Vous êtes sur la page 1sur 317

New Directions in the Philosophy of Science

SCIENTIFIC COMPOSITION AND


METAPHYSICAL GROUND
Edited by
Kenneth Aizawa and Carl Gillett

New Directions in the Philosophy of Science


Series Editor
StevenFrench
Department of Philosophy
University of Leeds
Leeds,United Kingdom

The philosophy of science is going through exciting times. New and


productive relationships are being sought with the history of science.
Illuminating and innovative comparisons are being developed between
the philosophy of science and the philosophy of art. The role of mathematics in science is being opened up to renewed scrutiny in the light
of original case studies. The philosophies of particular sciences are both
drawing on and feeding into new work in metaphysics and the relationships between science, metaphysics and the philosophy of science in general are being re-examined and reconfigured. The intention behind this
new series from Palgrave Macmillan is to offer a new, dedicated, publishing forum for the kind of exciting new work in the philosophy of science
that embraces novel directions and fresh perspectives. To this end, our
aim is to publish books that address issues in the philosophy of science
in the light of these new developments, including those that attempt to
initiate a dialogue between various perspectives, offer constructive and
insightful critiques, or bring new areas of science under philosophical
scrutiny.
More information about this series at
http://www.springer.com/series/14743

Kenneth Aizawa Carl Gillett


Editors

Scientific
Composition and
Metaphysical Ground

Editors
Kenneth Aizawa
Department of Philosophy
Rutgers, The State University of New Jersey
Newark, New Jersey, USA

Carl Gillett
Department of Philosophy
Northern Illinois University
DeKalb, Illinois, USA

New Directions in the Philosophy of Science


ISBN 978-1-137-56215-9
ISBN 978-1-137-56216-6
DOI 10.1057/978-1-137-56216-6

(eBook)

Library of Congress Control Number: 2016940986


The Editor(s) (if applicable) and The Author(s) 2016
The author(s) has/have asserted their right(s) to be identified as the author(s) of this work in accordance
with the Copyright, Designs and Patents Act 1988.
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made.
Cover illustration: Zoonar GmbH / Alamy Stock Photo
Printed on acid-free paper
This Palgrave Macmillan imprint is published by Springer Nature
The registered company is Macmillan Publishers Ltd.
The registered company address is: The Campus, 4 Crinan Street, London, N1 9XW, United Kingdom

Series Editors Preface

Modern metaphysics has had a bad rap in recent years. Philosophers of


science in particular have criticised and dismissed the whole field for being
irrelevant and out of touch with modern science. According to such critiques, metaphysical notions and devices such as grounding, dependence
and composition, for example, remain in thrall to a crudely Newtonian
picture of small colliding bits of matter which has long since been replaced
by the physics underpinning the Standard Model. However, and partly as
a result, the metaphysics of science has undergone something of a renaissance with numerous conferences and workshops springing up, as well as
major symposia organised by the leading philosophy of science societies.
At such events one can find not only old metaphysics being repurposed
for the current scientific environment but entirely novel metaphysical
frameworks constructed and proposed.
Aizawa and Gilletts exciting new collection focused on those metaphysical tools designed to capture vertical relations between entities
and systems of different kinds. One such device is grounding which has
become a ubiquitous umbrella term, covering relations between a variety
of entities from the scientific to the social. Schaffer and Wilson address
the issue whether there is a unified concept of grounding or whether it
breaks up into various distinct notions. Schaffer argues that the best way

vi

Series Editors Preface

to resolve this debate is to look to the relevant formalism of structural


equations and that this supports a form of grounding monism. Wilson
on the other hand rejects this methodological approach and maintains
that the kind of approach Schaffer advocates fails to distinguish crucial
features of the relations in question. Melnyk and Ney then add a further
dimension to the debate, with Melnyk suggesting that grounding is not
physicalistically acceptable and Ney defending it on the basis that it helps
to resolve certain debates in the philosophy of science and the philosophy
of mind.
Aizawa offers a bridge to the further issue of compositionality via
the notion of dimensioned realisation which allows us to develop noncausal, compositional explanations that go beyond bare grounding
claims. Gillett questions the very strategy of appropriating metaphysical
devices for the purposes of understanding scientific composition. Instead
he advocates an engagement methodology that begins by examining the
core features of scientific composition and then using this as the basis
for evaluating accounts of such features resulting from appropriation.
Harbecke also considers standard accounts of material constitution and
compares them to his regularity account of mechanistic constitution,
arguing that there is a fundamental incommensurability here and indicating possible ways this might impact on relevant ontological issues.
Franklin-Hall offers a critique of the new mechanistic approach to
explanation in science, arguing that it fails to meet certain plausible criteria, involving causation, parthood and depicting the system at the relevant level. And Pereboom argues that the vertical relations between the
mental and more fundamental levels can be captured by a non-reductive
made up of relation.
All these papers represent work that is at the very forefront of the
debate and they offer not just one but several new directions in metaphysically informed philosophy of science. As a result this volume stands
as another outstanding addition to our series.
Steven French
University of Leeds, UK

Acknowledgments

We would like to thank Dr. Jan Lewis, Dean, Faculty of Arts and Science,
Newark, and Dr. Raffaella De Rosa, Chair, Department of Philosophy,
both of Rutgers University, Newark, for financial support for the
Composition and Ground workshop held at Rutgers University, Newark,
April 1011, 2015. The papers at that workshop formed the nucleus of
the present volume. We would also like to thank Kit Fine and Ned Block
for contributing papers and commentary to the workshop, though they
could not contribute to this volume.

vii

Contents

1 Introduction: Vertical Relations in Science, Philosophy,


and the World: Understanding the New Debates
over Verticality
Kenneth Aizawa and Carl Gillett
Part I

Scientific Composition and the New Mechanism

39

2 New Mechanistic Explanation andtheNeed for


Explanatory Constraints
L. R. Franklin-Hall

41

3 Compositional Explanation: Dimensioned Realization,


New Mechanism, andGround
Kenneth Aizawa

75

4 Is Mechanistic Constitution aVersion of


Material Constitution?
Jens Harbecke

91

ix

Contents

5 Anti-Reductionism, Anti-Rationalism, andtheMaterial


Constitution oftheMental
Derk Pereboom
Part II

Grounding, Science, and Verticality in Nature

123

141

6 Ground Rules: Lessons fromWilson


Jonathan Schaffer

143

7 The Unity andPriority Arguments forGrounding


Jessica Wilson

171

8 The Metaphysics ofNature, Science, andtheRules


ofEngagement
Carl Gillett

205

9 Grounding andtheFormulation ofPhysicalism


Andrew Melnyk

249

10 Grounding inthePhilosophy ofMind: ADefense


Alyssa Ney

271

Index

301

Notes on Contributors

KennethAizawa holds a PhD in History and Philosophy of Science from the


University of Pittsburgh. He works primarily in the history and philosophy of
psychology and is the author of The Systematicity Arguments and, with Fred
Adams, The Bounds of Cognition. His most recent work has focused on embodied cognition and multiple realization. He has been Professor of Philosophy at
Rutgers University, Newark, since fall 2013.
L. R. Franklin-Hall is Associate Professor of Philosophy at New York
University. Franklin-Hall holds a PhD in Philosophy from Columbia University
and a BS in Biology from Stanford University. Franklin-Halls work, which has
appeared in Biology and Philosophy, Philosophy of Science, British Journal for the
Philosophy of Science, Philosophical Studies, among other venues, focuses on
experimental, explanatory, and classificatory practices in biology. Her research
on classification has aspired both to formulate principles that guide scientificcategorical practice and to evaluate the grounds for those principles. With
respect to scientific explanation, Franklin-Halls critical work has probed two
currently fashionable explanatory accountsexplanatory interventionism and
the new mechanistic accountwhile her constructive project has been to sketch
a theory of explanation designed to better accommodate the relatively abstract
explanations common in biology.
CarlGillett is Professor of Philosophy at Northern Illinois University. Gilletts
research specializations are in the areas of philosophy of science, metaphysics,
and philosophy of mind. Gillett has published broadly on the nature of physixi

xii

Notes on Contributors

calism, realization and multiple realization, reduction and emergence, and other
topics related to scientific composition.
Jens Harbecke is Professor of Theoretical Philosophy and Philosophy of the
Social Sciences at Witten/Herdecke University, Germany. He holds a PhD from
the University of Lausanne under the supervision of Michael Esfeld with a dissertation on the problem of mental causation. Subsequently, he worked as an
assistant professor and visiting scholar at the universities of Bern, Tel Aviv,
Jerusalem, and Washington University in St. Louis. He works on constitutive
explanations, causality in the metaphysics of mind, and counterfactual and regularity theories of causation. He is the project coordinator of a research project
funded by the European Commission on the philosophy of social science and
neuroscience with researchers at Witten/Herdecke, Helsinki, and Louvain-laNeuve. He also collaborates as a principal investigator within a philosophical
research project funded by the German-Israeli Foundation on causation and
computation in neuroscience with partners in Jerusalem and Cologne.
Andrew Melnyk is Professor of Philosophy at the University of Missouri,
where he has taught since 1991. He is interested in all aspects of the philosophy
of mind and in many aspects of philosophy of science. Much of his work is unified by his attempt to formulate, explore, and argue for a comprehensive thesis
of physicalism that invokes a carefully defined relation of realization. He has also
written about other minds, naturalism in philosophy, conceptual analysis, and
the inference from conceivability to possibility. His work has appeared in Journal
of Philosophy, Nos, Philosophy and Phenomenological Research, Mind, Australasian
Journal of Philosophy, Philosophical Studies, Philosophy of Science, and Synthese,
among others. His book, A Physicalist Manifesto: Thoroughly Modern Materialism,
was published in 2003. Melnyk was educated at St. Pauls School, London, and
Oxford University.
AlyssaNey is Associate Professor of Philosophy at the University of California,
Davis. She works in metaphysics, the philosophy of physics, and the philosophy
of mind. She is the author of Metaphysics: An Introduction (2014) and co-editor
of The Wave Function: Essays in the Metaphysics of Quantum Mechanics (2013).
DerkPereboom is the Susan Linn Sage Professor of Philosophy and Ethics at
Cornell University. He works mainly on philosophy of mind and on free will.
The physicalist position Pereboom proposes in philosophy of mind, initially
defended in a number of articles, is set out in detail in Consciousness and the
Prospects of Physicalism (2011). There he develops two responses to the leading

Notes on Contributors

xiii

arguments against physicalism. The first exploits the open possibility that introspective representations fail to represent mental properties as they are in themselves; more specifically, that introspection represents phenomenal properties as
having certain characteristic qualitative natures, which these properties might
actually lack. The second response draws on the proposal that currently unknown
intrinsic properties provide categorical bases for known physical properties and
also yield an account of consciousness. The books final theme is an anti-reductionist account of physicalism. In the view Pereboom defends, the fundamental
relation between the mental and the microphysical is constitution, where this
relation is not explicated by the notion of identity. Perebooms views on free will
are developed in Living without Free Will (2001), Four Views on Free Will (2007),
Free Will, Agency, and Meaning in Life (2014), and in his articles on this issue.
His overall position is that due to general facts about the nature of the universe,
we lack the sort of free will required for moral responsibilitythat is, for our
deserving, in a fundamental sense, to be blamed or punished for immoral action,
and to be praised or rewarded for morally right action. At the same time, he
contends that a conception of life without this type of free will would not undermine morality or to our sense of meaning in life, and in certain important
respects it may even be beneficial.
Jonathan Schaffer is a Distinguished Professor of Philosophy at Rutgers
University, New Brunswick. He holds a PhD from Rutgers in 1999, and has
since worked at the University of Houston, the University of MassachusettsAmherst, the Australian National University, and St. Andrews, before returning
to Rutgers in 2011. He works mainly in metaphysics, with interests in epistemology, language, mind, and science. His On What Grounds What (2009) is
widely regarded as one of the main sources of current interest in metaphysical
grounding.
Jessica Wilson is Associate Professor of Philosophy at the University of
Toronto. Her primary research interests are in general metaphysics (especially
metametaphysics, modality, and indeterminacy) and the metaphysics of science
(especially inter-theoretic relations). Wilsons recent publications include What
is Humes Dictum, and Why Believe It? (Philosophy and Phenomenological
Research, 2010), Fundamental Determinables (Philosophers Imprint, 2011), A
Determination-based Account of Metaphysical Indeterminacy (Inquiry, 2013),
and No Work for a Theory of Grounding (Inquiry, 2014). Her book
Metaphysical Emergence is forthcoming with Oxford University Press.

1
Introduction: Vertical Relations in Science,
Philosophy, and the World: Understanding
the New Debates over Verticality1
1

KennethAizawa andCarlGillett

Much of our knowledge comes through our successful explanations and it


is striking that among these explanations we have many, from very different areas, that are underpinned by what we neutrally term vertical relations.2 To avoid begging important questions about the characterization
of these phenomena that we will shortly see are at the heart of ongoing
debates, let us simply start by highlighting some examples.

For comments on an earlier draft, we are grateful to Geoff Pynn, Jonathan Schaffer, and Craig
Warmke.
2
One might think we could use the term non-causal relations, but as we outline below this would
beg some questions since one approach to verticality in nature takes it to be causal or of a kind with
causation. As we outline below, we also do not use the terms Grounding or realization, and so
on, because each of these terms is associated with one of the competing accounts of the nature of
verticality. We therefore plump for the neutral expression vertical relation.

K. Aizawa ()
Department of Philosophy, Rutgers University, Newark, NJ, USA
C. Gillett
Northern Illinois University, DeKalb, IL, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_1

K. Aizawa and C. Gillett

Perhaps most famously, across the sciences we find explanations that


explain higher-level entities (whether individuals, properties, or processes)
in terms of lower-level entities that scientists take to compose them and
hence these explanations use vertical relations. For example, we explain the
inheritance of traits between parent organisms and their offspring using
molecules taken to compose them. We explain the refractive index of a
crystal using the properties and relations of the atoms that compose it. Or
we explain the movement of the earths surface using the tectonic plates,
and currents of magma, taken to compose the earth. We use the term compositional explanation to refer to such explanation, though philosophers
have used various names for it.3 And we term the vertical relations that such
explanations posit scientific composition relations where this includes
relations between individuals, properties/relations, and also processes.4
We could provide many more such explanations in the sciences, but
scientists are not the only ones who use vertical relations in successful
explanations. For instance, in semantics and related fields we explain the
meaning of sentences or statements, such as Dogs have sharp teeth,
in terms of vertical relations to words, such as dogs and sharp. And
in logic, set theory, or mathematics we also find successful explanations
deploying vertical relations as well. Thus, for instance, we explain features
of sets using vertical relations to the entities treated as their members.
We have recently seen two prominent bursts of research on verticality. There is a surge of work on scientific composition, and the vertical
relations posited in scientific explanations, with the rise of the so-called
New Mechanism in philosophy of science and this has spawned what we
term the neo-Causal research tradition. And we have also recently seen
the sudden rise in analytic metaphysics of a research tradition focused on
vertical relations understood as so-called Grounding relations. However,
as well as these new developments, we should also mark that there have
also been other long-standing bodies of philosophical work on verticality. For example, there is a research tradition, over a number of decades,
in philosophy of mind and philosophy of science in what we term neo3

Other terms include reductive explanation, microstructural analyses, functional explanation,


constitutive explanation, or mechanistic explanation.
4
There are other vertical relations posited in the sciences (Healey 2013), but our primary focus here
is upon the vertical relations posited in compositional explanations.

Introduction: Vertical Relations in Science, Philosophy, and...

Functionalist accounts focused on vertical relations in nature, and science,


which understands them as so-called realization relations. And there are
still other research traditions in philosophy focused on vertical relations.5
The sudden increase of interest in verticality is an important sociological development in contemporary philosophy and it has a couple of
interesting features. Although we find theoretical accounts of the nature
of verticality, or what we term V-frameworks, across a number of areas
of philosophy, including philosophy of science, philosophy of mind, and
analytic metaphysics, it is striking that such work on verticality has been
what we may term siloed within particular areas. Apparently following
recent academic trends toward specialization, discussions of verticality
are usually confined to a particular area of philosophy and have proceeded either in total isolation from, or with very little contact with,
work in other areas of philosophy.
This situation is obviously unhealthy. For example, it leaves unanswered how research, and associated V-frameworks, in different areas
of philosophy relate to one another: Are the V-frameworks in different
philosophical areas simply focused on unrelated and independent issues?
Are such accounts compatible or even complementary? Or are some of
these V-frameworks competing, rival accounts of the same phenomena?
And what should we ultimately conclude about the nature of verticality
either universally or for particular particular kinds of phenomena (i.e., in
the sciences, in nature, in mathematics, etc.)?
The driving idea of the project that eventuated in this anthology was
to rectify this unhealthy situation and begin to address these important foundational questions. The main instrument was that of bringing
together prominent writers on verticality from various philosophical
research traditions. Initially, writers from various approaches, and areas
of philosophy, were brought together at a small workshop focused on discussion of pre-circulated papers.6 And this collection draws together the
results of these interactions in revised papers from the workshop.
5

To take just one example, there is the tradition in analytic metaphysics built around formal mereological systems adapted from work on sets. The latter tradition, like the Grounding traditions,
allows that the relata of vertical relations may be causally inert, so the points we make below for the
Grounding tradition also apply to this approach and show it offers one more competing kind of
V-framework for certain object phenomena involving verticality.
6
The workshop took place at Rutgers, Newark, in April 2015, and in addition to the papers collected in this volume also included presentations by Ned Block and Kit Fine.

K. Aizawa and C. Gillett

Bringing together the proponents of various research traditions has


been highly productive. Each of the coming Chapters makes important
contributions to discussions of verticality at various philosophical levels.
Furthermore, we suggest that the discussions at the workshops, and in
later Chapters, also make it clear we have now entered into an exciting new set of debates about verticality featuring a number of fresh, but
important, questions.
As with any significant dialectical transition, we are facing some metamethodological growing pains as we slowly come to appreciate the nature
of these new debates and their requirements. To ease this transition, our
goal in the remainder of this Introduction is therefore to provide a very
simple roadmap to the basic contours of the new debates, both to facilitate future discussion and to set up a number of important issues, and
threads of discussion, in later Chapters. We should explicitly note that, as
we hope will become clear, we do not seek to adjudicate which research
tradition and its V-frameworks should be favored for particular projects.
We aim simply to map out the shape of the new debates and let the contributors try to resolve the key issues in coming Chapters. Our focus is
solely focused on facilitating more productive work in future debates, so
we also largely avoid discussion of where past discussions and accounts
might have had problems or have been unproductive.
Briefly summarized, we are moving from the previous, siloed stage of
discussions of verticality to a new phase of debate where we have a range
of distinct V-frameworks, from different areas of philosophy, that make
diverging claims about the nature of verticality. Furthermore, as we highlight below, writers from each area of philosophy are also often offering
competing accounts of the same vertical relations or phenomena involving
verticality. We have universal accounts, such as Grounding V-frameworks,
that claim all vertical relations, from the natural across to the abstract
realms, are of a kind. In contrast, we have competing positions that claim
that the vertical relations we find in nature or science are unlike those in
the abstract realm or disciplines. We have positions, such as neo-Causal
V-frameworks, that claim vertical relations in science and nature are of a
kind with causation, and we have others, like neo-Functionalist accounts,
that claim verticality in science is unlike both causation and the vertical relations in the abstract realm of mathematics or set theory. These

Introduction: Vertical Relations in Science, Philosophy, and...

approaches all make competing claims, for example, about scientific conceptions of verticality and/or the vertical relations in nature.
Inter-twined with the latter disputes, we also have competition between
accounts pitched as treatments of the genus of verticality, such as some
Grounding V-frameworks, and other accounts focused on particular species of verticality, like many neo-Functionalist V-frameworks. Initially, one
might suppose that accounts of genus and species are pursuing different,
though interconnected, set of issues. However, in our new debates various
accounts focused on genus and species are each often intended by their
proponents to answer the same questionsthus still ending up as rivals.
For example, neo-Functionalists take their V-framework to articulate the
species of vertical relation that is best taken to back compositional explanation in science or underpin the formulation of physicalism. But proponents
of Grounding V-frameworks take their accounts to illuminate a genus relation of verticality that they in turn argue is the relation best taken to back
compositional explanation or formulate physicalism. So, we have debates
not just between V-frameworks offering different accounts of the vertical
relations in same object phenomena, but also differing over whether species or genus relations provide the best accounts of such phenomena.
Furthermore, each of these competing V-frameworks has also used to
offer distinct accounts of various issues connected to verticality. For example, with regard to verticality in nature, in our new debates we have distinct treatments of physicalism, reduction versus emergence, and more.
In the remainder of the Introduction, we focus primarily on the differing
V-frameworks themselves, but we highlight where disputes over applications
are important and a number of coming Chapters address such applications.
Finding that we have competing V-frameworks leads to a range of
meta-methodological changes. Since theory appraisal is comparative, for
instance, one can only ultimately justify ones favored account of verticality after comparing it to relevant rivals. It is thus no longer the case that
metaphysicians can only read other metaphysicians, letalone only writers
in their research tradition; or that philosophers of science can stick to
work by other philosophers of science; and so on. To justify ones favored
account of verticality with regard to some object phenomenon one must
critically engage rival accounts from different areasand ones account is
not justified until one shows ones V-framework is better than these rivals.

K. Aizawa and C. Gillett

Lastly, and just as importantly, we also suggest that we are moving to


debates looking in more depth at particular object phenomena involving
verticality, whether scientific conceptions of verticality or vertical relations in the mathematical realm, and so on. The new debates seek to construct V-frameworks to provide focussed accounts of such specific vertical
relations, paying special attention to their features, and assessing whether
competing V-frameworks succeed in providing adequate accounts of
these particular phenomena involving verticality. We are thus also moving into a new phase of the debates where V-frameworks are assessed by
contact with specific object phenomena.
Given our future-oriented goals, in Part 1 we provide a more precise picture of the new debates framing key points about them using
four theses which we fill out, and support, in subsequent sections. To
better orient newcomers to the debates, in Part 2, we provide minimal
sketches of work from the three research traditions that arguably figure
most prominently in recent debates and we support the claim that our
new debates involve competing accounts. In Part 3, we then outline the
kinds of account, types of project and methods of assessment, and metajustification, that we find in the new debates. We finish, in Part 4, by
providing the reader overviews of each of the coming papers, linking to
our survey where appropriate.

Part 1: ASimple Map oftheNew Debates:


Some (Widely Acceptable?) Theses
We want to frame the broad contours of the new debates using a few
target theses. We hope these theses will be widely acceptable, but at worst
they each provide a stalking horse in needed discussions about a number
of key meta-methodological issues.
Perhaps the most obvious conclusion from bringing together proponents of V-frameworks from different areas of philosophy is that many of
their accounts are often indeed rivals making competing claims about the
same phenomena. We thus frame this thesis:
(COMPETITIVE) There are multiple V-frameworks in distinct areas of
philosophy that offer diverging accounts of the nature of at least some of

Introduction: Vertical Relations in Science, Philosophy, and...

the same vertical relations, and/or offer conflicting answers to some of the
same questions about phenomena involving verticality, and these distinct
V-frameworks are rivals.

We should mark that some V-frameworks from different areas of philosophy, when used to pursue certain projects, may well be compatible
or even complementary. Below we give some examples. But we focus on
COMPETITIVE, since it corrects unhealthy assumptions or practices in
earlier debates and has substantial consequences.
One implication of COMPETITIVE is that researchers on verticality
can no longer continue to pursue the siloed approach. Theory appraisal
and justification are always comparative where a theory is assessed by
comparison to its relevant rivals. So defenders of V-frameworks need to
acknowledge and engage rival accounts from other areasand then show
their V-frameworks are better accounts of the relevant object phenomena
than these competing treatments. So a second thesis is also plausibly true
about the new phase of discussion:
(COMPARATIVE) A V-framework, offered as an account of certain object
phenomena involving verticality, is justified only if it is shown to provide a
better account of the object phenomena than its relevant rivals regardless of
the area of philosophy in which these rivals are offered.

The significance of COMPARATIVE should be clearto justify their


views of verticality as the best account of certain object phenomena
involving verticality writers must critically engage relevant rivals in other
areas.
It is far from mysterious how one might go about justifying ones
favored V-framework as the best account of some phenomena in these
new discussions. One simply has to engage relevant rivals and show ones
favored view provides a better account of the relevant object phenomena.
However, this point brings us to the issue of what work a V-framework
must do.
A V-framework is always offered as an account of certain object phenomena involving verticality. And we use the awkward term object phenomena involving verticality, since some accounts seek to illuminate
conceptions of verticality from some discourse, or body of explanation,

K. Aizawa and C. Gillett

and so on, while other accounts focus on vertical relations in the world or
some portion of it, and some accounts seek to do both. Whatever object
phenomena a V-framework is intended to cover, the primary work of this
account is to accommodate the nature, and various features, of the object
phenomena and related evidence such as the characteristics of associated explanations and formal systems. A corollary of this requirement is
that prior work is also necessary, since one must consequently provide a
detailed treatment of the nature of the specific vertical relations involved
in the object phenomenon under consideration, the characteristics of
associated explanations, and so on.
The new debates therefore address specific object phenomena in a
range of important ways. And we are therefore led to a third thesis about
our new discussions:
(FOCUSED) A central, and necessary, element in the assessment, and justification, of a V-framework offered as an account of certain examples of
object phenomena involving verticality is whether this account does, or
does not, capture the features of the relevant vertical relations and/or associated characteristics of the successful explanations that posit them and/or
other relevant evidence.

FOCUSED frames the detailed work on specific object phenomena


involving vertical relations that is the necessary basis of the new debates.
Lastly, we suggest that many writers in the new discussions are coalescing around a common meta-justification for accounts of certain object
phenomenathat is, treatments of vertical relations in the world, rather
than some conception of verticality. The basic idea behind this metajustification focuses on successful explanations positing vertical relations
about area X of the world, uses these conceptions of verticality underpinning such explanations as a guide to the vertical relations found in the
world in area X, and justifies the latter account of such vertical relations
in this portion of the world using the success of the relevant explanations.
We therefore get this fourth and final thesis:

Introduction: Vertical Relations in Science, Philosophy, and...

(META-SUCCESS) An account of the vertical relations in portion X of the


world is meta-justified if this account accurately reflects the conceptions of
vertical relations deployed in successful explanations about domain X.

The basic idea of the approach framed in META-SUCCESS is the contention that successful explanations are best explained by these explanations
being true. Consequently, the conceptions of verticality in the successful
explanations about area X are taken to provide a veridical picture of the
vertical relations in area X of the world itself. In coming sections, we will
fill out the nature of META-SUCCESS, and our other theses, and offer
support for these claims and our overall picture of the new debates over
verticality.

Part 2: TheState ofPlay About Verticality:


AMinimal Overview ofThree Research
Traditions
Given our limited space, we cannot provide a detailed characterization of
even one of the philosophical research traditions about verticality. Each of
these traditions has a rich history, is sophisticated in its treatment of verticality, has numerous proponents offering distinct variants, and includes
a range of competing versions of the relevant type of V-framework. Our
focus is therefore simply to provide a minimal characterization of each
research tradition for our purposes here, which are twofold. We simply seek, first, to give the reader an initial sense of the research tradition and, second, we seek to highlight one or two distinctive features
of verticality as it is characterized under the particular kind of account
offered by the research tradition. Such a minimal characterization thus
orients newcomers to the verticality debates and also allows us to support
COMPETITIVE by establishing that the differing research traditions are
offering rival accounts of at least some of the same object phenomena.
Given our purposes here, our survey thus sticks to providing minimal
answers to four questions for each tradition:

10

K. Aizawa and C. Gillett

I. What was the genesis of the research tradition offering this kind of
V-framework?
II. Who are some of the main proponents of this kind of V-framework
and what are some of the competing options within the research
tradition?
III. What object phenomena involving verticality are the intended focus of
the V-frameworks in the tradition?
And:
IV. What are some of the central features of verticality as it is characterized by the V-frameworks offered in this research tradition?
With these questions in mind, we start with philosophy of science, in
2.1, by sketching the rise of the New Mechanism and what we term the
neo-Causal research tradition that has grown along we dub neo-Functionalist V-frameworks in the metaphysics of mind, and philosophy of
mind/psychology, that now serve as both accounts of verticality in nature
and also increasingly figure in work on various issues in the philosophy
of science. Lastly, in 2.3, we look at the Grounding V-frameworks in
analytic metaphysics intended to provide universal accounts of all vertical
relations. We briefly summarize our findings, in 2.4, focusing in particular on how they support COMPETITIVE.

2.1: Philosophy ofScience: ATheoretical Vacuum,


theNew Mechanism, andNeo-Causal Approaches
Work focusing on the nature, and features, of compositional explanations in biology and other higher sciences was central to the critique of
the Positivists philosophy of science, including the Positivist accounts of
scientific explanation, reduction, and more. For example, writers such as
Jerry Fodor (1974) and Philip Kitcher (1984) used features of compositional explanations in the sciences to launch their famous criticisms of
the Positivists Nagelian model of reduction. However, although these

Introduction: Vertical Relations in Science, Philosophy, and...

11

writers used the vertical relations posited in such successful scientific


explanations, which they usually termed realization relations, this
important wave of early work neither provided a V-framework for realization or the associated phenomenon of multiple realization, nor did
this work provide detailed treatments of the distinctive nature of compositional explanation itself.
Post-Positivist philosophy of science was thus left with a double theoretical vacuum, and given the importance of compositional was unsurprisingly filled. In the next section, we detail how from the 1980s onwards
work on functionalism, cross-cutting philosophy of science and philosophy of mind, was co-opted to fill the need for a V-framework for realization and hence scientific composition. However, the need for better
accounts of compositional explanation was met by a later development
in the philosophy of science with the rise of the New Mechanism. In the
remainder of this section, we briefly highlight the New Mechanism and
how it has recently spawned a distinctive new research tradition focused
on vertical relations in the sciences.
The New Mechanists have made important contributions to the philosophy of science by offering highly detailed descriptive accounts of the
species of compositional explanation that these writers term mechanistic explanation [See, e.g., Bechtel and Richardson (1993), Craver
(2007), and Machamer, Darden, and Craver (2000), among others.]
New Mechanists have largely rejected the Positivists picture of explanation, as given in the Deductive-Nomological model, and any assumption that the crucial relation connecting explanans and explanadum is
logical in nature. Instead, New Mechanists have focused on the idea that
there is some sort of constitutive or compositional relation between
an explanandum and its explanans. [See, e.g., Bechtel and Abrahamsen
(2005), p. 426, Craver (2007), p. 153 and Machamer, Darden, and
Craver (2000), p.13.]
As Carl Cravers (2007) highlights, the underlying idea is that certain
lower-level processes compose a certain higher-level process and that this
compositional or constitutive relation underlies the scientific explanation
of the higher-level process. However, until quite recently, there have been

12

K. Aizawa and C. Gillett

few accounts offered within the New Mechanism of these key vertical
relation taken to posit in this way in mechanistic explanations.7
The latter situation has been increasingly addressed by writers in what
we term the neo-Causal research tradition on verticality whose driving
idea is that accounts of causation can be tweaked and applied to scientific
composition, or that techniques that offer sufficient conditions for causal
relations can be tweaked to supply sufficient conditions for scientific
composition.8 The central idea of neo-Causal V-frameworks for scientific
composition is thus that, very crudely put, compositional or constitutive
relations are either identical to causal relations or of a kind with causal
relations. Let us briefly note some of the versions of neo-Causalism.
The most prominent neo-Causal approach is offered in Craver (2007)
who is struck by the success of recent interventionist accounts of causation (Woodward 2003) that Craver takes to offer a way to avoid what
he thinks are the intractable ontological debates over the nature of
causation. Craver thus takes the manipulability relations outlined in
interventionism to offer a sufficient condition for causation, while leaving its ontology open. Consequently, Craver suggests that manipulability
relations drawn from the interventionist framework, with a number of
alterations, can provide a sufficient condition for the existence of constitutive or compositional relations between processes posited in mechanistic explanations, while again avoiding what Craver apparently takes to be
intractable ontological debates.
Strictly put, Craver is therefore not offering the standard kind of
V-framework, but only a sufficient condition for the vertical relations
posited in mechanistic explanations. Nonetheless, Craver still claims that
manipulability relations, in fact relations of mutual manipulability, suffice for vertical relations between processes of the kind posited in successful scientific explanations. Although amended in various ways, Craver
thus explicitly contends that just as one species of manipulability relation
7

See Kaiser and Krickle (2016) for a recent survey of such work.
A number of neo-Causal accounts often include conditions that compositional relations are not
causal, or hold between entities that are not logically distinct, but these accounts then all still go on
to use the machinery developed for causal relations with these conditions added to the machinery.
We therefore call these neo-Causal accounts and contend they treat causation and scientific composition as being of a kind.
8

Introduction: Vertical Relations in Science, Philosophy, and...

13

suffices for causation, so another kind of manipulability suffices for scientific composition relations between processes.
A variety of other neo-Causal accounts of scientific composition
have subsequently sprung up. However, unlike Cravers account, these
V-frameworks are explicitly intended to provide ontological accounts
of either the conceptions of composition posited in compositional
explanations and/or the vertical relations found between entities in
nature. These versions of neo-Causalism take existing accounts of
causation, tweak these accounts in various ways, and then offer the
resulting machinery as a V-framework for scientific composition. For
example, Jens Harbecke (2010, 2014a, 2014b) has developed a neoCausal V-framework by adapting regularity accounts of causation, and
Mark Couch (2011) has used Mackies INUS-based treatment to produce another neo-Causal V-framework. Still more explicit neo-Causal
approaches, such as that of Totte Harinen (2014), also now exist which
claim that scientific composition relations are quite literally identical to
causal relations.
To summarize, Neo-Causal approaches to verticality have grown out
of debates over the nature of a species of compositional explanation in
mechanistic explanation. Neo-Causal accounts are therefore intended
to offer V-frameworks for the conceptions of verticality used in certain
successful scientific explanations, and/or vertical relations in nature, but
are not intended to cover all vertical relations such as those posited in
mathematics or found in the abstract realm. The motivating thought
behind neo-Causal V-frameworks is that the vertical relations we find in
the sciences, and/or nature, are of a kind with causal relations, although
neo-Causalists diverge over what this involves. Neo-Causal accounts
therefore take vertical relations in scientific explanations, and/or nature,
either to be identical to relations of causation, regularity, mutual manipulability, sufficiency or counterfactual dependence, or take such relations
to suffice for such vertical relations.

14

K. Aizawa and C. Gillett

2.2: Philosophy ofMind/Metaphysics ofMind/


Metaphysics ofScience: TheFunctionalist Inheritance
A rich and important research tradition on verticality has grown out of
work in the 1970s and 1980s on functionalism. It is important to note
that there are a number of distinct, and very different, positions going
under the name of functionalism. This can in part be explained by the
tangled nature of the initial debates in which philosophers from very different backgrounds, with very different agendas, and focused on different object phenomena, all labeled themselves as functionalists. To note
just a few of the projects pursued in these debates, some of these writers
sought to offer accounts of the relations in compositional explanations in
the sciences, others sought to offer accounts of the mind-body relation,
still others sought to provide accounts of computational accounts in psychology, and some intended their accounts to speak to more than one of
these projects simultaneously.
Most problematically, all of these functionalist writers used the same
terms, such as causal role, functional property and realization,
to express distinct, proprietary notions. For example, some versions of
functionalism focused on realization relations between semantic
entities like sentences or statements and worldly entities, others focused
on relations of isomorphism whose relata are unconstrained in character, and still others concern ontological relations between causally individuated entities.9 It is the latter kind of account of realization, as an
ontological relation between causally individuated entities, that under9

To highlight the variety of concepts of realization in versions of functionalism, consider just


the three most familiar kinds highlighted by Endicott (2005). First, there is what we may term
M-realization relations, which are asymmetric, ontological determination relations between causally individuated property instances, or properties, instantiated by individuals located in space
time. Second, there are notions that we may term linguistic, or L-, realization holding between
entities in the world and some set of sentences. Famously, for example, the work of David Lewis on
topic-neutral Ramseyfication and theoretical terms uses a notion of L-realization. Crudely put, an
entity X L-realizes some theoretical term F when the entity X-satisfies the relevant Ramsey sentences for F. Third, there is a kind of computational or mathematical relation that we term
abstract, or A-, realization. Again putting things roughly, X is taken to A-realize Y if the elements of X map onto, or are isomorphic with, the elements of Y.This notion of realization is
commonly utilized with formal models and the relata of such realization relations are largely
unconstrained because A-realization simply holds in virtue of a mathematical mapping or isomorphism. Both L- and A-realization therefore contrast with M-realization which is an ontological

Introduction: Vertical Relations in Science, Philosophy, and...

15

pins the prominent research tradition in ongoing debates over verticality,


so having noted these other accounts of realization we leave them to
one side here.
What we term neo-Functionalist V-frameworks all focus on providing
accounts of vertical relations between entities in nature, and/or verticality
in scientific explanations, characterized as realization relations usually
between property instances. Perhaps the most prominent of these accounts
is what has variously been termed the Flat or Subset account of realization that takes one property instance X to realize another property instance
Y just in case the powers of Y are a subset of the power of X.Leading proponents of the Flat/Subset view are Lenny Clapp (2001), Sydney Shoemaker
(2001), and Jessica Wilson (1999, 2009) among others. Crucially, the Flat/
Subset view typically takes such vertical relations to be one-one relations
between property instances of the same individual that are qualitatively
similar because they overlap in the powers they contribute.
In contrast, a minority view about realization, dubbed the
Dimensioned account, takes realization to be a many-one relation
holding between the many property instances of constituents and a
property of the individual these constituents compose. Proponents of
the Dimensioned view arguably include Fodor (1968) and Dennett
(1969), and more recently Carl Gillett (2002, 2010), Derk Pereboom
(2002, 2011), and Kenneth Aizawa (2007). In contrast to the Flat/
Subset account, the Dimensioned view takes realization to be a manyone relation between property instances of constituent individuals and
a property instance of an individual that these parts constitute and such
that the realized and realizer properties are qualitatively distinct because
they share no powers.
One other account worth marking is offered by Andrew Melnyk
(2003) and it is instructive since it brings out what is central to the neoFunctionalist tradition. At the heart of Melnyks broad account of realization is the idea of a realized entity being characterized by a condition
C which is satisfied by its realizerthus Melnyks broad account does
not limit the nature of relata for his realization relation and his account
relation having as relata causally individuated entities in the world, often (though as we shall see,
not exclusively) property instances. The neo-Functionalist accounts are all focused on M-realization.

16

K. Aizawa and C. Gillett

potentially extends to all vertical relations regardless of whether we find


them in the natural or abstract portions of the world. However, one may
restrict the relevant condition C to a causal role or profile to produce a
neo-Functionalist account. And the latter point brings out one feature
that is common to all V-frameworks in the neo-Functional tradition: All
neo-Functional V-frameworks take vertical relations to require causally
individuated entities as their relata.
Neo-Functionalist accounts are routinely intended to provide treatments of the vertical relations in nature and/or the notions of verticality posited in successful scientific explanations, but are not usually
intended to extend to vertical notions in the abstract realm or disciplines like mathematics that deal with it. Recent work has broadened
neo-Functionalist accounts to cover the vertical relations of individuals (Gillett 2007, 2013; Pereboom 2011; Shoemaker 2007) and processes (Shoemaker 2007; Aizawa and Gillett, Unpublished) as well as
those of property instances. As we noted above, the work of writers
like Fodor and Kitcher in the philosophy of science left us without
a theoretical account of the realization relations posited in compositional explanations and this void has often been filled by one or other
neo-Functionalist V-framework. Consequently, neo-Functionalist
V-frameworks have also been increasingly used to provide accounts of
the compositional relations that underpin compositional explanation
and hence scientific composition, scientific reduction and emergence,
mental causation, and formulations of physicalism, among other work.
We should also mark that there is lively debate between proponents of
various V-frameworks about how well their competing accounts do at
these various projects.
To summarize, the neo-Functionalist research tradition is a longstanding one with a range of sophisticated V-frameworks and applications of them. Neo-Functionalist V-frameworks are offered as accounts
of vertical relations between the entities in nature and/or the conceptions of verticality posited in compositional explanations in the sciences.
A common feature of all neo-Functionalist V-frameworks is that they
require vertical relations to have causally individuated entities as their
relata.

Introduction: Vertical Relations in Science, Philosophy, and...

17

2.3: Analytic Metaphysics: Grounding andVerticality


Frameworks fromtheAbstract Realm
The newest research tradition on verticality is that of so-called Grounding
that has sprung up in the twenty-first century in analytic metaphysics.
Prominent proponents of Grounding include Kit Fine, (2001, 2012),
Gideon Rosen (2010), and Jonathan Schaffer (2009, 2016), among a
growing group. What should be noticeable is that these writers are logicians and/or analytic metaphysicians and it is in these areas of philosophy where Grounding V-frameworks have been developed. However,
the ambitions of its proponents for Grounding is not limited to these
areas, since Grounding theorists are characterized by claiming that their
V-framework cover all vertical relations whether in logic, mathematics,
science, nature, ethics, and so on. Once again, we offer just the most
minimal of characterizations given our goals here.10
Typically, advocates of Ground have declined to offer any definition
of Ground which they take to be a primitive notion. But proponents of
Grounding frameworks usually start with vertical relations between various abstract entities, whether sets or propositions, provide an account of
such relations often drawing upon our best formal systems characterizing
such relations, and use this account as the basis for their characterizations
of Grounding. Consequently, Grounding V-frameworks all routinely
take verticality as a relation that can hold between causally inert entities
and that has certain formal features, such as being asymmetric and irreflexive. In addition, proponents of Grounding V-frameworks also overlap
in intending their accounts to be of universal scope that covers all vertical
relations whether these relations are found in the natural or abstract portions of the world.
It is important to note that as well as these commonalities there are again
a range of differences among proponents of Grounding V-frameworks.
For instance, these writers differ over the relata of Grounding. Some such
as Fine take Grounding to hold between what they term facts and oth-

10
For introductory expositions, see Bliss and Trogdon (2014), Clark and Liggins (2012), or Raven
(2015).

18

K. Aizawa and C. Gillett

ers like Schaffer take its relata to be unconstrained and hence also to hold
between entities such as individuals, properties, and more.
It is important to note many writers take Grounding V-frameworks to
be articulating a genus relation covering various species of vertical relation. However, it is equally significant that proponents of Grounding
V-frameworks, such as Fine or Schaffer, also take this genus relation, in
Grounding, to be the relation that backs all metaphysical or constitutive explanationhence implying Grounding relations underpin the
compositional explanations we find in science. Proponents of Grounding
have also claimed it is the relation that underpins the formulation of
physicalism.
To summarize, Grounding V-frameworks have been developed
by writers in analytic metaphysics who have taken vertical relations
between entities of logic or mathematics as their exemplars, developed
V-frameworks for such relations, and then extended such Groundingbased V-frameworks to all vertical relations. Though Grounding
V-frameworks are often pitched as general accounts, articulating a
genus relation, it is also argued that all constitutive explanations are
best understood as being underpinned by this general relation. The
Grounding V-frameworks are thus intended to cover vertical relations
wherever they are found from logic and mathematics across to science
and nature. Among other characteristics, Grounding V-frameworks are
marked by allowing vertical relations to have causally inert entities as
relata.

2.4: Summary: TheSupport forCompetition


andtheDialectical Importance ofScientific
Composition
Even our minimal treatments of three vibrant research traditions in contemporary philosophy provide good reasons to believe that they offer
diverging accounts of at least some of the same object phenomena involving verticalityhence supporting our thesis COMPETITIVE. Most
obviously, we have seen that these research traditions are all either offering treatments of the vertical relations underpinning compositional

Introduction: Vertical Relations in Science, Philosophy, and...

19

explanations and/or that are found in nature. Scientific composition, and


verticality in nature, are thus dialectically important as a meeting point
for the various philosophical research traditions about verticality.
In philosophy of science, we have seen that neo-Causal V-frameworks
take existing treatments of causation and extend these accounts to vertical relations in scientific explanations and/or nature. Such accounts
take verticality to be identical to causation, regularity, mutual manipulability, sufficiency or counterfactual dependence, or take the latter to
suffice for vertical relations. In contrast to the neo-Causal tradition, neoFunctionalist and Grounding V-frameworks provide accounts that take
vertical relations to be of a different kind from causation.11 However,
neo-Functionalist and Grounding accounts also themselves diverge.
First, such view differ over whether a more absolute species or genus
relation underpins compositional explanation. Second, and more substantively, the accounts diverge over the nature of the vertical relations
backing compositional explanation. Neo-Functionalist accounts take
such vertical relations to require causally efficacious entities as relata and
this feature configures many of their substantive claims about the vertical relation they claim underpins compositional explanation. In contrast, Grounding V-frameworks claim that the vertical relation backing
compositional explanation, in Grounding, is one that does not require
causal entities as its relatahence directly contradicting the claims of the
neo-Functionalists.
There is far more to the sophisticated accounts of vertical relations
offered by each of these research traditions, but these few features suffice to
show these traditions offer diverging accounts of the nature of verticality
in the sciences and nature. We therefore conclude that COMPETITIVE
is indeed correct. Similar points no doubt hold about theories about the
vertical relations found in other portions of the world, and/or underpinning other successful explanations about the entities in these areas,
whether logic, mathematics, set theory, semantics, music, ethics, and so

11

As will become clearer below, Schaffer (2016, this volume) offers a Grounding V-framework that
may be a species of neo-Causalism. Nonetheless, other Grounding accounts still conflict with neoCausal views accounts of the relation backing compositional explanation, and all Grounding
accounts offer rivals to neo-Functionalist views.

20

K. Aizawa and C. Gillett

on. Let us therefore push on to explore the nature of the new discussions
that consequently ensue.

Part 3: Pressing Meta-Methodological Issues:


Types ofAccount, Project, Assessment and
Meta-Justification
Our work in the last section established that we are indeed in a new set
of debates over verticality, but also suggested that there are a number of
distinct projects to be pursued, and being pursued, using V-frameworks.
Sorting out these different projects, and hence the ways in which any
V-framework may be assessed or regard to such work, are a few of the
pressing issues that we address in this section.
We build on our minimal surveys, in 3.1, by sketching some of the distinct types of account, and various kinds of project, that a V-framework
might be used to address. We then look, in 3.2, at the ways in which
to assess V-frameworks and we provide support for COMPARATIVE
and FOCUSED. Lastly, in 3.3, we briefly discuss the issue of metajustification and outline one reason why many parties to the new debates
are coalescing around the approach framed in META-SUCCESS with its
focus on successful explanation.

3.1: Types ofV-Framework andTheir Intended Projects


Alongside the different accounts of verticality we have seen are offered
by competing V-frameworks, we also need to separate out various types
of object phenomena each of these views can be intended to provide an
account of. There is a range of projects for which a V-framework may be
used which differ in the object phenomena the V-framework is focused
upon. A V-framework may be offered of the concept(s) of verticality used
in some areawhether in a type of explanation, or in a certain kind of
theory or discourse, and so on. This might concern the concepts used in
the successful explanations of the sciences, or everyday discourse, or any
theory positing verticality.

Introduction: Vertical Relations in Science, Philosophy, and...

21

In contrast, one can also use a V-framework to provide an account of


the vertical relations found in some area of the world itself. Thus, one
can offer a V-framework as an account of vertical relations in the natural world or in the mathematical realm. When offering a V-framework
in pursuit of this project, the V-framework is offered as an account of
vertical relations that exist in the world.12 And, as we outline in section
3.3 below, there are reasons why some writers take their account of the
concepts of verticality used in the successful explanations offered about
some area X of the world to also provide such an account of the vertical relations in area X.So one can also intend a V-framework to serve
as an account of both the conceptions of verticality in certain successful
explanations and also as an account of associated vertical relations in the
world.
It is also useful to distinguish differences in the intended scope of
V-frameworks. What we will term a Local account is a V-framework
intended to capture the nature of some limited set of vertical relations
in some domain X, say relations of scientific composition or vertical relations in set theory. In contrast to Local Accounts, one can use a
V-framework as what we term a Global account that offers a framework
for all verticality. It is important to note one theorist may offer a certain
V-framework as a Local account of certain vertical relations, while someone else may offer the same V-framework as a Global account.
On top of the latter Local/Global distinction, we should also note that
one can intend a V-framework either as an account of more absolute or
ultimate species of vertical relation. Or one can offer this V-framework as
a more general account of such species of vertical relationswhether by
positing a unifying genus for such relations, or picking out commonali-

12

Given the latter points, there a number of different projects a Global Account (see below) could
be intended to pursue. One Global Account defends the claim that a V-framework captures the
nature of vertical relations as they appear in all successful explanations positing verticality and that
this V-framework also covers all vertical relations found in the world. However, one could also
offer a Global Account as a V-framework just covering all vertical relations in the world, but be
neutral about whether ones account covers conceptions of verticality or even accept that other
V-frameworks offer the best accounts of the concepts of verticality used in some successful
explanation.

22

K. Aizawa and C. Gillett

ties or resemblances among these absolute vertical relations.13 Both Local


and Global accounts can be focused on more absolute species of verticality, or more general relations, depending on the intentions of their
proponents.
Let us briefly note how some of these complexities play out in ongoing debates. For instance, with regard to scientific explanations we have
disputes between writers offering differing V-frameworks as the best
Local accounts of the conceptions of verticality posited in compositional
explanations. For example, in our new debates neo-Causal and neoFunctionalist V-frameworks are plausibly offering such competing Local
accounts of scientific notions of verticality and hence scientific composition. In addition, however, a defender of a Global account based on some
V-framework may also provide a competitor to such a Local account of
the notions of verticality in certain explanations or of the vertical relations in some area of the world. This is again plausibly the case in our new
debates where there appear to be Global views using certain Grounding
V-frameworks that entail accounts about either the conceptions of verticality used in scientific explanations, and/or the vertical relations in
nature, which we have seen are at odds with the diverging Local accounts
of either the neo-Causal and neo-Functionalist V-frameworks.
One might think that the latter Local and Global accounts are not
competitors, since the Local accounts of the neo-Functionalist are
focused on more absolute species of verticality, while the Global account
of Grounding V-frameworks is a more general account of such species. However, the Grounding V-frameworks intended as more general
accounts are still competitors with Local neo-Functionalist V-frameworks
intended to cover more absolute species of verticality, since both of the
latter accounts claim to articulate the vertical relation underpinning
compositional explanation and, as we outlined earlier, give conflicting
accounts of the nature of this vertical relation. (It is worth marking that
if Grounding V-frameworks were not claimed to illuminate the relation
backing compositional explanation, or underpinning physicalism, then
at least potentially the neo-Functionalist accounts focused on more abso13
See Koslicki (2015) for a more detailed discussion of some of the options for the type of general
accounts that might be intended by proponents of Grounding V-frameworks.

Introduction: Vertical Relations in Science, Philosophy, and...

23

lute species might be compatible with, or complementary to, Grounding


V-frameworks seeking generalities across such species.)
To summarize, there is diversity within both the kinds of object phenomena that one can use a V-framework to illuminate and also with
regard to the intended scope, and generality, a theorist intends for such a
V-framework. It should therefore be obvious that the same V-framework
can be used for a variety of different object phenomena and be intended
to have different scope or generality. Against this background, we cannot
emphasize too strongly how important it is to disambiguate the project
that a V-framework under discussion is intended to pursue both with
regard to the intended object phenomena, scope and generality of a
V-framework. Without such disambiguation then confusion result, for
one cannot clearly understand the hypothesis about verticality the relevant V-framework is being used to articulate. And we also consequently
cannot understand what it would be for the relevant account of verticality to succeed or fail. On this note, let us now turn to the work that a
V-framework needs to accomplish and the ways such accounts should be
assessed.

3.2: What Must Competing V-Frameworks Do


toBeSuccessful?
When we turn to illuminating the kind of work a V-framework must do,
we obviously need to clarify the project this account is intended to pursue with regard to its object phenomena, scope, and generality. So let us
consider a Local Account focused solely on the vertical relations underpinning compositional explanation in the sciences construed as more ultimate species of verticality. The primary obligation of such a V-framework
is to provide an adequate account of the features of the relevant vertical
relations, in this case scientific composition, as well as the characteristics
of associated explanations, in this case compositional explanations in the
sciences, and also to accommodate any other relevant evidence. On a first
pass, a Global account pitched at the same level of generality as some
Local account, or taking the same object phenomena as its focus, must
do the same work as the latter Local accountand all the other Local

24

K. Aizawa and C. Gillett

accounts of vertical relations found in other areas overlapping in object


phenomena with the Global account. Thus, given its character, such a
Global account must provide adequate accounts of the vertical relations
found in every area of the world and/or type of successful explanation.
We want to emphasize the transition in the type of work that the new
debates consequently require, since it contrasts with the kind of activity
often found in earlier debates. Rather than high level, often abstract,
discussion peppered with brief mention of specific phenomena, the
new debates require detailed engagement with specific phenomena
involving verticality. What is required is, first, a focus on articulating
the character of the relevant object phenomenon involving verticality
because we need to articulate the features of the relevant vertical relations, associated explanations, and so on that our V-framework seeks
to capture. Then, second, we construct a V-framework that is explicitly
focused on accommodating the features of this specific phenomenon
involving verticality. And, third, we assess the V-framework in a more
focused manner by how well it does at this latter task with regard to the
relevant object phenomenon.
We earlier framed the latter requirements in FOCUSED in order to
highlight how the new debates are moving us away from often unfocused discussions of verticality to accounts that are constructed, and
assessed, by careful, detailed examination of specific examples of vertical
relations and/or conceptions of verticality. It is important to emphasize
that FOCUSED only frames one important element in the assessment
of some V-framework, since theory appraisal is plausibly never done in
isolation. We always assess, and justify, theories by comparison to their
relevant rivals. We therefore also contend that COMPARATIVE frames
how the assessment of any V-framework that must also involve comparison to relevant rivals. And it is worth noting that we have already seen
such comparative assessments within particular areas of philosophy. For
example, as we briefly noted in 2.2 above, there are lively, and ongoing,
debates between competing neo-Functionalist V-frameworks about both
which provide the best accounts of the conceptions of verticality in compositional explanations and the vertical relations in the natural world.

Introduction: Vertical Relations in Science, Philosophy, and...

25

We have had some important initial salvoes in the wider comparative


assessment required by the new debates and we want to briefly sketch
one of the most important of these skirmishes given its prominence and
because it is continued in two of the coming Chapters. The area where
we have seen nascent discussion of the comparative merits of competing
V-frameworks is with regard to Grounding V-frameworks and their claim
to be Global Accounts, albeit at a more general level, covering all vertical
relations. In response, writers like Kathrin Koslicki (2015) and Jessica
Wilson (2014) have offered arguments challenging the plausibility, and
utility, of Grounding V-frameworks. We focus on Wilsons work to highlight these critiques, since this dialectical thread figures prominently in
coming Chapters.
One important line of argument offered by Wilson is that no
V-framework arguing for a general characterization of vertical relations,
and hence focusing on a genus relation, can be a better account of verticality in nature, or underpinning compositional explanation, than one
of the existing Local accounts positing more ultimate or absolute species
of vertical relation, including neo-Functionalist V-frameworks. Wilson
(2014) examines several possible needs for a genus relation of Ground
and argues that in all of these cases there is no reason to add such a
genus relation in understanding the relevant phenomena in addition
to the absolute species of vertical relation posited by the existing Local
V-frameworks. Wilson therefore argues that, with regard to verticality in
nature, there is no philosophical work for a generic relation of Grounding
to do in addition to the work done by ultimate species of verticality posited by the V-frameworks of various existing Local Accounts.
We have already begun to frame Wilsons argument using the framework
for the new debates we have outlined and we want to use this machinery
to further highlight what we take to be one of the underlying questions
in such debates. Once again, we in no way wish to adjudicate whether
Wilsons broad arguments against the utility of a generic Grounding relation are successful, or not, but we want to draw out another thread in
this nascent dispute. We suggest that we can now see that the issue is not
just about the utility of positing more absolute species versus accepting
genus relations and the general accounts that posit them. In addition,
there is also the issue of what kind of picture of verticality provides the

26

K. Aizawa and C. Gillett

best account of certain object phenomena involving verticality and hence


which kind of V-framework provides the best account of the relevant
object phenomena.
For example, whether construed as more or less absolute or general
relations, there is the question of whether a vertical relation of the kind
portrayed by neo-Functionalism, neo-Causalism or certain Grounding
frameworks provides the best account of the relation underpinning compositional explanation. As we have highlighted, neo-Functionalist frameworks take vertical relations to require causal entities as their relata, among
other features, hence offering one hypothesis about the vertical relation
that best accommodates the relevant evidence. In contrast, as we have
also noted, Grounding-based V-frameworks offer a conflicting treatment
of the relevant vertical relations. Similar points hold about neo-Causal
accounts. What we therefore have to examine, following FOCUSED,
is which of these rival V-frameworks provides the best account of this
object phenomenon.
Against this background, Wilson (2014) offers a range of arguments
that can be understood as pressing reasons why the conceptions of verticality in the Local neo-Functionalist accounts should be preferred to
account of vertical relations offered in Grounding V-frameworks. But it
appears that, following FOCUSED, we now need to articulate the feature of the relevant vertical relations, associated explanations, and related
evidence, and so on, and then to evaluate how well various V-frameworks
perform.
As we shall see in the Chapter by Schaffer, the latter work is exactly
what proponents of Grounding have now begun to pursue by articulating the relevant evidence and making a positive case that Grounding
V-frameworks best accomodate such evidence. On the other side,
the Chapter by Wilson directly engages this more detailed defense of
Grounding, while the Chapters by Aizawa, Gillett, and Melnyk all provide articulations of the features of compositional explanation, or related
phenomena, and reasons why their favored V-frameworks should still be
taken to provide the best accounts of verticality in the relevant object
phenomena rather than Grounding accounts. All of the latter Chapters
thus follow FOCUSED and COMPARATIVE to contribute to the new
debates in their own distinctive ways.

Introduction: Vertical Relations in Science, Philosophy, and...

27

3.3: Meta-Justifying V-Frameworks About Verticality


intheWorld? Understanding theAppeal
ofMeta-Success
Before we conclude our brief roadmap to the new debates, it is worthwhile turning to the meta-justification of V-frameworks intended to
provide accounts of vertical relations in the world, rather than this or
that conception of verticality. It was not so long ago that philosophical
accounts of vertical relations in the world were assessed by how well they
accorded with our intuitive judgments. So we would be asked to consider two quarter coins touching, or people shaking hands, think about
whether we would intuitively judge such cases to involve a vertical relation, and then assess a broached V-framework on this basis.
However, we should immediately note some concerns about such
approaches to justifying accounts of vertical relations in the world. Our
intuitive judgments are obviously relevant data for V-frameworks pursuing some projects: If one is offering a V-framework as account of our
commonsense or folk conception of verticality in nature, then such folk
judgments look like relevant data about the concept underlying such
judgment. However, in earlier debates, commonsense judgments have
been used to assess V-frameworks as accounts of verticality in the world,
rather than simply as treatments of our commonsense notions. But then
we face the thorny question of why we should think our commonsense or
folk conceptions reflect verticality in the world? Plausible answers to this
demand for a meta-justification of the methodology using folk judgements and concepts have not been forthcoming.
Against this background, it is interesting that many writers in recent
work on verticality, from across various areas, have begun to coalesce
around a different approach in their search for accounts of verticality in
the world which does have a clear meta-justification. The relevant methodology used in a search for accounts of vertical relations in a portion X
of the world is to focus on a conception of verticality, but it is the notion
of verticality used in successful explanations about area X rather than any
folk concept. But why this focus on successful explanations, and even their
conceptions of verticality, in the pursuit of accounts of verticality in the
world itself ?

28

K. Aizawa and C. Gillett

The central idea is the one we briefly marked earlier, namely the idea
that the best account of the success of an explanation is that the explanation is true or likely to be true. Consequently, successful explanations
about area X that posit vertical relations are likely true and so the conception of vertical relations used in these explanations either are, or are likely
to be, veridical representations of verticality in area X.We can therefore
see that there is a clear and compelling meta-justification for constructing accounts of verticality in some part of the world using our successful explanations about this portion of realityhence bringing us to the
approach we earlier framed in META-SUCCESS. We can thus begin
to see why it is no surprise to find V-frameworks increasingly offered as
Local accounts of both the conceptions of verticality offered in successful
explanations about domain X and also as accounts of the vertical relations
in domain X in the world.
We should note that META-SUCCESS is a meta-justification of
accounts of vertical relations in the world. V-frameworks offered in pursuit of other projects, such as illuminating this or that conception of
verticality, will likely pursue different methodologies and have different
meta-justifications. We should also note that we have framed METASUCCESS solely as a sufficient condition for a meta-justification of an
account of verticality in the world. However, we know of no other viable
meta-justifications of such accounts at present. So it is very much a live
question whether META-SUCCESS, or some version of it, should be
framed as a necessary and sufficient condition.

4: Summary ofComing Chapters


The contributed papers all make important contributions to the new
debates, often in multiple ways, so to further aid the reader we provide a
brief overview of each Chapter. We have made a very rough division of
the papers into two groups. Part I draws together papers that are primarily focused on discussions of vertical relations in science and/or nature as
they arise in the New Mechanism, Neo-Causal approaches, or in competing Neo-Functionalist accounts. The papers in Part II all engage, to
lesser or greater degrees, the new debates over verticality and Grounding

Introduction: Vertical Relations in Science, Philosophy, and...

29

frameworks. We should also note that some Chapters in Part I make


important points about Grounding frameworks, and various papers
in Part II offer arguments about neo-Causal and/or neo-Functionalist
accounts, so our division is only rough.
Part I starts with a paper by Laura Franklin-Hall, entitled New
Mechanistic Explanation and the Need for Explanatory Constraints,
which argues that the broader New Mechanist project has failed to live
up to its theoretical obligations with regard to mechanistic explanation.
Franklin-Hall offers an account of the New Mechanist picture under
which explanations are mechanistic models that satisfy three desiderata: they must represent causal relations, describe the proper parts, and
depict the system at the right level. However, Franklin-Hall argues
that even by its own lights the New Mechanist program currently fails
to provide an adequate treatment of mechanistic explanation because its
extant accounts still fail to distinguish good and bad examples of such
explanations.
Among other arguments, Franklin-Hall postulates that there is a carving error problem according to which New Mechanism provides no principled way of picking out the individuals that are, and are not, plausibly
the relevant proper parts. In addition, Franklin-Hall also postulates that
there is a zooming error problem according to which New Mechanism
gives no principled reason to pick out the appropriate level for a good
mechanistic explanation. Overall, for these and other reasons, FranklinHall suggests that even by its own lights the New Mechanism has yet to
provide a theoretically adequate treatment of mechanistic explanation of
the type it has taken as its focus.
In his Chapter, Compositional Explanation: Dimensioned
Realization, New mechanism, and Ground, Kenneth Aizawa uses the
Dimensioned account of realization to frame some of the distinctive features of compositional explanations and hence raise problems about the
theoretical treatments offered by the New Mechanism and also to press
concerns about the adequacy of Grounding accounts as treatments of
scientific composition.
With regard to the New Mechanism, Aizawa notes that the New
Mechanists have stereotypically focused on vertical relations between
processes often analyzed in terms of a two-part ontology of individu-

30

K. Aizawa and C. Gillett

als (or entities) and their processes (or activities). However, Aizawa
argues that the relata of the relevant vertical relations, and hence also
list of explananda and explanantia of mechanistic explanation, should
be expanded to include properties as well as individuals and processes.
Aizawa makes his case, in a manner following FOCUSSED, by looking at concrete scientific cases where individuals have properties that are
explained by properties of their parts. In making his case, Aizawa assumes
that the vertical relations between posited in the mechanistic explanation of properties are better characterized by the theory of Dimensioned
realization (Gillett, 2002, 2003), but Aizawa does not argue for this as
an alternative account of the compositional relation between properties.
As well as these contrasts with the New Mechanism, Aizawa presses the
point that scientific composition is not merely a logical or quasi-logical
relation. Aizawa contends instead that scientific composition involves
a non-logical, non-causal form of natural necessitation of the NeoFunctionalist variety outlined in Aizawa and Gilletts V-framework. And,
again following FOCUSED, Aizawa uses concrete cases of compositional
explanation to highlight how Grounding accounts do not accommodate
key features of scientific composition and also compositional explanation. In particular, Aizawa focuses on what he terms the love this expression of compositional explanation, in its feature of explaining entities
of one kind using qualitatively different kinds of entity, and argues that
Grounding frameworks fail to accommodate this key feature.
Jens Harbeckes Chapter, Is Mechanistic Constitution a Version of
Material Constitution, lays out Harbeckes sophisticated neo-Causal
V-framework for the vertical relations deployed in compositional
explanations in the sciences (Harbecke 2010, 2014a, b). Like Craver,
Harbecke begins with the thought that constitution is like causation,
but rather than starting with manipulability Harbecke begins with a
Humean regularity conception of causation and then transforms this
causal foundation into a constitutional relation by adding conditions
from classical mereology. The main work of Harbeckes Chapter is to
explore how his account of such mechanistic constitution relates to the
relation, which Harbecke terms metaphysical constitution, that we find
in puzzle cases in metaphysics such as the statue and the clay. Harbecke
defends a number of interesting substantive theses about the relations of

Introduction: Vertical Relations in Science, Philosophy, and...

31

mechanistic and metaphysical constitution and draws out their implications for a number of long-standing debates and positions in philosophy.
Broadening the discussion of accounts of vertical relations in science
and nature still further, Derk Perebooms paper, Anti-Reductionism,
Anti-Rationalism, and the Material Constitution of the Mental, makes
positive and negative contributions to the ongoing disputes between
competing Neo-Functionalist V-frameworks outlined briefly in 2.2
above. On the negative side, Pereboom further develops his important
critique of Subset/Flat accounts of vertical relations between properties
in nature. In past work, Pereboom (2011) has argued that under the
Flat/Subset view of Sydney Shoemaker (2001) or Jessica Wilson (1999)
the powers of realized and realizer property instances must be identical.
Perebooms paper now extends his critique by arguing that latter feature
means that realization under the Flat/Subset view is not the kind of asymmetric vertical relation that we plausibly find in nature.
On the positive side, Pereboom outlines his favored approach to the
vertical relations between individuals in nature developed in his previous
work (Pereboom 2011). Pereboom defends his account against a number of objections ranging from in-principle concerns about his primitive
made-up-of relation to Andrew Melnyks concern that Perebooms position is committed to physicalistically unacceptable posits. In response,
Pereboom outlines why his position does everything a naturalist would
require without falling into what he argues are the overly strong demands
of his critics.
As outlined above, the Chapters of Part II are all focused, to lesser or
greater degrees, on Grounding accounts. To begin the section, there is a
pair of papers that continue one prominent thread of the new debates
started by Wilson (2014). The final three Chapters consider, among other
issues, whether Grounding V-frameworks are suited to certain projects.
Jonathan Schaffers paper, Ground Rules: Lessons from Wilson,
explicitly engages Wilson (2014) which, as we noted earlier, presses
a broad critique of the utility and informativeness of Grounding
V-frameworks. Schaffer draws out what he takes to be two important
lessons from Wilsons critique, but then argues that, suitably modified,
his favored Grounding V-framework has learned these lessons and that
Wilsons favored account has not. The first lesson Schaffer draws from

32

K. Aizawa and C. Gillett

Wilson is that an account of Ground must not settle for bare Grounding
claims, such as X grounds Y, but must also make sense of the follow-up
inquiry into how a vertical relation makes sense of a certain explanatory
connection. The second lesson Schaffer draws from Wilson is that the
best account of certain object phenomena involving verticality should
follow a formalism that successfully unifies the putative features of verticality. We consequently see Schaffer offering ideas about how to articulate both FOCUSED and META-SUCCESS. Furthermore, Schaffer
argues that structural equation modelling, a technique drawn from the
sciences, provides the best formalism for verticality and then argues that
his Grounding V-framework is the account favored by this formalism.
On the other side, Schaffer argues that Wilsons favored account fails to
live up to her two lessons.
Jessica Wilsons Chapter challenges the adequacy of Schaffers new
Unity Argument for Grounding, then revisits a further Priority
Argument for Grounding. She summarizes Schaffers Unity Argument
as follows:
1. If some phenomena are aptly formally unified, then this provides
strong (albeit defeasible) reason to posit a unifier.
2. The diverse (small-c) causal relations are aptly formally unified by the
SEM framework.
Therefore, there is strong (albeit defeasible) reason to posit a general
notion of causation.
3. The diverse (small-g) grounding relations are just as aptly formally
unified by the SEM framework as the diverse (small-c) causal
relations.
Therefore, there is strong (albeit defeasible) reason to posit a general
notion of Grounding.
Wilson then provides multiple arguments that each of the premises
in Schaffers argument is false. One of her points against premise 1 is
that determinables are metaphysically and formally unified, but this has
not blocked the common view that determinables are schematic for or
reducible to determinates. Wilson thus challenges Schaffers approach to

Introduction: Vertical Relations in Science, Philosophy, and...

