A Minimalist Ontology of the

Natural World

This book seeks to work out which commitments are minimally sufficient to
obtain an ontology of the natural world that matches all of today’s well-
established physical theories. We propose an ontology of the natural
world that is defined only by two axioms. (1) There are distance relations
that individuate simple objects—namely, matter points. (2) The matter
points are permanent, with the distances between them changing. Every-
thing else comes in as a means to represent the change in the distance rela-
tions in a manner that is both as simple and as informative as possible. The
book works this minimalist ontology out in philosophical as well as math-
ematical terms and shows how one can understand classical mechanics,
quantum field theory and relativistic physics on the basis of this ontology.
Along the way, we seek to achieve four subsidiary aims: (a) to make a case
for a holistic individuation of the basic objects (ontic structural realism);
(b) to work out a new version of Humeanism, dubbed Super-Humeanism,
without natural properties; (c) to set out an ontology of quantum physics
that is an alternative to quantum state realism and that avoids any ontolog-
ical dualism of particles and fields; and (d) to vindicate a relationalist ontol-
ogy based on point objects also in the domain of relativistic physics.

Michael Esfeld has been full professor of philosophy of science at the Uni-
versity of Lausanne since 2002. His last book with Routledge is Conserva-
tive Reductionism (with Christian Sachse) (2008).

Dirk-André Deckert is leader of the junior research group Interaction of

Light and Matter in the Mathematical Institute of Ludwig Maximilians
University Munich. He is the author of Electromagnetic Absorber Theory –
A Mathematical Study (2010).
Routledge Studies in the Philosophy of
Mathematics and Physics
Edited by:
Elaine Landry
University of California, Davis, USA
and
Dean Rickles
University of Sydney, Australia

For a full list of titles in this series, please visit www.routledge.com

1 The Logical Foundation of Scientific Theories

Languages, Structures, and Models
Décio Krause and Jonas R. B. Arenhart

2 Einstein, Tagore and the Nature of Reality

Edited by Partha Ghose

3 A Minimalist Ontology of the Natural World

Michael Esfeld and Dirk-André Deckert
A Minimalist Ontology of the
Natural World

Michael Esfeld and Dirk-André Deckert

with Dustin Lazarovici, Andrea Oldofredi
and Antonio Vassallo
First published 2018
by Routledge
711 Third Avenue, New York, NY 10017
and by Routledge
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
Routledge is an imprint of the Taylor & Francis Group, an informa business
© 2018 Taylor & Francis
The right of Michael Esfeld and Dirk-André Deckert to be identified as
author of this work has been asserted by them in accordance with sections
77 and 78 of the Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced
or utilised in any form or by any electronic, mechanical, or other means,
now known or hereafter invented, including photocopying and recording,
or in any information storage or retrieval system, without permission in
writing from the publishers.
Trademark Notice: Product or corporate names may be trademarks or
registered trademarks, and are used only for identification and explanation
without intent to infringe.
Library of Congress Cataloging-in-Publication Data
Names: Esfeld, Michael, author. | Deckert, Dirk-André, 1979– author. |
Lazarovici, Dustin, 1985– author. | Oldofredi, Andrea, author. |
Vassallo, Antonio, author.
Title: A minimalist ontology of the natural world / by Michael Esfeld and
Dirk-André Deckert, with Dustin Lazarovici, Andrea Oldofredi, and
Antonio Vassallo.
Description: 1 [edition]. | New York : Routledge, 2017. | Series: Routledge
studies in the philosophy of mathematics and physics ; 3 | Includes
bibliographical references and index.
Identifiers: LCCN 2017031890 | ISBN 9781138307308
(hardback : alk. paper)
Subjects: LCSH: Quantum theory—Philosophy. | Physics—Philosophy.
Classification: LCC QC174.12 .E84 2017 | DDC 530.12—dc23
LC record available at https://lccn.loc.gov/2017031890

ISBN: 978-1-138-30730-8 (hbk)

ISBN: 978-1-315-14227-2 (ebk)

Typeset in Sabon
by Apex CoVantage, LLC
Contents

1 Introduction 1

2 Matter points and their dynamics 17

3 Minimalist ontology and dynamical structure

in classical and quantum mechanics 59

4 A persistent particle ontology for quantum field theory 99

5 Relationalism for relativistic physics 131

Bibliography 167
Index 179
1 Introduction

1.1 The aim of the book

Physics, being the study of nature (physis in Greek), and metaphysics, being
the study of the most fundamental and general traits of being (cf. Aristotle,
Metaphysics IV, 1003a21-22), can together be expected to answer the fol-
lowing three questions:

1. What is matter? What is space and time?

2. What are the laws of nature?
3. How does matter in space and time, being subject to certain laws,
explain the observable phenomena?

The question of what is matter is connected with the question of what is

space and time: one’s view of matter has implications for one’s view
of space and time, and vice versa. The stance that one takes with respect
to laws of nature, by contrast, is largely independent of the stance that
one takes with respect to matter, space and time. The question about the
laws has two aspects: what their content is and what their status of being
is. Finally, the position that one endorses with respect to matter, space
and time as well as laws is probed by the way in which one elaborates on
how this position as a whole explains the observable phenomena.
These questions are philosophical as well as scientific. Even if it may
seem too optimistic to hope that one day answers to these questions can
be given in full completeness, the mere attempt of finding sensible
answers and weighing their successfulness is already a fruitful enterprise
in building a view of nature and its workings. This enterprise can best be
characterized as natural philosophy, involving physics and philosophy in
a seamless manner.
This enterprise is probably as old as mankind, and possible answers
range, say, from quantum mechanics to religion. This is due to the fact
that it is not even clear in which sense we would be satisfied with a potential
answer. Since human thought is tied to concepts, one of the first questions
should be which concepts should we employ in our formulation? Are
2 Introduction
“particles, fields, strings, genes, trees, souls, devils, angels, and gods” good
choices?
Obviously, the story we will attempt to tell in finding answers to the
aforementioned questions will depend heavily on this choice, and some of
the chosen words should better relate to some of the natural things that
exist independently of our thoughts and language; otherwise, the aforemen-
tioned enterprise will be futile. On the one hand, the larger the vocabulary
that has to be taken as basic, the shorter the potential stories, but also the
weaker is their explanatory power as the many concepts in the vocabulary
cannot be scrutinized. On the other hand, the smaller the basic vocabulary,
the more can be scrutinized, but also the longer will the stories be that we
have to tell. Consider how Jackson (1994) describes the enterprise of
metaphysics:

Metaphysics, we said, is about what there is and what it is like. But of

course it is concerned not with any old shopping list of what there is
and what it is like. Metaphysicians seek a comprehensive account of
some subject matter—the mind, the semantic, or, most ambitiously,
everything—in terms of a limited number of more or less basic
notions. In doing this they are following the good example of physicists.
The methodology is not that of letting a thousand flowers bloom but
rather that of making do with as meagre a diet as possible. . . .
Because the ingredients are limited, some putative features of the
world are not going to appear explicitly in the story. The question
then will be whether they, nevertheless, figure implicitly in the story.
Serious metaphysics is simultaneously discriminatory and putatively
complete, and the combination of these two facts means that there is
bound to be a whole range of putative features of our world up for
either elimination or location.
(Jackson (1994), p. 25)

The same goes for the scientific enterprise in general. The sciences seek a
good balance between the complexity of the vocabulary and the length of
their explanations that should of course depend on the natural phenomena
under study. For instance, while molecules may be well described by using
the notion of atoms, it would be a tiresome endeavour to explain the func-
tions of a cell in terms of only the atomistic vocabulary. Such switches in the
basic concepts that are employed are quite common, even in mathematics.
However, they often make it difficult to relate one well-established theory
to another one. Consider, to mention just one example, how difficult it is
to specify how and when quantum mechanics can be well approximated
by Newtonian mechanics.
Nevertheless, all these theories have a common goal—namely, to describe
what there is. As it would be outrageous to think that what there is depends
on our theories about the world, there must be a sense in which all these
 Introduction 3
vocabularies have a, in some sense minimal, common set relating to the
things that exist. One convincing example that such a common set can be
found comes from statistical mechanics, which is able to relate the concepts
“temperature, pressure, volume and entropy” used in thermodynamics to the
concept “particle motion” used in Newtonian mechanics. A common set of
concepts must contain good candidates for building an ontology of the
natural world. Although we humans may never be able to fully infer what
there really is, we can at least ask the following question. What is a
minimal set of entities that form an ontology that matches today’s well-
established physical theories? In answering this question, we will carefully
distinguish within the vocabulary used in these physical theories between,
on the one hand, the concepts that relate to what there is in the sense of
this minimal set of entities and, on the other hand, the concepts that make
up what we call the dynamical structure of a physical theory, providing an
economic means of telling the necessary scientific stories.
In this sense, using parsimony as the guide for ontology, the aim of this
book is to develop a minimalist answer to the aforementioned questions: we
seek to work out which commitments are minimally sufficient to obtain an
ontology of the natural world that is empirically adequate. Generally speak-
ing, the reason for employing parsimony as the guide for ontology is that
for any candidate entity stemming from science—or common sense, or
intuitions—we need an argument for why one should endorse an ontolog-
ical commitment to that entity. Its being part of what is minimally sufficient
to obtain an ontology of the natural world that is empirically adequate is
the best argument for an ontological commitment. It is an illusion to
think that by abandoning parsimony and enriching the ontology, one
achieves explanations that are deeper than those that a parsimonious ontol-
ogy can yield; one thereby runs only into artificial problems and impasses,
as we shall show in this book.
In this vein, we start from the idea that given a plurality of objects, there
has to be a certain type of relations in virtue of which these objects make up
a world. The minimalist hypothesis then is that these relations also individ-
uate the objects, thus paving the way for the claim that there is nothing
more to these objects than standing in these relations. The objects thus
are simple, having no parts or any other internal structure. When it
comes to the natural world, relations providing for extension—namely,
distances—are the first and foremost candidate for the type of relations
that fulfills this task. Distances connect unextended and thus point-sized
objects. If they individuate these objects, they provide for variation
within a configuration of point-sized objects, with each of these objects
being distinct from all the other ones by at least one distance relation
that it bears to another object. In virtue of standing in distance relations,
these objects then are matter points (recall the sparse Cartesian conception
of the natural world as res extensa). In order to achieve empirical adequacy,
we furthermore have to stipulate that these relations change. We thus
4 Introduction
propose an ontology of the natural world that is defined by the following
two axioms, and only by these two axioms:

(1) There are distance relations that individuate simple objects—namely,

matter points.
(2) The matter points are permanent, with the distances between them
changing.

We submit that these two axioms prescribe the diet that is as meagre as pos-
sible in accounting for the natural world, to come back to the citation from
Jackson earlier. Everything else then comes in as a means to represent the
change in the distance relations that actually occurs in a manner that is
both as simple and as informative as possible.
We thus take up atomism and seek to develop it into a minimalist ontol-
ogy of the natural world. Atomism is the oldest and most influential tradi-
tion in natural philosophy, going back to the pre-Socratic philosophers
Leucippus and Democritus. The latter is reported as maintaining that

substances infinite in number and indestructible, and moreover without

action or affection, travel scattered about in the void. When they
encounter each other, collide, or become entangled, collections of
them appear as water or fire, plant or man.
(fragment Diels-Kranz 68 A57, quoted from Graham (2010), p. 537)

In a similar vein, Newton writes at the end of the Opticks,

It seems probable to me, that God in the Beginning form’d Matter in

solid, massy, hard, impenetrable, moveable Particles . . . the Changes
of corporeal Things are to be placed only in the various Separations
and new Associations and motions of these permanent Particles.
(Newton (1952), question 31, p. 400)

The attractiveness of atomism is evident from these quotations: on the one

hand, it is a proposal for a fundamental ontology that is most parsimonious
and most general. On the other hand, it offers a clear and simple explanation
of the realm of our experience. Macroscopic objects are composed of indivis-
ible particles. All the differences between the macroscopic objects—at a time
as well as in time—are accounted for in terms of the spatial configuration of
these particles and its change, which is subject to certain laws. That is why
Feynman famously writes at the beginning of the Feynman lectures on
physics,

If, in some cataclysm, all of scientific knowledge were to be destroyed,

and only one sentence passed on to the next generations of creatures,
what statement would contain the most information in the fewest
 Introduction 5
words? I believe it is the atomic hypothesis (or the atomic fact, or what-
ever you wish to call it) that all things are made of atoms—little parti-
cles that move around in perpetual motion, attracting each other when
they are a little distance apart, but repelling upon being squeezed into
one another. In that one sentence, you will see, there is an enormous
amount of information about the world, if just a little imagination
and thinking are applied.
(Feynman et al. (1963), ch. 1–2)

Whereas atomism is a purely philosophical proposal in Leucippus and

Democritus, it is turned into a precise physical theory by Newton. Accord-
ingly, classical mechanics—and classical statistical physics—are usually
seen as the greatest triumph of atomism. Nonetheless, atomism loses
nothing of its attractiveness when it comes to quantum physics. In the
first place, also in this domain, all experimental evidence is evidence of dis-
crete objects (i.e. particles)—from dots on a display to traces in a cloud
chamber. Entities that are not particles—such as waves or fields—come in
as figuring in the explanation of the behaviour of the particles, but they
are not themselves part of the experimental evidence: an electric field is
probed by the motion of a test charge subject to it (such as the electron
in the wire); the double slit experiment is made apparent by sufficiently
many particles hitting on a screen, etc.
Moreover, in quantum as in classical physics, there are good arguments
to maintain that we need an account of macroscopic objects—such as, for
instance, a cat or an apparatus with a pointer that points in a certain
direction—in terms of matter being arranged in a certain manner in physi-
cal space. In order to achieve such an account, one cannot only endorse the
quantum state, which is defined on a very high-dimensional mathematical
space (namely, the configuration space of the universe), but one has to con-
ceive that state as being the state of matter arranged in three-dimensional
space or four-dimensional space-time: we take the arguments to this
effect going back to Bell (2004, ch. 7) and elaborated on notably by
Maudlin (2010, 2015) to be convincing. In brief, according to these argu-
ments, it is not sufficient that one can find something in the quantum
state of the universe that functionally corresponds to cat-like or pointer-
like behaviour; for there to be a cat, or a pointer, there have to be basic
objects that compose a cat, or a pointer, if they are arranged in the right
manner in physical space such that the evolution of such a configuration
of basic objects then amounts to the motion of a cat, or a pointer, in space.
In other words, then, what has become known as a primitive ontology of
matter distributed in three-dimensional space or four-dimensional space-time
is a necessary condition to avoid the famous measurement problem of
quantum physics.1 To turn that necessary condition into a sufficient one,
one has to formulate a dynamic for the primitive ontology that excludes
superpositions of matter in space so that there always is a definite
6 Introduction
configuration of matter and Schrödinger’s cat paradox, among others, is
avoided, but that dynamic has to include entanglement to account for the
non-local correlations that are manifest, for instance, in the Einstein Podolsky
Rosen (EPR) experiment (for recent elaborations, see, notably, Allori et al.
(2008), Belot (2012) and Esfeld et al. (2014)). This reasoning applies not
only to non-relativistic quantum mechanics but also to quantum field
theory (QFT)—as well as a future theory of quantum gravity—since any
quantum theory is plagued by the measurement problem (cf. Barrett (2014)).
Its success notwithstanding, atomism faces three major problems. Our
transformation of atomism into a truly minimalist ontology of the
natural world seeks to address these problems.
(1) In the first place, Democritus as well as Newton set out atomism in
terms of a dualism of matter on the one hand and space and time on the
other: matter is conceived as being inserted in an absolute background
space and as evolving in an absolute background time. However, the justi-
fication of absolute space and time is debatable, and their ontological status
remains unclear in classical atomism. The commitment to absolute space
and time implies in any case a commitment to a surplus structure,
because absolute space and time reach, in any case, far beyond the actual
configuration of matter, with its being doubtful whether one obtains a
gain in explanation through that commitment.
Since Leibniz, relationalism about space and time is put forward to avoid
that dualism. We follow this tradition. We set out an ontology of the
natural world in relationalist terms, being committed only to distance rela-
tions among the atoms and deriving time from the change in these relations
as the order of that change. We show how such an ontology can match both
classical and quantum mechanics and how it remains a viable option in rel-
ativistic physics.
(2) Nonetheless, even if the dualism of matter on the one hand and space
and time on the other is removed, the question of what characterizes the
atoms as material substances remains. Democritus and Newton conceive
them as being equipped with a few basic intrinsic properties—that is, prop-
erties that belong to each atom taken individually, independently of all the
other ones, thus making up an intrinsic essence of each atom. The paradig-
matic example is mass in Newtonian mechanics. However, also in Newto-
nian mechanics, both inertial and gravitational mass are introduced
through their dynamical role—namely, as a dynamical parameter that
couples the motions of the particles to one another, as was pointed out
by Mach (1919, p. 241) among others. The same goes for charge, energy,
etc. When it comes to quantum mechanics, despite first appearances, prop-
erties such as mass and charge cannot be conceived as intrinsic properties of
the particles, but are situated on the level of their quantum state as repre-
sented by the wave function. In sum, as soon as atomism is worked out
as a precise physical theory, it turns out that anything that one might
regard as constituting an intrinsic essence of the atoms is in fact a dynamical
 Introduction 7
parameter, expressing a dynamical relation that couples the motions of the
atoms to one another. Hence, the question is what is the essence of the
atoms qua material entities?
We bring in ontic structural realism to answer this question: instead of
having an intrinsic essence, the atoms have a structural one. Standing in dis-
tance relations is their essence. Hence, although we propose an ontology of
atomism, we draw on holism to work that ontology out: the atoms are
holistically individuated in terms of the distances among them. We conceive
the distance relations as establishing the order of what coexists, thereby
taking up Leibniz’s relationalist definition of space: these relations are
able to distinguish the objects, thereby satisfying the principle of the iden-
tity of indiscernibles. There thus is a configuration of objects that is consti-
tuted by distance relations: by individuating the atoms, the distance
relations provide for variation within a given configuration of matter.
Over and above variation making up for a configuration of objects,
there is change, which hence is change in the relations that constitute the
configuration—that is, the distances. We follow Leibniz in conceiving time
as the order of that change, with that order being unique and having a direc-
tion. Mass, charge, energy, spin, wave function, etc., then, are dynamical
parameters that a physical theory introduces in order to obtain a law that
describes that change in a simple and informative manner. These parameters
sort the atoms into different particle species on the basis of salient patterns in
their relative motion. Consequently, the atoms are not intrinsically protons,
electrons, neutrons, etc., but are so described because their motion exhibits
certain contingent regularities. In a nutshell, some atoms do not move elec-
tronwise because they are electrons, but they can be classified as electrons
because they move electronwise.
Indeed, there is no need to admit physical properties at all. Relations do
all the work. It is a misconception to set out ontic structural realism as a
stance that is directed against object-oriented metaphysics (cf. Ladyman
and Ross (2007) and French (2014)). Ontic structural realism is opposed
to the property-oriented metaphysics that has dominated philosophy from
Aristotle to today’s analytic metaphysics. Of course, if there are relations,
there are objects that stand in the relations, but standing in the relations
is all there is to these objects—the relations are their essence (cf. the mod-
erate ontic structural realism set out in Esfeld (2004), Esfeld and Lam
(2008, 2011)).
In order to understand what physics says about the natural world, one
should not be misled by the subject-predicate form of ordinary language
to buying into an ontology of substances that are characterized by intrinsic
properties. The simple and elegant story that physics tells from its begin-
nings to this day is one of discrete objects—call them “matter points”—
standing in distance relations and dynamical parameters capturing the
change in these relations. The dynamical parameters—that is, our attempts
to conceive a dynamical structure that describes this change in a simple,
8 Introduction
elegant and informative manner—vary from one theory to another. By con-
trast, the commitment to simple, discrete objects standing in distance rela-
tions whose evolution these dynamical parameters seek to track remains
constant, from Leucippus and Democritus via Newton to today’s
quantum theories.
That is to say, the argument from theory change and pessimistic meta-
induction against scientific realism does not apply to our minimalist ontology
and neither does the argument from underdetermination. What is underde-
termined by the given evidence is the best formulation of a dynamical struc-
ture, but not the ontology of relative particle positions and motion—that is,
matter points individuated by distance relations and change in these rela-
tions. Consequently, we draw within any given physical theory a distinction
between the primitive variables that directly refer to what there is in the
world (i.e. the primitive ontology)—namely, relative particle positions and
their change—and the laws that describe the evolution of the primitive var-
iables (i.e. the dynamical structure of a physical theory), with all the other
variables being nomological in the sense that they fall on the side of the
laws instead of directly referring to something that there is in the world
over and above the primitive variables.
Hence, in a nutshell, there are distance relations individuating the
matter points and thereby constituting a configuration of them, and
there is change in these relations. That is all. In terms of Humeanism,
the distance relations among the matter points and their change through-
out the entire history of the universe are the Humean mosaic, and every-
thing else in the natural world supervenes on that mosaic in the sense that
it comes in as a means to describe that change in a manner that is both as
simple and as informative as possible. The argument for this sparse ontol-
ogy is its simplicity together with its empirical adequacy: less won’t do for
an ontology of the natural world; bringing in more creates new drawbacks
instead of providing additional explanatory value. This sparse ontology
hence amounts to a radical ontological reductionism: everything in the
natural world reduces to distance relations among matter points and the
change in these relations, in the sense that these relations and their
change make true all the true propositions about the natural world. (By
the natural world, we mean the physical, spatially extended world. We
have no intention here to apply this reductionism to the mind, conscious-
ness and normativity).
(3) However, an ontology of distance relations among point particles
(matter points) and their change only may appear as hopelessly old-
fashioned and therefore obviously wrong. Newtonian mechanics is not all
of classical physics. There also is classical electrodynamics, apparently
refuting atomism as a complete ontology of the natural world by replacing
the commitment to atoms only with a commitment to both particles and
fields being the stuff out of which the world is made. But this dualism
runs into an impasse: as Feynman stresses in his Nobel lecture, the field
 Introduction 9
spreads out to infinity, being defined everywhere in space-time, thus also in
regions where there will never be any particles whose motion it influences
(Feynman (1966), pp. 699–700). What then is the field? Is it some sort of
stuff filling all of space-time? Is it a property of space-time points, albeit
not a geometrical one? In the former case, we are committed to an extrav-
agant ontology with there being stuff in addition to the particles every-
where, although all the experimental evidence that we have is one of
particle stuff only, whose relative distances change. In the latter case, we
are back to the commitment to an absolute background space as the
carrier of the field properties (see Field (1985), pp. 40–42), which again
exist everywhere in space-time—that is, also in regions where they never
manifest themselves by influencing the motion of particles.
Apart from these questions there are physical and mathematical issues
that have not been solved till today. The influence that a particle exerts on
the electromagnetic field reacts back on the particle, leading to an
infinite force at the position of the particle and, hence, ill-defined equa-
tions of motion. These difficulties can as yet only be dealt with in a
perturbative sense by means of a so-called renormalization of ill-defined
corrections to the non-interacting theory. These problems are inherited
by QFT and string theory (to date, there is no non-perturbative QFT
describing a non-trivial relativistic interaction in 3 + 1 dimensions due
to this problem). The philosophical problem of the ontological status
of fields becomes even worse in QFT: the quantum state, represented by
the wave function, can no longer be considered as a field in four-
dimensional space-time as in the case of quantum mechanics of one
particle, but takes as arguments elements of the space of configurations
of fields.
The dualism of a quantum state (wave function) that is a field in config-
uration space and particles in ordinary space is even more unconvincing
than the classical dualism of particles and fields since it is mysterious
how objects in different spaces could interact. This implausible dualism is
since Everett (1957) the motivation for recognizing only the quantum
state (see most recently Albert (2015), chs. 6 and 7, for this motivation).
However, as the ongoing discussion shows, it is by no means obvious
how one can account for our experience and make sense of the quantum
mechanical probability calculus on the basis of recognizing only the
quantum state.2 Nonetheless, note that the parsimonious ontology advo-
cated in this book has a much wider scope than what is known as the prim-
itive ontology approach to quantum mechanics. Notably, it is not tied to
three-dimensional space; the distance relations defining this ontology are
not wedded to a particular geometry. Hence, from the perspective of this
ontology, the main objection to a quantum ontology of only a quantum
state in configuration space is not that this approach lacks a primitive ontol-
ogy in terms of objects that are localized in three-dimensional space. The
main objection does not concern the dimension of the underlying space,
10 Introduction
but the uneconomical dualism of a substantival space and material entities
(such as a wave function field) defined on that space.
The alternative move to a quantum ontology in terms of the quantum state
only is to draw on the other main ontology of quantum physics that has also
been pursued since the 1950s—namely, Bohm’s theory, which lays stress on
particles in ordinary space and thereby avoids from the start any form of a
quantum measurement problem: there always is a definite spatial configura-
tion of matter (no superpositions), composing the macroscopic objects with
which we are familiar. More generally speaking, given that all the evidence
in classical as well as quantum physics is evidence of relative particle posi-
tions and motion, the most straightforward explanation is the one that
takes this evidence literally in terms of a particle ontology—although, of
course, particles qua matter points individuated by distance relations are
theoretical entities. They are introduced because they constitute the most
parsimonious ontology that enables the formulation of a dynamics such
that the way in which the particles move then explains the available particle
evidence in the sense that this ontology is, taking everything into account,
overall more coherent than its rival, richer ontologies. In particular, we
show how an ontology of permanent particles that move on continuous tra-
jectories according to a deterministic law is also able to explain the experi-
mental evidence in QFT, including notably the appearance of particle
creation and annihilation events.
However, as this ontology is set out in Bohm (1952a) (and earlier in de
Broglie (1928)), it directly runs into the unconvincing dualism of particles
in ordinary space and a wave function in configuration space. Basing our-
selves on the dominant contemporary formulation of this theory known
as Bohmian mechanics by Dürr et al. (2013b) and their argument to
regard the wave function as nomological, we tackle this problem by delet-
ing the commitment to the wave function as a physical entity in addition to
and on a par with the particles: the wave function or quantum state is
nothing but a dynamical parameter introduced in the theory to capture
the evolution of the particles’ positions—that is, the change in their dis-
tances. In a nutshell, instead of overcoming the implausible dualism of
matter in ordinary space and quantum state in configuration space by rec-
ognizing only the quantum state, we submit that an ontology of matter in
physical space only, with that space being nothing but the distance relations
among matter points and the change in these relations, is sufficient and
indeed the best available option to account for the experimental evidence
in quantum as well as classical physics.
The central aim of this book, accordingly, is to elaborate on an ontology of
the natural world that is most parsimonious while being empirically ade-
quate: a commitment to less than distance relations among matter points
that are individuated by these very relations and change in these relations
would be insufficient. A commitment to more than that leads to trouble
instead of yielding deeper explanations. Along the way, we seek to achieve
 Introduction 11
four subsidiary aims: (a) to make a case for a holistic individuation of the basic
objects, combining ontic structural realism with relationalism about space;
(b) to work out a new version of Humeanism, dubbed Super-Humeanism,
that covers the whole of physics by doing without natural properties, taking
the Humean mosaic to consist only in basic relations among point objects
(i.e. distance relations among matter points) and the change in these relations;
(c) to set out an ontology of quantum physics, including QFT, that is a
complete alternative to Everett-style quantum state realism and that in
general avoids any dualism of particles and fields by relegating the latter
(including the quantum state) to the dynamical structure of a physical
theory whose only function is to capture the change in the distance relations
among the particles in ordinary space; and (d) to vindicate a relationalist
ontology based on distances also in the domain of relativistic physics.

1.2 The content of the book and its methodology

Apart from the introduction, the book has four chapters: in the second
chapter, we develop the minimalist ontology of matter points and their
dynamics in general philosophical as well as mathematical terms, indepen-
dently of any particular physical theory. We thereby set out a new argument
for relationalism about space and time, ontic structural realism about the
basic objects and Humeanism about the physical laws and the dynamical
parameters figuring in them. In Chapter 3, we show how this ontology
fits both classical mechanics of gravitation and quantum mechanics, as
long as the latter includes a primitive ontology of matter distributed in ordi-
nary space, as in Bohmian mechanics. We point out the similarities between
these two theories as well as their differences and explain how in both cases
one gets from an ontology and a dynamics defined for the universe as a
whole to a description of subsystems of the universe and a dynamics for
them.
Chapters 4 and 5 then expand on this ontology with respect to today’s
most advanced, established theoretical physics. In Chapter 4, we go into
QFT: in terms of the Dirac sea model for electrons, we show how an
ontology of persisting particles that move according to a deterministic
law can explain the characteristic phenomena of QFT—in particular the
appearance of particle creation and annihilation—and ground its empiri-
cal predictions. In Chapter 5, we consider relativistic physics, sketching
out a version of relationalism about space and time combined with
Super-Humeanism also for general relativistic physics on the basis of an
ontology of matter qua matter points that are individuated by the distance
relations in which they stand and characterized by the change in these
relations.
Our methodology is the one of natural philosophy or naturalized meta-
physics, treating physics and metaphysics as going together in an inseparable,
12 Introduction
seamless manner. We adopt an empiricist attitude in insisting on the fact that
all the experimental evidence consists in relative particle positions and
motion, working out a minimally sufficient ontology that accounts for this
evidence and arguing that any additional ontological commitment comes
with new drawbacks instead of providing a deeper explanation. However,
this book is opposed to a neo-positivist metaphysics as set out notably in
Ladyman and Ross (2007), in the sense of the idea that there is a one-way
road leading from physics to metaphysics, suggesting that one obtains meta-
physical claims from the mathematical structures of physical theories and
that only metaphysical claims thus established are valid.3 Consequently, we
reject Quine’s criterion of ontology in terms of, in brief, formulating a
theory in first-order logic and being committed to all the entities that the
theory thus formulated quantifies over (see Quine (1948)).
In general, on the one hand, this book is directed against the idea that
any metaphysical claims—be they positive, seeking to tell us what there
is, be they negative, seeking to rule out certain ontological options—can
be established directly on the basis of the mathematical structure of physical
theories. One cannot infer ontology directly from the mathematical struc-
ture of a physical theory, as, for instance, the futile attempts of making
ontological claims about trajectories, individuals, probabilities, etc., as
well as setting up various no-go theorems on the basis of the operator for-
malism of textbook quantum mechanics show (see Laudisa (2014) for a
recent criticism). But also a theory that goes beyond textbook quantum
mechanics, such as Bohmian mechanics, does not simply wear the ontology
on its sleeves, as the ongoing debate about the status of the wave function in
this theory illustrates. On the other hand, the a priori, armchair metaphys-
ics based on conceptual analysis that is widespread in today’s analytic phi-
losophy cannot on its own produce knowledge of the world, for it lacks the
means to cast ontological claims in a precise physical framework that is able
to predict and explain empirical phenomena. That is why we promote
natural philosophy—metaphysics becoming physics, or physics being
done on the basis of first ontological principles.
We advocate drawing a distinction between what simply exists according
to a given theory—that is, the primitive ontology—and what is introduced
in the theory through its functional (or dynamical) role for what there
simply is—that is, the dynamical structure of the theory, as expressed in
its mathematical formalism. Thus, a wave function in quantum mechanics
as well as parameters such as mass and charge in classical mechanics pre-
suppose a configuration of objects to which they are ascribed because
they are introduced in terms of their functional or dynamical role for the
evolution of these objects; by contrast, these objects—that is, the point par-
ticles—do not have any further function in the theory. They are simply
there.
In a second step, we move from primitive ontology within a theory to
ontology tout court. Natural philosophy seeks for an ontology of the
 Introduction 13
natural world that is not relative to particular physical theories. To our
mind, it is inappropriate to speak of the ontology of this or that physical
theory. Ontology is about what there is. It goes without saying that our
access to what there is comes through the representations that we conceive
in terms of physical theories. But this does not imply that ontology is rela-
tive to particular theories, because it is a mistake to seek to read ontology
off from the mathematical formalisms of physical theories. The measure for
ontology is simplicity (or parsimony) together with empirical adequacy. Of
course, parsimony is an a priori criterion of armchair metaphysics (cf.
Jackson (1994), p. 25). However, there is nothing wrong with using that
criterion as long as one can show in detail how it is possible to get from
the ontology formulated on that basis to the established physical theories
and, moreover, thereby solve the problems of interpretation of these theo-
ries (such as the quantum measurement problem). The argument for parsi-
mony then is that admitting more to the ontology than what is minimally
sufficient for the purpose of empirical adequacy does not amount to a
gain in explanation. It only leads to drawbacks (such as, e.g., the ones
pointed out by Leibnizians against absolute space), or creates artificial
problems (such as, e.g., the questions of how a particle can reach out to
other particles and accelerate them in virtue of properties that are intrinsic
to it, or how a wave function, being defined as a field on configuration
space, can push particles around in physical space).
Obviously, the measure for representation also is simplicity together
with empirical adequacy. However, simplicity in ontology and simplicity
in representation pull in opposite directions. That is why the neo-positivist
trend of using the mathematical structure of a physical theory as the guide-
line for ontology is wrong-headed. Employing only the concepts that
describe what there is on the simplest ontology (i.e. matter points individ-
uated by distance relations), the description of the evolution of the config-
uration of matter would not be simple at all, since one could not do much
better than listing all the change that actually occurs. Reading one’s onto-
logical commitments off from the simplest description—such as, e.g., New-
tonian mechanics—the ontology would not be simple at all: it would in this
case be committed to absolute space and time, to momenta, gravitational
masses, forces, etc.
To put this crucial issue differently, from the epistemological perspective,
in metaphysics as well as in physics or mathematics, we have to start from a
few basic notions taken as primitive. We can elucidate what these notions
mean—as we shall do with the notion of distances individuating point parti-
cles in section 2.1—but we cannot trace them back to other notions. These
notions, then, make up our proposal for the primitive ontology: what there
simply is. It is in principle possible to do the whole of physics with just the
notions of distances individuating point particles and change of these dis-
tances, but this would be uneconomical: no simple dynamical equation cap-
turing that change could be thus achieved. The reason is that we can define a
14 Introduction
configuration of matter and its change with these basic notions, but there is
nothing in a configuration of matter given in terms of the relative distances
between the matter points that provides information about how the change
of that configuration occurs. It is therefore reasonable to introduce further
notions that provide such information by being formulated in terms of their
functional role for the change in the configuration of matter so that the rep-
resentation of change becomes more simple without losing information.
These notions, which make up the dynamical structure of a physical
theory, vary as we acquire more data and make progress in constructing a
system that strikes as good as we can achieve a balance between being
simple and being informative about the change in the world.
However, it is a misunderstanding to think that by introducing further
notions one subscribes to ontological commitments that go beyond the
ones given in terms of the basic notions. This, again, is an instance of
the neo-positivist fallacy of inferring ontology from the mathematical
structure of a theory. Here is the reason Humeanism enters into this
book—namely, as a strategy to maintain scientific realism without build-
ing ontological commitments on the representational means that physical
theories employ. If we take the basic notions to define what is known as
the Humean mosaic, we can conceive all the further notions that we
need to achieve a simple theory—namely, the geometrical and the dynami-
cal ones—as being the means to obtain a representation of that mosaic
that is both simple and informative. We thereby stay neutral with
respect to the issue of whether or not there is something modal in the
world—a logos in the cosmos so to speak. We only claim that it is mis-
guided to conceive the varying means that we set up to conceptualize
the change that we perceive in a simple and informative manner as reveal-
ing that logos. Doing so only leads to artificial problems, such as the ones
involved in the surplus structure that comes with the commitment to abso-
lute space and time, or the ones showing up in the question of how a wave
function can push particles around in physical space.

1.3 Acknowledgements
The work on writing this book began when Michael Esfeld spent the aca-
demic year 2014–15 in Munich, having received the research award of
the Alexander von Humboldt Foundation. Michael Esfeld is grateful to
Stephan Hartmann and the Munich Center for Mathematical Philosophy
for their hospitality, as well as to the Humboldt Foundation whose
award and continuing support made this work possible. The overall frame-
work of this book was conceived by Michael Esfeld and Dirk-André
Deckert, with Michael Esfeld taking the lead for the philosophical argu-
ment and Dirk-André Deckert for the physical one. Chapter 3, section 2,
on identity-based Bohmian mechanics goes back to work by Dustin
 Introduction 15
Lazarovici, as do Chapter 3, section 4, and Chapter 5, section 2. Andrea
Oldofredi joined in for collaboration on Chapter 4 and Chapter 3,
section 4. Antonio Vassallo was the expert for setting out the relationalism
for both classical mechanics and general relativistic physics. These collabo-
rators are not to be held responsible for the overall framework of this book.
Dirk-André Deckert’s work was funded by the junior research group grant
Interaction between Light and Matter of the Elite Network of Bavaria.
Andrea Oldofredi’s and Antonio Vassallo’s contributions were supported
by the Swiss National Science Foundation, grant no. 105212-149650.
Dustin Lazarovici acknowledges funding from the Cogito foundation
(grant no. 15-106-R) as well as the Alexander von Humboldt Foundation
(Feodor Lynen research fellowship).
The book draws on material from the following papers: “Quantum
Humeanism, or: Physicalism without Properties”, by Michael Esfeld, pub-
lished in the Philosophical Quarterly 64 (2014), pp. 453–470; “The Ontol-
ogy of Bohmian Mechanics”, co-authored by Michael Esfeld, Dustin
Lazarovici, Mario Hubert and Detlef Dürr, published in the British Journal
for the Philosophy of Science 65 (2014), pp. 773–796; “From the Universe
to Subsystems: Why Quantum Mechanics Appears More Stochastic than
Classical Mechanics”, co-authored by Andrea Oldofredi, Dustin Lazarovici,
Dirk-André Deckert and Michael Esfeld, published in Fluctuations and Noise
Letters 15 (2016), Special issue Quantum and classical frontiers of noise, pp.
164002: 1–16; “The Physics and Metaphysics of Primitive Stuff”, co-
authored by Michael Esfeld, Dustin Lazarovici, Vincent Lam and Mario
Hubert, published in the British Journal for the Philosophy of Science 68
(2017), pp. 133–161; “What Is Matter? The Fundamental Ontology of
Atomism and Structural Realism”, co-authored by Michael Esfeld, Dirk-
André Deckert and Andrea Oldofredi, forthcoming in Anna Ijjas and Barry
Loewer (eds.): A Guide to the Philosophy of Cosmology, Oxford University
Press; “Relationalism about Mechanics Based on a Minimalist Ontology of
Matter”, co-authored by Antonio Vassallo, Dirk-André Deckert and
Michael Esfeld, published in the European Journal for Philosophy of
Science 7 (2017), pp. 299–318; “Leibnizian Relationalism for General Rela-
tivistic Physics”, co-authored by Antonio Vassallo and Michael Esfeld, pub-
lished in Studies in History and Philosophy of Modern Physics 55 (2016), pp.
101–107; “A Persistent Particle Ontology for QFT in Terms of the Dirac
Sea”, co-authored by Dirk-André Deckert, Michael Esfeld and Andrea Old-
ofredi, forthcoming in the British Journal for the Philosophy of Science, DOI
10.1093/bjps/axx018; “A Proposal for a Minimalist Ontology”, by Michael
Esfeld, forthcoming in Synthese, DOI: 10.1007/s11229-017-1426-8. We are
grateful to the publishers of these papers to leave us the right to use parts of
the material of these papers as a basis for this book.
We would like to thank the participants of the summer schools on the phi-
losophy of physics in Saig (Black Forest, Germany) in July 2014, July 2015,
July 2016 and July 2017 for discussions, in particular, David Albert,
16 Introduction
Jeff Barrett, Jeremy Butterfield, Detlef Dürr, Matthias Egg, Tiziano Ferrando,
Shelly Goldstein, Mario Hubert, Vincent Lam, Anna Marmodoro, Tim
Maudlin, Ward Struyve and Nino Zanghı̀, as well as Peter Pickl, Christian
Sachse, Barry Loewer, Harvey Brown, Oliver Pooley and Hans Christian
Öttinger. Last, but not least, we are grateful to three anonymous referees
of the draft of the book for their helpful comments and to Andrew Wecken-
mann for the professional handling of the book manuscript.

Notes
1. The term “primitive ontology” goes back to Dürr et al. (2013b, ch. 2), originally
published 1992. A forerunner of this notion can be found in Mundy (1989,
p. 46). Cf. also Bell’s notion of “local beables” in Bell (2004, ch. 7), originally
published 1975.
2. See Wallace (2012) as well as Wilson (2013a), Albert (2015), chs. 6 and 7, and
Ney (2015) for proposals in this sense. See, notably, Maudlin (2010, 2014, 2015)
as well as Dawid and Thébault (2014) and Dizadji-Bahmani (2015) for criticism.
3. See Ney (2012) for a discussion of neo-positivist metaphysics as well as a plea for
a moderate version of this project, in contrast to the one put forward by
Ladyman and Ross (2007). See, furthermore, Callender (2011), Morganti
(2013) and Guay and Pradeu (2017) for more balanced views of naturalized
metaphysics, as well as the papers in Ross et al. (2013), in particular, Chakra-
vartty (2013), Humphreys (2013) and Wilson (2013b).
2 Matter points and their
dynamics

2.1 Matter points individuated through distance relations

Let us come back to the quotations from Democritus and Newton in the
introduction. As is evident from these quotations, they both consider
atoms as the smallest, indivisible objects that compose all the other
objects and whose spatial configuration explains the observable features
of the macroscopic objects with which we are familiar and their change.
But what then are the atoms? Both Democritus and Newton are usually
taken to conceive them as being characterized by a few basic intrinsic
properties—that is, properties that each atom has in itself, independent of
its relationship to other atoms (see Langton and Lewis (1998) on intrinsic
properties, and Hoffmann-Kolss (2010), part 1, for an extensive discus-
sion). The paradigmatic example of such a property in classical physics is
mass in Newtonian mechanics, with a value of mass attributed to each par-
ticle taken individually.
However, it is doubtful whether this view holds up to scrutiny. Also in
Newtonian mechanics, both inertial and gravitational mass are introduced
through their dynamical role—namely, as dynamical parameters that
couple the motions of the particles to one another—as was pointed out
by Mach (1919, p. 241) among others: in short, mass tells us something
about how the particles move. Hence, mass is a parameter that expresses
a dynamical relation between the atoms. The same goes for charge. Conse-
quently, in expressing a dynamical relation, parameters such as mass and
charge presuppose objects given in terms of their spatial location to
which they are applied.
This argument against intrinsic properties is strengthened by taking
quantum mechanics into consideration. Over and above mass and
charge being parameters that express a dynamical relation between the
particles, quantum mechanics contradicts the view that these parameters
are situated at the particle locations and belong to the particles taken indi-
vidually. In the first place, the wave function is the central dynamical
parameter of quantum mechanics. Due to its being entangled, it is in
general not possible to attribute a wave function or quantum state to
18 Matter points and their dynamics
the particles taken individually, but only to their configuration as a whole.
Moreover, even in cases in which a many-particle wave function is separ-
able, quantities like mass and charge are not located at the position of the
particles, but spread over the support of the wave function, as far as they
figure in particle interactions. Experimental considerations involving
interference phenomena—for instance, in the context of the Aharonov-
Bohm effect and of certain interferometry experiments—show, in brief,
that mass and charge are effective at all the possible particle locations
that the quantum state admits (see, e.g., Brown et al. 1995, 1996 and ref-
erences therein; cf. most recently Pylkkänen et al. 2015). Consequently,
also dynamical parameters that always have a definite numerical value
in quantum mechanics—such as mass and charge—cannot be taken to
refer to intrinsic properties of the particles. In general, anything that
one may be inclined to regard as physical properties of the particles
over and above their spatial configuration is not a property that can be
considered as belonging to the particles taken individually, but is situated
on the level of their quantum state. The quantum state, in turn, is defined
on configuration space—that is, the mathematical space each of whose
points represents a possible configuration of matter in physical space.
Hence, again, the quantum state presupposes a configuration of objects
given in terms of their location in physical space to which it is applied.
Nonetheless, one may envisage upholding the commitment to intrinsic
properties by maintaining that parameters such as mass and charge are
introduced in physics through the dynamical role that they play for the
motion of the particles, but that the description in terms of a dynamical
role refers to an underlying intrinsic property. Thus, Jackson (1998)
writes,

When physicists tell us about the properties they take to be fundamen-

tal, they tell us what these properties do. It does not follow from this
that the fundamental properties of current physics, or of completed
physics, are causal cum relational ones. It may be that our terms for
the fundamental properties pick out the properties they do via the
causal relations the properties enter into, but that at least some of
the properties so picked out are intrinsic. They have, as we might put
it, relational names but intrinsic essences. One way to block this
result is holding that the nature of everything is relational cum
causal, which makes a mystery of what it is that stands in the causal
relations.
(Jackson (1998), pp. 23–24)

The argument hence is that at least some of the dynamical parameters that a
physical theory introduces have to be conceived as referring to intrinsic
properties of the objects to which these parameters are attributed,
because, otherwise, it would be a mystery as to what the objects are,
 Matter points and their dynamics 19
whose dynamics is described by means of parameters such as mass or
charge. This is an a priori argument in the vein of Aristotelian metaphysics
according to which objects or substances must have an intrinsic essence or
form (eidos). If physics does not deliver intrinsic properties, its descriptions
in terms of dynamical relations among objects should nevertheless be con-
ceived as referring to such properties. However, the problem with this argu-
ment is that, as mentioned earlier, dynamical parameters such as mass and
charge presuppose objects given in terms of their spatial location to which
they are applied so that these parameters are unsuitable to take the position
of something that refers to an intrinsic essence of these objects. Moreover, it
is a contingent matter of fact that in classical physics, such parameters are
attributed to the objects taken individually; in quantum physics, they are
situated on the level of their configuration. It is, therefore, worthwhile to
look for something else than intrinsic properties that can do the job of indi-
viduating the objects that stand in the dynamical relations without falling
into any sort of mystery.
Doing so does not automatically imply that one has to abandon the view
of an intrinsic nature of the physical objects. Apart from intrinsic proper-
ties, two other main philosophical possibilities to uphold this view are
available. The one possibility is to say that each particle has a primitive
thisness—that is, a primitive individual essence known as a haecceity.
This is not a qualitative feature: it is not constituted by any properties. It
is the fact of being a certain individual object, whereby the fact of being
that object is not tied to any characteristic features of the object. This is,
therefore, a very problematic metaphysical stance, which has no basis in
physics. It implies that there are ontological differences (permutation of
haecceities) that do not make any physical difference.
The other possibility is to maintain that the particles are bare substrata,
having a primitive stuff-essence (this view is usually traced back to Locke
(1690), book II, ch. XXIII, § 2). By contrast to haecceities, this is not an
individual essence, but a kind essence. Nonetheless, it remains unclear
what a primitive stuff-essence could be, since it could not be characterized
by any properties. In brief, thus, neither the idea of a primitive thisness nor
the one of bare substrata is a convincing way out to uphold the view of the
basic physical objects being constituted by an intrinsic essence, if physics
blocks the Aristotelian way of taking the basic objects to be endowed
with intrinsic properties. To paraphrase Jackson, a primitive thisness or a
bare substratum make a mystery of what the objects are that stand in the
dynamical relations described by parameters such as mass, charge and
wave function.
What we then arrive at is a commitment to naked particles so to speak—
that is to say, all there is to the particles is their spatial location. In fact,
Democritus and Newton conceive the atoms as being inserted into an abso-
lute background space, which is three-dimensional and Euclidean.
However, the commitment to an absolute space is at least as problematic
20 Matter points and their dynamics
as the commitment to intrinsic physical properties of the particles. There
are three central objections against this commitment:

1. Differences that do not make a physical difference: As Leibniz points

out in his famous objections to Newton’s substantivalism, there are
many different possibilities to place or to transform the whole configu-
ration of matter in an absolute space that leave the spatial relations
among the material objects unchanged so that there is no physical dif-
ference between them (see, notably, Leibniz’s third letter, §§ 5–6, and
fourth letter, § 15, in Gerhardt (1890), pp. 363–364, 373–374,
English translation Ariew (2000)).
2. Infinite expansion: Assume, as is well supported by physics, that the
configuration of matter consists in finitely many discrete objects, such
as point particles. If that configuration is embedded in an absolute
space, then that space will stretch out to infinity, unless an arbitrary
boundary is imposed (at least in a classical setting, since in general rel-
ativity the global matter distribution might determine a compact geom-
etry); in any case, it will stretch out far beyond the actual particle
configuration. However, all the experimental evidence is one of relative
particle positions and change of particle positions—that is, motion.
Thus, space is needed in physics only to describe the configuration of
matter and, notably, the change in that configuration. Consequently,
subscribing to the existence of an absolute space in which that config-
uration is embedded amounts to inflating the ontology.
3. Problem of what fills space: If one endorses a dualism of there being
both matter and space as distinct entities, one faces the question of
what the matter is that fills space (see Blackburn (1990)). What
makes up the difference between a point of space being occupied and
its being empty? However, none of the three possible answers of intrin-
sic properties, a primitive thisness or a bare substratum is convincing.
The failure to come up with a cogent answer to this question provokes
the objection that matter is inscrutable; this objection can then be
turned into an argument against the very existence of matter—that is,
an argument for idealism (see notably Foster (1982), ch. 4; cf. also Rob-
inson (1982), ch. 7).

The impasse of not being able to put forward a characterization of matter

that stands up to scrutiny arises only if one presupposes a dualism of
an absolute space and matter as that which fills space. Relationalism, by
contrast, is the view that there are only spatial relations—that is, distance
relations—among physical objects, but no space in which these relations
are embedded. Relationalism thereby avoids the three mentioned problems:
all transformations that leave the distance relations among the material
objects unchanged are mathematical representations of one and the same
physical configuration, so that there are no differences in the ontology
 Matter points and their dynamics 21
that do not make a physical difference. Furthermore, there is no space
beyond the actual configuration of matter. Most importantly for our pur-
poses, if there are only spatial relations among material objects, but no
points of space, then these very relations are available to answer the ques-
tion of what characterizes matter. In the following, we prefer the term “dis-
tance relations” to the term “spatial relations” in order to avoid the
connotation of an underlying space.
To recap, our claim is that atomism, if set out in terms of point particles
being inserted into an absolute space, fails to achieve the aim of being a par-
simonious ontology. The consequence of this failure is that atomism, thus
conceived, is unable to come up with a cogent answer to the question of
what the atoms are. To answer this question, one has to abandon points of
space and retain only point particles (matter points), with these point particles
standing in distance relations that individuate them. The point particles then
are simple in the sense that they have no parts or any other internal structure.
In a nutshell, to accomplish the task of setting out atomism as a minimalist
ontology of the natural world, only the following two axioms are required:

Axiom 1 There are distance relations that individuate simple objects—

namely, matter points.
Axiom 2 The matter points are permanent, with the distances between them
changing.

We submit that these two axioms are minimally sufficient to formulate an

ontology of the physical world that is empirically adequate, given that all
the empirical evidence comes down to relative particle positions and
change of these positions. Hence, relationalism about space and time is
motivated for us by the search for an ontology of matter that is most par-
simonious while being empirically adequate.
Why do we single out the distance relations? Without acknowledging a
plurality of objects in some sense—and be it so-called thin objects without
intrinsic properties (see French (2010))—it is difficult to see, to say the least,
how empirical adequacy could be achieved. If there is a plurality of objects,
there has to be a certain type of relations in virtue of which these objects
make up a configuration that then is the world. Generally speaking, one
can conceive different types of relations making up different sorts of
worlds. For instance, one may imagine thinking relations that individuate
mental substances making up a world of minds. Lewis’s hypothetical
basic relations of like-chargedness and opposite-chargedness, by contrast,
would not pass the test, since, as Lewis notes himself, these relations fail
to individuate the objects that stand in them as soon as there are at least
three objects (Lewis (1986a), p. 77).
When it comes to the natural world, the issue is relations that qualify as
providing for extension. That is the reason to single out distance relations.
In a future theory of quantum gravity, these relations may be conceived in
22 Matter points and their dynamics
a different manner than in our current and past physical theories. Nonethe-
less, we submit that relations providing for extension—namely, distances—
are, given the state of the art in both physics and philosophy, the first
choice for an ontology of the natural world that is to be empirically adequate.
Change in these relations then is sufficient to obtain empirical adequacy.
That is the reason to pose the two aforementioned axioms, and only these
two ones. Of course, all these claims are fallible. To repeat, we are after
minimal sufficiency; we make no claim about these axioms being necessary
or a priori.
To convey what axiom 1 means, we have to choose a representation. Let
us consider a universe consisting of a finite number of N 2 N matter points.
Taking the number of matter points to be finite is sufficient for empirical
adequacy and will make the following discussion much easier. In order to
obtain an ontology that is minimally sufficient for empirical adequacy,
we assume that the set O of all possible configurations of distance relations
between N 2 N matter points can be represented as follows:

Definition 1 Let M ¼ f1; 2; :::; Ng and E ¼ fði; jÞ j i; j 2 M; i 6¼ jg. The set O

comprises elements D ¼ ðDij Þði;jÞ2E that can be represented by numerical
assignments fulfilling the following requirements:

i. D ¼ ðDij Þði;jÞ2E is a N2 ðN  1Þ-tuple of positive values Dij 2 Rþ for each

ði; jÞ 2 E.
ii. For all ði; jÞ 2 E one has Δij = Δji.
iii. For all i; j; k 2 M, it is the case that Δij  Δik + Δkj.
iv. For all i; j 2 M, if i 6¼ j, then fDik j k 2 M; k 6¼ ig 6¼ fDjk j k 2
M; k 6¼ jg.

