craniology took its
legitimate place as one in a series of bodily measurements. One of
the earliest workers in measurements other than that of the skull was
Charles White (1728-1813).