33

META-SUCCESS through a FOCUSED observation. Her argument


against premise 2 is that structural equation modeling has not, in fact,
enfolded the diversity of causal relations. For example, Hall (2004) proposes that there are two concepts of causation: causation as counterfactual
dependence and causation as production. Structural equation modeling,
however, appears to capture the counterfactual type of causation and not
the productive type. Among the reasons she offers against premise 3 is the
observation that structural equation models do not appropriately model
the determinable/determinate relation, a prime illustration of Ground.
These arguments are thus instances of FOCUSED critiques.
The Chapter from Carl Gillett, The Metaphysics of Nature, Science
and the Rules of Engagement, provides a broader treatment of the
issue of how to construct, and assess, an adequate account of the vertical relations in science and nature. Gillett contends that compositional
explanations are basically our only successful explanations of the natural
world using vertical relations, so accounts of the notions of verticality in
these explanations must serve as the basis of any account of verticality in
nature. Gillett then lays out the strategy of what he terms Engagement
that flows from this point and the Rules of Engagement that frame
an appropriate methodologybasically, that accounts of verticality in
scientific explanations need to start with an examination of the features
of the notions in such explanations. In detail that all the prominent
philosophical accounts of vertivcality, including neo-Causal, Grounding
and neo-Functionlalist aproaches, do not derive from such Engagement,
but from what he terms the Appropriation of machinery developed for
other phenomena.
After outlining features of scientific composition, and two key
characteristics of compositional explanations, Gillett then shows that
the Appropriational views, including neo-Causal, Grounding and standard neo-Functionoats views, all fail to accommodate these features,
hence failing as accounts of verticality in science and nature. In contrast,
Gillett suggests that a minority treatment of verticality developed using
Engagement has no such defects. Gilletts broader methodological conclusion is that we need to follow Engagement, and to carefully consider
the character of compositional explanation, if we are to understand verticality either in science or in nature itself.

34

K. Aizawa and C. Gillett

Andrew Melnyks Chapter, Grounding and the Formulation of


Physicalism, also looks at how Grounding V-frameworks perform
at a task for which there are rival neo-Functionalist accounts. Melnyk
proposes that physicalism be formulated so as to satisfy two conditions.
It must characterize the narrowly physicalthose things that are physical in their own rightand the broadly physicalthose things that
are related to the narrowly physical such that they are nothing over and
above the narrowly physical. Melnyk offers three main arguments that
Ground does not provide a satisfactory account of the relation between
the broadly physical and the narrowly physical. He argues that, for one
thing, insofar as Ground is taken to be a primitive relation placing no
restrictions on the features of its relata, it cannot restrict its relata to the
physical. Thus, a physical X could, in principle, Ground a non-physical
Y.For a second thing, philosophers need not resort to a primitive notion
of Ground in order to formulate physicalism, since one might instead
formulate it in terms of a more robust non-primitive notion of realization along the lines formulated in Melnyk (2003). For a third thing,
Grounding itself does not seem to be a relation that is consistent with
physicalism. According to Melnyk, Grounding cannot be a broadly
physical relation, insofar as it is primitive, hence fundamental, hence not
Grounded. Nor can Grounding be a narrowly physical relation, since it is
not expressed by a two-place predicate of physics. Nor could Grounding
be defined by a complex construction of, say, physical terms, since that
would violate the putative primitive character of Ground. By contrast,
the notion of realization Melnyk invokes to formulate physicalism is consistent with physicalism.
The final paper is by Alyssa Ney, Grounding in the Philosophy of
Mind: A Defense, which again considers the suitability of a Grounding
V-framework for a certain application. Unlike Gillett and Melnyk, Ney
offers an enthusiastic defense of Ground for performing a certain task.
Specifically, Ney argues that the version of Grounding put forth in Fine
(2001, 2012) is a means by which to resolve long-standing disputes in
the philosophy of mind. Contrary to critics like Wilson (2014), Ney
consequently provides a FOCCUSED argument that Fines Grounding
V-framework is far from intellectually superfluous or unproductive.

Introduction: Vertical Relations in Science, Philosophy, and...

35

Ney narrows in on the aspect of Fines account of Grounding as a


truth-conveying relation between sentences, whose canonical form is Its
being the case that S consists in nothing more than its being the case that
T, U, where some sentence S gets to be true in virtue of the truth of
some Grounding sentences T, U, Ney argues that Fines Grounding
V-framework enables us to formulate a coherent position about sentences
of higher sciences under which they are still true, but not in virtue of the
entities that non-reductionists like Jerry Fodor have claimed in composed
entities.
Most importantly, Ney argues at length that this framework promises
to resolve a philosophical impasse between Fodors nonreductive physicalism and Kims reductionism. Fines Grounding framework enables us
to say, with Fodor, that psychological claims are true, factual, justified,
and important, because those psychological claims are Grounded in the
real. Ney claims that this gives Fodor what he wants most, namely, the
preservation of cognitive science as true, factual, and so forth. But, we do
not have to follow Fodor in saying that the intrinsic structure of reality
contains composed mental entities such as beliefs, desires, and pains. In
addition, Ney argues that Fines Grounding framework also enables us to
say, with Kim, that, in reality, there are no beliefs, desires, or pains. This
gives Kim what he wants most, namely, an austere ontology that avoids
problems of causal overdetermination or epiphenomenalism that go
along with composed mental entities such as beliefs, desires, and pains.
Overall, Ney thus argues that Fines V-framework earns its philosophical
keep by resolving a long-standing impasse in the philosophy of mind/
psychology.

References
Aizawa, K. (2007). The biochemistry of memory consolidation: A model system
for the philosophy of mind. Synthese, 155, 6598.
Bechtel, W., & Abrahamsen, A. (2005). Explanation: A mechanist alternative.
Studies in History and Philosophy of Science Part C: Studies in History and
Philosophy of Biological and Biomedical Sciences, 36(2), 421441.

36

K. Aizawa and C. Gillett

Bechtel, W., & Richardson, R. (1993). Discovering complexity. Princeton:


Princeton University Press.
Bliss, R., & Trogdon, K. (2014). Metaphysical grounding. The Stanford
Encyclopedia of Philosophy (Winter 2014 Edition). Edward N. Zalta (ed.),
URL = http://plato.stanford.edu/archives/win2014/entries/grounding/.
Clapp, L. (2001). Disjunctive properties: Multiple realizations. Journal of
Philosophy, XCVIII.
Couch, M. (2011). Mechanisms and constitutive relevance. Synthese, 183,
375388.
Craver, C. (2007). Explaining the brain. Oxford: Oxford University Press.
Craver, C., & Bechtel, W. (2007). Top-down causation without top-down
causes. Biology and Philosophy, 22(4), 547.
Dennett, D. (1969). Content and consciousness. London: Routledge Kegan Paul.
Fine, K. (2001). The question of realism. Philosophers Imprint, 1(1), 130.
Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),
Metaphysical grounding: Understanding the structure of reality (pp. 3780).
NewYork, NY: Cambridge University Press.
Fodor, J. (1968). Psychological explanation. NewYork: Random House.
Fodor, J. (1974). Special sciences (or: the disunity of science as a working
hypothesis). Synthese, 28(2), 97115.
Gillett, C. (2002). The dimensions of realization: A critique of the Standard
view. Analysis, 62(276), 316323.
Gillett, C. (2007). Hyper-extending the mind? Setting boundaries in the special
sciences. Philosophical Topics, 35, 161187.
Gillett, C. (2013a). Constitution, and multiple constitution, in the Sciences:
Using the neuron to construct a starting framework. Minds and Machines 23,
30937.
Gillett, C. (2013b). Understanding the Sciences through the Fog of
Functionalism(s). In P.Hunneman (ed.) Functions: Selection and Mechanisms
(pp. 159181). Dordrecht: Kluwer.
Glennan, S. (1996). Mechanisms and the nature of causation. Erkenntnis, 44(1),
4971.
Hall, N. (2004). Two concepts of causation. In J.Collins, N.Hall & L.Paul
(Eds.), Causation and counterfactuals (pp.225276). Cambridge, MA: MIT
Press.
Harbecke, J. (2010). Mechanistic constitution in neurobiological explanations.
International Studies in the Philosophy of Science, 24(3), 267285.
Harbecke, J. (2014a). The role of supervenience and constitution in neuroscientific research. Synthese, 191(5), 725743.

Introduction: Vertical Relations in Science, Philosophy, and...

37

Harbecke, J. (2014b). Regularity constitution and the location of mechanistic


levels. Foundations of Science, 20(3), 323338.
Harinen, T. (2014). Mutual manipulability and causal inbetweeness. Synthese.
Online pre-pubication.
Healey, R. (2013). Physical Composition. Studies in the History and Philosophy
of Science Part B, 44, 4862.
Kaiser, M., & Krickel, B. (2016). The metaphysics of constitutive mechanistic
phenomena. British Journal for the Philosophy of Science. Online
pre-publication.
Kim, J. (1992). Multiple realization and the metaphysics of reduction. Philosophy
and Phenomenological Research, 126.
Kitcher, P. (1984). 1953 and all that: A tale of two sciences. Philosophical Review,
93, 33573.
Koslicki, K. (2015). The coarse-grainedness of grounding. In K.Bennett and
D.Zimmerman (ed.), Oxford studies in metaphysics (pp. 306344). Oxford
University Press.
Machamer, P., Darden, L., & Craver, C. (2000). Thinking about mechanisms.
Philosophy of Science, 67(1), 125.
Melnyk, A. (2003). A physicalist manifesto: Thoroughly modern materialism.
Cambridge: Cambridge University Press.
Pereboom, D. (2002). Robust nonreductive materialism. Journal of Philosophy, 99,
pp.499531.
Pereboom, D. (2011). Consciousness and the Prospects of Physicalism. Oxford:
Oxford University Press.
Raven, M.J. (2015). Ground. Philosophy Compass, 10(5), 322333.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
R.Hale & A.Hoffman (Eds.), Modality: Metaphysics, logic, and epistemology
(pp.109136). NewYork, NY: Oxford University Press.
Schaffer, J. (2009). On what grounds what. In D. Chalmers, D. Manley &
R. Wasserman (Eds.), Metametaphysics (pp. 357383). Oxford: Oxford
University Press.
Schaffer, J. (2016). Grounding in the image of causation. Philosophical studies,
173, 49100.
Shoemaker, S. (2001). Realization and mental causation. In C. Gillett and
B. Loewer (eds.), Physicalism and its discontents (pp. 2333). Cambridge:
Cambridge University Press.
Shoemaker, S. (2003). Realization, micro-realization and coincidence. Philosophy
and Phenomenological Research, 67, 123.

38

K. Aizawa and C. Gillett

Shoemaker, S. (2007). Physical realization. NewYork: Oxford University Press.


Wilson, J. M. (1999). How superduper does a physicalist supervenience need to
be? Philosophical Quarterly, 49, 3352.
Wilson, J. M. (2009). Determination, realization, and mental causation.
Philosophical Studies, 149169.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57(56),
535579.
Woodward, J., 2003, Making Things Happen: A Theory of Causal Explanation,
NewYork: Oxford University Press.

Part I
Scientific Composition
and the New Mechanism

2
New Mechanistic Explanation
andtheNeed forExplanatory
Constraints
L.R.Franklin-Hall

Introduction
In the past decade and a half, a new movement (Glennan 2005: 443)
has arisen in the philosophy of biology, one called a revolution (Bechtel
2006: 280) with broad implications (ibid: 2) and which has met with
broad consensus (Campaner 2006: 15). On this hot topic (Robert
2004: 159), a vast literature has developed, within it one of the most
cited papers in Philosophy of Science (viz. Machamer etal. 2000).
What is the subject of such attention? It is the new mechanistic philosophy (Skipper and Millstein 2005: 327), articulated by a group of
philosophersincluding William Bechtel, Carl Craver, Lindley Darden,
Peter Machamer, and Stuart Glennaninterested in the nature of mechanisms, complex systems characterized most prominently as entities or

L.R. Franklin-Hall ()
Department of Philosophy, New York University, New York, NY, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_2

41

42

L.R. Franklin-Hall

activities organized such that they are productive of regular changes from
start or set-up to finish or termination conditions (Machamer et al.
2000: 3). Mechanisms are said to be worthy of attention largely because
they are central to a new and superior approach to scientific explanation, one truer to scientific practice than the long defunct deductivenomological (DN) view. It is also claimed that the mechanistic approach
has implications beyond explanation, as it transforms how one thinks
about a host of other issues in the philosophy of science (Bechtel and
Abrahamsen 2005: 426), including causality, laws, kinds, reduction, discovery, and scientific change.
Philosophical movements can be judged by their fruits. We can ask
of them: what problems does a movement offer solutions to? Judging
by both the language of the new mechanists and the influence of their
work, it would appear that the mechanistic approach had served up a
bevy of solutions. Yet I argue here that, at least with respect to its core
projectthat of elucidating the nature of scientific explanationappearances are deceptive: the mechanisms movement has not yet yielded the
advertised results. This is not because mechanisms advocates are committed to claims that are false. My critique is motivated instead by concerns that mechanistic explanatory accounts offered to dateeven in
their strongest formulationshave failed to move beyond the simple and
uncontroversial slogan: some explanations show how things work. In
particular, I argue that proposed constraints on mechanistic explanation
are not up to the task required of them: namely, that of distinguishing
acceptable explanations from those that, though minimally mechanistic,
are uncontrovertibly inadequate.
Sections The Mechanistic Explanatory Framework and Formulating
Explanatory Constraints sketch a version of the new mechanistic
explanatory account, one constructed by combining the most promising proposals from across the mechanistic corpus. After articulating three
principles at the heart of this pictureconcerning causation, parts, and
explanatory levelsections The Causal Standard through The Levels
Standard argue that these principles remain promissory notes. The chapter concludes in section Conclusion with an evaluation of the mechanistic explanatory program.

New Mechanistic Explanation andtheNeed forExplanatory...

43

The Mechanistic Explanatory Framework


Scientists in many disciplinesbut particularly in biologyfrequently
refer to mechanisms when describing the systems they investigate, provoking
a natural question: what is a mechanism? Answers to this question from
advocates of the new mechanistic philosophysuch as from Machamer
etal. (2000: 3), Bechtel and Abrahamsen (2005: 423), and Glennan (2002:
S344)differ more in language than content; all agree that a mechanism
is a physical system composed of at least somewhat organized parts whose
interactions either bring about or constitute some phenomenon.
Though mechanisms may be germane to various philosophical endeavors (Levy 2013; Nicholson 2012), most prominent is their central place
in a theory of explanation, one intended to apply to many of the biological sciences. According to that theory, explanations are explanatory in
virtue of communicating facts about how things work (Craver 2007b:
110) in the system that brings about, or constitutes, the phenomenon to
be explained. These facts should be communicated by a largely veridical
representationcalled a mechanistic modelof the system responsible
for the explanandum phenomenon (Bechtel and Abrahamsen 2005: 425;
Craver 2006; 2007b: vii; Glennan 2005: 446; Machamer etal. 2000: 3).
Mechanistic models need not take some canonical form, nor must
they be usable to derive a statement of the explanatory target (Bechtel
and Abrahamsen 2005: 430; Bechtel 2011: 537; Craver 2007b: 160;
Machamer etal. 2000: 23). What they must do is account for a systems
capacity to produce certain outputs in response to certain inputs. To do
this in a properly mechanistic style, they should describe the system as
having multiple parts that are organized in some respect and that change
through time according to dynamic principles, principles that might be
understood to reflect activities, laws, or some other species of regularity.
When such models bridge inputs and outputs as required, they can directly
explain systems-level capacities; they may also explain particular events
when supplemented by a statement of initial (i.e., activation) conditions.1
1

Beyond token capacities and events, mechanists also aspire to treat regularities. Though the details
are rarely made explicit, a given regularity can be explained via a mechanistic model jointly applicable to all of the particular systems underpinning a regularitys instances; to do this, such a model
must be at least somewhat abstract. For a discussion of how this might work, see Strevens First
Fundamental Theorem of Explanation (2008: Chap. 7).

44

L.R. Franklin-Hall

The most important variety of model that new mechanists judge as


unexplanatoryat least when deployed to explain the biological phenomena that interest them mostis the global model, one constituted by
a single dynamic principle stating that a system experiencing such-andsuch inputs will produce such-and-such outputs.2 Global models treat
systems as opaque black boxes; they fall short explanatorily in virtue of
failing to look under the hood and beneath the regularities couched
at the behavioral level to reveal underlying mechanisms (Kaplan and
Bechtel 2011: 442).

Formulating Explanatory Constraints


The explanatory framework sketched above is plausible yet incomplete.
The basic problem is that, for any candidate explanandum phenomenon
that the new mechanistic account aims to treat, there exist an enormous
range of models that satisfy the above-noted core mechanistic conditions, that is, by representing the system in terms of organized parts that
change according to dynamic principles. Yet, only a handful of these
models appear to be explanatorily apt. Thus, to fill out the account, we
must design constraints capable of distinguishing the good mechanistic
modelsthose that provide adequate explanationsfrom those that fall
explanatorily short.
To illustrate this challenge, and to motivate the new mechanistic contributions that might be used to meet it, I will describe four veridical, mechanistic models for a single phenomenon: a neurons capacity to release
neurotransmitters at its axon terminal when its dendrites are exposed to
neurotransmitters, and not otherwise.3 While the first model, called here
the Standard Model, is explanatorily acceptablebased as it is on textbook
2

Mechanists also judge unexplanatory phenomenological modelsthose that dont purport to


describe the inner workings of the system at issueas well as mechanistic models that are false
(even allowing for limited idealization) of the systems they purport to describe. As these exclusions
will be uncontroversial for any fan of causal explanation, they require no discussion.
3
Though this phenomenon is often modeled probabilistically, I treat it deterministically for the
sake of expository simplicity. This simplification is innocent; over-permissiveness would be found
equally on any probabilistic formulation.

New Mechanistic Explanation andtheNeed forExplanatory...

45

accountsthe other three will appear flawed. They each make a distinct
variety of explanatory error and can illustrate in the breach the constraints
that a mechanistic model must fulfill to be explanatorily acceptable.4
According to the Standard Model, a neurons capacity to release neurotransmitters when exposed to them is explained by describing the neuron as composed of a variety of somewhat organized macromolecular
parts, including membranes, channels within them, and ionic concentrations in the internal and external environmentall of which interact
according to dynamic principles, such as one stating that neurotransmitter
binding is followed by channel opening.5 Though these details could be
communicated in a variety of ways, they are most often presented in narrative form, as follows: neurotransmitter exposure leads neurotransmitter molecules to bind to ligand-gated receptors located in the dendrite
membrane. Upon binding, these channels open. Then sodium ions rush
into the cell, depolarizing the membrane locally. Next, a population of
voltage-gated membrane channels, located in the same region, also open
and more sodium enters. This begins a cascade of channel opening, depolarization, and further channel opening, that moves up the neuron until
it reaches the neurons axon terminal where voltage-gated calcium channels open and calcium enters the cell. Finally, vesicles containing neurotransmitters located near the axon terminal bind with the membrane,
releasing neurotransmitters to the extracellular environment.
To formulate a second kind of model that applies to the same explanandum, consider any regular side-effect of neurotransmitter binding, such
as the mild vibration of the cell membrane surrounding the receptor.
Presume that whenever the neuron is exposed to neurotransmitters, this
4

All four candidate models maintain that a neuron behaves thus because it is constituted in such a
way that (1) it does not release neurotransmitters absent neurotransmitter exposure, and (2) exposure initiates a cascade of events, one of which is neurotransmitter release. Yet, the first condition is
customarily taken for granted, and explanatory presentations focus on the second by describing the
relevant features of the constitution of the neuron, and how exposuregiven this constitution
has the specified result.
5
Just as the overall phenomenon might be treated either probabilistically or deterministically, so it
goes with this dynamic principle. Though I will not worry about the details, which sort of treatment is most apt will depend on how the channels are individuated. If single channels are separately
represented, a probabilistic treatment is most appropriate; if large collections of channels are treated
together, deterministic treatment will be preferred.

46

L.R. Franklin-Hall

vibration occurs, but it has no consequences on the remainder of the cell


depolarization process. Given this, we can formulate a model identical
to the Standard Model, except that it appeals to two alternative dynamic
principles, one relating neurotransmitter exposure and membrane vibration, and a second relating vibration and any later event genuinely relevant to neurotransmitter release, for example, the entry of calcium into
the axon terminal. With these principles and others, such an alternative
model might bridge inputs and outputs, stating first that neurotransmitter exposure is followed by membrane vibration, itself followed by cellular calcium entry, eventuating finally in neurotransmitter release.6 Like
the Standard Model, this model can appeal to organized parts changing
according to dynamic principles. Nevertheless, it is flawed in virtue of
making what I call a causation error.
A third kind of model correctly describes causal connections between
a systems parts, but individuates those parts in a non-standardand
explanatorily deficientway. Consider, for instance, a model that
describes just four connected parts of the neuron, large chunks of biomass extending about one-fourth of the way from dendrites to the axon
terminal, each capable of taking at least two states. This model might be
used to account for the target phenomenon as follows: neurotransmitter
exposure changes the state of the first part, which modifies the state of
the second part, in turn modifying the third in the same way, and then
finally the last hunk of neuronal materials, eventuating in the output
of interestneurotransmitter release. This model, however peculiar, is
also properly mechanistic: it describes multiple organized parts, changing
according to dynamic principles, and principles that themselves track the
causal order. Nevertheless, in virtue of its gerrymandered carving of the
system into quarter-neurons, it fails to reflect actual explanatory practice,
and is intuitively unexplanatory. It makes what I will call a carving error.

Some might suggest that this model isnt mechanistic at all, insisting that to be mechanistic a
model must satisfy a causal constraint. This would be to cut up the project slightly differently than
I have, but with no consequences for the overall argument. The task facing the new mechanist
would still be to cash out the causal constraint; it matters not whether that constraint is appealed
to in the definition of mechanistic models simpliciter, or (as in my exposition) in the characterization of explanatorily adequate mechanistic models.

New Mechanistic Explanation andtheNeed forExplanatory...

47

The fourth model characterizes both real causal connections and appeals
to natural, rather than gerrymandered, parts. Its distinctive explanatory shortcoming is that it describes the system at the wrong level, in
terms of organized atomic parts changing according to dynamic principles (in this case, principles aptly called laws) describing atomic interactions. Such a model will be so complex that, in contrast with the three
rehearsed already, it is not possible here to sketch the course of events it
would describe as following from neurotransmitter exposure. Yet such
a low-level model will still satisfy the requirements of the mechanistic
framework above: it describes organized parts that change over time
according to dynamic principles, collectively bridging inputs and outputs. By depicting the neuron in such detail, it makes what I call a zooming error, and should, as above, be censured by any explanatory account
that takes actual explanatory practice as its touchstone.7
The three flawed accounts just sketched were easy to design, and
equivalent alternatives are readily available for any explanandum you
might choose; they require no real creativity or insight. One starts with
the inputoutput relationship for which the mechanistic model must
account. These inputs and outputs, as the mechanists rightly emphasize,
will be underpinned, in any particular system, by a complex set of connections between that systems parts. To produce a model that errs causally,
describe at least some portion of the system underlying the explanandum
behavior in terms of correlationalnot causalprinciples. To produce
one that makes a carving error, describe that underlying system veridically, but use a peculiar set of terms, those that individuate the system in a
non-traditional way.8 Finally, to produce a model at the wrong level, either
zoom in on the parts of the system more than is explanatorily appropriateby describing, for example, the inner working of entities usually
treated as wholes by scientists accounting for the focal phenomenonor
fail to break the system into parts, thereby producing a global model.

A zooming error is a species of carving error, and they are separated largely for rhetorical purposes.
The first prototypically concerns using gerrymandered parts, while the second concerns otherwise
natural parts at too fine (or coarse) a grain, considering the explanandum phenomenon.
8
Though many peculiar sets will exist, not any will do: they must still be sufficiently expressive that
they can be used, in concert with some set of dynamic principles, to bridge inputs and outputs.

48

L.R. Franklin-Hall

At the core of the mechanistic explanatory account, as I reconstruct it,


stand three standards that rule out models that suffer from these three types
of errors. These should acteither individually or collectivelyas a kind
of sieve, sifting out the detritus, and revealing the explanatory nuggets.
The Causal Standard The dynamic principles that describe system
change should be causal. Different workers attempt to spell out this
requirement differently, sometimes drawing strategies from theories
of causation produced independently of the mechanisms movement.
For instance, some mechanists depend on Woodwards (2003) version
of the interventionist account of causation (Craver 2007b; Glennan
2002), while others develop their own activities theory (Bogen 2005;
Machamer 2004).
The Carving Standard Models should carve mechanisms at their joints,
describing them in terms of the appropriate set of parts (Craver 2006:
367; Bechtel and Abrahamsen 2005). They should not reflect the arbitrary differentiation of the parts of a whole (Bechtel 2008: 146). For
instance, parts appealed should be good parts, like macromolecules,
rather than bad parts, like quarter-neurons.
The Levels Standard Models should represent the system at the right
level, or grain, which in the judgment of many (though not all) new
mechanists will not be a fine-grained physical specification but will be in
various ways abstract (Levy and Bechtel 2013). In particular, some will
hold that an explanatory model should represent systems at the level just
below that of the explanandum phenomenon. Thus, it may be a mistake to explain neurotransmitter release at the axon terminal in terms
of atomic events, or even in terms of an influx of sodium into the terminal, rather than in terms of a comparatively coarse-grained event like
depolarization (Craver 2007: 23).
How successful are these standards? The burden of the next three sections is to argue that they are not yet up to the task assigned to them:
that of distinguishing the genuinely explanatorily models from the many
that fall short.

New Mechanistic Explanation andtheNeed forExplanatory...

49

The Causal Standard


According to the first standard, the dynamic principles embedded within
explanatory models must describe causal relations, not mere relations of
correlation. As Craver notes, analyses of explanation must include reference to causal relationships if they are to distinguish good explanations
from bad (2007a: 8).
This basic claim is highly plausible but requires elaboration. After all,
though causation is one of the most familiar features of our world, it is
also among the most obscure. What is this relation between cause and
effect, the basic material out of which a causal explanation is constructed?
Are causes related to effects, as Hume thought, just by their constant
conjunction? Or does causation involve a more metaphysically loaded
relation of dependence or necessitation? In that case, how are we to understand this more substantial connection, for instance, in terms of the
truth of certain counterfactuals, or in terms of some relationship between
universals?
Before discussing the new mechanistic approach to the causal relation, consider an alternative strategy that connects mechanisms and causation, pursued by an earlier generation whom we might call the old
mechanists. Peter Railton, Wesley Salmon, and J.L. Mackie aimed to
use mechanisms to contribute to our understanding of the causal relation, specifically to what distinguished causal connections from mere
correlations. Mackie, for example, hoped that what he called a mechanism might constitute the long-searched for link between individual
cause and effect (Mackie 1974: 228229). And both Salmon (1984)
and Railton (1978) attempted to give an account of causation in terms of
mechanism. Many believe that these accounts failed on their own terms
(Hitchcock 1995), though it was clear what these philosophers were up
to: they were using mechanisms to do battle with Humes Ghost, and
attempting to glimpse the secret connexion between cause and effect.
The relationship between this work and that of the new mechanists
has not always been transparent. Machamer etal. (2000) explicitly compared the new mechanists project to Salmons and Mackies but lamented
that it is unclear how to apply [Salmons and Mackies] concepts to our
biological cases (2000: 7). Glennan (2002) also suggests that the new

50

L.R. Franklin-Hall

mechanists approach was a successor project, writing that while philosophers of science typically associate the causal-mechanical view of scientific explanation with the work of Railton and Salmon, [.I] shall
argue that the defects of this view arise from an inadequate analysis of the
concept of mechanism (S342).
Yet a clear contrast exists between the old mechanists and the new, and
it may be misleading to see their projects as continuous. The key difference
concerns the relationship between cause and mechanism. The old mechanists were trying to reduce causation to mechanism; however, most new
mechanists use accounts of causation to understand the relations between
parts (or, properties of parts) of mechanisms. Speaking metaphorically,
old mechanisms were the causal glue, while new mechanisms are glued
together by causes. Along these lines, recent commentary calls for abandoning the idea that causation can be reduced to mechanism. On closer
inspection, it appears that the concept of mechanism presupposes that of
causation, far from being reducible to it (Kistler 2009: 599).
Given that mechanisms dont reduce causation but instead require an
account of it, what account should that be? Clearly, it must differentiate dynamic principles that reflect relations of correlation from those
of causation. To this end, two paths have been taken. The first is to tie
the mechanistic approach to an independent account of causation, one
that may lack any interestingly mechanistic character, for instance, to
Woodwards interventionism or Lewis counterfactual account. Craver
(2007), Glennan (2005), and Leuridan (2010) have pursued this strategy, adopting Woodwards (2003) account of causation, according to
which causal relations are those potentially exploitable for the purposes
of manipulation and control (Woodward 2003: 17). The second is to
develop an account of causation with mechanistic contexts in mind. For
example, Bogen (2008) and Machamer (2004) have pursued this option,
developing an activities view of causation.
Though the first approachthat of adopting an independent, nonmechanist account of causationis perfectly reasonable, I will not
explore it. Given the uncontroversial nature of the basic mechanistic conditionsat least for fans of causal explanationthose who fill out the
mechanistic picture by adopting a self-standing account of causation are
not much advancing the explanatory project. Needless to say, outsourcing

New Mechanistic Explanation andtheNeed forExplanatory...

51

causation may well be the right move for mechanists to make, and those
who do so may still contribute to our understanding of scientific explanation; however, their contributions must come from elsewhere, presumably from their elucidations of the other two constraints on mechanistic
explanationson parts and levelwhich will be explored in due course.
Some mechanists have attempted to make sense of the causal relation
via the notion of activities (see Bogen 2005, 2008; Machamer 2004;
Waskan 2011). Here is an early statement of the view:
An entity acts as a cause when it engages in a productive activity. [] A
mechanism is the series of activities of entities that bring about the finish
or termination conditions in a regular way. These regularities are nonaccidental and support counterfactuals to the extent that they describe
activities (Machamer etal. 2000: 68).

The basic idea is that X causes Y when related by an activity. Focusing in


this way on activities appears to provide a simple, scientifically informed
analysis of causation that avoids many of the thorny matterssuch as
the nature of laws, regularities, or counterfactualsthat consume those
more metaphysically minded. As Bogen puts it, [i]f the production of
an effect by activities which constitute the operation of a mechanism is
what makes the difference between a causal and a non-causal sequence of
events, mechanists need not include regularities and invariant generalizations in their account (Bogen 2005: 399).
This activities account, also called the actualist-mechanist theory
(Waskan 2011), is offered as one of many process or production theories
of causation (Hall 2004). In this case, what makes for a causal connection is an actual process of a certain type. Early advocates of the process
approach had empiricist sympathies: they were suspicious of the counterfactuals that seemed necessary to make sense of a dependence relation, and wanted to do without them. Their task was to distinguish, in
general terms, causal processes from what are sometimes called pseudoprocesses that may reflect merely correlated events, and all without a
counterfactual crutch.
There appear to be two ways of making activities part of a philosophically informative theory of causation. Most obviously, activities might

52

L.R. Franklin-Hall

be the special sauce that distinguishes the genuinely causal processes.


The philosophical task would be to describe these activities, characterizing precisely how they are special. The activity approach would, in this
case, be structurally similar to the old mechanists accounts, noted above,
which offered not activities but mechanisms, understood in terms of
the capacity of transmitting a local modification in structure (a mark)
(Salmon 1984: 147) or the exchange [or persistence] of a conserved
quantities (Dowe 1995: 323) as tools with which to separate the causal
wheat from the correlational chaff.
Second, the activities approach might, though refraining from the
above task, identify what the activities in fact are. This could be likened
to Descartes attempt to characterize the causally efficacious properties
such as extension and velocityas part of a quest to banish the substantial forms and final causes which Descartes contemporaries appealed
to, in his view, willy-nilly. Jon Elsters work on functions in the social
sciences also has this character. He emphasizes the importance of uncovering the nuts and bolts of social mechanisms because he believes that
absent a selection processthe functional properties that are appealed to
in socialscientific explanations are actually explanatorily empty (Elster
1989). This sort of project would be particularly well motivated if the
new mechanists suspected that biologists were likewise appealing to nonexplanatory, non-causal features.
Yet those developing an activities account of causation have refrained
from both of these tasks. Advocates dodge the first project by claiming
that activities have merely verbal unity. Scientists do somehow distinguish causally productive activities from those that are not, but the distinction cannot be captured informatively by any single account (Bogen
2008: 116). This is because there is no informative general characterization which discriminates causally productive activities from goings-on
which are not causally productive of the effect of interest (ibid: 113;
Machamer 2004). Mechanists also refrain from the second undertaking.
Unlike Descartes and Elster, they evince no general skepticism regarding
the activities appealed to by the competent scientists whose work they
study, noting instead that acceptable causal relations are those that our
scientific investigations reveal to us as how the world works (Machamer
2009: 4). And they claim, wisely enough, that there is no definitive list

New Mechanistic Explanation andtheNeed forExplanatory...

53

that philosophers might produce of the activities, and that it is the job of
scientists, in any case, to compile it.
The central feature of this account of causationthe activityis, from
a philosophical perspective, brute. Scientists identify activities, but they
have nothing generally in common; short of listing those taken seriously
by scientists at a given time, we cant say anything about what they are. It
remains possible that the quest to find a general account of causality like
Humes, Hempels, or Woodwards is misguided, and that wed be better off talking only of particular activities (Bogen 2008: 214). Yet, if we
take these claims seriously, the content of the first restriction on explanatory modelsthat they call on causal dynamic principlesis completely
opaque. Were I to offer a model containing a dynamic principle which
(intuitively) reflected relations of correlationsuch as the model above
that referred to membrane vibrationall that could be said is that such a
model is bad because it doesnt reflect activities, and that activities themselves were just the things that competent scientists talk about.

The Carving Standard


The second mechanistic explanatory standard insists that explanatory
models truck in the good parts of a mechanism. These are sometimes called
working parts (Bechtel 2008) or working entities (Darden 2008),
though they are most commonly labeled components (Bechtel and
Abrahamsen 2005: 425; Craver 2006: 369; 2007: 188), terms I use interchangeably. In contrast to a gerrymandered part or piece, which can
result from any conceivable decomposition, including those that slice,
dice or spiral cut a mechanism, component cut mechanisms at their
joints (Craver 2007b: 187188; see also 2007a: 5). As such, components
are not mere results of arbitrary differentiation (Bechtel 2008: 146).
Requirements on components aim to solve the carving problem. Though
all mechanistic explanations bridge the inputs and outputs of a system
with a veridical mechanistic model, there are multiple ways of decomposing a system into organized parts. Furthermore, multiple mechanistic
modelsthat is, those reflecting different decompositionscan bridge
inputs and outputs as required. Such alternative models describe the

54

L.R. Franklin-Hall

internal working of the same system(s) using different vocabularies. In


these alternative terms, the models package some of the same informationmost notably, information about how output states depend on
input states. Yet, none of these models can be censored for being nonmechanistic or false.
In the face of these false riches, the carving problem is that of providing a principle that distinguishes the good explanatory models from the
bad. On the one hand, it is very clear that, in explaining various goingson, scientists routinely carve mechanisms into good parts, rather than gerrymandered entities. But, on the other hand, it isnt transparent whatif
anythingthis practice is tracking. Fundamentalists may try to sidestep
the issue by asserting thatappearances asidethe only appropriate
explanations are those that carve systems into their fundamental physical constituents governed by physical laws. In contrast, however, many
new mechanists do embrace explanations appealing to non-fundamental
parts and properties. This gets them much closer to actual scientific practice, at the cost of then needing to specify which high-level mechanistic
models are appropriate.

Good Parts asComponents


In the context of addressing a variety of different topics, including but not
limited to the carving problem, Carl Craver has articulated a number of
features that good or real parts, also called components, must possess (2007: 128133, 187195).9 These features are a mix of epistemological and more metaphysical requirements. All are rather undisputed as
necessary conditions on the parts described by mechanistic models, and
are frequently mentioned by proponents of the mechanistic approach to
explanation.10
9

In particular, in addition to potentially addressing the carving problem, these conditions are
offered as standards for distinguishing models that appeal to real parts from those that describe
fictional posits(Craver 2007: 128133).
10
I focus on Cravers presentation because it is the most systematic available, but it is characteristic of
the new mechanist literature. For instance, compare Dardens (2008: 961962) discussion of working entities and Machamer etal.s (2000: 56) comments on individuation of entities and activities.

New Mechanistic Explanation andtheNeed forExplanatory...

55

1. Robustness: components should be detectable with a variety of causally and theoretically independent devices (2007: 132).
2. Manipulability: it should be possible to manipulate the entity in such
a way as to change other entities (2007: 132).
3. Plausibility: components should be physiologically plausible (2007:
132).
4. Stability: components should have a stable cluster of properties
(2007: 131) and should be loci of stable generalizations (2007: 190).
The first standard is that components be robust. Though some discussions of robustness have a more metaphysical cast, the variety of robustness at issue here is epistemic. To say that a component is robust is simply
to say that it is detectable by different kinds of devices, optimally those
operating on different principles. This standard is inspired by the usefulness of multi-device detection in helping scientists to distinguish genuine
features of a system from artifacts (Culp 1994).
Yet, robust detectability will not address the carving problem. First,
no device detects individuated parts as such, and consequently no part
component or otherwisecan be detected by more or fewer devices than
another. To illustrate, consider an electron micrograph of a cell. Such a
micrograph is (roughly) a representation of the electron density of material
in different regions. Patterns in the density revealed by electron microscopy
can provide evidence about the features of particular components, such as
the shape of a membrane channel. The micrograph itself, however, does
not detect which of the pieces are components; a carving into components is
something that the scientist brings to the micrograph to interpret it.
An alternative to insisting that components be detectable by different
devices is to suggest that the properties of components, as opposed to
parts, be so detectable. The problem with this alternative is that components and gerrymandered pieces will pass the test equally: we can detect
the properties of protein channels as well as quarter-neurons using a variety of normal neurophysiological devices. Thus, it does not appear that
robustness will contribute to solving the carving problem.
The second standard is manipulability. This standard requires that a
good part be itself manipulable in the service of affecting something else,
a constraint inspired by Ian Hackings (1983) famous call for entity

56

L.R. Franklin-Hall

realism, according to which we deem theoretical entities real when it


is possible to do things with them. As he put it, if you can spray [them]
then they are real (1983: 24). Craver explains his particular application of this idea as follows: [i]t should be possible to manipulate the
entity in such a way as to change other entities, properties, or activities
(132). Understood in this way, the quarter-neuron modelone of many
that we must censorwill pass the test, as it is perfectly possible to
manipulate a hunk of a neuron to affect something else. In consequence,
manipulability appears no better off than robustness in distinguishing
components from gerrymandered parts.
The third standard on good parts is called plausibility. Here, Craver
requires that components actually exist in the systems under consideration, rather than only under highly contrived laboratory conditions or
in otherwise pathological states (Craver 2007: 132). This suggestion is
designed to rule out models that describe parts not present in the systems
whose behavior is being explained. Here again, we have a principle that
does not help address the carving problem. Just as do components, gerrymandered parts can exist in non-pathological conditions, and are thus
plausible to treat as entities with respect to a behavior that a mechanistic model aims to explain.
This brings us to the final standard on components: that they have a
stable cluster of properties (Craver 2007: 131). In a related discussion,
Craver suggests that componentswhich themselves can be understood
as submechanisms composing larger mechanismsbe loci of stable generalizations (Craver 2007: 190). In contrast to the three conditions just
reviewed, there are prospects for developing this constraint in a way that
allows mechanists to address the carving problem.
The stability condition asserts that a parts status as a component
depends on its possessing a stable cluster of properties. A components
properties are stable, I will presume, if they would be maintained across
some range of background conditions. Any component with such a property cluster will be one about which we can frame generalizations that
are, to some degree, counterfactually stable. The virtues of such generalizations are legion, and a preference for them in explanatory contexts
is uncontroversial (Mitchell 2000; Woodward 2001). It thus appears to
make sense to carve systems, in explanatory contexts, in ways that allow
those systems to be described by stable generalizations.

New Mechanistic Explanation andtheNeed forExplanatory...

57

However, any appeal to stability to solve the carving problem must


provide more analysis than this. First, the two most straightforward interpretations of the requirement, which are positioned on opposite extremes,
either will fail to distinguish good carvings from bad or will be at odds with
other commitments of the new mechanistic program. On the one hand,
one cannot simply insist that components possess a cluster of properties
that is in some respects stable, since gerrymandered parts will meet this
minimal standard. Yet, on the other hand, mechanists also cannot say that
components are pieces with the most stable property clusters. This position
is unavailable because it is in direct tension with one of the animating
motivations of the mechanisms movement: the rejection of proper laws as
explanatorily central. The problem is straightforward: to insist on carving
mechanisms into components with the maximally stable cluster of propertiesthat which can be described in terms of maximally stable generalizationswould require modeling mechanisms in terms of basic physical
components, governed by causal dynamic principles which are physical
laws. But to explain system functioning in these terms is clearly not to the
mechanists tasteand for good reason. Scientists, particularly life scientists, explain systems functioning without appealing to proper laws, and
do so in terms of parts with property clusters that are often wildly unstable
from a physical point of viewfor example, proteins which denature in
all but a narrow range of pHs, or cell membranes which fragment in all
but specially tuned barometric circumstancesyet these models at least
appear to provide superior explanations to those provided by lower-level
physical accounts. In light of these complexities, mechanists require a version of the stability condition that is substantially more nuanced.
Such a nuanced requirement could be constructed in a number of ways
and which would be impossible to exhaustively survey here. Instead, consider one intermediate approach to the stability standard that seems in
line with the basic commitments of the new mechanistic program: provide principled guidelines on the range of background conditions over
which part properties must be stable, with that range being somewhere
between the minimal and maximal standards just considered. Parts with
properties stable over that range are components; those which are not
are mere pieces. This stability range can be extracted from the stability
properties of the explanandum. In particular, consider this constraint: a
part is a component of a mechanism for a behavior if the relevant proper-

58

L.R. Franklin-Hall

ties of the partin particular, those of its properties that underpin the
mechanisms behaviorare stable, at a minimum, throughout the range
of conditions over which the mechanisms overall behavior is stable.
Why might one want a parts property stability range to be determined
by that of the stability of the overall systems behavior? Arguably, because
it is only a part with this characteristic that could actually underpin the
behavior to be explained. After all, mechanism-level behaviorssuch
as the inputoutput relationships that are the target of most mechanistic explanationsthemselves have modal scope, holding in at least
some range of background conditions. If a mechanistic model is to fully
account for such a modally robust explanandum, the parts appealed to
in the model must themselves survivemaintaining their property clustersover that same range.11
In illustration, recall the explanandum behavior discussed above that
neurons release neurotransmitters when exposed to neurotransmitters.
This behavior holds of neurons over a range of conditionsin different
temperatures, different ionic environments, and so on. Among the neuronal components critical for the behavior are the ligand-gated ion channels located in the dendrite membrane. The channel properties relevant
to the overall mechanisms behaviormost notably, their disposition to
open in response to neurotransmitter bindingmust be stable over a
range of background conditions in order for the mechanistic model to
account for the stable systems behavior. Imagine, for instance, that in
some condition in which the system behavior was maintained, the ion
channel was denatured, and thus no longer possessed the property relevant to the behavior under analysis. Were this to be the case, one could
not model the behavior in terms of these parts.
How might this standard reject gerrymandered parts? The contrast
between the quarter-neuron model and the Standard Model can illustrate. Consider the range of background conditions over which the parts
11

There are situations more complicated than this. If a mechanism contains a variety of redundant
subsystemseach of which has a different range of stable functioningthe overall mechanism
behavior could have a range of stability greater than that of any particular component, or component pathway. Yet, this possibility doesnt undermine the more generic suggestion that some identifiable relationship exists between the stability of a mechanisms parts properties and the
mechanisms systems-level behavior, and that this connection might be used to determine the relevant stability range required of mechanism parts.

New Mechanistic Explanation andtheNeed forExplanatory...

59

represented in the quarter-neuron model would maintain their properties, as well as the range over which the macromolecules in the Standard
Model would do so. At first glance, in contrast to the macromolecules, it
may appear that the quarter-neuron will fail to possess properties as stable
as required. Its properties will change in a broad range of circumstances,
as the quarter-neuron will be modified in some way just in case any of its
proper parts is so modified. Thus, this revised standard maybe effective,
and the carving problem solved.
Unfortunately, the proposal just described is not strong enough to distinguish good parts from mere pieces, and can only be used to rule out
non-veridical models, not those reflecting inferior carvings. The problem
is that many gerrymandered pieces, correctly characterized, will in fact
possess properties that are just as stable as required by the constraint
that is, as stable as the behavior of the overall mechanism. This is because
only the properties that underpin the mechanisms behavior need to be
so stable, according to the standard under consideration here. Although
it is true that a relatively large partgerrymandered or otherwiselike
the quarter-neuron, will change in some ways in the face of a wide array
of background circumstances, it will not change as often with respect to
the properties that underpin mechanism behaviorthose determining
its capacity to bridge the relevant inputs and outputs. In fact, with the
caveat noted above, it will maintain these properties at least over the range
for which the system-level behavior is stable. One might be tempted to
reject such properties as peculiar or gerrymanderedand thus not those
whose stability is relevant for determining component-hood. However,
this would be to make ones account of good parts dependent upon a
substantive account of good properties, which mechanists dont provide. Thus, the tactic shows little promise. A stability constraintat least
in the version Ive proposedcannot solve the carving problem.

Good Parts asMutually Manipulable


A more sophisticated tool that might better address the carving problem is the mutual manipulability (MM) standard, proposed in Craver
(2007a, b). It aims to provide conditions for when a part is a component
in a mechanism (Craver 2007b: 141). Given that the term component

60

L.R. Franklin-Hall

is used in explicit contrast with mere pieces or parts (Craver 2007b:


188), the MM standard appears to be framed to solve the carving problem. It offers conditions for what are called relevant components via
two basic requirements on the relationship between a component and
a whole mechanism. These conditions require that something about the
whole mechanism depends on the features of the component, and conversely, that something about the component depends on the features
of the whole mechanism. More particularly, a part is a component of a
mechanism for a behavior if the following conditions are satisfied:
(A) Intervening to change the component can change the behavior of the
mechanism as a whole; and
(B) Intervening to change the behavior as a whole can change the behavior of the component. (Craver 2007b: 141)12
These conditions are loosely inspired by the interventionist account
of causation, and both (A) and (B) are counterfactual conditions.13 They
are either true or false depending on whether some ideal causal manipulationhere called an interventionwhich need not be possible to
actually carry out, would have the specified result. Depending on the
particular interventionresult pairing, this result might be a causal consequence of the change brought about by the intervention, or it could
follow constitutively from that change, just as an intervention to increase
the mass of my foot would change the mass of my whole body.
The first part of the MM condition, labeled (A) above, has two elements in need of refinement, one involving the intervention to change the
component, and the other the change in the behavior of the mechanism as a
whole. With respect to the first element, what would it mean to intervene

12

Craver sometimes presents the standard, quite reasonably, using his own symbolism. For instance,
another version of (A) requires that there is some change to Xs -ing that changes Ss -ing
(2007b: 153). Though these alternative statements are compatible with the interpretation I give of
the MM standard, and have informed my presentation, I do not use Cravers notation because it
would require too much space to adequately explain.
13
Though this statement is from Cravers (2007a), in explicating the view I am very influenced by
Cravers presentation in his (2007b). In correspondence, he reports that his presentation of the
standard there is particularly careful.

New Mechanistic Explanation andtheNeed forExplanatory...

61

to change the component? Note that use of the term intervention here,
though clearly inspired by its use by causal interventionists, should not be
understood in the precise technical sense defined by them (e.g., Woodward
2003) but instead as another sort of in-principle causal manipulation,
sometimes glossed simply as wiggling (Craver 2007b: 153; 2007a: 15).
With this in mind, there are two genres of change that might be intended.
First, the manipulation might change the input to the component. For
instance, in the case of a part like a ligand-gated ion channel, a change
might involve exposing the channel to neurotransmitters, something that
would have a variety of downstream effects, the most direct of which is the
opening of the channel. Second, such a change might be made to the features underpinning the inputoutput regularity realized by the component
itself. Again, focusing on the ligand-gated ion channel, a wiggling of the
inputoutput relationship could involve a modification of the channels
disposition to open upon neurotransmitter binding. A parallel ambiguity
faces the second half of the (A) conditionthat involving the resulting
change to the behavior of the mechanism as a whole. This could involve a
change (from some default) of the output produced in a particular circumstance, or a change to the overall inputoutput relationship that the
mechanism underpins.
In light of these alternative versions of the conditionboth with
respect to the feature intervened upon and the consequent changeI
distinguish between two versions of Cravers condition (A).
(Ai) intervening to change the input to a component (from a default
input)
changes the output of the mechanism as a whole (from a default
output).
(Aii) intervening to change the inputoutput relationship realized by the
component changes the inputoutput relationship realized by the
mechanism as a whole.
Some examples used to illustrate the MM standard fall under (Ai),
while others align more with (Aii). For instance, indicating the relevance
of the first version, Craver (2007a, b) suggests that what he calls activation experiments can (sometimes) test the fulfillment of the condition,

62

L.R. Franklin-Hall

experiments in which one activates a component, apparently by setting its


inputs in a certain way, and evaluates the consequences of this intervention on the system-wide output. On the other hand, indicating the relevance of the second version, Craver describes interference experiments.
In this case, the intervention can involve completely destroying, or more
subtly modifying, the characteristics of the candidate component, and
investigates change to the capacity of the whole mechanism. Fortunately,
it will not be necessary to determine which refinement of the (A) condition is most defensible. Instead, I will probe the efficacy of both versions.
The second half of the MM standard, (B), requires that intervening to
change the behavior as a whole can change the behavior of the component.
The most obvious uncertainty here concerns what it means to intervene
on the behavior of the whole. A prima facie worry is that one can only
intervene on the behavior of a whole by intervening on the behavior of
its parts (individually or in combination); if so, triviality threatens, since
there will always be some change to the behavior of the whole that changes
the behavior of the component, namely, an intervention that changes the
behavior of the whole just by changing the behavior of the component.
Fortunately, Craver suggests a more substantive reading of (B). An
intervention on the behavior of the whole is just one that sets the input
conditions on the mechanism in a certain way, that is, one that sets the
inputs to those required to bring about the particular system-wide output that is of interest (Craver 2007b: 146). The resulting change in the
behavior of the component is a change to its output (rather than to the
features underlying its capacity to produce certain outputs given certain
inputs). Thus, reconsider (B) as follows:
(B*) Intervening to change the input to the whole mechanism, such that
it will bring about a particular output of interest, can change the output of the component.
Can these standardsAi, Aii, and B*distinguish parts and components? According to (Ai), intervening to change the input to a component (from a default input) can change the output of the mechanism as
a whole (from a default output). This will not help rule out gerrymandered pieces, since some changes to the inputs to such piecessuch

New Mechanistic Explanation andtheNeed forExplanatory...

63

as quarter-neuronscan change the outputs to whole mechanisms. In


particular, changing the input to any of the quarter-neurons can lead a
neuron to release neurotransmitters. According to (Aii), intervening to
change the inputoutput relationship realized by the component should
be able to change the inputoutput relationship realized by the mechanism as a whole. Again, bad parts, such as quarter-neurons, pass this test
without event. After all, changes to the disposition of a quarter-neuron
can change the relevant disposition of the neuron as a whole. Finally,
consider (B*), which requires that intervening to change the input to
the whole mechanism, such that it will bring about a particular output
of interest, can change the output of the component. Again, this cuts
no ice against the bad parts. If we were to intervene on the whole
by setting the inputs to the whole system in the right way, perhaps by
exposing the system to neurotransmitters, the output of any of the quarter-neurons would change. Consequently, even bad partsthose wed
loathe to consider componentswill pass the MM test, and that test
proves not to be the constraint on components that was needed to fill
out the explanatory account.

Good Parts asScientifically Approved


Given the above difficulties, consider a very different kind of reaction to
the carving problem. This down-to-earth reply is inspired by the explanatory practice of scientists themselves. Scientists dont break up the world
any-which-way but rather have cultivated schemes of division which are
somewhat (though not entirely) uniform within subdisciplines. These
schemes award certain parts a scientific seal of approval. Such a practice might appear to provide a solution to the carving problem, one that
simply insists that it is to these only that mechanistic models must refer.
Machamer etal. (2000) gesture at such a proposal when they write that
the components [are those] that are accepted as relatively fundamental or taken to be unproblematic for the purposes of a given scientist,
research group, or field (13).
While this is a reasonable starting point for an inquiry into partitioning practices, as an answer to the carving problem it should be rejected

64

L.R. Franklin-Hall

as philosophically deflationary. Leaving the solution here is to make ones


philosophical account into a science-reporting task. The philosopher offering it has made little progress in explaining scientific explanatory activity
but has simply insisted thatwith respect to the parts describedgood
explanations are just what competent scientists offer as such. This no
more illuminates the nature of explanation than the cynics account of
speciesaccording to which, species are groups of organisms recognized
as species by taxonomistsilluminates the nature of kinds. The day may
come when philosophers, having failed to solve the carving problem,
should proclaim a cynics slogan. Yet this will be a retreat, and a major
concession with respect to the intelligibility of the scientific enterprise.

The Levels Standard


The final guideline on explanatory mechanistic models favors models that
describe systems at the right level, usually the one just below (in a sense
to be explored) the phenomenon to be explained. There are both reductive and (arguably) non-reductive dimensions to this suggestion. First, in
insisting that phenomena be explained in lower-level termsby describing organized components of mechanisms and their interactionsthe
mechanistic approach to explanation is, undoubtedly, somewhat reductive. However, the approach is also in some measure non-reductive, in
view of advocates resistance to what we might call fundamentalism,
according to which every phenomenon is best explained at the physical level, by a model referring exclusively to physical parts, properties,
and laws. Bechtel, for instance, explicitly contrasts his semi-reductive
mechanistic view with a fundamentalist account, suggesting that knowing how the components [of a mechanism] behave and understanding
how they are organized is sufficient for the purposes of explaining how
the mechanism as a whole behaves (Bechtel 2008: 151) and that, in
most cases, there is no incentive for performing further decomposition
(ibid). Similarly, Craver, while acknowledging reductive dimensions to
the mechanistic approach, still promises to provide mechanists with the
tools to challenge reduction as a normative model (Craver 2007b: 111).

New Mechanistic Explanation andtheNeed forExplanatory...

65

I will call this alternative to fundamentalism the cascade view.


According to it, whole-mechanism behaviors should be explained in terms
of the mechanisms immediate component parts and relations. While such
components can themselves be seen as even smaller mechanisms, and their
behaviors explained using mechanistic models describing each of their
own parts, relations, and dynamic principles, the cascade view denies that
explanations for the functioning of submechanisms (e.g., components)
can be plugged into the explanation for the functioning of the mechanism as a whole. Instead, successively lower-level mechanisms account
for different phenomena. Scientists construct a cascade of explanations,
each appropriate to its level and not replaced by those below (Bechtel
and Abrahamsen 2005: 426). If the cascade view is correct, an endeavor
to explain some phenomenon via a mechanistic model that describes
parts and relations located at a non-adjacent levelsay, one explaining
the regular cardiac rhythm by appeal to a mechanistic model that trucked
in atomic constituentswould blunder; its explanatory power would
be weaker than that of a comparatively high-level model. In this way,
the cascade view rules out the zooming errors from section Formulating
Explanatory Constraints.
This basic take on proper explanatory levels is enormously attractive,
as it appears to mesh perfectly with scientific explanatory practice, particularly in the life sciences. It seems that, with respect to level, scientists
offer just the kind of explanations that the cascade view would recommendreductive but almost invariably just below the phenomenon
to be explained, and far more abstract than fully fundamental ones. Yet,
the move from this feature of explanatory practice to the more ambitious normative claim about explanatory power stated in the previous
paragraphthough naturalis not irresistible. And fundamentalists will
resist it, partly by trying to make sense of this aspect of scientific practice
in pragmatic terms, explaining the fact that the explanations offered by
scientific papers and textbooks are high level while not conceding that
these explanations are objectively superior to fundamentalist ones. For
instance, perhaps full explanations are not offered simply because human
minds are too weak to grasp them at once. More important than the particulars of any error theory is the fact that reductionists will deny that

66

L.R. Franklin-Hall

the lack of fully spelled-out fundamentalist explanations in the scientific


literature should be explained by the fact that such explanations are not,
in principle, explanatorily optimal.
Under pressure from such an alternative, the cascade view requires articulation and defense. In particular, there are two (related) dimensionsone
descriptive and one normativealong which buttressing is mandatory.
The first and most pressing concern, the levels problem, involves simply
filling out the cascade proposal by making sense of what levels are,
including an adjacency relation between them. Though some philosophers can afford to remain silent on this topicand may even deny any
genuinely leveled aspect of nature that explanatory levels could track
(as in Heil 2005; Strevens 2008)the advocate of the cascade view cannot skate over it: it lies at the heart of her scheme.
The second topic, the stop problem, concerns the respect(s) in which
locally reductive explanations are better than those that describe systems in terms of even more basic parts, relations, and principles. At first
blush, fans of the cascade view may try to reject this question and to
shift the burden of proof back to the fundamentalist. Why not instead
insist that she defend her diabolical drive to explanatorily descend to the
basic physical level, rather than resting satisfied with what most scientists actually dole outlocally reductive explanations? While dialectically
tempting, this move is suspect. The cascader and fundamentalist are not
equivalently positioned, as the cascade view is distinctively threatened
with internal inconsistency. This is because the cascader has taken one
step toward reduction, believing as she does that global modelsthose
that treat systems as opaque black boxesare not explanatory, and that
the behavior of a complex system should be explained by breaking it into
organized lower-level parts and their interactionsbut then denies the
value of further deepening. Yet, whatever explanatory oomph the mechanist gets from analyzing systems in terms of their immediate components, it seems she would get even more from analyzing them into their
ultimate components. So, by her own lights, analysis all the way down
to the physical should be preferred. In consequence, mechanists must say
what is gained (or, at minimum, what is not lost) by stopping mechanistic explanations just one level down.

New Mechanistic Explanation andtheNeed forExplanatory...

67

Though I would prefer to explore both of these issues, for reasons of


space I restrict attention to the problem of levels.14 After all, to even
evaluate the mechanists solution to the stop problem, we would need to
know where we are advised to stop our mechanistic decomposition.
Though it is customary to see the world as leveled, just what this
involves is notoriously murky. When levels are judged to be features
of the world rather than features of the units or products of science
(Craver 2007b: 177), they may still be understood in a number of ways.
Lacking the space to consider all options, my focus here will be on the
view of ontological levels clearly ascendant in the new mechanist literature: levels of mechanisms.15 These levels are species of levels of composition, where the composites in question are whole mechanisms. Since
mechanisms are (at least often) embedded within one another, levels
of mechanism lend themselves to an adjacency relation: X is one level
below Y just in case X is an immediate component of the mechanism that is
Y.Founding figures in the mechanistic program have expressed sympathy for this view, with Glennan construing the layers that make up the
world in terms of nested mechanisms (2010a: 363), Craver seeing levels
as levels of mechanisms, in which lower levels are the components
in mechanisms for the phenomena at higher levels (2007b: 170), and
Bechtel sketching a largely comparable view of levels within a mechanism (2008: 147).
This view has three principal features. First, because each of a mechanisms immediate componentsthemselves understood as smaller
mechanismsmay have its own immediate components, which possess components likewise, mechanistic levels can be multiply embedded.
Second, all facts about the relative level of two things will be a joint
function of mind and world; thus, to call these levels ontological or
14

For a critique of the mechanists most promising response to the stop problem, that offered by
difference-making accounts of causal explanation as articulated by Woodward (2003, 2010), and
adopted explicitly by Craver (2007), see Franklin-Hall (2016). For my own positive proposal on
the stop problem, see Franklin-Hall (forthcoming). A recent paper on this problem that came out
too late for me to consider is Harbecke (2015).
15
For a detailed account of the different things philosophers have meant by level, see Craver
(2007, Chap. 5).

68

L.R. Franklin-Hall

features of the world would be, by my lights, to overreach. This follows from the fact that mechanisms themselvesand their componential
specificationsare only well defined (if at all) relative to some behavior.
And no behavior is delivered to us by the world, but must be picked
out by us. Third, even relative to a chosen behavior, questions about the
relative level of any two things can be ill-posed. Such questions are only
kosher when both entities in question are components (either immediate
or otherwise) of the mechanism in question.
The suggestion just sketched offers a more scientifically plausible, and
nuanced, understanding of our folk conception of levels than do the
global, flat stratifications advanced by Oppenheim and Putnam (1958).
And compared to levels defined in terms of the philosophically esoteric
laws, properties, and causeslevels of composition can appear innocent
and straightforward. Furthermore, and of central importance here, levels
of mechanistic composition can be naturally recruited to provide constraints on proper mechanistic explanation, as follows: for any phenomenon that one might want to explain, there is a mechanism responsible
for it, positioned at level n. To explain the phenomenon, an explanatory
mechanistic model should describe entitiesthat is, the immediate component parts of the mechanismat one level down, at n1.
Yet does this proposal address the levels problem, characterizing what
it means for one thing to be one level below another? If so, it is only by
way of a substantial promissory note. The problem follows immediately
from the difficulties already encountered in distinguishing components
or good parts, thus this discussion can be brief. Levels of mechanistic composition are only well defined if linked to an account of what
is required for a part to be an immediate component of a mechanism for
a behavior. Immediate components must themselves meet two conditions. First, they must be genuine components, not gerrymandered parts
or pieces. Second, these good parts must be, in some sense, just below
the mechanism as a whole (level n1), and not components of components (level n2). If supplied with a standard that, for any mechanism for
a behavior, specified all of its nested components, one could make sense
of which components were immediate; however, lacking a distinction
between parts and components, the immediacy requirement is impotent,
having no material on which to work. In light of this lacuna, even those

New Mechanistic Explanation andtheNeed forExplanatory...

69

willing to grant a response to the stop problem, and who see the cascade
view as normatively superior to explanatory fundamentalism, should not
yet consider it to be a genuine alternative; the levels standard cannot,
from the surplus of minimally adequate mechanistic models, tell the
good explanations from the bad.

Conclusion
Though attractive at first glance, none of the new mechanists explanatory guidelines have survived scrutiny, successfully discharging the work
assigned to them. This work, it is worth emphasizing, is extremely difficult. So, even granting that I am right that mechanists have yet to complete it, this hardly shows that their general framework, and particularly
their commitment to causal explanation, is mistaken. Rather, it suggests
that the mechanistic account is but a story half-told. Thus far, proponents
have labeled some important distinctionssuch as between causal and
correlational relationships, between components and mere pieces, and
between appropriate and inappropriate explanatory levels. But the task
of filling them out remains.
As I see it, the present shortcomings of the mechanistic explanatory
account are the flip side of an admirable feature of the mechanism movement, one which has had a salutatory influence on contemporary philosophy of biology (and science): that of taking science (and particularly
biology) seriously. I conclude by recalling the origins of the mechanists
explanatory project, in doing so noting both its merits and its limits.
From early writings to the present day, the new mechanists have been
struck by what appears to be an evident mismatch between the DN analysis of explanation and explanatory practice in the life sciences. On the DN
view, explanations are deductively valid arguments, in which a statement of
the explanatory target is derived from true sentences, including one stating
a law of nature. Reasonably enough, mechanists have found it difficult to
make this jive with what scientists actually did. Where were these supposed
arguments in scientific articles and textbooks? What were these strict laws
in a science like biology, where exceptions are more than just distracting
litter on a landscape of regularity? Something seemed to have gone wrong.

70

L.R. Franklin-Hall

In the face of this apparently nave philosophical precision, the new


mechanists returned to the basics. Rather than imposing a highly regimented account of explanation on the scienceone, quite typically,
reflecting the philosophers penchant for argument and logicwe were
encouraged to look with fresh eyes at the science.16 What kind of explanations did scientists really offer? Immediately clear was that explanations
often showed how things work. Yet in moving beyond that, the situation
became complicated. Scientists obviously provided explanations using a
large number of different representational schemes, with deductive logic
nowhere in view. They described causes but usually talked only of particular activities. And they talked frequently of these things they called
mechanisms but provided no account of what they (in general) were.
All of these are important observations. A rich, scientifically responsible philosophy of science must be accountable to what scientists do
they are our subjects, and their practices, our data. Thus, a mechanist
contribution has been in bringing interesting details of these practices
to philosophical attention, from cell biology to studies of metabolism,
neuroscience, and most recently to systems biology. But what tasks await,
once these phenomena are in view? To say, I will apply to philosophical
practice language that mechanists often use to describe scientific practice.
When studying explanation, philosophers aim not to explain how
things work in the physical world but instead how things work when
scientists show how things work. To do so, philosophers must, after
characterizing the surface features of explanatory practice, pry open its
black boxes, displaying the underlying mechanisms that account for
scientists very explanatory judgments. Is this just what new mechanists
have done? Have they looked under the hood of explanatory practice, and detailed its workings? The results of this inquiry suggest not.
Or, more sympathetically, it suggests that they have peaked under the
hood, but have not yet gotten their hands dirty taking the engine apart.
16

As Lindley Darden explains in her overview of the movement, [t]his work on mechanisms in
biology originated (primarily) not as a response to past work in philosophy of science but from
consideration of the work of biologists themselves, especially in molecular biology and neurobiology
and biochemistry and cell biology (2008: 958959). Similarly, Bechtel writes that these accounts
of mechanistic explanation attempt to capture what biologists themselves provide when they offer
explanations of such phenomena as digestion, cell division and protein synthesis (2007: 270).

New Mechanistic Explanation andtheNeed forExplanatory...

71

Inparticular, rather than opening the black boxes of the scientific enterprisewith respect to causation, part individuation, and explanatory
levelphilosophers have (largely) taken those practices for granted.17
Perhaps this results from a too-successful enculturation of philosophers
into the scientific mindset, making it difficult to achieve the critical distance needed to philosophize about science. If so, while mechanists may
be right that advocates of the DN account were too far from science to
say anything true about it, perhaps the new mechanists have remained too
close to science to say anything surprising about it.

References
Bechtel, W. (2006). Discovering cell mechanisms: The creation of modern biology.
Cambridge: Cambridge University Press.
Bechtel, W. (2007). Biological mechanisms: Organized to maintain autonomy.
In F.Boogerd (Ed.), Systems biology: Philosophical foundations. Amsterdam:
Elsevier.
Bechtel, W. (2008). Mental mechanisms: Philosophical perspectives on cognitive
neuroscience. NewYork: Routledge.
Bechtel, W. (2011). Mechanism and biological explanation. Philosophy of
Science, 78(4), 53357.
Bechtel, W., & Abrahamsen, A. (2005). Explanation: A mechanistic alternative.
Studies in History and Philosophy of Biological and Biomedical Sciences, 36,
42141.
Bogen, J. (2005). Regularities and causality; generalizations and causal explanation. Studies in History and Philosophy of Biological and Biomedical Sciences,
36, 397420.
Bogen, J. (2008). Causally productive activities. Studies in the History and
Philosophy of Science, 39, 11223.
Campaner, R. (2006). Mechanisms and counterfactuals: A different glimpse of
the (secret?) connexion. Philosophica, 77, 1544.
17

There is one mildly ironic exception to my general diagnosis. The only putative black box that
mechanists have opened is the scientists concept of mechanism. On reflection, this focus was
imprudent. Not every concept used by scientists is meaty, and not every term reflects a genuine
black box; mechanism is not a theoretical term within the science, but is a mere pointer, or placeholdersimilar perhaps to the philosophers term conception.

72

L.R. Franklin-Hall

Craver, C. F. (2006). When mechanistic models explain. Synthese, 153,


35576.
Craver, C.F. (2007a). Constitutive explanatory relevance. Journal of Philosophical
Research, 32, 320.
Craver, C.F. (2007b). Explaining the brain: Mechanisms and the mosaic unity of
neuroscience. Oxford: Clarendon.
Culp, S. (1994). Defending robustness: The bacterial mesosome as a test case.
PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association,
1, 4657.
Darden, L. (2008). Thinking again about biological mechanisms. Philosophy of
Science, 75, 95869.
Dowe, P. (1995). Causality and conserved quantities: A reply to Salmon.
Philosophy of Science, 62(2), 32133.
Elster, J. (1989). Nuts and bolts for the social sciences. Cambridge: Cambridge
University Press.
Franklin-Hall, L. R. (2016) High-level explanation and the interventionists
variables problem. British Journal for the Philosophy of Science, 67(2),
553577.
Franklin-Hall, L.R. (forthcoming). The causal economy account of scientific
explanation. In C.K. Waters & J.Woodward (Ed.), Minnesota studies in the
philosophy of science. Minneapolis, MN: University of Minnesota Press.
Glennan, S. (2002). Rethinking mechanistic explanation. Philosophy of Science,
69, S342S53.
Glennan, S. (2005). Modeling mechanisms. Studies in History and Philosophy of
Biological and Biomedical Sciences, 36, 44364.
Glennan, S. (2010a). Mechanisms, causes, and the layered model of the world.
Philosophy and Phenomenological Research, 81(2), 36281.
Hacking, I. (1983). Representing and intervening: Introductory topics in the philosophy of natural science. Cambridge: Cambridge University Press.
Hall, N. (2004). Two concepts of causation. In J.Collins, N.Hall, & L.A. Paul
(Eds.), Causation and counterfactuals (pp.225276). Cambridge, MA: MIT
Press.
Harbecke, J. (2015). Regularity constitution and the location of mechanistic
levels. Foundations of Science, 20(3), 323338.
Heil, J. (2005). From an ontological point of view. Oxford: Clarendon.
Hitchcock, C. (1995). Salmon on explanatory relevance. Philosophy of Science,
62, 30420.