Requirements (i) and (ii) state that the distance relation is, in mathematical
terms, irreflexive, symmetric and connex (meaning that all matter points in
a given configuration must be related with one another). Requirement (iii) is
the triangle inequality in virtue of which the relation in question is a dis-
tance relation. Requirement (iv) states that the distance relations individu-
ate the matter points: if matter point i is not identical with j, then the two
sets that list all the distance relations in which these points stand with
respect to all the other points in a configuration must differ in at least
one such relation. It is such differences in the way in which i and j relate
with the other points in the configuration that make it that i and j are dif-
ferent matter points. Consequently, by means of (iv), we avoid having to
accept the plurality of matter points as a primitive fact, which would
imply that the matter points are bare particulars or bare substrata.
Instead, they are individuated by the distance relations.
Formulating these requirements in terms of numerical assignments is a
means to express the features in virtue of which a relation is a distance rela-
tion. A minimal requirement in that respect is the triangle inequality (iii).
 Matter points and their dynamics 23
Furthermore, the ratios between the distances are part of the ontology (so
that the distances can individuate the matter points). By contrast, the numer-
ical assignments do not belong to the ontology, let alone the notion of abso-
lute scale that comes with them; they are just introduced for representational
purposes. Consequently, the fact that the values assignable to Δij are real
numbers does not smuggle in any infinity in this ontology: there is a finite
number of N matter points and, hence, finitely many distance relations. Con-
trary to a well known and long debated argument by Field (1980), in using
real numbers, we are not committed to mathematical Platonism in having to
subscribe to an ontological commitment to real numbers. In general, in using
something as a representational means—numbers, space and time, dynamical
parameters, laws, etc.—one is not committed to endorsing that means in
one’s ontology. Doing so requires an argument. The minimalist in ontology
sets out to rebut any such argument by claiming that (a) a minimalist ontol-
ogy as given by the two aforementioned axioms is sufficient as truth-maker
for all the true propositions about the world and that (b) additional ontolog-
ical commitments do not amount to a gain in explanation, but to commit-
ments to surplus structures that entail drawbacks.
Nonetheless, by introducing a labelling M of the matter points, this def-
inition can be taken to suggest that their numerical plurality is a primitive
fact. But this is just an artifact of the set-theoretical language. The elements
of a set M qua set-theoretical objects have to be numerically distinct for
M to be a well-defined set of N objects. However, the referents of this
formalism—the matter points qua physical objects—are individuated by
the distance relations given by Δ, so that these relations account for their
numerical plurality.
To emphasize the permutation invariance of the matter points in the for-
malism, it is possible to make the earlier definition of O independent of the
labelling by introducing the following equivalence relation:

Definition 2 Take Δ, Δ0 2 O, and consider SN as the set of all possible per-

mutations of elements of M. We define Δ ’ Δ0 if and only if there is a per-
mutation s 2 SN and a constant c 2 Rþ such that for all ði; jÞ 2 E it is the
case that D0ij ¼ cDsðiÞsðjÞ .
The set


~ ¼ O= ’:¼ ½D j D 2 O ;
O ½D’ ¼ fD0 2 O j D0 ’ Dg ð2:1Þ
’

then comprises all possible configurations of distance relations indepen-

dently of a labelling of the matter points and a scaling. The latter implies
that the relevant quantities are not the absolute values of the distance rela-
tions Δij, but only the ratios Δij/Δkl.
One way to envision an element of ½D’ 2 O ~ is by a representative Δ 2 O
that can be viewed as a coloured graph GðDÞ ¼ ðM; E; DÞ in which M are
the nodes, E are the edges, and to each edge ði; jÞ 2 E the colour Δij is
24 Matter points and their dynamics
attached. Also the graphs G(Δ) can be made label-independent by consider-
ing the equivalence classes [G(Δ)]’ = {G(Δ0 )jΔ0 ’ Δ} and treating G(Δ) only
as the corresponding representative of the class.
Accordingly, distances individuating point-objects that then are matter
points and change of these distances are the primitives of the minimalist
ontology that we propose. Indeed, using the material terms of unextended
substances being at a distance from each other with that distance changing
is ontologically more accurate to characterize this position than the formal
categories of objects and relations (cf. Mulligan (2012)). That notwith-
standing, we will continue to talk in terms of objects and relations to link
up with the literature.
In posing distance relations instead of an absolute space, we follow
Leibniz’s relationalism about space. According to Leibniz, distances make
up the order of what coexists (third letter, § 4, in Gerhardt (1890),
p. 363). Distances are able to distinguish objects, thus respecting Leibniz’s
principle of the identity of indiscernibles (although Leibniz himself does not
have a relational individuation of objects in mind). They thereby account
for there being a plurality of objects. Consequently, we do not presuppose
the numerical plurality of objects as given, which would be objectionable:
there would then be nothing which makes it that there is more than one
object. Quite to the contrary, in virtue of the matter points standing in dis-
tance relations that distinguish them, there is a plurality of them. In other
words, in virtue of these relations, there is a configuration of matter
points that is constituted through variation in the distance relations that
connect the matter points and that make it that these are matter points.
This conception furthermore accounts for the impenetrability of matter
without having to invoke a notion of mass: for any two matter points to
overlap, it would have to be the case that there is no distance between them.
Hence, matter is structurally individuated—namely, by the distances
among the material objects. In other words, we conceive relationalism as a
form of ontic structural realism, employing the relations that relationalism
about space acknowledges to individuate the relata of these relations. As
the literature on ontic structural realism has made clear, structures in the
sense of concrete physical relations—such as distances—can individuate
physical objects (see, e.g., Ladyman (2007)). Indeed, structures in this sense
can do exactly the same as properties are supposed to do: if one holds that
objects are bundles of properties, then the corresponding view is what is
known as radical ontic structural realism—namely, the view that objects
are constituted by relations, being the nodes in a network of relations (see
Ladyman and Ross (2007), ch. 2 and 3, and French (2014), ch. 5–7, for
an elaborate defense). If one thinks that there are underlying substances
that instantiate properties, then the corresponding view is what is known
as moderate ontic structural realism—namely, the view that objects and rela-
tions are on a par, being mutually ontologically dependent: relations require
relata in which they stand, but all there is to the relata is given by the
 Matter points and their dynamics 25
relations that obtain among them (see Esfeld (2004), section 3; Esfeld and
Lam (2011); this view is taken up in, e.g., Floridi (2008) and McKenzie
(2014)).
In any case, the fundamental objects do not have an intrinsic nature, but a
relational one. Relations are on the same footing as intrinsic properties in
that respect. To come back to the citation from Jackson (1998) earlier, if it
were mysterious what it is that stands in the relations (on the assumption
that all there is to a fundamental physical object are the relations in which
it stands to other such objects), then it would be mysterious in exactly the
same way what it is that instantiates the intrinsic properties that are supposed
to characterize the fundamental physical objects. To put it differently, in any
case, bare particulars are mysterious, and the commitment to bare particulars
is avoided by taking certain intrinsic properties, or certain relations to be
essential for the fundamental physical objects.
We advocate moderate ontic structural realism. To our mind, there is no
physical or metaphysical reason to conceive ontic structural realism as
being opposed to an object-oriented metaphysics: if there are relations,
there are objects that stand in the relations. In other words, ontic structural
realism can admit objects, as long as all there is to the objects are the rela-
tions among them. What ontic structural realism rejects is the property-
oriented metaphysics that dominates philosophy since Aristotle: the funda-
mental physical objects do not have an intrinsic essence. This is a concep-
tion of objects that stands on its own feet, being an alternative to both
the view of objects as bare substrata and the view of objects as bundles
of properties (or relations). There is a mutual ontological dependence
between objects and relations: as there cannot be relations without
objects that stand in the relations, so there cannot be objects without rela-
tions in which they stand. Hence, if one removed the distance relations,
there would not remain bare substrata, but there would then be nothing
(see Esfeld and Lam (2008, 2011)).
In order to obtain the result that the distance relations individuate the
matter points and thus distinguish them, we have to stipulate that these rela-
tions establish what is known as absolute discernibility in today’s literature:
each of the matter points distinguishes itself from all the other ones by at least
one distance relation that it bears to another matter point (see requirement
(iv) in definition 1). Being absolutely discernible, the matter points are indi-
viduals. Hence, the famous slogan “No entity without identity” coined by
Quine (1969, p. 23) applies to the matter points, although there is nothing
intrinsic about them: they do not have an intrinsic identity, but one provided
by distance relations that make them absolutely discernible entities.
What is known as weak discernibility in today’s literature would not be
enough, since weak discernibility does not avoid having to endorse a given
numerical plurality of objects: for weak discernibility to be satisfied, it is
sufficient that objects stand in an irreflexive relation, without there being
anything that distinguishes one object from the other ones. Hence, weak
26 Matter points and their dynamics
discernibility indicates that there is a numerical plurality of objects, but is
too weak to individuate the objects.1
Nonetheless, there is a tension in ontic structural realism between the fol-
lowing two claims:

1. The symmetries that physical theories exhibit are the guide to the ontic
structures.
2. The ontic structures individuate physical objects.

The tension consists in the fact that the higher a degree of symmetry a
structure exhibits, the less it is in the position to individuate objects (see
Keränen (2001)). In posing axiom (1), we propose to resolve this tension
by simply dropping the first of these two claims. What we need structures
in a parsimonious ontology for is to relate simple objects in such a way
that the structures individuate the objects. In other words, the rationale for
admitting structures in the ontology rests on the second of these two claims.
Indeed, any model satisfying axiom (1) has to include at least three
matter points and has to comply with requirement (iv) of definition 1. Sym-
metrical configurations are thus ruled out, as well as, for instance, the con-
figuration of an isosceles triangle. However, this is no objectionable
restriction: having empirical adequacy in mind, there is no need to admit
worlds with only one or two objects or entirely symmetrical worlds as phys-
ically possible worlds (and see Hacking (1975) and Belot (2001) for an
argument not to admit these as metaphysically possible worlds either).
The restriction formalized by (iv) does not rule out universal configurations
that are locally symmetric (i.e. having only a subset of matter points in a
symmetric configuration). What is excluded are cases in which all the
matter points in a universal configuration are arranged symmetrically.
Having in mind empirical adequacy, we note that this restriction poses no
problem for the ontology of physics: our universe definitely is not in a sym-
metric configuration. That notwithstanding, symmetries in physical theories
are very important to achieve a description of the evolution of our universe
that is both simple and informative, as we shall explain later: they entail a
great improvement in simplicity with only a little loss in information about
the actual particle configuration of the universe (which is not symmetric). In
short, the benefit of requirement (iv) is that it entitles us to conceive the dis-
tance relations as individuating the matter points, thus avoiding the com-
mitment to bare particulars; the cost is, in fact, none since (iv) by no
means calls the significance of symmetries in physical theories for the
description of the evolution of the universe into question.
Furthermore, there is no need to abandon absolute discernibility in
quantum physics either (since Bohmian mechanics solves the measurement
problem by, among other things, respecting the individuality of the
quantum particles; we will go into Bohmian mechanics in Chapter 3,
section 2). In brief, its position in the network of distance relations
 Matter points and their dynamics 27
distinguishes each matter point from all the other ones so that the matter
points are absolutely discernible, but they are permutation invariant in
the sense that labelling the matter points has no significance. To sum up
this important issue: to obtain objects that are structurally individuated
by the relations in which they stand, we need absolute discernibility; to
obtain absolute discernibility, we have to throw symmetries in the sense
of entirely symmetrical, global configurations out of the ontology; doing
so makes the ontology empirically adequate (since the actual configuration
of matter of the universe is not symmetrical) without infringing upon the
representational importance of symmetries.
Since all there is to the matter points are the distance relations in which
they stand, this is, like Cartesianism, a geometrical conception of matter.
However, it is not to be confused with super-substantivalism—that is, the
view that space (or space-time) is the only substance and matter a property
of space (see most recently Lehmkuhl (2016)). The main problem for this
view is to account for motion, if there are only points of space (or space-
time) and their topological and metrical properties, since these cannot
move. Indeed, Wheeler (1962) tried super-substantivalism out in his pro-
gramme of geometrodynamics, but failed in the attempt to reduce dynami-
cal parameters to geometrical properties of points of space-time (see Misner
et al. (1973), § 44.3–4, in particular, p. 1205). By contrast, if there are no
points of space or space-time, but only distance relations between sparse
points that hence are matter points, all the dynamical parameters that
figure in physical theories can then be construed in terms of the role that
they play in accounting for the change in these relations—that is, the
motion of these points. In short, in a geometrical conception of matter by
distance relations between sparse points, there is a clear sense in which
there is motion and dynamical parameters capturing motion.
However, the substantivalist who accepts a dualism of matter and space
can retort that by endorsing an absolute space that underlies the spatial con-
figuration of matter, the substantivalist ontology, although being less simple
than the relationalist one, gains in explanatory value. Thus, Maudlin
(2007a, pp. 87–89) takes length of a path in absolute space as the primitive
notion and derives the notion of distance of point particles from that notion
as the minimal path length connecting them, claiming that he is thereby able
to explain the constraints on the distance relation (such as the triangle
inequality). The concern, however, is that one does thereby not provide a
deeper explanation of the distance relation: in order to be able to define a
minimal path length in space, one has to presuppose a structure that is
rich enough to accommodate a metric—as the relationalist has to presup-
pose a relation that is rich enough to fulfill the triangle inequality in
order to count as distance relation. If one employed a primitive notion of
path that does not permit a definition of minimal length, then one could
not derive the distance relation from such a notion of path. In short, sub-
stantival space comes with a metric in terms of, for example, paths of
28 Matter points and their dynamics
geodesic motion, and any metric defining a physical space is such that it ful-
fills all the constraints of three-dimensional geometry. Hence, there is no
additional explanatory value here in comparison to the relationalist who
just presupposes that the relations admitted as primitive fulfill certain
requirements; there only is the disadvantage that substantival space con-
tains more structure than is needed to account for the empirical evidence,
which consists of relative particle positions and change of these positions.
In a nutshell, the substantivalist creates the illusion of giving a deeper expla-
nation of something that, actually, comes in a package with the postulation
of a substantival space.
Since the distance relations individuate the matter points, this view spells
out what Schaffer (2010a) calls “the internal relatedness of all things”,
because certain relations—namely distances—are constitutive of the funda-
mental physical objects. However, this view does not end up in monism
(pace Schaffer (2010b)): there is a plurality of objects—namely, a plurality
of matter points being at a distance of each other. Indeed, the matter points
can with good reason be considered as substances: although they are not
independent, since being tied together by distance relations is essential for
them, they are permanent. They do not come into existence, and they do
not go out of existence. The attractiveness of atomism depends on being
able to vindicate the matter points as substances in this sense: if their
number did not stay constant, atomism would be committed to the
absurd view that matter points come into existence out of nothing and
are literally annihilated, disappearing into nothing. As Parmenides
famously maintains,

What-is is ungenerated and imperishable, a whole of one kind, unper-

turbed and complete. Never was it, nor shall it be, since it now is, all
together, one, continuous. For what birth would you seek for it?
Where, whence did it grow? Not from what-is-not will I allow you
to say or to think; for it is not sayable or thinkable that it is not.
And what need would have stirred it later or earlier, starting from
nothing, to grow? Thus it must be completely or not at all.
(fragment Diels-Kranz 28 B8, quoted from
Graham (2010), pp. 215–217)

On atomism, what-is are the matter points (particles). They are therefore
substances.
As the matter points are permanent, so is the change in their distance
relations. To invoke again Parmenides, we noted that being is ungenerated
and imperishable, and change belongs to being. That is to say, as the matter
points do not come out of nothing and do not disappear into nothing, so
change does not arise out of a sudden and does not suddenly cease to
occur. If there are only matter points connected by distances, all change
is change in the distance relations among the permanent matter points.
 Matter points and their dynamics 29
Let us therefore now turn to axiom 2. Change, conceived as change in the
distances among matter points that are permanent, does not presuppose any
temporal notion. Nonetheless, change, thus conceived, is directed in the fol-
lowing sense: it goes from one particular state of the configuration of matter
points consisting in certain distances among the matter points to another par-
ticular state of that configuration consisting in other distances among some
matter points. Any such change may be reversible. Nevertheless, the actual
change in the configuration is directed by virtue of the fact that it goes
from one specific state of the configuration to another specific state of the
configuration. By contrast, there is no direction in the distance relations indi-
viduating matter points as given by axiom 1, since there is no spatial direction
as long as there are only distance relations, but no space into which these rela-
tions are embedded. Thus, in the ontology that we propose, matter points,
distance relations and the change of these relations are the only primitives.
If change is change in the distance relations, any change in any possible
model of this ontology has to be such that it satisfies the requirements
stated in the definition of the distance relation—that is, definition 1. In par-
ticular, by axiom 1 and requirement (iv), the distance relations individuate
the matter points and thereby establish their identity in any state of the con-
figuration of matter points. Given that, as on axiom 2, the matter points are
permanent, the change in the distance relations provides for the identity of
the matter points in the configuration across all change. In other words,
what makes (i) the same matter point through each state of the configuration
is the way in which the relations it bears with respect to all the other matter
points change. An appropriate manner to represent this identity, therefore,
consists in depicting the matter points as moving on continuous trajectories.
One may object that, if we are confronted with two separate snapshots
depicting two distinct states of the universal configuration of matter points,
then it might be impossible for us to identify the matter points across the
snapshots (that is, there is, in general, no unique way to order the snapshots
and connect them by means of continuous trajectories), contra our individ-
uation claim. However, the source of this impasse is epistemological rather
than ontological, since it just stems from the fact that no information is pro-
vided about the change that led from one snapshot to the other. Once we
have this information, no individuation worry arises, because the distance
relations and their change provide for the identity of the matter points.
Against this background, the change in the distance relations (and,
hence, the motion of the matter points) can be represented as a parame-
trized list of states of the configuration of matter—that is, a map
~
DðÞ : R ! O; l 7! ½Dl ’ ; ð2:2Þ

which, by means of λ ! 7 [Δλ]’, denotes the change of the distance relations in

the configuration of matter independently of a labelling. Again, by axiom 1,
it is ruled out that the universal configuration of matter points can evolve
30 Matter points and their dynamics
into a state that violates requirement (iv), such as the evolution into a sym-
metrical universal configuration (whereas the evolution into local symmetri-
cal configurations is by no means excluded). However, again, this does not
deny the importance of symmetries in physical theories in order to accom-
plish a simple representation of the change that actually occurs: although
the universe is by no means in a symmetrical state, employing symmetries
leads to a huge simplification in the description of the evolution of the uni-
verse (and its subsystems). Banning symmetrical universal configurations
from the ontology does not infringe upon the empirical adequacy of the
ontology and allows us to employ the distance relations and their change
to individuate the matter points in any given universal configuration and
to provide for their identity across change.
As the minimalist ontology that we propose does not imply absolutism
about space, so it does not imply absolutism about time: time derives
from change. Again, we follow Leibniz for whom time is the order of suc-
cession (see notably Leibniz’ third letter, § 4, and fourth letter, § 41 in Ger-
hardt (1890), pp. 363, 376). Hence, there is no time without change, but
change exhibits an order, and what makes this order temporal is that it is
unique and has a direction. This in particular implies that the notion
usually referred to as the “identity over time” of a particle is supervenient
on the distance relations and their change.
Although Leibnizian relationalism thus entails that the topology of
time induced by the unique ordering of the elements in O ~ is absolute,
this is perfectly consistent with relationalism about time, since time
depends on the change in the configuration of matter points. In Mach’s
words, “time is an abstraction, at which we arrive by means of the
change of things” (Mach (1919), p. 224). The important point is that
there is no external measure of time: the idea that the global dynamics
unfolds according to the ticking of a universal clock is meaningless. If
the entire universe could evolve at different external time rates, then
two such evolutions would be physically indistinguishable, given that
they would consist in the very same sequence of states of the universal
configuration of spatial relations. Hence, they would exhibit the very
same change in the distances among the matter points. In short, a com-
mitment to an universal external clock would introduce ontological dif-
ferences that would not make any physical difference (this point is made
also in, e.g., Barbour and Bertotti (1982), see especially pp. 296–297).
Consequently, there is no absolute metric of time. What we call
“time” in this context is just an arbitrary parametrization of the curve
λ 7! [Δλ]’ on O~ and not, as in the Newtonian case, an additional external
variable. For this reason, the only meaningful way to define a clock is to
choose a reference subsystem within the universe relative to which the
rate of change in distance relations is measured. An example of a
simple reference subsystem is the circular motion of a pointer on a dial
 Matter points and their dynamics 31
of a watch, the arc length drawn by the pointer being directly related to
the parameter λ in the definition of the map λ 7! [Δλ]’.
To repeat, the argument for the ontology being exhausted by the two
axioms of there being distance relations individuating matter points and
these distances changing is that one thus achieves a minimalist ontology
of the physical world that is empirically adequate and that any richer ontol-
ogy does not lead to a gain in explanation, but only to new drawbacks.
However, we obtained this ontology by starting from the paradigm of
atomism, working through an argument to the effect that atomism, when
elaborated on as a cogent ontology of matter, comes down to only these
two axioms. But atomism is not the only proposal for a fundamental ontol-
ogy of the physical world. The opposite view is to take matter to be one
continuous stuff, known as gunk, that fills all of space (note that this
view does not have to commit itself to points, neither matter points nor
points of space; see Arntzenius and Hawthorne (2005)). This view also
goes back to the first Presocratic natural philosophers: Thales, Anaximan-
der and Anaximenes can all be taken to embrace a stuff view of matter.
Let us call this view “ontological monism”, because it admits only one con-
tinuous substance, to mark the contrast with the ontology of atomism,
according to which there is a plurality of discrete substances. Like
atomism as set out in this section, this view is not tied to endorsing an abso-
lute background space: it can be construed as being committed to a contin-
uous stuff that is extended, but not to an absolute space that is distinct from
that stuff and into which that stuff is inserted.
In order to accommodate variation, gunk cannot be conceived as being
homogeneous throughout space. To take variation into account, one has
to maintain that there is more stuff in some parts of space and less stuff
in others. Atomism conceptualizes variation in terms of different distances
among the discrete matter points so that some matter points are situated
close to one another, whereas others are further apart: there are clusters
of matter points with distances among them that are smaller than the dis-
tances that these matter points bear to matter points outside such a
cluster. By contrast, the gunk ontology cannot accommodate variation
throughout space in terms of the concentration of primitive matter
points. It therefore faces the following question. What constitutes the fact
of there being more matter in some regions of space and less matter in
others? Consider what Allori et al. (2014) say in this respect when discuss-
ing the proposal for a primitive ontology of quantum physics in terms of a
continuous matter density that fills all of space, known as the GRWm
theory (we will go into this theory in Chapter 3, section 3):

Moreover, the matter that we postulate in GRWm and whose density is

given by the m function does not ipso facto have any such properties as
mass or charge; it can only assume various levels of density.
(Allori et al. (2014), pp. 331–332)
32 Matter points and their dynamics
Hence, the view of matter being gunk has to acknowledge as a further prim-
itive a variation of the density of gunk throughout space with gunk being
more dense in some parts of space and less dense in other parts. That is
to say, the primitive stuff admits of degrees, as expressed by the m function
in the GRWm formalism: there is more stuff in some parts of space than in
others, with the density of matter in the parts of space changing in time;
otherwise, the theory would not be able to accommodate variation. For-
mally, one can represent the degrees of density in terms of attributing a
value of matter density to the points of space (the m function as evaluated
at the points of space), although the matter density stuff, being gunk, is infi-
nitely divisible, and this ontology is not committed to the existence of
points of space. The main problem is that it remains unclear what could
constitute the difference in degrees of stuff at points of space, if matter
just is primitive stuff. The gunk theory thus is committed to the view of
matter being a bare substratum with its being a primitive fact that this sub-
stratum has various degrees of density in different parts of space. In a nut-
shell, there is a primitive stuff-essence of matter that furthermore admits of
different degrees of density.
In comparison to the gunk view of matter, atomism is the simpler and
clearer proposal for a fundamental ontology, because it avoids the
dubious commitment to a bare substratum or primitive stuff-essence of
matter with different degrees of density. Apart from this a priori argument,
its empirical success speaks in favour of atomism. All experimental evidence
in fundamental physics is evidence of relative particle positions and change
of these positions. Entities that are not particles—such as waves or fields—
come in as figuring in the explanation of the motion of the particles, but
they are not themselves part of the experimental evidence. Moreover,
science in general teaches us that macroscopic objects are composed of
chains of molecules and that these molecules are composed of atoms in
the sense of the chemical elements of the periodic table, which in turn are
composed of protons, neutrons and electrons. The ontology of atomism
can easily account for that composition: we finally get down to matter
points standing in distance relations. Everything else is composed of these
simple, discrete objects, consisting in certain clusters of matter points.
That is how we get from fundamental particle physics to statistical
physics of large ensembles of elementary particles, chemistry, molecular
biology and, finally, neuroscience. And that is why Feynman says at the
beginning of the Feynman lectures on physics, as quoted in the introduc-
tion, that the atomic hypothesis contains an enormous amount of informa-
tion about the world (Feynman et al. (1963), ch. 1–2).
Nonetheless, atoms qua matter points are theoretical entities. They are
not seen by the naked eye when one sees, for instance, dots on a screen as
outcomes of the double slit experiment in quantum mechanics. They are
admitted because they provide the best explanation of the observable
 Matter points and their dynamics 33
facts. The simplicity and parsimony of this proposal are part of the case
for its being the best explanation. To put it in a nutshell, particle evidence
is best explained in terms of particle ontology. However, this explanation
is not given by the ontology of matter points alone, but by this ontology
together with the dynamics that is put forward to describe the motion
of the matter points: it is the dynamics that provides for the stability of
the macroscopic objects with which we are familiar. For instance, an
ephemeral cat-shaped configuration of matter points would not be a cat;
only a stable such configuration is a cat. To obtain this stability, as we
will show in the following, no properties of the matter points over and
above the distances among them are needed, but a dynamics is that pro-
vides for trajectories such that there are stable macroscopic configurations
of atoms.
To conclude this section, let us come back to the term “primitive ontol-
ogy” that is used in the context of quantum physics to refer to the distribu-
tion of matter in physical space whose state the quantum formalism
describes (cf. Dürr et al. (2013b), ch. 2). We propose to endow this term
with a threefold substantial meaning:

1. The ontology is primitive in the sense of fundamental: the matter points

are not composed of anything, but they compose everything else.
2. It is primitive in the sense of referring to primitive objects: the matter
points have no intrinsic properties. However, they are not bare sub-
strata either. The distance relations in which they stand are their
essence.
3. It is primitive in the sense of factual: the configuration of matter points
is simply there. The matter points have no further function, whereas
everything else—the geometry of space-time, dynamical parameters
such as mass, charge, energy, etc.—is introduced through the function
that it has for the evolution of the configuration of matter points
(that is, its function for the change in the distances among the matter
points).

Consequently, we go beyond the term “primitive ontology” as it is

usually employed in the literature—namely, in the sense of the primitive
ontology of a given theory (cf. Allori et al. (2008))—by proposing a prim-
itive ontology tout court. However, this fundamental ontology is not the
foundation of knowledge. As mentioned earlier, matter points are theo-
retical entities. The justification of this ontology is a coherentist one:
matter points as fundamental objects and a dynamics for the evolution
of the distances individuating these objects is the way to achieve an
overall coherent system of our knowledge that provides for a clear and
elegant explanation of the experimental evidence (in terms of laws,
causes and/or unification).
34 Matter points and their dynamics
2.2 The account of change: why simplicity in ontology and
simplicity in representation pull in opposite directions
So far we have been concerned with simplicity in the sense of parsimony in
ontology: distance relations individuating matter points, with these rela-
tions changing, while the matter points are permanent, is the first and fore-
most candidate for the simplest ontology of the natural world that is
coherent and empirically adequate, given that all the experimental evidence
comes down to relative particle positions and their change. However, when
it comes to representing that change in a physical theory, the conceptual
means provided by this ontology are insufficient. Using only these concep-
tual means, we could not do much better than just listing the change that
actually occurs, but not formulate a simple law that captures that change.
The rationale for seeking for a law is simplicity in representation: in the
ideal case, the law is such that given an initial configuration of matter as
input, the law yields a description of all the—past and future—change of
the configuration as output. In this case, we have optimized both the sim-
plicity and the informational content of the representation: specifying an
arbitrary initial configuration of matter and putting the representation of
that configuration into the law, we obtain the complete information
about its change. However, in order to achieve such a law, we need more
parameters than distances individuating matter points. The reason is that
there is nothing about the distance relations in any given configuration of
matter that provides information about the—past and future—evolution
of these relations. That is why further parameters—both geometrical and
dynamical ones, over and above relative distances that change—have to
be attributed to the configuration of matter points to obtain a dynamical
law.
Consequently, simplicity in ontology and simplicity in representation
pull in opposite directions. Using only the concepts that describe what
there is on the simplest ontology (matter points individuated by distance
relations), the description of the evolution of the configuration of matter
would not be simple at all, since one could not do much better than
merely listing that change instead of capturing it in a simple and general
law. Reading one’s ontological commitments off from the simplest
description—such as, e.g., Newtonian mechanics—the ontology would
not be simple at all: it would in this case be committed to absolute
space and time, to momenta, gravitational masses, forces, etc. A similar
observation applies to the case of symmetrical universal configurations
of matter mentioned in the preceding section: working with such configu-
rations increases simplicity in representation. However, if one admitted
such configurations in the ontology, one would lose the elegance and par-
simony of the ontology, since, in this case, the matter points could not be
individuated by the relations in which they stand, but would have to be
conceived as bare particulars.
 Matter points and their dynamics 35
In general, our claim is that whatever geometry and dynamics a physical
theory employs, all this apparatus is there only as the means to achieve a
description of the change in the distance relations that is optimal with
respect to both simplicity and informational content about that change.
Hence, all the mathematical framework that we introduce here is only a
means, limited to our human abilities, to efficiently describe that change.
In this spirit, the apparatus we lay out in the following is surely not the
only possibility to represent change. Indeed, there may even be better
suited methods to do so. Furthermore, one description may only hold in
a certain physical regime and has to be adapted or developed anew to
include previously unknown physical phenomena, while the ontology
remains the same.
In this section, we set out the apparatus that we employ to represent the
change in the distance relations in general terms before discussing concrete
physical theories from Chapter 3 on. In the next section, we elaborate on
the philosophical position that allows us to both maintain minimalism in
ontology and to endorse the simplicity in representation that comes with
the geometry and dynamics that a physical theory uses.
Hitherto, we have formalized the mathematical counterparts of the ele-
ments of the ontology: the matter points, their distances, and the change
thereof. To achieve the formulation of a physical law that describes that
change, we need an additional structure that allows us to represent how
the change, encoded in the curve λ 7! [Δλ]’, occurs. One mathematical
framework to formulate such a law that has proven itself extremely power-
ful is calculus. A physical law can then be given by means of a velocity field
from which the entire change—that is, [Δλ]’ for all l 2 R—can be inferred
uniquely by specifying an initial configuration [Δ0]’ only.
One way to achieve a description through calculus is to find a differenti-
able manifold S equipped with a metric d and one strictly monotonic map
t : R ! R such that, for each l 2 R, the configuration [Δλ]’ with Dl ¼
ðDij;l Þði;jÞ2E can be represented in S by means of N points q1,t,q2,t,...,qN,t 2 S
parametrized by t 2 R that fulfill, for all ði; jÞ; ðk; lÞ 2 E and t 2 R,

dðqi;t ; qj;t Þ Dij;tðtÞ

¼ : ð2:3Þ
dðqk;t ; ql;t Þ Dkl;tðtÞ

This requirement stipulates that it must be possible to specify an embedding

of λ !7 [Δλ]’ in at least one geometry in form of a differentiable manifold such
that (i) the value d(qi,t,qj,t) at each edge ði; jÞ 2 E represents the distance rela-
tion Δij,τ(t) between two matter points qi,t and qj,t in S, and (ii) the change in
the distance relation Δij,τ(t) is represented by the change of the distance d(qi,t,
qj,t). This representation hence requires the introduction of many new degrees
of freedom that are not part of the ontology—namely (1) the scaling function
36 Matter points and their dynamics
of time τ, (2) the particular labels given to the matter points q1,t, q2,t, . . ., qN,t
and (3) the structure that comes along with the introduction of S and d.
Because of its simplicity as a representational means and its empirical ade-
quacy in the non-relativistic regime, we shall frequently employ the choice of
three-dimensional space S ¼ R3 with d being the Euclidean distance as an
example. Thus, by embedding the configuration of distance relations of
matter points into the geometry of Euclidean space S ¼ R3 , we introduce a
particular choice of orientation, rotation, translation, and scaling of the coor-
dinate axes and by means of the vector space structure also directions
qi,t  qj,t. However, to stress again, this is only a means of representation:
the configuration of distance relations [Δ]’ is by definitions 1 and 2 in the
preceding section independent of these additional degrees of freedom.
If such a geometrical embedding is possible, the purely relational change
λ 7! [Δλ]’ can be cast as change with respect to the embedding manifold S
in terms of the map

QðÞ : R ! S N ; t 7! Qt ¼ ðq1;t ;    ; qN;t Þ: ð2:4Þ

The additional degrees of freedom introduced when choosing a representa-

tion make it possible to encode change by means of a derivative as a velocity
field:

d
vt ðQt Þ :¼ Q; for t 7! Qt ; ð2:5Þ
dt t

provided the motion (2.2) and the chosen representations are sufficiently
smooth to allow for the evaluation of the derivative. The velocity field vt
assigns to each point Qt 2 S N an element in the tangent space of T Qt S N —
that is, the space of directions in which change may occur. Given a certain
motion t ! 7 Qt we could then infer the velocity field vt. However, if we
knew vt as a map S N ! TS N beforehand, all possible t ! 7 Qt could be inferred
uniquely from it as integral curves of vt starting at a certain initial value Q0.
It is then the task of physics to find a general law of motion in terms of vt
that represents the change in the distances among the matter points in an
efficient manner. Often this is done by induction on the basis of observed
velocity fields (2.5) that exemplify the general law well in certain regimes.
Obtaining a possible vt by means of a physical theory thus achieved, the
velocity field should also apply to more general motions. Hence, one can
turn the definition in (2.5) around and study all motions t 7! Qt that fulfill

d
Q ¼ vt ðQt Þ: ð2:6Þ
dt t

As already indicated, many degrees of freedom enter into such a represen-

tation. To make this explicit in a simple example, consider a possible
 Matter points and their dynamics 37
3
representation of λ 7! [Δλ]’ in Euclidean space S ¼ R . For t = τ(λ) and l 2 R
such a representation can be achieved as follows. In a first step, one arbi-
trarily chooses one matter point, let us call it the N-th matter point, and rep-
resents it by qN = 0. In a next step one has to choose an orientation of the
Euclidean coordinate axes and may then represent the rest of the matter
points inductively as q1;t ; :::; qN1;t 2 R3 such that the Euclidean distances
d(qi,t,qj,t) = jqi,tqj,tj fulfill (2.3), which allows for an arbitrary scaling. The
involved choices in the representation are (a) the labeling of the matter
points and the origin of the coordinate system and (b) the orientation of
the coordinate axes. Shifting the origin from one matter point to another
and choosing a different orientation of the axes in one time instant
amounts to the—possibly time dependent—translations and rotations,
respectively.
In general, for a given representation S, one may retrieve from a repre-
sentation t 7! Qt the actual configuration of distance relations λ 7! [Δλ]’ by
modding out these introduced degrees of freedom of the representation.
This shall be denoted by the map

~
P : SN ! O ð2:7Þ

such that, for all t 2 R, it holds that Δτ(t) = P(Qt). The change λ 7! [Δλ]’ is
then uniquely identified by t 7! Qt. Different choices in the degrees of
freedom, however, usually imply a change in the velocity field vt as it
depends on the differential structure of S N and, hence, the choices that
had to be made to introduce this structure.
As an example, let us consider the case of two different representations
of λ 7! [Δλ]’ in terms of t 7! Qt and t 7! Q0t in one and the same S. Let us
further denote their relationship by a time-dependent one-to-one map Tt :
SN ! SN such that Q0t ¼ T t ðQt Þ. Then the relationship between the velocity
fields vt guiding Qt and v0t guiding Q0t , respectively, must fulfill the following
diagram (Figure 2.1):

Figure 2.1 Different representations of the same motion.

38 Matter points and their dynamics
This diagram highlights that the fundamental state of affairs is encoded
only in the change of relational configurations λ 7! [Δλ]’; the way we
“lift up” such a change to S N is somewhat arbitrary at first. However,
the diagram also imposes restrictions on the possible velocity laws for a
fixed S. For instance, in the simple example S ¼ R3 , two different represen-
tations Qt and Q0t might differ by means of the orientation of the Euclidean
coordinate axes, in other words by a rotation, which can be encoded as a
linear and time-independent map T t ¼ T : R3 ! R3 . In this case, one can
infer the transformation rule
d 0 d d
v0t ðTQt Þ ¼ v0t ðQ0t Þ ¼ Qt ¼ TQt ¼ T Qt ¼ Tvt ðQt Þ: ð2:8Þ
dt dt dt
In addition to the introduced degrees of freedom for one particular geometry,
the many possible representations may come also with different geometries.
This implies that the distance relations in any given initial configuration of
matter points [Δ0]’ in general do not allow to fix all the geometrical facts
compatible with the entire history of relational change λ ! 7 [Δλ]’. For
instance, it may be the case that the distance relations in a given initial con-
figuration of matter points can be well described in Euclidean vector space
S ¼ R3 , but that the relations then develop in a way that does not allow
for an efficient description based on Euclidean geometry.
Considering such a vast amount of degrees of freedom, one may wonder
how to find a representation (2.4) of a motion (2.2) that enables us to con-
ceive a sensible description of change by means of (2.6) at all. One needs a
criterion to distinguish good from bad choices of representations such that,
for each case at hand, a well-suited representation can be found. The
obvious criterion is to strive for the simplest and most informative descrip-
tion of the history of relations λ 7! [Δλ]’. Recalling again the discussion of
the diagram in Figure 2.1 according to which a change in representation is
likely to imply a corresponding change in the velocity field, even though the
change to be described is one and the same, the application of this criterion
does not only depend on the representation alone but also on the corre-
sponding velocity field. In this spirit, one may in the first place guess an
embedding of the configuration of matter in a geometry and a representa-
tion of the dynamics in familiar terms and then check if the predictions
we get from (2.6) fit the empirical data. If at some point our dynamical evo-
lution diverges from the empirically observed one, it may be due to the fact
that vt in this representation cannot account accurately for the actual
change. Then we have to adjust the representation of λ 7! [Δλ]’.
Likewise, it may the case that a description in this representation is the-
oretically possible, but involves an utterly complicated velocity field, or vice
versa a very difficult representation but a trivial velocity field. The reason
may again be an inconveniently chosen representation, which then requires
further adaptation to find an optimal balance between representation and
 Matter points and their dynamics 39
the corresponding velocity field. Hence, we first have to make a choice with
respect to an initial configuration [Δ0]’. Then the strategy is to “fine tune”
the choice of representation in order to achieve the simplest and most infor-
mative description of the history of relations λ 7! [Δλ]’. Usually, the choice
of a three-dimensional geometry as embedding for the initial configuration
is a good starting point: less dimensions are not empirically adequate; more
dimensions risk to bring in extra structure that will not be needed. Thus, in
short, the geometry is just a matter of representation of the change in the
configuration of distance relations.
The pertinent question then is how the velocity field can be fixed in such a
way that specifying initial conditions at an arbitrary time and plugging them
into a law that fits into equation (2.6) is sufficient to yield the motion of the
matter points at any time as result. It is here that constant parameters—such
as mass, charge, total energy, constants of nature, etc.—or parameters at an
initial time—such as momenta, a wave function, fields, etc.—come into play.
They are the means to achieve this result: they are such that specifying a value
of them and inserting that value into a law of the type of (2.6) enables us to
capture the change in a given configuration of matter points in a manner that
is as simple and as informative as possible, obtaining as output the change in
the configuration of matter points—that is, the velocity field—for any time.
Let us call the parameters that allow us to achieve this result the dynamical
structure of a physical theory.
There is no reason to privilege dynamical parameters that are attributed
to the matter points individually. Furthermore, even if there are such
parameters, these do not represent intrinsic properties of the matter
points. Thus, in classical mechanics, mass is a parameter that is attributed
to the particles individually, but it is not an intrinsic property, as argued at
the beginning of section 2.1. In quantum mechanics, then, the parameter of
mass is situated on the level of the wave function. Due to entanglement, the
wave function usually applies only to a configuration of matter points. Fur-
thermore, already velocity is in classical mechanics not a parameter that can
be attributed to points taken individually. Consequently, our proposal is
not hit by the campaign of Butterfield (2006a,b) against pointillisme: But-
terfield’s pointillisme both against dynamics and geometry is situated on
the level of what we call the dynamical structure. Our proposal for ontol-
ogy is committed to matter points, but these are holistically individuated
by the distance relations in which they stand; they do not have intrinsic
properties.
In any case, a fundamental physical theory is such that it defines a dyna-
mical structure for the configuration of matter as a whole. Our proposal
makes this procedure intelligible: if the ontology is one of matter points
standing in distance relations, it is obvious that by changing any one dis-
tance relation one thereby changes in principle all the other ones. A
matter point cannot come closer to or get farther away from another
matter point without thereby also changing the distances that it bears to
40 Matter points and their dynamics
in principle all the other matter points in the configuration. Hence, if there
are only distance relations among matter points, there is no problem to con-
ceive interaction—that is, correlated motion of the matter points—and it is
to be expected that this correlation affects in principle also distant matter
points, without there being a problem of action at a distance or non-local
interaction.
In brief, if there are only distance relations and if consequently any
change affects in principle all the distance relations in a given configuration,
correlated motion is to be expected without it making sense to call for
something that transmits such “action at a distance” or non-local interac-
tion. What we have here can appropriately be called dynamical holism. Fur-
thermore, such a dynamical holism is not limited to the domain of
fundamental physics. For instance, if one maintains that the meaning of a
belief consists in its inferential relations to other beliefs (semantic holism,
see, e.g., Sellars (1956) and Davidson (1995)), then changing any one
belief affects in principle all the inferential relations in a given belief
system (see Esfeld (2001) for a comparison between holism in philosophy
of physics and philosophy of mind). Thus, although correlated motion is
the bedrock in the physical realm, one can give an explanation in terms
of unification (see Kitcher (1981)) by showing that such a holism obtains
also in other domains such as semantics.
As regards a physical theory, the task hence is to specify a dynamical struc-
ture such that, for any configuration of matter points given as initial condi-
tion, the dynamical structure fixes how the universe would evolve if that
configuration were the actual one. The dynamical structure thus goes
beyond the actual configuration of matter: it fixes for any possible
configuration of matter what the evolution of the universe would be like if
that configuration were the actual one. It thereby supports counterfactual
propositions. Furthermore, formulated in this way, there is a commitment
to determinism built into the dynamical structure of a successful fundamental
physical theory. However, there is nothing suspicious about determinism:
dynamical parameters figuring in laws that fix all the change, given an
initial configuration of matter points, are the simplest and most informative
way to capture change. It may turn out that, as a matter of fact, such a phys-
ical theory cannot be achieved. But if there are dynamical parameters that fix
only probabilities for how the configuration of matter points evolves, given
an initial configuration, there always remains the question open whether
one can do better—that is, find dynamical parameters that fix that change.
To put it differently, once one has obtained such parameters, one knows
that the work is done: one has achieved a description of the change that is
both simple and maximally informative in requiring only one configuration
of matter points as input in order to yield the whole past and future
change as output. The question that remains then only is whether that
description is empirically adequate and whether it can be further simplified
without losing informational content. By way of consequence, as far as
 Matter points and their dynamics 41
ontology is concerned, there is no reason to bring in probabilities. These
come in when one ignores what the exact actual configuration of matter
points is, as we will set out in the last section of Chapter 3.

2.3 The best of two worlds: Super-Humeanism

about geometry and dynamics
The spatial structure is permutation invariant: as explained in section 2.1,
although the distance relations individuate the matter points, labelling the
matter points has no significance. The dynamical structure, by contrast, is
in general not permutation invariant. It allows us to sort the matter
points into different kinds of particles: some matter points move like
charged particles, others like heavy or light particles, etc., so that they
can be described as electrons, neutrons, etc. Therefore, a permutation of
the particles’ velocities obviously leads to a different physical situation.
However, it would be unwarranted to conclude from this fact that some
matter points are intrinsically electrons, others intrinsically neutrons, etc.
In any fundamental physical theory, the dynamical structure is defined
by the configuration of the matter points as a whole: a law of the type of
(2.6) specifies a velocity field for the whole configuration of matter in the
universe, describing how that configuration changes by means of assigning
a velocity to each matter point and thus singling out a curve λ 7! [Δλ]’. This
is particularly evident in the case of the quantum mechanical wave function
whose evolution is defined on configuration space, with each point of that
space representing a possible configuration of matter as a whole; due to
entanglement, the wave function has, strictly speaking, to be assigned to
the configuration of matter as a whole. However, even if a physical
theory introduces dynamical parameters that are such that a value of
them is attributed to the matter points taken individually, these parameters
come in through their role in a dynamical structure that determines a veloc-
ity field for the configuration of matter as a whole. Thus, in the case of clas-
sical gravitation, for instance, the masses attributed to the particles are a
constant that couples the motions of the particles to one another. In
short, all the difference between the matter points originates in different
ways in which they move, which can be described in an efficient manner
by attributing values of various dynamical parameters—including constants
of nature—to them.
By way of consequence, as in the case of the matter points individuated
through the distances among them, so also when it comes to their dynamics,
there is no need to admit properties in the sense of intrinsic physical prop-
erties making up for an intrinsic essence of the matter points. Structure is all
there is—distance relations as far as the individuation of the fundamental
objects is concerned, dynamical ones as far as the change in their configu-
ration is concerned. We seek to capture the dynamical structure by
42 Matter points and their dynamics
conceiving various dynamical parameters, as we seek to capture the spatial
structure by conceiving, for instance, an Euclidean space into which the
matter points are embedded. However, as that space does not exist as an
absolute space in the world, but is our means to represent the distances
among the matter points, so the various dynamical parameters that physical
theories introduce do not exist in the world: they are our means to represent
the change in the distance relations among the matter points. As Hall
(2009, § 5.2) puts it,

The primary aim of physics—its first order business, as it were—is to

account for motions, or more generally for change of spatial configura-
tions of things over time. Put another way, there is one Fundamental
Why-Question for physics: Why are things located where they are,
when they are? In trying to answer this question, physics can of
course introduce new physical magnitudes—and when it does, new
why-questions will come with them. (So it is no part of the thesis we
are considering that physics is only concerned with explaining
motions; it is just that the other explanatory demands on it are, in a
certain sense, derivative on this one.)

What Hall alludes to in this quotation is the crucial distinction between

what we call the primitive ontology and the dynamical structure of a phys-
ical theory. The primitive ontology is the claim about what there simply is,
whereas the dynamical structure includes all the parameters that enter into
the theory in terms of their causal role for the evolution of what there
simply is. Matter points being individuated by distance relations that
change is our proposal for a primitive ontology that is independent of par-
ticular physical theories, being a proposal for a primitive ontology of the
whole of physics. That proposal is minimally sufficient for empirical ade-
quacy in a scientific realist spirit. It is backed up by the Cartesian argument
that the commitment to extension is the only substantial commitment in an
ontology of the natural realm, by the fact that positions of simple, discrete
objects (particles) are the central and primitive parameter in classical as well
as quantum physics and by the fact that all experimental evidence consists
of the relative positions of discrete objects and their change (motion).
Nonetheless, as explained at the beginning of section 2.2, positions of
discrete objects are not sufficient as parameter to obtain a simple law,
because a given configuration of objects described in terms of relative dis-
tances does not provide any information about how the distances change.
Hence, it cannot be sufficient as a basis for formulating a law of that
change (at least a law that satisfies common standards of being simple
and being informative). That is why further parameters have to be intro-
duced in terms of their causal role for that change—in particular, parame-
ters that are attributed to the matter points individually in a configuration
that is given as initial condition (such as mass and charge in classical
 Matter points and their dynamics 43
mechanics) or to the configuration as a whole (such as the wave function in
quantum mechanics). That is to say, these parameters have to be specified
as initial conditions in addition to the relative positions of the matter
points in order for a law of the type of (2.6) to yield the evolution of the
relative positions of the matter points (i.e. their motion).
Adding these parameters does as such not lead to an improvement in
simplicity. Fixing values for them may turn out to be as complicated and
messy as specifying initial relative positions of the matter points (consider
the case of the quantum mechanical wave function). The gain in simplicity
is that by figuring out these further parameters over and above the relative
particle positions in an initial configuration, one determines an initial con-
dition that can be inserted as input into a law of the type of (2.6) so that one
obtains as output (in the deterministic case) a description of the whole past
and future evolution of the particle configuration. It is therefore misleading
to call these parameters nomological: they are not laws. They are part of the
dynamical structure in that they are introduced in terms of their causal role
in the evolution of the distance relations, but they are a part of the dynami-
cal structure that has to be fixed as an initial condition in order to obtain a
simple law that describes the whole evolution of the configuration (instead
of a long list that merely enumerates how the configuration evolves) (cf. the
distinction between two types of physical state in Bhogal and Perry (2017)
—one type that consists only in the relative particle positions, and a further,
broader type that includes what has to be specified as initial condition in
addition to that).
However, their being part of the initial condition suggests prima facie to
subscribe to an ontological commitment to these parameters along with the
commitment to relative particle positions. In other words, at a first glance,
they look like stuff that there is in the world together with the distances
individuating the matter points. Consequently, the matter points appear
as being equipped with masses and charges, as well as existing in fields,
and be it a field like the quantum mechanical wave function. Nonetheless,
the distinction between primitive ontology and dynamical structure points
out that also for the scientific realist, there is no automatism from a param-
eter figuring in the dynamical structure of a well-established physical theory
to subscribing to an ontological commitment to that parameter along with
the elements of the primitive ontology. This issue has to be settled by phil-
osophical argument. The benchmark for evaluating these candidates for an
ontological commitment is whether including them in the ontology
enhances the overall coherence of the position in that it provides for an
understanding or an explanation of nature that is not available on the
basis of a commitment to a minimalist primitive ontology only.
In this section, we will first frame our proposal in terms of what is known
as Humeanism in contemporary metaphysics in order to show how laws
enter into the picture on the basis of a commitment to primitive ontology
only (the Humean mosaic). We will then argue that enriching the ontology
44 Matter points and their dynamics
by including those parameters that have to be specified as initial conditions
in addition to the relative particle positions does not lead to a gain in expla-
nation, but instead to new problems and serious drawbacks (thus decreas-
ing instead of increasing the overall coherence of the position).
On Humeanism, what is referred to as the primitive ontology in contem-
porary philosophy of physics is the entire ontology in the sense that every-
thing else supervenes on the primitive ontology: given the primitive
ontology, everything else in the natural world is also given. Supervenience
is here understood in the sense of reduction: everything reduces to the
primitive ontology. Hence, whatever else there may be, it is no addition to
being, it comes with the primitive ontology. In other words, the primitive
ontology is sufficient to make true all the true propositions about the
natural world (as mentioned in the introduction, we have no intention here
to apply this reductionism to the mind, consciousness and normativity).
The way in which the change in the configuration of matter that is acces-
sible to us occurs exhibits certain patterns or regularities. Conceptualizing
these patterns or regularities, according to what is known as the best
system account, the laws of nature are the theorems of the system that
achieves the best balance between being simple and being informative in
describing the evolution of the configuration of matter. A purely logical
system would be very simple, but not informative at all. An extremely
long inventory listing how the distances among the matter points change
would yield the complete information about that change, but would not
be simple at all. Laws of nature simplify and are informative at the same
time, striking the best balance between these two virtues (see notably
Lewis (1973b), pp. 72–75, Lewis (1994), section 3, Loewer (2007), Hall
(2009) and Cohen and Callender (2009)).
Strictly speaking, what the laws of nature are is fixed only by the entire
evolution of the configuration of matter—that is, the whole change in the
distance relations among the matter points (cf. Beebee and Mele (2002),
pp. 201–205). Of course, that whole change is not accessible to any
observer in the universe. The axioms and theorems that figure in our funda-
mental physical theories are the best conjectures about the laws of nature
that we can make by putting forward on the basis of the evidence that is
available to us a system that achieves the best balance between being
simple and being informative in accounting for that evidence. Conse-
quently, what we regard as the laws of nature changes as more evidence
becomes accessible to us.
Once laws are vindicated, Humeanism can handle causation in the same
manner: causal relations may simply be given in terms of the regularities
that the whole change in the configuration of matter exhibits (this was
Hume’s famous view see Hume (1739), book I, part III, and Hume
(1748), section VII; see Baumgartner (2013) for a contemporary regularity
theory of causation). Alternatively, the Humean can go for a sophisticated
regularity theory in terms of counterfactuals (see Lewis (1973a, 2004)). The
 Matter points and their dynamics 45
decisive issue then is that the truth-value of the counterfactuals expressing
causal relations supervenes on the configuration of matter and its change in
the actual world. To put it differently, no realism about other possible
worlds is required to obtain truth-makers for counterfactuals in Humean-
ism (see Armstrong (2004), p. 445, and Loewer (2007), pp. 308–316).
In a famous passage, Lewis states the ontology of Humeanism in this
way:

Humean supervenience is named in honor of the greater denier of nec-

essary connections [i.e. David Hume]. It is the doctrine that all there is
to the world is a vast mosaic of local matters of particular fact, just one
little thing and then another. (. . .) We have geometry: a system of exter-
nal relations of spatio-temporal distance between points. Maybe points
of spacetime itself, maybe point-sized bits of matter or aether or fields,
maybe both. And at those points we have local qualities: perfectly
natural intrinsic properties which need nothing bigger than a point at
which to be instantiated. For short: we have an arrangement of quali-
ties. And that is all. There is no difference without difference in the
arrangement of qualities. All else supervenes on that.
(Lewis (1986b), pp. ix–x)