New Mechanistic Explanation andtheNeed forExplanatory...

73

Kaplan, D. M., & Bechtel, W. (2011). Dynamical models: An alternative or


complement to mechanistic explanations? Topics in Cognitive Science, 3(2),
43844.
Kistler, M. (2009). Mechanisms and downward causation. Philosophical
Psychology, 22(5), 595609.
Leuridan, B. (2010). Can mechanisms really replace laws of nature? Philosophy
of Science, 77, 31740.
Levy, A. (2013). Three kinds of new mechanism. Biology & Philosophy, 28(1),
99114.
Levy, A., & Bechtel, W. (2013). Abstraction and the organization of mechanisms. Philosophy of Science, 80(2), 24161.
Machamer, P. (2004). Activities and causation: The metaphysics and epistemology of mechanisms. International Studies in the Philosophy of Science, 18, 2739.
Machamer, P. (2009). Explaining mechanisms. http://philsci-archive.pitt.
edu/5197/.
Machamer, P., Darden, L., & Craver, C.F. (2000). Thinking about mechanisms.
Philosophy of Science, 67, 125.
Mackie, J.L. (1974). The cement of the universe. Oxford: Clarendon.
Mitchell, S.D. (2000). Dimensions of scientific law. Philosophy of Science, 67,
24265.
Nicholson, D. J. (2012). The concept of mechanism in biology. Studies in
History and Philosophy of Science Part C: Studies in History and Philosophy of
Biological and Biomedical Sciences, 43(1), 15263.
Oppenheim, P., & Putnam, H. (1958). Unity of science as a working hypothesis. The philosophy of science, 405427.
Railton, P. (1978). A deductive-nomological model of probabilistic explanation.
Philosophy of Science, 45, 20626.
Robert, J. S. (2004). Embryology, epigensis, and evolution: Taking development
seriously. Cambridge: Cambridge University Press.
Salmon, W. (1984). Scientific explanation and the causal structure of the world.
Princeton, NJ: Princeton University Press.
Skipper, R.A., & Millstein, R.L. (2005). Thinking about evolutionary mechanisms: Natural selection. Studies in History and Philosophy of Biological and
Biomedical Sciences, 36, 32747.
Strevens, M. (2008). Depth: An account of scientific explanation. Harvard:
Harvard University Press.
Waskan, J. (2011). Mechanistic explanation at the limit. Synthese, 183(3),
389408.

74

L.R. Franklin-Hall

Woodward, J. (2001). Law and explanation in biology: Invariance is the kind of


stablity that matters. Philosophy of Science, 68(1), 120.
Woodward, J. (2003). Making things happen: A theory of causal explanation.
Oxford: Oxford University Press.

3
Compositional Explanation:
Dimensioned Realization,
New Mechanism, andGround
KennethAizawa

A hydrogen uoride (HF) molecule has a charge imbalance. It is more


negative on the uorine side than it is on the hydrogen side. This imbalance has both a direction and a magnitude, hence is a vector quantity. It
is the molecules dipole moment. The molecule has its dipole moment in
virtue of the properties of the hydrogen component, the uorine component, and the length of the bond between them. Intuitively speaking, and
simplifying somewhat, it is because of the higher density of protons in the
nucleus of the uorine component than in the nucleus of the hydrogen
component that leads the electrons to cluster on the uorine side of the
molecule thereby giving the HF molecule its dipole moment. The dipole
moment of HF gives us a simple example of the way in which a property of
a wholein this case a moleculeis non-causally explained by properties
of its partsin this case the atomic components of hydrogen and uorine.
There does not appear to be anything unique about dipole moments as
properties that deserve non-causal explanations. Why do corks oat on
K. Aizawa ( )
Department of Philosophy, Rutgers University, Newark, NJ, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientic Composition
and Metaphysical Ground, DOI 10.1057/978-1-137-56216-6_3

75

76

K. Aizawa

water? According to Robert Hooke, it was because they consist of cells


that contain air that prevents water from intruding into them. Again, a
property of a wholein this case a corkis non-causally explained by
properties of its partsin this case the properties of its constituent cells.
The Dimensioned view of realization (Gillett 2002, 2003) provides
a schematic that attempts to characterize the relation among properties
that is implicit in these explanations. Constructed on a version of the
causal theory of properties, the schema holds that
Property/relation instance(s) F1Fn realize an instance of a property G, in
an individual s under conditions $, if and only if, under $, F1Fn together
contribute powers, to s or ss part(s)/constituent(s), in virtue of which s has
powers that are individuative of an instance of G, but not vice versa.

The application of the schema to the dipole moment of HF is simple. G


is the dipole moment (1.91 debye), F1 is the electronegativity of hydrogen
(a dimensionless quantity 2.1), F2 is the electronegativity of uorine (4.0),
and F3 is the bond length between them (0.91 ). The symbol $ includes
background conditions, such as standard temperature and pressure. The HF
molecule is s and its parts are the hydrogen and the uorine components.
Dimensioned realization diers from causation in at least two important ways. First, causes must precede their eects, whereas realizer properties are contemporaneous with their realized properties. Once the
hydrogen and the uoride components with their respective electronegativities are bonded, there is no temporal delay in the production of the
dipole moment of the HF molecule. Second, causes must be wholly distinct from their eects, whereas the dipole moment is not wholly distinct
from the electronegativities. So, Dimensioned realization seems to be
involved in at least some non-causal explanations in science.
Science provides an abundance of instances in which properties of
wholes are thought to non-causally depend on the properties of their
parts. Nevertheless, this feature of science has received strikingly little
attention in two areas of philosophical investigation where it would seem
to be especially germane. First, philosophers of science working on compositional explanations have generally only mentioned properties in passing. Second, metaphysicians working on Grounding explanations have
often overlooked these sorts of cases. Consider these points in more detail.

Compositional Explanation...

77

The New Mechanists in the philosophy of science have devoted considerable attention to the explanation of biological processes, such as phototransduction. (See, e.g., Bechtel and Abrahamsen 2005, p. 423; Craver
2007; Craver and Darden 2013; Glennan 1996, 2002; Machamer et al.
2000; Thagard 2003.) How are photons impinging on the photoreceptors
converted into changes in neurotransmitter release? The answer, according
to many New Mechanists, involves a multitude of entities that engage in
activities. In the case of phototransduction, these entities (or individuals)
include molecules of rhodopsin, transducin, and cGMP phosphodiesterase. During phototransduction, the rhodopsin molecules capture photons,
change conformation, and dissociate into two components. These are the
activities of rhodopsin molecules. One of the components of the rhodopsin molecule, an opsin molecule, activates numerous transducin molecules.
Each transducin molecule next activates many cGMP phosphodiesterase
molecules, which in turn hydrolyze cGMP. In a familiar accounting, the
ontology of compositional explanations includes processes, entities, and
activities, with entities and activities explaining how processes are implemented.1 Properties are typically mentioned only in passing.
The New Mechanist reticence about properties is surprising, given the commonalities between compositional explanations of processes and properties, on
the one hand, and their dierences from causal explanations, on the other. 2
1

Kaiser and Kriekel (forthcoming), document a number of dierent proposals that have appeared in the
New Mechanistic literature regarding what the explananda of interlevel explanations are supposed to be.
2
The New Mechanists typically use the term mechanistic explanation, whereas the term compositional explanation is used here. There are two reasons for this terminological shift. They are,
somewhat paradoxically, that the term mechanistic explanation might be both more restrictive
and more expansive than what is intended here. This is not to say that any single philosopher
simultaneously uses mechanistic explanation both more restrictively and more expansively.
Instead, these usages are more like competing tendencies.
On the more restrictive side, some philosophers may wish to propose that mechanistic explanations just are what, say, Machamer et al. (2000), say they are. It is something like a conceptual
or analytic truth that a mechanistic explanation is an explanation of a process in terms of entities
and activities. Given this, there just cannot be a mechanistic explanation of, say, the properties of a
whole in terms of the properties of the wholes parts. To think otherwise is just a conceptual confusion. The use of compositional explanation is meant to sidestep this concern.
On the more expansive side, there is the fact that mechanistic explanation is now sometimes
used to include causal explanations. As one illustration, Craver and Darden (2013, Chap. 5),
claims that mechanisms produce, underlie, or maintain their phenomena. Producing and maintaining a phenomenon appear to be species of causing that phenomenon, whereas underlying a
phenomenon seems to be composing that phenomenon. (N.B., this use of producing is not the
technical sense of production from Hall 2004.)

78

K. Aizawa

First, recall that causes must precede their eects, but realizer properties
are contemporaneous with their realized properties. Similarly, the entities
and activities that compositionally explain a process are also contemporaneous with the explained process. Phototransduction begins with photon
capture by the photopigment molecules and ends with the change in
release of neurotransmitters. The entire cascade of biochemical reactions
plus neurotransmitter release need not run its course within the photoreceptor cell before phototransduction begins. Second, recall that causes
must be wholly distinct from their eects, whereas realizer properties are
not wholly distinct from the realized property. In like manner, the entities and activities that compositionally explain phototransduction are not
wholly distinct from the phototransduction. It is this lack of distinctness
that sometimes tempts reductionists to claim that a process, such as phototransduction, is nothing more than such and such entities engaging in
such and such activities.
Dimensioned realization, and indeed the whole of the New Mechanistic
literature, has essentially been ignored in the expansive Ground literature. It has been ignored even by advocates of Ground who conceive
of Ground as a generic non-causal determination relation that might
include as species interlevel relations among individuals, properties, processes, and so forth, in nature. Moreover, Ground has been claimed to be
the relation invoked in non-causal explanations. (Of course, Ground has
also been largely ignored in the New Mechanist literature as well.) So,
for example, Clark and Liggins (2012), mention that the brittleness of a
cup results from the conguration of its constituent atoms, then proposes
that Ground is closely related to explanation. Because the brittleness
of a cup results from the conguration of its constituent atoms, we can
explain why the cup is brittle by pointing out its atomic structure. This
is at least pointing in the direction of interlevel property relations of the
sort schematized by the theory of Dimensioned realization. Nevertheless,
no mention of Dimensioned realization or New Mechanism appears in
the literature on Ground. For example, Schaer 2016, proposes that
A second cluster of cases [of Ground] is that of the dependence of the
higher-level on the lower-level. So consider the physical state of Socrates
and his mental state. For the physicalist, the physicalist state realizes the
mental state. And this backs an explanation: Socrates is in this mental

Compositional Explanation...

79

state because he is in that physical state. (Schaer 2016, p. 3.) Again,


Dimensioned realization and the New Mechanist literature appear to be
relevant to the kind of determination relation invoked in the scientic
non-causal explanations sometimes envisioned in the Ground literature,
but these developments are not discussed.
The primary concern of this chapter is to bring Dimensioned realization to the attention of both New Mechanists and Grounders. For
both New Mechanists and Grounders, recognizing compositional explanations involving Dimensioned realization is an important step in the
development of more descriptively adequate accounts of non-causal,
compositional explanations. More specically, Dimensioned realization
shows how New Mechanists might embrace compositional explanatory
relations among properties. Moreover, Dimensioned realization suggests that, at least at times, one needs to move beyond bare grounding
claimsclaims such as X Grounds Yin order to develop a (good)
explanation.3 In principle, the recognition of Dimensioned realization
and its implications might be viewed as friendly additions to work on
New Mechanism and Ground.

New Mechanism andDimensioned Realization


In many quarters of the philosophy of science, there is strong resistance
to metaphysics and to ontological distinctions, such as that between entities (individuals), activities, properties, processes, and so forth. What
philosophers of science should be doing, it might be said, is attempting
to characterize scientic practice, such as how experiments are designed
or how data and hypotheses are represented. Closer to the topic of compositional explanation, what philosophers of science should be doing in
this area is providing descriptively adequate accounts of compositional
explanatory practices in science. How, one might wonder, could arcane
ontological details matter to understanding scientic practice?
3
Schaer, this volume, responds to the Wilson (2014), critique of Ground by, among other things,
agreeing that metaphysicians should want more than bare grounding claims. He also tries to provide something more by appeal to structural equation modeling.

80

K. Aizawa

To a signicant degree, the New Mechanists have already answered


this question. In order to understand certain scientic explanatory practices, one must attend to the ontological presuppositions embodied in
these explanatory practices. To understand how mechanisms gure in science, one needs an account of the ontology of mechanisms. One needs an
account of what mechanisms are.
According to one familiar New Mechanist account, the ontology presupposed in compositional explanations is that there are processes that
are explained by appeal to entities and activities.4 Once we have this processentityactivity ontology on the table, however, we might then ask
whether a more descriptively adequate account of scientic explanatory
practices would include properties. Examples such as the dipole moment
of HF and the oating of a cork suggest that the philosophy of science
does need such an account. Moreover, the Dimensioned view of realization attempts to provide such an account.5
As noted in the introduction, many of the familiar statements of the
ontology of what are here called compositional explanations do not mention properties. One frequently cited account alluded to above holds that
the mechanisms in compositional explanations are entities and activities
organized such that they are productive of regular changes from start
or set-up to nish or termination conditions. (Machamer et al. 2000,
p. 3; see also Craver and Darden 2013; Kindle locations 587588, and
Thagard 2003.) A somewhat dierent proposal is that A mechanism is a
structure performing a function in virtue of its component parts, component operations, and their organization. The orchestrated functioning of
the mechanism is responsible for one or more phenomena. (Bechtel and
Abrahamsen 2005, p. 423.) Component parts seem to be individuals,
component operations are similar to activities, and organization might
refer to the relations among the individuals and their activities.
This is not to say that properties are never mentioned by New
Mechanists. They often are, though only in passing. So, for example,
4

See, for example, Machamer et al. (2000), Craver and Darden (2013).
Strictly speaking, we should consider whether scientic explanation invokes both properties and
property instances, but for present purposes of bringing Dimensioned realization to the attention
of New Mechanists, we may forebear.
5

Compositional Explanation...

81

Machamer, Darden, and Craver write, Mechanisms are composed of


both entities (with their properties) and activities. Activities are the producers of change. Entities are the things that engage in activities. Activities
usually require that entities have specic types of properties. (Machamer
et al. 2000, p. 3.) Moreover, in unpublished work, Machamer provides a
nice example of the role of properties in mechanisms:
Experiments with a large number of odorants suggest that the shape of the
molecule, rather than its chemical composition, determines how it smells
(stereospecicity). The stereochemical theory of smell proposes that the
receptor sites on receptor cells have dierent shapes and that only molecules with a complementary shape t into that receptor site.
Here there is described a mechanism (albeit probable) that explains
how an odorant (or some kind of molecule that produces a particular
smell) binds to a receptor site and evokes an action potential which
explains the end state, here the arrival of the smell signal in the olfactory
bulb. The explanation proceeds by breaking down the olfactory system
into composing entities (molecules, receptor sites, cyclic AMP, olfactory
bulb, etc.), their properties (shape) and their activities (binding depolarization producing). (Machamer, unpublished, p. 11)

These points suggest that Machamer, Darden, and Craver, are, at least in
theory, open to an ontology that includes properties that gure in compositional explanations.
Of course, making passing references to properties and being open to
including them in mechanistic explanations is not the same thing as admitting compositional explanations based on a many-one non-causal determination relation from realizing properties to a realized property. One might
believe there are properties, but doubt that they stand in this sort of relation or that such relations form the basis for a distinct type of explanation.
Nevertheless, there is reason to embrace Dimensioned realization as a basis
for a type of compositional explanation that diers from a well-known
picture of the way in which compositional explanations work.
Craver (2007) has produced an extremely popular schema of the way
in which the process of an entity Ss engaging in an activity of -ing
is explained. One appeals to, say, an entity X1s engaging in an activity
of 1-ing causally inuencing an entity X2s engaging in an activity of

82

K. Aizawa

2-ing, and an entity X3s engaging in an activity of 3-ing, and so forth.


In this schema, it might be that the photopigment molecule X1 undergoes the activity 1 of capturing a photon, which causes X1 to undergo
the activity 2 of changing conformation, which causes X1 to undergo
the activity 3 of dissociating.
Contrast this with the compositional explanation of the dipole
moment of HF. The dipole moment is non-causally determined by the
electronegativity of the hydrogen component, the electronegativity of
the uorine component, and the bond length between them. There is,
however, no causal connection between the electronegativities or bond
length. The electronegativity of uorine does not modify the electronegativity of hydrogen, or vice versa. This is not to say, of course, that
there are no causal relations between, say, the electronegativity of the
hydrogen (uorine) component and any other property. Nor is it to say
that there are no causal interactions among any of the other properties
of the hydrogen and uorine components. There clearly are. Both the
hydrogen and uorine components have specic masses, which causally interact by way of gravitational forces. The present point is that the
dipole moment of HF is not determined by causal interactions among
the cited properties.
The same point might be made through another simple, yet still genuine, scientic example: Hookes discussion of the properties of cork. In
the Micrographia of 1665, he writes,
Next, it seemd nothing more dicult to give an intelligible reason, why
Cork is a body so very unapt to suck and drink in Water, and consequently
preserves it self, oating on the top of Water, though left on it never so
long: and why it is able to stop and hold air in a Bottle, though it be there
very much condensd and consequently presses very strongly to get a passage out, without suering the least bubble to pass through its substance.
For, as to the rst, since our Microscope informs us that the substance of
Cork is altogether lld with Air, and that that Air is perfectly enclosed in
little Boxes or Cells distinct from one another. It seems very plain, why
neither the Water, nor any other Air can easily insinuate it self into them,
since there is already within them an intus existens, and consequently, why
the pieces of Cork become so good oats for Nets, and stopples for Viols,
or other close Vessels. (Hooke 2003, Observation XVIII)

Compositional Explanation...

83

Hooke oered three properties of the cork to be explained: the cork does
not absorb water, it oats, and it has the capacity to stop a bottle. The
property of the cork is explained by the properties of its parts, the cells.
Each of the cells contains air and its walls are perfectly enclosed or
impermeable to air and water. The oating of the cork, for example, does
not depend on causal interactions between the impermeability of one
cell and the impermeability of another. There may, however, be causal
interactions between the impermeability of the individual cells and other
properties. And, of course, causal interactions between other properties
of the cells, such as their masses. What appears not to be relevant to the
explanation of the oating of the cork are causal interactions among the
properties invoked to explain the corks oating.
The upshot is that, in theory, New Mechanists should be willing
to accept that scientists believe in properties. Moreover, they should
be willing to accept that there are cases in which scientists use properties in compositional explanations wherein a property of a whole is
explained by (typically) many properties of the parts. The (small) price to
pay for accepting compositional explanations of the sort envisioned by
Dimensioned realization is that we must accept a picture of interlevel
compositional relations that diers somewhat from the one that is most
familiar in the New Mechanist literature.

Grounding andDimensioned Realization


Introductory comments in this chapter notwithstanding, it is no simple
matter to spell out the potential ramications of Dimensioned realization
relations and compositional explanation for the project of understanding
Ground, since there is diversity of opinion regarding what Ground is and
what features it has. So, for example, Kit Fine begins his important paper,
Guide to Ground, with the contention that
A number of philosophers have recently become receptive to the idea that,
in addition to scientic or causal explanation, there may be a distinctive
kind of metaphysical explanation, in which explanans and explanandum
are connected, not through some sort of causal mechanism, but through
some constitutive form of determination. (Fine 2012, p. 37.)

84

K. Aizawa

Fine does not explicitly consider any instances of compositional explanations in the sciences, and then reject them as not being instances of
Ground. Nevertheless, his map of the landscape seems to exclude the possibility of a non-causal, but nevertheless scientic, form of constitutive
(or compositional) determination found in compositional explanations.
What to say here? In principle, one might simply expand the category
of scientic explanations to include both causal and compositional explanations and continue to distinguish them from metaphysical explanations in terms of Ground. Dimensioned realization and compositional
explanation would then merely be part of a more expansive scientic and
philosophical project tangential to Ground.
One waybut, of course, not necessarily the only wayto defend this
position would be by making the case that scientic causal and compositional explanations invoke a natural necessity, whereas Grounding explanations invoke a metaphysical necessity.6 One might think that there is a
kind of natural necessity according to which the HF molecule has a particular dipole moment in virtue of the electronegativities of the hydrogen
and uoride components and their bond length, but still wonder why
it is that having these electronegativities and bond length leads to this
dipole moment. Thus, there would be a kind of explanatory gapno
a priori connection, saybetween the atomic facts about the hydrogen
and the uorine components and the molecular fact about HF. The interesting or important work for a theory of Ground would come from its
wielding a metaphysical necessity corresponding to the strictest form of
in virtue of relation.
The foregoing example suggests that, in theory, one might set aside the
investigation of compositional explanation and Dimensioned realization
as orthogonal to the investigation of Grounding explanation. In addition, there are, in practice, philosophers who are likely to take this option
in practice. In a nuanced regimentation of a theory of Ground, Audi
(2012), postulates a number of features of Ground that appear to hold
of Dimensioned realization. Both Ground and Dimensioned realization

See Fine (2012, p. 38) for comments that might invite this argumentation. The point here is not
to attribute or recommend this position to Fine, or anyone else, for that matter. Instead, it is merely
to note that there are options for the theorist of Ground.

Compositional Explanation...

85

involve a kind of determination relevant to explanation. Both appear to


be what he describes as worldly, rather than conceptual relations. Both
are asymmetric and irreexive relations. Nevertheless, for Audi, Ground
is explicitly to be distinguished from compositional, mereological conceptions such as Dimensioned realization.
Where [Jonathan Schaer and others] appear to think of grounding more
by analogy with constitution and composition, my emphasis on determination casts grounding as more kindred with causation and explanation. The
function of grounding, then, is to make sense not of a mereological or layered structure of the world, but of a certain kind of explanatory structure.
I posit grounding not as a bridge between the macro and the micro but
as necessary to account for the correctness of certain explanations that cannot be understood as causal explanations. The function of grounding, then,
is to make sense not of a mereological or layered structure of the world,
but of a certain kind of explanatory structure. (Audi 2012, pp. 7089)

So, for Audi, the Grounding project in metaphysics and the compositional project in the philosophical of science are orthogonal.
The foregoing two examples only touch the surface of the diversity
of views regarding Ground. This lack of consensus notwithstanding,
there are at least important segments of the Grounding project for which
Dimensioned realization and compositional explanations are quite germane. These are the more ambitious strains of Ground that attempt to
understand compositional explanations as a species of Grounding explanation. This is the strain apparently embraced by Schaer and more
explicitly by Clark and Liggins.
How might attention to Dimensioned realization and compositional
explanations in the sciences inform the theory of Ground? One thought is
that, in scientic explanations, there is more to having an explanationor
a good explanationthan merely having a putative explanans that determines an explanandum. There are principles of good explanations that
must also be respected. So, for example, some explanations are thought
to be in some sense defective because they are ad hoc. Roughly speaking,
such explanations rely on hypotheses that provide for a determination of
the explanandum by an explanans, but the explanans is defective since it
is tailored specically to handle only a single explanandum. As a second

86

K. Aizawa

example, there is something amiss in an attempt to explain the ability of


a pill to induce sleep by appeal to putative dormitive virtues. In cognitive science, one frequently encounters a version of this as the claim that
one cannot invoke homunculi to explain visual processing. (See, e.g.,
Gibson 1980, p. 60; Pylyshyn 2003, pp. 23.) In principle, there could
be a little person inside the head of an organism that is among the factors determining the organisms capacity to see. But, what, then, of the
common view that there is something inadequate or something missing
from such an explanation? It is prima facie incumbent upon the theorist
of Ground who wishes to assimilate compositional explanations as a species of Grounding explanation to grapple with some of these familiar
principles of good scientic explanation. This might be by denying them
or by providing the machinery needed to understand them.7
We might take the foregoing point a step further by showing how the
schema for Dimensioned realization suggests an analysis of homuncular
explanations.8 Consider a version of the dormitive virtues or homuncular fallacy as it applies to Hookes explanation of why corks oat.
Recall that Hooke explained why corks oat by observing that they are
made up of cells and postulating that the cells contain air and are impenetrable by water. He explained a property of the cork by appeal to qualitatively distinct properties of its constituent cells: oating is qualitatively
dierent from impenetrability, for example. In point of logic, however,
he might have tried to explain why corks oat by observing that they are
constituted by cells and postulating that these cells oat. He might have
claimed that a property of the whole cork is had in virtue of each of its
cellular parts having the same property. After all, the buoyancy of the
cork does appear to be Grounded in the buoyancy of the cells.
Hooke might have ventured an alternative explanation along this line,
but he did not. Why not? One answer might be that he simply failed to
think of explaining the property of the whole cork in terms of qualitatively

As noted in footnote 3 above, Schaer, for his part, is apparently inclined to the latter option.
The following analysis was inspired by Gillett (2010), which draws some important consequences
from the (typical) qualitative distinctness of the properties invoked in compositional explanations
for at views of realization. Credit should also be given to Gillett for helping me rene my articulation of how this account bears on the example from Hooke.
8

Compositional Explanation...

87

similar properties of cells, but that this alternative explanation is just as


good as Hookes actual explanation. Really, the buoyancy of the individual
cells determines the buoyancy of the cork and the buoyancy of the individual cells explains the buoyancy of the whole cork. So, maybe this alternative is really a good explanation.
But, there is, of course, another answer. Perhaps there was some principled reason Hooke had for passing over the explanation that corks oat
because they are made up of cells that oat. Perhaps there is more to a
good explanation than merely citing some facts that Ground some other
fact. Perhaps Hookes reason was that even if one accepts that the buoyancy of the cells Grounds the buoyancy of the cork, one then quickly
comes to wonder why cells oat. Corks oat, because they are made of
cells that oat. That is a bit of progress. But, why do cells oat? Faced
with this second question, one needs to appeal to something else. Perhaps
one needs to appeal to properties that are qualitatively distinct from oating or buoyancy, such as containing air and impermeability to water. The
mystery of oating might be displaced by moving from corks to cells,
but the mystery is not dispelled until one appeals to qualitatively distinct
properties.9 If this is right, there would seem to be more to good explanation than what is Grounded in what.
Here is another way of making the point. A cork has the property of
oating. If one explains this property by appeal to the oating of the
cells, then one does have an explanation of the corks oating. This much
might be captured by Grounding explanations according to which the
oating of a cork is Grounded in the oating of its constituent cells.
But, what is missing? What is missing is an explanation of why there
is a property such as oating, either for a cork or for a cell. To explain
this, one needs to appeal to properties that are qualitatively distinct from
oating. To explain why there is oating one must appeal to more than
bare claims of Ground. One must appeal to Grounds that are in some
sense qualitatively distinct from the Grounded. So, the Dimensioned

Gillett (2002, 2010), notes that the qualitative distinctness of realization properties from a realized property. In various lectures and Gillett (forthcoming), Gillett has described the dispelling of
the mystery idea in terms of the piercing explanatory power or PEP of compositional explanations. Compositional explanations get their PEP by appealing to qualitatively distinct properties.

88

K. Aizawa

realization provides us with the tools to move beyond mere intuitions


that there is something right about explaining the oating of a cork in
terms of the oating of its parts but also something missing in this explanation. Dimensioned realization enables us to say in clear terms what is
correct and what is missing. 10
Summarizing the foregoing, the natural sciences generally distinguish
among better and worse explanations. Ad hoc explanations and homuncular explanations, for example, are defective in some sense that bears
articulation. These kinds of explanations suggest that Grounders must
move beyond bare claims of Groundthat X Grounds Ysince the
defects in these explanations are not want of Ground, but want of something else. In the case of homuncular explanations, we have a concrete
analysis based on the theory of Dimensioned realization. Explanations in
terms of qualitatively distinct properties, as articulated by Dimensioned
realization, enable us to explain why an individual bears a property, but
also how something can come to bear such a property. Philosophers
of science working on compositional explanation have developed the
resources for handling this sort of case. Such resources merit attention
by Grounders.
10

Coincidentally, Craver and Darden (2013, Chapter 6), discuss homuncular explanation drawing
attention to the fact that, in such explanations, a property of the whole is explained by appeal to
the very same property of a part and that such explanations invite a further explanation of why the
part has that property.
Their analysis of what is wrong with homuncular explanations, however, diers from the one
oered here. They treat homuncular explanations as phenomenal models and oer the following
analysis of what is wrong with them:
Phenomenal models are supercial because they specify neither the internal components of
the mechanism nor the organizational and productive features by which the mechanism
works. Mechanistic models have depth. They reveal the internal structure of a mechanism.
This is why they are explanatory, not merely descriptive. (Craver and Darden 2013; Kindle
location 2011f )
By Craver and Dardens lights, what is missing in a homuncular explanation is a specication of the
internal components and the organizational and productive features by which the mechanism
works. Yet, in explaining the oating of a cork by appeal to the oating of its cells, one is, in fact,
specifying the internal components and the organizational and productive features by which the
oating comes about. In theory, however, Craver and Darden could embrace the analysis oered by
the Dimensioned realization. They accept the qualitative distinctness of the explaining processes
from the explained process.

Compositional Explanation...

89

Conclusion
The goal of this chapter has been to introduce Dimensioned realization
as a friendly addition to the toolbox of philosophers of science working
on mechanisms and metaphysicians working on Ground. Dimensioned
realization provides an account of non-causal, compositional relations
among properties that are invoked in scientic explanations of the properties of a whole in terms of the properties of their parts. This is analogous
to what many New Mechanists envision, namely, a non-causal, compositional relation that might be invoked in explanations of the processes of a
whole in terms of processes among its parts. Dimensioned realization also
provides one with the tools to analyze a particular kind of problematic
explanation in the sciences, namely, homuncular explanations. The existence of decient explanations suggests that, in order to explicate what
is decient in some explanations, Grounders need to move beyond bare
claims of Ground. Moreover, Dimensioned realization provides some of
the tools needed to analyze some of these decient explanations.11

References
Audi, P. (2012). Grounding: Toward a theory of the in-virtue-of relation. Journal
of Philosophy, 109(12), 685711.
Bechtel, W., & Abrahamsen, A. (2005). Explanation: A mechanist alternative.
Studies in History and Philosophy of Science Part C: Studies in History and
Philosophy of Biological and Biomedical Sciences, 36(2), 421441.
Clark, M. J., & Liggins, D. (2012). Recent work on grounding. Analysis, 72(4),
812823.
Craver, C. (2007). Explaining the brain. Oxford: Oxford University Press, USA.
Craver, C. F., & Darden, L. (2013). In search of mechanisms: Discoveries across the
life sciences. Chicago, IL: University of Chicago Press.

11

I have beneted in numerous ways from many discussions of these issues with Carl Gillett,
though any mistakes herein are my responsibility. Thanks to Carl and Kelly Trogdon for helpful
comments on earlier drafts of this chapter.

90

K. Aizawa

Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),


Metaphysical grounding: Understanding the structure of reality (pp. 3780).
New York, NY: Cambridge University Press.
Gibson, J. J. (1980). The ecological approach to visual perception. Hillsdale, NJ:
Lawrence Erlbaum Associates.
Gillett, C. (2002). The dimensions of realization: A critique of the standard
view. Analysis, 62(276), 316323.
Gillett, C. (2003). The metaphysics of realization, multiple realizability, and the
special sciences. Journal of Philosophy, 100, 591603.
Gillett, C. (2010). Moving beyond the subset model of realization: The problem
of qualitative distinctness in the metaphysics of science. Synthese, 177(2),
165192.
Gillet, C. (forthcoming). Reduction and emergence in science and philosophy.
Cambridge: Cambridge University Press.
Glennan, S. (1996). Mechanisms and the nature of causation. Erkenntnis, 44(1),
4971.
Glennan, S. (2002). Rethinking mechanistic explanation. Philosophy of Science,
69(S3), 342353.
Hooke, R. (2003). Micrographia, or some physiological descriptions of minute bodies made by magnifying glasses, with observations and inquiries thereupon:
Courier Corporation.
Kaiser, M., & Kriekel, B. (forthcoming). The metaphysics of constitutive mechanistic phenomena. British Journal for the Philosophy of Science.
Machamer, P. (unpublished). Explaining mechanisms. PhilSci Archives.
Machamer, P., Darden, L., & Craver, C. (2000). Thinking about mechanisms.
Philosophy of Science, 67(1), 125.
Pylyshyn, Z. W. (2003). Seeing and visualizing: Its not what you think. Cambridge,
MA: MIT Press.
Schaer, J. (2016). Grounding in the image of causation. Philosophical Studies,
173(1), 49100.
Thagard, P. (2003). Pathways to biomedical discovery. Philosophy of Science,
70(2), 235254.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57(56),
535579.

4
Is Mechanistic Constitution aVersion
ofMaterial Constitution?
JensHarbecke

Introduction
Recent philosophy of neuroscience has made substantial effort to develop
a conceptually sound and descriptively adequate account of mechanistic
constitution. The referent of this term is supposed to be the relation
that successful neuroscientific explanations declare to hold between the
to-be-explained cognitive or neural phenomenon and the explanatory
neural mechanisms underlying the phenomenon. Neuroscientists have
also referred to this relation by terms such as is responsible for (Bliss
and Lmo 1973, 331), gives rise to (Morris etal. 1986, 776), plays a
crucial role in (Davis etal. 1992, 32), contributes to, forms the basis
of (both Bliss etal. 1993, 38), underlies (Lmo 2003, 619; Frey etal.
1996, 703), or is constitutively active in [the phenomenon] (Malenka
etal. 1989, 556).

J. Harbecke (*)
Witten/Herdecke University, Witten, Germany
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_4

91

92

J. Harbecke

The philosophical literature contains several attempts to analyze the


notion conceptually. The most widely known is the mutual manipulability approach offered by Carl Craver (2007, 153). A more recent
one is the regularity account of mechanistic constitution that has been
developed in two versions by Mark Couch (2011) and myself (Harbecke
2010).1 The regularity approach characterizes the relation as reducible to
a particular kind of minimized regularity among mechanistic types. It has
been used for a scientifically informed formulation of supervenience (cf.
Harbecke 2014), for an account of levels in neuroscience and biology
(cf. Harbecke 2015a), and for the explication of a general methodology
for the establishment and integration of neuroscientific and biological
theories (cf. Harbecke 2015b). In this sense, the approach has been transformed into a comprehensive philosophical framework in the context of
neuroscientific explanation.
So far, however, the regularity account of mechanistic constitution
has not been systematically connected to the long-standing debate on
material constitution and the grounding problem (cf. Bennett 2004; Paul
2010; Wasserman 2015). This is at least partially due to a mismatch of
aims. Whereas the debate on mechanistic constitution is primarily concerned with questions about the norms of explanation in the sciences,
material constitution targets a metaphysical relation between individuals.
At the same time, however, for both debates, the notion of mereological
parthood plays a crucial role. Moreover, metaphysical questions about
reduction and autonomy are central to both. In light of these analogies,
it would be surprising if the two notions had no conceptual and logical
connection at all.
In this chapter, my aim is to unravel the similarities and differences,
and in particular the logical and conceptual connections, of the regularity account of mechanistic constitution and the standard accounts of
material constitution. The goal is to show that the two relations are of a
different logical order, and that this holds independently of whether one
chooses a pluralist or a monist interpretation of material constitution.
At the same time, there are several interesting analogies and connections
See also Gillett (2007, 2013), whose dimensioned view of realization and constitution concept is
closely linked to the regularity theory of constitution.
1

4 Is Mechanistic Constitution aVersion ofMaterial...

93

between the two notions. In a final step, I demonstrate that, once the
metaphysics presupposed by the regularity account is accepted, puzzles
such as the statue-and-lump case disappear. They receive an eliminativist
solution, which may be attractive in various respects.
The investigation proceeds as follows. In a first step, I reconstruct the
context in which the question about mechanistic constitution arises (section The Question of Mechanistic Constitution). I then review the
philosophical enquiry associated with mechanistic constitution (section
The Mechanistic Approach), which includes a discussion of the regularity theory (section What Is Mechanistic Constitution?) and of identity statements about phenomena and mechanisms (section Identity).
Subsequently, I review the problem of material constitution and the
grounding problem (section The Question of Material Constitution).
I then show that material constitution is to be distinguished from
mechanistic constitution (section Is Mechanistic Constitution Material
Constitution?) while there are various logical and conceptual connections between the two notions (section Connections). In a final step,
I suggest that the ontology presupposed by the regularity approach to
mechanistic constitution offers an informative eliminativist solution
to the problem of mechanistic constitution and grounding (section
Mechanisms and the Grounding Problem). The last section summarizes the argument and raises some open questions that the present chapter was unable to answer (section Conclusion).

The Question ofMechanistic Constitution


Taking explanatory practice in physics as a starting point, much philosophy of science in the twentieth century presupposed that generally
accepted explanations in the special sciences conform either to a deductivenomological (cf. Hempel and Oppenheim 1948; Hempel 1942,
1965) or to a unificationist model of explanation (cf. Friedman 1974;
Kitcher 1989). Both of these models assign laws of nature a central role
in explanatory practice in the sense that they demand any acceptable
explanans to contain at least one law of nature.

94

J. Harbecke

From the 1990s onwards, this general picture of explanation in the special sciences has been challenged, notably through the works of authors
such as Bill Bechtel and Bob Richardson (1993), Machamer etal. (2000),
Bill Bechtel and Adele Abrahamsen (2005), and Carl Craver (2002,
2007). What unites these authors is their interest in explanations in the
neurosciences. In their view, neuroscientific explanations essentially focus
on the mechanisms constituting a to-be-explained phenomenon. Lawlike
regularities subsuming the phenomenon play a secondary role in an explanation, if they play a role at all (however, cf. Fazekas and Kertsz 2011).
The views of these authors were inspired by close examinations of various cases of successful neuroscientific inquiry. The following case from
the cognitive neuroscience of memory is an example in this sense.

Representation ofHead Direction


Normal rats perform extremely well in repeated orientation, path finding, and path integration tasks. Among the behavioral phenomena
observed and studied during the last decades are successful orientation in
the radial-arm maze (cf. Eckerman etal. 1980), the Oasis maze (cf. Clark
etal. 2005), and the Morris water maze (cf. Morris 1984). The salient
ability of rats to solve orientation tasks in such environments is presumably due to the availability of an advanced representation and memory
system allowing storage of, and access to, various kinds of spatial information over longer periods of time.
In an attempt to connect the rats spatial representational and learning abilities to a neural basis, Morris et al. (1982) were able to show
that lesions in the rats hippocampus are correlated with an impairment of spatial memory acquisition. Subsequently, it was demonstrated that a selective inhibition of hippocampal N-methyl-D-aspartate
receptor (NMDA-receptor) activity leads to an impairment similar to
that induced by hippocampal lesions (Morris et al. 1986; Davis et al.
1992). Parallel to this research, attempts were made to develop models of the rats memory system based on the knowledge of the components of the representational system. A milestone in this respect was
the discovery by Ranck in 1984 of head direction cells (HD cells)

4 Is Mechanistic Constitution aVersion ofMaterial...

95

in the postsubiculum of the rat (Ranck in foreword to Wiener and Taube


2005). HD cells were subsequently investigated in detail by Rancks student Taube who identified HD cells also in the lateral mammillary nuclei
(LMN) (Taube etal. 1990a, b; Taube 1995). HD cells display a maximal
firing rate when the rats head points toward a specific direction in a given
environment. Typically, sloping levels of response can be identified over
45 on either side of the target.
Two models have been proposed concerning the neural networks that
may govern this kind of behavior of HD cells. A continuous attractor
network was presented as a potential candidate by Skaggs etal. (1995).
A state in a network of this kind is stable only if there is a single localized
cluster of active cells while all other cells are inactive. One way to design
such a network is to ensure strong excitatory connections between neighboring cells, and strong inhibitory connections between distant cells. The
problem with a network of this kind for the case at hand is that it assumes
recurrent connections which do not seem to exist in LMN.By focusing
on the primary sources of the head direction signal in the dorsal tegmental nucleus (DTN), Song and Wang (2005) constructed a computational
model with reciprocal DTNLMN connections, in which different
DTN neuron populations respectively inhibit LMN neurons on the right
and on the left. The activity of DTN neurons is updated by vestibular
inputs carrying information about angular head velocity (AHV). With
no AHV information, a local LMN activity is stable within the network.
With input about a change in AHV, the LMN activity moves in a way
that tracks the head direction. So far, this model has matched well onto
the available empirical data.

Between Mechanisms andPhenomena


The philosophical debate represented by Bechtel and Richardson (1993),
Machamer etal. (2000), Craver (2002, 2007), Glennan (1996, 2002),
and others has interpreted examples as the one above as violating the
norms of the deductivenomological ideal of explanation. The described
phenomenon of head direction representation is not explained by the
identification of an antecedent condition that connects to the phenom-

96

J. Harbecke

enon via a general law. Rather, the explanation is essentially based on the
identification, location, and analysis of the mechanisms that constitute a
to-be-explained phenomenon. The mechanists have taken this observation as the basis both for the formulation of a descriptive, or naturalist,
theory of successful and satisfactory explanations in neuroscience and for
the formulation of an explanatory ideal that is declared to be normatively
binding for future research in neuroscience.2
A conceptually transparent statement of both the descriptive theory
and the normative ideal requires a prior clarification of the following two
concepts: It has to be determined what a mechanism is, and it has to be
specified what it means for a mechanism to constitute a phenomenon.
It is here where the genuine philosophical work begins. A large part of
the recent debate on the mechanistic approach has focused on these two
issues, accordingly.

The Mechanistic Approach


As mentioned in the previous section, the mechanistic approach consists in the formulation of a descriptive theory of successful explanation in neuroscience as well as a normative ideal that explanations are
demanded to conform to. Both dimensions of the approach make central reference to the notion of a mechanism and to the relation of
constitution which is supposed to hold between a mechanism and
a to-be-explained phenomenon. This section offers some analyses of
the notions recently developed in the philosophical literature (sections
What Is a Mechanism? and What Is Mechanistic Constitution?)
and discusses questions of identity and reduction on their basis (section
Identity).
The descriptive thesis of successful and satisfactory explanations in neuroscience essential says that
Neuroscientists tend to accept only mechanistic explanations as genuine explanations. The normative thesis states that Neuroscientists should accept only mechanistic explanations as genuine
explanations. Note that answering these questions might bring up descriptive and normative
issues at a different conceptual level. One may have to answer the following questions: What do/
what should neuroscientists take mechanisms to be? and What does it/what should it mean for a
mechanism to constitute a phenomenon?
2

4 Is Mechanistic Constitution aVersion ofMaterial...

97

What Is aMechanism?
The most popular definitions of the term mechanism are the following
ones. Machamer etal. describe it as consisting of entities and activities
organized such that they are productive of regular changes from start or
set-up to finish or termination conditions (Machamer etal. 2000, 3).
Bechtel and Abrahamsen extend this definition by describing a mechanism as a structure performing a function in virtue of its component
parts, component operations, and their organization. The orchestrated
functioning of the mechanism is responsible for one or more phenomena (Bechtel and Abrahamsen 2005, 423).
Both formulations have their weaknesses. For instance, the distinction
between an entity and an activity can be tricky in scientific contexts as
many scientifically interesting entities are primarily defined over their
activities. Electrons, for instance, are often described as particles. At the
same time, in physics, they are primarily characterized by what they do.
But even if the absence of a clear distinction introduces some vagueness
into the definitions, they are generally considered as serving their purpose
well enough. In fact, they do seem to capture sufficiently well the metaphysical nature of the NMDA-receptor activity that is invoked to explain
head direction representation (cf. section The Question of Mechanistic
Constitution). And so, I will accept the definitions for the remainder of
the chapter.

What Is Mechanistic Constitution?


It is important to distinguish between what, in their definition of a
mechanism, Machamer etal. (2000, 3) call being productive of regular
changes or what Bechtel and Abrahamsen (2005, 423) characterize as
being responsible for one or more phenomena from causing a phenomenon or event. The latter relation holds between phenomena or events in
a diachronic manner. The former relation, in contrast, is believed to hold
between phenomena and underlying mechanisms. The idea seems to be
that the mechanisms synchronically realize, or instantaneously determine, the initially identified phenomenon such that the relata do not

98

J. Harbecke

have spatiotemporally distinct instances (cf. Craver and Bechtel 2007,


547).3 It is this relation that is now usually described as mechanistic
constitution.4
The philosophy of neuroscience literature currently contains two
main proposals for an analysis of the notion: a manipulationist theory
defended notably by Craver (2007), and a regularity theory proposed
in two versions by Couch (2011) and myself (Harbecke 2010).5 As I have
shown in the latter chapter (2010, 272), the manipulationist account
suffers from an aggravating unsatisfiability problem. The point is that,
when a mechanism in the same place and time as a given phenomenon
is manipulated, there is no fact of the matter as to whether the manipulation affected the phenomenon indirectly through the mechanism or
directly. However, the notion of an ideal intervention in the sense of
the manipulationist account essentially demands the possibility of distinguishing direct and indirect manipulations. Hence, the manipulationist
account is not satisfiable in constitutive contexts (cf. also Baumgartner
and Gebharter 2014). Due to this reason, I am mainly going to be concerned with the regularity account.
The regularity-based analysis of mechanistic constitution explicitly declares general constitution, that is, constitution between types of
mechanistic events, as the primary analysandum. Singular constitution,
that is, constitution between instantiations of mechanisms, is to be interpreted derivatively in terms of general constitution. Moreover, it commits
itself to an ontology that contains objects or entities in a weak sense only.
What the definitions above called components and their activities are

Not having distinct instances does not imply that the instances are identical; it merely says that
the spacetime regions instantiating the relata of mechanistic constitution overlap. Note that two
spacetime regions can overlap without being identical in the case that they do not perfectly overlap (i.e. if they are not both a mereological part of the respective other).
4
The term composition has been used by Machamer etal. (2000, 13), Bechtel and Abrahamsen
(2005, 426), and Craver (2007, 164); constitution occurs in Craver (2007, 153); constitutive
relevance is found in Craver (2007, 139). It is safe to say that the authors intend these terms
widely synonymously. For the sake of terminological unity, from now on, I will use the term constitution to denote the relation that is referred to by these expressions.
5
As mentioned in footnote 1, Gillett (2007, 2013) has developed a similar analysis for the notion
of realization.
3

4 Is Mechanistic Constitution aVersion ofMaterial...

99

transformed into compactivities so to speak.6 For instance, the type is


activated that is instantiated, among other things, by NMDA-receptors
becomes the type hosts an active NMDA-receptor that is instantiated by spacetime regions. Spacetime regions are the only individuals
accepted into the ontology of this framework.7
For the formal definition of general mechanistic constitution, I have
used Greek letters and to quantify over mechanistic types and
phenomena (i.e. and are second-order variables). Furthermore,
capital letters X, X1 X2 , Xn are used to express conjunctions of
types that can be co-instantiated (either in the same individual or in colocated individuals). The formulation goes as follows (cf. Harbecke 2010,
275278; to improve legibility, type conjunctions such as f X1 are
always abbreviated to X1):
Mechanistic ConstitutionA mechanistic type constitutes another
mechanistic type and/or phenomenon (written as C) if, and only
if:
(i)
 is contained in a minimally sufficient condition X1 of , such
that
(ii) X1 is a disjunct in a disjunction fX1 X 2 X n of type conjunctions minimally sufficient for , such that the disjunction is
minimally necessary for , such that
(iii) if and X1 are properly co-instantiated, then (a) their instances are
a mereological part of an instance of , and (b) this instance of is
a mereological part of the mentioned fused instances of and X1.

See also Kaiser and Krickel (2016) for the idea that the relata of mechanistic constitution are
hybrids between entities and activities.
7
This certainly is a non-standard way of speaking of mechanisms. Craver, for instance, likes to
characterize mechanisms as an activity -ing of one or more objects/entities denoted by x, y
and so on. In other words, mechanisms, according to Craver, are (kinds of ) events or processes
(where processes are spatiotemporally extended events). Nevertheless, since science is typically
interested not in events but in kinds of events, the commitment to an ontology consisting of space
time regions and mechanistic types may be more adequate to reconstruct pertinent explanations.
6

100

J. Harbecke

The main idea underpinning this definition is that mechanistic constitution is a relation between mechanistic types that are regularly, but
not redundantly, co-instantiated such that their instances are mereologically related. The mereological theory presupposed here is General
Extensional Mereology (GEM) as explicated by Varzi (2009). Since often
no single mechanism is sufficient for the occurrence of a given phenomenon, the definition makes reference to complex mechanisms involving
a range of mechanistic properties. Additionally, since sometimes more
than one mechanism can secure the occurrence of the given phenomenon, the definition also allows for alternative constitutive conditions.
The mereological requirement is introduced in order to ensure that the
phenomenon must occur (at least partially) in the same place and time
as the mechanisms that constitute it. All of these ideas are expressed by
conditions (i)(iii, a).
Condition (iii, b) is added for the following three reasons. First, it captures an intuitive relationship between the mechanisms realizing a given
phenomenon in the sense that an instantiation of the former is demanded
to occupy no less space or time than the phenomenon induced. It would
be strange to think that when a set of neural mechanisms constituting a
learning process by the rat occurs, that learning process partly takes place
outside the rat. Or if it actually does, condition (iii, b) demands that also
further lower-level mechanistic aspects outside the rats body must be
relevant. In particular, the condition excludes that there is such a thing as
a phenomenon partially or wholly occurring in the void (unless it itself is
a fundamental mechanism).8

It is an interesting question whether Craver (2007), as one of the most famous mechanists, would
accept conditions (iii, a) and (iii, b). He does say that mechanisms sometimes transcend or transgress the clear boundaries of a system. For instance, his view of the action potential is that it relies
crucially on the fact that some components of the mechanism are inside the membrane and some
are outside (141). At the same time, he points out that [o]ne cannot delimit the boundaries of
mechanismsthat is, determine what is in the mechanism and what is notwithout an account
of constitutive relevance (141). Since his definition of constitutive relevance explicitly demands a
mereological relationship of the mechanisms instance and the phenomenons instance, it seems
clear that Craver as well believes that no phenomenon can occur unless there is a mechanism that
occurs wherever the phenomenon occurs. Or in short, if the mechanism for a phenomenon transgresses the boundaries of a system, so does the phenomenon, and vice versa. Hence, the textual
evidence in Craver (2007) suggests that he accepts (an analogous version of ) condition (iii).
8

101

4 Is Mechanistic Constitution aVersion ofMaterial...

Second, without condition (iii, b), the mechanisms underlying a


given phenomenon P would not only constitute P but also many of Ps
effects. The reason is that, if the mechanisms are minimally sufficient for
P, they are typically also minimally sufficient for some of those phenomena for which P is minimally sufficient in a causal non-overlapping sense.
This consequence we should avoid, as it would render the constitution
relation too permissive.
The third reason is that, with condition (iii, b), Mechanistic
Constitution provides a criterion for reduction. Since mutual parthood
implies identity, mutual constitution, for instance, of a conjunction of
types X1 with a type ensures that X1 and are not only sufficient
and necessary for another but also coextensive. If nomological coextensiveness is accepted as sufficient for reduction, an identity of types can be
inferred (for more details, cf. section Identity).
Note that the regularity theory of mechanistic constitution subsumes
mechanisms and phenomena under the same ontological category,
namely the category of first-order types. This is desirable because what
is a mechanism for a given phenomenon can itself be considered a phenomenon whose underlying more detailed mechanisms can be analyzed
and explicated.
To see how the definition interprets explanations presented by working scientists, consider again how the representation of head direction
by rats was made understood by a specification of the underlying neural
mechanisms (cf. section The Question of Mechanistic Constitution). If
the second-order operator c is used to capture the criteria specified
by Mechanistic Constitution (from now on, we will sometimes speak of
a type conjunction being c-minimally sufficient for another type in this
sense) and Y1, Y2, , Yn are used to express disjunctions of conjunctions of properties all of which are minimally sufficient for the type on the
right-hand side of the conditional, the above-mentioned research results
on head direction representation in rats can be expressed by the following
proposition (HDR abbreviates head direction representation).

PHDR : ( FX1 Y1 c G ) ( GX 2 Y2 c H ) ( HX 3 Y3 c I ) ,

102

J. Harbecke

where, the type symbols are intended as follows:


I The representation of head directions
H 
The activation of a reciprocal DTNLMN network receiving
AHV inputs
G The firing of HD cells in the postsubiculum
F The activation of NMDA-receptors at the synapses of HD cells
Proposition PHDR says: (If NMDA receptor activation at the synapses
of certain neurons within a rats postsubiculum is instantiated together
with certain other properties in an appropriate way, then a firing of HD
cells in the postsubiculum occurs) and (If an activation of HD cells in the
postsubiculum occurs together with certain other properties, then) and
The central hypothesis associated with the regularity theory of constitution is that this is essentially the relationship that scientists have in mind
when they say that a mechanism is responsible for, gives rise to, plays
a crucial role in, contributes to, forms the basis of, underlies, or is
constitutively active in a phenomenon (cf. section Introduction).

Identity
An important aspect of the regularity theory of mechanistic constitution
is the fact thatunder the presupposition of a nomological coextensiveness principleit provides an empirical criterion for the identity of constitutively related mechanistic types. This is a valuable result in light of
the fact that, for instance, in the philosophy of mind and the philosophy
of the special sciences, it has sometimes been argued on purely conceptual grounds that higher-level types are (non)-reducible and (non)-identical to lower-level types.
The following example demonstrates how the regularity theory provides an empirical criterion for the identity of types. Suppose that the
mechanistic type F stands (as before) for the activation of NMDA-
receptors, F1 stands for calcium/calmodulin-dependent protein kinase
II (cf. Malenka etal. 1989), F2 stands for calcium-dependent protease,
and G stands for the phenomenon of the firing of HD cells. Moreover,

4 Is Mechanistic Constitution aVersion ofMaterial...

103

s uppose that the following constitutive hypothesis was established on the


basis of a series of empirical difference tests (cf. Harbecke 2015b):

G
PHDR
: FF1 F2 X1 Y1 c G

Suppose further that it is determined on the basis of further experiments that the type conjunction FF1F2 is c-minimally sufficient for G and
that, hence, X1 is in fact empty. In this case, the initial material theory
PHDRG will be transformed into the following one:

G
PHDR
: ( FF1 F2 Y1 c G )

The establishment of hypothesis PHDRG still leaves open whether Y1


is empty or not. In other words, PHDRG provides no information as to
whether the type conjunction FF1F2 is the only c-minimally sufficient
condition of G. Some researchers may find out that it is not empty as
there are non-NMDA-receptor-dependent firings of certain HD cells
perhaps in other hippocampal regions. However, it is also conceivable
that at some future point in the history of neuroscience, a feeling will
spread among researchers that everything is known about the hippocampus and head direction representation while no further c-minimally sufficient conditions for G have been identified. In such a case, the scientific
community might agree that Y1 is in fact empty9 and that PHDRG should
be transformed into the following hypothesis:

G
PHDR
: FF1 F2 c G

It should be mentioned that, since philosophers are part of the scientific community, it is not
entirely likely that the community will reach this conclusion for any phenomenon. Philosophers
such as Jerry Fodor might insist that there is always an open disjunctive list of possible constituters
for any phenomenon (cf. for instance 1974). Consequently, Y1 is never empty, even if in this world
no further complete constituting condition is found. In my view, this modal argument is irrelevant
for science as we know it. In particular, the proponents of this argument can still agree that the
emptiness of Y1 supports at least what can be called actual, or this-worldly, reduction. The latter, I
submit, is the only reduction of interest for science, and it is the one characterized here.
9

104

J. Harbecke

From PHDRG, it can be inferred that G and the type conjunction


FF1F2 are coextensive. The reason is that the mechanistic type conjunction FF1F2 is both minimally sufficient and minimally necessary for G (cf.
conditions (i) and (ii) of Mechanistic Constitution), and the individuals
instantiating FF1F2 are always a part of an individual instantiating G, and
vice versa (cf. conditions (iii) and (iv) of Mechanistic Constitution).
Hence, no individual instantiates G that does not also instantiate FF1F2,
and vice versa. Of course, the coextensiveness described in this way is a
this-worldly, or nomological, one. However, since lawful coextensiveness
is reasonably considered as sufficient for type identity (cf. Mellor 1977,
308309), it can be inferred that G = FF1 F2 .
On the other hand, should it turn out that Y1 of the original hypothesis is in fact not empty, the type conjunction FF1F2 is not the only c-
minimally sufficient condition for G. The type conjunction FF1F2 is then
not coextensive with G and, hence, not identical to G. This shows how
the theory of mechanistic constitution provides an empirical criterion for
the identity and non-identity of mechanistic types related by mechanistic
constitution.

The Question ofMaterial Constitution


Metaphysicians have long been interested in what today is known as the
problem of material constitution and the grounding problem. The
two problems are now standardly discussed with reference to the puzzle
of the statue and the clay and related puzzles (cf. Wasserman 2015). A
sculptor, so the puzzle goes, purchases a piece of clay on Monday. As it
happens, he calls it Lump. On Tuesday, he forms Lump into a statue
holding a lute and calls it David. The statue called David is now coincident with the lump of clay called Lump in the sense that the former
shares all material parts and properties with the latter. But is David identical to Lump, or are there reasons to believe that the two are coincident
but non-identical? And should David turn out to be non-identical to
Lump, what grounds the difference between the two? These are the questions philosophers have aimed to solve.

4 Is Mechanistic Constitution aVersion ofMaterial...

105

Pluralists have voted for a non-identity on the basis of the fact that,
even though the two objects share all their categorical properties, David
seems to have certain properties that Lump lacks. Among these may be
what are often called sortalish properties, which include (i) persistence
conditions, particularly modal properties like being essentially shaped
about like so, (ii) kind or sortal properties, and (iii) properties that things
have partially in virtue of their instantiation of properties in categories (i)
or (ii) (Bennett 2004, 341). In particular, David seems to instantiate the
properties is essentially a statue, could have partially been made out of
granite, did not exist on Monday, and so on, while Lump lacks these.
According to Leibniz law, a necessary condition for identity is the sharing
of all properties, and so the non-identity of the two objects follows.
The pluralists argument for a non-identity has the following general
form (where s and l are individual constants, x and y are individual variables, P(_,_) stands for the parthood relation, and is a
property variable):
Premise 1: Object s and object l share all of their parts, but there is at
least one property that s has and l lacks. (in short:
"x ( Pxs Pxl ) $f ( fs fl ) )
Premise 2: If two objects are identical, then they share all properties (in
short: "x"y ( x = y "f ( fx fy ) ) ).
Conclusion: Object s is not identical to l, even though the two share all
of their parts. (in short: s l "x ( Pxs Pxl ) )
The argument is formally valid. Hence, if the premises are true for
some s and l, the general result is established that it is possible for two
objects to occur in the same place and time and yet be non-identical. The
fact that one object is materially constituted by the other may imply that
the constituted object is not over and above the constituting object.
Nevertheless, the two are not the same thing. The general result obviously
contradicts the extensionality theorem that is part of the mereological
system, extensional mereology (EM), (and all systems containing it; cf.
Varzi 2009): "x"y ( x = y "z ( Pzx Pzy ) ) , and so it asks for a particular kind of mereology.

106

J. Harbecke

Different versions of the above argument use modal or temporal operators, such that s and l share all properties in this world or now, but one
out of s and l could lack, or will lack, a property that the other one has
and will continue to have. The conclusion remains the same. Moreover,
the argument can be developed for many other puzzles such as the persons
and their bodies puzzle or the ship of Theseus puzzle (cf. Wasserman 2015).
An interesting aspect of the metaphysical theory supported by versions of the above argument is the asymmetry that is typically associated
with material constitution. The sameness-of-parts relation as expressed in
Premise 1 is symmetric, and Lump clearly has certain sortalish properties that David lacks. But then why doesnt David materially constitute
Lump? And what is the further condition that one of the two objects
must satisfy to receive the primacy underlying the asymmetry?
These questions hint at the fact that the sentence The lump of clay
called Lump constitutes the statue called David is misleading. The pluralist cannot possibly mean that the relation of material constitution holds
between Lump and David. Rather, she (presumably) means that material
constitution holds between a set or collection of material clay parts, atoms
perhaps, and David, as well as between the set of clay parts and Lump. The
asymmetry then stems from the fact that both David and Lump can cease
to exist in this world, but the set or collection of material parts cannot.10
Monists have attacked the pluralists picture from several angles. First,
they have argued that it seems strange for two things to share all their
parts and yet fail to be identical. In virtue of what, so the monist asks,
does David have those further properties that Lump, or the collection
of clay parts, lacks? What grounds the sortalish properties if it is not the
categorical properties shared with Lump, or with the collection of material clay parts? If it is further material parts or properties not shared by
the collection, then the statue would fail to coincide with the collection
of clay parts. However, if no material parts or properties ground Davids
sortalish properties, then the collection of clay parts should not have any
different ones than the statue. Unless, of course, the sortalish properties
are simply projected onto the world by humans, or unless they are primitive properties.
See also Paul (2010, 583) and what she calls the mereology puzzle for material constitution.

10

4 Is Mechanistic Constitution aVersion ofMaterial...

107

If sortalish properties are projected onto the world, they dont seem
to be sufficiently ontic. There would not be a statue out there. Rather,
the properties would be in the eye of the beholder. On the other hand, if
Davids sortalish properties are primitive, the existence of a statue in the
same place and time as the collection of clay parts cannot be established
by any independent argument (but, see Bennett 2004).
Moreover, monists have claimed that the argument equivocates at least
one term, which would render the argument inconclusive (but, see Fine
2003). Finally, they have argued that, once we accept statues, the universe
will be overpopulated quickly. The point is that the statue and the collection of clay particles do not seem to be the only inhabitants of that space
time region. Also the arms-torso-head-legs object seems to be coincident
with the clay. Moreover, the forearm-upperarm-cribs-calves-feet-torso-head
object seems to coincide with it. But all of these objects are distinguished
from the others by at least some sortalish properties. In short, it is difficult for the pluralist to avoid becoming a bazillion-thinger (Bennett
2004, 358).11
As a reaction, monists have sometimes denied the existence of macroscopic objects altogether. What exists, they claim, are fundamental material things, perhaps atoms, perhaps more fundamental entities. These
entities can be configured and combined in various ways. But no such
configuration or combination gives birth to a new macroscopic object
(cf. Unger 1979). This position has sometimes been called mereological
nihilism or eliminativism.
A deflationary approach may open up a third way in between pluralism and monism. Such an approach points out that both pluralists and
monists agree on all the non-statue facts, such as the fact that there is a
particular collection or set of material entities present where the statue
is supposed to be present. Their argument over the existence and non-
existence of the statue on top of the collection of material entities is therefore mostly a verbal one. It changes nothing about the basic facts.
So, in short, material constitution is a relation that exists in an ontic
sense only in the eyes of pluralists. From this perspective, it is a relation
that holds between a set or collection of material things and an object
11

But see Rea (1998) for the position that this consequence is unproblematic.

108

J. Harbecke

such that the object has no material parts not contained in the collection.
The following definition captures the idea (where, x, y, z, x, and y are
individual variables, and P(_,_) signifies the parthood relation):
Material ConstitutionA collection of material things or particles x
materially constitutes an object y (written as Cxy) if, and only if:
(i) for all z, if z is a part of y, then it is an element of x, and
(ii) x is not identical to y, and
(iii) there is at least one collection of material things or particles that is x
and at least one object that is y.
In short
Cxy def "z ( Pzy z x ) ( x y ) $x $y ( ( x = x ) ( y = y ) )

It is important to see that also eliminativists and deflationists can


accept this definition. It is just that, in the eliminativists view, no collection of material things x and object y ever jointly satisfy all conditions
(i)(iii) of Material Constitution. In particular, if all material parts of an
existing object are contained in a collection of material things (condition
(i)), while the object is not identical with the collection (condition (ii)
perhaps because it has some sortalish properties that the collection lacks),
then the object does not exist (negation of condition (iii)).

Is Mechanistic Constitution Material


Constitution?
After the relations of mechanistic constitutionin the sense of the regularity theoryand of material constitution have been characterized in
sufficient detail, the question can be approached whether the two are
perhaps the same relation. As the following comparison shows, the
answer is clearly negative: they are neither the same nor of the same kind.
The most obvious reason (Reason 1) for distinguishing the two relations as defined above by Mechanistic Constitution and Material

4 Is Mechanistic Constitution aVersion ofMaterial...

109

Constitution is that they are relations of a different logical order.


Mechanistic constitution (expressed above by the two-place type of types
C(_,_)) was defined as a second-order relation between mechanistic
types, while material constitution (expressed above by the two-place type
C(_,_)) was characterized as a first-order relation between objects or
individuals (the presumption was that a collection of things is considered
as an individual as well).
One might perhaps argue that the distinction between first- and
second-order relations is a purely formal one. For instance, one might
presume that, on the basis of an ontology of tropes or abstract particulars, also the relation of material constitution could be defined as second
order.12
The problem with replacing classical properties by tropes and defining
individuals as collections of tropes is that one then needs to say something about sortalish properties. Are these tropes as well? If yes, then
David and the collection of clay particles no longer coincide. The statue
simply contains more tropes than the collection of clay particles. Or, are
sortalish properties not tropes? In that case, the statue would seem to
be wholly identical to the collection of clay particlesunless on top of
tropes, the ontology contains somehow classical properties as well among
which are the sortalish ones. In short, a trope solution may align mechanistic constitution and material constitution. But it does so either for the
price of material constitution collapsing into identity or for the price of
presupposing an ontology that is unduly abundant.
The second reason (Reason 2) for the non-equivalence of material
and mechanistic constitution is the fact that, at least according to the
above definitions, mechanistic constitution involves a minimization constraint (condition (i) of Mechanistic Constitution) that is irrelevant for
material constitution. Material Constitution demands that the constituted objects parts are all contained in the collection of material particles.
However, it does not demand that all elements of the collection be parts
of the constituted object.
This difference could be easily annihilated, of course. Even though
the definitions of material constitution found in the literature, at least to
12

Paul (2006) might be interpreted as choosing this option.

110

J. Harbecke

my knowledge, do not involve such a minimization constraint, the latter


could be introduced into Material Constitution without doing harm to
the general idea.
A third reason (Reason 3) for the difference between mechanistic and
material constitution is the fact that the respective questions about identity and reduction can be answered empirically for the relata of mechanistic constitution (cf. section Identity). In contrast, and as Ladyman
(2012) points out, it is hard even to imagine a way in which science
could be relevant to the debate about whether a statue is identical with
the lump of clay out of which it is fashioned (39). The only method that
seems to be available to decide about an identity or non-identity of the
statue and the collection of clay particles is one of conceptual analysis.
The fourth reason (Reason 4) for the non-equivalence is the fact that
the modal behavior of David and the collection of clay particles is usually considered central to the nature of material constitution. The statue
David, so the story goes, could have existed even if some of its clay particles would have been replaced by iron particles, as long as the surface
structure, color, texture, and so on of David would have remained the
same. Hence, only when also all the possible worlds inhabited by David
are examined can it be decided whether David is identical to the collection of clay particles or not.
The methodology associated with Mechanistic Constitution (cf.
Harbecke 2015b), in contrast, presupposes that, whether a mechanistic type is identical to another mechanistic type (or a conjunction of
types) is a matter of this-worldly coextensiveness. Behind this principle
lies the observation that Quines classical example of the allegedly coextensive properties has a heart, or is a cordate, and the property has
a kidney, or is a renate (Quine 1986, 8; cf. also Lewis 1997, 173)
is misleading. First of all, it is false that in the actual world, all renates
are cordates (some humans have an electrical pump instead of a heart).
However, even if this were the case: Science has various strategies at its
disposal to distinguish non-identical properties in this world, many of
which are based on manipulations of things and properties in the actual
world. For instance, when properties appear always in the same individuals but not in the same place and time within individuals, science reliably distinguishes them. With these observations in the background, the

4 Is Mechanistic Constitution aVersion ofMaterial...

111

sufficiency that the methodological theory associated with Mechanistic


Constitution ascribed to nomological coextensiveness with respect to
type identity looks more convincing.
The fifth reason (Reason 5) for a logical, conceptual, and, finally, ontic
difference between the two relations lies in the fundamental ontologies on
the basis of which the respective theories and definitions are formulated.
In the debate on material constitution, monists and pluralists disagree on
the existence of non-fundamental objects. However, they usually agree
on the existence of the same fundamental objects or individuals, such as
clay particles and atoms. These are taken to inhabit spacetime regions
and to have properties.
The regularity account of mechanistic constitution works in a systematic way only because it allows into its ontology only spacetime regions
as individuals and mechanistic properties as types (cf. section What Is
Mechanistic Constitution?). The latter kind of types are declared to be
hybrids of entities and activities. Whereas in natural and scientific language, the subject would typically characterize an individual such as a
NMDA-receptor that instantiates the type characterized by the predicate
is activated, the ontology used by Mechanistic Constitution accepts
only types that fuse entities and activities and that are instantiated by
spacetime regions: hosts an active NMDA-receptor that is instantiated by a region r. This way of speaking is discontinuous even with pertinent publications in the field. Scientists are happy to speak of receptors,
circuits, brain regions, and so on as objects. They usually do not speak of
mechanistic types being instantiated by spacetime regions.
However, the translation of the talk of receptors, circuits, brain regions,
and so on instantiating various activities into the mechanistic type ontology characterized above is pretty straightforward. The fact that an object
x (belonging to the class of Xs) instantiates a property P can always be
translated into the fact that the spacetime region s occupied by x instantiates the property of the P-ing of an X. Hence, everything science might
say about mechanisms is expressible on the basis of a mechanistic type
ontology.
At the same time, presuming such an ontology has several advantages.
For one thing, it is highly systematic and simple. It does not need to
bother itself with questions such as whether an electron is primarily to be

112

J. Harbecke

considered an object or an activity. Second, it avoids potential type/token


confusions. A sentence such as The firing of certain cells in the postsubiculum partially constitutes the representation of head direction in the
rat can in principle be read as a claim about this particular postsubiculum. However, the project of science should always target the general.
The mechanistic-type ontology by its very structure secures a focus on
types and general relationships.
An interesting consequence of these metaphysical commitments is
that the existence of clay particles or atoms as individuals is generally
denied. Atoms as well as clay particles are aspects of mechanistic types,
but they do not exist primarily, independently, or ipsa natura. Colloquial
individuals and objects including pieces of clay are rather more or less
artificial delineations of ingredients of mechanisms with a certain degree
of persistence. They may be instrumentally useful, but ultimately objects
in this substantial sense are mere projections.
As a consequence, neither the pluralist nor the monist position on the
puzzle of the statue and the clay can actually be formulated on the basis of
the ontology presupposed by Mechanistic Constitution. The relation of
material constitution as defined by Material Constitution is not satisfiable within such an ontology. In such a world, it is never instantiated.
As a side remark, note also that the parthood relation presupposed
by Mechanistic Constitution for spacetime regions does recognize
the extensionality theorem of EM (and all systems containing it, such as
GEM; cf. section What Is Mechanistic Constitution?). It is not clear to
me, however, whether the ontology presupposed by the debate on material constitution recognizes this principle as well for spacetime regions.
To my knowledge, the pertinent literature has not paid any attention to
this question.
To conclude, there are important differences between the relation of mechanistic constitution and material constitution (named
above Reasons 15), which suggests that the two relationsat least if
adequately characterized by Mechanistic Constitution and Material
Constitutionare fundamentally different in nature, and their relata
(if they exist at all) are fundamentally distinct. In fact, given the different

4 Is Mechanistic Constitution aVersion ofMaterial...

113

ontologies presupposed in the debates on the two notions and relations,


it is not even clear that each one is reconstructible in terms of the other.