Lewis describes the “vast mosaic of local matters of particular fact” not
only in terms of primitive point objects but also in terms of local qualities
in the sense of “perfectly natural intrinsic properties” instantiated by these
point objects. This commitment puts Humeanism in a vulnerable position.
Given that properties are introduced in physics through the role that they
play for the change in the configuration of matter, that role being expressed
in terms of parameters figuring in dynamical laws (cf. the quotation from
Jackson (1998) at the beginning of section 2.1), the Humean who takes
instantiations of natural properties to make up the local matters of partic-
ular fact has to maintain that this role is not essential to these properties.
What role a given property plays depends on the laws of nature, which in
turn depend on the whole evolution of the local matters of particular
fact. Hence, what that role is cannot be a local affair. Furthermore, if it
were essential for a given property to play a certain role, Humeanism
would no longer eschew the commitment to necessary connections: the
instantiation of such a role would impose modal, if not necessary connec-
tions on the local matters of particular fact. Thus, for instance, on Humean-
ism, the role that the property of mass plays for gravitational attraction in
the world is contingent instead of essential to that property so that, in
another possible world, mass can play a totally different role, depending
on what the mosaic of local matters of particular fact is like in that other
world. To illustrate this issue further, the properties of mass and charge
can swap their roles: there is another world possible, w*, in which the prop-
erty that we pick out as charge in the actual world w plays the role of the
46 Matter points and their dynamics
property that we pick out as mass in the actual world w, and vice versa. Of
course, w and w* would be indiscernible: in w and w* happen exactly the
same things, matter moves in exactly the same way in both worlds. None-
theless, w and w* are two different worlds on Lewis’s version of
Humeanism.
Consequently, Lewis’s Humeanism is committed to conceiving proper-
ties as pure qualities, known as quiddities. Furthermore, we cannot have
epistemic access to these qualities. Our epistemic access is limited to the
roles that the properties play for the evolution of the configuration of
matter as described by a physical theory. Thus, we can distinguish
between properties only insofar as they exercise different roles. There is
no means to distinguish between what is supposed to be different pure qual-
ities. The lack of epistemic access to properties conceived as pure qualities is
known as humility. Lewis (2009) bites the bullet of endorsing humility and
quidditism (that is, the commitment to properties as pure qualities).
However, the rather baroque metaphysics of pure qualities with its implica-
tion of possible worlds differing only in the pure qualities that are instanti-
ated in them is a heavy burden on Humeanism, in particular given that
Humeanism sees itself as a metaphysics that is close to science and empir-
icism, avoiding any sort of occult metaphysics (see notably Black (2000)).
Fortunately, there is no need to take the Humean mosaic to be built
up by instantiations of natural properties that then are pure qualities
(quiddities). Instead, it is reasonable to conceive it in terms of a primitive
ontology—namely, the spatial configuration of simple objects and its
change. As shown here, such a primitive ontology is not committed to
admitting space and time as substances, but can be formulated in terms
of relationalism about space and time. Consequently, the Humean mosaic
consists only in matter points individuated by distance relations and the
change of these relations.
If that mosaic contained pure qualities, these qualities could not have any
significance for the supervenience of laws on the mosaic. No purely quali-
tative variation could make up for a supervenience basis for laws. What
accounts for the mosaic of local matters of particular fact being a superve-
nience basis for laws is that there is a variation in it that amounts to a dis-
cernible difference between matter points and change in that variation. For
such a variation and change to obtain and for it to exhibit patterns or reg-
ularities such that laws can supervene on it it is sufficient that there are dis-
tance relations which individuate simple (= propertyless) objects and which
change. Consequently, abandoning Lewis’s natural properties does not
result in any arbitrariness in defining the Humean mosaic: these natural
properties do not have any significance for the supervenience basis
anyway. What is needed are fundamental relations that individuate
simple objects so that everything else can then supervene on this mosaic
of relations and their change. As argued in section 2.1, distance relations
individuating matter points perform that job when it comes to the
 Matter points and their dynamics 47
natural world (in other words, they are—the only—natural relations).
Hence, again, bringing in more—a commitment to mass, charge and the
like as natural properties that are pure qualities in this case—creates new
drawbacks (namely the commitments to quidditism and humility) instead
of having any explanatory value.
This new version of Humeanism can be dubbed Super-Humeanism. That
term is coined in analogy to the term “super-substantivalism” in the philos-
ophy of space-time: while the standard substantivalist holds that space or
space-time is a substance and that there is matter (particles, fields) filling
space-time, the super-substantivalist maintains that there only is space-
time, with what we take to be matter being reduced to the geometry of
space-time. By the same token, while the standard Humean holds that
there are spatial or spatio-temporal relations connecting points and
natural intrinsic properties instantiated at these points, the Super-
Humean maintains that there are only sparse points that then are matter
points with distance relations individuating these points, but neither is
there an underlying space nor are there natural intrinsic properties.2
Indeed, super-substantivalism can also be conceived as a Super-Humean-
ism: in this case, metrical relations individuate space-time points, with there
being a plenum of space-time points. These points do not instantiate intrin-
sic properties; all there is to them is their being connected by metrical rela-
tions. As regards Lewis himself, in being committed to “perfectly natural
intrinsic properties” instantiated at space-time points, he is also committed
to space-time substantivalism. However, as he notes in Lewis (1986a, p. 76
note 55), a dualism of both substantival space-time and matter is
uneconomical.
In fact, Lewis inclines towards resolving this dualism in terms of super-
substantivalism. It is not clear from Lewis (1986a, p. 76 note 55) whether
he is prepared to go beyond what Sklar (1974, pp. 166–167, 222–223) dis-
misses as a “linguistic trick”—namely, simply to attribute to space-time
points the properties that are commonly ascribed to material objects (such
as, e.g., mass and charge); such a move obviously cuts no ice, neither in
physics nor in philosophy. The view that cuts at least philosophical ice is
the project to reduce the properties commonly attributed to material
objects to genuine properties of space-time points—namely, their geometrical
ones, viz. the metrical relations between the space-time points. This can be
done in the framework of Humeanism by interpreting the properties com-
monly attributed to material objects as being part and parcel of the best
system, along with the laws, which then is the best system with respect to
the mosaic of spatio-temporal relations between points.3
However, this is just what the Super-Humean does: in virtue of standing
in distance relations, the points between which these relations hold are
matter points, because they fulfill the Cartesian criterion for matter—
namely, extension and motion. But these points then are sparse, and the
relations cut down to distances that leave the space-time geometry as a
48 Matter points and their dynamics
matter of representation. The commitment to a plenum of space-time points
(i.e. a substantival space-time) and fully-fledged spatio-temporal relations
(i.e. coming with a certain geometry) then is unmotivated; it only results
in a commitment to surplus structure, and it infringes upon the empirical
adequacy of this view (cf. the remarks in section 2.1). Hence, from a
Super-Humean perspective, super-substantivalism is cut down to a relation-
alism that only admits distance relations that individuate point objects,
which then are matter points in virtue of their standing in these relations.
If one seeks for properties, they come into the picture of Super-Humeanism
through the dynamical parameters that figure in the laws. Suppose, for the
sake of the argument, that the laws of classical mechanics and classical elec-
trodynamics are part of the best system. Mass, charge and other parameters
figure in these laws. Both inertial and gravitational mass as well as charge are
admitted in classical mechanics only because they perform a certain causal
role as described by the laws of gravitation and electrodynamics—namely,
to accelerate the particles in a certain manner. Hence, on Super-Humeanism,
parameters or properties like mass and charge are no addition to being. The
configuration of primitive objects and its change is all there is. Given that this
change exhibits certain patterns, laws can be formulated, and given the
parameters that figure in the laws and that are required, over and above posi-
tions, to specify an initial condition, one can attribute properties like mass
and charge to the primitive objects–that is, the particles in classical mechan-
ics. But the particles do not have these properties per se, as something
essential or intrinsic to them. They obtain them only through the regularities
that the change in the distance relations among them exhibits via the
attempt to capture these regularities in simple laws (see Hall (2009), § 5.2;
see also Martens (2017)). That is to say: it is not mass and charge qua
properties belonging to individual matter points that determine their
trajectories by means of the causal role that they play in the laws of classical
mechanics and electrodynamics; on the contrary, the trajectories that the
matter points take throughout the evolution of the universe make it that
parameters such as mass and charge figure in the dynamical laws such that
a value of mass and charge applies to the matter points taken individually
and is used to define the initial conditions that allow the application of the
laws.
It is therefore more appropriate to set this ontology out in terms of
predicates and truth-makers for predicates. The distance relations
among the matter points and the change in these relations make true
all the true propositions about the world, including in particular the
propositions expressing laws of nature. Hence, if the laws of classical
mechanics figure in the best system, predicates such as “mass” and
“charge” apply to the particles in virtue of the patterns that the particle
trajectories exhibit. These predicates as well as all the other ones appear-
ing in the propositions that are true about the world really apply, and the
propositions really are true, there is nothing fictitious about them. But
 Matter points and their dynamics 49
what there is and hence what makes them true is nothing over and above
matter points individuated by distance relations and the change in these
relations.4
In brief, on Super-Humeanism, not only the laws but also the dynamical
parameters that a physical theory introduces as well as the geometry of
space and time (see Huggett (2006) on geometry and inertial frames in clas-
sical mechanics) supervene on the evolution of the configuration of matter
in the universe, which is then characterized only in terms of the distance
relations connecting propertyless matter points. Super-Humeanism can
therefore be conceived as being inspired by the following quotation from
Russell (1927):

There are many possible ways of turning some things hitherto regarded
as “real” into mere laws concerning the other things. Obviously there
must be a limit to this process, or else all the things in the world will
merely be each other’s washing.
(Russell (1927), p. 325)

Thus, mass, charge, the geometry of space and time, etc., are “mere laws
concerning the other things”. The bedrock—the Humean mosaic—is
matter points individuated by distance relations that change. All the rest
are parameters introduced to capture that change in a manner that is
both simple and informative about that change, given that all the experi-
mental evidence consists in that change. Super-Humeanism thus is a
simple and parsimonious ontology that is close to science, being free of
the burden of the baroque metaphysics of quiddistic intrinsic properties
as in Lewis’s Humeanism.
By the same token, turning to quantum physics with particles as the
primitive ontology (as in Bohmian mechanics), on Super-Humeanism, it is
not the wave function that determines the trajectories of the particles; on
the contrary, given the motion of the particles, the regularities that their
motion exhibits make it that a wave function parameter figures in the dyna-
mical laws capturing that motion and has to be specified as initial condition
in order to apply the laws. Miller (2014), Esfeld (2014b), Callender (2015)
and Bhogal and Perry (2017) have worked this stance out independently of
one another as regards the wave function in quantum theories with a prim-
itive ontology (such as Bohmian mechanics) (see Dickson (2000) for a fore-
runner). There is a precise physical model in this sense by Dowker and
Herbauts (2005), namely how to conjecture the wave function from a
history of GRW flashes on a lattice. In this model, it is clear how the
wave function simplifies by figuring in a law for the evolution of the
GRW flashes. However, if it turned out that the universal wave function
were as complicated and messy as the entire history of the relative particle
positions (or the flashes), then the wave function would be inappropriate as
a parameter that enters into a law that describes that history; there would
50 Matter points and their dynamics
then be no gain in simplicity: the whole information about the physical
world is trivially contained in the entire history of the particle positions.5
On Super-Humeanism, in brief, there are no relations of entanglement in
nature over and above the distance relations among the matter points. The
quantum state does not mysteriously tie the motion of the Bohmian parti-
cles together by somehow guiding or piloting them in a coordinated
manner. There is no quantum state in nature, although all the propositions
formulated in terms of the quantum state are perfectly true. As a contingent
matter of fact about the actual universe, the way in which the particles
move is such that it manifests certain stable correlations so that, if we set
out to represent their motion in a manner that is both simple and informa-
tive, we have to write down an entangled quantum state and a law in which
an entangled wave function figures. Given that law, we then have a truth-
maker for the counterfactual propositions about what would happen to the
motion of other particles if one intervened on the motion of a specific par-
ticle in a certain manner, etc. But there are no mysterious influences going
from one particle to distant particles via a quantum state in nature.
That notwithstanding, Super-Humeanism is distinct from instrumental-
ism about the wave function: the wave function is the central dynamical
parameter, figuring in the law of motion for the particles, as well as
being itself subject to an evolution according to a law (i.e. the Schrödinger
equation). These laws then are linked, as we shall explain in Chapter 3, sec-
tions 2 and 4, with the probability calculus in which the wave function is
employed to calculate probabilities for measurement outcomes according
to Born’s rule. In brief, the first and foremost role of the wave function is
the one that it has in the law of motion for the primitive ontology
instead of being an instrument to calculate probabilities. In a nutshell,
Super-Humeanism allows us to have the best of two worlds: simplicity in
ontology achieved through parsimony and simplicity in description
achieved through buying into the simplest physical theory that is empiri-
cally adequate. Super-Humeanism thereby is a form of scientific realism.
Last but not least, Humeanism makes clear that there is nothing suspicious
about a physical theory being deterministic: there is nothing that determines
the evolution of the configuration of matter. That configuration develops in a
certain manner as a contingent matter of fact, exhibiting certain stable pat-
terns or regularities, which serve as truth-makers for laws of nature. But
this does not mean that the laws or any of the parameters figuring in them
have the power to determine the evolution of the configuration of matter,
or to make that evolution necessary. Considering any one given state of
the configuration of matter points consisting in certain distance relations
among the matter points, there is nothing in that state that poses a constraint
on how the distance relations can evolve (see notably Beebee (2000), Beebee
and Mele (2002), Beebee (2006) and Hall (2009), § 0). Humeanism thus
makes intelligible why physics strives for deterministic theories, because
these are the foremost candidates for theories that strike the best balance
 Matter points and their dynamics 51
between being simple and being informative. This is so even if it should turn
out that for any deterministic theory in physics—such as, e.g., Newtonian
mechanics—initial conditions are conceivable in which the determinism
breaks down. Humeanism can easily accommodate such marginal failures
of determinism, since it does not consider determinism as a modal feature
of the world, but only as a means to achieve an optimal balance between sim-
plicity and strength in a system that represents the evolution of the particular
matters of fact that make up the universe.
A fortiori, it is a misunderstanding to think that determinism in the dyna-
mical structure of a physical theory has implications for human free will.
The issue of such implications may arise only if one conceives the dynamical
structure as an additional element of the ontology—that is to say, it belongs
to the ontology in addition to the spatial structure, being a further primitive
over and above the matter points individuated by the distance relations in
which they stand and their change, such that, loosely speaking, the dynami-
cal structure pushes the configuration of matter in a certain direction. By
the way, if this were so, one realizes immediately that the ensuing conflict
with human free will depends in no way on the dynamical structure
being deterministic or stochastic: the tension with free will is the same
whether the physical laws make the configuration of matter (including
the matter of human brains) move in a certain direction with a certain prob-
ability that does not depend on human free will or whether the physical
laws do so in a deterministic manner (see Loewer (1996)). Again, reifying
the dynamical structure entails new problems.
However, the Super-Humean ontology provokes the objection that by
being most parsimonious, it loses explanatory value. We already considered
and rebutted in section 2.1 the claim of Maudlin (2007a, pp. 87–89)
according to which the substantivalist about space can explain the features
of the distance relation that the relationalist has to accept as primitive. In
brief, there is no gain of explanation in substantivalism, because the fea-
tures at issue come in a package with the postulation of an absolute
space. In general, where our minimalist ontology admits primitive distance
relations and primitive change of these relations that exhibits patterns such
that the relations and their change can be represented, for instance, in terms
of Euclidean geometry and masses, charges, forces, fields, wave functions,
etc., as they figure in the dynamical equations of a physical theory, a
richer ontology has an underlying space filled not only with point particles
but also with mass and charge as intrinsic properties of these particles as
well as forces, fields, or wave functions, etc.
The objection then is that thus enriching the ontology does not result in
an explanatory gain: these additional entities are dynamical parameters that
are defined through the functional or causal role that they play for the evo-
lution of the particle configuration. For instance, one does not give a deeper
explanation of attractive particle motion by admitting mass or the gravita-
tional force to the ontology, because these parameters are defined through
52 Matter points and their dynamics
the effect that they have (or can have or are the power to have) on the
motion of the particles. The only difference then is that what is contingent
on the minimalist ontology comes out as necessary on the richer ontology
(the Euclidean geometry is necessary given an underlying Euclidean
space, the attractive motion of the particles is necessary given masses and
the gravitational force, etc.). However, lifting the status of something
from contingent (because accepted as primitive matter of fact) to necessary
does not yield a deeper explanation. Taking explanations to end in the dis-
tance relations among matter points and their change endows our minimal-
ist ontology with all the explanatory value that one can reasonably demand,
namely to explain all the other phenomena in terms of the fundamental
physical entities.
The most prominent way to enrich the ontology is to admit properties
that are dispositions or powers—that is, the disposition or power to
change the motion of matter in a certain manner under certain conditions.
By contrast to Lewis’s conception of natural properties that are pure qual-
ities, properties conceived as dispositions or powers establish modal con-
nections in nature in the sense that it is necessary that a certain change in
the motion of the matter points occurs if such properties and conditions
are present. In contrast to Humeanism, the laws of nature then supervene
on these properties in the sense that the laws express which dispositions
or powers these properties are (see notably Bird (2007)). Consequently,
what the laws of nature are does not depend on the evolution of the config-
uration of matter. Quite to the contrary, the manner in which this evolution
occurs is brought about by these properties. Hence, if all the dynamical
parameters are instantiated by the configuration of matter at any given
time, what the laws of nature are is fixed by the properties that the config-
uration of matter has at any arbitrarily chosen time.
However, by admitting mass and charge as intrinsic properties of the
matter points over and above the distance relations among them, one
does not provide an explanation of the change in the distance relations
that is illuminating by contrast to accepting that change as a primitive
matter of fact as the Humean does. It is true that one traces the change
that occurs back to properties of the matter points. But these properties
are defined in terms of the causal role that they exert for the motion of
the matter points. Thus, one shifts what has to be accepted as primitive
from the change that occurs to the causal role of certain properties. This
is no gain in explanation. It is an instance of the scheme of which
Molière makes a caricature in Le malade imaginaire: one does not
explain why people fall asleep after the consumption of opium by attribut-
ing a dormitive power to opium—although, of course, mass and charge are
sparse, fundamental properties by contrast to the phenomenological prop-
erties of opium. Nonetheless, the Molière argument hits also these latter
properties: like the dormitive power of opium, they are defined in terms
of the effects that they bring about under certain conditions.
 Matter points and their dynamics 53
This situation does not change if one draws attention to the fact that
parameters such as mass and charge figure in dynamical laws that are con-
nected with the symmetries of space-time. The same goes for the position
that admits not only the space-time symmetries as primitive, coming with
the ontological commitment to a substantival space-time, but also the dyna-
mical laws as primitive (see notably Maudlin (2007a)). One can then obtain
beautiful explanations in terms of unification. But insofar as scientific
explanations are concerned, the Humean is entitled to vindicate them in
the same way as the opponent who endorses an ontological commitment
to space-time symmetries and modal entities in nature (dispositions,
powers, primitive laws). For the Humean, space-time geometry and the
laws including the parameters that figure in them come in as the package
that provides the most simple and most informative description of the con-
figuration of matter and its change, thus yielding scientific explanations
through unification (for the recent debate about explanations in Humean-
ism, see notably Loewer (2012), Lange (2013), Miller (2015) and Marshall
(2015)).
Admitting mass and charge as intrinsic properties of the matter points that
are dispositions or powers again introduces a surplus structure in the ontol-
ogy: the dispositions or powers are not always manifest, and there may be
dispositions or powers existing in the universe without the conditions for
their manifestation ever being fulfilled (the same goes for primitive laws:
they may not always be instantiated). Moreover, again, admitting disposi-
tions or powers brings in new drawbacks instead of adding explanatory
value. As reifying space leads into the impasse of having to answer the ques-
tion of what distinguishes matter from space elaborated on in section 2.1, so
upgrading the dynamical parameters that a physical theory introduces to
intrinsic properties of the physical objects leads into the impasse of having
to answer the question of how a physical object can reach out to other
objects and change their motion in virtue of properties that are intrinsic to
it—since the effect that a particle produces in virtue of instantiating proper-
ties such as mass and charge is a change in the motion of other particles.
The search for an answer to this question then often results in the view of
there being forces or fields that literally propagate in space by means of
which one particle changes the motion of other particles. Such views,
however, have been convincingly criticized as anthropomorphism by
Russell (1912) and others, and their complete breakdown is evident at
the latest when it comes to quantum non-locality: the correlated motion
of quantum particles cannot be explained in terms of a force or field prop-
agating in space, and the wave function is defined on configuration space
instead of being a field in physical space. In general, the ontological
status of such mediating entities remains unclear: Are forces or fields
some sort of stuff filling space in addition to the particles? Are they proper-
ties of something? But of what? Furthermore, they neither are dispositions
nor manifestations of dispositions: the manifestation of mass and charge
54 Matter points and their dynamics
qua dispositions of particles is the change in the motion of other particles,
not the force or the field, but the force or the field are not dispositions in
their own right either, since they depend on the mass and charge qua dispo-
sitions of the particles (we will go into fields in classical electrodynamics in
detail in section 5.2).
One can avoid these additional drawbacks by switching from intrinsic
properties to structures. That is to say, instead of admitting intrinsic prop-
erties that are dispositions or powers, one endorses dynamical relations
among the matter points that exist over and above the distance relations
and that are the disposition or power to change the distance relations in
a certain manner—that is, the disposition or power for a certain correlated
motion of the matter points. Thus, one maintains that the dynamical struc-
ture of a physical theory refers to a dynamical structure that exists in the
world and that is a modal structure, by contrast to the spatial structure.
The most prominent example of this strategy is admitting entanglement
relations among quantum particles as relations in which these particles
stand over and above the distance relations, with these relations being the
disposition or the power to manifest themselves in a certain correlated
motion of the quantum particles (as, for instance, in the correlated out-
comes of the EPR experiment). But one can also conceive the dynamical
structure of classical mechanics as referring by means of mass and
charge, conceived now as coupling parameters, as well as forces and
fields to dynamical relations in the world that are dispositions or powers
and that manifest themselves in attractive and repulsive motion of the
particles.
Indeed, it is common among the proponents of ontic structural realism to
regard the dynamical structures as irreducibly modal structures (see notably
Esfeld (2009) and French (2014), chs. 9 and 10). The resulting view then is
a structuralist conception of the laws of nature in distinction to the Humean
one: the laws are our manner to capture the dynamical structure that there
is in nature (cf. Cei and French (2014)). That structure is irreducibly modal
in that it constrains the evolution of the distance relations among the matter
points. Note, as already mentioned in section 2.1, that it is a misconception
to enquire whether these structures can individuate the objects that stand in
them and, in case they fail to do so, to conclude that there is a problem with
the basic physical objects being conceived as individuals: the structure of
entanglement, for instance, is encoded in the quantum state, which is
defined on configuration space—that is, the mathematical space each
point of which represents a possible particle configuration in physical
space. Consequently, the entanglement structure presupposes a particle
configuration to which it is applied and cannot be that what individuates
these particles.
However, as its counterpart in terms of admitting modal, intrinsic prop-
erties of the basic objects, so admitting modal relations in which these
objects stand over and above the distance relations does not provide
 Matter points and their dynamics 55
additional explanatory value, since this move is also subject to a Molière-
type objection: the modal structures are defined in terms of what they do
—or can do or are the power to do—for the particle motions. For instance,
one does not give a deeper explanation of quantum non-locality in terms of
a modal structure of quantum entanglement, because this structure is
defined in terms of correlating the possible ways in which the quantum
objects that instantiate this structure can evolve.
It is true that the commitment to modal, dynamical relations instead of
modal, dynamical intrinsic properties does not create the drawback of how
one particle can reach out to changing the motion of other particles in
virtue of properties that are intrinsic to it, since the particles are ab initio
dynamically related on the modal structural realist view. However, this
view faces another problem—namely, to vindicate the dynamical structure
as a network of concrete physical relations among the particles by contrast
to an abstract mathematical structure that represents the motions of the
particles in a simple and informative manner (as Humeanism maintains).
Prominent proponents of ontic structural realism simply refuse to acknowl-
edge that problem, being prepared to concede that it is not possible to dis-
tinguish concrete physical from abstract mathematical structures (see, e.g.,
Ladyman and Ross (2007), pp. 159–161, and French (2014), p. 230), but if
this were so, it would be a fatal blow to ontic structural realism as a pro-
posal for the ontology of physics by contrast to the ontology of mathemat-
ics (see, e.g., Briceno and Mumford (2016) for a recent criticism in that
sense).
It is also not sufficient to maintain, as one of us did in earlier publications
(Esfeld (2009)), that, for instance, the relations of quantum entanglement
are concrete physical relations by contrast to abstract mathematical ones,
because they are causal in the sense that they have an effect on the
motion of the particles (or are the disposition or the power to have such
an effect), since what is at issue precisely is whether there is anything
that has such an effect qua being a concrete physical relation. The distance
relations set the paradigm for concrete physical relations: they individuate
the matter points. As brought out most clearly by quantum entanglement,
if there is a dynamical relation of quantum entanglement over and above
the distance relations, putting a constraint on how these relations can
change, there is only one instantiation of such a relation that encompasses
all the matter points as its relata, since there is only one, non-separable
quantum state of the whole configuration of matter (as represented by a
point in the configuration space of the universe with the wave function
acting on that point).
In other words, the universal wave function in quantum mechanics is
entangled in the sense that it binds the temporal development of each par-
ticle to the temporal development of strictly speaking all the other parti-
cles. This fact brings out clearly the problem how to vindicate that
dynamical structure as a network of concrete physical relations by
56 Matter points and their dynamics
contrast to an abstract mathematical structure that is part and parcel of
the means to represent the real physical relations and their evolution—
namely, the distance relations and their change—in a simple and informa-
tive manner. In a nutshell, the objection is that by committing itself to a
modal, dynamical structure existing in nature, ontic structural realism
reifies what is the mathematical representation of the evolution of the
physical relations to a mysterious structure behind the scene of the chang-
ing distance relations that binds the motions of the matter points together.
Again, this is evidence of the general claim made in this book: bringing in
more than matter points individuated by distance relations and change in
these relations creates new drawbacks instead of providing additional
explanatory value.
However, our concern here is not the metaphysics of modality. Humean-
ism enters the proposal sketched out here only as a strategy to combine sim-
plicity in ontology with simplicity in representation—in other words, as a
strategy to maintain scientific realism without building ontological commit-
ments on the representational means that physical theories employ. The
proposed minimalist ontology is neutral about the metaphysics of modality.
Due to the axiom that the distance relations between the matter points
change, one may in fact receive this ontology as implying that it is in the
nature of the distance relations to change, or in the nature of the matter
points to change the distances among them. Such a formulation leaves
open whether there is anything in the nature of the matter points or the dis-
tance relations that makes certain ways of their change necessary (cf. the
interpretation of Strawson (1989) of the historical Hume according to
which Hume denies only our epistemic access to necessary connections,
but not their very existence). In other words, we remain agnostic about
whether there is a reason (logos) driving the evolution of the cosmos.
Thus, the answer to the question why we are so successful in representing
the change that is accessible to us in terms of maximally simple laws that
are maximally informative about that change may be that there is a
reason in the way in which change occurs, but the answer may also be
that since change is eternal, every possible sequence of change in the dis-
tance relations among the permanent matter points actually occurs, and
beings that are able to reason about that change can only exist in a sequence
of change (say from a big bang to a big crunch) that is highly regular so that
it enables a representation in terms of maximally simple and maximally
informative dynamical equations.
The only claim that we make is that it is misguided to take the geometry
and the dynamical parameters that enter into the laws of our physical theories
for anything more than our attempts to achieve a description of the change in
the distances among the matter points that is both simple and informative
about that change; reifying the geometry and the dynamical parameters
leads to artificial problems instead of satisfying the demand for a deeper
explanation. That is why the ontology of the natural world is fully specified
 Matter points and their dynamics 57
by the minimalism expressed in the two axioms of there being matter points
individuated by distance relations and change of these relations.

Notes
1. The debate about weak discernibility goes back to Saunders (2006). As regards
the distance relations, see the exchange between Wüthrich (2009) and Muller
(2011). On the shortcomings of weak discernibility, see, notably, Dieks and Ver-
steegh (2008, p. 926), Ladyman and Bigaj (2010, p. 130) and, for a recent over-
view and argument, Bigaj (2015).
2. Castañeda (1980, p. 106) uses the term “super-Humean world”, meaning a view
that does not regard energy (or forces) as something that exists in the world, but
there is no rejection of absolute space or natural intrinsic properties considered in
Castañeda. We are grateful to Gordon Belot for suggesting the term “Super-
Humeanism” for our view of a relationalism that rejects intrinsic properties of
the spatial relata.
3. In this case, one cuts philosophical ice. To cut physical ice, one would have to
show how to do physics in terms of the geometry of space-time only, in spite
of the failure of Wheeler’s programme of geometrodynamics in the 1960s men-
tioned in section 2.1.
4. LeBihan (2016) proposes a super-relationalism that eliminates the objects stand-
ing in the distance relations, but keeps natural properties. The problem with this
view is that all the candidates for natural properties from physics (such as mass,
charge, etc.) are defined in terms of their function for the motion of physical
objects such as point particles—that is, their function for the evolution of the dis-
tance relations among primitive objects that can be considered as being individ-
uated by these very relations. Again, properties can be dispensed with in a
parsimonious ontology, but not objects that stand in the basic relations.
5. Furthermore, there are attempts to formulate quantum mechanics without the
need of wave functions at all, which are often termed many-interacting-worlds
approaches; see Schiff and Poirier (2012), Hall et al. (2014) and Sebens
(2015). The basic idea in these papers is to use continuous or discrete and
finite collections of Bohmian trajectories to sample the wave function by
means of their distribution and their velocities so that the wave function itself
can be discarded in the dynamics. These trajectories may be interpreted as
many coexisting classical worlds that on top of their inherent classical interaction
are subject to a quantum interaction arising from the Bohmian potential, and
quantum expectation values can be computed by means of Laplacian averages.
3 Minimalist ontology and
dynamical structure in classical
and quantum mechanics

3.1 From ontology to physics: two strategies for

classical mechanics
Let us take up the representation of the distance relations between matter
points in terms of vectors in Euclidean space that we have introduced in
general terms in Chapter 2, section 2. This representation allows us to
specify a law that records the change of the distance relations in terms of
usual derivatives:

d
vt ðQt Þ :¼ Q; for t7!Qt : ð3:1Þ
dt t

This law defines a vector field vt, called the “velocity field”, such that all
possible motions t 7! Qt are integral curves of vt.
As explained in Chapter 2, the task of physics then is to find a general
law of motion that captures the change in the distances among the matter
points. Hence, one can turn the definition in (3.1) around and study all
motions t 7! Qt that fulfill

d
Q ¼ vt ðQt Þ: ð3:2Þ
dt t

For a configuration of matter points represented in Euclidean space S ¼ R3 ,

using the same notation as in Chapter 2, section 2, to denote the motion
λ ! [Δλ]’ by t 7! Qt = (q1,t, . . . , qN,t), one may set up a theory of classical,
pre-relativistic gravitation as follows, working with a representation of the
matter points as being inserted into an absolute Euclidean background
space and as evolving in an absolute background time, as in Newtonian
mechanics. The velocity field in (3.2) given as vt ¼ ðv1;t ; :::; vN;t Þ 2 R3N is
ruled by the equations

d 1
v ðQÞ ¼  r VðQÞ; for all k ¼ 1; :::; N; and Q 2 R3N ; ð3:3Þ
dt k;t Mk k
60 Minimalist ontology and dynamical structure
where the map V : R3N ! R is given explicitly by

1X N X
m k mj
VðQÞ :¼  G ; ð3:4Þ
2 k¼1 j6¼k jqk  qj j

and G; M1 ; m1 ; :::; MN ; mN 2 Rþ are additional parameters. Recall that

jqk,t  qj,tj equals δ(Δkj,τ(t)) defined in Chapter 2, section 2. Together with
further parameters denoted by q_ k;0 2 R3 , the equations in (3.3) uniquely
determine a velocity field that is given by
 
vt ðQt Þ ¼ v1;t ðQt Þ; :::; vN;t ðQt Þ ; for
Z t  ð3:5Þ
1
vk;t ðQt Þ :¼ q_ k;0 þ  rk VðQs Þ ds:
0 Mk

These additional parameters—that is, the relationships described in (3.3)

and (3.4), Newton’s constant of gravitation G, inertial and gravitational
masses Mk and mk, and the initial velocities q_ k;0 —make up the dynamical
structure of this version of classical gravitation. They are the only degrees
of freedom left in the choice of an admissible velocity field vt. Thanks to
(3.2), specifying these parameters together with the initial configuration
Q0 2 R3ðN1Þ determines the motion t 7! Qt uniquely.
Huggett (2006) shows how one can understand Euclidean geometry and
Newtonian mechanics in a package as the Humean best system for a world
of classical mechanics that consists only in distances among point particles
and their change.1 As explained in Chapter 2, section 3, the idea is that if
one considers the evolution of the distance relations in the configuration
of matter points of the universe, the spatio-temporal geometry best suited
to describe the universe is fixed together with the dynamical laws by the
evolution that these relations take throughout the history of the universe.
Given the fact that the change in the distance relations manifests certain
salient patterns, Huggett (2006) singles out the notion of inertial motion
as the idea of a particularly regular and simple motion. He then defines
the notion of adapted reference frame as an assignment of real numbers
at a given time—that is, a set of d-tuples such that (i) the origin of the
frame—the (0, 0, . . . , 0) tuple—corresponds to the “position” at time
zero of the matter point to which the frame is adapted and (ii) the distances
along the d axes correspond to the distances from that matter point. The
choice of d (the dimensionality of space) and of the distance—that is, the
labeling of spatial relations (see definition 1 in Chapter 2, section 1)—is
totally arbitrary. The only constraint is that this choice enables the best
physical description: in the classical case, the best choice is d = 3 with the
distance given by the usual Euclidean formula.
An adapted frame is a bona fide relational concept in the sense that it
does not depend on any absolute concept, such as that of position in
 Minimalist ontology and dynamical structure 61
absolute space. In particular, it is well-suited for a Leibnizian-Humean
stance. Given the fact that its definition relies only on the notion of
matter point and arbitrary numerical assignments, an adapted frame can
be easily constructed within our minimalist ontology. In fact, when we
say that, for example, in a frame adapted to a particle i the simplest and
most informative assignment that individuates another particle j at time t
is a triple (x, y, z), we rely only on an arbitrary labeling of the instantaneous
configuration the reference particle i is in (which is a well-defined concept
given our Leibnizian view of time as a labeling for the succession of
states of the universal configuration of matter points given in terms of the
relative distances among the matter points), as well as on an arbitrary label-
ing of the distance relation between i and j. In short, an adapted frame is a
mere representational means that requires nothing more than the relation-
alist Humean mosaic.
However, relying on adapted frames only might not be enough for recon-
structing classical mechanics. For example, it might be the case that no
matter point can be adapted to an inertial reference frame, thus blocking
Huggett’s strategy. In order to obviate this problem, unoccupied frames
can be relationally defined from adapted ones by means of continuous
rigid spatial translations. Roughly, a spatial translation from a frame O
into another frame O0 amounts to shifting the set of d-tuples constituting
O by a certain factor (defined, for instance, by a suitable continuous function
that takes a d-tuple in O as input and gives back a d-tuple in O0 ), such that
the relative distances between particles remain unaltered in the new frame.
Also in this case, no ontological commitment going beyond those included
in axioms 1 and 2 is called for: a spatial rigid translation so defined, in
fact, relies only on numerical assignments and, hence, does not require any
pre-existing geometrical notion such as that of affine connection.
Obviously, the notion of rigid spatial translation makes it possible to
relate any kind of frame, independently of whether they are occupied or
not. With this machinery in place, an inertial frame can then be defined
as the frame in which the dynamical laws that supervene on the history
of relations for the entire universe hold. In this way, the notion of inertial
motion neither presupposes a substantival affine structure that singles out
straight trajectories nor an absolute external time to which the uniformity
of motion should be referred. Rather, these two structures supervene on
purely relational facts. By the same token, absolute acceleration is
reduced to the history of change of the spatial relations holding between
an inertial and a non-inertial frame. Similarly, the regularities in the
history of relations make it that Euclidean geometry is the simplest and
most informative geometry representing that history. Such a framework
clearly vindicates Leibnizian relationalism about space (spatial relations
and their change are the ontological bedrock) and time (temporal facts
supervene on the history—that is, an ordered sequence—of instantaneous
distance relations).
62 Minimalist ontology and dynamical structure
In the same vein, as already mentioned in Chapter 2, section 3, Hall
(2009, § 5.2) sketches out how dynamical parameters such as both inertial
and gravitational mass as well as charge can be introduced as variables that
figure in the laws of classical mechanics achieving the simplest and most
informative description of the change in the relative particle positions
throughout the history of the universe. Thus, the applicability of Super-
Humeanism to Newtonian mechanics confirms that there is no point in
reading off one’s ontological commitments from the dynamical structure
of physical theories: one can be a scientific realist with respect to Newto-
nian mechanics and yet be committed only to distance relations individuat-
ing matter points and the change of these relations in spite of the fact that
Newtonian mechanics employs an absolute Euclidean background space
and an absolute background time as well as various further dynamical
parameters to represent that change.
To mention a particularly striking example of this fact, consider a model
of Newtonian mechanics with an angular momentum of the universe J that
is greater than zero in the centre-of-mass rest frame. Obviously, a rotating
universe is not conceivable in a relationalist ontology that admits only dis-
tance relations among matter points, but no space in which the configura-
tion of matter of the universe is embedded. However, also in a Newtonian
ontology of a universe rotating in absolute space, the rotation of the uni-
verse would manifest itself in certain changes in the distance relations
among the point particles (e.g. in inhomogeneities in the cosmic microwave
background with respect to our point of view). Consequently, the relation-
alist is free to interpret a value of angular momentum of the universe that is
greater than zero as a convenient means to capture those changes in a
simple and informative manner, without being committed to a space in
which the universe rotates: describing those changes by using only variables
for the change of relative distances would lead to a law of motion that is
extremely complicated. Introducing further variables—such as an angular
momentum of the universe that can have a value greater than zero—by con-
trast, would enable the formulation of a law of motion that is simple and
elegant, but that is there only to capture the change in the relative distances
among the point particles. In this manner, the relationalist can handle all
kinds of bucket-like challenges.
Nonetheless, this strategy cannot recognize all the possible mathemat-
ical solutions of the dynamical equations of a physical theory as describing
physically possible situations. Axiom 1 and requirements (i) to (iv) of the
minimalist ontology set out in Chapter 2, section 1, pose also a constraint
on the dynamics: for instance, an evolution of the distance relations
among the matter points that ends up in an entirely symmetrical configu-
ration of the matter points of the entire universe is excluded in the same
way as is a symmetrical initial configuration of the entire universe. Such
solutions are a mathematical surplus of the formalism; the corresponding
 Minimalist ontology and dynamical structure 63
points in configuration space do not represent physically possible config-
urations of matter points of the entire universe. As argued in in Chapter
2, section 1, this is no objectionable restriction: having empirical ade-
quacy in mind, there is no need to admit, for instance, entirely symmetri-
cal worlds as physically possible worlds. As also argued there, this refusal
by no means diminishes the importance of symmetries to obtain a simple
and informative representation of the universal configuration of matter
and its evolution.
The Super-Humean strategy is a purely philosophical one: In a nutshell,
on this strategy, there is a relationalist ontology, but a non-relationalist
physical theory. This is no problem, since the non-relationalist theory can
be interpreted in a cogent manner that is consistent with scientific realism
as being committed to no more than the relationalist ontology.
However, one may wonder whether buying into all the formal appara-
tus of, say, Newtonian mechanics is necessary to achieve a description of
the change in the distance relations that is both simple and informative.
This reflection opens up the way for another strategy to link the minimal-
ist ontology up with physics that can be dubbed alternative theory strat-
egy: instead of endorsing physical theories as they stand—such as
Newtonian mechanics—and refusing to take their dynamical structure
as guide to the ontology, one constructs alternative physical theories
whose formal apparatus stays as close as possible to the ontology of
there being only distance relations among point particles and their
change and that matches the standard theories in their testable predic-
tions. In a nutshell, this strategy consists in building a relationalist phys-
ical theory on a relationalist ontology.
It is important to be clear about what the alternative theory strategy
can and what it cannot achieve: even if the ontology is exhausted by dis-
tance relations individuating matter points and the change of these rela-
tions, when it comes to formulating a law capturing that change,
further dynamical parameters have to be introduced, since there is
nothing about the distance relations making up any given configuration
of matter points that contains information about the—past and future—
change of these relations. Consequently, the alternative theory strategy
has to admit dynamical parameters such as mass, constants of nature,
initial momenta, etc., and cannot but resort to the Super-Humean strategy
in order to ban these parameters from the ontology. However, when it
comes to space and time, the aim of this strategy is to avoid quantities
that are tied to absolute space and time, such as empty space-time
points, absolute velocities, absolute accelerations, or absolute rotations.
As regards classical mechanics, the alternative theory strategy goes back
at least to Mach (1919). In the last three decades, it has been worked
out by Barbour and collaborators (see Barbour and Bertotti (1982) for
the seminal paper that laid down the “best-matching” framework).
64 Minimalist ontology and dynamical structure
Furthermore, Belot (1999) and Saunders (2013) have also proposed each a
relationalist theory of classical mechanics.
Let us discuss Belot’s proposal first (see also Pooley and Brown (2002),
section 5, for an appraisal of this proposal). Belot starts by considering the
Hamiltonian formulation of classical mechanics given in the language of
symplectic geometry. For this we take Q to be the configuration space.
The tangent bundle TQ is the collection of tuples (q, q̇), where q 2 Q
and q̇ is a tangent vector (velocity vector) of a curve through q. The cotan-
gent bundle T  Q is, strictly speaking, the dual bundle of TQ (momenta are
described as linear functionals of velocity vectors), which is the mathemat-
ical formulation of the so-called phase space. For simplicity (since both
spaces are isomorphic), we will write (q, q̇) for the elements of T  Q,
where q̇ is to be understood as the corresponding momentum (see
Frankel (1997), section 2.3c, for technical details). Phase space comes
equipped with a smooth function H, called the Hamiltonian, which—
roughly—represents the total energy of the system, and a two-form ω,
called the symplectic form. Glossing over the technical aspects, ω
renders it possible to define a map H 7! XH that associates to the Hamil-
tonian a smooth vector field XH over T  Q: such a map is nothing but an
intrinsic representation of the usual Hamilton’s equations. Hence, by inte-
grating XH given some initial conditions ðq0 ; q_ 0 Þ, we get the unique curve
in T  Q that represents the dynamical evolution of the system under scru-
tiny. In the case of N gravitating particles, T  Q will be nothing but R6N .
The justification of this fact is straightforward, if we consider what it
takes to determine an initial condition ðq0 ; q_ 0 Þ: for each particle, we
have to specify three numbers that indicate its position and further three
numbers that give the velocity vector “attached” to it. We immediately
see in what sense this framework naturally fits a substantivalist under-
standing of space: two N-particle states that agree on all relational facts
about the configuration (not only the relative positions but also the rela-
tive orientations of the velocities), but disagree on how such a configura-
tion is embedded in Euclidean space, would count as physically distinct
possibilities. Then, the natural relationalist move would be to construct
a relational configuration space Q0 by quotienting out from Q all the
degrees of freedom associated with an embedding in Euclidean space,
such as rigid translations and rotations. If we call E(3) the set of isometries
of Euclidean three-space, then Q0 ¼ Q=Eð3Þ: this construction assures us
that distinct points in Q that represent the same relational configuration
“collapse” to the same point in Q0 . Note that (i) Q0 admits a well-
defined cotangent bundle T  Q0 , which is equipped with a well-behaved
symplectic structure, and (ii) that the starting Hamiltonian defined on Q
admits a smooth projection H0 to T  Q0 because it is invariant under the
action of E(3).
Belot’s theory qualifies as relational, since the ontological facts making
up a set of initial data do not encode any notion of position in absolute
 Minimalist ontology and dynamical structure 65
space or absolute velocity, and the laws of motion specify how these initial
data evolve; furthermore, the dynamical laws of the theory are fully defined
on relational phase space. However, there are at least three concerns that
one can raise about Belot’s proposal. In the first place, this theory still
has a notion of absolute time inherent in the dynamical laws. In fact,
there is nothing in the quotienting out procedure that leads from a dynamics
over Q to one over Q0 that eliminates the absolute temporal metric of New-
tonian mechanics, which means that the very same succession of purely
relational configurations can unfold at different rates depending on the
ticking of an universal external clock. Secondly, there is a clear sense in
which spatial relations are Euclidean from the beginning: they are just
equivalence classes of embedding degrees of freedom as encoded in E(3).
Thirdly, Belot’s theory, despite being very close to Newtonian mechanics,
is not as empirically predictive as its absolute counterpart. This is
obvious because, if we think about all the initial data that are needed in
the Newtonian theory, we realize that they must include the rate of
change in the orientation of the configuration of N particles with respect
to absolute space: in the passage from Q to Q0 , this information is simply
washed away. In particular, the Newtonian theory admits models with
non-vanishing total angular momentum J of the universe. We repeat that
this is not just a metaphysical aspect, but a physical one in the sense that
the condition J 6¼ 0 carries with it empirically testable consequences.
Belot’s proposal to overcome the problem is just to bite the bullet: his rela-
tional reduction recovers only part of the Newtonian one, but—given that
up to now we have reliable experimental evidence that the total angular
momentum J of the universe is zero—it recovers exactly the empirically ade-
quate part.
Let us now turn to Barbour’s proposal (see Barbour and Bertotti (1982),
Barbour (2003, 2012) for the original resources; Pooley and Brown (2002),
sections 6–7, give an excellent overview of the framework, together with
some cogent philosophical considerations). Barbour’s relationalist motiva-
tions are the same as Belot’s—that is, to eliminate all the spatial degrees
of freedom that produce no observable difference. However, Barbour
extends this requirement to temporal degrees of freedom as well. By way
of consequence, the construction of his framework involves two steps—
namely, the implementation of (i) spatial and (ii) temporal relationalism.
As regards the first step, Barbour adopts the same strategy as Belot: he
takes standard configuration space Q and quotients out all Euclidean isom-
etries, comprised of scale transformations, which means that he quotients
out also the degrees of freedom related to “stretchings” or “shrinkings”
of configurations that preserve the ratio of distances. This means that he
considers a wider group than E(3)—namely, the similarity group Sim(3).
Hence, his relational configuration space Q0 ¼ Q=Simð3Þ is aptly called
shape space, because each configuration in there is individuated by its
form and not by its size.
66 Minimalist ontology and dynamical structure
The second step is technically more complicated: firstly, Barbour defines
an “intrinsic” difference that measures how much two shapes are similar.
This difference is expressed in terms of “best-matching” coordinates. Intu-
itively, we imagine the two shapes laid down over two distinct Cartesian
coordinate grids O and O0 ; then we hold fixed the first shape and grid
and “move” the second by applying transformations in Sim(3) until the
two shapes are juxtaposed as close as possible. The best-matching coordi-
nates are then defined as the overlap deficit O  O0 between the two coor-
dinate grids. Secondly, he uses this intrinsic metric to define a Jacobi action,
thus setting a Jacobi variational principle on Q0 (see Lanczos (1970), pp.
132–140, for a technical introduction to the Jacobi principle in classical
mechanics). The Jacobi action is reparametrization invariant—that is, it
does not change whatever “time” parameter we choose.
With this machinery in place, carrying out the variation of the action
with respect to the best-matching coordinates, we obtain a set of general-
ized Euler-Lagrange equations whose integral curves are nothing but the
geodesics of Q0 . Given some initial conditions, one of these curves is
singled out, which represents the dynamical evolution of the system. This
evolution is given in fully relationalist terms: the curve singled out by the
equations plus the initial conditions represents a list of relational configura-
tions, which is parametrized by an arbitrary monotonically increasing
parameter: hence, there is no external clock that measures dynamical
change; on the contrary, it is the change in the list of configurations that
enables an (arbitrary) parametrization. The important point is that there
exists a particular parametrization of the curve for which the generalized
Euler-Lagrange equations take the usual Newtonian form. Thus, if we
adopt this (again, arbitrary) parametrization, we obtain a dynamical
description that matches the Newtonian one. In this sense, Newtonian
mechanics comes out of Barbour’s framework by means of something
closely resembling a gauge fixing. The descriptive simplicity of the Newto-
nian formulation then explains why, historically, classical physics was
framed in these terms.
There are at least three critical points about this framework worth being
highlighted. Firstly, no usual Newtonian potential is compatible with the
condition of scale-invariance. Even if it is always possible to reproduce
the form of the most usual classical potentials in the appropriate gauge
by a clever mathematical manoeuvre, still this mimicking strategy might
lead to unwanted physical restrictions, such as no angular momentum
exchange between subsystems (see Anderson (2013), section 5.1.2, for a
technical discussion of this point). Secondly, the implementation of a geo-
desic principle on a general shape space might not always be that straight-
forward. In general cases, in fact, the quotienting out procedure sketched
earlier leads to a shape space whose global geometry is that of a stratified
manifold, where each “stratum” is a sub-manifold that can differ from
the others in many respects, including the dimensionality. It is then quite
 Minimalist ontology and dynamical structure 67
intuitive to understand that, if Q0 is a stratified manifold, it is problematic
to account for a dynamical evolution given in terms of a geodesic trajectory
that hits different strata of Q0 (see Anderson (2015), section 9.4 and refer-
ences therein, for discussion). The moral is that Barbour’s framework
works well in a suitably small region of Q0 , but might break down on a
larger scale, depending on the particular geometrical structure of Q0 .
Thirdly, as a result of quotienting out the group of rotations from Q, one
gets the condition J = 0 as a constraint on shape space dynamics.
However, even if the actual universe satisfies the condition of J = 0, a rela-
tionalist theory should be able to account for observable consequences
ascribable in absolute terms to a non-vanishing total angular momentum
of the universal configuration of matter.
Finally, let us consider the proposal spelled out in Saunders (2013). Like
Belot and Barbour, Saunders’s aim is to dispense with absolute quantities
of motion. However, unlike the former two, he also seeks to save the core
conceptual structure of Newton’s Principia. In order to do so, he shows
that the Newtonian laws can be cast in terms of directed distances represent-
ing inter-particle separations. This is possible because the absolute notion of
“straight trajectory” needed to make sense of inertial motion—which, in
turn, is required for rendering Newton’s first and second law meaningful—
involves too much structure—namely, a privileged affine connection (that
of neo-Newtonian space-time). Instead, Newton’s laws can make perfect
sense even if we replace the talk of straight trajectories with that of relative
velocities not changing over time, and this can be accounted for not just by a
single preferred connection, but by a whole class of affine connections whose
time-like geodesics are mutually non-rotating. This is all that Saunders needs
in order to account for accelerations and rotations in relationalist terms: a
space-time manifold equipped with enough structure to allow for the com-
parison of spatial directions (and related angles) at different times (what he
calls “Newton-Huygens” space-time; see also Earman (1989), pp. 31–32,
for a formal characterization of this space-time). In the Newton-Huygens
space-time, unlike the neo-Newtonian one, it is meaningless to talk about
the absolute acceleration of a particle, or even its inertial motion (two
notions that are tied to a privileged affine connection), while it is perfectly
meaningful to ask questions about the change of orientation of a configura-
tion in time.
The huge virtue of this framework is that it is able to recover the full
spectrum of Newtonian models, thus accounting also for J 6¼ 0 cases,
without invoking absolute notions. Given, in fact, that differences in direc-
tion can be defined relationally by admitting a primitive notion of parallel-
ism, and then defining change in direction by comparison of spatial
relations at different times, Saunders’s theory can account for global rota-
tions in terms of relational quantities. However, the fact that this theory
makes it meaningful to compare spatial directions at different times repre-
sents a substantial weakening of the relationalist programme. This
68 Minimalist ontology and dynamical structure
framework is much less relationalist than Belot’s and Barbour’s, for which
the excision of any physical meaning attached to global rotations is a con-
stitutive feature. In this sense, Saunders’s theory is a “halfway house” form
of a relationalism, as he himself notes (Saunders (2013), p. 44).
In sum, there are well-grounded reservations whether these relationalist
theories fully implement a relationalist ontology. That ontology is relation-
alist both with respect to space and time, whereas Belot’s and Saunders’s
theories are relationalist only with respect to space. Furthermore, that
ontology is not tied to a particular geometry such as Euclidean geometry:
a configuration of matter points satisfying axiom 1 and definition 1 of
Chapter 2, section 1, does not carry with it any primitive geometrical
fact that singles out a distinguished space in which it has to be embedded.
By contrast, there is a clear sense in which Belot’s and Barbour’s spatial
relations are Euclidean from the outset: each point in Q0 can be seen as
an equivalence class of Euclidean configurations; there is no way, by
fixing a certain gauge, to end up with a configuration embedded in a
non-Euclidean space. Also Saunders’s relations are inherently Euclidean,
since Newton-Huygens space-time encodes the structure of a series of
instantaneous 3-dimensional affine spaces equipped with an Euclidean
metric. Moreover, all these theories rely on more primitive structure than
just distances. The very concept of shape requires primitive facts about
angles to be meaningful, so that Barbour’s ontology has to include a confor-
mal structure. Saunders’s ontology requires not only distances, but directed
distances, which means that some primitive geometrical facts have to be
postulated (especially those making up a standard of space-like parallel
transport), which are encoded in an affine structure.
At this point, one may legitimately ask whether it is possible to resort to
the Super-Humean strategy to argue that the additional structure of these
theories is just part of the package we get when seeking for the best
description of motion. The answer is that such a move, in this case,
raises substantial worries. Put simply, a Humean justification of the
surplus structure would imply that what these theories do is basically to
“embellish” a set of relational initial data Δ0 by embedding them in a
more structured set Q0 and then using the equations of motion cast in
terms of this surplus structure to evolve these data until reaching the
result Qt, from which the relational solution Δλ would be read off. But
that would have unwanted implications. The first of these is that, in this
way, neither of the earlier theories could be considered even mildly rela-
tionalist anymore, being parasitic on a dynamical description that involves
an irreducible surplus structure. In such a case, it would be awkward to
prefer these theories to Newtonian gravitation, given that Huggett’s
Humean strategy works perfectly for the purposes of a relationalist ontol-
ogy in that context. The second implication is that the strategy of evolving
relational data by “stealing a ride” to a non-relationalist dynamics and
then discarding the surplus structure as a mere representational means
 Minimalist ontology and dynamical structure 69
would suspiciously look like a trivial instrumentalist move, as discussed
notably by Earman (1989, p. 128) and Belot (2000, p. 10).
In sum, combining a minimalist relationalist ontology with the alterna-
tive theory strategy faces a dilemma: either one insists that angles or direc-
tions are just part of the Humean package, thus ending up with—to
paraphrase Earman (1989, p. 128)—a cheap instrumentalist rip-off of a
theory that in any case does not qualify as a genuine relationalist compet-
itor to Newtonian mechanics; or one bites the bullet and introduces more
primitive structure in the ontology (be it a conformal or an affine one),
thus betraying the original motivations for relationalism from ontological
parsimony. That is why we prefer the Super-Humean strategy, as imple-
mented for Newtonian mechanics by Huggett (2006) and Hall (2009,
§ 5.2).
In our discussion of Newtonian mechanics, we have assumed that matter
is made out of a collection of identical matter points that stand in spatial
relations with respect to each other. However, classical mechanics has
also brought forward very successful theories that describe matter by
means of continuous phase space densities evolving in time; consider, for
instance, the Vlasov, Navier-Stokes or Euler equations. This rightfully
raises the question whether a description of matter in terms of matter
points is adequate at all. This question has been answered in the affirmative
by Boltzmann who linked Newtonian mechanics and thermodynamics by
means of statistical mechanics—that is, a statistical analysis of Newtonian
dynamics. Gases, fluids and solids are fundamentally described by a large
collection of discrete particles that move according to Newtonian mechan-
ics. However, in the mathematical idealization of the so-called thermody-
namic limit,—that is, sending the volume as well as the total particle
number to infinity while keeping the density constant—it is much more
easy to disregard the particular microscopic and often irrelevant motion
of the respective particles and switch to a more coarse-grained description
of the motion in terms of phase space densities, which may then be regarded
as a description of a cloud of particles. This description, however, does not
assign a special ontological status to the phase space density as the latter is
just a mathematical quantity derived from the particle positions in order to
achieve a convenient and mathematically tractable description of the
motion of the particles.