Connections
The previous section focused on the differences between mechanistic constitution and material constitution as they have been defined in sections
What Is Mechanistic Constitution? and The Question of Material
Constitution. This section will highlight some interesting analogies
between the two relations.
The most striking one is clearly the fact that the metaphysical theory associated with Mechanistic Constitution faces a similar problem
as the one brought forward against pluralists in the debate on material
constitution. It seems difficult for the pluralist to resist the conclusion
that, for any collection of n fundamental entities or particles, there are
at least 2n objects (partially or wholly) materially constituted by it. For
instance and as pointed out already in section The Question of Material
Constitution, not only the statue seems to be constituted by the collection of clay particles. Also the arms-torso-head-legs object seems to be
coincident with the clay. But all of these additional objects are distinguished from the others by at least some sortalish properties. In short, it
is difficult for the pluralist to avoid becoming a bazillion-thinger.
Mechanistic constitution as defined above faces a similar problem of
ontological inflation. It seems that for almost any phenomenon p and any
mechanism m constituting p, another more fine-grained mechanism can
be described that constitutes both m and p. Consequently, a proponent of
Mechanistic Constitution and the metaphysical theory associated with
it will rapidly become a bazillion-mechanism theory with respect to
any investigated phenomenon.
The only solution to this problem seems to be a balance between different ontologies in terms of their simplicity and explanatory strength.
However, it is not clear at this point whether the problem can be
adequately addressed by the mechanistic approach based on Mechanistic
Constitution.

114

J. Harbecke

A second correspondence between mechanistic constitution and material constitution has already been mentioned in section Is Mechanistic
Constitution Material Constitution?. For both relations, the notion
of a spacetime region plays an important role. Phenomena and their
constituting mechanisms are characterized as sharing the same space
time region by definition. An object materially constituted by a collection of material entities is typically described as sharing the same place
and time with the latter. Mechanistic Constitution actually presupposes the applicability of GEM to spacetime regions. Above, it was left
open whether material constitution could follow along with this idea,
and thereby accept the extensionality theorem for spacetime regions at
least. Unfortunately, however, it was not even clear whether, for instance,
statue-and-clay pluralists would accept the characterization of spacetime
regions as individuals in the first place.
A third correspondence consists in the fact that both mechanistic
and material constitution are to be distinguished from supervenience.
As I have shown elsewhere (Harbecke 2014), mechanistic constitutional
claims are not equivalent to supervenience claims. Since material constitution is a relation between individuals, whereas supervenience is a relation between sets of types, it seems clear that the former is not equivalent
to the latter as well.

Mechanisms andtheGrounding Problem


The previous sections have shown that, despite certain analogies, mechanistic constitution and material constitution are not the same relation. In
particular, the two ontologies associated with them are radically different. This section aims to show, however, that the fundamental ontology
presupposed by the regularity definition of mechanistic constitution may
offer an interesting solution to the problem of material constitution: the
theory of object eliminativism.
As it was pointed out in section What is Mechanistic Constitution?,
the ontology presupposed by Mechanistic Constitution accepts only
spacetime regions and mechanistic types in its ontology, where the latter types involve an individual aspect at best. In other words, the ontology

4 Is Mechanistic Constitution aVersion ofMaterial...

115

does not recognize objects as primary at all. Objects are essentially projections on persisting mechanisms, and they do not enjoy any ontic status.
Perhaps surprisingly, such an ontology offers an eliminativist solution to
the problem of material constitution that still allows to make sense of the
pluralists and monists main claims. Whereas the compositional nihilists
(Paul 2010, 586) deny the existence of macroscopic objects and recognize
the existence of fundamental objects, an ontology as the one described
denies the existence of objects toto imperio. Not even the fundamental
objects exist qua objects.
Such a picture has sometimes been brought forward as an alleged
reductio argument against the eliminativist or nihilist stance (cf. Sider
1993; Zimmerman 1996; Schaffer 2003). However, the belief that our
everyday objects including statues and persons are in fact processes is less
absurd than it may seem at first. For one thing, metaphysical analysis
has found it notoriously difficult to clarify what an object actually is qua
object, since all objects that we encounter seem to be inseparable from at
least some of their properties.
Second, the boundaries of non-fundamental objects are typically
vague, and determining where a statue begins and ends is almost an
impossible task. Third, the traditional human hope to find at the bottom
of things atoms or some other kind of fundamental objects has been continuously disappointed by twentieth-century physics. As a consequence,
some philosophers of physics have reverted to a structural realist position
that accepts only relations as the fundamental building blocks of things
(cf. French and Ladyman 2003; Esfeld and Lam 2008). Hence, accepting
objects into ontology looks less appealing in the twenty-first century than
it might have looked 100 years ago.
The point is that, if statues, lumps of clay, and even clay particles considered as objects are merely more or less arbitrarily delineated chunks of
processes, the conceptual problem of material constitution disappears.
Whether the statue-process type is then identical to the collection-ofclay-particles-process type are identical becomes a problem open for
empirical investigation (cf. section Identity). Moreover, should the
strategy for choosing the best system out of a set of multilevel ontologies be successful, it may be possible to answer the question whether the
statue-process type has ontic status at all.

116

J. Harbecke

Of course, such a radicalized eliminativist solution to the problem


of material constitution would be very different from the traditional or
canonical accounts. In particular, modal properties would no longer play
any significant role for the puzzle. There would simply be no objects that
could have them. And when two mechanistic types constitute another,
according to the extensionality-and-identity principle it becomes irrelevant whether the one could have occurred without the other. The question is only whether it does occur sometime at some place without the
other.
Nevertheless, since it could turn out that the statue-process type is not
identical to the collection-of-clay-particles-process type, this solution can
indirectly make sense of the nagging intuition that has been driving the
entire debate on material constitution: that somehow the statue seems
to be a different thing than the collection of clay particles. Moreover, it
may explain why different people have reached very different conclusions
with respect to that problem: monists and eliminativists on the one side,
pluralists on the other. These two groups of people have simply made
different predictions about the empirical test of the statue type and the
clay type. At this point in history, however, neither can claim to have
conclusive evidence on their side. In other words, the solution offered
here can indirectly make sense of the intuitions and central claims of both
positions. It simply adjourns a final verdict on the truth of either one
until the day that the empirical question can be solved. In this sense, it
may be an interesting solution to puzzles such as the one about the statue
and the piece of clay.

Conclusion
The main aim of this chapter was to analyze the potential similarities
and differences of the regularity account of mechanistic constitution and
the standard accounts of material constitution. In a first step, a descriptive explanatory project from the neurosciences was reviewed in order to
frame the context in which the question about mechanistic constitution
is believed to arise. Subsequently, the regularity theory of mechanistic

4 Is Mechanistic Constitution aVersion ofMaterial...

117

constitution was reconstructed as a potential analysis of the relation. In a


third step, the problem of material constitution and the grounding problem were described, and a definition of the relation was offered that is
acceptable to both pluralist and monist solutions to these problems.
The comparison of the two relations of mechanistic and material constitution was at the focus of the next step. It was argued that the two
relations are of a different logical order. Moreover, the criteria used to
decide under which conditions the two relations collapse into identity
were described as fundamentally different. In a final step, it was shown
that the ontology presupposed by the regularity approach to mechanistic
constitution offers an informative eliminativist solution to the problem
of mechanistic constitution and grounding. Once the metaphysics presupposed by the regularity account is accepted, puzzles such as the lump/
statue case no longer occur.
Due to limits of space, the chapter had to leave certain further questions unanswered. Among these are the question about an independent
justification of the ontologies associated with Mechanistic Constitution
and Material Constitution. Is there a way to reject one of these ontologies and accept the other that is independent of the puzzles discussed
in this chapter? Moreover, is the coextensiveness-and-identity principle
characterized in section Identity feasible? These questions and problems will be the targets of future research on the mechanistic approach
and the problem of material constitution.

References
Baumgartner, M., & Gebharter, A. (2014). Constitutive relevance, mutual
manipulability, and fat-handedness. British Journal for the Philosophy of
Science. Advance online publication. doi:10.1093/bjps/axv003.
Bechtel, W., & Abrahamsen, A. (2005). Explanation: A mechanist alternative.
Studies in History and Philosophy of Biological and Biomedical Sciences, 36(2),
421441.

118

J. Harbecke

Bechtel, W., & Richardson, R. (1993). Discovering complexity: Decomposition


and localization as scientific research strategies. NewYork: Princeton University
Press.
Bennett, K. (2004). Spatio-temporal coincidence and the grounding problem.
Philosophical Studies, 118(3), 339371.
Bliss, T., Collingridge, G., etal. (1993). A synaptic model of memory: Long-
term potentiation in the hippocampus. Nature, 361(6407), 3139.
Bliss, T., & Lmo, T. (1973). Long-lasting potentiation of synaptic transmission
in the dentate area of the anaesthetized rabbit following stimulation of the
perforant path. The Journal of Physiology, 232(2), 331356.
Clark, R., Broadbent, N., & Squire, L. (2005). Hippocampus and remote spatial memory in rats. Hippocampus, 15(2), 260272.
Couch, M. (2011). Mechanisms and constitutive relevance. Synthese, 183(3),
375388.
Craver, C. (2002). Interlevel experiments and multilevel mechanisms in the
neuroscience of memory. Philosophy of Science, 69(3), 8397.
Craver, C. (2007). Explaining the brain. NewYork: Oxford University Press.
Craver, C. F., & Bechtel, W. (2007). Top-down causation without top-down
causes. Biology & Philosophy, 22(4), 547563.
Davis, S., Butcher, S., & Morris, R. (1992). The NMDA receptor antagonist
d-2-amino-5-phosphonopentanoate (d-AP5) impairs spatial learning and
LTP invivo at intracerebral concentrations comparable to those that block
LTP invitro. Journal of Neuroscience, 12(1), 2134.
Eckerman, D., Gordon, W., Edwards, J., MacPhail, R., & Gage, M. (1980). Effects
of scopolamine, pentobarbital, and amphetamine on radial arm maze performance in the rat. Pharmacology Biochemistry and Behavior, 12(4), 595602.
Esfeld, M., & Lam, V. (2008). Moderate structural realism about space-time.
Synthese, 160(1), 2746.
Fazekas, P., & Kertsz, G. (2011). Causation at different levels: Tracking the commitments of mechanistic explanations. Biology and Philosophy, 26(3), 119.
Fine, K. (2003). The non-identity of a material thing and its matter. Mind,
112(446), 195234.
Fodor, J. (1974). Special sciences (or: The disunity of science as a working
hypothesis). Synthese, 28(2), 97115.
French, S., & Ladyman, J. (2003). Remodelling structural realism: Quantum
physics and the metaphysics of structure. Synthese, 136(1), 3156.
Frey, U., Frey, S., Schollmeier, F., & Krug, M. (1996). Influence of actinomycin
D, a RNA synthesis inhibitor, on long-term potentiation in rat hippocampal
neurons invivo and invitro. The Journal of Physiology, 490(Pt 3), 703.

4 Is Mechanistic Constitution aVersion ofMaterial...

119

Friedman, M. (1974). Explanation and scientific understanding. The Journal of


Philosophy, 71(1), 519.
Gillett, C. (2007). Hyper-extending the mind? Setting boundaries in the special
sciences. Philosophical Topics, 34(1/2), 161187.
Gillett, C. (2013). Constitution, and multiple constitution, in the sciences:
Using the neuron to construct a starting framework. Minds and Machines,
23, 309337.
Glennan, S. (1996). Mechanisms and the nature of causation. Erkenntnis, 44(1),
4971.
Glennan, S. (2002). Rethinking mechanistic explanation. Philosophy of Science,
69(3), 342353.
Harbecke, J. (2010). Mechanistic constitution in neurobiological explanations.
International Studies in the Philosophy of Science, 24(3), 267285.
Harbecke, J. (2014). The role of supervenience and constitution in neuroscientific research. Synthese, 191(5), 725743.
Harbecke, J. (2015a). Regularity constitution and the location of levels.
Foundations of Science, 20(3), 323338.
Harbecke, J. (2015b). The regularity theory of mechanistic constitution and a
methodology for constitutive inference. Studies in History and Philosophy of
Science Part C: Studies in History and Philosophy of Biological and Biomedical
Sciences, 54, 1019.
Hempel, C.G. (1942). The function of general laws in history. The Journal of
Philosophy, 39, 3548.
Hempel, C.G. (1965). Aspects of scientific explanation. In C.G. Hempel (Ed.),
Aspects of scientific explanation and other essays in the philosophy of science
(pp.331496). NewYork: Macmillan.
Hempel, C., & Oppenheim, P. (1948). Studies in the logic of explanation.
Philosophy of Science, 15(2), 135175.
Kaiser, M., & Krickel, B. (2016). The metaphysics of constitutive mechanistic
phenomena. British Journal for the Philosophy of Science. Advance online publication. doi: 10.1093/bjps/axv058.
Kitcher, P. (1989). Explanatory unification and the causal structure of the world.
In P. Kitcher & W. Salmon (Eds.), Scientific explanation (pp. 410505).
Minneapolis: University of Minnesota Press.
Ladyman, J. (2012). Science, metaphysics and method. Philosophical Studies,
160(1), 3151.
Lewis, D. (1997). Modal realism at work: Properties. In D.Mellor & A.Oliver
(Eds.), Properties (Oxford Readings in Philosophy) (pp. 173187). Oxford:
Oxford University Press.

120

J. Harbecke

Lmo, T. (2003). The discovery of long-term potentiation. Philosophical


Transactions of the Royal Society. London. B.Biological Science, 358,
617620.
Machamer, P., Darden, L., & Craver, C. (2000). Thinking about mechanisms.
Philosophy of Science, 67(1), 125.
Malenka, R., Kauer, J., Perkel, D., Mauk, M., Kelly, P., Nicoll, R., & Waxham,
M. (1989). An essential role for postsynaptic calmodulin and protein kinase
activity in long-term potentiation. Nature, 340(6234), 554557.
Mellor, D.H. (1977). Natural kinds. British Journal for the Philosophy of Science,
28(4), 299312.
Morris, R. (1984). Developments of a water-maze procedure for studying spatial learning in the rat. Journal of Neuroscience Methods, 11(1), 4760.
Morris, R., Anderson, E., Lynch, G., & Baudry, M. (1986). Selective impairment of learning and blockade of long-term potentiation by an N-methyl-d-
aspartate receptor antagonist, AP5. Nature, 319, 774776.
Morris, R., Garrud, P., Rawlins, J., & OKeefe, J. (1982). Place navigation
impaired in rats with hippocampal lesions. Nature, 297(5868), 681683.
Paul, L. (2006). Coincidence as overlap. Nos, 40(4), 623659.
Paul, L. (2010). The puzzles of material constitution. Philosophy Compass, 5(7),
579590.
Quine, W.V. O. (1986). Philosophy of logic. Cambridge, MA: Harvard University
Press.
Rea, M. C. (1998). In defense of mereological universalism. Philosophy and
Phenomenological Research, 58(2), 347360.
Schaffer, J. (2003). Is there a fundamental level? Nos, 37(3), 498517.
Sider, T. (1993). Van inwagen and the possibility of gunk. Analysis, 53, 285289.
Skaggs, W., Knierim, J., Kudrimoti, H., & McNaughton, B. (1995). A model of
the neural basis of the rats sense of direction. In D.Touretzky & T.Leen
(Eds.), Advances in neural information processing systems (Vol. 7, pp. 173
180). Cambridge, MA: MIT Press.
Song, P., & Wang, X. (2005). Angular path integration by moving hill of activity: A spiking neuron model without recurrent excitation of the head-
direction system. The Journal of Neuroscience, 25(4), 10021014.
Taube, J. (1995). Head direction cells recorded in the anterior thalamic nuclei
of freely moving rats. The Journal of Neuroscience, 15(1), 7086.
Taube, J., Muller, R., & Ranck, J. (1990a). Head-direction cells recorded from
the postsubiculum in freely moving rats. I. Description and quantitative
analysis. The Journal of Neuroscience, 10(2), 420435.

4 Is Mechanistic Constitution aVersion ofMaterial...

121

Taube, J., Muller, R., & Ranck, J. (1990b). Head-direction cells recorded from
the postsubiculum in freely moving rats. II.Effects of environmental manipulations. The Journal of Neuroscience, 10(2), 436447.
Unger, P. (1979). There are no ordinary things. Synthese, 41(2), 117154.
Varzi, A. (2009). Mereology. In E.N. Zalta (Ed.), Stanford encyclopedia of philosophy (Summer 2009). http://plato.stanford.edu/entries/mereology/
Wasserman, R. (2015). Material constitution. In E.N. Zalta (Ed.), The Stanford
encyclopedia of philosophy (Spring 2015). http://stanford.library.usyd.edu.au/
entries/material-constitution/
Wiener, S., & Taube, J. (2005). Head direction cells and the neural mechanisms of
spatial orientation. Cambridge, MA: The MIT Press.
Zimmerman, D. W. (1996). Could extended objects be made out of simple
parts? An argument for atomless gunk. Philosophy and Phenomenological
Research, 56(1), 129.

5
Anti-Reductionism, Anti-Rationalism,
andtheMaterial Constitution
oftheMental
DerkPereboom

Ive set out and defended an account of the vertical relations between
the mental and more fundamental levels in terms of a theory of material constitution (Pereboom 2002, 2011). A controversial feature of
this account is that it rejects identity as the distinctive interlevel relation, by contrast with standard reductive positions and, perhaps surprisingly, with the rival nonreductive subset view. Instead, I appeal
to a fundamental made up of relation. Critics have argued that this
account is objectionable because this relation is obscure. Such a criticism is fueled by a rationalist presumption that accounts of this sort
can only appeal to conceptual analysis and logical relations such as
identity. I reject the obscurity criticism, and more generally, the rationalist presumption.

D. Pereboom ()
Cornell University, Ithaca, NY, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_5

123

124

D. Pereboom

Introduction
Physicalism about the mental requires that all mental entities be appropriately founded in microphysical entities. Supposing the truth of physicalism, it remains an open question whether the relation between the
microphysical and the mental is reductive or nonreductive. By contrast
with the mid-twentieth century, currently, most nonreductivists maintain that the main reason for accepting the nonreductive option is not
methodological but metaphysical. On the position I endorse, mental
natural kinds are not identical to natural kinds in microphysics because
mental causal powers are not identical to microphysical causal powers.
The fact that mental kinds are multiply realizable at the level of microphysical kinds provides an important reason to believe this is so. The view
I propose departs from other contemporary nonreductivisms insofar as
it rejects the token identity of mental and microphysical entities of any
kindincluding causal powers.
The most fundamental relation between the mental and the microphysical is material constitution, understood in such a way that it does
not feature identity (Pereboom 2002, 2011). Material constitution will
be a relation between material entities, and thus entities that can be
causally efficacious. It can be thought to qualify as a kind of grounding
(Fine 2001; Schaffer 2009) or building relation (Bennett 2011) on the
supposition that such relations are characterized in a general way, just by
asymmetry, irreflexivity, one-way necessitation, and as giving rise to generative explanations of the less fundamental by the more fundamental.
My particular conception of material constitution essentially involves
the made up of relation, which holds, for example, between a statue
and the lump of clay (Pereboom 2011, 2013a). The made up of relation
is itself asymmetric and irreflexive: the lattice is not made up of the
diamond, and the diamond is not made up of itself. It has a particular
directedness: the less fundamental is made up of the more fundamental,
and not vice versa. Crucially, at its core, the made up of relation is primitive and not characterizable in terms of more fundamental relations.
Here is a formal characterization of this notion of material constitution. Suppose that x and y are concrete physical entities. Entities x and y

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

125

are materially coincident just in case they, at some level, are made out of
the same parts. Then,
(C1) x materially constitutes y at t if and only if
(a) y is made up of and materially coincident with x at t;
(b) necessarily, if x exists at t, then y exists at t and is made up of and
materially coincident with x at t; and
(c) possibly, y exists at t and it is not the case that y is made up of and
materially coincident with x at t.
The last provision (c) rules out the identity of x and y (on the assumption of the necessity of identity), as does clause (a), since the made up of
relation is irreflexive.
Carl Gillett (2002, 2003) distinguishes between a flat notion of realization, in which properties of a thing are realized by properties of that
same thing, and a dimensioned conception, in which properties of a thing
are realized by properties of a distinct thing from which it is constituted,
and one might adapt this distinction to the notion of material constitution as Ive characterized it. Sydney Shoemaker (2003, 2007, 2014)
invokes a similar distinction between a variety of property realization in
which property F of X is realized by property G of the same thing X, and
microrealization, in which property F of X is realized by a microphysical
state of affairs, which consists in microphysical entities having certain
properties. Material constitution allows for Gilletts dimensioned view
and Shoemakers notion of microrealization. If M is constituted by P,
P might be a microphysical state that consists in microphysical entities
having certain properties. In my view, there will also be states of affairs at
a level higher than the microphysicalthe chemical and neural levels, for
examplethat constitute M.
Lynne Bakers discussion of constitution features several counterexamples that pose a challenge to clause (b) of this characterization, the necessitation of the constituted entity by its constitutor (Baker 2007: 1113,
106110). The existence of the dollar bill in my wallet is not necessitated
by its underlying microphysical arrangement, because its existence also
requires the United States Federal Reserve Bank and the laws governing

126

D. Pereboom

it. On an externalist view about mental content of the sort advocated


by Tyler Burge (1978), the existence of a token belief with some specific
content will not be necessitated by the existence of its microphysical constitutor, for the reason that in an appropriately different physical or social
environment, this same microphysical constitutor would not necessitate
a belief with that content. This type of phenomenon can be accommodated by a characterization similar to (C1), but in which, in accord with
Bakers (2000) account of constitution, (b) is altered to specify that the
existence of y is necessitated by the existence of x in an appropriate relational context, and (c) is similarly revised. Formally, suppose D designates the y-favorable circumstancesthe relational context required for
a constitutor to constitute y. Then:
(C2) x materially constitutes y at t if and only if
(a) y is made up of and materially coincident with x at t;
(b) necessarily, if x exists and is in D at t, then y exists at t and is made
up of and materially coincident with x at t; and
(c) possibly, y exists at t and it is not the case that y is made up of and
materially coincident with x in D at t.
Baker (2013) adduces another reason to prefer (C2) to (C1): its preferable, she argues, that the constitutor and the constituted be the same type
of entity; thus an object can be constituted of an object, but an object cant
be constituted by a state of affairs. The object head-and-hammerlets
call it HHwhich can exist even if its head-part and the hammer-part
are detached, would be a candidate for constituting the hammer, but HH
arranged hammer-wise would not be, for the reason that HH arranged
hammer-wise is a state of affairs and not an object. Im not sure Id want to
grant thisperhaps an object can be constituted of a microphysical state
of affairs. But if Bakers specification is granted, it seems that constitutors
may not necessitate the constituted thing, for the obtaining of HH alone
will not necessitate the existence of the hammer. An attractive response
is to specify that the hammer is indeed constituted just by HH, but only
when it has a hammer-wise arrangement. And this result can be secured
by adopting (C2), and allowing the y-favorable circumstances for which

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

127

D stands range over circumstances of part arrangement, and not only


over external sorts of relational contexts (Pereboom 2013b).
Finally, I prefer a model of the mental that is not functional in the
external-relations sense, on which the essence of a mental state consists
in its causal relations to entities external to themsensory inputs, behavioral outputs, and other mental states. Instead, I propose that mental
properties are identical to physical compositional properties, that is,
properties things have solely by virtue of intrinsic features of their parts,
either proper or improper, and the relations these parts have to each
other. This model accommodates the causal efficacy of mental states in
a way that external-relations functionalism cannotintuitively, purely
relational properties alone cannot be causally efficacious. This model also
accommodates nonreductivismmultiple realizability arguments show
that mental compositional properties would not be essentially neural or
microphysical. Given the identities it specifies between mental properties and compositional properties, this position can aptly be conceived
as a compromise with the typetype reductionist proposals of U.T.Place
(1956) and J.J.C.Smart (1959).
On a scientific realist conception, the temporally forward-looking
causal relata of natural kinds are explained not simply by external functional relations, but rather in central cases by properties intrinsic to those
kindsthat is, by properties intrinsic to every possible instance of the
kindand standardly by intrinsic properties at that kinds level of scientific description (Boyd 1999). Forward-looking causal relata of chemical compounds are explained in part by their compositional properties,
chemical properties intrinsic to those compound kinds. Polio symptoms
are explained partly by intrinsic biological properties of that disease type.
Analogously, properties intrinsic to mental state types will explain their
forward-looking causal relata.
This model can be illuminated by analogy to internally complex artifacts, for instance, a ball piston engine, a type of rotary internal combustion engine (Pereboom 2002, 2011). Characteristic of this type is having
parts with specific shapes and rigidities, and arranged in a particular way.
These features make up a compositional property of this engine type.
This property is multiply realizablethe parts of the engine can be made
of different sorts of materials, with the constraint that these materials can

128

D. Pereboom

play the roles at issue. Similarly, the various physical realizations of a dogs
and a humans belief that this fire is dangerous may feature a particular
compositional property. This property would be more abstract than any
specific neural compositional property, since it can be realized in different kinds of neural systems. Moreover, it may be that this compositional
property can also be realized in a silicon-based system, which could then
have the belief about danger. Imagine that computer engineers built such
a silicon-based system that mimicked the capacities of and interconnections among neurons in a human brain as exactly as could be, and then
activated it to replicate what happens in that human brain in the case
of the belief about danger. Its possible that the silicon-based state that
ensues would constitute that belief and have an internal structure similar
enough to the internal structure of the neural system for each to count as
instantiating the same compositional property.

The Exclusion Problem andDonaldsons Criticism


Ive argued (Pereboom 2002, 2011) that this account facilitates a response
to the causal exclusion problem for nonreductive physicalism (Kim 1989,
1992, 1998). John Donaldson (2015) points out that the contemporary
version of the causal and explanatory exclusion debate began with an
important statement of the exclusion problem by Norman Malcolm
(1968), whereupon Jaegwon Kim articulated it in ways widely regarded
as canonical. According to robust nonreductive physicalism, a mental
state M2 will be explained psychologically by some mental state M1,
which consists in certain mental properties being instantiated at some
time. Each of M1 and M2 will be microphysically realizedassuming
no specific theory of realizationwhere realization is just, following
Shoemaker (2007), to make real or to implement. Thus, there will be a
microphysical realization P2 of M2 and a microphysical realization P1 of
M1, such that P2 is microphysically explained by P1. But the psychological explanation of M2 by M1 will not reduce to the microphysical explanation of P2 by P1 (and, likewise, mutatis mutandis for states at various
other levels of description, such as the neural).

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

129

Underlying the irreducibility of the psychological explanation is the


fact that it appeals to the irreducibly mental causal powers of M1 to
account for M2, while the microphysical explanation appeals to microphysical causal powers of P1 to account for P2. Hence, the causal powers
of M1 will not be identical with those of P1, and those of M2 will not
be identical with those of P2. Furthermore, there will be a microphysical
explanation for P2 that appeals to the microphysical causal powers of P1,
and at the same time P2 (perhaps together with certain relational features) will be (noncausally, synchronically) sufficient for M2. As a result,
there will be a microphysical causal explanation for M2 that appeals to
the microphysical causal powers of P1. Because one standard way of
explaining an event causally is to cite a causal power whose exercise is a
sufficient cause of the event, citing the causal powers of P1 yields a causal
explanation of M2.
How are the microphysical and psychological explanations for M2
related? Given that both sorts of explanation invoke causal powers, what
is the relationship between the causal powers to which the microphysical
explanation appeals and those to which the psychological explanation
appeals? At this point, Jaegwon Kims challenge from causal exclusion
enters in. We just noted that since P1 yields a causal explanation of
microphysical realization P2 of M2, it will provide a causal explanation
of M2 itself. But what space then remains for a distinct psychological
causal explanation of M2 by M1?
Kim contends that its not credible that each of these distinct groups
of causal powers yields merely a partial cause of the event and that each
would thus be insufficient for the event to occur. He also contends that
its implausible that the psychological explanation appeals to causal powers whose activation is sufficient for occurrence of the event, while the
microphysical explanation appeals to distinct causal powers whose activation is also sufficient for its occurrence, and that in consequence the
event is causally overdetermined. The concern is that on this proposal,
every event that is mentally caused will be overdetermined in the way
that someone who is fatally hit simultaneously by each of two bullets shot
by two assassins is overdetermined, a result that is not credible. Call this
phenomenon redundant overdetermination.

130

D. Pereboom

Kims resolution of the problem is as follows. Real causal powers exist


only at the microphysical level, and so microphysical explanations will
refer to real microphysical causal powers. Only if mental causal powers are identical with microphysical causal powers will there be genuine
mental causation. Thus, psychological explanations that do not reduce to
microphysical explanations will fail to invoke causal powers. Such explanations may express regularities without at the same time referring to
causal powers.
This account, Kim maintains, solves the exclusion problem since if
the causal powers that a psychological explanation invokes are identical with those invoked by the underlying microphysical explanation, no
exclusion-generating competition between mental and microphysical
causes ensues. If a psychological explanation does not invoke causal powers at all, there will be no competition either. This account, which Kim
believes is the only viable solution to the exclusion problem, precludes
the nonreductivism about mental causal powers.
In Kims explication of the exclusion problem, the crucial issue for
nonreductivism is that it seems committed to widespread firing-squadstyle or as Donaldson (2015) puts it, flukey overdetermination. But
as Donaldson argues in his super-overdetermination statement of the
exclusion problem, the deeper challenge for the nonreductivist is to
provide a satisfactory dependence relation between the mental and the
physical, with the provision that the relation of identity is out of bounds.
This turns out to be difficult to do. For many avowed nonreductivists,
the core relation turns out to be identity after all, and thus they in a
fundamental sense reductionists. By contrast, my account of constitution
does not appeal to the identity relation, but rather to a primitive made up
of relation. While this does issue in a genuine form of nonreductivism,
Donaldsons concern is that such a relation is in a relevant sense obscure.

Identity
Despite their allegiance to nonreductivism, Jessica Wilsons (1999) and
Sydney Shoemakers (2003, 2007) subset view endorses a token-identity
thesis for mental and lower-level causal powers, although it does avoid

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

131

reductive type- and token-identity claims for mental states. The term
causal power refers to that by virtue of which an effect is produced
(Baker 2007: 98). Im attracted to the view that property instances are
the entities by virtue of which effects are produced, as L.A.Paul (2000)
advocates, and thus causal powers would be property instances. Causal
powers would then be, in the first place, token entities rather than types,
as is sometimes supposed.1
On Wilsons and Shoemakers view, the mental is realized by the neural
and in the microphysical because the forward-looking causal powers of
a mental state are a proper subset of the forward-looking causal powers
of the lower-level state. Here is Shoemakers most recent statement of his
position:
It is compatible with the claim that the instance of the higher-order property and that of its realizer are not identical that the former is part of the
latter. And that seems the right conclusion to draw from the fact that the
causal powers of the former are a proper subset of those of the latter. And
then it is open to us to say that while it is true that the instance of the realizer property causes the various effects we attribute to the realized property,
it does so because it includes as a part the instance of the realized property.
(Shoemaker 2014)

On Shoemakers account, it turns out that because the forward-looking


causal powers of a mental property instance are a subset of the forwardlooking causal powers of the realizing physical property instance, each
forward-looking causal power of the mental property instance will be
identical with a forward-looking causal feature of the physical property
instance. Wilson makes this feature of the subset view explicit:
What it is for a higher-level property to be realized by a lower-level property is for the set of forward-looking conditional causal powers associated
with the higher-level property to be a subset of the set of forward-looking
conditional causal powers of the lower-level property. Given this account
1

Paul argues for this view partly on the ground that aspects or property instances feature the right
fineness of grain, by contrast with events, at least on some accounts. I prefer to construe property
instances just as ways particular things are, as in David Robbs (2007) characterization, which he
contrasts with abstracta. Token causal powers will then also be ways particular things are.

132

D. Pereboom

of realization, it will never be the case that a given higher-level property has
a conditional causal power different from any of those of its realizer base
property. (Wilson 1999: 50)

Thus, at the most basic metaphysical level, when it comes to causal powers of properties, the relation between the mental and the microphysical
is identity. Types and tokens of mental states are not identical to types
and tokens of microphysical states. Yet at the level of causal powers, the
mental and the physical are identical.
Donaldson points out that on Stephen Yablos (1992) position according to which the relation between the mental and the physical is the
determinable/determinate relation, the most basic relation also turns out
to be identity. For on Yablos view, the relation between the determinable
and the determinate is also the subset relation, and the relation between
metaphysically basic mental features and physical features will be identity
as well. The relation between red and scarlet is that possible instances
of scarlet are a subset of possible instances of red. Within that subset,
such instances of scarlet are identical with instances of red. Similarly,
the relation between pain and a neural realization of pain is that possible
instances of the neural realization are a subset of possible instances of
pain, and within that subset, such instances of the neural realization are
identical with instances of pain.
However, if modal multiple realizability arguments count against
token-identity claims for mental states and properties, they will count
against token-identity claims for mental causal powers as well, no matter
what, metaphysically, one thinks causal powers are. Here is a modal multiple realizability argument that targets mental causal powers directly
(Pereboom 2011). Consider Annes belief at some particular time that
her parents live in Manhattana mental token, an instance of a mental
propertyand the causal power that it has or (in my preferred view)
with which it is identical. Suppose Anne is threatened with an illness that
would damage a small part of her brain that has a crucial role in realizing
this belief (but other parts of her brain have important roles as well).
Before this part is damaged by the illness, a neurosurgeon could remove
it and replace it with a sophisticated electronic microprocessorlets call
it a silicon prosthesis. Imagine that the illness never actually materializes, and Anne does not undergo the operation. Still, this token belief

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

133

would have retained its token mental causal power had she undergone
the operation, and had it thus at that time been realized by the token
neural-and-silicon causal power instead of the token neural power that
actually realizes it. As a result, the token mental causal power cannot be
identified with a token neural causal power, in particular not with the
token neural causal power that actually realizes it.
A temporal variant of this argument can be constructed if we are
allowed the supposition that belief tokens can persist over a significant
duration. Imagine that Annes illness does continue to threaten, and
before the part of the brain is damaged, the neurosurgeon removes it and
replaces it with the silicon prosthesis. After the operation, Anne retains
her token belief about where her parents live, and it possesses the token
mental causal power it had before the operation. This mental token causal
power is, however, no longer realized by the pre-operation neural token
causal power but rather by a neural-and-silicon token causal power. For
this reason, the token mental causal power cannot be identified with any
neural token causal power, specifically not with the token neural causal
power that realized it just before the operation.
For this reason, the dependence relation between the mental and the
physical, even at the basic metaphysical level, wont be identity. In my
account of this relation, I invoke the made up of relation instead. Theres
a connection here with the recent literature on grounding. The grounding relation, its advocates typically argue, is a general relation, of which
constitution is just one type, one on which the ground necessitates what
is grounded (perhaps given certain circumstances, as in my C2), and the
ground exists in virtue of the grounded (Schaffer 2009; Rosen 2010; Fine
2012). Jessica Wilson (2014) argues that there is no work for a theory
of ground to do, given that there are many metaphysical dependence
relations, such as type or token identity, functional realization, classical
mereological parthood relation, the set membership relation, the proper
subset relation, the determinable/determinate relation that can do this
work. Wilson makes a plausible case for her claim. But note that if the
grounding relation is conceived as irreflexive and asymmetric in addition (e.g. Bennett 2011), identity doesnt qualify. And, as Ive argued, the
proper subset relation and the determinable/determinate relation feature
identity at their core. So if the mental and the physical are related in
these sorts of ways, the relation between the two will not at its core be a

134

D. Pereboom

grounding relation. So it may be that if grounding is conceived as asymmetrical and irreflexive, grounding isnt so commonly invoked after all
as the relation between the mental and the physical, and that theres a
controversial philosophical issue at stake.

Two Objections
To this account, Andrew Melnyk (2014) objects that the primitive made
up of relation is not physicalistically acceptable, and he argues that the
same is true for the brute metaphysical necessitation that the account
appears to invoke, that is, necessitation that is not explicable in other
terms, such as the necessity of identity or analyticity. In response, even
if these two notions do not meet the strict sort of test that he has in
mind, and that he sets out in his (2003) book,2 there he himself suggests
a weaker test, that the posited entity in question and the role it plays
provide no encouragement whatever to any familiar sort of antiphysicalist (2003, p.26). I claim this weaker physicalist test-passing status for
the made up of relation. When a physics teacher says that diamonds are
made up of carbon atoms, toying with some familiar sort of antiphysicalism doesnt come to mind. And suppose that I am indeed forced to
say that the necessitation in question is brute and not explicable by the
necessity of analyticity or the identity of necessity. The reason Melnyk
cites for the incompatibility of brute necessitation and physicalism is
that if the necessitation were brute, it wouldnt be anything physical that
would make higher-level property ascriptions true: because this neces2

Here is Melnyks (2003: 2032, 2014) account of realization, which, in his view, secures
physicalism:
Let p name a particular actual physical state token, and m a particular actual mental state token.
Then p realizes m (in the intended sense) only if
(i) m is a token of a mental state type with a certain higher-order essence, that is, m is a token of a
mental state type M such that for a token of M to exist just is for there to exist a token of some
(lower-order) state type such that tokens of that (lower-order) state type play role RM, the role distinctive of M;
(ii) p is a token of a physical state type P such that, necessarily, given the physical laws, tokens of P
under physical circumstances C play role RM; and
(iii) the physical laws hold and physical circumstances C obtain.

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

135

sitation would be a primitive and irreducible modal fact, what would


make [the special-sciences property ascription] true would only be the
special-sciences property instance; nothing physical would be making the
ascription true (2003, p.60). But here one might also reply that a view
on which the (micro)physical brutely necessitates what would make the
ascription true would provide no encouragement to any familiar sort of
antiphysicalist, and that given this weaker test, the truthmaker criterion
Melnyk employs is too strong. Also, necessity and possibility are important notions in physics, and its an open question whether physical necessity is brute, and thus its open whether physics is committed to brute
metaphysical necessity.
Against my account of material constitution and in favor of reductionism, Donaldson objects that the made up of relation is in a pertinent sense obscure. While it is a dependence relation, and we know that
made of shares all of dependences properties what we dont know is
what made of has in addition to dependence (2015: 153). Schematically,
Donaldson proposes that for putative nonreductive physicalists, either
the relation between the mental and the physical is at its core identity,
and the view turns out to be a reductionist after all, or the relation is
brute and therefore unacceptable, or it is obscure. So reductionism wins
out no matter what.
In response, what the made up of relation adds to mere dependence
is revealed when we contrast various examples of dependence relations.
Donaldson construes mere dependence as asymmetrical necessitation.
But note that a numbers being a multiple of 4 asymmetrically necessitates its being even, but its being even isnt intuitively made up of its
being a multiple of 4. By contrast, somethings being made of steel parts
in such-and-such a configuration asymmetrically necessitates its being
a ball piston engine, while it is intuitive that the ball piston engine is
made of up that steel part configuration. The contrast between these
examples illuminates the difference between mere dependence and the
made up of relation: it displays the fact that made up of adds an element to the dependence between the steel part configuration and the
ball piston engine that is not present in the dependence between being
even and being a multiple of 4. True, this method doesnt provide a
conceptual or descriptive account of what made up of adds to depen-

136

D. Pereboom

dence. But at some point, we should expect to encounter conceptual


fundamentality, and its not unintuitive that made up of is conceptually fundamental.
Donaldsons main concern is that the claim that the relation between
the mental and the physical is being made up of, rather than, say, identity, is unacceptably obscure. Such a reaction is not atypical. In the
philosophical tradition and in analytic philosophy in particular, there
is a fairly widespread aversion to relations that are not in some broad
sense logical or mathematical. Relations that do not meet this criterion
are regarded as obscure, and a theory that posits such relations as fundamental is considered unattractive or deficient. A philosophical debate
that since the seventeenth century has prominently witnessed this reaction is the controversy about causal powers. Unreduced causal powers
are considered obscure, and to be replaced by strict regularity or counterfactual dependence. At the outset of his massively influential discussion of causal powers in his (1748/2000) Enquiry Concerning Human
Understanding, David Hume presents this general point of view. He
begins by contrasting the mathematical sciences over the moral sciences, that is, those that study human nature from the mentalistic
point of view. The mathematical sciences are distinguished by clear
and determinate ideas:
The great advantage of the mathematical sciences above the moral consists
in this, that the ideas of the former, being sensible, are always clear and
determinate, the smallest distinction between them is immediately perceptible, and the same terms are still expressive of the same ideas, without
ambiguity or variation. An oval is never mistaken for a circle, nor an hyperbola for an ellipsis. The isosceles and scalenum are distinguished by
boundaries more exact than vice and virtue, right and wrong. If any term
be defined in geometry, the mind readily, of itself, substitutes, on all occasions, the definition for the term defined: Or even when no definition is
employed, the object itself may be presented to the senses, and by that
means be steadily and clearly apprehended.

By contrast, the ideas of the moral sciencesand he later adds the metaphysical sciencesare subject to an absence of clarity, in the form of
ambiguity, and this lowers the quality of reasoning in those sciences:

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

137

But the finer sentiments of the mind, the operations of the understanding,
the various agitations of the passions, though really in themselves distinct,
easily escape us, when surveyed by reflection; nor is it in our power to recall
the original object, as often as we have occasion to contemplate it.
Ambiguity, by this means, is gradually introduced into our reasonings:
Similar objects are readily taken to be the same: And the conclusion
becomes at last very wide of the premises.

Hume concedes that the moral sciences have the advantage of shorter
and less complex inferences, and thus reasoning in those sciences is in
one sense easier.
The chief obstacle, therefore, to our improvement in the moral or metaphysical sciences is the obscurity of the ideas, and ambiguity of the terms.
The principal difficulty in the mathematics is the length of inferences and
compass of thought, requisite to the forming of any conclusion.

Hume then moves to the application of this methodological perspective to the idea of causal power. He claims that there are no ideas,
which occur in metaphysics, more obscure and uncertain, than those of
power, force, energy or necessary connexion, of which it is every moment
necessary for us to treat in all our disquisitions. And the point of the
section of the Enquiry on causation is to fix, if possible, the precise
meaning of these terms, and thereby remove some part of that obscurity, which is so much complained of in this species of philosophy.
Hume judges the idea of a fundamental causal power obscure, and
offers two influential proposals as substitutes, strict regularity: A causes
B just in case As are always followed by Bs; and counterfactual dependence: A causes B just in case: if As hadnt occurred, B wouldnt have
occurred.
However, we might ask: What is it, specifically, that recommends these
notions over the idea of causal power? Humes argument that the idea
of causal power has no corresponding impression in sensory experience
doesnt currently move many, and, moreover, he seems too quickly to set
aside the impression of ones own exertion of physical force as a source of
the idea. Many have argued that irreducible causal powers play an indispensable explanatory role, and all of this without the idea of causal power

138

D. Pereboom

being even broadly mathematical or logical. Can we say the same for the
idea of being made up of?
Thus there is a response to the obscurity charge against one genuinely
nonreductivist relation between the mental and the physical, namely,
material constitution construed as featuring the being made up of relation.
For there is a way to elucidate the difference between the being made up
of relation and mere dependence. One might agree that being made up of
is more obscure than identity, while resisting the claim that it is irremediably obscure. And a good case can be made that this relation has an indispensable explanatory role. Any mental/physical dependence relation that
issues in identity at its root is subject to multiple realizability arguments,
which are the standard type of argument against identity claims in this
general area of inquiry. This suggests that a relation other than identity is
required, and the material constitution construed as featuring the being
made up of relation is my candidate.3

References
Baker, L.R. (2000). Persons and bodies. Cambridge: Cambridge University Press.
Baker, L. R. (2007). The metaphysics of everyday life. Cambridge: Cambridge
University Press.
Baker, L. R. (2013). Perebooms robust nonreductive physicalism. Philosophy
and Phenomenological Research, 86, 736744.
Bennett, K. (2011). Construction area: No hard hat required. Philosophical
Studies, 154, 79104.
Boyd, R. (1999). Kinds, complexity, and multiple realization. Philosophical
Studies, 95, 6798.
Burge, T. (1978). Individualism and the mental. Midwest Studies in Philosophy
6, ed. P. French, T. Uehling, and H. Wettstein, Minneapolis: University of
Minnesota Press, 73121.
Donaldson, J. (2015). The superoverdetermination problem. Doctoral thesis,
University of Glasgow.
Fine, K. (2001). The question of realism. Philosophical Imprint, 1, 130.
3

Thanks to Louis DeRosset, David Christensen, Jessica Wilson, Ted Sider, Laurie Paul, Karen
Bennett, Nico Silins, Sydney Shoemaker, Lynne Baker, and Carl Gillett for valuable comments and
discussion.

Anti-Reductionism, Anti-Rationalism, andtheMaterial...

139

Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),


Metaphysical grounding (pp. 3780). Cambridge: Cambridge University
Press.
Gillett, C. (2002). The dimensions of multiple realization: A critique of the
standard view. Analysis, 62, 31623.
Gillett, C. (2003). The metaphysics of realization, multiple realizability, and the
special sciences. Journal of Philosophy, 100, 591603.
Hume, D. (1748/2000). An enquiry concerning human understanding. Oxford:
Oxford University Press.
Kim, J. (1989). The myth of nonreductive materialism. Proceedings and addresses
of the American Philosophical Association 63, pp.3147; reprinted in J.Kim,
Supervenience and mind. Cambridge: Cambridge University Press, 1993,
pp.265284.
Kim, J. (1992). Multiple realizability and the metaphysics of reduction.
Philosophy and phenomenological research 52, pp. 126; reprinted in
Supervenience and mind. Cambridge: Cambridge University Press, 1993,
pp.309335.
Malcom, N. (1968). The conceivability of mechanism. Philosophical Review 77
(1968), 4572.
Melnyk, A. (2003). A physicalist manifesto. Cambridge: Cambridge University
Press.
Melnyk, A. (2014). Perebooms robust nonreductive physicalism. Erkenntnis,
79(5), 11911207.
Paul, L.A. (2000). Aspect causation. Journal of Philosophy, 97, 23556.
Pereboom, D. (2002). Robust nonreductive materialism. Journal of Philosophy,
99, 499531.
Pereboom, D. (2011). Consciousness and the prospects of physicalism. NewYork:
Oxford University Press.
Pereboom, D. (2013a). Prcis of Consciousness and the Prospects of Physicalism.
Philosophy and Phenomenological Research, 86, 715727.
Pereboom, D. (2013b). Replies to Daniel Stoljar, Robert Adams, and Lynne
Baker. Philosophy and Phenomenological Research, 86, 753764.
Place, U.T. (1956). Is consciousness a brain process? British Journal of Psychology,
47, 4450.
Robb, D. (2007). Power essentialism. Philosophical Topics, 35, 34358.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
B.Hale & A.Hoffmann (Eds.), Modality: Metaphysics, logic, and epistemology
(pp.109136). Oxford: Oxford University Press.

140

D. Pereboom

Schaffer, J. (2009). On what grounds what. In D.J. Chalmers, D.Manley, &


R.Wasserman (Eds.), Metametaphysics: New essays on the foundations of ontology (pp.34583). Oxford: Oxford University Press.
Shoemaker, S. (2003). Realization, micro-realization, and coincidence.
Philosophy and Phenomenological Research, 67, 123.
Shoemaker, S. (2007). Physical realization. Oxford: Oxford University Press.
Shoemaker, S. (2014). Physical realization and mental causation. In S.C. Gibb
& J.Lowe (Eds.), The new ontology of the mental causation debate. Oxford:
Oxford University Press.
Smart, J.J. C. (1959). Sensations and brain processes. Philosophical Review, 68,
141156.
Wilson, J. M. (1999). How superduper does a physicalist supervenience need to
be? Philosophical Quarterly, 49, 3352.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57, 53579.
Yablo, S. (1992). Mental causation. Philosophical Review, 101, 24580.

Part II
Grounding, Science,
and Verticality in Nature

6
Ground Rules: Lessons fromWilson
JonathanSchaffer

Wilsons No Work for a Theory of Grounding (2014) offers an insightful critique of grounding-based approaches to metaphysical inquiry. She
argues that the notion of grounding is uninformative, disunified, and in
the end unhelpful. She then sketches a rival approach, which eschews the
notion of grounding, in favor of a plurality of small-g grounding-type
notions alongside a primitive notion of absolute fundamentality.
I think that Wilson is right to criticize many extant grounding-based
approaches for not being sufficiently informative. I just think that it is
possible for the grounding theorist to do better, and that my own (forthcoming) treatment in terms of structural equations (which are formal
models developed for understanding causal structure) does better in the
needed ways. I also think that her rival approach deserves serious consideration in its own right. But I argue that her approach is open to serious
criticisms, including every one of the criticisms she levels at the grounding theorist.

J. Schaffer (*)
Rutgers University, New Brunswick, NJ, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_6

143

144

J. Schaffer

My subtitle is Lessons from Wilson since I am saying that all theoristsincluding Wilson herselfshould draw the lesson that one needs
more informative conceptions of metaphysical structure, of the sort I
take structural equation models to provide.
My main title is Ground Rules, which I offer as a three-way pun.
First, one of the underlying issues between Wilson and myself is whether
the notion of grounding is sufficiently unified for useful work. Such issues
of unity arise not just with the notion of grounding, but with virtually all
interesting philosophical concepts. So I want to articulate some general
ground rules for unity debates. Second, I want to discuss a formalism to
model grounding, which will display the rules of grounding. These first
two ideas are related, insofar as I think that the general way to adjudicate
unity debates is by looking at the best formalism, and seeing whether or
not it enfolds the notion in a unified set of rules. Third, I want to express
my continued enthusiasm for the notion of grounding as one that can
be framed in a unified, informative, and helpful way, and so I exclaim:
ground rules!
Overview: In section A Brief Introduction to Grounding, I offer
a brief introduction to the notion of grounding. In sections Are
Grounding Claims Informative? and Are Grounding Claims Helpful?, I
take up Wilsons two main objections to grounding-based approaches
that bare grounding claims are uninformative, and that such claims are
unhelpfuland extract two main lessons. In section Wilsons Pluralistic
Framework, I critique Wilsons rival pluralistic approach for, among
other things, not taking up Wilsons own lessons. I conclude in section Structural Equation Models to the Rescue by explaining how an
approach based on structural equation models for grounding has a special
claim to adequacy.

A Brief Introduction toGrounding


Grounding has been championed by philosophers including Fine (2001),
Correia (2005), Schaffer (2009), Rosen (2010), and Bennett (2011). We
do not agree on all details, and accordingly I only claim to speak for
myself. Groundingas I understand itconnects more to less funda-

6 Ground Rules: Lessons fromWilson...

145

mental entities and thereby imposes structure over what there is. Some
entities are more fundamental than others (for instance, particles are
more fundamental than chemicals, and chemicals are more fundamental
than animals). Once one distinguishes more from less fundamental entities, it is natural to posit a relation linking certain more fundamental entities to certain less fundamental entities which derive from them. Grounding
names this directed linkage.
Grounding may be understood as the relation of dependence which
philosophers tried but failed to understand via the modal pattern of
supervenience. As Kim (1993: 167) notes:
Supervenience itself is not an explanatory relation. It is not a deep metaphysical relation; rather, it is a surface relation that reports a pattern of
property covariation, suggesting the presence of an interesting dependency
relation that might explain it.

Supervenience, after all, is a reflexive and (a fortiori) non-asymmetric relation, as well as a merely intensional relation that cannot distinguish features
of reality found at all the same possibilities. So grounding may be understood as the deep relation of dependence which shallow supervenience
analyses unsuccessfully targeted. In my view, one of the morals of the failure of the supervenience analysis is that the notion of metaphysical dependence is needed but unanalyzable, and hence best treated as primitive.1
Grounding then serves to back a distinctive sort of metaphysical explanation. If one wants to understand, for instance, why there is an H2O
molecule present, then one perfectly good sort of explanation for this fact
would involve the fact that an H, another H, and an O atom are arranged
and bonded in the right way. This is not a diachronic causal explanation, citing previous causes. (A diachronic causal explanation might for
instance cite the previous events in which hydrogen and oxygen gasses
were combined and exposed to a spark.) It is rather a synchronic metaphysical explanation, citing the more fundamental basis at the time. Just
as causation provides the direction and the linkage needed for causal
Of course I cannot prove the negative existential that there is no reductive analysis of the concept
of grounding to be found (though when has reductive conceptual analysis ever succeeded?); I only
mean to say that it is legitimate to use the concept regardless, without any such analysis to hand.
1

146

J. Schaffer

explanation, so grounding provides the direction and the linkage needed


for metaphysical explanation. The reason why there is an explanatory link
from the presence and arrangement of the H, H, and O to the presence
of the H2O is that the H, H, and O ground the H2O.In this vein, Audi
(2012: 104) says, The reason we must countenance grounding is that it
is indispensable to certain important explanations.2
Many of the most interesting debates in metaphysics can then be
understood as debates about what grounds what (and consequently
which facts explain which). For instance, disputants in the metaphysics
of mind usually agree that the physical and the mental both exist. They
disagree as to whether the physical grounds the mental (as the physicalist/
materialist thinks), or whether the mental grounds the physical (as the
idealist thinks), or whether both are independently fundamental aspects
of reality (as the dualist thinks).
I cannot offer a more detailed motivation for invoking such a notion
of grounding here (see Schaffer 2009; forthcoming), though this brief
sketch should be sufficient for the discussion to come. For Wilson thinks
that this entire picturealluring as it may seemis superficial and in the
end worthless, and she says that she can paint a better picture.

Are Grounding Claims Informative?


Wilson, on my reading, makes three main points, the first of which is
that bare grounding claimsclaims of the form this grounds thatare
uninformative. Thus, she (2014: 546) considers the metaphysicianlet
us name her Nataliewho says that the natural grounds the normative,
and points out just how much is left open:
[N]aturalists do not care only about whether, for example, normative
goings-on metaphysically depend on naturalistic goings-on: they also care
According to Salmon (1984), the lesson to be drawn from the failure of deductive-nomological
accounts of explanation is that explanation must be backed by causation, to make sense of the connections behind and the ordering of explanation. For metaphysical explanation then one also needs
a connecting and ordering relation, running not from cause to effect but from more basic to less
basic. This is grounding.
2

6 Ground Rules: Lessons fromWilson...

147

about whether normative goings-on exist; about whether, if they exist, they
are reducible or rather irreducible to (though still nothing over and above)
naturalistic goings-on; about how exactly normative goings-on are related
to naturalistic goings-on; about whether normative goings-on are efficacious and, if so, whether they are distinctively efficacious (that is, efficacious qua normative); and so on. Hence it is that naturalists almost never
rest with the schematically expressed locutions of metaphysical dependence, but rather go on to stake out different positions concerning how,
exactly, the normative or other goings-on metaphysically depend on the
naturalistic ones.

So Wilson (2014: 545) thinks that our friend Natalie has managed to tell
us almost nothing about how, exactly, normative and intentional goingson stand to naturalistic goings-on.
I think that there is something right and insightful here, but that it is
hard to identify exactly what. Or at least, I found Wilsons objection initially puzzling. For imagine a scientistlet us name him Sigmundwho
utters a bare causal claim, such as smoking causes cancer. Wilsons main
concerns about Natalie could equally be raised about our new friend
Sigmund. After allto mimic what Wilson saysscientists do not only
care about whether smoking causes cancer; they also care about whether
cancer exists, about how exactly smoking is related to cancer, and about
what the more fundamental physicochemical conditions underlying cancer are, and so on. So what? Surelywhatever the ultimate status of causation may bethese observations alone do not show that the notion
of causation is uninformative and unhelpful! So how could Wilsons
concerns possibly show that the notion of grounding is uninformative
and unhelpful?
Clearly, both Sigmund and Natalie have told us something informative. Sigmund has said something that rules out alternatives such as that
smoking is causally unrelated to cancer, or only related as a correlate of a
common cause (indeed decades of careful medical research, countered by
costly corporate propaganda, went into establishing his claim). Likewise
Natalie has said something that rules out alternatives such as those given
by certain forms of moral realism and by divine command theory (centuries of philosophical debate have centered on assessing her claim).

148

J. Schaffer

Both Sigmund and Natalie have also told us something helpful, insofar
as causation and grounding both serve to provide explanatory handles on
the world. Sigmund has said something that might help us understand,
for example, why a particular smoker has cancer, and likewise Natalie
has said something that might help us understand, for example, why a
particular natural situation is morally impermissible.
Of course neither Sigmund nor Natalie has said everything, but so
what? Given anything short of a maximally specific description of reality,
there will always be more information to add. What is the problem with
informative and helpful claims that merely leave some further questions
open? So when Wilson says that Natalie has said almost nothing, I want
to ask back, do you mean that Natalie has said nothing, or that she has said
something but should just say more? The former option strikes me as false
and the latter true but unobjectionable.3
Of course if Sigmund refuses to say anything further about the smokingcancer connection beyond smoking causes cancer, then s omething
has gone wrong. But the problem here is not with bare causal claims, nor
with the concept of causation they involve, but only with the strangely
silent theorist who refuses to do anything more than make such bare
causal claims. Likewise if Natalie should refuse to say anything further
about the naturalnormative connection, then something has gone
wrong. Wilson (2014: 549) speaks of the perversely uninterested metaphysician who only makes bare grounding claims and says nothing further. But the problem here is likewise not with bare grounding claims, nor
with the concept of grounding they involve, but only with the perversely
Two puzzling passages: Wilson (2014: 5445) says that it is not just that Grounding (failure of
Grounding) claims leave some interesting questions open; rather, it is that such claims leave open
questions that must be answered to gain even basic illumination about or allow even basic assessment of claims of metaphysical dependence, or associated theses such as naturalism. But I find this
puzzling since she does not say what she means by distinguishing merely interesting questions
from those that must be answered, or relatedly what she means by basic illumination. And I
think she is just wrong that bare grounding claims cannot be assessed. Natalies claim, for instance,
rules out alternatives such as divine command theory. So if divine command theory could be shown
to be true, Natalies claim would thereby be shown to be false.
Wilson (2014: 553; also 575) also takes up the analogy with bare causal claims. She allows that bare
causal claims are informative for entailing that their relata exist as distinct and causally connected
events, but admits no comparable value to bare grounding claims. But I find this puzzling as well
since, at least by my lights, grounding claims are informationally comparable: they entail that their
relata exist as non-distinct and grounding-connected entities.
3

6 Ground Rules: Lessons fromWilson...

149

uninterested theorist who refuses to do anything more than make such


bare grounding claims. One need only ask both Sigmund and Natalie to
say more.
What more should they say? It is worth distinguishing three respects
in which both Sigmund and Natalie should say more, which arise from
separating three of the follow-up questions Wilson poses: (1) whether
normative goings-on exist, (2) whether normative goings-on are efficacious, and (3) how exactly normative goings-on are related to naturalistic goings-on. As to questions such as (1) whether normative goings-on
exist, I think that this is a matter of settling conceptual entailments. I say
that both causation and grounding entail the existence of their relata, so
that an adequate conception of causation must have it that Yul Brynners
smoking caused his lung cancer entails the existence of Brynners smoking and his lung cancer, and that an adequate account of grounding must
have it that Marquis de Sades inflicting pain grounds his acting wrongly
entails the existence of de Sades inflicting pain and his acting wrongly.4
Soassuming that Sigmund thinks that there are episodes of smoking
causing cancerI say that he is thereby implicitly committed to saying
that cancer exists. And likewiseassuming that Natalie thinks that there
are episodes of the natural grounding the normativeI say that she is
thereby implicitly committed to saying that the normative exists.
As to questions such as (2) whether normative goings-on are efficacious, I think that this is a matter of recognizing conceptual separations.
I say that grounding is neutral as to the causal efficacy of its relata, so
that an adequate account of grounding may allow Marquis de Sades
inflicting pain grounds his acting wrongly to be consistent with both
the idea that de Sades acting wrongly is causally efficacious, and the idea
that de Sades acting wrongly is causally inert. So I say that Natalie has
so far remained non-committal on the causal efficacy of the normative.
One can make informative and helpful claims while still leaving other
questions open.
I have switched from generics (smoking causes cancer, pain grounds wrongness) to episodics,
since at least some types of generics can hold without episodes (this machine crushes oranges can
be true even if it never gets turned on). This matter has nothing to do with causation or grounding,
but purely with the semantics of generic constructions.
4

150

J. Schaffer

As to (3) how exactly normative goings-on are related to naturalistic


goings-on, I think that this is the most interesting follow-up question, insofar as it is a matter of appreciating conceptual embeddings. When Sigmund
says that smoking causes cancer, there are several further causal how questions he should try to say more about, where the right answer is neither
entailed by nor separated from the causal connection between smoking and
cancer. One sort of causal how question looming concerns the causal
mechanisms involved which mediate the link. (Is it the nicotine, the tar,
or perhaps the smoke and associated inflammatory reaction in the body?)
These questions can be resolved by further bare causal claims concerning
candidate intermediaries. (Does nicotine cause cancer? Does tar cause cancer? Can repeated inflammatory reactions in the body cause cancer?)
But a second and distinct sort of causal how question looming concerns the shape of the association between any given cause and effect.
(Is cancer risk linear with cigarettes per day, or does the risk peak at two
cigarettes per day and level off or even decline thereafter, or ?) These
latter questions are the more interesting questions, insofar as they push
one to embed causal claims in a deeper framework that posits not just an
onoff connection (cause or no cause?) but a more informative function
relating a range of values for the cause option (number of cigarettes per
day) to a range of values for the effect option (risk of cancer).
Likewise when Natalie says that the natural grounds the normative,
I think (section Structural Equation Models to the Rescue) that one
should ask her to embed her grounding claim in a deeper framework that
posits not just an onoff connection (ground or no ground?) but a function relating a range of values for the more basic option (different states
of nature, such as those differing in the pleasure-to-pain ratio) to a range
of values for the derivative option (different normative statuses, such as
position in the preferability rank). The grounding framework would then
include information about how exactly normative goings-on are related
to naturalistic goings-on. Ideally, one should want to know the precise
rule taking certain aspects of the natural state of the situation as input,
and delivering the normative status of the situation as output.5
Caveat: Wilson herself may be understanding her own how exactly question in a weaker way,
since she (2014: 5467) only asks the grounding theories to choose between options such as the
following:
5

6 Ground Rules: Lessons fromWilson...

151

So I agree with Wilson that Natalies bare grounding claim leaves open
how the grounding pattern works, and think that there is an important
lesson to be drawn from this. But I think that the lesson is not to discard
the notion of grounding but to develop it further, in ways that allow one
to go beyond bare grounding claims and add even more information
about the underlying pattern. Thus, I think that Wilson is best understood as offering the following lesson:
Wilsons first lesson: An account of grounding must give one more than just
the bare ideology of this grounds that; it must also allow one to make
sense of follow-up inquiry into how the connection runs, in terms of the
specific rule mapping the more basic inputs to the less basic output.

Are Grounding Claims Helpful?


Wilsons second main point, following on her first point that bare grounding claims are not sufficiently informative, is that the metaphysician will
inevitably be driven to say more, and in particular will be driven to speak
of more specific small-g grounding-type relations in explaining how
the grounding connection works. Thus, she (2014: 540) says:
Grounding, like supervenience, is too coarse-grained to do the work of appropriately characterizing metaphysical dependence on its own, failing to distinguish importantly different (eliminativist, reductionist, non-
reductionist,
emergentist) accounts of such dependence, Investigations into metaphysical dependence cannot avoid appealing to the specific small-g grounding
[A] naturalist might be a role functionalist, maintaining that normative state types are
characterized by functional or causal roles played by naturalistic state types. Or a naturalist
might maintain that normative state types and/or tokens stand in something like the determinable/determinate relation to naturalistic goings-on. Or a naturalist might maintain that
normative state types and/or tokens are appropriately seen as proper parts of naturalistic state
types and/or tokens.
Now when one specifies a function, one may specify the function in extension as a list of ordered
pairs, or specify it in intension as a rule which maps input to output. I think of Wilsons options as
coarse-grained types of rules, and am saying that one should ultimately want the specific rule
involved.

152

J. Schaffer

relations that are capable of answering these crucially basic questions about
the existential, ontological, metaphysical, and causal status of metaphysically
dependent goings-on.