3.2 Bohmian quantum mechanics

When it comes to quantum mechanics, a discussion of ontology makes sense
only if one spells out to which version of quantum mechanics one is commit-
ted—that is, how one solves the measurement problem. As we mentioned in
the introduction, quantum theories that are committed to a distribution of
matter in physical space that accounts for measurement outcomes are
70 Minimalist ontology and dynamical structure
known as primitive ontology theories (see Allori et al. (2008)). The quantum
theory going back to de Broglie (1928) and Bohm (1952a) is the most prom-
inent of them. It is based on a primitive ontology of permanent point particles
as in classical mechanics. In the following, we draw on the predominant con-
temporary formulation of Bohm’s theory, known as Bohmian mechanics (see
Dürr et al. (2013b)). This theory conceives a dynamical law for the motion of
the particles in physical space (and not just for the quantum state, as does the
Schrödinger equation). It can be cast as a quantum theory of classical grav-
itation (since, however, the relevance of gravitation in the realm of light
elementary particles might seem questionable—and we use it only for point-
ing out the parallels with its classical version—one should rather think of this
theory as describing a gas of electric charges by replacing the gravitational
constant G with Coulomb’s constant (4π0)1 and the gravitational masses
mk with the electric charges ek).
In Bohmian mechanics, the t-dependence of υt is given as functional of a
map CðÞ : R  R3N ! C; ðt; qÞ7!Ct ðqÞ by
ℏ r C ðQ Þ
vk;t ðQt Þ :¼ Im k t t ; ð3:6Þ
Mk Ct ðQt Þ

where Ct is requested to be a square-integrable solution to the Schrödinger

equation
iℏ@ t Ct ¼ HCt ; ð3:7Þ

for the operator

X
N
ℏ2
H¼  D þ V: ð3:8Þ
k¼1
2Mk k

Here, ℏ 2 Rþ denotes an additional constant while G, Mk and V are the

same mathematical entities as in the preceding section. For a sufficiently
regular C0, equation (3.7) admits a unique solution
Ct :¼ eitH C0 ð3:9Þ

fulfilling Ct|t=0 = C0 and, therefore, together with (3.6), yields a unique

velocity field
 
vt ðQt Þ ¼ v1;t ðQt Þ; :::; vN;t ðQt Þ for
ð3:10Þ
ℏ r ðe C0 ÞðQt Þ
itH
vk;t ðQt Þ :¼ Im k itH :
Mk ðe C0 ÞðQt Þ

These additional parameters—that is, the relationships described in (3.6),

(3.7), and (3.4), Planck’s constant ℏ, Newton’s gravitational constant G,
the inertial masses Mk, and the initial wave function C0—make up the
 Minimalist ontology and dynamical structure 71
dynamical structure. They are the only degrees of freedom left in the choice
of a velocity field υt admissible in this Bohmian setup. Again, thanks to
(3.2), specifying these parameters together with the initial configuration
Q0 2 R3N determines the motion t 7! Qt uniquely.
Note the similarity between classical mechanics and Bohmian quantum
mechanics as set out here. Both theories fit into the framework of the
first order differential equation (3.2), which requires the specification of
the velocity field υt. All admissible candidates for υt can be found with
the help of only a few additional parameters and the relations—cf. (3.3)
and (3.6)–(3.7), respectively—which rule the t-dependence of υt  cf.
(3.5) and (3.10), respectively. Specifying further parameters—here the
initial wave function C0 instead of the initial velocities q_ k;0 —uniquely deter-
mines the velocity field υt, which in turn, for a given initial configuration
Q0, uniquely determines the motion t 7! Qt as solution of (3.2).
The fact that in classical mechanics υt itself is usually defined by spec-
ifying its derivative as in (3.3), which effectively results in a second order
equation, concerns only the way of formulating the dynamical structure
but does not affect the ontology. The same is possible for Bohmian
mechanics: Bohm (1952a) conceived the velocity law for the particles as
a second order equation. Doing so does not change the role of the
introduced quantities: given υt, the fundamental law is (3.2). This law
describes the change in the configuration of matter points Qt. The
ontology that we propose consists in the configuration of matter points
individuated by distance relations and the change of these relations. The
dynamical parameters that a physical theory introduces are mathematical
variables that are employed in order to find a good candidate for υt. Thus,
beside the other constants that have to be fixed, the initial velocities in
classical mechanics are just parameters that help to easily find an admis-
sible υt thanks to (3.5). The role of the initial velocities is taken over by
the initial wave function in quantum theory. In the same vein, besides
the other constants, it determines the admissible υt by (3.10). Therefore,
also the initial wave function just is a mathematical variable that helps
to carry out the task of finding a suitable candidate for υt. In the second
order formulation of Bohmian mechanics, the initial velocities q_ k;0
appear as additional parameters. But these parameters are redundant,
since, in order to reproduce the statistical predictions of quantum mechan-
ics, they have to be fixed in accordance with (3.6); there is no empirical
data that would justify or require a different choice.
On this view, hence, forces are not explanatory in the sense that they
describe agent-like entities that push the particles around. Consequently,
from the perspective of a parsimonious ontology of matter points and
change in their configuration, it makes no sense to argue that the second
order formulation of Bohmian mechanics in terms of forces, including a
specific quantum force, is explanatory superior to the first order formula-
tion of Dürr et al. (2013b) on which we rely here (see Belousek (2003)
72 Minimalist ontology and dynamical structure
for that debate). In a nutshell, the formulation of Bohmian mechanics in
terms of forces does not add explanatory value, but, to the contrary,
creates new drawbacks, since the forces would have to coordinate the
motion of the particles instantaneously all over space instead of propagat-
ing in space (cf. Chapter 2, section 3, and in particular the combination of
Bohmian mechanics and Humeanism about dynamical structure set out
there; by contrast, see Suárez (2015) for a proposal for an ontology for
Bohmian mechanics in terms of a multitude of dispositions instantiated
by each individual particle).
Nonetheless, the wave function Ct may strike one as an odd dynamical
parameter in an ontology of matter points that are characterized only by
their distance relations, so that any dynamical parameter is there to
capture the change in the distance relations: the wave function does not
care about the actual particle positions—that is, the distance relations
among the matter points. Moreover, in quantum physics, all the dynamical
parameters are situated on the level of the wave function, including mass
and charge. That notwithstanding, as is evident from equation (3.10), the
wave function has the job to yield a velocity field along which the particles
move as output, given any possible configuration of matter points as repre-
sented by a point in configuration space as input. In other words, in being
defined for any possible particle configuration admitted in configuration
space, the wave function enables us to obtain the velocity of the actual par-
ticle configuration as output, if that configuration is given as input. Conse-
quently, although all the dynamical parameters are situated on the level of
the wave function, it would be wrong-headed to say that the interaction
takes place on the level of the wave function: interaction is the correlated
change of the particles’ velocities. That correlated change in physical
space is represented by means of the wave function in configuration space
and the dynamical parameters that are situated on its level. The wave func-
tion then serves to determine the velocity of the particles, as described in
equation (3.10), whereby the interactions between the matter points enter
into the theory via the Hamiltonian operator (3.8). Since the equation of
motion for the particles is a first order differential equation, it has a
unique solution for any initial particle configuration (assuming the vector
field generated by the wave function is sufficiently “nice”). In this sense,
one can say that Bohmian mechanics replaces the specification of an
initial velocity in classical mechanics with the specification of an initial
wave function. Given the wave function and the particle configuration at
a certain time, their evolution is fixed for all times.
Let us now briefly consider how the Super-Humean and the alternative
theory strategy fare in Bohmian mechanics. As regards the former,
we already mentioned the Humean treatment of the wave function in
Chapter 2, section 3. Space-time in Bohmian mechanics poses no problem
for Humeanism, given that the laws of Bohmian dynamics, although
being different from the Newtonian ones, are nonetheless formulated
 Minimalist ontology and dynamical structure 73
over an absolute Newtonian spatio-temporal background: the manner in
which Huggett (2006) deals with space-time in the Newtonian case can
simply be applied to the Bohmian case. As regards the alternative theory
strategy, the proposal of Belot (1999) would require forcing the Bohmian
formalism in a Hamiltonian context; this is possible, but results in a quite
unreasonably complicated formalism with a huge amount of surplus
descriptive structure, which is not needed by Bohmian dynamics (see
Holland (2001a,b) for a decently worked out Hamiltonian version of
Bohmian mechanics). Barbour’s framework, by contrast, can in a natural
way be applied to Bohmian mechanics (see Vassallo (2015), Vassallo and
Ip (2016)). The same goes for the milder relationalist approach of Saunders
(2013) since this approach would basically amount to re-write the Bohmian
theory in a way that makes it meaningless to refer distances and directions
to any point taken as “origin”.
Bohmian mechanics illustrates very clearly the point made at the beginning
of Chapter 2, section 1—namely, that even the classical parameters mass and
charge are situated on the level of the wave function. Consider the following
experiment: imagine a charged particle whose (effective) wave function is of
the form ψ = ϕA + ϕB, where ϕA and ϕB are of equal size and shape but con-
centrated on two distant regions of space that we denote by A and B, respec-
tively. (These regions could be surrounded by infinite high potential walls—
or, more simply put, a box—to keep the wave function from spreading.) The
particle is located in one of these regions, let’s say in A. Not surprisingly, the
trajectory of a second charged particle passing near A is affected by the elec-
tromagnetic interactions and deflected towards A, if it has opposite charge, or
away from A if it has equal charge as particle one. However, if that second
particle were passing near region B, it would be affected in the very same
way, no matter how far away that is from the actual position of the other
particle. This scenario demonstrates, firstly, the explicitly non-local character
of Bohmian mechanics. It also shows that it would be wrong to think of
charge in the familiar way as something localized at the position of the par-
ticles. A similar reasoning would apply to the particle mass, in so far as grav-
itational interactions play a role in quantum mechanics (see Brown et al.
1995, 1996; Pylkkänen et al. 2015).
Furthermore, Bohmian mechanics illustrates that it is a contingent
matter of fact that the dynamical structure represents the matter points as
being sorted into different particle species. One can take the contingency
of this fact into account by formulating the physical theory in such a way
that not only the primitive ontology but also the dynamical structure is
not committed to different particle species from the outset. Goldstein
et al. (2005a,b) have shown how to do this for Bohmian mechanics. They
reformulate Bohmian mechanics in such a way that its dynamics reflects
the ontological commitment to propertyless particles, treating all particles
as identical. To appreciate what this means and how the reformulation is
carried out, let us begin with the following observation. If we insist that
74 Minimalist ontology and dynamical structure
particles are distinguished only by their positions instead of by intrinsic
properties, we note that the configuration space R3N has too much mathe-
matical structure in that it cares about permutations of the particle labels.
That is to say the following: the nature of the Bohmian law of motion
(being a first-order differential equation on configuration space) is such
that it determines at every time t the change of the system’s spatial config-
uration depending on the current configuration Q(t). However, there are
neither intrinsic properties nor relations distinguishing the configuration
represented by the tuple (Q1, Q2, . . ., QN) from, say, the configuration rep-
resented by the tuple (Q2, Q1, . . ., QN) with the particles 1 and 2 inter-
changed. It is thus understood that—for so-called identical or
indistinguishable particles—the natural configuration space of an N-parti-
cle system is not R3N , but
N 3


R :¼ S  R3 j ]S ¼ N ; ð3:11Þ

which is the set of all subsets of R3 containing exactly N elements.

Consequently, the wave function of the system should now be defined on
the configuration space N R3 as well, which in fact can be done (see Gold-
stein et al. (2005a), section 4). Nevertheless, it is still more convenient, in
general, to represent the quantum state as a function on R3ðNÞ (which can
be regarded, mathematically, as the universal covering space of N R3 ). As
long as we consider a system in which all particles are associated with
the same mass and charge, the demand of consistency then leads immedi-
ately to a wave function that is symmetric or anti-symmetric under permu-
tations of the particle coordinates and hence to the famous boson/fermion
alternative. In Dürr et al. (2006), N R3 was thus already introduced as the
configuration space of identical or indistinguishable particles, referring to
a single species of particles, and it is shown how the quantum statistics of
identical particles thus arise in the Bohmian theory (see also Dürr and
Teufel (2009), ch. 8.5).
However, we now note that as soon as we have to admit more than
one value for the parameters Mk, the standard formulation of Bohmian
mechanics breaks down. That is because equation (3.6) no longer
defines a law of motion on N R3 , since it discriminates different particles
by their associated mass, while configurations represented on N R3 do not
do so. The basic idea of Goldstein et al. (2005a,b) is thus to symmetrize
equation (3.6) in order to get a permutation invariant equation, because
any permutation invariant equation on R3N defines, in a canonical way, a
law of motion on N R3 , the configuration space of identical particles. In
this way, they show that we can treat all particles as identical, while
still accounting for the empirical data that, as usual, are explained in
terms of a particle “zoo”.
To preserve equivariance of the law—i.e. the conservation of total prob-
ability by the Bohmian flow—the symmetrization has to be done in the
 Minimalist ontology and dynamical structure 75
following way. For the rest of this section, we adopt the notation of Gold-
stein et al. (2005a,b) to avoid mixing the aforementioned label-dependent
definitions with the label-independent approach here. The standard
guiding equation (3.6) can be written in the form
dQ jðQðtÞÞ
¼ ; ð3:12Þ
dt rðQðtÞÞ

where
r ¼ c c

is the probability density and j = (j1, . . ., jN) with

ℏ
ji ¼ Im c ri c
Mi

the probability current corresponding to the system’s wave function ψ. In

equation (3.12), numerator and denominator have to be symmetrized inde-
pendently by summing over all possible permutations of the particle labels
1, . . ., N. Hence, we get a new, permutation invariant guiding equation,
which reads
P
dQk s2S jsðkÞ s
¼ P N ðQðtÞÞ: ð3:13Þ
dt s2SN r s

Here, the sum goes over all elements of the permutation group SN and

sQ :¼ Qsð1Þ ; :::; QsðNÞ

means that every coordinate Qi is assigned a new index Qσ(i), changing the
order in the N-tupel.
In this theory, which Goldstein et al. (2005a,b) dubbed identity-based
Bohmian mechanics, we do not attribute a priori any mass to any specific
particle. The law of motion merely determines N trajectories for N parti-
cles, and it is a characteristic of this law that one of those trajectories
happens to behave—at least in the relevant circumstances—like the trajec-
tory of a particle with mass M1, another like the trajectory of a particle with
mass M2, and so on, depending only on the (contingent) initial conditions of
the system, respectively the universe.
To illustrate how this works, let us discuss an example given in Goldstein
et al. (2005a, section 3), which compares the standard formulation of
Bohmian mechanics with the identity-based version. Consider a two-parti-
cle universe consisting of an electron with mass Me and a muon with mass
Mμ. Suppose, for simplicity, that they are in a non-entangled state C(q1, q2)
= ϕ(q1)χ(q2) (note that we could symmetrize this wave function, though this
76 Minimalist ontology and dynamical structure
would be redundant when plugged into the symmetrized guiding equation).
Then, the standard guiding law (3.6) leads to the following equations of
motion:
dQ1 ℏ rðQ1 Þ
¼ Im ;
dt Me ðQ1 Þ
ð3:14Þ
dQ2 ℏ rwðQ2 Þ
¼ Im :
dt Mm wðQ2 Þ

In contrast, the symmetrized guiding equation (3.13) reads

2 2
dQ1
ℏ
Me
jwðQ2 Þj Im ð ðQ1 ÞrðQ1 ÞÞ þ ℏ
Mm
jðQ2 Þj Im ðw ðQ1 ÞrwðQ1 ÞÞ
¼ 2 2 2 2 ;
dt jðQ1 Þj jwðQ2 Þj þ jðQ2 Þj jwðQ1 Þj
ð3:15Þ
2 2
dQ2
ℏ
Mm
jðQ1 Þj Im ðw ðQ2 ÞrwðQ2 ÞÞ þ ℏ
Me
jwðQ1 Þj Im ð ðQ2 ÞrðQ2 ÞÞ
¼ 2 2 2 2 :
dt jðQ1 Þj jwðQ2 Þj þ jðQ2 Þ wðQ1 Þj

We see that equation (3.14) can easily be taken to suggest that there is an
intrinsic mass and thus a distinct type to every particle: particle 1, described
by the coordinates Q1, is the electron with mass Me, while particle 2,
described by the coordinates Q2, is the muon with mass Mμ. In equation
(3.15), by contrast, neither Q1 nor Q2 is designated as the position of the
electron, respectively the muon. A priori, the two particles are distinguished
only by the position that they have at time t. However, if we consider a
situation in which ϕ and χ have disjoint support, say, when one wave
packet is propagating to the left and the other one to the right, one of
the two sums in the nominators and denominators will be zero, so that
the equation of motion effectively reduces to equation (3.14) (possibly
with the indices 1 and 2 interchanged). This is to say, in particular, that
in situations where the two-particle wave function is suitably decohered,
one of the particles will play the role of the electron—being effectively
described by equations (3.6) and (3.7) with the parameter Me, while the
other one will play the role of the muon—being effectively described by
equations (3.6) and (3.7) with the parameter Mμ.
Which trajectory turns out to be guided by which part of the wave func-
tion thereby depends only on the law of motion and the (contingent) initial
conditions of the system, rather than on intrinsic properties of the particles.
In fact, if both parts of the wave function were brought back together and
then separated again, one and the same particle could switch its role from
being the electron to being the muon, and vice versa. Hence, to be an elec-
tron, a muon, or a positron, etc., is nothing more than to move—in the rel-
evant circumstances—electronwise, muonwise, or positronwise, and so
forth.
 Minimalist ontology and dynamical structure 77
Apart from such circumstances in which the different parts of the wave
function are well separated, one could say that the particles in the previous
example are guided by a superposition of (what one would usually call) an
electron wave function and a muon wave function. However, it would be
misleading to claim that this amounts to a superposition of being an elec-
tron and being a muon. Ontologically, there are no superpositions of any-
thing, only propertyless particles moving on definite trajectories. Rather,
the labels “electron”, “muon”, etc., are meaningless in the general case.
One obvious objection to the move proposed by Goldstein et al. (2005a,
b) is that the guiding law (3.13) is much more contrived than the one in
standard Bohmian mechanics. This is again an illustration of the fact that
simplicity in ontology and simplicity in representation pull in opposite
directions. That notwithstanding, a few things can be said to address the
concern of buying into too complicated a formalism. First, one should
note that the apparent complexity of equation (3.13) is really just the
price for expressing a law of motion for configurations in N R3 on the coor-
dinate space R3N and does not automatically amount to more complicated
physics. Second, it should be noted that (modulo some subtleties discussed
by Goldstein et al. (2005a,b)), the symmetrized theory will give rise to the
familiar statistical description of subsystems in terms of effective wave func-
tions, which is all that matters for most practical purposes.
In this context, it should also be noted that, given the universal wave
function, the “right” statistical description of subsystems—that is, the
one agreeing with the predictions of standard quantum mechanics, arises
for typical initial conditions in terms of the particle configuration—that
is, in quantum equilibrium (see section 4). Hence, the emergence of differ-
ent particle types as empirically observed in nature is not attributed to
special initial conditions (quite the opposite), though it is explained by
the particular form of the universal wave function.
Finally, concerning the (empirical) content of the proposed theory, it
should be emphasized that the trajectories described by identity-based
Bohmian mechanics will in general differ from those obtained from stan-
dard Bohmian mechanics, but that the statistical predictions for experimen-
tal outcomes are the same. In this sense, the symmetrized theory is
empirically equivalent to Bohmian mechanics and hence empirically equiv-
alent to standard quantum mechanics.

3.3 The GRW quantum theory

When it comes to the ontology of quantum physics, the argument for
endorsing Bohmian mechanics is not that its ontology matches the one of
classical mechanics, but that it provides the best solution to the measure-
ment problem. As mentioned in the introduction, we take the arguments
by Bell (2004, ch. 7) and Maudlin (2010, 2015) among others for solving
78 Minimalist ontology and dynamical structure
this problem in terms of what is known as a primitive ontology of a spatial
configuration of matter and its evolution in physical space to be convincing.
This strategy accounts for the experimental evidence in terms of there
always being a well-defined configuration of matter in physical space; the
quantum state then is the means to represent the evolution of that configu-
ration (see notably Allori et al. (2008)). In other words, there are no super-
positions of anything in physical space, although, of course, the dynamics of
the evolution of the configuration of matter in physical space is not classi-
cal. Bohmian mechanics is the most well known and the most elaborate
implementation of this strategy.
Nonetheless, if one pursues this strategy, there are other options avail-
able than endorsing a primitive ontology of permanent particles. In partic-
ular, instead of subscribing to a discrete ontology of matter points, one can
try out a continuous ontology of gunk—that is, a continuous matter density
field. Let us therefore consider now what is known as the GRWm theory.
This theory combines the quantum dynamics proposed by Ghirardi,
Rimini and Weber (GRW) (see Ghirardi et al. (1986)) and further devel-
oped by Ghirardi et al. (1990) (see also Gisin (1984, 1989)) with the prim-
itive ontology of a matter density field stretching out throughout an
absolute background space (see Ghirardi et al. (1995)).
In GRW, the evolution of the wave function Ct is given by a modified
Schrödinger equation. The latter can be defined as follows: the wave func-
tion undergoes spontaneous jumps at random times distributed according
to the Poisson distribution with rate Nλ. Between two successive jumps
the wave function Ct evolves according to the usual Schrödinger equation.
At the time of a jump the kth component of the wave function Ct undergoes
an instantaneous collapse according to
1=2
ðLxxk Þ Ct ðx1 ; :::; xk ; :::; xN Þ
Ct ðx1 ; :::; xk ; :::; xN Þ 7! 1=2
; ð3:16Þ
k ðLxxk Þ Ct k

where the localization operator Lxxk is given as a multiplication operator of

the form

1 1 2
Lxxk :¼ e2s2 ðxk xÞ ; ð3:17Þ
ð2ps Þ 2 3=2

and x, the centre of the collapse, is a random position distributed according

1=2
to the probability density pðxÞ ¼k ðLxxk Þ Ct k2 . This modified Schrödinger
evolution captures in a mathematically precise way what the collapse pos-
tulate in textbook quantum mechanics introduces by a fiat—namely, the
collapse of the wave function so that it can represent localized objects in
physical space, including in particular measurement outcomes. GRW
thereby introduce two additional parameters, the mean rate λ as well as
 Minimalist ontology and dynamical structure 79
the width σ of the localization operator, which can be regarded as new con-
stants of nature whose values can be inferred from (or are at least bounded
by) experiments (such as chemical reactions on a photo plate, double slit
experiments, etc.). An accepted value of the mean rate λ is of the order
of 1015s1. This value implies that the spontaneous localization process
for a single particle occurs only at astronomical time scales of the order
of 1015s, while for a macroscopic system of N * 1023 particles, the collapse
happens so fast that possible superpositions are resolved long before they
would be experimentally observable. Moreover, the value of σ can be
regarded as localization width; an accepted value is of the order of
107m. The latter is constrained by the overall energy increase of the
wave function of the universe that is induced by the localization processes.
However, it is obvious that modifying the Schrödinger equation is, by
itself, not sufficient to solve the measurement problem: to do so, one has
to answer the question of what the wave function and its evolution repre-
sents. One therefore has to add to the GRW equation a link between the
evolution of the mathematical object Ct in configuration space and the dis-
tribution of matter in physical space in order to account for the outcomes of
experiments and, in general, the observable phenomena. Ghirardi et al.
(1995) accomplish this task by taking the evolution of the wave function
in configuration space to represent the evolution of a matter density field
in physical space. This then constitutes what is known as the GRWm
theory and amounts to introduce in addition to Ct and its time evolution
a field mt(x) on physical space R3 as follows:

X
N Z
3 3 2
mt ðxÞ ¼ Mk d x1 : : :d xN d3 ðx  xk ÞjCt ðx1 ; :::; xN Þj : ð3:18Þ
k¼1

This field mt(x) is to be understood as the density of matter in physical space

R3 at time t (see Allori et al. (2008), section 3.1). Hence, on this theory,
despite its formulation in terms of particle numbers, there are no particles
in the ontology. More generally speaking, there is no plurality of fundamen-
tal physical systems. There is just one object in the universe—namely, a
matter density field that stretches out throughout space and that has
varying degrees of density at different points of space, with these degrees
of density changing in time.
By introducing two new dynamical parameters—lambda and sigma—
whose values have to be put in by hand, the GRW theory abandons the sim-
plicity and elegance of the Schrödinger equation and the Bohmian guiding
equation, without amounting to a physical benefit (there is of course a
benefit in comparison to stipulating the collapse postulate by a simple
fiat, but doing so is no serious theory). Indeed, there is an ongoing contro-
versy whether the GRWm ontology of a continuous matter density field that
80 Minimalist ontology and dynamical structure
develops according to the GRW equation is sufficient to solve the measure-
ment problem. The reason is the so-called problem of the tails of the wave
function. This problem arises from the fact that the GRW theory mathemat-
ically implements spontaneous localization by multiplying the wave func-
tion with a Gaussian, such that the collapsed wave function, although
being sharply peaked in a small region of configuration space, does not
actually vanish outside that region; it has tails spreading to infinity. In
the literature starting with Albert and Loewer (1996) and P. Lewis
(1997), it is therefore objected that the GRW theory does not achieve its
aim—namely, to describe measurement outcomes in the form of macrophy-
sical objects having a definite position. However, there is nothing indefinite
about the positions of objects according to GRWm. It is just that an
(extremely small) part of each object’s matter is spread out through all of
space. But since the overwhelming part of any ordinary object’s matter is
confined to a reasonably small spatial region, we can perfectly well
express this in our (inevitably vague) everyday language by saying that
the object is in fact located in that region (see Monton (2004), pp. 418–
419, and Tumulka (2011)). Thus, the GRWm ontology offers a straightfor-
ward solution to what Wallace (2008, p. 56) calls the problem of bare tails.
However, there is another aspect, which is known as the problem of
structured tails (see Wallace (2008), p. 56). Consider a situation in which
the pure Schrödinger evolution would lead to a superposition with equal
weight of two macroscopically distinct states (such as a live and a dead
cat). The GRW dynamics ensures that the two weights do not stay equal,
but that one of them (e.g. the one pertaining to the dead cat) approaches
unity while the other one becomes extremely small (but not zero). In
terms of matter density, we then have a high-density dead cat and a low-
density live cat. The problem is that it seems that the low-density cat is
just as cat-like (in terms of shape, behaviour, etc.) as the high-density cat,
so that in fact there are two cat-shapes in the matter density field, one
with a high and another one with a low density. There is an ongoing con-
troversy about this problem: Maudlin (2010, pp. 135–138) takes it to be a
knock down objection against the GRW matter density ontology, whereas
others put forward reasons that aim at justifying to dismiss the commitment
to there being a low density that is as cat-like as the high-density cat in the
matter density field (see notably Wallace (2014), Albert (2015), pp. 150–
154, and Egg and Esfeld (2015), section 3).
Be that as it may, there arguably is another, more important drawback of
the GRW dynamics that concerns the meaning of the spontaneous localiza-
tion of the wave function in configuration space for the evolution of the
matter density field in physical space. To illustrate this issue, consider a
simple example—namely, the thought experiment of one particle in a box
that Einstein presented at the Solvay conference in 1927 (the following pre-
sentation is based on de Broglie’s version of the thought experiment in de
Broglie (1964), pp. 28–29, and on Norsen (2005)): the box is split in two
 Minimalist ontology and dynamical structure 81
halves which are sent in opposite directions, say from Brussels to Paris and
Tokyo. When the half-box arriving in Tokyo is opened and found to be
empty, there is on all accounts of quantum mechanics that acknowledge
that measurements have outcomes a fact that the particle is in the half-
box in Paris.
On GRWm, the particle is a matter density field that stretches over the
whole box and that is split in two halves of equal density when the box
is split, these matter densities travelling in opposite directions. Upon inter-
action with a measurement device, one of these matter densities (the one in
Tokyo in the example given earlier) vanishes, while the matter density in the
other half-box (the one in Paris) increases so that the whole matter is con-
centrated in one of the half-boxes. One might be tempted to say that some
matter travels from Tokyo to Paris; however, since it is impossible to assign
any finite velocity to this travel, the use of the term “travel” is inappropri-
ate. For lack of a better expression let us say that some matter is delocated
from Tokyo to Paris (this term has been proposed by Matthias Egg, see Egg
and Esfeld (2014), p. 193); for even if the spontaneous localization of the
wave function in configuration space is conceived as a continuous process
as in Ghirardi et al. (1990), the time it takes for the matter density to dis-
appear in one place and to reappear in another place does not depend on
the distance between the two places. This delocation of matter, which is
not a travel with any finite velocity, is quite a mysterious process that the
GRWm ontology asks us to countenance.
On Bohmian mechanics, by contrast, in this example, there always is one
particle moving on a continuous trajectory in one of the two half-boxes,
and opening one of them only reveals where the particle was all the time.
In other words, Bohmian mechanics provides a local account of the case
of a particle in a box. However, when moving from Einstein’s thought
experiment with one particle in a box (1927) to the EPR experiment (Ein-
stein et al. (1935)), even Bohmian mechanics can no longer give a local
account, as proven by Bell’s theorem (Bell (2004), ch. 2; see also notably
chs. 7 and 24). On the GRWm theory, again, the measurement in one
wing of the experiment triggers a delocation of the matter density, more
precisely a change in its shape in both wings of the experiment, so that,
in the version of the experiment by Bohm (1951, pp. 611–622) the shape
of the matter density constitutes two spin measurement outcomes. On
Bohmian mechanics, fixing the parameter in one wing of the EPR experi-
ment influences the trajectory of the particles in both wings via the wave
function of the whole system, which consists of the measured particles as
well as of the particles that make up the measuring devices.
Hence, in this case, it clearly comes out that according to the Bohmian
velocity equation (3.10), the velocity of any particle depends strictly speak-
ing on the position of all the other particles. However, each particle always
moves with a determinate, finite velocity so that its motion traces out a con-
tinuous trajectory, without anything jumping—or being delocated—in
82 Minimalist ontology and dynamical structure
physical space. The best conjecture for a velocity field that captures this
motion that we can make—namely (3.10), requires acknowledging that
the motions of these particles are correlated with each other, but this
does not imply a commitment to there being some spooky agent or force
in nature that instantaneously coordinates the motions of all the particles
in the universe. Quantum physics just teaches us that it is a fact about
the universe that when we seek to fill in (3.2) with a simple and general
law that accounts for the empirical evidence, we have to write down a
law that represents the motions of the particles to be correlated with one
another (cf. the remarks and references on Bohmian Super-Humeanism in
Chapter 2, section 3).
Again, our proposal for a minimalist ontology makes intelligible why
correlated motion is in principle to be expected: if all the change is
change in the distance relations among matter points in a given configura-
tion of matter points, one cannot change the distance between two matter
points without thereby also affecting the distances among in principle all
the other matter points in the configuration. Thus, as mentioned in
Chapter 2 at the end of section 2, this ontology accounts for non-local cor-
relations without it making sense to call for something that transmits
instantaneous particle interaction across space. Nonetheless, this is a
general account of correlated motion, not a specific account of the
quantum correlations. Since our minimalist ontology accepts the change
in the distance relations as a primitive matter of fact (axiom 2 in
Chapter 2, section 1), it takes explanations to end in distance relations
and their change, arguing that anything beyond distance relations and
their change that one endorses in the ontology creates new drawbacks
instead of providing deeper explanations.
However, Einstein (1948) is certainly right in pointing out that a com-
plete suspension of the principles of separability and local action would
make it impossible to do physics: a theory that says that the motion of
any object is effectively influenced by the position of every other object in
the configuration of matter of the universe would be empirically inadequate
and rule out any experimental investigation of nature. In order to meet Ein-
stein’s requirement, it is not necessary to rely on a dynamics of the collapse
of the wave function, as does GRW. Bohmian mechanics fulfills this condi-
tion because decoherence will in general destroy the entanglement between
large and/or distant systems, allowing to treat them, for all practical pur-
poses, as evolving in an independent manner. Moreover, while accounting
for all phenomena of non-relativistic quantum mechanics, Bohmian
mechanics is able to recover classical behaviour in the relevant regimes
(see Dürr et al. (2013b), ch. 5). Since Bohmian mechanics is a theory
about the motion of particles, this classical limit does not involve or
require any change in the ontological commitment, but consists in the prop-
osition that typical Bohmian trajectories look approximately Newtonian on
macroscopic scales (if the characteristic wave length associated to ψ is small
 Minimalist ontology and dynamical structure 83
compared to the scale on which the interaction potential varies). Alto-
gether, the Bohmian theory, against the background of an ontology of
matter points that are characterized only by their relative positions and a
dynamics for the change of these positions, illustrates that there is
nothing suspicious about a non-local dynamics.
Apart from the matter density ontology, there is another ontology for the
GRW theory available. This ontology goes back to Bell (2004, ch. 22, orig-
inally published 1987): whenever there is a spontaneous localization of the
wave function in configuration space, that development of the wave func-
tion in configuration space represents an event occurring at a point in phys-
ical space. These point events are known as flashes; the term “flash”,
however, is not Bell’s, but was coined by Tumulka (2006, p. 826). Accord-
ing to the GRW flash theory (GRWf), the flashes are all there is in physical
space. Macroscopic objects are, in the terms of Bell (2004, p. 205), galaxies
of such flashes. Consequently, the temporal development of the wave func-
tion in configuration space does not represent the distribution of matter in
physical space. It represents the objective probabilities for the occurrence of
further flashes, given an initial configuration of flashes. Hence, there is no
continuous distribution of matter in physical space—namely, no trajectories
of particles—and no field such as a matter density field either. There only is
a sparse distribution of single events. Although GRWf and GRWm are rival
proposals for an ontology of the same formalism (the GRW quantum
theory), there also is a difference between them on the level of the formal-
ism: if one endorses the GRWm ontology, it is reasonable to develop the
GRW equation into a formalism of a continuous spontaneous localization
of the wave function (as done in Ghirardi et al. (1990)); by contrast, if one
subscribes to the GRWf ontology, there is no point in doing so.
Whereas Bohmian mechanics vindicates an ontology of discrete objects
(matter points, particles) in quantum physics and GRWm vindicates an
ontology of gunk, it may seem that GRWf vindicates an ontology of
super-substantivalism. In order to answer the question of what the flashes
are, one may be inclined to say that they are properties of space: in brief,
space flashes at some points. However, this is just a linguistic trick in the
sense of Sklar (1974, pp. 166, 222–223): one simply changes the language
from using substantives that characterize objects in space to verbs that are
predicated of points of space. However, the verb “to flash” does not express
a bona fide property of space, with topological, affine and metrical proper-
ties setting the paradigm for what bona fide properties of points of space
are. It is a placeholder for a characterization of matter that would have
to be filled in.
However, one cannot apply to GRWf the answer to this question that we
propose for matter points—namely, that their essence are the distance rela-
tions in which they stand, since the flashes are ephemeral instead of perma-
nent. Consequently, there is no such thing as the change in distance
relations among flashes as the objects that stand in these relations. To
84 Minimalist ontology and dynamical structure
put it differently, the GRW flashes are the Bohmian particles deprived of
their trajectories, so that there is no possibility to conceive them as the
objects with respect to which one represents changing distance relations,
with these relations (and their change) making up the essence of these
objects. Instead, there is an absolute physical space with flashes coming
into existence in it out of nothing and disappearing into nothing.
In any case, the account that the original GRW theory envisages for mea-
surement interactions does not work on the flash ontology—in other words,
this ontology covers only the spontaneous appearance and disappearance of
flashes, but offers no account of interactions: on the original GRW pro-
posal, a measurement apparatus is supposed to interact with a quantum
object; since the apparatus consists of a great number of quantum
objects, the entanglement of the wave function between the apparatus
and the measured quantum object will be immediately reduced due to the
spontaneous localization of the wave function of the apparatus.
However, even if one supposes that a measurement apparatus can be con-
ceived as a galaxy of flashes (but see the reservations of Maudlin (2011), pp.
257–258), there is on GRWf nothing with which the apparatus could inter-
act: there is no particle that enters it, no mass density and in general no field
that gets in touch with it either (even if one conceives the wave function as a
field, it is a field in configuration space and not a field in physical space
where the flashes are). There only is one flash (standing for what is
usually supposed to be a quantum object) in its past light cone, but there
is nothing left of that flash with which the apparatus could interact.
In sum, thus, as the ontology of permanent discrete objects (substances in
the guise of particles or matter points) stands out as the best proposal for
ontology already on metaphysical reasons alone as argued in Chapter 2,
section 1, so the quantum dynamics that Bohmian mechanics bases on
this ontology stands out as the most convincing proposal for a solution
to the measurement problem in the framework of primitive ontology theo-
ries of quantum physics (see Esfeld (2014a) for a detailed comparison of
these theories). Of course, this assessment of the dynamics would change
if experimental tests of collapse theories like GRW against theories that
exactly produce the predictions of textbook quantum mechanics—such as
Bohmian mechanics—were carried out successfully and confirmed the col-
lapse theories where they deviate from the standard predictions (see Cur-
ceanu et al. (2016) for such experiments).

3.4 Dynamical structure from the universe to subsystems

The dynamical structure that a physical theory introduces is in any case
defined for the universe as a whole. To solve equation (3.2)—and the
corresponding equations in classical as well as quantum Bohmian mechan-
ics (equations (3.5) and (3.10))—one has to put in initial data for the
 Minimalist ontology and dynamical structure 85
whole configuration of matter points. That is to say, the dynamical struc-
ture correlates in principle the motion of any matter point with the one
of any other matter point in the universe. As already mentioned in
Chapter 2 at the end of section 2, it is therefore appropriate to speak of
a dynamical holism.
In classical mechanics, as is evident from equation (3.5), one calculates
the velocity for each matter point relative to each other matter point sepa-
rately, and that velocity depends on the distance between the two matter
points. Nonetheless, to obtain the correct velocity for any given matter
point, one would have to take into account its relation to all the other
matter points. For instance, as soon as there are dynamical parameters
whose value is conserved throughout the universe—such as the total
energy—there are global correlations in the motions of the matter points.
Thus, the difference between classical and quantum mechanics by no
means concerns a difference between a local dynamics and a dynamical
holism. It is this one: the quantum mechanical wave function is defined
only for the configuration of matter as a whole; it correlates the motion
of any matter point with in principle any other matter point without that
correlation having to depend on the distance between the matter points.
One can therefore say that the dynamical holism is more obvious in
quantum mechanics than it is in classical mechanics: in quantum mechan-
ics, there is a single dynamical parameter—the wave function—that corre-
lates the motion of all the matter points; in classical mechanics, by contrast,
one attributes dynamical parameters to the matter points taken individually
and figures out their correlated motion pairwise, depending on the distance
between them. However, this difference does not justify regarding quantum
mechanics as endorsing the configuration space on which the wave function
is defined, by contrast to three-dimensional space, as the space in which the
physical reality is situated (see Albert (2015), pp. 142–143, for such a
view). In both cases, dynamical parameters are introduced in order to deter-
mine the change in the distance relations among matter points in such a way
that specifying an initial value of these dynamical parameters together with
an initial configuration of matter points is sufficient to fix the whole evolu-
tion of the distance relations among the matter points—independently of
whether this is done for the whole configuration at once, or pairwise for
the matter points. In any case, in brief, if the configuration of the universe
Q0 and the correct velocity field vt were known, one could deduce from
equation (3.2) a unique motion t 7! Qt of the entire configuration.
That notwithstanding, note that the ontology defined by the two axioms
in section 2.1 has a much wider scope than what is known as the primitive
ontology approach to quantum physics. Notably, it is not tied to three-
dimensional space; the distance relations defining this ontology are not
wedded to a particular geometry. Thus, from the perspective of this ontol-
ogy, the main objection to the ontology of Albert (2015, chs. 6 and 7) of
only a wave function in configuration space is not that the wave function
86 Minimalist ontology and dynamical structure
only ontology lacks a primitive ontology in terms of objects that are local-
ized in three-dimensional space. From the perspective of parsimony, the
main objection is the uneconomical dualism of a space defined at least by
topological relations and material entities (such as a wave function field)
existing on that space that are defined in terms of some intrinsic features,
which hence do not come out of the relations constituting that space.
In any case, laws that describe the evolution of the universe as a whole
are useless for calculations. Even if we know the types of parameters that
enter into the dynamical structure, both classical mechanics of gravitation
and Bohmian quantum mechanics suggest not only one but many possible
velocity fields vt, given that for example the initial velocities or the initial
wave function, respectively, are free parameters. We cannot know the
initial conditions for the configuration of matter as a whole. At best we
know a little about a subsystem and nearly nothing about the rest of the
universe. Moreover, even if we had precise knowledge of initial conditions,
the complexity of calculation increases exponentially with the number of
particles of a given system, as is clearly brought out already by the three
body problem in classical mechanics. Furthermore, the evolution of a
given configuration of matter points may be extremely sensitive to pertur-
bations on its initial conditions, so that a slight error about the initial con-
ditions may lead to a great error in predicting the evolution of the system. In
brief, on the one hand, we seek for universal physical theories and have
such theories at our disposal; on the other hand, these theories are useless
as they stand when it comes to solving the equations.
What is the way out of this dilemma? Let us use the notation Q = (Qsys,
Q ) for which Qsys comprises the distance relations that constitute the
env

subsystem under investigation and Qenv those of its environment—that is,

all the rest of the universe. Instead of asking, given a particular Q =
(Qsys, Qenv) and a velocity field vt, what is the precise motion t7!Qsys t to
be conducted, we can ask more humbly, with the knowledge that we
have, what kind of motion we can expect to happen in most cases—that
is, for most of the initial configurations Qenv and admissible velocity
fields vt. For instance, when flipping a coin n times, it is generically impos-
sible to predict the individual outcomes and thus to predict the exact
sequence of heads and tails, although this sequence is determined
completely by Q0 and vt. Nevertheless, it is possible to derive statements
like this one: in most cases, the number of heads will be almost equal to
the number of tails provided that the number of throws n is large
enough. Although we cannot infer precise deterministic statements from
our physical theory because of our ignorance, it is still possible to infer
probabilistic ones.
To derive such statements, we first need a measure that tells us what
“most” means. This has obviously to be a measure on the space of all pos-
sible configurations as well as of the unknown parameters of the dynamical
structure. To obtain this measure, we have to define what “most” shall
 Minimalist ontology and dynamical structure 87
mean at an arbitrary time t. This definition has to be such that it respects a
principle of stationarity under transport of the equations of motion to any
other time t: the notion of “most” must not change in t. Such a measure
then enables us to turn predictions of our physical theory concerning sub-
systems which, due to our ignorance we could not infer, into random var-
iables whose distributions are determined by this measure. Such a measure
can be called typicality measure. (The term “random variables”, however,
is somewhat misleading: there is nothing random about these variables,
since the macrostate of a system—defined in terms of such variables—is
completely determined by its microstate).
To illustrate this procedure, let us come back to classical mechanics and
consider a system of N point particles. Denoting by qi and pi the position,
respectively the momentum of the i’th particle, let us call X(t) = (q1(t), . . .,
qN(t);p1(t), . . ., pN(t)) the microstate of the system at time t. The space of all
possible microstates, here G :¼ R3N  R3N , is called phase space. Let us now
employ the Hamiltonian formulation of the microscopic laws of motion,
which take the form
8
>
> @H
>
> q_i ¼
< @pi
; ð3:19Þ
>
> @H
>
>
: p_i ¼ 
@qi

with
XN
p2i
Hðq; pÞ ¼ þ Vðq1 ; :::; qn Þ: ð3:20Þ
i¼1
2mi

More compactly, this can be written as

ðq_i ; p_i Þ ¼ vH ðq; pÞ; ð3:21Þ

where vH denotes the vector field on Γ generated by the Hamiltonian H.

These equations give rise to a Hamiltonian flow Ft,0 such that X(t) =
Ft,0(X) for any initial microstate X. In equation (3.20), mi denotes the
mass of the i’th particle and V the interaction potential, which can be
split into
X
Vðq1 ; :::; qn Þ ¼ V int ðqi  qj Þ þ V ext ðq1 ; :::; qN ; tÞ: ð3:22Þ
i<j

Vint then corresponds to a pair-interaction among the particles (e.g. gravi-

tation) and Vext is an external potential, summarizing the influences of
the environment. Of course, if the N-particle system is the entire universe,
then Vext = 0, since there is nothing outside the universe.
88 Minimalist ontology and dynamical structure
If Vext is zero (or at least time-independent), such a Hamiltonian system
has several nice properties. For one, it conserves the total energy, meaning
that H = const. along any solution of (3.19). Furthermore, by Liouville’s
theorem, the Hamiltonian flow conserves phase space volume. This is to
say that the uniform Lebesgue measure λ is a stationary measure on Γ in
the sense that for all t
0 and any Borel set A  Γ,

lðFt;0 AÞ ¼ lðAÞ: ð3:23Þ

For fixed E 2 R, it is usually convenient to consider the reduced phase space

ΓE :={X 2 Γ : H(X) = E} to which a system with total energy E is confined by
virtue of energy conservation. λ then induces a stationary measure λE on the
hypersurface ΓE, which is called the microcanonical measure. By conven-
tion, we normalize this measure to λE(ΓE) = 1. The Hamiltonian formula-
tion of classical mechanics thus brings out that the Lebesgue measure is
distinguished as the simplest stationary measure on phase space.
Making use of this formalism, let us consider the stock example of an
ideal gas in a box (with perfectly reflecting walls) that will serve as our
toy-model for the universe. The number of particles in such a macroscopic
system is of the order of Avogadro’s constant—that is N * 1023. Clearly,
determining the actual configuration and / or predicting the trajectories for
so many particles is a hopeless task, even if the particles are non-interacting
as in our example.
Nevertheless, it is possible to make meaningful predictions about this
system. For instance, we can ask the following: what is the rate of particles
that have a velocity in the x-direction that is approximately v0 (where v0 is
some arbitrary, positive number)? We can formalize this in terms of the
random variable:

1X N
FðXÞ :¼ w ðXÞ: ð3:24Þ
N i¼1 fvi;x 2½v0 d;v0 þdg

Here, δ > 0 is a small positive number (giving precise meaning to “approx-

imately v0”) and χ is the indicator function, i.e. χ{vi,x2[v0δ,v0+δ]} equals
one if vi;x ¼ m1 pi;x lies in the interval [v0δ, v0+δ] and zero if it does not.
As mentioned earlier, calling such a variable a “random variable” is some-
what misleading, since the state of the system is always fully determined by
its microstate. The point is that a great number of microstates X will in
general correspond to (approximately) the same value of such a variable
F so that F does not disclose the microstate of the system.
Fixing the mean energy per particle to NE ¼ 32 kB T (kB is the Boltzmann
constant and T can later be identified as the temperature of the system),
 Minimalist ontology and dynamical structure 89
it is a mathematical fact that
 
1 mv2
Z exp 

 b
kB T 2
lim lE X 2 GE : vi;x 2 ½a; b ¼  3=2 dv : ð3:25Þ
E ¼3k T
N!1;N 2 B a 2pkB T
m

From this, one can conclude that for any  > 0,

08
 

91
>
1 mv2 >
>
>
Z b exp 
>
>
B<
1 X N
=C
B
kB T 2
C
lE B X :
wfvi;x 2½a;bg ðXÞ   3=2 dv
>  C
@>
>
N 2pk T
>
>A
>
:
i¼1 a B
>
; ð3:26Þ

m

! 0; N ! 1:

The derivation of this result is a more or less elementary exercise in measure

theory. The more profound question, however, is what this result actually
means. 
2
The function rðvÞ / exp  kB1T mv2 is called the Maxwell distribution. It
is a probability measure, describing a distribution of particle velocities.
Note that there is actually nothing random about the velocities of particles
in a gas. The velocity (as well as the position) of every single particle is com-
prised in the microstate X whose evolution is described by a deterministic
equation of motion. There are possible X for which the actual distribution
of velocities in the gas differs significantly from that described by the
Maxwell distribution. For instance, there are microstates X for which all
particles move with one and the same velocity. Or microstates X for
which a few very fast particles account for almost the entire kinetic
energy, while all the others are nearly at rest. But these states are (obvi-
ously) very special ones. The crucial and remarkable fact expressed by
equation (3.26) is that, for large N, the overwhelming majority of possible
microstates is such that the distribution of velocities in the gas is (approx-
imately) Maxwellian (cf. Boltzmann (1896), p. 252). The “overwhelming
majority of microstates” is thereby defined in terms of the stationary
measure λE. In this sense, the Maxwell distribution constitutes a prediction
of the microscopic particle theory as a statistical regularity manifested for
typical (initial) configurations.
Note that the role of the microcanonical measure in this argument is only
to give precise meaning to “by far the largest number of all possible
states”—that is, to provide a well-defined notion of typicality. The
Maxwell distribution, in contrast, refers to actual statistical patterns—
that is, relative frequencies in typical particle ensembles. Hence, it is impor-
tant to appreciate the fact that while two measures appear in the mathemat-
ical equation (3.26), their status is very different (cf. Goldstein (2012)). To
90 Minimalist ontology and dynamical structure
make this point clear, let us add the following observations (see Lazarovici
and Reichert (2015) for a detailed argument):

1. Since the box in our example exists only once—even more so if it is sup-
posed to be a model for the universe—probabilistic statements about its
(initial) microstate have no empirical meaning. The Maxwellian ρ refers
to an actual distribution of velocities that exists in the box. The micro-
canonical measure does not refer to an ensemble of boxes, but pertains
to a way of reasoning about the box and the physical laws describing it,
allowing us to establish that the observed velocity distribution is
typical.
2. Also, the microcanonical measure is not supposed to quantify our
knowledge and / or ignorance about the microstate of the gas. While
it is correct to say, in some sense, that randomness in a deterministic
theory is only due to our ignorance regarding initial conditions, it is
important to note the very limited degree to which knowledge, informa-
tion, credences or other subjective notions play a role in the analysis. It
is an objective fact that for the great majority of microstates, the distri-
bution of velocities in an ideal gas is (approximately) Maxwellian, and
it is this objective fact that we take to be explanatory.
3. With respect to a typicality measure, only sets of very large ( 1) or
very small ( 0) measure are meaningful. Therefore, a probability
measure has actually too much mathematical structure and the
meaning of “typical” would not change, if we changed our measure
in a more or less continuous fashion.

An analogous reasoning can be applied to more mundane examples like the

before mentioned coin toss. It is a statistical regularity found in our universe
that the relative frequency of heads or tails in a long series of fair coin tosses
is approximately 1/2. If we agree that a coin toss is guided by the same laws
as all other physical processes in the world, this statistical regularity has to
be explained on the basis of the fundamental microscopic theory (here: clas-
sical mechanics).
Let us denote by Fi the outcome of the i’th coin toss in a long series of N
coin tosses. We say that Fi = 1 if the outcome is heads and Fi = 0 if the
outcome is tails. Since classical mechanics is deterministic, the outcome
of every single coin toss is actually determined, through the fundamental
laws of motion, by the initial state of the universe. Hence, we have Fi =
Fi(X) for X 2 Γ the initial microstate of the Newtonian universe. The func-
tions Fi are obviously (very) coarse-graining. We do not care about the
exact configuration of atoms making up the coin; we do not even care
about the exact position or orientation of the coin; we only ask which
side is up as the coin lands on the floor. This defines our macroscopic
observables.
 Minimalist ontology and dynamical structure 91
There are possible initial configurations conceivable that would give rise
to a universe that looks pretty much like ours, but in which the relative fre-
quency of heads is very different from 1/2: there are possible initial config-
urations for which every coin ever to be tossed will land on heads, or for
which tails will come out two out of three times. But such initial conditions
are very special ones. In contrast, typical initial conditions of the universe—
compatible with there being coins and coin tossers in the first place—are
such that the relative frequency of heads or tails in a long series of fair
coin tosses is approximately 1/2. Formally, the claim is that for any  > 0,


!

1 XN
1

l
F ðXÞ 
>  ! 0; N ! 1: ð3:27Þ

N i¼1 i 2

This is to say that if N is sufficiently large, the set of initial conditions for
which the relative frequency of heads deviates significantly from 1/2 is
extremely small. Such initial conditions are thus not impossible, but atypi-
cal. (3.27) is a law of large numbers statement. The law of large number is
what connects probabilities to relative frequencies in typical ensembles. The
distinction between the typicality measure and the probability distribution
is here, once again, crucial in order to avoid the usual redundancy of
explaining probabilities in terms of probabilities.
On this account, probabilities are objective. They apply to patterns in the
world instead of subjective beliefs. It is a matter of fact that, as the number
of coin tosses N becomes very large, almost all sequences of coin toss out-
comes manifest the pattern of an approximately equal frequency of heads
and tails. This matter of fact is independent of what agents believe about
the outcomes (although both are linked: it is of course rational to adapt
one’s beliefs to the patterns in the world).
However, there are many situations in classical mechanics that are not
like the coin toss or the molecules in a gas. For example, when we
compute the trajectory of a stone thrown on earth, we can, in general,
use a simple deterministic equation without being embarrassed by our igno-
rance regarding the exact initial microstate of the stone or its environment.
There are two conditions satisfied here that allow us to do that:

1. The external forces—that is, the influence of the rest of the universe
neglected in the computations is very small compared to the attraction
between the stone and the earth. This is usually the case because other
gravitating bodies are either very far away or have very small mass
compared to the earth.
2. The evolution of the relevant macroscopic variable—here, the centre of
mass of the stone—is reasonably robust against variations in the
92 Minimalist ontology and dynamical structure
microscopic initial conditions. In other words, small changes in the
microscopic initial conditions have only small effects on the trajectory
of the stone. This is why our ignorance about the exact position and
momentum of every single particle constituting the stone (or the
earth, or the person/apparatus throwing the stone) does not prevent
us from making reliable predictions about the motion of its centre of
mass.

Nonetheless, even in this case, our prediction for the trajectory of the stone
is strictly speaking a typicality result. Atypical events in the environment or
fluctuations of the particles constituting the stone could lead to very differ-
ent outcomes. Hence, to be precise, we would have to cast our result about
the trajectory of the stone in a form that looks quite similar to the proba-
bilistic statements (3.26) or (3.27). For instance, denoting by x(t) the com-
puted trajectory (depending on the initial position and momentum of the
stone) and by x ~ ðtÞ the actual trajectory of the stone (depending on the
initial condition X of the universe), we could write:

 
lð X : sup j~
x ðtÞ  xðtÞj >  Þ  0: ð3:28Þ
0tT

Still, the stone throw example points to a striking difference between clas-
sical and quantum mechanics. In classical mechanics, we often encounter
situations in which correlations between a subsystem and its environment
become negligible, allowing for a (more or less) deterministic description
of the subsystem. In quantum mechanics, by contrast, the generic situation
is much more similar to the coin toss or the molecule in a gas, where pre-
dictions of statistical patterns are the best we can hope for. This is so
even if we settle for a formulation of quantum mechanics with an ontology
of particles and a deterministic law of motion, such as Bohmian mechanics.
Let us therefore now discuss how probabilities enter Bohmian mechanics.
For a statistical analysis of Bohmian mechanics, we need (a) a sensible
typicality measure defined on configuration space and (b) a procedure to
get from the fundamental, universal description in terms of the universal
wave function to a well-defined description of Bohmian subsystems. In
the following, we will largely rely on the accomplishment of these tasks
in Dürr and Teufel (2009) and Dürr et al. (2013b) (see Callender (2007)
and Maudlin (2007b) for a philosophical analysis). Given the universal
wave function, the appropriate notion of typicality for particle configura-
tions is given in terms of the measure with density |C|2. The crucial
feature of this measure is that it is equivariant, assuring that typical sets
remain typical and atypical sets remain atypical under the Bohmian time
evolution. More precisely, if FCt;0 is the flow on configuration space
 Minimalist ontology and dynamical structure 93
induced by the guiding equation (3.6), then
Z Z
2 2
PC ðAÞ :¼ jC0 j d3N q ¼ jCt j d3N q
ð3:29Þ
A FC
t;0
ðAÞ

holds for any measurable set A  R3N . Equivariance is thus the natural gen-
eralization of stationarity for a non-autonomous (time-dependent) dynam-
ics. The |C|2-measure can be proven to be the unique equivariant measure
for the Bohmian particles dynamics that depends only locally on C or its
derivatives (see Goldstein and Struyve (2007)). In this sense, it is even
more strongly suggested as the correct typicality measure for Bohmian
mechanics than the Lebesgue measure is in classical mechanics.
Let us now have a closer look at how Bohmian mechanics treats subsys-
tems of the universe. Suppose that the subsystem consists of n  N parti-
cles. We then split the configuration space into R3N ¼ R3n  R3ðNnÞ , so
that, writing q = (x, y), the x-coordinates describe the degrees of freedom
of the subsystem and the y-coordinates describe the possible configurations
of the rest of the universe. Analogously, we split the actual particle config-
uration into Q = (Qsys, Qenv) = (X, Y), with Qsys = X, the configuration of
the subsystem under investigation and Qenv = Y the configuration of its
environment.
In passing from the fundamental, universal theory to a description of the
subsystem, we can just take the universal wave function Ct(q) = Ct(x, y)
and plug into the y argument the actual configuration Y(t) of the rest of
the universe. The resulting

cYt ðxÞ :¼ Ct ðx; YðtÞÞ ð3:30Þ

is now a function of the x-coordinates only. It is called the conditional wave

function. In terms of this conditional wave function, the equation of motion
for the subsystem takes the form

_ rx cYt ðxÞ

XðtÞ / Im ð3:31Þ
cYt ðxÞ
x¼XðtÞ

to be compared with (3.6). However, since the conditional wave function

depends explicitely on Y(t), its time evolution may be extremely compli-
cated and not follow any Schrödinger-like equation. Fortunately, in many
relevant situations, the subsystem will dynamically decouple from its envi-
ronment. The subsystem has an effective wave function φ if and only the
universal wave function takes the form

Cðx; yÞ ¼ φðxÞwðyÞ þ C? ðx; yÞ; ð3:32Þ

94 Minimalist ontology and dynamical structure
where χ and C? have disjoint y-support and Y 2 supp χ, so that in partic-
ular C?(x, Y) = 0 for almost all x. (Note that this is much weaker than
assuming that C has a product structure, which is in general not the
case.) This means that we can effectively forget about the empty wave
packet ψ?(x, y) and describe the subsystem in terms of its own independent
wave function φ. If we can furthermore assume that the interaction between
subsystem and environment is negligible for some time—that is,
V ext ðx; yÞφðxÞwðyÞ  0; ð3:33Þ

the effective wave function will satisfy

R its 2own, autonomous Schrödinger
evolution. Such a φ—normalized to |φ(x)| dx = 1—is the Bohmian coun-
terpart of the usual quantum mechanical wave function. It is these effective
wave functions that physicists manipulate in laboratories and for which
Born’s rule is formulated.
For our statistical analysis, we start by considering the conditional
measure
2
jCððx; YÞÞj dn x 2
PC ðfQ ¼ ðX; YÞ; X 2 dn xgjYÞ ¼ R 2 ¼ jcY ðxÞj dn x: ð3:34Þ
jCððx; YÞÞj dn x

In the special situations described by (3.32), the conditional wave function

ψY on the right-hand side becomes the effective wave function φ. For prac-
tical purposes, though, conditioning on the configuration Y is much too spe-
cific, since we have only very limited knowledge of Y. However, many
different Y will yield one and the same effective wave function for the sub-
system. Collecting all those Y, and using the fact that by yielding the same
effective wave function they also yield the same conditional measure (3.34),
a simple identity for conditional probabilities yields
2
PC ðfQ ¼ ðX; YÞ; X 2 dn xgjcY ¼ φÞ ¼ jφj dn x: ð3:35Þ

From this formula, one can now derive law of large numbers estimates of
the following kind: at a given time t, consider an ensemble of M identically
prepared subsystems with effective wave function φ. Denote by Xi the
actual configuration of the i’th subsystem. Let A  R3n consider the corre-
sponding indicator function χ{Xi2A}, which is 1, if the configuration Xi is
in A and 0 otherwise. Then it holds for any  > 0 that
(
Z
)!