Among the small-g grounding relations, she (2014: 535) includes


type and token identity, functional realization, the classical mereological parthood relation, the causal composition relation, the set membership relation, the proper subset relation, the determinable/determinate
relation, and so on. Butshe adds (2014: 553; also 576)once the
small-g grounding relation is specified (as it must be), the need to
speak generically of grounding is lost: But insofar as appeal to specific
small-g grounding relations is required to gain even basic illumination
about metaphysical dependence, what if any point is there moreover to
positing Grounding? And so she says that the notion of grounding is
unhelpful, because it is inevitably superseded.6
I agree with Wilson that the metaphysician will be driven to make
sense of how the grounding connection works in these sorts of ways. (I
do not fully agree with Wilsons list of species of grounding relationsfor
instance, I would not include identity as a grounding relationbut the
matter is not directly relevant here.) Indeed I think that the metaphysician will be driven to take the matter still further, and try to articulate the
specific rule linking the more fundamental input to the less fundamental
output. That is essentially Wilsons first lesson. But I do not see how it follows from the fact that one may be driven to decide which species of a
given genus is found, and that the genus notion is thereby unhelpful and
to-be-discarded. (Otherwise one is headed toward a radical elimination
of every genus notion!) Again I think that there is something right and
insightful here, but that it takes some work to identify exactly what.
Return again to the guiding case of causation. Most would agree that
there are species of causal relations. But from the fact that one may be
Here I am simplifying Wilsons argument considerably. As I read her, the grounding is inevitably
superseded claim comes on p.553, and much of the rest of her paper rebuts various reasons one
might give for saying that grounding is still worth positing. In Wilsons terms, I am probably best
classified as defending what she (2014: 567) calls the general unifier of the specific grounding
relations rationale, thoughas comes out in section Wilsons Pluralistic FrameworkI also
endorse what she (2014: 558) calls the fix the direction of priority rationale. (My thanks to Jessica
Wilson for helpful clarifications.)
6

6 Ground Rules: Lessons fromWilson...

153

driven to decide the detailed sorts of causal connections involved in the


(quite complicated) relationship between smoking and cancer, it does not
follow that it is unhelpful to have the notion of causation, or to use it in
saying that smoking causes cancer. (I am not assuming that there is a unified genus notion of causation worth positing. I think that there is such a
unified genus, but that is a substantive claim to be argued for. I am only
saying that it does not follow that there is no unified genus of causation,
simply from the truism that one may be driven to further specify the type
of small-c causal relation involved in a given case.)
There is a very general issue looming, which arises for virtually all philosophically interesting concepts, which is when there is a unified notion
worth positing. For virtually any candidate notion, there will be monists
who think that there is a single unified concept to be posited, surrounded
on the one side by nihilists who deny that there is even one meaningful
concept being invoked, and on the other side by pluralists who say that
there are many distinct concepts being conflated.7 Indeed every philosopher will deploy some concepts in a monistic spirit, dismiss other concepts in a nihilistic spirit, and divide other concepts in a pluralistic spirit.
So everyone should be interested in principled ways of deciding when to
be a monist, a nihilist, or a pluralist, for any given target concept.
It seems to me that the best principled way to decide when to be a
monist, a nihilist, or a pluralist for a given concept is to construct the best
formalism one can for the concept. If there is no meaningful concept,
this should show up in a lack of any clear formalism, and if there are
many, this should show up in a need for a formal distinction. But if one
winds up with a clear and precise formalism that embeds the concept in a
unified way, then this is a good sign that there is a single unified concept.
I offer this as a general ground rule for unity debates: let the best formalism decide (Rosen 2010: 114; Schaffer forthcoming, Sect. 4.4).8
In the grounding literature, Fine, Rosen, and I are paradigm monists (though Fine is hard to classify since he also distinguishes metaphysical, natural, and normative ground), Hofweber (2009)
and Daly (2012) are nihilists, and Wilson (2014) and Koslicki (2015) are pluralists. In the causation literature, monism has been the dominant view, but the early Russell (1912; though not 1948)
is a nihilist, and Anscombe (1975), Cartwright (2007), and Hall (2004) are pluralists.
8
Some may wish to add the requirement that there is a single guiding idea behind the formalism.
For instance, there is an ongoing debate about set theory, as to the extent to which the ZermeloFraenkel axioms (ZFC) can be seen as guided by the iterative conception of sets (cf. Boolos 1971).
7

154

J. Schaffer

To illustrate, consider causation yet again. What would constitute a


decent test for causal monism? I offer this: develop a formalism for causation, and see whether nothing clear and informative emerges (a sign of
nihilism), or whether one must distinguish red arrows for one type of
causal connection from blue arrows for another type of causal connection (a sign of pluralism), or whether one can in the end succeed in saying
something clear and informative using only a single colored arrow (a sign
of monism).
This ground rule is intended as a decent rule of thumb, and neither as a
mechanical nor as an infallible procedure. There can be dispute over what
the best formalism is, there can be questions about when a formalism is
embedding a concept in a unified way, and it is at least conceivable that
our best formalism could attribute more or less unity to a given concept
than is really present.9 I just think that in the cases at handnamely the
cases of causation and groundingthe best formalism for both makes use
of structural equation models, and clearly does not need to draw different
colored arrows for different flavors of dependence (section Structural
Equation Models to the Rescue). If so, then the notion of grounding has
exactly the same claim to unity as the notion of causation.
With this ground rule in mind, I return to the question of why the
notions of grounding and causationand genus notions generallymay
still be helpful, even if there is always a possible follow-up question as
to which species is found. I offer two answers, my first of which is that
without the genus notion, one may miss relevant generalizations. These
generalizations are part of what the best formalism for the notion must
capture. The best formalism should include rules involving the notion,
which encode the generalizations one would lose without the notion.
For instance, assuming that Brynners smoking 100 cigarettes per day
caused his lung cancer, one should want a framework for causation that
The structural equations model approach I advocate (section Structural Equation Models to the
Rescue) comes out very well on this score, as it has a strong claim to be guided by the idea of
directed dependency. (My thanks to Jon Litland on this point.)
9
Wilson (2014: 568) offers determinable properties as a potential case where the formalism attributes more unity than philosophers commonly assume. But first, Wilson herself thinks that this
is the wrong view of determinables (cf. Wilson 2012). Second, I do not think it is clear what the
best formalism for determinatedeterminable structure is. (I would be content to let the best formalism decide in this case as well.) Third, I do not claim the test to be perfect anyway, but just a
decent guide. I do not know if Wilson would contest this more modest claim.

6 Ground Rules: Lessons fromWilson...

155

allows us to infer Brynners cancer risk given that he smoked 100 cigarettes per day, to counterfactually consider what Brynners cancer risk
would have been had he smoked 0 cigarettes per day (or 20, or 200),
and to underwrite explanatory claims as to why Brynner developed lung
cancer. Such a framework would display the relevant connections between
causation, inference, counterfactuals, and explanation, which constitute
some of the generalizations that make causation worth positing. Likewise,
assuming that de Sades acting to inflict pain grounds his acting wrongly,
one should want a framework for grounding that allows us to infer the
wrongness of de Sades action given that he acted to inflict pain, to counterfactually consider the normative status of alternative courses of action,
and to underwrite explanatory claims as to why de Sade has acted wrongly.
My second answer as to why genus notions may still be helpful is that
without the genus notion one may be unable to enumerate the species. For
instance, a theorist who refused the general notion of causation would have
no clear way to enumerate her own small-c causal relations. To illustrate, let us imagine that she starts off by invoking some more specific
causative notions like baking, making, waking How can she continue?
She cannot say and all other species of causation because that would
be cheating (explicitly invoking the very notion of causation that she has
foresworn). And she cannot just say and so on because what could
that mean for her (besides serving as a device to implicitly invoke the very
concept of causation that she has foresworn)? Likewise, the theorist who
refused the general notion of grounding would have no clear way to enumerate her own preferred menu of small-g grounding relations. Wilson
herself (2014: 535) resorts to and so on when listing her own open-ended
plurality of small-g grounding relations, and so one must wonder how
she understands her own list to continue, if not in terms of listing further
species of the very genus notion that she has foresworn, namely grounding.
So I think that what is right and insightful in Wilsons complaint is
that the best way to determine whether a genus notion is helpful is to embed
the notion in a formalism which treats the notion in a unified way, and
reveals the generalizations one would miss without the notion. For without
such an embedding, one has no rules governing the notion. Accordingly,
I think that Wilson is best understood as offering the following further
lesson:

156

J. Schaffer

Wilsons second lesson: An account of grounding should be embedded in a


formalism that outfits the notion with unified rules and so reveals useful
generalizations one would miss without the notion.

Wilsons Pluralistic Framework


Having argued that bare grounding claims are not sufficiently informative, and that the metaphysician will inevitably be driven to speak of more
specific small-g grounding relations in explaining how the grounding
connection works, Wilson thenthirdsketches an alternative pluralist framework that eschews the general notion of grounding and only
uses the more specific small-g concepts. She (2014: 576) endorses the
interest in questions of what grounds what (cf. Schaffer 2009):
Proponents of Grounding are correct that metaphysicians should be concerned with the question of what metaphysically depends on what; and
they are correct that the idioms of metaphysical dependence are not properly interpreted in semantic, epistemic, causal or merely modal terms.

But she (2014: 576) claims that the notion of dependence is best understood as schematic, merely standing in for some yet-to-be-specified one
of her many small-g grounding relations:
[S]uch idioms should rather be taken just to advert, schematically or otherwise, to one or other of the specific metaphysical relationstype and
token identity, the functional realization relation, the classical mereological
part-whole relation, the causal composition relation, the set membership
relation, the proper subset relation, the determinabledeterminate relation,
etc.already on the scene.

So far it might seem as if Wilson was not positing anything metaphysically new, but on the contrary advocating that the new questions of
dependence should be interpreted schematically against the backdrop of
a plurality of old relations.
Not so: there is a twist. For Wilson (2014: 55862) considers an argument for groundingwhich she credits to Fine and Hellie, and labels the

6 Ground Rules: Lessons fromWilson...

157

fix the direction of priority argumentthat grounding is needed to make


sense of questions as to which end of one of her small-g relations is the
more fundamental end. For instance, even given that this particle is a part of
the cosmos, there is a remaining question as to whether the whole depends
on the part or the part on the whole (see generally Schaffer 2010). Wilson
acknowledges the force of the objection and thus sees fit to posit her own
distinctive hyperintensional primitive structuring concept of metaphysical
inquiry, that of (absolute) fundamentality. So she (2014: 561) says:
Which entities are in the fundamental base is primitive; this primitive specification then fixes the direction of priority (assuming there is one, as there
may not be in cases of self- or mutual grounding, or cases of entities having
nothing to do with one another) associated with a given specific small-g
grounding relation, when applied to goings-on in the base; effectively, fundamentality is hyperintensional. For example, if the One is primitively fundamental, then proper parts of the One will be non-fundamental; if the
Many are primitively fundamental, then fusions of the Many will be
non-fundamental.

So Wilson also posits her own primitive structuring concept of metaphysical inquiry, to help orient her plurality of small-g grounding relations in the right direction.
I pause to note howrhetoric asideWilson winds up largely agreeing with friends of grounding. The underlying point of agreement is that
the metaphysician needs a new primitive hyperintensional notion to go
beyond merely listing what exists so as to characterize the structure of reality. The main residual disagreement is whether this primitive should be
one of being-absolutely-fundamental, or the relative and linking notion of
grounding (/being-relatively-more-fundamental-than-and-linked-to). These
notions are respectively analogous to being-causally-initial and causing
(being-relatively-causally-earlier-than-and-linked-to).10 This is not to say
that there is no disagreement between Wilson and friends of grounding
Both grounding and causation are notions of a directed linkage, which is why they are both apt
to back explanation (section A Brief Introduction to Grounding). Note that the and-linked-too
bit is needed. This H atom on Earth is relatively more fundamental than that H2O molecule on
Mars just because atoms are generally more fundamental than molecules, even though this atom is
not linked to that molecule (cf. Bennett forthcoming: Chap. 5). Likewise this event on Earth yesterday is relatively causally earlier than that event on Mars today just because of the overall causal
10

158

J. Schaffer

such as myself (cf. Wilson 2014: 5623), but only that there is a merely
internecine disagreement between those friends of primitive hyperintensional notions of metaphysical structure such as Rosen and I who opt
to take grounding as primitive, those friends such as Wilson (also Sider
2011) who prefer to take being-absolutely-fundamental as primitive, and
those such as Fine who claim a need for multiple primitive notions in the
neighborhood.
That saiddegree of real disagreement asideI think that Wilsons
framework is clearly worthy of serious consideration. I have three objections, however, the first of which is that I think Wilsons framework is
impoverished compared to the grounding framework. It seems to me that
absolute fundamentality can easily be defined in terms of grounding (the
fundamental is that which has no deeper grounds), and so a framework
using grounding as a primitive can easily be used to say everything one
wants to say via absolute fundamentality. But there is no obvious definition to be found in the other direction, and so it is not at all obvious that
using absolute fundamentality as a primitive will allow one to say everything one wants to say in terms of relative fundamentality, or in the even
stronger linking terms of the grounding connection.
This impoverishment makes trouble for Wilson in scenarios in which
there is no fundamental level at all, but just a limitless descent of ever-
deeper structure. If such a scenario is metaphysically possible,11 it is trouble for Wilson, for her framework can attribute no metaphysical structure
to it. After all, when nothing is metaphysically fundamental, her primitive gives no guidance. But the friend of relative fundamentality can still
make sense of metaphysical structure in such scenarios, including the
guiding idea that things are getting ever more fundamental without limit.
This relative impoverishment also makes trouble for Wilson, with
respect to making sense of structure among non-fundamental entities.
temporal order, even though these two events are not linked. (My thanks to Ross Cameron for
insightful comments which prompted these clarifications.)
11
I myself have wavered over whether such a scenario is metaphysically possible. If parts are always
more fundamental than wholes, and if gunky structures with limitless descending chains of parthood are possible, then one seems to get scenarios with no fundamental entities at all (Schaffer
2003). But if one does not assume that parts are always more fundamental than wholes, then no
such argument looms, and one may be able to respect the intuition that there needs to be an ultimate ground of being (Schaffer 2010: Sect. 2.4).

6 Ground Rules: Lessons fromWilson...

159

Suppose that what is fundamental are just particles in the void, and consider the following three non-fundamental entities: my whole body, my
whole body minus my left shoulder, and my heart. Holding fixed that
particles in the void are fundamental, and holding fixed the mereological
and other small-g relations among these three entities, there still seems
to be a residual question as to the direction of fundamentality (and one
not so different in spirit from the question of whether the ultimate parts
or the ultimate whole is basic, which inspired Wilson to add a primitive
notion of fundamentality in the first place). Again Wilsons view seems to
give no guidance.12 So overall I do not think that Wilson has successfully
blocked the fix the direction of priority argument for grounding.
My second and third objections concern whether Wilsons own framework is equally liable to the criticisms she herself levels against grounding
theorists. It seems to me that Wilsons own framework does not adequately take up her own lessons, which (to repeat) were:
Wilsons first lesson: An account of grounding must give one more than just
the bare ideology of this grounds that; it must also allow one to make
sense of follow-up inquiry into how the connection runs, in terms of the
specific rule mapping the more basic inputs to the less basic output.
Wilsons second lesson: An account of grounding should be embedded in
a formalism that outfits the notion with unified rules and so reveals useful
generalizations one would miss without the notion.

As to Wilsons first lesson, while she goes beyond the bare ideology of
this grounds that, she also stops short of saying exactly how the grounding connection works, and instead settles for some in-between resting
point involving her small-g relations. So by Wilsons lights, if Natalie
just says that the natural grounds the normative, then what she has said
Wilson herself (2014: 5646) takes up a similar example, but I am afraid that I do not understand her reply. I read her as saying that the answer turns on whether one treats the entities involved
as fusions or as functionally defined entities. But I do not see how either treatment makes a difference within Wilsons framework, unless one also has some general principle of relative fundamentality for fusions or for functionally defined entities (etc.) For suppose that my whole body, my
whole body minus my left shoulder, and my heart are all understood as fusions, and that particles
are fundamental. I see no way to extract any conclusion as to relative fundamentality for these
fusions, without some general principle connecting parthood to relative fundamentality.
12

160

J. Schaffer

is uninformative and unhelpful, perversely uninterested, and not even


giving basic information, partly for failing to specify how the grounding connection works. But if a rival metaphysicianlet us name him
Patrickjust adds (as per Wilson 2014: 547) that normative state tokens
are proper parts of natural state tokens, then by Wilsons lights Patrick
has suddenly said something informative and helpful. I find this baffling.
That is, I find it baffling to think that there is such a vast gap between
what Natalie started with and what Patrick added. By my lights, both
have said informative and potentially helpful things, though both should
say more about how exactly the connection works, and they should do so
by staring the connecting rule as precisely as possible.
The guiding analogy with causation is again helpful. Imagine that a rival
scientistlet us name her Renatastarts with Sigmunds bare causal claim
that smoking causes cancer, and adds some information about the species of
small-c causal relation involved. Perhaps Renata has a specific causative
notion of inflaming or perhaps she only has the specific causative notions
of production and dependence (cf. Hall 2004). So perhaps she says
that smoking inflames cancer, or perhaps she says that smoking produces
cancer. If someone were to say that Sigmunds original claim was uninformative and unhelpful, but that Renatas addition suddenly crosses over to
the informative and helpful, I would be baffled. With causation, what one
needs in the end is not to move to more specific causatives (though sometimes that can be a helpful step), but to specify the underlying pattern of
association as precisely as possible, which Renata has not yet done.
Wilson herself (2014: 548) asks rhetorically: [S]hould metaphysicians resist being as articulate as their metaphysical means allow in characterizing what depends on what? Insofar as Wilsons view settles for
some in-between resting point involving her small-g relations, without specifying the exact connecting rule involved, it seems to me that she
has not heeded her own wise advice.
As to Wilsons second lesson, while she is a pluralist about the concept of
grounding and dismisses it as disunified, she deploys her own preferred
conceptsincluding both her primitive notion of fundamentality, and her
various posited small-g relations such as identity, parthood, and causal
compositionin a monistic spirit. Wilson does not offer general criteria for
when a given concept is unified, nor does she give any defense of the unity

6 Ground Rules: Lessons fromWilson...

161

of her preferred concepts. So I cannot guess why she thinks that her preferred concepts are any better off than grounding. But it seems to me that
every concern she raises against grounding being too uninformative and
needing to be superseded by more specific relations could have been raised
with equal force against virtually every notion that she herself deploys.
Consider Wilsons primitive posit of fundamentality. Could the metaphysician rest with bare claims of the form this is fundamental (/this
is not fundamental)? Of course not. With the posit of fundamentality
will come the need to settle certain framework questions (e.g., does fundamentality entail existence?), and to integrate the machinery of fundamentality into the machinery of Wilsons small-g relations (e.g., Can
entities related by proper parthood both be fundamental? Can entities
related by set formation both be fundamental?) Andperhaps most relevantly given the current dialecticthere will be the question (one which
Wilson especially should face) as to whether there is a single unifed notion
of fundamentality, as opposed to a merely schematic notion standing in
for some yet-to-be-specified small-f status, such as being mereologically
atomic and being set theoretically elemental.
Or consider causation, as involved in Wilsons own notion of causal
composition. As I have argued throughout the preceding discussion, parallel issues of unity arise for both grounding and causation. Overall I find
it puzzling that Wilson dismisses the notion of grounding as disunified,
but then goes on to deploy notions such as fundamentality and causation
with no concern as to their unity. At minimum, she owes a reason for
thinking that her preferred notions are any better off.
So I conclude that Wilsons interesting and original view is not just
impoverished but also fails to heed her main lessons. Wilsons framework does not succeed in saying how exactly the grounding connection
works (at most it says something slightly more specific, using some intermediate determinables instead of specifying precise rules), and Wilsons
framework is not associated with any formalism by which the unity of
concepts is judged in a principled way (I see no stable general conception
of conceptual unity behind it, nor any reason to regard Wilsons preferred
notions such as fundamentality and causal composition as being any better off). I conclude that Wilsons alternative view is a step in the wrong
direction, even by her own lights.

162

J. Schaffer

Structural Equation Models totheRescue


So far I have extracted two lessons from Wilsons insightful critique of
grounding, and argued that Wilsons own alternative view is not just
impoverished but also fails to heed her own main lessons. Can one do
better?
I should first acknowledge that I think that previous accounts of
groundingincluding my ownalso fail to heed Wilsons lessons, and
in that respect her critique is successful. In earlier work (Schaffer 2009), I
basically spoke of a binary this grounds that relation, to which I attributed little structure beyond that of inducing a partial ordering. I included
nothing that made sense of follow-up inquiry into how the grounding
connection works, and did virtually nothing to embed grounding into
a formalism that displays its informative general structural features.
(Lessons learned!)
I equally think that Fines and Rosens accounts (Wilsons other main
targets) fall afoul of her lessons. Both work with a simple on/off grounding connection, without any natural connection to follow-up how
questions. Both do more to embed grounding into a formalism (indeed
Fine 2012 presents an exquisitely developed logic for grounding), but still
do not connect grounding to the most important surrounding notions of
inference, counterfactuals, and explanation.13
Yet in the case of causation, the technology already exists to go further
and say how exactly the causal pattern works, and to connect causation
to inference, counterfactuals, and explanation, via structural equation
models (see generally Pearl 2000; Spirtes etal. 2000). Structural equation models are systems of mathematical representation developed for
understanding causal structure, but they readily extend beyond causal
structure to any directed dependency structure. There is nothing specifically causal in the math. (The big picture: Explanation is about tracking
real dependencies (cf. Kim 1994), causation and grounding are both spe-

Indeed my main criticism of these views (Schaffer forthcoming: Sects. 4.14.3) is that they conflate grounding with metaphysical explanation, which is tantamount to conflating causation with
causal explanation.
13

6 Ground Rules: Lessons fromWilson...

163

cies of dependence, and structural equations are our best technology for
modeling dependence.)
In what remains, I briefly sketch the structural equation models I now
favor for understanding grounding (Schaffer forthcoming), and explain
how this approach fully respects Wilsons lessons (in a way already known
to work quite well for causation) and thereby has a special claim to adequacy. I am not saying that such models are perfect or that there is no
further work to be done. I am only saying that the technology already
exists to provide a fairly informative conception of grounding, and so
absorb Wilsons lessons.14
It is useful to think of a structural equation model as constructed in
three stages (Halpern 2000). First, one is trying to model some portion of
reality, so one sets up some variables to represent the system under study. In
a structural equation model, one starts off by dividing these variables into
exogenous (/independent) variables representing the basis conditions, and
endogenous (/dependent) variables representing the resulting conditions,
with all of these variables allotted a contrast space of values serving as the
options under consideration. Second, one adds in a dynamics for the system, whichin the deterministic caseconsists of specifying dependence
functions which say, for each endogenous variable, what value it takes
as output given input values for certain other variables (which thereby
count as parent variablesno parenthood loops are permitted). Third,
one adds in an assignment, whichin the deterministic casespecifies
a unique value for each exogenous variable. Once one has specified the
assignment (/set the initial conditions) and the deterministic dynamics,
the value of every other variable is uniquely determined.
To illustrate, suppose that one is trying to model how the truth-value
of a conjunction p&q depends on the truth-values of its conjuncts p and
q. Then a natural classical model would take the system under study S*
to consist of a pair of exogenous variables P and Q, and one endogenous
variable R, each allotted 0 and 1 as options to represent falsity and truth.
The dynamics L* would say that the value of R is determined by the min
See Koslicki forthcoming for criticism of my use of structural equation models for understanding
grounding. Wilson (2014: 5705) claims that self-grounding and other grounding loops are possible, which would also constitute a line of criticism to structural equation models (at least in the
form I present them). I am not convinced but I lack the space to engage with the examples here.
14

164

J. Schaffer

function on the pair {P, Q}. And the assignment A* would set P and
Q to their actual truth-values, which I will suppose is P=1 and Q=1.
Formally, this may be stated as:

S* = < {P,Q} , {R} , R* >, where R * maps all variables to {0, 1}

L* = < S*, {R <= min ( P,Q )} >

M * = < L*, { P, 1 , Q, 1 } >

Or suppose that one is trying to model how the mass of an H2O molecule depends on the masses of its atomic parts: the H, the other H,
and the O.Working in a Newtonian regime and approximating a bit, a
natural model would take the system under study S** to consist of three
exogenous variables, H1, H2, O, and one endogenous variable, H2O,
each mapped to the positive reals (R+) to represent Daltons of mass. The
dynamics L** would say that the value of H2O is determined by the addition function on {H1, H2, O} (mass is additive in Newtonian systems).
And the assignment A** would set H1 to 1, H2 to 1, and O to 16 (these
are the approximate atomic masses). Formally, this may be stated as:

{ }
{ }

< H1, {R} >, < H 2, R + >,


S * * = < {H1, H 2, O} , {H 2O} ,
>
+
+
< O, R >, < H 2O, R >

{ }

L * * = < S * *, {H 2O = H1 + H 2 + O} >


M * * = < L * *, {< H1, 1 >, < H 2, 1 >, < O, 16 >} >

Structural equation models go beyond bare causal and grounding


claims, by including a dynamics that codifies how exactly the connection works, in terms of a specific rule mapping the prior inputs to the
posterior output. We do not just have a connection. What we have is
a connection as specified by a particular function. On the causal side, a
model representing Brynners cigarette intake and cancer risk would
need to specify a dependence function relating cigarette intake to cancer

6 Ground Rules: Lessons fromWilson...

165

risk, which might be a linear function, or one that peaks at 20 cigarettes


per day, etc. Likewise with the model of conjunction-dependence just
displayed, one can not only say that the conjunct truth-values ground
the conjunction truth-value, but also say how: the pattern is as given by
the min function. (That marks the difference between conjunction-type
dependence and disjunction-type dependence whose pattern is given by
the max function.) Likewise with the model of mass-inheritance just displayed, one can not only say that the atomic masses ground the molecular
mass, but also say how: the pattern is as given by the addition function.
And so:
First lesson learned: Structural equation models give one more than just the
bare ideology of this grounds that; they allow one to make sense of follow-up inquiry into how exactly the connection runs by including specific
rules mapping the more basic inputs to the less basic output.

Structural equation models also allow one to engage in inference,


counterfactual reasoning, and explanation, in part via the dependence
functions that specify how exactly the connection works. Given that
Brynner actually smoked 100 cigarettes per day, one can infer his actual
cancer risk as the output of the relevant dependence function, given the
input of 100 cigarettes per day. Having this function also allows one
to engage in counterfactual reasoning as to what Brynners cancer risk
would have been had he smoked 0, or 20, or 200 cigarettes per day, and
to underwrite explanatory claims as to why Brynner developed lung cancer.15 Likewise with the model of conjunction-dependence: one can infer
that the conjunction is true given the actual truth of its conjuncts, one
can reason counterfactually that the conjunction would have been false if
a given conjunct had been false, and one can underwrite an explanation
as to why the conjunction is true. And likewise with the model of mass-
inheritance: one can infer that the mass of the H2O molecule is 18Da
There is dispute as to how best to treat causal explanation within the structural equation model
formalism. See, for instance, Woodward and Hitchcock 2003, and also Halpern & Pearl 2005. This
dispute concerns how to use the resources of the formalism to best capture the idea of causal explanation. But what is not in dispute is that the resources needed to make sense of causal explanation
are in place.
15

166

J. Schaffer

given the actual masses of its atomic parts, one can reason counterfactually as to what mass the molecule would have had if the atomic parts had
had different masses, and one can underwrite an explanation as to why
the H2O molecule has a mass of 18Da.
Notice that the formalism itself, and the connections to matters such
as inference, counterfactuals, and explanation, is indifferent as to which
species of causal relation is present. One does not need red arrows for
pushing or production, and blue arrows for pulling or dependence or
some other type of causal connection. The mathematics works the same
regardless, and the connections to inference, counterfactuals, and explanation are the same regardless. In this way, structural equation models
justify causal monism, by outfitting the notion of causation with uniform
rules and thereby allowing one to say something informative and worthwhile about causal relations generally.16
Exactly the same case can be made for grounding monism, given a
structural equations model treatment. The mathematics does not care
which of Wilsons small-g grounding relations is present. One does
not need red arrows for composition and blue arrows for realization
or some other type of metaphysical connection. The mathematics works
the same regardless, and the connections to inference, counterfactuals,
and explanation are the same regardless. And so:
Second lesson learned: Structural equations models are embedded in a formalism that outfits grounding with unified rules and so reveals useful generalizations one would miss without the notion.

Corollary: the notion of grounding has exactly the same claim to unity as the
notion of causation.
Putting this together, in the case of causation the technology already
exists to go beyond bare causal claims, and to articulate exactly how cause
and effect are connected, as well as to embed causation in a formalism
which treats the notion in a unified way and reveals useful generalizations
about inference, counterfactuals, and explanation which one would miss
Pearl (2010: 72) offers exactly this style of reply to the pluralist Cartwright (2007), challenging
her to cite a single example that does not fit his unitary structural equations formalism.
16

6 Ground Rules: Lessons fromWilson...

167

if one refused the notion. This technology smoothly extends to grounding. And so one finds a ready-made way to take up Wilsons lessons and
to reach an informative conception of grounding, while making sense
of the deep analogies between causation and grounding as relations of
directed dependency, and while understanding the special power both
relations have of giving us explanatory handles on the world. So I conclude that Wilson has lessons to teach the grounding theorist, but equally
the grounding theorist has ways to learn these lessons.17

References
Anscombe, G. E. M. (1975). Causality and determinism. In E. Sosa (Ed.),
Causation and conditionals (pp.6381). Oxford: Oxford University Press.
Audi, P. (2012). Grounding: Toward a theory of the in-virtue-of relation. The
Journal of Philosophy, 109, 685711.
Bennett, K. (2011). Construction area (no hard hat required). Philosophical
Studies, 154, 79104.
Bennett, K. (forthcoming). Making things up. Oxford: Oxford University Press.
Boolos, G. (1971). The iterative conception of set. The Journal of Philosophy, 68,
21531.
Cartwright, N. (2007). Hunting causes and using them: Approaches in philosophy
and economics. Cambridge: Cambridge University Press.
Correia, F. (2005). Existential dependence and cognate notions. Munich:
Philosophia Verlag.
Daly, C. (2012). Scepticism about grounding. In F. Correia & B. Schnieder
(Eds.), Metaphysical grounding: Understanding the structure of reality
(pp.81100). Cambridge: Cambridge University Press.
Fine, K. (2001). The question of realism. Philosophers Imprint, 1, 130.
Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),
Metaphysical grounding: Understanding the structure of reality (pp. 3780).
Cambridge: Cambridge University Press.

Thanks especially to Jessica Wilson, and also to Kenneth Aizawa, Ross Cameron, Janelle Derstine,
Kit Fine, Kathrin Koslicki, Jon Litland, Meghan Sullivan, and audiences at the 2014 Eastern APA,
Fordham, and the Composition and Ground Workshop (2015) at Rutgers-Newark.
17

168

J. Schaffer

Hall, N. (2004). Two concepts of causation. In J.Collins, N.Hall, & L.Paul


(Eds.), Causation and counterfactuals (pp. 22576). Cambridge: The MIT
Press.
Halpern, J. (2000). Axiomatizing causal reasoning. Journal of Artificial
Intelligence Research, 12, 31737.
Halpern, J., & Pearl, J. (2005). Causes and explanations: A structural-model
approach. British Journal for the Philosophy of Science, 56, 84387.
Hofweber, T. (2009). Ambitious, yet modest, metaphysics. In D. Chalmers,
D.Manley, & R.Wasserman (Eds.), Metametaphysics: New essays on the foundations of ontology (pp.26089). Oxford: Oxford University Press.
Kim, J. (1993). Postscripts on supervenience. In J.Kim (Ed.), Supervenience and
mind: Selected philosophical essays (pp. 16174). Cambridge: Cambridge
University Press.
Kim, J. (1994). Explanatory knowledge and metaphysical dependence.
Philosophical Issues, 5, 5169.
Koslicki, K. (2015). The coarse-grainedness of grounding. Oxford Studies in
Metaphysics, 9, 30649.
Koslicki, K. (forthcoming). Where grounding and causation part ways:
Comments on Jonathan Schaffer. Philosophical Studies.
Pearl, J. (2000). Causality: Models, reasoning, and inference. Cambridge:
Cambridge University Press.
Pearl, J. (2010). Nancy cartwright on hunting causes. Economics and Philosophy,
26, 6994.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
B.Hale & A.Hoffmann (Eds.), Modality: Metaphysics, logic, and epistemology
(pp.10936). Oxford: Oxford University Press.
Russell, B. (1912). On the notion of cause. Proceedings of the Aristotelian Society,
7, 126.
Russell, B. (1948). Human knowledge: Its scope and limits. London: George Allen
& Unwin.
Salmon, W. (1984). Scientific explanation and the causal structure of the world.
Princeton, NJ: Princeton University Press.
Schaffer, J. (2003). Is there a fundamental level? Nous, 37, 498517.
Schaffer, J. (2009). On what grounds what. In D. Chalmers, D. Manley, &
R.Wasserman (Eds.), Metametaphysics: New essays on the foundations of ontology (pp.34783). Oxford: Oxford University Press.
Schaffer, J. (2010). Monism: The priority of the whole. Philosophical Review,
119, 3176.

6 Ground Rules: Lessons fromWilson...

169

Schaffer, J. (forthcoming). Grounding in the image of causation. Philosophical


Studies.
Sider, T. (2011). Writing the book of the world. Oxford: Oxford University Press.
Spirtes, P., Glymour, C., & Scheines, R. (2000). Causation, prediction, and search
(2nd ed.). Cambridge, MA: The MIT Press.
Wilson, J. M. (2012). Fundamental determinables. Philosophers Imprint, 12, 117.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57, 53579.
Woodward, J., & Hitchcock, C.R. (2003). Explanatory generalizations. Part I:
A counterfactual account. Nous, 37, 124.

7
The Unity andPriority Arguments
forGrounding
JessicaWilson

Introduction
Grounding, understood as a primitive posit operative in contexts where
metaphysical dependence is at issue, is not able on its own to do any substantive work in characterizing or illuminating metaphysical dependenceor
so I have argued (Wilson 2014). Such illumination rather requires appeal to
one or other of the specific metaphysical relationstype or token identity,
functional realization, the determinabledeterminate relation, the mereological partwhole relation, and so ontypically at issue in these contexts.
In that case, why posit big-G Grounding in addition to the small-g
grounding relations already in the metaphysicians toolkit? The best reasons for doing so stem from the Unity argument, according to which the
further posit of Grounding is motivated as an apt unifier of the specific
relations, and the Priority argument, according to which Grounding is

J. Wilson ( )
Department of Philosophy, University of Toronto, Toronto, ON, Canada

The Author(s) 2016


K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_7

171

172

J. Wilson

needed in order to fix the direction of priority of the specific relations.


In my (2014), I considered versions of these arguments, and argued that
they did not succeed; in two recent papers, however, Jonathan Schaffer
(2016 and this volume) aims to develop a better version of the Unity argument, and offers certain objections to my reasons for rejecting the Priority
argument. In this paper, I consider and respond to these new motivations for Grounding. I start with some clarificatory remarks concerning
the dialectical import of Grounding, its assumed relata, and how I take
the ideology/ontology distinction to be relevant to the discussion; I then
present and respond to Schaffers new versions of the Unity and Priority
arguments.

Preliminaries
The Dialectical Import ofGrounding
Why posit Grounding? Certain of the original proponents, including Fine (2001), Schaffer (2009), and Rosen (2010), initially motivate
Grounding as a neo-Aristotelian corrective to overly Quinean or empiricist approaches to metaphysical theorizing: rather than ignore metaphysical dependence or treat it, unsuccessfully, in empiricist-friendly
terms like entailment or supervenience, we should return to a traditional
Aristotelian concern with what is fundamental, and what depends on
what, understood in metaphysically substantive terms as involving a distinctive, primitive notion or relation of Grounding operative in contexts
where metaphysical dependence is at issue.
The rhetoric of revolutionary revival here is inspiring, and frequently
reproduced. But it is misleading, in ways that obscure Groundings dialectical import. To start, the initial motivation for Grounding is enthymematic. As Ive previously pointed out:
Attention to metaphysical dependence is not new: many, perhaps most,
contemporary metaphysicians have spent their careers investigating forms
of such dependence, typically assumed to go beyond merely modal or

The Unity andPriority Arguments forGrounding

173

causal notions, in service of developing or assessing comprehensive theses


such as physicalism [] or of developing or assessing accounts of some
phenomenaevents, properties, possible worlds, persons, objects, laws,
causes, artifacts, institutions, and seemingly indeterminate states of affairs,
among many othersin terms of some others presumed (as a working,
speculative, or antagonistic hypothesis) to be more fundamental. These
investigations take the idioms of metaphysical dependence (in virtue of ,
nothing over and above, grounded in) to be schematic placeholders for
specific metaphysical relations [] that we have independent reason to
accept, and which serve, against the backdrop of some presumed more
fundamental base, to characterize diverse forms of metaphysical dependence in a genuinely explanatory and illuminating way. These specific relationscall them (small-g) grounding relationsinclude type identity,
token-but-not-type identity, functional realization, the classical mereological partwhole relation, the causal composition relation, the set membership relation, the proper subset relation, and the determinabledeterminate
relation, among others. (Wilson 2014, 539)

Given existing substantive accounts of metaphysical dependence, there


is no direct route from the failure of supervenience or other empiricistfriendly conceptions of such dependence to a distinctive, much less primitive, posit of Grounding.1 Proponents need some alternative motivation
for this posit.
Increasing the pressure here is that there is no hope of dispensing with
attention to small-g relations in favor of attention (only) to Grounding in
investigations into metaphysical dependence. These investigations largely
and crucially proceed by considering what implications a given smallg relation has for the existence, distinctness, efficacy, and so on, of the
goings-on whose status as dependent is at issue (as per usual: against the
backdrop of some presumed more fundamental base). Hence it is, for
example, that the primary positions in the physicalism debates vis--vis
the status of the mental as metaphysically dependent (or not) on the
1

Nor does the rhetoric of a neo-Aristotelian revival of concern with metaphysical dependence
make sense, for Aristotle operated with a variety of small-g relations, differently applied in different
cases, rather than with a primitive big-G conception. The point here isnt merely (anti-)rhetorical,
but also indicates that no ready appeal to an Aristotelian notion in good historical standing is available to proponents of primitive Grounding.

174

J. Wilson

physicalreductive physicalism, non-reductive physicalism, eliminativism, epiphenomenalism, strong emergentismare defined in terms of
their answers to these questions. Moreover, specific versions of these views
crucially appeal to features of specific small-g relations (or the lack of any
appropriate such relation) as motivating the answers at issue. Hence it is,
to focus just on a few non-reductive physicalist accounts, that Putnam
(1967) argues that taking mental states to be functional states accommodates the multiple realizability of the mental, that Yablo (1992) argues
that taking mental states to be determinables of physical determinates
accommodates the distinctive efficacy of the mental, and that Wilson
(1999) argues that taking mental states to have a proper subset of the
token powers of their physical realizers guarantees the distinctness and
physical acceptability of the mental.
As a primitive posit, however, Grounding is too abstract, on its own,
to provide answers to such questions, much less illuminating answers.
Suppose that the mental is Grounded in the physical. Does the mental
exist? Is it distinct from the physical? Is it epiphenomenal or not? If it is
efficacious, is it distinctively efficaciousefficacious qua mental? As is
reflected in the discussions of the original proponents, who express inclinations toward realism (Schaffer), anti-realism (Fine), and agnosticism
(Rosen) about Grounded goings-on, no clear answer even to the question
of existence follows from a Grounding claim. Similarly for failures of
Grounding claims. Suppose that the mental is not Grounded in the physical. Is this due to the mentals being a case in point of strong emergence,
substance dualism, eliminativism, expressivism, or what? Here again, no
answers follow just from attention to Grounding. Nor is there hope of
overcoming this underdetermination by supplementing Grounding with
general presuppositions entailing specific answers to such questions, not
just because such presuppositions will fail to accommodate various live
accounts of metaphysical dependence (a point to which I will return
down the line), but because such presuppositions will necessarily fail to
provide the fine-grained explanations of how these answers are generated
that attention to small-g relations is able to provide.
We are now in position to see past the rhetoric to the real dialectical
import of Grounding. The debate over Grounding is not over whether
metaphysicians should be concerned with dependence and prioritythey
clearly already are. Nor is it over whether investigations into dependence

The Unity andPriority Arguments forGrounding

175

and priority can or should dispense with attention to small-g relations in


favor of attention to Groundingthey clearly cant do so. What is rather at
issue is the rhetorically less revolutionary question: should we posit a primitive notion or relation of Grounding in addition to the small-g relations
that are an existing and indispensible part of the metaphysicians toolkit?
As above, in what follows, Ill consider Schaffers recent arguments for
a positive answer to the less revolutionary question. First, however, two
preliminary clarifications.

The Relata
Proponents of Grounding differ somewhat as regards the metaphysical
category of this notion. Most commonly, Grounding claims are taken to
express the holding of a relation, but there is disagreement about whether
the relata of Grounding are facts, understood as states of affairs or
Russellian propositions (Fine 2001; Rosen 2010; Audi 2012), or rather
entities of diverse ontological categories (Cameron 2008; Schaffer 2009).
There is also disagreement about the adicity of the Grounding relation
(see Jenkins 2011; Schaffer 2012), and about whether there is one or
rather multiple primitive relations of Grounding, associated with metaphysical, nomological, and normative areas of inquiry (see Fine 2012).2
Accounts on which the relata of Grounding are broadly representational, or which aim to neutrally regiment claims about metaphysical
dependence via appeals to sentential or propositional operators, reflect
a conception of Grounding as entering into explanations, suited to be
reasoned with (as in Fines logic of ground).3 As I discuss in my (2014),
my view is that in specifying the relata relevant to grounding explanations, metaphysicians should talk about the worldly goings-on directly:
compare causation and causal explanation, where theorizing cuts to the
metaphysical chase; Schaffer (2012) makes a similar point. Hence, I will
follow Schaffer in characterizing Grounding as a relation whose relata are
worldly entities, perhaps supplemented (though here too I think there is
2

Note that the specific Grounding relations here, as well as the constituent and feature-based
forms of dependence discussed in Koslicki 2012, are distinct from the specific small-g relations
Ive flagged.
3
Hence Fine (2001, 15) says, We take ground to be an explanatory relation.

176

J. Wilson

an inappropriate admixture of worldly and representational phenomena)


to allow for the contrastive account of Grounding discussed in his (2012)
and (2016).

Ideology andOntology
As well see, Schaffers pitch is sometimes presented in terms of our possessing a general concept of Grounding, rather than in terms of there
being a general relation of Grounding, even though his ultimate aim is
to motivate the latter posit. This may reflect the supposition that general
concepts bring general metaphysical posits in their wake. I think representation and reality can come apart, howeverin particular, I think that
even if there were a general concept of Grounding, it wouldnt immediately follow that there is a correspondingly general metaphysical posit,
since the concept might be given a deflationary treatment, as schematic
for or reducible to some specific metaphysical posit(s). Im interested in
the metaphysical question, so Ill pitch my remarks accordingly. Down
the line I will revisit the question of whether and when a general concept
(or associated general term) should be taken to motivate a correspondingly general metaphysical posit.

The Unity Argument


Again, the question before us is: why posit Grounding in addition to the
diverse small-g relations already on the metaphysical scene? In his (2009),
Schaffer suggests that the posit of Grounding is motivated as a unifier
of the small-g grounding relations; that is, as tracking certain important
features held in common among all the diverse forms of metaphysical
dependence:
I digress to consider a possible objection, according to which there are
many distinct notions of grounding, united only in name. [] By way of
reply, I see no more reason to consider this a case of mere homonymy,
than to consider various cases of identity as merely homonymous. In both
cases, there is a common term, and the same formal structure. This is

The Unity andPriority Arguments forGrounding

177

some evidence of real unity. At the very least, I would think it incumbent
on the objector to provide further reason for thinking that the general
term grounding denotes no unified notion. (377)

Here the primary motivation for unity stems from taking the smallg grounding relations to share the formal structure of a strict partial
orderthat is, to each be irreflexive, asymmetric, and transitive.
Schaffer has come to believe that a better Unity argument is needed.
This reflects, in part, that he now thinks that some cases of metaphysical
dependence are not transitive, and asymmetry and irreflexivity seem too
thin a reed upon which to hang a unified general posit of Grounding.4
This also reflects that, in response to my previously stated concerns, he
has come to appreciate the need for Grounding to provide a basis for
answering certain core questions about the status of Grounded entities
(or to say why a general notion of Grounding doesnt need to answer
them), and moreover to provide an account (though presumably one
more general than those provided by specific small-g relations) of how,
exactly, some goings-on metaphysically depend upon some others.
Schaffers new unification strategy, as per his contribution to this volume, as described in Schaffer (this volume), is threefold. First, he offers a
rulelet the best formalism decidefor determining when considerations of unity (defeasibly) support positing a general concept:
It seems to me that the best principled way to decide [whether to posit a general
notion] is to construct the best formalism one can for the concept. If there is no
meaningful concept, this should show up in a lack of any clear formalism, and
if there are many, this should show up in a need for a formal distinction. But if
one winds up with a clear and precise formalism that embeds the concept in a
unified way, then this is a good sign that there is a single unified concept. I offer
this as a general ground rule for unity debates: let the best formalism decide
(Rosen 2010; Schaffer forthcoming, Sect. 4.4). (153; emphasis in original)
4

See Schaffer (2012). Interestingly, Schaffers main case illustrating intransitivity (whereby a balls
surface being dented partly grounds the balls having a specific shape, and the balls having a specific
shape grounds its being more-or-less spherical, but the balls surface being dented doesnt ground
the balls being more-or-less spherical) involves mixing two different small-g relations (mereological
parthood and the determinable-determinate relation). A more straightforward case (see Wilson
2014) adverts to set membership: sets metaphysically depend on their members, but set membership is not transitive.

178

J. Wilson

Second, by way of addressing the concern that bare Grounding claims


underdetermine basic questions concerning the existence, distinctness, and
efficacy of Grounded entities, he suggests that as a matter of conceptual
entailment, Grounded entities both exist and are distinct from their associated Grounding entities, and that as a matter of conceptual exclusion, the
causal status (as efficacious, or distinctively efficacious) of Grounded entities
is appropriately left open. Third, he suggests that suitably informative yet
suitably general answers to the question of how, exactly Grounded entities stand to Grounding entities in a given case of small-g dependence may
be provided by considering the patterns of broadly counterfactual dependence made salient by modeling cases of dependence using the apparatus
of structural equations modeling (SEM) (more on this shortly). Putting
these three factors together: given that the diverse forms of metaphysical
dependence are all appropriately (and best) modeled using the SEM formalism, in ways which preserve the conceptual entailments and exclusions,
and which illuminate the general but still informative how, exactly patterns, then, Schaffer claims, this provides good (albeit defeasible) evidence
that there is a single unified concept of Grounding.
In implementing this new unity strategy, Schaffer aims to duplicate,
for the case of Grounding, unity-based motivations relevantly similar to
those he takes to support the posit of a general notion of causation:
To illustrate, consider causation [...] What would constitute a decent test for
causal monism? I offer this: develop a formalism for causation, and see
whether [...] one must distinguish red arrows for one type of causal connection from blue arrows for another type of causal connection (a sign of
pluralism), or whether one can in the end succeed in saying something clear
and informative using only a single colored arrow (a sign of monism). (154)5

Here Schaffer highlights, by analogy to causation, a second potentially


informative kind of answer to the how exactly question, besides the
5

Note that here Schaffers remarks target a general relation of causation as opposed to (just) a general concept of causation. Schaffer goes on to discuss two other reasons for endorsing genus-level
notions of causation and Grounding, respectively, according to which, first, the SEM formalism
enables one to make useful explanatory and predictive generalizations, and second, that such general notions enable one to speak open-endedly about the species-level notions or relations. Ill discuss these further motivations down the line when considering whether and how considerations of
formal or other forms of unity motivate an associated general metaphysical posit.

The Unity andPriority Arguments forGrounding

179

fine-grained answers offered by the small-g relations. For causal claims


such as smoking causes cancer, one sort of answer to the how, exactly
claim proceeds by specifying the underlying causal mechanisms, but
another sort of answer, associated with the SEM framework developed
(in particular) by Pearl (2000) and Spirtes et al. (1993), proceeds by
specifying the shape of the association between cause and effect, allowing us to embed causal claims in a deeper framework that posits not
just an on-off connection (cause or no cause?) but a more informative
function relating a range of values associated with one option (number
of cigarettes per day) to a range of values for the other (risk of cancer) (150). On the SEM framework, such patterns of dependence are
represented via structural equations specifying how the value of a given
(endogenous) child variable is determined as a function of the values of
certain (exogenous) parent variables (e.g., one linking an independent
variable representing cigarettes per day with a dependent variable representing cancer risk); these functional dependencies then constitute pattern-based information about how exactly cause and effect are related.
Schaffer suggests that Grounding claims should be similarly embedded
in a deeper framework that posits not just an on-off connection (ground
or no ground?) but a function relating a range of values for the one
option [] to a range of values for the other (150). For example, one
might model a case where the normative is Grounded in the natural, by
specifying a function from (natural) pleasure/pain ratios to (normative)
preference rankings. Indeed, Schaffer suggests that the analogy between
causation and Grounding runs so deep that we should similarly take
Grounding claims to be informatively embedded in the SEM framework,
though with some provisos to preserve the distinction between causation
and Grounding. Supposing so, then the proponent of Grounding can
take on board what Schaffer nicely calls Wilsons first lesson, according to which, as he puts it, An account of grounding must give us more
than just the bare ideology of this grounds that, and in particular must
allow us to make sense of follow-up inquiry into how exactly the grounding connection works, in terms of the specific rule mapping the more
fundamental input to the less fundamental output (151). Moreover, by
lights of the let the best formalism decide criterion, embedding specific
cases of (small-g) grounding in the SEM framework provides unity-based
reason to posit Grounding:

180

J. Wilson

[I]n the cases at handnamely the cases of causation and groundingthe


best formalism for both makes use of structural equation models, and
clearly does not need to draw different colored arrows for different flavors
of dependence [...] [This is a strong (albeit defeasible) indicator of unity.]
If so, then the notion of grounding has exactly the same claim to unity as
the notion of causation. (154)

The bracketed sentence in the quotation above reflects a sentence originally in


this passage; I include it here as explanatory of my presentation of Schaffers
argument, below. Nothing deep hinges here on whether the unity-based
motivation for Grounding is pitched in these terms or in terms of formal unifications providing a good sign that there is a unified notion of Grounding.
We are now in position to lay out Schaffers new Unity argument:
(1) If some phenomena are aptly formally unified, then this provides
strong (albeit defeasible) reason to posit a unifier. (let the best formalism decide)
(2) The diverse (small-c) causal relations are aptly formally unified by
the SEM framework.
Therefore, there is strong (albeit defeasible) reason to posit a general
notion of causation.
(3) The diverse (small-g) grounding relations are just as aptly formally
unified by the SEM framework as the diverse (small-c) causal relations.
Therefore, there is strong (albeit defeasible) reason to posit a general
notion of Grounding.
Ill now argue that each of premises 13 are false.

Against Premise 1
If some phenomena are aptly formally unified, does this in itself provide
strong (albeit defeasible) reason to posit a unifier? It seems not.
The cases of determinables and determinates and of special science
entities are illustrative. Diverse determinates are formally unified in ways

The Unity andPriority Arguments forGrounding

181

that would be constitutive of general determinables, were determinables to irreducibly exist; moreover, we have terms for, and concepts of,
determinables. But these considerations alone are not taken to provide
strong (albeit defeasible) reason to posit determinables. On the contrary,
the most common treatments of determinables are along deflationary
anti-realist or reductionist lines, according to which (on an anti-realist
view) terms for/concepts of determinables are taken to be schematic for
determinate terms/concepts, or (on a reductionist view) determinables
are taken to be identical with disjunctions of determinates. Similarly for
special science entities: diverse lower-level physical goings-on are formally
unified in ways that would be constitutive of special science entities, were
special science entities to irreducibly exist; moreover, we have terms for,
and concepts of, special science entities. But these considerations alone
are not taken to provide strong (albeit defeasible) reason to posit special
science entities. On the contrary, the most common treatments of special
science entities are in deflationary anti-realist or reductionist terms.
Why is it that formal unity alone isnt typically seen as strongly (albeit
defeasibly) motivating the posit of a general unifier? The obvious reason is that parsimony considerationsgood old Ockhams Razorpush
toward give deflationary treatments of formal unity, wherever possible. Moreover, the methodological force of parsimony is not that of a
defeaterit is not as if unity considerations first motivate general metaphysical posits, which are then potentially defeated by considerations of
parsimony. Rather, parsimony considerations are first on the scene: thou
shalt not posit entities beyond necessity. So premise (1) is false.
Similar remarks apply to a version of premise (1) understood as incorporating two other broadly unity-based considerations: first, that the
SEM formalism enables one to make useful generalizations, pertaining
to prediction, counterfactual reasoning, and explanation; second, that
a general notion or relation provides a basis for speaking open-endedly
about the species-level notions or relations.6 Here again, attention to
the standard treatments of determinables and special science entities is
6

As Schaffer (this volume) says, the theorist who refused the general notion of grounding would have
no clear way to enumerate her own prefed menu of small-g grounding relations. Wilson herself
(2014: 535) resorts to and so on when listing her own open-ended plurality of small-ggrounding
relations, and so one must wonder how she understands her own list to continue, if not in terms of
listing further species of the very genus notion that she has foresworn, namely grounding (155).

182

J. Wilson

informative, for the formal treatments of these entities also provide a


basis for useful generalizations, pertaining to predictions, counterfactual reasoning, and explanation, and for speaking open-endedly about
determinates or lower-level natural phenomena. But the standard accommodation of these features is again in deflationary terms, according to
which the availability of explanatory generalizations and use of general
concepts/terms is taken to reflect not distinctively general features of
reality but rather inexact resemblances between determinate or lowerlevel physical goings-on, which resemblances track certain patterns in
lower-level phenomena that, were we better epistemically, perceptually,
or theoretically situated, could be omitted without loss of metaphysical
generality. And here again the methodology is immediately and in the
first instance driven by parsimony considerations.7
To be sure, sometimes it is appropriate to posit a general unifier of
some diverse phenomena. But given Ockhams razor, this requires providing reasons for thinking that the commonalities at issue are tracking a distinctively unspecific level of ontological grain which cannot be
accommodated in deflationist (schematic, reductionist) terms. Hence,
for example, I argue for a non-reductive treatment of determinables on
grounds that these are associated, on a given occasion, with a proper
subset of the token powers of their realizers on that occasion, and that
candidate reductive treatments of determinables (notably: in terms of
disjunctions of determinates) fail to accommodate this proper subset
relation (since instances of disjunctions are associated with all the token
powers of the disjunct instanced on the occasion).8 And I argue for a
non-reductive treatment of certain special science entities on grounds,
first, that these have strictly fewer degrees of freedom DOF (independent
parameters required to specify their law-governed properties and behavior) than are needed to specify the (aggregative system of ) lower-level
entities upon which they depend; and second, that given the loss of information associated with such eliminations in (e.g., spin) DOF, the special
7

There are other motivations for deflationary strategies in these and other cases, including George
Berkeleys concerns about the coherence of general concepts or abstract ideas, and Jaegwon Kims
concerns about causal overdetermination.
8
See Wilson (1999) and (2009). The irreducibility at issue here is compatible with determinables
being posterior to determinates, as is usually assumed.

The Unity andPriority Arguments forGrounding

183

science entities at issue cannot be identified with any (aggregative system


of ) lower-level entitieseffectively, because the lower-level (e.g., quantum mechanical) laws cant operate without the eliminated information.9
So far as Im aware, proponents of Grounding havent offered reasons
for rejecting deflationary means of accommodating what unity exists
among the small-g relations. In any case, the moral of determinables and
special science entities remains: given Ockhams razor, if there is a strong
(albeit defeasible) presumption in the vicinity of formal or other unity, it
is in favor of deflationary rather than inflationary accounts of such unity.
Premise (1) is false.

Against Premise 2
Are the diverse (small-c) causal relations aptly formally unified by the
SEM framework? It seems not.
To start, we need to get clear about what sort of unification of the
small-c relations is needed, if the SEM-based motivation for a unified notion of causation is going to serve as an analogue for a unified
posit of Grounding. Schaffer suggests that the SEM framework formally
unifies diverse causal relations, such as baking, making, waking. But
these sorts of small-c causal relations are not relevantly analogous to
small-g grounding relations. The diverse small-g relations are of importantly different forms of dependence that might be at issue in a given
caseagain, type and token identity, the determinable/determinate relation, the set membership relation, and so on. As such, if the unity associated with the SEM formalism is to provide a model for the unity of the
small-g relations, what is in the first instance required is that the SEM
formalism unify any diverse forms of causal relation that we have reason
to think exist. Here there is room for dispute, since some accounts of
causation are presented, rightly or wrongly, as competitors. As such, we
might not require that the SEM framework formally unify all candidate

9
See Wilson 2010. The irreducibility at issue here is compatible with the special science entities and
laws being posterior to the lower-level (e.g., quantum mechanical) entities and laws, as physicalists
assume.

184

J. Wilson

forms of causation, which include (among other contenders) regularity or nomological sufficiency accounts (on which causation is a matter
of instantiation of a causal law), dispositional essentialist accounts (on
which causation involves the manifestation of a disposition), transference accounts (on which causation involves the transfer of a conserved
quantity), and counterfactual accounts (on which causation is tracked by
certain counterfactual dependencies). After all, some of these might be
wrong.
Still, independent of the end of metaphysical causal inquiry, thanks
to Hall (2004) we have good reason to think that there are (at least) two
fundamentally different, and incompatibly applicable, forms of causal
relation: first, causation as production (covering regularity, transference, and powers-based accounts), and second, causation as (counterfactual) dependence. Both sorts of accounts are needed, Hall compellingly
argues, to accommodate various theses about causationthat causation
is transitive (Transitivity), that cause and effect are connected by spatiotemporally continuous processes (Locality), that the character of a causal
relation is determined by its intrinsic features in combination with the
laws (Intrinsicness), that counterfactual dependence between wholly
distinct events is sufficient for causation (Dependence), and that omissions can be causes and effects (Omissions)that are individually true
but jointly incompatible, as is illustrated by what theses are required in
order to handle (in particular) cases of double prevention. Halls somewhat conservative interpretation of this incompatibility is that there are
at least two notions of causation associated with different of the true
theses, which are operative in different cases of causal relation:
[T]he five theses I have mentioned are, I claim, all true. Given the deep and
intractable tensions between them, that can only be because they
characterize distinct concepts of causation. Events can stand in one kind of
causal relationdependencefor the explication of which the counterfactual analysis is perfectly suited (and for which omissions can be perfectly
suitable relata). And they can stand in an entirely different kind of causal
relationproductionwhich requires an entirely different kind of analysis (and for which omissions are not suitable relata). Dependence and
Omissions are true of the first of these causal relations; Transitivity, Locality,
and Intrinsicness are true of the second. (226)

The Unity andPriority Arguments forGrounding

185

Now, the SEM formalism does not unify these two fundamentally
different small-c forms of causal relation. On the contrary, as James
Woodward (see, e.g., his forthcoming) and others have noted, the SEM
formalism aims to accommodate causation as counterfactual dependence,
and does not aim to accommodate causation as production. This is no
surprise, since SEM accounts incorporate and model the driving intuition
behind counterfactual dependence accounts, according to which causes
make a difference to their effects: counterfactually wiggle the cause, and
the effect wiggles, too. As Schaffer put it in a previous draft of his paper,
structural equation models are our best technology for understanding
difference-making relations.10
Since the SEM framework models (at best) counterfactual dependence
accounts of causation and clearly does not model production accounts
of causation (much less diverse forms of such accounts), this framework
does not aptly formally unify the diverse small-c relations, in the relevant
sense. Premise (2) is false.

Against Premise 3
The falsity of premises (1) and (2) undermines the sub-conclusion of
Schaffers argument, according to which there is strong (albeit defeasible)
reason to posit a general notion of causation: the SEM framework does
not in fact formally unify the relevant small-c causal relations, and even
if it did, Ockhams razor would push toward trying to accommodate
such unity in deflationary terms, antecedent to positing a general unifier.
These results in turn technically undermine the value of premise (3)
according to which the diverse small-g grounding relations are just as
aptly formally unified by the SEM framework as the diverse (small-c)
causal relationsas generating the desired unity-based motivation for
Grounding, since at this point the truth of premise (3) is compatible with
the SEM frameworks not aptly formally unifying the small-g relations.
10

As such, the formal unity afforded by the SEM framework in modeling, for example, baking,
making, and waking is at best a unification of different applications of a counterfactual dependence account. But the formal unity associated with different applications of a single small-c relation is beside the point of motivating Grounding as a unifier of the diverse small-g relations.

186

J. Wilson

Let us put aside the analogy to small-c relations, however, and independently ask: does the SEM framework aptly formally unify the small-g
relations? It seems not.
We can start by observing that for the SEM framework to properly
model grounding relations, grounded goings-on must counterfactually
depend on grounding goings-on: wiggling the ground must wiggle the
grounded. Thats the whole point of the SEM-based approachto identify
or model dependence relations, of whatever sort, as reflected in counterfactual dependencies. But, for reasons Ill discuss shortly, many grounded
goings-on are not counterfactually dependent on grounding goings-on.
For example, and to start, suppose one wants to model the dependence
of determinables on determinates, against the backdrop assumption that
fundamental reality is maximally determinate. More specifically, suppose
one wants to model the metaphysical dependence of a shirts being red, at
a time, on the shirts being maroon, at that time. On the face of it, there
is no counterfactual dependence in this case; for both intuitively and on
every similarity-based account of counterfactuals (i.e., on nearly every
live account), the counterfactual if this shirt werent maroon, it wouldnt
be red is false, since in the closest worlds where the shirt isnt maroon, it
is some other determinate of red. Koslicki (2016) precisifies this observation by constructing an SEM model for this case, on which the exogenous
variables represent determinate states of the shirtMaroon, Crimson,
Navy, and so on; the endogenous variables represent determinable states
of the shirtRed, Blue, and so on; and the structural equations connect
these variables in the obvious ways (Maroon = Red, Navy = Blue, and so
on). Having done so, she notes:
There is now reason to doubt whether [] the model at hand actually
encodes how the shirts determinate shade sets its determinable color, as
Schaffer claims []. Given that Maroons being set to 0 leaves open, for
example, whether Crimson should be set to 1in the scenario in question
[it] would therefore be incorrect to define [the relevant structural equation] in such a way that it assigns 0 to Red whenever 0 is assigned to
Maroon; for the scenario in question may be one in which the shirt is
nevertheless red, only in some other way, for example, by being crimson
rather than maroon. This result presents a counterexample to Schaffers

The Unity andPriority Arguments forGrounding

187

slogan, wiggle the ground, and the grounded wiggles (Schaffer 2016,
Sect. 3.2): for in a case in which we wiggle the ground by imagining the
shirts color to be changed from maroon to crimson, say, it is not the case
that thereby the grounded wiggles as well, since the shirt continues to be
red, only in a different way. (107)

Koslicki also helpfully observes that the concern here is a Grounding


variation on the theme of causal preemption:
When the structural equation model is applied to an alleged case of determinable/determinate grounding, the grounding scenario in question is in
fact more aptly compared to a causal scenario involving massive causal
preemption, i.e., a scenario in which a single effect can be brought about by
multiple alternative causes, each of which is individually sufficient to bring
about the effect in question and each of which occurs only if none of the
others occur. As it stands, it is not clear, even in the causal case, how the
structural equation model, as described by Schaffer, would produce the
correct results in a case of massive causal preemption. At most, then, we are
dealing with a situation in which a supposedly clear case of grounding is
comparable to a problematic case of causation, one which has led to headaches for extant theories of causation including, by Schaffers own admission, the structural equation model of causation. (108)

I agree with Koslicki both that the SEM formalism does not appropriately model the determinable/determinate case, and that the difficulty here is of a piece with notorious difficulties that difference-making
accounts of causation have in modeling cases of causal preemption. In
what follows, I want to defend, develop, and generalize these concerns.
To start, Schaffer would presumably reject Koslickis specific model
of the determinable/determinate case, on grounds that, like causation,
grounding is a contrastive notion, such that grounding claims have the
general form a rather than b grounds c rather than d. To be sure, relative
to the fine-grained contrast class of the exogenous variables in Koslickis
model, where Maroon contrasts not just with Navy but with other determinates of red, the counterfactual dependence between determinate and
determinable will not be in place, but, Schaffer might say, thats not a failure of the SEM modelits a failure of the modeler to correctly specify

188

J. Wilson

an appropriately coarse-grained contrast (wiggle) class.11 Its a model,


after all, and the variables can and should be set up in whatever way
is deemed perspicuous. On Schaffers proposed model, there is a single
exogenous variable (Determinate), which is set to 1 if the shirt is maroon,
and to 0 if it is navy, and a single endogenous variable (Determinable),
which is set to 1 if the shirt is red and to 0 if the shirt is blue. On this
model, wiggling the determinate color of the shirt (from maroon to navy)
does wiggle the determinable color of the shirt (from red to blue).
Ultimately, however, the appeal to a coarse-grained contrast class fails
to show that the SEM framework aptly models the determinate/determinate case. For an apt model of this case should be able to capture
the metaphysical dependence of the shirts being red on the shirts being
maroon, independent of the contrast between red and blue. After all, the
world might have been one according to which everything was one or
other shade of red. In such a world, the usual reasons for thinking that
determinable instances metaphysically depend on determinate instances
might have remained in placeit might have been reasonable to suppose
that fundamental reality is maximally determinate, and that determinable instances are ontological abstractions from determinate instances.
But if we were denizens of such a world, then like a close cousin of Black
and White Mary, we would not have possessed the concepts of determinable colors other than red. We might not have even contemplated the
bare possibility of other determinables, or we might have rejected their
bare possibility, concluding (rightly or wrongly, given further details) that
the redness of things was metaphysically necessary. Either way, we would
be unable to model the determinable/determinate dependence at issue by
means of the SEM framework.
Indeed, we dont need to go to an all-red world to generate the difficulty here. We just need to consider cases of metaphysically necessary
determinables, such as shape. Just as the shirts being red depends on the
shirts being maroon, so does the shirts being shaped depend on its being
some determinate shape. But here there is, of metaphysical necessity, no
11

Nor, I think he would say, need we require (as Koslicki seems to assume) that the variables in the
determinable/determinate case exactly mirror the variables in the throw/shattering case that he
originally discusses, in representing the onoff obtaining of events or states of affairs.

The Unity andPriority Arguments forGrounding

189

contrast class to which one might appeal in order to set up any counterfactual dependencies, for whatever determinate shape properties the shirt
has, it will still be shaped.12
Ultimately, then, the SEM framework does not aptly model the metaphysical dependence of determinables on determinates, since gaining the
requisite counterfactual dependence requires an appeal to coarse-grained
contrast classes that may be either conceptually unavailable or metaphysically impossible. But, as Schaffer grants, one might well maintain that
determinables metaphysically depend on determinates.13 Hence, the
SEM framework does not formally unify (all) the small-g relations, and
premise (3) is false.
This result can be pressed further by pinpointing and generalizing the
underlying reason why the SEM framework goes wrong in the determinabledeterminate case. As above, Koslicki observes that the failure
of determinables to counterfactually depend on their associated determinates is analogous to the failure of effects, in cases of preemption, to
counterfactually depend on their associated causes: in both cases, the
dependence at issue is strongly immune to counterfactual variation,
effectively due to there being a manyone structure between dependence
base and dependent goings-on. In the determinabledeterminate case,
however, we can say more; for here the manyone structure is generated by the determinabledeterminate relations being, on the operative
understanding, an abstraction relation taking more specific to less specific
goings-on. It is the washing away of determinate-level details that makes
it the case that wiggling the determinate often and sometimes necessarily fails to wiggle the determinable.
That the underlying concern in the determinabledeterminate case is
generated by this relations taking more to less specific goings-on spells further trouble for the claim that the SEM formalism aptly unifies the smallg relations. For many paradigm cases of small-g relations are such that the
associated grounded goings-on are ontologically and causally less specific
than their grounding goings-on. This is true, for example, of many of the
12

Indeed, even if the shirt is a vague object, it will still be (indeterminately) shaped.
Indeed, one might well maintain this even if, as I argue in Wilson 2012, determinables can be
fundamental; for as I discuss in Wilson 2014, metaphysical dependence can be symmetric.
13

190

J. Wilson

small-g grounding relations posited by physicalists as holding between


special science and lower-level physical goings-on, including accounts on
which special science goings-on are characterized by abstract functional
roles that can be implemented by diverse lower-level realizers, accounts
on which special science goings-on are associated with proper subsets of
the powers of their diverse lower-level realizers, and accounts on which
special science goings-on have strictly fewer DOF than are had by their
diverse lower-level realizers. Indeed, it is even clearer in these cases than
it is in the case of the determinable/determinate relation that the whole
point of the relation, from a metaphysical point of view, is to characterize
dependent goings-on whose comparatively abstract nature enables them
to float largely free of actual or counterfactual variations in the goings-on
upon which they depend. It is this insensitivity that accommodates, for
example, the multiple realizability of special science properties, and the
compositionally flexible persistence conditions of special science entities.
These relations are in the business of same-making, not difference-making.
Hence it is that, in cases of dependence involving these relations, difference-making considerations involving counterfactually nearby states of
affairs are typically not in place, and difference-making considerations
involving counterfactually distant states of affairs are, even if in place,
irrelevant to characterizing the same-making form of dependence at issue.
It is moreover worth noting that the abstractionist character of these
and other small-g relations blocks a potential line of response on the
part of Schaffer and others14 who take Grounding to be appropriately
modeled by the SEM framework. Here the suggestion starts with the
claim that the analogy between Grounding and causation runs deep,
with it being difficult, in each case, for counterfactual dependence
approaches to appropriately model manyone cases of dependence,
with the tricky cases of small-g relations being relevantly similar to
cases of causal preemption and cases where a given effect is characterized in general terms; the further suggestion is then that variations on
available strategies for overcoming these difficulties in the causal case
may overcome the difficulties in the grounding case. But granting
14

Notably, Alistair Wilson, in his (in progress).

The Unity andPriority Arguments forGrounding

191

(as a generalization of Koslickis observation) that the failures in both


cases are relevantly similar, there is no hope that the strategies for accommodating the problematic cases of causal relation will carry over. For the
strategies for gaining back counterfactual sensitivity in cases of causal
preemption or of general effects crucially rely on characterizing the effect
in more fine-grained terms; this is, for example, the recommended strategy for accommodating the full range of cases of preemption in Lewis
(2000) and Paul (2000). But while there is no barrier to reconceiving
of preempted or overly general effects in more fine-grained terms, this
strategy is not available for cases of abstractionist grounding, since in
such cases the holding of the relation requires that the grounded be less
specific than the ground.
Summing up: not just the determinabledeterminate relation but many
paradigm cases of small-g grounding relations fail to be aptly modeled by
the SEM formalismand insuperably so. As such, the SEM framework
does not aptly formally unify the small-g relations. Premise (3) is false.
Ive now argued that all three premises of Schaffers Unity argument
are false. I conclude that considerations of formal and other unity, of the
sort associated with the SEM formalism, in particular, provide no reason
to posit a general relation of Grounding.

The Priority Argument forGrounding


Another proposed motivation for Grounding is that Grounding is
required in order to fix the direction of priority of small-g relations, since
in some cases these relations fail to do this on their own.15 For example,
given just that some entities stand in the partwhole relation, nothing
follows about whether the parts depend on the whole, or vice versa. In
that case, the suggestion goes, something more is needed to determine
what is prior to whatnamely, Grounding. My (2014) response to this
argument is to grant that something more is needed, but to deny that
the additional component is Grounding. Rather, I argue, by attention to
15

Perhaps some small-g relations are able to fix the direction of priority on their own; in cases of set
membership, for example, perhaps members are always prior to their containing sets.

192

J. Wilson

standard metaphysical methodology, that in the first instance what more


is needed for the small-g relations to do their work is a primitive specification of the fundamental.16 Schaffers indirect support for the Priority
argument targets my proposed alternative, in ways I will consider after
presenting my alternative and illustrating its application.

The Primitive Fundamentality Framework


There are two cases where the direction of priority associated with the
holding of a given small-g relation might be at issue: first, cases where the
relation connects fundamental to non-fundamental goings-on; second,
cases where the relata are each non-fundamental.
For the first sort of case, I argue that, as attention to standard metaphysical methodology shows, what more is needed is specification of
what is presumed, as a speculative, antagonistic, or working hypothesis,
to be fundamental. Hence, given that the Whole is fundamental (as per
monism), then proper parts of the Whole are non-fundamental; given
that atoms are fundamental (as per atomism), fusions of the atoms are
non-fundamental; given that the fundamental goings-on are maximally
determinate (as physicalists commonly assume), then determinables of
these goings-on are non-fundamental; and so on. So Grounding is not
needed in order for the small-g relations to fix the direction of priority
between fundamental and non-fundamental goings-on.
Before continuing, one might ask: what if the fundamental is understood as that which is not itself grounded? If so, then since on my proposal
the small-g relations (typically) do their work only after the fundamental is specified, the notion of the un-grounded operative in characterizing the fundamental must appeal to Grounding, after all. But, I argue,
we should not understand the fundamental as the un-Grounded, both
because doing so inappropriately metaphysically characterizes basic entities in non-basic (indeed, relational negative) terms, and because such
a characterization rules out of court various live metaphysical views on

16

Or of what serves as fundamental; see my 2014, note 64, and below.

The Unity andPriority Arguments forGrounding

193

which the fundamental goings-on are self-grounding (as per, e.g., a selfsustaining god) or mutually grounding (as per, e.g., Leibnizian monads).
Rather, we should metaphysically characterize the fundamental in primitive, metaphysically neutral termsafter all, if anything is fundamental,
its the fundamental! Though the fundamental is primitive, we can say
more about this notion; namely, that it follows from what goings-on are
fundamental at a world that these, individually or together, provide a
ground (nota bene: in one or other small-g fashion) for all goings-on at
the world. Such a conception encodes the intuitive, commonly registered
understanding of the fundamental in terms of all God had to do to create the world. Perhaps, in drawing to attention that what is fundamental
sets the valence for certain priority relations, I may be seen (as Schaffer
suggests) as here introducing a new hyperintensional primitive notion;
but really, I think I am just making explicit the presuppositions of standard metaphysical methodologyas in, for example, Schaffers (2010)
descriptions of monism and pluralism:
The monist holds that the whole is prior to its parts, and thus views the
cosmos as fundamental, with metaphysical explanation dangling downward from the One. The pluralist holds that the parts are prior to their
whole, and thus tends to consider particles fundamental, with metaphysical explanation snaking upward from the many. Just as the materialist and
idealist debate which properties are fundamental, so the monist and pluralist debate which objects are fundamental. (31)

Indeed. And the way the monist and the pluralist go about debating
which objects are fundamental is, again, to first assume (as a working,
speculative, or antagonistic hypothesis) one or the other fundamental
base, and then go on to explore which such base best accommodates the
rest of the reality, by appeal (in sometimes complex fashion) to various
small-g relations understood as holding between fundamental and nonfundamental goings-on.
Moving now to the second case: what about priority relations between
goings-on each or all of which are non-fundamentalsay, between
hands and bodies? A specification of the fundamental wont, in itself,
always fix the direction of priority between such non-fundamenta: for

194

J. Wilson

example, atomists might agree about what is fundamental, but disagree


about whether hands are prior to bodies, or vice versa. So, how are priority relations between non-fundamenta, presumably also involving smallg relations, determined?
My treatment of the second case again encodes standard metaphysical
methodology. To start, investigating into dependence relations between
non-fundamental goings-on (non-fundamenta) requires that one be in
possession of fairly specific accounts of the non-fundamenta in terms of
fundamental goings-on (fundamenta)else one wouldnt be in position
to characterize the former goings-on as non-fundamental. As per the treatment of the first case, such accounts of non-fundamenta in terms of fundamenta appeal to the holding of various small-g relations, where one of the
relata is fundamental and the other is non-fundamental. These accounts
of non-fundamenta, in turn (more specifically, their metaphysical correlates), provide a basis, along with further suppositions or associated facts
about non-fundamenta and their relations, for priority relations between
non-fundamenta. As well see, there is considerable room for debate about
which further suppositions and associated priority relations are (or are not)
in place, even holding fixed the operative account of whatever non-fundamenta are at issue, but in any case, no appeal to Grounding is required.
By way of illustration, suppose that a form of atomism is true, on
which hands and bodies are mereological fusions of atoms, and hands
are mereological proper parts of bodies. Which is prior: hands, bodies,
or neither? The answer depends on which further facts are (assumed
to be) in place. A mereological atomist might maintain that hands are
prior to bodies, on grounds that a hand fusion can exist without any
body fusion existing, but not vice versa. Alternatively, a mereological
atomist might maintain that hands and bodies are on a par, prioritywise: they overlap; one is bigger; littler ones can exist without bigger
ones existing and not vice versabut so what? Or suppose that atomism is true, but hands and bodies are functionally realized by atomic
aggregates. Which is prior: hands, bodies, or neither? Again, the answer
depends. A functionalist atomist might maintain that bodies are prior
to hands, on grounds that a bodys function can be implemented in
the absence of a hand, but not vice versa. Alternatively, a functionalist atomist might maintain that hands are prior to bodies, on grounds
that a bodys function sensitively depends on the functions of its parts,

The Unity andPriority Arguments forGrounding

195

including its hands. As these examples suggest, no appeal to Grounding


is needed for the associated non-fundamenta to (be reasonably taken to)
stand in priority relations (or not to stand in any such relations, as the
case may be). What is rather needed, in addition to the suppositions/
facts about what is fundamental and which small-g relations are in place
between the fundamental and the non-fundamental goings-on, are suppositions/facts about the natures of the non-fundamenta and how (via
one or other small-g relation) the non-fundamenta stand to one another.
Summing up: on the primitive fundamentality framework, facts about
what is (or serves as) fundamental, coupled with the holding of diverse
small-g relations, determines what is non-fundamental; these facts,
coupled with facts about the natures of the non-fundamenta and how
the non-fundamenta stand to one another, determine priority relations
between non-fundamenta (if such there be, in a given case). No appeal to
Grounding is required.

Three Concerns withthePrimitive Fundamentality


Framework
Schaffer (this volume) thinks that the primitive fundamentality framework is worthy of serious consideration. But he offers three reasons for
thinking that this framework is worse off than one rather appealing to
primitive Grounding, understood as a primitive hyperintensional notion
of relative fundamentality. If Schaffer were correct, that would resurrect the Priority argument for Grounding. In what follows, I argue that
Schaffer is not correct.

Expressive Power
Schaffer claims that the primitive fundamentality framework is expressively impoverished as compared to the Grounding framework:
It seems to me that absolute fundamentality can easily be defined in terms
of relative fundamentality (the fundamental is that which has no
deeper grounds), and so a framework using relative fundamentality as a
primitive can easily be used to say everything one wants to say via absolute

196

J. Wilson

fundamentality. But there is no obvious definition to be found in the other


direction, and so it is not at all obvious that using absolute fundamentality
as a primitive will allow one to say everything one wants to say in terms of
relative fundamentality. (158)

I respond by restating my (2014) position that the notion of the fundamental17 should not be defined in terms of relative fundamentality (the
fundamental is that which has no deeper grounds); for we should not
metaphysically define or characterize the fundamental in non-basic, theoretically loaded terms. In particular, it is not true that a framework using
relative fundamentality as a primitive can easily be used to say everything
one wants to say via absolute fundamentality, since the relative fundamentality framework will not allow one to express a number of currently live
metaphysical theses, including views positing a self-sustaining God, mutually dependent monads, and so on. And while it is true that on my account
there is no definition of relative fundamentality just in terms of fundamentalityor, more to the metaphysical point, that the facts about relative fundamentality are not generated just by the facts about fundamentalityon
my view the facts about relative priority, if such exist in a given case, are
generated by facts about what is fundamental, coupled with off-the-shelf
resources about small-g relations and their features and implications in specific circumstances. Since my framework can say everything Schaffers can
say and then some, it is not a fundamentality framework but a Grounding/
relative fundamentality framework that is expressively impoverished.

Priority intheAbsence ofaFundamental Level


Schaffer claims that a primitive fundamentality framework faces difficulties in cases where there is no fundamental level:
[The need for primitive fundamentality] makes trouble for Wilson in scenarios in which there is no fundamental level at all, but just a limitless
descent of every deeper structure. If such a scenario is metaphysically possible, it is trouble for Wilson, for she can attribute no metaphysical structure
17

Contra Schaffers exegesis, I do not use the expression absolute fundamentality, or appeal to any
such notion, for reasons that I discuss in my 2014 (note 64), and upon which I will expand shortly.

The Unity andPriority Arguments forGrounding

197

to itwhen nothing is metaphysically fundamental, Wilsons primitive


gives no guidance. But the friend of relational fundamentality can still make
sense of metaphysical structure in such scenarios, including the guiding idea
that things are getting ever more fundamental without limit. (158)

I respond, to start, by noting two ways in which priority might be fixed,


on my account, in worlds with infinite chains of dependence (hence, I
reject Schaffers characterization of my view as one requiring an absolutely fundamental level). The deeper point illustrated by these possibilities is that fixing the direction of priority does not require an absolutely
fundamental level; rather, an appropriately principled metaphysical
asymmetry that enables some goings-on (which may or may not exist, as
per the cases below) to act as a fundamental level will do.
First, priority might be fixed in the absence of a fundamental level
if there is convergence on a fundamental level, and non-fundamental
goings-on depend on goings-on in that limit. The suggestion here
extends Monteros (2006) observation that even successive decompositions can still bottom out into something fundamental. For example, just
as the infinite decreasing sequence of numbers 1/2, 1/3, 1/4 is still
bounded below by zero, there could be infinite descending sequences of
decompositions, with fundamental entities below them all (179). What
I furthermore add (or take away) from Monteros line of thought is that
even if the goings-on in the limit do not exist, the valence of priority may
still be established by reference to goings-on in this limit, much as the
thermodynamic properties and behavior of a gas are properly modeled
as non-fundamental features of statistical mechanical collections in the
thermodynamic limit, as the number of particles and the volume each
approach infinity.18 In other words, goings-on in the limit may act as a
fundamental level.
Second, priority might be fixed in the absence of a fundamental level
if there is a level at which the archeology of further dependence relationsceases to be relevant to priority relations at or above that level.

18

The possibility broached here might also be seen as a metaphysical version of the temporal supertask discussed in Cameron 2008, 9.

198

J. Wilson

Such a level acts as a fundamental base for what lies above; hence, for
example, the physical level might operate as a fundamental level for purposes of understanding priority relations among broadly scientific phenomena, even if the physical entities are non-fundamental relative to
some deeper level of reality.19 The existence of this sort of structure would
also provide a principled basis for fixing the valence of priority; hence
it is that Montero (2006) suggests that the physicalist commitment to
the priority of the physical over the mental can be accommodated in an
infinitely decomposable world as the thesis that all mental properties
are eventually determined by non-mental properties such that all further
determinations of these properties, if any, are non-mental (187).
Again, these possibilities indicate that associating a direction of priority with a given small-g relation (assuming that the relation does not do
this on its own) does not require an absolutely fundamental level; rather,
an appropriately principled metaphysical asymmetry in the structure of
reality, enabling some goings-on to act as a fundamental level, will do.
But what if there is no fundamental level, no convergence on a fundamental level, and no level at which deeper archeology ceases to matter?
In that case, one may reasonably deny that it makes sense to posit any
priority relations between non-fundamenta (besides those fixed just by
the relation alone, as might be the case with the set membership relation).
As Leibniz said in his correspondence with Arnaud,
Where there are only beings by aggregation, there are no real beings. For
every being by aggregation presupposes beings endowed with real unity,
because every being derives its reality only from the reality of those beings
of which it is composed, so that it will not have any reality at all if each
being of which it is composed is itself a being by aggregation, a being for
which we must still seek further grounds for its reality, grounds which can
never be found in this way, if we must always continue to seek for them.
(1686/1989, 85)

19

In my 2014, I referred to such a level as relatively fundamental, which reference is not to be


confused with Schaffers use of relatively fundamental.

The Unity andPriority Arguments forGrounding

199

Here, as Cameron (2008) puts it, the thought is that if everything were
dependent, there would be no grounding to being (67). In further
support of this thought, consider again the all God had to do to create
the world heuristic that underlies the conception of the primitively fundamental in my framework. In cases where there is no convergence, and
no level of goings-on that fix the priority relations at higher levels, then
to create the world God would have to bring into being all the levels, and
all the goings-onthat is, God would have to do, or create, everything.
Hence, on the operative understanding of the fundamental (and again,
modulo any fixed directions of priority there might be) everything would
be on a par, priority-wiseand thats just to be expected. Indeed, Schaffer
(2010) agrees: There must be a ground of being. If one thing exists only
in virtue of another, then there must be something from which the reality
of the derivative entities ultimately derives (37).20

Relative Fundamentality
Schaffer claims that a primitive fundamentality framework has difficulty
accommodating priority relations between non-fundamenta:
[Wilsons framework has trouble] in making sense of structure among nonfundamental entities. Suppose that what is fundamental are just particles in
the void, and consider the following three non-fundamental entities: my
whole body, my whole body minus my left shoulder, and my heart. Holding
fixed that particles in the void are fundamental, and holding fixed the
mereological and other small-g relations among these three entities,
there still seems to be a residual question as to the direction of fundamentality (and one not so different in spirit from the question of whether the
ultimate parts or the ultimate whole is basic, which inspired Wilson to add
a primitive notion of fundamentality in the first place). [] Wilsons view
seems to give no guidance. (1589)
20

Schaffers commitment to a fundamental level opens the door to a fourth response to his objectionnamely, to deny that it makes sense to posit a world without any fundamental base. Hence
it is that in his (2010) he offers as an advantage of monism that it can accommodate both infinite
decomposition and the reasonable assumption that dependence relations require a ground of
being.

200

J. Wilson

Schaffer notes, more specifically, that I provide an account of relative fundamentality, which aims to treat the sort of case he mentions, but says:
I am afraid that I do not understand her reply. I read her as saying that
the answer turns on whether we treat the entities involved as fusions or
as functionally defined entities. But I do not see how either treatment
makes a difference, within Wilsons framework, unless we have some
general principle to hand of relative fundamentality for fusions or for functionally defined entities (etc.). For suppose that my whole body, my whole
body minus my left shoulder, and my heart are all understood as fusions,
and that particles are fundamental. I see no way to extract any conclusion
as to relative fundamentality for these fusions, without some general principle connecting parthood to relative fundamentality. (159, note 12)

I respond that Schaffer has not read me correctly, though I think I


could have been clearer in my original exposition. On my view, the direction of priority between non-fundamenta is not assumed to follow just
from the (small-g) facts about what is fundamental coupled with facts
about how non-fundamenta stand to fundamentahence it is that even
the mereological atomist has options so far as understanding the priority
relations between hand and body. Even more options arise if the nonfundamenta fusions are functionally characterized, if more complex parthood relations (of the sort discussed by, e.g., Fine 1999; McDaniel 2001;
Paul 2002; Koslicki 2008) are at issue, if there is disagreement about the
persistence conditions of the non-fundamenta, if endurantist or perdurantist conceptions of objects are presupposed, and so on. Relatedly, nor
would it make sense to assume that priority between non-fundamenta
of a given variety (e.g., fusions) is algorithmically determined with the
help of some general principle. Rather, what emerges from attention to
metaphysical methodology is that relative fundamentality is a matter of
suppositions/facts about what is fundamental and how the non-fundamenta small-g depend on the fundamenta, along with (not general principles, but) suppositions/facts about the natures of the non-fundamenta
and how (via one or other small-g relation) the non-fundamenta stand to
one another. So that my view does not entail or encode general principles
of relative fundamentality is a feature, not a bug.

The Unity andPriority Arguments forGrounding

201

Here it is also worth pointing out that what priority relations hold
between non-fundamenta is further complicated by its being the case that
the dependence of a given non-fundamental entity on another is typically
wrapped up in not one, but a number of relations. For example, even
supposing that there is a clear sense in which my hand qua functionally
specified entity depends on my body qua functionally specified entity,
but not vice versa, there is an equally clear sense in which the weight of
my body, at least at the present moment of writing, is in part metaphysically dependent on the weight of my hand, in that the weight of my body
is clearly an additive function of the weight of its parts, and certainly the
things that my body, even if abstractly functionally specified, can doits
powers, so to speak, depend to some extent on the powers of my hand.
That relations of relative fundamentality are not properly seen as subject to uniformly applicable general principles, for even a single small-g
relation, much less (as the proponent of Grounding supposes) for all such
relations, spells deep and to my mind insuperable trouble for a framework that appeals to primitive Grounding/relative fundamentality. As I
previously rhetorically asked:
Is all this complexity supposed to involve numerous Grounding relations,
primitively pointing in different directions? The idea is just plain silly, and
suggests that, even if there were some problem (which there is not) with the
specific relations not being themselves up to the task of fixing directions of
priority among non-fundamental goings-on, the posit of additional
Grounding relations would not be of any help. (Wilson 2014, 566)

Non-rhetorically: no. Nor, given that the SEM framework does not formally unify the small-g relations, can this framework be appealed to as
a general basis for rendering the primitive pointings of Grounding any
more substantive. We can do no better, in such investigations, than to
work closely with the diverse relations that are plausibly taken to hold
between non-fundamental goings-on, as informed by the accounts of
these non-fundamenta in terms of fundamental goings-on (or goings-on
that properly serve as fundamental), making explicit in the process what
assumptions are guiding our claims that one or another of these is, in a
given case, operating as a grounding relation (or not).

202

J. Wilson

Summing up: the primitive fundamentality framework is not subject to any of Schaffers three concerns: it is not expressively impoverished, it has resources for dealing with the absence of a fundamental
level, and it accommodates relative fundamentality in an appropriately
fine-grained and articulate way. This is no surprise, since the framework
encodes the usual suppositions and strategies of standard metaphysical
investigations into dependence and priority. Moreover, and by way of
contrast, the primitive Grounding/relative fundamentality framework is
both expressively impoverished andfor all that Schaffer has yet establisheddeeply inarticulate as regards characterizing the diverse and
complex network of relations of relative fundamentality. I conclude that
considerations of priority provide no reason to posit a general relation
of Grounding.
Acknowledgements Thanks to Jonathan Schaffer, an interlocuter par excellence, and to the Dr. Martin R. Lebowitz and Eve Lewellis Lebowitz Foundation,
along with the Phi Beta Kappa Society and the American Philosophical
Association, for making possible the fruitful debate that led to this paper. Thanks
also to members of audiences at the Eastern APA and Fordham Lebowitz Prize
lectures, the University of Buffalo, the University of Notre Dame, the University
of Tennessee, the University of Edinburgh Workshop on Grounding, and the
Rutgers Newark Workshop on Composition and Ground, for helpful comments
and questions.

References
Audi, P. (2012). Grounding: Toward a theory of the in-virtue-of relation. Journal
of Philosophy, 109, 685711.
Cameron, R.P. (2008). Turtles all the way down: Regress, priority and fundamentality. Philosophical Quarterly, 58, 114.
Fine, K. (1999). Things and their parts. Midwest Studies in Philosophy, 23,
6174.
Fine, K. (2001). The question of realism. Philosophers Imprint, 1, 130.
Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),
Metaphysical Grounding (pp. 3780). Cambridge: Cambridge University
Press.

The Unity andPriority Arguments forGrounding

203

Hall, N. (2004). Two concepts of causation. In J.Collins, N.Hall, & L.Paul


(Eds.), Causation and counterfactuals (pp.225276). Cambridge, MA: MIT
Press.
Jenkins, C.S. (2011). Is metaphysical dependence irreflexive? The Monist, 94,
267276.
Koslicki, K. (2008). The Structure of Objects. Oxford: Oxford University Press.
Koslicki, K. (2012). Varieties of ontological dependence. In F. Correia &
B.Schnieder (Eds.), Metaphysical Grounding: Understanding the Structure of
Reality (p.186). Cambridge: Cambridge University Press.
Koslicki, K. (2016). Where grounding and causation part ways: Comments on
Jonathan Schaffer. Philosophical Studies, 173, 101112.
Leibniz, G. W. (1686/1989). Philosophical Essays. Ariew, R. and Garber, D.
(Eds.). Indianapolis: Hackett.
Lewis, D. (2000). Causation as influence. The Journal of Philosophy, 97,
18297.
McDaniel, K. (2001). Tropes and ordinary physical objects. Philosophical Studies,
104, 269290.
Montero, B. (2006). Physicalism in an infinitely decomposable world.
Erkenntnis, 64, 177191.
Paul, L.A. (2000). Aspect causation. The Journal of Philosophy, 97, 235256.
Paul, L.A. (2002). Logical parts. Nos, 36, 57896.
Pearl, J. (2000). Causality. Cambridge: Cambridge University Press.
Putnam, H. (1967). The nature of mental states. In W.H. Capitan & D.D.
Merrill (Eds.), Art, Mind, and Religion (pp.1223). Pittsburgh: Pittsburgh
University Press.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
B.Hale & A.Hoffmann (Eds.), Modality: Metaphysics, logic, and epistemology
(pp.10936). Oxford: Oxford University Press.
Schaffer, J. (2009). On what grounds what. In D.Manley, D.J. Chalmers, &
R. Wasserman (Eds.), Metametaphysics: New Essays on the Foundations of
Ontology (pp.47383). Oxford: Oxford University Press.
Schaffer, J. (2010). Monism: The priority of the whole. Philosophical Review,
119, 3176.
Schaffer, J. (2012). Grounding, transitivity, and contrastivity. In F.Correia &
B.Schnieder (Eds.), Metaphysical Grounding: Understanding the Structure of
Reality (pp.122138). Cambridge: Cambridge University Press.
Schaffer, J. (this volume). Ground Rules: Lessons from Wilson.
Schaffer, J. (2016). Grounding in the image of causation. Philosophical Studies,
173, 49100.

204

J. Wilson

Spirtes, P., Glymour, C., & Scheines, R. (1993). Causation, Prediction, and
Search. NewYork, NY: MIT Press.
Wilson, J.M. (1999). How superduper does a physicalist supervenience need to
be? The Philosophical Quarterly, 49, 3352.
Wilson, J. M. (2009). Determination, realization, and mental causation.
Philosophical Studies, 145, 149169.
Wilson, J. M. (2010). Non-reductive physicalism and degrees of freedom.
British Journal for the Philosophy of Science, 61, 279311.
Wilson, J. M. (2012). Fundamental determinables. Philosophers Imprint, 12,
117.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57, 145.
Wilson, A.(in progress). Metaphysical causation.
Woodward, J. (forthcoming). Causation in science. In P. Humphreys (Ed.), The
Oxford Handbook of the Philosophy of Science. Oxford: Oxford University
Press.
Yablo, S. (1992). Mental causation. The Philosophical Review, 101, 245280.

8
The Metaphysics ofNature, Science,
andtheRules ofEngagement
CarlGillett

The Scientific Revolution was powered, at least in large part, by explanations that pierced the manifest image of common sense by explaining its
level of everyday individuals, properties, and processes using qualitatively
distinct, lower-level entities taken to compose them. And such explanations have now been iterated through all the levels of nature. For example,
we take the corrosive action of glaciers to be explained by the movement
of the ice molecules that we take to compose glaciers. We explain the
motility of cells using the properties and relations of the molecules that we
take to compose them. We understand why kidneys clean blood in terms
of the properties and relations of the cells taken to compose them. And we
could easily go on, and on, through such explanations across the sciences.1

This paper makes explicit the methodology used in Gillett (2016) and overlaps with some of the
arguments offered in Chap. 2 of the latter.

C. Gillett ()
Northern Illinois University, DeKalb, IL, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_8

205

206

C. Gillett

Given their nature, such explanations are plausibly termed compositional


explanations, since they are founded around showing how lower-level entities of one kind (whether individuals, properties, or processes) compose entities of very different kinds at higher levels.2 Philosophers of science have used
a range of other terms for compositional explanation, including reductive
explanation, functional explanation, or mechanistic explanation, and
there is a substantial body of work on the nature of such explanation, including a recent burst of research.3 Oddly, however, a couple of the key features of
compositional explanations have not received much philosophical attention.
First, compositional explanations allow us to explain one kind of entity,
such as a cell or its moving, in terms of the qualitatively different kinds of
entity taken to compose it, like molecules or molecular processes of polymerization, and this hence results in what I term the Piercing Explanatory
Power, or PEP, of compositional explanations. Second, we should mark
that once we have successfully supplied a compositional explanation of
certain entities in terms of certain others that compose them, then we have
established that these entities are in some sense the same. Most importantly, a successful compositional explanation consequently shows that the
mass-energy, or force, associated with a certain entity just is the massenergy, or force, of certain component entities. This is what I will term the
Ontologically Unifying Power, or OUP, of compositional explanations.
Our vast array of compositional explanations in the sciences, from
fundamental physics to condensed matter physics or materials sciences,
on to chemistry or biochemistry, through cytology and physiology, and
now even beginning to encompass the neurosciences and psychology,

2
Analytic metaphysicians sometimes complain that I am using composition in a way different
from their usage, but it is important to emphasize that I am simply following a long-standing, and
continuing, scientific usage of the terms composition and compose that stretches back at least
to Sir Isaac Newton. Throughout the paper, unless I indicate otherwise, my usage of composition
therefore solely refers to the relations and concepts deployed in compositional explanations. If there
are still further concepts of composition used in the sciences (Healey 2013), then I do not discuss
them here.
3
Contemporary work on compositional explanation, including the species of it relating processes
in so-called mechanistic explanation, goes back at least to early work by Dennett (1961) and
Fodor (1968), through Wimsatt (1974) and Cummins (1983), down to more recent work such as
Bechtel and Richardson (1993), Glennan (1996), Machamer et al. (2000) and Craver (2007),
among many others.

The Metaphysics ofNature, Science, andtheRules...

207

have had intellectual impacts in all kinds of ways. For instance,


compositional explanations have been central to the centuries-old unification project in physics that has now established that there are no special
forces, or energies, and that the only fundamental forces and energies are
all microphysical in character. Connected to the later finding, and again
driven by compositional explanations, working scientists now routinely
assume the global claim that everything in nature is either identical to a
microphysical entity or is composed by microphysical entities. (I will call
this thesis physicalism here.)
Against this scientific background, if one accepts that successful explanations are likely to be true, then one will accept that compositional
explanations are true and hence illuminate the structure of nature. But
such explanations utilize compositional concepts, what I term scientific
composition, and hence offer the obvious place to look for insight about
what are variously termed vertical or making-up relations in nature
such as the partwhole relations of individuals or the compositional relations of properties or processes. For if compositional explanations are
true, then their compositional notions presumably limn the vertical structure of nature. We thus also get a compelling approach to understanding
what I dub here the metaphysics of nature: that is the character of the
vertical relations between entities in nature and any set of philosophical
issues involving such vertical relations, such as the formulation and truth
of physicalism, reduction versus emergence, and so on. 4
The latter is a methodological strategy not just for understanding
scientific conceptions of verticality, but also vertical relations in nature.
I call this methodology Engagement: engage our hugely successful compositional explanations in the sciences that form our core, and plausibly
even our sole, body of successful explanations about vertical relations in
nature both to understand scientific notions of composition and also the
compositional relations in nature itself. Let me briefly outline the three
broad phases of Engagement.

There is a whole lot more to the ontology of nature than what I am terming the metaphysics of
nature that is limited to issues about verticality, but framing this limited area will facilitate my
discussion here. Many of my points can plausibly be extended to other ontological issues ontological
issues about nature falling outside of the metaphysics of nature as I have defined it here.

208

C. Gillett

In the first step of Engagement, one pursues the project of looking at


concrete cases from the sciences to articulate and describe the features of
compositional explanations and their notions. Let me frame this phase thus:
(Step 1Descriptive Engagement) Examine successful compositional explanations in the sciences to illuminate the features of their compositional
concepts, the characteristics of such explanations themselves, and any associated evidence.

Building on the results of Step 1, we can then turn to the next phase
which is that of constructing theoretical accounts for scientific notions
of composition whose success is judged by how well they do at capturing
the features of such concepts, and the explanations using them, illuminated in Step 1. We can roughly frame this phase as follows:
(Step 2Constructive Engagement) Formulate theoretical accounts of scientific notions of composition, whether the partwhole relations of individuals, realization of property/relation instances, or implementation of
processes, to accommodate their features and other evidence outlined in
Step 1, where such frameworks are judged by how well they succeed or fail
at accommodating such characteristics illuminated in Step 1.

The results of Step 2 are accounts of the compositional concepts used so


successfully in compositional explanations, so the resulting frameworks
initially contribute to a narrow project in the philosophy of science. But
compositional explanations are successful and apparently the only set of
explanations about verticality in nature that have such a measure of success. Consequently, the results of Step 2 plausibly provide the basis of the
obvious approach to metaphysics of nature itself. We thus get this phase:
(Step 3Reflective Engagement) Pursue the theoretical issues involving the
metaphysics of nature by reflecting on the implications of compositional
explanations using the theoretical accounts of scientific composition produced in Step 2 as well as the descriptive findings outlined in Step 1.

It is tempting to label this last phase Philosophical Engagement,


but that would be misleading for two reasons. First, the other two

The Metaphysics ofNature, Science, andtheRules...

209

steps are plausibly philosophical. And, second, theoreticians other than


philosophers actively pursue Step 3. Thus, for instance, so-called scientific reductionists and emergentists, such as Steven Weinberg (1994),
Philip Anderson (1972) and Robert Laughlin (2007), are all actively
pursuing Engagement and are focused on Step 3 although they are not
philosophers.5 And the latter example also reminds us that one may use
the results of Step 2in various ways, so, for example, the scientific reductionist the scientific reductionist uses these accounts to argue there are no
compositional relations in nature itself.
One might think Engagement would be widely pursued by philosophers as an approach to understanding either scientific notions of composition or the metaphysics of nature. However, one of my initial goals in
this chapter, in Part 1, is to document how all the prominent approaches
to scientific composition, and/or to verticality in nature, from the philosophy of science, philosophy of mind, and analytic metaphysics presently
diverge from Engagement to varying degrees by following what I term an
Appropriational strategythat is, rather than pursuing Steps 1 and 2,
these accounts take machinery developed for other phenomena and shoehorn scientific composition into these alien frameworks.
Even accepting the latter situation, one might contend there is no
problem because one or other of the Appropriational approaches in
philosophy still captures the nature of scientific composition. However,
another goal of my chapter is to show that the dominant accounts of
vertical relations in philosophy of science, philosophy of mind, and analytic metaphysics all provide inadequate treatments of scientific composition. To establish this conclusion, I start by briefly looking at a couple
of examples of compositional explanation in Part 2, and highlight some
of the key features of their compositional notions in Part 3. In particular, I outline features of scientific composition underpinning both the

Scientific reductionists argue that reflection on such explanations shows there are no composed
entities or compositional entities in nature, since the scientific reductionist argues that the truthmakers for compositional explanations are very different in character. So a wide array of positions
on metaphysics of nature can be, and are, adopted in Step 3.

210

C. Gillett

distinctive PEP and OUP of compositional explanations. Building on


this work, in Part 4, I then show that all the prominent philosophical approaches fail to accommodate key features of scientific notions
of composition and explanations, including those characteristics plausibly underpinning either PEP or OUP.I thus establish that pursuing an
Unengaged approach leads to failure both in understanding scientific
notions, and also plausibly in illuminating the metaphysics of nature. In
contrast, in Part 5, I highlight a minority approach to scientific composition that does follow the rules of Engagement and promises to provide a
fully adequate account.
Initially, one might not be too bothered that prominent philosophical
accounts of verticality fail to provide adequate accounts of notions of verticality used in our spectacularly successful scientific explanations. Surely,
this is just a failing in a narrow descriptive project in one small area of
philosophy of science? However, as my earlier remarks highlight, these
findings are indeed important because we have a clear, and compelling,
meta-justification, briefly outlined above, for taking the vertical concepts
of compositional explanations to reflect the structure of nature. And we
are yet to be given a compelling meta-justification for why we should take
Unengaged accounts to reflect the metaphysics of nature. Consequently,
we have substantive reasons to think that because the prominent philosophical accounts of verticality provide inadequate accounts of scientific
composition, these frameworks also provide inaccurate accounts of the
structure of nature itself.
Overall, my final conclusion is that we need to obey the rules of
Engagement if we are to be successful in either understanding scientific
composition or the metaphysics of nature. The compositional relations
we find in science and nature are singular relations which must be understood on their own terms, rather than through the lens of machinery
developed for other phenomena. However popular in contemporary philosophy, Unengaged and Appropriational approaches that do not start
with our successful scientific explanations are plausibly doomed to failure. In contrast, sticking to the rules of Engagement, and relying upon
our core evidence about verticality in nature, promises a more successful
approach.

The Metaphysics ofNature, Science, andtheRules...

211

Part 1Appropriational Accounts in Analytic


Metaphysics, the Philosophy of Science, and
the Metaphysics of Mind
I look at prominent approaches to vertical relations from three different areas of philosophy that are increasingly used to provide accounts
of scientific composition. In section 1.1, I provide a brief overview of
work in analytic metaphysics on Grounding; in section 1.2, I outline
what I term neo-Causal treatments from philosophy of science of constitutive relations and explanations; and, in section 1.3, I survey standard functionalist frameworks from the philosophy of mind. Although
differing in various ways, I suggest all these views, when considered as
treatments of scientific composition, are Unengaged to varying degrees
because each of these positions does not construct its account of scientific
composition through the detailed examination of compositional explanations. How then are these various views constructed? I show that work
in all of these areas actually pursues the Appropriational strategy: each
account appropriates machinery developed for other phenomena.

1.1 Grounding Frameworks fromAnalytic


Metaphysics: AnUnengaged Approach
toall Verticality inNature andScience
There are a number of prominent proponents of Grounding who
develop it in various ways, including Kit Fine (2001, 2012), Gideon
Rosen (2010) and Jonathan Schaffer (2009, 2016), among others. The picture of verticality underlying Grounding frameworks has
been formulated by focusing primarily on the vertical relations
between semantic, logical, abstract, and/or mathematical entities
(orwhat I term SLAM entities).6 And using the various formal systems
6

I do not deny that on the lists of Grounding relations in some articles on Grounding we occasionally find what appear to be scientific examples. However, what I do deny is that we ever find a
detailed examination of such compositional explanations in the sciences informing accounts of
Grounding.

212

C. Gillett

covering SLAM entities and their relations, formal features have been
outlined to capture the resulting notion of Grounding and associated
concepts.
Given the latter approach, Grounding accounts of scientific composition are pretty clearly Appropriational and Unengaged because they
were not directly developed by looking at compositional explanations
in the sciences, but were developed for vertical relations in the abstract
realm and then extended to other vertical relations. This approach is
less surprising when one realizes that many defenders of Grounding
appear to accept a division among the types of explanation, and relations underpinning them, across science and metaphysics. Thus, we find
Fine telling us Ground, if you like, stands to philosophy as cause stands
to science (Fine 2012, p. 40). Furthermore, Fine elaborates on this
picture:
A number of philosophers have recently become receptive to the idea that,
in addition to scientific or causal explanation, there may be a distinctive
kind of metaphysical explanation, in which explanans and explanandum
are connected, not through some sort of causal mechanism, but through
some form of constitutive determination. (Fine 2012, p.38)

The resulting view is one that takes the sciences and their explanations
to always focus on relations of causation, while taking metaphysics (philosophy?) to focus on constitutive explanations based around
Grounding. It is therefore unsurprising that in formulating their accounts
of vertical relations, proponents of Grounding frameworks do not use
compositional explanations from the sciences because many proponents
of Grounding are unsure there are any such explanationsa stance we
see in the next section is also increasingly common in the philosophy of
science as well. However, proponents of Grounding take this relation to
be the vertical relation underpinning all constitutive explanations, thus
presumably including compositional explanation in the sciences.
Accounts of Grounding are not easy to articulate in detail, since
Grounding is taken by its proponents to be a primitive notion that is not

The Metaphysics ofNature, Science, andtheRules...

213

open to definition using other terms. And there are a range of differences
between proponents of Grounding over further issues such as the exact
relata of Grounding, whether Grounding is a worldly relation or not,
and more. However, what we can safely say is that Grounding has the
features of the relations between the SLAM entities used as exemplars of
such relations, including allowing causally inert entities as their relata,
and that Grounding is taken to have certain formal features captured by
logical systems.
To make our discussion less abstract, I focus on the work of Schaffer
who allows the relata of Grounding relations to include entities like
properties, individuals, and processeshence offering the most promising approach to the kinds of relation we will see in the next section
are posited in compositional explanations in the sciences. Schaffer takes
Grounding relations to be well-founded and to take the form of a partial
ordering. And Schaffer also explicitly states that Grounding relations do
not require causal entities as relata. In addition, Schaffer is like other
writers on Grounding in claiming his account of Grounding covers all
vertical relations wherever we find them, including those in the natural
world. More pertinently, Schaffer claims that Grounding is the relation
that backs constitutive explanations wherever one finds them (Schaffer
this volume)thus taking scientific composition, the relation that underpins or backs compositional explanation, to be identical to Grounding.
Furthermore, in another important development of his account Schaffer
(2016, this volume) claims that we should favor an account of vertical
relations that fits the best formalism, that structural equation modeling drawn from the sciences is the best formalism for verticality in various explanations, and that such structural equation modeling best fits his
Grounding account.
In coming sections, I am going to assess whether Grounding, and
Schaffers account in particular, provides an adequate account of the
vertical relations underpinning compositional explanations in the
sciences. For, as we have seen, its proponents claim Grounding is
the vertical relation that backs all constitutive explanation wherever
we find it.

214

C. Gillett

1.2 Manipulability-Based andother Neo-Causal


Accounts fromPhilosophy ofScience: AnEngaged
but Appropriational Approach
Interestingly, the assumption that all scientific explanations are ultimately
causal, or involve relations of a kind with causation, is increasingly widespread in philosophy of science. For example, James Woodward (2014)
in his Stanford Encyclopedia of Philosophy entry on scientific explanation
explicitly wonders whether all explanations in the sciences are causal.
And the impetus for such a view about scientific explanation is easy to
find, since recent scientific techniques for illuminating causal relations,
including structural equation modeling, are spectacularly successful.
Against this background, the natural intellectual tendency to extrapolation leads to the conclusion that all scientific explanation is captured
by such techniques. Consequently, frameworks inspired by these scientific techniques, such as Woodwards own interventionist account
(Woodward 2003) using manipulability relations, have been extended to
increasingly greater numbers of explanations beyond the obviously causal
ones. One immediate question is about the mechanistic explanation
that has been the focus of so much recent attention in philosophy of
science. Mechanistic explanation appears to be both compositional and
non-causal (and I am counting it as species of compositional explanation
positing compositional relations between processes), so how does this fit
with such a Monistic stance about all scientific explanation?
However, rather than theoretical work on the nature of mechanistic
explanation providing a base of opposition to the claim that all scientific
explanation is causal, such work increasingly provides a manifestation of
this tendency. Much of the work actually seeking to provide a theoretical account of mechanistic explanation assumes either that the compositional notions in mechanistic explanations are really causal notions or
that composition is causation-like and captured by frameworks developed for causal relations with some tweaks. These are what I term neoCausal approaches to scientific composition.
Whether this work in the philosophy of science is plausibly Engaged is
a difficult question, since it is unclear how deeply or carefully such writers
pursue Step 1. Writers on mechanistic explanation are starting with real

The Metaphysics ofNature, Science, andtheRules...

215

cases of compositional explanation, but the neo-Causal approaches to scientific composition are again still Appropriational. Neo-Causal accounts
are plausibly not constructed following Steps 1 and 2 to build frameworks tailored to capture the nature of scientific composition after examining actual cases. Instead, the neo-Causal approaches take machinery
already developed for causation and then revise this machinery to putatively accommodate scientific composition, so this is plausibly another
Appropriational strategy.
I cannot sketch all of the neo-Causal approaches, so let me briefly note
a number of these accounts and then focus on the most prominent example of such a view. For instance, some versions of neo-Causalism take
existing accounts of causation, alter these accounts in various ways, and
produce frameworks for scientific composition. Jens Harbecke (2010,
2014a, b) uses regularity accounts of causation in this way and Mark
Couch (2011) takes Mackies INUS-based account to underpin his treatment. And writers like Totte Harinen (forthcoming) claim that scientific
composition relations are quite literally identical to causal relations.
However, the most prominent example of such an account is found in
the work by Carl Craver (2007) so I consider his framework in detail, but
the critical assessment I later provide of Cravers account plausibly carries over to the other neo-Causal approaches. Craver has popularized the
strategy of using interventionist approaches to causal explanation, and
their manipulability relations, as a way to characterize the compositional,
or as Craver terms them constitutive, relations between processes that
underpin compositional explanations. For my purposes, the interesting
feature of Cravers approach is that he is charitably interpreted as providing a sufficient condition for the composition of processes that is built
primarily around the mutual manipulability of processes.
Craver is quite explicit that his approach is an extension to compositional relations and explanations of a framework developed for causal
relations and explanations. And Craver takes this feature to be a merit of
his view. Craver tell us:
The mutual manipulability account is a plausible condition of constitutive
relevance because it fits well with experimental practice and because it is an
extension of the view of etiological [causal] relevance. (Craver 2007, p.162)

216

C. Gillett

Craver outlines the heart of his account of the composition of processes in this passage (where -ing and -ing are processes grounded
by the relevant individuals X and S):
My working account of constitutive [i.e. implementational] relevance is as
follows: a component is relevant to the behavior of a mechanism as a whole
when one can wiggle the behavior of the whole by wiggling the behavior of
the component and one can wiggle the behavior of the component by wiggling the behavior as a whole. The two are related as part and whole and
they are mutually manipulable. More formally: (i) X is a part of S; (ii) in the
condition relevant to the request for explanation there is some change to
Xs -ing that changes Ss -ing; and (iii) in the condition relevant to the
request for explanation there is some change to Ss -ing that changes Xs
-ing. (Craver 2007, pp.1523. Original emphasis)

The first condition demands partwhole relations between the individuals that ground the composed and composing processes, that is, between
the - and -ing individuals. The other two conditions require a relation of mutual manipulability. It is important to emphasize that Craver
deploys the sophisticated machinery of interventionism to articulate
these core ideas and that Craver also adds a number of further nuances to
his final account in order to address phenomena he highlights in scientific practice.7 However, I am going to leave these nuances to one side in
my discussion, since it is the adequacy of the core mutual manipulability
conditions that I am interested in.8 The basic idea is that mutual manipulability, suitably qualified, suffices for the composition of processes in the
sciences.
In this case, Craver is not importing a metaphysical framework developed outside the sciences in order to characterize scientific phenomena.
Nonetheless, Cravers, and the general neo-Causalist, strategy is

For example, Craver adapts his conditions so they take account of redundancy. See Chap. 4 of
Craver (2007), and particularly section 8, for the details.
8
The type of objections to Cravers account that I offer in the next chapter also apply to Cravers
more sophisticated, amended account.

The Metaphysics ofNature, Science, andtheRules...

217

Appropriation, for these approaches again seek to fit compositional


relations into machinery developed for other phenomena, in this case
causal relations. As we saw above, however, Craver is quite explicit in this
approach and one can thus charitably take Craver as staking out neoCausalism as a philosophical position, hence taking causation and composition to be of a kind. In the coming sections, I again assess whether
the wagers of neo-Causalism pay off.

1.3 The Standard Functionalist Machinery


ofPhilosophy ofMind: AnAppropriational
Approach totheComposition ofProperties
intheSciences
As anyone familiar with philosophy of mind will know, there has been
a large body of work, spanning many decades, on functionalism and
related treatments of the realization between properties. This work
clearly concerns verticality in nature. Furthermore, increasing numbers
of writers who work on scientific composition, connected issues in the
philosophy of science such as reduction/emergence, as well as on metaphysics of nature, continue to use an account from this area in what I
term the standard functionalist framework. Does this body of work in
naturalistic philosophy provide the type of Engaged accounts I have
suggested it is natural to pursue?
Many of the proponents of standard functionalism appear to think
that it does and hotly defend such a position. But once we explicitly
outline Steps 13 of Engagement, this happy picture about the healthy
state of naturalistic metaphysics actually starts to look implausible
and the status quo looks more sickly. I suggest there are plausible reasons to conclude that philosophers in this tradition have rarely been
Engaged in the ways framed in Step 1that is, by actually looking carefully at real compositional explanations from the science in constructing
their frameworks. Furthermore, it appears that standard functionalist
accounts were developed using Appropriationmachinery developed
for very different projects was then applied to scientific composition.
Let me briefly offer my take on the tangled genesis of functionalism to
support these claims.

218

C. Gillett

Philosophy of mind and philosophy of psychology came into existence


in their present form during the 1960s, 70s, and 80s which saw an amazing spasm of creative work on a range of foundational questions in the
philosophy of mind/psychology, the philosophy of science, and what was
viewed as naturalistic metaphysics. One of the products of this activity were the versions of functionalism, and the proprietary technical
machinery associated with them, which are widely, but mistakenly, taken
to have a shared metaphysical basis. For all the versions all the versions
of functionalism talk of functional properties, causal roles, realization, and so on. Unfortunately, the different versions of functionalism were crafted to serve different purposes and have distinct proprietary
understandings of functional properties, causal roles, realization,
and so on. The result is a very confused, and ultimately damaging, situation in naturalistic philosophy. Drawing out the nuances of these problems requires more space than I can devote to it here, but I can outline
the genesis of this situation and highlight reasons to be suspicious that
standard functionalist machinery does not comfortably reconstruct scientific notions of composition.9
One tradition of functionalism, including writers like Fodor (1968)
and Dennett (1969, 1978), was focused on using scientific examples of
compositional explanations to provide a model for work in cognitive science or computational psychology. In contrast, another group of writers on functionalism, like David Lewis (1972), hailed from analytic
areas of philosophy and focused on commonsense concepts using more
traditional philosophical tools. Work in these traditions was seldom kept
separate despite their distinct goals and methodologies.
The empirically oriented work on functionalism was directly and
explicitly focused on concrete compositional explanations and their
compositional notions. For example, focusing on such concrete examples, Dennett emphasized that compositional relations in the sciences
are often manyone relations involving teams of component individuals
and properties that together compose some higher-level individual or
property. Dennett explicitly accepted that the composed and component

I provide a detailed overview of these problems in Gillett (2007c, 2013b).

The Metaphysics ofNature, Science, andtheRules...

219

individuals/properties are at different levels of individuals and qualitatively


distincthence accommodating the PEP of the associated explanations.10
And a similar view of scientific composition plausibly inspired Fodors
positive account of the structure of psychology and other higher sciences,
including their distinctive realization relations, as well as his critique of
Nagelian reduction using so-called multiple realization.11
Unfortunately, in what one may in hindsight regard as one of the great
tragedies of naturalistic philosophy, pioneering writers like Fodor and
Dennett working simultaneously in the philosophy of science and the
philosophy of psychology, never constructed a theoretical framework for
scientific notions of composition used in compositional explanations
such as the realization of properties. And the resulting theoretical vacuum was inevitably filled. The machinery of topic-neutral Ramseyfication
developed in the analytic tradition of functionalism by Lewis for understanding semantic relations of predicates was unfortunately co-opted to
frame ontological notions from the sciences that were the focus of the
empirical traditions like that of a realization relations between property instances. Matters were made still more complicated by the still later
alterations made to Lewiss machinery inspired by the singular features of
computational explanations in psychology and other influences.12
These appropriations of Lewiss machinery plausibly spawned a collection of new technical concepts, like that of a second order property,
now widely used by philosophers. And the new conceptions built around
topic-neutral Ramseyfication, whether of a functional property, realization, causal role, and so on, were taken to cover all the versions
of functionalism and their proprietary versions of the latter notions,
though I suggest below that it is dubious whether they fit the notions of
the empirically oriented tradition of functionalism.

10

Dennett utilized these distinctive features in articulating his influential methodology for a scientific psychology. Consider Dennetts example of how we explain a capacity for face recognition as
composed by a team of lower-level entities. Famously, in such cases, Dennett clarifies that we can
understand richly intentional entities as composed by teams of entities with more rudimentary, and
hence qualitatively distinct, intentionality, which in turn can be understood as composed by teams
of entities with still more rudimentary intentionality.
11
Fodor (1968, 1974).
12
See Piccinini (2004) for a survey of these developments and their results.

220

C. Gillett

What I am terming the standard functionalist framework was developed


by writers working within these appropriations of Lewiss machinery. Most
prominently, Jaegwon (1992, 1998), Jessica Wilson (1999, 2009), Larry
Shapiro (2004) and Sydney Shoemaker (2001) have worked within this
framework to shape an ontological account of the compositional relation
between property/relation instances that is termed the Flat or Subset
view of realization. The Subset/Flat view takes realization to be a oneone
relation between qualitatively identical or qualitatively similar properties of
the same individual.13 And this view of realization is now routinely used
in work in the philosophy of science to articulate scientific notions of
composition involving properties, or the issues involving such notions like
Kims (1999) reduction-by-functionalization account, Shapiros (2000)
account of multiple realization in the sciences, or Wilsons (2010, forthcoming) treatment of emergence and related issues.
Given its genesis, I take the Subset/Flat view of realization, and the
other accounts of standard functionalism, to be Unengaged and the
results of Appropriation. I know many take this to be highly contentious,
but the key point is that standard functionalist machinery was plausibly
not produced by engaging concrete cases of compositional explanation.
And, whatever one thinks about the genesis of Subset/Flat view, we can
now see that there are real worries about whether the Flat/Subset covers the notions developed in the empirical tradition of functionalism
of Fodor and Dennett or scientific notions of compositions themselves.
Such problems have plausibly been masked by the convoluted history of
functionalism. But recall that Dennett highlights how the compositional relations he finds in scientific explanations are onemany relations
of teams of qualitatively different realizer properties instantiated in distinct individuals from the higher-level individual instantiating the realized property. In contrast, we have just seen that the Subset/Flat account
takes realization to be a oneone relation between qualitatively similar
properties of the same individual.
The latter points are important because whatever one concludes about
whether standard functionalism was constructed by following Step 1 of
Engagement, many of the proponents of the Subset/Flat view intend it as an
13

See Gillett (2002, 2003, 2010) for an outline, and critical evaluation, of the basic features of the
Subset/Flat view of realization. The latter critique is extended below.

The Metaphysics ofNature, Science, andtheRules...

221

account of the scientific notion of composition between property/relation


instances in compositional explanations in the sciences or, at the least, as an
account of some of the vertical relations between entities in nature. In coming sections, I therefore also assess the Flat/Subset views merits as an account
of scientific composition and hence as a basis for the metaphysics of nature.

1.4 Assessing theAppropriational Accounts


The frameworks we have surveyed all follow the path of Appropriation
and use machinery developed for other purposes to capture scientific composition. A looming question is whether it matters that these
accounts bypass Engagement and the combination of Steps 1 and 2? To
answer this question, we need to assess the success of Appropriational
frameworks as accounts of scientific composition. In the next section, I
briefly outline a couple of compositional explanations to guide our discussion. Then, in the following section, Part 3, I draw out some of the
features of the compositional concepts we find in such cases. Finally, in
Part 4, I examine how well the various Appropriational views actually do
as accounts of scientific composition.

Part 2Compositional Explanation in the


Sciences: Explaining Cellular Motility and
Dendritic Spine Growth at the Molecular Level
As our target, I am going to use our molecular explanations of the movement, and dendritic spine growth, in neurons at the cellular level. Many
young neurons crawl from the place of their birth to their final locations in
the brain. And, like all cell movement, the movement of neurons is based
around a number of steps: first, the cell protrudes from its surface, then,
second, the cell attaches at the protruded surface leaving itself stretched
and in tension, given its attachment at the back of the cell; and, third, the
cell releases the attachment at the back of the cell and the tension within
itself makes the cell contract to pull the cell forward to the new point of
attachment in the protrusion. At this point, the cycle begins again and in

222

C. Gillett

this way the neuron slowly moves. Given its complexity, I am going to
focus solely on the first step of the neurons movement in its protrusion.
The basis for our molecular explanation of cellular movement is complex, but at its core is the behavior of the actin molecules found in the cell
and particularly its cytoskeleton. Narrowing down just to focus onto our
molecular explanation of cellular protrusion, the crucial molecular processes are easily outlined. Crucially, the cell is filled with monomers of
globular actin (G actin) in the form of unchained actin molecules. One
important feature of actin is that it can polymerize swiftly in long filaments (F actin). And this is what we find in neuronal movement. When
stimulated in a certain direction, many filaments of actin are all formed
within the neuron pushing the molecules composing the cell membrane in
a certain direction, given the rigidity of the molecules in the cells cytoskeleton, until other molecules attach the protrusion to the surface. We consequently explain the protrusion of the cell in large part using these directed
polymerizations of monomers of G actin into many filaments of actin that
press on the membrane in the direction in which the cell is travelling.
Notice that in this case scientists take actin and other molecules to be
parts or constituents of the neuron. Furthermore, we have compositional explanations of the property of motility of the cell using properties and relations of the constituent molecules taken to compose this
property instance. And we have a compositional explanation of the cells
protruding using the molecular processes of polymerization, and other
processes, taken to compose this cellular process. So we not only have
compositional relations posited between individuals, or partwhole
relations, but also compositional relations between property instances
in realization relations, and between processes in relations of implementation. Table 8.1 summarizes these compositional relations and my
favored terms for them. There are all manner of interesting characteristics
of such compositional relations that I note in the next section, but let me
briefly outline a related group of compositional explanations.
The same molecular phenomena are also used to explain a property, and
associated process, of adult neurons that can swiftly grow dendritic spines
in the direction of an electrical stimulus. Our compositional explanation,
at the molecular level, of how the neuron grows dendritic spines, and so
quickly, is based upon the same components. The neuron has receptors

The Metaphysics ofNature, Science, andtheRules...

223

Table 8.1 Table of the various kinds of compositional relation, the category of
entity they relate, and my terms for them.
Relata

Compositional Relation

Processes

Lower-level processes together implement a


higher-level process
Lower-level individuals together constitute a
higher-level individual
Lower-level properties together realize a
higher-level property
Lower-level powers together comprise a
higher-level power

Individuals
Properties
Powers

in its membrane that react to electrical stimulation and send messenger


proteins that stimulate monomers of actin to polymerize into interconnected rods of F actin in the direction of the stimulus. Given the rigidity
of the neurons cytoskeleton, these polymers of actin together push out
the membrane of the neuron to form the new dendritic spine and hence
implement the neurons swiftly growing a new dendrite.
Notice that in this example we again have a higher-level process based by
the constituted individual, the neuron, in growing a dendritic spine. Again,
the process based by the constituted individual is composed by, that is, is
implemented by, a range of processes based by the neurons constituents,
including filaments of actin. Crucially, these component individuals once
more form a team of inter-related individuals, in filaments of actin, and many
other proteins, that are spatially contained within the constituted individual
in a certain neuron. Lastly, mark that the various compositional relations in
both of our examples hold only under certain background conditions.

Part 3Key Features of Scientific Composition


and Compositional Explanation
I could easily supply many similar examples of compositional explanation from other disciplines, and other pairs of levels of nature, but these
examples focused on the neuron are sufficient for my purposes here. The
richness and sophistication of the notions we find in such compositional
explanations poses serious presentational challenges, so let me simply

224

C. Gillett

supply a list of fifteen common features of the compositional relations


in our cases and examples like them that I have cataloged and defended
in detail elsewhere (Gillett 2016, Chap.2). I then examine a handful of
these features in more detail in order to guide my discussion here.
I contend that the vertical notions in compositional explanations in
the sciences are all such that they concern:
(i) relations that have working entities as relata, that is such relations
have as relata entities that are, at least in part, individuated by their
relation to processes associated with them;
(ii) non-productive determination relations that are synchronous,
between entities that are in some sense the same and which are not
identical either to the manifestation of powers or to the transfer of
energy and/or mediation of force;
(iii) mass-energy neutral relations, that is, their relata have mass-energy
but the overall mass-energy of the relata equals the mass-energy of
the entities on one side of the relation;
(iv) asymmetric relations;
(v) transitive relations;
(vi) irreflexive relations;
(vii) relations having qualitatively different relata;
(viii) manyone relations with teams of entities composing another
entity;
(ix) relations such that, under the conditions, components naturally
necessitate, that is, suffice for, the composed entity.
(x) relations allowing cases of multiple composition, such as the multiple realization of properties, multiple constitution of individuals,
and so on.
(xi) relations only holding under background conditions, where the
entities treated as background conditions for the compositional
relation do not have powers that comprise the powers of the composed entity and do not base processes implementing the processes
grounded by the composed entities;
(xii) relations always involving teams of individuals which spatially
overlap and bear constitution/parthood relations where the con-

The Metaphysics ofNature, Science, andtheRules...

225

stituent individuals bear spatio-temporal, powerful, and/or productive relations to each other and hence form collectives;
(xiii) relations always involving comprising powers;
(xiv) relations always involving realizing properties;
and:
(xv) relations always involving implementing processes or the potential
for them.
This list is somewhat overwhelming and is likely not exhaustive. But it
should really be no surprise that the compositional notions used in some
of our most successful scientific explanations should be complex, sophisticated, and nuanced.
Given space limitations, we cannot assess how well the three kinds of
account we surveyed in Part 1 do with all of these features. And it might
also reasonably be argued that we should only be focused upon explanatorily salient characteristics. So I propose to focus on a handful of key
features from this list in features (i)(iii) and (vii)(ix), since the latter
are important general characteristics of scientific composition and also
the features that plausibly underlie the PEP and OUP of compositional
explanations themselves. Let me therefore briefly outline characteristics
(i)(iii) and (vii)(ix), and their links to PEP and OUP.
To begin, it is important to note a quite simple, but also foundational,
feature of the entities that are the relata of scientific notions of composition framed in (i). The relata of scientific relations of composition are
what I term working entities. That is, the various kinds of compositional relations in the sciences all relate entities that are individuated, at
least partially, by the processes with which they are associated and hence
by what I term roles. As we shall see shortly, this simple point has wideranging implications.
Given its centrality to the nature of compositional explanations, it is
important to mark another foundational feature of scientific composition: under the conditions, components naturally necessitate the composed entitythat is, the components suffice for the composed entity
in the relevant circumstances. This feature of compositional relations,

226

C. Gillett

(ix) on the list, underpins their central role in compositional explanation


that plausibly operates by representing these ontological relations taken
to hold between entities in nature. When we have successfully identified
certain components under the relevant conditions, then we have identified lower-level entities in the explanans that non-productively suffice for,
and hence explain the existence of, the higher-level entity, under those
circumstances, which is the explanadum.
Let me turn to a couple features of scientific composition underpinning the OUP of compositional explanation. To start, we should note
that, as (ii) frames, vertical compositional relations are plausibly not
a species of the productive or causal determination that holds horizontally across time between wholly distinct entities. Productive determination is identical to the triggering and manifestation of powers, is
temporally extended, occurs between wholly distinct entities, and usually
involves the mediation of force and/or the transfer of energy. In contrast,
the relations posited in our compositional explanations are not identical
to the triggering and manifestation of powers (although implementation
has as relata processes that are identical to such manifestations), are all
synchronous, occur between entities that are in some sense the same and
do not involve the mediation of force and/or the transfer of energy.
By itself, this characteristic of compositional relations of having relata
that are in some sense the same provides some ontological unification,
but if we dig into why composition does not involve the transfer of
energy, or mediation of force, then we can better understand how they
underlie Ontological Unifying Power. Having relata that are in some
sense the same is a slippery characteristic, but luckily the sciences concretize this feature, and others, in the connected feature (iii) that I term
the mass-energy neutrality of compositional relations. I take a relation to
be mass-energy neutral when its relata have mass-energy, but the overall
mass-energy of the relata equals the combined mass-energy of the entities on just one side of the relation. As we can see in our examples, the
mass-energy of a composed individual, in the neuron, and its properties/
relations, and/or processes, is not additional to, or subtractive from, the
mass-energy of the relevant component individuals, in the various constituent proteins, and their properties/relations, and/or processes. Thus,
for example, the mass-energy of the cell just is the mass-energy of its parts
and their relations/properties.

The Metaphysics ofNature, Science, andtheRules...

227

Mass-energy neutrality is a singular feature of scientific composition.


For example, in the absence of composition, when a new entity moves into,
or comes into being within, a region occupied by certain other entities,
then the mass-energy within this region is equal to the mass-energy of the
initial entities combined with the mass-energy of the new entity. However,
when a new entity is brought into being within a region of spacetime
through a compositional relation, then the mass-energy within the region
simply equals the mass-energy of the component entities alone. Roughly
put, the mass-energy of the composed entity just is the mass-energy of its
components because these entities are not wholly distinct.
We can thus see how compositional explanations have the Ontological
Unifying Power that has made them so important in the sciences, especially with regard to the unification program in physics. Once we show
that, for instance, the mass-energy of biological entities just is the massenergy of their molecular components, then we have established that
there are no special biological energies. So the mass-energy neutrality of
scientific composition, in feature (iii), is central, along with (ii), in underpinning the OUP of compositional explanation.
Next let us now turn to the features of scientific composition linked to
the PEP of compositional explanation. Mark that the entities bearing compositional relations in our case, and the many like it in the sciences, are usually qualitatively distinct and hence have feature (vii). For example, it is very
common for higher-level individuals to have novel properties and powers
not had by any of their components. For example, the neuron has the property of growing dendritic spines, but molecules do not, and so on. That the
vertical relations posited in compositional explanations have feature (vii)
appears to be a necessary condition of such explanations having PEP.
It may initially be puzzling how entities of one kind could compose
entities of very different kinds and hence how we can compositionally
explain one kind of entity in terms of qualitatively distinct components.
So it is important to note a final, apparently connected, feature of scientific composition. As Dennett emphasized in earlier discussions, composition is a joint affair where teams of many individuals, properties,
powers, and processes studied by lower-level sciences together compose
the qualitatively different individuals, properties, powers, and processes
studied by higher-level sciences. So we have many polymerizing actin filaments, and other molecular processes, implementing the cells protruding

228

C. Gillett

or growing a dendritic spine. As my examples thus highlight, and as (viii)


captures, compositional relations in the sciences are thus usually manyone with many components and one composed entity. Although each
component is qualitatively distinct from the composed entity, the deeper
point is that numerous component entities as a team jointly and nonproductively result in the qualitatively different entity they are taken to
compose. It thus appears that feature (viii) of scientific composition, in
its manyone character, as well as feature (vii), each play a central role in
underpinning the distinctive PEP of compositional explanations.

Part 4Illuminating the Failure of the


Dominant Philosophical Approaches with
Scientific Composition
Our work in Parts 2 and 3 allows us to assess the success of the Appropriational
accounts of scientific composition we surveyed from analytic metaphysics, philosophy of science, and philosophy of mind. Remember that such
accounts succeed or fail, in large measure, depending on how well they
accommodate the features of scientific composition and compositional
explanation. So, focusing solely on the six features (i)(iii) and (vii)(ix)
highlighted in the last section, I look at how well Cravers manipulability account, the Subset/Flat view of realization, and Schaffers Grounding
account, each does in accommodating these characteristics. I am going to
argue that these Appropriational accounts fail to accommodate one or more
of these features. Consequently, I also document how each account fails to
accommodate either the PEP, or OUP, of compositional explanations.

4.1 Why Manipulability (or Regularity or Counterfactual Dependence or Supervenience or


Sufficiency) and other Neo-Causal Accounts Fail:
Vitalism, Mass-Energy Neutrality, and OUP
Cravers manipulability account can be interpreted as seeking to provide
a sufficient condition for the compositional relations between processes.
To start the assessment of this account I am going to focus on a type

The Metaphysics ofNature, Science, andtheRules...

229

of hypothesis from historical debates in the sciences that is a paradigm


of a position rejecting composition and denying the existence of compositional, including mechanistic, explanation. In debates in nineteenthcentury biology, there had been long-standing problems in providing
compositional accounts of biological entities. Consequently, one of the
prominent positions of this period was that of so-called Vitalists of
different kinds who offered hypotheses under which various categories of
biological entities were uncomposed by cellular and molecular entities,
but under which such uncomposed biological entities were still nomologically connected to these cellular and molecular entities contained
within them. It bears emphasis that in their time, Vitalist theories were
mainstream scientific theories held by prominent scientists.
There were a wide variety of differing Vitalist accounts, but for my
purpose of critically engaging Cravers manipulability account let me
stipulate the following as our Vitalist hypothesis. This Vitalist account
takes organisms or organs to be biological individuals that are composed
by cellular or molecular individuals that are spatially contained within
these biological individuals. To this end, we may assume that these biological individuals have properties, for example their rigidity or hardness,
realized by the properties/relations of cellular or molecular individuals.
And we may further assume that processes based by the biological individuals realized properties are implemented by processes based by the
properties of the cells or molecules serving as realizers.
However, our Vitalist hypothesis also assumes that there are some biological properties, like be able to digest or metabolize, of organs or organisms that are unrealized and hence uncomposed. Call these uncomposed
biological properties B-properties. And let us take the B-properties
of organs or organisms to base what we may term B-processes of the
organs of the organisms, such as digesting or metabolizing. Our Vitalist
also takes B-processes to be uncomposed, and hence unimplemented,
by processes grounded by the properties and relations of cells or moleculesalthough the latter individuals and processes are spatially contained within the former individuals and processes. Lastly, this Vitalist
hypothesis assumes that there are brute laws of nature such that the
B-properties and B-processes come into existence with, and only exist
alongside, certain cellular and molecular properties and processes of
organs or organisms. Furthermore, the Vitalist hypothesis also assumes

230

C. Gillett

that there are brute laws of nature that when there are changes in the
B-properties and B-processes of organs or organisms, then there are
changes in the properties and processes of the cellular or molecular individuals contained within the relevant organ or organism, and vice versa,
even though the B-properties and B-processes are not composed by these
cellular and molecular properties and processes.
Although rather elaborate, this kind of Vitalist hypothesis has the
virtue of highlighting important concerns about Cravers or other
manipulability-based accounts of scientific composition. Notice that
under the Vitalist hypothesis, cells and molecules plausibly are parts of
the organ and organism under a spatial containment account of parthood that Craver apparently favors. So Cravers first clause demanding
partwhole relations between the relevant individuals involved in the
processes is satisfied. And so too are Cravers other conditions demanding
mutual manipulability. Given the brute laws of nature at play, changing
the properties and processes of the spatially contained cells or molecules
results in changes in the B-properties and B-processes of the organ or
organism, and vice versa.
Given the latter points, Cravers manipulability-based sufficient condition for the composition of processes is satisfied in the scenario outlined by
the Vitalist hypothesis and counts the B-processes as composed, and implemented, by the processes of molecules or cells. The obvious problem is that
the Vitalist hypothesis frames a paradigm example where we lack any such
scientific composition of processes, so Cravers manipulability criterion
offers a mistaken account of the composition of processes in the sciences.
Simply offering counter-examples to an account is rarely fully satisfying or convincing, but our earlier work also illuminates the deeper
flaws of the manipulability account that lead to these mistaken attributions. As I highlighted earlier, scientific composition involves relata that
are in some sense the same, noted in (ii), where such relations are also
mass-energy neutral as (iii) frames. But a manipulability-based condition can be satisfied by entities that are wholly distinct and fails to guarantee that we have relations that are mass-energy neutral. This is why
Cravers manipulability-based condition implies there is a compositional
relation between the B-processes, and the processes of the cells or molecules, which are wholly distinct processesthus failing to accommodate

The Metaphysics ofNature, Science, andtheRules...

231

feature (ii). In this case, the B-property is an uncomposed property with


its own fundamental biological energy that underpins the uncomposed
B-process, so we do not have a mass-energy neutral relation. The latter feature of the broached hypothesis was why so many Vitalists explicitly endorsed fundamental, and hence uncomposed, vital energies and/
or forces. So Cravers account also fails to accommodate feature (iii) of
scientific composition.
Given the latter findings, we can also see that the manipulability relations underpinning Cravers neo-Causal account do not suffice for the
Ontological Unifying Power of compositional explanations. Having
mutual manipulability relations just does not entail that we have entities
that are in some sense the same and whose mass-energy, or force, is the
same. The manipulability account also fails to accommodate an important
feature of compositional explanation in its Ontological Unifying Power.
Cravers manipulability-based account thus also fails as a criterion
for, or positive account of, compositional relations in the sciences. And
Vitalist hypotheses of various configurations also plausibly establish,
for similar reasons, the inadequacy of accounts of compositional relations in the sciences that either take them to be identical to, or have a
sufficient condition for their existence framed in terms of, relations of
causation (under various philosophical treatments of its nature), regularity, counter-factual dependence, sufficiency, or supervenience. Each of
these accounts can also be satisfied by relations holding between entities
that are wholly distinct where the relations between these entities are not
mass-energy neutral. Consequently, these kinds of neo-Causal, and other,
accounts of scientific composition also fail to accommodate key features
of scientific composition and hence also the Ontological Unifying Power
of compositional explanation.
One lesson from the failure of these accounts is that appropriating
frameworks for causal relations and trying to shoehorn scientific composition into such machinery, even with revisions, fail to accommodate the
singular features of compositional relations in the sciences. And we have
seen such failings have wider ramifications, since they leave it a mystery
how compositional explanation has its distinctive Ontological Unifying
Power and has so successfully driven the program of unifying the forces
and energies we find in nature.