1 XN

C
2

Pt ¼ Q:
w ðQÞ  jφðxÞj
<  ! 0; N ! 1: ð3:36Þ

N i¼1 fXi 2Ag A

This is to say that for typical configurations of the universe, the particles in
an ensemble of subsystems with effective wave function φ are distributed
 Minimalist ontology and dynamical structure 95
2
according to |φ| . Thus, Born’s rule holds in typical Bohmian universes—
that is, in quantum equilibrium.
Once again, we emphasize that the |C|2-measure given in terms of the
universal wave function is only used to define typicality. It is not supposed
to describe an actual distribution of configurations—that is, an ensemble of
universes, because the universe exists only once. By contrast, the |φ|2-
measure on the right-hand side, defined in terms of the effective wave func-
tion, does refer to actual particle distributions in a typical ensemble of iden-
tically prepared subsystems. Born’s rule is thus predicted and explained by
Bohmian mechanics.
Comparing equation (3.35) to (3.26) (and recalling the reasoning that
led to the respective equations) we recognize the analogy between the deri-
vation of Maxwell’s distribution in classical mechanics and Born’s rule in
Bohmian mechanics. In essence, it is Boltzmann’s statistical mechanics
applied to two different theories. The status of probabilities and the role
of typicality is the same in both cases, although the dynamical laws are
strikingly different. On the one hand, this illustrates the deepness and uni-
versality of Boltzmann’s insights. On the other hand, it shows that there is
no need to look for a fundamentally new kind of randomness in the
quantum realm. If the microscopic laws and the ontology of the theory
are clear, probabilities in quantum mechanics are no more mysterious
than they are in classical mechanics.
However, why then does the quantum realm appear to us so much more
random and unpredictable than the classical realm? The answer to this
question is in part trivial. Quantum mechanics is usually employed to
make predictions about microscopic systems, while classical mechanics is
most often employed to make predictions about macroscopic systems and
coarse-grained observables. The latter is bound to be more robust against
our ignorance regarding microscopic initial conditions. Furthermore, our
ability to describe a particular subsystem and the level of detail that we
can thereby achieve depends heavily on the strength of correlations
between the investigated subsystem and the rest of the universe. Newtonian
mechanics is a non-local theory, though only in a rather mild sense. Forces
fall off quickly with increasing distance (and gravity is very weak to begin
with) so that parts of the universe can often be described as autonomous
Newtonian systems for all practical purposes.
In quantum mechanics, as clearly brought out by Bohmian mechanics,
however, there are correlations possible that do not depend on the
distance of the systems. An example is the two particles in the EPR exper-
iment which produce the famous anti-coincidences. This difference stems
from the different manners in which these theories define the velocity field.
In classical mechanics, as mentioned earlier, the time evolution of the
velocity is given separately for each component k = 1, . . ., N in (3.5).
In Bohmian mechanics, the time evolution of the velocities (3.6) is deter-
mined by the wave function Ct whose temporal development is given by
96 Minimalist ontology and dynamical structure
the Schrödinger evolution (3.7). Both are defined on the whole configura-
tion space. That is why correlations that are independent of the distance of
the systems are admitted in quantum physics. Consequently, it is much
more difficult to consider any proper part of a Bohmian universe as
“isolated”, while ignoring the influence of the rest of the universe. As a
matter of fact, it is often possible to provide an autonomous Bohmian
description of a Bohmian subsystem in terms of an effective wave func-
tion. This autonomy, however, can be somehow deceiving, because the
effective wave function still depends implicitly on the configuration of
the environment (e.g. on the procedure used to prepare that state in an
experimental situation).
More precisely (and more profoundly), our possible knowledge about
the particle configuration in a Bohmian subsystem is restricted by the
theorem of absolute uncertainty, which has no analogue in classical
physics (see Dürr et al. (2013b), ch. 2). Absolute uncertainty is a direct con-
sequence of the conditional probability formula (3.34): all our records
about the particle positions—brain states, computer prints, pointer posi-
tion, etc.—are included in the configuration Y of the rest of the universe.
Hence, all possible correlations between these records and the configuration
of the subsystem are already taken into account in equations (3.34) and
(3.35) that yield Born’s rule for the distribution of particle positions.
This connection between our epistemic state and the effective wave func-
tion of the subsystem then works in two ways. On the one hand, it means
that given a Bohmian subsystem with effective wave function φ, our infor-
mation about the particle configuration cannot be more precise than what is
given by the |φ|2-distribution. On the other hand, it means that if we
perform additional measurements to determine the particle positions with
greater accuracy, the system’s effective wave function becomes more and
more peaked. Hence, the gradients in the velocity formula (3.6) induce
higher and higher possible velocities, depending on the precise initial config-
uration of the particles. Less uncertainty about the initial particle positions
thus implies more uncertainty about the (asymptotic) velocities—this is the
source of Heisenberg’s uncertainty principle. Even small deviations in the
initial configuration will thus lead to large deviations of the resulting
Bohmian trajectories. Our rapidly increasing uncertainty about the particle
positions is then mirrored by the quick spreading of the wave function
under the Schrödinger time evolution. Therefore, computing any one partic-
ular trajectory does not lead to useful information about the real trajectory
of the system. Hence, the manifestly non-local nature of quantum mechan-
ics is such that a system becomes immediately more chaotic as we try to
minimize our ignorance regarding microscopic initial conditions. As a con-
sequence, we have to resort to probabilistic reasoning much earlier than
is often the case in classical physics. For a quantum system, Born’s rule
provides—provably—as good a description as we can get in a universe in
quantum equilibrium.
 Minimalist ontology and dynamical structure 97
Coming back to the comparison between Bohmian mechanics and the
GRW matter density quantum theory in the preceding section, consequently,
although the GRW matter density field is specified by the wave function,
whereas the Bohmian particle positions are not specified by the wave func-
tion, the implications for our knowledge are similar. The reason for going
for a primitive ontology theory of quantum physics is that one solves the
measurement problem by endorsing a configuration of matter in physical
space. But the consequence of doing so is that there is an epistemic limit to
the accessibility of that configuration. Such a limited accessibility applies to
any primitive ontology theory of quantum physics, independent of whether
one takes the configuration of matter in physical space to consist of discrete
objects (particles), as in Bohmian mechanics, or in one continuous field, as in
the GRW matter density theory (see Cowan and Tumulka (2016)). In other
words, one does not avoid what may seem to be a drawback of subscribing to
so-called hidden variables—namely, a limited epistemic accessibility—by
amending the Schrödinger equation and taking the wave function as it
figures in such an amended Schrödinger equation to represent the configura-
tion of matter in physical space. In a nutshell, if one endorses a commitment
to a configuration of matter in physical space in quantum physics, one has to
endorse a commitment to a limited epistemic accessibility of that configura-
tion, in whatever way one spells out the theory of that configuration and its
evolution.
By way of consequence, in any primitive ontology theory of quantum
mechanics, probabilities come in through our ignorance of the exact config-
uration of matter in physical space. This ignorance implies that we imme-
diately have to resort to probabilistic descriptions in quantum physics as
elaborated on earlier, independently of whether the dynamical law for
the evolution of the configuration of matter in physical space is determinis-
tic (as in Bohmian mechanics) or stochastic (as in the GRW theory).

Note
1. See Belot ((2011), pp. 60–77) for a criticism that spells out how the general
objections against Humeanism apply in this case.
4 A persistent particle ontology
for quantum field theory

4.1 The requirements for an ontology of QFT

Any proposal for an ontology of quantum physics has to provide a solution
to the measurement problem by setting out what happens in nature on the
level of individual quantum systems instead of merely making statistical
predictions for measurement outcomes. As regards non-relativistic
quantum mechanics, we argued in Chapter 3, sections 2 and 3, that
Bohmian mechanics is the most convincing proposal to solve the measure-
ment problem. Bohmian mechanics supplements the wave equation (i.e. the
Schrödinger equation) with a guiding equation that yields trajectories for
individual particles in three-dimensional space (what is known as the prim-
itive ontology) and explains the states of macroscopic systems as well as
their stability in terms of these trajectories.
In doing so, Bohmian mechanics fits well with the minimalist ontology
pursued in this book: the Bohmian particles are matter points without
any intrinsic features. Their being individuated by their position in space
means that they are individuated by the distance relations among them,
which make up the configuration of matter of the universe. The wave func-
tion then is not an additional element of the physical ontology. It is a dyna-
mical parameter employed in the laws that achieve the best combination of
simplicity and information in describing the change in the distance relations
among the particles (i.e. their trajectories) throughout the entire history of
the universe (cf. the quantum Super-Humeanism mentioned in Chapter 2,
section 3). In this chapter, we base ourselves on this minimalist ontology
including Super-Humeanism applied to the dynamical parameters and
laws of quantum physics, without always explicitly mentioning this back-
ground in the text and in the formalism. Our aim now is to show that
the Bohmian approach does the same service for QFT that it does for
quantum mechanics.
When it comes to QFT, the requirement for an ontology is the same as in
quantum mechanics: the measurement problem plagues QFT in the same
way as quantum mechanics (see Barrett (2014)); any proposal for its solu-
tion has to explain what happens in nature on the level of individual
100 A persistent particle ontology
processes instead of merely making statistical predictions. On the one hand,
a particle ontology is generally regarded as excluded for QFT, given that
one of the most striking phenomena predicted by QFT is the so-called cre-
ation and annihilation of particles: if there is not a definite number of par-
ticles that persist, particles cannot be what is fundamental (see, e.g.,
Halvorson and Clifton (2002); Kuhlmann (2010), ch. 8; Ruetsche (2011),
chs. 9–11). On the other hand, it is by no means clear that the ontology
of QFT is one of fields, since QFT does not admit fields that have definite
values at the points of four-dimensional space-time (see Baker (2009)). In
general, if one advocates an ontology in terms of fields, the problem is to
formulate a dynamics that accounts for measurement outcomes in a con-
vincing manner, given that most measurement outcomes consist explicitly
in definite positions of something—be it the tracks caused by cosmic radi-
ation on Victor Hess’s photo plates, the electron and positron tracks in
Carl Anderson’s cloud chamber, or the muon tracks recorded by the
ATLAS detector at CERN that became famous for the Higgs-particle detec-
tion. In a nutshell, in QFT as in any other area of physics, the experimental
evidence is particle evidence. Fields, waves and the like come in to explain
that evidence, but are not themselves part of the evidence.
Against this background, we submit that also when it comes to QFT,
the particle ontology set out in this book—matter points being individu-
ated by the distance relations in which they stand and the change in
these relations—is the simplest ontology that is empirically adequate:
less will not do; bringing in more leads to drawbacks instead of providing
additional explanatory value. In particular, this ontology comes with a
clear and straightforward solution to the measurement problem: there
always is a definite spatial configuration of matter points. In order to vin-
dicate this ontology for QFT, we show in the following that the Bohmian
approach works for QFT in the same way as for quantum mechanics: as it
is a non sequitur to take particle trajectories to be ruled out in quantum
mechanics due to the Heisenberg uncertainty relations, so it is a non sequi-
tur to take permanent particles moving on definite trajectories according
to a deterministic law to be ruled out in QFT due to the statistics of par-
ticle creation and annihilation phenomena. In both cases, such an under-
lying particle ontology is in the position to explain the statistics of
measurement outcomes.
Bohm himself in his original paper Bohm ((1952b), appendix A) already
discussed a possible extension of his theory to the electromagnetic field.
Indeed there is a field version of Bohmian QFT (see Struyve and
Westman (2006, 2007) as well as Valentini (1992), ch. 4). As regards the
particle ontology, the first Bohmian QFT can be traced back to Bell
(2004, ch. 19, first published 1986) (see Dürr et al. (2005) for a continuum
version of Bell’s proposal on a lattice; see Struyve (2010) for an overview of
all the Bohmian proposals). Bell’s proposal grants the entities that figure in
the particle pair-creation and annihilation processes of QFT the status of
 A persistent particle ontology 101
real, fundamental particles, which are, accordingly, created and annihilated
at random times and positions. Between these random events, the particles
evolve according to a Bohmian law of motion.
We do not adopt this proposal. In the first place, since these entities
depend on the choice of an initial reference vacuum state that is not
unique (see Fierz and Scharf (1979)), instead of being objects that simply
exist, we consider it to be inappropriate to grant them the status of funda-
mental objects in the ontology. More generally speaking, the Bohmian
approach is motivated by finding an underlying ontology that explains
the statistics of measurement outcomes. Although Bell’s proposal achieves
an explanation of the statistics of measurement outcomes, we think that
it is worthwhile to pursue the Bohmian approach only if one is prepared
to go all the way down to an ontology of fundamental objects that are
simply there—that is, that do not come into existence out of nothing and
that do not disappear into nothing—and that evolve according to a
simple, deterministic law (that is, an evolution not interrupted by random
jumps), much in the spirit of the quote from Parmenides given in Chapter
2, section 1.
That is why we take up the old idea going back to Dirac (1934): there is
no particle creation or annihilation. There are only conditions under which
particle motion becomes experimentally accessible or fails to be so, and
these conditions are not unique; they may even depend on the state of
motion of the experimental devices (as it is the case for the Unruh effect).
We build an ontology of permanent particles for QFT on this idea. In
this task, we face a shortcoming of the modern formulation of QFT,
which is given only in terms of perturbative scattering theory. The mathe-
matical difficulties involved in the computation of scattering matrix ele-
ments are already so severe that in the renormalization group
formulation of QFT, the question about a fundamental equation of
motion is not even thematized. In ontology, however, we have to make
explicit what the laws are according to which individual processes occur;
for what there is in nature are individual events and processes. Techniques
developed to calculate measurement outcome statistics do not reveal how
these processes evolve from an initial to a final state.
Despite its success, the standard model of QFT—as to date any other
known non-trivial, Lorentz covariant, interacting classical and quantum
field theory—is plagued by ill-defined equations of motion. Two central
obstacles are the ultraviolet divergences of the bosonic as well as the fermi-
onic fields. While the former type of divergence is inherited from the corre-
sponding classical field theory of point charges, the origin of the latter can
be understood best through the representation of quantum states in QFT as
seas of infinitely many particles, where “infinity” would have to be con-
ceived as a mathematical idealization, as in the thermodynamic limit of
classical statistical mechanics. In such representations, quantities like the
total current, then given as the sum of the respective infinitely many
102 A persistent particle ontology
one-particle currents, generically diverge. It is therefore the hope that when
working with relative instead of absolute quantities, these mathematical
difficulties can be overcome: for instance, although the currents belonging
to two quantum states may diverge individually, their difference may still
be well-defined; often, the difference is already the physically relevant quan-
tity (see for instance Scharf (1995), Deckert et al. (2010), Gravejat et al.
(2013) and Mickelsson (2014) for recent mathematical rigorous attempts
to arrive at a well-defined time evolution and/or vacuum polarization
current). Nevertheless, except in quantum chromodynamics, in the inter-
play with bosonic fields, this divergence leads to the infamous Landau
pole and the breakdown of renormalization theory already at a finite,
though very large, energy.
These obstacles on the level of the equations of motion that persist since
the 1930s notwithstanding, major progress has been made during the past
80 years on the level of scattering theory. When representing matrix elements
of the scattering operator in terms of informal Taylor series in the coupling
parameter of the interaction, it was shown that order by order all occurring
infinities in each of the summands can be removed algorithmically by sym-
bolic manipulation. Despite the lack of a rigorous mathematical understand-
ing, the first orders of perturbative corrections agree astonishingly well with
the experiments and speak for themselves. The higher orders are, however,
untrustworthy as it is unknown whether the renormalized series of matrix
elements are summable. In fact, toy models suggest that these renormalized
series are asymptotic series only—that is, informal series that are divergent
but may give a good approximation for small coupling parameters when
only taking the first few lower order summands into account. As an
example of an asymptotic series motivated by the Euclidean φ4-theory of sta-
tistical mechanics, consider the following function
Z 1
2 4
f ðaÞ :¼ ex ax dx ð4:1Þ
1

and observe that while for any a 2 R the sequence

Z 1 X n
N
ðx2  ax4 Þ
f N ðaÞ :¼ dx; N 2 N; ð4:2Þ
1 n¼0 n!

converges to f(α) for N ! 1 with infinite radius of convergence, the

sequence
N Z
X 1 n
ðx2  ax4 Þ
gN ðaÞ :¼ dx; N 2 N; ð4:3Þ
n¼0 1 n!

diverges. Nevertheless, plotting jf(α)  gN(α)j reveals that there is an optimal

N*(α) such that for all N  N*(α) the approximation gets better, while for
 A persistent particle ontology 103

Figure 4.1 This graph shows the logarithmic absolute difference between f(α) and
its asymptotic series gN(α) for the particular orders N. The optimal
error lies roughly around N*(α)  22. Until N = N* the approximation
given by the asymptotic series improves steadily; for N > N* it becomes
worse and the error diverges.

N > N*(α) the approximation gets worse again; see Figure 4.1. Series with
properties such as gN(α) often arise from informal perturbation theory and
are referred to as asymptotic series. In this sense, it might well turn out that
next-generation experiments with sufficiently high accuracy reveal that the
higher order perturbative corrections actually worsen the celebrated accuracy
of the perturbative predictions of the standard model instead of improving
them. The predictive power of the perturbative formulation is, hence, unclear.
The mathematical problems left aside, when it comes to the ontology of
QFT, one has to be concerned with what is going on in the world according
to the theory. That is to say, one has to consider the actual law of motion of
the theory and not only the mapping between incoming and outgoing
asymptotic particle states as it is done in scattering theory. Consequently,
for the purposes of ontology, one cannot rely on the modern renormaliza-
tion machinery that allows to compute finite summands of the informal and
potentially divergent perturbation expansion of the scattering matrix. For
our purpose, we therefore propose to return to the early attempts of the
1930s to define equations of motion for QFT. In doing so, we will of
course not obtain better predictions of measurement outcomes than stan-
dard QFT. The situation is the same as in Bohmian mechanics: the
104 A persistent particle ontology
guiding equation does not yield better predictions; accordingly, it does not
have to be solved to calculate statistical predictions of measurement out-
comes. It figures in the explanation of these statistics by answering the ques-
tion of what is really going on in the world. Hence, the purpose in returning
to Dirac (1934) is not to pursue outdated physics, but to make progress
with respect to the ontology of QFT. In modern formulations of QFT, a
mathematical description in terms of the Dirac sea, although being canon-
ically equivalent on the level of wave functions, has been abandoned in
favour of a more economic description involving particles and anti-particles
as well as their creation and annihilation. It is not our intention to reintro-
duce the Dirac sea as computational device. On the contrary, we will argue
that after it served its ontological purpose, it suggests itself to develop com-
putational methods that let go the bulk of the Dirac sea, replacing it by a so-
called vacuum state, and only track its excitations. What one gains in for-
mulating the fundamental equations of motion in terms of the Dirac sea is
an ontology that explains the appearances of particle and anti-particle cre-
ation and annihilation.
In the following, we examine the so-called Dirac sea model or hole
theory, which was the first one to predict the phenomenon of electron-pos-
itron pair-creation. Although it is conceived only for the electron sector of
the standard model of QFT, it is applicable to all fermionic matter. As Bell
(2004, ch. 19) argued, the commitment to fermions is sufficient to account
for the empirical data: they all consist in a spatial arrangement of fermions.
Hence, if the Dirac sea model can cover all fermionic matter, this is all that
can be reasonably demanded for it to serve as a proposal for the ontology of
QFT. We should emphasize once more that the reliance on fermions being
the matter points that account for the empirical data is a choice that is
sufficient and suits our search for a minimalist ontology of the natural
world. There are other approaches tried out in the literature: thus,
Struyve and Westman (2006) consider a proposal that entirely relies on
bosons. Furthermore, ontologies with a commitment to both fermions
and bosons are of course also conceivable. Pursuing a minimalist ontology,
however, it is suitable to choose one or the other. The simplicity by which
the fermions can account for particle detection in physical space singles
them out as the first and foremost candidate for the matter points in the
ontology. By contrast, there is no obvious manner to account for particle
detection in terms of bosons, considering, for instance, the problems con-
nected to the position operator. Denying the bosons a place in the ontology
does not make them irrelevant in our theory: they make up the dynamical
structure by which the interaction between the fermions is mathematically
described.
Apart from these ontological considerations, the Dirac sea model can be
given a rigorous mathematical treatment in certain regimes. It is therefore
suitable for probing the compatibility of QFT with an ontology of persistent
particles. Colin (2003) and Colin and Struyve (2007) have carried out an
 A persistent particle ontology 105
investigation in that sense in the physics literature (see also the brief
remarks in Bohm et al. (1987), pp. 373–374, and Bohm and Hiley
(1993), ch. 12.3). In the philosophical literature, the Dirac sea model is
hitherto largely ignored, apart from a brief assessment in Saunders (1999,
pp. 78–79, 86–88).
In the next section, we introduce the Dirac sea model (section 4.2). Then,
we define the state of equilibrium for the particles—that is, a state in which
nothing can be observed, by physicists usually referred to as the “vacuum
state” (section 4.3)—and small excitations from equilibrium, which can
be observed and which appear as if there were a creation and annihilation
of particles. This then is the basis on which we show how the formalism
that effectively describes these excitations leads naturally to the standard
QFT formalism, which employs creation and annihilation operators. We
discuss how, due to the ontology of persistent particles, it is possible to
give a clear meaning to these operators as well as to the vacuum state (sec-
tions 4.4 and 4.5). We conclude with showing how this ontology and
dynamics explains the measurement outcomes, arguing that it does for
QFT what Bohmian mechanics does for quantum mechanics.

4.2 The Dirac sea model

The Dirac sea model can be inferred from the standard model by imposing
the following restrictions:

1. Restriction to the electron sector of the standard model;

2. Restriction to direct electrodynamic interaction and neglect of
radiation;
3. Modelling interaction with all other fermion sectors of the standard
model only effectively by a time-dependent “external” interaction.

The resulting model is on the one hand simple enough for our discussion.
On the other hand, it has sufficient structure to describe the phenomenon
of electron-positron pair-creation. In terms of the particle ontology, restric-
tions one to three are to say that we consider only a part of the dynamical
structure of the standard model—namely, only those parameters that assign
to some matter points the dynamical role of being electrons (that is, move
electronwise)—and neglect all other dynamical parameters. More precisely,
we assume that the dynamical structure introduces a distinction of the
matter points in two groups: the first group of matter points 1, . . ., N for
N < M shall be referred to as electrons, while the rest of the matter points
is assigned the role of other fermionic particles by the dynamical structure.
Without restrictions one to three, the model would become unnecessarily
opaque. Nonetheless, we emphasize again that the restriction to electrons
is arbitrary, since the Dirac sea model and hence our argument is also
106 A persistent particle ontology
applicable to the other fermion sectors and the other types of interaction in
the standard model.
Electrons in direct electrodynamic interaction exclusively repel each other.
If this were the only interaction, the spatial extension of this cloud of N elec-
trons would inevitably grow larger and larger without any bound. This is an
artifact of restrictions one to three, as we model only the electron sector
directly—though we allow an indirect, effective interaction to other fermion
sectors. To avoid this unphysical behaviour (which is due to the simplicity
of the considered model) we stipulate further that, due to the motion of all
the other particles and their interaction by means of an “external interaction”
(which then includes also attractive interactions, maybe even gravitation,
although the standard model is not yet able to describe it), the spatial exten-
sion will always be bounded (which might also be suggested by general rela-
tivity and supported by cosmological evidence). Furthermore, we assume that
this “external” interaction is reasonably well behaved in the following sense:
neither will it dampen the motion of the electrons to such an extent that all
electron motion comes to a rest, nor will it drive the electron velocities arbi-
trarily close to the speed of light. Thus, there is no infinite energy transfer. The
idea behind this assumption is that motion should be somewhat conserved
among all fermion sectors. Neither does motion arise from nothing nor
does it cease to exist, it only varies over the individual particles.
Posing these two additional restrictions allows us to avoid the discussion
of mathematical problems—namely, the infamous infrared and ultraviolet
divergences—which to this day prevent the formulation of a well-defined
equation of motion in QFT. Put mathematically, we assume that the
initial data and the “external” interaction in our theory are such that:

4. Infrared cut-off: The spatial extension of the universe is restricted to a

finite volume Γ in 3d-space R3 .
5. Ultraviolet cut-off: The electron momenta are restricted to be lower than
some finite ultraviolet cut-off Λ.

In a yet to be found well-defined standard model of QFT, the earlier assump-

tions must not be imposed a priori, but have to come out a posteriori. In par-
ticular, the gained mathematical well-definedness of these ad hoc stipulations
comes at the cost of a violation of Lorentz invariance and the introduction
of seemingly arbitrary parameters. Accordingly, the main objective of
modern renormalization theory is the removal of these unwanted cut-offs.
As yet, however, a removal is only achieved order by order in the informal
(and as discussed most likely asymptotic) expansion series of the scattering
matrix elements. When it comes to defining an equation of motion, the intro-
duction of the aforementioned cut-offs is to date unavoidable in QFT. The
ultraviolet cut-offs allow for a self-adjoint interaction Hamiltonian in the cor-
responding second-quantized Schrödinger equation, while the infrared cut-off
is for example needed in order to find Fock space representations of the
 A persistent particle ontology 107
far-field of massless particles (due to a gauge symmetry violation, all the parti-
cles except for the Higgs boson are at first assumed to be massless—the masses
are later generated by the spontaneous symmetry breaking induced by the
Higgs mechanism). Nevertheless, it is the well-founded hope amongst physi-
cists that, provided the infrared and ultraviolett cut-offs are, respectively,
small and large enough, and carrying out an appropriate renormalization
scheme, the effects on the dynamics are only small in certain regimes of interest.
In fact, given the cut-offs, the formulation presented here is equivalent to the
textbook formulation on the level of wave functions. Furthermore, it is possible
to include all gauge fields of electroweak interaction to allow for a description
of radiation effects in order to remove assumptions one to three altogether (see
Colin and Struyve (2007)). In general, the Dirac sea model applies to any
fermion sector and potentially any interaction between those fermions
(except maybe gravitation whose quantum nature is unclear). In short, the
earlier assumptions are no objectionable limitations of our endeavour. They
rather underline the general point that the rationale of the Bohmian approach
is not to obtain better physics, but an ontology for the existing physics.
The Dirac sea model was originally proposed by Dirac (1934, 1947) in
Hartree-Fock approximation in terms of density matrices. Without this
approximation it was later included in QFT by means of a second quanti-
zation procedure of the one-particle Dirac equation yielding the so-called
Dirac field operators on Fock space. We discuss the Dirac sea model in
terms of a straightforward quantum-mechanical, N-particle wave equation,
a formulation which, for finite Γ and Λ, is equivalent to the textbook
version by means of second quantization of the one-particle Dirac equation
with Coulomb pair interaction (see section 4.4). Additionally, we conceive
the Dirac sea model in terms of Bohmian mechanics.
The Bohmian velocity field vt guiding the motion of N particles can be
given in the form

vt ðXÞ ¼ cðjtðkÞ ðXÞ=rt ðXÞÞk¼1;:::;N ; ð4:4Þ

 
rt ðXÞ ¼ Ct ðXÞ Ct ðXÞ; jðkÞ
t ðXÞ ¼ Ct ðXÞ 1
ðk1Þ
a 1 ðNkÞ Ct ðXÞ; ð4:5Þ

where X = (x1,. . .,xN) 2 ΓN, and ρt and jðkÞ

t for k = 1,. . .,N are the probability
density and currents generated by a wave function Ct and c stands for the
speed of light. Here, Ct denotes an anti-symmetric, square-integrable, N-par-
ticle spinor-valued function on configuration space G  R3N , in short Ct 2
H^N for H ¼ L2 ðG; C4 Þ, that solves the wave equation

iℏ@ t Ct ðx1 ; :::; xN Þ ¼ H N Ct ðx1 ; :::; xN Þ;

X
N
  ð4:6Þ
HN ¼ H 0k ðxk Þ þ V k ðt; xk Þ þ H Ik ðxk Þ :
k¼1
108 A persistent particle ontology
The Hamiltonian HN is made of the free Dirac Hamiltonian H k0 ðxÞ; the exter-
nal influences Vk(x) are given by

H 0k ðxÞ ¼ 1 ðk1Þ H 0 ðxÞ 1 ðNkÞ ; H 0 ðxÞ ¼ ica  rx þ bmc2 ;

ð4:7Þ
V k ðt; xÞ ¼ 1 ðk1Þ Vðt; xÞ 1 ðNkÞ ; for some external potential Vðt; xÞ;

where m stands for the electron mass and denotes the tensor product (note
that C is spinor-valued). Furthermore, β and the components of the vector α =
(α1,α2,α3) are C44 matrices that fulfill the anti-commutator relations {αj, αk} =
0 = {β, αj} and b2 ¼ 1 ¼ a2j for j ¼
6 k, j,k 2 {1,2,3}. Moreover, HI denotes the
Hamiltonian modelling the interaction between the particles. Since we
neglect radiation, the interaction is mediated directly by
1X
H Ik ðxÞ ¼ Uðx  xj Þ; ð4:8Þ
2 j6¼k

where the electric interaction between the N electrons can be taken to be the
e2 1
Coulomb potential UðxÞ ¼ 4p 0
jxj . Here, 0 is the electric constant and e <
0 the charge of an electron. Note that H Ik depends also on xj, which is sup-
pressed in our notation. Finally, the composite, effective interaction of all
particles on electron k is given through a time-dependent potential Vk(t,x),
stemming for example from the Coulomb field of a present ion, etc.
In sum, the velocity field (4.4) and the wave equation (4.6) define what
we call the Dirac sea model—an N-particle version of Bohmian mechanics
with Dirac’s instead of Schrödinger’s dispersion relation. Given an initial
configuration Q0 and wave function C0, the corresponding solutions of
these equations yield a unique trajectory of configurations of the N elec-
trons parametrized by time. In principle, there is nothing more to say
about the ontology: there are N persistent matter points that move electron-
wise, that is, evolve according to the deterministic law of motion given by
(4.4)–(4.6). However, we face the following situation:

I. Generically, we do not have complete knowledge about the initial data

(Q0,C0).
II. Even if we did, if N is large (which is the interesting case, as we assume
large cut-offs), it is in general neither analytically nor numerically fea-
sible to compute solutions Ct of the wave equation (4.6).

Problem (I) is generic to any theory that applies to the universe as a whole.
Predictions about subsystems have to be inferred from a detailed statistical
analysis of what is to be expected in most situations with respect to a mean-
ingful measure. In Bohmian mechanics, this is made possible by the special
form of the velocity law (4.4), which ensures that if the initial data Q0 is
distributed at random according to |C0(X)|2d3NX, configuration Qt is
 A persistent particle ontology 109
2 3N
distributed according to |Ct(X)| d X. This feature is known as equivar-
iance. The statistical analysis of subsystems has been carried out in Dürr
et al. (2013b, ch. 2) for Bohmian mechanics. We discussed it in Chapter
3, section 4. This analysis applies here as well. Its bottom line is a proof
that Born’s rule holds true: if the effective wave function of the subsystem
is given by φt(x1,. . .,xn) for n  N, then the distribution of the subconfi-
guration of particles q(t) = (q1(t),. . .,qn(t)) is given by |φt(x)|2dnx.
Problem (II) stems from the fact that the pair potential U strongly entan-
gles all tensor components of the wave function Ct during the time evolu-
tion. Even a perfect initial anti-symmetric product state will therefore
immediately lose its product structure due to (4.6). The complexity of
this entanglement increases exponentially with N. Even today’s super com-
puters fail to solve the Schrödinger equation in general situations in three
spatial dimensions for more than three particles. This is already due to
memory requirements, not to mention the necessary computational
power. Consequently, one must hope to find special non-trivial situations
in which the very complicated N-particle dynamics becomes simple
enough so that it can be approximated by studying the motion of a few par-
ticles only.
The original motivation for Dirac’s hole theory was not the complexity of
its solutions, but stemmed from the attempt to make sense of Dirac’s equa-
tion describing only one particle (see Saunders (1991) for the context of
Dirac’s theory). One quickly realized that the solutions showed strange
behaviour—for instance the Zitterbewegung (Schrödinger (1930)) and
Klein’s paradox (Klein (1929)—which made Dirac’s equation hard to inter-
pret physically. The mathematical reason for this strange behaviour is the
presence of a negative energy spectrum of the free Dirac Hamiltonian
H0(x) as given in (4.7). Therefore, the starting point of Dirac’s hole theory
was to somehow suppress this negative spectrum. As the electrons move
like fermions, this can be accomplished by filling all the negative energy
states with a sea of particles—and, granted the cut-offs four to five, thus
ending up with an N-particle wave equation as noted earlier. The Pauli exclu-
sion principle then prevents the wave function of any additional electron
from growing negative energy components, because all negative energy
states are already occupied. The consequence is of course that one ends up
with the same complexity problem (II) as well, albeit for a different reason.
Our starting point is different. We seek a theory about the universe as a
whole, based on an ontology of permanent particles. We bring in Dirac’s
hole theory as a suitable means to implement that ontology. This model
together with its ontology is spelled out concisely in this and the preceding
section. However, even though we only treat a simplified model of that
theory and only describe electrons while modeling the rest of the particle
interactions effectively, it is natural to start with not only one but all the
electrons of the universe. The model can thus only be considered as a
serious contender, if it is also possible to analyze it mathematically
110 A persistent particle ontology
despite of the complexity problem (II), which is why we anyhow have to
cope with a large number of electrons N. This is the content of the rest
of this chapter of the book: our aim is to show how on the basis of this
model we can explain the statistical predictions of standard QFT.
Before doing so, let us compare our approach with the discussion of the
Dirac sea model in a Bohmian framework by Colin and Struyve (2007).
They also assume a finite fermion number; they are committed to positions
for fermions for both positive and negative energy particles, whereas there
are no bosons in the ontology. Anti-particles are defined as holes in the sea
of the negative energy particles. Whereas Colin and Struyve (2007) define
the fermion number as the total number of particles minus the number of
negative energy particles plus an infinite constant, we are committed to
N matter points that de facto coincide with the fermions. Nonetheless,
even though in our model there is no ontologically significant difference
between positive and negative energy particles, one may say with the
usual jargon that the fermion number can be defined as the number of pos-
itive energy particles plus the number of negative energy particles, which
remains constant. Colin and Struyve (2007) propose an equation of
motion for fermions that defines the vector velocity field for configurations
of particles, which is dependent on the wave function of the system accord-
ing to the usual Bohmian recipe; the expectation value of the fermion
number density is related to the position density: intuitively, the number
of fermions in a given region of space corresponds to the fermion positions
in that region. This model is regularized with the introduction of ultraviolet
momentum cut-offs and finite space; these constraints ensure that it is a
mathematically well-defined theory. Equivariance guarantees that the
empirical predictions of the standard model are reproduced, exactly as in
our case.
We emphasize again that so far the standard model of QFT only permits
to remove the introduced cut-offs when computing informal perturbative
corrections of scattering amplitudes with respect to the non-interacting
QFT, while the respective equations of motion for the standard model
become ill-defined with any attempt to remove the cut-offs. Since both
the theories of Colin and Struyve (2007) and the one developed in this
book are concerned with the dynamics and not with scattering theory,
they can only be compared to the standard model with cut-offs—until even-
tually a formulation of the equations of motion of the standard model
without cut-offs is found. Nevertheless, this does not infringe upon the
experimental adequacy of the models with cut-offs. On the contrary,
there is a vast spectrum of experimental evidence, starting from the histor-
ical Lamb shift calculation and ranging to recent results, such as the inde-
pendence of the effective particle masses of the infrared cut-off. In short,
these models agree with the experiments in large regimes, provided the
cut-offs are sufficiently large and an appropriate renormalization scheme
is applied.
 A persistent particle ontology 111
4.3 Equilibrium states and the vacuum
How can we tackle problem (II)—that is, find manageable approximations
of the in general very complicated dynamics of the Dirac sea model?
Clearly, this will be possible only in special situations—that is, for a
certain class of initial quantum states. Furthermore, such an approximation
cannot be carried out without coarse-graining the level of detail of the
information that is to be inferred about the system.
Consider, as an analogy, a classical gas of N particles confined to the
volume Γ. Although it is in principle possible to infer the actual motion
of the individual particles by solving Newton’s law of motion, for large
N, this would be a hopeless endeavour and for many practical purposes
unnecessary. For instance, even in equilibrium, the microscopic, actual
Newtonian motion of the particles may be very intricate. However, effec-
tively the net result is that, macroscopically, the gas density is almost cons-
tant with the variation around this constant density being small. The
smallness requirement depends on the practical purpose as it determines
which effects will be visible and which ones will drown in the fluctuations.
Hence, it defines what is meant by macroscopic. In the same vein, one can
introduce further macroscopic parameters like volume, pressure and tem-
perature. For most engineering purposes their relationship constitutes a
satisfactory description of equilibrated gases—that is, the theory of thermo-
dynamics. Such a coarse grained description is not restricted to equilibrium
states only. For example, in certain regimes it is possible to describe the
mediation of a perturbation created in the gas by an external influence in
terms of an effective equation for pressure (sound) waves that excite the ini-
tially equilibrated gas, without knowledge of the actual microscopic New-
tonian motion of the individual particles. Using this classical example as a
guide, the idea of describing complex dynamics in terms of small deviations
from an equilibrium whose time evolution can be understood in more
simple terms will be the key ingredient.
In quantum mechanics, much progress has been made in deriving such
coarse-grained or approximate descriptions with full mathematical rigour
in the case of, for instance, Bose-Einstein condensates and large molecules
(for recent works on Bose and Fermi gases see, e.g., Pickl (2011), Benedik-
ter et al. (2014)). The strategy underlying this type of analysis is to show
that in certain regimes—such as large particle number N, high or low
density, etc.—the reduced one-particle density matrix of the N-particle
system described by Ct can be well approximated by the one of a
product state, although the actual wave function Ct is far from being a
product state. The product state has a much simpler time evolution.
This comes at the cost of the level of detail that can be inferred from
such an approximation, which is then restricted to questions that can
be answered by knowing the reduced density matrices only. In many sit-
uations this is, however, all that is needed. The mechanism that is usually
112 A persistent particle ontology
exploited to derive such approximations rigorously is that if the variation
of the interaction operator—e.g. such as H Ik in (4.6)—is sufficiently small,
then it can be replaced by its expectation value for a certain time interval
without accumulating too much of an error. In the following, we set up a
similar strategy for the N-particle system of the Dirac sea, although we
have to concede that the contemporary mathematical techniques do not
yet allow us to control the approximation made in full rigour.
In this respect, we depart from the approach of Colin (2003) and Colin
and Struyve (2007). They, too, analyze the motion of the N particles in the
Dirac sea by comparing it to the motion that the particles would conduct in
a special state—namely, the one physicists call the vacuum state. Colin and
Struyve (2007) argue in the spirit of Dirac (1934) that, as the latter gives
rise to on average very uniform distributions of particles, deviations from
uniformity are observable provided the fluctuations around the uniform
motion in the vacuum state are sufficiently small. The distinction of the
observable motion against that uniform background is then based on the
expected particle densities and their fluctuations, for which explicit esti-
mates are given. We on the contrary look directly for approximate solutions
to (4.6). In doing this, we similarly track the motion of the Dirac sea of N
particles in terms of deviation from the motion of a reference state of the
Dirac sea. However, we do so already on the level of wave functions.
This will then naturally lead to the standard QFT formalism in terms of cre-
ation and annihilation operators. Furthermore, we argue that the choice of
reference state is not unique but rather a matter of choice of “good coordi-
nates” for a representation of states with respect to which one can econom-
ically arrive at approximate solutions to (4.6).
Let us pursue the idea of describing complex dynamics in terms of small
deviations from an equilibrium whose time evolution can be understood in
more simple terms. At first, we only consider V(t,x) in equation (4.6) to be
zero. That is to say, we stipulate that there are no external influences: the
motion of the electrons can be represented in terms of (4.4) and (4.6)—
that is, subject to the Fermi repulsion (due to the anti-symmetry of Ct)
and the Coulomb repulsion only. In analogy to the classical gas, the
simple approximate solutions inferred in the following play the role of
the equilibrium states. In the next section, we then consider the case V(t,
x)6¼0, which leads to excitations of the equilibrium—the analogue to the
pressure waves in the classical example.
As already indicated, the term creating problem (II) is the Coulomb pair-
interactions encoded in H Ik given by (4.8). Although the general dynamics
may be very complicated, there may be special initial conditions that lead
to a particularly simple dynamics, and thus, form a good reference
system to study more complicated dynamics. We therefore define a class
of initial quantum states that allow an approximation in which this interac-
tion term effectively vanishes and call it the class of equilibrium states—that
is, our reference states. This class shall consist of all states C 2 HN that
 A persistent particle ontology 113
fulfill the following two conditions: (a) The quantum mechanical expecta-
tion of the interaction operator H Ik is approximately constant—that is,
there is a constant EI such that the average interaction approximately ful-
fills
* +
 1X
E  C; H Ik ðxÞC ¼
I
C; Uðx  xj ÞC for all x 2 R3 and k ¼ 1; :::; N; ð4:9Þ
2 j6¼k

and (b) the fluctuation around this constant expectation value is sufficiently
well behaved throughout the time evolution. Condition (a) states that the
pair-interaction between the particles effectively averages out initially,
while condition (b) implies that this feature is preserved over time.
In view of the weak law of large numbers, these conditions (a) and (b)
can only be met for sufficiently large N. As the theory so far is meant to
apply to the total number of electrons in the universe, N is naturally
large. The exact sense of “approximatively” and “sufficiently” depends
on the practical purpose: conditions (a) and (b) are there to make sure
that the solutions to the fundamental equation of motion (4.6) for V(t,x)
= 0, given an initial state C0 in this class of equilibrium states, can for all
practical purposes be sufficiently well approximated—e.g. in the sense of
reduced density matrices—by a solution to the much simpler effective equa-
tion of motion
X
N
 
iℏ@ t Ct ðx1 ; :::; xN Þ ¼ H 0k ðxk Þ þ EI Ct ðx1 ; :::; xN Þ; ð4:10Þ
k¼1

replacing the complicated interaction H Ik by the constant EI.

Regarding the precise mathematical requirements of conditions (a) and (b)
and the precise sense of the approximation we are purposely vague, since the
exact behaviour of the fluctuation needed to carry out the rigorous mathe-
matics is not entirely settled in the fermionic case (unlike the bosonic case,
where for example Gross-Pitaevski and mean-field approximations can be
rigorously derived). If an initial state C0 fulfills (a) and (b) in a sufficiently
strong sense, the corresponding fully interacting time evolution (4.6) is
close to the non-interacting one (4.10) as the errors which depend on
the accumulated fluctuations (b) around the expectation value (a) can be
controlled—at least in an appropriate sense, such as the one of reduced
density matrices and for large enough N and gas densities. However, while
fermionic N-particle states that fulfill (a) are known (e.g. the non-interacting
fermionic ground state, see (4.15)), it is unknown whether there are ones that
fulfill also condition (b) in a sufficiently strong sense. In fact, it is currently
conjectured that (b) might be too strong and that demanding sufficiently
small fluctuations for only those particles with momenta below some thresh-
old might already suffice. This would mean that condition (b) could even be
weakened and we would still be able to show the closeness of (4.6) to the
114 A persistent particle ontology
approximate time evolution (4.10) for such states. The exact notion is
however irrelevant to our discussion. The important point is that (4.10)
does not have to be assumed, which would be highly questionable for an
interacting Fermi gas, but can be derived from much simpler and plausible
conditions such as (a) and (b).
Before considering a pertinent example of such an equilibrium state,
let us put the motivation for focussing on the class of equilibrium
states in other terms. Assuming that there are only electrons subject to
Fermi and Coulomb repulsion, the ground state Cgs of such an electron
gas is expected to be one in which for almost all initial Q0 with respect to
the relevant measure |Cgs(X)|2d3NX, the electrons are very homoge-
neously distributed. Then, if the measure |Cgs(Q)|2d3NX gives rise to a
homogeneous distribution, our defining condition (4.9) of the class of
equilibrium states is a consequence, and the net effect of H Ik is a constant
potential, canceling out all interactions: on the level of wave functions,
the electrons in such a state effectively do not “take notice” of each
others presence. Effectively, they move as if they were in a vacuum.
Hence, the dynamics generated by (4.6) is very simple, and, within the
bounds of the approximation, it leaves the class of equilibrium states
invariant.
A natural representative for a state in the equilibrium class would be the
actual ground state Cgs of the interacting system. Its mathematical struc-
ture, however, is extremely complicated and to date not accessible due to
the discussed entanglement induced by the H Ik terms. Consequently, we
have to find a simpler candidate that replaces Cgs in the sense of the afore-
mentioned approximation. Physicists usually choose the ground state Cgs  of
the corresponding approximate equation (4.10) given by
X
N
H 0k ðxk Þ Cgs
 ¼ E C ;
gs gs
ð4:11Þ
k¼1

0
where Egs is the lowest eigenvalue. Since the H k ðxk Þ commute pairwise, C
gs

0
can be found by studying the spectrum of H (x) only.
Due to the finite volume Γ and momentum cut-off Λ, the momenta are
restricted to p 2 P LG ¼ fk ¼ ðk1 ; k2 ; k3 Þ j k k k L; ki ¼ 2p=Lni ; ni 2 Z; i ¼
1; 2; 3g—in the case of Γ being a cube of length L = Γ1/3—; for each admis-
sible momentum value there are four one-particle eigenstates
H 0 þp;" ¼ þEp þp;" ; H 0 þp;# ¼ þEp þp;# ; H 0 p;" ¼ Ep p;" ; H 0 p;# ¼ Ep p;# : ð4:12Þ

These are characterized by the sign of the eigenvalue whose modulus is

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
determined by Ep ¼ p2 þ m2 and by a mathematical entity referred to
as spin, which is denoted by the subscript " or #. The one-particle Hilbert
space H ¼ L2 ðG; C4 Þ can then be split into two subspaces spanned by the
eigenstates of positive and negative eigenvalues, denoted by Hþ and H
 A persistent particle ontology 115
respectively. The ground state of the eigenvalue equation (4.11) can subse-
quently be given by the anti-symmetric tensor product

 ¼ φ1 ^ φ2 ^ ::: ^ φN ;
Cgs ð4:13Þ

where φn 2 fsk;k j s ¼ ; k 2 P LG ; k ¼"; #g, n = 1,2,. . .,N, are eigenstates

(4.12) with lowest possible eigenvalues. Due to the anti-symmetric tensor
product ^, these states must be pairwise distinct for Cgs to be non-zero;
however, their particular enumeration with respect to n does not play a
role. The total ground state energy Egs  given in (4.11) equals the sum of
the corresponding eigenvalues.
Among all such N-dependent ground states Cgs, the one for N ¼ 2jP LG j,
in which the total number of matter points equals twice the number of
admissible momenta in the set P LG (since there are two spin values), is dis-
tinguished. In this case, Cgs
 is built solely out of all eigenstates with negative
eigenvalues. Adding another eigenstate, which then must have a positive
eigenvalue, increases the total ground state energy Egs  . Taking one eigen-
state away also increases the ground state eigenvalue Egs  , since the corre-
sponding negative eigenvalue must be subtracted. Because of this fact,
one usually considers the case of a sea of N ¼ 2jP LG j Vmany particles and
denotes the corresponding ground state Cgs  by O :¼ k2P L ;s¼";# k;s , which

G
in physics is referred to as the Dirac sea of the vacuum as previously men-
tioned. Note that this is only a convenient mathematical idealization (and
might not be fully obeyed in nature as we seem to see slightly more
matter than anti-matter). Due to the anti-symmetric tensor product ^ one
may also choose (φn)n = 1,. . .,N to be any other orthonormal basis of the sub-
space H  H. This changes the definition of O at most by a constant phase
factor, which is irrelevant to the dynamics. Therefore, neglecting the phase
factor, we will represent O in the future as
O ¼ φ1 ^ φ2 ^ ::: ^ φN ð4:14Þ

for an arbitrary orthogonal basis (φn)n=1,. . .,N of H .

It can now be checked directly that for C = O in the equilibrium condi-
tion (4.9), one gets
Z
1 1X N
2 3
ð4:9Þ ¼ hO; H Ik ðxÞOi ¼ ðN  1Þ Uðx  yÞjφn j ðyÞ d y; ð4:15Þ
2 N n¼1

which is constant (note that |φj| = 1/|Γ|1/2). This agrees with Dirac’s heuristic
picture that, in an equilibrium state, the electrons are uniformly distributed
as, according to Born’s rule, the right-hand side of (4.15) can be interpreted
P
as expectation value of 12 j6¼k Uðx  xj Þ for (N  1) uniformly distributed
random variables xj. Beside this, there are unfortunately only few rigorous
results on the quality of the approximation of (4.10) to (4.6)—that is,
116 A persistent particle ontology
whether O also fulfills condition (b) in an appropriate sense to qualify as a
state in equilibrium. Nevertheless, the indication that it holds in certain
physically interesting regimes comes from the overwhelming accuracy
achieved by predictions obtained by means of using (4.10) whose justifica-
tion requires conditions (a) and (b) as a priori assumptions.
Before continuing our investigation of O, let us once more come back to
the discussion of the introduced cut-offs four and five. It is a mathematical
fact that when removing one of them (or both), the number of states in the
negative spectral subspace H —that is 2jP LG j, becomes countable infinite.
Hence, filling completely H is at tension with a primitive ontology of
finitely many persistent particles. It is therefore no option for our approach.
Since we have only a finite number of particles in the ontology, we can only
fill finitely many states in H , leaving the remaining infinitely many ones
unoccupied.
One may therefore raise the following concern: if one couples these par-
ticles to an open dynamical system such as the Maxwell field without
further constraints, an electron (even deep down in the sea) may pump
energy into the Maxwell field degrees of freedom by means of radiation
and thus acquire an ever more negative kinetic energy state. Since
without cut-offs there is no lower bound in H , this process may go on
forever so that it can eventually cause the entire dynamical system to
become unstable; this is the so-called radiation catastrophe (see Greiner
and Bromley (2000)). However, an open system with an energy sink at
spatial infinity into which the radiated energy may simply disappear
without ever interacting again with the sources is of course no viable can-
didate for a fundamental theory (not to mention the effects of the escaping
energy in a general relativistic setting). Hence, on top of the mathematical
difficulties discussed earlier, removing the cut-offs must be done in a phys-
ically sensible way in order to avoid a dynamic instability.
One way to achieve this is to conceive the interaction between the fermi-
onic fields and the boson fields as requiring that all radiation that has been
emitted by a source at a time instant has to be absorbed again by some
other source at another time instant. Such a requirement removes by definition
the discussed dynamic instability. If an electron were to radiate and thereby to
acquire a more negative kinetic energy state, another electron would have to
absorb the radiated energy and acquire a more positive kinetic energy state.
However, the larger the energy gap between the electrons is, the less likely
is an interaction between them. This means that depending on the initial
state of the system, there are effective upper and lower bounds on the
energy of the electrons so that the system is dynamically stable.
Nonetheless, until today, the mathematically status of such a require-
ment is unknown, not to mention a proof of the dynamic stability. That
notwithstanding, the given argument is a sound indication that one does
not have to abandon an ontology of finitely many particles when removing
 A persistent particle ontology 117
the cut-offs. Furthermore, as a mathematical idealization, it is of course
completely admissible to pass over to the thermodynamic limit (as done
in classical statistical mechanics), in the sense of infinitely many particles
and an infinite volume at constant density, whenever the corresponding
dynamical structure thereby becomes mathematically more tractable. As
in classical statistical mechanics, however, this limit has no bearing on an
ontology of finitely many persistent particles.
In Bohmian mechanics, all there is to the particles comes down to their
position in space only and the evolution of this position as given by a
guiding equation, in our case equation (4.4). In other words, in Bohmian
mechanics, particles are primitive objects in the sense that they do not
have intrinsic features, but are individuated only by the relative distances
among them. All the other parameters including mass, charge, spin, etc.,
are not additional elements of the ontology characterizing the particles,
but dynamical parameters employed to describe the evolution of the relative
distances among the particles. Against this background, we emphasize
again that an energy value is nothing but a dynamical parameter capturing
a particular form of motion. Hence, admitting negative energy values poses
no problem in an ontology that is committed only to particles as character-
ized by their position in space—in other words, matter points as individu-
ated by distance relations and their change. Negative energy values then are
only a bookmark to keep track of a certain evolution of these distance rela-
tions. The same goes for positrons and anti-matter in general: all these are
matter points moving in a certain manner.
In sum, we have defined a class of initial quantum states—namely, the
equilibrium class—that allow us to solve the, in general, very complicated
dynamics (4.6) approximately by the much simpler dynamics (4.10). Further-
more, we have found a particularly simple representative O of this class.
Solving the approximate equation (4.10) for the initial value C0 ¼ O yields
the explicit and particular simple evolution Ct ¼ Ot for
gs I
Ot ¼ eiðE þNE Þt O; ð4:16Þ

which can now be taken as approximate solution Ct to equation (4.6) for the
same initial value C0 = O. Such a state only gives rise to a trivial dynamics of
a sea of N electrons in which none of them “takes notice” of the rest as if
they were in a “vacuum”—hence the name vacuum state. Nevertheless,
this state O will provide the basis for studying more interesting dynamics
in the next section.
Finally, it has to be emphasized that even with Ct and Ot being close ini-
tially for C0 = O in the sense of, for instance, reduced density matrices, little
information is provided about the closeness of the velocity fields (4.4) gen-
erated by Ct and Ct , respectively. In general, the respective velocity fields
and the corresponding trajectories differ. However, the results of their sta-
tistical analysis agree due to the mode of the approximate agreement in
118 A persistent particle ontology
terms of, for instance, the reduced density matrices. In other words, the
price that one has to pay to overcome the complexity problem (II) by
means of an approximation in terms of solvable equations is that one has
to abandon the hope of obtaining a calculation of actual trajectories in
favour of a statistical analysis. This is, however, the same situation as in
the example of the classical gas discussed at the beginning of this section.
Furthermore, it is the same situation as in Bohmian mechanics: there also
is no point in calculating individual trajectories, since tiny deviations in
the initial configuration will lead to large deviations of the resulting
Bohmian trajectories. Consequently, our knowledge of subsystems of the
universe is limited to what can be obtained from Born’s rule, as explained
in Chapter 3, section 4.