232

C. Gillett

A second lesson is that although manipulability, regularity, counterfactual dependence, sufficiency, or supervenience relations may underpin
methods for discovering, or illuminating, the existence of specific relations
of scientific composition, it is a mistake to take compositional relations
in the sciences to be relations of causation, manipulability, regularity,
counter-factual dependence, sufficiency, or supervenience, or to take
such relations to suffice for scientific composition. Scientific composition
is a singular, and ontologically richer, and more complex, relation than
the latter relations. Unsurprisingly, appropriating frameworks for these
different relations consequently, fails to provide an adequate account of
scientific composition or compositional explanation.

4.2 Why Grounding Accounts Fail: Working Parts,


Mass-Energy Neutrality, andOUP
Let us turn to Grounding frameworks with which we can now immediately discern an issue. We have just seen that bare sufficiency does not
suffice for scientific composition and that this hence undercuts accounts
that identify scientific composition relations and bare sufficiency relations. Scientific relations of composition are relations of natural necessitation, and hence sufficiency relations, but they are a distinctive kind
of sufficiency relation among working entities such that certain of these
entities suffice for others where these entities are not wholly distinct and
their mass-energy, or force, is in some sense the same. Given these points,
Grounding accounts face difficulties, since similar criticisms to those
offered against neo-Causal accounts establish that the vertical relation
underpinning compositional explanation, in scientific composition, is
not Grounding.
Although Schaffers Grounding account, and others, offers a general
framework apparently applicable to composition between any category
of entity, for illustrative purposes I limit my discussion to Grounding
frameworks applied to the concepts of composition between individuals, that is, the partwhole relations between individuals, posited in the
sciences.

The Metaphysics ofNature, Science, andtheRules...

233

First, I should note that apparently reflecting their development for


productively inert and unlocated SLAM entities, relations of Grounding
do not require working entities as their relata. So we immediately have
a feature of scientific composition, in (i), that is not captured by a
Grounding account. Initially, this might seem like a small problem, but
this foundational difficulty manifests itself in other problems as well.
Second, Grounding accounts of scientific composition have a problem with mass-energy neutrality and feature (iii). To be a Grounding
relation it is not required to be a mass-energy relation, nor is it required
that the mass-energy, or force, of the Grounded entity be, in some sense,
the mass-energy, or force, of the Grounding entity or entities. However,
scientific composition relations are always mass-energy neutral relations
and Grounding accounts of scientific composition hence fail to accommodate feature (iii). So we see a second reason why scientific composition
is not Grounding. And notice this is regardless of whether Grounding is
a general relation, rather than a more ultimate species of verticality, for
however we consider Grounding it lacks one of the features had by the
relation underpinning compositional explanation.
Unsurprisingly, Grounding accounts of scientific composition also
give the wrong answers about whether we have scientific composition in
the kinds of cases broached by Vitalists. This time consider a case where
we have the same scenario we sketched for Cravers account, but where
it is also the case that the biological individual is taken to be an uncomposed individual in an entelechy that by brute laws of nature comes
into existence with, and only exists alongside, certain inter-related cells
or individuals that it spatially contains. And where the same brute laws
outlined earlier also hold. As we have seen, in such cases we would
have dependences between entities, whether individuals or properties/
processes, and hence a partial ordering, so we do have a situation that
suffices for Grounding relations of the kind that Schaffer posits between
the entelechy and molecules or cells. So in such Vitalist cases, Grounding
accounts of scientific composition take the biological individual, in the
entelechy, to be composed by the cells or moleculesbut all sides agree
this is not the case. So we have counter-examples to the Grounding
account accompanying its failures to accommodate key features of sci-

234

C. Gillett

entific composition in (i) and (iii).14 In addition, the Grounding account


of scientific composition also plausibly fail to accommodate the OUP of
compositional explanations because we have now seen that having entities related by Grounding, as in the Vitalist scenario, does not unify these
entities in the ways we have seen successful compositional explanations
unify the entities in their explanans and explanadum.
It is again worth noting that the latter problems arise whether, or not,
we take Grounding to be a general relation accompanied by more absolute species of vertical relation. For the claim of a Grounding theorist like
Schaffer is that Grounding is the relation underpinning all constitutive
explanation including compositional explanation. But we have found that
Grounding fails to have the right features to be this relation in the case
of compositional explanation. It may still be that Grounding relations
accompany the relations that do underpin compositional explanations,
but such relations are not identical to Grounding relations. Furthermore,
it is not in virtue of the properties they have by falling under some general relation of Grounding that allows the relation to underpin, for example, the OUP of compositional explanation. Scientific composition is its
own singular relation that relates working entities that are not wholly
distinct and where the latter relations are mass-energy neutral relations.
But Grounding requires none of the latter features and we again see that
the kind of determination (or sufficiency) relation posited to capture the
nature of scientific composition matters a lot. Appropriating an account
developed for the vertical relations of inert, unchanging, massless, unlocated SLAM entities unsurprisingly fails to capture the character of the
vertical relations of the productive, dynamic, located entities with massenergy that we find in nature.
What about Schaffers recent suggestion that structural equation modeling favors his Grounding account? I cannot engage Schaffers rich and
interesting suggestion in anything like the detail it deserves, but let me
make some brief points connected to our foregoing discussion.
14

I suspect that Grounding accounts also fail to cover feature (ii) and and do not entail entities that
are in some sense the same as their relata. However, I do not rely on this objection since I am unsure
about the commitments of Grounding accounts with regard to this kind of characteristic of their
relata.

The Metaphysics ofNature, Science, andtheRules...

235

First, it bears emphasis that structural equation modeling is in the


same family of techniques as the interventionist framework used by
Cravers neo-Causal account. And to my knowledge, such techniques
are used in the sciences almost exclusively to illuminate horizontal causal
relations and not vertical ones. Second, we should also note that there
is ongoing controversy in the philosophy of science about whether such
techniques can be successfully adapted to the vertical relations in nature
from their home in application to horizontal causal relations.15 However,
I should mark that such debates have largely focused on the conditions of
Woodwards (2003) particular framework, whereas Schaffer is focused on
his own treatment of structural equation modeling. So perhaps the recent
disputes do not apply to Schaffers approach, although this remains to be
seen. However, third, and more importantly, I outlined the problems
that Vitalist hypotheses pose for the related interventionist framework
Craver sought to apply to scientific composition. And I highlighted reasons to believe that the underlying manipulability relations, that appear
to underpin the whole family of techniques including structural equation modeling, simply do not on their own suffice for scientific composition. Similar concerns therefore appear to apply to the claim that the
related approach of structural equation modeling suffices for scientific
composition. I have thus already given prima facie plausible reasons to
dispute Schaffers underlying claims that formal techniques like structural equation modeling capture scientific composition or vertical relations in nature. But I should again emphasize that such a conclusion is
only an initial pass and more careful consideration needs to be given to
confirm its truth.
Once we become Engaged, and highlight the detailed features of the
relations of composition posited in the sciences, we consequently see that
a second Appropriational account of scientific composition fails. My conclusion in this section is that Grounding is not the relation that underpins
compositional explanation. Grounding accounts of scientific composition, adapted from work in analytic metaphysics, provide flawed accounts
of scientific concepts of componency in the sciences because they fail to
15

See Baumgartner (2010). For a response, see Woodward (2015).

236

C. Gillett

accommodate the working nature of parts, and other components in the


sciences, and hence also fail to capture a range of related characteristics of
scientific composition, including its mass-energy neutrality. As a result,
Grounding accounts also plausibly fail to accommodate the distinctive
Ontological Unifying Power of compositional explanation.

4.3 Why Standard functionalist Accounts Fail: RolePlaying, Qualitative Distinctness, andaLack ofPEP
In marked contrast to the previous two accounts, standard functionalist
treatments from the philosophy of mind appear to offer more promising treatments of features (i)(iii) because they have the great virtue of
being role-based. Let me briefly highlight how standard functionalist
accounts successfully accommodate the troublesome features of scientific
composition that trip up both neo-Causal and Grounding accounts.
Recall that the Flat/Subset view of the realization of properties is based
upon what I termed role-playing in a oneone relation between property
instances that are qualitatively the same, or similar, in either matching
or overlapping in their contribution of powers under some condition.16
Notice, first, that standard functionalist accounts take properties to be
individuated by their contributions of powers which are entities that
when triggered manifest in certain processes. So the Flat/Subset view
plausibly relates working entities and so satisfies (i). Second, the realizer
and realized properties instances under the Flat/Subset view are plausibly
entities that are in some sense the same, since the powers and processes of the one just are powers and processes of the other, so the view
accommodates (ii). Third, role-playing is also plausibly a mass-energy
neutral relation, covering (iii), because the processes that result from
the powers of the realized property instance just are, or are among, the
processes that result from the powers of the realizer property instance
hence the work of the realized instance just is the work of the realizer. Finally, and fourth, role-playing is a species of natural necessitation
16
If the arguments of Pereboom (2011, this volume) are correct, then the powers of realized and
realizer property instance are actually identical under the Flat/Subset view.

The Metaphysics ofNature, Science, andtheRules...

237

because, under the conditions, the role-playing suffices for the property/
relation instance whose role it plays, thus satisfying (ix).
Given the failures we have seen other approaches have had with features (i)(iii) one can consequently better appreciate the recent popularity of standard functionalist accounts from the philosophy of mind
in application to various topics in the philosophy of science involving scientific composition. And it bears emphasis that the Flat/Subset
account also unsurprisingly accommodates the OUP of compositional
explanation. Unfortunately, once we delve more carefully into the success of the Flat/Subset view with the other central features of scientific
composition, and compositional explanation, then we find a different
set of problems.
To begin, notice that role-playing is a oneone relation and hence fails
to accommodate the manyone character of scientific composition and
feature (viii). And this is linked to a still more troubling problem. By its
nature, role-playing cannot have qualitatively distinct relata and hence
accommodate feature (vii). Role-playing works through powers, processes, and roles that match exactly or are overlappingso role-playing
necessarily has relata that are qualitatively the same or similar and also
fails to cover feature (vii).17
Sometimes proponents of the Subset/Flat view seek to save their frameworks by challenging my descriptive account of cases of compositional
explanation and their compositional notions as involving manyone relations or even qualitatively distinct relata. I contend such alternative interpretations are descriptively defective and lead to all manner of problems
which I have documented in various ways elsewhere. However, I can most
swiftly highlight the deeper difficulties here by accepting for arguments
sake that scientific composition is neither a manyone relation nor one
with qualitatively distinct relata, for we immediately face a glaring problem.
The resulting concern is that we are left committed to a PEP-less
characterization of compositional explanation. If we take compositional
explanations to be based upon realization relations that do not relate
qualitatively distinct entities, then these explanations do not explain
17

If Perebooms arguments noted above are correct, then we have identity of powers, processes, and
roles.

238

C. Gillett

composed entities using qualitatively distinct entities. But PEP is one of


the most central, and distinctive, features of compositional explanations,
so we see that the proposed alternative descriptive interpretation of the
scientific cases is plausibly mistaken. Furthermore, it is also clear that the
standard functionalist accounts of scientific composition fail to accommodate the PEP of compositional explanation.
To summarize, although standard functionalist approaches in the philosophy of mind, through their role-based approach, have hit on key
insights about scientific composition, the latter points show that the kind
of role-based relation is again crucially important. As result of the twists
and turns of debates over the mind-body problem in philosophy, standard
functionalist accounts have ended up focused on role-playing, but this
relation fails to accommodate central features of scientific composition
and compositional explanation. The Flat/Subset view and other standard
functionalist accounts not only fail to accommodate key characteristics of
scientific composition in (vii) and (viii), but also consequently fail to provide an account of the PEP of compositional explanation in the sciences.

Part 5An Engaged Account of Scientific


Composition: Working Components and their
Joint Role-Filling
Finding that Appropriational, Unengaged accounts of scientific composition have deep problems does not yet suffice as a case for Engagement,
since perhaps these Appropriational accounts are the best we can. In order
to properly support Engagement, I therefore want to very briefly sketch a
type of account that results from Engagement and briefly outline why it
promises to do better than the Appropriational accounts. Since this type of
account is produced through full Engagement, it unsurprisingly has deep
affinities with the positions of Fodor and Dennett. And there are a number contemporary accounts of this kind developed in the work of writers
like Ken Aizawa (2007), Derk Pereboom (2002, 2011), Sydney Shoemaker
(2007) and myself (Gillett 2002, 2007a, 2013a) and Aizawa and Gillett
unpublished).18 Here I simply focus on the ontological backbone of these
18

I count the framework, or parts of it, in Sydney Shoemakers (2007) if it is interpreted as involving manyone relations of role-filling. On the other hand, if Shoemaker (2007) is taken to focus

The Metaphysics ofNature, Science, andtheRules...

239

types of account in what I term joint role-filling relations, since the nature
and significance of this type of relation has not been properly appreciated
and because this suffices to highlight the promise of this Engaged approach.
This type of account is constructed by reference to the features of scientific composition relations we find in actual cases. Given the finding
that entities in the sciences are plausibly working entities, it assumes that
such entities are all individuated by roles that, in various ways, connect back to their associated processes. Slightly different kinds of role
individuate the different ontological categories of entity, in individuals,
properties, powers, and processes, given their differences from each other,
but I proceed below with a generic notion of role to articulate the general framework.
Given that the entities that are components and composed in the
sciences are all working entities, this type of account also assumes that
their componency is based around the work or role of component
entities sufficing for the individuative work or role of the composed
entity. Putting the general idea of working components in terms of roles,
it appears that working components are entities whose roles mean that
these entities suffice, under the circumstances, for the role that is individuative of the composed entity. However, as we have begun to see, we
now have reason to conclude that a role-based account using role-playing
provides an inadequate account of scientific composition.
However, focusing on the features we find in concrete cases in the sciences, and especially their manyone character, this alternative Engaged
account takes scientific composition to involve what I am terming joint
role-filling where scientific components are members of collectives, or
teams, of working entities spatially contained within the relevant composed entity and inter-related such that they jointly result in the role
of the composed entity by jointly filling it. So, for example, a team of
inter-related molecules, including actin, jointly fills the role of the neuron. Or various processes of polymerization, and other molecular processes, jointly fill the role of protruding in the cell.
Notice the contrasts between joint role-filling and the role-playing that
underlies the Flat/Subset account of realization and related accounts. In
role-playing, the component entity plays the very role of the composed,
on oneone relations, basically role-playing relations, then Shoemakers account taken as a treatment of scientific composition is flawed in the ways laid out above.

240

C. Gillett

and hence we have a oneone relation between entities that are qualitatively identical or similar. However, in joint role-filling, none of the
components itself plays the role of the composed entity, since in this
manyone relation the components together fill the role of the composed.
Consequently, the entities jointly filling the relevant role are qualitatively
distinct from the composed entity.
Although abstractly framed, I now want to detail why these differences
between role-playing and joint role-filling have important implications
for accommodating the features of scientific compositionand both the
Ontological Unifying Power and also the PEP of compositional explanations. So let us consider whether an account of scientific composition
built around joint role-filling accommodates the characteristics (i)(iii)
and (vii)(ix) we have targeted in our discussion here.
To start, the relata of a joint role-filling relation are working entities
that do have mass-energy. Satisfaction of (i) by the joint-filling accounts
is therefore built into them given their focus on working entities. And
joint role-filling is also a non-productive relation passing (ii), since it is
synchronous, has relata that are in some sense the same, does not involve
the transfer of energy and/or mediation of force, and is not identical to
the manifestation of powers. Furthermore, the work of the composed
entity, in its individuative role, just is the combined work of the rolefillers. And the mass-energy of the entity whose role is filled just is the
combined mass-energy of the role-fillers. Consequently, joint role-filling
is plausibly a mass-energy neutral relation and covers feature (iii) that so
troubled neo-Causal, and Grounding, frameworks.
We thus accommodate three important features of scientific composition by taking it to be a joint role-filling relation between working
components. But it should also be obvious that we can cover another
central feature scientific composition, in the manyone character framed
in (viii), since we saw that composition is a manyone relation and this is
central to joint role-filling which is also a manyone relation. The latter is
important because it allows the joint role-filling framework to accommodate qualitatively distinct relata and hence (vii). Joint role-fillers may all
each be qualitatively distinct in their roles from the entity whose role they
fill. That is, none of the role-fillers itself plays the role of the composed
entity. Instead, the role of the composed entity is only jointly filled by the

The Metaphysics ofNature, Science, andtheRules...

241

role-fillers where their roles together suffice for the qualitatively distinct
role of the composed entity. Consequently, two more central features
of scientific composition are accommodated under the joint role-filling
framework.
Next it is important to note that joint role-filling is a non-productive
relation of natural necessitation satisfying (ix). Under certain conditions,
we have inter-related entities that can serve as role-fillers and which hence
suffice for, and as we saw above non-productively result in, the role that is
the composed entity. Spatio-temporally arrayed teams of entities related
by powerful and/or productive relations are such that their relations mean
that these entities do, or would, jointly fill the productive role of the
composed entity. But this role is individuative of the composed entity.
Hence the existence of these lower-level entities, under the conditions at
this time, non-productively determines that we also have the composed
entity at this time, so joint role-filing accommodates (ix) because it is a
relation of natural necessitation.
Given the foregoing points, the framework of working entities bearing
joint role-filing relations plausibly accommodates the key features of scientific composition we have focused upon in this chapter. And elsewhere,
I have shown the framework covers the other features of scientific composition as well.19 We thus already have good reasons to think the Engaged
framework of joint role-filling does better than the Appropriational
accounts. At this point, let me therefore step back and outline why the
account also offers a promising account of compositional explanation.
Let me assume what I term ontic representationalism about compositional explanation in the view that some scientific explanations work by
representing ontological relations between entities in the world. Given
this assumption, first, I have already outlined how joint role-filling is
a relation of natural necessitation. So, when one has joint role-filling,
under the conditions, then this naturally necessitates that one has the
composed entity whose role is filled. When we have certain molecular
processes of polymerization, alongside other molecular processes, under
the conditions, then one must have a cell protruding. Thus, we see how
the explanans of a compositional explanation explains its explanadum
under the joint role-filing account. Second, the joint role-filling account
19

Gillett (2016), Chap. 2.

242

C. Gillett

accommodates the PEP of compositional explanations. Joint role-filling


has qualitatively distinct relata, so if such relations underpin compositional explanations we can see how compositional explanation can
have PEP. And, third, if we start with entities X1Xn and Y that we
think are independent, but then successfully show that X1Xn jointly
fill the role of Y, we show that Y and X1Xn are in some sense the same.
Furthermore, establishing such a joint role-filling relation shows that
mass-energy, and force, of Y just is the mass-energy, and force, of X1
Xn. So joint role-filling also accommodates the OUP of compositional
explanation.
We can therefore see that the joint role-filling approach offers a promising account of compositional explanation as well as scientific composition. However, I should emphasize that I have only sketched the broad
virtues of this framework and it remains to provide detailed accounts of
particular compositional relations in the sciences. But a range of such
accounts of specific compositional relations in the sciences do now exist,
including treatments of the realization of properties (Gillett 2007a,
2016), the partwhole relations of individuals (Gillett 2007a, 2013a,
Pereboom 2011) and there is also promising work on the implementation of processes (Shoemaker 2007; Aizawa and Gillett unpublished). I
refer interested readers to this work to assess whether the promise of joint
role-filling accounts is borne out in these detailed treatments of specific
compositional relations.
To summarize, the joint role-filling framework is a fully Engaged
approach that does better than the Appropriational frameworks, since
it accommodates all the features of scientific composition. Furthermore,
this fully Engaged approach also does better than Appropriational
accounts by accommodating both the PEP and OUP of compositional
explanation. Much else needs to be done to develop such a fully Engaged
treatment, but I hope I have done enough here to show we achieve superior results when we follow the rules of Engagement.

The Metaphysics ofNature, Science, andtheRules...

243

Conclusion: Engaging theSciences


andtheMetaphysics ofNature
When it comes to the honest toil of investigating the principles governing
what objects are parts of others, it appears that most ontologists have been
following the paradigm of abstract algebra when it would have been better
to take a lead from sciences such as geology, botany, anatomy, physiology,
engineering, which deal with the real. (Simons 2006, p.611)

I have offered further support for Simonss conclusion, since we have


seen that frameworks appropriated from the abstract realm do poorly in
application to scientific composition and the metaphysics of nature, and
also that pursuing an Engaged approach following the disciplines that
deal with the real is more productive. And I have established similar
points about approaches that appropriate machinery developed for causation or work in the philosophy of mind. In each of these cases, we have
seen that we need to treat the singular vertical relations we find in science
and nature on their own terms, rather than trying to shoehorn them into
frameworks developed for distinct phenomena.
It is worth emphasizing that nothing I have shown warrants pursuing
issues in ontology outside of the metaphysics of nature using Engagement
and its focus on successful scientific explanation. Presumably, the way
to understand, for example, the vertical relations of SLAM entities is
to look at successful explanations utilizing such relations, rather than
compositional explanations in the sciences whose relations have working
entities as relata. So an analog of Engagement for these singular relations,
rather than Engagement itself, is plausible for each area in which we find
verticality.
Someone might still argue that despite everything I have shown, it is
safe to pursue issues in metaphysics of nature, say formulating physicalism or illuminating the global structure of nature, or solutions to the relation of mind and brain, or providing frameworks to understand debates
over reduction and emergence in the sciences, using the Grounding,
neo-Causal, or standard functionalist machinery. However, I know of no
plausible meta-justification that supports our taking such treatments of
the metaphysics of nature, or issue involving it our taking as trustworthy

244

C. Gillett

such treatments. At present, the only plausible meta-justification for an


account of the metaphysics of nature follows Simonss direction to deal
with the real by engaging the compositional explanations that are our
only widely successful explanations of verticality in nature. And I have
shown that Grounding, neo-Causal, or standard functionalist accounts
all mischaracterize scientific compositionand thus plausibly the metaphysics of nature.
Consequently, to understand scientific composition and vertical relations in nature itself, we plausibly need to follow the rules of
Engagement and develop a proprietary account of relations scientific
of composition developed specifically for such singular relations. As I
have shown, one violates the rules of Engagement at ones peril, while
sticking to these rules offers us a productive way to address verticality in
nature and the sciences, and the array of important issues intertwined
with it.

References
Aizawa, K. (2007). The biochemistry of memory consolidation: A model system
for the philosophy of mind. Synthese, 155, 6598.
Aizawa, K., & Gillett, C. (unpublished): The parts of Science: Scientific composition and compositional explanation.
Anderson, P. (1972). More is different: Broken symmetry and the nature of the
hierarchical structure of science. Science, 177, 39396.
Baumgartner, M. (2010). Interventionism and epiphenomenalism. Canadian
Journal of Philosophy, 40, 359384.
Bechtel, W., & Richardson, R. (1993). Discovering complexity. Princeton:
Princeton University Press.
Couch, M. (2011). Mechanisms and constitutive relevance. Synthese, 183,
375388.
Craver, C. (2007). Explaining the brain. Oxford: Oxford University Press, USA.
Dennett, D. (1969). Content and consciousness. London: Routledge Kegan Paul.
Dennett, D. (1978). Brainstorms. Montgomery, VT: Bradford Books.
Fine, K. (2001). The question of realism. Philosophers Imprint, 1(1), 130.

The Metaphysics ofNature, Science, andtheRules...

245

Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),


Metaphysical grounding: Understanding the structure of reality (pp. 3780).
NewYork, NY: Cambridge University Press.
Fodor, J. (1968). Psychological explanation. NewYork: Random House.
Fodor, J. (1974). Special sciences (or: the disunity of science as a working
hypothesis). Synthese, 28(2), 97115.
Gillett, C. (2002). The dimensions of realization: A critique of the standard
view. Analysis, 62, 31623.
Gillett, C. (2003). The metaphysics of realization, multiple realizability and the
special sciences. Journal of Philosophy, 100, 591603.
Gillett, C. (2007a). Hyper-extending the mind? Setting boundaries in the special sciences. Philosophical Topics, 35, 161187.
Gillett, C. (2007b). Understanding the new reductionism: The metaphysics of
science and compositional reduction. The Journal of Philosophy, CIV,
193216.
Gillett, C. (2007c). A mechanist manifesto for the philosophy of mind: A third
way for functionalists. Journal of Philosophical Research, 32, 2142.
Gillett, C. (2010). Moving beyond the subset model of realization: The problem of
qualitative distinctness in the metaphysics of science. Synthese, 177, 165192.
Gillett, C. (2013a). Constitution, and multiple constitution, in the sciences:
Using the neuron to construct a starting framewor. Minds and Machines,
23(3), 309337.
Gillett, C. (2013b). Understanding the Sciences through the Fog of
Functionalism(s). In Hunneman (Ed.), Functions: Selection and mechanisms.
Dordrecht: Kluwer.
Gillett, C. (2016). Reduction and emergence in science and philosophy. Cambridge:
Cambridge University Press.
Gillett, C., & Loewer, B. (Eds.). (2001). Physcialism and its discontents.
Cambridge: Cambridge University Press.
Harbecke, J. (2010). Mechanistic constitution in neurobiological explanations.
International Studies in the Philosophy of Science, 24(3), 267285.
Harbecke, J. (2014a). The role of supervenience and constitution in neuroscientific research. Synthese, 191(5), 725743.
Harbecke, J. (2014b). Regularity constitution and the location of mechanistic
levels. Foundations of Science, 20(3), 323338.
Harinen, Totte. (forthcoming). Mutual manipulability and causal inbetweeness.
Synthese.
Healey, R. (2013). Physical composition. Studies in the History and Philosophy of
Science Part B, 44, 4862.

246

C. Gillett

Kim, J. (1992). Multiple realization and the metaphysics of reduction. Philosophy


and Phenomenological Research, 126.
Kim, J. (1998). Mind in a physical world. Cambridge, MA: MIT Press.
Laughlin, R. (2007). A Different Universe: Reinventing Physics from the
Bottom Down. New York: Basic Books.
Lewis, D. (1972). Psychophysical and theoretical identifications. Australasian
Journal of Philosophy, 24958.
Machamer, P., Darden, L., & Craver, C. (2000). Thinking about mechanisms.
Philosophy of Science, 67, 125.
Pereboom, D. (2002). Robust nonreductive materialism. Journal of Philosophy,
99, 499531.
Pereboom, D. (2011). Consciousness and the prospects of physicalism. Oxford:
Oxford University Press.
Piccinini, G. (2004). Functionalism, computationalism, and mental states.
Studies in the History and Philosophy of Science, 35, 811833.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
R.Hale & A.Hoffman (Eds.), Modality: Metaphysics, logic, and epistemology
(pp.109136). NewYork, NY: Oxford University Press.
Schaffer, J. (2009). On what grounds what. In D. Chalmers, D. Manley, &
R. Wasserman (Eds.), Metametaphysics (pp. 357383). Oxford: Oxford
University Press.
Schaffer, J. (2016). Grounding in the image of causation. Philosophical studies
173(1), 49100.
Shapiro, L. (2004). The mind incarnate. Cambridge, MA: MIT Press.
Shoemaker, S. (2001). Realization and mental causation. In C. Gillett &
B. Loewer (Eds.), Physicalism and its discontents. Cambridge: Cambridge
University Press.
Shoemaker, S. (2007). Physical realization. NewYork: Oxford University Press.
Simons, P. (2006). Real wholes, real parts: Mereology without algebra. Journal of
Philosophy, xx, 597613.
Weinberg, S. (1994). Dreams of a final theory. NewYork: Random House.
Wilson, J. M. (1999). How superduper does a physicalist supervenience need to
be? Philosophical Quarterly, 49, 3352.
Wilson, J. M. (2009). Determination, realization, and mental causation.
Philosophical Studies, 145, 149169.
Wilson, J. M. (2010). Non-reductive physicalism and degrees of freedom.
British Journal for the Philosophy of Science, 61, 279311.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57, 145.

The Metaphysics ofNature, Science, andtheRules...

247

Wilson, J. M. (forthcoming). Metaphysical Emergence: Weak and Strong. In:


Bigaj and Wuthrich (Ed.), Metaphysics in contemporary physics. Poznan Studies
in the Philosophy of the Sciences and the Humanities.
Woodward, J. (2003). Making things happen: A theory of causal explanation.
NewYork: Oxford University Press.
Woodward, J. (2014). Scientific explanation. The Stanford Encyclopedia of
Philosophy (Winter 2014 Edition), Edward N.Zalta (Ed.). http://plato.stanford.edu/archives/win2014/entries/scientific-explanation/
Woodward, J. (2015). Interventionism and causal exclusion. Philosophy and
Phenomenological Research, 91, 303347.

9
Grounding andtheFormulation
ofPhysicalism
AndrewMelnyk

There are many sciences, and each science, to the extent that it gets things
right, uses its own characteristic theoretical vocabulary to describe a characteristic domain of entities. But how are the many sciences related to
one another? And how is the domain of entities proprietary to each science related to the domains of entities proprietary to the others? To try to
answer these questions is to address what I once called the problem of the
many sciences (Melnyk 1994, 222224; 2003, 12).
The problem of the many sciences looks like a promising candidate for
the sort of philosophical problem that naturalistic metaphysics should
addresswhere naturalistic metaphysics seeks to answer questions
that (i) ask what the world is like, albeit at a very high level of abstraction, that (ii) apparently dont fall within the province of the sciences
(as traditionally understood), but that (iii) creatures like us are capable in

A. Melnyk ( )
University of Missouri, Missouri, USA
The Author(s) 2016
K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_9

249

250

A. Melnyk

principle of answering (Melnyk 2013). Because I aspire to be a naturalistic metaphysician, I ask in this chapter whether an appeal to the relation
of grounding posited recently by certain philosophers might be useful
in one kind of approach to the problem of the many sciencesa physicalist approach. Jonathan Schaffer has explicitly proposed appealing to
grounding to formulate physicalism, albeit very briefly (Schaffer 2009,
364). Gideon Rosen has suggested formulating naturalism, a close relative of physicalism, by appeal to grounding (Rosen 2010, 111112). And
Shamik Dasgupta has recently tried to remove one obstacle to formulating physicalism by appeal to grounding (Dasgupta 2014).1 The prospects
of a grounding formulation of physicalism are also worth investigating
simply because of the remarkable level of current philosophical interest
in the putative relation of grounding.
The putative grounding relation that my question concerns is not
meant to be a generic relation under which such familiar relations as
supervenience, realization, and composition fall as species. Rather, it is
supposed to be a relation on a par with such relations; and it might be
posited either in addition to, or as a replacement for, such relations and
their kin (Wilson 2014, passim). It is also supposed, at least by three of its
leading proponents, to be a primitive relation (Schaffer 2009, 364; Rosen
2010, 113114; Fine 2012, 7879).2 Not all philosophers sympathetic
to grounding take it to be primitive. Dasgupta, for example, identifies
grounding with a certain sort of explanation: to say that some facts
ground another is just to say that the former explain the latter, in a particular sense of explain (Dasgupta 2014, 558). He therefore leaves open
the possibility that the particular sense of explain could be spelledout, yielding an account of what grounding is; indeed, he states what is
in effect a non-trivial sufficient condition for the grounding relation to
hold.3 In this chapter, however, I shall only consider a supposedly primitive grounding relation.
1

Daniel Stoljar devotes a subsection of his Stanford Encyclopedia of Philosophy entry on physicalism
to what he calls grounding physicalism, as if it were a standard approachwhich it isnt (Stoljar
2015, 10.3).
2
Rosen says we must accept it as primitive at least for now.
3
It is this: It is essential to ground that for any Xs and any Y, if the Xs obtain and if a fact about
the essence of a constituent of Y implies that the Xs are materially sufficient for Y, then the Xs

Grounding andtheFormulation ofPhysicalism

251

If an appeal to grounding is to be useful in a physicalist approach to


addressing the problem of the many sciences, then that will be because
it can play the starring role in formulating physicalism, understood as a
comprehensive doctrine about the world which accords to physics and
the physical a certain descriptive and metaphysical primacy among the
many sciences and their domains. How exactly to formulate physicalism
in this sense is a hard question (see, e.g., Melnyk 2003, 2014, Chaps. 1
and 2). But a formulation of physicalism must indisputably do at least
the following two things. First, it must characterize a relatively narrow
class of physical entities that are, as it were, physical in their own right;
it might characterize them, for example, as those entities expressible
in the proprietary vocabulary of physics (for elaboration and defense
of this option, see Melnyk 2003, 1120; 223237; for a useful survey
of other options, see Ney 2008). Call these entities narrowly physical.
Second, it must specify a relation R such that, necessarily, if an entity
which isnt narrowly physical (e.g., a chair or a zebra) stands in R to
an entity which is narrowly physical, then the former entity is nothing over and above the narrowly physical entity in the intuitive sense
required for physicalism. Call such an entity broadly physical. If an
appeal to grounding is to be useful in formulating physicalism, it will
be because grounding can be taken, and with advantages, to be relation
R in the characterization of the broadly physical.4 A rough statement of
physicalism would then be that everything is either narrowly physical
or broadly physical.
In this chapter, I caution against a gadarene rush to a grounding formulation of physicalism; and I do so by giving three reasons why we
should hesitate to take R in a formulation of physicalism to be grounding.5 Each reason occupies its own section.

ground Y (Dasgupta 2014, 588).


4
Here I follow Schaffer in taking concrete states of affairs to be possible relata of the grounding
relation (Schaffer 2010, 36).
5
For wide-ranging skepticism regarding the theoretical desirability of positing a relation of grounding, see Wilson (2014).

252

A. Melnyk

Grounding andtheBroadly Physical


The first reason for hesitating to take R to be grounding concerns whether
it is even capable of doing the job in a formulation of physicalism that
proponents of a grounding formulation of physicalism need it to do. For
it to do that job, the following conditional must be true: necessarily, if an
entity which isnt narrowly physical is grounded in an entity which is narrowly physical, then the first entity is nothing over and above the second
entity in the sense required for physicalism. I first want to argue that we
have no warrant for thinking that this conditional is true. I will end this
section by arguing that it is false.
Obviously it doesnt follow merely from ones needing grounding to
do a certain job in a formulation of physicalism that it is capable of
doing it. Nor is it at all obvious that it can do it: even if grounding is
indeed the primitive structuring conception of metaphysics (Schaffer
2009, 364), it doesnt follow a priori that, necessarily, if X grounds Y,
then Y is nothing over and above X in the sense required for physicalism: the primitive structuring conception of metaphysics might turn
out just not to be like that. One might claim to know by intuition
that, necessarily, if X grounds Y, then Y is nothing over and above X in
the sense required for physicalism. But it is quite implausible to claim
that one has reliable intuitions regarding a merely posited primitive
relationjust as it would have been implausible, when the neutrino
was first posited, for someone to claim to have reliable intuitions about
the properties of neutrinos.
But what if one knew independently that, necessarily, if X grounds Y,
then X metaphysically necessitates Y? Wouldnt that be enough to show
that, necessarily, if X grounds Y, then Y is nothing over and above X in
the sense required for physicalism? It wouldnt, for
being metaphysically necessitated by the narrowly physical does not entail
being nothing over and above the narrowly physical in the sense required
for physicalism
and not just because the metaphysical necessitation might result
from bizarre possibilities like occasionalism. Here is a novel argument

Grounding andtheFormulation ofPhysicalism

253

intended to demonstrate this failure of entailment; it invites us to consider a series of three cases.6
Suppose, first, that a state-token x of one state-type nomically necessitates a later state-token y of an entirely different state-type. For vividness,
think of x as a neural state and y as a pain state:
(1) x at t1 nomically necessitates y at t2.
Suppose also that this nomic necessitation is brute, not in the sense
that it has no explanation at all (for it may have a theistic explanation in
terms of a divine will), but in the sense that it has no explanation in terms
of more basic nomic generalizations: it has no same-level explanation in
terms of states of other types that intervene between x and y; and it has
no lower-level explanation in terms of underlying states that constitute x
and y. Clearly, the brute nomic necessitation of y by x does not entail that
y is nothing over and above x.
Now consider a second case exactly similar to the first except that the
necessitating state x and the necessitated state y are now simultaneous,
so that the brute nomic necessitation of y by x is synchronic rather than
diachronic:
(2) x at t1 nomically necessitates y at t1.
Surely, the brute nomic necessitation of y by x still doesnt entail that y
is nothing over and above x. For it didnt entail this in the first case, and
the second case differs from the first only in the changed relation between
the time of x and the time of y. It is very implausible to think that we
could move the time of x arbitrarily close to the time of y while y continues to be something over and above x, but that the moment we make the
times identical, y becomes nothing over and above x. To think that would
be to attribute magical powers to time.
Consider, finally, a third case which is exactly the same as the second, except that now the brute necessitation is not nomic but rather
metaphysical:
(3) x at t1 metaphysically necessitates y at t1.

For earlier arguments with the same goal, see (Melnyk 2003, 5770; Wilson 2005).

254

A. Melnyk

Now, the nomic (brute, synchronic) necessitation of y by x didnt entail


that y is nothing over and above x; and theres no reason to think that
the change from nomic (brute, synchronic) necessitation to metaphysical
(brute, synchronic) necessitation could make any relevant difference. So
the metaphysical (brute, synchronic) necessitation of y by x still doesnt
entail that y is nothing over and above x. But, of course, brute, synchronic
metaphysical necessitation is still metaphysical necessitation. So what this
third case shows is that it can happen that x metaphysically necessitates y
without ys being nothing over and above x.
Since being metaphysically necessitated by the narrowly physical
doesnt entail being nothing over and above the narrowly physical in the
sense required for physicalism, it seems that any possible warrant for
thinking that an entity which is grounded in a narrowly physical entity
must be nothing over and above the physical entity would have to arise
from whatever it is that holds in addition to metaphysical necessitation
when a narrowly physical entity grounds an entity that isn't narrowly
physical. But, precisely because this additional factor is primitive, so
that nothing can be said about its nature, we cant use premises about its
nature to derive the conclusion that it ensures the acceptability to physicalism of whatever is grounded in the narrowly physical.
So far in this section I have been arguing that we have no warrant for
believing the conditional claim that, necessarily, if an entity which isnt
narrowly physical is grounded in an entity which is narrowly physical,
then the first entity is nothing over and above the second entity in the
sense required for physicalism. But proponents of a grounding formulation of physicalism might respond that they dont need to provide any
such warrant, because the conditional claim is true by stipulation: when
they utter tokens of grounding they are to be understood as referring
to a certain primitive relation that has the property of being such that,
necessarily, if an entity which is narrowly physical stands in that relation
to an entity which isnt narrowly physical, then the second entity is nothing over and above the first in the sense required for physicalism; and if
a relation lacks this property, then it simply isnt what they are referring
to when they utter tokens of grounding. On this proposal, a first-pass
formulation of physicalism would say that every entity is either narrowly
physical or else stands in a certain relation, to be called grounding, to an

Grounding andtheFormulation ofPhysicalism

255

entity that is narrowly physical, where grounding is defined, as above,


to be primitive and to have the property of being such that
Now a stipulative definition, unlike a descriptive definition, cannot be
faulted for being inaccurate; but it can be faulted for being inconsistent,
and the stipulative definition of grounding mooted in the previous
paragraph is, I shall now argue, defective in just this way.7 To begin, let us
briefly consider the relation of being taller than. It can only hold between
relata that have a certain characteristic, namely, a height; it cannot possibly hold between items that cannot have heights. Why this restriction
on the relations possible relata? Plausibly, the restriction follows from
the nature of being taller than, a nature into which height enters: for X
to be taller than Y just is for X to have a certain height and for Y to have
a certain (lesser) height. But the restriction can follow from the nature
of being taller than only because that nature is complex. Grounding,
by contrast, is stipulated to refer to a relation that is primitive: when it
holds between two items, it doesnt do so in virtue of anything elseit
just holds. Because grounding has no complex nature, its nature places
no restriction on what items it could (metaphysically) relate. So it could
(metaphysically) hold between any two items, whatever their respective
natures or features. It could hold, for instance, between a narrowly physical X and a Y that is a veritable paradigm of non-physicalityor, for that
matter, a Y that is physical all right, but patently something over and
above X.But this possibility entails that the relation cannot be such that,
necessarily, if an entity which is narrowly physical stands in it to an entity
which isnt narrowly physical, then the second entity is nothing over and
above the first in the sense required for physicalism. The upshot, then, is
that the stipulative definition of grounding mooted above is inconsistent: its stipulation of grounding as a primitive relation conflicts with
its stipulation of grounding as a relation such that, necessarily, if an
entity which is narrowly physical stands in it to an entity which isnt narrowly physical, then the second entity is nothing over and above the first
in the sense required for physicalism.

Hence I think Wilson concedes too much when she writes that A Grounding claimeffectively
stipulates nothing over and above-ness (Wilson 2014).

256

A. Melnyk

More importantly, however, the argument of the preceding paragraph


also serves to show that its untrue (not just unwarranted) that anything
that is grounded in a narrowly physical entity would be bound, metaphysically, to be nothing over and above that entity in the sense required
for physicalism. For, precisely because the relation of grounding is primitive, and therefore not constituted even in part by facts about its relata,
nothing rules out the metaphysical possibility of its holding between a
narrowly physical entity, on the one hand, and, on the other, something
that isnt nothing over and above it in the sense required for physicalism.
Grounding looks to be the wrong tool for doing the job that proponents
of a grounding formulation of physicalism need it to do.
The official task of this section is now complete; but if it is indeed
untrue that anything that is grounded in a narrowly physical entity would
be bound, metaphysically, to be nothing over and above that entity in the
sense required for physicalism, then this helps to support an important
epistemological conclusion. The conclusion is that the broadly empirical
methods of acquiring knowledge used in the sciences cant be deployed to
support claims that this grounds that.
Consider a concrete case in which we are trying to decide whether
some state X grounds a certain state Y.How could we come to know that
the physical state grounds the mental state? What, in principle, would
constitute evidence that the physical state grounds the mental state? That
X grounds Y is surely a fact over and above the sheer spatio-temporal
contiguity of X to Y, butcruciallyit makes no difference to the causal
powers of either X or Y; and this makes it hard to see how the fact that
X grounds Y could be known by direct observation. But could the claim
that X grounds Y still be supported by an inference to the best explanation? Not in the obvious way, for if the fact that X grounds Y makes no
difference to the causal powers of either X or Y, then it can play no direct
role, at any rate, in explaining any observable feature of the world. Still,
a claim can form part of the best explanation of certain observable facts,
not because the claim itself explains those facts, but because it makes
possible an increase in parsimony that is relevant to the assessment of the
hypothesis of which it is a part as the best available explanation of the
facts in question. Suppose, then, that we observe that Xs are sufficient
for Ys. Could the hypothesis that the Xs ground the Ys constitute a better

Grounding andtheFormulation ofPhysicalism

257

explanation of the observed regularity than, say, the rival hypothesis that
the Xs are lawfully sufficient for the Ys, because it is more parsimonious
than this rival? No. The hypothesis that the Xs ground the Ys would
be more parsimonious than the rival only if it entailed that the Ys were
nothing over and above the Xs. But, I have argued in this section, there is
no reason to think that it does, and one reason to think that it does not.

The Dispensability ofGrounding


My second reason for hesitating to take relation R (in a formulation of
physicalism) to be grounding is that, other things being equal, we should
not formulate physicalism by positing a new primitive relation (e.g.,
grounding) if we can do so without positing a new primitive relation.
And we can formulate physicalism without positing a new primitive relationby appealing instead to a carefully spelled-out relation of realization. So, other things being equal, we should not formulate physicalism
by positing the primitive relation of grounding.
I claim that physicalism can be formulated to a first approximation as
the view that every entity (better: entity-token) is either narrowly physical or else is realized, in a carefully defined sense of realized, by some or
other narrowly physical entity (for elaboration, see Melnyk 2003, 611;
2032).8 To make my claim plausible, I will have to explain in some
detail what that sense is. So let p name a particular actual physical statetoken, and m a particular actual mental state-token. Then p realizes m
(in the intended sense) only if
(i) m is a token of a mental state-type M with a certain higher-order
essence: for a token of M to exist just is for there to exist a token of
some (lower-order) state-type such that tokens of that (lower-order)
state-type play role RM, the role distinctive of M;

To a first approximation only, because the formulation leaves various questions unanswered.
Should the entities quantified over include abstracta? Or necessary existents? To what categories
should the entities belongstates, events, properties, objects, facts, truths? See Melnyk (2003,
611; 2032).

258

A. Melnyk

(ii) p is a token of a physical state-type P such that, necessarily, given


thephysical laws and physical circumstances C, tokens of P play role
RM; and
(iii) the laws of physics hold and physical circumstances C obtain.
Claims (i) through (iii) are necessary for p to realize m, but they are
not sufficient. They jointly entail that some token of mental state-type
M exists. But they do not entail that this token of mental state-type
M is one and the same as the particular token of mental state-type
M that we have called m. Hence, claims (i) through (iii) do not
jointly entail that p realizes (not just any old token of M but) m in
particular. However, if claims (i) through (iii) are conjoined with the
further mental-to-mental (not mental-to-physical) identity claim that
(iv) the token of mental state-type M whose existence is entailed by
claims (i) through (iii) = m,
then all four claims together do entail that p realizes m in particular. Thus,
p realizes m (in the intended sense) if and only if claims (i) through (iv)
are true.
Four glosses on claims (i) through (iv) are required for their full comprehension. First, the identity claim implicit in claim (i)that mental
state-type M = so-and-so higher-order state-typeis metaphysically necessary and, in almost all cases, a posteriori. Second, the word necessarily
in claim (ii) is meant to express the idea that the claim that tokens of P
play role RM is in principle derivable from statements of the laws of physics plus the claim that physical circumstances C obtain. Third, claim (i)
speaks of playing a role only for the sake of role-playings familiarity; it
would be better to speak, more broadly, of meeting a condition, where the
condition could indeed be met by playing a causal role, but could also be
met in other ways, for example, by standing in certain spatio-temporal
relations or having a certain history or having a certain bio-function
(Melnyk 2003, 3742). Finally, the term higher-order is used in claim
(i) instead of the standard functional. This is partly because the connotations of functional are unnecessarily narrow (see the third gloss), but
mostly because higher-order draws attention to the metaphysical heart
of this definition of realization and its associated formulation of physicalism: its construal of broadly physical state-types as higher-order types.

Grounding andtheFormulation ofPhysicalism

259

A few features of realization understood in this way deserve emphasis


in the context of the present chapter. First, realization, unlike grounding, is not a primitive relation, for the holding of the realization relation
between p and m just is the holding of the four conditions described by
claims (i) through (iv). Second, and for the same reason, realization is
an internal relation in the sense that its holding is the holding of certain
conditions regarding p and m none of which is the holding of a relation
between p and m. (Being non-primitive is a necessary condition for a
relation to be internal but not a sufficient one: a relation constituted
by the holding of other, simpler relations between its relata would on
that account not be primitive, but neither would it be internal.) Third,
to say that a physical state-token realizes a mental state-token in the
sense defined is to say more than that a certain modal correlation holds
between a type of physical state and a type of mental statewhich is
all that claims of mental-on-physical supervenience do. Realization is in
fact a hyperintensional relation, at least in the sense that it slices more
finely than metaphysical necessitation: p might metaphysically necessitate m (given the physical facts) but not realize itor it might metaphysically necessitate m (given the physical facts) because it realizes it. This
feature of realization ought to appeal to fans of grounding, who object
to supervenience claims precisely on the ground that they merely report
modal correlations, and who claim that grounding has the advantage over
supervenience of being a hyperintensional relation (e.g., Schaffer 2009,
364). Finally, to claim that a physical state-token realizes a mental statetoken in the sense defined is to commit oneself to a particular view about
the nature of the mental state-type of which the mental state-token is a
token: the view that the mental state-type has what I have called a higherorder essence. In this regard, realization contrasts sharply with grounding,
which, as we saw at the end of the section Grounding and the Broadly
Physical, imposes no constraint upon the nature of its relata. Indeed, I
conjecture that any relation fit to serve as relation R in a formulation of
physicalism must take some stand on the nature of broadly physical statetypeswhich is to say that no relation could make tokens of absolutely
any kind of state-type physicalistically acceptable.
The possibility of formulating physicalism by appeal to the relation
of realization as defined above shows that we can formulate physicalism

260

A. Melnyk

without positing a new primitive relation of grounding, and hence that,


other things being equal, we should not formulate physicalism by positing
grounding. Are other things equal? Here I can only assert that realization
physicalism can claim several virtues: it keeps faith with certain intuitions
about the content of physicalism; it entails, and arguably explains, the
supervenience of the non-narrowly physical on the narrowly physical and
the (closely related) necessitation of the non-narrowly physical by the narrowly physical; and it helps to solve the various problems of (generalized)
mental causation (see Melnyk 2003, 33; 5960; 4970; 123174, esp.
134139 and 159164).

The Consistency ofGrounding


withPhysicalism Itself
My third reason for hesitating to take relation R in a formulation of
physicalism to be grounding is that it is hard to see how instances of
grounding could themselves be consistent with physicalism (when formulated in terms of grounding). Thus, grounding apparently falls foul of a
requirement on formulating physicalism that was first clearly articulated
by Michael Lynch and Joshua Glasgow in 2003: any candidate for R
must be a relation whose instantiation is itself consistent with physicalism (Lynch and Glasgow 2003).9 To see why, let us consider the possible
ways in which instances of grounding could be rendered consistent with
physicalism (formulated in terms of grounding).

In a recent paper, Dasgupta appears to raise exactly the same problem for a grounding formulation
for physicalism (Dasgupta 2014, 561562). But a careful reading reveals that in fact he doesnt. The
appearance arises because he raises his problem by presenting an argument that Xs grounding Y
has no purely physical ground (Dasgupta 2014, 571). But the reason Xs grounding Y has no
purely physical ground, for Dasgupta, is that it is partly grounded in some ungrounded connection between Xs and Ys (Dasgupta 2014, 569); the reason is not that the grounding relation itself
is problematically non-physicala possibility, indeed, that the paper nowhere mentions. And
Dasguptas solution to his problem is to argue that physicalism can allow the sort of connection
between Xs and Ys that he has in mind to have no ground at all, and hence no physical ground
(Dasgupta 2014, 575).

Grounding andtheFormulation ofPhysicalism

261

Might instances of grounding be consistent with physicalism because


they are broadly physicalbroadly physical because themselves grounded
in narrowly physical entities (pace the contention of the section Grounding
and the Broadly Physical that an entitys being grounded in something
narrowly physical doesnt suffice to make it broadly physical)? I think not.
First, grounding is meant to be a primitive relation, and it seems that
nothing primitive is even a candidate to be grounded. For anything primitive is fundamental, and nothing fundamental is grounded (save perhaps
in itself, but clearly self-grounding is no help here). Secondly, whenever
an instance of grounding is itself grounded in something physical, there
must then be a second instance of groundingthe first instance itself
being groundedwhich also has to be grounded in something physical;
and so on. Such a series of groundings cannot terminate, since the final
instance of grounding would, in that case, fail to be grounded in something narrowly physical and hence would not be consistent with physicalism.10 Neither is it plausible, however, that the series should continue
indefinitely, since this would require infinitely many physical entities for
the infinitely many instances of grounding to be grounded in.11
If instances of grounding do not achieve consistency with physicalism
because they are broadly physical, might they do so because they are narrowly physical? But obviously grounding is not a physical relation in the
sense of a relation expressed by a simple two-place predicate of physics.
Nor can grounding be a relation expressed by some complex construction
of physical terms, or of physical plus topic-neutral terms, because that
would make grounding analyzable and hence not primitive.
It may be, however, that the category of the narrowly physical should
be expanded, and that grounding belongs to the expanded category.
Wemight, that is, also want to count something as narrowly physical if

10

The hence is justified because in this paragraph I am assuming that every instance of grounding
would be made consistent with physicalism in the same way, by being grounded in something narrowly physical.
11
Dasgupta argues that an infinite series of grounded grounding facts is harmless (Dasgupta 2014,
587589). But he is not talking about an infinite series of physically grounded grounding facts, as I
am; and in any case an infinite series of grounded grounding facts is harmless on his account only
if grounding is not primitive.

262

A. Melnyk

(i) it is topic-neutral in the sense of being no more associated with any


one branch of science than any other, and (ii) its existence is presupposed by a complete physical description of the worlda description
of the world that tells us everything about the world that physics has to
tell. And, though the matter is highly controversial, it may be that, for
example, causation would count as narrowly physical by meeting conditions (i) and (ii). But what about grounding? Alas (for the prospects
of grounding physicalism), a complete physical description of the world
seems not to be committed to the grounding relation.12 For the one place
where one might suspect such a commitment is in what physics says
about the relationship between large physical systems and small ones
say, between gross matter and molecules. But physicists seem to think
that the orthodox language of physics has the expressive power in principle to characterize any physical system whatsoever, no matter how large
or complex; and if they are right, as I assume they are, then a complete
physical description of the world is not committed to grounding.
So far I have been assuming that the requirement proposed by Lynch
and Glasgow (that tokens of R be consistent with physicalism) is a
genuine requirement on formulating physicalism. But Tom Polger has
recently argued that it is not, on the grounds that it cannot in principle
be met by any view, physicalist or not, according to which all things exist
solely in virtue of certain other things (Polger 2013, 8485). Now I certainly agree that one can make the mistake of formulating physicalism
too strongly.13 But I deny that the LynchGlasgow requirement cannot
be met. It can be met, when R in a formulation of physicalism is taken

12

Committed to the grounding relation cannot just mean the same as logically entails that the
grounding relation has instances. In the sense I intend, a complete physical description of the
world is committed to grounding iff (i) the complete physical description is possibly true, (ii) the
claim that there exist instances of grounding is possibly false, and (iii) it is logically necessary that,
if the complete physical description is true, then there exist instances of grounding. Conditions (i)
and (ii) serve to rule out degenerate cases of entailment, in which the complete physical description
is necessarily false or the conclusion necessarily true.
13
For example, physicalism is formulated too strongly if it is formulated as saying that all facts or
truths hold in virtue of physical facts or truths. This formulation is too strong because, if physicalism is true, then its a fact that physicalism is true; but the fact that physicalism is truethe fact
that nothing exists that is neither narrowly nor broadly physicaldoesnt hold in virtue of physical
facts alone. See Melnyk (2003, 2526; 97 n.17).

Grounding andtheFormulation ofPhysicalism

263

to be the relation of realization as I defined it above. The existence of


instances of realization is consistent with physicalism, because realizations holding between physical state-token p and mental state-token m
just is the holding of the four conditions described by claims (i) through
(iv), as we saw in the section The Dispensability of Grounding; and the
holding of each of these four conditions is consistent with physicalism,
as the following paragraphs will show.
Claims (ii) and (iii) are consistent with physicalism because they
describe only narrowly physical conditions. Claims (i) and (iv) are (metaphysically) necessary identity claims. So they are committed to the relation of identity, and to the entities they claim to be (self-)identical.14 That
identity (i.e., metaphysically necessary self-identity) is consistent with
physicalism can be argued for in two ways, of which the first is this. So
long as a complete physical description of the world says that anything at
all exists, whether it be a particle, a field, spacetime, or even the physical
universe as a whole, the physical description is thereby committed to the
claim that the thing is metaphysically necessarily self-identical. Hence
identity is something to which a complete physical description of the
world is itself committed. Given the suggestion made two paragraphs
ago, that something counts as narrowly physical if (i) it is topic-neutral,
and (ii) its existence is presupposed by a complete physical description of
the world, it follows that identity is consistent with physicalism.
The second way to argue that identity is consistent with physicalism restricts the scope of physicalism to contingent reality. In line with
this restriction, it claims that realization physicalism requires that only
instances of relations (or properties) that are contingent must be either
narrowly physical or else realized by something narrowly physical. An
instance of a relation is contingent in the intended sense if there is a
possible world at which, though its actual-world relata exist there, it fails
to hold between them.15 Instances of identity in the actual world, however, are not contingent in this sense: if a = b, then in no possible world
do a and b exist but a b, that is, identity is necessary in Kripkes sense

14
15

All claims of identity are claims of self-identity, of course.


So its not enough for contingency that there be a possible world at which a = b is not true.

264

A. Melnyk

(Kripke 1980, 109). Since this is so, realization physicalism does not have
to require that instances of identity be narrowly physical or realized by
something narrowly physical; and so instances of identity are consistent
with physicalism even if they are neither narrowly physical nor realized by
something narrowly physical.
At this point, enthusiasts for a grounding formulation of physicalism
might ask whether what I have just said about identity is true also of
grounding, so that grounding too is consistent with physicalism (contrary to my earlier contention). The answer to this question is affirmative,
of course, only if its true that, if X actually grounds Y, then theres no
possible world in which X and Y exist but X does not ground Y.But I see
no reason to think that this is true. The two standard arguments for the
necessity of identity, which appeal, respectively, to the necessity of selfidentity and to the claim that non-descriptive referring expressions are
rigid designators (Kripke 1980, 104), look most unlikely to carry over to
support an analogous thesis of the necessity of grounding. Also, as noted
in the section Grounding and the Broadly Physical, the grounding relation, because it is primitive, doesnt hold between two items in virtue of
any other facts; a fortiori, it doesnt hold in virtue of other facts about
the relata; a fortiori again, it doesnt hold solely in virtue of other facts
about the relata. So we cannot reason that, just because in some world
w the relata exist, in w the grounding relation must hold between them.
Finally, even if it is granted that grounding is a species of metaphysical
necessitationin a sense which implies that, necessarily, if X grounds
Y, then X metaphysically necessitates Yit doesnt follow that in every
world in which X and Y exist X grounds Y.It does indeed follow that in
every such world X metaphysically necessitates Y, given the transitivity of
inter-world accessibility; but there is (we are assured) more to grounding
than metaphysical necessitation.16

16

Might what I said about identity be true also of realization? Not if the first relatum is taken to be
physical state-token p, or p plus physical conditions C.For the actual worlds laws of physics dont
hold in all possible worlds in which p, or p plus C, exists. But if the first relatum is taken to be p
plus C and the holding of the actual worlds laws of physics, then perhaps yes. I dont know which
view of the first relatum is correct.

Grounding andtheFormulation ofPhysicalism

265

What about the entities said by claims (i) and (iv) to be (self-)identical?
Claim (iv) speaks of mental state-token m, which is by hypothesis physically realized and therefore consistent with physicalism. Claim (i) speaks
of the mental state-type M. But because (one might reasonably suppose)
there are no untokened types, the existence of state-type M just is the
existence of its state-tokens. So for M to be consistent with physicalism,
it is enough if each of its tokens is physically realizedwhich they are
if realization physicalism is true. In short, realization physicalism says
that every contingent entity-token is either narrowly physical or realized
by something narrowly physical, and tokens of realization are consistent
with physicalism because they are in part narrowly physical and in part
realized by something narrowly physical.
My argument that the existence of instances of realization is consistent with physicalism relies on the tacit premise that the existence of X
is consistent with physicalism if (i) the existence of X just is (i.e., =) the
existence of Y1, Y2, Y3,Yn, and (ii) each of the Yi is narrowly physical
or realized by something narrowly physical. It might therefore seem as if
I have modified the formulation of realization physicalism by introducing a new way in whicha new relation in virtue of whichan instance
of a property (or relation) that isnt, or isnt wholly, narrowly physical can be rendered consistent with physicalismwhich would then
immediately prompt the same question that we asked about instances of
realization, mutatis mutandis, namely, the question of how instances of
this new relation can be rendered consistent with physicalism. But this
appearance is illusory. To be sure, one could express realization physicalism as follows:
Every contingent instance of a property (or relation) is
either narrowly physical
or realized by something narrowly physical
or is one and the same as the existence of instances I1, I2, I3,In (of
properties or relations P1, P2, P3,Pn, respectively), every one of which is
either narrowly physical or realized by something narrowly physical.17

17

This rough formulation is good enough, I hope, for the present purpose.

266

A. Melnyk

But such a formulation is unnecessary and misleading. It obscures the


fact that, if the existence of X just is (i.e., =) the existence of Y1, Y2, Y3,
Yn, then realization physicalists may stop speaking of X as such without
reducing their ontological commitments; and if they may do so, and the
existence of Y1, Y2, Y3,Yn is consistent with physicalism for independent reasons, then they do not need a third disjunct in their formulation
of physicalism to handle X.In the case at hand, if realizations holding
between physical state-token p and mental state-token m just is the holding
of the four conditions described by claims (i) through (iv) in the section
The Dispensability of Grounding, then it is open to realization physicalists to acknowledge the existence of those four conditions but then to
shut upor at least to utter nothing containing the term realize. And
since each of the four conditions is consistent with physicalism, no further problem remains for realization physicalists. You might worry that,
if this move is possible, realization cant be much of a relation.18 And I
would entirely agree; but it gets the job done.
But if the last five paragraphs are correct, what in that case is wrong
with Polgers argument for thinking that the LynchGlasgow requirement on a formulation of physicalism (that tokens of R be consistent
with physicalism) cant in principle be met? Polgers argument is dilemmatic, and the relevant portion is this:
If the R-relation linking the Ps [physical entities] and the Ms [non-narrowly
physical, e.g., mental entities] is among the Ms, then either it depends on
and is determined by the Ps or it does not. If it does not, then the claim
that all Ms depend on and are determined by the Ps is falsified But if it
is so dependent, then we will need to know by what R-relation it so
depends, and we are off and running on a regress. So it seems that the
R-relation cannot be among the Ms. (2013, 84)

When the R-relation is realization, I have tried to divide and conquer, taking it to be partly among the Ps and partly among the Ms. To
the extent that it is among the Ms, I have acceptedof coursethat it
depends on and is determined by the Ps, and have said specifically that
it is realized by the Ps. Polger thinks that a regress must now ensue, but
18

By the light of day.

Grounding andtheFormulation ofPhysicalism

267

does not say why. He may think a regress inevitable because we need a
second R-relation to make the first R-relation consistent with physicalism,
and a third to make the second consistent with physicalism, and so forth.
And in response to the contrary suggestion that further R-relations are
not needed because the first R-relation could make itself consistent with
physicalism, he may mean to object that, even so, there would still ensue
a regress of tokens of the first R-relation.
What should we make of Polgers charge of regress? Here, for convenience, are the four crucial claims from above that together define realization in my sense:
(i) m is a token of a mental state-type M with a certain higher-order
essence: for a token of M to exist just is for there to exist a token of
some (lower-order) state-type such that tokens of that (lower-order)
state-type play role RM, the role distinctive of M;
(ii) p is a token of a physical state-type P such that, necessarily, given the
physical laws and physical circumstances C, tokens of P play role
RM; and
(iii) the laws of physics hold and physical circumstances C obtain.
(iv) the token of mental state-type M whose existence is entailed by
claims (i) through (iii) = m.
I have claimed two things: (1) the holding of the realization relation
between physical state-token p and mental state-token m just is the
holding of the four conditions described by claims (i) through (iv); and
(2) everything required for claims (i) through (iv) to be true is physical or physically realized, hence consistent with physicalism (when formulated by appeal to realization). But in so claiming, have I appealed
to a second R-relation? If I have, then it is with (1), and the second
R-relation is identity; but I have already argued, in two ways, that
identity is consistent with physicalismand neither way appeals to a
third R-relation. In fact, however, I need not be construed as having
appealed, in (1), to identity as a second R-relation. For, as noted two
paragraphs ago, I am at liberty simply to cease speaking of realization as
such, while retaining the substance of my realization physicalismto
replace p realizes m with the claim that the four conditions described
by claims (i) through (iv) hold.

268

A. Melnyk

So much for appealing to a second R-relation. It remains to ask


whether, in endorsing (1) and (2), I am committed to a regress of tokens
of realization. Clearly not in (1). In (2)? No. The claim, made by (2), that
everything required for claims (i) through (iv) to be true is physical or
physically realized entails a single claim of physical realization: the claim
that m is physically realized. But m, of course, is the state-token whose
physical realization we were originally concerned with. So (2) doesnt
introduce even a second token of realization, letalone an infinite series of
them. And there is no circularity here either: the claim that p realizes m is
no part of the analysis of the claim that p realizes m. For the conjunction
of (1) and (2) makes a meta-claim, a claim about ps realizing m; it doesnt
purport to define ps realizing m.
I tentatively conclude that, as Lynch and Glasgow claimed, any candidate for R in a formulation of physicalism must be a relation whose
instantiation is itself consistent with physicalism; and that a grounding
formulation of physicalism does not meet this requirement. The conclusion of this section, together with those of the sections Grounding and the
Broadly Physical and The Dispensability of Grounding, make a strong
case that much philosophical work remains to be done if physicalism is
to be formulated by appeal to grounding.19

References
Dasgupta, S. (2014). The possibility of physicalism. The Journal of Philosophy,
111(9), 557592.
Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),
Metaphysical grounding: Understanding the structure of reality (pp. 3780).
Cambridge: Cambridge University Press.
Kripke, S. (1980). Naming and necessity (p.1980). Cambridge, MA: Harvard
University Press.

19

For helpful comments on an earlier draft I thank Ken Aizawa, Carl Gillett, and the other participants in the Composition and Ground Workshop at Rutgers University, Newark, April 1011,
2015.

Grounding andtheFormulation ofPhysicalism

269

Lynch, M., & Glasgow, J. (2003). The impossibility of superdupervenience.


Philosophical Studies, 113(3), 201221.
Melnyk, A. (1994). Being a physicalist: How and (more importantly) why.
Philosophical Studies, 74, 221241.
Melnyk, A. (2003). A physicalist manifesto: Thoroughly modern materialism.
Cambridge: Cambridge University Press.
Melnyk, A. (2013). Can metaphysics be naturalized? And if so, how? In D.Ross,
J.Ladyman, & H.Kincaid (Eds.), Scientific metaphysics (pp.7995). Oxford:
Oxford University Press.
Melnyk, A. (2014). Perebooms robust non-reductive physicalism. Erkenntnis,
79(5), 11911207.
Ney, A. (2008). Defining physicalism. Philosophy Compass, 3, 103348.
Polger, T. (2013). Physicalism and Moorean supervenience. Analytic Philosophy,
54(1), 7292.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
B.Hale & A.Hoffmann (Eds.), Modality: Metaphysics, logic, and epistemology
(pp.109136). Oxford: Oxford University Press.
Schaffer, J. (2009). On what grounds what. In D. Chalmers, D. Manley, &
R.Wasserman (Eds.), Metametaphysics: New essays on the foundations of ontology (pp.34783). Oxford: Oxford University Press.
Schaffer, J. (2010). Monism: The priority of the whole. The Philosophical Review,
119, 3176.
Stoljar, D. (2015). Physicalism. The Stanford encyclopedia of philosophy (Spring
2015 edition), E.N. Zalta (Ed.), http://plato.stanford.edu/archives/spr2015/
entries/physicalism/
Wilson, J.M. (2005). Supervenience-based formulations of physicalism. Nos,
39(3), 426459.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry: An
Interdisciplinary Journal of Philosophy, 57(56), 535579.

10
Grounding inthePhilosophy
ofMind: ADefense
AlyssaNey

Introduction
One of the major trends in metaphysics in recent years has been in the
development and application of novel conceptual frameworks for representing facts about realism, fundamentality, and metaphysical priority.
Of particular interest have been the concepts of grounding (proposed by
Paul Audi (2012), Kit Fine (2001), Gideon Rosen (2010), and Jonathan
Schaffer (2009), among others)1, the concept of the real (proposed by
Fine (2001)), and that of metaphysical structure (proposed by Ted Sider
(2011)). All of these have been proposed as new primitive concepts, and

Witmer et al. (2005) defend a related in virtue of notion. Bennett (2011) speaks of
building.

A. Ney ()
Department of Philosophy, University of California, Davis, CA, USA
271

The Author(s) 2016


K. Aizawa, C. Gillett (eds.), Scientific Composition
and Metaphysical Ground, DOI10.1057/978-1-137-56216-6_10

272

A. Ney

often their introduction is motivated by the argument that other notions


metaphysicians use in order to frame their positions are inadequate for the
task of characterizing the important metaphysical issues. Formulations of
metaphysical problems and views in terms of existence, quantification,
and modal notions should be replaced (Fine, Schaffer) or supplemented
(Audi, Rosen, Sider) with formulations in terms of these new distinctively metaphysical notions.
Schaffer is especially direct. He complains that contemporary metaphysics, insofar as it has been inspired by the Quinean task [of determining what exists], has confused itself with trivialities (2009, p.361). This
confusion about what the important issues are is tied to not having the
conceptual tools to represent the issues that matter. The deep questions are
not whether there are such things, but how. We want to know not what
exists, but what is grounded in what. Sider, in his Writing the Book of the
World, does not dismiss the importance and challenge of settling existence
questions, but similarly emphasizes that this is not what metaphysics at bottom is about. It is instead about finding out how the world fundamentally
is, as opposed to how we ordinarily speak or think about it (2011, p.1,
my emphasis). Fine (2001, 2009, 2012) and Rosen (2010) too argue that
the important metaphysical questions cannot be formulated using only
those representational tools accepted as legitimate by most contemporary
metaphysicians: physical, logical (including quantificational), mereological, and modal concepts.2 Part of the remedy, the prescribed metaphysical
fix, is to embrace the use of new primitive metaphysical concepts such as
that of the real, grounding, or metaphysical structure.
These suggestions have been met with mixed reactions in philosophical
circles. Some of those working on first-order metaphysical problems were
quick to see the utility of these notions. For example, Jamie Dreier noted
soon after the publication of Fines 2001 defense of unanalyzed grounding and reality concepts that these were precisely the concepts needed
to clarify long-standing disputes between noncognitivists and realists
in meta-ethics in the face of creeping minimalism. However, in the
philosophy of mind, these proposals have been met with skepticism.
A commonly voiced complaint over the past decade, one heard frequently
2
Why have metaphysicians limited themselves to physical, logical, mereological, and modal
notions? Daniel Nolan discusses this issue in his (2014).

10

Grounding inthePhilosophy ofMind: A Defense

273

in conversation if not so often in print (though see Jessica Wilsons


(2014)), is that these metaphysical concepts are philosophically superfluous; they add nothing to the concepts philosophers of mind have had in
their toolboxes for years. Even if Fine, Schaffer, and the others are correct
that some metaphysicians have neglected issues of ontological priority
and fundamentality, this is far from the case in the philosophy of mind
where the central issue, the mindbody problem, is not one about existence, but rather ontological dependence. Philosophers of mind rarely
debate whether we have minds or mental states, whether many claims in
psychology are true. Rather, the primary issue concerns the minds relation to a more fundamental physical reality. Over the past several decades,
philosophers of mind and science have worked hard to distinguish and
make precise many notions (of reduction, supervenience, realization, and
emergence), positions (such as reductive and nonreductive versions of
physicalism, functionalism, and various versions of emergentism), even
principles governing what is real (such as Alexanders Dictum3) that serve
to characterize the relevant notions of ontological priority and fundamentality. To say that we need to introduce new concepts, new primitive
notions of fundamentality or grounding, in order to characterize these
issues, one must somehow be ignorant of these developments, or worse,
willfully neglecting this important work.
Now I believe this reaction to the proponents of primitive grounding and fundamentality notions is natural and to be expected, especially
given the rhetorical choices of those who defend grounding. However,
my goal here is to demonstrate some important uses for these metaphysical notions in the philosophy of mind and why I believe they add something extremely useful to the discussion.4
Confusions about the proper use of fundamentality and grounding
notions in metaphysics in general and the metaphysics of mind in particular come from several sources. One is, as I have already suggested, the
way that proponents of grounding have sometimes chosen to characterize
what is motivating the introduction of these concepts. These metaphysical
3

This is a principle frequently appealed to by Jaegwon Kim and others in the metaphysics of mind.
It says that for something to be real it must possess causal powers.
4
In other work, I apply these resources to debates in the philosophy of causation and mental
causation.

274

A. Ney

revisionists emphasize the distinctive virtues of using these notions over


the concepts of supervenience, necessitation, reduction via conceptual
analysis, and so on. This has (quite naturally) led philosophers of mind
to believe that grounding is being proposed as a replacement for not only
these, but also all of the other notions that are often appealed to in the
philosophy of mind such as identity and realization. This, I will show, is
not necessarily the case. The grounding framework gives us more resources
not fewer and it has room to incorporate those that have already been
developed (even if this is not usually acknowledged).
Second, different philosophers have proposed different concepts using
the common term grounding. Since the differences between these proposals are subtle, the nuances are often neglected and a vague or obscure notion
of grounding is quickly rejected without attention to the individual virtues
of the different proposals. I will argue that Fines framework has distinctive
advantages but to see this it needs to be carefully teased apart from the others. As I hope to show, Fines framework may be useful as a foundation for
developing an approach to the mindbody problem that can resolve and
clarify debates. I hope to show that by utilizing Fines distinctions, we are
able to offer novel, conciliatory positions allowing us to move past some
debates that have been carrying on in the philosophy of mind for decades.

Grounding andAnti-Realism
In order to see what may be added by an appeal to grounding, lets start
by noting something about the metaphysical relations that are typically
discussed in the philosophy of mind such as type- or token-identity, constitution, and property- or event-realization.5 We are apt to talk about the
obtaining of one of these relations when we have some entities (or types
or ways of conceiving some entities or types) that are already assumed to
exist. Our interest isnt in whether or not these entities or types exist, but
rather in characterizing the metaphysical relationship of one to the other.
5

Of course, there exists a diverse variety of ways of understanding the constitution and realization
relations. The differences among them will not matter for what follows. Note that I will not discuss
supervenience and necessitation as these notions have been widely recognized for years in the philosophy of mind to be insufficient to characterize the sense in which mental phenomena may be
ontologically dependent on physical phenomena. See Kim (1984) and Wilson (2005).

10

Grounding inthePhilosophy ofMind: A Defense

275

A common view about grounding is that it is some kind of primitive


relation among entities that is something like identity, constitution, or
realization but more abstract, a sort of generalized ontological dependence or priority relation.6 This is true on some conceptions of grounding, for example, Jonathan Schaffers and the related building-relations
discussed by Karen Bennett (2011), but this is not so for all proposals.
For example, Fine introduces grounding (and his notion of reality) not
so that we may better understand cases in which we have some entities
each of which is antecedently assumed to exist. Rather, he introduces
grounding to deal with situations in metaphysics in which one wants to
deny the existence of a class of entities. The central topic of Fines 2001
paper introducing grounding was how philosophers who wish to endorse
anti-realist positions about various domains can have overall consistent
views. This emphasis is continued in his more recent work:
If [the anti-realist] wishes to deny the reality of the mental, for example,
then he must explain or explain away the appearance of the mentalThe
question now is: how is this explanatory challenge to be construed? What
is it to explain the appearance of a world with minds in terms of a mindless
world or the appearance of a world with value in terms of a purely naturalistic world? My own view is that what is required is that we somehow
ground all of the facts which appear to presuppose the reality of the mental
or of value in terms of facts which do not presuppose their reality. (2012,
p.41, my emphasis in bold)

Here is one of the main examples Fines paper takes up. Suppose some
philosopher is a nominalist about abstract entities. Then it looks like she
will hold, as one of her main philosophical claims:
(1) Numbers do not exist.
And yet, even if she is a nominalist, it seems clear that she should not
want to deny something all of us accept, the simple mathematical fact
expressed by:

Here I am using dependence to indicate a relation such that when it obtains, it need not imply
that one of the relata is more fundamental than the other. (Ontological dependence is not an asymmetric relation.) When I speak of the obtaining of an ontological priority relation, what is prior is
thereby implied to be more fundamental than what it is prior to.

276

A. Ney

(2) There is a prime number between 2 and 5.


That is to say, she wont want to be a skeptic about mathematical
truth. But of course (2) trivially entails:
(3) There are numbers. (Numbers exist.)
And so, if the nominalist does not wish to be a skeptic about mathematical truth, it looks as if she is faced with an inconsistent set of beliefs.
Fine notes that this isnt a distinctive problem for nominalism but is a
general problem for various species of metaphysical anti-realism. The
moral noncognitivist will find her position in tension with the basic
moral truths she holds; the presentist will find her position in tension
with mundane past-tensed truths (that the North won the US Civil War,
that dinosaurs existed before humans); and so on.
Fine (2001) considers various resolutions to this problem, but for our
purposes it will only be necessary to present his favored solution which
gives the anti-realist who is not skeptical about mathematical truth a way
to have an overall consistent view. First, the non-skeptical nominalist will
not deny (2) since it states a basic mathematical fact and is a mathematical truth if anything is. Nonetheless it is just this, a mathematical truth.
That is to say, when the mathematician or student of mathematics states
such a fact, that there is a prime number between 2 and 5, we should recognize that she is not intending to assert something that is a particularly
deep metaphysical truth, a fact about what the metaphysical structure of
the world is and what kinds of entities are real or not real. And yet we
(who think metaphysics is a worthy task, who sometimes assert views
about such matters) sometimes make claims that do concern the deep
metaphysical structure of the world. Fine thus proposes the introduction of a primitive operator on sentences or propositions: In reality. This
operator is a component of sentences or propositions7 that are intended
to state claims reflecting fundamental metaphysics. And of course, this is

Fines official view is that this operator (and the grounding operator to be described shortly)
should be taken to apply to sentences (2012, p. 46), but he sometimes speaks of propositions
grounding other propositions. I will sometimes speak loosely as well of facts grounding other facts.
This should be understood as indicating the grounding of a sentence describing one fact in some
sentences describing some other facts.

10

Grounding inthePhilosophy ofMind: A Defense

277

the sort of claim the nominalist intends. Recognizing this, we can say that
really there is no tension between the nominalists main claim and (2) (or
even (3)). For what the nominalist intends to assert is not (1) that numbers do not exist, but rather that in reality, numbers do not exist.8 And as
long as we do not conflate this reality locution with quantificational idioms like there exists, we can see that there is no contradiction between:
(4) In reality, numbers do not exist, and:
(3) There are numbers. (Numbers exist.)
Thus, the tension introduced by non-skeptical anti-realism is resolved.
To this, some have objected that they dont have a grip on what Fine
means by in reality.9 As noted, the reality operator is officially introduced as a primitive, but this shouldnt lead one to worry we have no grip
on what it adds to a sentence. Fine gives a positive characterization of it
as follows:
One might think of the world and of the propositions by which the world
is described as each having its own intrinsic structure; and a proposition
will then describe how things are in themselves when its structure corresponds to the structure of the world. Thus it is this positive idea of the
intrinsic structure of reality that should be taken to inform the relevant
conception of what is fundamental or real. (2001, p.25)

The real propositions are those that describe the intrinsic structure of
reality. This isnt to say that propositions that do not describe the intrinsic structure of reality may not be true. They of course may be. But it is
the goal of the metaphysician (at least some of the time) to make claims
that are not just true but also real, that do describe the intrinsic structure
of reality.
Some will complain that this positive characterization of the reality
operator doesnt address the concern since it relies on a further esoteric
notion, the intrinsic structure of reality. Since it is difficult or impossible

See Fine (2009) for an explanation of the distinction.


This is a critique raised by Thomas Hofweber (2009), Chris Daly (2012), and (preemptively)
Carnap (1950).
9

278

A. Ney

to provide a definition of these metaphysical notions, perhaps it is worth


providing a simple example to try to better capture what a non-skeptical
anti-realism is supposed to look like. If one can understand how a nonskeptical anti-realism looks when applied to a mundane case in which we
philosophers dont already have theoretical commitments, perhaps it will
be clearer how the framework may be implemented to state new and coherent positions in the philosophy of mathematics or mindbody debate.
Consider any ordinary situation in which some person sincerely and
with good (external) reasons asserts:
(5) There is a mess in the kitchen.
In that situation, must we assume the speaker intends to make a deep
metaphysical claim? A claim that there are in reality such things as messes
and that one of these messes is located in the kitchen? Of course not.10
And yet even if the claim is not intended to track metaphysical structure,
it can still be perfectly true. Ive asked you after all to imagine such a
situation, one in which the claim is true. I submit this isnt very puzzling.
A non-skeptical anti-realist could give many alternative accounts of mess
talk, ways for (5) to be reasonable to assert, indeed true, while there are
no such things in reality as messes. Here are three such accounts. First,
perhaps There is a mess is just an idiomatic way of saying that things
are arranged in a way someone does not like, so that the word mess
is somewhat like the word sake, a noun that contributes meaning to
a sentence while always lacking denotation. Another possibility is that
there are messes, but whether something is a mess is a subjective matter.
Messes arent objective existents. Rather what is a mess depends on what
most normal people are apt to consider a mess. So, although a speaker of
(5) may say something true, it will not be something that describes how
things are in reality. When we say that there is something in reality, we
mean after all that there is something that exists in the mind-independent
world, not merely from the point of view of one perspective or another.
But even if there are objective facts about when There is a mess is true,
this still wouldnt entail that a speaker of (5) expresses the claim that in
10
One indication of this is that one could have expressed the same thing by saying instead of There
is a mess in the kitchen, The kitchen is a mess. If she was trying to express something about the
deep metaphysics of the situation, these would not be equivalent.

10

Grounding inthePhilosophy ofMind: A Defense

279

reality, there are such things as messes and one of them is in the kitchen.
Lets pause a moment to see a third account according to which (5) may
express an objective truth and yet still not be real.11
On a functionalist understanding of mess, a claim like (5) is true
when there are some things or other in the kitchen capable of playing
a certain causal role, whatever is the causal role associated with our concept, mess. Ill tentatively work with: being a collection of things that
persists in a location without good reason that is apt in the circumstances
to cause obstruction and annoyance. Suppose in our imagined situation
what plays that role is a pile of dirty dishes in the sink. Functionalists may
take different approaches when they consider (5)s connection to reality
in this situation. One approach would be to adopt a functionalist realism
about messes. Then one will say that in reality there are messes, one such
mess is the pile of dishes in the sink, and since the sink is in the kitchen,
the sentence expresses a truth about what there is in reality.12 However,
one might be concerned about this approach for several reasons. One
is that the concept of mess permits multiple realization and so it seems
wrong to think that messes just are piles of dishes in sinks. Ill however
focus on another reason for rejecting this functionalist realism about
messes. This is that a pile of dishes in a sink is only properly counted
as a mess in a particular context, namely one in which the dishes are
there without good reason and apt to cause annoyance and obstruction.
In other circumstances, such as when one places a pile of dishes in a sink
in order to promptly clean them, a pile of dishes doesnt count as a mess
but a means to an end. So it isnt right to think of the pile of dishes itself
as the mess. Piles of dishes arent the right kinds of things on their own
to be messes. Nor would it be correct to think of the larger mereological
sum consisting of the pile of dishes, the sink, and all of the things apt to
be annoyed orobstructed in the situation as the mess. Thats not a mess
11

In outlining this third way for how it may be that a sentence is true, yet not true in reality, I am
departing from Fines official view. Fines descriptions of cases involving sentences that are not true
in reality generally involve subjectivity such as we find in the first two accounts above. However, as
I will now argue, this third way also constitutes a way in which a sentence may be true while not
correctly describing the intrinsic structure of a given situation.
12
This would be to endorse what is usually called an occupant or realizer functionalism about
messes. Messes are the things that occupy the mess-role.

280

A. Ney

either. (Anyway, if that were the right account, it would make (5) false,
since that object is too big to be in the kitchen.) Better, one sympathetic
to a functionalist approach to (5) should deny that (5)s truth depends
on the existence of any one kind of thing, a mess, but rather depends
on a particular kind of situation being instantiated.13 There being a mess
amounts to a situation that may involve not only dishes but also a variety
of other kinds of objects, but only in the larger circumstance in which the
objects play the causal role associated with the concept of mess. This is
an account in which the truth of the sentence (5) is objective in the sense
that its truth does not depend in any way on someones perspective. It
depends merely on what kinds of things there are in the world and how
they are arranged. But the sentence is still not true in reality because interpreted as a claim about what there is in reality, it would make a false claim
that in reality there are such things as messes that are located in kitchens.
According to this anti-realist functionalist account, sentences like (5) latch
onto the world in a more complicated way, by referring to a more spatially
extended situation, a causal network. Is there a mess in the kitchen? Yes,
but not because in reality there are messes and one of them is located in
the kitchen. Rather, there is a mess in the kitchen because there are in
reality many kinds of things, dishes and sinks and people, interacting in
the right way to make this sentence true.
So there are many ways it could turn out that when someone asserts
(5), they are not making a claim accurately tracking the kinds of things
there are in reality. This isnt to say that one is thereby speaking figuratively or not expressing a fact or saying something that isnt true, interesting, or justified. Not all assertions, not even all true, justified, and
interesting assertions need to mark out the kinds of things there are in
reality in a way that would interest a metaphysician. I hope Ive indicated
some ways this could work out for the everyday case in which one says
someone has made a mess.
We can now see where the notion of grounding enters in Fines framework. It is precisely here, to show how those sentences that are not tracking
the intrinsic structure of reality may yet be true. Formally, grounding is a
13

A good predecessor to what I am suggesting here is Ryles (1949) discussion of category


mistakes.

10

Grounding inthePhilosophy ofMind: A Defense

281

two-place sentential operator acting on a sequence of sentences <Q, R, >


(the grounds) and a target sentence P (the grounded) where: <Q, R, >
grounds P. Although again Fine does not analyze this notion, offering it
rather as an ideological primitive, he clarifies the notion thus:
If the truth that P is grounded in other truths, then they account for its
truth; Ps being the case holds in virtue of the other truths being the case.
(2001, p.15)

To say that some true sentences may not themselves be real, yet nonetheless be grounded in the real is to say that while these truths do not
themselves track metaphysical structure, they have an explanatory basis
in truths that do. The details of this basis might itself be complicated
(involving facts about individual perspectives or, as Ive argued, causal
networks). But ultimately a true statement will have a set of facts that
explains its connection to reality. Once these are provided, there is no
longer any explanatory gap left over regarding why the grounded sentence is true, or why the fact it describes obtains.
In addition to the primitive notions of reality and ground, Fine also
introduces a third notion that is defined in terms of the notions of the
real and ground. This will be useful in what follows. It is the notion of
reduction:
The true proposition P reduces to the propositions Q, R, iff (i) P is not
real; (ii) P is grounded in Q, R, ; and (iii) each of Q, R, is either real
or grounded in what is real. (2001, p.26)

Fine argues that this definition of reduction is superior to those that


have been proposed previously. In particular, it is superior to accounts
of reduction in terms of supervenience or other modal notions in that it
is explicit in this account that what is reduced is not real (2001, p.11).
Here, it is worth emphasizing again that Fine is interested in a notion
of reduction that will be useful for formulating anti-realist positions.
I started this section by acknowledging that in many debates in the philosophy of mind, the assumption is that we do not want to be anti-realists
or eliminativists about the mental. Mental phenomena are assumed to be

282

A. Ney

real, the question is rather what relation they bear to physical phenomena. For example, in a situation in which Tom is in pain, the reductionist,
nonreductive physicalist, and emergentist will typically all agree the mental state is real. Their debate concerns rather whether Toms being in pain
is identical to a physical state, is realized by, but not identical to a physical
state, or is instead caused by rather than constituted by a physical state
of Toms. In a debate that has this form, we may note that Fines notion
of reduction will not be particularly useful since it doesnt capture any of
these three options. As one might already guess, I do not agree that the
debate should be understood in this way, as limited to a choice between
these three options. One of the main points I want to make is that the
grounding framework gives us a way of framing views on the mindbody
problem that are (in at least some domains) more reasonable than those
that have previously been articulated. We will come back to this when we
examine the case of phenomenal and other psychological states in more
detail momentarily.
As a final exegetical point, note that this framework and all that has
been said up until now leaves open the possibility that a sentence may be
true, grounded in other sentences, and yet itself be real. This would be
a case of grounding without reduction, where what is grounded reflects
metaphysical structure as well as its grounds. The distinction between the
case of grounding with reduction and grounding without reduction will
play a role in the applications below.

Grounding Withor Without Reduction


Although Fines framework makes room for cases of grounding without
reduction, where the grounded is real just as the ground, this situation is
treated in his work as something of an obscurity, brought up mainly to
solve a puzzle that might arise in cases of infinite mereological (or otherwise explanatory) descent. Indeed, Fine is explicit that with grounding
there is a presumption that the grounded is not real (2001, p.27). This
is a place where we can see a clear difference between Fines (as well as
Rosens) grounding framework and those of others, for example of Audi
(2012) and Schaffer (2009).

10

Grounding inthePhilosophy ofMind: A Defense

283

Suppose again we are discussing the status of the fact that Tom is in
pain. In Audis framework, if this fact is grounded in some fact about
physical states (say that Toms C-fibers are firing), this is to say that there
is a kind of noncausal determination relation obtaining between these
facts, one that arises due to an essential connection between the properties that constitute these facts. In this way of thinking about grounding, grounding isnt even compatible with reduction in Fines sense. Pain
has to be real for there to be an essential connection between it and the
physical property figuring in the grounding claim. In Schaffers framework as well, we find that where a grounding relation obtains, there is
no presumption that the grounded is not real. Instead for Schaffer, quite
the opposite, anything that is grounded must be real. Grounding is a
relation that obtains, not like for Audi between facts or for Fine between
sentences, but instead between entities of any ontological category. So,
Schaffer might speak of Toms pain being grounded in some physical feature of Toms brain or body. To say for some entity that it is grounded is
just to say that it has the status of a derivative entity (2009, p.373), which
entails that it is an entity and hence real. I propose that it is an advantage
of the Finean framework that it does not have this consequence of Audi
and Schaffers frameworks that what is grounded is automatically real or
an element of ones ontology.
But why? Why should it be an advantage of the account rather than a
cost that it permits this flexibility? Wilson (2014, pp.244248) has argued
this is a liability for grounding approachesthere is what she calls the
metaphysical underdetermination problem for theories of grounding. This
is that knowing a grounding fact obtains leaves completely open all of the
interesting questions we care about when we raise questions of existence,
ontological dependence, priority, and fundamentality. This is a problem
because if the argument for introducing a new grounding primitive was
that the logical, mereological, and modal relations metaphysicians were
previously using were inadequate to capture the metaphysical relations
between (say) mental phenomena and physical phenomena or mathematical phenomena and observable phenomena, then an appeal to grounding
doesnt offer an improvement in this respect and so is unmotivated.
This is an important worry. To respond, we may start by noting that
nothing in the framework I have discussed here suggests that a bare

284

A. Ney

appeal to grounding can answer the question of the precise nature of the
metaphysical relation between (for example) mental and physical phenomena by itself, nor even questions about their existence. When we just
say a truth is grounded, we do not say what its grounds are. But although
simply saying a truth is grounded wont answer all of the metaphysical
questions that interest us (and so close the explanatory gaps a grounding
claim is supposed to close, according to Fine), stating what those grounds
are will. To see this return to:
(7) Tom is in pain.
Here are four proposals for the grounds of (7):
(8) Toms C-fibers are firing.
(9) Toms C-fibers are firing. The firing of ones C-fibers is typically
caused by tissue damage and typically causes withdrawal behavior. One
is in pain if one is in the kind of state that in the relevant circumstances
is typically caused by tissue damage and typically causes withdrawal
behavior.
(10) Tom is in an internal state of the kind that in the circumstances
is typically caused by tissue damage and typically causes withdrawal
behavior. One is in pain if one is in the kind of state that in the relevant
circumstances is typically caused by tissue damage and typically causes
withdrawal behavior.
(11) Tom believes he is in pain.
We may also consider a fifth possibility that while (7) is true and real,
the event it describes is caused by the event described by (8), but (7) is
not grounded in anything.14
What we see here are different candidate grounds for (7) (or the denial
of a ground altogether) that correspond to different ways of answering the question of the relation between pain facts and physical facts.
These correspond to five canonical views in the metaphysics of mind:
(brute, i.e., nonfunctionalist) type identity theory, occupant functionalism, causal role functionalism, subjectivism, and emergentist dualism. This of course by no means exhausts the range of available options.
14

A complication arises here in that (7) refers to Tom, a human being, and that we should not think
of facts about human beings as generally ungrounded. Lets postpone this issue and just ask the
question of whether (7) is grounded relative to its ascription of pain.

10

Grounding inthePhilosophy ofMind: A Defense

285

But it suffices to demonstrate how even though merely saying that something is grounded does nothing to eliminate metaphysical underdetermination, saying in what it is grounded will.
The question that next naturally arises and is indeed pressed by Wilson
is then why the concept of grounding is needed when we already have at
our disposal these concepts of identity, realization, causation, as well as
mereological notions.15 Cant we accommodate all of the options mentioned above without also using the notion of ground? We may first
remark that what we have in effect shown is how (using the grounding
framework), we may bypass any explicit mention of identity, realization,
etc. while still seeing a diversity of metaphysical options via the variety of
grounds possible for (7). But the more important point is again to insist
on being careful about which grounding framework we are considering.
For some frameworks (e.g. Schaffers), Wilsons concern would be justified. We may dispense with the grounding notion in favor of identity,
realization, and the like (assuming we also have available a way of saying
which entities are real or fundamental). Grounding is just a less specific
way of describing the ontological dependence of some entities on some
other entities. But Fine is interested as we have seen in accommodating a
form of non-skeptical anti-realism. In the Tom case, this would amount
to a view according to which (7) is true, but not real. The view is antirealist in the sense that its proponent is denying the reality of mental states
without claiming that sentences like (7) are false. A view like that cannot
be accommodated using the frameworks of identity, realization, or emergence. The view rejects the existence of identity, realization, causal, and
mereological relations between pain and physical phenomena, because it
rejects the reality of mental phenomena. Yet, it is not the eliminativism
of Paul or Patricia Churchland (1981, 1986) either. According to their
eliminativism, it is not just that mental states are not real, but that all psychological claims that would appear to be about them are false. Using the
grounding framework, we can say that many psychological statements are
true even though they are not real, so long as they are grounded in what is
real. This is accommodated because grounding is not a relation between
15

There is a question about whether we need the concept of real as well. Wilson allows that we need
at least something like this, a concept of fundamentality.

286

A. Ney

entities that must exist (like identity, realization, mereological relations,


etc.) but rather an operator acting on sentences, sentences that may or
may not have constituents corresponding to features of the world. The
next sections will show why accommodating this as a coherent position
will be useful for resolving long-standing debates.
We can now answer the question raised at the beginning of this section: why is it a good thing that the grounding framework permits different accounts, some according to which what is grounded is real and some
according to which what is grounded is not real (cases of Finean reduction)?
The reason is because psychological and other statements vary in the way
they track metaphysical structure. When we see the grounds, we can see in
virtue of what a given sentence is true. And then by examining these particular grounds, we can tell whether or not what is grounded is real or not.
We have now introduced enough of the framework to be able to see
how those who worry that grounding (at least in Fines sense) is just a
vague way of getting at the ontological dependence relations philosophers of mind have been discussing for decades are misunderstanding the
proposal. Because ground is an operator on entire sentences, not individual entities, it plays a different role than most of these notions (certainly
identity, realization, and causation, which link an entity or entities). It
allows us to discuss cases in which the target sentence uses noun phrases
that do not correspond to genuine ontology. This opens up an expanded
range of positions about a topic of discourse. In the next sections, we will
focus on the case of psychology.

The Special Sciences Debate


What I want to suggest is that some debates in the philosophy of mind
can be resolved (not merely clarified) by appeal to this framework. The
debate I will discuss is the one that has essentially determined the main
divide we now see in the field between, on the one hand, reductive physicalists (reductionists), and, on the other, nonreductive physicalists, particularly functionalists. In his paper, Special Sciences: Or, the Disunity
of Science as a Working Hypothesis, Jerry Fodor argued that the special sciences (psychology, biology, economics, etc.) were in an important sense autonomous from fundamental physics. Although the subject

10

Grounding inthePhilosophy ofMind: A Defense

287

matter of the special sciences consists entirely of physical things of one


sort or another, the explanations provided by the special sciences cite
distinctive properties not reducible to physical properties (1974, p.103).
If we consider any special science lawFodors example in the paper was
Greshams law, the law that good money drives out bad moneyit will
be capable of covering a physically diverse and heterogeneous variety of
objects: silver or copper coins, strings of wampum, and someones writing a check. Fodor draws on Putnams earlier (1967) point about multiple realization. Although each instantiation of a special science law will
involve the instantiation of a physical type, this type will vary from one
instantiation to another. This shows, according to Fodor that [n]ot all
natural kinds (not all the classes of things and events about which there
are important, counterfactual-supporting generalizations to make) are, or
correspond to, physical natural kinds (1974, p.113). Applying this point
to the case of pain, Fodors claim is that when there are true, justified,
and important psychological laws involving psychological predicates like
pain, since these laws may be instantiated by creatures possessing a heterogeneous variety of physiologies, we cannot identify this psychological
kind with any particular physical kind.16 At best, we might try to say that
pain is identical to some wild disjunction of physical kinds. But such a
heterogeneous and unsystematic disjunction (1974, p.108) would not
be the sort of kind that would appear in our scientific theories and so is
not the sort of kind in which we should believe.
One influential critique of Fodors position was given by Jaegwon Kim
who argued that Putnam and Fodors claims about the heterogeneous
multiple realization of special science kinds actually undermines the
possibility of our possessing genuine laws in the special sciences. Kim
asked the following question:
If pain is nomically equivalent to [the property of possessing one of the
physical realizers of pain], the property claimed to be wildly disjunctive
and obviously non-nomic, why isnt pain itself equally heterogeneous and
non-nomic as a kind? (1992, p.323)
16

And recall Putnam (1967) on pain: Consider what the brain-state theorist has to do to make
good his claims. He has to specify a physical-chemical state such that any organism (not just a
mammal) is in pain if and only if (a) it possesses a brain of a suitable physical chemical structure;
and (b) its brain is in that physical-chemical state.

288

A. Ney

Kim presents this concern as an issue about the projectibility of these


irreducible special science kinds. As Kim points out, for concepts to be
useful in science, they should denote kinds whose instantiations lead to
stable behavior. For any scientific kind K, we should expect the existence
of some true generalizations saying that if K is instantiated, then some
particular kind of behavior will follow. This is what it means to say that
K is projectible. Kim raises the point that if special science kinds are
not realized by fairly univocal physical kinds but instead may be realized
on different occasions by varied and heterogeneous sorts of underlying
physical processes (and so are nomically equivalent to wildly disjunctive
physical kinds), then this threatens their projectibility. For then we cannot expect all instantiations of the kind to lead to similar behavior. What
behavior will result will depend on the specific realization we find on
that occasion. And if this is right, it threatens the putative special science
laws ability to support counterfactuals. Kim thus argues that we should
believe that in any science, fundamental or otherwise, the concepts that
are employed should be such as to pick out univocal physical kinds. This
raises a skeptical worry about the physically irreducible kinds Fodor says
that psychology and the other special sciences describe.
Another worry Kim famously raised for Fodors position is a metaphysical worry concerning causal overdetermination. In Mind in a
Physical World and elsewhere, Kim suggests that the irreducible special
science kinds posited by Fodor would at best be epiphenomena since on
any given occasion (assuming physicalism, a position Fodor explicitly
endorses) there will always be a complete physical causal explanation for
any occurrence. There doesnt seem to be room then for Fodors irreducible special science kinds to have any causal impact on what happens, and
so positing them seems explanatorily superfluous. This is of course what
is commonly known as the causal exclusion problem for nonreductive
physicalism.
Kims proposal in the end is not to reject that we may use psychological
predicates like pain to refer to genuine kinds. Rather, he simply insists
that we must be using these terms to refer to physical kinds with univocal
causal profiles. If what we call pain in humans is quite different physiologically from what we call pain in mollusks so that we cant speak of
a common physical type between them, then humans and mollusks will

10

Grounding inthePhilosophy ofMind: A Defense

289

not instantiate any common psychological kind and the psychological


laws for humans will differ from the psychological laws for mollusks.
What resulted and has continued to this day is a disagreement between
reductionists like Kim and nonreductive physicalists like Fodor. My claim
is that the grounding framework can let us state a view that lets both sides
have at least most of what they want (and all of what was reasonable in
the two positions). We start by assuming what neither party here denies
(since both are physicalists): that there are some fundamental physical
features and so a set of real claims about the instantiation of these features. The solution comes when we are able to recognize that the two parties are arguing past each other because they aim to capture different sort
of facts. On the one side (Kims) are those who are predominantly interested in issues of metaphysicswhat kinds of entities are real, what kinds
of properties do they have, are there causal relations and if so between
which entities do these relations obtain? On the other side are those (like
Fodor) who are predominantly interested in something elseestablishing certain claims as true or explanatory in a given scientific context. If
we allow the possibility of a non-skeptical anti-realism of the kind outlined above, then we do not need to reinterpret psychological claims that
appear to be tracking diverse realizations as claims about physical kinds
to make sense of how they may be true. But nor need we see them as
referring to irreducible special science kinds. We may adopt a view that
is metaphysically reductionist, one denying the reality of special science
kinds, while allowing the truth of special science claims.
So first lets be explicit how this gets Fodor what he wants. We can
immediately concede as he insists that many claims in psychology and
(perhaps) economics are true and factual, justified and important. This
being so, we must also say they are grounded in what is real. So to
understand what it is in virtue of which they are true (in the metaphysical
sense), we have to understand these grounds. But given Fodors interest
in capturing special science laws in the sense they are intended, there is
no reason to say that these special science claims will themselves be real,
to say that they correspond directly to the intrinsic structure of reality. To
use one of Fodors own cases, to make a financial claim is not to attribute
an intrinsic feature to some piece of metal or paper, but instead to capture
a complex web of causal relations. Although some special science claims

290

A. Ney

may be intended to track the intrinsic structure of reality (as Ill discuss
below), the point of emphasizing the autonomy of the special sciences
from fundamental physics in Fodors work seems precisely to emphasize
that many psychological claims are not intended as claims of fundamental metaphysics, about what kinds of things there are, but rather intend to
track causal patterns that may be instantiated by a broad variety of things.
Fines framework allows us to say that even if this is so, the special science
claims in which Fodor is interested may be true. And this seems to be precisely what Fodor most wants. Greshams law, good money drives out bad
money, can be true. But for this to be so, the world need not be carved
up into little things that are money.17 As in the third account of mess talk
above, the truth of such claims may be explained by facts that do track
reality, but not what is suggested by the grammar of the financial claim.
If functionalism is correct, we may see monetary truths as grounded not
in facts about a particular ontic kind that is instantiated in wallets and
banks, but rather truths about a complex web of causal facts. Then we
should give up the claim that money is a kind of thing altogether. But
again, this doesnt undermine the truth of the claim, just its success at
reflecting metaphysical structure.
Seeing things this way also gives Kim most of what he wants. To say
that a special science claim is true and justified does not require saying its
predicates refer to real kinds. So it does not require we posit the existence
of additional higher level kinds any more than the truth of There is a
number between 2 and 5 (in any normal mathematical context) requires
the existence of numbers. As such, there is no threat of overdetermination
or epiphenomena. Special science claims, when true, will be grounded in
real claims (seeing how is an important first-order project in the metaphysics of science in which the various conceptual tools of realization and
constitution may be brought to bear), but to be so grounded does not
require the existence of potentially overdetermining higher-order kinds.

17

It is clear that many who have followed Fodor in adopting nonreductive physicalism want more,
want to say that many special sciences claims arent just true and justified but also that they refer to
additional higher-level special science kinds. However, there does not seem to be any justification
for this further ontological claim and there are reasons (those noted by Kim) against it.

10

Grounding inthePhilosophy ofMind: A Defense

291

Kims point about projectibility, on the other hand, is trickier. It is


not clear to me (and I dont think it has been clear to many) why projectibility should require that a special science concept denote a kind
with a univocal physical causal profile. One way to ensure that a claim
be projectible is for it to track a physical kind, but it is not clear why this
is a necessary condition. Scientists may succeed in tracking consistent
behavior, capturing interesting patterns, even where no underlying metaphysical unity in the objects can be found and we should allow for that.18
What we seem to have in such cases is disunity in the kinds of objects
there are and unity in broader causal patterns. Thus, while the framework
I am proposing eschews realism about the psychological kinds that may
appear to be the denotations of psychological predicates, it finds a way to
ground the truth (and projectibility) of claims involving these predicates
in real facts about causal patterns.
Just as it appears questionable whether Kim is correct that psychological claims that are projectible and support counterfactuals must refer to
univocal kinds, it is similarly doubtful whether Fodor is correct that psychological laws always make claims that track physically heterogeneous
sets of circumstances. As philosophers of neuroscience have been keen to
emphasize (Bechtel and Mundale 1999), psychological claims are diverse.
It is a virtue of the framework I am developing here that it has room to
distinguish the variety of psychological claims in a perspicuous way.
Let me say a bit more about what I mean. The reductionism/nonreductive physicalism debate is most of the time presented as if it is all or
nothing. One must either be a reductionist about all of psychology or a
nonreductive physicalist tout court. But psychological statements vary
in their connection to underlying neurophysiology and so functionalism (and nonreductive physicalism) may be more plausible in some cases
than in others. Recognizing this, some are content to be functionalists
about propositional attitudes like beliefs and desires, but reductionists
about phenomenal states (e.g. Ned Block). In the present framework,
this would amount to adopting an anti-realism about propositional attitude ascriptions, but a realism about phenomenal ascriptions. Since there

18

See Loewer (2009) for a discussion of this issue.

292

A. Ney

is nothing incoherent in this combination of positions, we see another


virtue of having an approach to the mindbody problem that does not
presuppose an answer yes or no to the question of whether a sentence that
is true and grounded must therefore be automatically real (or not real).
And even when we look more closely within the realm of phenomenal
or intentional phenomena, the metaphysical connection of psychological
statements to neurophysiology may vary.
Finally, to close, lets come back to the example of Tom is in pain and
see how the variety of grounds for
(7) Tom is in pain,
may suggest a form of realism or anti-realism along the lines defended
here. Above I presented four options for how one might understand the
grounds of (7).19 We may set aside (8) as both Fodor and Kim (and most
others) would deny that this suffices to close the explanatory gap between
the mental and physical. We may also set aside (11) for our purposes.
Although it is an interesting and important reductive option, typically
reductive and nonreductive physicalists are not happy to ground claims
about phenomenal states in claims merely about our beliefs.20 The disagreement, if there is one, rather would be between (9) and (10):
(9) Toms C-fibers are firing. The firing of ones C-fibers is typically
caused by tissue damage and typically causes withdrawal behavior. One
is in pain if one is in the kind of state that in the relevant circumstances
is typically caused by tissue damage and typically causes withdrawal
behavior.
(10) Tom is in an internal state of the kind that in the circumstances
is typically caused by tissue damage and typically causes withdrawal
behavior. One is in pain if one is in the kind of state that in the relevant
circumstances is typically caused by tissue damage and typically causes
withdrawal behavior.

19

Here, I am considering proposals for the full, as opposed to partial, grounds for (7). See the distinction in Fine (2012).
20
This represents a position in the ballpark of what is proposed in Dennett (1991). The fact that
this view shows how Dennetts position is able to accommodate true phenomenal claims is also a
virtue of the account, but one I do not have the space to explore here.

10

Grounding inthePhilosophy ofMind: A Defense

293

We may allow that (9) and (10) are both real and also that either may
in principle explain the truth of (7). But there is a question of which of
(9) or (10) gives the best explanation of (7) as it is asserted in a given
context. Which is the correct grounding explanation for (7) has ramifications for whether one should take a realist or anti-realist position about
pain. (9) does, while (10) does not, explain the truth of (7) in terms of
the existence of a particular kind of state that is a pain state.21 (10) reveals
pain talk to be metaphysically grounded not in the instantiation of a
particular kind of state, but rather a broad causal nexus. Those who take
(10) to be the correct view about what grounds (7) will thus (on the picture I have sketched here) adopt a non-skeptical anti-realism about pain.
This is what I have argued is the reasonable position if one accepts with
Fodor (and Putnam and the very many other nonreductive physicalists)
that psychological statements using the concept pain do not refer to a
homogenous physical kind, but accepts the metaphysical points of Kim
(and Lewis and other reductionists).

A Comment onReality
As a sidebar, it is worth acknowledging a bit of awkwardness in Fines
terminological framework as applied to the mindbody debate. In the
framework I am proposing, only psychological statements that track univocal physical kinds make claims about reality. Those that do not may be
true, justified, and important, but not real. Some have asked whether we
really want to say that in all cases in which functionalism is motivated,
that is, all attitude ascriptions or (if Fodor is right about pain) pain
ascriptions, these claims are not real. In every sense of real that matters to us for most but the most esoteric purposes, one can see how this
is really not very helpful. The use of not real here is too easily confused
with a way of rejecting a statement. After all, does a metaphysician really
want to offer the diagnosis that someones pain is not real?

21

Recall the discussion of various views about mess talk.

294

A. Ney

One response would be to note that the claim of nonreality is made as a


point about metaphysical structure only, nothing else. But here we might
try to avoid such confusions by seeking out alternative terminology. One
option would be to replace the word real with fundamental. However,
I can imagine similar complaints brought to bear. Fundamental is a
technical term used by metaphysicians, but it also has connotations to
the general public of relative importance and do we want to imply that
statements about pain are somehow less important than statements of
neurophysiology? It is better if we can sidestep that confusion. Anyway,
if fundamental means has no further explanation or brute, then this isnt
the word we are looking for. As already noted, to say a sentence is real
does not entail it is not grounded. To cite a simple example, conjunctions have grounding explanations in terms of their conjuncts, but many
would allow that a conjunction tracks reality no worse than its conjuncts
taken together.
Another option would be to replace real with intrinsic. Making this
move from real to intrinsic, we would remove the implication that a
pain ascription fails to be correct, or grounded in reality, or important.
But I am not sure if using intrinsic does any better at removing confusion than real. We wouldnt want to imply a claim that is intrinsic in
this sense may not be a claim about relations, for example, The mass
of the proton is greater than the mass of the electron does appear to
express a claim about reality. But we would have to be clear that the
intrinsic/real statements are those that make a claim about the properties or relations instantiated by the entities referred to by the sentence,
rather than describing how they are with respect to other things or individual perspectives or nothing at all. So in describing a relation between
two objects (as in The mass of the proton is greater than the mass of
the electron) we are thereby attributing an intrinsic feature to the pair
(cf. Lewis 1986).
Perhaps a more technical-sounding term like ontological would be
best. But we could raise further concerns about that too, since ontological fails to track all kinds of metaphysical distinctions there might be in
nature. In lieu of finding more satisfactory terminology, Ill continue to
use Fines real in the remainder of this chapter.

10

Grounding inthePhilosophy ofMind: A Defense

295

Other Frameworks
After all of this, one might still be wondering why we need to use this new
framework of grounding to make the distinctions Ive wanted to make
in this chapter, in particular in order to introduce the conciliatory position I discussed above. Several have asked why we cant simply say that
we should analyze psychological statements in terms of statements about
causal networks and leave it at that. Talk of analysis is something metaphysicians (and philosophers of mind) have long been comfortable with
and doesnt require introducing new terminology like real and ground.
This is prima facie a reasonable point, but there are many problems
with this approach that metaphysicians are by now well familiar with.
First, it is not possible in many cases to provide the relevant analyses.
But more generally, even if we had them, analyses dont tell us in virtue
of what in the world a claim is true, what our ontological commitments
ought to be if we accept that claim, only in what circumstances it is true.
This point was articulated years ago by William Alston in a paper critical of Quinean approaches to ontological commitment (Alston 1958).
Suppose a nominalist about universals wishes to allow that sentences like
Patience is a virtue may be true while denying the existence of universals. She may then analyze Patience is a virtue in terms of some sentence
that doesnt quantify over universals, something like Patient people are
virtuous people. The trouble Alston noted is that the result of the analysis is simply the claim that these two sentences mean the same thing. But
agreeing the sentences are semantically equivalent doesnt entail anything
about the reality or unreality of universals. Rather we then see the sentences assert the existence of patience as much as they assert the existence
of patient people. What Alston argued, and what Fine and Rosen and I
are pressing, is that if one wants to say that Xs are not real, but Ys are,
and that the X-truths that are grounded in Y-truths, then one should just
come out and say this and use this language, since such metaphysical
claims cannot be replaced by talk of analyses.
One interesting view in many ways similar to what I am proposing
here but formulated without a grounding framework was developed in
2007 by Carl Gillett. This view, which he calls compositional reduction-

296

A. Ney

ism, also aims to reconcile the different insights of reductionism and nonreductive physicalism. The central idea of compositional reductionism is
that for reasons of ontological parsimony we should reject the existence
of psychological kinds that are not identical to causally univocal physical
kinds, but we should also allow that the claims made by psychologists are
not intended to track physically univocal kinds. Gilletts compositional
reductionist accommodates these points by saying that a sentence like my
Tom is in pain, may have associated with it two sets of truth conditions:
Tom is in pain is true iff Tom instantiates a particular physical type (say,
his C-fibers are firing).
Tom is in pain is true iff Tom instantiates the higher-order property of
instantiating a physical type that in the circumstances plays the pain-role.

Psychological statements like Tom is in pain may be true because they


satisfy something like the first set of truth conditions. They cannot satisfy
the second set of truth conditions according to the compositional reductionist, because there are no such things as higher-order properties. This
in many ways looks very similar to the non-skeptical anti-realism I have
proposed here. But there are problems.
The first problem with Gilletts strategy is similar to what was just
noted of the proposal to replace ground and reality talk with talk simply
of analyses. Merely stating truth conditions doesnt necessarily tell us in
virtue of what metaphysically a given claim is supposed to be true, rather
than stating a semantic equivalence.
But the starkest problem with this proposal is that although it is advertised as a way to reconcile the insights of reductionism and nonreductive physicalism, the position does not give the nonreductive physicalist
what he wants at all. And this is because it makes what psychologists say
consistently false. They are, according to the view, trying to assert claims
meeting the second set of truth conditions and yet, according to the
view, there are no such higher-order properties.22 According to Gillett,
22
In fairness to Gillett, he isnt defending compositional reductionism in his paper, only aiming to
set it out as an interesting view worthy of consideration. He presents the complaint I just made as
a puzzle that those who want to advocate the position would have to solve. I am arguing that it is
not a problem for the different view I propose here.

10

Grounding inthePhilosophy ofMind: A Defense

297

although we may use psychological statements like Tom is in pain so as


to track the first set of truth conditions, he agrees with Fodor that psychologists usually do not. Instead, Gillett argues they use psychological
sentences to express claims about the instantiation of irreducible psychological kinds. This means that the kinds of claims psychologists generally
make when asserting their hypotheses, making predictions and providing
explanations, are, according to the compositional reductionist, false.
Gillett is assuming that for a sentence like Tom is in pain to be true,
it must be tracking a kind, the referent of pain. What the grounding
framework gives us is a way of seeing how sentences may be true in virtue of how reality is structured without requiring that the true sentences
directly mirror this structure. The key point, the position I am defending, relies upon is that one truth (or set of truths) may explain another
even while demonstrating why the explananda may mislead as to realitys
structure.
Now, I should note that there are other metaphysical frameworks that
have been developed that are also able to accommodate something like
the position I articulate in this chapter and dont suffer the problems just
noted. For example, John Heil develops quite a similar position in Chap.
9 of his most recent book (2012). Heils approach differs from mine in
that he uses the framework of truthmaking rather than grounding to
make his point but the broad sketches of the two approaches are the
same.23 In the truthmaking framework, there is a basic set of facts about
genuine ontology and then a truthmaking relation is postulated to obtain
between this ontology and sentences (or other truthbearers). Unlike Heil,
I believe the grounding framework of Fine is superior to the truthmaking
framework. In my view, it is going to be important to see some truths
as grounded in what is not real and this is essentially nonsense in the
truthmaking framework. I have also been influenced by Fines critique of
truthmaking presented in his (2012).
Siders metaphysical framework using the fundamental notion of
structure to replace Fines real and the introduction of a concept of
a metaphysical semantics (distinct from that of a linguistic semantics)
to replace talk of grounding is also an interesting alternative framework
23

They were developed independently.

298

A. Ney

that might allow one to state the kind of non-skeptical anti-realist view
I defend here. I havent argued against Siders framework in this chapter either. My goal rather has been to argue that a desirable approach
to solving the mindbody problem should be able to capture situations
in which a sentence is true, its truth is grounded or made true by facts
about the world, and yet it misleads on matters of ontology. If one prefers to adopt Siders framework or perhaps the truthmaking framework
to accommodate this, fine. One is still thereby acknowledging that one
must move beyond the tools for presenting metaphysical positions that
philosophers of mind have traditionally allowed themselves. And this is
what needs to be recognized.

Conclusion
I hope to have shown here how at least one grounding framework may be
useful in the philosophy of mind, providing us especially with a range of
anti-realist views that do not reject the truth, factuality, importance, or
justification we have for claims in psychology. Psychological claims may
possess all of these honorifics without undermining the search for a unified, sparse, and nonredundant underlying metaphysics.24

References
Alston, W. (1958). Ontological commitments. Philosophical Studies, 9, 817.
Audi, P. (2012). Grounding: Toward a theory of the in-virtue-of relation. Journal
of Philosophy, 109, 685711.
Bechtel, W., & Mundale, J. (1999). Multiple realizability revisited: Linking
cognitive and neural states. Philosophy of Science, 66, 175207.
Bennett, K. (2011). Construction Area (No Hard Hat Required). Philosophical
Studies, 154, 79104.
Carnap, Rudolf. (1950). Empiricism, semantics, and ontology. In Meaning and
necessity. Chicago: University of Chicago Press.
24

I thank Ken Aizawa, Louise Antony, Jamin Asay, Carrie Figdor, Kit Fine, Carl Gillett, Jens
Harbecke, Kerry McKenzie, Kelly Trogdon, and especially Jessica Wilson for comments and criticism that led to substantial improvements of this chapter.

10

Grounding inthePhilosophy ofMind: A Defense

299

Churchland, P. (1981). Eliminative materialism and the propositional attitudes.


Journal of Philosophy, 78, 6790.
Churchland, P. (1986). Neurophilosophy. Cambridge: MIT Press.
Daly, C. (2012). Scepticism about grounding. In F. Correia & B. Schnieder
(Eds.), Metaphysical grounding (pp. 81100). Cambridge: Cambridge
University Press.
Dennett, D. (1991). Consciousness Explained. New York: Little, Brown, and Co.
Dreier, J. (2004). Meta-ethics and the problem of creeping minimalism.
Philosophical Perspectives, 18, 2344.
Fine, K. (2001). The question of realism. Philosophers Imprint, 1, 130.
Fine, K. (2009). The question of ontology. In D. Chalmers, D. Manley, &
R.Wasserman (Eds.), Metametaphysics. Oxford: Oxford University Press.
Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.),
Metaphysical grounding (pp.3780). Cambridge: Cambridge University Press.
Fodor, J. (1974). Special sciences. Synthese, 28, 97115.
Gillett, C. (2007). Understanding the new reductionism: The metaphysics of
science and compositional reduction. Journal of Philosophy, 104, 193216.
Heil, J. (2012). The Universe as we find it. Oxford: Oxford University Press.
Hofweber, T. (2009). Ambitious, yet modest, metaphysics. In D. Chalmers,
D.Manley, & R.Wasserman (Eds.), Metametaphysics (pp.260319). Oxford:
Oxford University Press.
Kim, J. (1984). Concepts of supervenience. Philosophy and phenomenological
research, 45, 153176.
Kim, J. (1992). Multiple realization and the metaphysics of reduction. Philosophy
and Phenomenological Research, 52, 126.
Kim, J. (1998). Mind in a physical World. Cambridge: MIT Press.
Lewis, D. (1986). On the plurality of Worlds. Oxford: Blackwell.
Loewer, B. (2009). Why is there anything except physics? Synthese, 170,
217233.
Nolan, D. (2014). Hyperintensional metaphysics. Philosophical Studies, 171,
149160.
Putnam, H. (1967). Psychological predicates. In W.H. Capitan & D.D. Merrill
(Eds.), Art, mind, and religion (pp. 3748). Pittsburgh: University of
Pittsburgh Press.
Rosen, G. (2010). Metaphysical dependence: Grounding and reduction. In
R.Hale & A.Hoffman (Eds.), Modality: metaphysics, logic, and epistemology
(pp.109136). Oxford: Oxford University Press.
Ryle, G. (1949/2002). The concept of mind. Chicago: University of Chicago
Press.

300

A. Ney

Schaffer, J. (2009). On what grounds what. In D. Chalmers, D. Manley, &


R. Wasserman (Eds.), Metametaphysics (pp. 347383). Oxford: Oxford
University Press.
Sider, T. (2011). Writing the book of the World. Oxford: Oxford University Press.
Wilson, J. M. (2005). Supervenience formulations of physicalism. Nos, 39(3),
426459.
Wilson, J. M. (2014). No work for a theory of grounding. Inquiry, 57, 145.
Witmer, G., Butchard, W., & Trogdon, K. (2005). Intrinsicality without naturalness. Philosophy and Phenomenological Research, 70, 326350.

Index

A
absolute fundamentality, 143, 157,
158, 195, 196, 196n17
activation experiments, 612
activities account of causation, 503,
81, 97
actualist-mechanist theory, 51
agnosticism, 174
Alexanders Dictum, 273
analytic metaphysics, 2, 3, 3n5, 10,
1718, 206n2, 209, 21113,
228, 235
angular head velocity (AHV), 95
anti-realism, 174, 27482, 285, 289,
291, 292, 296
anti-rationalism, 12338
anti-realists, 181, 275, 276, 278,
281, 285, 293, 298
functionalist account, 280

anti-reductionism, 12338
Appropriational accounts, 209,
238
assessing, 22132
manipulability-based and neoCausal accounts, 21417
philosophy of mind, standard
functionalist machinery,
21721
appropriation of machinery, 33
A-realization relations, 14n9

B
bazillion-mechanism theory, 107,
113
big-G Grounding relations, 171,
173n1
B-processes, 22931

The Author(s) 2016


K. Aizawa, C. Gillett (eds.), Scientic Composition
and Metaphysical Ground, DOI 10.1057/978-1-137-56216-6

301

302

Index

B-properties, 22930
broadly physical relation, 34, 2527,
261, 264

C
carving error problem, 29, 46
carving standard models, 48, 5364
cascade view of explanatory
mechanism, 656
causal composition relation, 152,
156, 160, 161, 173
causal explanation, 49, 50, 67n14, 69,
77, 77n2, 83, 85, 129, 145,
165n15, 175, 212, 215, 288
causal power, 124, 1367
dened, 131
forward-looking, 127, 1312
microphysical, 12930
token, 131n1,1323
causal preemption, 187, 190, 191
causal relation. See relations
causal standard models, 4853
causation, 12, 13, 19, 303, 48, 49,
137, 1458, 1525, 157n10,
1603, 1667, 178180,
1835, 187, 190, 214, 215,
262
activities account of causation,
503, 81, 97
error, 46
distinguished from Dimensioned
realization, 76
C-bers, 284, 292
classical mereological parthood
relation, 152
classical mereological partwhole
relation, 156, 171, 173, 191,
207, 208, 216, 222, 230, 232,
242

compositional explanation, 2, 10,


7589, 98n4, 206, 207, 210,
2224
compositional reductionism, 296
constitutive explanation, 2n3, 18,
212, 213, 215, 234
Constructive Engagement, 208
counterfactuals, 13, 19, 50, 51, 60,
155, 162, 165, 166, 181,
1847, 18991, 287, 288,
291
dependence, 33, 136, 137, 178,
1848, 189, 190

D
deductivenomological (DN) model
of explanation, 11, 42, 93, 95,
146n2, 154n9
deationists, 108, 182
dependence, 49, 51, 78, 130, 133,
135, 145, 154, 156, 160,
1636, 17980, 190, 194,
197, 201, 202, 233, 275n6
counterfactual, 33, 136, 137, 178,
1846, 189, 190
metaphysical, 145, 147, 148n3,
151, 152, 156, 1718, 186,
188, 189
ontological, 273, 275, 275n6,
283, 285, 286
depolarization, 45, 46, 48, 81
Descriptive Engagement, 208
determinable/determinate relations,
33, 132, 133, 151n1, 152,
154n9, 156, 171, 173,
177n4, 1801, 183,
18691
dialectical importance of scientic
composition, 1820

Index

dialectical import of Grounding,


1725
Dimensioned realization, 15, 7589,
125
distinguished from causation, 76
Grounding explanation and,
838
New Mechanism and, 7983
dispositional essential accounts, 184
divine command theory, 147, 148n3
dormitive virtues, 86
dorsal tegmental nucleus (DTN)
head-direction signal in, 95

E
eliminativism, 107, 114, 174, 285
eliminativists, 93, 108, 11517, 151,
281
emergentism, 174, 273
Engagement, 33, 2079, 217, 221,
238, 2434
phases of, 208
rules of, 210, 242, 244
entailment, 149, 172, 178, 253,
262n12
entity, entities, 81, 97
microphysical, 207
realism, 556
epiphenomenalism, 35, 174
exclusion problem, for nonreductive
physicalism, 12830
explanandum phenomenon, 11, 435,
47, 48, 57, 58, 85, 212
explanans, 11, 83, 85, 93, 212, 226,
234, 241
explanation
causal, 49, 50, 67n14, 69, 77,
77n2, 83, 85, 129, 145,
165n15, 175, 212, 215, 288

303

compositional, 2, 10, 7589,


98n4, 206, 207, 210, 2226
constitutive, 2n3, 18, 212, 213,
215, 234
deductivenomological model of,
11, 42, 93, 95, 146n2,
154n9
functional, 2n3, 206
mechanistic, 2n3, 11, 5464,
77n2, 96, 206n3, 214
metaphysical, 1456, 146n2,
162n13
non-causal, 756, 789
reductive, 2n3, 66, 206
unicationist model of, 93
explanatory constraints, formulating,
448
expressive power, 1956
extensional mereology (EM), 105
extensionality theorem of, 112

F
lament actin (F actin), 222, 223
Flat/Subset view of the realization,
15, 2369
formalism, 1536, 159, 161, 162,
166, 177, 213
structural equation modeling,
323, 79, 144, 150, 1627,
17881, 178n5, 18391, 201,
214, 2345
formulation of physicalism, 250, 251
consistency of grounding, 2608
dispensability of grounding,
25760
grounding and broadly physical,
2527
forward-looking causal power, 127,
131

304

Index

functional explanation, 2n3, 206


functionalism, 11, 14, 1415n9,
127, 21721, 273, 290, 291,
293
causal role, 284
homuncular, 219n10
occupant or realizer, 279n12, 284
functionalist atomist, 1945
functionalist inheritance, 1416
functionalist realism, 279
functional realization, 133, 152, 156,
171, 173
fundamentalism, 64
fundamentality, 273
absolute, 143, 157, 158, 195,
196, 196n17
primitive, 161, 192202
relative, 158, 159n12, 195, 196,
199202

G
General Extensional Mereology
(GEM), 100, 112
Global account, 212, 23, 24
globular actin (G actin), 222
good parts
as components, 549
as mutually manipulable, 5963
as scientically approved, 634
Greshams law, 287, 290
Grounding, 1n2, 2, 3n5, 1446,
171202, 2714
and anti-realism, 27482
and broadly physical relation, 34,
2527, 261, 264
consistency of, 2608
dialectical import of, 1725
dispensability of, 25760

explanation, and Dimensioned


realization, 838
frameworks, 21113, 2326
helpfulness of informative claims,
1516
informative claims, 14651
and narrowly physical relation, 34,
251, 252, 2546, 257, 2605
in philosophy of mind, 27198
primitive fundamentality
framework, 192202
priority argument for, 1912
reality, 2934
with/without reduction, 2826
relata of, 1756
rules of, 14367
special sciences debate, 28693
unity argument for, 17691
V-frameworks, 4, 5, 1719, 22,
25, 312

H
head direction cells (HD cells),
945, 102, 103
head direction representation
(HDR), 945, 1014
higher-order state-type, 2579
homuncular fallacy, 86, 88n10
homuncular functionalism, 219n10
hydrogen uoride (HF) molecule,
75, 76, 82, 84

I
identity
mechanistic constitution, 1024
material constitution of the
mental, 12338

Index

ideology, 176
inference, 137, 162, 165, 166
inaming, 160
informative claims, grounding and,
14651
interference experiments, 62
intervention, 60, 61
interventionism, 12, 50, 216
intrinsicness, 184

J
joint role-lling, 23942

L
lateral mammillary nuclei (LMN)
HD cells in, 95
Leibniz law, 105
levels standard models, 48, 649
Local account, 212, 23, 25
locality, 184
L-realization relations, 14n9

M
made-up-of relation, 31, 1236,
130, 1336, 138
manipulability relations, 5566,
21416, 22832
manipulationist theory, 98
mass-energy neutrality, 226, 227
material constitution
formal characterization of, 1245
and mechanistic constitution,
connections between,
11314
and mechanistic constitution,
dierences between, 10813

305

of the mental (see material


constitution of the mental)
mereology puzzle for, 106n10
solutions to grounding problem,
11416
material constitution of the mental,
12338
brute metaphysical necessitation,
1345
exclusion problem and
Donaldsons criticism,
12830
identity and, 1304
obscurity criticism, 12830,
1358
see also material constitution
mechanism, dened, 97
mechanistic constitution, 301,
91117
dened, 97102
head direction, representation of,
945
identity and, 1024
and material constitution,
connections between,
11314
and material constitution,
dierences between, 10813
question of, 934, 1048
regularity theory of, 92, 98, 101,
102, 108
solutions to material constitution
problems, 11416
mechanistic explanation, 2n3, 11,
5464, 77n2, 96, 206n3, 214
mechanistic explanatory framework,
434
mental state-token (m), 2579, 266,
267

306

Index

mereological atomist, 194


mereological nihilism, 107
mereological parthood, 177n4
mereological requirement, 100
mereology puzzle, for material
constitution, 106n10
metaphysical dependence, 171
metaphysical explanation, 1456,
146n2, 162n13
metaphysical framework, 297
metaphysical inquiry, 157
metaphysical structure, 271, 272,
276
metaphysicians, 5, 17, 76, 79n3, 89,
104, 146, 148, 151, 152, 156,
157, 160, 161, 172, 174, 175,
206n2, 250, 272, 273, 277,
280, 283, 2935
metaphysics, 30, 79, 93, 117, 137,
146, 20544
analytic, 2, 3, 3n5, 10, 1718,
206n2, 209, 21113, 228,
235
of mind, 1416, 2713, 275,
276, 284, 289, 290
of nature, 20710, 2434
of science, 1416
microphysical causal power, 124,
12930
microrealization, 125
microstructural analyses, 2n3
molecular explanations, of cellular
movement, 221, 222
monism, 107, 153n7, 154, 166, 178,
192, 193, 199n20
monists, 92, 106, 107, 111, 112,
11517, 153, 160, 193
moral realism, 147
M-realization relations, 1415n9

mutual manipulability (MM)


standard, 5963, 92, 215, 216,
230, 231

N
narrowly physical relation, 34, 251,
252, 2546, 257, 2605
necessitation, 30, 49, 124, 125,
1345, 232, 236, 241, 2524,
259, 260, 264, 274
neo-Causal approach, 1213, 22, 26,
28, 21415, 22832
neo-Causal research tradition, 2, 4,
12, 16
neo-Functionalist account, of vertical
relations, 35, 19, 26
neo-Functionalist V-frameworks,
1516, 22, 25
new mechanistic philosophy, 1112,
4171
carving standard, 5364
causal standard, 4953
Dimensioned realization, 7983
explanatory constraints,
formulating, 448
levels standard, 649
mechanistic explanatory
framework, 434
nihilism, 107, 154
nihilists, 115, 153
nomic necessitation, 253, 254
non-causal explanation, 756, 789
non-causal relations, 1n2
non-reductive physicalism, 174, 273,
290n17, 291, 292, 296
exclusion problem for, 12830,
288
nonreductivism, 124, 127, 130

Index

O
Ockhams razor principle, 1813,
185
omissions, 184
ontic representationalism, 241
Ontologically Unifying Power
(OUP), 206, 210, 2257, 231,
234, 240
ontology, ontological, 1213, 14n9,
29, 67, 77, 7981, 93, 98,
99, 99n7, 101, 109,
11115, 117, 175, 176,
182, 188, 189, 207n4, 219,
220, 226, 2389, 241, 243,
266, 273, 274n5, 275,
275n6, 283, 285, 286,
2948
overdetermination, 129, 130
redundant, 129
super-overdetermination, 130

P
particles
role in mechanism, 97
phenomenal models, 88n10
philosophers of mind, 273, 274
Philosophical Engagement, 208
philosophy of mind, 2, 11, 1416,
102, 209, 211, 243
Grounding in, 27198
standard functionalist machinery
of, 21721, 2368
philosophy of science, 2, 3, 1014,
19, 70, 77, 80, 93, 20810,
212, 214, 21820, 228, 235
manipulability-based and neocausal accounts, 21417

307

phototransduction, 77, 78
physicalism, 124, 250, 251
broadly physical relation, 34,
2527, 261, 264
consistency of grounding, 2608
dispensability of grounding,
25760
formulation of, 345
grounding and broadly physical,
2527
narrowly physical relation, 34,
251, 252, 2547, 2605
non-reductive, 12830, 174,
273
reductive, 174, 273
physicalist approach, 250, 251
physical state-token (p), 2579, 266,
267
piercing explanatory power (PEP),
30, 87n9, 206, 210, 219, 225,
23742
plausibility, 55, 56
pluralism, 107, 154, 193
pluralists, 92, 1057, 11117, 153,
156, 160, 193
Positivist(s)
Nagelian model of reduction, 10
philosophy of science, 1011
power
causal (see causal power)
expressive, 1956
Ontologically Unifying Power,
206, 210, 2257, 231, 234,
240
piercing explanatory power, 30,
87n9, 206, 210, 219, 225,
23742
powers-based accounts, 184

308

Index

primitive fundamentality framework,


192202
expressive power, 1956
priority in absence of fundamental
level, 1969
relative fundamentality, 199202
priority argument, for Grounding,
1912
priority relations, in absence of
fundamental level, 1969
production, 160, 184
projectibility, 288, 291
proper subset relation, 133, 152,
156, 173, 174, 182, 190
properties
role in mechanisms, 813

Q
qualitative distinctness, 86n8,
87n9, 23641
Quinean approaches, 272, 295

R
realism, 174
anti-realism, 174, 27482, 285,
289, 291, 292, 296
entity, 556
functionalist, 279
moral, 147
reality, 145, 146, 148, 163, 176,
182, 186, 188, 193, 198,
199, 263, 272, 273,
27581, 285, 289, 290,
2937
realization
A-realization, 14n9
Dimensioned, 15, 7589, 125

Flat/Subset view of, 15, 2369


functional, 133, 152, 156, 171,
173
L-realization, 14n9
microrealization, 125
M-realization, 1415n9
relations, 1, 2, 11, 1415
Subset/Flat view of, 220, 228, 237
reduction
grounding with/without, 2826
notion of, 281, 282
reductionism, 135, 291, 296
anti-reductionism, 12338
compositional, 296
reductive explanation, 2n3, 66, 206
reductive physicalism, 174, 273
redundant overdetermination, 129
Ref lective Engagement, 208
regularity or nomological suciency
accounts, 184
regularity theory of mechanistic
constitution, 92, 98, 101, 102,
108
relata, 3n5, 14, 14n9, 1619, 26, 30,
34, 97, 98n3, 99n6, 110,
112, 127, 148n3, 149, 172,
1756, 184, 192, 194, 213,
2226, 230, 233, 237,
240, 243, 255, 256, 259,
263, 264
relations
A-realization, 14n9
big-G Grounding, 171, 173n1
broadly physical, 34, 2527, 261,
264
causal composition, 152, 156,
160, 161, 173
classical mereological parthood,
152

Index

classical mereological partwhole,


156, 171, 173, 191, 207,
208, 216, 222, 230, 232,
242
determinable/determinate, 32,
132, 133, 151n1, 152, 154n9,
156, 171, 173, 177n4, 1801,
183, 18691
L-realization, 14n9
made-up-of, 31, 1236, 130,
1336, 138
manipulability, 5566, 21416,
22832
M-realization, 1415n9
narrowly physical, 34, 251, 252,
2546, 257, 2605
non-causal, 1n2
priority relations, in absence of
fundamental level, 1969
proper subset, 133, 152, 156,
173, 174, 182, 190
R-relation, 2668
set membership, 133, 152, 156,
173, 177n4, 183, 191n15,
198
small-c causal, 153, 155, 160,
180, 1835, 186
small-g grounding, 143, 1512,
1557, 15961, 166, 171,
173, 174, 17680, 181n6,
185, 18996, 198, 200
vertical relations, 1, 1n2, 2, 4, 5,
25
relative fundamentality, 158,
159n12, 195, 196, 199202
robustness, 55
role-playing, 23640, 258
R-relation, 2668

309

S
scientic composition, 2, 1113, 30,
20712, 214, 221, 22544
dialectical importance of, 1820
Scientic Revolution, 205
semantic, logical, abstract, and/or
mathematical (SLAM)
entities, 21113, 233, 234,
243
set membership relation, 133, 152,
156, 173, 177n4, 183,
191n15, 198
small-c causal relations, 153, 155,
160, 180, 1835, 186
small-g grounding relations, 143,
1512, 1557, 15961, 166,
171, 173, 174, 17680,
177n4, 181n6, 185, 18996,
198, 200
spacetime region
role in mechanistic and material
constitution, 114
special sciences debate, 28693
stability, 557
standard functionalist framework,
21721, 23642
standard model, 4469
carving, 48, 5364
causal, 4853
levels, 649
stipulation, 254, 255
stop problem, 66, 67, 67n14, 69
structural equation modeling (SEM),
323, 79, 144, 150, 154,
1627, 17881, 183, 18591,
201, 213, 214, 2345
Subset/Flat view of realization, 220,
228, 237

310

Index

suciency, 111, 232, 234


supervenience, 92, 114, 145, 151,
172, 173, 231, 232, 250, 259,
260, 281

T
theory of object eliminativism, 114
token causal powers, 131n1, 1323
token identity, 152
topic-neutral Ramseycation, 219
transference accounts, 184
transitivity, 184
type identity, 104, 111, 133, 173,
284
typetype reductionist proposals,
127

U
Unengaged approach, 21013, 238
unexplanatory phenomenological
models, 44, 44n2
unicationist model of explanation,
93
unity argument, for Grounding,
17691

V
verticality. See vertical relations
vertical relations, 1, 1n2
general characterization of, 25
neo-Functionalist account of, 35
V-frameworks, 36, 11
comparative, 7, 20, 24, 26
competitive, 6, 9, 10, 18, 19
focused, 8, 20, 24, 26, 30, 32, 33
Grounding, 4, 5, 1719, 2223,
25, 312
meta-justifying, 278
meta-success, 9, 20, 27, 33
neo-Causal, 4, 12, 13
neo-Functionalist, 1516, 25
types of, 203
Vitalism, 22831, 233, 235

W
Wilsons pluralistic framework, 15661
working parts. See good parts

Z
zooming error problem, 29, 47,
47n7, 65