4.4 Excitations of the vacuum and the Fock space formalism

More interesting dynamics will take place if we allow V(t, x) to be non-
zero, thus including external influences. Under the action of V(t, x), an
initial vacuum state O, as defined in (4.14), may evolve into an excited
state, as for instance
F ¼ w ^ φ2 ^ ::: ^ φN ; ð4:17Þ

where the element φ1 of the orthonormal basis of H is replaced by a wave

function w 2 Hþ , a so-called electron-position pair. For the sake of simpli-
city, let us assume that after the transition from O to F, the external influ-
ence vanishes again. As in the preceding section we strive to find an
economic effective description of the time evolution of F to a Ft, t > 0
that fulfills the complicated dynamics (4.6)—that is,
i@ t Ft ¼ H N Ft ; ð4:18Þ

where HN is the N-particle Hamiltonian defined in (4.6).

In analogy to the pressure waves in the classical gas example mentioned
in the previous section, the leading idea is to describe the complex evolution
of Ft by the evolution of the excitation only as compared to the simple evo-
lution of the reference state Ot given in (4.16). In mathematical terms this
can be done by the following ansatz
X
N Z
3
Ft ðx1 ; :::; xN Þ ¼
kþ1
ð1Þ d yk Zt ðxk ; yk ÞOt ðx1 ; :::; yk ; :::; xN Þ: ð4:19Þ
k¼1

The summation and the factor (1)k+1 are to ensure the anti-symmetry of
the wave function. For the choice
Ot¼0 ¼ O; Zt¼0 ¼ w φ1 ; ð4:20Þ
 A persistent particle ontology 119

one obtains F ¼ F due to the orthonormality of the states χ, φ1,. . .,φN.
t¼0
The excitation is hence encoded by a two-particle wave function ηt(x, y). Its
x tensor component, which we shall refer to as electron component, tracks
the evolution of the initial excitation χ. Its y tensor component, which we
shall refer to as hole component, tracks the evolution of the corresponding
state in the reference Ot that is missing.
As these components might entangle during the evolution, they cannot
be described by separate one-particle wave functions. In order to find a
good effective evolution equation for ηt, one has to study the evolution of
the difference Ft  Ft , which due to the unitarity of the time evolution
amounts to gain good control of
d
F ; F ¼ hiH N Ft ; Ft i ð4:21Þ
dt t t

* X
N
Z   +
kþ1 d
þ Ft ; ð1Þ dyk Z ðx ; y Þ Ot ðx1 ; :::; yk ; :::; xN Þ ð4:22Þ
dt t k k
k¼1

* X
N
Z +
kþ1
þ Ft ; ð1Þ dyk Zt ðxk ; yk ÞðiH ÞOt ðx1 ; :::; yk ; :::; xN Þ :
N
ð4:23Þ
k¼1

We observe that
Z
 
dyk H N Zt ðxk ; yk ÞOt ðx1 ; :::; yk ; :::; xN Þ  Zt ðxk ; yk ÞH N Ot ðx1 ; :::; yk ; :::; xN Þ ð4:24Þ

Z
(
¼ dyk ððH 0k ðxk ÞZt ðxk ; yk Þ  Zt ðxk ; yk Þ H 0k ðyk Þ  Uðxk  yk ÞZt ðxk ; yk ÞÞ ð4:25Þ

0 1
BXN X
N C
B C
þB Uðxk  xm Þ þ Uðxk  yk Þ  Uðyk  xm ÞÞZt ðxk ; yk ÞÞÞOt ðx1 ; :::; yk ; :::; xN Cð4:26Þ
@ A
m¼1 m¼1
m¼
= k m¼
= k

(
holds, where H 0 ðyÞ ¼ ia  ry þ bm with the gradient acting to the left.
The terms in (4.26) have two roles. On the one hand, they describe how,
due to the presence of the excitation ηt, the excited Dirac sea deviates
from the equilibrium condition (4.9)—a phenomenon that is referred to
as vacuum polarization. On the other hand, they describe the back reaction
on the electron and hole components of ηt. Note the relative signs: the first
two terms in (4.26) exert a repulsive interaction on the electron component
120 A persistent particle ontology
of ηt stemming from the x1,. . .,yk,. . .,xN components of the Dirac sea,
while the third one exerts an attractive interaction on the hole component
of ηt due to the xn, n 6¼ k, components of the Dirac sea—exactly as one
would expect, since the hole component describes the absence of one elec-
tron in the sea. Hence, it gives rise to the same charge of an electron,
however, with opposite sign. According to Born’s rule, which applies
thanks to the discussed mode of approximation, the position distribution
of these two charges is given by |ηt(x, y)|2d3xd3y.
If N is large and, as in our case, the number of excitations is sufficiently
small compared to N, one may argue that the sums in the first and third
term in (4.26) may be replaced by their respective mean-field values,
which are constant due to the nature of Ot. This approximation is of
course only justified for small coupling constants e2, as starting from
order e4 also the state Ot will be perturbed due to the polarization terms.
However, since our focus is on the ontology and not on finding the best
approximation, we shall simply neglect this higher order polarization
effect in the following. Hence, the remaining term (4.25) dictates the evo-
lution of the electron-hole wave function ηt according to
(
i@ t Zt ðx; yÞ ¼ H 0k ðxk ÞZt ðxk ; yk Þ  Zt ðxk ; yk ÞH 0k ðyk ÞUðxk  yk ÞZt ðxk ; yk Þ: ð4:27Þ

Although a rigorous mathematical justification is lacking, one may have the

well-founded hope that for the ansatz (4.19) with the approximated evolu-
tion of the vacuum state (4.16) and initial values (4.20), the right-hand side
of (4.21) could in principle be estimated with Grönwall type estimate using
our computation (4.24) to give a bound on the difference between Ft and
Ft —that is, the quality of our approximation for example in terms of a
trace norm of the respective reduced density matrices.
In sum, we arrive at an economic two-particle description given by
(4.20), (4.16) and (4.27) that approximates the actual N-particle dynamics
of the initial excitation F in (4.17). As the initial vacuum state essentially
does not change over time—cf. (4.16)—and since due to (4.9) the
vacuum behaves as if there are no electrons, for all practical purposes,
one may be inclined to forget about O entirely. The vacuum state O then
simply encodes a state without excitations. A transition from O to Ft by
means of (4.6), however, requires the introduction of the two-particle
wave function ηt. Figure 4.2 illustrates these two views.
Further excitations require the introduction of further multi-particle
excitations and so on. However, as long as the number of excitations is
small compared to N, those excitation wave functions are much less
complex objects to keep track off than the actual N-particle dynamics of Ct.
To streamline the mathematics according to this idea, it makes sense to
formulate the N-particle wave function dynamics (4.6) not on H^N , the
Figure 4.2 Both illustrations depict the same excitation of the wave function of the
universe. It may have started out as the vacuum O with all negative
energy states occupied, cf. (4.14), and then subject to the action
of the external influence V, evolved into the excited universal wave
function F given in (4.17). The top figure gives an account of this
process in terms of the wave function of the universe. The approximate
description of the evolution of the universal wave function Ft given by
means of Ft , cf. (4.19), is one in which the vacuum states Ot, cf. (4.16),
and the excitation ηt, cf. (4.27), are evolved separately. As Ot is constant
up to a phase, one may simply neglect it in the description and only
account for the additional positive energy electron states and the
states that have to be taken out of Ot in order to create the appropriate
holes in the sea. This view of the universe is depicted in the bottom
figure. As discussed in the text, this naive picture is only valid as long
the vacuum polarization can be neglected—that is, the reaction of the
created excited states back on the sea states that accumulates over
time. Nevertheless, the view in the bottom figure can be maintained
even when using for a reference state Ot, another one whose time evo-
lution incorporates these polarization effects. This is also the view that
is adopted in the standard formulation of QFT—see our comment
(4.29) about the O-dependence of the representation.
 122 A persistent particle ontology
Hilbert space of the N matter points that move electronwise, but on a space
that only keeps track of the varying number of wave function excitations
with respect to a reference state such as O much in the spirit of the
bottom illustration in Figure 4.2. For this purpose one introduces the so-
called Fock space F O , which can be defined by employing the so-called cre-
ation and annihilation operators. Algebraically the creation operator ψ* is
given by

c ðf Þ φ1 ^ φ2 ^ :::φN ¼ f ^ φ1 ^ φ2 ^ :::φN ð4:28Þ

for any choice of f ; φk 2 H. The annihilation operator ψ is defined as the

corresponding adjoint ψ. Due to the anti-symmetric tensor product ^
these operators fulfill the anti-commutation relations
fcðf Þ; c ðgÞg ¼ hf ; gi; fcðf Þ; cðgÞg ¼ 0 ¼ fc ðf Þ; c ðgÞg for all f ; g 2 H;
ð4:29Þ
and; furthermore; cðwÞO ¼ 0 ¼ c ðφÞO for all w 2 Hþ ; φ 2 H :

The Fock space F O can then be defined as the tensor product F O ¼

F eO F hO of the electron Fock space F eO and hole Fock space F hO , which
are spanned by finite linear combinations of the form ψ*(f1). . .ψ*(fn)O
and ψ(g1). . .ψ(gn)O for f k 2 Hþ and gk 2 H , respectively. In this algebraic
construction the specific reference state O in H^N is hidden as O is encoded
by |Oi = 1 1. It has to be emphasized that the independence of |Oi from O
in this notation (as also suggested by the bottom illustration in Figure 4.2),
however, is an illusion, since O is encoded in the relations (4.29)—acknowl-
edging the fact that O is the Dirac sea with all states in H occupied. By
construction, there is a well-known canonical isomorphism between H^N
and the N-particle sector of F NO —that is, the subspace of F O spanned by
states with equal numbers of electron and hole excitations. It reads

i : HN ! F NO ; iðf 1 ^ f 2 ^ ::: ^ f N Þ :¼ c ðf 1 Þ:::c ðf N Þcðφ1 Þ:::cðφN ÞjOi ð4:30Þ

for f k 2 H. Thus, the approximate excitation Ft for (4.19) is represented in

F O by
Z Z
3
jFt i ¼ d x dy Zt ðx; yÞc ðxÞcðyÞjOi: ð4:31Þ

Using these translation rules implied by ι, we can recast the actual N-par-
ticle dynamics (4.6) generated by the Hamiltonian HN in terms of creation
and annihilation operators as
i@ t jCt i ¼ H~ jC i
t
Z Z Z ð4:32Þ
 
~ ¼
H
3 3 3
d x c ðxÞ H 0x þ V t ðxÞ cðxÞ  d x d y c ðxÞc ðyÞUðx  yÞcðyÞcðxÞ:
 A persistent particle ontology 123
This expression is the standard second-quantized Hamiltonian of quantum
electrodynamics (QED) in Coulomb gauge when neglecting radiation. Note
that in H~ the creation and annihilation operators appear in pairs, which
ensures that for an initial state in the N-particle sector of F O the generated
dynamics also remains there.
In conclusion, for an initial wave function Ct¼0 2 H^N , there is a unique
initial Fock vector |Ct=0i := ι(Ct=0) such that the time evolutions ðCt Þt2R on
Hilbert space H^N and ðjCt iÞt2R on Fock space F O generated by the Ham-
iltonians HN in (4.6) and H, ~ respectively, fulfill |Cti := ι(Ct) for all times
t. In other words, the N-particle formalism is equivalent to the Fock
space formalism when restricting to the N-particle sector. While the
former represents the dynamics absolutely, the latter represents the dynam-
ics with respect to a reference state O (see again Figure 4.2). The latter is, as
we described above for the case of the two-particle excitation, tailor-made
for finding approximations of excitations of O.

4.5 The appearance of particle creations and annihilations

In using the second-quantized formalism in section 4.4 instead of the N-par-
ticle formalism in (4.6), one is naturally led to a mode of language in which
excitations are “created” and “annihilated”. When an initial reference
vacuum state O evolves into an excited state like F in (4.17), an excitation
ηt is created, which effectively evolves according to (4.27). As can be seen
from (4.27), the interaction between the electron and hole components of
ηt is attractive, which may lead to a recombination unless an external influ-
ence V(t, x) prevents it. In such a process the evolved excitation Ft decays
back to the reference state O resulting in the annihilation of the excitation.
One may be inclined to refer to the electron component of ηt as describing
one created “electron” and the hole component as describing one created
“hole” or, synonymously, “positron”. This, however, is an abuse of the
well-defined term “electron” given by the introduced ontology—namely, a
persistent particle that moves electronwise. Even if the state Ft in (4.19)
were not only an approximation but the actual N-particle wave function,
by the form of the velocity law (4.4) and anti-symmetry, the electron compo-
nent of ηt would not influence only one electron, but all of them. It is, there-
fore, the collective motion of all electrons that causes the excitation ηt to
appear. In order to develop a heuristic picture about this, let us go back to
our initial analogy of the pressure waves in a classical gas. As we have dis-
cussed already, the pressure wave in the gas corresponds to the excitation
ηt. The phenomenon of the pressure wave is perceived by the corresponding
increase of density due to the configuration of gas molecules. In the same
way, at a particular time t, it is the configuration of electrons in the Dirac
sea that makes up the position statistics of the excitation ηt in the Dirac
sea. Here, it is, however, important to note that the gas molecules do not
124 A persistent particle ontology
necessarily move along with the dispersing pressure wave in the gas; instead,
they oscillate. At different times, the phenomenon of the pressure wave is due
to different molecules of the gas. Similarly, it is not necessarily the same elec-
trons that make up the position statistics encoded in the excitation ηt. It
would thus be a misconception to interpret ηt as guiding a two-particle
system when the word “particle” is intended to refer to matter points existing
in nature.
On the mathematical level of wave functions, the hole component of ηt
simply encodes what has to be taken out of the reference state Ot, while the
electron component of ηt encodes what has to be added to Ot in order to get
a good approximation of the actual quantum state Ct of the Dirac sea. This
convenient approximation in terms of O comes at the price of losing direct
information about the actual particle trajectories: as mentioned, in this case
only the statistics about particle positions encoded in the wave function are
under control. Hence, all that can be said about ηt is that against the
uniform distribution of a sea of electrons described by Ot—due to the equi-
librium condition (4.9)— jηt(x, y)j2d3xd3y is the probability of finding an
additional negative charge and the absence of a negative charge (or mathe-
matically equivalently, the presence of a positive charge) in the volumes d3x
at x and d3y at y, respectively.
Nevertheless, the excitation ηt has the same properties as wave-packets
and would in principle be capable of guiding two particles. To emphasize
the difference between such excitations and actual particles, the term
quasi-particles is usually employed in physics (as, e.g., phonons being the
quasi-particles describing quantum excitations of crystal lattice oscilla-
tions). In this view, electron and hole components of ηt describe electron
and hole quasi-particles having exactly the same properties except for the
opposite charge, as can be seen from (4.27). The bottom line is that at all
times there are N electrons in the Dirac sea. None of them were ever
created or destroyed. It is their collective motion, which may deviate
from the vacuum motion, that gives rise to what we refer to as “electron-
positron pair-creation”.
One can emphasize the quasi-particles character of the excitations
already in the mathematical formalism. For this it is convenient to split
the field operators ψ*, ψ into electron and positron creation and annihila-
tion operators
 
c ðf Þ ¼ b ðf Þ þ c ðf Þ with b ðf Þ ¼ c ðPþ f Þ; c ðf Þ ¼ c ðP f Þf 2 H; ð4:33Þ

where P± are the orthogonal projectors on H to H , respectively. The number

of electrons and hole quasi-particles can then be given by
X  X
Ne ¼ b ðwn Þbðwn Þ; Nh ¼ cðφn Þc ðφn Þ; ð4:34Þ
n n

using any orthonormal bases (χn)n=1,. . .,N in Hþ and (φn)n=1,. . .,N in H .

 A persistent particle ontology 125
Note that by virtue of (4.29), for instance,

N e jOi ¼ 0 ¼ N h jOi; ð4:35Þ

while for F in (4.17), NejFi = 1 = NhjFi as expected.

However, much caution has to be taken when interpreting the particle
number operators in (4.34). First, these operators do not yield a definite
number before taking, for instance, their expectation value. Even then,
rather than an absolute number, these operators record only the number of
wave-packet excitations—such as, for instance, ηt—relative to the chosen ref-
erence state Ot, as can be seen from (4.35)—that is, (4.29). Furthermore,
recall that in the preceding section when considering the case of zero external
influence—V(t, x) = 0—we argued that the ground state of the free theory, O,
is a good candidate to approximate the vacuum. Let us stress again that this
choice stemmed from the need to find a good approximation and was by no
means unique (which is why we regarded a whole class of reference equilib-
rium states—recall that O was just a representative in the case of no external
influence V(t, x) = 0). Depending on the kind of approximation in mind, one
choice of reference state O might be better than another. Changing the refer-
ence state, however, naturally changes the meaning of the number operators,
as excitations are only defined with respect to it.
For instance, when an external influence V(t, x) is present, the equilib-
rium conditions—that is, conditions (a) and (b) discussed in (4.9) and
later in this chapter—naturally change. When V(t, x) only changes very
slowly, physicists usually employ the so-called Furry picture: for each
time t one chooses as reference O the state in which all negative energy
states of the Hamiltonian H0(x) + V(t, x) are filled. However, it is shown
in Fierz and Scharf (1979) that even this seemingly canonic construction
of reference states (4.34) is not Lorentz invariant—in the sense that what
might appear as an empty vacuum in one reference frame may contain
many quasi-particles in another. This is because the spectrum, employed
to define O, transforms like time under Lorentz boosts. Hence, it does
not give rise to an invariant.
The choice of a reference state O will ultimately depend on physics in the
sense that it concerns an experiment on the appropriate detector model that
specifies what the equilibrium conditions (4.9) are, and hence, which exci-
tations cause clicks. At temperature zero, a good candidate for a reference
state fulfilling this condition surely is the interacting ground state corre-
sponding to the given experimental setup. This is because the Fermi exclu-
sion principle and the Coulomb repulsion will naturally act to distribute the
electrons in a sufficiently uniform manner in order to meet (4.9). Mathe-
matically, such states are very difficult to compute, as, due to the pair-
interaction, these states are highly entangled instead of being product
states. However, as is evident from the success of QED, in many situations
these highly entangled ground states can for all practical purposes be
126 A persistent particle ontology
approximated by product states such as O (e.g. in the sense of reduced
density matrices as discussed earlier). It has to be emphasized that this
need for a reference state is only an artifact of our approximation in
order to deal with the complexity problem (II) of section 4.2. The Dirac
sea model is independent of observers or detector models: there are
always N electrons in the sea moving according to the laws (4.4) and (4.6).
Consequently, the choice of O should rather be seen as a choice of suita-
ble coordinates with respect to which one may track the complicated N-
particles dynamics (4.6) most conveniently. In principle, any choice is pos-
sible, since the isomorphism (4.30) between the N-particle formalism (4.6)
and the formalism in terms of creation and annihilation operators (4.32)
does not depend on particular properties of O. The number operators
that are defined with respect to this choice O, however, can be assigned a
physical interpretation only with respect to that particular O. For
example, if O fulfills the equilibrium conditions (a) and (b) discussed in
(4.9) and later in this chapter, the net sum of all pair potentials between
the electrons vanishes and each electron effectively feels a constant potential
and, thus, moves freely as if it were in a vacuum (a prime example of what
could be called dark matter, as effectively there is no interaction except for
maybe gravitation). The presence of an excitation of O disturbs those con-
ditions, resulting in an effective interaction between all electrons in the sea
with the electron-hole excitation ηt, which dictates the form of (4.27). The
density jηtj2 then describes the distribution of an additional negative and
positive charge with respect to the uniform background distribution of O.
Only in this sense do the particle number operators record the number of
in principle physically measurable excitations. This is what is gained
from the introduction of an ontology of persisting particles: it is capable
of explaining the import of all the relevant mathematical quantities of the
theory.
The Unruh effect (Unruh (1976)) can also be understood in this way: the
equilibrium conditions were chosen to find a good approximation of the
complex N-particle dynamics and are at best Lorentz invariant. Switching
to an accelerated frame will, in general, violate those conditions. Conse-
quently, states O and O0 describing a vacuum in a rest frame and an accel-
erated frame, respectively, may differ. Thus, O may appear in the
accelerated frame as excitation of O0 and a detector in the accelerated ref-
erence frame may therefore click. In sum, excitations cannot be given an
absolute but only a relative meaning with respect to a certain equilibrium
state chosen for our effective description. This is the reason why there is
no point in according an ontological significance to these excitations in
the form of particle number operators or quasi-particles—which are the
ontological particles in the proposal of Bell (2004, ch. 19)—since there is
no canonical choice for O.
Finally, let us briefly come back one last time to the removal of the cut-
offs and the mathematical idealization of the thermodynamic limit of
 A persistent particle ontology 127
infinitely many particles. It is a fact that when V(t, x) is, for instance, given
by an electromagnetic four-potential with non-zero spatial components, in
this limit the time evolution cannot be implemented on Fock space F O with
respect to the vacuum O as defined in (4.14) (see Ruijsenaars (1977)). Even
a gauge transformation for a gauge four-potential with non-zero spatial
components cannot be implemented as a map on F O into itself. A way
out of this dilemma is to construct Fock spaces F OVðtÞ for all times t with
respect to reference states OV(t) that are appropriately chosen according
to the external influence V(t, x), instead of the fixed vacuum O. A detailed
discussion of this fact goes beyond the scope of this book. In brief, the
upshot is that the time evolution has to be implemented on time-varying
Fock spaces in terms of a map F OVðt Þ ! F OVðt Þ for initial and final times
0 1
t0, t1 (see Deckert et al. (2010)). The restriction on the choice of admissible
OV(t) depends on V(t, x); it is by no means unique. Hence, the particle
number operators in the Fock spaces F OVðtÞ refer to excitations of different
Ot. This shows again that there is not one vacuum O and no absolute par-
ticle number operator, but only a “choice of coordinates” in terms of refer-
ence states OV(t), which makes the mathematical discussion of the Dirac sea
tractable. However, this poses no problem, as the physical meaning of what
we call excitations is as clear as the physical meaning of the reference state,
which is specified by a physical consideration such as the equilibrium con-
dition discussed above—or more concretely, a detector model.

4.6 The merits of the Bohmian approach

Before concluding this chapter, two points are worth being addressed. In
the first place, one may gain the impression that the introduced formalism
based on a reference state O, which fills the spectral subspace H with
actual particles that move electronwise and leave Hþ empty, violates one
of the much-celebrated symmetries of QED—namely, the so-called charge
symmetry (a symmetry that is broken in the weak interaction). This is a mis-
conception. Charge symmetry simply demands that the equations of motion
remain invariant when the signs of the elementary charge e and the one of
the external influence are inverted, which is the case (see (4.6) and (4.27) for
the effective equation).
Furthermore, it has to emphasized that even in the non-interacting case,
while the statistics about the particle positions (which are encoded in the
wave functions of the excitations thanks to Born’s law) are Lorentz invari-
ant, the actual particle trajectories generated by the velocity law (4.4) are
not. It is a well-known fact that the latter depend on a preferred foliation
of Minkowski space-time—in the setup above, we have chosen equal
time hypersurfaces. This circumstance is not an artifact of our model; it
is generic to all relativistic versions of Bohmian mechanics. The reason is
the manifestly non-local nature of quantum mechanics as demonstrated,
128 A persistent particle ontology
for instance, by Bell’s theorem and the subsequent experiments. Instead of
adding this extra structure of a preferred foliation by hand, it is possible to
introduce it in a Lorentz-covariant manner by taking it to be determined by
the N-particle wave function itself, or equivalently the Fock space vector
(see Dürr et al. (2013a)). We did not do so because this would have
made the definition of the Dirac sea model and our arguments for a particle
ontology unnecessarily opaque. Furthermore, the actual trajectories of indi-
vidual electrons figure only in the ontology of our theory, while, due to
problem (I) mentioned in section 4.2, they alone do not allow to infer pre-
dictions. It is rather the collective motion of all particles in the Dirac sea
that generates the correct position statistics, which are encoded in the exci-
tation wave functions by Born’s law; the latter are accessible and Lorentz-
invariant. Again, the lack of Lorentz invariance of the particle trajectories
in Bohmian mechanics should not be counted against the theory, since
these are necessary to obtain definite measurement outcomes, and no
one has developed a dynamics of quantum systems that is both Lorentz
invariant and that accounts for definite measurement outcomes (see also
section 5.5).
Despite the still open desiderata, we have demonstrated that the predic-
tions of the standard model in the electron sector with cut-offs and neglect
of radiation arise naturally from a theory of N persistent particles. Creation
and annihilation appear only in an effective description with respect to a
chosen reference state. This programme is not only applicable to the elec-
tron, but to all fermion sectors.
In conclusion, let us stress again the parallelism with Bohmian mechanics
for quantum mechanics: in both cases, the ontology is exactly the same—a
fixed, finite number of permanent point particles that are characterized only
by their positions—that is, the relative distances among them. Furthermore,
the dynamics is of the same type—that is, a deterministic law describing the
motion of the particles on continuous trajectories without any jumps. In
both cases, the objective is to answer the questions of what there is in
nature on the fundamental level (i.e. permanent matter points) and how
what there is evolves (i.e. provide a dynamics for the individual processes
in nature). From this ontology and dynamics then follow the formalisms
of statistical predictions of both standard quantum mechanics and QFT.
In this sense, these predictions are explained by this underlying ontology
and dynamics.
The same goes for the solution to the measurement problem that hits
QFT in the same way as quantum mechanics: matter points like the
Bohmian point particles are admitted because they explain the measure-
ment outcomes. For instance, the dots on the screen as recorded in the
double slit experiment are made up by these point particles. However,
this explanation is not achieved by the (primitive) ontology of particles
alone, but by this ontology together with the dynamics (i.e. the guiding
equation): it is the dynamics that provides for the stability of macroscopic
 A persistent particle ontology 129
objects, including pointer positions and dots on a screen. As Dickson (2000)
convincingly argued, to obtain this stability, no properties of the particles
over and above their position in space are needed, but a dynamics is that
provides for trajectories such that there are stable macroscopic particle
configurations.
Bearing these facts in mind, the solution to the measurement problem
provided by Dirac sea Bohmian QFT is of the same type: here again, the
Bohmian point particles are admitted because they explain the measure-
ment outcomes. There always is a definite, finite number of particles
moving on continuous trajectories, with some of these particles making
up the macroscopic phenomena that we see. As discussed, we only have
to be more careful to relate those particles and their dynamics to what
we see with the naked eye such as the trajectories recorded in cloud
chamber detectors. The latter are generated by deviations from the collec-
tive motion of the Bohmian particles in an equilibrium state such as the
vacuum. Thus, to stress again, what explains these phenomena is not the
mere fact of there being particle configurations, but the particle dynamics:
it consists in this case of excitations from a sea of particles in an equilibrium
state, with these excitations, as elaborated on earlier, implying a change of
the motion of in principle all the particles in the sea. To sum up, what
explains the traces in the cloud chamber are these excitations as affecting
the motion of the particles in the sea.
In sum, the Bohmian approach works for QFT in the same way as for
quantum mechanics, and the argument for this approach is the same in
QFT and in quantum mechanics: a solution to the measurement problem
that tells us what happens in nature on the level of individual processes
and that spells out how what thus happens explains the macroscopic phe-
nomena including measurement outcomes. The Bohmian approach vindi-
cates the minimalist ontology of there being only matter points
individuated by distance relations and the change of these relations also
when it comes to our currently most advanced physical theory of matter—
namely, the standard model of QFT.
5 Relationalism for relativistic
physics

5.1 The challenge from relativistic physics

The minimalist ontology that we pursue in this book is based on the follow-
ing two axioms:

(1) There are distance relations that individuate simple objects—namely,

matter points.
(2) The matter points are permanent, with the distances between them
changing.

The distance relations are conveniently represented in terms of a three-

dimensional geometry such as Euclidean geometry, although neither the
geometry being three-dimensional nor its being Euclidean are necessary.
The change in the distance relations enables a representation in terms of
an order that is unique and directed, with time being that order, although
its metric depends on the choice of a subsystem of distance relations
within the universal configuration of matter relative to which change is
measured.
A physical theory introduces a dynamical structure in terms of dynamical
laws (usually formulated as differential equations) in which various dyna-
mical parameters figure that are attributed either to the matter points
taken individually (e.g. mass) or to an entire configuration of them (e.g.
an entangled wave function), in the last resort the whole configuration of
matter points of the universe (e.g. the universal wave function). The task
of the dynamical structure is to capture the change in the distance relations
among the matter points in a manner that is both simple and informative
about that change. Neither the geometry nor the dynamical structure
belong to the ontology. The ontology is completely given by the two men-
tioned axioms.
This ontology rests on a crucial distinction between variation (axiom 1)
and change (axiom 2). There is variation within a configuration in virtue
of the distance relations being such that they discern—and thereby
individuate—the matter points. And there is change in the distance
132 Relationalism for relativistic physics
relations, with that change providing for an intertemporal identity of the
matter points, enabling a representation of that change in terms of the
matter points moving on continuous trajectories. The argument for this
ontology is its simplicity together with its empirical adequacy: through
axiom 1, this ontology is most simple. If there is a plurality of objects,
there has to be a relation of a certain type that unifies the world. If that
relation also individuates the objects—so that no commitment to intrinsic
properties, a primitive thisness or bare substrata is called for—the ontol-
ogy is most simple. The distance relation satisfies this condition. It is at
least not obvious what other type of concrete physical relation could do
so as well. Through axiom 2, this ontology is empirically adequate. If
there were only relations that individuate the fundamental objects, but
no change, the ontology would be most simple. However, it would have
no chance of being empirically adequate. The simplest way to introduce
change is in terms of change of the relations that are there anyway to indi-
viduate the objects. This is sufficient to obtain empirical adequacy: all the
empirical data consist in relative positions of objects and change of these
positions (i.e. motion). That is why no ontological commitment neither to
the geometry nor to the dynamical parameters that a physical theory
employs in order to represent that change is necessary (which would
require at least one additional axiom). The two mentioned axioms consti-
tute a minimally sufficient ontology that is empirically adequate. In a nut-
shell, as argued in the preceding chapters, less won’t do for an ontology of
the natural world; bringing in more creates new drawbacks instead of pro-
viding additional explanatory value.
The geometry and dynamical structures considered in Chapter 3 (clas-
sical mechanics of gravitation, Bohmian quantum mechanics) and in
Chapter 4 (Dirac sea Bohmian QFT) are such that the configuration of
matter is represented as being inserted into a three-dimensional Euclidean
space and as evolving in time such that the evolution of the configuration
at a time t is fixed by dynamical parameters that are attributed to that con-
figuration at that very t (e.g. particle masses and charges, or an entangled
wave function). Nonetheless, for the geometry and the dynamical struc-
ture to achieve a description of the change in the configuration of
matter that is most simple and informative, this representation has to be
based on the change in the distance relations that actually occurs. In
other words, the change comes first, and then come the geometry and
the dynamical laws as means of representing that change. If the geometry
and the dynamical structure are no addition to being, then the truth-maker
for the propositions that ascribe geometrical as well as dynamical param-
eters such as particle masses and charges or a wave function to the config-
uration of matter points at a time t cannot be the distance relations at that
very t. What makes these ascriptions true is the change that actually
occurs in the distance relations among the matter points. The way in
which that change occurs manifests certain patterns or regularities.
 Relationalism for relativistic physics 133
Because there are such patterns, a physical theory can achieve a descrip-
tion of that change that is most simple and informative by attaching geo-
metrical and dynamical parameters to the configuration of matter points.
It is a gain in simplicity if by attributing dynamical parameters to the con-
figuration of matter points at a given time t and inserting these parameters
as initial conditions into a dynamical law, one achieves a description of
the whole past and future evolution of that configuration, although the
truth-maker for this ascription is not located in the configuration of
matter at that very t.
However, it is a contingent issue whether the patterns that the change in
the distance relations among the matter points exhibits are such that this
change can be represented in terms of a dynamical structure that requires
only parameters attributed to the configuration of matter points at a
given time as initial condition. As regards the ontology, the change in any
one distance relation between two matter points entails a change of in prin-
ciple all the other distance relations in the configuration. But this does not
imply that when it comes to representing change, it is possible to obtain a
representation that is both simple and informative about the change that
actually occurs on the basis of initial conditions that consist in parameters
attributed to the configuration of matter points at a given time. The corre-
lated motions of quantum particles represented by means of an entangled
wave function enable such a representation and are the reason why we
have written down the fundamental dynamics of QFT in a non-relativistic
manner in Chapter 4 (in the sense that this dynamics requires a notion of
objective simultaneity).
Relativity physics challenges this manner of representing change. In rel-
ativistic interactions, it is dynamical parameters situated in the past (and
possibly also in the future) that capture the evolution of a given configura-
tion, whereby the distance relations in the given configuration are not rele-
vant for the dynamical structure. That is why relativity physics not only
introduces a new dynamical structure but also challenges the ontology as
given by the two mentioned axioms, because this ontology is built on
matter points that are individuated by the distance relations in a given con-
figuration and the change in these relations. In this chapter, we address this
challenge. We first consider classical electrodynamics, which is the first rel-
ativistic theory, arguing that fields are part of the dynamical structure of
this theory instead of new elements of the ontology and discussing an alter-
native theory of classical electrodynamics that works with direct particle
interactions (i.e. the Wheeler-Feynman theory) (section 2). We then argue
in general terms that the minimalist ontology pursued in this book
remains a cogent stance also when it comes to relativistic physics
(section 3) and show how Super-Humeanism combined with this minimalist
ontology is a valid option also for general relativistic physics (section 4).
Finally, we briefly go into the relationship between quantum entanglement
and relativity (section 5).
134 Relationalism for relativistic physics
5.2 Fields and relativistic laws in classical electrodynamics
Since the advent of classical electrodynamics, the field concept has con-
quered modern physics, which today is to a large extent field theory.
However, while fields have proven to be extremely successful as effective
devices, physics and philosophy run into an impasse when they accord an
ontological significance to these devices and buy into a dualistic ontology
of fields and particles, or an ontology that replaces particles with fields.
On closer examination, the concept of fields as mediators of particle inter-
actions turns out to be philosophically unsatisfying and physically problem-
atic, as it leads, in particular, to self-interaction singularities (see Lange
(2002) for a good introduction to the debate about fields in physics and phi-
losophy). Against this background, we argue that the true significance of
fields is that of bookkeeping variables, summarizing the effects of diachro-
nic relativistic interactions in order to obtain an efficient description of sub-
systems in terms of initial data (cf. Mundy (1989), p. 45). Indeed,
objections that match the ones against absolute space in Chapter 2,
section 1, apply to an ontological dualism of particles and fields as well:
fields also stretch out to infinity far beyond where any particles are
whose motion they could influence (see e.g. Feynman (1966), pp. 699–
700). More importantly, the ontological status of fields is unclear. Are
they properties—namely, properties of space-time points? Or are they
some sort of stuff filling space-time?
If fields are properties of space-time points, one falls back to the commit-
ment to an absolute space or space-time whose points instantiate the field
properties (this commitment is clearly brought out by Field (1980), p. 35,
and Field (1985), pp. 40–42). Over and above the drawbacks that the com-
mitment to an absolute space or space-time implies, geometrical properties
are bona fide properties of space-time, but it is rather odd to attribute in
addition to the geometrical properties also causal properties influencing
the motion of certain particles—namely, the charged ones—in the guise
of field properties to the space-time points. To put it differently, if proper-
ties of space-time points are causal in the sense that they influence the
motion of matter, then they should have an effect on the motion of all par-
ticles, not just the charged ones; after all, it is precisely the universality of
gravitation that motivates its geometrical account in general relativity
theory.
Moreover, if one conceives the field properties as dispositions, one faces
the following problem: the manifestation of the original dispositional prop-
erty of the particles—namely, their charge—which consists in the accelera-
tion of other particles, then is mediated by further dispositional properties
—namely, the field properties. Thus, a disposition (such as the charge of a
particle) produces in the first place a further disposition (the field proper-
ties), and both these dispositions then manifest themselves in the accelera-
tion of particles. Consequently, in any interaction mediated by a field,
 Relationalism for relativistic physics 135
there are two distinct dispositional properties that bring about the same
manifestation.
If one conceives fields as a stuff filling space-time, one faces the draw-
backs of the gunk view of matter discussed at the end of Chapter 2,
section 1—namely, the commitment to a bare substratum of matter that,
moreover, admits different degrees at different points of space-time as a
primitive matter of fact. In addition to that, in the case of the electromag-
netic field, there can be points or regions of space-time where the field value
is zero. Is there no field stuff in these regions? Or does the field stuff exist
everywhere and merely exerts no force on particles in these regions?
Whereas the property view of fields usually goes with a dualistic ontol-
ogy of particles and fields, the gunk view of fields is usually conceived as
replacing particles with fields. However, as the discussion of the GRWm
dynamics in Chapter 3, section 3, has shown, we do not have a convincing
dynamics at our disposal that explains the experimental evidence (which is
evidence of particle positions and motion) in terms of a field ontology, not
to mention the atomistic constitution of matter from elementary particles
via the chemical elements to molecules. This assessment holds indepen-
dently of the issue of a dynamics in terms of wave function collapse for
quantum mechanics: there simply is no clear-cut dynamics for a field ontol-
ogy to explain the experimental evidence; the GRWm dynamics is the most
advanced proposal in that respect. In sum, again, enriching the ontology—
with fields in this case—leads to new drawbacks instead of providing addi-
tional explanatory value.
Consequently, the electromagnetic field has a dynamical role, but not an
ontological status. That is to say: as we developed the Super-Humean strat-
egy for geometrical space, mass, charge, etc., in classical mechanics and the
wave function in quantum mechanics, so this strategy applies to fields as
they appear in classical electrodynamics. They are part of the dynamical
structure of the theory, being a means to describe the change in the particle
positions (i.e. the change in their relative distances) in a manner that
achieves the best combination of being simple and being informative
about that change.
The proposal of a pure particle ontology runs counter to the intuition
that the existence of at least certain forms of electromagnetic fields is
somehow obvious. After all, electromagnetic fields are obviously there
when we turn on the radio. Moreover, there seems to be a quite literal
sense in which all we actually see is light. And light can obviously be manip-
ulated: it can be reflected, refracted, polarized, absorbed . . . Against this
background, our claim that light does not exist, since the electromagnetic
field does not exist, may appear absurd. In fact, what we propose is to
quine light, to borrow an expression from Dennett (1988) coined for the
philosophy of mind (“Quining qualia”).
Our claim is that propositions using the concepts of light, radiation, elec-
tromagnetic signals, etc., are perfectly true, but their truth-maker are
136 Relationalism for relativistic physics
particle motions only. This, again, is the Super-Humean strategy spelled out
in Chapter 2, section 3. The “electromagnetic signals” that we pick up
when we turn on the radio thus refer to a particular kind of interaction
between our receiver and coherently oscillating particles at the broadcasting
station. Our impression of “red light” refers to a particular kind of interac-
tion between the observed object and our visual receptors. The “reflection
of light” refers to a series of interactions involving at least a source S, a
mirror M and a target T so that certain counterfactuals about the strength
of the effect on T depending on the presence/absence of M and the geometry
of the setup are true, etc. It is somewhat tedious, but quite straightforward
to spell out the details.
In any case, science pushes us in the direction of this eliminative path.
Our physical theories do not contain any concepts of redness or blueness
or greenness. To the contrary, sense impressions involving the perception
of colour are related to certain wave-lengths in the electromagnetic field.
To eliminate the electromagnetic field as well in favour of an ontology of
direct particle interactions then merely means to cut out the middle-man:
the frequencies that we usually identify with red light or blue light or
green light are taken to refer directly to accelerations of particles. Adopting
the terminology of Sellars (1962), we can say the following: the manifest
image and the scientific image of the world connect at other points than
those ones proposed in the field theory, but this does not make the manifest
image less accurate or the theoretical account less compelling.
Note that we are not concerned here with the phenomenology of light
and colour sensation (that is, we are not concerned with “quining
qualia”). Our point only is that the field ontology and the sparse particle
ontology of the natural world are in the same boat when it comes to the
link between the scientific image of the world and our sense impressions.
Field ontology is not in a better position to establish that link than a
sparse particle ontology, but, to the contrary, faces the mentioned draw-
backs. Generally speaking, enriching the ontology of the natural world
with entities that are not needed for a sparse and empirically adequate
ontology does not help when it comes to the link between the natural
world and the phenomenology of mental states, but only entails problems
such as the mentioned ones in the case of an ontological commitment to
fields.
One can also formulate gravitation in Newtonian mechanics in terms of
a gravitational field. But in this case, it is obvious that the gravitational field
(or potential) is just a useful mathematical tool to represent the direct par-
ticle interactions rather than a candidate for ontology. The reason is that
the degrees of freedom contained in the gravitational field at any time t
can be immediately reduced to the particle configuration at the same
time. The electromagnetic field at time t, by contrast, depends on the trajec-
tories of the particles in the past (and possibly also the future) of t, because
the sources affect the field in a retarded (and possibly also an advanced)
 Relationalism for relativistic physics 137
way, reflecting the light cone structure of relativistic space-time. This fact is
no reason to elevate the electromagnetic field to a different ontological
status than the gravitational field in Newtonian mechanics. But it explains
why the field is harder to dispense with in the relativistic case: it is a book-
keeper of past events—and possibly also future events—rather than a
summary of co-present events.
The retardation furthermore explains the appearance that electromag-
netic effects propagate at finite speed as if they had sit venia verbo a life
of their own. After all, when we look up to the night-sky, we see many
stars that have long ceased to exist. So, one might say, whatever interacts
with us to create the impression of the star is not the star itself (which
exploded millions of years ago) but something else—namely, electromag-
netic radiation, which had been emitted from the star and then traveled
all the way to earth. However, while it is a natural way of speaking to
say that electromagnetic effects “travel” or “propagate”, the ontological
conclusion that there must thus be something, in addition to the particles,
that actually propagates through time and space is unwarranted. The regu-
larities in a world may be such that the present motion of subsystems of the
universe is correlated with the present motion of distant subsystems (as in
Newtonian mechanics), or that it is correlated with the past (and possibly
future) motion of other subsystems intersecting the past (and possibly
future) light cones of the subsystems (as in relativistic theories). This
depends on what the laws turn out to be that provide for the best system,
achieving the best balance between being simple and being informative
about the actual motion conducted by the systems in the universe. But all
this supervenes on the Humean mosaic of particle distances and their
change only.
Another important difference between the electromagnetic field and the
gravitational field in Newtonian mechanics is that the Maxwell equations
allow for a variety of non-trivial and non-equivalent vacuum solutions—
that is, solutions that do not have charged particles as field sources.
These so-called free fields thus cannot be expressed in terms of the
charge-trajectories; they do however affect the electromagnetic forces and
hence the particle motions. The existence of free fields is sometimes
voiced as an objection against the viability of a pure particle ontology for
classical electrodynamics. However, to emphasize again, the Super-
Humean strategy is not committed to mathematically reducing all dynami-
cal structures to the primitive variables. Rather, dynamical structures such
as free fields can enter as part of the best system description of the Humean
mosaic, while the latter consists only of particle distances and their change.
That said, the existence of free fields can hardly be maintained on empir-
ical grounds. Free fields are certainly necessary for providing an efficient
description of subsystems. When we set out to describe the dynamics of
N charges in a space-time region M, we have to account for the influence
of charges outside of M by specifying appropriate boundary conditions
138 Relationalism for relativistic physics
for the electromagnetic field (the “incoming radiation”). These boundary
conditions will in general contain free fields, or rather external fields, in
the sense of fields that have no sources in M (because they have sources
outside of M). However, on the fundamental level—that is, when M is
the whole of space-time—the assumption of genuinely free fields coming
in “from infinity” seems unwarranted. Empirically, we will never be able
to determine that some observed radiation is truly source-free. In fact,
good scientific practice is to assume that it is not and look for—or simply
infer—the existence of material sources. Empirical evidence is fully consis-
tent with the assumption that the model of classical electrodynamics that
best fits our universe is one in which all field degrees of freedom can be ulti-
mately reduced to the history of charged matter. However, the models of
classical electrodynamics (without self-interaction) in which all field
degrees of freedom can be reduced to the history of matter are also
models of a corresponding direct interaction theory. On the fundamental
level, it is thus unnecessary and unwarranted to buy into free fields as a
physical possibility.
Whereas the philosophical problems that an ontological commitment to
fields entails can be removed by considering fields as parts of the dynamical
structure instead of the ontology of classical electrodynamics, there are phys-
ical problems in the Maxwell-Lorentz theory that cannot be solved in that
way, requiring in fact new physics. The situation hence is different from
the one in classical mechanics: Leibnizian relationalism can be vindicated
as an ontology for Newtonian mechanics, which is a perfectly consistent
physical theory, by adopting Super-Humeanism with respect to the geomet-
rical and dynamical structure of Newtonian mechanics. One can also develop
a relationalist alternative physical theory (as Barbour does), but this is by no
means mandatory; in any case, also an alternative physical theory requires
more conceptual means than the ones provided by the sparse ontology of
Leibnizian relationalism. In the Maxwell-Lorentz theory of electromagne-
tism, to the contrary, there is a physical problem—namely the one of
self-interactions—which constitutes a physical motivation to search for an
alternative physical theory, independently of the problems that come with
an ontological commitment to fields.
The Maxwell-Lorentz theory of classical electrodynamics—as the name
already suggests—consists of two parts. The Maxwell equations describe
the evolution of the electromagnetic field and the coupling of the field to
charges and currents. The Lorentz force equation describes the motion of
a charged particle in the presence of an electromagnetic field:

€m ¼ e Fmn x_ n :
mx ð5:1Þ

Here, Fμν is the field tensor, comprising electric and magnetic components,
m denotes the mass and e the charge of the particle with space-time trajec-
tory xμ(τ) and a dot denotes a derivative with respect to eigentime τ.
 Relationalism for relativistic physics 139
The Maxwell field equations can be separated again into homogeneous
and inhomogeneous equations, where the first involve only field degrees
of freedom. The homogenous equations tell us that the anti-symmetric
field tensor Fμν (a two-form) can be written as the exterior derivative of a
potential (a one-form) Aμ—that is, as
Fmn ¼ @ m An  @ n Am : ð5:2Þ

The inhomogeneous Maxwell equations couple the field degrees of freedom

to matter—that is, they tell us how charges influence the electromagnetic
field. Fixing the gauge-freedom contained in (5.2) by demanding @ μAμ(x)
= 0 (Lorentz gauge), the remaining Maxwell equations take the particularly
simple form:
□Am ¼ 4pjm ; ð5:3Þ

with □ = @ μ@ μ the d’Alembert operator and jμ the four-current density,

which for N point charges is:
XN XN Z
j ðxÞ ¼
m
ji ¼
m
ei d4 ðx  zi ðti ÞÞz_ mi ðti Þ dti : ð5:4Þ
i¼1 i¼1

Now, given the trajectories zi(τi)i = 1, . . ., N of the particles, the solutions of

(5.3) are well known. By the linearity of equation (5.3), we can sum the
contribution of each particle. A special solution is given by the advanced
and retarded Liénard-Wiechert potentials:
z_ mi ðti Þ
Ami; ðxÞ ¼ ei ; ð5:5Þ
ðxn  zni ðti ÞÞz_ i;n ðti Þ

where x denotes a space-time point and tþi ðxÞ and ti ðxÞ are the advanced
and retarded times given as implicit solutions of
ðxm  zmi ðtÞÞðxm  zi;m ðtÞÞ ¼ 0: ð5:6Þ

This means that the field equation connects events with zero Minkowski
distance. In other words, it means that the advanced / retarded Liénard-
Wiechert field at x depends on the charge-trajectories at their points of
intersection with the future, respectively the past light cone of x.
To any such solution, we can further add any solution of the free wave
equation
□Am ¼ 0; ð5:7Þ

corresponding to the aforementioned free fields.

The problem with classical electrodynamics is that a self-consistent
description of an N-particle system requires us to solve (5.3) and (5.1)
140 Relationalism for relativistic physics
together. But this set of coupled differential equations is ill-defined. The
Liénard-Wiechert field (5.5) is singular precisely at the points where it
has to be evaluated in (5.1)—namely, on the worldlines of the particles.
This is the notorious problem of the electron self-interaction: a charged par-
ticle generates a field, the field acts back on the particle, and since the field-
strength is infinite at the position of the particle, the interaction blows up.
As mentioned at the beginning of Chapter 4, this problem carries on to QFT
in the guise of the ultraviolet divergences. Note that already the 1-body
problem is ill-defined in the Maxwell-Lorentz theory. It is not the interac-
tion between particles, but the duality of particle and field that leads to sin-
gularities. The reason why classical electrodynamics still works so well for
most practical purposes is that physicists, in general, solve either the
Maxwell equations for a given charge-distribution or the Lorentz equation
for a given electromagnetic field, but not both together. Nonetheless,
strictly speaking, the Maxwell-Lorentz field theory is an inconsistent
theory.
One prominent way to cope with the self-interaction problem is imple-
mented in the Lorentz-Dirac theory. The equation that is believed to
describe all radiative phenomena correctly is the Lorentz-Dirac equation
2
xm ¼ eFmn x_ n þ e2 ðx_ m x
m€ €n x
€n  x⃛m Þ; ð5:8Þ
3

where Fμν does not contain the self-field. The four-vector

2
Gm :¼ e2 ðx_ m x
€n x
€n  x⃛m Þ ð5:9Þ
3

is known as the Schott-term and includes, in particular, the radiation fric-

tion. Note that this equation of motion depends on the third derivative of
x(τ)—that is, the time-derivative of the acceleration.
In his seminal paper on the classical electron theory, Dirac showed that
(5.8) can be derived from Maxwell’s equations together with the principle
of energy-momentum conservation; analogous results can also be obtained
by considering the point-particle limit for a spherical charge-distribution
(Dirac (1938); see Rohrlich (1997) for a good historical overview). These
derivations, however, have to rely on a highly dubious mass renormaliza-
tion procedure: an infinite, negative bare mass for the point-particle has
to be assumed in order to cancel a diverging inertial term arising from
the infinite self-energy. Moreover, while these derivations suggest that the
self-interaction is the origin of the radiation reaction, this is not a consistent
interpretation of the Lorentz-Dirac theory. The right-hand side of (5.8) is
divergence-free at the position of the particle and hence does not contain
any self-field according to Maxwell’s equations (cf. Wheeler and
Feynman (1945)). Nevertheless, in the end, we do not have to take the deri-
vation seriously in order to accept the result. If equation (5.8) proves to be
 Relationalism for relativistic physics 141
meaningful and empirically adequate, we may accept it as the correct law of
motion for the classical electron regardless of its logical relationship to the
original Maxwell-Lorentz theory.
Unfortunately, also the Lorentz-Dirac theory faces severe issues: (i)
except for very fine-tuned initial conditions, the infamous triple-dot
term in equation (5.8) leads to runaway solutions that diverge in finite
time due to particles (self-)accelerating to the speed of light (see Spohn
(2000) and the discussion in Frisch (2005)). In this sense, the infinite
self-energy—respectively, the negative bare mass introduced to cancel
it—still manifests as the source of dynamical instabilities. (ii) It is essential
that the self-field of the particle does not appear on the right-hand side of
(5.8). Consequently, the Lorentz-Dirac theory involves not just one elec-
tromagnetic field, but a different type of electromagnetic field for each
particle in the universe. The field created by particle j must carry some
property that makes it interact with all the other charges, but not with j
itself. This leads to an enormous increase in mathematical complexity of
the N-particle problem (we have to specify N initial fields and track the
evolution of each one separately) as well as to a grotesque inflation of
the physical ontology.
This situation motivates the search for a formulation of the dynamical
structure of classical electrodynamics without the electromagnetic field.
This idea goes back to Gauss (1867) and was further explored by Schwarzs-
child (1903), Tetrode (1922) and Fokker (1929). Wheeler and Feynman
(1945) showed in their absorber theory that a time-symmetric direct inter-
action formulation of classical electrodynamics is able to account for all
radiative phenomena. Henceforth, the respective direct interaction theory
is commonly known as Wheeler-Feynman electrodynamics. This theory
involves no self-interactions from the beginning. The Lorentz-Dirac equa-
tion including, in particular, the radiation friction can be derived from a
statistical analysis as a phenomenological description. Moreover, there
are good indications that the Wheeler-Feynman theory is free of runaway
solutions (Bauer (1997)), which is to say that the solutions of the
Lorentz-Dirac theory that are also approximate solutions of Wheeler-
Feynman are automatically the good ones that do not lead to runaway sin-
gularities. In a nutshell, the Wheeler-Feynman theory captures precisely the
physical content of the field theory while avoiding the unphysical artifacts.
A retarded direct interaction theory, such as the one proposed by Ritz
(1908), cannot explain the phenomenon of radiation damping—that is,
the fact that accelerated charges lose energy-momentum (unless one
admits an ad hoc modification of the equations of motion like Mundy
(1989)). In the absence of self-fields, this radiative reaction can only come
from interactions with other charges, so that the damping effect would be
considerably delayed in a purely retarded theory. This leaves us with the
time-symmetric direct interaction theory proposed by Fokker (1929) and
Wheeler and Feynman (1945). This theory is in a better position to
142 Relationalism for relativistic physics
explain the radiation damping, because advanced reactions to retarded
actions arrive instantaneously.
Another important virtue of the time-symmetric theory is that it can be
defined by a principle of least action for what is arguably the simplest rel-
ativistic action for interacting particles:
" Z qffiffiffiffiffiffiffiffiffiffi Z Z #
X 1X  2 m
S¼ mi z_ i z_ i;m dli 
m
ee d ðzi  zj Þ z_ i z_ j;m dli dlj : ð5:10Þ
i
2 i6¼j i j

The equations of motion for an N particle system then read as follows:

X 1 
€mk ¼
mk x ek ej ðjÞ Fmn
ret þ
ðjÞ mn
Fadv x_ k;n ; ð5:11Þ
j6¼k
2

where Fmn mn
ret and F adv correspond to the retarded and advanced Liénard-Wie-
chert fields, respectively. These equations contain no self-interaction, hence
no singularities, and no mysterious free fields. The forces acting on a parti-
cle k are completely determined by the trajectories of other particles. This
Wheeler-Feynman theory of classical electrodynamics faces notably two
challenges—namely, to account for (i) the phenomenon of radiation
damping and (ii) for the radiative arrow of time—that is, the fact that we
observe only retarded radiation.
Wheeler and Feynman (1945) present a statistical account of the radia-
tive reaction. They assume that an accelerated charge interacts with a
large, homogeneous, spherically symmetric charge-distribution in the
future. In the absence of external disturbances, the net force exerted by
this so-called absorber, surrounding our local, low-entropy environment,
is assumed to be approximately zero. Then Wheeler and Feynman show
in a series of three computations of increasing generality that, if the
absorber particles are disturbed by the retarded forces from the accelerated
charge, their advanced reaction will correspond, in form and magnitude, to

1
ðF  Fadv Þ: ð5:12Þ
2 ret

A test-particle in the vicinity of the accelerated charge will thus experience a

net effect of

1 1
ðFret þ Fadv Þ þ ðFret  Fadv Þ ¼ Fret ; ð5:13Þ
2 2

as if the accelerated charge produced a purely retarded force. Moreover, at

the location of the charge, the difference 12 ðFret  Fadv Þ corresponds precisely
to the radiation reaction term (5.9), as was already shown by Dirac. In
 Relationalism for relativistic physics 143
particular, the accelerated particle will thus experience a damping force as a
result of its interaction with the absorber.
Apart from the justification of the temporal asymmetry thus introduced,
the Wheeler-Feynman analysis of radiation reaction could have—and
perhaps should have—ended here. The fact that it did not exposed their
work to a lot of unnecessary criticism. The absorber response (5.12) is,
notably, independent of any detailed properties of the absorber like mass
or charge or the exact arrangement of its constituting particles. For this
reason, Wheeler and Feynman—having already derived the absorber
response in a series of hands-on computations—suggest that it should be
possible to conclude the same result from first principles. Hence, they go
on to present a remarkably simple and elegant argument based on the so-
called absorber condition:
X ðkÞ ðkÞ

Fret  Fadv ¼ 0; ð5:14Þ
k

where the sum goes over all particles in the universe. This amounts to the
assumption that the distribution of charges in the universe form a complete
P ðkÞ P ðkÞ
absorber, so that Fret ðxÞ ¼ 0 and Fadv ðxÞ ¼ 0 hold separately, at any
k k
space-time point x.
One can imagine universes in which this assumption is valid to a good
approximation, such as a compact spatial geometry with a homogeneous
distribution of charges. But it is easy to point the finger at equation
(5.14) and ask: “Why should we believe in that?”. Many commentators
have thus rejected the Wheeler-Feynman account, suggesting that the deri-
vation of the radiative reaction rests on the validity of equation (5.14) (see,
e.g., Zeh (2010), pp. 36–37, and Earman (2011), p. 368). This suggestion,
however, is dubious, because—as shown by Wheeler’s and Feynman’s
(1945) series of three computations and as further emphasized in Bauer
et al. (2014)—the radiative reaction follows from statistical assumptions
alone that are much weaker and much more robust than the infamous
absorber condition.
While the calculations of Wheeler and Feynman seem sound and reason-
able, one worry that immediately arises is that the same arguments apply
also in the opposite time direction. If we assumed that the accelerated
charge interacts (by advanced forces) with an absorber in the distant
past, the retarded absorber response would correspond to 12 ðFadv  Fret Þ,
resulting in a net force of Fadv on surrounding particles and an anti-
damping force on the accelerated charge.
Since radiation damping is a many-particle phenomenon in Wheeler-
Feynman electrodynamics, it makes sense to look for a thermodynamic
explanation of the asymmetry. Unfortunately, for reasons explained
next, the state space of the theory is not yet understood well enough to
spell out such an account in detail. The general argument proposed by
144 Relationalism for relativistic physics
Wheeler and Feynman (1945), which seeks to rule out a retarded response
from the past absorber on probabilistic grounds, is generally not consid-
ered to be successful (see Arntzenius (1994), pp. 40–41, Price (1996),
pp. 65–73, and Frisch (2005), ch. 6, for a critique, as well as Bauer et
al. (2014) for a more positive assessment). However, while we do not
yet have a conclusive account of the radiative asymmetry in Wheeler-
Feynman electrodynamics, there are clear indications how the retrocausal
effects manifested on the microscopic level can be reconciled with our
macroscopic experience and the perceived arrow of radiation. This puts
the Wheeler-Feynman theory in a position that is not worse than that of
the field theory, where the explanation of the radiative asymmetry is
subject to debate as well (for the current state of that debate see, e.g.,
Earman (2011)). Moreover, even if the justification of the asymmetry is
an open question, the statistical derivation of the radiation reaction is
already a significant success that the field theory cannot match, since
there, one has to rely on a highly unphysical mass renormalization proce-
dure in order to obtain the analogous result.
Arguably, the real reason why the Wheeler-Feynman theory is not appre-
ciated by working physicists is this one: its equation of motion is not the
kind of ordinary differential equation that physicists and mathematicians
are trained to solve, but a so-called delay differential equation. As we can
see from the action functional (5.10) (or, alternatively, from (5.11) together
with (5.5)), the force acting on a particle at some space-time point x
depends on the trajectory of the other particles at their points of intersection
with the past and future light cones originating in x. The force is hence not
determined by an instantaneous state of the system, where “instantaneous
state” means the configuration of the physical system on a suitable space-
like hypersurface that includes x.
As a result, the Wheeler-Feynman laws of motion cannot be posed as
initial value problems. As of today, it is not known what kind of initial
or boundary conditions one has to impose in order to ensure existence
and uniqueness of solutions. The best conjecture is that initial data for a
Wheeler-Feynman system comprise entire segments of trajectories, rather
than instantaneous Cauchy data on a space-like hypersurface (see Bauer
et al. (2013), Deckert et al. (2014), Deckert and Hinrichs (2016) and
Bauer et al. (2016) for the current state of the solution theory). From this
point of view, it is precisely the role of electromagnetic fields to introduce
additional degrees of freedom that enable the formulation of relativistic
laws as initial value problems. Indeed, there is no cogent reason to expect
Cauchy data in a relativistic theory. Relativistic space-time, in contrast to
Newtonian space-time, does not come equipped with a foliation into
instantaneous spatial geometries. The dynamical relations in relativistic the-
ories are typically spatio-temporal ones. Hence, we have no grounds to
expect that the dynamical state of the universe can be completely described
by physical data on a space-like hypersurface.
 Relationalism for relativistic physics 145
The field theory allows us to trade a diachronic, spatio-temporal descrip-
tion in terms of particle histories for a synchronic description in terms of an
infinite number of field degrees of freedom that encode the history of
charge-trajectories in their spatial dependencies. One can therefore say
that the central significance of the fields, their raison d’être, is to serve as
bookkeepers for the particle histories in order to save the concept of instan-
taneous dynamical states and the Newtonian paradigm of laws as initial
value problems in a relativistic setting. Ultimately, though, this remains
unsuccessful, since the initial value problem is not well-posed in the
Maxwell-Lorentz theory either, even if self-interactions are neglected (see
Deckert and Hartenstein (2016)). The reason is, simply put, that the
initial fields specified on a spacelike hypersurface would have to include,
in particular, the fields created by the charges in the past. However, we
do not know what these fields are, unless we compute them from the
charge trajectories in the first place. As Deckert and Hartenstein (2016)
show, a generic choice of initial data—compatible with the Maxwell con-
straints—will lead to singularities in the field solutions which, in turn,
produce singularities or discontinuities in the charge trajectories. Hence,
even in the case of the field theory, the best way to obtain physical solutions
of the fundamental laws is to abandon the initial value formulation and
solve a delay differential equation in terms of the particle trajectories, of
which the Wheeler-Feynman law (5.11) is a special case.
For practical purposes, the Wheeler-Feynman theory also allows us to
introduce fields as bookkeeping variables in the sense discussed earlier.
For a given distribution of charges, the effective description of a subsystem
then corresponds to a boundary value problem within the usual Maxwell-
Lorentz theory (respectively the Lorentz-Dirac theory, if the radiation fric-
tion is taken into account). The Wheeler-Feynman theory can therefore
explain the success of textbook electrodynamics. On a fundamental level,
the theory reminds us however that in relativistic physics, we have to
expect the dynamics to depend on past and future particle motion rather
than an instantaneous dynamical state of the universe.
Consequently, since the Wheeler-Feynman laws of motion draw on the
geometric structure of relativistic space-time, with particles interacting
along light cones, this raises the question how such a theory can be compat-
ible with an ontology of distance relations that denies the fundamentality of
both relativistic space-time and spatio-temporal relations. If relativistic
laws of the Wheeler-Feynman type arise as part of the Humean best
system, this means in the first place that the summary of the Humean
mosaic that strikes the optimal balance between simplicity and strength is
one in which the change of the particle configuration at one time depends
on the particle configuration at former and later times (with “time” here
referring only to the order of change in the configurations). This depen-
dence is, however, only descriptive and does not entail or presuppose any
real physical connection between particle configurations at different times.
146 Relationalism for relativistic physics
Furthermore, the fact that relativistic laws of the Wheeler-Feynman type
connect precise particle events with zero Minkowski distance means that
there exists an embedding of the distance relations into a three-dimensional
space, respectively a four-dimensional space-time, and a parameterization
of the change of the distance relations (providing a temporal metric) such
that the best system description relates precisely those particle events—
that is, those vertices of the embedded network of relations for which the
ratio of spatial distance and temporal distance is constant (that constant
being the speed of light). It is plausible that a geometric representation
for which this ratio is constant for dynamically related events would turn
out to be more simple (and at least equally informative) as a representation
in which it is not.
Hence, there is an embedding of the history of distance relations into a
four-dimensional space-time geometry such that the best system laws
connect those and only those particle events that are, in the usual nomen-
clature, light-like separated. Such laws will then naturally turn out to be
invariant under symmetries that leave light-like connections—that is, the
light cones—invariant. These symmetries are, of course, exactly those
described by the Lorentz group. The space-time substantivalist will
usually claim to provide an explanation of such symmetries in terms of
the underlying space-time manifold. However, as mentioned already in
Chapter 2, section 1, this does not amount to any deeper explanation,
because the substantivalist, in turn, has to accept other structures and enti-
ties such as space-time points and the Minkowski metric as primitive.
In any case, one important consequence of the Lorentz symmetries is that,
given a complete solution of the laws, there are an infinite number of equiv-
alent possibilities—corresponding to different Lorentz frames—to slice up
the space-time trajectories into a history of spatial relations. Any such folia-
tion of four-dimensional space-time into three-dimensional geometries will
identify different instantaneous configurations with different spatial distances
between the matter points. However, basing oneself on an ontology of dis-
tance relations, at most one of these foliations can accurately represent the
real state of affairs, that is, identify the correct co-present configurations
and a family of three metrics such that the spatial distances between
matter points correspond to the fundamental distance relation. The best
system laws do not identify the “correct” frame of reference.

5.3 Relativistic dynamical structure and minimalist ontology

When it comes to the relationship between on the one hand ontology and
on the other hand geometry and dynamical structure, the situation in rela-
tivistic physics is the same as in Newtonian mechanics: simplicity in ontol-
ogy and simplicity in representation pull in opposite directions. If one starts
from the dynamical structure of Newtonian mechanics, being the paradigm
 Relationalism for relativistic physics 147
example of a simple and universal theory, and uses the dynamical structure
as guideline to the ontology, one gets to an ontology of point particles, pos-
sessing mass and moving in an absolute space and an absolute time, because
inertial motion in Newtonian mechanics is defined as motion with respect
to an absolute space in an absolute time. However, as Leibniz, Mach and
others point out, this ontology faces serious drawbacks, notably the three
ones discussed in Chapter 2, section 1 (differences that do not make a dif-
ference, infinite expansion and problem of what characterizes matter as
filling space). Given that all the empirical evidence consists in relative par-
ticle positions and change of these positions, the relationalist ontology as
stated in the two axioms on which this book is based is the more simple
ontology that is empirically adequate and that avoids these drawbacks,
which originate in the surplus structure that comes with endorsing an abso-
lute space and time.
In relativistic physics, the situation is similar. Starting from a relativistic
dynamical structure as given by the Maxwell-Lorentz theory of classical
electrodynamics or the Wheeler-Feynman theory and implemented in the
geometry of the special theory of relativity, there are now not only two,
but three main candidates for an ontology that clarifies the relationship
between matter and space-time.
(i) In the first place, there is the option corresponding to the Newtonian
one. There are point particles moving in an absolute space, with each par-
ticle being individuated by its position in absolute space. Consequently, dis-
tance relations to other particles in a given configuration are not relevant
for the individuation of the particles. These relations derive from the geom-
etry of the underlying space. The particles are unified by being situated in
one absolute space. That space is constituted by the topological, affine
and metrical connections among its points. That ontology hence is the
same as in Newtonian mechanics. The availability of this ontology high-
lights again the importance to distinguish between what individuates the
particles and what concerns the dynamical structure that captures their
motion. On this ontology, the individuation is carried out by the particle
position in absolute space so that simultaneity plays no role for the individ-
uation of the particles. Hence, absolute space has a crucial function for the
particle ontology, whereas the nature of time is an issue of the dynamical
structure.
As regards the dynamical structure, in the relativistic case by contrast to
the Newtonian one, the change of the position of a given particle is not
fixed by the simultaneous positions and velocities of the other particles in
a given configuration (e.g. via a law of gravitation based on attributing a
mass and an initial velocity to the particles). In a field theoretical setting,
the change of the position of a given particle is determined by the field con-
figuration in its immediate past light cone (via a field law based on attrib-
uting a charge to the particle in addition to its initial velocity and inertial
mass, with that charge reacting to the field configuration in its immediate
148 Relationalism for relativistic physics
past, and that field configuration, in turn, being fixed in the last resort by
other charged particles in the past light cone—leaving aside free fields, as
argued in the preceding section). In a setting of direct particle interaction
as in the Wheeler-Feynman theory, the change of the position of a given
particle is determined by the past and future motions of other particles—
namely, the motions of the other particles in its past and future light
cones (via a law of electrodynamics based on attributing a charge over
and above an inertial mass and an initial velocity to the particles, but no
field configuration). In this latter case, there is empty space between the par-
ticles as in Newtonian mechanics, whereas in the former case, space is rep-
resented as being filled with fields. Consequently, in both cases, there is a
unique and directed temporal order of the evolution of each particle
taken individually, but there is no absolute time, because there is no
unique temporal order of the motion of all the particles taken together,
although each particle moves with respect to absolute space. Hence, in
this setting, whether or not there is absolute time is an issue of the dynami-
cal structure, by contrast to the question of whether or not there is absolute
space.
The drawbacks of this position are the same as the ones of the Newto-
nian ontology—namely, the commitment to the surplus structure of an
absolute space expanding to infinity and, consequently, the impasse into
which the subsequent question leads regarding what the particles are,
filling space. More precisely, against the background of what we have dis-
cussed in Chapter 2, section 1, this impasse amounts to this option being
committed to regarding the particles, occupying the points of absolute
space, as bare substrata. Hence, this ontology does not meet the require-
ment of simplicity. It is an illustration of how bringing in more than dis-
tance relations individuating matter points creates drawbacks instead of
providing additional explanatory value.
(ii) Furthermore, there is a genuine relativistic option for the ontology.
This option consists in replacing the distance relation with a distance-tem-
poral relation—that is, a spatio-temporal relation. Instead of matter points
that are each separated by a distance from all the other ones, with that dis-
tance individuating the matter points, there are continuous lines (i.e. worl-
dlines) with the points on each line being distinguished from the points on
each other line by a spatio-temporal interval, which is usually represented
in terms of a four-dimensional geometry. All these spatio-temporal intervals
exist at once.
This ontology can be cast in terms of a space-time relationalism. No
commitment to a space-time that exists in distinction from matter, with
matter filling space-time, is called for. In this case, the objections against
an absolute space—that is, the objections against the first option—do not
apply. By contrast to the relationalist ontology pursued in this book,
according to this ontology, instead of matter points being substances that
are individuated by distance relations, with these relations changing,
 Relationalism for relativistic physics 149
there are continuous sequences of matter point events, forming continuous
lines (worldlines) that are individuated by the spatio-temporal intervals
between them. These intervals do not change. They all exist at once.
That is why this ontology is known as block universe: all the events
throughout the history of the universe exist at once. Again, we stress that
the block universe does not have to be conceived in terms of a four-
dimensional space-time existing and being filled with matter. It can also
be set out in terms of all the spatio-temporal relations between matter
point events existing at once.
The main challenge for this ontology is to distinguish between variation
within a given configuration and change of that configuration. If one
replaces point particles with point events forming continuous lines and a
three-geometry with a four-geometry to represent the relations between
them, one obtains variation, but no change: the relations represented in
terms of a four-geometry provide for variation within the block universe.
However, since these relations exist all at once, there is no change. Only
when we cut three-dimensional slices through the block and compare
them, we can define change in terms of the differences between such
slices. But this change concerns only an—arbitrary—description and not
the ontology, since the ontology is the four-dimensional block (Geach
(1965), in particular, p. 323, is the locus classicus of this objection).
To include change, one, therefore, has to stipulate the following over and
above the global geometrical order characterizing the block universe (what-
ever the number of dimensions may be in which that geometry is formu-
lated): the points on each worldline are ordered according to earlier and
later, with that order being unique and directed. Then there is a local
time for each object (worldline), but no global time. Comparing different
points on each worldline, one can introduce change in terms of different
spatio-temporal intervals between different points on two worldlines. In
this way, the spatio-temporal relations provide for variation and change
at once against the background of the points on each worldline being
ordered according to earlier and later.
However, in this case, temporal order has to be presupposed as a
primitive—that is, the ordering of the points on a worldline according
to earlier and later. That ordering obtains as a primitive matter of fact, inde-
pendently of whether or not there is change in the spatio-temporal intervals
between these two lines. Hence, temporal order is not derived from
change, but change is derived from the order of the points on each worldline
being temporal as a primitive matter of fact. What is the difference between
ordering the points on such a line according to, say, below and above
and ordering them according to earlier and later? That difference is
primitive. Consequently, one cannot do without endorsing a primitive tempo-
ral, metrical requirement on the block universe view, even if one casts the
block universe view in relationalist terms instead of endorsing
an absolute space-time: well-defined spatio-temporal intervals between
150 Relationalism for relativistic physics
non-simultaneous point events have to be presupposed as primitive. By con-
trast, the ontology of matter point substances individuated by distance rela-
tions does not rely on any primitive metrical requirement regarding the
change of the distance relations—in fact, there is no primitive metrical
requirement in this ontology apart from the triangle inequality figuring in
the definition of the distance relation (see definition 1 in Chapter 2, section 1).
The question then is whether endorsing a primitive temporal order of the
points on each worldline is sufficient for empirical adequacy. There is no
possibility to introduce becoming and the passage of time: there is a local
temporal order of the points on each world line, but since these points
exist all at once, there is no such thing as local becoming (see by contrast
Dieks (2006)). The points on any such worldline are ordered according
to earlier and later as a primitive matter of fact, but they all exist at
once. Becoming and the passage of time then are merely occurring in our
experience of the world. As Weyl (1949, p. 116) famously put it, “The
objective world simply is, it does not happen”. Consequently, it is doubtful
whether this is a conception of change that stands up to scrutiny. In other
words, it is doubtful whether paying the price of a primitive temporal order
is worthwhile, since doing so arguably does not give us the phenomenology
of time. These brief considerations are not intended to be conclusive (see
Skow (2015) for a recent defense of the block universe view). Their
purpose is to underline again that the ontology of the natural world is a
matter of philosophical argument and cannot be settled by referring to
the mathematical structure of a physical theory.
(iii) Let us therefore consider how the ontology pursued in the first three
chapters of this book fares when it comes to relativistic physics. As stated in
the first axiom, there are matter points individuated by the distance rela-
tions among them. This ontology thus implements a primitive spatiality
requirement in the guise of distances between points. Furthermore, as
stated in the second axiom, these distances change, while the matter
points are permanent. Since a change in any one distance relation implies
a change in all of them (the whole configuration), the order of that
change is unique and exhibits a direction. However, there is no primitive
temporal requirement. Any measure of that change requires picking out a
sub-configuration of distance relations relative to which change is
measured.
This hence is an important advantage of this conception: it does not pre-
suppose any primitive temporal requirement. More precisely, it trades a
primitive spatiality requirement (distances) with primitive change of these
distances for a primitive temporal requirement. The physical theories of rel-
ativity then simply bring out clearly the consequences of this move that are
already there for any physical theory, from classical mechanics on: it is
crucial for this position to distinguish between the ontology on the one
hand and the geometry and dynamics on the other. The latter come in as
a package to represent the change in the distance relations in a manner
 Relationalism for relativistic physics 151
that is most simple and informative. In classical mechanics of gravitation,
the geometry of Euclidean space and force laws built on attributing param-
eters to the matter points such as inertial and gravitational mass constitute
that package. In physical theories of relativity from classical electrodynam-
ics on, the geometry of (flat or curved) four-dimensional space-time and
field laws or laws formulated in terms of direct retarded and advanced par-
ticle interaction, both based on attributing parameters to the matter points
such as inertial mass and charge, constitute that package. As there is no
three-dimensional Euclidean space out there in the world and the matter
points do not have properties such as mass or charge per se, so there is
no (flat or curved) four-dimensional space-time out there in the world.
The package of geometry and dynamics is only our means to represent
the variation and the change in the world in the most simple and informa-
tive manner. Of course, we have access to the world only through our
means of representation. However, it is inappropriate to read off the ontol-
ogy from the means of representation. Doing so ignores their function—
namely, to yield a simple and informative representation and to employ
whatever variables are suitable for this purpose. In other words, the justifi-
cation for these variables (geometrical as well as dynamical ones) is that
they are a convenient means to achieve this aim, not that they are what
there is out there in the world.
The standard for the ontology is simplicity together with empirical ade-
quacy. Thus, the argument for the ontology of matter points individuated
by distance relations that change is its simplicity together with its empirical
adequacy. The argument for its simplicity in the context of comparing this
ontology to the four-dimensional, spatio-temporal ontology of a block uni-
verse (which implements not only a relativistic dynamics but also a relativ-
istic ontology) is that this ontology does not rely on any primitive temporal
requirement. The argument for its empirical adequacy in comparison to
that ontology is that it has no problem of accounting for the phenomenol-
ogy of time (becoming, passage of time), because it endorses change in the
fundamental relations as a primitive. Admitting change as primitive—but
doing so without endorsing a primitive metrical requirement for that
change—just is what enables this ontology to accommodate temporal
becoming and the passage of time. In a nutshell, this ontology achieves
the best of two worlds: the Parmenedian world of eternal being and the
Heraclitean world of change. The matter points are substances because
they are permanent; they do not come into being, and they do not go out
of being. But they are individuated by relations that change as a primitive
matter of fact.
If one puts this ontology into the framework of terms that are common
in contemporary analytic metaphysics, the substances—the matter points—
are endurants because they persist without having temporal parts (but they
do not have spatial parts either; they do not have any parts at all). Further-
more, since what exists is a spatial configuration of these substances
152 Relationalism for relativistic physics
constituted by the distances among them, this is an ontology of presentism.
However, employing these terms is quite misleading, since they presuppose
time as a primitive. In this ontology, there is no time, but only change, and
change is what enables this ontology to give an account of the phenomenol-
ogy of becoming and the passage of time. What exists is a configuration of
matter points individuated by distance relations that change. That change
exists, but not a whole ordered stack of configurations of matter points—
this only is a manner of representation of change. There only is one config-
uration of matter points of the universe, with the relations that individuate
the elements of that configuration changing. This is the only reason why this
ontology is akin to presentism. In any case, presentism, thus conceived, is
the most simple and parsimonious ontology, since only one configuration
exists.

5.4 Super-Humeanism for general relativistic physics

Against this background, let us now apply Leibnizian relationalism includ-
ing Super-Humeanism to general relativistic physics. Whereas the Super-
Humean strategy has been tried out for classical and quantum mechanics
in the recent literature, it has as yet not been considered for general relativ-
ity theory (GR). The metric formulation of GR involves a set of tensor-fields
defined over a four-dimensional semi-Riemannian manifold. Using a local
(i.e. component-based) language, we can cast the field equations of the
theory in the following form (κ being an appropriate constant):
Gij ½gij  ¼ kT ij ½φ; gij : ð5:15Þ

For the sake of simplicity, we omit the term involving the cosmological
constant. The indices i and j range from 0 to 3. The left hand side of
(5.15), being dependent on the metric tensor gij, contains information
about the geometry of the manifold, while the right-hand side conveys
information about the physical properties (e.g. energy density, momentum
flux) of the material sources mathematically modeled by the portmanteau ϕ,
such as the electromagnetic field.
This formulation makes clear that Leibnizian relationalism combined
with the Super-Humean strategy faces what appears to be two knock
down objections, which we have addressed in general terms in the two pre-
ceding sections; we shall now rebut them insofar as they concern GR in par-
ticular: in the first place, space-time in GR is inherently four-dimensional.
There is no objective—that is, non-arbitrary—way to distinguish spatial
from temporal relations holding between events. This in turn implies that
there is no well-defined notion of spatial configuration in a Leibnizian
sense, which undermines the possibility to recover time in the way
Huggett (2006) does when applying Super-Humeanism to Newtonian
 Relationalism for relativistic physics 153
mechanics. Moreover, GR is considered as the paradigm example of a field
theory. Accordingly, insofar as there is a relationalism worked out for GR
in the literature, it is a relationalism that replaces spatial with spatio-
temporal relations and particles with fields (see Rovelli (1997) for a
prominent example as well as Dieks (2001), section 6; see Pooley (2001)
and Pooley (2013), section 7, for a philosophical assessment). The famous
hole argument, going back in its contemporary form to Earman and
Norton (1987), can be seen as lending further support to such a relationalism.
Following this field relationalism, there is a plenum of fields instead of
distance relations between sparse points. However, one may wonder
whether this relationalism is worth its name, because there no longer is a
principled difference with substantivalism. The reason is the ambiguous
status of the metrical field in GR: the metrical field implements the geome-
try of space-time, but it also carries energy and momentum. For the sub-
stantivalist, the metrical field has a special status, being space-time, since
it contains the geometry of the universe. For the field relationalist, it is a
field interacting with other fields. Both endorse the metrical field as an
entity sui generis, the substantivalist calling it “space-time”, the field rela-
tionalist regarding it as a field among other fields. In any case, therefore,
this field relationalism relies on a huge amount of geometrical structure,
since the metrical field is part of the ontology, even if that field is considered
as consisting in metrical relations between point events. Not only is this
field ontology thereby committed to primitive geometrical facts, it also
gives up the simplicity and coherence of a relationalism that is based only
on distance relations: if fields are what there is, they are not properties of
anything, but have to be conceived as some sort of a primitive, extended
stuff. However, as pointed out in section 2 as well as before in Chapter
2, end of section 1, this implies the commitment to a primitive stuff-
essence of matter qua fields in contrast to the individuation of matter qua
sparse, unextended points through distance relations on the sparse relation-
alism proposed in this book.
The relationalist can resist the move to such a field relationalism: in GR
as in any other field theory, fields are tested by the motion of particles.
There is no direct evidence of fields. All the evidence is one of particle
motion. Thus, all empirical determinations of the gravitational field
amount to observations of the motion of bodies in the sense of change in
the instantaneous spatial relations they stand in. As Einstein puts it,

The gravitational field manifests itself in the motion of bodies. There-

fore the problem of determining the motion of such bodies from the
field equations alone is of fundamental importance.
(Einstein and Infeld (1949), p. 209)

Consequently, the field equations are there to determine the motion of

bodies. This opens the door for applying the Super-Humean strategy also
154 Relationalism for relativistic physics
to GR: instead of buying into the dualism of gravitational field and material
bodies as suggested by Einstein and Infeld in this quotation, one can main-
tain that the ontological bedrock—the Humean mosaic—consists in the
motion of bodies only. The gravitational field is a mere representational
means that enables us to describe the overall motion of the bodies in a
manner that optimizes simplicity and information about that motion.
In practice, though, GR is almost never used as a “closed” theory of
interacting point particles. One reason is that, very much like in classical
electrodynamics, the N-body problem is ill-defined in GR, because the
space-time metric is singular along trajectories of point particles.
Another, more pragmatic reason is that on astronomical or even cosmolog-
ical scales, on which the theory is most often used, the ripples in the space-
time produced by individual particles are much too fine grained to be obser-
vationally relevant. Instead, the matter content of the relevant systems (be it
a galaxy or the entire universe) is usually described in terms of continuous
fields, with the energy-momentum tensor entering the field equations being
that of an “incompressible fluid” or “cosmic dust”. In the literature, this is
often interpreted as entailing an ontological commitment to fields rather
than particles. Such an interpretation, however, mistakes an effective
description for a fundamental one. One can with good reason maintain
that the ontological commitment of GR is still with point particles. The
relationship between the fundamental particle description and the effective
field description (e.g. in terms of fluid dynamics) is one of coarse-graining;
cf. our remark on Vlasov and fluid dynamics in the framework of classical
mechanics at the end of section 3.1.
This reasoning also applies to the account of purely gravitational phe-
nomena, such as space-time singularities or gravitational waves. Also in
these cases all that is physically observable is the change of spatial relations
among material bodies. The metric field carrying the information of the
configuration of black holes or the propagating gravitational waves is a
convenient mathematical means for the dynamical description, which has
a very similar role as the electromagnetic field in the theory of classical elec-
trodynamics. However, the fact that the mathematical description makes
use of a field by no means implies or even requires the commitment to a
field in the ontology, if in the end all that has to be accounted for is the
change of spatial relations among matter points. For example, the LIGO
experiment setup that detected gravitational waves in 2016 is a precision
measurement of a 4 km distance by means of laser interferometry that is
able to detect changes of length on the order of one-thousandth the
charge diameter of a proton. What was monitored was a certain type of
change of a spatial relation between matter points, which can mathemati-
cally be conveniently described by a wave rippling through in the metric
field.
For simplicity, in the following, we will focus on “pure gravity”—that is,
neglect non-gravitational forces—such that all particle motion is geodesic
 Relationalism for relativistic physics 155
motion according to GR. Again, as mentioned in Chapter 2, section 1 in
general terms, this Leibnizian stance cannot recognize all mathematically
possible solutions of in this case the field equations of GR as describing
physically possible worlds. In particular, (5.15) allows for empty cosmolog-
ical solutions—that is, models in which the universe is totally deprived of
matter—yet space-time has, say, ripples and lumps. These solutions have
to be dismissed as mathematical surplus structure of the theory; taking
them ontologically serious would amount to inflating the ontology with
the gravitational field as a substance sui generis existing over and above
matter. Again, dismissing these solutions is no problem, since the world
we live in undoubtedly is not an empty universe. By the same token, also
the solutions lacking (global) Cauchy hypersurfaces—that is, three-surfaces
that are intersected only once by any non-space-like curve—have to be dis-
missed, since these solutions would depict cosmological models in which it
is even not possible to give a physical meaning to a space/time split.
Although all these solutions can be dismissed as mathematical surplus
without ontological significance and although the objection from GR
being a field theory can be countered on the basis of all empirical evidence
consisting in the motion of bodies, the fact remains that the fundamental
relations in GR seem to be spatio-temporal ones between events rather
than distance relations between particles that change. This comes out
clearly even if one focuses only on models of space-times that admit
space-like Cauchy surfaces. Consider a solution of (5.15) consisting in a
four-geometry that includes word-lines of material bodies, such that the
manifold possesses a product topology S3  R. Then, we can reduce the
physical description of this four-geometry to the sum of descriptions on a
pile of three-dimensional Cauchy surfaces S3 that cut the manifold in the
space-like direction. The manifold can thus be foliated by a series of
space-like leaves. Nonetheless, the inherent four-dimensionality of GR
shows up in the fact that the choice of the foliation is not unique; for any
choice of three-foliations, we always recover the same four-dimensional
physical representation including the worldlines of the material bodies.
Again, however, the arguments spelled out in the preceding section
suggest taking the four-geometry as a means of representation instead of
endorsing it as constituting the ontology of GR in the guise of a block uni-
verse. It is true that the distances making up the configuration of matter
points and their change cannot be specified uniquely in GR. However,
this just shows that fundamental space-like facts can be described in differ-
ent yet equivalently simple and strong ways. We should not conclude from
this descriptive underdetermination that there are no fundamental space-
like, relationalist facts, on pain of losing the simplicity, coherence and
empirical adequacy of the ontology.
This situation is in a certain sense analogous to the one of Super-
Humeanism in the area of quantum mechanics: on Super-Humeanism
applied to Bohmian mechanics, for instance, the particle positions and
156 Relationalism for relativistic physics
their change make up the Humean mosaic. However, due to the universe
being in quantum equilibrium, our knowledge is limited to what can be
obtained by applying Born’s rule. Furthermore, there are two different
formulations of Bohmian mechanics—the standard one and the identity-
based one—that agree on the accessible facts, but disagree on the particle
trajectories (see Chapter 3, section 2). Thus, the Humean mosaic cannot
be uniquely specified in the Bohmian primitive ontology approach.
In general, a limit of accessibility applies to any primitive ontology pro-
posed for quantum mechanics, as pointed out at the end of Chapter 3.
Nonetheless, the argument for these theories is that they solve the mea-
surement problem by providing an ontology and a dynamics that make
unique measurement outcomes intelligible. Hence, the issue of what is
the Humean mosaic is one of ontological argument and not whether or
not the Humean mosaic can be uniquely described or is accessible to
observation.
Recent progress in physics provides a further argument to vindicate the
Leibnizian perspective on space and time in general relativistic physics.
Gomes et al. (2011) show that there is an alternative theory defined on
the phase space of GR. Such a theory is not just alternative, but dual to
GR in the sense that, under the appropriate choice of gauge, the dynamical
trajectories of the two theories coincide. This dual theory is Barbour’s shape
dynamics (see Gomes and Koslowski (2013) for a concise presentation and
see the discussion of Barbour’s theory of classical mechanics in Chapter 3,
section 1).
Leaving aside technical considerations, the important philosophical con-
sequence of this duality in classical (that is, non-quantum) gravity is that
there are two metaphysical stances with respect to space and time that
are compatible with the empirical predictions of general relativistic
physics. The first one is the irreducibly four-dimensional perspective of
GR; the second one is the three-dimensional perspective of shape dynamics.
In this theory, dynamics can be depicted as a succession of (conformal)
three-geometries. Unlike GR, this theory is not invariant under change of
foliation. Instead, this characteristic symmetry of GR is traded for a local
conformal symmetry, which means that the geometry defined on each
folio lacks a privileged notion of scale—it is not size, but shape that matters.
In shape dynamics, there is a well-defined notion of instantaneous spatial
configuration. This fact does not imply that instants are referred to an exter-
nal clock ticking in the background. What exists is a succession of spatial
configurations that can be arbitrarily labeled by a monotonically increasing
parameter, but this succession is not in time. On the contrary, it defines a
“time-like” direction for the unfolding of the dynamics. Hence, shape
dynamics is compatible with Leibniz’ (and Mach’s) view of space as the
order of coexistence (i.e. bodies related by spatial relations) and time as a
bookkeeping device to describe the order of change (i.e. the succession of
instantaneous spatial configurations).
 Relationalism for relativistic physics 157
We do not claim that shape dynamics fully implements the ontology of
Leibnizian relationalism. First of all, also this theory is a field theory. Sec-
ondly, at this stage, the theory has been worked out only for pure gravity
(but see Gomes and Koslowski (2012) for some first results for gravity-
matter coupling). Thirdly, this theory includes irreducible geometric facts
regarding angles (that is, a conformal structure) contra our fundamentally
geometry-less ontology. However, as argued in Chapter 3, section 1, it
would be a misunderstanding to require that a relationalist ontology has
to be vindicated by a physical theory that employs only relationalist repre-
sentational means. The fact that Barbour’s theory, which is the most
detailed relationalist physical theory for both classical mechanics and
general relativistic physics as yet, has to rely on a primitive conformal struc-
ture for the very concept of shape to be meaningful just underlines again
that when it comes to a physical theory, more representational means are
needed than what is provided by the sparse ontology of distance relations
only, although this ontology is fully empirically adequate.
The point at stake here is that, to put it in the terms of Gryb and Thé-
bault (2016), gravity in general relativistic physics is “Janus-faced”: it is
compatible both with a block universe where there is variation of local
matters of fact but no change (and, hence, no time ordering change), and
with a dynamical universe where there is change in the guise of change in
the fundamental relations that connect the fundamental objects, on which
temporal facts supervene in the form of a monotonically increasing param-
eter labelling the succession. In short, the duality of GR and shape dynamics
again shows that one cannot read the ontology off from the mathematical
structure of a physical theory, since one would in this case end up with two
proposals for the ontology of general relativistic physics that contradict
each other, based on two different formalisms for general relativistic
physics. The ontology has to be settled by criteria such as parsimony and
coherence together with empirical adequacy.
Against this background, taking the distance relations among the point
particles and their change throughout the history of the universe to be
the Humean mosaic, let us now exploit for GR the Super-Humean strategy
that Huggett (2006) developed in order to vindicate a Leibnizian relation-
alist ontology for Newtonian mechanics. To briefly recall (see Chapter 3,
section 1, for details), Huggett (2006) employs the concept of adapted
frame, that is, a reference frame tied to a body (e.g. a material particle)—
otherwise said, an assignment of N-tuples of real numbers over time such
that the origin (0, 0, . . ., 0) corresponds to the body to which the frame
is adapted. Inertial frames then are those frames in which an observer
would describe the history of relations in the simplest and strongest way
in terms of Newton’s laws. Note that, in order to exploit this characteriza-
tion, we do not necessarily need any body to be actually tied to an inertial
frame. This is because, once we have a class of adapted frames, unoccupied
frames can be defined as those related to the occupied ones by means of a
158 Relationalism for relativistic physics
spatial rigid translation. As pointed out in Chapter 3, section 1, the notions
of adapted frame and spatial translation do not require any ontological
commitment over and above what is included in axioms 1 and 2: they
just presuppose matter points and change of spatial relations among
them. Once this characterization of inertial frame is in place, absolute accel-
eration can be reduced to the history of change of the spatial relations
holding between an inertial and a non-inertial frame: acceleration is part
of the description of how much the pattern that constitutes the history of
a non-inertial frame deviates from the regularities encoded in the inertial
pattern as seen in an inertial frame. By the same token, any other absolute
quantity of motion such as, for example, angular momentum, can be shown
to supervene on the Humean mosaic.
When it comes to GR, as in the classical case, we can define an adapted
frame of reference as one in which a certain particle is always at rest at the
origin, so that all distances along the axes are just the distances from that
particle (i.e. the arbitrary way in which we label spatial relations). We
can then relate all reference frames by means of rigid spatial translations
(i.e. continuous functions that send N-tuples of coordinates in one frame
to N-tuples of coordinates in another frame in a way that preserves our cri-
terion to assign labels to spatial relations). We look as usual for regularities
in the history that admit a particularly simple yet strongly informative
description. However, the problem that we face in this respect is that
there is no such thing as classical inertial motion in GR: this theory
merges inertial structure and the gravitational field. Consequently, there
is no unique way to separate them, let alone find a case where gravitational
effects can be neglected in a region bigger than a point. Hence, the descrip-
tion that we search for cannot be encoded in a law of the form x €i ¼ 0 (a dot
indicates the usual differentiation with respect to an arbitrary time-like
parameter defined along a given pattern); in this case, we would be
dealing with a Newtonian world rather than a general relativistic one.
Nonetheless, even if we cannot make use of the concept of inertial
motion in GR, still we can employ a generalization of this concept: the sim-
plest and most informative description of the regularities in a general rela-
tivistic world is that of geodesic motion (more precisely, non-space-like
geodesic motion)—that is, x €i þ Gijk x_ j x_ k ¼ 0, using an appropriate (affine)
parametrization. Geodesic motion is a well-defined concept in GR, since
the theory stipulates that the motion of freely-falling bodies follows geode-
sic trajectories. Here we immediately see that the (Levi-Civita) connection
with coefficients Gijk is not a geometrical feature crafted so to speak in
the Humean mosaic: it is just a tool that helps to formalize how the descrip-
tion will change among two different observers moving geodesically—in
other words, a description of how two frames adapted to freely-falling
bodies and related by a rigid spatial translation differ.
Note that we do not presuppose the existence of a space-time structure
(in particular, a connection) that defines what it is for a motion to be
 Relationalism for relativistic physics 159
geodesic, but, rather, the other way round: we define geodesic motion as a
particularly simple pattern in the history of relational change. Then we con-
struct a connection as a bookkeeping device that accounts for the relational
differences among different geodesics: there are no primitive geometrical
facts, just descriptive tools taken from the language of (differential) geom-
etry. Similar to the classical case, here there might actually be no body
moving geodesically: what we require is just that any other adapted
frame can be related to a geodesic one by means of a spatial rigid transla-
tion. If this cannot be achieved, then the possible world we are dealing with
does not admit the laws of general relativistic physics as theorems of the
simplest and strongest system.
It is worth noting that, in the Newtonian case, the coefficients of the con-
nection do not arise, because all inertial motions are identical, which is con-
sistent with the fact that Newtonian space-time admits a flat connection (i.
e. the simplest and strongest description of inertial motion is one for which
Gijk ¼ 0). In the general relativistic case, by contrast, geodesic motions
might in fact differ even when adopting the same affine parametrization
(what is usually called “geodesic deviation”, which is at the root of the
explanation of how tidal forces arise). From this point on, all the relevant
geometrical and dynamical features of the theory can be shown to super-
vene on the Humean mosaic in the usual manner. For example, curvature
is a simple and informative way to describe relational change between
freely-falling bodies; more generally, the metrical structure gij of space-
time supervenes on geodesic motions through the usual relation Gijk ¼
g gkh;j þ ghj;k  gjk;h (a comma indicates standard differentiation with
1 hi
2
respect to the subsequent index). Geodesic trajectories also suffice to fix
the topology of space-time, as shown in Malament (1977). Note that, by
construction, the geometrical description encoded in Gmns and gμν does not
inhere within single trajectories, but supervenes on the totality of geodesic
motions. In this sense, the continuous large-scale geometry of space-time
results from a coarse-grained description of the history of relational
change among particles.
In general, all fields—including, as we have seen, the metrical one—are
in this framework shortcuts to describe certain forms of particle motion.
Hence, the relationship between particles and fields is not physical—parti-
cles do not generate fields, nor are they pushed by fields—but descriptive: a
particle-vocabulary can be coarse grained to a field one in order to simplify
the description of gravitational phenomena without loss of physical infor-
mation. Then, quantities such as momentum densities, energy flows and
the like can be constructed out of this field-vocabulary. This issue is
crucial: if the ontology is the one of Leibnizian relationalism, quantities
such as the stress-energy tensor do not refer directly to the underlying par-
ticle motions, but express their field-like description; they are, so to speak, a
description of how a description changes. Clarifying this point is very
important because if we took particles and fields on a par and, in particular,
160 Relationalism for relativistic physics
regarded particles as field sources, then singularities would arise. This is all
the more true for GR. In this theory, a Schwarzschild radius is associated to
any extended body, which depends on the body’s mass: roughly speaking, if
the body’s extension drops below this radius, then the body collapses into a
black hole singularity. It is then obvious why, in GR, this notion of massive
point-particle makes little sense besides some particular approximations.
However, it should be clear that our notion of particle is not this notion
of massive point-particle.
What has been said so far shows that there is nothing strange in claiming
that a large-scale continuum theory such as GR (i.e. a theory employed to
model continuous material systems at astrophysical or cosmological scales)
admits an extremely sparse ontology in terms of distance relations individ-
uating matter points. Nonetheless, these results do not show that the central
law of GR—namely, the field equations (5.15)—can be recovered from the
Humean mosaic as a theorem of the system that strikes the best balance
between simplicity and strength in describing the change in the distance
relations among the point particles that occurs throughout the history of
the actual world. Indeed, thus recovering the field equations (5.15) of GR
is outright impossible for the following reason: if the Humean mosaic are
distance relations among sparse matter points and their change, we can
obtain geometry and dynamics in a package as the system that optimizes
simplicity and information in the description of that change. However,
geometry and dynamics remain distinct, as in the classical case. In other
words, the dynamical laws that describe such a Humean mosaic are those
of a theory formulated over a fixed (albeit non-Euclidean) background.
There is nothing in the regularities of motion of such a Humean mosaic
that could lead to a higher order description where geometry is coupled
to matter in a dynamical way. The ontology of there being only distance
relations and their change is too meagre to get (5.15) from such a possible
world by Humean means: the most we can get is a possibly very complex,
but still non-dynamical geometry compatible with the geodesic law of
motion. This is because geometry supervenes on the mosaic by means of
the law of geodesic motion, while non-gravitational fields come out as a
way to describe non-geodesic motion: given this construction, there is no
way to conjecture any inter-dependence between geometry and material
fields.
That notwithstanding, this result can be accommodated in an ontology
of there being only distance relations among sparse matter points and
their change that does justice to general relativistic physics. Consider the
class of cosmological solutions of (5.15) (modulo those that we have previ-
ously discarded as mathematical surplus): each of these models describes a
possible world in which GR holds. These models are each characterized by
a certain metric gij (compatible with the Levi-Civita connection) and a
certain stress-energy tensor Tij. The main difference that this cluster of pos-
sible worlds bears with respect to a cluster of Newtonian worlds is that the
 Relationalism for relativistic physics 161
Newtonian worlds always feature the same spatial and temporal metrical
structures. By contrast, in general, gμν co-varies with Tμν from world to
world in the GR case. This is one possible sense in which we can understand
what is known as the background independence of GR: the same spatio-
temporal structures of Newtonian mechanics appear in all Newtonian
models, so that they are nomologically necessary; spatio-temporal struc-
tures in GR, by contrast, are nomologically contingent.
Against this background, what (5.15) represents is the manner in which
in any possible world of GR, spatio-temporal structures described by some
metric tensor gij are correlated with material structures described by a
stress-energy tensor Tij. That is to say, in any possible world of the ontology
of distance relations among sparse matter points and their change in which
GR is valid, the system that strikes the best balance between simple and
being informative about a particular such world is one in which a dynamics
describes the evolution of the configuration of matter over a fixed back-
ground—that is, a particular cosmological solution of (5.15) describing a
particular possible world of GR. The general law (5.15) then is not a
theorem that generalizes the regularities within a world, but a trans-
world generalization: it expresses the relationship between the mathemati-
cal structures formulating the geometry and the mathematical structures
formulating the dynamics in any possible world where GR is the best
system describing the change in the distance relations among the point par-
ticles. In sum, we propose a generalization of the Super-Humean account
that Huggett (2006) developed for Newtonian mechanics that shows that
geometry and dynamics supervene on the Humean mosaic of Leibnizian
relations at each world by means of a geodesic law; we then obtain Ein-
stein’s field equations as the simplest and strongest description of the corre-
lation between geometry and dynamics for the whole class of these worlds.
We consider only classical gravitation (in a Newtonian as well as general
relativistic setting), but not quantum gravity, since, by contrast to the stan-
dard model of QFT, there is no established quantum theory of gravitation
as yet. The discussion of Leibnizian relationalism applied to general relativ-
istic physics in this section and the elaboration of a (Bohmian) particle
ontology for QFT in Chapter 4, both conceived in the Super-Humean
setting, complete our case for the minimalist ontology set out in Chapter
2 being able to cover all existing physics, in fact being the best proposal
for an ontology of the natural world that is based on our scientific knowl-
edge in fundamental physics as a whole.
Nonetheless, let us add a few sketchy remarks with respect to the claims
about the alleged emergence of space-time in a future theory of quantum
gravity (see, e.g., Huggett and Wüthrich (2013) for a sympathetic overview
and, e.g., Esfeld and Lam (2013) for voicing scepticism):

1. As in any quantum theory, so also in a future theory of quantum gravity,

one cannot settle ontological issues on the basis of an operator
162 Relationalism for relativistic physics
formalism. One has to spell out an ontology that solves the measurement
problem and thereby establishes the link between the theory and the
empirical evidence (which is supposed to confirm the theory).
2. The emergence of time is nothing new and poses no problem. Indeed, the
Leibnizian relationalism proposed in this book includes what can be
described as the emergence of time by deriving time from change. Thus,
the emergence of time can be studied already for classical mechanics.
More importantly, Barbour’s framework, which includes the emergence
of time on the basis of a commitment to fundamental spatial relations,
can be put to work also in the domain of quantum gravity (see, e.g.,
Barbour et al. (2014) and Gryb and Thébault (2016)). Hence, one has
to distinguish between the emergence of time and the emergence of
space: the former, by contrast to the latter, is well studied.
3. As elaborated on in Chapter 2, section 1, without acknowledging a plu-
rality of objects in some sense—be it so-called thin objects without
intrinsic properties (see French (2010))—it is difficult to see how empir-
ical adequacy could be achieved. If there is a plurality of objects, there
has to be a certain type of relations in virtue of which these objects
make up a configuration that then is the world. This type of relations
has furthermore to be such that it characterizes the world as a physical
or material world, by contrast to, say, a hypothetical world of Carte-
sian minds that are individuated by thinking relations. That is the
reason to single out a type of relations providing for extension—
namely, the spatial relations. These relations are able to constitute a
physical or material world by individuating simple objects that then
are matter points and to lead to empirically adequate theories
through including their change as second axiom in the ontology and
formulating a dynamical structure that describes the change in the
spatial relations. No one has as yet worked out a theory in which
another type of relations (from which spatial relations may then
emerge) performs that task.

Of course, there is no a priori argument independent of science for spatial

relations being the fundamental relations that unify the world. To put it in
the spirit of Popper, the ontology set out here is an audacious proposal that
is open to falsification. Future science may teach us other lessons. However,
instead of simply announcing such lessons on the basis of not yet existing
physical theories, one should work out how the criteria (1) and (3) can
be met in a way that is not based on fundamental distance relations.

5.5 Relativistic physics and quantum entanglement

Instead of making the effort to argue that Leibnizian relationalism together
with Super-Humeanism stands also firm as the best proposal for the
 Relationalism for relativistic physics 163
ontology of general relativistic physics, one can try to exploit the tension
between relativistic physics on the one hand and quantum mechanics and
QFT on the other hand in order to vindicate an ontology that is based on
spatial relations rather than spatio-temporal ones. That tension is
brought by Bell’s theorem. This theorem is based on the following two
premises:

1. No conspiracy: the choice of the variables to be measured on a physical

system is independent of the past state of the system.
2. Locality: an event can only be influenced by events that are located in
its past light cone.

Bell’s theorem then proves that no theory that endorses these two premises
can reproduce the statistical predictions of measurement outcomes of
quantum mechanics, more precisely the statistics of the measurement out-
comes on entangled quantum systems (see Bell (2004), in particular, ch.
2, 7 and 24; see, furthermore, Seevinck and Uffink (2011)). Since any
future physical theory will have to reproduce the statistical predictions of
measurement outcomes of quantum mechanics that are confirmed by exper-
iment, Bell’s theorem poses a constraint not only on hidden variable theo-
ries of quantum mechanics but also on any future physical theory.
The no conspiracy premise is a general premise that concerns all exper-
imental investigations of nature: if the past state of the system under inves-
tigation influenced the choice of the variables to be measured on the system,
then no reliable information about the system could be obtained through
experimental investigation. This premise is compatible with determinism;
it is, for instance, satisfied in Bohmian mechanics (see Esfeld (2015)).
Since the no conspiracy premise is a general premise of natural science
that is not specific for quantum physics, or relativity physics, the locality
premise is the one that Bell’s theorem rules out. More precisely, Bell’s
theorem is widely taken to establish non-locality in the sense that events
that are separated by a space-like interval influence each other, such as
the events of the measurement settings and measurement outcomes across
both wings of the EPR experiment (see the seminal monograph of
Maudlin (2011)). Note that such an influence does not automatically
mean that there is direct causation between space-like separated events.
The correlation between these events can also be accounted for in terms
of a common cause, but that common cause then is non-local. Both
Bohmian mechanics and the GRW matter density theory can be received
as providing such a common cause explanation of the correlated measure-
ment outcomes in the EPR experiment (see Egg and Esfeld (2014)).
Recent research on the tension between quantum non-locality and rela-
tivity physics has shown that it is possible to formulate the dynamical struc-
ture of quantum physics in the primitive ontology framework in a Lorentz
invariant manner if one considers the entire spatio-temporal distribution of
164 Relationalism for relativistic physics
the elements of the primitive ontology throughout the history of the uni-
verse as a whole; in particular, Tumulka (2006, 2009) has established
this result in the framework of the GRWf theory with flash-events as the
primitive ontology (see Bedingham et al. (2014) for a similar result for
the GRWm theory with a field as the primitive ontology). However, this
result does not obtain if one considers the evolution of the configuration
of the elements of the primitive ontology starting with an arbitrary initial
configuration, not to mention the interaction between the elements of the
primitive ontology: there is no means to represent that evolution (and the
interaction) in a Lorentz invariant manner available—that is, without
including influences that connect space-like separated events in the dynam-
ics so that a definite temporal order of these events has to be assumed—
although that temporal order is not accessible in experiments (see Esfeld
and Gisin (2014) and Barrett (2014)). Again, this is a feature of the mea-
surement problem: everything is fine with respect to Lorentz invariance as
long as one considers only possible evolutions of a given initial configura-
tion of, say, GRW flashes and assigns probabilities to these possible evolu-
tions. However, one cannot understand the realization of one such
evolution (corresponding to the occurrence of particular measurement out-
comes) in a Lorentz invariant manner.
By way of consequence, the recent research on Lorentz invariance in the
framework of the GRWf and the GRWm theories constitutes no argument
to prefer the ontology of flashes or a field with a GRW-type collapse
dynamics to the ontology of permanent particles with a Bohmian dynamics.
The argument of Chapter 3, section 3, for Bohmian mechanics providing
the best solution to the measurement problem stands firm also in view of
these recent results. As we briefly mentioned at the end of Chapter 4, the
dynamical structure of Bohmian mechanics as well as Bohmian QFT
requires the committed to a preferred foliation of flat Minkowski space-
time into space-like hypersurfaces, but this foliation does not have to be
introduced by hand: it is possible to consider the universal wave function
as determining that preferred foliation. Nevertheless, the theory then is
Lorentz invariant on the level of empirical predictions. It is not possible
to discover that preferred foliation through experiments (see Dürr et al.
(2013a) and the brief discussion in Chapter 4, section 6). In any case, prob-
lems that primitive ontology theories of quantum physics may face with
respect to the dynamical structure of special and GR theory cannot be
counted as an argument against these theories: they solve the measurement
problem. No one has produced a solution to this problem that (a) acknowl-
edges determinate measurement outcomes and (b) is a relativistic theory of
interactions, including in particular measurement interactions (see again
Barrett (2014)).
These results as well as the general argument for the tension between
quantum non-locality and relativity physics are conceived in the framework
of a field formulation of the dynamics and including the background
 Relationalism for relativistic physics 165
assumption that advanced action is ruled out. The field formulation of clas-
sical electrodynamics situates every influence in the immediate past of a
given event by representing interaction as being transmitted by fields.
Accordingly, we have formulated the locality premise that enters Bell’s
theorem in such a way that locality means local influences being situated
in the past light cone of a given event. Bell’s theorem then establishes
that quantum physics violates this locality principle (if one takes the
general no conspiracy premise for granted). Against the background of a
field formulation of the dynamics, it is obvious why this locality principle
breaks down in the quantum case: the wave function is a field parameter.
However, it is not a field parameter on ordinary space, but on configuration
space, being a field following a local dynamics only on configuration space.
By way of consequence, in contrast to relativity physics, the wave function
correlates the motion of spatially distant particles (and does so indepen-
dently of their distance in physical space).
Nonetheless, as in the case of classical electrodynamics, so also in the
quantum case, one can conceive a formulation of the dynamical structure
that is based on both retarded and advanced particle interaction (see
Cramer (2016) for the most prominent physical model and notably Price
(1996), ch. 9, Dowe (1996) and Corry (2015) for philosophical endorse-
ments; there even is a version of Bohmian mechanics with advanced
action, see Sutherland (2008)). Generally speaking, whenever there is a
dynamical structure of a physical theory in terms of direct interaction
between spacelike separated events, one can in principle also conceive a
structure that transforms that direct interaction into an advanced action
of future events. By contrast to the case of classical electromagnetism as
illustrated by the Wheeler-Feynman theory of direct particle interactions,
in the quantum case, the advanced action option, however, has to include
the field parameter, working with both retarded and advanced waves in
the dynamical structure. That difference notwithstanding, by buying into
advanced action, one avoids quantum non-locality in the sense of influences
that connect space-like separated events (be it in terms of direct causation,
be it in terms of a non-local common cause). In the case of advanced action,
all influences are represented in terms of waves that go exclusively through
the light cones, but in addition to influences from the past light cone on a
given event, there also is an influence coming from the future light cone
of the event in question.
Consequently, by taking the advanced action option, one cannot restore
locality in the sense of the locality premise formulated above. There is no
question of a dynamical structure that situates every influence on a given
event in the (immediate) past light cone of the event in question. Admitting
advanced action infringes furthermore upon the no conspiracy premise:
there is no independence between the past state of the system and the
choice of the measured parameter. However, there is no direct influence
from the past state of the system on the choice of the measured parameter.
166 Relationalism for relativistic physics
Both are correlated only through the influence from the future measurement
outcomes. Any advanced action theory entails such a correlation. This cor-
relation established through the influence from the future does not necessar-
ily result in a paradox or a conspiracy (see, e.g., Price (1996), ch. 9, and
Lazarovici (2014)). In brief, a direct interaction dynamics can in principle
be empirically adequate. It provides a relativistic dynamics also for the
quantum case, albeit one that includes advanced in addition to retarded
action.
The in principle availability of a dynamical structure in terms of both
retarded and advanced action in the case of relativity physics at least as
far as classical electromagnetism is concerned as well as in the quantum
case at least as far as quantum mechanics is concerned shows that there
is not necessarily a conflict between the dynamical structures of relativity
physics and quantum physics. By way of consequence, it is short-sighted
to base an argument for an ontology of spatial rather than spatio-temporal
relations on the tension between quantum physics and relativity.
The discussion of relativistic physics rather underlines again the need to
distinguish between the ontology of the natural world and the dynamical
structure used in the formulation of a physical theory: there is no point
in reading the ontology off from the dynamical structure that one
chooses. We need criteria for the ontology that are independent of consid-
erations of the dynamical structure—that is, criteria that are not touched by
the underdetermination of dynamical structure and the change of dynami-
cal structure in the history of physics. Parsimony—together with empirical
adequacy—is the central criterion for ontology, because one has to justify
why one should admit an entity to the ontology, and its figuring in the
dynamical structure of a physical theory does not provide such a justifica-
tion. Parsimony leads to the question of what is an ontology of the
natural world that is minimally sufficient for empirical adequacy. Leibniz-
ian relationalism combined with Super-Humeanism—that is, an ontology
of distance relations individuating matter points and the change of these
relations—is the answer to this question. As set out in this book, we thus
achieve an ontology that covers all known physics while being most
parsimonious.
Bibliography

Albert, D. Z. (2015). After physics. Cambridge, Massachusetts: Harvard University

Press.
Albert, D. Z. and Loewer, B. (1996). Tails of Schrödinger’s cat. In Clifton, R. K.,
editor, Perspectives on quantum reality, pages 81–91. Dordrecht: Kluwer.
Allori, V., Goldstein, S., Tumulka, R., and Zanghı̀, N. (2008). On the common
structure of Bohmian mechanics and the Ghirardi-Rimini-Weber theory.
British Journal for the Philosophy of Science, 59(3):353–389.
Allori, V., Goldstein, S., Tumulka, R., and Zanghı̀, N. (2014). Predictions and prim-
itive ontology in quantum foundations: a study of examples. British Journal for
the Philosophy of Science, 65(2):323–352.
Anderson, E. (2013). The problem of time and quantum cosmology in the relational
particle mechanics arena. arXiv:1111.1472v3 [gr-qc].
Anderson, E. (2015). Configuration spaces in fundamental physics. arXiv:1503.
01507v2 [gr-qc].
Ariew, R., editor (2000). G. W. Leibniz and S. Clarke: Correspondence. Indianap-
olis: Hackett.
Armstrong, D. (2004). Going through the open door again: counterfactual versus
singularist theories of causation. In Collins, J., Hall, N., and Paul, L. A.,
editors, Causation and counterfactuals, pages 445–457. Cambridge, Massachu-
setts: MIT Press.
Arntzenius, F. (1994). Electromagnetic arrows of time. In Horowitz, T. and Janis,
A., editors, Scientific failure. Rowman & Littlefield, Maryland.
Arntzenius, F. and Hawthorne, J. (2005). Gunk and continuous variation. The
Monist, 88:441–465.
Baker, D. J. (2009). Against field interpretations of quantum field theory. British
Journal for the Philosophy of Science, 60:585–609.
Barbour, J. (2003). Scale-invariant gravity: particle dynamics. Classical and
Quantum Gravity, 20:1543–1570.
Barbour, J. (2012). Shape dynamics. An introduction. In Finster, F., Müller, O.,
Nardmann, M., Tolksdorf, J., and Zeidler, E., editors, Quantum field theory
and gravity, pages 257–297. Basel: Birkhäuser.
Barbour, J. and Bertotti, B. (1982). Mach’s principle and the structure of dynamical
theories. Proceedings of the Royal Society A, 382:295–306.
Barbour, J., Koslowski, T., and Mercati, F. (2014). The solution to the problem of
time in shape dynamics. Classical and Quantum Gravity, 31(15):155001.
Barrett, J. A. (2014). Entanglement and disentanglement in relativistic quantum
mechanics. Studies in History and Philosophy of Modern Physics, 48:168–174.
Bauer, G. (1997). Ein Existenzsatz für die Wheeler-Feynman-Elektrodynamik.
München: Herbert Utz Verlag.
168 Bibliography
Bauer, G., Deckert, D. A., and Dürr, D. (2013). On the existence of dynamics in
Wheeler-Feynman electromagnetism. Zeitschrift für angewandte Mathematik
und Physik, 64(4):1087–1124.
Bauer, G., Deckert, D.-A., Dürr, D., and Hinrichs, G. (2014). On irreversibility and
radiation in classical electrodynamics of point particles. Journal of Statistical
Physics, 154(1):610–622.
Bauer, G., Deckert, D.-A., Dürr, D., and Hinrichs, G. (2016). Global solutions to
the electrodynamic two-body problem on a straight line. arXiv:1603.05115
[math-ph].
Baumgartner, M. (2013). A regularity theoretic approach to actual causation.
Erkenntnis, 78:85–109.
Bedingham, D., Dürr, D., Ghirardi, G. C., Goldstein, S., Tumulka, R., and Zanghı̀,
N. (2014). Matter density and relativistic models of wave function collapse.
Journal of Statistical Physics, 154:623–631.
Beebee, H. (2000). The non-governing conception of laws of nature. Philosophy and
Phenomenological Research, 61:571–594.
Beebee, H. (2006). Does anything hold the universe together? Synthese, 149:509–533.
Beebee, H. and Mele, A. (2002). Humean compatibilism. Mind, 111:201–223.
Bell, J. S. (2004). Speakable and unspeakable in quantum mechanics. Second
edition. Cambridge: Cambridge University Press.
Belot, G. (1999). Rehabilitating relationalism. International Studies in the Philoso-
phy of Science, 13:35–52.
Belot, G. (2000). Geometry and motion. British Journal for the Philosophy of
Science, 51(4):561–595.
Belot, G. (2001). The principle of sufficient reason. Journal of Philosophy, 98:55–74.
Belot, G. (2011). Geometric possibility. Oxford: Oxford University Press.
Belot, G. (2012). Quantum states for primitive ontologists: a case study. European
Journal for Philosophy of Science, 2(1):67–83.
Belousek, D. W. (2003). Formalism, ontology and methodology in Bohmian
mechanics. Foundations of Science, 8:109–172.
Benedikter, N., Porta, M., and Schlein, B. (2014). Mean field evolution of fermionic
systems. Communications in Mathematical Physics, 331(3):1087–1131.
Bhogal, H. and Perry, Z. R. (2017). What the Humean should say about entangle-
ment. Noûs, 51(1):74–94.
Bigaj, T. (2015). Dissecting weak discernibility of quanta. Studies in History and
Philosophy of Modern Physics, 50:43–53.
Bird, A. (2007). Nature’s metaphysics: laws and properties. New York: Oxford
University Press.
Black, R. (2000). Against quidditism. Australasian Journal of Philosophy, 78:87–104.
Blackburn, S. (1990). Filling in space. Analysis, 50:62–65.
Bohm, D. (1951). Quantum theory. Englewood Cliffs: Prentice-Hall.
Bohm, D. (1952a). A suggested interpretation of the quantum theory in terms of
“hidden” variables. 1. Physical Review, 85(2):166–179.
Bohm, D. (1952b). A suggested interpretation of the quantum theory in terms of
“hidden” variables. 2. Physical Review, 85(2):180–193.
Bohm, D. and Hiley, B. J. (1993). The undivided universe. An ontological interpre-
tation of quantum theory. London: Routledge.
Bohm, D., Hiley, B. J., and Kaloyerou, P. N. (1987). A causal interpretation of
quantum fields. Physics Reports, 144:349–375.
Boltzmann, L. (1896). Vorlesungen über Gastheorie. Leipzig: Barth.
Briceno, S. and Mumford, S. (2016). Relations all the way down? Against ontic
structural realism. In Marmodoro, A. and Yates, D., editors, The metaphysics
of relations, pages 198–217. Oxford: Oxford University Press.
 Bibliography 169
Brown, H. R., Dewdney, C., and Horton, G. (1995). Bohm particles and their detec-
tion in the light of neutron interferometry. Foundations of Physics, 25(2):329–347.
Brown, H. R., Elby, A., and Weingard, R. (1996). Cause and effect in the pilot-wave
interpretation of quantum mechanics. In Cushing, J. T., Fine, A., and Goldstein,
S., editors, Bohmian mechanics and quantum theory: an appraisal, volume 184
of Boston Studies in the Philosophy of Science, pages 309–319. Dordrecht:
Springer.
Butterfield, J. (2006a). Against pointillisme about geometry. In Stadler, F. and Stölz-
ner, M., editors, Time and history. Proceedings of the 28th International Ludwig
Wittgenstein Symposium, pages 181–222. Frankfurt (Main): Ontos.
Butterfield, J. (2006b). Against pointillisme about mechanics. British Journal for the
Philosophy of Science, 57:709–753.
Callender, C. (2007). The emergence and interpretation of probability in Bohmian
mechanics. Studies in History and Philosophy of Modern Physics, 38:351–370.
Callender, C. (2011). Philosophy of science and metaphysics. In French, S. and
Saatsi, J., editors, The Continuum companion to the philosophy of science,
pages 33–54. London: Continuum.
Callender, C. (2015). One world, one beable. Synthese, 192(10):3153–3177.
Castañeda, H.-N. (1980). Causes, energy and constant conjunctions. In van
Inwagen, P., editor, Time and cause. Essays presented to Richard Taylor,
pages 81–108. Dordrecht: Reidel.
Cei, A. and French, S. (2014). Getting away from governance: a structuralist
approach to laws and symmetries. Methode, 4:25–48.
Chakravartty, A. (2013). On the prospects of naturalized metaphysics. In Ross, D.,
Ladyman, J., and Kincaid, H., editors, Scientific metaphysics, pages 27–50.
Oxford: Oxford University Press.
Cohen, J. and Callender, C. (2009). A better best system account of lawhood. Phil-
osophical Studies, 145(1):1–34.
Colin, S. (2003). Beables for quantum electrodynamics. Fondation Louis de Broglie.
Annales, 29(1–2):273–296.
Colin, S. and Struyve, W. (2007). A Dirac sea pilot-wave model for quantum field
theory. Journal of Physics A, 40(26):7309–7341.
Corry, R. (2015). Retrocausal models for EPR. Studies in History and Philosophy of
Modern Physics, 49:1–9.
Cowan, C. W. and Tumulka, R. (2016). Epistemology of wave function collapse in
quantum physics. British Journal for the Philosophy of Science, 67:405–434.
Cramer, J. G. (2016). The quantum handshake. Entanglement, nonlocality and
transactions. Cham: Springer.
Curceanu, C. and alteri (2016). Spontaneously emitted x-rays: an experimental signa-
ture of the dynamical reduction models. Foundations of Physics, 46(3):263–268.
Davidson, D. (1995). The problem of objectivity. Tijdschrift voor Filosofie, 57:203–
220.
Dawid, R. and Thébault, K. P. Y. (2014). Against the empirical viability of the
Deutsch-Wallace-Everett approach to quantum mechanics. Studies in History
and Philosophy of Modern Physics, 47:55–61.
de Broglie, L. (1928). La nouvelle dynamique des quanta. Electrons et photons. Rap-
ports et discussions du cinquième Conseil de physique tenu à Bruxelles du 24 au
29 octobre 1927 sous les auspices de l’Institut international de physique Solvay,
pages 105–132. Paris: Gauthier-Villars. English translation in Bacciagaluppi, G.
and Valentini, A. (2009). Quantum theory at the crossroads. Reconsidering the
1927 Solvay conference, pages 341–371. Cambridge: Cambridge University Press.
de Broglie, L. (1964). The current interpretation of wave mechanics. A critical study.
Amsterdam: Elsevier.
170 Bibliography
Deckert, D.-A., Dürr, D., Merkl, F., and Schottenloher, M. (2010). Time-evolution
of the external field problem in quantum electrodynamics. Journal of Mathemat-
ical Physics, 51(12):122301.
Deckert, D.-A., Dürr, D., and Vona, N. (2014). Delay equations of the Wheeler-
Feynman type. Journal of Mathematical Sciences, 202(5):623–636.
Deckert, D.-A. and Hartenstein, V. (2016). On the initial value formulation of clas-
sical electrodynamics. Preprint: arXiv:1602.0468.
Deckert, D.-A. and Hinrichs, G. (2016). Electrodynamic two-body problem for pre-
scribed initial data on a straight line. Journal of Differential Equations, 260
(9):6900–6929.
Dennett, D. C. (1988). Quining qualia. In Marcel, A. and Bisiach, E., editors, Con-
sciousness in contemporary science, pages 43–77. Oxford: Oxford University Press.
Dickson, M. (2000). Are there material objects in Bohm’s theory? Philosophy of
Science, 67(4):704–710.
Dieks, D. (2001). Space-time relationalism in Newtonian and relativistic physics.
International Studies in the Philosophy of Science, 15(1):5–17.
Dieks, D. (2006). Becoming, relativity and locality. In Dieks, D., editor, The ontol-
ogy of space-time, pages 157–176. Amsterdam: Elsevier.
Dieks, D. and Versteegh, M. A. M. (2008). Identical quantum particles and weak
discernability. Foundations of Physics, 38:923–934.
Dirac, P. A. M. (1934). Théorie du positron. In Rapport du 7ème Conseil Solvay de
physique: structure et propriétés des noyaux atomiques, pages 203–212. Paris:
Gauthier-Villars. Reprinted in Julian Schwinger (ed.): Selected papers on
quantum electrodynamics. New York: Dover 1958. Chapter 7, pp. 82–91.
Dirac, P. A. M. (1938). Classical theory of radiating electrons. Proceedings of the
Royal Society A, 167(929):148–169.
Dirac, P. A. M. (1947). The principles of quantum mechanics. Third edition.
Oxford: Oxford University Press.
Dizadji-Bahmani, F. (2015). The probability problem in Everettian quantum
mechanics persists. British Journal for the Philosophy of Science, 66:257–283.
Dowe, P. (1996). Backwards causation and the direction of causal processes. Mind,
105:227–248.
Dowker, F. and Herbauts, I. (2005). The status of the wave function in dynamical
collapse models. Foundations of Physics Letters, 18:499–518.
Dürr, D., Goldstein, S., Norsen, T., Struyve, W., and Zanghı̀, N. (2013a). Can
Bohmian mechanics be made relativistic? Proceedings of the Royal Society A,
470:2162.
Dürr, D., Goldstein, S., Taylor, J., Tumulka, R., and Zanghı̀, N. (2006). Topolog-
ical factors derived from Bohmian mechanics. Annales Henri Poincaré, 7(4):791–
807. Reprinted in Dürr et al. (2013b, ch. 8).
Dürr, D., Goldstein, S., Tumulka, R., and Zanghı̀, N. (2005). Bell-type quantum
field theories. Journal of Physics A: Mathematical and General, 38(4):R1–R43.
Dürr, D., Goldstein, S., and Zanghı̀, N. (2013b). Quantum physics without
quantum philosophy. Berlin: Springer.
Dürr, D. and Teufel, S. (2009). Bohmian mechanics: the physics and mathematics of
quantum theory. Berlin: Springer.
Earman, J. (1989). World enough and space-time. Absolute versus relational theo-
ries of spacetime. Cambridge, Massachusetts: MIT Press.
Earman, J. (2011). Sharpening the electromagnetic arrow(s) of time. In Callender,
C., editor, The Oxford handbook of philosophy of time, chapter 16, pages
360–391. Oxford: Oxford University Press.
Earman, J. and Norton, J. (1987). What price spacetime substantivalism? The hole
story. British Journal for the Philosophy of Science, 38:515–525.
 Bibliography 171
Egg, M. and Esfeld, M. (2014). Non-local common cause explanations for EPR.
European Journal for Philosophy of Science, 4:181–196.
Egg, M. and Esfeld, M. (2015). Primitive ontology and quantum state in the GRW
matter density theory. Synthese, 192(10):3229–3245.
Einstein, A. (1948). Quanten-Mechanik und Wirklichkeit. Dialectica, 2:320–324.
Einstein, A. and Infeld, L. (1949). On the motion of particles in general relativity
theory. Canadian Journal of Mathematics, 1:209–241.
Einstein, A., Podolsky, B., and Rosen, N. (1935). Can quantum-mechanical
description of physical reality be considered complete? Philosophical Review,
47:777–780.
Esfeld, M. (2001). Holism in philosophy of mind and philosophy of physics. Dor-
drecht: Kluwer.
Esfeld, M. (2004). Quantum entanglement and a metaphysics of relations. Studies in
History and Philosophy of Modern Physics, 35:601–617.
Esfeld, M. (2009). The modal nature of structures in ontic structural realism. Inter-
national Studies in the Philosophy of Science, 23:179–194.
Esfeld, M. (2014a). The primitive ontology of quantum physics: guidelines for an
assessment of the proposals. Studies in History and Philosophy of Modern
Physics, 47:99–106.
Esfeld, M. (2014b). Quantum Humeanism, or: physicalism without properties. Phil-
osophical Quarterly, 64(256):453–470.
Esfeld, M. (2015). Bell’s theorem and the issue of determinism and indeterminism.
Foundations of Physics, 45:471–482.
Esfeld, M. and Gisin, N. (2014). The GRW flash theory: a relativistic quantum
ontology of matter in space-time? Philosophy of Science, 81:248–264.
Esfeld, M. and Lam, V. (2008). Moderate structural realism about space-time.
Synthese, 160(1):27–46.
Esfeld, M. and Lam, V. (2011). Ontic structural realism as a metaphysics of objects.
In A. and Bokulich, P., editors, Scientific structuralism, pages 143–159. Dor-
drecht: Springer.
Esfeld, M. and Lam, V. (2013). A dilemma for the emergence of spacetime in canon-
ical quantum gravity. Studies in History and Philosophy of Modern Physics,
44:286–293.
Esfeld, M., Lazarovici, D., Hubert, M., and Dürr, D. (2014). The ontology of
Bohmian mechanics. British Journal for the Philosophy of Science, 65(4):773–796.
Everett, H. (1957). “Relative state” formulation of quantum mechanics. Reviews of
Modern Physics, 29(3):454–462. Reprinted in DeWitt, B. S. and Graham, N.,
editors (1973). The Many-Worlds Interpretation of Quantum Mechanics,
pages 141–149. Princeton: Princeton University Press.
Feynman, R. P. (1966). The development of the space-time view of quantum electro-
dynamics. Nobel Lecture, December 11, 1965. Science, 153:699–708.
Feynman, R. P., Leighton, R. B., and Sands, M. (1963). The Feynman lectures on
physics. Volume 1. Reading (Massachusetts): Addison-Wesley.
Field, H. H. (1980). Science without numbers. A defence of nominalism. Oxford:
Blackwell.
Field, H. H. (1985). Can we dispense with space-time? In Asquith, P. D. and
Kitcher, P., editors, Proceedings of the 1984 Biennial Meeting of the Philosophy
of Science Association. Volume 2, pages 33–90. East Lansing: Philosophy of
Science Association.
Fierz, H. and Scharf, G. (1979). Particle interpretation for external field problems in
QED. Helvetica Physica Acta. Physica Theoretica, 52(4):437–453.
Floridi, L. (2008). A defence of informational structural realism. Synthese,
161:219–253.
172 Bibliography
Fokker, A. D. (1929). Ein invarianter Variationssatz für die Bewegung mehrerer
elektrischer Massenteilchen. Zeitschrift für Physik, 58(5):386–393.
Foster, J. (1982). The case for idealism. London: Routledge.
Frankel, T. (1997). The geometry of physics. Cambridge: Cambridge University
Press.
French, S. (2010). The interdependence of objects, structure and dependence. Synth-
ese, 175:89–109.
French, S. (2014). The structure of the world. Metaphysics and representation.
Oxford: Oxford University Press.
Frisch, M. (2005). Inconsistency, asymmetry, and non-locality. A philosophical
investigation of classical electrodynamics. New York: Oxford University Press.
Gauss, C. F. (1867). Brief an Wilhelm Weber 19. März 1845. In Werke. Mathema-
tische Physik. Band 5, pages 627–629. Göttingen: Universitätsdruckerei.
Geach, P. (1965). Some problems about time. Proceedings of the British Academy,
51:321–336.
Gerhardt, C. I., editor (1890). Die philosophischen Schriften von G. W. Leibniz.
Band 7. Berlin: Weidmannsche Verlagsbuchhandlung.
Ghirardi, G. C., Grassi, R., and Benatti, F. (1995). Describing the macroscopic
world: closing the circle within the dynamical reduction program. Foundations
of Physics, 25(1):5–38.
Ghirardi, G. C., Pearle, P., and Rimini, A. (1990). Markov processes in Hilbert
space and continuous spontaneous localization of systems of identical particles.
Physical Review A, 42:78–89.
Ghirardi, G. C., Rimini, A., and Weber, T. (1986). Unified dynamics for micro-
scopic and macroscopic systems. Physical Review D, 34(2):470–491.
Gisin, N. (1984). Quantum measurements and stochastic processes. Physical Review
Letters, 52:1657–1660 (and reply p. 1776).
Gisin, N. (1989). Stochastic quantum dynamics and relativity. Helvetica Physica
Acta, 62:363–371.
Goldstein, S. (2012). Probability in physics, chapter Typicality and notions of prob-
ability in physics, pages 59–71. Berlin: Springer.
Goldstein, S. and Struyve, W. (2007). On the uniqueness of quantum equilibrium in
Bohmian mechanics. Journal of Statistical Physics, 128(5):1197–1209.
Goldstein, S., Taylor, J., Tumulka, R., and Zanghı̀, N. (2005a). Are all particles
identical? Journal of Physics A: Mathematical and General, 38(7):1567–1576.
Goldstein, S., Taylor, J., Tumulka, R., and Zanghı̀, N. (2005b). Are all particles
real? Studies in History and Philosophy of Modern Physics, 36(1):103–112.
Gomes, H., Gryb, S., and Koslowski, T. (2011). Einstein gravity as a 3D confor-
mally invariant theory. Classical and Quantum Gravity, 28:045005.
Gomes, H. and Koslowski, T. (2012). Coupling shape dynamics to matter gives
spacetime. General relativity and gravitation, 44(6):1539–1553.
Gomes, H. and Koslowski, T. (2013). Frequently asked questions about shape
dynamics. Foundations of Physics, 43(12):1428–1458.
Graham, D. W. (2010). The texts of early Greek philosophy. The complete frag-
ments and selected testimonies of the major Presocratics. Edited and translated
by Daniel W. Graham. Cambridge: Cambridge University Press.
Gravejat, P., Hainzl, C., Lewin, M., and Séré, E. (2013). Construction of the Pauli–
Villars-regulated Dirac vacuum in electromagnetic fields. Archive for Rational
Mechanics and Analysis, 208(2):603–665.
Greiner, W. and Bromley, D. A. (2000). Relativistic quantum mechanics. Third
edition. Wave equations. Berlin: Springer.
Gryb, S. and Thébault, K. P. Y. (2016). Time remains. British Journal for the Phi-
losophy of Science, 67:663–705.
 Bibliography 173
Guay, A. and Pradeu, T. (2017). Right out of the box: How to situate metaphysics
of science with regard to other metaphysical approaches. Synthese, forthcoming.
Hacking, I. (1975). The identity of indiscernibles. Journal of Philosophy, 72:249–
256.
Hall, M. J. W., Deckert, D.-A., and Wiseman, H. M. (2014). Quantum phenomena
modelled by interactions between many classical worlds. Physical Review X,
4:041013.
Hall, N. (2009). Humean reductionism about laws of nature. Unpublished manu-
script. http://philpapers.org/rec/halhra.
Halvorson, H. and Clifton, R. K. (2002). No place for particles in relativistic
quantum theories? Philosophy of Science, 69:1–28.
Hoffmann-Kolss, V. (2010). The metaphysics of extrinsic properties. Frankfurt
(Main): Ontos.
Holland, P. (2001a). Hamiltonian theory of wave and particle in quantum mechan-
ics I: Liouville’s theorem and the interpretation of the de Broglie-Bohm theory. Il
Nuovo Cimento B, 116:1043–1070.
Holland, P. (2001b). Hamiltonian theory of wave and particle in quantum mechan-
ics II: Hamilton-Jacobi theory and particle back-reaction. Il Nuovo Cimento B,
116:1143–1172.
Huggett, N. (2006). The regularity account of relational spacetime. Mind, 115
(457):41–73.
Huggett, N. and Wüthrich, C. (2013). Emergent spacetime and empirical (in)coher-
ence. Studies in History and Philosophy of Modern Physics, 44:276–285.
Hume, D. (1739). A treatise of human nature.
Hume, D. (1748). Enquiries concerning human understanding and concerning the
principles of morals.
Humphreys, P. (2013). Scientific ontology and speculative ontology. In Ross, D.,
Ladyman, J., and Kincaid, H., editors, Scientific metaphysics, pages 51–78.
Oxford: Oxford University Press.
Jackson, F. (1994). Armchair metaphysics. In Michael, M. and Hawthorne, J. O.,
editors, Philosophy in mind. The place of philosophy in the study of mind,
pages 23–42. Dordrecht: Kluwer.
Jackson, F. (1998). From metaphysics to ethics. A defence of conceptual analysis.
Oxford: Oxford University Press.
Keränen, J. (2001). The identity problem for realist structuralism. Philosophia
Mathematica, 9(3):308–330.
Kitcher, P. (1981). Explanatory unification. Philosophy of Science, 48:501–531.
Klein, O. (1929). Die Reflexion von Elektronen an einem Potentialsprung nach der
relativistischen Dynamik von Dirac. Zeitschrift für Physik, 53(3-4):157–165.
Kuhlmann, M. (2010). The ultimate constituents of the material world. In search of
an ontology for fundamental physics. Frankfurt (Main): Ontos.
Ladyman, J. (2007). On the identity and diversity of objects in a structure. Proceed-
ings of the Aristotelian Society. Supplementary Volume, 81(1):23–43.
Ladyman, J. and Bigaj, T. (2010). The principle of the identity of indiscernibles and
quantum mechanics. Philosophy of Science, 77:117–136.
Ladyman, J. and Ross, D. (2007). Every thing must go: metaphysics naturalized.
New York: Oxford University Press.
Lanczos, C. (1970). The variational principles of mechanics. Fourth edition. Univer-
sity of Toronto Press.
Lange, M. (2002). An introduction to the philosophy of physics: locality, fields,
energy and mass. Oxford: Blackwell.
Lange, M. (2013). Grounding, scientific explanation, and Humean laws. Philosoph-
ical Studies, 164:255–261.
174 Bibliography
Langton, R. and Lewis, D. (1998). Defining ‘intrinsic’. Philosophy and Phenomeno-
logical Research, 58:333–345.
Laudisa, F. (2014). Against the ‘no-go’ philosophy of quantum mechanics. Euro-
pean Journal for Philosophy of Science, 4:1–17.
Lazarovici, D. (2014). A relativistic retrocausal model violating Bell’s inequality.
Proceedings of the Royal Society A, 471:DOI 10.1098/rspa.2014.0454.
Lazarovici, D. and Reichert, P. (2015). Typicality, irreversibility and the status of
macroscopic laws. Erkenntnis, 80(4):689–716.
LeBihan, B. (2016). Super-relationalism: combining eliminativism about objects and
relationism about spacetime. Philosophical Studies, 173:2151–2172.
Lehmkuhl, D. (2016). The metaphysics of super-substantivalism. Noûs, DOI
10.1111/nous.12163.
Lewis, D. (1973a). Causation. Journal of Philosophy, 70:556–567.
Lewis, D. (1973b). Counterfactuals. Oxford: Blackwell.
Lewis, D. (1986a). On the plurality of worlds. Oxford: Blackwell.
Lewis, D. (1986b). Philosophical papers, volume 2. Oxford: Oxford University
Press.
Lewis, D. (1994). Humean supervenience debugged. Mind, 103(412):473–490.
Reprinted in Lewis, D. (1999). Papers in Metaphysics and Epistemology, pages
224–247. Cambridge: Cambridge University Press.
Lewis, D. (2004). Causation as influence. In Collins, J., Hall, N., and Paul, L. A.,
editors, Causation and counterfactuals, pages 75–106. Cambridge, Massachu-
setts: MIT Press.
Lewis, D. (2009). Ramseyan humility. In Braddon-Mitchell, D. and Nola, R.,
editors, Conceptual Analysis and Philosophical Naturalism, pages 203–222.
Cambridge, Massachusetts: MIT Press.
Lewis, P. J. (1997). Quantum mechanics, orthogonality, and counting. British
Journal for the Philosophy of Science, 48:313–328.
Locke, J. (1690). An essay concerning human understanding.
Loewer, B. (1996). Freedom from physics: quantum mechanics and free will. Phil-
osophical Topics, 24(2):91–112.
Loewer, B. (2007). Counterfactuals and the second law. In Price, H. and Corry, R.,
editors, Causation, physics, and the constitution of reality. Russell’s republic
revisited, pages 293–326. Oxford: Oxford University Press.
Loewer, B. (2012). Two accounts of laws and time. Philosophical Studies, 160:115–
137.
Mach, E. (1919). The science of mechanics: a critical and historical account of its
development. Fourth edition. Translation by Thomas J. McCormack. Chicago:
Open Court.
Malament, D. B. (1977). The class of continuous timelike curves determines the
topology of spacetime. Journal of Mathematical Physics, 18(7):1399–1404.
Marshall, D. (2015). Humean laws and explanation. Philosophical Studies,
172:3145–3165.
Martens, N. (2017). Regularity comparativism about mass in Newtonian gravity.
Philosophy of Science, 84(5).
Maudlin, T. (2007a). The metaphysics within physics. New York: Oxford Univer-
sity Press.
Maudlin, T. (2007b). What could be objective about probabilities? Studies in
History and Philosophy of Modern Physics, 38:275–291.
Maudlin, T. (2010). Can the world be only wave-function? In Saunders, S., Barrett,
J., Kent, A., and Wallace, D., editors, Many worlds? Everett, quantum theory,
and reality, pages 121–143. Oxford: Oxford University Press.
 Bibliography 175
Maudlin, T. (2011). Quantum non-locality and relativity. Third edition. Chichester:
Wiley-Blackwell.
Maudlin, T. (2014). Critical study of David Wallace, The emergent multiverse:
quantum theory according to the Everett interpretation, Oxford University
Press 2012. Noûs, 48(4):794–808.
Maudlin, T. (2015). The universal and the local in quantum theory. Topoi, 34:349–
358.
McKenzie, K. (2014). Priority and particle physics: ontic structural realism as a fun-
damentality thesis. British Journal for the Philosophy of Science, 65:353–380.
Mickelsson, J. (2014). The phase of the scattering operator from the geometry of
certain infinite-dimensional groups. arXiv:1403.4340 [math-ph].
Miller, E. (2014). Quantum entanglement, Bohmian mechanics, and Humean super-
venience. Australasian Journal of Philosophy, 92:567–583.
Miller, E. (2015). Humean scientific explanation. Philosophical Studies, 172:1311–
1332.
Misner, C. W., Thorne, K. S., and Wheeler, J. A. (1973). Gravitation. San Francisco:
Freeman.
Monton, B. (2004). The problem of ontology for spontaneous collapse theories.
Studies in History and Philosophy of Modern Physics, 35(3):407–421.
Morganti, M. (2013). Combining science and metaphysics. Contemporary physics,
conceptual revision and common cause. Basingstoke: Palgrave Macmillan.
Muller, F. A. (2011). How to defeat Wüthrich’s abysmal embarassment argument
against space-time structuralism. Philosophy of Science, 78:1046–1057.
Mulligan, K. (2012). Beyond objects, properties and relations. Handout of a lecture
given at the University of Lausanne.
Mundy, B. (1989). Distant action in classical electromagnetic theory. British Journal
for the Philosophy of Science, 40(1):39–68.
Newton, I. (1952). Opticks. Edited by I. B. Cohen. New York: Dover.
Ney, A. (2012). Neo-positivist metaphysics. Philosophical Studies, 160:53–78.
Ney, A. (2015). Fundamental physical ontologies and the constraint of empirical
coherence: a defense of wave function realism. Synthese, 192(10):3105–3124.
Norsen, T. (2005). Einstein’s boxes. American Journal of Physics, 73:164–176.
Pickl, P. (2011). A simple derivation of mean field limits for quantum systems.
Letters in Mathematical Physics, 97(2):151–164.
Pooley, O. (2001). Relationalism rehabilitated? II: Relativity. http://philsci-archive.
pitt.edu/221/.
Pooley, O. (2013). Substantivalist and relationalist approaches to spacetime. In Bat-
terman, R., editor, The Oxford handbook of philosophy of physics, pages 522–
586. Oxford: Oxford University Press.
Pooley, O. and Brown, H. (2002). Relationalism rehabilitated? I: Classical mechan-
ics. British Journal for the Philosophy of Science, 53:183–204.
Price, H. (1996). Time’s arrow and Archimedes’ point. New directions for the
physics of time. Oxford: Oxford University Press.
Pylkkänen, P., Hiley, B. J., and Pättiniemi, I. (2015). Bohm’s approach and individ-
uality. In Guay, A. and Pradeu, T., editors, Individuals across the sciences,
chapter 12, pages 226–246. Oxford: Oxford University Press.
Quine, W. V. O. (1948). On what there is. Review of Metaphysics, 2:21–38.
Quine, W. V. O. (1969). Ontological relativity and other essays. New York: Coulm-
bia University Press.
Ritz, W. (1908). Recherches critiques sur l’électrodynamique générale. Annales de
chimie et de physique, 8(13):145–209.
Robinson, H. (1982). Matter and sense. A critique of contemporary materialism.
Cambridge: Cambridge University Press.
176 Bibliography
Rohrlich, F. (1997). The dynamics of a charged sphere and the electron. American
Journal of Physics, 65(11).
Ross, D., Ladyman, J., and Kincaid, H., editors (2013). Scientific metaphysics.
Oxford: Oxford University Press.
Rovelli, C. (1997). Halfway through the woods: contemporary research on space
and time. In Earman, J. and Norton, J., editors, The cosmos of science, pages
180–223. Pittsburgh: University of Pittsburgh Press.
Ruetsche, L. (2011). Interpreting quantum theories. The art of the possible. Oxford:
Oxford University Press.
Ruijsenaars, S. N. M. (1977). Charged particles in external fields. I. Classical theory.
Journal of Mathematical Physics, 18(4):720–737.
Russell, B. (1912). On the notion of cause. Proceedings of the Aristotelian Society,
13:1–26.
Russell, B. (1927). The analysis of matter. London: Routledge.
Saunders, S. (1991). The negative-energy sea. In Saunders, S. and Brown, H. R.,
editors, The philosophy of vacuum, pages 65–109. Oxford: Oxford University
Press.
Saunders, S. (1999). The ‘beables’ of relativistic pilot-wave theory. In Butterfield, J.
and Pagonis, C., editors, From physics to philosophy, pages 70–89. Cambridge:
Cambridge University Press.
Saunders, S. (2006). Are quantum particles objects? Analysis, 66:52–63.
Saunders, S. (2013). Rethinking Newton’s Principia. Philosophy of Science, 80
(1):22–48.
Schaffer, J. (2010a). The internal relatedness of all things. Mind, 119(474):341–376.
Schaffer, J. (2010b). Monism: the priority of the whole. Philosophical Review, 119
(1):31–76.
Scharf, G. (1995). Finite quantum electrodynamics: the causal approach. Second
edition. Berlin: Springer.
Schiff, J. and Poirier, B. (2012). Communication: quantum mechanics without
wavefunctions. Journal of Chemical Physics, 136(3):031102.
Schrödinger, E. (1930). Über die kräftefreie Bewegung in der relativistischen Quan-
tenmechanik. Sitzungsberichte der Preußischen Akademie der Wissenschaften.
Physikalisch-Mathematische Klasse, pages 418–428.
Schwarzschild, K. (1903). Zur Elektrodynamik. ii. Die elementare elektrodyna-
mische Kraft. Nachrichten von der Gesellschaft der Wissenschaften zu Göt-
tingen, Mathematisch-Physikalische Klasse, 1903:132–141.
Sebens, C. (2015). Quantum mechanics as classical physics. Philosophy of Science,
82:266–291.
Seevinck, M. P. and Uffink, J. (2011). Not throwing out the baby with the bath-
water: Bell’s condition of local causality mathematically ‘sharp and clean’. In
Dieks, D., Gonzalez, W. J., Hartmann, S., Uebel, T., and Weber, M., editors,
Explanation, prediction and confirmation, pages 425–450. Dordrecht: Springer.
Sellars, W. (1956). Empiricism and the philosophy of mind. In Feigl, H. and Scriven,
M., editors, The foundations of science and the concepts of psychology and psy-
choanalysis. Minnesota Studies in the Philosophy of Science. Volume 1, pages
253–329. Minneapolis: University of Minnesota Press.
Sellars, W. (1962). Philosophy and the scientific image of man. In Colodny, R.,
editor, Frontiers of science and philosophy, pages 35–78. Pittsburgh: University
of Pittsburgh Press.
Sklar, L. (1974). Space, time and spacetime. Berkeley: University of California Press.
Skow, B. (2015). Objective becoming. Oxford: Oxford University Press.
Spohn, H. (2000). The critical manifold of the Lorentz-Dirac equation. EPL (Euro-
physics Letters), 50(3):287.
 Bibliography 177
Strawson, G. (1989). The secret connexion. Causation, realism, and David Hume.
Oxford: Oxford University Press.
Struyve, W. (2010). Pilot-wave approaches to quantum field theory. Journal of
Physics: Conference Series, 306:012047.
Struyve, W. and Westman, H. (2006). A new pilot-wave model for quantum field
theory. AIP Conference Proceedings, 844:321–339.
Struyve, W. and Westman, H. (2007). A minimalist pilot-wave model for quantum
electrodynamics. Proceedings of the Royal Society A, 463:3115–3129.
Suárez, M. (2015). Bohmian dispositions. Synthese, 192:3203–3228.
Sutherland, R. I. (2008). Causally symmetric Bohm model. Studies in History and
Philosophy of Modern Physics, 39:782–805.
Tetrode, H. (1922). Über den Wirkungszusammenhang der Welt. Eine Erweiterung
der klassischen Dynamik. Zeitschrift für Physik, 10(1):317–328.
Tumulka, R. (2006). A relativistic version of the Ghirardi-Rimini-Weber model.
Journal of Statistical Physics, 125(4):821–840.
Tumulka, R. (2009). The point processes of the GRW theory of wave function col-
lapse. Reviews in Mathematical Physics, 21:155–227.
Tumulka, R. (2011). Paradoxes and primitive ontology in collapse theories of
quantum mechanics. arXiv:1102.5767 [quant-ph].
Unruh, W. G. (1976). Notes on black-hole evaporation. Phys. Rev. D, 14:870–892.
Valentini, A. (1992). On the pilot-wave theory of classical, quantum and subquan-
tum physics. Trieste: International School for Advanced Studies, PhD thesis.
Vassallo, A. (2015). Can Bohmian mechanics be made background independent?
Studies in History and Philosophy of Modern Physics, 52:242–250.
Vassallo, A. and Ip, P. H. (2016). On the conceptual issues surrounding the notion
of relational Bohmian dynamics. Foundations of Physics, 46(8):943–972.
Wallace, D. (2008). Philosophy of quantum mechanics. In Rickles, D., editor, The
Ashgate companion to contemporary philosophy of physics, pages 16–98. Alder-
shot: Ashgate.
Wallace, D. (2012). The emergent multiverse. Quantum theory according to the
Everett interpretation. Oxford: Oxford University Press.
Wallace, D. (2014). Life and death in the tails of the GRW wave function.
arXiv:1407.4746 [quant-ph].
Weyl, H. (1949). Philosophy of mathematics and natural science. Princeton: Prince-
ton University Press.
Wheeler, J. A. (1962). Geometrodynamics. New York: Academic Press.
Wheeler, J. A. and Feynman, R. P. (1945). Interaction with the absorber as the
mechanism of radiation. Reviews of Modern Physics, 17:157–181.
Wilson, A. (2013a). Objective probability in Everettian quantum mechanics. British
Journal for the Philosophy of Science, 64:709–737.
Wilson, M. (2013b). What can contemporary philosophy learn from our ‘scientific
philosophy’ heritage? In Ross, D., Ladyman, J., and Kincaid, H., editors, Scien-
tific metaphysics, pages 151–181. Oxford: Oxford University Press.
Wüthrich, C. (2009). Challenging the spacetime structuralist. Philosophy of Science,
76:1039–1051.
Zeh, H. D. (2010). The physical basis of the direction of time. Fifth edition. Berlin:
Springer.
Index

absolute, 6, 9, 13, 14, 20, 21, 23–5, Black, R., 46

27–8, 31, 34, 42, 51, 59, 61–2, Blackburn, S., 21
64–5, 67–8, 73, 78, 84, 96, 102–3, block universe, 149–51, 155, 157
123, 125–7, 134, 147–8, 149–50, Bohm, D., 10, 70, 71, 81, 101, 105
158 Bohmian mechanics, 10–12, 15, 27,
absorber, 142 49, 70–8, 81, 83–5, 92, 93,
absorber condition, 143 95–7, 99, 104, 105, 107–9, 117–8,
advanced action, 137, 139, 142–3, 128, 156, 163–5
151, 165–6 Boltzmann, L., 89
Albert, D., 10, 16, 80, 85–6 Briceno, S., 55
Allori, V., 6, 31, 33, 70, 78–9 Bromley, D. A., 116
Anaximander, 31 Brown, H., 64, 65
Anaximenes, 31 Butterfield, J., 39
Anderson, E., 67
Aristotle, 1, 7, 25 Callender, C., 16, 44, 49, 93
Armstrong, D., 45 Cartesian, 4, 27, 42, 48, 66, 162
Arntzenius, F., 31, 144 Castañeda, H.-N., 57
arrow of time, 142 Cei, A., 54
asymptotic particle states, 104 Chakravartty, A., 16
asymptotic series, 102–3 Clifton, R. K., 100
atomism, 4–7, 9, 15, 21, 28, 31–2 Cohen, J., 44
Colin, S., 105, 107, 110, 112
Baker, D. J., 100 consciousness, 8, 44
Barbour, J., 31, 64–5 Corry, R., 165
Barrett, J., 6, 100, 164 Cowan, C. W., 97
Bauer, G., 141 Cramer, J., 165
Baumgartner, M., 45 Curceanu, C., 84
Bedingham, D., 164
Beebee, H., 44, 50 Davidson, D., 40
Bell’s theorem, 81, 128, 163, 165 Dawid, R., 16
Bell, J. S., 5, 16, 78, 81, 83, 101, 104, de Broglie, L., 10, 70, 81
126, 163 Deckert, D.-A., 15, 144–5
Belot, G., 6, 26, 64, 67, 69, 73, 97 Democritus, 4–8, 17, 20
Belousek, D., 72 Dennett, D., 135
Benedikter, N., 122 determinism, 41, 50–1, 163, 171
Bertotti, B., 31, 64–5 Dickson, M., 49, 129
Bhogal, H., 43, 49 Dieks, D., 56, 150, 153
Bigaj, T., 56 Dirac sea, 11, 15, 104–8, 110–12, 115,
Bird, A., 52 120, 122, 124, 126–9, 133, 169
180 Index
Dirac, P., 101, 104, 107, 112, 141 Gauss, C. F., 141
discernibility, 25–7, 47, 56, 168 Geach, P., 149
dispositions, 52–4, 72, 135 Ghirardi, G. C., 78–9, 81, 83
Dizadji-Bahmani, F., 16 Gisin, N., 78, 164
Dowe, P., 165 Goldstein, S., 90, 93
Dowker, F., 57 Gomes, H., 156–7
Dürr, D., 10, 16, 33, 70, 72, 74, 83, Gravejat, P., 102
93, 96, 101, 109, 128, 164, 170 Greiner, W., 116
dynamical structure, 3, 8, 11–2, 14, GRW theory, 31–2, 57, 77–84, 97,
39, 40–3, 51, 54–6, 59, 60, 62–3, 135, 163–4
71–2, 74, 84–7, 104, 106, 117, Gryb, S., 157, 162
131, 133, 135, 137–8, 141, 146–8, Guay, A., 16
162, 164–5, 166 gunk, 31–2, 78, 83, 135

Earman, J., 67, 69, 143–4, 153 Hacking, I., 26

Egg, M., 80, 81, 163 Hall, M. J. W., 57
Einstein, A., 83, 153–4 Hall, N., 42, 44, 48, 50, 62, 69
electrodynamics, 9, 48, 53, 123, Halvorson, H., 100
133–5, 137–8, 140–2, 144–5, Hartenstein, V., 145
147–8, 151, 154, 165 Hawthorne, J., 31
electron sector, 104, 105–6, 128 Heraclitus, 151
electron-positron pair, 104, 106, 124 Herbauts, I., 57
electron-positron pair-creation, 104, Hiley, B. J., 105
106, 124 Hoffmann-Kolss, V., 17
entanglement, 6, 39, 41, 50, 54–5, holism, 7, 40, 85
83–4, 109, 114, 133, 162, 175 Holland, P., 73
Esfeld, M., 7, 15, 25, 40, 49, 54–5, Huggett, N., 49, 60, 69, 73, 153,
80–1, 84, 161, 163–4 158, 161
Euclidean geometry, 38, 51–2, 60–1, Hume, D., 45
68, 131 Humean mosaic, 8, 11, 14, 44, 46–7,
Everett, H., 10 49, 61, 137, 145, 154, 156, 158–61
Humeanism, 8, 11, 14–5, 41, 44–50,
Feynman, R., 9, 134, 141–2, 144 52–3, 55–7, 62, 72–3, 82, 97, 99,
field, 2, 5–6, 9–11, 13, 32, 35–7, 39, 133, 138, 152–3, 156, 163, 166, 171
41, 43, 45, 47, 51, 53–4, 59–60, humility, 46–7, 174
64, 70–2, 78–87, 96–7, 99–100, Humphreys, P., 16
101–2, 107–8, 110, 114, 116, 118, Huygens, C., 67–8
120, 124, 129, 133–42, 144–5,
148, 151–5, 157–8, 160–1, 163–5 identical particles, 74, 172
field theory, 6, 9–11, 99, 102, 129, individuals, 12, 25, 54
133–4, 136, 140–1, 144–5, 153, Infeld, L., 153
155, 157, 161, 163–4 infrared divergence, 106
Field, H., 9, 23, 134 intrinsic properties, 7, 8, 17–9, 21, 25,
Fierz, H., 101, 125 33, 39, 45–7, 49, 51–5, 57, 74, 76,
flashes, 57, 83–4, 164 132, 162
Floridi, L., 25 Ip, P. H., 73
Fock space, 107, 118, 122–3, 127–8
Fokker, A. D., 141–2 Jackson, F., 2, 13, 18, 25, 46
Foster, J., 21
Frankel, T., 64 Keränen, J., 26
free will, 51, 174 Kitcher, P., 40
French, S., 7, 21, 25, 54–5, 162 Klein’s paradox, 109
Frisch, M., 141, 144 Klein, O., 109
 Index 181
Koslowski, T., 156–7 Newtonian mechanics, 2–3, 7, 9,
Kuhlmann, M., 100 13, 17, 34, 50, 59, 60, 62–3,
65–6, 69, 95, 137–8, 147–8,
Ladyman, J., 7, 12, 16, 25, 55–6 158, 161
Lam, V., 7, 15, 25, 161 Ney, A., 16
Lanczos, C., 66 non-locality, 53–4, 163–5, 175
Landau pole, 102 Norsen, T., 81
Lange, M., 53, 134 Norton, J., 153
Langton, R., 17
Laudisa, F., 12 Oldofredi, A., 15
Lazarovici, D., 90, 166 ontic structural realism, 7, 11, 25–6,
LeBihan, B., 57 54–6, 175
Lehmkuhl, D., 27
Leibniz, G. W., 167 Parmenides, 28, 101
Leucippus, 4–5, 8 parsimony, 3, 13, 33–4, 50, 69, 86,
Lewis, D., 17, 21, 44–7 157, 166
Lewis, P., 80 particle creation/annihilation, 10–11,
light, 41, 70, 84, 106–7, 135–7, 139, 100–5, 112, 122–4, 126, 128
141, 144–6, 148, 163, 165–6, 169 particle number operators, 125–7
Locke, J., 19 permutation invariance, 23
Loewer, B., 15, 44, 45, 51, 53, 80 Perry, Z., 43, 49
Lorentz, H., 102, 107, 125–6, 128, perturbative QFT, 102
138–41, 145–7, 164, 176 Pickl, P., 112
Poirier, B., 57
Mach, E., 7, 17, 31, 64, 147 Pooley, O., 16, 64–5, 153
Marshall, D., 53 positron, 76, 100, 104, 106, 117,
Martens, N., 48 124, 170
matter density field, 78–9, 80–1, Pradeu, T., 16
83, 97 presentism, 152
Maudlin, T., 5, 16, 28, 51, 53, 78, 80, Price, H., 144, 165–6
84, 93, 163 primitive ontology, 6, 8, 10–14,
Maxwell, J. C., 89–90, 95, 116, 16, 31, 33, 42–4, 46, 49–50,
137–41, 145, 147 70, 74, 78, 84, 86, 97, 99, 116,
McKenzie, K., 25 156, 164
measurement problem, 6, 10, 13, 27, primitive thisness, 19, 21, 132
70, 78–80, 84, 97, 99–100, 129, probabilities, 12, 41, 50, 83, 91–2,
156, 162, 164 94–5, 97, 164, 175
Mele, A., 44, 50 Pylkkänen, P., 18, 73
Mickelsson, J., 102
Miller, E., 49, 53 quantum gravity, 6, 22, 161–2, 171
Misner, C, 27 quidditism, 46–7, 168
modality, 56 Quine, W. V. O., 12, 25
Molière, J.-B., 52, 54
Monton, B., 80 radiation, 100, 105, 107–8, 116, 123,
Morganti, M., 16 128, 136–8, 140–5
Muller, F. A., 56 radiation catastrophe, 116
Mulligan, K., 24 radiation damping, 142, 144
Mumford, S., 55 reductionism, 8, 44, 173
Mundy, B., 16, 134, 142 Reichert, P., 90
relationalism, 6, 11, 15, 21, 24–5,
natural philosophy, 1, 4, 12–3 31, 46, 48, 57, 61, 65, 68–9,
negative energy states, 109, 121, 125 138, 149, 152–3, 157, 160–3,
Newton, I., 4 166, 174
182 Index
relativistic physics, 6, 11, 15, 131, Sutherland, R. I., 165
133, 145, 147, 150, 152, 156–7, symmetry, 26, 107, 112, 127, 156
159, 161–3, 166, 170
renormalization, 9, 101–2, 104, 107, Tetrode, H., 141
110, 141, 144 Teufel, S., 74, 93
retarded action, 142, 166 Thébault, K. P. Y., 16, 157, 162
Rimini, A., 78 Thales, 31
Ritz, W., 142 thermodynamics, 3, 69, 111
Robinson, H., 21 truth-maker, 23, 45, 49–50,
Rohrlich, F., 141 133, 136
Ross, D., 7, 12, 16, 25, 55 Tumulka, R., 80, 83, 97, 164
Rovelli, C., 153 typicality, 87, 90–3, 95
Ruetsche, L., 100
Ruijsenaars, S. N. M., 127 Uffink, J., 163
Russell, B., 49, 53 ultraviolet divergence, 102, 106, 140
underdetermination, 8, 155, 166
Saunders, S., 56, 64, 67–8, 73, 105, 109 Unruh effect, 101, 126
Schaffer, J., 28 Unruh, W. G., 126
Scharf, G., 101–2, 125
Schiff, J., 57 vacuum, 101–2, 104–5, 111–2,
Schrödinger, E., 109 114–5, 117–8, 120–1, 123–7,
Schwarzschild, K., 141 129, 137, 176
Sebens, C., 57 vacuum excitations, 104, 112, 118
Seevinck, M., 163 vacuum polarization, 102, 120–1
self-interaction, 134, 138, 140–2, 145 vacuum polarization current, 102
Sellars, W., 40, 136 Valentini, A., 101
shape dynamics, 156–7, 172 Vassallo, A., 15, 73
simplicity, 8, 13, 26, 33–6, 43, 50, 56, Versteegh, M. A. M., 56
64, 66, 76–7, 80, 99, 104, 106, 118,
132–3, 145, 147–8, 151–5, 160 Wallace, D., 16, 80
Sklar, L., 47, 83 wave function, 7, 9–10, 12–4, 18–9,
Skow, B., 150 39, 41, 43, 49–51, 53, 55–7,
spin, 7, 81, 107–8, 115, 117 71–81, 83–6, 93–7, 99, 104,
Spohn, H., 141 107–10, 112, 114, 118–24, 128,
standard model of QFT, 104, 107, 131, 133, 135,
110 164–5
Strawson, G., 56 weak, 2, 26, 56, 68, 94, 95, 113–4,
structural realism, 7, 11, 15, 25–6, 54, 127, 143, 170
55–6, 175 Weber, T., 78
Struyve, W., 16, 93, 101, 104–5, 107, Westman, H., 101, 104
110, 112 Weyl, H., 150
Suárez, M., 72 Wheeler, J. A., 27, 141–4
substantivalism, 21, 27, 47–8, 51, 83, Wheeler-Feynman electrodynamics,
153, 174 141, 144
substratum, 19, 21, 32, 135 Wilson, A., 16
Super-Humeanism, 11, 41, 47–50, 57, Wilson, M., 16
62, 82, 99, 133, 138, 152–3, 156, Wüthrich, C., 56, 161
163, 166
super-substantivalism, 27, 47–8, Zeh, H. D., 143
83, 174 Zitterbewegung, 109